Quantitative Microbial Risk Assessment to
Estimate Illness in Freshwater Impacted by
Agricultural Animal Sources of Fecal Contamination
U.S. Environmental Protection Agency
Office of Water
December 2010
EPA 822-R-10-005
-------
U.S. Environmental Protection Agency
DISCLAIMER
Mention of commercial products, trade names, or services in this document or in the references
and/or endnotes cited in this document does not convey, and should not be interpreted as
conveying, official EPA approval, endorsement, or recommendation.
December 2010 ii
-------
U.S. Environmental Protection Agency
FOREWARD
Managing and minimizing the public health threat associated with fecal pollution in recreational
water are important aspects of policy development and regulation for the U.S. Environmental
Protection Agency (EPA) Office of Water. The Beaches Environmental Assessment and Coastal
Health Act of 2000 (BEACH Act) amended the Clean Water Act Sections 104 (v) and 304(a)(9)
to require EPA to conduct studies associated with pathogens and human health, and to publish
new or revised Recreational Water Quality Criteria (RWQC). To meet these requirements, EPA
is conducting a series of studies that will inform the development of new or revised RWQC.
This document describes a quantitative microbial risk assessment (QMRA) that was conducted to
to estimate illness in freshwater impacted by agricultural animal sources of fecal contamination.
This assessment was based on the EPA/International Life Sciences Institute Framework for
Microbial Risk Assessment (ILSI, 1996), and the structure follows EPA's peer-reviewed
Microbiological Risk Assessment (MRA) Tools, Methods, and Approaches for Water Media
(MRA Tools document), which has been peer-reviewed by renowned microbial risk assessors
and the EPA Science Advisory Board.
December 2010 Hi
-------
U.S. Environmental Protection Agency
CONTENTS
LIST OF TABLES VII
LIST OF FIGURES IX
ACRONYMS AND ABBREVIATIONS XI
1. EXECUTIVE SUMMARY 1
1.1. QMRA METHODS 2
1.1.1. Forward QMRA methods 2
1.1.2. Relative QMRA methods 3
1.2. RESULTS 3
1.2.1. Risk of illness associated with recreation at a freshwater beach impacted by
agricultural animal sources of fecal contamination 3
1.2.2. Comparison of risks for freshwater beaches impacted by agricultural animal and
human sources of fecal contamination 5
1.3. CONSIDERATIONS 5
2. PROBLEM FORMULATION 7
2.1. STATEMENT OF CONCERN 7
2.2. PURPOSE AND CONTEXT 7
2.2.1. Purpose 7
2.2.2. Context for using QMRA to estimate recreational water risks 7
2.2.3. Prior use of QMRA to estimate risks associated with waterborne pathogens 11
2.3. SCOPE AND RISK RANGE 13
2.3.1. Hazards 13
2.3.2. Reference pathogens 14
2.3.3. Livestock-impacted sites 19
2.3.4. Human-impacted sites 20
2.3.5. Shorebird-impacted sites 21
2.4. POPULATIONS INCLUDED IN THE RISK ASSESSMENT MODEL 21
2.5. REFERENCE HEALTH OUTCOMES 22
2.6. UNITS OF EXPOSURE AND ROUTE OF CONCERN 22
2.7. TARGET RISK LEVEL 23
2.8. SCENARIOS MODELED 23
2.9. QUESTIONS TO BE ADDRESSED 24
2.10. CONCEPTUAL MODELS 24
2.10.1. Top-tier models 24
2.10.2. Sub-tier model: model parameter form and estimation 27
2.10.3. Sub-tier model: animal-impacted water pathogen-loading model 28
2.10.4. Sub-tier model: reference pathogen dose-response models 31
2.10.5. Sub-tier model: volume of water ingested during recreational activities 33
2.10.6. Sub-tier model: secondary infections 34
2.11. SUMMARY OF QMRA METHODS 34
December 2010 iv
-------
U.S. Environmental Protection Agency
2.11.1. Risk of illness associated with recreation at a beach impacted by agricultural animal
sources of fecal contamination 35
2.11.2. Comparison of animal-impacted water risks with POTW-impacted water 36
2.12. ENVIRONMENTAL SAMPLING 38
2.13. TOOLS USED IN THE QMRA 40
2.14. SUMMARY OF ASSUMPTIONS 40
2.15. SOURCES OF VARIABILITY AND UNCERTAINTY 41
2.15.1. Variability 42
2.15.2. Sources of uncertainty 43
2.16. FACTORS AND DATA NOT INCLUDED IN THE QMRA 44
2.17. IDENTIFIED GAPS IN THE KNOWLEDGE BASE 44
3. ANALYSIS 46
3.1. EXPOSURE 46
3.1.1. Prevalence and abundance of reference pathogens in livestock 47
3.1.2. Abundance of reference pathogens in disinfected secondary effluent 51
3.1.3. Abundance of FIB in livestock manures 52
3.1.4. Ability of livestock-derived reference pathogens to infect humans 53
3.1.5. Mobilization of reference pathogens and FIB 58
3.1.6. Factors used to convert densities of pathogens on land to densities in runoff 60
3.1.7. Volume of water ingested 60
3.1.8. Exposure profile 61
3.2. HEALTH EFFECTS 61
3.2.1. Health endpoint 61
3.2.2. Dose-response relationships 62
3.2.3. Morbidity 64
3.2.4. Health effects profile 66
4. RISK CHARACTERIZATION 67
4.1. RISK OF ILLNESS ASSOCIATED WITH RECREATION AT A BEACH IMPACTED BY
AGRICULTURAL ANIMAL SOURCES OF FECAL CONTAMINATION 67
4.
4.
4.
.1. Methods 68
.2. Base analysis cattle results 73
.3. Base analysis pig results 76
4. .4. Base analysis chicken results 78
4. .5. Base analysis comparison of results 79
4.1.6. Sensitivity analysis results for alternate dose-response relationships 81
4.1.7. Sensitivity analysis results for alternate ingestion 84
4.2. RELATIVE QMRA FOR ANIMAL-IMPACTED WATER AND HUMAN-IMPACTED WATER 85
4.2.1. Methods 86
4.2.2. Relative QMRA results 90
4.3. DISCUSSION 93
4.3.1. Interpretation of results 93
4.3.2. Considerations and caveats 97
4.4. CONCLUSIONS 100
December 2010 v
-------
U.S. Environmental Protection Agency
REFERENCES 102
APPENDICES A-l
APPENDIX A. SELECTED PEER-REVIEWED QMRAs FOR RECREATIONAL WATER EXPOSURE A-l
APPENDIX B. DATA SUMMARY REFERENCE PATHOGENS IN LIVESTOCK AND HUMAN WASTE B-l
APPENDIX C. SHOREBIRDS AND STORMWATER REFERENCE PATHOGEN LITERATURE REVIEW C-1
APPENDIX D. EPA ENVIRONMENTAL MONITORING PROGRAM D-l
APPENDIX E. PATHOGEN AND FIB MOBILIZATION FRACTIONS DUE TO RAINFALL E-l
APPENDIX F. MICROBIAL RISK ASSESSMENT INTERFACE TOOL SIMULATION IMAGES F-l
ANNEXES (each under separate cover)
ANNEX 1 STATE-OF-THE-SCIENCE REVIEW OF QUANTITATIVE MICROBIAL RISK ASSESSMENT:
ESTIMATING RISK OF ILLNESS IN RECREATIONAL WATERS
ANNEX 2 DEVELOPMENT OF A QMRA MODEL TO EVALUATE THE RELATIVE IMPACTS TO HUMAN
HEALTH RISKS FROM ANIMAL-IMPACTED RECREATIONAL WATERS
ANNEX 3 DISTRIBUTION AND PREVALENCE OF SELECTED ZOONOTIC PATHOGENS IN U.S.
DOMESTIC LIVESTOCK
December 2010 vi
-------
U.S. Environmental Protection Agency
LIST OF TABLES
Table 1. Estimated annual illnesses in the United States from known pathogens 15
Table 2. Pathogenic organisms in animal waste of concern to human health 16
Table 3. Estimated densities of reference pathogens in disinfected secondary effluent 21
Table 4. Variable parameters and underlying causes for their variations 42
Table 5. Salmonella serotype prevalences 55
Table 6. Valid Cryptosporidium species and associated major and minor hosts 57
Table 7. Cryptosporidium spp. of humans and domestic animals 57
Table 8. Cryptosporidiumparvum dose-response parameter estimates 58
Table 9. Abundance of reference pathogens in agricultural animal sources 69
Table 10. Prevalence of infection (% of animals shedding reference pathogens at any point in time 70
Table 11. Human infectious potential 70
Table 12. Mobilization fractions for land applied fecal wastes (logic values) 71
Table 13. Dose-response models and morbidity 73
Table 14. Summary of infection and illness risks from recreation in cattle manure-impacted water 74
Table 15. Summary of infection and illness risks from recreation in pig slurry-impacted water 76
Table 16. Summary of infection and illness risks from recreation in chicken-litter impacted water 78
Table 17. Alternate ingestion: Cryptosporidium infection and illness from pig-impacted runoff 84
Table 18. Abundance of fecal indicator bacteria in fecal sources 88
Table 19. Mobilization of fecal indicator bacteria for animal fecal sources 89
Table 20. Relative QMRA illness risks from exposure to agricultural animal-impacted water 90
Table 21. Synopsis of selected peer-reviewed QMRAs of recreational water exposure A-1
Table 22. Reported Salmonella densities in livestock feces and other matrices B-l
Table 23. Reported Campylobacter spp. densities in livestock manure and other matrices B-3
Table 24. Reported Cryptosporidium spp. densities in livestock manure and other matrices B-5
Table 25. Reported Giardia spp. densities in livestock manure and other matrices B-7
Table 26. Reported E. coll O157:H7 densities in livestock manure and other matrices B-8
Table 27. Reported rotavirus densities in treated sewage B-9
Table 28. Reported adenovirus densities in treated sewage B-9
Table 29. Reported norovirus densities in treated sewage B-10
Table 30. Reported Salmonella densities in treated sewage B-10
Table 31. Reported Campylobacter spp. densities in treated sewage B-10
Table 32. Reported Cryptosporidium spp. densities in treated sewage B-ll
Table 33. Reported Giardia spp. densities in treated sewage B-ll
December 2010 vii
-------
U.S. Environmental Protection Agency
Table 34. Reported E. coll O157:H7 densities in treated sewage B-12
Table 35. Avian species associated with Cryptosporidium and Giardia C-3
Table 36. Reported reference pathogen densities in stormwater-dominated water C-5
Table 3 7. Percent solids of poultry and cattle manure applied to experimental plots D-3
Table 3 8. Organisms and methods used for analysis of water and manure samples D-6
Table 39. Manure application rates E-2
Table 40. Method-organism combinations and data availability E-3
Table 41. Mobilization fraction ranges and means for pathogens E-4
Table 42. Comparison of typical and experimental manure FIB densities E-5
Table 43. Runoff FIB densities for plots with and without manure application E-6
Table 44. Mobilization and abundance distributions, alternative 1 E-10
Table 45. Mobilization and abundance distributions, alternative 2 E-10
December 2010 viii
-------
U.S. Environmental Protection Agency
LIST OF FIGURES
Figure 1. Summary of forward QMRA results 4
Figure 2. Summary of relative QMRA results 5
Figure 3. Flowchart for forward QMRA 9
Figure 4. Flow chart for a reverse QMRA 10
Figure 5. Anchoring a QMRA using observed pathogen densities and health effects 12
Figure 6. Non-foodborne illnesses in the United States 16
Figure 7. Forward QMRA conceptual model 25
Figure 8. Relative QMRA conceptual model 26
Figure 9. Reverse QMRA conceptual model 27
Figure 10. Transport of pathogens and indicators to swimmers from livestock manure 29
Figure 11. Conceptual model of paths for livestock pathogens reaching recreation sites 31
Figure 12. Ingested volumes for the combined data (children and adults) 33
Figure 13. Ingested volumes, child and adult data separated 34
Figure 14. Schematic exposure diagram for recreation at agricultural animal-impacted waterbody 47
Figure 15. Salmonella enterica prevalence in humans and livestock 56
Figure 16. Interaction between health effects and risk characterization components 66
Figure 17. Detailed conceptual model for forward QMRA 68
Figure 18. Probability of infection and illness from recreation in cattle-impacted water 74
Figure 19. Cumulative probability illness risk plot for cattle manure-impacted water 75
Figure 20. Probability of infection and illness from recreation in pig slurry-impacted water 76
Figure 21. Cumulative probability illness risk plot for pig slurry-impacted water 77
Figure 22. Probability of infection and illness from recreation in chicken litter-impacted water 78
Figure 23. Cumulative probability illness risk plot for chicken litter-impacted water 79
Figure 24. Comparison of illness risks from recreation in agricultural animal-impacted runoff. 80
Figure 25. Probability density for illness from recreation in animal-impacted water 80
Figure 26. Cumulative probability plot: evaluation of alternative dose-response for Cryptosporidium ... 82
Figure 27. Cumulative probability plot: evaluation of alternative dose-response for Campylobacter 82
Figure 28. Cumulative probability plot: evaluation of alternative dose-response for E. coli O157 83
Figure 29. Alternate ingestion: Cryptosporidium infection and illness from pig-impacted runoff 85
Figure 30. Relative QMRA approach 1 probability of illness boxplot 91
Figure 31. Relative QMRA approach 2 probability of illness boxplot 91
Figure 32. Probability density for illness risks from E. coli relative QMRA approach 2 92
Figure 3 3. Probability density for illness from recreation in disinfected secondary effluent 95
December 2010 ix
-------
U.S. Environmental Protection Agency
Figure 34. Routes by which bird-origin FIB and pathogens reach recreation sites C-l
Figure 35. Flistogram of cumulative runoff volumes from plots subject to the design rain event E-l
Figure 36. Flistogram of mobilization fractions for enterococci from plots treated with cattle manure.. E-7
Figure 37. Flistogram of enterococci mobilization fractions from plots treated with swine slurry E-7
Figure 38. Flistogram of mobilization fractions for enterococci from plots treated with poultry litter ... E-8
Figure 39. Flistogram of E. coll (via Colilert) mobilization fractions for cattle manure plots E-8
Figure 40. Flistogram of E. coli (via Colilert) mobilization fractions for swine slurry plots E-9
Figure 41. Flistogram of E. coli (via Colilert) mobilization fractions for poultry litter plots E-9
December 2010
-------
U.S. Environmental Protection Agency
ARS
AWQC
BEACH Act
BMP
BPW
CD
CPU
CWA
DNA
EHEC
EPA
EtOH
FIB
FSIS
GI
GM
HCGI
ILSI
IMS
IU
mEI
mL
MPN
MRAIT
ND
NEEAR
NM
NRC
NRCS
PBS
PBW
PCR
PFU
POTW
QMRA
qPCR
ACRONYMS AND ABBREVIATIONS
Agricultural Research Service
ambient water quality criteria
Beaches Environmental Assessment and Coastal Health Act of 2000
best management practice
buffered peptone water
Consent Decree
colony forming units
Clean Water Act
deoxyribonucleic acid
enterohemorrhagic E. coli
U.S. Environmental Protection Agency
ethyl alcohol
fecal indicator bacteria
Food Safety and Inspection Service
gastrointestinal
geometric mean
Highly Credible Gastrointestinal Illness
International Life Sciences Institute
immunomagnetic separation
infectious units
membrane-enterococus indoxyl-B-D-glucoside (agar)
milliliters
Most Probable Number
Microbial Risk Assessment Interface Tool
non-detect
National Epidemiological and Environmental Assessment of Recreational
(Water Study)
not measured
National Research Council
Natural Resources Conservation Service
phosphate buffered saline
phosphate buffered water
polymerase chain reaction
plaque forming units
publicly owned (sewage) treatment works
quantitative microbial risk assessment
quantitative polymerase chain reaction
December 2010
xr
-------
U.S. Environmental Protection Agency
RT-PCR
RWQC
SA
SD
STEC
TSC
U.K.
U.S.
USDA
UV
VBNC
WHO
WWTP
reverse transcriptase polymerase chain reaction
recreational water quality criteria
Settlement Agreement
standard deviation
Shiga toxin (producing) E. coli
tryptose sulfite cycloserine (agar)
United Kingdom
United States
U.S. Department of Agriculture
ultraviolet (light)
viable but non-culturable
World Health Organization (United Nations)
wastewater treatment plant
December 2010
xn
-------
U.S. Environmental Protection Agency
1. Executive Summary
Under the Beaches Environmental Assessment and Coastal Health Act of 2000 (BEACH Act),
EPA committed to "conduct quantitative microbial risk assessment (QMRA) (based on
measurement of pathogenic organisms and indicators^) to estimate illness at a freshwater beach
9
impacted by agricultural animal sources of fecal contamination." This report documents EPA's
activities to meet this commitment and addresses the following two questions:
1. What is the risk of illness associated with recreation at a freshwater beach impacted by
agricultural animal (cattle, swine, and chicken) sources of fecal contamination? and
2. How do those risks compare to risks associated with freshwater beaches impacted by
human sources of fecal contamination?
The QMRA characterizes risks on a single recreation event basis for the general population and
is defined by the following assumptions: (1) fresh cattle manure, pig slurry, and poultry litter
(fecal materials) are land-applied at standard agronomic (maximum U.S. allowable) rates
adjacent to a freshwater beach; (2) the fresh fecal materials contain fecal indicator bacteria (FIB)
and reference pathogens consistent with levels reported in the peer-reviewed literature; (3) FIB
and reference pathogens from the fresh land-applied fecal materials reach the freshwater beach
via runoff from an intense rainfall event; (4) FIB and reference pathogens are mobilized during
the rainfall event at levels consistent with those observed during the EPA environmental
monitoring studies; (5) primary contact recreation (e.g., swimming) occurs in the undiluted
runoff; and (6) exposure to reference pathogens occurs through water ingestion during
recreation. This scenario is intentionally formulated to result in health-protective estimates of
risk3 (conservative).
The QMRA indicates that the median risk of illness from recreational exposure to the cattle-
impacted waterbody is equivalent to the risk associated with the 1986 (current) recreational
water quality criteria (RWQC)4 (USEPA, 1986). The median risk of illness from exposure to the
pig-impacted waterbody is approximately four-times lower than the risk associated with the
current RWQC, and the median risk of illness from exposure to the chicken-impacted waterbody
is approximately 300-times lower than the risk associated with the current RWQC.
1 Fecal indicator bacteria provide an estimation of the amount of feces, and indirectly, the presence and quantity of fecal
pathogens in the water (NRC, 2004).
2 Case 2;06-cv-04843-PSG-JTL Document 159-3 Files 08/08/2008 Page 3 of 15,
http://www.epa.gov/waterscience/criteria/recreation/pdf/sa.pdf
3 "Conservative" is used here to note that risk estimates will err on the side of a higher value and thus be more protective of
human health.
4 The 1986 RWQC were based on the results of a series of epidemiology studies conducted in human fecal matter-impacted
(human impacted) water and establish a level of health protection in recreational fresh waters at 8 cases of Highly Credible
Gastrointestinal Illness (HCGI) per 1000 recreation events.
December 2010 1
-------
U.S. Environmental Protection Agency
In comparing risks in waterbodies that contain FIB at the current RWQC from land-applied
agricultural animal fecal material, the predicted median risks of illness are at least 20- to 30-
times lower than the risk associated with human-impacted water (risks for cattle and chicken
impacted waters are lower depending on the FIB used). If FIB are present at the current RWQC
from fecal material deposited directly into a waterbody, pig- and chicken-impacted water risks
are similar to the land-applied risks, whereas cattle-impacted water risks are similar to the
current RWQC.
1.1. QMRA Methods
This QMRA follows the EPA and International Life Sciences Institute Framework for Microbial
Risk Assessment (ILSI, 1996) and employs peer-reviewed tools and approaches (USEPA, 2010).
A traditional forward QMRA characterizes the risk of illness associated with recreation at a
freshwater beach (first question above). A relative QMRA provides a comparison of the
estimated risks at the current RWQC from recreation in water impacted by agricultural sources
of fecal contamination to those associated with human-impacted water (second question above).
In this QMRA, we use a probabilistic framework and characterize each model parameter using a
statistical distribution where the parameters of those distributions account for variability and/or
uncertainty.
1.1.1. Forward QMRA methods
For each of the animal sources (fresh cattle manure, swine slurry, and poultry litter), the density
of reference pathogens5 in the runoff (USEPA, 2009b) is calculated based on data (see Appendix
B) describing the reference pathogen density in land-applied fecal material, the prevalence of
infection (percent of infected animals), the human infectious potential of the reference pathogens
from the agricultural animals, and the proportion of the applied reference pathogens that run-off
following a rain event (based on data collected specifically for this risk assessment; see
Appendix D for further information). These data are referred to hereafter as the EPA
environmental monitoring program).
That density is multiplied by the volume of water ingested during recreational activities to
estimate the "dose" of pathogens for this exposure scenario. That dose is input to the appropriate
dose-response relationship resulting in a probability of infection. The probability of infection is
multiplied by a morbidity factor to estimate a probability of illness. The risk associated with
5 In this report, a set of reference pathogens for the EPA recreational water QMRA work was established and is described herein
that represents a large proportion of illnesses in the United States, are representative of the fate and transport of waterborne
pathogens of concern, are present in human and animal waste and recreational waters, can survive in the environment, and have
corresponding dose-response relationships in the peer-reviewed literature. For animal-impacted waters, the reference pathogens
are Cryptosporidium, Giardia, Salmonella enterica, Campylobacterjejuni, and£. coli O157:H7. Other pathogens were also
considered for inclusion as reference pathogens (e.g., Hepatitis E virus, Listeria monocytogenes, orLeptospira)', however, by
comparison, these pathogens are thought to cause few illnesses from recreational water exposure and/or do not have available
dose-response relationships based on human data.
December 2010 2
-------
U.S. Environmental Protection Agency
each fecal contamination source is characterized as the total probability of gastrointestinal (GI)
illness from each source-specific reference pathogen.
7.7.2. Relative QMRA methods
For the relative QMRA, previously developed methods for direct fecal contamination (fecal
material deposited directly into a waterbody) (Schoen and Ashbolt, 2010; Seller et al., 2010b)
are extended by including land application of fecal material, and FIB and reference pathogen
mobilization (proportion of FIB and reference pathogens that run-off) during rainfall events. The
estimated risks are calculated for a hypothetical waterbody that contains geometric mean FIB
densities at the U.S.-recommended RWQC for recreational freshwaters (33 colony forming units
[CFU] 100 mU1 enterococci and 126 CFU 100 mL"1 E. coli, respectively). We provide separate
calculations for each fecal source/FIB combination.
Pathogen dose is calculated based on observed and literature-based ranges of pathogen and FIB
densities in fecal waste, the prevalence of infection, the fraction of human-infectious strains, and
the proportion of the FIB and pathogens that mobilize during a rain event. Similar to the forward
QMRA, doses are input to the appropriate dose-response relationship resulting in a probability of
infection. The probability of infection is multiplied by a morbidity factor to produce a
probability of illness. The risk associated with each fecal contamination source is characterized
as the total probability of GI illness from each source-specific reference pathogen. The resulting
risk distributions are then compared to benchmark risks for human-impacted waters.
1.2. Results
7.2.1. Risk of illness associated with recreation at a freshwater beach impacted by
agricultural animal sources of fecal contamination
The forward QMRA predicts risk of illness from recreational exposure to the animal-impacted
waterbodies during and immediately after an intense rain event. The forward QMRA simulation
results for the cattle manure, pig slurry, and chicken litter-impacted recreational water are
presented in boxplot format in Figure I.6'7
6 In Figure 1 and subsequent boxplots, the edges of the box represent the 25th and 75th percentiles of the simulation results
(probability of infection or illness), the line in the center of the box is the median value, the whiskers represent the 10th and 90th
percentiles, and the diamonds below and above the whiskers represent the 5th and 95th percentiles, respectively.
In Figure 1 and several subsequent figures, a reference line labeled "Current geometric mean RWQC" is provided. This line
represents an estimate of the GI illness risk associated with the FIB densites that are specified by the geometric mean RWQC
(USEPA, 1986). Simulation median values can be compared to this line to evaluate how the simulation results compare to the
level of risk associated with the current RWQC.
December 2010 3
-------
U.S. Environmental Protection Agency
These results can be summarized as follows:
• The predicted median cumulative risk of illness from recreational exposure to the cattle-
impacted waterbody is effectively equivalent to the risk of illness that is associated with
the current RWQC.
• The predicted median cumulative risk of illness from recreational exposure to the pig-
impacted waterbody is approximately 4-times lower than the risk of illness that is
associated with the current RWQC.
• The predicted median cumulative risk of illness from recreational exposure to the
chicken-impacted waterbody is approximately 300-times lower than the risk of illness
that is associated with the current RWQC.
• E. coli 0157 is the predicted dominant risk agent in cattle-impacted water, followed by
Campylobacter and Cryptosporidium. For pig-impacted water, Campylobacter and
Cryptosporidium are the predicted dominant risk agents, followed by Giardia. For
chicken-impacted water, Campylobacter is the predicted dominant risk agent.
• The predicted variability is greatest for chicken-impacted water and least for pig-
impacted water.
Current geometric
3. slurry Chicken litter
Figure 1. Summary of forward QMRA results
December 2010
-------
U.S. Environmental Protection Agency
1.2.2. Comparison of risks for freshwater beaches impacted by agricultural animal and
human sources of fecal contamination
The relative QMRA simulation results for the cattle manure, pig slurry, and chicken litter-
impacted recreational water are presented in Figure 2.
These results can be summarized as follows:
• At the current geometric mean RWQC, the predicted median risk of illness from
recreational exposure to the cattle-impacted waterbody is approximately 25- to 150-times
lower than risk of illness associated with human sources of contamination.
• At the current geometric mean RWQC, the predicted median risk of illness from
recreational exposure to the pig-impacted waterbody is approximately 30-times lower
than the risk of illness that is associated with human sources of contamination.
• At the current geometric mean RWQC, the predicted median risk of illness from
recreational exposure to the chicken-impacted waterbody is approximately 20- to 5000-
times lower than risk of illness that is associated with human sources of contamination.
Current G. mean
RWQC equivalent
cattle enterococci
pig enterococci
chicken enterococci
cattle £. co//
pig £. co//
chicken E. co//
Figure 2. Summary of relative QMRA results
1.3. Considerations
Like any scientific study, this work has a number of important conceptual constraints. In this
risk assessment, we consolidated a vast range of disparate data and information to support an
improved understanding about risks to human health that would have been difficult or impossible
to characterize through an observational (e.g., epidemiology) study.
December 2010
-------
U.S. Environmental Protection Agency
To facilitate this risk assessment, we necessarily made several simplifying, health-protective
assumptions to limit the scope of the assessment to ensure it could be completed defensibly and
in a timely manner. The most important conceptual constraints and caveats are that (1) the
analyses only considered one intentionally limited exposure scenario; (2) FIB and pathogen
mobilization was modeled on a simulated intense rain event in a single location—we used a
modest set of reference pathogens that represent a large proportion of illnesses in the United
States, however, it is possible that animal-impacted water could contain pathogens of potential
public health concern that we did not evaluate; and (3) we summarized our results to facilitate
comparison to the existing (1986) RWQC and, as such, do not describe the risks associated with
extreme or rare events.
Risk assessment is widely used by governmental and regulatory agencies worldwide to protect
public health from exposure to a myriad of contaminants through numerous routes of exposure.
Air pollution regulations, protection of the food supply chain, and drinking water regulations are
large-scale examples that illustrate the effective use of risk assessment methodologies within a
environmental regulatory context. To date, epidemiology studies have been the primary tool
used to characterize human health risks from exposure to recreational water. Those
epidemiology studies have generally focused on waters impacted by wastewater (human sewage)
effluent. Substantial progress has been made in improving the quality of wastewater effluent in
the United States. However, greater attention is being paid to other contamination sources. In
fact, non-point fecal contamination is one of the most common reasons that U.S. waterbodies are
classified as impaired with respect to their recreational use. Epidemiology studies are not likely
to be effective in characterizing risks in many waters of this type due to technical, logistical,
and/or financial constraints. As illustrated in this report, QMRA is a viable and valuable
complement to epidemiology for waters where epidemiology data are not available, do not apply,
or are impractical to collect. Finally, the data, results, and caveats of this study provide context
for understanding recreational risks in diverse waterbodies, and could help to facilitate
implementation of upcoming new or revised RWQC.
December 2010
-------
U.S. Environmental Protection Agency
2. Problem Formulation
2.1. Statement of Concern
Managing and minimizing the public health threat associated with fecal pollution in recreational
water are important aspects of policy development and regulation for the U.S. Environmental
Protection Agency (EPA) Office of Water. Human exposure to recreational water impacted by
fecal contamination is known to cause a variety of adverse health effects including
gastrointestinal (GI) and respiratory illness (Craun et al., 2005; NRC, 2004; Parkhurst et al.,
2007). Microbial hazards in recreational water contaminated by feces include pathogenic
bacteria, viruses, and parasitic protozoa of human and animal origin. Risks to swimmers may
differ depending on the source (human or animal) of the excreta because (1) the pathogens in
animal manure differ in type, occurrence, and abundance from those in human sewage (WHO,
2004b); and (2) the routes by which human-infectious pathogens of animal origin (zoonoses)
reach swimmers can differ from human enteric pathogens (e.g., intermittent rainfall transport as
compared to wastewater treatment plant effluent with relatively constant flow).
2.2. Purpose and Context
2.2.7. Purpose
This quantitative microbial risk assessment (QMRA) estimates human GI illness associated with
recreation at a freshwater beach contaminated by fecal material from agricultural animal sources
(livestock). It compares those risks to those associated with recreation in water impacted by
human sewage sources. The assessment follows the EPA/International Life Sciences Institute
peer-reviewed microbial risk assessment framework (ILSI, 1996) and employs peer-reviewed
microbial risk assessment tools and approaches (USEPA, 2010).
2.2.2. Context for using QMRA to estimate recreational water risks
The Beaches Environmental Assessment and Coastal Health Act of 2000 (BEACH Act) and the
associated Consent Decree (CD) and Settlement Agreement (SA) require EPA to publish new or
revised recreational water quality criteria (RWQC) by October 2012. To meet these
requirements, EPA is conducting a series of studies as part the Critical Path Science Plan for
Development of New or Revised Recreational Water Quality Criteria (science plan) to form the
technical basis of new or revised RWQC (USEPA, 2007). This QMRA was conducted to meet
the SA requirement to "conduct QMRA (based on measurement of pathogenic organisms and
indicators) to estimate illness at a freshwater beach impacted by agricultural animal sources of
o
fecal contamination."
8 Case 2;06-cv-04843-PSG-JTL Document 159-3 Files 08/08/2008 Page 3 of 15,
http://www.epa.gov/waterscience/criteria/recreation/pdf/sa.pdf
December 2010
-------
U.S. Environmental Protection Agency
Epidemiology studies have linked swimming-associated illnesses with FIB densities in point
source human-impacted recreational water (see reviews by Priiss, 1998; Wade et al., 2003;
Zmirou et al., 2003). For a more recent review, see WERF (2009). In these epidemiology
studies, FIB were used to detect the possible presence of microbial contamination from human
waste (NRC, 2004).
Although several epidemiology studies have considered non-point sources of contamination,
these studies do not specifically link FIB densities to risks from agricultural animals. At a given
level of FIB, risks for animal-impacted water may differ from human-impacted water because
the mix and densities of pathogens in animal manure are different from those in human excreta.
Another important distinction is that pathogen loading to recreational water from animal manure
typically differs (event-driven) from wastewater outfall loading (continuous). Because of these
issues, it would be technically and logistically difficult to conduct epidemiology studies on
predominately agricultural animal-impacted waters. QMRA provides a scientifically defensible
mechanism to characterize risks for agricultural animal-impacted water.
QMRA applies risk assessment principles (NRC, 1983) to approximate the consequences from
exposure to selected infectious pathogens (Haas et al., 1999). For recreational water contact,
QMRA can be used to
• estimate the risk of GI illness for recreational water where no epidemiological data are
available (forward);
• understand which pathogens caused GI illnesses in epidemiological studies (reverse);
• compare the relative levels of risk to human health associated with fecal contamination
from various sources (relative); and
• harmonize QMRA models with epidemiology studies (anchoring).
The QMRA presented in this report uses both forward (traditional) and relative approaches. The
forward QMRA quantifies risks associated with specific animal waste (cattle, swine, and
chicken) runoff scenarios. The relative QMRA compares risks associated with cattle, swine, and
chicken-impacted water to risks associated with recreation in human-impacted water with FIB
densities at the current RWQC levels (USEPA, 1986).
2.2.2.1. Forward QMRA
The pathogen densities to which swimmers are exposed depend upon myriad factors, the most
important of which is the primary source of fecal pollution at the site (Dorevitch et al., 2010;
Schoen and Ashbolt, 2010; Seller et al., 2010a; 2010b; WERF, 2009). In a forward QMRA,
knowledge of prevalence of infection and abundance of pathogens in sources is used to predict
risks of infection or illness associated with recreational activities. In this traditional QMRA
approach, an exposure assessment (statistical analyses of pathogen occurrence, ingestion
volumes, and abundance in sources and fate and transport modeling) is used to estimate the
December 2010 8
-------
U.S. Environmental Protection Agency
pathogen density in water and the volume of water ingested. An estimate of human health risk is
then computed based on pathogen specific dose-response relationships (Figure 3). Thus, forward
QMRA assesses risks based on particular site or exposure features (e.g., water primarily affected
by runoff events). The QMRA described in this report includes a forward QMRA.
Forward QMRAs provide a mechanism to link animal manure exposure with human health risks.
The few epidemiology studies based on inland water affected by animal waste have not produced
risk estimates associated with recreation or data linking FIB with risk. Epidemiology studies
may be limited in this regard due to (1) the temporally sporadic pathogen loading to recreational
water from animal manure, and (2) a decoupling between the FIB that are traditionally used in
epidemiology studies and the mix of pathogens present in animal manure.
Revise model or
model parameters
Estimate pathogen density
(exposure) via modeling
Yes
QM RA: Estimate risk and
confidence interval for Gl
illness
Report risk of Gl
illness, confidence
interval
Figure 3. Flowchart for forward QMRA
December 2010
-------
U.S. Environmental Protection Agency
2.2.2.2. Reverse QMRA
In a reverse QMRA, infection or illness rates associated with recreational activities are used in
conjunction with knowledge about contamination source to make inferences about likely
pathogen presence in a waterbody. The same components are used as with the forward QMRA
approach (the volume of water ingested, pathogen specific dose-response relationships). The
reverse QMRA output (i.e., pathogen densities) is the starting point for forward QMRA (and vice
versa) (Figure 4).
Observed FIB
densities
Observed
relationship
between health
effects and FIB
QMRA: Derive pathogen
densities that are consistent
with observed health effects
and primary sources
Revise estimates for
pathogen mixture and
density
Conduct uncertain!
sensitivity analv
Yes
Are data
available to
refine risk
estimates?
No
Report pathogen
mixture, density, and
confidence intervals
Figure 4. Flow chart for a reverse QMRA
In the absence of direct pathogen monitoring, reverse QMRA provides a mechanism to infer
pathogen densities in a recreational waterbody from specific sources of fecal contamination and
can clarify epidemiology study results (Seller et al., 2010a). For example, Seller et al. (2010a)
used reverse QMRA to identify human enteric viruses, particularly norovirus, as the likely
causes of the observed illnesses from the epidemiology studies conducted in 2003 to 2004 on the
Great Lakes in the United States as part of EPA's National Epidemiological and Environmental
Assessment of Recreational (NEEAR) Water Study.
December 2010
10
-------
U.S. Environmental Protection Agency
Because of the lack of epidemiological data associating illness caused by swimming in livestock-
impacted water with water quality measures, a reverse QMRA was not conducted as part of this
report.
2.2.2.3. Relative QMRA
The relative QMRA approach compares risks associated with recreation in water affected by
human and non-human sources of fecal pollution. The relative QMRA approach allows direct
comparison of the risks for various sources because the approach assumes that each source
contributes a given level of FIB (e.g., 33 CPU enterococci/100 mL). Estimated distributions of
FIB and reference pathogens in each source are used to calculate the relative levels of risk
(Schoen and Ashbolt, 2010; Seller et al., 201 Ob). This report includes a relative QMRA.
2.2.2.4. QMRA anchoring
QMRA anchoring harmonizes QMRA models with epidemiology studies. These assessments
require both water quality information (as measured by FIB) and epidemiological data for a
given site. The anchoring process compares health impacts predicted using QMRA based on
water quality data with observed health effects from epidemiology studies. Next, QMRA model
parameters are adjusted to improve agreement between observations and predictions. QMRA
anchoring can be used to extend QMRA models to sites where epidemiological studies are
impractical or unavailable (Figure 5). This QMRA approach has only recently been proposed
(WERF, 2009) and has not yet appeared in the literature.
2.2.3. Prior use of QMRA to estimate risks associated with water borne pathogens
EPA conducted a detailed literature review to document the use of QMRA to estimate the risks
associated with recreational water impacted by cattle, swine, or poultry waste in (Annex 1). That
review established the QMRA state-of-the-science for waterborne contaminants and provides
insight into the techniques available for use in a QMRA of animal-impacted water. The
literature search yielded approximately 300 QMRA studies and was used to
• identify the pathogens that QMRAs most commonly address;
• identify how QMRA studies address variability and uncertainty;
• assess how often QMRAs include secondary transmission;
• identify which QMRA elements support RWQC; and
• compare the methods used for sensitivity analyses and risk characterization.
Sixteen of those studies estimated risks associated with waterborne recreation and all but one
was a forward QMRA. Appendix A provides a synopsis of the 16 studies. The literature review
indicates that QMRA has been used in a variety of scenarios and is useful when other techniques
such as epidemiology studies are impossible or cost-prohibitive.
December 2010 11
-------
U.S. Environmental Protection Agency
Pathogen
densities
Figure 5. Anchoring a QMRA using observed pathogen densities and health effects
Several observations may be drawn from these studies. First, the studies focused on a small
subset of pathogens that may be important in waterborne exposure during recreational activities.
The two pathogens analyzed most frequently—rotavirus and Cryptosporidium—are important
contributors to risk of GI illness, primarily due to their high infectivity, frequent occurrence in
sewage, and relatively high persistence in environmental matrices. Other human enteric viruses,
particularly noroviruses, have been implicated in numerous outbreaks since the 1950s (Sinclair et
al., 2009), making their absence in QMRA studies notable. However, a recently published dose-
response relationship for norovirus (Teunis et al., 2008a) has helped address this gap (Schoen
and Ashbolt, 2010; Seller et al., 2010a).
Second, a lack of comprehensive data on pathogen occurrence in sources and pathogen fate and
transport characteristics limits the ability to model variability in pathogen sources. In the
literature review, the two most common methods to account for source variability were (1) using
empirical distributions for pathogen density based on relatively limited data, and (2) assuming
log-normally distributed pathogen densities. The QMRA effort described in this document
explicitly models variability and uncertainty based on FIB and pathogen density data drawn from
the peer-reviewed literature, and from EPA field studies conducted specifically for this risk
assessment.
December 2010
12
-------
U.S. Environmental Protection Agency
Third, these studies used limited dose-response modeling, and most did not account for
variability and uncertainty in dose-response model parameters because high quality and diverse
dose-response model data are limited. Our QMRA analyses rely on dose-response models from
peer-reviewed studies. While those models represent the state-of-the-art, they may not
adequately describe risks associated with susceptible sub-populations. Understanding this
limitation, we used a probabilistic QMRA framework9 to address susceptibility to the extent
possible.
Finally, most risk estimations do not account for secondary transmission and immunity. Several
QMRA studies show that infectious disease transmission attributes can influence risk in
unintuitive ways (Eisenberg et al., 2004, 2008; Riley et al., 2003; Seller et al., 2006, 2009);
however, our QMRA analyses do not explicitly address these parameters, this is because
previous work indicates that they are unlikely to substantially affect the estimated risks, given
the pathogens present in livestock manure, and the relatively infrequent exposure to recreational
water (via incidental ingestion of water during recreation) (Seller and Eisenberg, 2008).
2.3. Scope and Risk Range
2.3.7. Hazards
Although human and animal waste can contain numerous pathogenic microbes, recreational
water monitoring data, public health reporting, epidemiology studies, outbreak reports, and dose-
response studies suggest that a modest subset of these pathogens are representative of the
majority of hazards in human and livestock-impacted recreational water (reference pathogens).
The use of reference pathogens to represent the infectivity and the likely environmental fate and
transport of each microbial group (WHO, 2004a) is a widely accepted practice in the field of
QMRA (Roser et al., 2007; Seller et al., 201 Ob).
Reviews of waterborne transmission of zoonotic pathogens identified pathogens of primary
concern based on their occurrence in water, abundance in animal feces, and persistence and
ability to multiply in the environment (Bicudo and Goyal, 2003; Goss and Richards, 2008;
Rosen, 2000; USEPA, 2009a, 2009b). Based on those criteria, the protozoans Cryptosporidium
and Giardia and the bacterial pathogens E. coli O157:H7, Salmonella, and Campylobacter are
the primary pathogens of concern in livestock waste (Bicudo and Goyal, 2003; Goss and
Richards, 2008; Rosen, 2000). Pathogens and diseases of secondary concern include Yersinia
enterolitica (Bicudo and Goyal, 2003), brucellosis, and leptospirosis (Rosen, 2000).
Transmission of fecally-associated viruses of animal origin to humans is considered rare (Rosen,
2000; Sobsey et al., 2006), but is an emerging issue.
9 In this QMRA, all parameters, including the dose-response relationship, are characterized by statistical distributions to the
extent that data were available to support the use of a distribution.
December 2010 13
-------
U.S. Environmental Protection Agency
Data published in 1999 indicated that known pathogens accounted for an estimated 38.6 million
illnesses each year in the United States, including 5.2 million due to bacteria, 2.5 million due to
parasites, and 30.9 million due to viruses (Table 1) (Mead et al., 1999). Of those illnesses, 13.8
million were thought to be foodborne, leaving 24.8 million illnesses of which some portion was
due to waterborne exposures (including, but not limited to, recreational water contact).
Several researchers have developed illustrative lists of waterborne pathogens to consider as
reference pathogens for QMRAs of recreational water (Olivieri and Seller, 2002; Rosen, 2000;
Seller et al., 2010b). For example, Rosen (2000) compiled a list of pathogens in human and
animal waste and ranked them in terms of their risk to human health (Table 2).
2.3.2. Reference pathogens
For the EPA recreational water QMRA described in this report, we established a set of eight
reference pathogens that (1) cause a large proportion of non-foodborne illnesses in the United
States from Mead et al. (1999) (Figure 6 and Table 1); (2) are representative of the fate and
transport of other waterborne pathogens of concern (Ferguson et al., 2009); (3) are present in
human and animal waste and recreational water (USEPA, 2009b); (4) can survive in the
environment; and (5) have corresponding peer-reviewed dose-response relationships (USEPA,
2010). The reference pathogens are
• Norovirus
• Rotavirus
• Adenovirus
• Cryptosporidium spp.
• Giardia lamblia
• Campylobacter spp.
• Salmonella10
• E. coli O57:H7.
10 In keeping with the usual convention, in this report, Salmonella refers to Salmonella enterica spp. enteric, except in specific
reference to a different Salmonella species.
December 2010 14
-------
U.S. Environmental Protection Agency
Table 1. Estimated annual illnesses in the United States from known pathogens (SOURCE: adapted
from Mead et al, 1999)
Pathogen
Class
Bacteria
Parasitic
Viral
Pathogen
Bacillus cereus
Botulism, foodborne
Brucella spp.
Campylobacter spp.
Clostridium perfringens
Escherichia coli O157:H7
E. coli, non-0157 (Shiga toxin-
producing E. coli [STEC])
E. coli, enterotoxigenic
E. coli, other diarrheogenic
Listeria monocytogenes
Salmonella Typhi
Salmonella, non-typhoidal
Shigella spp.
Staphy loco ecus food poisoning
Streptococcus, foodborne
Vibrio cholerae, toxigenic
V. vulnificus
Vibrio, other
Yersinia enterolitica
Subtotal
Cryptosporidium parvum
Cyclospora cayetanesis
Giardia lamblia
Toxoplasma gondii
Trichinella spiralis
Subtotal
Norwalk-like virus (norovirus)
Rotavirus
Astrovirus
Hepatitis A
Subtotal
Total
Total
Estimated
Annual Cases
27,300
58
1554
2,453,926
248,520
73,480
36,740
79,420
79,420
2518
824
1,412,498
448,240
185,060
50,920
54
94
7880
96,368
5,204,934
300,000
16,264
2,000,000
225,000
52
2,541,316
23,000,000
3,900,000
3,900,000
83,391
30,883,391
38,629,641
%
Foodborne
100
100
50
80
100
85
85
70
30
99
80
95
20
100
100
90
50
65
90
10
90
10
50
100
40
1
1
5
#
Foodborne
27,360
58
77
1,963,141
248,520
62,458
31,229
55,594
23,826
2493
659
1,341,873
89,648
185,060
50,920
49
47
5122
86,731
4,175,565
30,000
14,638
200,000
112,500
52
357,190
9,200,000
39,000
39,000
4170
9,287,170
13,814,924
#Non-
foodborne
0
0
111
490,785
0
11,022
5,511
23,826
55,594
25
165
70,625
358,592
0
0
5
47
2758
9637
1,029,369
270,000
1626
1,800,000
112,500
0
2,184,126
13,800,000
3,861,000
3,861,000
79,221
21,601,221
24,814,717
December 2010
15
-------
U.S. Environmental Protection Agency
Table 2. Pathogenic organisms in animal waste of concern to human health (SOURCE: adapted from Rosen, 2000)
Type of Oganism
Pathogens of Primary Concern
Pathogens of Secondary
Concern
Protozoa
Cryptosporidium spp. (C. parvum, C. hominis)
Giardia spp.
Cryptosporidium spp. (others)
Toxoplasma gondir
Balantidium coli
Bacteria
Campylobacter spp. (C. jejuni, C. coli)
E. coli O157:H7
E. coli, non-0157 STEC
E. coli, enterotoxigenic
E. coli, other diarrheogenic
Listeria monocytogenes
Salmonella enterica (particularly serotypes
associated human infection, including enteritidis,
newport, typhimurium)
Shigella spp.
Vibrio cholerae, toxigenic
Yersinia enterocolitica^
Brucella spp.
Leptospirosis interrogates
Viruses
Adenovirus'
Astrovirus^
Calciviruses1^
CoxsackievuW
EchoviiW
Hepatitis A1^
Hepatitis E
NorovitW
Rotavirus
Bovine rotavirus
T Not known to originate
* Not considered a major
from livestock sources
source of waterborne infection
All
others Campylobacter E. coli, all types Salmonella
0.1%
Hepatitis
0.3%
Spp.
2.0%
Astrovirus
15.6%
Rotaviru
15.6%
0 3% /^~ nontyphoidal
' "/ 0.3%
// Shigella spp.
-^^~ 1.4% Cryptosporichum
parvum
1.1%
^ Giardia lamblia
7.3%
Toxoplasma gondii
0.5%
Norwalk-like
viruses
55.6%
Figure 6. Non-foodborne illnesses in the United States (SOURCE: adapted from Mead et al., 1999)
December 2010
16
-------
U.S. Environmental Protection Agency
These eight reference pathogens adequately represent the risk from pathogens potentially present
in fecal matter and in the diverse range of U.S. recreational waters. In addition, their selection is
consistent with previous EPA work that suggested prioritizing standard methods and recreational
and drinking water guidelines for Salmonella spp., Campylobacterjejuni, E. coll O157:H7,
Cryptosporidium, Giardia, and selected viral contaminants (USEPA, 2005b).
In this QMRA, we use the bacterial and parasitic protozoan reference pathogens to characterize
the risk associated with animal-impacted waters (since the viruses are primarily species-specific,
we do not need them to estimate the risk of human GI illness from animal-based water
contamination). In previous related work, we used viral, bacterial, and parasitic protozoan
reference pathogens to evaluate risks associated with human-impacted waters (Schoen and
Ashbolt, 2010; Seller et al., 2010a, 201 Ob).
Appendix B presents data on the occurrence of the reference pathogens in the fecal pollution
sources of interest, and the analysis chapter (3) describes dose-response models for each of the
reference pathogens. Justification follows for inclusion of each of the eight microorganisms as a
reference pathogen for the EPA recreational water QMRA work.
2.3.2.1. Norovirus
Noroviruses are an important cause of human enteric infection and illness. They are estimated to
cause approximately 23,000,000 illnesses in the United States annually (Mead et al., 1999) and
are associated with up to 90% of the epidemic nonbacterial gastroenteritis (GI illness) worldwide
(Lindesmith et al., 2003). Norovirus illness is not limited to young children (Dolin, 2007);
however, a portion of the general population appears to be immune to infection from specific
norovirus genotypes, perhaps due to memory immune response (Lindesmith et al., 2003). Teunis
et al. (2008a) recently published a norovirus dose-response study that expresses dose in terms of
quantitative polymerase chain reaction (qPCR) genome equivalents. Noroviruses are resistant to
water treatment (Haramoto et al., 2006; Laverick et al., 2004; Lodder and de Roda Husman,
2005; Pusch et al., 2005; van den Berg et al., 2005), and remain infective for prolonged periods
of time in the environment (Allwood et al., 2005; Lee et al., 2008). Strains of norovirus also
exist that are uniquely associated with animals (Mattison et al., 2007). Direct zoonotic
transmission appears to be rare, but genetic mixing of animal and human viruses seems plausible
with the finding that common human strains replicate in pigs and cattle (Koopmans, 2008).
2.3.2.2. Adenovirus
Adenoviruses are primarily of human origin, although some animals are known to be infected by
and shed host-specific variants. Adenovirus types vary widely in their pathology, with strains 40
and 41 causing enteric infections in young children, and with secondary contributions by strains
2 and 31 (Heirholzer, 1992; Jiang, 2006). A significant limitation in the use of adenovirus as a
reference pathogen is that no dose-response relationship has been published for the ingestion
route of exposure. Experimental studies of adenovirus 4 and 7 with healthy adult volunteers
December 2010 17
-------
U.S. Environmental Protection Agency
indicate that inhalational exposure results in a higher rate of infection at the same dose than
intranasal and oral exposure (Couch et al., 1969). Thus, the inhalation dose-response model
appears to be a conservative estimator for ingestion. Adenoviruses are detected frequently in
sewage, surface water (e.g., Xagoraraki et al., 2007), and surface water affected by stormwater
(e.g., Jiang etal., 2001).
2.3.2.3. Rotavirus
Rotavirus is the leading cause of GI morbidity and mortality among young children and is of
greater public health concern to young children and immunocompromised persons and
populations than the general population. Reinfection of adults is common, but is usually
asymptomatic (Molyneaux, 1995). Dose-response studies indicate that a low dose of rotavirus
(<10 focus forming units) is sufficient to infect a significant proportion of the exposed
population (Haas et al., 1993; Ward et al., 1986). Large numbers of rotavirus (on the order of
1010 organisms/g) can be shed in stool (McNulty, 1978), and rotavirus can survive for weeks on
fomites and in environmental waters (Boone and Gerba, 2007). Although pigs also shed
rotavirus, those strains appear to be host-adapted and not likely to pose a significant risk to
humans (Martella et al., 2010).
2.3.2.4. Cryptosporidium and Giardia spp.
Cryptosporidium and Giardia spp. have been implicated in many U.S. and international
waterborne disease outbreaks. Dose-response models are available for both protozoa, and both
parasites can infect a significant proportion of the exposed population at low doses. The dose-
response characteristics of Cryptosporidium, however, may vary among isolates—C. parvum and
C. hominis are the two species of primary importance in human infections (Messner et al., 2001;
Teunis et al., 2002; USEPA, 2005a). Cryptosporidium and Giardia spp. are frequently isolated
from publicly owned treatment works (POTW) effluent, stormwater, and livestock manure, and
their respective oocysts and cysts can survive for extended periods of time in the environment.
The high environmental loading of potentially human infectious Cryptosporidium in calves
makes Cryptosporidium of particular interest in estimating risk related to livestock sources of
fecal pollution.
2.3.2.5. Campylobacter spp.
Two species of Campylobacter—C. jejuni and C. coli—cause most Campylobacter infections in
humans, with the majority caused by C. jejuni. Several dose-response relationships for C. jejuni
have been published (Medema et al., 1996; Teunis et al., 2005). Campylobacter spp. is prevalent
in livestock, particularly poultry and sheep, has been implicated in outbreaks associated with
consumption of milk, and is present in levels as high as 79,000 Most Probable Number
(MPN)AOO mL in wastewater treatment plant (WWTP) effluent.
December 2010 18
-------
U.S. Environmental Protection Agency
2.3.2.6. E.coliO157:H7
E. coli O157:H7 is representative of Shiga toxin producing E. coli (STEC), possesses the
potential for serious adverse health outcomes, and has been implicated in waterborne outbreaks.
A peer-reviewed dose-response model is available (Teunis et al., 2008b). E. coli O157:H7 is
frequently isolated from cattle manure, often in very high densities, but less often from swine
manure and seldom from poultry manure (Appendix B). E. coli O157:H7 can potentially grow
in soil, sediment, water, and possibly other environmental matrices—all of which emphasizes its
potential to be found in POTW- and livestock-impacted waters.
2.3.2.7. Salmonella
The most heterogeneous of the reference pathogens is Salmonella, whose serotypes have adapted
to a wide variety of host-specific environments. Because Salmonella serotypes also vary widely
in their ability to infect humans, dose-response modeling of Salmonella can be somewhat
complex. Salmonella are of particular interest as reference pathogens because they have many
sources; have been associated with outbreaks (primarily foodborne); occur in abundance in
chicken, cattle, and swine manure; and because some serotypes pose serious human health
hazards (Berg, 2008; O'Reilly et al., 2007). Salmonella can persist in soils for 180 days or
longer (Holley et al., 2006), depending on several factors including soil moisture, presence of
manure, and clay content. Salmonella densities may increase in manures and manure-soil
mixtures (You et al., 2006). In surface waters, Salmonella can be detected throughout the year,
with densities and serotype diversity typically higher during summer months than winter months
(Haley et al., 2009).
2.3.3. Livestock-impacted sites
The reference pathogens in livestock manure are primarily bacterial and protozoan (Appendix
B). Among human viruses of potential concern, only hepatitis E is associated with livestock
operations (Banks et al., 2004; Legrand-Abravanel et al., 2009; Rutjes et al., 2009; Sinclair et al.,
2009; Takahashi et al., 2009). Although the presence of Hepatitis E antibodies in pigs is notable
(Meng et al., 1999; Smith, 2001), including Hepatitis E in QMRA is limited by the lack of dose-
response relationships available to estimate risks to humans. In this regard, experiments with
monkeys indicate that oral inoculation with hepatitis E is inefficient in producing disease. In
addition, in countries with well-developed sewage treatment facilities and practices, the
prevalence of Hepatitis E in environmental waters is relatively low (Smith, 2001). Therefore,
using bacterial and protozoan reference pathogens to evaluate livestock-impacted water with
QMRAs is more appropriate at this stage of our understanding.
Livestock-derived pathogens reach surface water primarily through runoff from land with fresh
or treated manure during and immediately after rainfall events. This mechanism requires
pathogens to be in fecal material when the manure is applied to land (occurrence), present in
sufficient numbers to contaminate runoff, and carried in runoff to receiving water (mobilization).
December 2010 19
-------
U.S. Environmental Protection Agency
Such processes vary between different livestock handling practices and within a particular
livestock manure type. For example, the proportion of animals that are infected by a specific
pathogen (shedding pathogens in their feces) is variable in time and space; the level of
storage/land treatment varies among farms and even between applications on a given farm; and
mobilization of pathogens depends on the rain event (and the antecedent rainfall), the slope of
land where manure is applied, the groundcover and soil characterization of the application site,
among other factors (Ferguson et al., 2007).
Because the pathogens in livestock manure are not necessarily the same species or serotypes that
cause human illness, estimating the proportion of human-infectious strains of each reference
pathogen in each animal source is important. For example, the overlap between Salmonella
serotypes prevalent in humans and livestock can be used to develop a lower bound on the
potential loading of human-infectious Salmonella from livestock (see Chapter 3 for further
information).
Appendix B summarizes the prevalence and abundance of pathogen shedding from animal
sources, including cattle, pigs, chickens, and gulls.
2.3.4. Human-impacted sites
At sites affected by humans, pathogen sources include treated sewage and other human-based
sources such as on-site septic systems and swimmers (Elmir et al., 2007; Loge et al., 2009). A
literature review identified representative concentrations of the reference pathogens in
disinfected secondary sewage effluent (Appendix B). Table 3 provides an overview of that
review.
Although all of the reference pathogens are found at substantial levels in human wastewater,
research indicates that relatively few reference pathogens accounted for the vast majority of
swimming-associated GI illnesses observed in EPA's 2003 to 2004 NEEAR epidemiology
studies (Wade et al., 2006, 2008) conducted at POTW-impacted recreational sites on the Great
Lakes (Seller et al., 2010a). The scenario evaluated in this QMRA does not cover human-
impacted water specifically; however, EPA's literature review and preliminary QMRA actitivies
included human-impacted sites (Schoen and Ashbolt, 2010; Seller et al., 2010a, 201 Ob).
December 2010 20
-------
U.S. Environmental Protection Agency
Table 3. Estimated densities of reference pathogens in disinfected secondary effluent
Reference
Pathogen
Rotavirus
Norovirus
Adenovirus
Cryptosporidium
Giardia
Campylobacter
Salmonella
E. coli O157:H7
Estimated Density in Chlorinated
Secondary Effluent
10 plaue forming units (PFU)/L
1000 qPCR genomes/L
10 virions/L
40 oocysts/L
13 cysts/L
100 MPN/L
100 MPN/L
2.5 stx2 gene carrying bacteria/L
Summary Justification (Citation[s])
Rao et al. (1987)
Lodder and de Roda Husman (2005); Katayama
et al. (2008)
Irving and Smith (1981); He and Jiang (2005);
MWRDGC (2008)
McCuin and Clancy (2006)
Rose et al. (2004); Seller et al. (2007b)
Stampietal. (1993)
Koivunen et al. (2001); Lemarchand and Lebaron
(2003); Jimenez-Cisneros et al. (2001)
Garcia- Aljaro et al. (2004)
2.3.5. Shorebird-impacted sites
Pathogens in shorebird feces are primarily bacterial and to a lesser degree protozoan, including
the reference pathogens. Thus, using bacterial and protozoan reference pathogens is appropriate
for QMRAs evaluating shorebird-impacted recreational waters. Although the scenario evaluated
in this QMRA does not cover shorebird-impacted water, EPA's literature review and preliminary
work included shorebirds (Schoen and Ashbolt, 2010; Seller et al., 2010b). Appendix B
summarizes the prevalence of reference pathogens in shorebird feces.
2.4. Populations Included in the Risk Assessment Model
The QMRA analyses characterize risks of illness from a single recreation event for the general
population. In this case, the recreational "event" is defined to be as consistent as possible with
exposures that occurred during EPA's water epidemiology studies (USEPA, 1986; Wade et al.,
2006, 2008). Those studies reported statistically relevant relationships between FIB and GI
illness in subjects engaging in self-reported body contact recreation. Here, we assume that water
ingestion (consistent with the ingestion rates reported by Dufour et al. [2006]) is conservative
and representative of the body-contact recreation activities that occurred during the EPA's water
epidemiology studies (USEPA, 1986; Wade et al., 2006, 2008).
11
Sub-populations can have variable risks because of differences in water contact times, water
ingestion rates, and susceptibility to infection for some pathogens (Gerba et al., 1996). However,
conducting QMRA for specific sub-populations is not currently feasible given the uncertainty in
the differences between susceptible populations and the general population (Parkin et al., 2003),
1' New or revised RWQC will provide a specified level of public health protection to the population, as defined by the tolerable
or acceptable level of risk, hi the 1986 RWQC, this level of protection was specified not to exceeed 8 cases of HCGI per 1000
recreation events. Thus, RWQC are not designed to provide a specific level of public health protection to an individual during
any specific recreation event.
December 2010
21
-------
U.S. Environmental Protection Agency
and a lack of data on dose-response relationships for specific subpopulations (USEPA, 2010).
However, because this risk assessment uses a stochastic framework, susceptible sub-populations
are accounted for to the extent that reported variations in the values of dose-response model
parameters reflect response variations among different sub-populations.
2.5. Reference Health Outcomes
Water recreation can cause adverse health outcomes including GI illness, respiratory infection
and illness, skin infection and disease, conjunctiva infection and disease, and ear infections and
disease (e.g., "swimmer's ear"). Although swimmers might suffer from any of these outcomes,
epidemiology studies indicate that water quality as measured by FIB is generally predictive of GI
illness12 (Priiss, 1998; Wade et al., 2003; Zmirou et al., 2003) and less frequently respiratory
illness (Fleisher et al., 1996, 2010). Moreover, QMRA-compatible exposure data are strongest
for the ingestion route of exposure and most dose-response relationships are consistent with a GI
infection endpoint. Therefore, to ensure that the QMRA analyses described in this report are as
compatible as possible with the water epidemiology studies, the reference health outcomes in the
QMRA include (1) infection via exposure to reference pathogens through ingesting surface water
during recreation, and (2) GI illness conditional on infection.
Evaluating infection differs by pathogen, based on how the dose-response models defined
infection. Available models mostly defined infection as seroconversion or shedding pathogens
in feces. Similarly, GI illness definitions varied among studies, but were generally related to
diarrhea, and/or vomiting (Colford et al., 2002; Payment et al., 1991, 1997).
2.6. Units of Exposure and Route of Concern
The units of exposure are the number of pathogens ingested per recreation event. The number of
pathogens is estimated based on the volume of water ingested during recreation and the
estimated pathogen densities in the ingested water. Bacteria measurement units are usually MPN
or CPU; for most viruses, PFU, although for norovirus, units are qPCR genome copies. For
protozoa, the units are oocysts or cysts.
The route of concern is ingestion of water during recreational activities. This QMRA does not
include aerosol exposure, ingestion of sediment or soil, or skin, eye, or ear exposures. It is
important to note that EPA's water epidemiology results (USEPA, 1986; Wade et al., 2006,
2008) are based on self-reported body contact recreation, which does not necessarily require
12 Several different definitions of GI illness have been used in water epidemiology studies. For example, the 1986 EPA AWQC
are based on HCGI that was defined as a symptom category including any one of the following unmistakable or combinations of
symptoms: (1) vomiting; (2) diarrhea with fever or a disabling condition (remained home, remained in bed or sought medical
advice because of the symptoms); and (3) stomachache or nausea accompanied by a fever. In the 2003/2004 NEEAR Great
Lakes epidemiology studies, GI illness was defined as any of the following: diarrhea (3 or more loose stools in a 24-hour period),
vomiting, nausea and stomachache, and nausea or stomachache that affects regular activity (inability to perform regular daily
activities). This recent definition of GI illness occurs more frequently as it excludes the requirement of fever. It is also consistent
with GI illness definitions used in other recent epidemiology studies (e.g., Colford et al., 2002; Payment et al., 1991,1997).
December 2010 22
-------
U.S. Environmental Protection Agency
water ingestion. For this QMRA, we assume that the observed water contact recreation-
associated illnesses observed during the water epidemiology studies were a result of water
ingestion. Based on the results of a recent reverse QMRA reported by Seller et al. (2010a),
water ingestion during recreation is a reasonable approximation of the epidemiology study
exposure metric.
2.7. Target Ri sk Level
The target risk level defined by the 1986 AWQC for fresh water was 8 cases of HCGI per 1000
exposures (swimming events), which was established based on epidemiology and water quality
studies conducted by EPA (USEPA, 1986). However, the more recent water epidemiology
studies (Wade et al., 2006, 2008) use GI illness rather than HCGI as the target health outcome
(see Footnote 12 above). For this QMRA, we use an estimated equivalent risk target based on
GI illness to the 8 cases of HCGI per 1000 exposures that provides a similar overall level of
public health protection. Based a preliminary review of the available epidemiology information,
a target risk level of 30 GI illnesses per 1000 exposures is used as a preliminary equivalent
benchmark. This estimate takes into account the more frequent occurrence of GI illness
compared to HCGI.
2.8. Scenarios Modeled
We selected the QMRA scenario to evaluate illnesses resulting from recreation at a freshwater
beach impacted by agricultural animal sources of fecal contamination. As part of this effort,
several assumptions were made to limit the scope of the scenario and to ensure that the QMRA
results would be conservative, including the following:
• Exposure is via primary contact recreation (e.g., swimming) at a freshwater beach.
• Water ingestion is the predominant route of exposure during primary contact recreation.
The scenario does not evaluate the risk of ingesting water from activities such as wading,
boating, or fishing; however, we expect those risks to be less than those associated with
primary contact recreation.
• Fresh cattle manure, pig slurry, and poultry litter are applied at agronomic rates (the
highest rate at which manure should be applied in the United States) to land adjacent to
the freshwater beach to minimize the uncertainty and variability associated with
environmental fate and transport of FIB and reference pathogens.
• The cattle manure, pig slurry, and poultry litter contain FIB and reference pathogens
consistent with levels reported in the peer-reviewed literature.
• The fresh, solid, untreated fecal contamination from cattle, pigs, and chickens reaches the
freshwater beach via runoff from an intense rainfall event and undergoes minimal
dilution in receiving water. An intense rainfall event produces a higher load of pathogens
December 2010 23
-------
U.S. Environmental Protection Agency
and FIB than a less intense precipitation event. This results in a conservative risk
estimate via forward QMRA.
• The FIB and reference pathogens are present in cattle manure, pig slurry, and poultry
litter runoff at levels consistent with the observed mobilization of FIB and reference
pathogens from the EPA environmental monitoring studies (see Section 2.12 and
Appendix D for further information).
• Assuming that the mobilized pathogens and FIB are applied adjacent to receiving waters
presents the highest exposure and produces a more conservative risk estimate relative to
conditions with best management practices.
2.9. Questions to be Addressed
The QMRA is designed to address the following primary and secondary questions:
1. What is the risk of illness associated with recreation at a freshwater beach impacted by
agricultural animal (cattle, swine, or chicken) sources of fecal contamination during or
immediately after a rain event?
2. How do those risks compare to risks associated with freshwater beaches impacted by
human sources of fecal contamination (effluent from a POTW)?
The forward QMRA estimates the risk of illness associated with recreation (swimming) at a
beach impacted by agricultural animal sources of fecal contamination. Numerical simulations
are used in which the pathogens in land-applied manure are selected from ranges derived from
the literature; pathogen mobilization (from the manure) proportions are based on observed
mobilization rates for each pathogen and manure type; and the swimmers are assumed to be
exposed to untreated and undiluted runoff.
The relative QMRA compares risks from recreation in the animal-impacted water to those
associated with human-impacted water. To achieve this, we extended previous related work that
evaluated the estimated human health risks from exposure to recreational water impacted directly
by fecal contamination from human and non-human sources (Schoen and Ashbolt, 2010; Seller
et al., 2010b). This relative QMRA extends that evaluation to include land application of manure
that contains FIB and reference pathogens and their mobilization (proportion of FIB and
reference pathogens that run-off) during and immediately after rainfall events based on the
results of EPA environmental monitoring studies.
2.10. Conceptual Models
2.10.1. Top-tier models
Figure 7 and Figure 8 illustrate the top-tier conceptual models for the QMRA. Forward QMRA
(Figure 7) is used to answer question 1 (above) while the relative QMRA approach (Figure 8) is
used to address question 2. A previously conducted reverse QMRA (Figure 9) provides context
December 2010 24
-------
U.S. Environmental Protection Agency
about human-impacted water and justification for the use of specific reference pathogens for
human-impacted waters (Seller et al., 2010a).
1 Fecal pollution source
characterization (mixture
of pathogens)
Disease rate for
reference pathogens
(mixture of pathogens)
Volume of water
ingested
Pathogen dose-
responses
model parameters
• Infection models
• Illness models
Forward QMRA
Agronomic rate land application
Mobilization during runoff:
specific to each fecal pollution
source
Pathogen densities in recreational
waters impacted by a specific
fecal pollution source
Dose of
pathogen
t ingested ^
>
f
Dose-response
relationship:
L Infection or Illness J
^
-**
.'
Illnesses attributable to
individual pathogens
Total Gl illness rate per 1000
recreation events
Figure 7. Forward QMRA conceptual model
In Figure 7, the input data characterize pathogens present in fecal pollution source(s), the fraction
of human-infectious pathogenic strains in each fecal source of interest, the prevalence of
infection in the non-human source (proportion of animals shedding the pathogen), ingested
volumes, dose-response models and parameters, and pathogen mobilization. Output of the
forward QMRA model is the probability of infection and illness associated with exposure to
water during recreation.
December 2010
25
-------
U.S. Environmental Protection Agency
Density of each reference
pathogen and indicator in land
applied manure
Mobilization during
runoff event
s~~ ~"^
Proportion ofanimals
shedding reference
pathogen
Density of each reference
pathogen in runoff from plots
with land-applied manures
.j
1
1
Indicator densities in
runoff from plots with
land-applied manures
h~
>
Density of F IB in
water: existing
AWQC levels for
ENT and EC
Density of human
infectious reference
pathogens in recreational
water that has FIB present
atAWQC levels
Volume of water
ingested
f" ~^\.
Pathogen dose-
response
model parameters
• Infection models
• Illness models
I
Dose of pathogen
ingested
^
r
Dose-response
relationship:
Infection or Illness
I
Fraction of human
infectious reference
pathogen strains
from each source
Comparative
Risk QMRA
Illnesses attributable to
individual pathogens
Total G 1 illness rate from all
reference pathogens (per 1000
recreation events)
Figure 8. Relative QMRA conceptual model
In Figure 8, input data are somewhat different than those used in the forward QMRA. For the
relative QMRA, we assume a specific level of FIB is in the waterbody—in this case, the current
RWQC levels for enterococci or E. coli. These FIB levels are used in conjunction with the FIB
and reference pathogen levels in the land-applied material, the fraction of human-infectious
pathogenic strains in each fecal source of interest, the prevalence of infection in the non-human
source (proportion ofanimals shedding the pathogen), the proportions of FIB and reference
pathogens that mobilize during a rain event, and the volume of water ingested (Schoen and
Ashbolt, 2010; Seller et al., 2010b). The relative QMRA model output is the probability of
infection and illness associated with exposure to water during recreation for each source of
interest referenced to the chosen level of FIB.
December 2010
26
-------
U.S. Environmental Protection Agency
Fecal pollution source
characterization (mixture
ofpathogens)
Disease rate for
reference pathogens
(mixture of pathogens)
Dilution and attenuation
Variability in
exposure (ingested
volume)
• Children
• Adults
Pathogen densities in recreational
waters impacted by a specific
fecal pollution source
Pathogen dose-
responses
model parameters
• Infection models
• Illness models
Ingested
volume
Net close-response
• Infection
• Illness
Indicator density
at recreation site
Epidemiologkal
relationship
between
indicator density
and (31 illness
\ X
ft
Secondary
transmission
Reverse QMRA
Number of attributable 61
illnesses per 1000
Illnesses attributable to
individual pathogens
Figure 9. Reverse QMRA conceptual model
13
In Figure 9, input data characterize the pathogens present in fecal pollution source(s) (either
based on the relative abundance ofpathogens in the fecal pollution source or on the observed
health effects for each reference pathogen); ingested volumes; dose-response models and
parameters; and observed illness rate (i.e., number of illnesses per day per 1000 swimmers). The
output of the reverse QMRA model is an estimate of pathogen densities at a recreation site with a
known fecal pollution source (Seller et al., 2010a).
2.10.2. Sub-tier model: model parameter form and estimation
In this report, we modeled parameter uncertainty and variability explicitly by treating each
parameter as a random variable. In cases with insufficient data to justify a specific statistical
13 Again, no reverse QMRA was conducted as part of this study. The discussion of reverse QMRA and Figure 9 are provided
because EPA previously conducted reverse QMRA, and in this study it provides context about the relative importance of the
reference pathogens in human-impacted recreational waters.
December 2010
27
-------
U.S. Environmental Protection Agency
distribution, we used point estimates. This approach is consistent with previous QMRAs
(Eisenberg et al., 1996; 1998; Seller et al., 2010b). The stochastic parameters in the model
include the following:
• reference pathogen density (abundance) in animal manure;
• FIB density in animal manure;
• prevalence of reference pathogen shedding in the animal-source;
• proportion of FIB and reference pathogens that mobilize during a rainfall event;
• volume of water ingested during recreation;
• dose-response parameters (to the extent that peer-reviewed literature supports it); and
• morbidity fraction (proportion of infections that result in illness).
Abundance, prevalence, and mobilization of all pathogens differ with manure type. Different
distributions are used for each pathogen-manure type combination.
2.10.3. Sub-tier model: animal-impacted water pathogen-loading model
FIB and reference pathogen loading to a recreational waterbody can occur through direct or
indirect contamination (Figure 10). Direct contamination occurs when fresh undiluted fecal
material is deposited into a waterbody. Indirect contamination occurs during transport from
adjacent land into a waterbody via rainfall runoff. Seller and colleagues (201 Ob) reported the
risks from direct fecal contamination from agricultural animals into a recreational waterbody.
Those results indicated that the GI illness risks associated with exposure to recreational
waterbody directly impacted by fresh cattle feces might not differ substantially from water
impacted by human sources; however, the risks associated with exposure to recreational water
directly impacted by gull, chicken, and pig wastes appear to be lower than those impacted by
human sources (Seller et al., 2010b). The QMRA described in this report extends that work by
considering indirect contamination (described below). These two routes represent reasonable
conservative stream loading scenarios for livestock fecal pollution.
December 2010 28
-------
U.S. Environmental Protection Agency
Manure handling
Pathogens • Indicators
Source
•i i
+j I
G
C
5
S
c
8
Si
•5
•
Land application
Pathogens • Indicators
t
I
Inactivation
Pathogens • Indicators
Best management practices
• Pathogens • Indicators
8
'
Q
Dilution + mixing
Receptor
Figure 10. Transport of pathogens and indicators to swimmers from livestock manure
An important distinction between direct and indirect contamination is that the source material of
concern for direct contamination is feces from an individual or individuals; for indirect
contamination the source material of interest is effectively a composite sample of fecal material.
The data used to characterize the abundance of FIB and reference pathogens for direct (Schoen
and Ashbolt, 2010; Seller et al., 20lOb) and indirect (present QMRA) contamination reflect this
distinction. Using data from individual fecal samples to characterize abundance in the previous
work corresponded to an exposure that was assumed to be in close proximity to manure
deposited directly into recreational water. In this QMRA, data from an intensive environmental
sampling program conducted by EPA (see Section 2.12) were used to characterize the
mobilization of FIB and reference pathogens due to rainfall and subsequent runoff.
Pathogen loading from agricultural animal sources depends on the prevalence of animals infected
by reference pathogens, the abundance of reference pathogens in fresh manure, manure handling
practices (particularly storage time and timing of application), time between application and
rainfall, and the path by which pathogens reach receiving water. During transport to receiving
water, pathogens may be inactivated or removed in buffer strips or other physical barriers.
December 2010
29
-------
U.S. Environmental Protection Agency
Figure 11 provides a conceptual model showing how indirect FIB and reference pathogen
loading can occur for agricultural animal sources: (1) they may be deposited directly on fields
during grazing and be mobilized during rainfall and transported, (2) they may be in treated or
untreated manure that is spread on fields and mobilized or transported during a rain event, (3)
they may run-off from feedlot pens, or (4) they may escape storage due to an extreme rainfall
event or mishap.
EPA conducted a preliminary literature review to evaluate the important factors associated with
animal-impacted waterbodies (USEPA, 2009a), developed preliminary exposure models for the
pathways described above, and conducted exploratory analyses to determine which model
parameters most strongly affected QMRA output (Annex 2). Salient findings from the exposure
modeling and preliminary QMRA work included the following:
• Collecting and storing fecal material on site can be an effective barrier to pathogen
mobilization. Depending on storage time, land application may cause short-term
pathogen risk spikes immediately following application. These spikes can be roughly
equivalent to the risk associated with open grazing operations.
• Provided sufficient time is provided, storage can effectively reduce pathogens.
• Managing land application to avoid periods of high rainfall reduces risk.
• Prevalence of a herd's infection changes over time.
• Understanding the prevalence (and dynamics) of human infectious pathogenic strains in
animal-impacted water is important to estimate its risk.
• Pathogen densities in manure are highly uncertain.
• Pathogen super-shedders have the potential to drive the risk during a rainfall event.
• Environmental inactivation rates of pathogens are highly uncertain. Therefore, reducing
pathogens through uncontrolled environmental processes is not feasible unless extended
residence times are guaranteed.
Based on the goals of the QMRA and the available data, the microbial loading model used in this
QMRA addresses land application of manure and subsequent mobilization of FIB and reference
pathogens at the beach that is adjacent to the animal-impacted runoff.
December 2010 30
-------
U.S. Environmental Protection Agency
Potential pathways for faecal material to be mobilised to local waterways
Open Grazing-Direct deposition of faecal material
Some portion of
overland flow may
not be intercepted
by the pond and may
flow directly to stream
Pens - animals not fully housed, faecal material
on ground and may be mobilised by rainfall
Shed - animal; fully housed, Spillaf
faecal material collected
overland transport
recreational
wimming site
Figure 11. Conceptual model of paths for livestock pathogens reaching recreation sites
2.10.4. Sub-tier model: reference pathogen dose-response models
The QMRA dose-response models for reference pathogens come from peer-reviewed studies
(see Section 3.2.2 for further information). A brief overview of dose-response modeling is
presented below, while the EPA MRA Tools document (USEPA, 2010) provides a more
comprehensive review of this topic.
The infection process requires that a person ingests pathogens, at least one pathogen initiates an
infection, and a proportion of infections proceed to illness. All three of these processes can be
described with probability distributions.
When the probability of ingesting a dose of pathogens is Poisson-distributed and all of the
ingested pathogens have an equal probability of initiating infection, the exponential dose-
response model is appropriate:
December 2010
31
-------
U.S. Environmental Protection Agency
[1]
where Anfect is the individual probability of infection, dis dose (number of pathogens), and r is a
parameter of the distribution equal to the probability that an individual pathogen initiates
infection.
When the probability of ingesting pathogens is Poisson-distributed and the probability that
individual pathogens initiate infection is beta-distributed, the beta-Poisson model is appropriate:
[2]
where a and /? are parameters of the Beta distribution and \F\ denotes a confluent
hypergeometric function. A commonly used approximation to the beta-Poisson may be used
when p » 1 and (3 » a. This approximation is:
[3]
When pathogens are aggregated (are no longer Poisson-distributed), a Poisson-stopped log-
normal distribution can describe the distribution of doses in an inoculum. When dose is assumed
to follow a Poisson-stopped log-normal distribution, and the ability of individual ingested
pathogens to initiate infection is beta-distributed, the resulting dose-response model is:
^^ (d; a, /?, a) = l-2Fi («, teH a + A-te)) [4]
where the parameter a is related to the degree of aggregation in the pathogen dose, a and (3 are
parameters of the Beta distribution, and jF\ denotes a hypergeometric function. Note that
equation 4 reduces to the exact beta-Poisson dose-response model as a — » 0.
Published studies have used empirical dose-response models (which cannot be derived using
assumed distributions for exposure and infection initiation) based on fitting those models to data
or that are based on those models mimicking observed patterns of infections among humans.
Among these empirical models, the Gompertz-log model (equation 5) describes response
(illness) of humans to doses of Salmonella of numerous serotypes:
l-e-~ [5]
where a and b are parameters of the distribution that take on different values for different
Salmonella serotypes, and/Wss denotes the individual probability of illness.
Two types of models describe the progression of illness to infection — a constant rate model and
a dose-dependent model. The constant rate model, which is the most common in published
QMRA studies, assumes a fixed proportion of individuals infected by a given pathogen
December 2010 32
-------
U.S. Environmental Protection Agency
progresses to illness. The proportion progressing to illness may be estimated based on results of
feeding studies or epidemiology studies. The dose-dependent model assumes the conditional
probability of illness given infection is given by:
/'(illness infection;J, 77, /r) = 1 - (l + 77f(d))
[6]
where 77 and /rare parameters of the distribution andf(d) is a function of dose. Studies have
explored functions of dose for progression of infection to illness and have identified data sets in
which progression was independent of dose, was dependent on the inverse of dose, and was
dependent on dose.
2.10.5. Sub-tier model: volume of water ingested during recreational activities
Results reported by Dufour et al. (2006) characterize the volume of water ingested during
recreational activities for swimming episodes of 45 minutes duration. The data can be fit to log-
normal distributions for children and adults combined (Figure 12) or individually (Figure 13)
(Seller et al., 2007b). In this QMRA, we use the distribution for children and adults combined to
represent water ingestion during recreational activities for the general population.
100 -
0)
f
0>
2 1
o
0.1
• Quantifiable volumes
o Volumes below detectable limits
-e-
12 5 10 20 30 40 50 60 70 80 90 95 98 99
Percent of Observations Less Than Correponding Value
Figure 12. Ingested volumes for the combined data (children and adults)
December 2010
33
-------
U.S. Environmental Protection Agency
100 -.
10 7
55
-------
U.S. Environmental Protection Agency
2.11. 1. Risk of illness associated with recreation at a beach impacted by agricultural
animal sources of fecal contamination
We use a traditional forward QMRA approach to characterize the risk of illness associated with
recreation at a freshwater beach. The methodology for the QMRA analyses is a Monte Carlo-
based approach with model parameters characterized as statistical distributions, whenever
possible. Separate Monte Carlo analyses were conducted for each agricultural animal source.
The analysis begins with literature review-based data (Appendix B) to characterize the densities
and prevalences of reference pathogens in fecal waste for each animal source (solid fresh cattle
feces, liquid fresh swine feces, and fresh poultry litter). We assume that these materials are
applied to land at agronomic rates14 and are mobilized during a 100-year return period storm
(referred to hereafter as intense rain event) for the Piedmont region of Georgia. 15 The results
from the EPA Office of Research and Development's environmental monitoring program
(Section 2.12 and Appendix D) characterize the proportion of the land-applied pathogens that
run-off following this type of rain event. Specifically, the density of FIB and reference
pathogens in water running off is proportional to the number of land-applied organisms with a
variable proportionality constant for different manure types and conditions (e.g., plot slope,
antecedent soil moisture). Mathematically,
where
NJ is density of organism /' in runoff water (organisms/volume);
VRQ is net runoff during the event (volume);
fitRO is the proportion of organisms mobilized during the entire event;16
^ is the density of organism / in the land-applied manure; and
^-manure is the mass of manure applied to the plot generating runoff volume VRO.
Thus, for each of the animal sources, the density of reference pathogens in the runoff is
calculated as the product of the reference pathogen density in land applied fecal waste, the
14 Manures were applied at agronomic rates based on measured nutrient concentrations in samples of the land-applied manure.
These application rates are specific to the manures and ground cover used in the study and may differ from other manures whose
nitrogen densities are different or for other plots where the nutrient requirements are different, either because of residual nutrients
in the soil or because the ground cover has a different nutrient uptake. Assuming pathogens and indicator organisms are well-
mixed in manures, the pathogen and indicator loads scale linearly with manure application rate and other sites may have manure
indicator and pathogen loads signficantly different from those in the current study.
15 In the literature, mobilization is often assumed to be a function of runoff, not rainfall. In this study, the rainfall applied to the
plots was fairly uniform and based on a 100-year return period storm event. The runoff was variable between plots and was a
function of location of the plot, slope, soil characteristics, etc.
16 The termfiRO is a random variable with range based on mobilization fractions in the EPA plot-scale experiments. Because the
EPA experiments used a single rainfall intensity and rate, the dependence offiRO on event characteristics is unknown.
December 2010 35
-------
U.S. Environmental Protection Agency
prevalence of infection (percent of animals shedding), the human infectious potential of the
pathogen, and the proportion of the applied reference pathogens that run-off following a rain
event divided by volume of the runoff for the event. In this exposure scenario, recreation is
assumed to occur at the edge of the recreational waterbody, where the runoff enters the
waterbody. Therefore, the dose of pathogens for this exposure scenario is the product of the
volume of water ingested during recreational activities and the density of each pathogen in the
runoff. That dose is input to the appropriate dose-response relationship resulting in a probability
of infection. The probability of infection is multiplied by a morbidity factor to produce a
probability of illness. This scenario is intentionally conservative (i.e., developed to produce
health-protective estimates of risk), including an intense runoff event, no attenuation of
pathogens between runoff and entry into receiving water, and ingestion of undiluted runoff.
The risk associated with each fecal contamination source is characterized as the total probability
of GI illness, Pfn , using the probability of illness from each source-specific pathogen in a
manner that is parallel to computing annual risks of infection by combining daily risks (Regli et
al., 1991):
rp
This process is repeated 10,000 times for each fecal contamination source to generate a
distribution of risk.
2.11.2. Comparison of animal-impacted water risks with POTW-impacted water
The second analysis, which uses the relative QMRA approach, provides a relative comparison of
the estimated risks from recreation in water impacted by agricultural sources of fecal
contamination to those associated with human-impacted water. Previously developed methods
(Schoen and Ashbolt, 2010; Seller et al., 2010b) form the basis for this analysis, but were
extended by including land-application of FIB and reference pathogens, and mobilization
(proportion of FIB and reference pathogens that run-off) during rainfall events based on the
results of the EPA environmental monitoring studies (Section 2.12).
In this analysis, the estimated risks are calculated for a hypothetical waterbody that contains
geometric mean FIB densities at the current (USEPA, 1986) RWQC for freshwater (33 CFU
enterococci/100 mL and 126 CFUE1. coli /100 mL, respectively). We conduct separate
calculations for each source of fecal contamination (cattle, pigs, and chicken). The current
RWQC were established to provide a level of health protection equivalent to approximately 8
cases of HCGI per 1000 recreation events for water impacted by treated effluent. As noted
previously, recent recreational water epidemiology studies use a definition of GI illness that
excludes fever as a required symptom (Colford et al., 2007; Wade et al., 2006, 2008). This more
recent health metric occurs more frequently than GI illness. In this QMRA, we use an
benchmark risk of 30 cases of GI illness per 1000 recreation events as an estimate of the
December 2010 36
-------
U.S. Environmental Protection Agency
equivalent level of GI illness to the currently acceptable level of 8 cases of HCGI per 1000
recreation events. This estimate is based on a preliminary evaluation of the data reported by
Wade et al. (2006).
Reference pathogen doses are derived from the density of the FIB from each source (Schoen and
Ashbolt, 2010; Seller et al., 2010b). Pathogen dose is calculated based on independent Monte
Carlo samples from the observed ranges of pathogen and FIB densities in fecal waste and the
proportion of these organisms that mobilize during a rain event. Note that this sampling scheme
does not require a specific relationship between the FIB and pathogen in the fecal waste or in the
receiving water. The dose of each reference pathogen from each source is calculated as follows:
"rp T,,"? ., ~~ rp rrp rp
where
S is the fecal contamination source;
CFIB is the waterbody density of enterococci or E. coli (CFU/100 mL);
nS
FIB is the density of FIB in runoff from plots with land-applied manure (CFU/100
mL) or in sewage (CFU/L);
Rs
rp is the density of pathogen species in runoff from plots with land-applied manure
number of pathogens or genomes 100 mL"1) or in sewage (number of pathogens
or genomes L"1);
rp is the fraction of human-infectious pathogenic strains from source S;
rp is the prevalence of infection in the non-human source 17 (proportion of animals
shedding the pathogen); and
V is the volume of water ingested (mL).
This relation is similar to previously cited methods (Schoen and Ashbolt, 2010; Seller et al.,
201 Ob), except here, the densities of reference pathogens and FIB in water ingested during
recreation are based on the density of the organisms in the land-applied manure and the
proportion of the organisms running off during rain events.18
17 For previous work conducted on human sources, Ir was assumed to be 1.0 because the FIB and pathogen data are from
sewage not individual fecal samples, and therefore already accounts for the pathogen prevalence.
18 In these analyses, we assume that the FIB and reference pathogens derive from the source being evaluated. In reality, there can
be numerous sources of FIB in a waterbody, including sources that do not contribute pathogens. The relative QMRA analyses
developed in this report are conservative for waterbodies that also contain non-pathogenic sources—non-pathogenic sources
would cause FIB levels to be relatively higher compared to pathogen levels.
December 2010 37
-------
U.S. Environmental Protection Agency
Similar to the forward QMRA described in the previous section, doses are input to the
appropriate dose-response relationship resulting in a probability of infection. The probability of
illness is conditional on infection and is calculated via a morbidity fraction for each reference
pathogen. The total probability of illness for each fecal contamination source is computed as
described above. This process is repeated 10,000 times for each source to generate a distribution
of risk. Those distributions of risk are then compared to the benchmark risks for human-
impacted water described above.
2.12. Environmental Sampling
The objective for EPA's environmental monitoring and sampling study was to generate primary
data to characterize zoonotic enteric pathogens and FIB densities in surface water affected by
agricultural activities. In addition, the study emphasized overland transport inputs and processes.
The monitoring design included rain simulation experiments in small plots amended with solid
cattle manure, swine slurry from a lagoon, and un-composited litter from a chicken operation.
These matrices were selected because higher pathogen densities are associated with fresh fecal
material, and pathogen removal efficiencies vary both between and within treatment processes
(Bicudo and Goyal, 2003; Goss and Richards, 2008; Heinonen-Tanski et al., 2006; Larney and
Hao, 2007; Letourneau et al., 2010; Martens and Bohm, 2009; Peu et al., 2006; Topp et al., 2009;
Vanotti et al., 2005; Vinneras, 2007; Wong and Selvam, 2009; Ziemer et al., 2010). Using fresh
fecal material promoted conservative modeling (because of the assumption that relatively high
pathogen densities are present in land-applied material) and robust results (because the model
does not rely on assumptions regarding the degree of removal during treatment prior to land
application).
The rainfall simulation experiments were designed to
1. estimate pathogen and FIB mobilization rates from manure-impacted plots to surface
water, and
2. provide data to characterize pathogen and FIB densities in overland runoff.
The monitoring study used rainfall simulators instead of natural rainfall events to enhance
reproducibility, allow greater control over important independent variables, and better
characterize mobilization and loading rates of pathogens and FIB through greater sample size.
Previous experiments targeting microbial transport from land-applied manure focused on the
behavior of FIB from cattle manure and produced limited pathogen information (Collins et al.,
2004; 2005; Guber et al., 2007a; Muirhead et al., 2006; Sinton et al., 2007).
The rainfall simulation experiments were held at 36 plots (0.75 x 2 m ) on U.S. Department of
Agriculture (USDA)-owned land in Oconee County, Georgia (33° 47'N, 83°23'W) (Butler et al.,
2008). The experimental plots were delineated with galvanized sheet metal (23 cm width)
placed into the ground to a depth of 18 cm. Consistent with previous work, Tlaloc 3000 rainfall
December 2010 38
-------
U.S. Environmental Protection Agency
simulators (Joern's Inc., West Lafayette, IN) were used (Soupir, 2003; Soupir et al., 2006).
Simulated rainfall was applied to the plots at a set rate intended to simulate both an intense rain
event for the Georgia Piedmont region and to be sufficient to produce surface runoff. Runoff
was collected at the lower end of each plot. Baseline simulations were conducted to determine
background pathogen, FIB, and nutrient levels. Histograms were used to identify frequency
distributions of baseline runoff volumes and to select plots within a specific range of volumes.
Manure applications followed a randomized split plot design. The treatments consisted of
manure applications from three animal types—pigs (liquid manure), beef cattle (solid manure),
and broiler poultry (litter)—and a control treatment (no manure application). Each treatment had
three replicates (plots) and three manure applications timed relative to rainfall simulations. The
timing for rainfall application was one hour, one week, and two weeks after manure application.
The type (mixed fescue/Bermuda crop) and height (10 cm) of the vegetation cover was the same
for all plots. Each type of manure was analyzed for pathogen and FIB loading prior to
application via randomized composite samples. During each rainfall simulation run, samples
were collected every five minutes for the duration of the event to account for the cumulative
runoff volume. Six runoff samples from selected intervals (5, 10, 20, 30, 60 minutes, and total
composited) were analyzed for both E. coli and enterococci total densities. Samples were split
into separate containers for non-microbial analyses, including total suspended solids, dissolved
organic carbon, and nutrients. Two composited samples (10 L) were collected per run for
pathogen analysis (30-minute composite and total composite). Samples were analyzed for
Cryptosporidium, Giardia, Salmonella, E. coli O157, and Campylobacter, depending on the type
of manure applied. Manure from the various sources was applied at agronomic rates following
USDA Natural Resources Conservation Services (NRCS) guidelines (Midwest Plan Service,
2004), and based on the nitrogen requirement of the type of crop and the nutrient concentration
in the manure being applied.
In the plots where rainfall was not applied immediately after manure application (1-week and 2-
week treatments), plastic covers were placed on the plots to protect them from natural rain
events. These covers were placed well above the vegetation cover to allow air circulation and
heat exchange. The type of plastic selected allowed for 75 to 80% of the UV light to penetrate.
This experiment was conducted three times, (October 2009, March 2010, and June 2010) to
obtain sufficient data points and to account for varying climatic conditions.
During the first simulation, it was determined that the levels of pathogens of interest in the
applied material were too low to detect in the runoff. Therefore, for the second and third rainfall
simulation runs, the manure was seeded with surrogate pathogens to determine the mobilization
rates of pathogens. Surrogate pathogens were all non-virulent species that did not pose an
environmental risk or of infection to project personnel. The following surrogate pathogens were
used: (1) E. coli O157:H7 B6914 #87, which was added to cattle feces, and swine slurries, and
poultry litter (the latter only during the March simulation); (2) UV-inactivated Giardia and
Cryptosporidium, which were added to cattle manure and swine slurries; and (3) Salmonella
December 2010 39
-------
U.S. Environmental Protection Agency
X3985, which was added to cattle manure, swine slurries, and poultry litter. Preliminary
calculations determined the concentration of surrogate pathogens needed for each type of manure
to increase the likelihood of detection in runoff. The final surrogate pathogen concentrations
considered recovery methodology, decay of the organisms in manure and during transport,
literature for leaching rates for pathogens or FIB from livestock manure, maximum pathogen
levels observed in livestock manure, and maximum number of organisms that could be produced
to use for spiking. Mixtures containing the different combination of surrogate pathogens were
then prepared and seeded in the appropriate type of manure. Manure seeding was conducted in
the laboratory the same day the manure was applied to the plots and transported to the field on
ice. Analysis of both surrogate pathogens and wild-type pathogens was conducted in all manures
and runoff as described above.
2.13. Tools Used in the QMRA
The software used to implement the forward QMRA is MathCad (Mathsoft Corp.). A previously
developed MathCad worksheet, the Microbial Risk Assessment Interface Tool (MRAIT) was
used as a base package and modified to accept appropriate input (Seller et al., 2007a). The R
programming language (R Development Core Team, 2009) and the Python programming
language (Python Software Foundation, 2009) were used for the relative QMRA analyses.
Previous code in R (Schoen and Ashbolt, 2010; Seller et al., 2010b) was adapted to account for
FIB and reference pathogen mobilization during rain events. The Analytica™ computational
environment (Lumina Decision Systems) was used to develop initial QMRA models for
livestock-impacted sites.
2.14. Summary of Assumptions
Assumptions underlying the QMRA include the following:
• GI illness is the health outcome of primary concern in this QMRA. Infection from the
reference pathogens and subsequent illness result in GI illness. Based on epidemiological
investigations, skin infection and disease, conjunctiva infection and disease, and ear
infections and disease are assumed not to be correlated with FIB (Priiss, 1998; Wade et
al., 2003; Zmirou et al., 2003). Although FIB might predict respiratory infection and
illness (Fleisher et al., 1996), GI illness occurs more frequently; therefore, GI illness rates
predicted by the QMRA are assumed to be conservative and protective for respiratory
illness.
• Water ingestion during recreational activities is the exposure route of interest. Other
routes of exposure, such as inhalation and dermal contact, do not substantially add to the
risk associated with ingestion.
December 2010 40
-------
U.S. Environmental Protection Agency
• The risk of illness at a freshwater beach impacted by agricultural animal sources of fecal
contamination is adequately characterized by the risk associated with the bacterial and
parasitic protozoan reference pathogens.
• Loss of pathogen virulence due to passage through non-human hosts or exposure to a
non-enteric environment can be characterized as "high," "medium," and "low" in the
QMRA based on the relative occurrence of species that infect humans and strains and
serotypes present in typical livestock wastes.
• Human dose-response models adequately predict infection or illness risks for reference
pathogens, regardless of the source (though variability in host-pathogen system response
can be included in dose-response modeling).
• Data collected from the EPA environmental monitoring program in conjunction with data
from the peer-reviewed literature can be used to estimate microbial water quality at a
freshwater beach impacted by agricultural animal sources of fecal contamination.
• Use of a static, individual-level QMRA model is reasonable—secondary transmission
and immunity do not substantially modify risks.
• For the general population, body contact recreation (as self-reported in water
epidemiology studies) involves water ingestion volumes consistent with the recreational
activities reported by Dufour et al. (2006).
• Removal or die-off of reference pathogens and FIB after mobilization from fields and
prior to ingestion by swimmers is limited.
• Recreation at the assumed point of exposure produces a conservative estimate of risk and
is protective compared to other potential exposure points (downstream, diluted, or
contamination scenarios that are older).
• The mobilization fractions observed during the EPA simulated rain events are
representative of the highest mobilization fractions realized during actual rain events.
2.15. Sources of Variability and Uncertainty
One particularly attractive attribute of QMRA is its ability to account for both variability and
uncertainty. In this QMRA, we use a probabilistic framework and characterize each model
parameter using a statistical distribution19 where the parameters of those distributions account for
variability and/or uncertainty. Although it is desirable to treat variability separately from
uncertainty in QMRAs (USEPA, 2006), the available data were insufficient to do this for this
risk assessment.
19 In cases where data are sparse, we use a uniform distribution and specify lower and upper feasible bounds. If those bounds
span more than two orders of magnitude, we use a log-uniform distribution.
December 2010 41
-------
U.S. Environmental Protection Agency
2.15.1. Variability
Substantial variability is anticipated in both the QMRA exposure assessment and health effects
components. We account for the known variability to the extent possible and reasonable. Table
4 summarizes the variable parameters and the underlying causes of the variation.
Table 4. Variable parameters and underlying causes for their variations
QMRA
Component
Variable Parameter
Causes
FIB and pathogen density
Exposure
Assessment
Temporal and spatial heterogeneity
Sporadic loading (epidemics, super-shedders)
Rainfall and runoff (intensity, depth, antecedent conditions)
Waves and currents
Solar radiation
Tides
Season
Ingested volume
• Exposure duration
• Age
Gender
Dose-response
Health Effects
Differences in immune system competency
Prior exposure
Vaccination
Age
Other heterogeneity in host response
Intra-species, intra-strain, intra-serotype, and intra-isolate
heterogeneity in pathogen virulence
Health end-point measured
Secondary transmission
and immunity
• Population-level immune status and background infection
rate
• Heterogeneous contact patterns
• Age
The most significant variability in the exposure assessment is due to temporal and spatial
heterogeneity in reference pathogen and FIB densities. FIB and pathogen densities change by
orders of magnitude over short time periods (Boehm et al., 2002, 2007; Curriero et al., 2001).
Detection methods also impart variability—selective media and injured cells are important issues
underlying heterogeneity. In addition to variability during nominal conditions, non-standard
events can cause extreme variability. Such events include super-shedding of pathogens,
combined sewer overflows, and extreme rainfall/runoff. When extreme variability exists in
pathogen or FIB densities, traditional statistical fitting to distributions such as the log-normal
distribution may be inappropriate (Petterson et al., 2007, 2009; Pouillot et al., 2004; Signer et al.,
2007).
December 2010
42
-------
U.S. Environmental Protection Agency
In health effects modeling, variation among pathogens occurs in their ability to infect humans
and variation among humans occurs in their susceptibility to infection. The choice of dose-
response model can account for these variations to some extent; for example, the beta-Poisson
model accounts for heterogeneity in the ability of pathogens to initiate infection. However,
because dose-response models are based on studies typically performed with healthy adult
volunteers and with a limited number of isolates, the available models might not capture the full
variability in human response or health endpoint. To address this variability, dose-response
model parameters can be treated as random variables or as part of a meta-distribution that can be
estimated via Bayesian inference (Englehardt and Swartout, 2004; Messner et al., 2001; Teunis
et al., 2008b). Susceptible sub-populations include immunocompromised and elderly,
individuals with prior exposure to a pathogen or related microorganism (Balbus et al., 2004;
Balbus and Embrey, 2002; Gerba et al., 1996), or, as in the case of norovirus, persons lacking a
specific antigen (Lindesmith et al., 2003).
2.15.2. Sources of uncertainty
The primary sources of uncertainty in QMRAs include
• enumeration (through microbiological analyses) estimation (through modeling) of
pathogen densities,
• choice of distributional form for FIB and pathogen densities,
• choice of distributional form and range of mobilization fractions (fraction of organisms
applied in manure that run-off during a rain event),
• uncertainty in dose-response model parameters,
• intensity of secondary infections, and
• model uncertainty.
The uncertainty in exposure assessment has two components: (1) FIB and pathogen density
estimates, and (2) volume of water ingested. For example, MPN estimates of microorganism
density are far more uncertain than those from membrane filtration techniques (Gronewold et al.,
2008; Gronewold and Wolpert, 2008). Membrane filtration results can be interpreted as Poisson-
distributed estimates around the true mean density (Gronewold and Wolpert, 2008). When
pathogens aggregate (e.g., via clumping or attaching to particles), this assumption is not valid,
and a Poisson-stopped logarithmic distribution (Teunis et al., 2008a) or discrete growth
distribution (Englehardt et al., 2009) may be more appropriate. Although not well characterized
in the literature, enumerations from qPCR methods are associated with uncertainty because of
the small volumes amplified, cycle-to-cycle variations in amplification efficiency, inhibition, and
other matrix effects (Ruijter et al., 2009; Rutledge and Cote, 2003). Uncertainty also arises when
comparing FIB and pathogen density estimates from qPCR and membrane filtration methods
because of differences in their ability to detect viable, viable but non-culturable cells (VBNC),
December 2010 43
-------
U.S. Environmental Protection Agency
dead cells, or extra-cellular DNA (Haugland et al., 2005; Liu et al., 2009; Nocker and Camper,
2006).
Dose-response models are developed based on data from (usually healthy) humans or animals
exposed to a known or estimated number of pathogens of a particular strain or subgroup.
Uncertainty arises for several reasons. First, the parameter estimates are based on limited data.
Second, sub-populations that were not represented in the studies might respond differently.
Third, the responses of homogenous groups of volunteers might differ from those of the general
population. Fourth, it is assumed that ingested doses, both in the dose-response experiments and
during recreation are homogenously (Poisson) distributed, but pathogens may clump, resulting in
differing actual ingested doses.
2.16. Factors and Data not Included in the QMRA
Factors not included in the QMRA include the following:
• Illnesses other than GI infections. Most other potentially water-related adverse health
outcomes do not correspond with FIB in recreational waters.
• Routes of exposure other than ingestion. Rates of inhalation and hand-to-mouth activities
are expected to be much lower than ingestion for swimming.
• Pathogens other than the reference pathogens. However, reference pathogens account for
the majority of potentially waterborne illnesses and are representative of other pathogens
that could potentially be in agricultural-animal impacted water.
• Potential loss of pathogen virulence during extra-enteric transport. Because of a lack of
data, assuming no loss is conservative and health protective.
• Growth of FIB and pathogens during transport. Growth of FIB and (bacterial) pathogens
is variable. A data-rich site-specific assessment would need to account for these factors.
• Die-off or attenuation of FIB and pathogens is beyond the scope of the scenario.
2.17. Identified Gaps in the Knowledge Base
Through extensive literature reviews and the EPA environmental monitoring studies (Section
2.12), we have assembled sufficient data to conduct a QMRA to estimate illness at a freshwater
beach impacted by agricultural animal sources of fecal contamination. Outstanding gaps in the
data include the following:
• Fate and transport of FIB and pathogens to estimate risks downstream (temporally and
spatially) from the source. This data gap results from a lack of understanding of transport
and survival processes for extra-enteric organisms and the variety of sites and conditions
under which pathogens and FIB move from fecal pollution sources to receiving water.
December 2010 44
-------
U.S. Environmental Protection Agency
• Effects of best management practices (BMPs). Basic research on the efficacy of best
management practices to remove pathogens from receiving water is ongoing. How well
BMPs perform is expected to vary between operations, seasons, and loading conditions.
Furthermore, new practices are in development and will likely be implemented at some
livestock operations. Although EPA has collected data on the efficacy of manure
treatment systems and BMPs, such as the presence of vegetative filter strips and fenced
areas for calves, those data cannot be used to characterize general conditions. This risk
assessment assumes that there is no treatment or other attenuation of pathogens in
livestock waste other than retention in manure matrixes and soil.
• Dose-response relationships. The dose-response for Salmonella likely does not account
for variability between the environmentally-relevant strains. The existing dose-response
relationship that accounts for strain variability is unstable at the low pathogen densities
that are relevant in recreational water. The dose-response for adenovirus20 is based on an
inhalational route of exposure for adenovirus 4; whereas, waterborne GI illness probably
results from ingested adenovirus 40/41.
• Animal-impacted recreational waters could contain pathogens of public health concern
that were not evaluated. As described previously, we selected our reference pathogens
based on robust criteria. The reference pathogens for agricultural animal-impacted
waters are assumed to be bacterial and protozoans as human infectious virus are typically
not associated with agricultural animals. However, Hepatitis E virus is associated with
livestock operations (Banks et al., 2004; Legrand-Abravanel et al., 2009; Rutjes et al.,
2009; Sinclair et al., 2009; Takahashi et al., 2009) and pigs shed rotavirus, but those
strains appear to be host-adapted and not likely to pose a significant risk to humans
(Martella et al., 2010). Although using bacterial and protozoan reference pathogens to
evaluate livestock-impacted water with QMRAs is appropriate at this stage of our
understanding, it is possible that future research could provide sufficient information that
a reference virus could also be included for agricultural animal-impacted water QMRAs.
Adenovirus is not used in this animal-impacted waters QMRA as it is generally species-specific. It was, however, used in the
reverse QMRA that was previously conducted to determine which pathogens are the most likely to cause illnesses in human-
impacted recreational waters (Seller et al., 2010a).
December 2010 45
-------
U.S. Environmental Protection Agency
3. Analysis
The analysis phase of a QMRA is the technical evaluation of data related to the potential
exposure to microbial contaminants, host characterization, and human health effects. It also
includes quantification of the dose-response relationship for contaminants in water media.
Although the problem formulation phase may partially address these issues, the analysis phase
provides more detail and quantitative analysis (USEPA, 2010).
The two components of the analysis phase are exposure characterization and human health
effects characterization of (ILSI, 1996). Characterization of exposure and human health effects
are iterative and interrelated processes because they must be compatible with the risk
characterization phase of the QMRA (Chapter 4). The analysis phase culminates with an
exposure profile (Section 3.1.8) and a host-pathogen profile (Section 3.2.4). Calculations using
these data are conducted within the risk characterization phase of the risk assessment
(Chapter 4).
3.1. Exposure
Figure 14 illustrates the processes leading to human exposure to pathogens at a freshwater beach
impacted by fecal contamination from agricultural sources. EPA conducted a literature review to
characterize the parameters associated with these processes (Annex 3) (see also Seller et al.,
201 Ob). The sections that follow highlight this literature review. Appendix B provides a tabular
summary of the literature review for FIB and reference pathogen levels in cattle, pig, and
chicken fecal source materials, as well as for chlorinated secondary effluent. Additional data that
may be useful to characterize reference pathogens levels in shorebird feces and urban runoff is
summarized in Appendix C.
December 2010 46
-------
U.S. Environmental Protection Agency
Runoff event
•Rainfall
•Runoff
Shedding abundance -Antecedentsoil moisture
•Pathogens Land application -Soil type
•Indicators
•Incorporation
•Distribution
/ Transport to
s tream
•Mobilization
•Overland flow
Mixing with
ream
Exposure
Figure 14. Schematic exposure diagram for recreation at agricultural animal-impacted waterbody
3.1.1. Prevalence and abundance of reference pathogens in livestock
Based on the systematic literature searches conducted as part of this report, we collated the
prevalence and abundance ranges for this risk assessment. Two types of prevalence data are
relevant to QMRAs for livestock-impacted water—sample-level prevalence and herd-level
prevalence. Sample-level prevalence is the proportion of fecal samples from a specific operation
or group of operations where a specific microorganism is detected. Herd-level prevalence is the
proportion of herds studied in which at least one sample is positive for a specific microorganism.
Because these QMRAs explore representative risks that animal operations pose to swimmers,
sample-level prevalence was used to calculate risk.
Abundance is the number of organisms per mass or volume of applied manure. The abundance
data used in this risk assessment were based on reported average pathogen densities for fresh
solid cattle manure, solid poultry litter, and liquid pig manure. Note that the use of average
values to characterize pathogen abundance is different than in prior work (Seller et al., 201 Ob),
which used abundances from individual fecal samples. Those abundances were used for an
exposure that was assumed to be in close proximity to manure deposited directly into
recreational water, which was deemed appropriate for that context. The use of average values in
this current risk assessment is appropriate because the land-applied material is effectively a
composite sample from multiple individual samples. Thus, the average value represents an
unbiased estimate for the expected value of pathogen density in the land-applied material. The
land-applied composite material is comprised of fecal material from shedding and non-shedding
animals, with the proportion of manure containing pathogens determined by the sample-level
prevalence of a particular organism. Therefore, the average density of a given pathogen in land-
applied fecal material is the average abundance scaled by the sample-level prevalence.
December 2010
47
-------
U.S. Environmental Protection Agency
3.1.1.1. Salmonella
Large-scale studies of Salmonella prevalence in pigs exhibited high year-to-year and herd-to-
herd variability, with reported prevalence generally falling in the 8 to 15% range (Foley et al.,
2008; Hutchison et al., 2004); however, prevalence among pigs appears to increase with age
(Dorr et al., 2009). Salmonella infection in cattle differed between dairy and beef cattle, as well
as with age, season, and herd size (Callaway et al., 2005; Edrington et al., 2004; Huston et al.,
2002; Kunze et al., 2008; Warnick et al., 2003; Wells et al., 2001). Large-scale studies of
Salmonella infection in both dairy and beef cattle (Fossler et al., 2005; Hutchison et al., 2004)
indicate prevalence in the 5 to 18% range, with higher prevalence reported for some herds.
Prevalence in chicken flocks (both layers and broilers) was highly variable and dependent on the
age of the chickens (Byrd, 1998; Martin et al., 1998) and possibly on the geographic region (Ebel
et al., 1992; Garber et al., 2003). Based on the high variability of Salmonella observed in these
studies, we selected a prevalence range of 0 to 95% as representative of Salmonella shedding
among chickens.
Among pig manure samples positive for Salmonella, two studies (Boes et al., 2005; Hutchison et
al., 2004) indicate a range of Salmonella fecal abundance from 102 8to 1049 organisms g"1 feces.
Slurries differ from fresh fecal deposits because the conditions under which wastes are stored
impact density range. The range 105 to 106'5 organisms/100 mL for abundance in swine slurry
was selected based on reported densities from a study with relatively high abundance taken from
a lagoon with fresh manure (Vanotti et al., 2005). Salmonella abundance in cattle feces was
reported in the range of 10°6 to 105'8 organisms g"1 feces. The range used for average abundance
in solid cattle feces is 1026 to 1046 organisms g"1 feces based on the findings of Fegan et al.
(2004) and Hutchison et al. (2004). Cattle from different production systems (grass vs.
concentrate fed) did not exhibit significantly different shedding densities (Fegan et al., 2004).
Average abundance of Salmonella in feces of chickens appears to be independent of bird age and
inoculation/ingestion dose (Byrd, 1998), with representative average densities in the range of 10"
1 tolO44 organisms g"1 of fresh chicken excrement (Kraft et al., 1969). Both studies used to
establish the poultry Salmonella density range based density estimates on multiple samples taken
from each house.
3.1.1.2. Campylobacter
Campylobacter spp. are frequently found in pig slurry lagoons (McLaughlin et al., 2009) and pig
feces (Dorner et al., 2004; Weijtens et al., 1997), with prevalence generally increasing with the
age of the animal. Given the high prevalence and increased prevalence with age, the pig
Campylobacter prevalence is estimated to be in the range of 46 to 98%. Cattle Campylobacter
prevalence differs among beef and dairy cattle, with feedlot cattle generally exhibiting higher
prevalence than cattle on pasture, and with prevalence increasing with the length of time cattle
occupy feedlots (Besser et al., 2005). Considering the different prevalence among operations
and between age cohorts, a representative range of prevalence for Campylobacter among all
December 2010 48
-------
U.S. Environmental Protection Agency
cattle is 5 to 38% (Hoar et al., 2001; Wesley et al., 2000). Chicken-shedding prevalence for
Campylobacter also tends to increase with age (Luangtongkum et al., 2006), and flocks
frequently approach 100% infection rates (Cox et al., 2002). Campylobacter shedding is nearly
universal among chicken houses and within-house rates are high and increase with bird age. A
representative range of Campylobacter prevalence in chickens is 57 to 69% (Cox et al., 2002; El-
Shibiny et al., 2005).
Studies reporting Campylobacter abundance in solid pig fecal samples (Hutchison et al., 2005;
Weijtens et al., 1999) suggest a representative density range of 102'° to 105'7 organisms g"1 feces;
whereas, a single study of slurry densities reported the range of 103'3to 103'7 organisms/100 mL
(McLaughlin et al., 2009). Studies of cattle Campylobacter abundance (Hutchison et al., 2005;
Inglis et al., 2004; Moriarty et al., 2008; Stanley et al., 1998) reported diverse results. The range
of average abundance we selected (101'8 to 104'5 organisms g"1 feces) spanned the averages in all
reported studies and fell within the full range of abundances observed in individual samples in
the study reporting the greatest variability (10L2to 107'3 organisms g"1 feces). Studies on
Campylobacter abundance in chicken feces (Bull et al., 2006; Cox et al., 2002; Hutchison et al.,
2005; Whyte et al., 2001) were in general agreement, with a representative range of 102'8 to 106'5
organisms g"1 feces. As with poultry Salmonella abundance data, the poultry Campylobacter
studies reported average abundances of groups of samples taken from floors of individual
houses.
3.1.1.3. E.coliO157:H7
E. coli O157:H7 infection and shedding occurs frequently among cattle and pigs, but is very
uncommon in chickens (Doane et al., 2007). Several studies report relatively low infection rates
among pigs (Chapman et al., 1997; Cornick and Helgerson, 2004; Feder et al., 2003; Hutchison
et al., 2004) with prevalence differing among types of operations and ages of animals—typically
in the range of 0.1 to 12%. Cattle E. coli O157:H7 prevalence and shedding are difficult to
characterize, given wide differences among age cohorts and animals on different types of
operations. E. coli O157 prevalence appears to differ between calves and adult cattle and
between cattle before and after their arrival on feedlots. E. coli O157 infection peaks in young
cattle between 3 to 18 months of age, and declines thereafter (Ellis-Iversen et al., 2009). In a
large study of beef cattle, LeJeune et al. (2004) observed a general increase in prevalence of E.
coli O157:H7 among animals with increased time spent in the feedlot.
Pig shedding of E. coli O157:H7 is highly variable, and a representative range of abundances
among all feces appears to go from none detected to 107 organisms g"1 feces (Cornick and
Helgerson, 2004), with animals shedding more intensely during early infection. We found no
data that estimated average density in swine slurry, so we conservatively estimated an average
density based on the reported density range and assuming feces were diluted to a slurry with a
4% solids fraction.
December 2010 49
-------
U.S. Environmental Protection Agency
Assessment of the available studies onE. coli O157:H7 in cattle (e.g., Berry et al., 2007;
Hutchison et al., 2004) led to estimates of prevalence and abundance ranges of 9.7 to 28% and
103'* to 1084 organisms g"1, respectively. The high end of the cattle E. coli O157:H7 abundance
range is very high and was taken from a large, systematic study that did not account for animal
age or super-shedding. To avoid biasing our estimate of cattle E. coli O157:H7 range by
including data from super-shedders or other samples that are not representative of land-applied
manure, we estimated the range of average cattle E. coli O157:H7 densities based on analysis by
Hutchison et al. (2004). Using their reported geometric mean and maximum densities, and
assuming abundances were log-normally distributed, the log-mean and standard deviation of the
average abundance of E. coli O157:H7 were estimated at 3.08 and 1.49, respectively.
3.1.1.4. Cryptosporidium
Estimates of ranges of prevalence and abundance of Cryptosporidium in livestock and other
wastes are based on a comprehensive review by Ferguson et al. (2009) and supplemented with
additional studies. Cryptosporidium shedding is sporadic among pigs, and individual herd
prevalence may be low with a characteristic range of 0 to 45% (Heitman et al., 2002; Hutchison
et al., 2005; Xiao et al., 2006). As for E. coli O157:H7, young cattle (<3 months) exhibit much
higher prevalence of Cryptosporidium than older cattle (Wade et al., 2000), as well as the
prevalence of genotypes that are more infectious to humans (Chalmers and Giles, 2010). A
representative range for Cryptosporidium prevalence in cattle, inclusive of all age groups, is
estimated to be 0.6 to 23%. Cryptosporidium shedding has been observed among chickens,
though the species excreted are generally not infectious to humans (Xiao et al., 2004). An older
study by Ley et al. (1988) reported Cryptosporidium prevalence among chickens to be between 6
to 27%.
For abundance in solid manure, a representative range of Cryptosporidium shedding rates among
pigs is 101'7 to 103'6 oocysts g"1 (Hutchison et al., 2004). Reinoso and Becares (2008) reported
the range of Cryptosporidium densities in swine slurry to be 104'2 to 105'4 oocysts/L. Cattle-
shedding rates for Cryptosporidium vary for calves and adults, with adults sporadically shedding
low densities of oocysts and calves shedding very high densities. To avoid making unnecessarily
subjective assumptions about the proportion of animals that are calves and the management
practices associated with calves and their manure, we excluded densities that were reported
based only on calf samples from the range of averages for cattle Cryptosporidium. Notably, this
choice led to the exclusion of data from the study by Wade et al. (2000), in which average
density among samples from calves positive for Cryptosporidium was 21,090 oocysts/g. A
1 ^9 1
representative range for average manure oocyst density of 10" to 10 oocysts g" was selected
based on data from Sturdee et al. (2003), where the low end of the range is based on a low
detection limit for Cryptosporidium in manure and the known tendency of adult cattle to shed
oocysts at low densities (Atwill et al., 2006). No studies allowed for the estimation of a range of
abundances of Cryptosporidium in chicken feces, though Hutchison et al. (2004) searched
December 2010 50
-------
U.S. Environmental Protection Agency
unsuccessfully for Cryptosporidium in fresh chicken manure as part of a large-scale study of
pathogens in livestock manure.
3.1.1.5. Giardia
Estimates for the prevalence of Giardia in pig feces are primarily drawn from Heitman et al.
(2002), Xiao et al. (2006), and Hutchison et al. (2004). The range of Giardia prevalence in pig
manure is estimated to be 3.3 to 18%. In cattle, Giardia prevalence varies with animal age, with
infection peaking when calves are relatively young and the probability of infection of an
individual within its lifetime approaching 100% in some operations (Olson et al., 1997; Ralston
et al., 2003; Wade et al., 2000). Two large-scale studies (Payer et al., 2000; Wade et al., 2000)
indicate a prevalence range for Giardia among cattle of 0.2 to 37%.
Wide ranges of shedding densities of Giardia among both pigs and cattle were observed, with
pig feces abundance in the range 10° to 106'8 cysts g"1 (data presented graphically in Maddox-
Hyttel et al., 2006). A single study reporting a slurry density of 103'5 cysts/L was provided by
Reinoso and Becares (2008). The range of average Giardia density for cattle was selected to be
10°'18 to 103'5 cysts g1 (Heitman et al., 2002; Wade et al., 2000).
3.1.2. A bundance of reference pathogens in disinfected secondary effluent
While the scenario evaluated in this QMRA does not cover human-impacted water specifically,
the results from the agricultural animal-impacted water QMRA are compared to waters impacted
by disinfected secondary effluent, as well as EPA's literature review and preliminary QMRA
work that included human-impacted sites (Schoen and Ashbolt, 2010; Seller et al., 2010a;
2010b). Estimating ranges of pathogen abundance in human fecal pollution is complicated by
the episodic occurrence of pathogens in sewage, large differences in removal of the pathogens
for different wastewater treatment processes, and differences in disinfection doses and contact
times. A summary of the literature review is provided below and a tabular summary is provided
in Appendix B.
None of the bacterial reference pathogens (E. coli O157:H7, Campylobacter, Salmonella spp.)
reportedly appear in significant densities in chlorinated secondary effluent (Garcia-Aljaro et al.,
2005; Lemarchand and Lebaron, 2003; Stampi et al., 1993) as they are Gram-negative species
that are very susceptible to disinfection. Reported densities of Cryptosporidium in secondary
effluent are relatively low, even in the absence of disinfection (Bonadonna et al., 2002; Bukhari
et al., 1997; Castro-Hermida et al., 2008; Payment et al., 2001; Scott et al., 2003). A
representative range of Cryptosporidium densities in secondary effluent that accounts for
episodes of natural variability in raw sewage and treatment process performance is 10"1 ° to 101'5
oocysts L"1 (Rose et al., 2004).
Reported Giardia densities in wastewater treatment plant effluent are somewhat higher than
Cryptosporidium densities, though Giardia is also subject to episodic loading and variations in
December 2010 51
-------
U.S. Environmental Protection Agency
removal depending on treatment processes (Bukhari et al., 1997; Carraro et al., 2000; Castro-
Hermida et al., 2008; Payment et al., 2001; Scott et al., 2003). Similar to the approach used for
Cryptosporidium, we selected the range of Giardia abundance in chlorinated secondary effluent
based on the widest reported range and estimate it to be 10"1 ° to 102'1 cysts L"1 (Rose et al.,
2004), not accounting for method recovery. Giardia cyst levels in chlorinated secondary effluent
are only slightly higher than Cryptosporidium levels despite substantially higher densities in raw
sewage and undisinfected secondary effluent because Giardia is inactivated to a greater degree
with chlorine than Cryptosporidium (USEPA, 2005a). A wide range of norovirus densities in
secondary effluent has been reported (Haramoto et al., 2006; Katayama et al., 2008; Laverick et
al., 2004; Lodder et al., 1999; Lodder and de Roda Husman, 2005; Pusch et al., 2005; van den
Berg et al., 2005). Based on these data, the range of norovirus abundance in chlorinated
secondary effluent is in the range 10"2 to 106 genomic copies L"1 (Haramoto et al., 2006;
Katayama et al., 2008) (triangular distribution with mode of 4 logs). We estimate the removal
range from treatment to be 1.0 to 4.0 logs (triangular distribution with mode of 2.5 logs).
3.1.3. Abundance of FIB in livestock manures
The FIB E. coli and enterococci are members of the normal intestinal microbiota of cattle, pigs
and poultry and are assumed to be present in 100% of their fecal samples.
For cattle, an important determinant of the shedding intensity for E. coli is diet. Berry et al.
(2006) observed different shedding intensities for cattle fed grass and cattle fed concentrate, with
the cattle on concentrate shedding E. coli at a significantly higher density. Other large studies
(e.g., Moriarty et al., 2008; Sinton et al., 2007; Thurston-Enriquez et al., 2005; Weaver et al.,
2005) reported E. coli fecal densities consistent with the range reported by Berry and colleagues
(including cattle on both grass and concentrate). Based on those studies, we use the full range of
E. coli fecal densities reported by Berry et al. (2006), 105'° to 106'7 CFU/g, for the E. coli density
in solid cattle manure QMRA simulations.
Reported pig slurry E. coli abundances fall within a narrow range (Coehlo et al., 2007; Flill and
Sobsey, 2003; Marti et al., 2009; Peu et al., 2006), given the variety of holding times, lagoon
designs, and environmental conditions associated with pig slurries at different farms. For the
average density, we used the highest and lowest average slurry densities reported in the
literature, resulting in a range of densities of 105'0 to 106'7 CFU/100 mL. Poultry E. coli densities
reported in the literature span a much wider range than those for solid cattle feces and pig
slurries, likely because chicken litter is a mixture of bedding, feathers, feces, and other materials,
and is more heterogeneous than solid cattle manure and swine slurry. Further, significant time
may pass between excretion of chicken feces and sampling of the litter from the chicken house
floor (not direct fecal deposits). The chicken litter E. coli density range for the relative QMRAs
was based on those observed by Terzich et al. (2000), because that study was large (operations in
12 states) and included assays of litter, not feces. The chicken litter E. coli density range was
105'°to 1010'9 CFU/g.
December 2010 52
-------
U.S. Environmental Protection Agency
Solid cattle manure enterococci densities appear more variable than those of E. coli., with
shedding density differing by season (Moriarty et al., 2008), type of operation (Weaver et al.,
2005), and other factors. The range of 102'4 to 106'8 CFU/100 mL was selected for solid cattle
manure average enterococci density based on data from Moriarty et al. (2008) and Thurston-
Enriquez et al. (2005). Average densities reported in other studies of solid cattle manure fall
within this range (Sinton et al., 2007). As for E. coli, the range of reported average enterococci
densities in swine slurry fell within a relatively narrow range (Bradford et al., 2008; Coehlo et
al., 2007; Hill and Sobsey, 2003; Peu et al., 2006; Vanotti et al., 2005). The average range used
in the relative QMRAs is 105'°to 105'9 CFU/100 mL, based on the studies by Peu et al. (2006)
and Bradford et al. (2008). Only two studies of chicken litter enterococci density were identified
(Brooks et al., 2009; Kelley et al., 1995). Those studies yielded a chicken litter enterococci
density range of 104 to 106 CFU/100 mL, with the upper end of the range estimated based on
data presented graphically by Brooks et al. (2009).
3.1.4. A bility of livestock-derived reference pathogens to infect humans
The relative fraction of human infectious strains in each reference pathogen in non-human
sources is important but highly uncertain. The available data are insufficient to quantitatively
characterize this attribute within a QMRA context. Thus, we assign categorical values of low
(L), medium (M), or high (H) to describe the ability of the livestock-derived reference pathogens
to infect humans based on (1) the overlap of species, strains, and serotypes known to infect
humans and to be present in the manure of the livestock species; (2) the prevalence of the
pathogen species and types most likely to infect humans as a proportion of the overall prevalence
of the pathogen in manure of a specific livestock host; and (3) review articles describing disease
transmission and host-specificity for the diseases associated with each pathogen. The mid-points
of the ranges of 0 to 33% for L, 33 to 66% for M, and 67 to 100% for H, were then used as point
estimates in this analysis.
3. L 4.1. Campylobacter spp.
Ketley (1997) designated C.jejuni and C. coli as the species playing a major role in human
infections (80 to 90% of Campylobacter infections), but notes that other species have the
potential for initiating human infections. For all livestock hosts, the prevalence of
Campylobacter species or subtypes of species varies between farms and regions, with age of
animal, season, between isolates from fecal samples and isolates from other environmental
reservoirs (e.g., trough water), and probably with other factors (El-Shibiny et al., 2005; Hakkinen
and Hanninen, 2009; Minihan et al., 2004; Weijtens et al., 1999; Wesley et al., 2000). C.jejuni
and C. coli are prevalent among cattle, pigs and chickens, with chickens exhibiting higher
incidence of C. coli shedding (as a percentage of all Campylobacter-positive samples) than that
of cattle and pigs (El-Shibiny et al., 2005).
December 2010 53
-------
U.S. Environmental Protection Agency
C.jejuni and C. coli are also the most often isolated from humans and their feces, animal hosts
and their feces, and environmental samples. Devane et al. (2005) reported a ratio of 90:10 for
C. jejuni isolates to C. coli isolates in human feces samples from New Zealand. These
researchers also observed that the two most common human isolates accounted for 43.6% of
isolates from beef cattle feces, 32.2% of isolates from dairy cattle feces, and lesser fractions of
isolates from other animals. The concordance between subtypes observed in humans and those
observed in beef cattle, dairy cattle, and sheep feces and sheep offal was confirmed in
subsequent work by Garrett et al. (2007). Furthermore, the dose-response characteristics C.
jejuni appear to differ among fresh cultures and laboratory cultures (Chen et al., 2006).
Given this lack of species-specific prevalence data and the absence of a general dose-response
model for human infection with C. coli, we know little about the potential for C. coli to infect
humans. Based on these observations, cattle and swine Campylobacter were assessed as having
high infectious potential for humans, while chicken Campylobacter were assessed to have
medium human infectious potential.
3.1.4.2. Salmonella
The relative risk posed by Salmonella serotypes in animals is inferred by comparing the
serotypes prevalent in different animal hosts and humans. The U.S. Centers for Disease Control
and Prevention (CDC, 2006) identified the serotypes from human Salmonella isolates between
1996 and 2006. The USDA Food Safety and Inspection Service (USDA FSIS, 2009) identified
the serotypes for Salmonella isolates identified in broilers, market hogs, steer and heifers, and
cows and bulls between 1998 and 2007. Collectively, these data indicate that the prevalence of
serotypes within a given host changes significantly from year to year, though for humans, the
serotypes typhimurium and enteriditis were consistently among the top three isolated. The
overlap between serotypes prevalent in humans and in livestock is used to estimate the potential
transmission of human-infectious Salmonella from livestock.
Table 5 and Figure 15 summarize the 24 most common serotypes of non-typhoid Salmonella
from human isolates. Serotype prevalence (as a percent of total isolates) for broilers,
steers/heifers, cows/bulls, and market hogs are also presented. They also show the overlap
between the most common human and animal Salmonella serotypes, with all animals exhibiting
relatively high prevalence of human-infecting serotypes Typhimurium, Newport, Saint-Paul,
Infantis, Anatum, and Mbandaka, and all hosts except pigs subject to infection with the
Montevideo serotype.
December 2010 54
-------
U.S. Environmental Protection Agency
Table 5. Salmonella serotype prevalences
Serotype
Typhimurium (w/ var. Copenhagen)
Enteriditis
Newport
Heidelberg
Javiana
Montevideo
Muenchen
Oranienburg
Saintpaul
Infantis
Thompson
Braenderup
Agona
I, 4, [5], 12:i-
Hadar
Mississippi
Typhi
Paratyphi B var L(+) tartrate (+)
Poona
Berta
Stanley
Anatum
Bareilly
Mbandaka
Other or not identified
Human
21.6
17.8
8.4
5.2
3.4
2.4
2.0
.7
.6
.5
.5
.0
.0
.2
.1
.0
.0
.0
0.8
0.6
0.6
0.6
0.5
0.5
20.5
Broiler
10.6
6.8
17.4
2.4
0.9
1.2
2.2
1.2
0.3
0.8
56.1
Steer/ Heifer
2.3
5.8
3.5
5.8
1.2
2.3
4.6
2.3
2.3
2.3
2.3
4.6
1.2
59.8
Cow/Bull
9.8
0.7
13.5
1.1
8.4
1.1
0.4
3.6
0.4
5.8
0.4
2.6
52.4
Market Hog
14.0
3.2
0.3
4.5
7.4
1.4
1.3
9.5
0.4
58.2
December 2010
55
-------
U.S. Environmental Protection Agency
25
S.
a 20
0.
ra
HJ
•s 15
U
B
0 10 -
a
c
-------
U.S. Environmental Protection Agency
3.1.4.4. Cryptosporidium spp.
Cryptosporidium species have widely varying public health significance and appear to have
adapted to specific hosts or groups of hosts. For example, Xiao et al. (2004) associated
Cryptosporidium species to major and minor hosts (Table 6). Among the more than 16 species
of Cryptosporidium identified to date, C. parvum and C. hominis are believed to cause the
majority of human infections among immunocompetent hosts. Other animals considered major
hosts for C. parvum and C. hominis include cattle, sheep, goats, and monkeys (Xiao et al., 2004,
2006). Humans are minor hosts for other Cryptosporidium species, including C. muris, C.
meleagridis, C. felis, and C. canis (Table 7). Even among C. parvum, however, the ability of
individual isolates to infect varies as illustrated in Table 8 (Messner et al., 2001).
Among livestock species, cattle more often carry Cryptosporidium species that infect humans,
while swine Cryptosporidia less often infect humans, and poultry Cryptosporidia appear to
infect humans rarely (Xiao et al., 2006). Consequently, the human infectious potential of cattle
and swine Cryptosporidia is assessed as high (given the occurrence of human infectious
Cryptosporidia in swine, but not the occurrence of C. suis in humans), and the human infectious
potential of chickens is considered as low.
Table 6. Valid Cryptosporidium species and associated major and minor hosts (SOURCE: adapted from Xiao et
al., 2004)
Species
C. muris
C. andersoni
C. parvum
C. hominis
C.felis
C. canis
C. meleagridis
C. baileyi
C. galli
Major Host
Rodents, Bactrian camels
Cattle, Bactrian camels
Cattle, sheep, goats, humans
Humans, monkeys
Cats
Dogs
Turkeys, humans
Chicken, turkeys
Finches, chicken, capercalles, grosbeaks
Minor Host
Humans, rock hyrax, mountain goats
Sheep
Deer, mice, pigs
Dugongs, sheep
Humans, cattle
Humans
Parrots
Cockatiels, quails, ostriches, duck
—
Table 7. Cryptosporidium spp. of humans and domestic animals (SOURCE: adapted from Xiao et al., 2004,
2006)
Host
Human
Cattle
Pig
Chicken
Major Parasites
C. hominis, C. parvum
C. parvum, C. andersoni
Pig genotype I
C. baileyi
Minor Parasites
C. meleagridis, C. felis, C. canis, C. muris, corvine genotype, pig
genotype I
Bovine genotype B, deer-like genotype, C. bovis, C. felis
Pig genotype II
C. meleagridis, C. galli
December 2010
57
-------
U.S. Environmental Protection Agency
Table 8. Cryptosporidiumparvum dose-response parameter estimates (SOURCE: adapted from Messner et al.
2001)
Isolate
UCP
IOWA
TAMU
Parameter Estimate (r)
Traditional
0.000336
0.00526
0.0571
Bayesian (80% Credible Interval)
0.000339 (0.000231, 0.000556)
0.00488 (0.00342, 0.00752)
0.0370 (0.0208, 0.0833)
3.1.4.5. Giardia spp.
Different researchers have called the species of Giardia that cause the majority of human
illnesses G. lamblia, G. duodenalis, or G. intestinalis (e.g., Adam, 2001; Thompson et al., 2004).
Thompson and colleagues noted that Giardia isolates from humans fall into one of two major
genotype assemblages, and that some Giardia genotypic groupings are confined to specific
animal hosts. Based on a listing of the most important Giardia species and genotypes and their
associated hosts (Adam, 2001), cattle and pigs appear to have the potential for shedding Giardia
that pose risks to humans, while chickens do not appear to be a significant source of human-
infectious Giardia cysts. Therefore, cattle and swine Giardia are assigned a high human
infectious potential and chicken Giardia are assessed as low.
3.1.5. Mobilization of reference pathogens and FIB
The mobilization of reference pathogens and FIB due to rainfall is estimated based on data from
the EPA environmental monitoring program (see Section 2.12 and Appendix D21).
The fraction of microorganisms mobilized during a rain event is primarily a function of the
following:
• The organism, soil type, particle size distribution, and the strength of attachment of
organisms to soil or manure matrices (Bradford and Schijven, 2002; Gargiulo et al., 2008;
Guber et al., 2005; Guber et al., 2007b; Guzman et al., 2009; Hodgson et al., 2009;
McLaughlin et al., 2003);
• The rainfall intensity and duration (Davies et al., 2004; Trask et al., 2004); and
• Groundcover, tillage, and slope (Davies et al., 2004; Guber et al., 2006; Harrigan et al.,
2004; Trask et al., 2004).
21 Mobilization fractions were computed based on data from the October 2009 and March 2010 Runs. As of September 2010, the
June 2010 run was complete, but the data were not yet available for these analyses. All of the raw data are available upon request
from Dr. Marirosa Molina, EPA.
December 2010 58
-------
U.S. Environmental Protection Agency
Models to estimate mobilization fractions are generally of the following form (Benham et al.,
2006; Pachepsky et al., 2009):
[10]
where AMR is the number of organisms released during some period At (e.g., over a specified
runoff event); MS is the number of bacteria in the manure storage layer prior to the runoff event;
a and b are empirically derived constants; and A<2 is runoff yield during the time interval At.
This general model form uses the ratio of the number of organisms appearing in the runoff
during a defined event and the number of organisms applied during the event.
In this risk assessment, we calculate the mobilization fraction for rain events using manure
densities from samples taken prior to simulated rain events, and FIB and pathogen densities in
composite samples collected during similar rain/runoff events, as follows:
_ Number of organisms occurring in the runoff from the plot during an event
Number of organisms applied to the plot in manure
This approach is similar to that of Miller and Beasley (2008), who assessed mobilization based
on flow weighted mean runoff density. Spiked and unspiked manures were sampled prior to
application on experimental plots, and manure pathogen and FIB densities were determined in
the manures with the methods described in Appendix D. For pathogens, composite runoff
samples were assembled by compositing all runoff originating from each plot and sampling the
composited runoff at 30 and 60 minutes after the initiation of runoff. For FIB, grab samples
were collected in addition to the 60-minute composite samples. The grab sample densities were
not used in mobilization fraction estimates in this risk assessment. Using these parameters, the
mobilization fraction (equation 1 1) is calculated as follows:
/,=
VROC,
V C
manure m,i
manure slurries
V C
R0 ' [12]
where, mmanure is the mass of solid manure applied to the plot; Vmanure is the volume of slurry
applied to the plot; A is the density of microorganism /' in the solid manure (number of
organisms/g); VRQ is the cumulative runoff volume for the rain event; and Ct is the density of
organism /' in a composite sample of all the runoff from the site (i.e., an event flow-weighted
average concentration, dimensions of number organisms per unit volume).
Appendix E describes the data and approach used to compute the pathogen and FIB mobilization
fractions for this risk assessment.
December 2010 59
-------
U.S. Environmental Protection Agency
3.1.6. Factors used to convert densities of pathogens on land to densities in runoff
As described above, mobilization fraction is the proportion of organisms applied to plots in
manure that is mobilized and transported in runoff. In the QMRA, FIB density in runoff water is
used with ingested volume to compute doses of pathogen exposure to use in dose-response
models to estimate risk. To estimate the density in the runoff water, the mobilization fraction
needs to be scaled as follows to ensure consistency of units:
solid manures
liquid manures L J
where C^RO is the density of organism /' in runoff water (dimensions of organisms/L3);/ is
mobilization fraction for organism /; mmanure is the per-plot manure application rate (dimensions
of mass per plot); Vsiurry is the per-plot slurry application rate (dimensions of L3); and VRO is the
cumulative runoff volume from the plot for the rain event (dimensions of L3).
This expression is used in the forward QMRA calculations in which we estimate risk associated
with a land application and runoff event (Section 4.1). Note that this expression is not needed in
the relative QMRA approach (Section 4.2) because the FIB and pathogens mobilized from
manure are both diluted by the same volume of runoff water.
The average runoff volume and per-plot manure application rates used are as follows:
• average cumulative runoff volume: 57.7 L;
• cattle manure application rate: 1600 g/plot;
• swine slurry application rate: 2670 mL/plot; and
• poultry litter application rate: 670 g/plot.
These conversions result in runoff densities with units of organisms per L of runoff.
3.1.7. Volume of water ingested
The volume of water ingested during recreational activities is characterized as a log-normal
distribution with a geometric mean of 18.5 mL and standard deviation of the logic transformed
data of 0.628 (Dufour et al., 2006; Seller et al., 2007b). This distribution is based on the reported
combined data for children and adults. These data are the most quantitative data available for
characterizing the volume of water ingested during recreational activities.
For comparison, previous QMRAs for recreational exposure have used ingestion volumes of 100
mL (Gerba et al., 1996; Steyn et al., 2004; Wong et al., 2009), 50 mL (Ashbolt and Bruno,
2003), 30 mL (van Heerden et al., 2005a), and an empirical distribution of ingested volumes with
a range of 0 to 190 mL (specific to sports divers) (Schijven and de Roda Husman, 2006).
Alternative ingestion values were evaluated via sensitivity analyses in this QMRA.
December 2010 60
-------
U.S. Environmental Protection Agency
3.1.8. Exposure profile
The exposure profile distills the most important information and data developed during the
exposure component of the analysis phase. Each component of the exposure analysis describes
the information available on that specific topic. The exposure profile includes only the
information that will be used in conjunction with the human health characterization for the risk
characterization.
For this risk assessment, the exposure-related data that will be used in the calculations include
the following:
• literature-based data characterizing the densities of reference pathogens in fecal source
material (abundance) for each animal source (solid fresh cattle manure, fresh swine
slurry, and fresh poultry litter);
• literature-based and EPA environmental monitoring program-based data characterizing
the FIB densities (abundance) in each animal source (solid fresh cattle manure, fresh
swine slurry, and fresh poultry litter);
• literature-based data characterizing the prevalence of infection from reference pathogens
in each animal source (cattle, pigs, and chicken);
• a qualitative interpretation of the literature-based data describing the relative fraction of
human infectious strains of each of the reference pathogens in non-human sources;
• EPA environmental monitoring program-based data characterizing the proportion of the
land-applied FIB and pathogens that mobilize (mobilization fraction) following a rain
event and runoff to a recreational waterbody; and
• literature-based data characterizing the volume of water ingested during recreational
activities.
Tabular summaries of the specific data that are used in the QMRA calculations are provided in
Sections 4.1 and 4.2.
3.2. Health Effects
3.2.1. Health endpoint
The health effect of interest in this QMRA is GI illness. Other health outcomes have been
excluded for reasons described previously. Thus, the reference health outcomes in the QMRA
analyses are
• Infection through ingesting surface water contaminated with reference pathogens during
recreation, and
• GI illness conditional on infection.
December 2010 61
-------
U.S. Environmental Protection Agency
As described in Section 2.3, the reference pathogens for this risk assessment are
Cryptosporidium spp., Giardia lamblia, Campylobacter spp., Salmonella enterica, andE1. coll
O157:H7.
3.2.2. Dose-response relationships
As described previously, dose-response relationships for the reference pathogens are taken from
the peer-reviewed literature and are for an infection endpoint. Definitions of infection most
often used in dose-response models were seroconversion and/or shedding of pathogens in feces.
Likewise, the definition of reference pathogen illness varied (summarized in Section 3.2.3
below), but was generally related to the incidence of diarrhea, and/or vomiting. The following
are descriptions of and justifications for the reference pathogen dose-response relationships for
this recreational water QMRA effort. Again, only bacterial and protozoan reference pathogen
are used in this QMRA.
3.2.2.1. Cryptosporidium dose-response model
The dose-response model for Cryptosporidium in the QMRA is based on analysis for the Long
Term 2 Enhanced Surface Water Treatment Rule (LT2ESWTR) (USEPA, 2005a). In the
experimental dose-response studies, human response varied widely to different isolates of
Cryptosporidiumparvum (Messner et al., 2001; Okhuysen et al., 1999, 2002). With analyses
based on those of Messner et al. (2001), the LT2ESWTR Cryptosporidium dose-response model
was developed using Bayesian analyses of individual and combined data sets for different
isolates and outbreak data. The LT2ESWTR dose-response model is exponential with model
parameter r = 0.09. Uncertainty within the dose-response model is evaluated by allowing the
model parameter to vary uniformly across the range of 0.04 to 0.16, consistent with the range
reported in the LT2ESWTR (USEPA, 2005a).
3.2.2.2. Giardia dose-response model
The Giardia dose-response model was developed based on data from human feeding studies with
Giardia lamblia dose over a range of 1 to 106 cysts (Rendtorff, 1954a, 1954b). Response data
corresponding to infection (endpoint was shedding cysts in feces) were fit to an exponential
dose-response model with parameter r = 0.0199 (Rose et al., 1991).
3.2.2.3. Campylobacter spp. dose-response model
We evaluated two dose-response models for Campylobacter. The first is based on a feeding
study conducted by Black et al. (1988). The resulting dose-response relationship is fit to a beta-
Poisson dose-response relationship with parameters a= 0.144 and /?= 7.59 (Medema et al.,
1996). The second is based on outbreak data associated with exposure to contaminated milk
(Teunis et al., 2005). An exact beta-Poisson dose-response model with parameters a = 0.024
and (3 = 0.011 provided the best fit to the outbreak data.
December 2010 62
-------
U.S. Environmental Protection Agency
3.2.2.4. E. coli O15 7:H7 dose-response model
The E. coli O157:H7 dose-response model was derived using data from eight outbreaks (Teunis
et al., 2008b) and from an assumption that doses ingested in each of those outbreaks were
Poisson-gamma distributed. The exposure model was refined by adjusting the gamma-
distribution parameter for exposure to reflect the dispersion associated with each outbreak. An
exploration of various models led Teunis and colleagues to select a beta-Poisson dose-response
model (infection endpoint). Dr. Teunis developed and made available 10,000 pairs of dose-
response parameters. We use two approaches with these data. In the first, median values from
those pairs are used as point estimates (a= 0.4 and (3= 37.6). In the second approach,
uncertainty in the dose-response parameter space is evaluated through the use of the individual
dose-response parameter pairs in the Monte Carlo simulations.
3.2.2.5. Salmonella dose-response model
Salmonella occurrence and infectivity varies widely with serotype (McCullough and Eisele,
195 la, 195 Ib). To account for this, the dose-response model for Salmonella was chosen to be
representative of the overall incidence of infection when individuals are exposed to the range of
serotypes that could reasonably occur in recreational water. We evaluated two Salmonella dose-
response models, a beta-Poisson model (Haas et al., 1999) and a Gompertz-log model (Coleman
and Marks, 1998, 2000; Seller et al., 2007b). The Haas and colleagues dose-response model is
based on infection data for multiple serotypes of Salmonella., with outlier data excluded from
analysis. The best fit model for the pooled data set is the beta-Poisson model, with parameters a
= 0.3126 and /? = 2884. The log-Gompertz model (for an illness endpoint) evaluation showed
that the model parameters took on a range of values for the serotypes for which human dose-
response data were available. Assuming that the infectivity of environmentally relevant
serotypes are uniformly distributed over the observed range from the feeding study, the dose-
response parameter In (a) is estimated to vary uniformly between 29 and 50, and b = 2.148.
3.2.2.6. Rotavirus dose-response model
The rotavirus dose-response model was developed using data from human feeding studies (Ward
et al., 1986). Volunteers in the study were adult males, 18 to 45 years old. Overall, the ratio of
ill-to-infected individuals was 0.67, and the progression of infection to illness did not appear to
be dose-dependent. The approximate beta-Poisson model with parameters a = 0.2531 and/? =
0.4265 (Haas et al., 1993) provided the best fit to the data. An issue unresolved in the peer-
reviewed literature is that the viral units used in the feeding studies were reported as focus
forming units rather than individual viral particles. It is, therefore, possible that the most
commonly used assumption—that PFUs of rotavirus are equivalent to the focus forming units
from the feeding study—results in an overestimation of risk associated with rotavirus.
December 2010 63
-------
U.S. Environmental Protection Agency
3.2.2.7. Norovirus dose-response model
The dose-response model (infection endpoint) for the QMRA studies is an exact beta-Poisson
model with parameters a = 0.04 and /? = 0.055. The norovirus dose-response model was
developed from human feeding studies conducted with healthy adult volunteers (Teunis et al.,
2008a). In this volunteer study, the virus was aggregated in the inoculum and so the dose-
response model had to be flexible enough to account for this aggregation. Because norovirus
particles are expected to be dilute in recreational water, this QMRA assumes that aggregation of
viral particles will be minimal. Given this assumption, the aggregated dose dose-response model
(Teunis et al., 2008a) simplifies to an exact beta-Poisson model.
3.2.2.8. Adenovirus dose-response model
The adenovirus dose-response model is based on dose-response data of adult human exposure to
aerosols of adenovirus type 4 (Couch et al., 1966, 1969). For aerosol exposure and an infection
endpoint, the best fit dose-response model for adenovirus is an exponential model with parameter
r = 0.4172 (Crabtree et al., 1997). Use of the inhalation dose-response model yields conservative
estimates for infection rates, because infection among adults is initiated with higher probability
at lower doses via aerosol exposure than via other routes. The use of the inhalation adenovirus 4
dose-response model for predicting GI infection via oral exposure is established in the literature
(Crabtree et al., 1997; Teunis et al., 1999; van Heerden et al., 2005a, 2005b).
The mismatch between this dose-response model and an ingestion route of exposure is likely to
make risk predictions from adenovirus more uncertain than those for other reference pathogens.
A significant fraction of the non-infant population may have a level of immunity to GI infection
with adenovirus.
3.2.3. Morbidity
For this analysis, morbidity refers to the proportion of infections that progress to a symptomatic
response (illness). For each of the reference pathogens, morbidity is expressed as a range to the
extent that supporting data are available. Justification is provided below for the morbidity ranges
used in the QMRA analyses.
In the dose-response study for Campylobacter, the proportion of infections progressing to illness
Q 8
was dose-dependent with best fit parameter estimates of K =3.63 x 10 and 77 = 2.44 x 10
(refer to equation 6). In a human feeding study (Black et al., 1988), there was no apparent trend
with dose for the proportion of infections progressing to symptomatic illness, and approximately
18% of infected volunteers became symptomatic (fever, diarrhea, or both). In this QMRA, the
morbidity ratio is assumed to be dose-independent because that assumption yields more
conservative estimates of illness at low doses and reflects the uncertainty we believe is present
for the Campylobacter dose-response model for low doses. Based on the data from the feeding
December 2010 64
-------
U.S. Environmental Protection Agency
study, the progression from infection to symptomatic illness for Campylobacter is assumed to
occur in the range of 0.1 to 0.6.
The progression from infection to symptomatic illness for E. coll O157:H7 is assumed to be in
the range of 0.2 to 0.6 based on outbreak data (Bielaszewska et al., 1997); the percentage of
symptomatic and asymptomatic individuals who were household contacts of hemolytic uremic
syndrome patients (Werber et al., 2008); and the occurrence of anti-Stx2 IgG (Ludwig et al.,
2002). This range is consistent to the proportion of illnesses reported in an analysis of an E. coli
O157:H7 outbreak (Teunis et al., 2004).
The progression from infection to symptomatic illness for Salmonella varied from zero to one
during the feeding studies, with low morbidity (0%) in most cases (McCullough and Eisele,
195 la, 1951b). Given the wide variability and high proportion of relatively low morbidity in the
feeding studies, a point estimate of 20% is used to characterize the progression from infection to
symptomatic illness.
The progression from infection to symptomatic illness for Cryptosporidium is based on EPA's
research from development of the LT2ESWTR (USEPA, 2006). In that analysis, EPA analyzed
available literature and identified studies with applicable data. DuPont et al. (1995) found that
39% of those infected had clinical cryptosporidiosis. Haas et al. (1996) provided information
based on the same data also suggesting a morbidity rate of 39%, but computed 95% confidence
limits of 19% and 62%. More recently, a study found that after repeated exposure to C. parvum
(IOWA strain), the morbidity rate was the same as for the initial exposure in re-infected subjects
(Okhuysen et al., 1998). Okhuysen et al. also found that 58% of their subjects who received
Cryptosporidium doses developed diarrhea, which is an underestimate of morbidity because
symptoms other than diarrhea contribute to the morbidity rate. Based on these data, the
progression from infection to symptomatic illness for Cryptosporidium is assumed to range from
0.2 to 0.7.
Giardia infection is often asymptomatic, with asymptomatic cases representing as much as 50%
to 75% of infected persons (Mintz et al., 1993). In a study at the Swiss Tropical Institute, 27%
of 158 patients who had Giardia cysts in their feces exhibited symptoms (Degremont et al.,
1981). Based on these data, the progression from infection to symptomatic illness for Giardia is
assumed to be in the range of 0.2 to 0.7.
The progression from infection to symptomatic illness for norovirus is assumed to be in the
range of 0.3 to 0.8 based on feeding study data (Teunis et al., 2008a). In that study, the
conditional probability of illness among infected subjects appears to show dose dependence.
However, dose independence is assumed using the lowest and highest proportion of ill patients
for the various doses studied as the lower and upper bounds of the morbidity range, respectively.
December 2010 65
-------
U.S. Environmental Protection Agency
3.2.4. Health effects profile
Similar to the exposure profile, the health effects profile is a distillation of the most important
information and data that are developed during the health effects component of the analysis
phase. The health effects profile is a relatively brief summary of only those pieces of
information that will be used in conjunction with the exposure characterization for the risk
characterization phase of the assessment.
For this risk assessment, the health effects-related data that will be used in the calculations
include the health endpoint of interest (GI illness), the dose-response relationships for the
reference pathogens, and the fraction of infections that lead to illness (morbidity). Sections 4.1
and 4.2 provide tabular summaries of the specific data used in the QMRA calculations. The
interaction between these components and the risk characterization phase of the assessment is
illustrated schematically in Figure 16.
Compy/obocterspp.
• Ingested dose
• Dose-response, infection endpoint
Campylobacter
• Morbidity ratio
• GI illness endpoint
Salmonella
' Ingested dose
• Dose-response, infection endpoint
Salmonella
• Morbidity ratio
• GI illness endpoint
E. Coli 0157
• Ingested dose
• Dose-response, infection endpoint
E. Coli 0157
• Morbidity ratio
• GI illness endpoint
Cryptosporidium parvum
• Ingested dose
• Dose-response, infection endpoint
C. parvum
• Morbidity ratio
• GI illness endpoint
Giardiaspp.
• Ingested dose
• Dose-response,
infection endpoint
Giardiaspp.
• Morbidity ratio
• GI illness endpoint
Net risk of GI illness
• Sum of risks for individual
pathogens
• Distribution of risks via
Monte Carlo simulation
Figure 16. Interaction between health effects and risk characterization components
December 2010
66
-------
U.S. Environmental Protection Agency
4. Risk Characterization
Risk characterization combines the methods outlined in the problem formulation phase and
the data compiled in the analysis phase to compute and convey the overall potential risk to
humans for the scenario(s) under consideration (USEPA, 2010). EPA's policy statement
on risk characterization prescribes a clear, transparent, and reasonable process that is
consistent with other assessments of similar scope prepared across EPA programs (USEPA,
2000). This phase of the assessment identifies and discusses all the major issues associated
with determining the nature and extent of the risk. It also provides commentary on any
constraints limiting interpretation of the results. The nature of a risk characterization
depends on the data, information, and resources available and the regulatory application of
the assessment.
Risk characterization, which can include both qualitative and quantitative data, summarizes the
extent and weight of evidence and the results, major points of interpretation, and rationale. It
also describes the strengths and weaknesses of the evidence, and discusses uncertainties,
variability, and potential effects of alternative assumptions. Scenarios, model parameters, and
analysis options that deserve further consideration are identified, so that assessment results can
inform decision-making.
As described in the problem formulation chapter (2), this QMRA addresses the following two
questions:
1. What is the risk of illness associated with recreation at a freshwater beach impacted by
agricultural animal (cattle, swine, and chicken) sources of fecal contamination? and
2. How do those risks compare to risks associated with freshwater beaches impacted by
human (POTW) sources of fecal contamination?
A described below, we used two distinct QMRA risk characterization approaches to answers
these questions. The first question is addressed through forward QMRA and the second question
by relative QMRA. The forward QMRA provides a conservative estimate of risk associated with
each of the fecal pollution sources based on the scenario. The relative approach normalizes risks
to a specific FIB level to allow a direct comparison of risks among sources.
4.1. Risk of Illness Associated with Recreation at a Beach Impacted by Agricultural Animal
Sources of Fecal Contamination
To characterize the risk of illness associated with recreation at a freshwater beach impacted by
cattle, pig, and chicken sources of fecal contamination, we used a traditional forward QMRA
approach. The general methodology for the QMRA is a Monte Carlo simulation-based approach
with model parameters characterized as statistical distributions. The primary benefit of the
Monte Carlo simulation approach compared to a simpler point-estimate approach is that the
December 2010 67
-------
U.S. Environmental Protection Agency
inherent variability in the modeled system is accounted for explicitly and the resulting
distribution provides a more nuanced view of predicted risks. Separate Monte Carlo analyses are
conducted for each agricultural animal source. Sensitivity analyses evaluate the effects of
alternative assumptions and parameter values on the model outputs and complement the base
QMRA results.
4.1.1. Methods
A detailed schematic diagram for the forward QMRA is presented in Figure 17.
Density of reference
pathogens in source
material (abundance)
Infection rate for reference
pathogens (prevalence)
Human infectious potential
for reference pathogens
Agronomic rate land application
Pathogen mobilization during runoff
(specific to each fecal pollution
source)
Pathogen densities in recreational
waters impacted by a specific fecal
pollution source
Volume of water
ingested
Pathogen dose-
response parameter
values (generally
infection endpoint)
Dose of
pathogen
ingested
I
Dose-response
relationship:
Infection or Illness
Morbidity
Infections and illnesses
attributable to individual
reference pathogens
Total G I illness rate per 1000
recreation events
Figure 17. Detailed conceptual model for forward QMRA
December 2010
68
-------
U.S. Environmental Protection Agency
The risk characterization begins with literature-based data describing the average densities of
reference pathogens (abundance) in each of the animal sources under consideration (solid fresh
cattle manure, fresh pig slurry, and fresh poultry litter). Table 9 summarizes of the data used for
this purpose (see Chapter 3 for further information about the data). We selected cattle manure,
pig slurry, and chicken litter abundance data to represent average values (typified by land-
applied composited manure) from operations in the United States. The abundance data are also
based on studies with wide geographic range, long duration, and large numbers of samples,
where possible. This approach resulted in narrower abundance ranges than those based on
individual fecal samples (Seller et al., 2010b) or uncommon events (e.g., super-shedding cattle,
abundance ranges based on samples collected from young animals, operations without shedding
animals). This approach minimizes potential bias in interpreting the literature-based data
describing the number and ages of animals producing manure, or animal/manure management
practices such as handling calf manure separately from adult animal manure.
Table 9. Abundance of reference pathogens in agricultural animal sources
Pathogen
E. co//O157:H73
Campylobacter
Salmonella
Cryptosporidium
Giardia spp.
Cattle1
Low High
Log-normal (3.08, 1.49)
1.8 4.5
2.6 4.6
-0.3 3.2
0.2 3.5
Low
0
3.3
5
4.2
3.5
Pigs2
High
5.6
3.7
6.8
5.4
3.5
Chicken
Low High
0 0
2.0 6.3
0.5 4.4
0 0
0 0
1. Density in solid manures (cattle manure and chicken litter): units of Iog10 organisms/g wet weight
2. Density in liquid manures (pig slurry): units are Iog10 organisms/100 mL
3. Log-normal distribution used in place of log uniform to account for low probability events with very high abundances. Values
shown are log mean and log standard deviation values
Abundance ranges from the literature for reference pathogens are characterized by log-uniform
distributions in this analysis. E. coli O157:H7 in cattle had to be treated differently because the
available abundance data indicate that average abundances are strongly influenced by infrequent
shedding of high levels of pathogens. A log-normal distribution was used to account for this
characteristic because a log-uniform distribution would have over-estimated the likelihood of an
extreme event.22
We also used literature-based data to characterize the prevalence of infection in each of the
animal sources (cattle, pigs, and chicken). In this analysis, prevalence represents the average
proportion of animals that are shedding the reference pathogens at any point in time. As
22 In a uniform distribution, all values between the minimum and maximum are equally likely (the log-uniform refers to the fact
that values shown are Iog10 values; i.e., a value of 3 corresponds to 1000). ForE. coli O157:H7 in cattle, use of a uniform
distribution (for the Iog10 values) would have resulted in too many values at the high end of the range, therefore an alternative
distribution was used that fit the literature-based data more closely.
December 2010
69
-------
U.S. Environmental Protection Agency
described in Chapter 3, this prevalence is different than herd-level prevalence, which quantifies
the fraction of herds that have at least one shedding individual during a specified period. In this
report, we conservatively assume that shedding is occurring (i.e., herd-level prevalence is 100%).
For all of the reference pathogens, relatively high herd-lev el prevalences are reported,
particularly for Campylobacter in all of the livestock types and for E. coli O157:H7 and
Cryptosporidium among cattle. A summary of the data employed for this purpose is provided in
Table 10.
Table 10. Prevalence of infection (% of animals shedding reference pathogens at any point in time
Pathogen
E. co//O157:H7
Campylobacter
Salmonella
Cryptosporidium
Giardia
Cattle
Low High
9.7 28
5 38
5 18
0.6 23
0.2 37
Pigs
Low High
0.1 12
46 98
7.9 15
0 45
3.3 18
Chicken
Low Ligh
0
57
0
6
0
0
69
95
27
0
1. The apparent mismatch between chicken Cryptosporidium abundance in Table 9 and prevalence in Table 10 results from the
enumeration of all Cryptosporidia in the study on prevalence and only specific species in the study on abundance. As noted in
Section 3.1.1.4, there is no overlap in the Cryptosporidium species for which humans and chickens are major hosts.
The relative fraction of human infectious strains of each of the reference pathogens from non-
human sources is highly uncertain, and the literature did not have sufficient data to confidently
assign quantitative values or ranges to this model parameter. However, not all strains of
pathogens that animals shed infect humans. Section 3.1 describes attempts to quantify the
overlap in pathogenicity of animal and human strains. Those data indicate variation in the
human health risk posed by pathogens originating from cattle, swine, and chickens. Values of
low (L), medium (M), or high (H) human infectious potential are assigned to each reference
pathogen for each fecal source based on the prevalence of known human-infectious
species/strains/serotypes/isolates in animal feces (Seller et al., 2010b). The mid-points of the
ranges of 0 to 33% for L, 33 to 66% for M, and 67 to 100% for H, were then used as point
estimates in this analysis (Table 11).
Table 11. Human infectious potential
Pathogen
E. co//O157:H7
Campylobacter
Salmonella
Cryptosporidium
Giardia
Cattle
H
H
M
H
H
Pigs
H
H
M
L
H
Chicken
—
M
M
—
—
December 2010 70
-------
U.S. Environmental Protection Agency
The fresh fecal materials are assumed to be applied to land at agronomic rates. This choice is
consistent with common practice and provides a conservative estimate of the pathogen load
available for mobilization, because the agronomic rate is the highest rate at which manures
should be applied. The results from the EPA environmental monitoring program (summarized in
Section 2.12) are used to characterize the proportion of the land-applied pathogens that mobilize
(mobilization fraction) following an intense rain event and runoff to a recreational waterbody.
The density of organisms (FIB and pathogens) in water running off manure-applied plots is
proportional to the number of land-applied organisms, with a different proportionality constant
for each organism/manure type combination. Appendix E describes the specific methods used to
derive these mobilization fractions. Table 12 summarizes the mobilization fractions.
Table 12. Mobilization fractions for land applied fecal wastes (logio values)
Pathogen
E. coli O157:H7
Campylobacter
Salmonella
Cryptosporidium
Giardia spp.
Low
-3.65
-4.85
-5.57
-4.46
-6.40
Cattle
High
-0.20
-1.46
-1.26
-0.18
-0.39
Low
-3.01
-2.20
-3.85
-3.90
-4.58
Pigs
High
-1.50
-1.01
-2.40
-1.48
-0.06
Chicken
Low High
-4.01 -2.21
-8.60 -1.74
-3.68 -2.65
Not tested
Not tested
For the risk characterization, it is assumed that mobilization of pathogens due to a specific runoff
event are correlated to each other; that is, an event which mobilizes one pathogen to a relatively
high degree (within its observed mobilization range) also mobilizes other pathogens to a similar
degree (within the observed range for those pathogens). This approach is implemented
numerically by generating a random number between zero and one (for each iteration in the
simulation) and treating that number as a percentile of the mobilization distributions for each
reference pathogen present in the land-applied manure (each of which is log-uniform). In each
simulation iteration, the mobilization fractions for each pathogen are computed from that
percentile of the corresponding distribution for the microorganism-manure combination. This
process is repeated for each of 10,000 iterations in each simulation.23
Using the data summarized above, we calculate (1) the density of each reference pathogen in
runoff water as the product of the reference pathogen abundance in land-applied fecal waste from
infected animals; (2) the prevalence of infection in each animal source; (3) the human infectious
potential of each pathogen, (4) a proportionality constant (specific to the fecal source type and
the rain event to which the calculations are referenced) that is used to convert organisms applied
to organisms in runoff water; and (5) the proportion of the applied reference pathogens that run-
off following a rain event (i.e., the mobilization fraction).
23 For example, for a particular iteration a random number between 0 and 1 is drawn—assume 0.15. Next, for that iteration, we
assume that the "rain event" causes runoff at the 15th percentile of each of reference pathogens. Although the mobilization
fractions will vary from pathogen to pathogen depending on the reported ranges (low and high in Table 12), the relative fraction
mobilized for each event is driven by the intensity of the event (as determined by the random number).
December 2010 71
-------
U.S. Environmental Protection Agency
The recreational waterbody is assumed to be adjacent to the plots where the fresh fecal material
is applied and has a small water volume relative to the volume of runoff it receives during the
rain event. Recreation is assumed to occur at the point where the runoff meets the adjacent
recreational waterbody. Therefore, the dose of pathogens for this exposure scenario is the
product of the volume of water ingested during recreational activities and the density of each
pathogen in the runoff. This scenario yields a conservative estimate of risk in that dilution would
occur to some degree when the runoff meets the receiving water, and the dilution would reduce
the density of the pathogens in the water ingested the further downstream that ingestion occurs.
The volume of water ingested in the base analysis is modeled as a log-normal distribution with a
log mean and log standard deviation of 2.92 and 1.43 mL, respectively (equivalent to a GM of
18.5 mL) (Dufour et al., 2006). Sensitivity analyses evaluate the implications of alternative
volumes of water ingested—specifically, evaluations of 1 mL and 50 mL ingestion values
consider minimal- and high-intensity water contact activities, respectively.
The computed doses to the appropriate dose-response relationships were calculated and resulted
in a probability of infection. Table 13 summarizes the dose-response relationships used in the
base analyses. Sensitivity analyses evaluate alternative dose-response relationships that account
for uncertainty. The alternative dose-response relationships were selected because they represent
the best available alternatives that allow consideration of uncertainty based on available
information from the literature (USEPA, 2010). Specifically, the exponential dose-response
model parameter r is varied uniformly between 0.04 and 0.16 for Cryptosporidium spp. (USEPA,
2005a, 2006). The hypergeometric dose-response model parameters of a = 0.011 and /? = 0.024
are evaluated for Campylobacter jejuni (Teunis et al., 2005); note that individual beta-Poisson
alpha and beta pairs supplied by Dr. Teunis were used for E. coli O157 (Teunis et al., 2008b)
rather than the median of those values (as used in the base analyses). Lastly, a Gompertz-log
distribution was evaluated for Salmonella enterica (illness) with a uniformly distributed ln(a)
parameter with values ranging from 29 to 50 and parameter b equal to 2.148 (Coleman and
Marks, 1998, 2000; Seller et al., 2007b).
The probability of infection from each reference pathogen is multiplied by a pathogen-specific
morbidity ratio (Table 13) to produce a probability of illness. The risk associated with each fecal
contamination source (cattle, pigs, and chicken) is then characterized as the total probability of
GI illness, based on the probability of illness from each of the reference pathogens, as described
previously.
The forward QMRA risk calculations are conducted with a modified version of MRAIT (Seller
et al., 2007a), which was originally designed to characterize risks associated with exposure to
pathogens in reclaimed water. It was modified for use in this risk assessment to accept input
parameters consistent with this exposure scenario. The MRAIT dose-response section was also
updated for E. coli O157:H7 to accommodate new peer-reviewed information (Teunis et al.,
2008b).
December 2010 72
-------
U.S. Environmental Protection Agency
Using the parameters and values described above, MRAIT generated 10,000 iterations for each
reference pathogen in each fecal contamination source. This approach required 12 simulations
for the base analysis (5 for cattle, 5 for pigs, and 2 for chicken because E. coli O157,
Cryptosporidium, and Giardia are not found in chicken litter) and 6 simulations for the
sensitivity analyses (alternative dose-response for Campylobacter, E. coli O157,
Cryptosporidium, and Salmonella, and two alternative ingestion values to represent minimal- and
high-intensity water contact activities). Thus, a total of 18 simulations (each of which comprised
10,000 iterations) were performed. Appendix F illustrates a MRAIT screen image from one of
the simulations. Results were saved, exported to text files, and subsequently used to generate
tabular and graphical summaries.
Table 13. Dose-response models and morbidity
Reference
Pathogen
Cryptosporidium
spp.
Giardia lamblia
Campylobacter
jejuni
E. coli O157:H7
Salmonella
enterica
Published Dose-
Response Model
Exponential
(USEPA, 2005a, 2006)
Exponential
(Haas et al, 1999; Rose et
al, 1991)
Beta-Poisson
(Medemaetal, 1996;
Teunis et al., 1996; 2005)
Beta-Poisson
(Teunis et al., 2008b)
Beta-Poisson
(Haas et al., 1999)
Model
Parameters
0.09
0.0199
0.145
7.59
0.4
37.6
0.3126
2884
Infectious
Doseso
8 oocysts
35 cysts
800 CPU
207 CPU
23,600
CPU
Morbidity
(% of Infections
Resulting in Illness)
20-70%
20-70%
10-60%
20-60%
20%
Health
Endpoint
Infection
Infection
Infection
Infection
Infection
4.1.2. Base analysis cattle results
The base analysis QMRA simulation results for fresh cattle manure based on all five of the
bacterial and protozoan reference pathogens are summarized and presented in Table 14, Figure
18 (boxplot format), and Figure 19 (cumulative probability format). In Figure 18 (and
subsequent boxplots), the edges of the box represent the 25th and 75th percentiles of the
simulation results (probability of infection or illness), the line in the center of the box is the
median value, the whiskers represent the 10th and 90th percentiles, and the diamonds below and
above the whiskers represent the 5th and 95th percentiles, respectively.
December 2010
73
-------
U.S. Environmental Protection Agency
Table 14. Summary of infection and illness risks from recreation in cattle manure-impacted water
Infection Risks
Pathogen
E. coli O157
Campylobacter
Salmonella
Cryptosporidium
Giardia spp.
Cumulative
5th %ile
3.9E-05
7.4E-06
7.5E-08
6.0E-07
1.3E-08
2.2E-04
10th %ile
1.3E-04
2.1E-05
2.3E-07
2.6E-06
5.2E-08
6.4E-04
Median
1.6E-02
1.6E-03
2.6E-05
5.2E-04
3.7E-05
4.6E-02
90th %ile
6.2E-01
7.9E-02
2.5E-03
9.7E-02
2.4E-02
7.7E-01
95th %ile
8.1E-01
1.6E-01
7.1E-03
3.5E-01
9.8E-02
9.1E-01
5th %ile
1.5E-05
2.3E-06
1.5E-08
2.6E-07
5.6E-09
8.4E-05
Illness Risks
10th %ile
4.8E-05
6.6E-06
4.6E-08
1.1E-06
2.2E-08
2.3E-04
Median
5.8E-03
5.0E-04
5.1E-06
2.2E-04
1.5E-05
1.8E-02
90th %ile
2.3E-01
2.7E-02
4.9E-04
4.1E-02
l.OE-02
3.3E-01
95th %ile
3.2E-01
5.4E-02
1.4E-03
1.5E-01
4.3E-02
4.5E-01
Current geometric
mean RWQC
ECO157inf
ECO157MI
Campy inf
Campy ill
Sal inf
Sal ill
Crypto inf
Crypto ill
Giardia inf
Giardia ill
Cum inf
Cum ill
Figure 18. Probability of infection and illness from recreation in cattle-impacted water
December 2010
74
-------
U.S. Environmental Protection Agency
Current geometric
mean RWQC
equivalent
R coli 0157 ill
Campylobacter
Salmonella
Cryptosporidium
Giardia
Prob. vs. Cum. ill
10.2 0.5 1 2
10
70 80 90 95 98 99 99.89.9
Percent of predictions less than corresponding value
Figure 19. Cumulative probability illness risk plot for cattle manure-impacted water
Collectively, the data presented in Table 14, Figure 18, and Figure 19 indicate the following:
• The greatest predicted median risk of illness from recreational exposure to the
hypothetical cattle-impacted waterbody is associated with E. coli O157.
• The predicted median risks of illness associated with Campylobacter and
Cryptosporidium are approximately an order of magnitude below that of E. coli O157.
• The predicted median risks of illness associated with Giardia and Salmonella are
approximately two to three orders of magnitude below that associated with E. coli O157.
• The predicted cumulative median risk of illness from recreational exposure to the cattle-
impacted waterbody, as specified in the QMRA scenario, is slightly lower than, but
effectively equivalent to the risk of illness that is associated with the current geometric
mean RWQC based on water impacted by human sources of contamination (USEPA,
1986).24
• The predicted 90th percentile risk of illness associated withE1. coli 0157 is the highest of
the reference pathogens, followed by Cryptosporidium, Giardia, and Campylobacter—
each of which are approximately one order of magnitude lower.
24 This can be seen from the data presented in the following two ways: (1) compare RWQC geometric mean GI illness equivalent
risk (0.03) to the cumulative risk from Table 14 (0.018), and (2) compare the 50th percentile of the cumulative risk line in Figure
19 to the geometric mean RWQC line.
December 2010
75
-------
U.S. Environmental Protection Agency
• The predicted 95th percentile risk of illness associated with E. coli O157 and
Cryptosporidium are the highest of the reference pathogens, followed by Campylobacter
and Giardia, which are approximately one half of an order of magnitude lower.
4.1.3. Base analysis pig results
The base analysis QMRA simulation results for pig slurry based on all five of the bacterial and
protozoan reference pathogens are presented in Table 15, Figure 20, and Figure 21.
Table 15. Summary of infection and illness risks from recreation in pig slurry-impacted water
Infection Risks
Pathogen
E. coli 0157
Campylobacter
Salmonella
Cryptosporidium
Giardia spp.
Cumulative
5th %ile
2.2E-08
7.2E-04
7.1E-06
6.5E-05
1.4E-06
l.OE-03
10th %ile
6.9E-08
1.3E-03
1.5E-05
1.6E-04
3.3E-06
1.9E-03
Median
1.7E-05
1.1E-02
2.4E-04
4.4E-03
3.2E-04
2.1E-02
90th %ile
4.7E-03
7.3E-02
3.9E-03
9.7E-02
2.6E-02
2.1E-01
95th %ile
1.4E-02
1.1E-01
8.1E-03
2.0E-01
6.5E-02
3.5E-01
5th %ile
7.7E-09
2.0E-04
1.4E-06
2.6E-05
5.6E-07
3.4E-04
Illness Risks
10th %ile
2.4E-08
3.7E-04
6.9E-08
6.7E-05
1.4E-06
6.4E-04
Median
6.0E-06
3.5E-03
1.7E-05
1.9E-03
1.3E-04
7.6E-03
90th %ile
1.7E-03
2.5E-02
4.7E-03
4.3E-02
1.2E-02
8.5E-02
95th %ile
4.9E-03
3.8E-02
1.4E-02
9.0E-02
2.9E-02
1.6E-01
Current geometric
mean RWQC
8
~s
.a
f
.
.8
ff
ECO157 inf
ECO157 ill
Campy inf
Campy ill
Sal inf
Sal ill
Crypto inf
Crypto ill
Giardia inf
Giardia ill
Cum inf
Cum ill
Figure 20. Probability of infection and illness from recreation in pig slurry-impacted water
December 2010
76
-------
U.S. Environmental Protection Agency
a
fin
10-s
Current geometric
mean RWQC
equivalent
K coli O157
Campylobacter
Salmonella
Cryptosporidium
Giardia
Cumulative
0.10.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 99.89.9
Percent of predictions less than corresponding value
Figure 21. Cumulative probability illness risk plot for pig slurry-impacted water
The data in these tables and figures reveal the following:
• The greatest predicted median risks of illness from recreational exposure to the
hypothetical pig-impacted waterbody are associated with Campylobacter and
Cryptosporidium.
• The predicted median risk of illness associated with Giardia is approximately an order of
magnitude below that of Campylobacter and Cryptosporidium.
• The predicted median risks of illness associated with Salmonella and E. coli O157 are
approximately two to two-and-a-half orders of magnitude below those of Campylobacter
and Cryptosporidium.
• The predicted cumulative median risk of illness from recreational exposure to the
hypothetical pig-impacted waterbody is approximately four-times lower than the risk of
illness that is associated with the current geometric mean RWQC (0.0076 compared to
0.03).
• The predicted 90th percentile risk of illness associated with Cryptosporidium,
Campylobacter, and Giardia are the highest of the reference pathogens, followed by
E. coli O157 and Salmonella, which are approximately one order of magnitude lower.
• The predicted 95th percentile risk of illness associated with Cryptosporidium,
Campylobacter, and Giardia are the highest of the reference pathogens. The 95th
percentile risk of illness associated with Salmonella is slightly lower, followed by E. coli
December 2010
77
-------
U.S. Environmental Protection Agency
O157, which is approximately one order of magnitude lower than the reference pathogens
exhibiting the highest risks.
4.1.4. Base analysis chicken results
The base analysis QMRA simulation results for fresh chicken litter based on two of the bacterial
reference pathogens are presented in Table 16, Figure 22, and Figure 23. Inspection of those
data highlights that chicken litter simulations were not conducted for E. coli O157,
Cryptosporidium or Giardia (the literature review described in Section 3.1.3 and summarized in
Table 9 indicated that the abundance of these references pathogens is minimal or zero, thus
simulations were not conducted for these reference pathogens).
Table 16. Summary of infection and illness risks from recreation in chicken litter-impacted water
Infection Risks Illness Risks
Pathogen 5th %ile 10th %ile Median 90th %ile 95th %ile 5th %ile 10th %ile Median 90th
%ile 95th %ile
E. coli O157 ..........
Campylobacter 2.2E-08 1.2E-07 1.9E-04 1.7E-01 3.3E-01 6.9E-09 3.6E-08 6.0E-05 5.2E-02 1.1E-01
Salmonella 2.9E-08 8.0E-08 5.1E-06 3.4E-04 9.0E-04 7.2E-09 2.0E-08 1.3E-06 8.5E-05 2.2E-04
Cryptosporidium ..........
Giardia spp. ----------
Cumulative 4.1E-07 1.5E-06 3.5E-04 1.7E-01 3.3E-01 1.1E-07 4.4E-07 l.OE-04 5.2E-02 1.1E-01
Probability of adverse effect
300000000
o o
o o
_^ _^ Current G.
i — ' — i i — — i equivalent
/ ° /
'/
!{!•
o o _L
.j7 f .°w 3 .g 3 .s 3 .s § .§ § ^ Campy inf
& <*9 & ~5' & ~5' P" ~& 5-> ~& 5-> ~& \ / A Csmpyl ill
<$ •=? <$ -^ ^ -^ ^ -^ ^ -^ ^ •=? ^^ 7
a /\ a i* .0 as .O fy .O .as .V ^> 1 1 o8l IDT
VJQ^
-------
U.S. Environmental Protection Agency
2 10-5-
10-s
Current geometric
mean RWQC
equivalent
Campylobacter
Salmonella
Cumulative
0.10.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 99.89.9
Percent of predictions less than corresponding value
Figure 23. Cumulative probability illness risk plot for chicken litter-impacted water
The data presented in Table 16, Figure 22, and Figure 23 indicate the following:
• The predicted median Campylobacter risk of illness from recreational exposure to the
chicken litter-impacted waterbody is greater than that associated with Salmonella by
approximately one-and-a-half orders of magnitude.
• The predicted cumulative median risk of illness from recreational exposure to the
hypothetical chicken-impacted waterbody is approximately 300-times lower than the risk
of illness that is associated with the current geometric mean RWQC (0.0001 compared
to 0.03).
• The predicted 90l and 95l percentile risks of illness associated with Campylobacter are
approximately two to three orders of magnitude greater than those associated with risks
from Salmonella.
4.1.5. Base analysis comparison of results
Comparisons of the QMRA simulation results for the cattle manure, pig slurry, and chicken
litter-impacted recreational water are presented in Figure 24 and Figure 25. Note that Figure 24
consolidates the cumulative illness risks from Figure 18, Figure 20, and Figure 22 into a single
boxplot. Figure 25 presents the probability densities for the simulation results.
December 2010
79
-------
.£ Environmental Protection Agency
M
0>
_c
-------
U.S. Environmental Protection Agency
Figure 24 and Figure 25 reinforce the following interpretations of the base QMRA simulation
results:
• The predicted median cumulative risk of illness from recreational exposure to the cattle-
impacted waterbody is effectively equivalent to the risk of illness that is associated with
the current geometric mean RWQC based on water contaminated by human sources.
• The predicted median cumulative risk of illness from recreational exposure to the pig-
impacted waterbody is approximately four-times lower than the risk of illness that is
associated with the current geometric mean RWQC.
• The median cumulative risk of illness from recreational exposure to the chicken-impacted
waterbody is approximately 300-times lower than the risk of illness that is associated
with the current geometric mean RWQC.
• The predicted variability is greatest for chicken-impacted water and least for pig-
impacted water.
• A substantial portion of the simulations for cattle-impacted water resulted in risks that
appear to be relatively high (for example, greater than 100 illnesses per 1000 recreation
events). A smaller but still substantial fraction of the simulations for pig and chicken-
impacted water also resulted in apparently high risks.25
4.1.6. Sensitivity analysis results for alternate dose-response relationships
We used pig slurry-impacted water to conduct the sensitivity analysis for alternative dose-
response relationships to represent all three sources and to maximize the likelihood that any
differences would be apparent (because cattle-impacted water risks are higher, the potential to
observe substantial changes in simulation output is lower). Similarly, because chicken-impacted
water risks are substantially lower, changes in simulation output may not represent changes in
simulation output for cattle-impacted water. Finally, pig slurry risks include all reference
pathogens, whereas, chicken litter risks include only a subset of the reference pathogens.
As indicated in Section 4.1.1, alternative dose-response simulations were conducted for
Cryptosporidium, Campylobacter, E. coli O157, and Salmonella enterica. The results from
those simulations are presented in probability plot format in Figure 26 (Cryptosporidium), Figure
27 (Campylobacter), and Figure 28 (E. coli O157). The alternative simulations for Salmonella
enterica resulted in illness risks that were extremely low (below 10"9), so are not presented
graphically.
' The parameter combinations causing these high risk outcomes are discussed in Section 4.3.
December 2010 81
-------
U.S. Environmental Protection Agency
100
10-1-
O 10-3-
c
g
"§ 1°-4-
M—
c
"S 10-5-
"ro 10-6-
o
^ 10-7 H
10-8-PL
Current geometric
mean RWQC
equivalent
Cryptosporidium infection
Cryptosporidium infection alternate dose response
Cryptosporidium illness
Cryptosporidium illness alternate dose response
0.01 0.06.10.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 9929.9
Percent of predictions less than corresponding value
Figure 26. Cumulative probability plot: evaluation of alternative dose-response for Cryptosporidium
100 ,
10-1-
-------
U.S. Environmental Protection Agency
10°
10-1-
10-2 H
5 10-3-
c
g
I KHH
^c
"5 10-5-
ro
.0
o
10-6-
D- 10-7-
10-8
Current geoometric
mean RWQC
equivalent
Infection
Infection alternate dose response
Illness
Illness alternate dose response
0.10.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 99.89.9
Percent of predictions less than corresponding value
Figure 28. Cumulative probability plot: evaluation of alternative dose-response for E. coli O157
The sensitivity analysis indicates that the QMRA results are not substantially impacted by
including uncertainty in dose-response^br Cryptosporidium (Figure 26) or E. coli O157 (Figure
28). This can be seen by comparing the "illness" to "illness alternative dose-response" curves in
the corresponding figures. The results of the QMRA simulations are, however, sensitive to the
selection of dose-response relationships for Campylobacter jejuni and Salmonella.
The alternative dose-response relationship for Salmonella was a Gompertz-log relationship that
was developed to account for strain variability (Coleman and Marks, 1998, 2000; Seller et al.,
2007b). The Gompertz-log model is an empirical dose-response model that is based on a good
fit to the experimental data collected in volunteer feeding studies from the 1940s (McCullough
and Eisele, 195 la, 195 Ib). Those studies, however, used doses of Salmonella that were several
orders of magnitude greater than the predicted doses used in this QMRA. Because the
Gompertz-log model is an empirical model, and the doses under consideration are outside of the
range that provided a good fit, the extent to which this dose-response relationship may be used to
extrapolate to low-dose risk predictions is not known.
The alternative dose-response for C. jejuni is a hypergeometric (exact beta-Poisson) function
(Teunis et al., 2005). This dose-response relationship amends the previous dose-response
relationship by Medema et al. (1996) to account for low-dose human response to C. jejuni
exposure shown in two contaminated milk outbreaks. This relationship exhibits higher levels of
infection at low doses and a steeper increase with dose than the previous function, which was
December 2010
83
-------
U.S. Environmental Protection Agency
based only on the human feeding study. However, other outbreak-based dose-response studies
indicate that the dose-response relationship can shift based on the type of contaminated media
(e.g.,type of food, water) (Bollaerts et al., 2008). Given that this alternative dose-response is
based on exposure (primarily in children) to C.jejuni in milk, and this risk assessment is for the
general population exposed to Campylobacter spp. via animal-impacted recreational water, use
of the base analysis Campylobacter dose-response is reasonable. However, the additional
uncertainty associated with the applicability of the alternative dose-response relationship should
be taken into consideration during the risk management process.
4.1.7. Sensitivity analysis results for alternate ingestion
Sensitivity analysis simulations were also conducted using alternative rates of water ingestion of
pig slurry-impacted water—1 mL and 50 mL (compared to the base analysis that used a log-
normal distribution of water ingestion with geometric mean of approximately 18 mL). For these
simulations, Cryptosporidium was the model reference pathogen.
The 1 mL ingestion is used to evaluate the potential risks associated with low-contact activities
such as wading, beachcombing, fishing, and others. Similarly, the 50 mL ingestion is used to
evaluate the potential risks associated with prolonged exposure to water or vigorous water play.
The base analysis was designed to be consistent with the self-reported body-contact recreation in
EPA's water epidemiology studies.
The results from these simulations are summarized in Table 17 and presented in boxplot format
in Figure 29. These alternative ingestion volume simulations indicate that the median risks scale
linearly with volume ingested, within the evaluated ranges. Furthermore, the 5th and 10th
percentiles of the risk distributions are impacted to a lesser degree than the median risk values by
the selection of a point estimates rather than the use of a log-normal distribution. The 90l and
95th percentiles of the risk distributions are impacted to a greater degree than the median risk
values by the selection of a point estimate rather than the use of a log-normal distribution.
Table 17. Alternate ingestion: Cryptosporidium infection and illness from pig-impacted runoff
Ingestion
1 mL Point estimate
Lognormal distribution
50 mL point estimate
5th %ile
7.0E-06
6.5E-05
3.5E-04
Infection Risks
10th %ile Median 90th %ile
1.5E-05
1.6E-04
7.4E-04
2.4E-04
4.4E-03
1.2E-02
3.4E-03
9.7E-02
1.6E-01
95th %ile
6.3E-03
2.0E-01
2.7E-01
5th %ile
2.9E-06
2.6E-05
1.5E-04
10th %ile
6
6
3
2E-06
7E-05
1E-04
Illness Risks
Median 90th %ile
9.9E-05
1.9E-03
4.9E-03
1.5E-03
4.3E-02
6.9E-02
95th %ile
2.8E-03
9.0E-02
1.2E-01
December 2010
84
-------
U.S. Environmental Protection Agency
10° -,
lo-1-
"5 10"2 -
^ i n-4 -
O
~r
I
i
_L
o
o
-f-
1
1
_L
O
O
— 1 —
— 1 —
///
i /// O
/ / y — 1 —
1 / / /
/ / . \
o / /. / ///
Y/s
///
A ///
L
o
— 1 — Current G
/// equiva en
///
///
0
io-6
io-7
io-8
l/ /l
Infection std ingeston
Infection 1ml ingestion
Infection 50 ml ingestion
Illness std ingestion
illness 1 ml ingestion
Illness 50 mL ingeston
Figure 29. Alternate ingestion: Cryptosporidium infection and illness from pig-impacted runoff
4.2. Relative QMRA for Animal-Impacted Water and Human-Impacted Water
A relative QMRA approach was used to compare the risk of illness associated with recreation at
a freshwater beach impacted by cattle, pig, and chicken sources of fecal contamination and
human-impacted recreational water. This approach complements the forward QMRA approach
by normalizing risks from specific livestock fecal pollution sources to a specified FIB density
and by facilitating a comparison of risks between different fecal pollution sources. However,
this approach requires more assumptions and data than used in the forward QMRA approach,
such as the range of FIB densities in fecal pollution sources and an assumption that FIB
mobilization has the same driving forces as pathogen mobilization.
For animal-impacted water, FIB and pathogen loading to a recreational waterbody can occur via
direct or indirect (runoff) contamination. Previous studies developed a methodology, model, and
set of literature to evaluate the estimated human health risks from exposure to recreational waters
impacted by human and direct fresh non-human sources of fecal contamination (Schoen and
Ashbolt, 2010; Seller et al., 2010b). The results from those relative QMRA studies indicate that
at a given level of FIB in a waterbody, the GI illness risks associated with recreational exposure
impacted by direct cattle contamination might not be substantially different from those impacted
by human sources. However, the risks associated with exposure to recreational water impacted
December 2010
85
-------
U.S. Environmental Protection Agency
by direct gull, chicken, and pig fecal wastes appear to be substantially lower than those impacted
by human sources (Seller et al., 2010b).
The relative QMRA conducted for this study extends the previous work by considering indirect
loading of pathogens and FIB via land application of livestock waste to a waterbody. Rather
than assuming livestock wastes are deposited directly into receiving water, the revised model
assumes that fresh livestock wastes are applied to land at agronomic rates, and pathogens and
FIB are mobilized and transported to receiving water during an intense rain event. Although
manure handling practices differ between operations, we assume land application, which reflects
national practices. This approach does not preclude a QMRA for sites that handle manure
differently. In addition, down-slope processes from land-applied wastes also vary between sites.
Here, we conservatively assumed that runoff is introduced into the receiving water with no
buffer strips or other best management practices in place. Finally, estimates of FIB and pathogen
mobilization during and subsequent to rainfall events are based on the results of the EPA
environmental monitoring studies.
Similar to the forward QMRA described above, the general methodology for this relative QMRA
is a Monte Carlo simulation-based approach with model parameters characterized as statistical
distributions. The simulation code used by Schoen and Ashbolt (2010) and Seller et al. (2010b)
was adapted for this relative QMRA to include mobilization of land-applied pathogens and FIB
due to a rainfall event.
4.2.1. Methods
In these analyses, the estimated risks are calculated for a hypothetical waterbody that contains
FIB densities from fresh cattle manure, fresh pig slurry, and fresh poultry litter at the current
geometric mean RWQC (USEPA, 1986) for freshwater (33 CFU /100 mL enterococci and 126
CFU /100 mL E. coli, respectively). Separate analyses were performed for each source of fecal
contamination based on each of the FIB.
The conceptual diagram for the relative QMRA was presented previously (see Figure 8).
Reference pathogen doses are derived as a function of the density of the FIB from each of the
specific sources (Schoen and Ashbolt, 2010; Seller et al., 2010b). Specifically, pathogen dose is
calculated based on independent Monte Carlo samples from observed or literature-based ranges
of pathogen and FIB densities in fecal waste, the prevalence of infection, the fraction of human-
infectious strains, and the proportion of the FIB and pathogens that mobilize during a rain event.
This sampling scheme does not require a specific relationship between the FIB and pathogens in
the fecal waste or in the receiving water. However, the mobilization of pathogens and FIB are
related to each other, as Section 3.1.6 describes. The dose of each reference pathogen from each
source is calculated as follows:
December 2010 86
-------
U.S. Environmental Protection Agency
a
^^-x(7£xA^)x/£x/;xF [14]
where
5" is the fecal contamination source;
CFIB is the waterbody density of enterococci or E. coli (CFU 100 mL"1);
nS
FIB is the density of FIB in land-applied manure (g or CFU lOOmL" );
MpIB is the mobilization fraction of the FIB for the fecal source (#/100mL
runoff)/(#/g manure) or (#/mL manure runoff/#/mL manure slurry);
Rs
rp is the density of pathogen species in land-applied manures with pathogens
(number of pathogens or genomes (g or 100 mL)"1);
Msrp is the mobilization fraction of the pathogen species for the fecal source;
s
rp is the fraction of human-infectious pathogenic strains from source S;
Is
rp is the prevalence of infection in the non-human source (proportion of animals
shedding the pathogen); and
V is the volume of water ingested (mL).
Although this equation is similar to that used in previous related QMRAs (Schoen and Ashbolt,
2010; Seller et al., 201 Ob), except here the pathogen and FIB densities in water ingested during
recreation are a function of the organisms in the land-applied manure and the mobilization
fractions of the organisms during rain events. In the previous studies, (1) mobilization fractions
of the organisms during rain events were not included, and (2) direct contamination occurred
from cattle, pig, and chicken feces rather than indirect contamination from cattle manure, pig
slurry, and chicken litter.
Similar to the forward QMRA described above, doses are input to the appropriate dose-response
relationship resulting in a probability of infection. Probability of illness is computed using the
morbidity fractions for each reference pathogen. The total probability of illness for each
contamination source is computed as described previously (i.e., one iteration). Each simulation
includes 10,000 iterations for each fecal contamination source/pathogen combination. The
resulting distributions of risk are compared to a benchmark risk for human-impacted water
(based on the current geometric mean RWQC) and to the risk results for direct agricultural
contamination as reported by Seller et al. (201 Ob).
December 2010 87
-------
U.S. Environmental Protection Agency
The animal source pathogen abundance, prevalence, human infectious potential, mobilization
fractions, morbidity fractions, and dose-response models used in this analysis are the same as
were used in the forward QMRA presented above (Table 9, Table 10, Table 11, Table 12, and
Table 13, respectively). Table 18 summarizes the literature-based and observed data used to
characterize the FIB densities (abundance) in solid fresh cattle manure, fresh pig slurry, and fresh
poultry litter.
Table 18. Abundance of fecal indicator bacteria in fecal sources
Indicator
Enterococci
E. coli
Literature
Observed
Literature
Observed
Cattle
Low High
2.4 6.8
4.7 5.5
5.0 6.7
6.7 8.3
Pigs
Low High
5.0 5.9
0.2 2.0
5.0 6.7
0.7 3.1
Chicken
Low high
4.0
3.8
5.0
2.7
6.0
5.8
10.9
4.4
1. Density in solid manures: units of Iog0 (#/g wet weight)
2. Density in liquid manures: units are logo (#/100 mL)
Similar to the forward QMRA, we used EPA environmental monitoring program (see Section
2.12 and Appendix D) results to characterize the proportion of the land-applied FIB that mobilize
and run-off to a recreational waterbody (mobilization fraction) following a typical rain event.
The density of FIB (E. coli and enterococci) in water running off manure-applied plots is
assumed to be proportional to the number of land-applied organisms, with a different
proportionality constant for each organism/manure combination.
Several alternative indicator organism-detection method combinations were used to monitor
indicator density in manure and runoff in the EPA environmental monitoring program. Those
alternatives included anE1. coli O157 surrogate with soil and manure matrix affinities and runoff
characteristics that were assumed to be similar to those of generic E. coli. The alternatives with
sufficient data for characterizing mobilization of FIB were enterococci via culture on membrane-
enterococus indoxyl-B-D-glucoside (mEI) agar, E. coli via the Colilert MPN method, and E. coli
O157 surrogate strain via membrane filtration. The observed mobilization fractions for E. coli
via Colilert and enterococci via culture on mEI ager were analyzed to determine whether the
mobilization distributions appeared uniform or triangular (described in Appendix E). Table 19
summarizes the mobilization distributions used for FIB in the relative risks analyses.
December 2010
88
-------
U.S. Environmental Protection Agency
Table 19. Mobilization of fecal indicator bacteria for animal fecal sources
Organism
Enterococci
E. coli
E. coli O157 surrogate
Cattle
Distriubtion Values
Uniform (-2.8, 0.3)
Triangular (-5.0, -2.75, -2.0)
Uniform (-3.7, -0.20)
Pigs
Distriubtion Values
Uniform (-1.0,2.5)
Uniform (-2.0,1.0)
Uniform (-3.0, -1.5)
Chicken
Distriubtion Values
Triangular (-1.25, -0.25, 0.32)
Triangular (-2.75, 0.25, 1 .25)
Uniform (-4.0, -2.2)
1. Mobilization fractions reported as Iog10 values
2. Parameters of the Uniform distribution are (min, max)
3. Parameters of the Triangular distribution are (min, mode, max)
4. Mobilization values greater than 0 represent an increase in indicator in runoff compared to the land applied material
For the relative QMRA, we conducted two complementary sets of analyses using the data
described above. In the first approach (Approach 1), the enterococci andE. coli abundance and
mobilization distributions were based on observed data from the EPA environmental monitoring
program ("Observed" data in Table 18; "enterococci" and "E. coir data in Table 19). The
results from these analyses are likely specific to the observed abundances of the FIB in the land-
applied materials and site features such as the densities of the FIB in the soil (before land
application).
The mobilizations reported for enterococci and E. coli depend on their abundance in the land-
applied material. As described in Appendices D and E, use of the pig slurry data in Approach 1
is highly questionable because the manure pathogen densities are drawn from distributions
developed based on a literature review and intended to reflect typical conditions in the United
States. In contrast, manure FIB densities were selected from manures specific to the EPA
environmental monitoring program experiments. Furthermore, the FIB densities in runoff from
control plots were high relative to runoff from pig manure-treated plots and some chicken-
manure-treated plots—particularly for enterococci (values substantially greater than 0 in Table
19).
In Approach 2, the enterococci and E. coli abundances were literature-based (Table 18), the
E. coli mobilization distributions were based on the E. coli O157 surrogate data, and the
enterococci mobilization distributions were based on the observed data from the EPA
environmental monitoring program (Table 19). The use of the pig slurry enterococci
mobilization data is inappropriate in this case because the mobilizations reported for enterococci
and E. coli are dependent on their abundance in the land-applied material. The abundance and
mobilization data for this approach represent average values that are substantially less dependent
on the enterococci andE. coli levels observed during the environmental monitoring program.
December 2010
89
-------
U.S. Environmental Protection Agency
4.2.2. Relative QMRA results
The relative QMRA simulation results are summarized in Table 20 and shown in boxplot format
in Figure 30 for Approach 1 and Figure 31 for Approach 2. A probability density plot for the
E. co//-based results from Approach 2 is presented in Figure 32.
The predicted Approach 1 enterococci and E. co//'-based results for swine slurry, and E. Co/A-
based results in chicken litter were driven by the low observed levels of FIB and observed
mobilization fractions greater than one. As indicated above, use of the pig slurry data in this
analysis is questionable because the pig slurry used in the EPA environmental monitoring
program had significantly lower counts of enterococci and E. coli than those reported in the
literature and is likely substantially different than the pig slurries used to estimate pathogen
abundances. For these reasons, we believe that the Approach 2 results are more robust.
Table 20. Relative QMRA illness risks from exposure to agricultural animal-impacted water
Data Used
Approach 1
Approach 2
Indicator
Enterococci
E. coli
Enterococci
E. coli
Fecal Source
Cattle manure
Pig slurry
Chicken litter
Cattle manure
Pig slurry
Chicken litter
Cattle manure
Pig slurry
Chicken litter
Cattle manure
Pig slurry
Chicken litter
5th %ile
9.1E-06
2.2E-01
7.8E-06
9.7E-08
1.2E-01
6.1E-04
2.6E-06
4.7E-06
4.0E-06
4.7E-05
1.2E-09
Illness Risks
10th %ile Median
1.9E-05 3.2E-04
3.3E-01 6.7E-01
2.1E-05 2.3E-03
2.3E-07 4.9E-06
2.0E-01 6.3E-01
1.7E-03 5.7E-02
8.0E-06 1.1E-03
Not conducted
1.4E-05 1.4E-03
9.6E-06 2.1E-04
9.1E-05 l.OE-03
6.7E-09 5.2E-06
90th %ile
5.8E-03
8.2E-01
6.8E-02
1.2E-04
8.1E-01
1.6E-01
8.0E-02
5.3E-02
5.4E-03
1.1E-02
4.5E-03
95th %ile
1.6E-02
8.4E-01
9.6E-02
3.4E-04
8.4E-01
1.8E-01
1.6E-01
8.0E-02
1.4E-02
2.0E-02
1.7E-02
December 2010
90
-------
U.S. Environmental Protection Agency
Probability of Illness
O-LjOiL/i-LuJtO^-'J,
« A ^^
O v
RWQC equivalent
~
0
-p
0 0
T
o
rO^ rO&* rO^ f cP* f cP* f cP* ^™ Cattle 6nt
«^* <-«^ <-«^ ^' •<*' *$' CZ3 swine ent
^ ^5> ^ ^jSf -OV5? vE^ „ „ , . ,
^o . ^ ^% O> v rV-° ^ / 1 chicken ent
^ (j!^0 ^H cattle ec
^^D swine ec
^^D chicken ec
Figure 30. Relative QMRA approach 1 probability of illness boxplot
Current G. mean
RWQC equivalent
v
I
o o
1 cattle enterococci
1 pig enterococci
] chicken enterococci
] cattle E co//
] pig E. co//
] chicken E. co//
Figure 31. Relative QMRA approach 2 probability of illness boxplot
December 2010
91
-------
U.S. Environmental Protection Agency
o
^
o
a
o
s
600 -,
500 -
400 -
300 -
200 -
100 -
Current geometric mean
RWQC equivalent
0 -
Cattle manure
Pig slurry
Chicken litter
-11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1
Log illness rates
Figure 32. Probability density for illness risks from E. coll relative QMRA approach 2
Inspection of the data presented in Table 20, Figure 30, Figure 31, and Figure 32 indicates the
following:
• The enterococci-based and E. co//'-based results for Approach 2 indicate that the
predicted median risk of illness from recreational exposure to the cattle-impacted
waterbody is 25 to 150-times lower than risk of illness associated with human sources of
contamination at the current geometric mean RWQC (1.1 x 10"3 and 2.1 x 10"4
respectively compared to 0.03).
• The E. coli -based results for Approach 2 indicate that the predicted median risk of illness
from recreational exposure to the pig-impacted waterbody is approximately 30-times
lower than the risk of illness associated with human sources of contamination at the
current geometric mean RWQC (1.0 x 10"3 compared to 0.03).
• The enterococci-based and E. co//'-based results for Approach 2 indicate that the
predicted median risk of illness from recreational exposure to the chicken-impacted
waterbody is approximately 20- to 5000-times lower than the risk of illness associated
with human sources of contamination at the current geometric mean RWQC (1.4 x 10"3
and 5.2 x 10"6, respectively compared to 0.03), depending on the FIB used.
December 2010
92
-------
U.S. Environmental Protection Agency
• The predicted variability is greatest for chicken-impacted water and least for pig-
impacted water.
• A small portion of the simulations resulted in risks that were greater than the current
geometric mean equivalent risk under Approach 2.
For cattle-impacted water, the enterococci-based results from Approach 1 are similar to those
from Approach 2. The Approach 1 E. co//-based results are lower than those from Approach 2
by two to three orders of magnitude. This result likely occurred because the E. coli densities
observed during the EPA environmental monitoring program were well above the literature-
based results.
For chicken-impacted water, the enterococci-based results from Approach 1 are similar to those
from Approach 2. The Approach 1 E. co//'-based results are less credible, as indicated above.
The predicted relative risks of illness are highly dependent on the FIB used. Relative QMRA
results are generally higher for enterococci than for E. coli. The median Approach 2 risks of
illness for cattle-impacted water based on enterococci are higher than those for E. coli by
approximately one order of magnitude. The median Approach 2 risks of illness for chicken-
impacted water based on enterococci are higher than those for E. coli by approximately two-and-
a-half orders of magnitude.
4.3. Discussion
4.3.1. Interpretation of results
The purpose of this QMRA was to estimate the human GI illness risk associated with recreation
at a freshwater beach impacted by fecal contamination from agricultural animal sources. Again,
the analysis addresses the following two questions: (1) What is the risk of illness associated with
recreation at a freshwater beach impacted by agricultural animal (cattle, swine, and chicken)
sources of fecal contamination?, and (2) How do those risks compare to risks associated with
freshwater beaches impacted by human sources?
Two complementary QMRA approaches were used. A traditional forward QMRA approach
characterizes the risk of illness associated with recreation at a freshwater beach impacted by
agricultural animal sources of fecal contamination. A relative QMRA compares the estimated
risks from recreation in water impacted by agricultural sources of fecal contamination to those
associated with human-impacted water.
The forward QMRA results estimate risk of illness in runoff within the context of the exposure
scenario evaluated. We made several simplifying assumptions to limit the scope of the exposure
scenario and ensure that the evaluation results would protect health relative to uninvestigated
conditions. Some of the most important assumptions were
December 2010 93
-------
U.S. Environmental Protection Agency
• primary contact recreation occurs at a freshwater beach adjacent to land that has fresh
cattle manure, pig slurry, or chicken litter applied at agronomic rates;
• GI illness is the health outcome of primary concern and GI illness rates are protective for
respiratory illness;
• ingestion of water is the primary exposure route of interest;
• FIB and pathogens reach the beach via runoff from an intense rainfall event;
• results of the EPA environmental monitoring program can be used to estimate
mobilization fractions of FIB and pathogens to the recreational water; and
• recreation at the assumed point of exposure is health protective compared to other
potential exposure points (downstream, diluted, or aged contamination scenarios).
The forward QMRA base analyses indicate that the predicted median risk of illness from
recreational exposure to the cattle-impacted waterbody during and immediately after an intense
rain event is effectively equivalent to the risk of illness associated with the current geometric
mean RWQC (USEPA, 1986). The predicted median risk of illness from recreational exposure
to the pig-impacted waterbody is approximately 4-times lower than the risk of illness that is
associated with the current geometric mean RWQC (i.e., 0.03/0.0076 = 4), and the predicted
median risk of illness from recreational exposure to the chicken-impacted waterbody is
approximately 300-times lower than the risk of illness associated with the current geometric
mean RWQC. E. coli O157 is the predicted dominant risk agent in cattle-impacted water,
followed by Campylobacter and Cryptosporidium. For pig-impacted water, Campylobacter and
Cryptosporidium are the predicted dominant risks agents, followed by Giardia. For chicken-
impacted water, Campylobacter is the predicted dominant risk agent. To anchor the results, we
compared these QMRA results to a summary of the literature on recreational water outbreaks
with animal-related sources (USEPA, 2009a). The outbreak literature indicates that the pathogen
source in the majority of recreational water-related outbreaks remains unknown. However there
are several examples of recreational water outbreaks where cattle were the principal source of
contamination (Cransberg et al., 1996; Feldman et al., 2002; Ihekweazu et al., 2006). In those
outbreaks, E. coli O157 was the etiologic agent, which is consistent with the QMRA results. No
outbreak reports are available for pig- or chicken-impacted waters.
For all three animal sources, there were combinations of model parameters resulted in predicted
risks that are substantially higher than the median risks (refer to Figure 25). At first glance, this
observation may appear to suggest that risks from animal-impacted waters may be of greater
concern than the median risk values suggest. However, this trend is not specific to agricultural-
animal impacted water, and in fact, the same observation may be made about predicted risks
from recreational exposure to pathogens in disinfected secondary effluent (Figure 33) (Seller et
December 2010 94
-------
U.S. Environmental Protection Agency
al., 2010a, 2010b).26 These high-risk parameter combinations warrant careful risk management
consideration, as they may represent specific environmental conditions under which the risk of
illness may be unacceptably high. Moreover, understanding the drivers of these high risk
conditions could provide opportunities for meaningful risk reductions. For instance, some
pathogens exhibit seasonality or life-cycle dependencies, which could be exploited by targeting
the high-prevalence time periods through best management practices.
The relative QMRA compares the estimated risks from recreation in water impacted by
agricultural sources of fecal contamination to those associated with human-impacted water. In
these analyses, we assume that sufficient agricultural animal-impacted runoff occurs so that the
freshwater beach contains geometric mean FIB densities (enterococci and E. colt) equivalent to
the current RWQC. In essence, this approach considers the relative level of risk from the various
fecal sources at a fixed level of FIB. By selecting the current geometric mean RWQC FIB levels
as the comparison point, risks in human impacted waters inherently serve as a reference, because
the current RWQC were established to provide a known level of public health protection in
human-impacted water (i.e., 8 cases of HCGI per 1000 recreation events, or in this risk
assessment, an equivalent risk of 30 cases of GI illness per 1000 recreation events).
Current geometric mean
RWQC equivalent
400 -
o
"3 300
£ 20
O
* 100 -
Median
risk
-7 -6 -5 -4 -3 -2 -1
Log1Q illness rate
Figure 33. Probability density for illness from recreation in disinfected secondary effluent
26 For disinfected secondary effluent, the density in the tail of the distribution likely occurs when noro virus densitites in the raw
wastewater are high and attenuation through wastewater treatment is low. Refer to Section 3.1.2 and Appendix B for further
information.
December 2010
95
-------
U.S. Environmental Protection Agency
The results from the relative QMRA reflect both specific conditions from the EPA
environmental monitoring study and average values that are substantially less dependent on the
FIB levels observed during the EPA studies. In general, we believe the most representative
results of the relative QMRA studies are those from Approach 2. The principal findings from the
Approach 2 relative QMRA are that at the current geometric mean RWQC:
• the predicted median risk of illness from recreational exposure to the cattle-impacted
waterbody is 25- to 150-times lower than the risk of illness associated with human
sources of contamination;
• the predicted median risk of illness from recreational exposure to the pig-impacted
waterbody is approximately 30-times lower than the risk of illness associated with human
sources of contamination; and
• the predicted median risk of illness from recreational exposure to the chicken-impacted
waterbody is approximately 20- to 5000-times lower than the risk of illness associated
with human sources of contamination.
The results from Approach 1 indicate clearly that FIB levels in fecal material from a specific
location can have a strong influence on the relative QMRA. In Approach 1, several
combinations of input parameters resulted in output that likely is not widely representative. This
set of conditions highlights the need to carefully and appropriately match FIB and pathogen
datasets in relative QMRA studies. In Approach 1, using pathogen abundance data in swine
slurry from the literature and FIB data from aged slurries resulted in output that diverged from
previous and current work.
The relative QMRA presented in this report uses the same methodology and set of peer-reviewed
literature that was developed to evaluate risks from exposure to recreational water impacted by
direct non-human contamination (Schoen and Ashbolt, 2010; Seller et al., 2010b). This work
extends those previous related QMRAs to include land application of fecal material, and FIB and
pathogen mobilization during and after rainfall events. These additions generally resulted in
decreased risks for cattle, but not for pigs or chicken.
The EPA environmental monitoring program data indicate that environmental conditions
underlying the data (nature, age, and level of treatment of the source material and levels of native
E. coli and enterococci in soil, etc.) can strongly influence FIB densities in the source material
(Table 18). We can only speculate on the extent to which the same conditions influence
reference pathogen densities; future monitoring and modeling studies could address this
knowledge gap.
In comparing our results from this risk assessment (Approach 2) to the results from the direct-
contamination scenario (Seller et al., 2010b), the risks associated with indirect contamination are
decreased for cattle and essentially unchanged for pig- and chicken-impacted water. This
December 2010 96
-------
U.S. Environmental Protection Agency
comparison is analogous to one that can be made for waters impacted by POTW effluent
compared to raw or poorly treated sewage. For the indirect contamination scenario, risks in
cattle-impacted water appear to be similar to those from pig or chicken-impacted waters. In the
case of direct contamination, risks in cattle-impacted waters are higher than those from pig or
chicken-impacted waters (Seller et al., 2010b).
Finally, similar to the forward QMRA results, combinations of model parameters for all three
animal sources result in predicted risks that are substantially higher than the median risks (Figure
32), highlighting the need for careful risk management of the types of conditions that could lead
to these high-risk outcomes.
4.3.2. Considerations and caveats
Like any scientific study, this work has a number of important considerations and conceptual
constraints. In this report, we compiled a vast range of disparate data and information to provide
an improved understanding about risks that would be difficult or impossible to characterize
through an observational study. Risk assessment is used in this way by governmental and
regulatory agencies worldwide to protect public health from exposure to a myriad of
contaminants through numerous routes of exposure (e.g., air pollution, food protection, drinking
water). To facilitate the conduct of this risk assessment, we necessarily made several
simplifying, health-protective assumptions to limit the scope of the assessment. In this regard,
several of the most important considerations and conceptual constraints are discussed below.
Exposure scenario is limited. The analyses only considered one exposure scenario, and which
was intentionally limited. Several important attributes of the exposure scenario might make it
difficult to extend the results from these analyses to a diverse range of recreational sites and
situations. The chain of events leading to human exposure to agricultural animal-derived
pathogens from recreational water is complex, and numerous processes can impact the predicted
risks. For example, manure handling practices before land application can greatly influence FIB
and pathogen levels in the land-applied material. This risk assessment evaluated the most
common minimum manure handling processes used in the United States; however, the pig slurry
FIB data from the EPA environmental monitoring program indicated the potential for substantial
variability. Similarly, BMPs (e.g., post-land application, pre-runoff) could greatly change the
abundance of FIB and pathogens in runoff, and dilution of the runoff water with uncontaminated
water would change the relative abundance of FIB and pathogens in recreational water.
Given the myriad exposure-related conditions that could reasonably occur in agricultural animal-
impacted water and the substantial variability related to exposure, we chose a relatively simple
and health protective exposure scenario for this analysis. If a more comprehensive exposure
model was implemented that included manure treatment, attenuation of pathogens and FIB prior
to and after runoff, and dilution of runoff water, the resulting forward QMRA would certainly
yield lower risk estimates. For example, our exposure scenario specifies recreation in undiluted
December 2010 97
-------
U.S. Environmental Protection Agency
runoff that is impacted by land applied manure at current U.S. agronomic rates. Risks would
certainly be lower in surface waters that are less impacted (due to less loading, dilution, or
attenuation due to die-off between runoff and recreation).
On the other hand, the impact of a more comprehensive model on the relative QMRA is less
certain. Pathogens and FIB are attenuated by the same processes, though by different amounts
and at different rates. This differential attenuation could produce fewer FIB than pathogens and
result in a higher risk associated with a given FIB level than if no treatment or best management
practices were undertaken. This possibility does not indicate that treatment or BMPs are
undesirable—just that the interpretation of FIB must consider all relevant processes in the
exposure profile.
We did not account for super-shedding exposure scenarios in this analysis (Arthur et al., 2009;
Chase-Topping et al., 2008). Risks to human health would be greater than those reported here if
super-shedding increased levels of pathogens in feces compared to a relatively constant level of
indicator data (see Annex 2). For example calves shed high levels of Cryptosporidium during
defined periods (Bryan et al., 2009; Chase-Topping et al., 2008). The general approach in this
QMRA could be used to evaluate specific conditions that could lead to higher risks such as
defined animal populations, soil types, rainfall patterns, dilution of receiving water, or the
presence of super-shedding animals.
FIB and pathogen mobilization. FIB and pathogen mobilization was modeled on a simulated
intense rain event in a single location (Georgia, USA). The experimental work produced the first
reported estimates for mobilization of both Campylobacter and Salmonella and valuable data for
assessing the runoff of the other organisms, but the extent to which the mobilization results apply
to other types of rain events at this location is not known. Furthermore, because soil
characteristics vary substantially across the United States, the mobilizations are likely specific to
the soil at the study location. FIB in soil can strongly influence mobilization rates, depending on
the relative levels of the FIB in the applied source material and FIB already present in the soil.
For example, during the EPA environmental monitoring study, the densities of enterococci and
E. coli in the pig slurry and chicken litter were relatively low (compared to levels reported in the
literature), but some levels of these FIB in the runoff were greater than the levels in the source
material. The predicted mobilization fraction of the FIB, therefore, was reported as greater than
one (logic value of 0), indicating that the FIB in the runoff water originated from the soil rather
than the source material. These observations have clear implications for interpreting FIB data
that are soil-based as compared to fecal source-based for agricultural animal-impacted water.
Furthermore, the prevalence of infection and the associated implications on pathogen abundance
in land-applied fecal source material is undoubtedly more complicated than this model addresses.
Because a given animal is either infected or not at any point in time, the variability in fecal
source abundance could be greater than this analysis suggests because we used average
abundances and assumed that at least one animal contributing to land-applied manure is shedding
December 2010 98
-------
U.S. Environmental Protection Agency
at any given time. A more rigorous characterization of this variability would mostly result in
lower risks, but would result in greater risk when infection is present. Identifying the frequency
and conditions leading to those high-risk periods could present opportunities for risk reductions.
Other zoonotic pathogens. Animal-impacted water could contain pathogens of public health
concern that were not evaluated and that might not fit into the FIB paradigm that is used to
regulate recreational water quality in the United States. We selected our reference pathogens
because they comprise an overwhelming proportion of all known pathogens that cause non-
foodborne illness in the United States (Mead et al., 1999), are representative of the fate and
transport of other pathogens of potential concern from a waterborne route of exposure (Ferguson
et al., 2009), are present in human and animal waste and recreational water (USEPA, 2009b),
possess the potential for extra-enteric survival, and have corresponding dose-response
relationships in the peer-reviewed literature (USEPA, 2010). The scientific understanding of
zoonotic pathogens is continually evolving, and based on evolving information, recreation in
agricultural animal-impacted water could cause illnesses that would otherwise be considered
rare. For example, Hepatitis E virus, Listeria monocytogenes, or Leptospira are pathogens that
are present in agricultural animal waste but which are thought to cause few illnesses from
recreational water exposure.
• Hepatitis E is a virus that can cause serious liver disease. Although, Hepatitis E is
uncommon in the United States,27 it is associated with livestock operations (Banks et al.,
2004; Legrand-Abravanel et al., 2009; Rutjes et al., 2009; Sinclair et al., 2009; Takahashi
et al., 2009). Further, the presence of Hepatitis E in pigs (Feagins et al., 2007; Meng et
al., 1999; Smith, 2001) and an emerging virus related to Hepatitis E in chickens
(Haqshenas et al., 2001) are of particular note.
• Listeria monocytogenes can cause a serious disease mainly in elderly persons, pregnant
women, newborns, and immunocompromised adults.28 Listeria monocytogenes is found
in soil and water, and animals can carry the bacterium without appearing ill. In the
United States, an estimated 2500 persons become seriously ill with listeriosis each year
(Mead et al., 1999).
• Leptospira occurs worldwide and is an important zoonosis, in part due to its prolonged
survival in water (Levett, 2001; Meites et al., 2004). Zoonotic reservoirs include
livestock (pigs and cattle), domestic pets (dogs), and wildlife (Levett, 2001). The source
of Leptospira infection in humans usually results from dermal contact with the urine of
an infected animal.
Therefore, if exposure to animal-impacted water was widespread, illnesses from non-reference
zoonotic pathogens could occur at higher rates than would otherwise be expected.
http://www.cdc. go v/hepatitis/HEV/index.htm
28 http://www.cdc. go v/nczved/divisions/dfbmd/diseases/listeriosis/
December 2010 99
-------
U.S. Environmental Protection Agency
Low probability events. In assessing the relative risks associated with fecal pollution sources,
we used median and other percentile values for describing risks. This choice is appropriate for
the purpose, but does not describe the risks associated with extreme, rare, or low probability
events. Although risks associated such types of extreme events are difficult to characterize, they
are important in the overall risk management context.
4.4. Conclusions
The risk assessment described in this report addresses two questions: (1) What is the risk of
illness associated with recreation at a freshwater beach impacted by agricultural animal (cattle,
swine, and chicken) sources of fecal contamination?; and (2) How do those risks compare to
risks associated with freshwater beaches impacted by human sources?
For our exposure scenario (runoff-induced pathogen mobilization from land-applied fecal
material) the median risk of illness from recreational exposure to the cattle-impacted waterbody
is equivalent to the risk of illness associated with the current (1986) geometric mean RWQC; the
median risk of illness from recreational exposure to the pig-impacted waterbody is
approximately four-times lower than the risk of illness associated with the current geometric
mean RWQC; and the median risk of illness from recreational exposure to the chicken-impacted
waterbody is approximately 300-times lower than the risk of illness associated with the current
geometric mean RWQC.
In comparing animal-impacted water to human-impacted water, the most representative results
come from literature-based FIB and pathogen abundances combined with mobilizations from the
EPA environmental monitoring program. These results indicate that at the current geometric
mean RWQC, the predicted median risk of illness from recreational exposure to each of the
animal-impacted water are at least 20 to 30-times lower than risk of illness associated with
human-impacted water. These risks are similar to or lower than those associated with direct
agricultural animal contamination.
Risk assessment is widely used by governmental and regulatory agencies worldwide to protect
public health from exposure to a myriad of contaminants through numerous routes of exposure.
Air pollution regulations, protection of the food supply chain, and drinking water regulations are
large-scale examples that illustrate the effective use of risk assessment methodologies within a
environmental regulatory context. To date, epidemiology studies have been the primary tool
used to characterize human health risks from exposure to recreational water. Those
epidemiology studies have generally focused on waters impacted by wastewater effluent (i.e.,
human sewage-impacted waters). Substantial progress has been made in improving the quality
of wastewater effluent in the United States in recent decades. Now more attention is being paid
to other sources of fecal contamination. In fact, non-point fecal contamination is one of the most
common reasons that waterbodies in the United States are classified as impaired with respect to
their use as recreational waters. Epidemiology studies are not likely to be effective in
December 2010 100
-------
U.S. Environmental Protection Agency
characterizing risks in many waters of this type due to technical, logistical and/or financial
constraints. As illustrated in this report, QMRA is a viable and valuable complement to
epidemiology studies for waters where epidemiology studies are not available, do not apply, or
are impractical. Finally, the data, results, and caveats of this study provide context for an
improved understanding of recreational risks in diverse waterbodies, and could help to facilitate
implementation of upcoming new or revised RWQC.
December 2010 101
-------
U.S. Environmental Protection Agency
References
Adam, R.D. (2001) Biology ofGiardia lamblia. Clinical Microbiology Reviews 14(3): 447-475.
Alley, M, Connolly, I, Fenwick, S., Mackereth, G., Leyland, M, Rogers, L., Haycock, M, Nicol, C., and Reed, C.
(2002) An epidemic of salmonellosis caused by Salmonella Typhimurium DT160 in wild birds and humans in New
Zealand. New Zealand Veterinary Journal 50(5): 170-176.
Allwood, P.B., Malik, Y.S., Maherchandani, S., Hedberg, C.W., and Goyal, S.M. (2005) Effect of temperature on
the survival of F-specific RNA coliphage, feline calicivirus, and Escherichia coli in chlorinated water. International
Journal of Environmental Research and Public Health 2(3-4): 442-446.
Arnone, R. (2005) Evaluating Cryptosporidium and Giardia in urban stormwater. In 2005 Watershed Management
Conference, July 19-22, 2005. Williamsburg, VA, United States: American Society of Civil Engineers, pp. 1153-
1154.
Arthur, T.M., Keen, J.E., Bosilevac, J.M., Brichta-Harhay, D.M., Kalchayanand, N., Shackelford, S.D., Wheeler,
T.L., Nou, X., and Koohmaraie, M. (2009) Longitudinal study of Escherichia coli O157:H7 in a beef cattle feedlot
and role of high-level shedders in hide contamination. Applied and Environmental Microbiology 75(20): 6515-6523.
Ashbolt, N.J., and Bruno, M. (2003) Application and refinement of the WHO risk framework for recreational waters
in Sydney, Australia. Journal of Water and Health 1(3): 125-131.
Atwill, E.R., Hoar, B., das Gracas Cabral Pereira, M., Tate, K.W., Rulofson, F., and Nader, G. (2003) Improved
quantitative estimates of low environmental beading and sporadic periparturient shedding of Cryptosporidium
parvum in adult beef cattle. Applied and Environmental Microbiology 69(8): 4604-4610.
Atwill, E.R., Pereira, M.D.G.C., Alonso, L.H., Elmi, C., Epperson, W.B., Smith, R., Riggs, W.A., Carpenter, L.V.,
Dagartz, D.A., and Hoar, B. (2006) Environmental load of Cryptosporidium parvum oocysts from cattle manure in
feedlots from the central and western United States. Journal of Environmental Quality 35(1): 200-206.
Balbus, I, Parkin, R., Makri, A., Ragain, L., Embrey, M., and Hauchman, F. (2004) Defining susceptibility for
microbial risk assessment: results of a workshop. Risk Analysis 24(1): 197-208.
Balbus, J.M., and Embrey, M. A. (2002) Risk factors for waterborne enteric infections. Current Opinion in
Gastroenterology 18(1): 46-50.
Banks, M., Bendall, R., Grierson, S., Heath, G., Michell, I, andDalton, H. (2004) Human and porcine hepatitis E
virus strains, United Kingdom. Emerging Infectious Diseases 10(5): 953-955.
Bates, I, Goddard, M.R., and Butler, M. (1984) The detection of rotaviruses in products of wastewater treatment.
Journal of Hygiene 93(3): 639-643.
Benham, B.L., Baffaut, C., Zeckoski, R.W., Mankin, K.R., Pachepsky, Y.A., Sadeghi, A.M., Brannan, KM., Soupir,
M.L., and Habersack, M. J. (2006) Modeling bacteria fate and transport in watersheds to support TMDLs.
Transactions of'the ASABE 49(4): 987-1002.
Berg, R. (2008) The Alamosa Salmonella outbreak: a gumshoe investigation. Journal of Environmental Health
71(2): 54-58.
Berg, R.W., and Anderson, A.W. (1972) Salmonellae and Edwardsiella tarda ingullfeces: a source of
contamination in fish processing plants. Applied and Environmental Microbiology 24(3): 501-503.
December 2010 102
-------
U.S. Environmental Protection Agency
Berry, E.D., Wells, J.E., Archibeque, S.L., Ferrell, C.L., Freetly, H.C., and Miller, D.N. (2006) Influence of
genotype and diet on steer performance, manure odor, and carriage of pathogenic and other fecal bacteria. II.
Pathogenic and other fecal bacteria. Journal of Animal Science 84(9): 2523-2532.
Berry, E.D., Woodbury, B.L., Nienaber, J.A., Eigenberg, R.A., Thurston, J.A., and Wells, IE. (2007) Incidence and
Persistence of Zoonotic Bacterial and Protozoan Pathogens in a Beef Cattle Feedlot Runoff Control Vegetative
Treatment System Journal of Environmental Quality 36(6): 1873-1882.
Besser, T.E., LeJeune, J.T., Rice, D.H., Berg, I, Stilborn, R.P., Kaya, K., Bae, W., and Hancock, D., D. (2005)
Increasing prevalence of Campylobacter jejuni in feedlot cattle through the feeding period. Applied and
Environmental Microbiology 71(10): 5752-5758.
Betancourt, W.Q., and Rose, J.B. (2005) Microbiological assessment of ambient waters and proposed water sources
for restoration of a Florida wetland. Journal of Water and Health 3(2): 89-100.
Bicudo, J., and Goyal, S. (2003) Pathogens and manure management systems: a review. Environmental Technology
24(1): 115-130.
Bielaszewska, M., Janda, J., Blahova, K., Minarikova, H., Jikova, E., Karmali, M.A., Laubova, J., Sikulova, J.,
Preston, M.A., Khakhria, R., Karch, H., Klazarova, H., andNyc, O. (1997) HumanEscherichia coli O157:H7
infection associated with the consumption of unpasteurized goat's milk. Epidemiology and Infection 119(3): 299-
305.
Black, R.E., Levine, M.M., Clements, M.L., Hughes, T.P., and Blaser, M.J. (1988) Experimental Campylobacter
jejuni infection in humans. Journal of Infectious Diseases 157(3): 472-479.
Blaser, M.J., Hardesty, H.L., Powers, B., and Wang, W.-L.L. (1980) Survival of Campylobacter fetus subsp. jejuni
in bio logical milieus. Journal of Clinical Microbiology 11(4): 309-313.
Boehm, A.B., Grant, S.B., Kim, J.H., Mowbray, S.L., McGee, C.D., Clark, C.D., Foley, D.M., and Wellman, D.E.
(2002) Decadal and shorter period variability of surf zone water quality at Huntington Beach, California.
Environmental Science and Technology 36(18): 3885-3892.
Boehm, A.B. (2007) Enterococci concentrations in diverse coastal environments exhibit extreme variability.
Environmental Science and Technology 41(24): 8227-8232.
Boes, J., Alban, L., Bagger, J., Mogelmose, V., Baggesen, D.L., and Olsen, IE. (2005) Survival of Escherichia coli
and Salmonella Typhimurium in slurry applied to clay soil on a Danish swine farm. Preventive Veterinary Medicine
69(3-4): 213-228.
Bofill-Mas, S., Albinana-Gimenez, N., Clemente-Casares, P., Hundesa, A., Manzano-Rodriguez, I, Allard, A.,
Calvo, M., and Girones, R. (2006) Quantification and stability of human adenovirus and polyomavirus JCPyV in
wastewater matrices. Applied and Environmental Microbiology 72(12): 7894-7896.
Bollaerts, K., Aerts, M., Faes, C., Grijspeerdt, K., Dewulf, I, and Mintiens, K. (2008) Human salmonellosis:
estimation of dose-illness from outbreak data. Risk Analysis 28(2): 427-440.
Bonadonna, L., Briancesco, R., Ottaviani, M., and Veschetti, E. (2002) Occurrence of Cryptosporidium oocysts in
sewage effluents and correlation with microbial, chemical and physical water varaibles. Environmental Monitoring
and Assessment 75: 241-252.
Boone, S.A., and Gerba, C.P. (2007) Significance of fomites in the spread of respiratory and enteric viral disease.
Applied and Environmental Microbiology 73(6): 1687-1696.
December 2010 103
-------
U.S. Environmental Protection Agency
Bradford, S.A., and Schijven, J. (2002) Release of Cryptosporidium and Giardia from dairy calf manure: impact of
solution salinity. Environmental Science and Technology 36: 3916-3923.
Bradford, S.A., Segal, E., Zheng, W., Wang, Q., and Hutchins, S.R. (2008) Reuse of concentrated animal feeding
operation wastewater on agricultural lands. Journal of Environmental Quality 37(5 Suppl): S97-S115.
Brooks, J.P., Adeli, A., Read, J.J., and McLaughlin, M.R. (2009) Rainfall simulation in greenhouse microcosms to
assess bacterial-associated runoff from land-applied poultry litter. Journal of Environmental Quality 38(1): 218-229.
Bryan, B.A., Kandulu, J., Deere, D.A., White, M, Frizenschaf, J., and Grossman, N.D. (2009) Adaptive
management for mitigating Cryptosporidium risk in source water: a case study in an agricultural catchment in South
Australia. Journal of Environmental Management 90(10): 3122-3134.
Bukhari, Z., Smith, H.V., Sykes, N., Humphreys, S.W., Paton, C.A., Girdwood, R.W.A., and Fricker, C.R. (1997)
Occurrence of Cryptosporidium spp. oocysts and Giardia spp. cysts in sewage influents and effluents from
treatment plants in England. Water Science and Technology 35(11-12): 385-390.
Bull, S.A., Allen, V.M., Domingue, G., Jorgensen, F., Frost, J.A., Ure, R., Whyte, R., Tinker, D., Corry, J.E.L.,
Gillard-King, J., and Humphrey, T.J. (2006) Sources of Campylobacter spp. colonizing housed broiler flocks during
rearing. Applied and Environmental Microbiology 72(1): 645-652.
Butler, D.M., Franklin, D.H., Cabrera, M.L., Tasistro, A.S., Xia, K., and West, L.T. (2008) Evaluating aeration
techniques for decreasing phosphorus export from grasslands receiving manure. Journal of Environmental Quality
37(3): 1279-1287.
Butterfield, J., Coulson, J., Kearsey, S., Monaghan, P., McCoy, J., and Spain, G. (1983) The herring gallLarus
argentatus as a carrier of salmonella. J ournal of Hygiene (London) 91(3): 429-436.
Byrd (1998) Horizontal transmission of Salmonella typhimurium in broiler chicks. Journal of Applied Poultry
Research 7(1): 75-80.
Callaway, T.R., Keen, J.E., Edrington, T.S., Baumgard, L.H., Spicer, L., Fonda, E.S., Griswold, K.E., Overton, T.R.,
VanAmburgh, M.E., Anderson, R.C., Genovese, K.J., Poole, T.L., Harvey, R.B., and Nisbet, D.J. (2005) Fecal
prevalence and diversity of Salmonella species in lactating dairy cattle in four states. Journal of Dairy Science
88(10): 3603-3608.
Callaway, T.R., Edrington, T.S., Anderson, R.C., Byrd, J.A., and Nisbet, D.J. (2008) Gastrointestinal microbial
ecology and the safety of our food supply as related to Salmonella Journal of Animal Science 86(E. Supplement):
E163-E172.
Carducci, A., Morici, P., Pizzi, F., Battistini, R., Rovini, E., and Verani, M. (2008) Study of the vrial removal
efficiency in an urban wastewater treatment plant. 58(4): 893-897.
Carraro, E., Fea, E., Salva, S., and Gilli, G. (2000) Impact of a wastewater treatment plant on Cryptosporidium
oocysts and Giardia cysts occurring in a surface water. Water Science and Technology 41: 31-37.
Casanovas, L., de Simon, M., Ferrer, M., Arques, J., andMonzon, G. (1995) Intestinal carriage of campylobacters,
salmonellas, yersinias and listerias in pigeons in the city of Barcelona. J ournal ofApplied Bacteriology 78(1): 11-13.
Castro-Hermida, J.A., Garcia-Presedo, L, Almeida, A., Gonzalez-Warleta, M., Correia Da Costa, J.M., and Mezo,
M. (2008) Contribution of treated wastewater to the contamination of recreational river areas with Cryptosporidium
spp. and Giardia duodenalis. Water Research 42: 3528-3538.
CDC (2006). Salmonella annual summary. URL www.cdc.gov/ncidod/dbmd/phlisdata/salmonella. htm
December 2010 104
-------
U.S. Environmental Protection Agency
Chalmers, R.M., and Giles, M. (2010) Zoonotic cryptosporidiosis in the UK - challenges for control. Journal of
Applied Microbiology 109(5): 1487-1497.
Chapman, P.A., Siddons, C.A., Cerdan Malo, A.T., and Harkin, M.A. (1997) A 1-year study of Escherichia coli
O157 in cattle, sheep, pigs and poultry. Epidemiology and Infection 119: 245-250.
Chase-Topping, M., Gaily, D., Low, C., Matthews, L., and Woolhouse, M. (2008) Super-shedding and the link
between human infection and livestock carriage of Escherichia coli O157. Nature Reviews. Microbiology 6(12):
904-912.
Chen, L., Geys, H., Cawthraw, S., Havelaar, A., and Teunis, P. (2006) Dose-response for infectivity of several
strains ofCampylobacterjejuni in chickens. Risk Analysis 26(6): 1613-1621.
Cheng, H.-W.A., Lucy, F.E., Graczyk, T.K., Broaders, M.A., Tamang, L., and Connolly, M. (2009) Fate of
Cryptosporidium parvum and Cryptosporidium hominis oocysts and Giardia duodenalis cysts during secondary
wastewater treatments. Parasitology Research 105: 689-696.
Chern, E.G., Tsai, Y.-L., and Olson, B.H. (2004) Occurrence of genes associated with enterotoxigenic and
enterohemorrhagic Escherichia coli in agricultural waste lagoons. Applied and Environmental Microbiology 70(1):
356-362.
Cizek, A., Literak, I., and Scheer, P. (2000) Survival of Escherichia coli O157 in faeces of experimentally infected
rats and domestic pigeons. Letters in Applied Microbiology 31(5): 349-352.
Cizek, A.R., Characklis, G.W., Krometis, L.-A., Hayes, J.A., Simmons, O.D.I., DiLonardo, S., Alderisio, K.A., and
Sobsey, M.D. (2008) Comparing the partitioning behavior of Giardia and Cryptosporidium with that of indicator
organisms in stormwater runoff. Water Research 42(17): 4421-4438.
Claudon, D.G., Thompson, D.I., Christenson, E.H., Lawton, G.W., and Dick, E.G. (1971) Prolonged Salmonella
contamination of a recreational lake by runoff waters. Applied Microbiology 21(5): 875-877.
Coelho, B., B.R., Roy, R.C., Topp, E., and Lapen, D.R. (2007) Tile water quality following liquid swine manure
application into standing corn. Journal of Environmental Quality 36: 580-587.
Coleman, M.E., and Marks, H.M. (1998) Topics in dose-response modeling. Journal of Food Protection 61(11):
1550-1559.
Coleman, M.E., and Marks, H.M. (2000) Mechanistic modeling of salmonellosis. Quantitative Microbiology 2: 227-
247.
Colford, J.M., Rees, J.R., Wade, T.J., Khalakdina, A., Hilton, J.F., Ergas, I.J., Burns, S., Benker, A., Ma, C., Bowen,
C., Mills, D.C., Vugia, D.J., Juranek, D.D., and Levy, D.A. (2002) Participant blinding and gastrointestinal illness in
a randomized, controlled trial of an in-home drinking water intervention. Emerging Infectious Diseases 8(1): 29-36.
Colford, J.M., Jr., Wade, T.J., Schiff, K.C., Wright, C.C., Griffith, J.F., Sandhu, S.K., Burns, S., Sobsey, M.,
Lovelace, G., and Weisberg, S.B. (2007) Water quality indicators and the risk of illness at beaches with nonpoint
sources of fecal contamination. Epidemiology 18(1): 27-35.
Collins, R., Donnison, A., Ross, C., and McLeod, M. (2004) Attenuation of effluent-derived faecal microbes in grass
buffer strips. New Zealand Journal of Agricultural Research 47(4): 565 -574.
Collins, R., Elliott, S., and Adams, R. (2005) Overland flow delivery of faecal bacteria to a headwater pastoral
stream. Journal of Applied Microbiology 99(1): 126-132.
December 2010 105
-------
U.S. Environmental Protection Agency
Cornelius, L. (1969) Field notes on salmonella infection in greenfinches and house sparrows. Journal of Wildlife
Diseases 5(3): 142-143.
Cornick, N.A., and Helgerson, A.F. (2004) Transmission and infectious dose of Escherichia coli O157:H7 in swine.
Applied and Environmental Microbiology 70(9): 5331-5335.
Couch, R.B., Cate, T.R., Douglas, R.G.J., Gerone, P.I, and Knight, V. (1966) Effect of route of inoculation on
experimental respiratory viral disease in volunteers and evidence for airborne transmission. Bacteriological Reviews
30(3): 517-529.
Couch, R.B., Knight, V., Douglas, R.G.J., Black, S.H., andHamory, B.H. (1969) The minimal infectious dose of
adenovirus type 4; the case for natural transmission by viral aerosol. Transactions of the American Clinical and
Climatological Association 80: 205-211.
Cox, N.A., Stern, N.J., Musgrove, M.T., Bailey, J.S., Craven, S.E., Cray, P.P., Buhr, R.J., and Hiett, K.L. (2002)
Prevalence and level of Campylobacter in commercial broiler breeders (parents) and broilers. Journal of Applied
Poultry Research 11(2): 187-190.
Crabtree, K.D., Gerba, C.P., Rose, J.B., and Haas, C.N. (1997) Waterborne adenovirus: a risk assessment. Water
Science and Technology 35(11-12): 1-6.
Cransberg, K, van den Kerkhof, J.H., Banffer, J.R., Stijnen, C., Wernars, K, van de Kar, N.C., Nauta, I, and Wolff,
E.D. (1996) Four cases of hemolytic uremic syndrome—source contaminated swimming water? Clinical
Nephrology 46(1): 45-49.
Craun, G.F., Calderon, R.L., and Craun, M.F. (2005) Outbreaks associated with recreational water in the United
States. International Journal of Environmental Health Research 15(4): 243-262.
Cruickshank, J.C., and Smith, H.W. (1949) Isolation of Salmonellae from dogs, cats, and pigeons. British Medical
Journal 2(4639): 1254-1258.
Curriero, F.C., Patz, J.A., Rose, J.B., and Lele, S. (2001) The association between extreme precipitation events and
waterborne disease outbreaks in the United States, 1984-1994. American Journal of Public Health 91(8): 1194-
1199.
Davies, C.M., Ferguson, C.M., Kaucner, C., Krogh, M, Altavilla, N., Deere, D.A., and Ashbolt, N.J. (2004)
Dispersion and transport of Cryptosporidium oocysts from fecal pats under simulated rainfall events. Applied and
Environmental Microbiology 70(2): 1151-1159.
Degremont, A., Starchier, D., Wolfensberger, E., and Osterwalder, B. (1981) Clinical and therapeutic study of a
group of 217 patients with intestinal giardiasis and amebiasis [in German]. SchweizMed Wochenschr 111(52):
2039-2046.
Desmonts, C., Minet, I, Colwell, R., and Cormier, M. (1990) Fluorescent-anitbody method useful for detecting
viable but non-culturable Salmonella spp. in chlorinated wastewater. Applied and Environmental Microbiology
56(5): 1448-1452.
Devane, M.L., Nicol, C., Ball, A., Klena, J.D., Scholes, P., Hudson, J.A., Baker, M.G., Gilpin, B.J., Garrett, N., and
Savill, M.G. (2005) The occurrence of Campylobacter subtypes in environmental reservoirs and potential
transmission routes. Journal of Applied Microbiology 98: 980-990.
Doane, C.A., Pangloli, P., Richards, H.A., Mount, J.R., Golden, D.A., and Draughon, F.A. (2007) Occurrence of
Escherichia coli O157:H7 in diverse farm environments. Journal of Food Protection 70(1): 6-10.
Dolin, R. (2007) Norovirus - vhallenges to control. The New England Journal of Medicine 357(11): 1072-1073.
December 2010 106
-------
U.S. Environmental Protection Agency
Dorevitch, S., Ashbolt, N.J., Ferguson, C.M., Fujioka, R., McGee, CD., Seller, J.A., and Whitman, R.L. (2010)
Meeting report: knowledge and gaps in developing microbial criteria for inland recreational waters. Environmental
Health Perspectives 118(6): 871-876.
Dorner, S.M., Huck, P.M., and Slawson, R.M. (2004) Estimating potential environmental loadings of
Cryptosporidium spp. and Campylobacter spp. from livestock in the Grand River Watershed, Ontario, Canada.
Environmental Science and Technology 38: 3370-3380.
Dorr, P.M., Tadesse, D.A., Zewde, B.M., Fry, P., Thakur, S., and Gebreyes, W.A. (2009) Longitudinal study of
Salmonella dispersion and the role of environmental contamination in commercial swine production systems.
Applied and Environmental Microbiology 75(6): 1478-1486.
Dufour, A.P., Evans, O., Behymer, T., and Cantu, R. (2006) Water ingestion during swimming activities in a pool: a
pilot study. Journa of Water and Health 4(4): 425-430.
Duncan, R., Stroud, R., and Locke, L. (1983) Salmonella enteritidis isolated from an eared grebe (Podiceps
nigricollis). Journal of Wildlife Diseases 19(1): 63-64.
DuPont, H.L., Chappell, C.L., Sterling, C.R., Okhuysen, P.C., Rose, J.B., and Jakubowski, W. (1995) The
infectivity of Cryptosporidium parvum in healthy volunteers. New England Journal of Medicine 332(13): 855-859.
Ebel, E.D., David, M.J., and Mason, J. (1992) Occurrence of Salmonella enteriditis in the U.S. commerical egg
industry: report on a national spent hen survey. Avian Diseases 36: 646-654.
Edrington, T.S., Hume, M.E., Looper, M.L., Schultz, C.L., Fitzgerald, A.C., Callaway, T.R., Genovese, K.J.,
Bischoff, K.M., McReynolds, J.L., Anderson, R.C., and Nisbet, D.J. (2004) Variation in the faecal shedding of
Salmonella and E. coli O157:H7 in lactating dairy cattle and examination of Salmonella genotypes using pulsed-
field gel electrophoresis. Letters in Applied Microbiology 38(5): 366-372.
Eisenberg, J.N.S., Seto, E.Y.W., Olivieri, A.W., and Spear, R.C. (1996) Quantifying water pathogen risk in an
epidemic logical framework. Risk Analysis 16(4): 549-563.
Eisenberg, J.N.S., Seto, E.Y.W., Colford, J.M., Olivieri, A.W., and Spear, R.C. (1998) An analysis of the
Milwaukee crypto sporidio sis outbreak based on a dynamic model of the infection process. Epidemiology 9(3): 255-
263.
Eisenberg, J.N.S., Seller, J.A., Scott, J., Eisenberg, D.M., and Colford, J.M.J. (2004) A dynamic model to assess
microbial health risks associated with beneficial uses of biosolids. Risk Analysis 24(1): 221-236.
Eisenberg, J.N.S., Moore, K., Seller, J.A., Eisenberg, D., and Colford, J.M.J. (2008) Microbial risk assessment
framework for exposure to amended sludge projects. Environmental Health Perspectives 116(6): 727-733.
El-Shibiny, A., Connerton, P.L., and Connerton, I.F. (2005) Enumeration and Diversity ofCampylobacters and
Bacteriophages Isolated during the Rearing Cycles of Free-Range and Organic Chickens. Applied and
Environmental Microbiology 71(3): 1259-1266.
Ellis-Iversen, J., Cook, A., Smith, R., Pritchard, G., and Nielen, M. (2009) Temporal patterns and risk factors for
Escherichia coli O157 and Campylobacter spp, in young cattle. Journal of Food Protection 72(3): 490-496.
Elmir, S.M., Wright, M.E., Abdelzaher, A., Solo-Gabriele, H.M., Fleming, L.E., Miller, G., Rybolowik, M., Peter
Shih, M.T., Pillai, S.P., Cooper, J. A., and Quaye, E. A. (2007) Quantitative evaluation of bacteria released by bathers
in a marine water. Water Research 41(1): 3-10.
Englehardt, J., Swartout, J., and Loewenstine, C. (2009) A new theoretical discrete growth distribution with
verification for microbial counts in water. Risk Analysis 29(6): 841-856.
December 2010 107
-------
U.S. Environmental Protection Agency
Englehardt, J.D., and Swartout, J. (2004) Predictive population dose-response assessment for Cryptosporidium
parvum: infection endpoint. Journal of Toxicology and Environmental Health-Part A-Current Issues 67(8-10): 651-
666.
Payer, R., Trout, J.M., Graczyk, T.K., and Lewis, E.J. (2000) Prevalence of Cryptosporidium, Giardia andEimeria
infections in post-weaned and adult cattle on three Maryland farms. Veterinary Parasitology 93: 103-112.
Feagins, A.R., Opriessnig, T., Guenette, D.K., Halbur, P.O., andMeng, X.J. (2007) Detection and characterization
of infectious Hepatitis E virus from commercial pig livers sold in local grocery stores in the USA. Journal of
General Virology 88(Pt 3): 912-917.
Feder, I.E., Wallace, M, Gray, J.T., Fratamico, P., Fedorka-Cray, P.J., Pearce, R., Call, IE., Perrine, R., and
Luchansky, J.B. ( 2003) Isolation of Escherichia coli O157:H7 from intact colon fecal samples of swine. Emerging
and Infectious Diseases 9: 380-383.
Fegan, N., Vanderlinde, P., Higgs, G., and Desmarchelier, P. (2004) Quantification and prevalence of Salmonella in
beef cattle presenting at slaughter. Journal of Applied Microbiology 97: 892-898.
Feldman, K.A., Mohle-Boetani, J.C., Ward, J., Furst, K., Abbott, S.L., Ferrero, D.V., Olsen, A., and Werner, S.B.
(2002) A cluster of Escherichia coli O157: nonmotile infections associated with recreational exposure to lake water.
Public Health Reports 117(4): 380-385.
Fenlon, D.R. (1981) Seagulls (Lams spp.) as vectors of salmonellae: an investigation into the range of serotypes and
numbers of salmonellae in gull faeces. Journal ofHygeine 86(2): 195-202.
Fenlon, D.R. (1983) A comparison of Salmonella serotypes found in the faeces of gulls feeding at a sewage works
with serotypes present in the sewage. Journal ofHygeine 91(1): 47-52.
Ferguson, C.M., Davies, C.M., Kaucner, C., Krogh, M., Rodehutskors, J., Deere, D.A., and Ashbolt, N.J. (2007)
Field scale quantification of microbial transport from bovine faeces under simulated rainfall events. Journal of
Water and Health 5(1): 83-95.
Ferguson, C.M., Charles, K., and Deere, D.A. (2009) Quantification of microbial sources in drinking-water
catchments. Critical Reviews in Environmental Science and Technology 39(1): 1-40.
Fleisher, J.M., Kay, D., Salmon, R.L., Jones, F., Wyer, M.D., and Godfree, A.F. (1996) Marine waters contaminated
with domestic sewage: nonenteric illnesses associated with bather exposure in the United Kingdom. American
Journal of Public Health 86(9): 1228-1234.
Fleisher, J.M., Fleming, L.E., Solo-Gabriele, H.M., Kish, J.K, Sinigalliano, C.D., Piano, L., Elmir, S.M., Wang,
J.D., Withum, K, Shibata, T., Gidley, M.L., Abdelzaher, A., He, G., Ortega, C., Zhu, X., Wright, M., Hollenbeck,
J., and Backer, L.C. (2010) The BEACHES Study: health effects and exposures from non-point source microbial
contaminants in subtropical recreational marine waters. International Journal of Epidemiology.
10.1093/ije/dyq084dyq084.
Foley, S.L., Lynne, A.M., and Nayak, R. (2008) Salmonella challenges: prevalence in swine and poultry and
potential pathogenicity of such isolates. Journal of Animal Science 86(14 suppl): E149-E162.
Fong, T.T., Phanikumar, M.S., Xagoraraki, L, and Rose, J.B. (2010) Quantitative detection of human adenoviruses
in wastewater and combined sewer overflows influencing a Michigan river. Applied and Environmental
Microbiology 76(3): 715-723.
December 2010 108
-------
U.S. Environmental Protection Agency
Fossler, C.P., Wells, S.J., Kaneene, J.B., Ruegg, P.L., Warnick, L.D., Eberly, L.E., Godden, S.M., Halbert, L.W.,
Campbell, A.M., Bonn, C.A., and Geiger Zwald, A.M. (2005) Cattle and environmental sample-level factors
associated with the presence of Salmonella in a multi-state study of conventional and organic dairy farms.
Preventive Veterinary Medicine 67: 39-53.
Foster, G., Evans, I, Knight, H.I., Smith, A.W., Gunn, G.J., Allison, L.J., Synge, B.A., and Pennycott, T.W. (2006)
Analysis of feces samples collected from a wild-bird garden feeding station in Scotland for the presence of
verocytotoxin-producing Escherichia coli O157. Applied and Environmental Microbiology 72(3): 2265-2267.
Fricker, C. (1984) A note on salmonella excretion in the black headed gull (Lams ribibundus) feeding at sewage
treatment works. Journal of Applied Bacteriology 56(3): 499-502.
Garber, L.P., Smeltzer, M., Fedorka-Cray, P., Ladely, S., and Ferris, K. (2003) Salmonella enterica serotype
enteritidis in table egg layer house environments and in mice in U.S. layer houses and associated risk factors. Avian
Diseases 47(1): 134-142.
Garcia-Aljaro, C., Bonjoch, X., and Blanch, A.R. (2005) Combined use of an immunomagnetic separation method
and immunoblotting for the enumeration and isolation of Escherichia coli O157 in wastewaters. Journal of Applied
Microbiology 98(3): 589-597.
Garcia-Aljaro, C., Muniesa, M., Jofre, I, and Blanch, A.R. (2004) Prevalence of the stx2 gene in coliform
populations from aquatic environments. Applied and Environmental Microbiology 70(6): 3535-3540.
Gargiulo, G., Bradford, S.A., Simunek, I, Ustohal, P., Vereecken, H., andKlumpp, E. (2008) Bacteria transport and
deposition under unsaturated flow conditions: the role of water content and bacteria surface hydrophobicity. Vadose
Zone Journal 7(2): 406-419.
Garrett, N., Devane, M.L., Hudson, J.A., Nicol, C., Ball, A., Klena, J.D., Baker, M.G., Gilpin, B.J., and Savill, M.G.
(2007) Statistical comparison ofCampylobacterjejuni subtypes from human cases and environmental sources.
Journal of Applied Microbiology 103(6): 2113-2121.
Gerba, C.P., Rose, J.B., and Haas, C.N. (1996) Sensitive populations: who is at the greatest risk? International
Journal of Food Microbiology 30(1-2): 113-123.
Girdwood, R., Fricker, C., Munro, D., Shedden, C., and Monaghan, P. (1985) The incidence and significance of
salmonella carriage by gulls (Lams spp.) in Scotland. Journal of Hygiene (London) 95(2): 229-241.
Goss, M., and Richards, C. (2008) Development of a risk-based index for source water protection planning, which
supports the reduction of pathogens from agricultural activity entering water resources. Journal of Environmental
Management S'l'(4): 623-632.
Graczyk, T.K., Payer, R., Trout, J.M., Lewis, E.J., Farley, C.A., Sulaiman, I., and Lai, A.A. (1998) Giardia sp. cysts
and infectious Cryptosporidium parvum oocysts in the feces of migratory Canada geese (Branta canadensis).
Applied and Environmental Microbiology 64(7): 2736-2738.
Graczyk, T.K., Majewska, A.C., and Schwab, K.J. (2008) The role of birds in dissemination of human waterborne
pathogens. Trends in Parasitology 24(2): 55-59.
Grant, S.B., Pendroy, C.P., Mayer, C.L., Bellin, J.K., and Palmer, C.J. (1996) Prevalence of enterohemorrhagic
Escherichia coli in raw and treated municipal sewage. Applied and Environmental Microbiology 62(9): 3466-3469.
Gronewold, A.D., Borsuk, M.E., Wolpert, R.L., and Reckhow, K.H. (2008) An assessment of fecal indicator
bacteria-based water quality standards. Environmental Science and Technology 42(13): 4676-4682.
December 2010 109
-------
U.S. Environmental Protection Agency
Gronewold, A.D., and Wolpert, R.L. (2008) Modeling the relationship between most probable number (MPN) and
colony-forming unit (CPU) estimates of fecal conform concentration. Water Research 42(13): 3327-3334.
Guber, A.K., Shelton, D.R., and Pachepsky, Y.A. (2005) Effect of manure onEscherichia coli attachment to soil.
Journal of Environmental Quality 34(6): 2086-2090.
Guber, A.K., Shelton, D.R., Pachepsky, Y.A., Sadeghi, A.M., and Sikora, L.J. (2006) Rainfall-induced release of
fecal conforms and other manure constituents: comparison and modeling. Applied and Environmental Microbiology
72(12): 7531-7539.
Guber, A.K., Karns, IS., Pachepsky, Y.A., Sadeghi, A.M., VanKessel, J.S., andDao, T.H. (2007a) Comparison of
release and transport of manure-borne Escherichia coli and enterococci under grass buffer conditions. Letters in
Applied Microbiology 44(2): 161 -167.
Guber, A.K., Pachepsky, Y.A., Shelton, D.R, and Yu, O. (2007b) Effect of bovine manure on fecal conform
attachment to soil and soil particles of different sizes. Applied and Environmental Microbiol ogy 73(10): 3363-3370.
Guzman, J.A., Fox, G., and Perm, C. (2009) Sorption of E. coli from liquid swine manure in natural and artificial
soils. In 2009 ASABE Annual International Meeting. ASABE (ed). Reno, Nevada.
Haas, C.N., Rose, J.B., Gerba, C., and Regli, S. (1993) Risk assessment of virus in drinking water. Risk Analysis
13(5): 545-552.
Haas, C.N., Crockett, C.S., Rose, J.B., Gerba, C.P., and Fazil, A.M. (1996) Assessing the risks posed by oocysts in
drinking water. Journal of the American Water Works Association 88(9): 131-136.
Haas, C.N., Rose, J.B., and Gerba, C.P. (1999) Quantitative Micmbial Risk Assessment (eds), J.W. Wiley, Inc.,
Hakkinen, M., and Hanninen, M.-L. (2009) Shedding of Campylobacter spp. in Finnish cattle on dairy farms.
Journal of Applied Microbiology 107(3): 898-905.
Haley, B.J., Cole, D.J., andLipp, E.K. (2009) Distribution, diversity, and seasonally of waterborne Salmonellae in a
rural watershed. Applied and Environmental Microbiology 75(5): 1248-1255.
Haqshenas, G., Shivaprasad, H.L., Woolcock, P.R., Read, D.H., andMeng, X.J. (2001) Genetic identification and
characterization of a novel virus related to human hepatitis E virus from chickens with hepatitis-splenomegaly
syndrome in the United States. Journal of General Virology 82(Pt 10): 2449-2462.
Haramoto, E., Katayama, H., Oguma, K., Yamashita, H., Tajima, A., Nakajima, H., and Ohgaki, S. (2006) Seasonal
profiles of human noroviruses and indicator bacteria in a wastewater treatment plant in Tokyo, Japan. Water Science
and Technology 54(11-12): 301-308.
Haramoto, E., Katayama, H., Oguma, K., and Ohgaki, S. (2007) Quantitative analyis of human enteric adenoviruses
in aquatic environments. Journal of Applied Microbiology 103: 2153-2159.
Harrigan, T.M., Northcott, W., Kravchenko, A., and Thomas, A. (2004) Time-series analysis of sediment and
bacterial transport from manure applied land following aeration tillage. In 2004 ASAE/CSAE Annual International
Meeting. ASABE (ed). Ottawa, Ontario, Canada.
Haugland, R.A., Siefring, S.C., Wymer, L.J., Brenner, K.P., andDufour, A.P. (2005) Comparison of Enterococcus
measurements in freshwater at two recreational beaches by quantitative polymerase chain reaction and membrane
filter culture analysis. Water Research 39(4): 559-568.
He, J. W., and Jiang, S. (2005) Quantification of enterococci and human adenoviruses in environmental samples by
real-time PCR. Applied and Environmental Microbiology 71(5): 2250-2255.
December 2010 110
-------
U.S. Environmental Protection Agency
He, L.-M, Lu, I, and Shi, W. (2007) Variability of fecal indicator bacteria in flowing and ponded waters in
southern California: implications for bacterial TMDL development and implementation. Water Research 41(14):
3132-3140.
Heinonen-Tanski, H., Mohaibes, M, Karinen, P., and Koivunen, J. (2006) Methods to reduce pathogen
microorganisms in manure. Livestock Science 102(3): 248-255.
Heirholzer, J.C. (1992) Adenoviruses in the immunocomproraised host. Clinical Microbiological Reviews 5(3): 262-
274.
Heitman, T.L., Frederick, L.M., Viste, J.R., Guselle, N.J., Morgan, U.M., Thompson, R.C., and Olson, M.E. (2002)
Prevalence of Giardia and Cryptosporidium and characterization of Cryptosporidium spp. isolated from wildlife,
human, and agricultural sources in the North Saskatchewan River Basin in Alberta, Canada. Canadian Journal of
Microbiology 48(6): 530-541.
Hejkal, T.W., Smith, E.M., and Gerba, C.P. (1984) Seasonal occurrence of rotavirus in sewage. Applied and
Environmental Microbiology 47(3): 588-590.
Hill, V.R., and Sobsey, M.D. (2003) Performance of swine waste lagoons for removing Salmonella and
entericmicrobial indicators. Transactions of the American Society of Agricultural and Biological Engineers 46(3):
781-788.
Hoar, B., Atwill, E.R., Elmi, C., andFarver, T.B. (2001) An examination of risk factors associated with beef cattle
shedding pathogens of potential zoonotic concern. Epidemiology and Infection 127(1): 147-155.
Hodgson, C., Bulmer, N., Chadwick, D., Oliver, D., Heathwaite, A., Fish, R., and Winter, M. (2009) Establishing
relative release kinetics of faecal indicator organisms from different faecal matrices. Leters in Applied Microbiology
49(1): 124-130.
Holley, R.A., Arras, K.M., Ominski, K.H., Tenuta, M., and Blank, G. (2006) Salmonella survival in manure-treated
soils during simulated seasonal temperature exposure. Journal of Environmental Quality 35(4): 1170-1180.
Hubalek, Z. (2004) An annotated checklist of pathogenic microorganisms associated with migratory birds. Journal
of Wildlife Diseases 40(4): 639-659.
Huston, C.L., Wittum, T.E., Love, B.C., and Keen, J.E. (2002) Prevalence of fecal shedding of Salmonella spp in
dairy herds. Journal of the American Veterinary Medicine Association 220(5): 645-649.
Hutchins, S.R., Walters, L.D., Avery, S.M., Synge, B.A., and Moore, A. (2004) Levels of zoonotic agents in British
livestock manures. Letters in Applied Microbiology 39: 207-214.
Hutchison, M.L., Walters, L.D., Avery, S.M., Munro, F., and Moore, A. (2005) Analyses of livestock production,
waste storage, and pathogen levels and prevalences in farm manures. Applied and Environmental Microbiology
71(3): 1231-1236.
Hutchison, S.R., Walters, L.D., Avery, S.M., Synge, B.A., and Moore, A. (2004) Levels of zoonotic agents in
British livestock manures. Letters in Applied Microbiology 39: 207-214.
Ihekweazu, C., Barlow, M., Roberts, S., Christensen, H., Guttridge, B., Lewis, D., and Paynter, S. (2006) Outbreak
of E. coli O157 infection in the southwest of the UK: risks from streams crossing seaside beaches. Eurosurveillance
11(4): 128-130.
ILSI (International Life Sciences Institute) (1996) A conceptual framework to assess the risks of human disease
following exposure to pathogens. Risk Analysis 16(6): 841:848.
December 2010 111
-------
U.S. Environmental Protection Agency
Inglis, G., Kalischuk, L., andBusz, H. (2004) Chronic shedding of Campylobacter species in beef cattle. Journal of
Applied Microbiology 97(2): 410-420.
Irving, L.G., and Smith, F.A. (1981) One-year survey of enteroviruses, adenoviruses, and reoviruses isolated from
effluent at an activated-sludge purification plant. Applied and Environmental Microbiology 41(1): 51-59.
Ito, K., Kubokura, Y., Kaneko, K., Totake, Y., and Ogawa, M. (1988) Occurrence of Campylobacter jejuni in free-
living wild birds from Japan. Journal of Wildlife Diseases 24(3): 467-470.
Jiang, J., Alderisio, K. A., and Xiao, L. (2005) Distribution of Cryptosporidium genotypes in storm event water
samples from three watersheds in New York. Applied and Environmental Microbiology 71(8): 4446-4454.
Jiang, S., Noble, R., and Chu, W. (2001) Human adenoviruses and coliphages in urban runoff-impacted coastal
waters of Southern California. Applied andEnvironmentalMmicrobiology 67(1): 179-184.
Jiang, S. (2006) Human adenovirus in water: occurrence and health implications: a critical review. Environmental
Science and Technology 40(23): 7132-7140.
Jimenez-Cisneros, B.E., Maya-Rendon, C., and Salgado-Velazquez, G. (2001) The elimination of helminth ova,
faecal conforms, Salmonella and protozoan cysts by various physicochemical processes in wastewater and sludge.
Water Science and Technology 43(12): 179-182.
Jolis, D., Pitt, P., and Hirano, R. (1999) Risk assessment for Cryptosporidium parvum in reclaimed water. Water
Research 33(13): 3051-3055.
Jones, K. (2001) Campylobacters in water, sewage and the environment. Journal of Applied Microbiology 90: 68S-
79S.
Julian, T.R., Canales, R.A., Leckie, J.O., and Boehm, A.B. (2009) A model of exposure to rotavirus from nondietary
ingestion iterated by simulated intermittent contacts. Risk Analysis 29(5): 617-632.
Kapperud, G., andRosef, O. (1983) Avian wildlife reservoir of Campylobacter fetus subsp. jejuni, Yersinia spp., and
Salmonella spp. in Norway. Applied and Environmental Microbiology 45(2)'. 375-380.
Karaguzel, A., Koksal, L, Baki, A., Ucar, F., Gok, L, and Cirav, Z. (1993) Salmonella and Shigella carriage by gulls
(Larus sp.) on the east Black Sea region of Turkey. Microbios 74(299): 77-80.
Kassa, H., Harrington, B.J., and Bisesi, M.S. (2004) Cryptosporidiosis: a brief literature review and update regarding
Cryptosporidium infeces of Canada geese (Branta canadensis). Journal of Environmental Health 66(7): 34-39.
Katayama, H., Haramoto, E., Oguma, K., Yamashita, H., Tajima, A., Nakajima, H., and Ohgaki, S. (2008) One-year
monthly quantitative survey of noroviruses, enteroviruses, and adenoviruses in wastewater collected from six plants
in Japan. Water Research 42(6-7): 1441-1448.
Kelley, T.R., Pancorbo, O.C., Merka, W.C., Thompson, S.A., Cabrera, M.L., and Barnhart, H.M. (1995) Bacterial
pathogens and indicators in poultry litter during re-utilization. Journal of Applied Poultry Research 4: 366-373.
Ketley, J.M. (1997) Pathogenesis of enteric infectionby Campylobacter. Microbiology 143(1): 5-21.
Kinzelman, J., McLellan, S., Amick, A., Preedit, J., Scopel, C., Olapade, O., Gradus, S., Singh, A., and Sedmak, G.
(2008) Identification of human enteric pathogens in gull feces at Southwestern Lake Michigan bathing beaches.
Canadian Journal of Microbiology 54(12): 1006-1015.
Kirk, J., Holmberg, C., and Jeffrey, J. (2002) Prevalence of Salmonella spp in selected birds captured on California
dairies. Journal of the American Veterinary Medical Association 220(3): 359-362.
December 2010 112
-------
U.S. Environmental Protection Agency
Kirkpatrick, C. (1986) Isolation of Salmonella spp. from a colony of wading birds. Journal of Wildlife Diseases
22(2): 262-264.
Klein, M, Brown, L., van den Akker, B., Peters, G.M., Stuetz, R.M., and Roser, D.J. (2010) Monitoring bacterial
indicators and pathogens in cattle feedlot waste by real-time PCR. Water Research 44(5): 1381-1388.
Koenraad, P.M.F.J., Hazeleger, W.C., vanderLaan, T., Beumer, R.R., andRombouts, P.M. (1994) Survey of
Campylobacter spp. in sewage plants in The Netherlands. Food Microbiology 11: 65-73.
Koivunen, I, Lanki, E., Rajala, R.L., Siitonen, A., and Heinonen-Tanski, H. (2001) Determination of salmonellae
from municipal wastewaters Water Science and Technology 43(12): 221-224.
Koopmans, M. (2008) Progress in understanding norovirus epidemiology. Current Opinion in Infectious Diseases
21(5): 544-552.
Kraft, D.J., LOlechowski-Gerhardt, C., Berkowitz, I, and Finstein, M.S. (1969) Salmonella in wastes produced at
co mmercial poultry farms. Applied Microbiology 18(5): 703-707.
Kudva, I.T., Blanch, K, andHovde, C.J. (1998) Analysis of Escherichia coli O157:H7 survival in ovine or bovine
manure and manure slurry. Applied andEnvironmental Microbiology 64(9): 3166-3174.
Kuhn, R.C., Rock, C.M., and Oshima, K.H. (2002) Occurrence of Cryptosporidium and Giardia in wild ducks along
the Rio Grande River Valley in southern New Mexico. Applied and Environmental Microbiology 68(1): 161-165.
Kunze, D.J., Loneragan, G.H., Platt, T.M., Miller, M.F., Besser, T.E., Koohmaraie, M., Stephens, T., and Brashears,
M.M. (2008) Salmonella enterica burden in harvest-ready cattle populations from the Southern High Plains of the
United States. Applied and Environmental Microbiology 74(2): 345-351.
Langeland, G. (1982) Salmonella spp. in the working environment of sewage treatment plants in Oslo, Norway.
Applied and Environmental Microbiology 43(5): 1111-1115.
Larney, F.J., and Hao, X. (2007) A review of composting as a management alternative for beef cattle feedlot manure
in southern Alberta, Canada. Bioresource Technology 98(17): 3221-3227.
Laverick, M.A., Wyn-Jones, A.P., and Carter, M.J. (2004) Quantitative RT-PCR for the enumeration of noroviruses
(Norwalk-like viruses) in water and sewage. Letters in Applied Microbiology 39(2): 127-136.
Lee, I, Zoh, K., and Ko, G. (2008) Inactivation and UV disinfection of murine norovirus with TiO2 under various
environmental conditions. Applied and Environmental Microbiology 74(7): 2111-2117.
Legrand-Abravanel, F., Mansuy, J.-M., Dubois, M., Kamar, N., Peron, J.-M., Rostang, L., and Izopet, J. (2009)
Hepatitis E virus genotype 3 diversity, France. Emerging Infectious Diseases 15(1): 110-114.
LeJeune, J.T., Besser, T.E., Rice, D.H., Berg, J.L., Stilborn, R.P., and Hancock, D.D. (2004) Longitudinal study of
fecal shedding of Escherichia coli O157:H7 in feedlot cattle: predominance and persistence of specific clonal types
despite massive cattle population turnover. Applied and Environmental Microbiology 70(1): 377-384.
Lemarchand, K., and Lebaron, P. (2003) Occurrence of Salmonella spp. and Cryptosporidium spp. in a French
coastal watershed: relationship with fecal indicators. FEMSMicrobiology Letters 218(1): 203-209.
Lenahan, M., O'Brien, S.B., Byrne, C., Ryan, M., Kennedy, C.A., McNamara, E.B., Fanning, S., Sheridan, J.J., and
Sweeney, T. (2009) Molecular characterization of Irish E. coli 0157 :H7 isolates of human, bovine, ovine and
porcine origin. Journal of Applied Microbiology 107(4): 1340-1349.
December 2010 113
-------
U.S. Environmental Protection Agency
Letourneau, V., Duchaine, C., Cote, C., Letellier, A., Topp, E., and Masse, D. (2010) Presence of zoonotic
pathogens in physico-chemically characterized manures from hog finishing houses using different production
systems. Bioresource Technology 101(11): 4048-4055.
Levesque, B., Brousseau, P., Simard, P., Dewailly, E., Meisels, M., Ramsay, D., and Joly, J. (1993) Impact of the
ring-billed gull (Lams delawarensis) on the microbiological quality of recreational water. Applied and
Environmental Microbiology 59(4): 1228-1230.
Levett, P.N. (2001) Leptospirosis. Clinical Microbiology Reviews 14(2): 296-326.
Ley, D.H., Levy, M.G., Hunter, L., Corbett, W., and Barnes, H.J. (1988) Cryptosporidia-positive rates of avian
necropsy accessions determined by examination of auramine O-stained fecal smears. Avian Diseases 32(1): 108-
113.
Lillehaug, A., Monceyron Jonassen, C., Bergsjo, B., Hofshagen, M., Tharaldsen, I, Nesse, L., and Handeland, K.
(2005) Screening of feral pigeon (Colombo livia), mallard (Anas platyrhynchos) and graylag goose (Anser anser)
populations for Campylobacter spp., Salmonella spp., avian influenza virus and avian paramyxovirus. Ada
Veterinaria Scandinavica 46(4): 193-202.
Lindesmith, L., Moe, C., Marionneau, S., Ruvoen, N., Jiang, X., Lindbland, L., Stewart, P., LePendu, J., and Baric,
R. (2003) Human susceptibility and resistance to Norwalk virus infection. Nature Medicine 9(5): 548-553.
Liu, Y.J., Zhang, C.M., and Wang, X.C. (2009) Simultaneous detection of enteric bacteria from surface waters by
QPCR in comparison with conventional bacterial indicators. Environmental Monitoring and Assessment 158(1-4):
535-544.
Locke, L.N., Shillinger, R.B., and Jareed, T. (1973) Salmonellosis in passerine birds in Maryland and West Virginia.
Journal of Wildlife Diseases 9(2): 144-145.
Lodder, W., Vinje, J., Heide, V.d., Husman, R., Leenen, E., and Koopmans, M. (1999) Molecular detection of
norwalk-like caliciviruses in sewage. Applied and Environmental Microbiology 65(12): 5624-5627.
Lodder, W.J., and de Roda Husman, A.M. (2005) Presence of noroviruses and other enteric viruses in sewage and
surface waters in The Netherlands. Applied and Environmental Microbiology 71(3): 1453-1461.
Loge, F.J., Lambertini, E., Borchardt, M.A., Basagaoglu, H., and Ginn, T.R. (2009) Effects of etiological agent and
bather shedding of pathogens on interpretation of epidemic logical data used to establish recreational water quality
standards. Risk Analysis 29(2): 257-266.
Luangtongkum, T., Morishita, T.Y., Ison, A.J., Huang, S., McDermott, P.P., and Zhang, Q. (2006) Effect of
conventional and organic production practices on the prevalence and antimicrobial resistance of Campylobacter spp.
in poultry. Applied and Environmental Microbiology 72(5): 3600-3607.
Ludwig, K., Sarkim, V., Bitzan, M., Karmali, M.A., Bobrowski, C., Ruder, H., Laufs, R., Sobottka, L, Petric, M.,
Karch, H., and Muller-Wiefel, D.E. (2002) Shiga toxin-producing Escherichia coli infection and antibodies against
Stx2 and Stxl in household contacts of children with enteropathic hemolytic-uremic syndrome. Journal of Clinical
Microbiology 40(5): 1773-1782.
Maddox-Hyttel, C., Langkjeer, R.B., Enemark, H.L., andHakan, V. (2006) Cryptosporidium and Giardia indifferent
age groups of Danish cattle and pigs—occurrence and management associated risk factors. Veterinary Parasitology
141(1-2): 48-59.
Martella, V., Banyai, K., Matthijnssens, J., Buonavoglia, C., and Ciarlet, M. (2010) Zoonotic aspects of rotaviruses.
Veterinary Microbiology 140(3-4): 246-255.
December 2010 114
-------
U.S. Environmental Protection Agency
Martens, W., and Bohm, R. (2009) Overview of the ability of different treatment methods for liquid and solid
manure to inactivate pathogens. Bioresource Technology 100(22): 5374-5378.
Marti, R., Dabert, P., and Pourcher, A.-M. (2009) Pig manure contamination marker selection based on the influence
of biological treatment on the dominant fecal microbial groups Applied and Environmental Microbiology 75(15):
4967-4974.
Martin, S.A., McCann, M.A., and Waltman, W.D. (1998) Microbiological survey of Georgia poultry litter. Journal
of Applied Poultry Research 7(1): 90-98.
Mattison, K., Karthikeyan, K., Abebe, M., Malik, N., Sattar, S.A., Farber, J.M., and Bidawid, S. (2007) Survival of
calicivirus in foods and on surfaces: experiments with feline calicivirus as a surrogate for norovirus. Journal of Food
Projection70(2): 500-503.
McCuin, R.M., and Clancy, J.L. (2006) Occurrence of Cryptosporidium oocysts in US wastewaters. Journal of
Water and Health 4(4): 437-452.
McCullough, N., andEisele, C. (195 la) Experimental human salmonellosis. I. pathogenicity of strains of Salmonella
meleagridis and Salmonella anatum obtained from spray-dried whole egg. Journal of Infectious Diseases 88(3):
278-289.
McCullough, N., and Eisele, C. (195 Ib) Experimental human salmonellosis. III. pathogenicity of strains of
Salmonella newport, Salmonella derby, and Salmonella bareilly obtained from spray-dried whole egg. Journal of
Infectious Diseases 89(3): 209-213.
McDonough, P.L., Fogelman, D., Shin, S.J., Brunner, M.A., and Lein, D.H. (1999) Salmonella enterica serotype
Dublin infection: an emerging infectious disease for the Northeastern United States. Journal of Clinical
Microbiology 37(8): 2418-2427.
McLaughlin, M.R., Brooks, J.P., and Adeli, A. (2009) Characterization of selected nutrients and bacteria from
anaerobic swine manure lagoons on sow, nursery, and finisher farms in the mid-South USA. Journal of
Environmental Quality 38(6): 2422-2430.
McLaughlin, S.J., Kalita, P.K., Cooke, R.A., Fox, S.P., and Kuhlenschmidt, M.S. (2003) Adsorption kinetics of
Cryptosporidium parvum to soils. In 2003 ASAE Annual International Meeting. ASAE (ed). Las Vegas, NV.
McNulty, M.S. (1978) Rotaviruses. Journal of General Virology 40: 1-18.
Mead, P.S., Slutsker, L., Dietz, V., McCaig, L.F., Bresee, J.S., Shapiro, C., Griffin, P.M., and Tauxe, R.V. (1999)
Food related illness and death in the United States. Emerging Infectious Diseases 5(5): 607-625.
Medema, G.J., Teunis, P.P., Havelaar, A.H., and Haas, C.N. (1996) Assessment of the dose-response relationship of
Campylobacterjejuni. International Journal of Food Microbiology 30(1-2): 101-111.
Meites, E., Jay, M.T., Deresinski, S., Shieh, W.J., Zaki, S.R., Tompkins, L., and Smith, D.S. (2004) Reemerging
leptospirosis, California. Emerging Infectious Diseases 10(3): 406-412.
Meng, X.J., Dea, S., Engle, R.E., Friendship, R., Lyoo, Y.S., Sirinarumitr, T., Urairong, K., Wang, D., Wong, D.,
Yoo, D., Zhang, Y., Purcell, R.H., and Emerson, S.U. (1999) Prevalence of antibodies to the hepatitis E virus in pigs
from countries where hepatitis E is common or is rare in the human population. Journal of Medical Virology 59(3):
297-302.
Messner, M.J., Chappell, C.L., and Okhuysen, P.C. (2001) Risk assessment for Cryptosporidium,'. a hierarchical
bayesian analysis of human dose-response data. Water Research 35(16): 3934-3940.
December 2010 115
-------
U.S. Environmental Protection Agency
MWRDGC (Metropolitan Water Reclamation District of Greater Chicago) (2008) Dry and Wet Weather Risk
Assessment of Human Health: Impacts of Disinfection vs. No Disinfection of the Chicago Area Waterways System.
Prepared by Geosyntec Consultants.
Midwest Plan Service (2004) Manure Characteristics. MWPS Ames, IA.
Miller, I, I, and Beasley, B., W. (2008) Influence of livestock manure type on transport of Escherichia coli in
surface runoff. Water Quality Research Journal of Canada 43(2/3): 129-136.
Minihan, D., Whyte, P., O'Mahony, M., Fanning, S., McGill, K., and Collins, J.D. (2004) Campylobacter spp. in
Irish feedlot cattle: a longitudinal study involving pre-harvest and harvest phases of the food chain. Journal of
Veterinary Medicine, Series B 51(1): 28-33.
Mintz, E.D., Hudson-Wragg, M., Mshar, P., Cartter, M.L., and Hadler, J.L. (1993) Foodborne giardiasis in a
corporate office setting. Journal of Infectious Diseases 167(1): 250-253.
Mitchell, T.R., and Ridgwell, T. (1971) The frequency of salmonellae in wild ducks. Journal of Medical
Microbiology 4(3): 359-361.
Molyneaux, P.J. (1995) Human immunity to rotavirus. Journal of Medical Microbiology 43(6): 397-404.
Moriarty, E.M., Sinton, L.W., Mackenzie, M.L., Kearki, N., and Wood, D.R. (2008) A survey of enteric bacteria and
protozoans in fresh bovine faeces on New Zealand dairy farms. Journal of Applied Microbiology 105: 2015-2025.
Muirhead, R.W., Collins, R.P., and Bremer, P.J. (2006) Numbers and transported state of Escherichia coli in runoff
direct from fresh cowpats under simulated rainfall. Letters in Applied Microbiology 42(2): 83-87.
Nocker, A., and Camper, A.K. (2006) Selective removal of DNA from dead cells of mixed bacterial communities by
use of ethidmrnmonoazide. Applied and Environmental Microbiology 72(3): 1997-2004.
NRC (National Research Council) (1983) Risk Assessment in the Federal Government: Managing the Process.
National Academy Press, Washington, DC.
NRC (2004) Indicators for Waterborne Pathogens. The National Academies Press, Washington, D.C.
O'Reilly, C.E., Bowen, A.B., Perez, N.E., Sarisky, J.P., Shepherd, C.A., Miller, M.D., Hubbard, B.C., Herring, M.,
Buchanan, S.D., Fitzgerald, C.C., Hill, V., Arrowood, M.J., Xiao, L.X., Hoekstra, R.M., Mintz, E.D., and Lynch,
M.F. (2007) A waterborne outbreak of gastroenteritis with multiple etiologies among resort island visitors and
residents: Ohio, 2004. Clinical Infectious Diseases 44(4): 506-512.
Okhuysen, P.C., Chappell, C.L., Sterling, C.R., Jakubowski, W., and DuPont, H.L. (1998) Susceptibility and
serologic response of healthy adults to reinfection with Cryptosporidium parvum. Infection and Immunity 66(2):
441-443.
Okhuysen, P.C., Chappell, C.L., Crabb, J.H., Sterling, C.R., and DuPont, H.L. (1999) Virulence of three distinct
Cryptosporidium parvum isolates for healthy adults. Journal of Infectious Diseases 180(4): 1275-1281.
Okhuysen, P.C., Rich, S.M., Chappell, C.L., Grimes, K.A., Widmer, G., Feng, X., and Tzipori, S. (2002) Infectivity
of a Cryptosporidium parvum isolate of cervine origin for healthy adults and interferon-gamma knockout mice.
Journal of Infectious Diseases 185(9): 1320-1325.
Olivieri, A., and Seller, J. (2002) Evaluation of the Public Health Risks Concerning Infectious Disease Agents
Associated with Exposure to Treated Wastewater Discharged by the City of Vacaville, Easterly Wastewater
Treatment Plant. Prepared by EOA, Inc. for the City of Vacaville.
December 2010 116
-------
U.S. Environmental Protection Agency
Olson, M.E., Thorlakson, C.L., Deselliers, L., Morck, D.W., and McAllister, T.A. (1997) Giardia and
Cryptosporium in Canadian farm animals. Veterinary Parasitology 68(4): 375-381.
Pachepsky, Y., Guber, A., Shelton, D., and McCarty, G. (2009) Size distributions of manure particles released under
simulated rainfall. Journal of Environmental Management 90(3): 1365-1369.
Palmgren, H., Aspan, A., Broman, T., Bengtsson, K., Blomquist, L., Bergstrom, S., Sellin, M., Wollin, R., and
Olsen, B. (2006) Salmonella in Black-headed gulls (Lams ridibundus); prevalence, genotypes and influence on
Salmonella epidemiology. Epidemiology andInfecttion 134(3): 635-644.
Parkhurst, D.F., Craun, G., and Seller, J. (2007) Conceptual bases for relating illness risk to indicator
concentrations. In Statistical Framework for Recreational Water Quality Criteria and Monitoring. Wymer, L. (ed):
John Wiley & Sons Ltd., England.
Parkin, R., Seller, J., and Olivieri, A. (2003) Incorporating susceptible subpopulations inmicrobial risk assessment:
pediatric exposures to enteroviruses in river water. Journal of Exposure Science and Environmental Epidemiology
13(2): 161-168.
Payment, P., Richardson, L., Siemiatycki, J., and Dewar, R. (1991) A randomized trial to evaluate the risk of
gastrointestinal disease due to consumption of drinking water meeting current microbiological standards. American
Journal of Public Health 81(6): 703-708.
Payment, P., Siemiatycki, J., Richardson, L., Renaud, G., Franco, E., andPrevost, M. (1997) A prospective
epidemic logical study of gastrointestinal health effects due to the consumption of drinking water. International
Journal of Environmental Research and Public Health 7(1): 5-31.
Payment, P., Plante, R., and Cejka, P. (2001) Removal of indicator bacteria, human enteric viruses, Giardia cysts,
and Cryptosporidium oocysts at a large wastewater primary treatment facility. Canadian Journal of Microbiology
47(3): 188-193.
Petterson, S.R., Signer, R.S., and Ashbolt, N.J. (2007) Incorporating method recovery uncertainties in stochastic
estimates of raw water protozoan concentrations forQMRA. Journal of Water and Health 5(Supplement 1): 51-65.
Petterson, S.R., Dumoutier, N., Loret, J.F., and Ashbolt, N.J. (2009) Quantitative Bayesian predictions of source
water concentration for QMRA from presence/absence data for E. coli O157:H7. Water Science and Technology
59(11): 2245-2252.
Peu, P., Brugere, H., Pourcher, A.-M, Kerouredan, M., Godon, J.-J., Delgenes, J.-P., andDabert, P. (2006)
Dynamics of a pig slurry microbial community during anaerobic storage and management. Applied and
Environmental Microbiology 72(5): 3578-3585.
Pouillot, R., Beaudeau, P., Denis, J.B., and Derouin, F. (2004) A quantitative risk assessment of waterborne
cryptosporidiosis in France using second-order monte carlo simulation. Risk Analysis 24(1): 1-17.
Priiss, A. (1998) Review of epidemic logical studies on health effects from exposure to recreational water.
International Journal of Epidemiology 27(1): 1-9.
Pusch, D., Oh, D.Y., Wolf, S., Dumke, R., Schroter-Bobsin, U., Holme, M., Roske, L, and Schreier, E. (2005)
Detection of enteric viruses and bacterial indicators in German environmental waters. Archives of Virology 150(5):
929-947.
Python Software Foundation (2009) Python Programming Language. Hampton, NH.
Quessy, S., and Messier, S. (1992) Prevalence of Salmonella spp., Campylobacter spp. and Listeria spp. in ring-
billed gulls (Lams delawarensis). Journal of Wildlife Diseases 28(4): 526-531.
December 2010 117
-------
U.S. Environmental Protection Agency
R Development Core Team (2009) R: A Language and Environment for Statistical Computing. Vienna, Austria.
Rajal, V.B., McSwain, B.S., Thompson, D.E., Leutenegger, C.M., and Wuertz, S. (2007) Molecular quantitative
analysis of human viruses in California stormwater. Water Research 41(19): 4287-4298.
Ralston, B.I, McAllister, T. A., and Olson, M.E. (2003) Prevalence of infection pattern of naturally acquired
giardiasis and cryptosporidiosis in range beef calves and their dams. Veterinary Parasitology 114: 113-122.
Rao, V.C., Metcalf, T.G., and Melnick, J.L. (1987) Removal of indigenous rotaviruses during primary settling and
activated-sludge treatment of raw sewage. Water Research 21(2): 171-177.
Regli, S., Rose, J.B., Haas, C.N., and Gerba, C.P. (1991) Modeling the risk from Giardia and viruses in drinking-
water. J ournal of the American Water Works Association 83(11): 76-84.
Reinoso, R., and Becares, E. (2008) The occurrence of intestinal parasites in swine slurry and their removal in
activated sludge plants. Bioresource Technology 99: 6661-6665.
Rendtorff, R.C. (1954a) The experimental transmission of human intestinal protozoan parasites. I. Giardia lamblia
cysts given in capsules. American Journal of Hygiene 59(2): 209-220.
Rendtorff, R.C. (1954b) The experimental trnamission of human intestinal protozoan parasites. I. Attemps to
transmit Endamoeba coli and Giardia lamblia by water. American Journal of Hygiene 60(3): 327-328.
Riley, S., Fraser, C, Donnelly, C.A., Ghani, A.C., Abu-Raddad, L.J., Hedley, A.J., Leung, G.M., Ho, L.M., Lam,
T.H., Thach, T.Q., Chau, P., Chan, K.P., Lo, S.V., Leung, P.Y., Tsang, T., Ho, W., Lee, K.H., Lau, E.M., Ferguson,
N.M., and Anderson, R.M. (2003) Transmission dynamics of the etiological agent of SARS in Hong Kong: impact
of public health interventions. Science 300(5627): 1961-1966.
Roberts, J.D., Sibergeld, E.K., and Graczyk, T. (2007) A probabilistic risk assessment ofCryptosporidium exposure
among Baltimore urban anglers. Journal of Toxicology and Environmental Health, Part A 70(18): 1568-1576.
Robertson, L.J., Paton, C.A., Campbell, A.T., Smith, P.O., Jackson, M.H., Gilmour, R.A., Black, S.E., Stevenson,
D.A., and Smith, H.V. (2000) Giardia cysts and Cryptosporidium oocysts at sewage treatment works in Scotland,
UK. Water Research 34(8): 2310-2322.
Robertson, L.J., Hermansen, L., and Gjerde, B.K. (2006) Occurrence of Crypto sporidium oocysts and Giardia cysts
in sewage inNorway. Applied and Environmental Microbiology 72(8): 5297-5303.
Rose, J.B., Mullinax, R.L., Singh, S.N., Yates, M.V., and Gerba, C.P. (1987) Occurrence of rotaviruses and
enteroviruses in recreational waters of Oak Creek, Arizona. Water Research 21(11): 1375-1381.
Rose, J.B., Haas, C.N., and Regli, S. (1991) Risk assessment and control of waterborne giardiasis. American Journal
of Public Health 81(6): 709-713.
Rose, J.B., Huffman, D.E., Riley, K, Farrah, S.R., KLukasik, J.O., and Hamann, C.L. (2001) Reduction of enteric
microorganisms at the upper Occoquan Sewage Authority water reclamation plant. Water Environment Research
73(6): 711-720.
Rose, J.B., Nowlin, H., Farrah, S.R., Harwood, V., Levine, A., Lukasik, J., Menendez, P., and Scott, T.M. (2004)
Reduction of Pathogens, Indicator Bacteria, and Alternative Indicators by Wastewater Treatment and Reclamation
Processes. Water Environment Research Foundation Report 00-PUM-2T. Alexandria, VA.
Rosen, B.H. (2000) Waterborne Pathogens in Agricultural Watershes. Watershed Science Institute, National
Resouces Conservation Service Ithaca, NY.
December 2010 118
-------
U.S. Environmental Protection Agency
Roser, D.J., Ashbolt, N.J., Davies, C.M., Glamore, W.C., Hawker, K.M., and Miller, B.M. (2007) Application of
TMDL and risk assessment principles for pathogen management at an urban recreation lake. In Watershed
Management to Meet Water Quality Standards and TMDLS, 4th Conference Proceedings. San Antonio, TX, pp. 420-
426.
Ruijter, J.M., Ramakers, C., Hoogaars, W.M.H., Karlen, Y., Bakker, O., vandenHoff, M.J.B., and Moorman,
A.F.M. (2009) Amplification efficiency: linking baseline and bias in the analysis of quantitative PCR data. Nucleic
Acids Research 37(6): e45.
Rutjes, S.A., Lodder, W.J., Lodder-Verschoor, F., van den Berg, H.H.J.L., Vennema, H., Duizer, E., Koopmans, M.,
and de Roda Husman, A.M. (2009) Sources of hepatitis E virus genotype 3 in the Netherlands. Emerging and
Infectious Diseases 15(3): 381-387.
Rutledge, R.G., and Cote, C. (2003) Mathematics of quantitative kinetic PCR and the application of standard curves.
Nucleic Acids Research 3(16): e93.
Schijven, I, and de Roda Husman, A.M. (2006) A survey of diving behaviour and accidental water ingestion among
Dutch occupational and sport divers to assess the risk of infection with waterborne pathogenic microorganisms.
Environmental Health Perspectives 114(5): 712-717.
Schoen, M.E., and Ashbolt, N.J. (2010) Assessing pathogen risk to swimmers at non-sewage impacted recreational
beaches. Environmental Science and Technology 44(7): 2286-2291.
Scott, T.M., McLaughlin, M.R., Harwood, V.J., Chivukula, V., Levine, A., Gennaccaro, A., Lukasik, I, Farrah,
S.R., and Rose, J.B. (2003) Reduction of pathogens, indicator bacteria and alternative indicators by wastewater
treatment and reclamation processes. Water Science and Technology: Water Supply 3(4): 247-252.
Sedmak, G., Bina, D., MacDonald, I, and Couillard, L. (2005) Nine-year study of the occurrence of culturable
viruses in source water for two drinking water treatment plants and the influent and effluent of a wastewater
treatment plant in Milwaukee, Wisconsin (August 1994 through July 2003). Applied and Environmental
Microbiology 71(2): 1042-1050.
Shere, J.A., Bartlett, K.J., andKaspar, C.W. (1998) Longitudinal study of Escherichia coli O157:H7 dissemination
on four dairy farms in Wisconsin. Applied and Environmental Microbiology 64(4): 1390-1399.
Signer, R.S., and Ashbolt, N.J. (2006) Pathogen monitoring offers questionable protection against drinking-water
risks: a QMRA (quantitative microbial risk analysis) approach to assess management strategies. Water Science and
Technology 54(3): 261-268.
Signer, R.S., Ashbolt, N.J., and Roser, D.J. (2007) Microbial risk implications of rainfall-induced runoff events
entering a reservoir used as a drinking water source. Journal of Water Supply: Research and Technology -AQUA
56(8): 515-531.
Sinclair, R.G., Jones, E.L., and Gerba, C.P. (2009) Viruses in recreational water-borne disease outbreaks: a review.
Journal of Applied Microbiology 107(6): 1769-1780.
Sinton, L. W., Braithwaite, R.R., Hall, C.H., and Mackenzie, M.L. (2007) Survival of indicator and pathogenic
bacteria in bovine feces on pasture. Applied and Environmental Microbiology 73(24): 7917-7925.
Smith, E.M., and Gerba, C.P. (1982) Development of a method for detection of human rotavirus in water and
sewage. Applied and Environmental Microbiology 43(6): 1440-1450.
Smith, H., Brown, J., Coulson, J., Morris, G., and Girdwood, R. (1993) Occurrence of oocysts of Cryptosporidium
sp. mLarus spp. gulls. Epidemiology andInfecttion 110(1): 135-143.
December 2010 119
-------
U.S. Environmental Protection Agency
Smith, J.L. (2001) A review of hepatitis E vims. Journal of Food Protection 64(4): 572-586.
Sobsey, M.D., Khatib, L.A., Hill, V.R., Alocilja, E., and Pillai, S. (2006) Pathogens in Animal Wastes and the
Impacts of Waste Management Practices on their Survival, Transport and Fate. In Animal Agriculture and the
Environment: National Center for Manure and Animal Waste Management White Papers. Rice, J.M., Cldwell, D.F.,
and Humenik, F.J. (eds). St. Joseph, MI: ASABE.
Seller, J.A., Olivieri, A., Crook, I, Parkin, R., Spear, R., Tchobanoglous, G., and Eisenberg, J.N.S. (2003) Risk-
based approach to evaluate the public health benefit of additional wastewater treatment. Environmental Science and
Technology 31 (9): 1882-1891.
Seller, J.A., Eisenberg, J.N.S., DeGeorge, J.F., Cooper, R.C., Tchobanoglous, G., and Olivieri, A.W. (2006) A
public health evaluation of recreational water impairment. Journal of Water and Health 4(1): 1-19.
Seller, J.A., Seto, E., and Olivieri, A.W. (2007a) Microbial Risk Assessment Interface Tool, Final Project Report,
Water Environment Research Foundation Project 04-HHE-3(eds).
Seller, J.A., Seto, E.Y., and Olivieri, A.W. (2007b) Application of microbial risk assessment techniques to estimate
risk due to exposure to reclaimed waters. WateReuse Foundation.
Seller, J.A., and Eisenberg, J.N.S. (2008) An evaluation of parsimony for microbial risk assessment models.
Environmetrics 19(1): 61-78.
Seller, J.A. (2009) Potential implications of person-to-person transmission of viral infection to US EPA's
Groundwater Rule. Journal of Water and Health 7(2): 208-223.
Seller, J.A., Bartrand, T., Ashbolt, N.J., Ravenscroft, J., and Wade, T.J. (2010a) Estimating the primary etiologic
agents in recreational freshwaters impacted by human sources of faecal contamination. Water Research 44(16):
4736-4747.
Seller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J., and Ashbolt, N.J. (2010b) Estimated human health risks
from exposure to recreational waters impacted by human and non-human sources of faecal contamination. Water
^esearc/z:doi:10.1016/j.watres.2010.1006.1049.
Soupir, M.L. (2003) Release and transport of bacteria and nutrients from livestock manure applied to pasturelands,
M.S. thesis, Virginia Polytechnic Institure and State University. Blacksburg, VA.
Soupir, M.L., Mostaghimi, S., and Yagow, E.R. (2006) Nutrient transport from livestock manure applied to
pastureland using phosphorus-based management strategies. Journal of Environmental Quality 35(4): 1269-1278.
Stampi, S., Varoli, O., Zanetti, F., and De Luca, G. (1993) Arcobacter cryaerophilus and thermophilic
campylobacters in a sewage treatment plant in Italy: two secondary treatments compared. Epidemiology and
Infection 110(3): 633-639.
Stanley, K.N., Wallace, J.S., Currie, J.E., Diggle, P.J., and Jones, K. (1998) The seasonal variation of thermophilic
campylobacters in beef cattle, dairy cattle and calves. Journal of Applied Microbiology 85: 472-480.
Steyn, M., Jagals, P., and Genthe, B. (2004) Assessment of microbial infection risks posed by ingestion of water
during domestic water use and full-contact recreation in a mid-southern African region. Water Science and
Technology 50(1): 301-308.
Sturdee, A.P., Bodley-Tickell, A.T., Archer, A., and Chalmers, R.M. (2003) Long-term study of Cryptosporidium
prevalence on a lowland farm in the United Kingdom. Veterinary Parasitology 116: 97-113.
December 2010 120
-------
U.S. Environmental Protection Agency
Suwa, M, and Suzuki, Y. (2001) Occurrence of Cryptosporidium in Japan and countermeasures inwastewater
treatment plants. Water Science and Technology 43(12): 183-186.
Takahashi, K., Okamoto, H., Abe, N., Kawakami, M., Matsuda, H., Mochida, S., Sakugawa, H., Suginshita, Y.,
Watanabe, S., Yamamoto, K., Miyakawa, Y., and Mishiro, S. (2009) Virulent strain of hepatitis E virus, genotype 3,
Japan. Emerging infectious diseases 15(5): 704-709.
Teltsch, B., Kedmi, S., Bonnet, L., Borenzstajn-Rotem, Y., and Katzenelson, E. (1980) Isolation and identification
of pathogenic microorganisms at wastewater-irrigated fields: ratios in air and wastewater. Applied and
Environmental Microbiology 39(6): 1183 -1190.
Terzich, M., Pope, M.J., Cherry, T.E., and Hollinger, J. (2000) Survey of pathogens in poultry litter in the United
States. Journal of Applied Poultry Research 9(3): 287-291.
Teunis, P.P., van der Heijden, O.G., van der Giessen, J.W.B., and Havelaar, A.H. (1996) The Dose-Response
Relation in Human Volunteers for Gastro-intestinal Pathgens. RIVM, The Netherlands.
Teunis, P.P., Nagelkerke, N.J., and Haas, C.N. (1999) Dose-response models for infectious gastroenteritis. Risk
Analysis 19(6): 1251-1260.
Teunis, P.P., Chappell, C.L., and Okhuysen, P.C. (2002) Cryptosporidium dose-response studies: variation between
isolates. Risk Analysis 22(1): 175-183.
Teunis, P.P., Takumi, K., and Shinagawa, K. (2004) Dose-response for infection by Escherichia coli O157:H7 from
outbreak data. RiskAnafysis24(2): 401-407.
Teunis, P.P., Van den Brandhof, W., Nauta, M., Wagenaar, J., Van den Kerkhof, H., and Van Pelt, W. (2005) A
reconsideration of the Campylobacter dose-response relation. Epidemiolofy andInfecttion 133(4): 583-592.
Teunis, P.P., Moe, C.L., Liu, P., S, E.M., Lindesmith, L., Baric, R.S., Le Pendu, J., and Calderon, R.L. (2008a)
Norwalk virus: how infectious is it? Journal of Medical Virology 80(8): 1468-1476.
Teunis, P.P., Ogden, I.D., and Strachan, N.J.L. (2008b) Hierarchical dose-response of E. coli O157:H7 from human
outbreaks incorporating heterogeneity in exposure. Epidemiology and Infection 136: 761-770.
Thompson, R.C.A. (2004) The zoonotic significance and molecualr epidemiology ofGiardia and giardiasis.
Veterinary Parasitology 126: 15-35.
Thurston-Enriquez, J., Gilley, J., and Eghball, B. (2005) Microbial quality of runoff following land application of
cattle manure and swine slurry. Journal of Water and Health 3(2): 157-171.
Till, D., McBride, G., Ball, A., Taylor, K., and Pyle, E. (2008) Large-scale freshwater microbiological study:
rationale, results and risks. Journal of Water and Health 6(4): 443-459.
Topp, E., Scott, A., Lapen, D.R., Lyautey, E., and Duriez, P. (2009) Livestock waste treatment systems for reducing
environmental exposure to hazardous enteric pathogens: Some considerations. Bioresource Technology 100(22):
5395-5398.
Trask, J.R., Kalita, P.K., Kuhlenschmidt, M.S., Smith, R.D., and Funk, T.L. (2004) Overland and near-surface
transport of Cryptosporidium parvum from vegetated and nonvegetated surfaces. Journal of Environmental Quality
33: 984-993.
USEPA (U.S. Environmental Protection Agency) (1986) Ambient Water Quality Criteria for Bacteria, Office of
Water, EPA440/5-84-002,Washington, DC,
December 2010 121
-------
U.S. Environmental Protection Agency
USEPA (2000) Science Policy Council Handbook: Risk Characterization. Prepared by the Office of Science Policy,
Office of Research and Development, Washington, DC.
USEPA (2005a) Appendices to the Occurrence and Exposure Assessment for the Final Long Term 2 Enhanced
Surface Water Treatment Rule. Office of Water, Washington, DC.
USEPA (2005b) Detecting and Mitigating the Environmental Impact of Fecal Pathogens Originating from Confined
Animal Feeding Operations: Review, EPA/600/R-06/021. National Risk Management Research Laboratory
Cincinnati, OH.
USEPA (2006) National Primary Drinking Water Regulations: Long Term 2 Enhanced Surface Water Treatment
Rule (LT2ESWTR); Final Rule. 40CFR Parts 9, 141 and 142, Volume 71, Number 654,
USEPA (2007) Critical Path Science Plan for the Development of New or Revised Recreational Water Quality
Criteria, 823-R-08-002, Office of Water, Office of Research and Development.
USEPA (2009a) Review of Published Studies to Characterize Relative Risks from Different Sources of Fecal
Contamination in Recreational Water, EPA 822-R-09-001, Office of Science and Technology, Washington, DC.
USEPA (2009b) Review ofZoonotic Pathogens in Ambient Waters, EPA 822-R-09-002, Office of Science and
Technology, Washington, DC.
USEPA (2010) Microbiological Risk Assessment Tools, Methods, and Approaches for Water Media. In preperation,
Office of Science and Technology, Washington, DC.
USDA FSIS (U.S. Department of Agriculture Food Safety and Inspection Service) (2009). Serotypes Profile of
Salmonella Isolates from Meat and Poultry Products. January 1998 through December 2007. URL
www.fsis.usda.gov/Science/Serotvpes Profile Salmonella Isolates/index.asp
van den Berg, H., Lodder, W., van der Poel, W., Vennema, H., and de Roda Husman, A.M. (2005) Genetic diversity
of noroviruses in raw and treated sewage water. Research in Microbiology 156(4): 532-540.
vanHeerden, I, Ehlers, M.M., Vivier, J.C., and Grabow, W.O.K. (2005a) Risk assessment of adenoviruses detected
in treated drinking water and recreational water. Journal of Applied Microbiology 99: 926-933.
vanHeerden, I, Ehlers, M.M., and Grabow, W.O.K. (2005b) Detection and risk assessment of adenoviruses in
swimming pool water. Journal of Applied Microbiology 99: 1256-1264.
Vanotti, M.B., Milkier, P.O., Hunt, P.O., and Ellison, A.Q. (2005) Removal of pathogen and indicator
microorganisms from liquid swine manure in multi-step biological and chemical treatment. Bioresource Technology
96(2): 209-214.
Vereen, E.I, Lowrance, R., Cole, D.I, and Lipp, E.K. (2007) Distribution and ecology of Campylobacters in coastal
plain streams (Georgia, United States of America). Applied and Environmental Microbiology 73(5): 1395-1403.
Vinneras, B. (2007) Comparison of composting, storage and urea treatment for sanitising of faecal matter and
manure. Bioresource Technology 98(17): 3317-3321.
Wade, S.E., Mohammed, H.O., and Schaaf, S.L. (2000) Prevalence of Giardia sp., Cryptosporidium parvum, and
Cryptosporidium muris (C. andersoni) in 109 dairy herds in five counties of southeastern New York. Veterinary
Parasitology 93: 1-11.
Wade, T.J., Pai, N., Eisenberg, I, and Colford, J.M. (2003) Do US EPA water quality guidelines for recreational
waters prevent gastrointestinal illness? A systematic review and meta-analysis. Environmental Health Perspectives
111(8): 1102-1109.
December 2010 122
-------
U.S. Environmental Protection Agency
Wade, T.J., Calderon, R.L., Sams, E., Beach, M, Brenner, K.P., Williams, A.H., and Dufour, A.P. (2006) Rapidly
measured indicators of recreational water quality are predictive of swimming-associated gastrointestinal illness.
Environmental Health Perspectives 114(1): 24-28.
Wade, T.J., Calderon, R.L., Brenner, K.P., Sams, E., Beach, M., Haugland, R., Wymer, L., and Dufour, A.P. (2008)
High sensitivity of children to swimming-associated gastrointestinal illness: results using a rapid assay of
recreational water quality. Epidemiology 19(3): 375-383.
Ward, R.L., Bernstein, D.I., Young, B.C., Sherwood, J.R., Knowlton, D.R., and Schiff, G.M. (1986) Human
rotavirus studies in volunteers: determination of infectious dose and serological response to infection. Journal of
Infectious Diseases 154(5): 871-879.
Warnick, L.D., Kaneene, J.B., Ruegg, P.L., Wells, S.J., Fossler, C.P., Halbert, L.W., and Campbell, A.M. (2003)
Evaluation of herd sampling for Salmonella isolation on midwest and northeast US dairy farms. Preventive
Veterinary Medicine 60: 195-206.
Weaver, R.W., Entry, J.A., and Graves, A. (2005) Numbers of fecal streptococci and Escherichia coli in fresh and
dry cattle, horse, and sheep manure. Canadian Journal of Microbiology 51(10): 847-851.
Weijtens, M.J.B.M., van der Plas, I, Bijker, P.G.H., Urlings, H.A.P., Koster, D., van Logtestijn, J.G., and Huis in't
Veld, J.H.J. (1997) The transmissio n of campy lobacter in piggeries; an epidemic logical study. Journal of Applied
Microbiology 83(6): 693-698.
Weijtens, M.J.B.M., Reinders, R.D., Urlings, H.A.P., and Van der Plas, J. (1999) Campylobacter infections in
fattening pigs; excretion pattern and genetic diversity. Journal of Applied Microbiology 86(1): 63-70.
Wells, S.J., Fedorka-Cray, P., Dargatz, D.A., Ferris, K., and Green, A. (2001) Fecal shedding of Salmonella spp. by
dairy cows on farm and at cull cow markets. Journal of Food Protection 64(1): 3-11.
Werber, D., Mason, B. W., Evans, M.R., and Salmon, R.L. (2008) Preventing household transmission of Shiga toxin-
producing Escherichia coli O157 infection: promptly separating siblings might be the key. Clinical Infectious
Disease 46(8): 1189-1196.
WERF (Water Environment Researf Foundation) (2009) Report on the Expert Scientific Workshop on Critical
Research and Science Needs for the Development of Recreational Water Quality Criteria for Inland Waters. Water
Environmet Research Foundation. Project PATH4W09, Alexandria, VA.
Wesley, I.V., Wells, S.J., Harmon, K.M., Green, A., Schroeder-Tucker, L., Glover, M., and Siddique, I. (2000) Fecal
shedding of Campylobacter wAArcobacter spp. in dairy cattle. Applied and Environmental Microbiology 66(5):
1994-2000.
Wetzel, A.N., and LeJeune, J.T. (2007) Isolation of Escherichia coli O157:H7 strains that do not produce Shiga
toxin from bovine, avian and environmental sources. Letters in Applied Microbiology 45: 504-507.
WHO (World Health Organization) (2004a) Guidelines for Drinking-Water Quality. Third edition. Volume 1.
Geneva, Swizerland.
WHO (2004b) Waterborne Zoonosos: Identification, Causes, and Control. J. Cotruvo, A. Dufour, G. Rees, J.
Bartram, R. Carr, G.F. Craun, R. Payer, and Gannon, V.P.J. (eds), London, UK, IWA Publishing.
Whyte, P., Collins, J.D., McGill, K., Monahan, C., and O'Mahony, H. (2001) The effect of transportation stress on
excretion rates of Campylobacters in market-age broilers. Poultry Science 80(6): 817-820.
Wobeser, G., and Finlayson, M. (1969) Salmonella typhimurium infection in house sparrows. Archives of
Environmental Health 19(6): 882-884.
December 2010 123
-------
U.S. Environmental Protection Agency
Wong, J.W.C., and Selvam, A. (2009) Reduction of indicator and pathogenic microorganisms in pig manure through
fly ash and lime addition during alkaline stabilization. Journal of Hazardous Materials 169(1-3): 882-889.
Wong, M, Kumar, L., Jenkins, T.M., Xagoraraki, I., Phanikumar, M.S., and Rose, J.B. (2009) Evaluation of public
health risks at recreational beaches in Lake Michigan via detection of enteric viruses and a human-specific
bacteriological marker. Water Research 43(4): 1137-1149.
Xagoraraki, L, Kuo, D.H.-W., Wong, K., Wong, M., and Rose, J.B. (2007) Occurrence of human adenovirus at two
recreational beaches of the Great Lakes. Applied and Environmental Microbiology 73(24): 7874-7881.
Xiao, L., Payer, R., Ryan, U., andUpston, S. (2004) Cryptosporidium taxonomy: recent advances and indications
for public health. Clinical Microbiology Reviews 17(1): 72-97.
Xiao, L., Moore, IE., Ukoh, U., Gatei, W., Lowery, C.J., Murphy, T.M., Dooley, J.S., Millar, B.C., Rooney, P.J.,
and Rao, J.R. (2006) Prevalence and identity of Cryptosporidium spp. in pig slurry. Applied and Environmental
Microbiology 72(6): 4461-4463.
You, Y., Rankin, S.C., Aceto, H.W., Benson, C.E., Toth, J.D., and Dou, Z. (2006) Survival of Salmonella enterica
serovar Newport in manure and manure-amended soils. Applied and Environmental Microbiology 72(9): 5777-5783.
Zhou, L., Kassa, H., Tischler, M.L., and Xiao, L. (2004) Host-adapted Cryptosporidium spp. in Canada geese
(Branta canadensis). Applied and Environmental Microbiology 70(7): 4211-4215.
Ziemer, C.J., Bonner, J.M., Cole, D., Vinje, J., Constantini, V., Goyal, S., Gramer, M., Mackie, R., Meng, X.J.,
Myers, G., and Saif, L.J. (2010) Fate and transport of zoonotic, bacterial, viral, and parasitic pathogens during swine
manure treatment, storage, and land application. Journal of Animal Science 88: E84-E94.
Zmirou, D., Pena, L., Ledrans, M., and Letertre, A. (2003) Risks associated with the microbiological quality of
bodies of fresh and marine water used for recreational purposes: summary estimates based on published
epidemiological studies. Archives of Environmental Health 58(11): 703-711.
Zuckerman, U., Gold, D., Shelef, G., and Armon, R. (1997) The presence of Giardia and Cryptosporidium in
surface waters and effluents in Israel. Water Science and Technology 35(11-12): 381-384.
December 2010 124
-------
U.S. Environmental Protection Agency
APPENDICES
Appendix A. Selected Peer-Reviewed QMRAs for Recreational Water Exposure
Table 21. Synopsis of selected peer-reviewed QMRAs of recreational water exposure
Study
Ashbolt
and Bruno
(2003)
Gerba et
al. (1996)
Jolis et al.
(1999)
Risk of Interest
Risk of GI illness
and respiratory
illnesses
associated with
recreational water
Risk of rota virus
infection from
recreational and
drinking water
exposures
Risk of
cryptosporidiosis
associated with
exposure at parks
and golf courses
irrigated with
tertiary reclaimed
water.
Microorganism(s)
Enteric viruses
Adenovirus
Rotavirus
Cryptosporidium
parvum
Pathogen Concentration
and Variability
Ratio of pathogens to
enterococci assumed
relatively constant - data
on enterococci collected
during the study and
reported as number of
samples meeting a
compliance criterion
Drinking water
concentrations estimates
assumed to be 0.004 PFU/L
and 100 PFU/L, based on
review of the occurrence of
rotavirus in drinking water
and surface water and
assuming 99.99% removal
in treatment. Surface water
concentrations estimated to
be0.24/Land29/L(the
occurrence range).
Concentration of
Cryptosporidium parvum in
tertiary effluent set to the
arithmetic mean of six
samples (variability not
reported or considered).
Concentration in treated
secondary effluent taken as
2 logs less than the mean of
three samples of secondary
effluent.
Ingested Volume
or Mass
50 mL fixed
volume assumed
Ingested volumes
used were 100 mL
for recreational
exposure, 2 L for
child and adult
drinking water
exposure, and 4 L
for elderly
drinking water
exposure.
Assumed golfer
and park user
ingested volume
of 1 mL per
outing
Dose-Response
Exponential dose-
response model with r =
1 for enteric viruses.
Adenovirus dose-
response model (r =
0.4 17) for respiratory
illness associated
viruses
Beta-Poisson dose-
response model (a=
0.26, Ar50 = 5.62) used
for risk of infection.
Risk of clinical illness
assumed 0.5x risk of
infection. The fraction
of illnesses progressing
to mortality assumed
0. 1 % for the general
population and 1 .0% for
the elderly.
Exponential
Cryptosporidium
parvum model (r =
0.00467, 95%
confidence interval
O.00195, 0.0962>,no
information on
distributional form
assumed for r)
Ratio of illness to
infection set at 0.5.
Secondary
Transmission
Not considered
Secondary
transmission rates
discussed, but
details on
calculations not
provided
Not considered
Sensitivity Analysis
Not reported
Risks corresponding to
high and low
concentrations in
drinking water and
recreational water
presented
Not reported. Authors
critically assessed
findings in their study
and characterized the
study as preliminary.
December 2010
A-l
-------
U.S. Environmental Protection Agency
Study
Julian et
al. (2009)
Parkin et
al. (2003)
Roberts et
al. (2007)
Risk of Interest
Risk of rota virus
infection from
multiple exposure
routes for a child 6
years of age or
younger; exposure
routes were
fomite-to-mouth,
fomite-to-hand,
and hand-to-mouth
Risk of entero virus
infection to
sensitive
population via
recreation in water
receiving WWTP
effluent; study
was a data
collection and
problem
formulation effort
Risk of
cryptosporidiosis
associated with
fishing in an
urbanized stream
reach
Microorganism(s)
Rotavirus
Coxsakievirus A
andB
Echoviruses Human
entero viruses
Polioviruses
Cryptosporidium
Pathogen Concentration
and Variability
Virus density on fomite
assumed uniformly
distributed (0.00 1-10
virus/cm2). Inactivation
rate on fomite and hands
assumed normally
distributed (different mean
and standard deviation for
fomite and hand
distributions)
Anecdotal data on virus
occurrence in swimming
water reported, but no
characterizations of
temporal variation in
viruses found in a literature
search
Number of oocysts ingested
per month via hand-to-
mouth transmission or in
consumption of fish was
assumed Poisson-
distributed; distribution
parameters estimated using
occurrence of oocysts in
hand- washings and on fish
Ingested Volume
or Mass
Transfer
efficiency from
fomite to mouth
and hand-to-
mouth assumed
normally
distributed with a
mean of 41% and
aSDof25%.
Transfer
efficiency from
fomite to hand
assumed normally
distributed with a
mean of 36% and
SD deviation of
26%.
Not considered
Not calculated
separately from
pathogen
concentration
estimate
Dose-Response
Beta-Poisson dose-
response model (a=
0.26,^0 = 5.62) used
for risk of infection
Epidemiology studies
indicate that children at
greater risk than adults
for entero virus
infection; the effects of
dose-response and
exposure not
differentiated; authors
noted there are no
known dose-response
relationships for
children
Exponential (r =
0.00419). The dose-
response parameter was
treated as a random
variable, although the
distributional form used
is not reported.
Secondary
Transmission
Not considered
Not considered
Not considered
Sensitivity Analysis
Model was run with a
parameter set to either
the 25th or 75th
percentile value of its
distribution and all
other parameters at the
median value.
Sensitivity to a
parameter is assessed
based on the ratio of the
p25 to the p75 estimated
risks.
Not relevant
Sensitivity analysis
results reported, but
details of the method
not provided
December 2010
A-2
-------
U.S. Environmental Protection Agency
Study
Schijven
andde
Roda
Husman
(2006)
Schoen
and
Ashbolt
(2010)
Risk of Interest
Risk of infection
for occupational
and sport divers in
fresh and marine
water
Risk of GI illness
from swimming in
human- and gull-
impacted surface
water
Microorganism(s)
Campylobacter
jejuni
Enteroviruses
Salmonella
Campylobacter
Giardia
Cryptosporidium
Noro virus
Pathogen Concentration
and Variability
Both pathogens assumed
log-normally distributed,
with the reported lowest
and highest values (in the
literature) assumed to be
the 99% confidence
interval values.
Study introduced the
relative risks QMRA
approach in which
pathogen densities are
drawn from distributions
based on reports for
specific fecal pollution
sources and referenced to
indicator levels for the
same fecal pollution
sources. Pathogens in gull
wastes were
Campylobacter and
Salmonella. All pathogens
assumed present in human
sewage.
Ingested Volume
or Mass
Ingested water
depended on diver
status
(recreational vs.
occupational),
setting (marine vs.
fresh vs.
swimming pool)
and on equipment
used, especially
mask type.
Reported ingested
volumes ranged
from 0-1 90 mL.
Number of dives
per year drawn
from an empirical
distribution.
Relative doses of
pathogens and
indicators are
used
Dose-Response
Hypergeometric (exact
beta Poisson) model
with a= 0.145 and ft =
8.007 was used for
dose-response for C.
jejuni
The rota virus
hypergeometric model
with a= 0.167 and fi =
0.191 used for dose-
response for
entero viruses; note that
this is an extremely
conservative assumption
Norovirus: Poisson-
stopped logarithmic
series
Salmonella: Gompertz
model for serotype
Bareilly
Campylobacter. two
alternative
parameterizations of the
exact beta-Poisson
model
Cryptosporidium :
exponential model
Giardia: exponential
model
Secondary
Transmission
Not considered
Not considered
Sensitivity Analysis
Annual risk of infection
differed significantly
with diver status
(occupational vs.
recreational),
equipment used, and
setting
Stochastic framework
used
December 2010
A-3
-------
U.S. Environmental Protection Agency
Study
Signer
and
Ashbolt
(2006)
Seller at
al. (2003)
Risk of Interest
Human exposure
to pathogens via
drinking water
when routine
pathogen
monitoring is
conducted
Risk of viral
gastroenteritis
associated with
recreational and
non-recreational
use of a river
downstream of a
wastewater
treatment plant
discharge. Two
wastewater
treatment
scenarios were
compared.
Microorganism(s)
Cryptosporidium
spp.
Model enteric virus
with clinical
features of rotavirus
Pathogen Concentration
and Variability
During base flow
conditions, untreated water
Cryptosporidium density is
log-normally distributed
with mean and standard
deviation of log-
transformed densities equal
to 3. 11 and 1.28,
respectively.
During event (rainfall)
conditions, untreated water
Cryptosporidium density is
log-normally distributed
with mean and standard
deviation of log-
transformed densities equal
to 5.27 and 0.61,
respectively.
Bacteriophage
concentration in raw
wastewater assumed
uniformly distributed in the
range Ixl04-5xl04.
Removal modeled for
treatment and removal and
mixing processes modeled
for discharged effluent. The
ratio of model enteric virus
concentration to
bacteriophage
concentration assumed log-
uniform distributed in the
range 0.001-1.0.
Ingested Volume
or Mass
Ingested (oral)
volume log-
normally
distributed with
mean and
standard deviation
of log-
transformed
densities equal to
-0.046 and 0.535,
respectively
Exposure factor
was a random
variable chosen
from uniform
distributions
whose ranges
were selected
based on observed
recreational use
by month and day
of the week
(weekday v.
weekend).
Dose-Response
Exponential model, r =
0.00419
Beta-Poisson (presented
in study in modified
form) with a assumed
uniformly distributed in
the range 0.1 5-0.42 and
Pin the range 0.3-2. 3.
Secondary
Transmission
Not considered
Dynamic
population-based
model, including
individuals
infected from
activities other
than use of river
for recreation
Sensitivity Analysis
Model sensitivity was
assessed via
comparison of three
sampling scenarios
Univariate sensitivity
analyses for input
parameters
December 2010
A-4
-------
U.S. Environmental Protection Agency
Study
Seller et
al. (2006)
Seller et
al. (2010a)
Risk of Interest
Risk of infection
during full-body
contact recreation
in a non-POTW
impacted estuarine
recreational
waterbody
Risk of GI illness
during primary
contact with water
during recreation
Microorganism(s)
Rotavirus as a
representative
pathogen
Salmonella
Campylobacter
Giardia
Cryptosporidium
E. coli O15T.H7
Noro virus
Adenovirus
Rotavirus
Pathogen Concentration
and Variability
Rotavirus density was
based on a model calibrated
with empirical coliphage
data. The relationship
between coliphage density,
expected rotavirus density,
and fraction of total
pathogen load comprised
by rotavirus not presented
explicitly.
Study introduced the
reverse QMRA approach in
which pathogen densities
are inferred from QMRA
conducted with known
sources and illness rates.
Pathogen densities relative
to each other based on ( 1 )
observed relative densities
in POTW effluent and (2i)
the proportion of U.S. non-
food GI illness
Ingested Volume
or Mass
Hourly rate of
water ingestion
assumed;
swimmers were in
the water at
different times
and for different
durations
Point estimate
based on
arithmetic mean
of log-normal
distribution of
values reported by
Dufour et al.
(2006)
Dose-Response
Beta-Poisson (presented
in study in modified
form) with a assumed
uniformly distributed in
the range 0.125-0.5 and
P in the range 0.21-
0.84;
probability of
symptomatic response
range 0.1-0.45
Adenovirus:
Exponential model, r =
0.4172
Rotavius: beta-Poisson
model, a= 0.2531, 0=
0.4265
Norovirus; beta-Poisson
model a = 0 04 B =
0 055
Salmonella'.
approximate beta-
Poisson model, a=
0.04, B = 2884
E. co/z0157:H7:
approximate beta-
Poisson model, a= 0.4,
0=45.9
Campylobacter. exact
beta-Poisson model, a=
0.024, B= 0.011
Cryptosporidium :
Exponential model, r =
0.09
Giardia'. exponential
model, r= 0.01 99
Secondary
Transmission
Secondary
transmission
modeled via a
deterministic time-
dependent
transmission
model accounting
for the immune
status of the
population
Not considered
Sensitivity Analysis
Sensitivity analyses
performed for several
variables; variables set
to low, medium and
high values to
determine whether their
variation changed the
study findings
Stochastic framework
used and model results
validated via
comparison of modeled
time to illness onset
distribution with
observed time to illness
onset distribution
December 2010
A-5
-------
U.S. Environmental Protection Agency
Study
Seller et
al.
(2010b)
Steyn et
al. (2004)
Risk of Interest
Risk of GI illness
during primary
contact with water
during recreation
Risk of infection
via drinking water
or water-borne
recreation
Microorganism(s)
Salmonella
Campylobacter
Giardia
Cryptosporidium
E.coliOl57:R7
Noro virus
Adenovirus
Rotavirus
Salmonella
Pathogen Concentration
and Variability
Study used the relative
risks QMRA approach in
which pathogen densities
were drawn from
distributions based on
reports for specific fecal
pollution sources and
referenced to indicator
levels for the same fecal
pollution sources
Salmonella density
determined during
monitoring; calculations
performed for the GM
value (167 CFU/lOOmL),
the minimum value (36)
and the maximum value
(883)
Ingested Volume
or Mass
Relative doses of
pathogens and
indicators used
For full contact
recreation,
ingested volume
assumed 100 mL
Dose-Response
Adenovirus:
Exponential model, r =
0.4172
Rotavius: beta-Poisson
model, a= 0.2531, 0=
0.4265
Norovirus: beta-Poisson
model, a=Q.Q4,fi =
0.055
Salmonella:
approximate beta-
Poisson model, a=
0.04, p = 2884
E. co/z0157:H7:
approximate beta-
Poisson model, a= 0.4,
ft =45. 9
Campylobacter. exact
beta-Poisson model, a=
0.024, p= 0.011
Cryptosporidium :
exponential model, r =
0.09
Giardia: exponential
model, r= 0.01 99
Approximate beta
Poisson dose-response,
with a=Q. 3126 andN50
= 23,600
Secondary
Transmission
Not considered
Not considered
Sensitivity Analysis
Stochastic framework
used and model results
validated via
comparison of modeled
time to illness onset
distribution with
observed time to illness
onset distribution
Not reported
December 2010
A-6
-------
U.S. Environmental Protection Agency
Study
van
Heerden
etal.
(2005a)
Wong et
al. (2009)
Risk of Interest
Risk of human
adeno virus
infection via
drinking water or
recreational water
exposure
Risk of enteric
virus infection
associated with
swimming at
coastal beaches
impacted by
POTW discharges
Microorganism(s)
Adenovirus
Adenovirus
Pathogen Concentration
and Variability
Adenovirus density
assumed Poisson-
distributed (in time, not
space) with the distribution
mean determined from
frequency of positive
determinations among
drinking water and surface
water samples. Mean
adeno virus densities (in
viruses per 100 mL) were
0.0014 and 0.00245 for two
drinking water, 0.0546 for
a river water, and 0.0097
for water behind a dam.
Experimental distribution
for adeno virus occurrence
based on Regression on
Order Statistics to account
for non- detect observations.
Ingested Volume
or Mass
Drinking water
consumption rate
fixed at 2 L per
capita p er day
and recreational
water
consumption rate
fixed at 30 mL per
capita per day
100 mL/day
Dose-Response
Exponential model was
used for adenovirus
dose-response; the
model parameter was
not explicitly provided,
although based on the
citation provided in the
study, it can be inferred
to be that for inhalation
of adenovirus aerosols, r
= 0.417
Exponential model, r =
0.417 (based on data for
inhalation of adenovirus
aerosols)
Secondary
Transmission
Not considered
Not considered
Sensitivity Analysis
Univariate sensitivity
analyses conducted to
assess the impact of
consumption rates,
dose-response
parameters and
recovery rates on risk
estimates
Not reported
December 2010
A-7
-------
U.S. Environmental Protection Agency
Appendix B. Data Summary Reference Pathogens in Livestock and Human Waste
B.I Reference Pathogens in Livestock Manure
Table 22. Reported Salmonella densities in livestock feces and other matrices
Study
Boes et al.
(2005)
Byrd(1998)
Haley et al.
(2009)
Hutchison et
al. (2004)
Hutchison et
al. (2004)
Sournce/Media
Swine manure slurry
from 62 herds; Danish
farms
Fecal material and
poultry litter from
hatcheries
Stream water
Fresh pig manure
Fresh chicken manure
Serotype
Typhimurium
Typhimurium
spp.
spp.
spp.
Description
Samples drawn from swine
manure slurry and from soil
after application of swine
manure slurry
Day-old chicks challenged with
100, W4 or \06 Salmonella
typhimurium by gavage; litter
and cecal contents monitored for
17 days
Water samples from a mixed use
(livestock, on- site septic system,
small community) watershed
sampled for Salmonella
Multiple commercial farms
Multiple commercial farms
Abundance
Salmonellae detected in all slurry samples. Average
Salmonella Typhimurium density was 0.2 CFU/g
(note - not Iog10 CFU); maximum density estimated
to be 2500 CFU/g for a sub-clinically-infected herd;
observed abundance among 112 slurry samples 33%
of samples with < 0. 1 MPN, 1 3% of samples between
0.1 and 1 MPN, 28% between 1 and 10 MPN, 12%
between 10 and 1 10 MPN, and 14% > 100 MPN.
Pens containing chicks inoculated with 100
Salmonellae: 2.05 to 3.03 Iog10 CFU/g litter (n = 10)
Pens containing chicks inoculated with 104
Salmonella: 2.39 to 4.55 Iog10 CFU/g litter (n = 10)
Pens containing chicks inoculated with 106
Salmonella: 3.65 to 4.42 Iog10 CFU/g litter (n = 10)
Geometric mean of Salmonella in water did not vary
greatly among sampled sites; the highest and lowest
mean densities were 0.746 MPN/100 mL and 0.496
MPN/lOOmL
Geometric mean of 600 CFU/g (n = 10); maximum
observation of 78,000 CFU/g
Geometric mean of 220 CFU/g (n = 12); maximum
observation of 22,000 CFU/g
Notes
Authors proposed a
polynomial survival model
for Salmonella in soil
Fecal colonization rate and
Salmonella count in fecal
contents varied according to
challenge dose; the number
of chicks inoculated (5%,
10%, 25%, and 50% of
chicks in a pen) did not
influence the overall
incidence of infection in the
pen
Wastes taken from farms
throughout Great Britain and
results are believed
representative of overall
prevalence in the region
Wastes taken from farms
throughout Great Britain and
results believed
representative of overall
prevalence in the region
December 2010
B-l
-------
U.S. Environmental Protection Agency
Study
Hutchison et
al. (2004)
Sournce/Media
Fresh cattle manure
Serotype
spp.
Description
Multiple U.K. commercial farms
Abundance
Geometric mean of 2100 CFU/g (n = 62); maximum
observation of 580,000 CFU/g
Notes
Wastes taken from farms
throughout Great Britain and
results believed
representative of overall
prevalence in the region.
Salmonella density higher in
stored manure than fresh
manure
December 2010
B-2
-------
U.S. Environmental Protection Agency
Table 23. Reported Campylobacter spp. densities in livestock manure and other matrices
Study
Blaseretal. (1980)
Cox et al. (2002)
Domer et al. (2004)
Domer et al. (2004)
Domer et al. (2004)
El-Shibiny et al. (2005)
Hutchison et al. (2005)
Hutchison et al. (2005)
Hutchison et al. (2005)
Moriarty et al. (2008)
Source/Media
Human feces
Chicken feces
Poultry (broiler)
feces
Nursing or weaner
pigs
Sows and gilts
Poultry
Cattle
Swine
Poultry
Dairy cattle
Strain
jejuni
spp.
spp.
spp.
spp.
spp.
spp.
spp.
spp.
spp.
Description
C. jejuni recovered and enumerated from
stools from 8 persons with suspected
campylobacteriosis; C. jejuni identified
an all 8 samples
Results are the composite of samples
taken from 35 commercial broiler farms;
results segregated by age of chicken
Abundance data from multiple studies
were pooled and fit to a gamma
distribution
Abundance data from a single study
(Weijtens et al., 1999) fit to a gamma
distribution
Abundance data from two studies
(Weijtens et al., 1997; Weijtens et al.,
1 999) fit to a gamma distribution
Estimates based on multiple published
studies
Composite samples of manure from pens
collected
Composite samples of manure from pens
collected
Composite samples of manure from pens
collected
Samples taken from 4 farms considered
to span conditions in New Zealand
Abundance
Median: 2.8xl08 CFU/g. Range:
6xl06-lxl09CFU/g
Breeders: 2.8-3.9 Iog10 CFU/g
feces
Broilers: 3.5-6.5 Iog10 CFU/g
feces
Gamma-distributed abundance,
distribution parameters (a, f!) =
(27.78,0.2558)
Gamma-distributed abundance,
distribution parameters (a, p) =
(4.419,0.6319)
Gamma-distributed abundance,
distribution parameters (a, p) =
(4.207,0.8859)
106-109 CFU/g excreta
320 CFU/g for fresh feces
530 CFU/g for stored feces
3 10 CFU/g for fresh feces
1600 CFU/g for stored feces
260 CFU/g for fresh feces
590 CFU/g for stored feces
For all seasons: median 430
CFU/g, range 15-1.8xl07 CFU/g.
Notes
Campylobacter less prevalent in
broilers (offspring) than
breeders, but shedding
(colonization) higher in broilers
than breeders
Prevalence of C. jejuni and C.
coli reported, but not related to
abundance in manure;
Campylobacter abundance bi-
modally distributed among
samples
December 2010
B-3
-------
U.S. Environmental Protection Agency
Study
Stanley etal. ( 1998)
Stanley etal. (1998)
Weijtens etal. (1997)
Weijtens etal. (1999)
Whyte etal. (2001)
Source/Media
Beef cattle feces
Dairy cattle feces
Sow feces at one
week prior to
delivery
Fattening pig feces
from 10 weeks of
age to 25 weeks
Poultry feces
Strain
spp.
spp.
spp.
spp.
spp.
Description
Fresh beef cattle sampled at slaughter
Fresh dairy cattle manure samples
collected in pens of 4 dairy herds in the
United Kingdom
Sow feces sampled and bacteria
enumerated 1 week prior to delivery
For each sampling event, six feces
samples were collected per pig; pigs were
monitored from birth and housed with 16
pigs each on an experimental farm
Fecal samples from sacrificed chickens
from 10 Irish farms enumerated for
Campylobacter; though samples were
analyzed before, during, and after
transport to a processing facility, the only
values quoted are for before transport;
studies conducted in Ireland
Abundance
610 MPN/g feces
Adult cows: 69.9 MPN/g feces
(SD3)
Calves: 33,000 MPN/g (SD 170)
5.0+1.1 logic CFU/g (farm 1, n =
5) and 3.6+0.4 Iog10 CFU/g (farm
2,w = 5)
At 13 weeks: mean fecal
Campylobacter density 4.1+0.7
logio CFU/g (n = 8 pigs, average
of 6 fecal samples per sampling
event per pig)
At 19 weeks: mean fecal
Campylobacter density 3.3+1 .0
Iog10 CFU/g (n = 8 pigs, average
of 6 fecal samples per sampling
event per pig)
At 25 weeks: mean fecal
Campylobacter density 2.0+0.1
logio CFU/g (n = 8 pigs, average
of 6 fecal samples per sampling
event per pig)
6.1 1+0.37 logio CFU/g feces for 5
farms and 6.61+0.38 Iog10 CFU/g
feces for 5 additional farms
Notes
Two peak periods (seasonal) of
shedding noted
Prevalence data for sows and
piglets also collected at 1 week,
4 weeks and 8 weeks post-
delivery
The abundance (and prevalence)
of Campylobacter varied weekly
and between fecal samples on a
given sampling event. Several
pigs had periods of non-
detectable fecal Campylobacter
between periods of high fecal
Campylobacter abundance.
Abundance was highest shortly
after colonization and generally
decreased with age.
December 2010
B-4
-------
U.S. Environmental Protection Agency
Table 24. Reported Cryptosporidium spp. densities in livestock manure and other matrices
Study
Atwill et al. (2003)
Atwill et al.
(2006)
Berry et al. (2007)
Heitman et al.
(2002)
Hutchison et al.
(2004)
Hutchison et al.
(2005)
Hutchison et al.
(2004)
Moriarty et al.
(2008)
Source/Media
Beef cow (>24 months)
feces, California
Beef cattle feces from
feedlot
Beef cattle feces from
feedlot
Manure from dairy cattle
Cattle manure
Fresh and stored pig
manure
Fresh and stored chicken
manure
Dairy cattle manure
Species
C. parvum
C. parvum
spp.
C. parvum
C. parvum
C. parvum
C. parvum
Spp.
Description
Manure samples from preparturient and
postparturient beef cows on three
California farms sampled and C. parvum
was enumerated via a sensitive method
Manure from 22 feedlots in 7 western
and central states sampled in the period
8/2000 to 1/2002
Manure from beef feedlots was sampled
(composite samples) each 4 weeks during
a 26 month study
Manure from two dairy operations
collected from pasture
Manure samples collected from
throughout Great Britain
Composite samples from fresh and stored
manure were collected between April
2000 and December 2002
Composite samples from fresh and stored
manure collected between April 2000
and December 2002
Samples taken from freshly-deposited
manure
Abundance
For samples positive for C.
parvum, the arithmetic mean
oocyst density was 3.38 oocysts/g
feces and SD 2.64 oocysts/g feces
Among samples positive for C.
parvum, the geometric mean was
447 oocysts/g manure (range 203-
7702 oocysts/g)
Average: 14 oocysts/g
Range: 0.5 oocysts/g manure to
1510 oocysts/g manure
Mean densities in manure from the
two farms were 18.8 and 490
oocysts/g (considering only
positive samples)
For fresh manure GM density 1 9
oocysts/g (n = 44)
Maximum density 3500
For stored manure, GM density 10
oocysts/g (n = 12)
Maximum density 480
GM densities 58 for fresh manure,
and 33 for stored manure
No C. parvum identified in any
chicken samples
Among positive samples,
Cryptosporidium density ranged
from 1-25 oocysts/g feces.
Notes
No significant difference in
prevalence or shedding of C.
parvum between preparturient
and postparturient cows
C. parvum detected in only 0.2%
of samples; abundance data fit
with a negative binomial
distribution
Cryptosporidium spp. identified
in 58% of composite manure
samples collected over a 26-
month study
C. muris not detected in any
fecal samples
Prevalence low in the herds
studied
December 2010
B-5
-------
U.S. Environmental Protection Agency
Study
Sturdee et al.
(2003)
Wade et al. (2000)
Wade et al. (2000)
Source/Media
Cattle feces
Dairy cattle feces
Dairy cattle feces
Species
C. parvum
C. muris
C. parvum
Description
Rectal and recently-deposited fecal
samples collected at a farm with beef and
dairy cattle and calf rearing operations
Fecal samples collected rectally from
dairy cattle at 109 farms in southeastern
New York; data were stratified by cattle
age
Fecal samples collected rectally from
dairy cattle at 109 farms in southeastern
New York
Abundance
Description
Bull beef
Dairy cow
Calf, home-
bred
Calf, bought-in
Mean
(oocysts/g)
1371
1778
107,025
24,448
Mean: 24,413 oocysts/g feces
Range: 1 to 100,000 oocysts/g
feces
Mean: 21,090 oocysts/g feces
Range: 1 to 79,040 oocysts/g feces
Notes
Highest observed density was
280,000 oocysts/g feces for a
home-bred calf
C. muris recovered from animals
with a wide range of ages
C. parvum was recovered only
from calves <30 days of age
December 2010
B-6
-------
U.S. Environmental Protection Agency
Table 25. Reported Giardia spp. densities in livestock manure and other matrices
Study
Hutchison et al. (2004)
Hutchison et al. (2004)
Ralston et al. (2003)
Heitman et al. (2002)
Heitman et al. (2002)
Wade et al. (2000)
Source/Media
Cattle farmyard
manures and slurries
Swine farmyard
manures and slurries
Range beef calf and
dam manures
Dairy cattle manure
Pig manure
Dairy cattle manure
Species
G. intestinalis
G. intestinalis
spp.
spp.
spp.
spp.
Description
Resultsfor samples collected
throughout Great Britain
Results for samples collected
throughout Great Britain
Fecal samples collected from calves
and dams from range operations in
Canada
Fecal samples collected from farms
in Canada
Fecal samples collected from farms
in Canada
Fecal samples collected from 2 12
farms in southeastern New York
Abundance
Geometric mean and maximum
cyst densities 10 and 5000 cysts/g,
respectively
Geometric mean and maximum
cyst densities 68 and 160,000
cysts/g, respectively
Giardia abundance in feces varied
with animal age group. Density
ranged from 0 at 1 week of age to a
maximum of 2230 cysts/g (range
0-574,933 cysts/g of feces) of
feces at 5 weeks of age. The
geometric mean decreased after
week 5 to a low of 2 cysts/g at 25-
27 weeks of age
Mean cyst range 1 .5-29.9 cysts/g
Mean cyst density 16.1 cysts/g
1-85,217 cysts, mean of 3039
cysts/g feces
Notes
December 2010
B-7
-------
U.S. Environmental Protection Agency
Table 26. Reported E. coli O157:H7 densities in livestock manure and other matrices
Study
Cornick and
Helgerson
(2004)
Hutchison et
al. (2004)
Kudva et al.
(1998)
Kudva et al.
(1998)
Kudva et al.
(1998)
Animal
Swine
Swine
Sheep
Cattle
Cattle
Source/Media
Feces
Manure
Manure pit
Manure
Manure slurry
Description
3-month old pigs challenged
with graded doses ofE. coli
O157:H7; pigs housed indoors
on concrete floors or decks;
experiments were conducted in
Iowa
Samples were collected from
multiple commercial farms in
the UK
Composite samples (from
manure pits receiving waste
from multiple animals)
collected and enumerated for
E. coli O157:H7; experiments
were conducted in Idaho
Composite samples (from
manure pits receiving waste
from multiple animals) were
collected and enumerated for
E. coli O157:H7; experiments
conducted in Idaho
Untreated slurries and treated
slurries (the retentate post-
storage and separation) were
sampled and enumerated for E.
coli O157:H7; experiments
conducted in Idaho
Abundance
Shortly after inoculation fecal E.
coli density ranged between 10
and!07CFU/g.
Two weeks after inoculation,
fecal E. coli Ol 57:H7 density
ranged from 50 to 1000 CFU/g.
Two months after inoculation,
fecal E. coli density ranged from
non-detect to 104 CFU/g.
Geometric mean of 3900 CFU
E. co/zO157/g(« = 15).
Highest observed density was
750,000 CFUE. coli O157/g
1 . 1 5 x 108 CFU/g feces from a
composite sample
Two samples yielded 2.04x10
CFU/g feces and 4.35xl08
CFU/g feces
Two samples of untreated slurry
yielded 1.02xl06 CFU/mL and
2.36xl06CFU/mL
A single sample of treated slurry
yielded 2.35xl06 CFU/mL
Notes
Swine infectious dose ofE. coli O157:H7 is
higher than that of cattle, resulting in lower
incidence of transmission ofE. coli O157:H7
between pigs than between cattle. Shedding
duration was dose-dependent, with shedding
lasting at least 2 weeks for all challenged
animals and for >2 months for some animals.
Prior to shedding, sheep experimentally
inoculated with£. coli O157:H7; some of the
animals contributing to the manure pit were
not infected.
Prior to shedding, cattle experimentally
infected withE. coli O157:H7; some of the
animals contributing to the manure pit not
infected
Prior to shedding, cattle were experimentally
infected with£. coli O157:H7; some of the
animals contributing to the manure pit not
infected
December 2010
B-8
-------
U.S. Environmental Protection Agency
B.2 Reference Pathogens in Treated Sewage
Table 27. Reported rotavirus densities in treated sewage
Setting
Activated sludge effluent
Secondary effluent
Unchlorinated secondary
effluent
Chlorinated secondary effluent
Secondary sewage effluent
Range
0-1500 /L (average 740)
1-21 fluorescent foci/L (GM =
9.8)
48-3228 (average 1012)/L
0-32 (average 9.6)/L
7.5-374 (GM = 41)/L
Study
Bates et al. (1984)
Hejkaletal. (1984)
Rao et al. (1987)
Rao et al. (1987)
Smith and Gerba (1982)
Table 28. Reported adenovirus densities in treated sewage
Setting
Secondary effluent
Treated wastewater
Secondary effluent
Secondary effluent
Unchlorinated secondary
effluent
Secondary effluent
Chlorinated secondary effluent
Chlorinated secondary effluent,
multiple plants
WWTP effluent
Lake Michigan water
Secondary effluent
Range
594-9030 genome copies/L
2400 genome copies/mL
(relatively stable with season)
Ixl03-4xl04 genome copies/L
ND-54000 PCR detection units/L
(mean =3 90)
0-600 infectious units(IU)/L
(GM=250)
6.1x10-1.4x10 viral genome
copies/L (0 genome copies in
chlorinated secondary effluent)
0-1150IU/L(GM=300)
Mean of 7000 reverse
transcription-(RT-PCR) units/L
0-4000 MPN/L (estimated from
results presented graphically)
7-3800 viral particles/L
ND-2.5 MPN/L
Study
Bofill-Mas et al. (2006)
Carducci et al. (2008)
Fongetal. (2010)
Haramoto et al. (2007)
Irving and Smith (1981)
He and Jiang (2005)
Irving and Smith (1981)
Katayama et al. (2008)
Sedmak et al. (2005)
Xagoraraki et al. (2007)
MWRDGC (2008)
December 2010
B-9
-------
U.S. Environmental Protection Agency
Table 29. Reported norovirus densities in treated sewage
Setting
Treated wastewater
Chlorinated secondary effluent,
multiple plants
Treated wastewater
Treated sewage
Treated sewage
Treated sewage and river water
samples
Range
ND-0.64 copies/mL for genotype
I; below dectection-2.6
copies/mL for genotype 2
47-2900 RT-PCR units/L
2.2-3.0 logs of removal for
secondary treatment
0-1,650,000 DNA copies/L
896-7499 PCR-detectable units/L
1.8xl04-9.7xl07 genetic
equivalents/L
Study
Haramoto et al. (2006)
Katayama et al. (2008)
Haramoto et al. (2006)
Laverick et al. (2004)
Lodder and de Roda Husman (Lodder and
de Roda Husman, 2005)
Pusch et al. (2005)
Table 30. Reported Salmonella densities in treated sewage
Setting
Chlorinated secondary effluent
Treated sewage disinfected with
peracetic acid
WWTP effluent
Treated wastewater
Secondary effluent
Treated wastewater
Range
7.5xl05-8.5xl06 (viable only)
MPN
30 CFU/100 mL
43-460 MPN/100 mL
ND-9 MPN/L
3-573 MPN/L (mean 110)
0-60 MPN
Study
Desmont et al. (1990)
Jimenez-Cisneros et al. (2001)
Koivunen et al. (2001)
Langeland (1982)
Lemarchand and Lebaron (2003)
Teltschetal. (1980)
Table 31. Reported Campylobacter spp. densities in treated sewage
Setting
WWTP effluent
WWTP effluent
WWTP secondary effluent
WWTP disinfected secondary
effluent
Receiving water for WWTP
effluent
Range
262-79,000 organisms/100 mL
(paper reviews data from other
studies and enumeration
technique is not stated)
ND-3000 MPN/100 mL
(estimated based on graphical
data)
0-9MPN/100 mL
0
ND-0,500 CFU/100 mL
Study
Jones (2001)
Koenraad et al. (1994)
Stampietal. (1993)
Stampietal. (1993)
Vereen et al. (2007)
December 2010
B-10
-------
U.S. Environmental Protection Agency
Table 32. Reported Cryptosporidium spp. densities in treated sewage
Setting
Tertiary effluent
Treated wastewater
Teritary effluent
Secondary effluent
Secondary effluent
Secondary effluent
Secondary effluent
Secondary effluent (multi-region
study)
Treated wastewater
Secondary effluent
Secondary effluent
Tertiary effluent
Treated wastewater
Treated wastewater
Range
Mean 37 oocysts/L (SD = 9)
<10-60 oocysts/L
Mean = 0.21 oocysts/L (SD =
0.06)
2-390 oocysts
4-8 oocysts/L
0.03-9.6 oocysts/L
ND-209 oocysts/L (mean = 0.91)
< 0.1 to 40. 8 oocysts/L
1-120 oocysts/L (GM = 4)
ND-343 oocysts/L
100^4,500 oocysts/L
Mean of 0.0003 oocysts/L
0.06-1. 15 oocysts/L
8.3-8.05 oocysts/L (n = 3)
Study
Bonadonna et al. (2001)
Bukharietal. (1997)
Carraro et al. (2000)
Castro-Hermida et al. (2008)
Cheng et al. (2009)
Ferguson et al. (2009)
Lemarchand and Lebaron (2003)
McCuin and Clancy (2006)
Payment etal. (2001)
Robertson et al. (2000)
Robertson et al. (2006)
Rose etal. (2001)
Suwa and Suzuki (2001)
Zuckerman et al. (1997)
Table 33. Reported Giardia spp. densities in treated sewage
Setting
Secondary effluent from 7
wastewater treatment plants in
England
Secondary effluent from a large
Italian WWTP
Non-disinfected secondary effluent
numerous Spanish WWTPs
Settled non-disinfected secondary
effluent from 4 plants in Ireland
Effluent from a large Canadian
WWTP employing phyisco-
chemical treatment
Effluent from multiple Norwegian
WWTPs
Chlorinated secondary effluent from
multiple plants in the United States
Combined data for raw sewage and
WWTP effluent for Israeli plants
Range
< 10-720 cysts/L
0.77-2.4 cysts/L
2-6000 cysts/L
0-3 cysts/L (mean densities from
4 plants)
2-898 cysts/L
100-5 1,333 cysts/L
0.1-1.4xl02 cysts/L (mean= 12.8
cysts/L)
0-300 cysts/L
Study
Bukharietal. (1997)
Carraro et al. (2000)
Castro-Hermida et al. (2008)
Cheng et al. (2009)
Payment etal. (2001)
Robertson et al. (2006)
Rose et al. (2004))
Zuckerman et al. (1997)
December 2010
B-ll
-------
U.S. Environmental Protection Agency
Table 34. Reported E. coli O157:H7 densities in treated sewage
Setting
Influent, primary effluent, and
secondary effluent (PCR of EHEC
toxicity factor)
Cow waste lagoon water
(PCR of 3 EHEC toxicity factors)
Municipal sewage treatment plant
serving, 400,000 (st^-carrying
bacteria)
Municipal sewage treatment plant
serving 5000 (.^-carrying bacteria)
Municipal sewage treatment plant
servingl,400,000 (str2-carrying
bacteria)
Municipal sewage treatment plant
serving 1500 (.^-carrying bacteria)
Municipal sewage treatment plant
serving 1500 (s/^-carrying bacteria)
Raw sewage (.str2-carrying bacteria)
Secondary effluent and tertiary
effluent (.str2-carrying bacteria)
Human wastewater (E. coli O157)
Range
0-1 (volume tested not available, only 1
sample [influent] had one PCR positive
signal)
68-2.3 x 104 MPN/lOOmL
1.6(±0.3) log(MPN + l)/mL
2(±0.4) log(MPN + l)/mL
1.9(±0.4) log(MPN + l)/mL
2.3(±0.2) log(MPN + l)/mL
1.2(±0.2) log(MPN + l)/mL
2.6 log(MPN + l)/mL
Below detection limit
10 to 100 CFU/100 mL
Study
Grant et al. (1996)
Chern et al. (2004)
Garcia- Aljaro et al. (2004)
Garcia- Aljaro et al. (2004)
Garcia- Aljaro et al. (2004)
Garcia- Aljaro et al. (2004)
Garcia- Aljaro et al. (2004)
Garcia- Aljaro et al. (2004)
Garcia- Aljaro et al. (2004)
Garcia- Aljaro et al. (2004)
December 2010
B-12
-------
U.S. Environmental Protection Agency
Appendix C. Shorebirds and Stormwater Reference Pathogen Literature Review
Although not modeled in this QMRA study, shorebirds and Stormwater are other non-point
sources of fecal pollution posing risks to humans. In this appendix, the routes by which fecal
pollution from these sources reaches recreation sites are described and the hazards posed by
those fecal pollution sources are summarized.
C. 1 Overview of Pathogen and FIB Loads Attributable to Shorebirds
Pathogen and indicator loads attributable to waterfowl can reach recreational water via multiple
routes and in significant densities. Routes by which waterfowl fecal indicators and pathogens
may reach the waters at recreational sites include the following (Figure 34):
• direct deposition as feces into the water column;
• direct deposition via mechanical transfer (e.g., carried to receiving water on the legs of
birds wading in sewage or sewage-impacted water) into the water column;
• resuspension of deposited organisms from sediment or suspension of organisms growing
in sediment;
• runoff of organisms (either deposited or progeny of deposited organisms) from soil,
vegetation or impervious areas near the recreation area; and
• advection of bird-origin FIB or pathogens from stocks (e.g., in wetlands hydraulically
connected to recreation site waters during high tides, as noted by He et al., 2007) when
the stocks become hydraulically connected to the recreational water during tides or
flooding.
Direct-deposition:
• In fecal material
• Mechanical
Advection from
offsite
• y v- ':•
Resuspension O *'")
from sediments-
Figure 34. Routes by which bird-origin FIB and pathogens reach recreation sites
December 2010
C-l
-------
U.S. Environmental Protection Agency
The densities of FIB and pathogens in bird feces depend upon whether or not birds are infected
(prevalence) and the abundance of the FIB and pathogens in the feces. Assuming that the
contamination is reasonably fresh, the ratio of pathogen to FIB densities in recreational water
primarily impacted by birds is roughly the same as that in the bird feces. This assumption does
not account for temporal variations in the FIB to pathogen ratio that might arise from
• growth of FIB (but not pathogens with the known exception of E. coli O157, and the
possible exceptions of Salmonella or Campylobacter, though no reports of growth of the
latter two pathogens were identified during preparation of this report) in sands and
sediment;
• different die-off rates for specific pathogens and FIB; and
• differences in FIB to pathogen ratio among the rainfall-driven loads (advection, runoff)
and loads that are relatively steady (direct deposition and resuspension).
In general, these effects are relatively minor compared to the impact of loading rates of
pathogens and FIB.
C.2 Cryptosporidium and Giardia in Shorebird Feces
To identify the pathogens found in bird feces, EPA conducted an initial literature review. The
results of that review indicate that protozoan and bacterial reference pathogens have been
isolated from birds frequently (e.g., see Hubalek, 2004) and viral reference pathogens have not
been reported in bird feces as summarized below.
Many waterfowl are known to carry Cryptosporidium and Giardia. Graczyk et al. (2008)
reviewed the open literature and reported numerous avian species known to harbor human-
infectious Cryptosporidium and Giardia. Theyreported the density of the oocysts and cysts in
some bird feces (Table 35) (based on Table 1 in Graczyk et al., 2008). Clearly, Cryptosporidium
and Giardia including species implicated in human infections, are prevalent and abundant in
gulls (Larus sp.), ducks (Anas sp.), Canadian geese (Branta canadensis), and other bird species
known to be prevalent near recreational water sites.
December 2010 C-2
-------
U.S. Environmental Protection Agency
Table 35. Avian species associated with Cryptosporidium and Giardia (SOURCE: adapted from Graczyk et al., 2008)
Pathogen
Species
Cryptosporidium
spp.
Cryptosporidium
spp.
Cryptosporidium
spp.
C. parvum
C. hominis
C. parvum
Giardia spp.
Giardia spp.
Avian Species
Larus spp.
Anas discour
A. cerca carolinensis
A. platyrhynchos
A. americana
Lophodytes cucullanus
Mergus merganser
Branta canadensis
B. canadensis
B. canadensis
Anas discour
A. cerca carolinensis
A. platyrhynchos
A. americana
Lophodytes cucullanus
Mergus merganser
B. canadensis
Comments
5% of fecal and 22% of cloacal lavage samples
positive; 64% and 83% of oocysts, respectively,
were viable
Migratory ducks; 49% of birds positive; PCR did
not confirm C. parvum; oocyst concentration range
0-2182/gfeces; mean 47.53+270 oocysts/g
Residential and migratory geese; 81 and 90% of
fecal samples from collection sites positive
Residential and migratory geese; 2.4% of samples
positive; novel avian genotypes identified; oocysts
acquired from local unhygienic sources
Migratory geese; oocysts infectious to neonatal
geese; oocyst concentration range 670-6900/g
feces; mean 3700 oocysts/g feces
Migratory ducks; 49% of birds positive; PCR did
not confirm G. lamblia; cyst concentration range 0-
29,293/g feces; mean 436+3525 oocysts/g
Migratory geese; cyst concentration range 750-
7900 cysts/g feces; mean 4 100 cysts/g feces
Reference
Smith et al.
(1993)
Kuhn et al.
(2002)
Kassa et al.
(2004)
Zhou et al.
(2004)
Graczyk et al.
(1998)
Kuhn et al.
(2002)
Graczyk et al.
(1998)
C.3 Pathogenic Bacteria in Shorebird Feces
Campylobacter species including the human-infectious C. jejuni and C. coll have been reported
for gulls (Hubalek, 2004; Kinzelman et al., 2008; Quessy and Messier, 1992); crows, magpies
and starlings (Ito et al., 1988); and domestic pigeons (Ito et al., 1988; Lillehaug et al., 2005),
They are likely common in other bird species. Reported prevalences are as high as 25%,
indicating the likelihood that large numbers of birds might be infected simultaneously and that
those birds have the potential to generate sufficiently high densities of Campylobacter to pose a
credible human health hazard.
Salmonella has been documented to occur in many birds in many settings (Alley et al., 2002;
Berg and Anderson, 1972; Butterfield et al., 1983; Casanovas et al., 1995; Cornelius, 1969;
Cruickshank and Smith, 1949; Duncan et al., 1983; Fenlon, 1981; Fricker, 1984; Girdwood et al.,
1985; Kapperud and Rosef, 1983; Karaguzel et al., 1993; Kirk et al., 2002; Kirkpatrick, 1986;
Levesque et al., 1993; Locke et al., 1973; McDonough et al., 1999; Mitchell and Ridgwell, 1971;
Palmgren et al., 2006; Quessy and Messier, 1992; Wobeser and Finlayson, 1969). As with
Salmonellae from animal operations, the hazard these bird-origin pathogens pose to humans is
December 2010
C-3
-------
U.S. Environmental Protection Agency
related to the serotypes that are present. A brief survey of the literature on bird-borne
Salmonella is presented below, with the primary intent of demonstrating the carriage of human
infectious serotypes of Salmonella by birds that might be present at recreational water sites.
Similar to Salmonella in livestock and humans, serotype prevalence among birds appears to vary
temporally and spatially. Fenlon (1983) found that gulls nesting near a sewage treatment plant
and feeding on sewage had a 55% carriage rate of Salmonella. There was general (though not
perfect) concordance between serotypes present in the raw and treated sewage and the serotypes
found in the gull feces. Interestingly, given practical limits on sampling frequency for sewage
effluent and the likelihood that prevalence of different Salmonella serotypes vary with time, the
gull feces may yield a more complete picture of Salmonella serotype presence in sewage than
individual samples drawn from the effluent. Palmgren et al. (2006) observed a 2.7% prevalence
of Salmonella spp. in Black-headed gulls at a site in Sweden, with the dominant serotype (> 50%
of isolates) being Typhimurium—a serotype important in human salmonellosis. The authors
found the S. Typhimurium DT195 isolates from gulls were related to those isolated from
domestic animals and humans, and hypothesized that Black-headed gulls might play a role in the
spread of S. Typhimurium in Sweden. Shorebirds other than gulls may be sources of human-
infectious Salmonella, as shown by Kirkpatrick (1986), who isolated S. Newport and S.
Typhimurium var Copenhagen from droppings in black-crowned night heron nests. The author
noted that, during the time period of the study, S. Newport and S. Typhimurium were the two
most common serotypes in human infections in the vicinity of the study site (Ocean County,
New Jersey, USA) and speculated that the herons were infected via sewage-impacted marine
water.
An estimate for density ofSalmonellae in gull droppings is provided by Levesque et al. (1993).
Among ring-billed gulls (Lams delawarensis) nesting along the St. Lawrence River in the
vicinity of Quebec City, mean concentration of Salmonellae on 3 different sampling days was
150, 230, and 12,000 CFU/g feces; the ratios of Salmonellae to fecal coliforms on those three
days were 6.25xlO"5, 2.09 xlO"4, and 2.31xlO"3, respectively. Among typed Salmonella isolates,
several serotypes potentially pathogenic in humans (brandenberg, agona, hadar, Stanley, and
Typhimurium) were identified.
Few studies have shown a connection between birds andE. coli O157 contamination of
recreational water, though two routes—mechanical transmission (attached to birds) and
transmission via fecal material of infected birds are possible. Hubalek (2004), in a review of
literature on pathogens in birds, noted that pathogenic strains of E. coli, such as E. coli O157:H7,
have been isolated from both healthy and diseased birds (both resident and migrant) including
Ardea cinerea (the grey heron), Branta canadensis (Canadian geese), Cygnus columbianus
(tundra swans), Uria aalge (the common murre), and Columbapalumbus (wood pigeons).
Cizek et al. (2000) achieved experimental infection of pigeons withE. coli O157. The infected
pigeons appeared asymptomatic, yet shed the pathogens for 14.8 ±3.4 days when infected with a
December 2010 C-4
-------
U.S. Environmental Protection Agency
dose of 105 CPU and 20.2 ±5.2 days when infected with a dose of 109 CPU. Based on this
finding, the authors considered it credible that pigeons may play a role in E. coli O157 infection
transmission. Shere et al. (1998) also hypothesized that birds may play a role in E .coli O157:H7
on dairy farms, based on genetic similarity between E. coli O157:H7 isolated from cattle and a
pigeon found on the same dairy farm. Foster et al. (2006) isolated STEC O157 from droppings
at a bird feeding station in Scotland. Potential hosts (known to feed at the station) include
blackbirds, greenfinches, chaffinches, house sparrows, or unobserved species. STEC O157 was
isolated from only 1 of 231 composite samples, which indicates that STEC O157 occurrence in
birds is relatively rare.
In summary, all of the bacterial reference pathogens occur in feces of birds. Campylobacter and
Salmonella of strains and types pathogenic to humans are prevalent in a variety of bird species.
E. coli O157:H7 has been observed in bird feces, but appears much less prevalent than
Campylobacter and Salmonella. The Campylobacter species and Salmonella serotypes observed
in bird feces often are similar to those prevalent in adjacent human populations.
C.4 Reference Pathogens in Storm water
In conducting the literature review for reference pathogens in animal and human-impacted water,
numerous articles were obtained with information describing the occurrence and densities of
reference pathogens in stormwater. Although these data are not used explicitly in the QMRAs
described in this report, these data are potentially valuable for future consideration. A summary
of the data that were found are summarized below (Table 36).
Table 36. Reported reference pathogen densities in stormwater-dominated water
Study
Arnone et al.
(2005)
Betancourt and
Rose (2005)
Cizek et al.
(2008)
Jiang et al.
(2005)
Pathogen
Cryptosporidium
spp.
Cryptosporidium
spp.
Cryptosporidium
spp.
Cryptosporidium
spp.
Prevalence
0-100%
25% (1/4
samples)
NA
88% (determined
by PCR) or 56%
(determined by
microscopy
Abundance
0-31oocysts/100L
<2-287
oocysts/lOOL (GM
= 72)
50-180
oocysts/lOOL
(based on
arithmetic means of
samples)
Not determined
Notes
Samples taken from five locations,
with features ranging from urban
high-density to wooded/pervious;
highest prevalence and abundance of
Cryptosporidium was in runoff from
the wooded area
Samples collected in Florida and
designated as "stormwater"; drainage
not described
Samples collected in five tributaries
to a drinking water reservoir; data
presented graphically as densities in
stormwater
Samples collected from streams
during rain events; genotypes
indicated nearly all isolates likely of
non-human origin
December 2010
C-5
-------
U.S. Environmental Protection Agency
Study
Till et al. (2008)
Arnone et al.
(2005)
Betancourt and
Rose (2005)
Till et al. (2008)
Betancourt and
Rose (2005)
Rose et al.
(1987)
Rajal et al.
(2007)
Rajal et al.
(2007)
Till et al. (2008)
Claudon et al.
(1971)
Pathogen
Cryptosporidium
spp.
Giardia spp.
Giardia spp.
Giardia spp.
Enteric viruses
Rotavirus
Adenovirus 40/41
Enteroviruses
Adenovirus
Salmonella
Prevalence
3% and 5% of
samples from
municipal and
forested
drainages,
respectively
0-100%
0% (0/4 samples)
7% of samples
from forested
and municipal
drainages
100% (4/4
samples)
2 out of eight
sites studied
lout of 61
samples (2%)
0 out of 61
samples (2%)
3 1% of samples
from a forested
drainage and
28% of samples
from a municipal
drainage
4/12 samples
(33%)
Abundance
Data only presented
graphically
0-377 oocysts/lOOL
Data reported
graphically
0.48-4.4 MPN/100
L (GM = 2)
0.237-0.25 MPN
PFU/L
230 genomes/L
ND
Not determined
Notes
Samples taken from five locations;
with features ranging from urban
high-density to wooded/pervious;
highest prevalence and abundance of
Giardia in runoff from the wooded
area
Samples collected in Florida and
designated as "stormwater"; drainage
not described
Giardia occurrence appeared
relatively insensitive to land use
Samples collected in Florida and
designated as "stormwater"; drainage
not described
Samples collected from recreational
water in regions without suspected
impacts from POTWs or animal
operations; authors speculated that
pathogens may have been of
swimmer origin
The authors speculated that the
estimated adenovirus density is an
underestimate.
High adenovirus occurrence in the
forested drainage attributed to a
single known source
Sample sites loctaed in a separate
storm sewer system upstream of
discharge from an experimental
animal operation; serotypes were, in
general, consistent with those
commonly causing human infection
December 2010
C-6
-------
U.S. Environmental Protection Agency
Appendix D. EPA Environmental Monitoring Program
The objective for EPA's environmental monitoring and sampling effort was to generate primary
data to characterize recreator exposure to fecal pathogens and FIB in surface water impacted by
agricultural activities through the analyses of overland transport inputs and in-stream processes.
The study design includes conducting rain simulation experiments in small plots amended with
either beef cattle manure, swine slurries, or poultry broiler litter. This appendix provides a
detailed description of the experiments, which are summarized in Section 2.12.
D.I Rainfall Simulation
D. 1.1 Experimental design, plot, and event description
The rainfall simulation experiments were conducted on 18 1.5 x 2 m plots divided in halves,
providing for a total of 36 0.75 x 2 m treatment plots. The plots located in USDA-owned land in
Oconee County, GA (33° 47'N, 83°23'W) are described in Butler et al. (2008). Each treatment
plot was delineated with galvanized sheet metal (23 cm width) placed into the ground to a depth
of 18 cm. The vegetation cover was maintained at 10 cm in height and consisted of a mixed crop
of fescue and bermuda grasses. The slopes for the treatment plots ranged from 8 to 12%. Two
rainfall simulators (Tlaloc 3000 type, Joern's Inc., West Lafayette, IN), were placed each on top
of one double plot. This type of rainfall simulator has been commonly used for nutrient and
pathogen transport studies (Soupir, 2003; Soupir et al., 2006). Baseline simulations were
conducted to determine background pathogen, FIB, and nutrient levels. Histograms were
initially used to identify frequency distributions of baseline runoff volumes and allowed us to
select plots within a specific range of volumes. During the rainfall simulation event, rainfall was
applied to 4 plots per day, 3 days per week, for a total of 12 plots per week for 3 consecutive
weeks after manure application.
Treatments consisted of manure applications from the following three animal types: swine
(liquid manure), beef cattle (solid manure), and poultry (broiler litter)—and a control treatment
(no manure application). Each treatment had three replications (plots) and three manure
application timings relative to rainfall simulation time. The manure was applied to the plots in a
completely randomized split plot design taking into consideration the type of manure and the
rainfall application regime (1 hour, 1 week, and 2 weeks after manure was applied to the plots).
The rainfall application rate was set at 6.125 pounds per square inch, which resulted in 2 to 4
inches of rain per hour. This rate was equivalent to a precipitation return period of < 100 years
for the Georgia piedmont area and sufficient to produce a surface runoff event in a reasonable
timeframe (30 minute to 3.5 hours, depending on the moisture conditions of the soil). After
runoff was produced, rainfall continued to be applied for 60 minutes. In the plots where rainfall
was not applied immediately after manure application (1-week and 2-week treatments), plastic
December 2010 D-l
-------
U.S. Environmental Protection Agency
covers were placed on the plots to protect against natural rain events. These enclosures were
placed well above the vegetation cover to allow for air circulation and heat exchange. The type
of plastic selected allowed for 75 to 80% of the UV light to penetrate. This experiment was
conducted three times, (October 2009, March 2010, and June 2010) to obtain sufficient data
points and to account for varying climatic conditions. Soil moisture was determined during the
March and June simulations. During the March 2010 simulation (Run B), it varied from 0.271 ±
0.042 m3/m3 prior to start the simulation to 0.466 ± 0.032 m3/m3 during the simulation.
D.I.2 Sample collection
Runoff was collected at the lower end of each plot by means of a stainless steel flume, at 5-
minute intervals for the duration of the event and was composited in a 40 gallon container to
determine cumulative runoff volumes. The color of the runoff varied depending on the type of
manure applied. Poultry and cattle produced runoff of a deep brown (poultry) to greenish color
(cattle), very high in suspended solids, while swine and control treatments produced light brown
runoff. Five samples from selected intervals (5, 10, 20, 30, and 60 minutes) were collected
directly from the flume (-500 mL) to determine E. coll and enterococci total densities. An extra
sample (1 L) was collected at 15 minutes for Clostridium analysis. After the microbial sample
was obtained at the selected time point, the remainder of the runoff was added to the 40 gallon
container. After each five minute addition, the container was weighed to determine the
cumulative runoff volume. Two composited samples (10L) were collected per run from the 40
gallon container for pathogen and FIB analysis (30-min composite and total composite). These
samples were analyzed for E. coli, enterococci, Clostridium spp., Cryptosporidium, Giardia,
Salmonella, E. coli O157, and Campylobacter, depending on the type of manure applied.
Composited samples (10 L bladders) and individual Clostridium samples (1 L) were shipped the
same day of collection to an independent laboratory on ice by overnight courier. Temperature
inside the coolers was monitored during transport with individual digital thermometers (i-
buttons).
D.2 Manure Description and Plot Application
As noted previousloy, manures were obtained from cattle, swine, and poultry. The total amount
of manure to be used on all plots was collected directly from farms a day in advance of the first
day of the study. Cattle manure was obtained from a beef cattle farm operated by
USD A/Agricultural Research Service (ARS) in Watkinsville, GA, by collecting fresh pats from
the pasture site where cattle were grazing. Broiler litter was obtained from a poultry farm
operator. The litter was obtained directly from the inside of the chicken house from the top layer
of litter. Litter composition was considered to be typical of this type of operation, and consisted
of a mixture of chicken manure, wood chips, and feathers. Swine manure was obtained from two
different sources because the first operator (University of Georgia) temporarily discontinued
swine operations during the course of the study. During the first two simulations, swine manure
December 2010 D-2
-------
U.S. Environmental Protection Agency
was obtained from a lagoon receiving manure flushed from the swine pits. During the third
simulation, the swine manure was obtained from a commercial facility housing over 2500 pigs.
The manure was obtained directly from the pipe as the house was being flushed before it actually
mixed with the lagoon material. Once collected, all manures were transported to the laboratory
and stored at 4 °C until the day of application, which consisted of 24 hours, 1 day, or 2 days.
Holding times depended on the experimental design. The solids fraction of the applied solid
manures is provided in Table 37.
Table 37. Percent solids of poultry and cattle manure applied to
experimental plots
Type of Manure
Poultry
Cattle
Poultry
Cattle
Simulation Run
A
A
B
B
% Solids
69.3
11.7
75.6
11.9
Each type of manure was sampled for pathogen and FIB loadings prior to application via
randomized composite samples. Manures were weighed in the laboratory into individual
containers on the day of application and transported to the field on ice. Application of manures
was scheduled for Tuesday, Wednesday, and Thursday during the first week of the event.
Rainfall was applied to the plots accordingly to the day of application following 1 hour, 1 week,
and 2 weeks after manure was applied. Cattle manure was applied in small pats evenly
distributed across the plots. Poultry litter and swine slurry were poured directly and evenly
across the plots.
Poultry and swine manures were applied at agronomic rates (100 and 300 Ib/acre, respectively)
following USDA Natural Resources Conservation Service (NRCS) guidelines based on the
nitrogen requirement of the type of crop and the nutrient concentration in the manure being
applied. Cattle manure was applied at 10% of the total daily manure produced by grazing beef
cattle.
D.3 Seeding of Manures with Surrogate Pathogens
During the first simulation (Run A), it was determined that the natural concentration of the
pathogens of interest was too low in the manure being applied to detect in the runoff water.
Therefore, it was decided to seed the manures with surrogate pathogens to determine the
leaching rates of pathogens from the applied manures. The surrogate pathogens selected were all
non-virulent species that did not pose a risk of infection to project personnel or the environment.
Manures were spiked for both Run B and Run C.
December 2010
D-3
-------
U.S. Environmental Protection Agency
D. 3.1 Surrogate pathogens description
The surrogate organisms used to seed the manures, as well as the type of manure that they were
added to, is provided below.
• E. coli O157:H7 B6914 #87 was added to cattle feces, poultry litter (only during the
March 2010 simulation), and swine slurries;
• Salmonella X3985 was added to cattle manure, poultry litter, and swine slurries; and
• UV-inactivated Cryptosporidium and Giardia were added to cattle manure and swine
slurries.
D. 3.2 Stock surrogate cocktails
Seeding experiments were conducted by an external laboratory to determine the concentration of
surrogate pathogens to add to the different types of manures to increase the likelihood of
detection in runoff water. The calculation for the final surrogate pathogen concentrations took
into consideration the recovery of the methodology used for analysis, the decay of the organisms
in manure as well as during transport, values from literature for previously observed leaching
rates for pathogens or FIB from livestock manures, maximum pathogen levels observed in
livestock manure, and maximum number of organisms that could be produced to use for spiking.
Stock surrogates suspensions were shipped by overnight courier to EPA and stored at 4 ± 1° C
until the day of use. Each suspension was vortexed for 2 minutes before removing an aliquot for
the stock enumeration or preparing the spike cocktail.
On each day of spiking, aliquots (500 jiL) of each individual surrogate suspension were
aseptically removed from the stock tubes after 2 minutes of vortexing. Each suspension volume
was transferred to an individual labeled, sterile 2 mL tube with a screw-cap. Vials were stored in
the refrigerator until analysis.
Spike cocktails were prepared each day for each manure type. A chart designating the volumes
of each surrogate suspension to be used to prepare the cocktail for each manure type was
provided with the stock surrogates that were shipped to the EPA laboratory. Enough volume of
each surrogate cocktail was provided so that a 1.5 mL (1500 jiL) subsample was removed from
each cocktail for enumeration. The 1.5 mL spike cocktail was aseptically transferred to a 2 mL
labeled, sterile tube with a screw cap.
Individual plots to be seeded with swine slurry received 1 x 109 UV-irradiated Cryptosporidium
parvum oocysts, 1><107 UV-irradiated Giardia lamblia cysts, 1 x 1010E. coli O157:H7 strain
B6194, and 1 x 1010Salmonella X3985. For plots to be amended with cattle manure, 5 x 107 C.
parvum oocysts, 1 x 107G. lamblia cysts, 1 x 1010 Salmonella X3985, and 1 x 1010E. coli
O157:H7 strain B6194 was applied. Surrogate spiking levels for each poultry litter plot received
1 x 1010E. coli O157:H7 strain B6194 (March 2010 only) and 1 x 1010Salmonella X3985.
December 2010 D-4
-------
U.S. Environmental Protection Agency
D.3.3 Manure spiking procedure
Swine. Each day the total number of plots to have swine slurry was determined using the
experimental design diagram. The total volume of slurry applied to plots was calculated and was
aseptically transferred to a sterile 20 L carboy containing a sterile Teflon-coated stir bar. After
adequate mixing, volumes of the un-spiked slurry required for background analysis were
removed and transferred to sterile containers that were stored at 4 °C until analysis. The spike
cocktail was then aseptically added to the container, under continuous stirring. The container
housing the spike cocktail was rinsed with sterile phosphate buffered saline (PBS) and added to
the container. The spiked slurry was stirred on a stir plate for 30 minutes. After mixing, spiked
manure subsamples were removed. The remaining slurry was measured into individual 4 L
sterile containers for application to the plots. After application to the plots, the container was
rinsed with sterile PBS and this rinse water was also applied to the plots to ensure a complete
transfer of the spiked manure.
Poultry. Poultry litter was measured into 5 gallon plastic pails equipped with a cover (rinsed
with 70% ethyl alcohol [EtOH], inverted on clean foil or bench protectors, and air-dried
overnight). For each plot that received poultry litter, 1.05 kg was weighed and transferred to an
individual 2.5 kg container. Before adding the spike cocktail, un-spiked subsamples were
removed for analyses. The spike cocktail was then added in three portions to each container
using a sterile pipette and shaking for 5 minutes after the addition of each portion. After the final
portion was added, the spike cocktail container was rinsed with PBS and added to the poultry
litter. The covered pail was shaken for 5 minutes and allowed to stand for 30 minutes. Spiked
subsamples were removed from the container for additional analyses. Spiked subsamples were
directly added to the fields. Containers were rinsed at the end as described above.
Cattle. The total amount of cattle manure was measured into a 5 gallon plastic pail (rinsed as for
poultry litter above). For each plot that receives cattle manure, 2.4 kg was applied. On each day
of manure application, the total mass of cattle manure needed for the day was calculated and
additional amounts were included for additional assays. The total amount of cattle manure was
added to a clean, sterile container and un-spiked samples were removed for subsequent analyses.
The spike cocktail was added in three portions using a sterile pipette and then mixed with a clean
mixing device after the addition of each portion. After the final portion was added, the spike
cocktail container was rinsed with PBS and added to the cattle manure. The manure was mixed a
final time for 5 minutes and allowed to stand for 30 minutes. Spiked manure subsamples were
removed for analysis. Subsamples for plot application were weighed into sterile containers using
disposable sterile scoops or equivalent.
D.4 Microbial Analysis
Samples were coded as C for cattle, S for swine, P for poultry, and X for control. Samples were
analyzed for the presence and concentration of E. coll O157:H7, Salmonella spp.,
December 2010 D-5
-------
U.S. Environmental Protection Agency
Campylobacter spp., Clostridium spp., Giardia cysts, and Cryptosporidium oocysts and
infectious Cryptosporidium spp. In addition, the sample were also analyzed for the presence and
concentration of the surrogate pathogens E. coli O157:H7 B6914 #87 and Salmonella X3985.
Percent solids in cattle and poultry manure samples were determined using Standard Methods
2540B. The swine manure was treated as a water sample. A 1:10 dilution (5 g or mL of samples
in 45 mL phosphate buffered water [PBW]) was prepared for all manure samples. Each volume
was analyzed in triplicate for the MPN and Clostridium assays. A summary of the organisms
analyzed and the methods employed is provided in Table 38.
Table 38. Organisms and methods used for analysis of water and manure samples
Organism
E. coli
Enterococcus
E. coli 01 57
E. coli 0157
Salmonella
Salmonella
Salmonella X3985
E. co//B6-194
Crypto/Giardia
Campylobacter
Campylobacter
Clostridium
Method
Culture
Culture
Culture
qPCR
Culture
qPCR
Culture
Culture
Microscopy
Culture
qPCR
Culture
Description
(Colilert)
Method 1600
Broth tube enrichment MPN
Gene targets: stxl, stx2, eae
Broth tube enrichment MPN
Broth tube enrichment MPN
Direct plating onto TSA-A.
EPA Method 1623
Broth tube enrichment MPN
Modified TSCF
D.4.1 qPCR Assays
Each cubitainer containing runoff water was shaken to re-suspend settled particles and then
aliquoted to individual sterile containers for various analyses. For quantification of bacterial
pathogens using qPCR, 400 mL of each water sample was centrifuged at 4000* g at 4 °C for 30
min, the supernatant was discarded and the pellet was re-suspended in PBS. Concentrated
samples were stored at -80 °C until DNA purification and qPCR assays.
D. 4.2 Fecal indicator assays
Enterococci concentrations were determined in both manure and runoff samples by membrane
filtration following EPA Method 1600. A 1:10 dilution using PBW was prepared for all solid
manures (cattle and poultry), while swine was analyzed as a water sample. For manure, dilutions
from 10"2 to 10"5 (g or mL) were prepared. For runoff samples, dilutions ranged from 1 mL to
10"3 mL, depending on the type of manure. Defined substrate technology (Colilert®, Idexx) was
used to determine concentrations ofE. coli in both manure and runoff water. For manure, the
same dilution range as used for enterococci was used; for runoff water, the volumes used
December 2010 D-6
-------
U.S. Environmental Protection Agency
included 1, 10, 50 and 100 mL, depending on the type of manure or fecal aging time. Samples
were analyzed in duplicate.
D.4.3 MPN assays
Concentrations of E. coli O157:H7, Salmonella spp., and Campylobacter spp. in runoff samples
were determined using the MPN technique. Samples were analyzed in triplicate, at volumes of
10, 1, and 0.1 mL in the enrichment step by pipeting the volume directly to the tubes. In
subsequent weeks, the highest volume analyzed for each sample was first concentrated by
membrane filtration onto a 0.45-|im cellulose nitrate membrane and subsequently transferred to
the enrichment medium. The volume assayed by membrane filtration varied (between 30 and
100 mL), depending on the amount of particulates present in each sample. The other volumes
analyzed, 5 and 0.5 mL, were pipetted directly into the tubes.
For E. coli O157:H7 and Salmonella spp. MPN assays, buffered peptone water (BPW) was
inoculated and incubated at 35 to 37 °C for 20 to 24 hours. A 10 jiL portion of each BPW
bottle/tube enrichment was then streaked onto HardyCHROM™ O157 agar. Presumptive E. coli
colonies were tested for a positive indole reaction and a negative fluorescence. Colonies with the
appropriate response were then tested with antiserum against E. coli O157 antigen using a
commercially available latex agglutination kit. For Salmonella spp., 100 jiL from each BPW
bottle/tube enrichment was inoculated into a 10 mLtube of Rappaport-Vassiliadis enrichment
broth and incubated for 24 hours at 43°C. A 10 jiL portion of each enrichmentbroth tube was
then streaked on Salmonella and Shigella xylose lysine deoxycholate agar biplate and incubated
for 18 to 24 hours at 35 to 37°C. Isolated presumptive positive colonies were inoculated to
Enterotubes and incubated at 35 to 37 °C for 18 to 24 hours for biochemical confirmation of
Salmonella spp.
The enrichment step for samples analyzed for Campylobacter spp. included the same volumes as
analyzed for E. coli O157:H7 and Salmonella spp. Water or membrane filters were inoculated in
Bolton Broth and incubated at 35 to 37 °C for 4 hours and then transferred to 42 ± 1°C for 20 to
44 hours. After the incubation period, a 10 jiL portion from each enrichment tube/bottle was
streaked for isolation onto Campylobacter blood free selective agar. The plates were incubated
in a microaerophilic atmosphere (5 to 6% oxygen, 10% carbon dioxide and 85 to 85% nitrogen)
at 37°C for 48 hours. Following incubation, each plate was inspected for presumptive-positive
Campylobacter growth. Next, each plate was tested for positive latex agglutination using a
commercially available kit. Plates with colonies resulting in autoagglutination were scored
positive for MPN calculations and retained for qPCR assay.
D. 4.4 Surrogate pathogens assays
E. coli O157:H7 strain B6914 was analyzed by direct plating onto trypticase soy agar. Colonies
that fluoresced green under UV light were enumerated; positives were confirmed with E. coli
December 2010 D-7
-------
U.S. Environmental Protection Agency
O157 latex agglutination kit. Salmonella surrogate was determined using the same MPN assay
described above.
D. 4.5 EPA Method 1623: Giardia and Cryptosporidium
For enumeration of Giardia cysts and Cryptosporidium oocysts, 2 to 9 L of each sample was
analyzed using EPA Method 1623 with samples concentrated by centrifugation or filtration.
Packed pellet volumes were measured and no more than 0.5 mL packed pellet volume was
analyzed in one immunomagnetic separation (IMS) reaction as prescribed in the method.
Isolated cysts and oocysts were enumerated as prescribed in the method using epifluorescence
microscopy.
In cattle manure samples, 10 mL of the diluted manure sample (5 g wet weight in 45 mL PBW)
was analyzed in one IMS reaction. Recovered cysts and oocysts were enumerated using
epifluorescence microscopy. For the swine manure, 50 mL was concentrated by centrifugation
and analyzed as described above. Poultry manure samples were not analyzed for Giardia cysts
and Cryptosporidium oocysts.
D. 4.6 Infectious Cryptosporidium oocysts by foci detection method
Cryptosporidium oocysts were isolated from interfering debris using IMS as described in EPA
Method 1622. The isolated bead-oocyst complex was rinsed with 10 mL PBS to remove the
IMS buffers, which are toxic to the human ileocaecal adenocarcinoma (HCT) monolayers. The
rinsed bead-oocyst complex was quantitatively transferred to 1.5-mL Eppendorf tube using PBS.
The bead-oocyst complex was rinsed and the supernatant discarded. The bead-oocyst complex
was re-suspended in 150 jiL of Hank's balanced salts solution (HBSS) or PBS and an equal
volume of Hank's Balanced Solution, pH 2.0, containing 2% trypsin and incubated at 37 °C for 1
hour. Every 15 minutes, the tubes were vortexed for 10 seconds. After 1 hour, the tubes were
prepared for magnetic particle concentration. Inoculation medium (300 jiL) was added to each
tube, gently mixed, and centrifuged at 10,000x g for 2 minutes. The supernatant was aspirated to
50 jiL and a fresh 500 jiL aliquot of inoculation medium was added and gently mixed. The tube
was centrifuged at 10,000* g for 2 min and the supernatant aspirated to 50 jiL. Each sample
concentrate was re-suspended in 350 jiL of inoculation medium and inoculated to a single well
of an 8-welled chamber slide containing a monolayer of HCT-8 cells (ATCC CCL-244) and 100
jiL of inoculation medium. Chamber slides were incubated for 65 to 72 hours at 35° C in a
humid (5% CO2) atmosphere. After the incubation period, the growth medium was aspirated
from each well and the monolayers were rinsed with pre-warmed PBS to remove unattached
oocysts. Wells were then fixed with absolute methanol for 8 minutes and then rehydrated for 30
minutes with PBS containing 2% goat serum and 10% of a 0.002% solution of Tween 20.
Infections were detected by staining monolayers with a fluorescein labeled polyclonal rat
immunoglobulin G antibody for detection of the intracellular reproductive stages of
December 2010 D-8
-------
U.S. Environmental Protection Agency
Cryptosporidium parvum oocysts. Enumerations of infection sites were observed using
epifluorescence microscopy.
D. 4.7 Microscopy for Giardia and Cryptosporidium assays
A Zeiss Axioskop fluorescence microscope, equipped with a blue filter block (excitation
wavelength, 490 nanometer (nm); emission wavelength, 510 nm) was used to detect labeled
oocysts at a magnification of 360*. DAPI staining characteristics were observed at 640x
magnification using a UV filter block (excitation wavelength, 400 nm; emission wavelength, 420
nm). The internal morphology of oocysts and intracellular reproductive stages of C. parvum
oocysts was observed by using Nomarski DIG microscopy at 640 to 1600x magnification.
D.4.8 Clostridium spp. assays
Samples were analyzed for Clostridium spp. densities using a modification of the SCA/NHS
method for detection of Clostridium spp. on tryptose sulfite cycloserine (TSC) agar. Water
samples were filtered through 0.45-|im cellulose nitrate filters and aseptically applied to agar
plates. Plates were incubated at 44.5 °C for 48 hours. Brown to black colonies were counted as
Clostridium spp. Due to the presence of high levels of particulate matter, fluorescence was not
assessed on these samples.
December 2010 D-9
-------
U.S. Environmental Protection Agency
Appendix E. Pathogen and FIB Mobilization Fractions Due to Rainfall
The degree to which microorganisms run-off from soil depends on myriad factors. Rather than
attempt characterizing each of those factors separately, all experiments and analyses were
conducted for a defined typical rain event. Experimental conditions for plot-scale runoff
experiments were selected based on an intense (< 100 year return period) rain event for the
Georgia Piedmont region.
In plot-scale experiments rainfall was applied to plots at an average rate of 6.89 cm/hourr (a=
0.62 cm/hour) for a rain event duration of 60 minutes. Average cumulative runoff volume from
plots was 57.7 L (cr= 16.1 L). A histogram showing the distribution of runoff volumes for
individual plots is presented in Figure 35. The wide variability in runoff volume despite the low
variability in rain intensity and duration arises from differences between plots including
antecedent soil moisture, slope, location, soil type and grain size distribution and other factors.
c
Q)
CT
0>
1
0 2
0
4
0
6
0
8
0
1C
Cumulative Runoff Volume (L)
Figure 35. Histogram of cumulative runoff volumes from plots subject to the design rain event
In addition to application of a prescribed rainfall at a typical intensity, the design event simulated
in plot-scale experiments entailed land application of manures to plots at an agronomic rate. The
agronomic application rate is the mass or volume per unit area with a nutrient content equal to
the nutrient requirement for the vegetation on the plot (Midwest Plan Service, 2004). The
agronomic rate accounts for the type of vegetation on the plot, the nutrient content of the soil
prior to application of the manure, and the nutrient content of the manure. The application rates
selected for solid cattle manure, swine slurry, and poultry litter are presented in Table 39.
December 2010
E-l
-------
U.S. Environmental Protection Agency
Table 39. Manure application rates
Manure
Solid cattle manure
Swine slurry
Poultry litter
Application
Rate per Plot
1.6kg
2.7 L
0.667 kg
Application Rate per
Area (plot dimensions:
2 m x 0.75 m)
1.07kg/m2
1.78L/m2
0.445 kg/m2
Data for both manure density and runoff density are required to calculate mobilization fraction
via equation 12. As described in Appendix D, manures and runoff composite samples were
assayed for numerous pathogens and FIB, in some cases via multiple methods. For use in
equation 12, there must be sufficient data for a given organism-method combination for both
manure and runoff samples. Criteria for selecting data for use in equation 12 included the
following:
• only plots with manure sample densities above detection limits were used;
• data indicating more pathogens running off than applied were assumed anomalous
(perhaps resulting from contamination of plots or cross-contamination in the laboratory)
and excluded; and
• at least 5 paired data for runoff and manure samples were available for a particular
manure-organism combination.
FIB data yielding mobilization fractions greater than one were not excluded from analyses, since
background levels of FIB were relatively high on control plots. However, two alternative
methods for estimating indicator runoff fraction were used so that the impact of background
indicator runoff could be assessed. The occurrence of FIB in control plots is described below
and implications with respect to QMRA modeling are described. For Campylobacter, there were
insufficient culture data to estimate the range of mobilization fractions; thus, it was assumed that
mobilization fraction calculated using qPCR manure and runoff density is equivalent to that
calculated using MPN counts in the manure and runoff. This assumption is consistent with the
correlation in qPCR and culture counts of organisms in fresh manures as observed by Klein et al.
(2010). The methods employed and data sets with data meeting the criteria for use in estimating
mobilization fraction are summarized in Table 40.
December 2010
E-2
-------
U.S. Environmental Protection Agency
Table 40. Method-organism combinations and data availability
Organism
Method
Data for
Manure?
Data for
Runoff?
Pathogens
Campylobacter
Cryptosporidium
Giardia
E. coli O157
E. coli O157 surrogate
E. coli stxl
E. coli stx2
E. coli eae
Salmonella (wild type)
Salmonella surrogate
Salmonella
MPN
Simplates
qPCR
EPA 1623
EPA 1623
MPN
Membrane filtration
qPCR
qPCR
qPCR
MPN
MPN
qPCR
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Fecal indicator bacteria
Clostridium perfringens
E. coli
Enterococci
Enterococcus
Total conforms
Culture (mCP)
Culture (TSC)
MPN (Colilert)
Culture (mEI)
qPCR
MPN (Colilert)
0
0
0
0
0
0
0
0
0
0
Equation 12 was evaluated using the manure application rate data (Table 39), cumulative runoff
volumes for each plot, and the manure and runoff organism densities (Table 40) to determine the
mobilization fraction for each plot. Results for pathogens are summarized in Table 41. For all
pathogens except Campylobacter mobilization fractions are based on a single organism-method
combination. For Campylobacter., MPN data were insufficient to develop mobilization fractions
for poultry litter and qPCR data were used. For nearly all pathogens the mobilization fraction
ranges spanned several orders of magnitude, despite the relatively uniform rainfall treatment
applied to each plot. The E. coli O157 surrogate was observed in control plot runoff. This
observation indicates the potential for contamination of control plots or runoff water samples
from control plots.
December 2010
E-3
-------
U.S. Environmental Protection Agency
Table 41. Mobilization fraction ranges and means for pathogens
Pathogen
Campylobacter
E. coli O157
Salmonella
Cryptosporidium
Giardia
Manure
Cattle
Swine
Chicken
Cattle
Swine
Chicken
Cattle
Swine
Chicken
Cattle
Swine
Chicken
Cattle
Swine
Chicken
Logio of the
Minimum
Mobilization
Fraction
-4.85
-2.20
-8.60
-3.65
-3.01
-4.01
-5.57
-3.85
-3.68
-4.46
-3.90
Logio of the
Minimum
Mobilization
Fraction
-1.46
-1.01
-1.74
-0.200
-1.501
-2.214
-1.26
-2.40
-2.65
-0.179
-1.48
Geometric Mean
of Mobilization
Fraction
0.000373
0.0495
8.52E-07
0.0159
0.00664
0.00118
000235
0.000781
0.000556
0.00272
0.00201
Not tested
-6.40
-4.58
-0.387
-0.0617
4.72e-05
0.00481
Not tested
Basis/Method
MPN
MPN
qPCR
CFU, surrogate
CFU, surrogate
CFU, surrogate
MPN, surrogate
MPN, surrogate
MPN, surrogate
EPA 1623
EPA 1623
EPA 1623
EPA 1623
EPA 1623
EPA 1623
FIB mobilization fractions were generally calculated as described above, but details of the
calculations are presented below separately to address mobilization fractions greater than one.
E. coli and enterococci densities observed in manures used in plot-scale experiments are
compared to "typical" ranges derived from reports in the literature in Table 42. The distributions
of experimental FIB densities appear skewed for both FIB and for all three manure types. The
range of cattle manure densities in manures used in the plot experiments is higher than that from
literature studies. Because fresh manures from an operational cattle facility were used in
experiments, we believe the experimental manures are typical of manures in the United States for
similar types of operations and manure handling practices. Therefore, the range of manure FIB
densities considered "typical" may be too narrow. Both enterococci and E. coli were much less
abundant in the poultry litter than in values reported in the literature. Because poultry litter is a
heterogeneous mixture of feces and other materials, it less clear how typical the experimental
manure densities are. Swine slurry FIB densities are much lower than typical values for both
enterococci and E. coli. A plausible explanation for those low densities and indication that these
are not typical that the slurries were taken from an operation with very few pigs contributing to
the slurry lagoons during the second round of experiments.
December 2010
E-4
-------
U.S. Environmental Protection Agency
Table 42. Comparison of typical and experimental manure FIB densities
Logio(min)
Logio(max)
Arithmetic
Mean
Entero cocci
Cattle
Swine
Chicken
Literature
Experiment
Literature
Experiment
Literature
Experiment
2.0
4.70
5.3
0.176
5.0
3.78
5.1
5.46
7.2
2.02
7.0
5.81
147,000
46.5
219,000
E. coli
Cattle
Swine
Chicken
Literature
Experiment
Literature
Experiment
Literature
Experiment
5.0
6.69
6.1
0.70
5.1
2.69
6.7
8.32
7.3
3.07
10.9
4.36
55,600,000
546
9012
Runoff indictor densities from plots with manures and control plots are summarized in Table 43.
Densities in control plot runoff were variable, and in some instances high relative to swine and
poultry runoff densities. Other mobilization studies have also reported significant densities in
runoff from control plots (Miller and Beasley, 2008; Thurston-Enriquez et al., 2005). E. coli
O157 surrogate data are included in Table 43 because mobilization association with that
organism is similar to that of generic E. coli and there is no background E. coli O157 surrogate in
the soil to confound estimation of mobilization fraction from manure-borne organisms. There
were several instances ofE. coli O157 occurrence in runoff from control plots, potentially due to
transport ofE. coli O157 from up-slope to down-slope plots.
Histograms of the FIB mobilization fractions for enterococci andE1. coli for each manure type
are presented in Figures 36 to 41. Cattle mobilization fractions are within anticipated ranges,
though mobilization fractions greater than one (more enterococci running off the plot than
applied in the manure) occurred for three plots. Mobilization fraction was significantly greater
than for many of the plots treated with swine slurry. This, along with the very low densities of
FIB in swine slurries applied to the plots and relatively high densities of FIB in the runoff from
the control plots, indicates that the majority ofE. coli and enterococci in the runoff from plots
with swine slurry applied did not originate from the swine slurry. These findings indicate a
disconnection between the FIB and pathogen densities in the runoff. Poultry litter FIB
mobilization results are between those for cattle and swine. Mobilization fractions from plots
with applied poultry litter are generally less than one and in the instances in which mobilization
fraction exceed one the fraction is not as high as those observed for runoff from plots treated
with swine manures.
December 2010
E-5
-------
U.S. Environmental Protection Agency
Table 43. Runoff FIB densities for plots with and without manure application
Enterococci
Parameter
Logio(Minimum mobilization fraction)
Logio(Maximum mobilization fraction)
Arithmetic mean of mobilization fraction
Geometric mean of mobilization fraction
Minimum runoff density (CFU/100 mL)
Maximum runoff density (CFU/100 mL)
Arithmetic mean of runoff density (CFU/100 mL)
Geometric mean of runoff density (CFU/100 mL)
Cattle
-2.784
0.262
0.336
0.0480
580
560,000
100,200
20,050
Swine
-0.694
2.315
30.2
7.13
0.5 (DL)
56,000
8440
782
Chicken
-1.168
1.141
1.316
0.497
500
3,600,000
378,000
45,500
Control
—
—
—
—
0.5
192,000
27,280
1313
E. coli
Parameter
Logio(Minimum mobilization fraction)
Logio(Maximum mobilization fraction)
Arithmetic mean of mobilization fraction
Geometric mean of mobilization fraction
Minimum runoff density (CFU/100 mL)
Maximum runoff density (CFU/100 mL)
Arithmetic mean of runoff density (CFU/100 mL)
Geometric mean of runoff density (CFU/100 mL)
Cattle
-5.13
-1.98
0.001988
0.000575
520
1203300
317,000
208,500
Swine
-1.55
0.972
2.62
0.917
0.5
310.6
65.7
9.17
Chicken
-2.94
1.17
3.02
0.711
0.5
54,750
9390
1298
Control
0.5
6630
553
12.6
E. coli O157 surrogate
Parameter
Logio(Minimum mobilization fraction)
Logio(Maximum mobilization fraction)
Arithmetic mean of mobilization fraction
Geometric mean of mobilization fraction
Minimum runoff density (CFU/100 mL)
Maximum runoff density (CFU/100 mL)
Arithmetic mean of runoff density (CFU/100 mL)
Geometric mean of runoff density (CFU/100 mL)
Cattle
-3.65
-0.20
0.141
0.0159
230,700
47,600,000
8,560,000
2,760,000
Swine
-3.01
-1.50
0.0116
0.00664
2130
508,000
147,550
32,100
Chicken
-4.01
-2.21
0.0026
0.00118
118
1,402,000
334,000
7590
Control
1429
December 2010
E-6
-------
U.S. Environmental Protection Agency
I I \ I I \ \ I
-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5
Log10(Mobilization Fraction)
Figure 36. Histogram of mobilization fractions for enterococci from plots treated with cattle manure
i r
~ i
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
Log10(Mobilization Fraction)
Figure 37. Histogram of enterococci mobilization fractions from plots treated with swine slurry
December 2010
E-7
-------
U.S. Environmental Protection Agency
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Log10(Mobilization Fraction)
Figure 38. Histogram of mobilization fractions for enterococci from plots treated with poultry litter
-4 -3
Log10(Mobilization Fraction)
Figure 39. Histogram of E. coll (via Colilert) mobilization fractions for cattle manure plots
December 2010
E-8
-------
U.S. Environmental Protection Agency
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
LoglOfMobilization Fraction)
Figure 40. Histogram of E. coll (via Colilert) mobilization fractions for swine slurry plots
LogtOjMobilization Fraction)
Figure 41. Histogram of E. coll (via Colilert) mobilization fractions for poultry litter plots
Inspection of the mobilization histograms suggests two alternative treatments for modeling
mobilization and runoff of FIB. In the first alternative, manure FIB abundances and
mobilizations are based on observations from the experiments, with mobilization fraction
distributions based on inspection of the histograms in Figures 36 to 41. Parameters
corresponding to this alternative are presented in Table 44. In QMRA calculations conducted
December 2010
E-9
-------
U.S. Environmental Protection Agency
using these parameters (see Section 4.2), pathogen abundances drawn from the literature are
paired with FIB abundances taken from experimental data. This mismatch renders results from
alternative 1 specific to the experimental conditions of the mobilization experiments and not
necessarily representative of general livestock runoff occurrences.
Table 44. Mobilization and abundance distributions, alternative 1
Mobilization (Experimental)
Distribution
Min
Max
Mode
Manure Abundance
(Experimental)
Min
Max
Entero cocci
Cattle
Swine
Chicken
Uniform
Uniform
Triangular
-2.8
-1.0
-1.25
0.26
2.5
0.32
-0.25
4.7
0.176
3.8
5.5
2.0
5.8
E. coli
Cattle
Swine
Chicken
Triangular
Uniform
Triangular
-5.0
-2.0
-2.75
-2.0
1.0
1.25
-2.75
0.25
6.7
0.70
2.7
8.3
3.1
4.4
In the second alternative, E. coli O157 surrogate mobilization is used instead of mobilization
distributions observed for E. coli and Experimental mobilization distributions are used for
Enterococcus for cattle and poultry. Because no suitable data were available for characterizing
Enterococcus mobilization in swine slurry no calculations were performed for that manure-
inidcator combination. In this alternative, indicator abundances are based on the observations
presented in the literature. Parameters corresponding to this alternative are presented in Table
45. This alternative has the advantage over alternative 1 of using consistent sets of abundances
for pathogens and FIB and of basing mobilization fractions on only the organisms originating in
the manures.
Table 45. Mobilization and abundance distributions, alternative 2
Mobilization (Experimental for ENT, E. coli
O157 surrogate values for E. coli)
Distribution
Min
Max
Mode
Manure Abundance
(Literature)
Min
Max
Entero cocci
Cattle
Swine
Chicken
Uniform
Triangular
-2.8
-1.25
0.26
0.32
2.0
4.6
5.0
5.1
4.8
7.0
E. coli
Cattle
Swine
Chicken
Uniform
Uniform
Uniform
-3.65
-3.0
-4.0
-0.20
-1.5
-2.2
5.0
6.1
5.1
6.7
7.3
10.9
December 2010
E-10
-------
U.S. Environmental Protection Agency
Appendix F. Microbial Risk Assessment Interface Tool Simulation Images
The forward QMRA calculations were performed using MRAIT, a tool originally developed for
estimating risks related to biosolids application. The tool was substantially modified for use in
estimating risks associated with runoff from land-applied agricultural wastes and is in
development for use in additional QMRAs. This appendix provides the output of MRAIT.
These results are included in this report both to demonstrate the methodology used in the forward
QMRA calculations and as an illustration of MRAIT.
Compared with other QMRA tools, MRAIT is intended for relatively easy use by users informed
in QMRA methodologies but without extensive programming experience. As illustrated in the
information below, users may rely on default assumptions for dose-response model parameters,
source prevalences and abundances, and mobilization parameters. More advanced users may
choose parameters based on additional data or on site-specifc data and knowledge.
December 2010 F-l
-------
Version 1.0
Microbial Risk Assessment Interface Tool
Developed by EOA, Inc. through
WERF projects OO-PUM-3 and 04-HHE-3
© Copyright 2007 by the Water Environment Research Foundation. All rights reserved. Permission to copy must be
obtained from the Water Environment Research Foundation.
MRAITmodifed September 2010 by Seller Environmental, LLC for US EPA to conduct P4 QMRA
Introduction
The Microbial Risk Assessment Interlace Tool esimates the population risk of microbial infection or illness
associated with ingestion of reclaimed water from various exposure scenarios.
This modified version is configured to estimate risk of infection or illness associated with recreational activities in agricultural
impacted waters and POTW impacted recreational waters. The mobilization of microbes is based on field studies
conducted by EPA in Athens, GA.
To run the model, below you will be asked to provide pathogen concentration data, and to specify a set of parameters
that define the model. For an example case study and guidance in specifying the parameters, please refer to the
user documentation that accompanies this tool.
A. Specify pathogen. The risk assessment will be performed for the following pathogen:
pathogen :=
^orovirus
Cryptosporidium
Siardia
Salmonella
Note that MRAIT was not set up to run Campylobacter or Norovirus.
To run Camplylobacter select "E coli O157:H7" and change the dose
response parameters below in section F.
MRAIT has been modified to run Norovirus. To run norovirus, select it
from the list to the left, then follow instructions below in Section C.
B. Input data. Specify an input file with concentration data for the above selected pathogen.
The file must be a text file with no header, and contain one column of concentrations (units of pathogens per liter).
Abundance of reference pathogens in fecal sources (logs #g or #/100mL)
Pathogen
£ coli O157:H7
Campylobacter
Salmonella
Cryptosporidium
Giardia spp.
Norovirus 1
Norovirus attenuation 2
1. Density in raw sewage: trianai
Cattle
low high
0.5 5.65
0 5.176
0.9 5.76
0 3.9
0 4.93
_
NA NA
low
ND
3.3
5
4.2
3.5
_
NA
liar distribution with min, mode and
Pigs
high
5.6
3.7
6.8
5.4
3.5
_
NA
Chicken
low high
0 0
2.8 6.5
0.3 4.8
0 0
0 0
_
NA NA
Secondary Effluent
low high
-
-
-
-1 1.5
-1 2.1
-2 6
1 4
max = (-2,4,6) loos
2. Attenuation through secondary disinfected effluent: triangular distribution with min, mode and max = (1,2.5,4} logs
3. Density in solid manures: units oflog(#/g wet weight)
4. Density in liquid manures: units are log w (#/100 mL)
Cattle abundace Pig abundance Chicken abundance
AC_EC_1 := 0.5 AC_EC_h := 5.65 AP_EC_1 := 0 AP_EC_h := 5.6
AC_Camp_l:=0 AC_Camp_h := 5.176 AP_Camp_l := 3.3 AP_Camp_h := 3.7 ACh_Camp_l := 2.8 ACh_Camp_h := 6.5
AC_S_1 := 0.9 AC_S_h = 5.76 AP_S_1 = 5 AP_S_h := 6.8 ACh_S_l = 0.3 ACh_S_h := 4.8
AC_Cryp_l := 0 AC_Cryp_h := 3.9 AP_Cryp_l := 4.2 AP_Cryp_h := 5.4
AC G 1 := 0 AC G h := 4.93 AP G 1 := 3.45 AP G h := 3.5
December 2010
F-2
-------
Prevalence (%)
Pathogen
£ CO// O157:H7
Campylobacter
Salmonella
Crypto
Giardia
Cattle
low high
9.7 28
5 38
5 18
0.6 23
0.2 37
Pigs
low high
0.1 12
46 98
7.9 15
0 45
3.3 18
Chicken
low high
57 69
0 95
6 27
Cattle prevalence
Pig prevalence
PC_EC_1 := 9.7
PC_Camp_l := 5
PC_S_1 := 5
PC_Ciyp_l := 0.6
PC G 1 := 0.2
PC_EC_h := 28
PC_Camp_h := 38
PC_S_h := 18
PC_Ciyp_h := 23
PC G h := 37
PP_EC_1:=0.1
PP_Camp_l := 46
PP_S_1 := 7.9
PP_Ciyp_l := 0
PP G 1 := 3.3
PP_EC_h := 12.0
PP_Camp_h := 98
PP_S_h := 15
PP_Ciyp_h := 45
PP G h := 18
Chicken prevalence
PCh_Camp_l := 57 PCh_Camp_h := 69
PCh_S_l := 0 PCh_S_h := 95
PCh_Ciyp_l := 6 PCh_Ciyp_h := 27
Human Infectious Potential
Pathogen
E CO// O157:H7
Campylobacter
Salmonella
Crypto
Giardia
Cattle
H
H
M
H
H
Pigs
H
H
M
L
H
Chicken
M
M
L
Cattle Infectious potential Pig Infectious potential Chicken Infectious potential
HC_EC := 0.835
HC_Camp := .835
HC_S := 0.5
HC_Cryp := .835
HC G:= .835
HP_EC := .835
HP_Camp := .835
HP_S := 0.5
HP_Cryp := .165
HP G:= .835
HCh_Camp := 0.5
HCh_S := 0.5
HCh_Cryp := .165
Pathogen
E. coli O157:H7
Campylobacter
Salmonella
Cryptosporidium
Giardia
Norovirus
Morbidity (%)
low high
MECJ := 0.2
MCamp_l:=0.1
MCiypJ := 0.2
MG 1 := 0.2
20
10
20
20
20
30
MEC_h := 0.6
MCamp_h := 0.6
MCryp_h := 0.7
MG h := 0.7
60
60
70
70
MNJ := 0.3 MN_h := 0.8
MS := 0.2
December 2010
F-3
-------
Mobilization fraction of pathogens (log 10 mobilization fractions)
Pathogen
E coliO157:H7
Campylobacter
Salmonella
Cryptosporidium
Giardia spp.
Cattle
low high
-3.65 -0.2
-4.85 -1.46
-5.57 -1.26
-4.46 -0.18
-6.4 -0.39
Pigs
low high
-3.01 -1.5
-2.2 -1.01
-3.85 -2.4
-3.9 -1.48
-4.58 -0.06
Chicken
low high
-4.01 -2.21
-8.6 -1.74
-3.68 -2.65
Not tested
Not tested
Note that mobilizations for each simulation are generated in a separate worksheet based on the above data.
Those values are written to a file and read in below. It is assumed that pathogens are mobilized for each simulation
at a random percentile of the log uniform distribution
Conversion factors to convert percent mobilizations to organisms per liter
cat_convert := 27.64 pig_convert := .6909 chick_convert := 17.27
Use input variables from above for these definitions
highlighted in yellow
Vals := 10000
Number of random input values ("Vals") specified here is set at 10000 to
be at least as great as the number of simulations that will be run.
land_applied_abun := runif(Vals,AC_EC_l,AC_EC_h)
PC EC 1 PC EC lT
Prevalence := runif Vals,
Human inf := HC EC
100
100
Morbidity := runif(Vals, MECJ, MECJi)
cattle mobil :=
For Salmonella, set
Morbidity = 1 because
dose response is based
on illness not infection
pig_mobil :=
chick mobil :=
convert := cat convert
mobilization := submatrix(cattle_mobil,0,9999,0,0)
Need to select appropriate variable (cattle_mobil,
pig_mobil, or chick_mobil, and set the third andfourth
values to
0, 1, 2, 3, or 4 for
EC, Campy, Salmonella, Crypto Giardia, respectively
„- I, _land applied abun TT ^mobilization)
runoff :=\10 - •Prevalence-convert-Human_inf-10 /
cone := runoff
December 2010
F-4
-------
[7] Pathogen Details -
C. Wastewater treatment. Specify what treatment processes will be applied. If the input data correspond to final effluent
concentrations (i.e., no additional wastewater treatment applied), select "None" below:
treatmenttype :=
Secondary treatment (activated sludge), filtration, and disinfection
'iteration and disinfection
disinfection (combined chlorine)
Dther - User provides log reduction
The default inactivation and/or
removal distribution for this
treatment - pathogen
combination is as follows in
units of log reduction:
Distribution
Check below if you
wish to override the
defaults to the left:
l~ override defaults
Distribution:
••formal
Jniform
Triangular
3oint Estimate
Gamma
Negative Binomial
For the animal impacted
waters QMRA set the
WWTP to "None".
For human impacted water
norovirus assessment
select "Other" and use a
triangular distribution with
values of (1, 2,5, 4) logs
log removal:
D. Fit concentration data to a statistical distribution. Specify the form of the statistical distributional that will be used to
fit the input concentration data specified above in Section A (default is Lognormal) OR specify that the input data are to be
used directly
cone choice :=
^ognormal
Weibul
For the animal impacted waters QMRA
set "conc_choice" to "Use input data directly"
[JJ Exposure Details
E. Specify an exposure scenario: There are three alternative exposure scenarios built into this interlace:
k1. Crop irrigation assumes that exposure to pathogens occurs via ingestion of crops irrigated via reclaimed water.
2. Recreation assumes that exposure to pathogens occurs via ingestion of reclaimed water through recreational activities
in an unrestricted impoundment.
3. Golf Course/Landscape Irrigation assumes that exposure to pathogens occurs via incidental or accidental ingestion
of reclaimed water from a golf course or park.
exp_scenario :=
Crop Irrigation
Golf Course/Landscape Irrigation
For the animal impacted waters QMRA
set exposure to "recreation"
December 2010
F-5
-------
3arameters for Recreation:
Recreation Exposure Pathway: (Note if recreation is not selected above as the
exposure route of interest the following section is inactive and cannot be changed)
Specify the volume that is ingested per exposure event: (units of ml)
The default ingestion volume for an
exposure event associated with
recreational exposure:
Check below if you
wish to override the
defaults to the left:
Choose a distribution:
[7] Other Exposure Routes
•luuuon c j —
l~ override defaults
|bg p92
fog JL43
1 1
••formal
Jniform
Triangular
3oint Estimate
3amma
Negative Binomial
distribution values in
units of ml:
(or In(ml) if log normal
distribution)
1 1
1 1
1 1
December 2010
F-6
-------
[7] Dose Response Details -
F. Specify dose-response function:
Reference
Pathogen
Oypfosporid/um
spp.
G/ard/a /amb//a
Campy/obacfer
jejuni
E. co//O157:H7
Salmonella
enferica
Published dose-
response model
Exponential
(U.S. EPA, 2005a,
2006)
Exponential
(Haas etal., 1999;
Rose etal., 1991)
Beta-Poisson
(Medema et al.,
1996;Teunis et al.,
1996:2005)
Beta-Poisson
(Teunis et al.,
2008b)
Beta-Poisson
(Haas etal., 1999)
Model
parameters
0.09
0.0199
0.145
7.59
0.4
37.6
0.3126
2884
IDso
8 oocysts
35 cysts
800 CPU
207 CPU
23,600
CPU
Morbidity
(%of
infections
resulting in
illness)
20-70%
20-70%
10-60%
20-60%
20%
Health
Endpoint
Infection
Infection
Infection
Infection
Infection
Note that the dose
response
- parameters that
were in the orignial
M RAIT have been
updated to reflect
those shown
to the left
default functional form and paramaters
for this pathogen are:
Functional form JBeta-Poisson
poi
oint
poi
oint
|value:
Check below if
you wish to
override the
defaults to the left:
l~ override defaults
Choose a functional form:
ixponential
rlypergeometric
3ompertz-log
December 2010
F-7
-------
If you are overriding the function above you need to provide dose-response parameters:
plead from file (to the right)
Specify parameters manually (below)
If reading dose-reponse parameters from a
file, Specify the file here (see user documentation
for guidance on the formating of thisfile):
y
(right-click on disk icon, and
select properties to change the
path and filename)
If specifying dose-response parameters manually, set them here:
Choose a distribution:
Specify parameters:
••formal
Jniform
Triangular
jog Normal
jamma
Negative Binomial
1
1
1
Choose a distribution:
Specify parameters:
Jniform
P • I
^og Normal
3amma
Negative Binomial
1
r
i
G. Model selection suggestion: Based on values for specified exposure, dose, the dose-response function, and the following
tolerance for error, the program will suggest either the static or dynamic model.
Specify your tolerance for error: Difference in predicted incidence between static and dynamic models
per year
-------
Histogram of runoff density for: |E. coli0157
1-10
MO3
100
10
1
Summary statistics for run off density:
Number of input points: cone n = 10000
mean(conc) = 13075.96 min(conc) = 0.002
median(conc) = 62.24 ma^conc) = 1.32 x 10
5-10 1-10
Density(#/L)
Histogram of exposure:
Histogram of dose:
1-10"
Linear scale
=«t 5000
MO5 2-105 3-105
Dose-response curve:
0.8|
Dose
0.6
0.4
0.2
ni r
1-10 Vio 3 o.oi o.i i 10 100 i-io3
4000
2000
50 100 150
Ingestion (ml/event)
Log scale
3000
2000
1000
l-lQ-Ia-lo 5.010.1 1 10 10(D-lQ3ia4lQ5-106
Dose
exponential_DR(d,r) := 1 - exp(-r-d)
i \ (A
betapoisson DR(d,a,B) := 1 - 1 + —
I P
hypergeom_DR(d,a,p) := 1 -mhyper(a,a + p,-d)
gompertz_log_DR(d,x,y) := 1 - exp(-exp(-x+ y-ln(d)))
Dose
December 2010
F-9
-------
Risk result:
Based on thel
model the distribution of the
probability of infection per exposure event is:
Note: Be careful in comparing static vs dynamic, and individual vs population-based
results, asthe description of risk shown in the box above will vary by scenario, depending . i
upon static vs dynamic, individual vs population-based, and in some caseswhetherthe 1Uth /olle. p.ze-j
dose-response for the pathogen of interest is expresed in terms of probability of illness or ^^^^^
infection.
Median: |l.2e-2
Mean:
|l
.5e-l
90th %ile: |5.0e-l
Note: if the risk estimates are below 1e-9 per event (or per 100,000 fora population level assessment), they are simply reported as<1e-9.
Linear scale
8000
a
o
j| 6000
1
^ 4000
0
^H
•g 2000
H
A
1
—
-
-
>M
0.5
Risk
Illness Risk
-6
percentile(illness_prob,0.05) = 2.9 x 10
percentile(illness_prob,0.1) = 1.2 x 10
median(illness_prob) = 4.388 x 10
mean(illness_prob) = 0.059
percentile(illness_prob,0.9) = 0.221
percentile(illness_prob,0.95) = 0.301
y
Log scale
3000
2000
1000
0
\ I I I I
1-10 f-101?-10 f-10 f-1014-10 30.01 0.1 1
Risk
a
C3
1
I
1-10 f-101?-10 f-10 f-1014-10 30.01 0.1 1
Risk of Illness per ev=nt
These commands send the simulation output to
a file for archiving
infection_prob
illness_prob
December 2010
F-10
-------
ANNEX 1
State-of-the-Science Review of Quantitative Microbial Risk
Assessment: Estimating Risk of Illness in Recreational Waters
(Final Report - August 2010)
For
Quantitative Microbial Risk Assessment to
Estimate Illness in Freshwater Impacted by
Agricultural Animal Sources of Fecal Contamination
U.S. Environmental Protection Agency
December 2010
-------
U.S. Environmental Protection Agency
STATE-OF-THE-SCIENCE REVIEW
OF QUANTITATIVE MICROBIAL RISK ASSESSMENT:
ESTIMATING RISK OF ILLNESS IN RECREATIONAL WATERS
U.S. Environmental Protection Agency
Office of Water
Office of Science and Technology
Health and Ecological Criteria Division
August 2010
-------
U.S. Environmental Protection Agency
DISCLAIMER
Mention of commercial products, trade names, or services in this document or in the references
and/or endnotes cited in this document does not convey, and should not be interpreted as
conveying official EPA approval, endorsement, or recommendation.
August 2010 ii
-------
U.S. Environmental Protection Agency
TABLE OF CONTENTS
Disclaimer ii
Table of Contents iii
List of Tables iv
List of Figures iv
Acronyms v
Executive Summary 1
I Introduction 3
I.I Background 3
1.2 Purpose 3
1.3 Report Context 4
1.4 Report Organization 5
II QMRA Background 6
III QMRA Literature Review: State-of-the-Science 9
III.l Literature Search Strategy and Summary of Results 9
III.2 Developing the Literature Database 9
III.3 QMRA: State-of-the-Science 10
IV Novel and Cutting-Edge QMRA-Related Techniques 38
IV. 1 Exposure Assessment 38
IV. 1.1 General Description 38
IV.1.2 Cutting-Edge Exposure Assessment Techniques 39
IV.1.3 Summary: Cutting-Edge Techniques for Exposure Modeling 46
IV.2 Health Effects Modeling 46
IV.2.1 Dose-Response Modeling 46
IV.2.2 Accounting for Susceptible Populations 58
IV.2.3 Secondary Transmission 60
IV.3 Risk Characterization 62
IV.3.1 Sensitivity Analysis 62
V References 66
August 2010 in
-------
U.S. Environmental Protection Agency
LIST OF TABLES
Table 1. Literature Search Strategy 10
Table 2. Topics Evaluated in QMRA Studies Used to Establish the State-of-the-Science 12
TableS. Summary of Select Published QMRA Studies 14
Table 4. Select Empirical Microbial Dose-Response Models 48
Table 5. Comparison of Bayesian Dose-Response Studies 52
Table 6. Elements that May be Included in Risk Characterization 62
Table 7. Sensitivity Analysis Methods and Techniques 64
LIST OF FIGURES
Figure 1. EPA/ILSI Generalized Framework for Assessing the Risks of Human Disease
Following Exposure to Food- and Waterborne Pathogens 7
Figure 2. Elements of the Analysis Phase within the EPA/ILSI QMRA Framework 8
Figure 3. Factors Affecting the Viability of Pathogens and Indicators Along with Pathways.... 39
Figure 4. DGD, Negative Binomial, and Poisson Probability Distribution Illustration 45
Figure 5. States and Flowpaths in a Dynamic Disease Transmission Model 61
August 2010 iv
-------
U.S. Environmental Protection Agency
ALOR
Ab+
AD
ANOVA
AWQC
BEACH Act
BIC
CART
CDC
CPU
CI
CSA
CSO
CWA
DALY
DEC
DGD
DIC
DSA
EHEC
EPA
FIB
FMD
GI
HAV
ILSI
mL
MCMC
MRA
NEEAR
NRC
NRSA
PBBK
PCR
PFU
POTW
QMRA
RWQC
TCID
TMDL
TSS
USDA
UV
ACRONYMS
difference in log odds ratio
antibody positive
automatic differentiation
analysis of variance
ambient water quality criteria
Beaches Environmental Assessment and Coastal Health Act of 2000
Bayesian information criterion
classification and regression tree
U.S. Centers for Disease Control and Prevention
colony forming units
confidence interval
conditional sensitivity analysis
combined sewer overflow
Clean Water Act
daily adjusted life years
diarrhegenic (E. coif)
discrete growth distribution
deviance information criterion
differential sensitivity analysis
enterohemorrhagic E. coli
U.S. Environmental Protection Agency
fecal indicator bacteria
foot and mouth disease (virus)
gastrointestinal
Hepatitis A virus
International Life Sciences Institute
milliliters
Markov Chain Monte Carlo
microbial risk assessment
National Epidemiological and Environmental Assessment of Recreational
(Water Study)
National Research Council
nominal range sensitivity analysis
physiologically-based biokinetic
polymerase chain reaction
plaque forming units
publicly owned (sewage) treatment works
quantitative microbial risk assessment
recreational water quality criteria
tissue culture infectious dose
total maximum daily load
total suspended solids
U.S. Department of Agriculture
ultraviolet (light)
August 2010
-------
U.S. Environmental Protection Agency
WHO World Health Organization (United Nations)
WWTP wastewater treatment plant
WQS water quality standard(s)
August 2010 vi
-------
U.S. Environmental Protection Agency
EXECUTIVE SUMMARY
This report provides a "state-of-the-science" review of quantitative microbial risk assessment
(QMRA) techniques for estimating the risk of illness from exposure to pathogenic
microorganisms in recreational waters. QMRA is one component in a comprehensive toolbox
being developed by U.S. Environmental Protection Agency (EPA) to support the implementation
of new or revised recreational water quality criteria (RWQC).
INTRODUCTION
The Beaches Environmental Assessment and Coastal Health Act of 2000 (BEACH Act) requires
EPA to publish new or revised RWQC. Historically, RWQC have been based on the results of
epidemiological studies. These studies provide quantitative relationships between indicator
organism densities and adverse health outcomes at those locations. To meet the requirements of
the BEACH Act, EPA conducted several new epidemiological studies in coastal marine and
freshwaters. EPA is evaluating the extent to which the relationships from these studies broadly
apply to other waters covered under the Clean Water Act (CWA).
QMRA is one tool EPA could use to evaluate the applicability of the new epidemiology studies
to other waters. To date, a limited number of QMRAs have been performed specifically for
recreational waters. EPA's Office of Science and Technology (Health and Ecological Criteria
Division) and Office of Research and Development requested this state-of-the-science review to
consolidate and summarize the scientific literature on QMRA techniques applicable for
recreational waters.
REPORT CONTENT
This report provides a detailed review of the technical literature associated with QMRA
emphasizing recreational waters impacted by pathogens from cattle, swine, and/or poultry waste,
including
• An overview of QMRA including a description of the features that differentiate microbial
risk analysis from chemical risk analysis.
• A summary of the literature search strategy.
• A summary and comparison of the available QMRA studies of waterborne pathogens for
recreational water exposures. These studies establish the current state-of-the-science
with respect to QMRA for waterborne contaminants and provide insights into the
techniques currently available for use in a QMRA of animal-impacted waters.
• A description of cutting-edge or novel techniques for use in exposure assessment, health
effects modeling, and risk characterization. These techniques could expand the
boundaries of the current state-of-the-science for QMRA in the near to mid-term future.
August 2010
-------
U.S. Environmental Protection Agency
LITERATURE REVIEW RESULTS
Approximately 300 QMRA studies, review papers, and related literature were identified. Studies
of limited relevance to recreational settings or of lower quality were excluded from detailed
review. Drinking water studies focusing on the role of treatment process efficacy for
determining relative risks were also excluded. This process resulted in 40 key studies that
• establish a list of key QMRA elements for supporting RWQC;
• identify the pathogens most commonly addressed in QMRAs;
• identify how variability has been addressed in QMRA studies;
• assess the frequency that secondary transmission is included in QMRAs; and
• compare the methods used for sensitivity analyses and risk characterization.
Additionally, the literature indicates that there are numerous techniques specific to exposure
assessment, health effects modeling, and risk characterization that are novel or beyond the
approaches typically used in QMRA studies. For each of these areas, this report provides an
overview of "traditional" QMRA approaches followed by a summary and comparison of these
techniques.
KEY FINDINGS
QMRA has been used under a wide variety of settings and scenarios and is useful in conditions
where epidemiological studies are difficult, impractical, or cost-prohibitive. Moreover, QMRAs
typically consider variability more comprehensively than other techniques for evaluating
potential health hazards. Collectively, the following key findings can be drawn from these
studies:
1. Most QMRAs focus on a limited number of waterborne pathogens. The two pathogens
analyzed most frequently, human enteric viruses and Cryptosporidium, are both believed
to be major contributors to risk of waterborne GI illness.
2. Temporal and spatial variability of pathogen density is difficult to characterize because of
limited data.
3. A limited number of dose-response models are available and most studies do not account
for variability and uncertainty in their dose-response models.
4. Secondary transmission of infection and immunity are typically not accounted for in risk
estimates.
5. Most QMRAs do not differentiate between average (nominal) conditions and rare events.
Rare events can be associated with higher levels of human health risk than nominal
conditions.
August 2010
-------
U.S. Environmental Protection Agency
I INTRODUCTION
I.I BACKGROUND
A central goal of the Clean Water Act (CWA) is to protect and restore waters of the United
States for swimming and other recreational activities. A key component in the CWA framework
for protecting and restoring recreational waters is State adoption of water quality standards
(WQS) to protect the public from illnesses associated with microbes in water. In this regard, one
of EPA's key roles is to recommend ambient water quality criteria (AWQC) for recreational
waters under Section 304(a) of the CWA) for subsequent adoption by States.
Historically, U.S. Environmental Protection Agency (EPA or the Agency) recommended AWQC
have been based on fecal indicator bacteria densities. In the 1960s, the Federal government
recommended fecal coliforms as the basis for AWQC for recreational waters. In 1986, EPA
recommended enterococci andE1. coli as the basis for the current criteria (U.S. EPA, 1986).
These fecal indicator organisms do not cause human illness themselves (i.e., they are not human
pathogens); rather, they are indicators of fecal contamination and therefore indicators of the
potential presence of human pathogenic organisms (NRC, 2004).
It has been almost 25 years since EPA last issued AWQC for recreational waters. The science
related to AWQC development and implementation has advanced significantly during this time.
EPA believes that new scientific and technical advances need to be considered, if feasible, in the
development of new or revised 304(a) criteria by 2012. To this end, EPA has been conducting
research and assessing relevant scientific and technical information to provide the scientific
foundation for the development of new or revised criteria. The enactment of the Beaches
Environmental Assessment and Coastal Health (BEACH) Act of 2000 provided EPA with an
opportunity to conduct new studies and provided additional impetus to issue new or revised
criteria for coastal recreational waters (specifically, for Great Lakes and coastal marine waters)
to replace or amend the 1986 EPA recommended criteria. EPA believes that the new or revised
criteria must be scientifically sound, implementable for broad CWA purposes, and provide for
improved public health protection over the 1986 criteria.
1.2 PURPOSE
As one aspect of developing new or revised AWQC, the Agency would like to consider
extending the observed relationships between indicator organisms and adverse health outcomes
as determined from discrete series of epidemiology studies to the broader set of waters covered
under the CWA so that all waters of the United States are equivalently protective of public
health. Additionally, once new or revised recreational AWQC are published, the Agency would
like to provide States guidance on using quantitative microbial risk assessment (QMRA) as part
of an implementation toolbox that could be used to ensure State water quality standards (WQS)
are appropriately protective for local conditions and/or for developing WQS. To support that
effort, the Health and Ecological Criteria Division within the Office of Water, in conjunction
with the Office of Research and Development, requested the development of a "state-of-the-
science" review for QMRA for estimating risk of illness resulting from exposure to fecal
material of a variety of sources with a particular emphasis on animal-derived (cattle, swine, and
poultry) waste. This report documents the results of that effort.
August 2010
-------
U.S. Environmental Protection Agency
The approach employed to develop this report was to conduct a review of the technical literature
to establish the state-of-the-science of QMRA with an emphasis on development of QMRAs for
recreational waters impacted by pathogens from cattle, swine, and/or poultry waste.
1.3 REPORT CONTEXT
The attributes of QMRA to support the implementation of new or revised RWQC that may favor
its use in conjunction with other health-based approaches include the following:
• QMRA methods explicitly account for variability and uncertainty in pathogen
occurrence, exposure rates, and human health response;
• QMRA models may be used to evaluate alternative scenarios or potential management
options; and
• risk estimates from QMRA models are accompanied by confidence intervals and
sensitivity information that may be used to support risk management.
For the case of human exposure to animal-derived pathogens from recreational waters, the
following features are consistent with the use of QMRA:
• the occurrence of pathogens varies widely with time, location in a water body, region,
land use, and myriad other factors;
• the human health effects related to animal-origin pathogens may vary significantly
between pathogen strains, serotypes, or isolates; and
• there are a large number of settings where exposure to waterborne pathogens of animal
origin is possible and those settings may be widely diverse (e.g., some settings may be
impacted primarily from dairy cattle while others by swine).
Although QMRA is a promising approach for analyzing risks associated with recreational water
use, developing QMRAs of animal-impacted waters presents numerous challenges, some of
which may be alleviated through application of the techniques and data assembled in this report.
Challenges in developing QMRAs for animal-impacted waters arise in all aspects of the QMRA
framework (ILSI, 1996, 2000), which includes exposure assessment, health effects modeling,
and risk characterization.
Exposure assessment is complicated by the many and complex physical processes comprising the
production and transport of animal-derived pathogens from farms to receiving waters and
ultimately to humans. For example,
• pathogen production rates vary significantly between farms as well as among individual
animals at a given farm;
• manure handling differs from farm to farm and from season to season—these practices
have a profound effect on the availability of pathogens for transport to receiving waters;
• factors governing overland and subsurface transport of pathogens are, at present, not
entirely understood, and rates are generally highly variable;
August 2010
-------
U.S. Environmental Protection Agency
• inactivation (or growth) rates of pathogens varies with pathogen, media, and
environmental conditions, and the data that are available to describe these processes are
incomplete; and
• there may be multiple pathways by which pathogens reach receiving waters, some of
which may be complex.
Furthermore, in developing a QMRA for animal-impacted waters, the complexity of flow paths
by which pathogens reach receiving waters will require innovative modeling. Models must be
sufficiently detailed to include factors to which risk estimates are sensitive, but simple enough
that they can provide information to support risk management.
Complications in health effects modeling for QMRAs of animal-impacted waters relate to dose-
response model data gaps and the potential secondary infections. Even among the select
reference pathogens on which this report focuses, there is strain-to-strain variability, substantial
uncertainty in low-dose infection rates, and variability in person-to-person sensitivity to
infection. It is also possible that pathogens of animal origin differ from those of human origin
(e.g., in sewage) in their ability to infect humans. Although the importance of considering
sensitive sub-populations in QMRA models is well recognized, dose-response models that would
allow differentiation between these groups are generally lacking. While techniques for including
secondary transmission in QMRA models are relatively well established, as discussed in the
survey of published QMRAs, they are seldom included nor are parameter values well described.
The techniques and metrics used in risk characterization for QRMAs of animal-impacted waters
will be critical for the effective use of QMRA risk estimates by risk managers and the scientific
community at large. As described below, sensitivity analyses are often overlooked in risk
assessments. It is expected that models for risk from animal-impacted waters will rely on many
parameters and that multiple models will need to be evaluated. Studies reviewing techniques for
sensitivity analysis and suggesting best practices are reviewed in this report to facilitate
sensitivity analysis for animal-impacted waters.
1.4 REPORT ORGANIZATION
Section II of this report provides a brief overview of QMRA and focuses on the fundamental
components of QMRA and the features that differentiate microbial risk analysis from chemical
risk analysis. Section III first summarizes the literature search strategy and then reviews and
compares QMRA studies related to waterborne recreational exposure. These studies establish
the current state of the science with respect to QMRA for waterborne contaminants and provide
insights into the techniques currently available for use in a QMRA of animal-impacted waters.
Lastly, Section IV provides a description of cutting-edge or novel techniques for use in exposure
assessment, health effects modeling, and risk characterization as related to QMRA. These
techniques are likely to expand the boundaries of the current state of the science for QMRA in
the near- to mid-term future. In this section, particular emphasis is placed on exposure
assessment techniques that might be employed in the analysis of the complex and variable
processes leading to ingestion of pathogens originating from animal feces.
1 In this report, use of the term "infectious dose" is avoided because it is considered to be ambiguous. Where possible, the
relative ability of pathogens to initiate infection is expressed in terms of specific doses such as IDi (the dose at which 1% of the
exposed population is expected to become infected) or ID50 (the median infectious dose; the dose at which 50% of the exposed
population is expected to become infected).
August 2010
-------
U.S. Environmental Protection Agency
II QMRA BACKGROUND
Quantitative microbial risk assessment (also known as MRA and pathogen risk assessment) is a
process that evaluates the likelihood of adverse human health effects that can occur following
exposure to pathogenic microorganisms or to a medium in which pathogens occur (ILSI, 1996b,
2000). To the extent possible, the QMRA process includes evaluation and consideration of
quantitative information; however, qualitative information is also employed as appropriate
(WHO, 1999). QMRA methodologies have been applied to evaluate and manage pathogen risks
for a range of scenarios including from food (Bollaerts et al., 2009; Nauta et al., 2005; Seto et al.,
2007), sludge/biosolids (Dowd et al., 2000; Eisenberg et al., 2004, 2008; Flemming et al., 2009)
,drinking water (Astrom et al., 2007; Medema et al., 1995; Regli et al., 1991; Seller, 2009),
recycled water (Asano et al., 1992; Westrell et al., 2003) and recreational waters (Ashbolt and
Bruno, 2003; Seller et al., 2003, 2006, 2010a,b).
The principles, processes, and methods for conducting risk assessments for chemical agents were
formalized in the early 1980s by the National Research Council (NRC) resulting in a four step
process or framework (NRC, 1983). These are hazard identification, dose-response assessment,
exposure assessment, and risk characterization. Many of the earliest MRAs employed the NRC
conceptual framework to provide a structure from which the assessments could be conducted
(Haas, 1983; Regli et al., 1991; Rose et al., 1991).
As the field of microbial risk assessment developed, it became clear that there were some
complexities associated with modeling the infectious diseases that are unique to pathogens.
Thus, there are features of microbial risk that necessitate use of techniques and data in different
ways than in assessment of chemicals and other risks, and include:
• variations in the ability of individual organisms in a population of pathogens to initiate
infection;
• wide variations in the susceptibility of human and animal hosts to infection—and even
wider variation in expression of disease symptoms;
• risks of secondary (person-to-person) transmission of pathogens;
• growth of pathogens in vivo and for a smaller subset in the environment;
• high variability (spatial and temporal) in the occurrence of pathogens in the environment;
and
• difficulty in recovery and enumeration of pathogens.
Therefore, the conceptual framework for chemicals may not always be appropriate for the
assessment of risk of human infection following exposure to pathogens. To address this concern,
EPA's Office of Water developed a conceptual framework in conjunction with the International
Life Sciences Institute (ILSI) to assess the risks of human infection associated with pathogenic
microorganisms (ILSI, 1996b). The Office of Water is in the process of developing a
comprehensive document that describes tools, methods, and approaches for microbial risk
assessment to support human health protection for water-based media. The EPA/ILSI
framework emphasizes the iterative nature of the risk assessment process (Figure 1), and allows
wide latitude for planning and conducting risk assessments in diverse situations. This framework
consists of the following three principal components: problem formulation, analysis, and risk
characterization. The analysis phase is further subdivided into the characterization of exposure
and human health effects.
August 2010
-------
U.S. Environmental Protection Agency
Figure 1. EPA/ILSI Generalized Framework for Assessing the Risks of
Human Disease Following Exposure to Food- and Waterborne Pathogens
(SOURCE: Adapted from ILSI, 1996)
The problem formulation stage is used to identify (1) the purpose of the risk assessment, (2) the
critical issues to be addressed, and (3) how the results might be used to protect public health.
Once identified, initial descriptions of the exposure and potential health effects are described and
then a conceptual model is developed. This conceptual model is used as a starting point for the
analysis phase of the risk assessment and later as an interactive tool along with components
developed in the analysis phase to initiate the risk characterization.
In the analysis stage, information about both the exposure and the health effects is compiled and
summarized. This compilation of quantitative and qualitative data, expert opinion, and other
information results in exposure and host/pathogen profiles that explicitly identify the data to be
integrated into the risk characterization and the associated assumptions and uncertainties. These
two elements, while separate, must also be sufficiently interactive to ensure that the results are
compatible. Specific features of the analysis phase are shown in Figure 2.
The final stage, risk characterization, results in a statement of the likelihood, types, and/or
magnitude of effects likely to be observed in the exposed population under the expected
exposure scenario, including all of the inherent assumptions and uncertainties. Often, the risk
characterization phase includes data integration through parameterization of a mathematical
model, numerical simulation, and interpretation.
August 2010
-------
U.S. Environmental Protection Agency
Figure 2. Elements of the Analysis Phase within the EPA/ILSI QMRA Framework (SOURCE:
Adapted from ILSI, 1996)
August 2010
-------
U.S. Environmental Protection Agency
III QMRA LITERATURE REVIEW: STATE-OF-THE-SCIENCE
A literature search (as detailed below) was performed to identify and obtain studies relevant to
QMRA or its components—exposure assessment, human health effects assessment (including
dose-response modeling), and risk characterization—as related to waterborne contaminants and
with a particular emphasis on risks posed by animal-derived pathogens. The literature search
also sought to identify the following:
• studies describing QMRAs or commonly used QMRA techniques relevant to
characterizing risks from recreational water exposure; and
• studies describing novel or cutting-edge QMRA-related techniques.
The literature search strategy and results of the literature review are summarized below.
III.l LITERATURE SEARCH STRATEGY AND SUMMARY OF RESULTS
A summary of the literature search strategy is presented in Table 1. Searches were made in the
specified databases for the primary keywords with results narrowed, if necessary, by secondary
keywords. Similarly, citation searches by the authors listed in Table 1 AND any one of the
primary keywords were conducted in the Web of Science database, in some cases using
secondary keywords.
Consistent with the approach used in a prior QMRA literature review (Seller et al., 2004), titles
and abstracts of studies identified from the literature search were assessed to determine which
studies were highly relevant. Highly relevant studies were assigned to the following categories
according to the study type or the element(s) of QMRA studied:
• QMRA studies (i.e., a specific risk is estimated or a methodology for estimating a
specific risk was explored);
• overviews and reviews;
• exposure assessment;
• dose-response modeling;
• sensitivity analysis;
• risk characterization; and
• application of Bayesian techniques.
III.2 DEVELOPING THE LITERATURE DATABASE
The literature search yielded more than 350 highly relevant studies. These studies were retrieved
and reviewed. Based on review of these articles, more than 160 additional studies were
identified for inclusion in the literature database. The additional studies were identified based on
key references in papers reviewed during the first round of literature survey or to fill data gaps
left open in the first round. Of the over 500 total studies acquired, more than 300 were related to
the state-of-the-science of QMRA (QMRA studies, papers reporting new or advanced
techniques, review papers, etc.). More than 200 contained data, analyses, or reviews related to
the occurrence, abundance, fate, or hazard of animal-derived pathogens or to manure handling.
The latter studies are not summarized in this report.
August 2010
-------
U.S. Environmental Protection Agency
Table 1. Literature Search Strategy
Databases
Authors
Primary Keywords
Secondary Keywords
Current Contents
Web of Science
Pubmed
Highwire
ASCE Civil
Engineering
Database
Environmental
Engineering abstracts
Water Resources
abstracts
Ashbolt, N.
Bollaert, K.
Buchanan, R.L.
Cassin, M.H.
Edberg, S.C.
Eisenberg, J.N.S.
Englhardt, J.
Gale, P.
Gerba, C.P.
Haas, C.N.
Koopman, J.
McBride, G.
Medema, G.J.
Messner, M.
Olivieri, A.
Parkin, R.
Petterson, S.R.
Pouillot, R.
Regli, S.
Rose, J.B.
Roser, D.
Seller, J.A.
Teunis, P.P.M.
QMRA
MRA
Microbial risk
Risk assessment
Exposure assessment
Dose response
Risk characterization
Fecal pollution
Indicator bacteria
Manure
Disease transmission
Secondary transmission
Sensitivity analysis
Variability
Susceptibility
Recreational water
Infection
Pathogen*
Microbial*
Fecal
Salmonella
E. co/; O157*
Campylobacter
Cryptosporidium
Listeria monocytogenes
III.3 QMRA: STATE-OF-THE-SCIENCE
The results of the literature review for QMRA and related studies are reported in this section. To
make this task tractable, it was necessary to prioritize the review so that the most relevant and
highest quality QMRA studies were examined in greatest detail. Narrowing the criteria for
including studies meant excluding some high-quality studies in the food literature, such as
studies primarily concerned with post-slaughter processes, the preparation of food products. For
example, the large number of studies of Listeria monocytogenes growth in deli meat storage was
not directly relevant to recreational waterborne exposure and was excluded. However, several
studies providing novel techniques for incorporating growth of Listeria monocytogenes into
exposure assessment are reviewed below (exposure assessment techniques). Drinking water
studies that focused on the role of treatment process in determining risk were also excluded,
although several studies that assessed the role of source water quality in finished drinking water
risk were included.
The following objectives were established for reviewing the selected QMRAs:
• development of a list of QMRA studies from which to draw elements of future study
designs;
August 2010
JO
-------
U.S. Environmental Protection Agency
• identification of the pathogens addressed in QMRAs and potential reasons the study
authors selected those pathogens;
• identification how variability has been addressed in QMRA studies, particularly source
variability and consumption variability;
• assessment of the tendency for risk analysts to include secondary transmission in the
estimate of overall risk; and
• comparison of the practices used by different QMRA researchers and practitioners,
particularly sensitivity analyses and risk characterization.
As noted in the preceding section, studies were added to the database that were cited in other
studies but not identified in the initial literature survey. Doing so was important, given the
tendency of authors to justify their choice of model structure and parameterization on the work
and choices of prior researchers. A total of 40 studies meeting the criteria described above were
selected for detailed review. The main exposure scenarios, indicators and pathogens evaluated,
and dose-response models employed in these 40 studies are summarized in Table 2, while Table
3 includes synopses of these studies.
Based on the results of the literature review, it can be generally inferred that the utility of QMRA
has been clearly demonstrated in a wide variety of scenarios. Moreover, (1) QMRA has been
used in conditions where analysis by other techniques such as epidemiological methods would
have been difficult or cost-prohibitive, and (2) QMRAs typically consider variability more
comprehensively than alternative techniques for assessing potential health hazards. For example,
many of the QMRAs reviewed for this report accounted for the variability in pathogen or
indicator density by treating them as stochastic variables.
Several specific observations may be drawn from comparison of the QMRA studies. First, the
assembled studies focused on a small subset of the pathogens potentially important in waterborne
exposure during recreation. The two pathogens analyzed most frequently, rotavirus and
Cryptosporidium, are both believed to be important contributors to risk of GI illness, primarily
due to their IDio (or other measure of low-dose infection), frequent occurrence in sewage, and,
particularly for Cryptosporidium, relatively high persistence in environmental matrices. As
reported in a recent review of outbreaks caused by waterborne viruses (Sinclair et al., 2009),
norovirus and echovirus (along with adenovirus) have been implicated in the majority of
outbreaks associated with waterborne viruses since the 1950s, making their absence from the list
of pathogens analyzed in the QMRAs in Table 3 conspicuous. Another potential reason for
frequent selection of rotavirus and Cryptosporidium is the availability of peer-reviewed dose-
response models based on oral ingestion. Numerous studies (Bastos et al., 2008; Eisenberg et al.,
2004, 2006; Hamilton et al., 2006; Ottoson and Stenstrom 2003; Petterson and Ashbolt, 2001b;
Seller et al., 2003, 2006) used rotavirus as a surrogate for enteric viruses. When considering the
general public, this approach is conservative, given that rotavirus has a higher probability that a
single organism can initiate infection than all other enteric viruses with known dose-response.3
When considering children, use of a dose-response model developed based on experiments on
adults may not yield a conservative estimate of risk.
2 QMRAs have recently begun to address risk associated with norovirus (e.g., see Schoen and Ashbolt, 2010; Seller et al.,
2010a,b). However, these studies are not summarized here as they were published after the literature review was conducted.
3 Comparison with norovirus is not made here, as the recent dose-response model (Teunis et al., 2008) uses dose units in cell
equivalents that precludes direct comparison with dose-response models based on culturable units.
August 2010 11
-------
U.S. Environmental Protection Agency
Table 2. Topics Evaluated in QMRA Studies Used to Establish the State-of-the-Science
Exposure
Scenario
Water, non-
drinking
Water, drinking
Food
Soil
Aerosol
Fomite
No. of
Studies
16
15
8
8
2
1
Pathogen
Cryptosporidium
Rotavirus
Giardia
Campylobacter spp.1
Salmonella enterica2
Enterovirus
Adenovirus
C. jejuni
E. coli O1 57
Fecal indicator
bacteria3
E. coif
Other5
No. of
Studies
17
15
8
7
6
5
4
3
2
2
2
1
Dose-Response
Model
Exponential
Exact beta-Poisson
Approx. beta-
Poisson
Beta-binomial
Empirical function of
enterococci density
Log-normal
No. of
Studies
19
2
17
2
2
2
1 Includes all Campylobacter studies; studies dealing specifically with C. jejuni are also included
2 Includes all Salmonella studies; the sole study dealing with Salmonella enterica Typhimurium is also
included
3 Not pathogenic
4 Excludes E. coli O157
5 Ascaris lumbricoides, coronaviruses, diarhheagenic E. coli, enterohemorrhagic E. coli, foot and mouth
diseases (FMD) virus, hepatitis A virus (HAV), Listeria monocytogenes, Salmonella enterica
Typhimurium, sanitary conditions.
The mismatch between pathogens potentially present in a particular setting and those chosen for
analysis in QMRA will only be resolved if the scientific community develops dose-response
models for other enteric viruses. Given the costs and ethical concerns related to human (feeding
trial) studies, it is unknown if additional studies will be conducted. Thus, further insights may be
limited to animal feeding studies (pending improved understanding of interspecies differences in
dose-response) or on the novel dose-response model development techniques described in
Section IV below.
Second, modeling variability in pathogen source density appears to be hampered by scarcity of
both data and analysis techniques. The two most common methods for accounting for source
variability among the studies are (1) use of empirical distributions for pathogen density based on
relatively short time series, and (2) assumption of log-normal distribution of pathogen densities.
Drawbacks to use of empirical distributions are inconsistency in sampling strategies used to
develop databases, frequent non-detects, and, most importantly, constraint of pathogen densities
to those observed in a limited number of samples. In sampling from a set of observations to
account for pathogen density variability, the estimates for pathogen density are constrained to the
highest and lowest observed values. This constraint prevents consideration of rare events
associated with potentially high risk, such as severe/chronic adverse health outcomes.
Use of distributions to describe pathogen density in sources overcomes the constraints associated
with use of empirical distributions. Among the studies reviewed for this report, many studies
August 2010
12
-------
U.S. Environmental Protection Agency
employed point estimates for pathogen source density. Among studies using distributions to
describe pathogen variability, the following distributions were employed: normal, triangular, log-
normal, negative binomial, uniform, and Poisson. Use of distributions to characterize pathogen
source variability raises difficulties in accounting for frequent non-detects of pathogens.
Pathogen occurrence tends to be tied to sporadic events such as rainfall events of a particular
magnitude (Signer and Ashbolt, 2006) or the occurrence of outbreaks among humans or animal
populations. As such, pathogen time series are characterized by frequent non-detects. One study
that accounted for non-detects in development of a distribution for pathogen occurrence
(Petterson and Ashbolt, 200Ib) divided pathogen observations into non-detects and detects and
then fit the pathogen densities from samples in which pathogens were detected separately.
Third, most of the studies reviewed in this report employ identical dose-response models but do
not account for variability and uncertainty in dose-response model parameters. As with
variability in exposure, this observation highlights the fact that high quality and diverse dose-
response model data are not available. The use of a small number of dose-response models may
indicate that some QMRA modelers choose the pathogens to model based on the availability of
dose-response models. Lack of dose-response models for many pathogens of public health
concern and for differing routes of exposure is a major data gap. The need for dose-response
models corresponding to different exposure routes (i.e., ingestion, inhalation, etc.) arises from
the ability of some waterborne pathogens (e.g., adenovirus) to initiate infection via multiple
routes. The exact beta-Poisson relationship was seldom used, as were empirical models popular
in food dose-response studies (e.g., as described in Buchanan et al., 2000; Moon et al., 2004).
Variability in dose-response model parameters or in response of the exposed population are
rarely considered or addressed. A likely cause for the latter is that dose-response model studies
do not consistently provide confidence intervals for model parameters and seldom present
quantitative information on the distribution form for parameter estimates. Another technique for
including variability of population response into risk estimation would be to assume population
response to be binomially or beta-binomially distributed. Alternatively, different dose-response
models might be used for sensitive and non-sensitive populations, as demonstrated for
Cryptosporidium by Pouillot et al. (2004).
Finally, secondary transmission and (temporary) immunity are often neglected in risk estimation.
Several studies (Eisenberg et al., 2004, 2008; Seller et al., 2006; Seller and Eisenberg, 2008;
Seller, 2009) have demonstrated that consideration of secondary transmission and immunity can
significantly influence overall risk associated with exposure to pathogens and often in unintuitive
ways.
August 2010 13
-------
U.S. Environmental Protection Agency
Table 3. Summary of Select Published QMRA Studies
Study
An et al.
(2007)
Ashbolt and
Bruno
(2003)
Risk of Interest
Occupational risk
associated with
reuse water in rice
paddies
Risk of Gl illness
and respiratory
illnesses associated
with recreational
waters
Microorganism(s)
£ co//
Enteric viruses and
adenovirus
Pathogen
Concentration and
Variability
£ co//
concentrations used
in Monte Carlo
simulations drawn
from distributions of
data from
experimental
studies; the authors
indicate £ co//
concentration
distribution was
normal without
justification or details
in parameter
estimation. Number
and timing of
irrigation events not
stated.
The ratio of
pathogens to
enterococci was
assumed relatively
constant; data on
enterococci
collected during the
study and reported
as number of
samples meeting a
compliance criterion.
Ingested Volume
or Mass
Not adequately
described; based
on 1000-fold and
10,000-fold
reduction in
volumes
associated with
"direct ingestion"
50 ml_ fixed
volume assumed
Dose-Response
Beta-Poisson
model for £. co//
(Haasetal., 1999)
(a =0.1 778, A/50 =
8.60x107)
50% of infected
persons assumed
to develop illness
Exponential dose-
response model
with r = 1 for
enteric viruses.
Adenovirus dose-
response model (r
= 0.41 7) for
respiratory illness-
associated viruses
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Not reported
Not reported
August 2010
14
-------
U.S. Environmental Protection Agency
Study
Bastos et al.
(2008)
Charles et
al. (2003)
Risk of Interest
Risk of infection
associated with
consumption of
crops irrigated with
reclaimed
wastewater
Reduction in risk in
occurrence of
pathogens in raw
drinking water
associated with
buffer distances
between septic
system and
receiving waters
Microorganism(s)
Rotavirus,
Campylobacter,
Giardia,
Cryptosporidium
Enteric viruses
(adenovirus,
enterovi ruses,
reoviruses,
norovirus, rotavirus,
HAV virus),
pathogenic
protozoa
Pathogen
Concentration and
Variability
Point estimates for
risk made using
ranges of pathogens
drawn from
measurement of
concentration in
treatment effluent
and using empirical
relations for
retention of bacteria
on food crops.
Range of
concentrations 0.1 to
1 organisms per 105
£. co// bacteria for
rotavirus and
Campylobacter and
0.01 to 0.1
organisms per 105 £
.co// bacteria for
Giardia and
Cryptosporidium.
Distributions based
on data from
monitoring
Ingested Volume
or Mass
Based on
statistics drawn
from official
Brazilian sources.
The authors
discriminated
between low and
high income
persons.
Not considered
Dose-Response
Beta-Poisson
model for rotavirus
(a =0.253, A/5o =
6.17)
Beta-Poisson
model for
Campylobacter (a
= 0.145, A/so = 896)
Exponential model
for Giardia (r =
0.0199) and
Cryptosporidium (r
= 0.0042).
Not considered
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Not reported
Sensitivity
analyses were
planned; the
impact of
factors such
as septic
system
management
and
disinfection on
risk to be
evaluated.
August 2010
15
-------
U.S. Environmental Protection Agency
Study
Dialloet al.
(2008)
Dowd et al.
(2000)
Risk of Interest
Risk of Gl illness
due to direct
ingestion of canal
waters during
recreation or other
activities, or due to
consumption of raw
crops irrigated with
canal waters
Risk of infection for
workers and near-
neighbors during
application of
biosolids
Microorganism(s)
Cryptosporidium,
Giardia, and
diarrhegenic £ coli
(DEC)
Rotavirus,
coronavirus,
Salmonella spp., £
coli
Pathogen
Concentration and
Variability
8% of measured £.
coli assumed DEC;
all pathogen
distributions
assumed triangular,
with median, lowest,
and highest values
based on data
collected in the
study
Pathogen
concentration at
point of ingestion
estimated based on
Gaussian dispersion
models for point
sources and areal
sources; release
rates from sources
based on
experimental
measurements of
aerosol transport
and abundance of
pathogens in
aerosols of
biosolids.
Ingested Volume
or Mass
Ingested volumes
of water 100 ml_
and 50 ml_for2
scenarios; soil
ingested masses
10 and 100 mg
Based on an
assumed normal
inhalation rate of 8
m3/day
Dose-Response
Beta-Poisson
model for DEC (a =
1778, A/so =
8.60x107)
Exponential model
for C. parvum (r =
Of\f\ ylO~7\
.00467)
Exponential model
for Giardia (r =
0.0198)
Exponential model
for Coxsackievirus
B3(r = 0.2532) and
beta-Poisson
model for
Salmonella typhi (a
= 0.3126, A/50 =
2.3x104)
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Multiple point
values
corresponding
to different
assumptions
assessed
Point
estimates
corresponding
to a range of
distances
from the
source, wind
velocities, and
durations of
exposure
August 2010
16
-------
U.S. Environmental Protection Agency
Study
Eisenberg
et al. (2004)
Eisenberg
et al. (2008)
Risk of Interest
Risk of infection
(including secondary
transmission and
immunity) from
direct ingestion of
soils amended with
biosolids
Risks associated
with direct ingestion,
aerosol exposure,
and groundwater
exposure to
pathogens
associated with
applied biosolids
Microorganism(s)
Enterovirus
Rotavirus (as a
surrogate for
enteric viruses)
Pathogen
Concentration and
Variability
Pathogen density in
raw sludge was not
reported; %
pathogens present
in biosolids (applied)
ranged from 0.1% to
0.5% of pathogens
in raw sludge;
simulations
considered
enterovirus removal
in treatment and
subsequent survival
in the environment
Raw sludge enteric
virus concentration
assumed to be log-
normally distributed;
pathogen removal in
treatment calculated
based on models of
treatment processes
and found to be
linearly-related to
retention time
Ingested Volume
or Mass
1 to 200 mg of soil
per day or less
(three point
values)
Direct ingestion
rate assumed 100
mg/day. An
ingestion rate of
1.L per capita per
day assumed for
groundwater
ingestion. An
average breathing
rateof0.83m3/hr
and exposure time
of 8 hours used
for aerosol
exposure.
Dose-Response
Approximate beta-
Poisson, with a
ranging from 0.126
to 0.5 and/?
ranging from 0.21
to 0.84
Beta-Poisson (a =
0.26, A/so = 5.62;
originally reported
as/?= 0.42 plaque
forming units (PFU)
Secondary
Transmission
Secondary
transmission
considered,
including the
possibility that
individuals
were in an
immune state
Secondary
transmission
considered;
secondary
transmission
estimated via a
deterministic
compartmental
transmission
model
Sensitivity
Analysis
Simulations
run for high,
medium and
low values of
most model
parameters.
Results
analyzed via
CART
(classification
and
regression
tree analysis)
Estimated risk
associated
with 3
different
sludge
treatments
compared
August 2010
17
-------
U.S. Environmental Protection Agency
Study
Flemming et
al. (2009)
Risk of Interest
Human health risk
from oral or aerosol
ingestion of soils
amended with fresh
or stored biosolids
Microorganism(s)
Campylobacter
spp., Salmonella
spp.,
Cryptosporidium
spp., Giardia spp.,
Clostridium
perfringens
Pathogen
Concentration and
Variability
Pathogen data fit to
log-normal
distributions via
censored regression
Ingested Volume
or Mass
Soil ingestion rate
for children: log-
normal distribution
with geometric
mean of 35 mg/d
and standard
deviation of log-
transformed
ingested volumes
3.94
Aerosol ingestion
based on average
breathing rate of
0.83 m3/h and
aerosol
concentration
based on
modeling
Dose-Response
Gompertz-log
model for
salmonellosis (see
Table 4) with ft =
2.148 and a
distributed
uniformly from 29
to 50.
Exponential model
for Cryptosporidium
infection, with r
uniformly
distributed from
0.04 to 0.16
Exponential model
for Giardia infection
with r = 0.01 99
Lognormal model
for Clostridium
perfringens with
mean of lognormal
= -24.7 and
standard deviation
of the lognormal =
2.32.
Secondary
Transmission
Not considered
Sensitivity
Analysis
Stochastic
model used
and risk
estimates
presented
with CIs
ivioaei
sensitivity
tested via
assessment
of alternate
scenarios
.
importance of
pathogens in
overall risk
assessed
August 2010
18
-------
U.S. Environmental Protection Agency
Study
Gale (2005)
Gerba et al.
(1996)
Risk of Interest
Risk of infection
through
consumption of root
crops grown on
agricultural lands
where treated
sewage sludge is
applied
Risk of rotavirus
infection from
recreational and
drinking water
exposures
Microorganism(s)
Non-typhi
salmonellas,
Campylobacter
jejuni (strain
A3249), Listeria
monocytogenes,
Escherichia coli
O157,
Cryptosporidium
parvum, Giardia,
enterovirus
Rotavirus
Pathogen
Concentration and
Variability
Arithmetic mean
values for pathogen
concentration in raw
sewage used along
with an event tree
approach to
estimate post-
treatment pathogen
concentrations.
Removal in all
phases of treatment
assumed known and
removal fractions for
the processes based
on data drawn from
published studies.
The exposure model
accounts for decay
after application and
removal via washing
prior to
consumption.
Drinking water
concentrations
estimates were
0.004 PFU/L and
100PFU/L, based
on review of the
occurrence of
rotavirus in drinking
waters and surface
waters and
assuming 99.99%
removal in
treatment; surface
water concentrations
estimated to be
0.24/L and 29/L (the
occurrence range)
Ingested Volume
or Mass
Proportion of the
population
consuming treated
sewage irrigated
crops and mean
consumption
masses taken
from European
Union and United
Kingdom
published data
Ingested volumes
used 100 ml_ for
recreational
exposure, 2 L for
child and adult
drinking water
exposure, and 4 L
for elderly drinking
water exposure
Dose-Response
Beta-Poisson
model for non-typhi
salmonellas (a =
0.3136, A/50 =
24,420),
Campylobacter (a
= 0.15, A/so = 795),
Listeria
monocytogenes (a
= 0.17, A/50 =
2.1x106),
Escherichia coli
O1 57 (a =0.1 6, A/so
= 1130), and
rotavirus (a-
0.265, A/so = 5.6).
Exponential model
for Giardia (r =
0.0199) and
Cryptosporidium (r
= 0.00419).
Beta-Poisson dose-
response model (a
= 0.26, A/so = 5.62)
used for risk of
infection. Risk of
clinical illness was
assumed 0.5 x risk
of infection.
Fraction of
illnesses
progressing to
mortality assumed
0.1% for the
general population
and 1.0% for the
elderly
Secondary
Transmission
Not considered
Secondary
transmission
rates
discussed, but
details on
calculations
not provided
Sensitivity
Analysis
No formal
sensitivity
analysis
performed;
assessments
made of the
model's major
uncertainties
Risks
corresponding
to high and
low
concentration
s in drinking
water and
recreational
waters
presented
August 2010
19
-------
U.S. Environmental Protection Agency
Study
Giannoulis
et al. (2005)
Hamilton et
al. (2006)
Risk of Interest
Risk of
contamination of a
groundwater
drinking water
source
Risk of enteric virus
infection associated
with consumption of
raw vegetables with
non-disinfected
secondary treated
reclaimed water
Microorganism(s)
Fecal conforms;
sanitary survey
results were also
factored into risk
determination
Rotavirus (as a
surrogate for
enteric viruses)
Pathogen
Concentration and
Variability
Risk based on
frequency
distribution of
combined sanitary
survey and fecal
coliform monitoring
results
Virus concentration
in irrigation water
assumed log-
normally distributed,
with the mean and
standard deviation
based on
experimental data.
Volume of irrigation
water retained after
irrigation was
assumed log-logistic
distributed for
broccoli, normally
distributed for lettuce
and cucumbers, and
based on empirical
data for cabbages.
Inactivation of
viruses between
irrigation and
harvest was
assumed to follow
first-order kinetics;
two estimates for
inactivation rate
parameter were
used.
Ingested Volume
or Mass
Not applicable
Ingested mass
based on
empirical
probability density
function for
consumption of 4
foods drawn from
U.S. Department
of Agriculture
(USDA) reports
Dose-Response
Risk considered
low when sanitary
survey score less
than 3 (scale of 1 to
10, with 10 being
high risk) and fecal
coliform count was
in category "B" or
"A" (scale of A to E,
with E being the
highest fecal
coliform count)
Beta-binomial
model (a=0. 167, ft
= 0.191; based on
fits of data to the
beta-Poisson
model)
Secondary
Transmission
Not applicable
Not considered
Sensitivity
Analysis
Not reported
Spearman
rank
correlation for
input
variables
August 2010
20
-------
U.S. Environmental Protection Agency
Study
Jolis et al.
(1999)
Julian et al.
(2009)
Risk of Interest
Risk of
cryptosporidiosis
associated with
exposure at parks
and golf courses
irrigated with tertiary
reclaimed water.
Risk of rotavirus
infection from
multiple exposure
routes fora child 6
years of age or
younger; exposure
routes were fome-to-
mouth, fome-to-
hand, and hand-to-
mouth
Microorganism(s)
Cryptosporidium
parvum
Rotavirus
Pathogen
Concentration and
Variability
Concentration of
Cryptosporidium
parvum in tertiary
effluent set to the
arithmetic mean of 6
samples (variability
not reported or
considered).
Concentration in
treated secondary
effluent taken as 2
logs less than the
mean of 3 samples
of secondary
effluent.
Virus density on
fomite assumed
uniformly distributed
(0.001 to 10
virus/cm2);
Inactivation rate on
fomite and hands
assumed normally
distributed (different
mean and standard
deviation for fomite
and hand
distributions).
Ingested Volume
or Mass
Assumed golfer
and park user
ingested volume
of 1 ml_ per outing
Transfer efficiency
from fome to
mouth and hand-
to-mouth
assumed normally
distributed with a
mean of 41% and
a standard
deviation of 25%.
Transfer efficiency
from fome to hand
was assumed
normally
distributed with a
mean of 36% and
a standard
deviation of 26%.
Dose-Response
Exponential
Cryptosporidium
parvum model (r =
0.00467, 95% Cl
<0.00195,0.0962>,
no information on
distributional form
assumed forr)
Ratio of illness to
infection set at 0.5.
Beta-Poisson dose-
response model (a
= 0.26, A/so = 5.62)
used for risk of
infection.
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Not reported;
authors
critically
assessed
findings in
their study
and
characterized
the study as
preliminary
Model run
with a
parameter set
to either the
25th or 75th
percentile
value of its
distribution
and all other
parameters at
the median
value.
Sensitivity to
a parameter is
assessed
based on the
ratio of the
p25 to the p75
estimated
risks.
August 2010
21
-------
U.S. Environmental Protection Agency
Study
Makri et al.
(2004)
Mara et al.
(2007)
Risk of Interest
Risk of
cryptosporidiosis
from ingestion of
drinking water
Risk of rotavirus
infection associated
with use of
wastewater for
restricted and
unrestricted crop
irrigation
Microorganism(s)
Cryptosporidium
parvum
Rotavirus,
Campylobacter,
Cryptosporidium
Pathogen
Concentration and
Variability
Concentration in
source water
assumed Poisson-
distributed. Oocyst
viability assumed
beta-distributed.
Oocyst recovery in
monitoring assumed
beta-distributed.
Concentrations used
in risk estimation
based on E. coli
occurrence and ratio
of E. coli to
Cryptosporidium and
Campylobacter.
Low and high values
for E. coli used for
point estimates. For
mechanized
agriculture, low and
high values 105 and
KTE. co// perl 00 g
soil and for labor-
intensive agriculture,
low and high values
104and 10s E. coli
per 100 g soil.
Ingested Volume
or Mass
Water
consumption
taken from
empirical data; 5th
and 95th
percentiles used
Ingestion
assumed for soil
particles from
lands irrigated
with reclaimed
wastewater. In
highly mechanized
agriculture,
ingestion rates
assumed 1 to 10
mg /day for 150
days. In labor
intensive
agriculture
ingestion 10 to
100 mg/day for
300 days.
Dose-Response
Lognormal dose-
response for
infection endpoint
(jU=-5A8,
-------
U.S. Environmental Protection Agency
Study
Ottoson and
Stenstrom
(2003)
Parkin et al.
(2003)
Risk of Interest
Risk of infection
from reclaimed
water in direct
consumption, use of
fields irrigated with
reclaimed water and
use of groundwater
influenced by
reclaimed water
Risk of enterovirus
infection to sensitive
population via
recreation in waters
receiving
wastewater
treatment plant
(WWTP) effluent;
study was a data
collection and
problem formulation
effort.
Microorganism(s)
Rotavirus,
Salmonella
typhimurium,
Campylobacter
jejuni, Giardia
lamblia,
Cryptosporidium
parvum
Coxsakievirus A
and B, echoviruses,
human
enteroviruses and
poliovi ruses
Pathogen
Concentration and
Variability
Concentrations of
pathogens in
grey water were
assumed
proportional to
measured
coprostanol
concentration with
the proportionality
derived from
epidemiology
studies.
Coprostanol
concentration
assumed log-
normally distributed
in greywater. Die-off
of pathogens
assumed first order
in all pathways
Anecdotal data on
virus occurrence in
swimming waters
reported, but no
characterizations of
temporal variation in
viruses found in a
literature search.
Ingested Volume
or Mass
Accidental
ingestion of 1 ml_
of untreated
greywater or 1 ml_
treated greywater;
ingestion of 1mL
per day for 26
days/yr;
consumption of
groundwater
(volume not
reported).
Not considered
Dose-Response
Beta-Poisson
model for rotavirus
(a =0.265, A/50 =
5.6),
Campylobacter
jejuni (a = 0.145,
A/so = 896) and
Salmonella
typhimurium (a =
0.3126, A/50 =
23,600).
Exponential model
for Giardia (r =
0.0199),
Cryptosporidium (r
= 0.00419) and
fecal enterococci (r
= 0.00565).
Epidemiology
studies indicate
that children are at
greater risk than
adults for
enterovirus
infection. The
effects of dose-
response and
exposure not
differentiated. The
authors noted there
are no known dose-
response relations
for children.
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Regression
analysis of
risks
predicted
using two
models
Not relevant
August 2010
23
-------
U.S. Environmental Protection Agency
Study
Petterson et
al.(2001a)
Risk of Interest
Risk of viruses on
lettuce irrigated
municipal secondary
treatment effluent
Microorganism(s)
Enterovirus (with
rotavirus as a
representative
virus)
Pathogen
Concentration and
Variability
Positive (detected)
secondary effluent
enteric virus
concentrations were
fit to log-normal and
Gaussian-kernel
type distributions.
27% of samples
were below the
detection limit.
Decay of viruses on
lettuce assumed to
follow first order
decay with the
inactivation rate
assumed normally
distributed;
distribution
parameters based
on measurements
with B40-8 phage.
Ingested Volume
or Mass
Mass ingested per
lettuce
consumption
event 100 g.
Dose-Response
Rotavirus dose-
response model
used; parameters
and assumptions
regarding their
distributional form
not provided
Secondary
Transmission
Not considered
Sensitivity
Analysis
Not reported;
stochastic
analysis
August 2010
24
-------
U.S. Environmental Protection Agency
Study
Pouillot et
al. (2004)
Roberts et
al. (2007)
Risk of Interest
Risk of
cryptosporidiosis via
consumption of
treated drinking
water
Risk of
cryptosporidiosis
associated with
fishing in an
urbanized stream
reach
Microorganism(s)
Cryptosporidium
parvum
Cryptosporidium
Pathogen
Concentration and
Variability
In the first scenario,
C. parvum density in
source water fixed at
values reported by
local health
authorities (in
oocysts per 100 L
finished water (1, 0,
0, 0,4,2,0, 1, 0,77,
1) and density
assumed Poisson
distributed. In a
second scenario, C.
parvum density in
source water
assumed negative-
binomially
distributed
Number of oocysts
ingested per month
via hand-to-mouth
transmission or in
consumption offish
assumed Poisson-
distributed.
Distribution
parameters
estimated using
occurrence of
oocysyts in hand-
washings and on
fish.
Ingested Volume
or Mass
Distribution of
ingested volumes
based on
empirical data
drawn from a
study of French
participants
Not calculated
separately from
pathogen
concentration
estimate
Dose-Response
Immunocompetent
population:
exponential dose-
response model, r
= 5.26x10'3, 95%
Cl [2.88x10~3,
10.9x10"3]
Immuno-
compromised
population:
exponential dose-
response model, r
= 0.354, 95% Cl
[0.221, 0.612]
Probability of
illness given
infection assumed
beta-distributed,
withp = beta(9,11),
based on
experimental data.
Exponential (r =
0.00419); the
dose-response
parameter was
treated as a
random variable,
though the
distributional form
used is not
reported
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
A second-
order Monte
Carlo method
used to
estimate risk
and
confidence
intervals for
all risks; this
allowed
evaluation of
sensitivity to
uncertain
parameters
Sensitivity
analysis
results
reported, but
details of the
method not
provided
August 2010
25
-------
U.S. Environmental Protection Agency
Study
Ryu and
Abbaszadeg
an (2008)
Risk of Interest
Risk of
cryptosporidiosis
and giardiasis from
drinking treated
surface waters
Microorganism(s)
Cryptosporidium,
Giardia
Pathogen
Concentration and
Variability
Oocyst and cyst
concentrations
equated to mean
daily concentrations
for seasonal and
annual estimates
Ingested Volume
or Mass
2 L per capita per
day
Dose-Response
Exponential model
for Giardia (r =
0.0199),
Cryptosporidium (r
= 0.00419).
Fraction of
pathogens
recovered capable
of initiating infection
0.41 for
Cryptosporidium
and 0.22 for
Giardia.
Secondary
Transmission
Not considered
Sensitivity
Analysis
Not reported
August 2010
26
-------
U.S. Environmental Protection Agency
Study
Schijven et
al. (2005)
Risk of Interest
Risk of foot and
mouth disease
(FMD) virus infection
in cattle consuming
surface waters into
which partially or
untreated milk
contaminated with
FMD has been
discharged
Microorganism(s)
FMD virus
Pathogen
Concentration and
Variability
The route of
infection assumed
discharge of
contaminated milk
into a sewer system,
inactivation of FMD
virus, discharge prior
to treatment or
discharge after
waste water
treatment, and
consumption of
surface waters by
cows.
Inactivation rate in
various media drawn
from prior studies;
concentration of
FMD virus in
untreated milk
assumed 160
TCIDso/mL.
Additional
calculations
performed at a milk
FMD density of
1.6x106 TCIDso/mL.
Ingested Volume
or Mass
Cows assumed to
drink 50 L per
capita per day
Dose-Response
Exponential
models; for calves
exposed to FMD
aerosols, r = 0.03
(95% Cl 0.01 7 to
0.051)
For pigs exposed to
p p r j"\ c n 1 c f —
dCIUoUlo, , / —
0.0016 (95% Cl
0.00074 to 0.003)
For pigs given oral
doses, , r= 4.1x10"
7(95%CI2.0x10"7
to 7.5x10"7)
Secondary
Transmission
Not considered
Sensitivity
Analysis
Several
scenarios and
dose-
response
models were
evaluated
(model
sensitivity was
evaluated)
August 2010
27
-------
U.S. Environmental Protection Agency
Study
Schijven
and de
Rod a
Husman
(2006)
Risk of Interest
Risk of infection for
occupational and
sport divers in fresh
and marine waters
Microorganism(s)
Campylobacter
jejuni,
enterovi ruses
Pathogen
Concentration and
Variability
Both pathogens
assumed log-
normally distributed
with the reported
lowest and highest
values (in the
literature) assumed
to be the 99% Cl
values.
Ingested Volume
or Mass
Ingested water
depended on diver
status
(recreational vs.
occupational),
setting (marine vs.
fresh vs.
swimming pool)
and on equipment
used, especially
mask type.
Reported ingested
volumes ranged
from 0 mLto 190
ml_. Number of
dives per year
was drawn from
an empirical
distribution.
Dose-Response
The exact beta
Poisson model with
oc= 0.145 and p =
8.007 used for
dose-response for
C. jejuni.
The rotavirus exact
beta Poisson model
with a = 0.167 and
f}= 0.191 used for
dose-response for
enteroviruses.
Secondary
Transmission
Not considered
Sensitivity
Analysis
Annual risk of
infection
differed
significantly
with diver
status
(occupational
vs.
recreational),
equipment
used, and
setting.
August 2010
28
-------
U.S. Environmental Protection Agency
Study
Seidu et al.
(2008)
Risk of Interest
Risk of (1)
accidental
consumption of
wastewater by
farmers (2)
accidental
consumption of
contaminated soils
by farmers; (3)
accidental
consumption of both
wastewater and soil
by farmers and (4)
consumption of
wastewater-irrigated
lettuce collected
from the farm,
wholesaler or retail
market.
Microorganism(s)
Rotavirus, Ascaris
lumbricoides
Pathogen
Concentration and
Variability
Rotavirus density
estimated as fecal
coliform density *
10"5; rotavirus
assumed log-
normally distributed
in stream and ditch
water, in soils and
on lettuce; rotavirus
assumed uniformly
distributed in piped
water.
Estimates for A.
lumbricoides density
distribution in
various media drawn
from published
studies. Like
rotavirus, distribution
of A. lumbricoides in
all media, except
pipe-borne water,
assumed log-
normal. The
distribution of A.
lumbricoides in
piped water
assumed uniform.
Ingested Volume
or Mass
Accidental
irrigation water
ingestion
assumed
uniformly
distributed from 1
to 5 mL/day for
150d/yr;
accidental
ingestion of soil
assumed
uniformly
distributed from 10
to 100 mg soil/d
for 150 days/yr.
Lettuce
consumption
assumed
uniformly
distributed from 10
to 12 g per
serving, 208
exposures/yr
Dose-Response
Beta-Poisson
model for rotavirus
(a =0.253, A/50 =
6.17)
An exponential
dose-response
model with r = 1
used for A.
lumbricoides',
model chosen
because there no
peer-reviewed
dose-response
model available
and because it is
the most
conservative
estimate
Secondary
Transmission
Not considered
Sensitivity
Analysis
Not reported
August 2010
29
-------
U.S. Environmental Protection Agency
Study
Shuval et al.
(1997)
Risk of Interest
Risk of infection
associated with
consumption of
uncooked
vegetables from
fields irrigated with
wastewater
Microorganism(s)
HAV
Pathogen
Concentration and
Variability
Fecal coliform
concentration in
applied wastewater
is assumed to be
107CFU perl 00 g
of vegetable. Die off
before harvest
assumed 3 logs and
ratio of hepatitis A
virus to fecal
conforms is
assumed 1:105.
Ingested Volume
or Mass
Daily average
consumption of
lettuce and
cucumbers
assumed to be
100 g and number
of days both
vegetables are
consumed
assumed to be
150 days/yr.
Dose-Response
Beta-Poisson
model for HAV
median infectious
dose estimated
based on
unattributed data
(A/so = 30 to 1 000
PFU)and a= 0.2
(with no
justification).
Attack rate
estimated as 0.5 for
lettuce
consumption and
0.25 for
cucumbers.
Secondary
Transmission
Not considered
Sensitivity
Analysis
Risks
reported for a
high and low
estimates of
IDso and for
two levels of
wastewater
treatment
August 2010
30
-------
U.S. Environmental Protection Agency
Study
Signer and
Ashbolt
(2006)
Risk of Interest
Human exposure to
pathogens via
drinking water when
routine pathogen
monitoring
conducted.
Microorganism(s)
Cryptosporidium
spp.
Pathogen
Concentration and
Variability
During baseflow
conditions, untreated
water
Cryptosporidium
density log-normally
distributed with
mean and standard
deviation of log-
transformed
densities equal to
3.11 and 1.28,
respectively.
During event
(rainfall) conditions,
untreated water
Cryptosporidium
density log-normally
distributed with
mean and standard
deviation of log-
transformed
densities equal to
5.27 and 0.61,
respectively.
Ingested Volume
or Mass
Ingested (oral)
volume log-
normally
distributed with
mean and
standard deviation
of log-transformed
densities equal to
-0.046 and 0.535,
respectively
Dose-Response
Exponential model,
r = 0.00419
Secondary
Transmission
Not considered
Sensitivity
Analysis
Model
sensitivity
assessed via
comparison of
three
sampling
scenarios
August 2010
31
-------
U.S. Environmental Protection Agency
Study
Signer et al.
(2007)
S meets et
al. (2007)
Risk of Interest
Risk of infection
from drinking water
following a
rainfall/runoff event
Risk of infection
from
Cryptosporidium in
treated drinking
water
Microorganism(s)
Cryptosporidium,
Giardia,
Campylobacter
Cryptosporidium
Pathogen
Concentration and
Variability
Drinking water
reservoir influent
pathogen
concentration data
corresponding to wet
and dry weather fit
to log-normal and
gamma distributions.
Parameter
uncertainties were
quantified via
Markov Chain Monte
Carlo (MCMC)
analyses.
Cumulative
distribution function
for Cryptosporidium
density in treated
drinking water taken
from empirical
distribution with low
concentrations
extrapolated from
data. Distribution of
Cryptosporidium in
finished drinking
water also estimated
based on stochastic
model of drinking
water treatment.
Ingested Volume
or Mass
1.1 L/day
Number of 190 ml_
glasses of drinking
water consumed
assumed Poisson-
distributed, with
mean equal to
0.53 L
Dose-Response
Beta-Poisson
model for
Campylobacter (a
= 0.145, A/50 =
896). Exponential
model for Giardia (r
= 0.0199),
Cryptosporidium (r
= 0.00419).
Cryptosporidium
dose-response
model beta-
Poisson with oe =
0.115and/? =
0.176
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Influences of
uncertainties
(quantified via
MCMC) and
variability
(estimated via
Bayesian
techniques)
on risk
assessed via
factor
sensitivity
analysis
(worst-case
scenario
determination)
Analysis
stochastic and
risk results
presented as
statistical
distribution
August 2010
32
-------
U.S. Environmental Protection Agency
Study
S meets et
al. (2008)
Seller at al.
(2003)
Risk of Interest
Risk of infection
from Campylobacter
in treated drinking
water
Risk of viral
gastroenteritis
associated with
recreational use of a
river downstream of
a WWTP discharge.
Two wastewater
treatment scenarios
compared.
Microorganism(s)
Campylobacter
Model enteric virus
with the clinical
features of
rotavirus
Pathogen
Concentration and
Variability
Source water
Campylobacter
density was taken
from historical raw
water density
measurements and
CIs developed using
bootstrap analysis.
Removal in
treatment estimated
via stochastic model
of all treatment
processes.
Bacteriophage
concentration in raw
wastewater
assumed uniformly
distributed in the
range 1x104 to
5x104. Removal
modeled for
treatment and
removal and mixing
processes modeled
for discharged
effluent. The ratio of
model enteric virus
concentration to
bacteriophage
concentration
assumed log-
uniform distributed in
the range 0.001 to
1.0.
Ingested Volume
or Mass
0.177 L unboiled
water / d
Exposure factor
was a random
variable chosen
from uniform
distributions
whose ranges
were selected
based on
observed
recreational use
by month and day
of the week
(weekday vs.
weekend)
Dose-Response
beta-Poisson dose-
response model
with
-------
U.S. Environmental Protection Agency
Study
Seller et al.
(2006)
Steyn et al.
(2004)
Risk of Interest
Risk of infection
during full-body
contact recreation
Risk of infection via
drinking water or
waterborne
recreation
Microorganism(s)
Rotavirus as a
representative
pathogen
Salmonella
Pathogen
Concentration and
Variability
Rotavirus density
based on a model
calibrated with
empirical coliphage
data. The
relationship between
coliphage density,
expected rotavirus
density and fraction
of total pathogen
load made up by
rotavirus based on
literature review.
Salmonella density
determined during
monitoring.
Calculations
performed for the
geometric mean
value (167 CFU/100
ml_), the minimum
value (36) and the
maximum value
(883)
Ingested Volume
or Mass
Hourly rate of
water ingestion
assumed.
Swimmers were in
the water at
different times and
for different
durations.
For full contact
recreation,
ingested volume
assumed 100 ml_
Dose-Response
Beta-Poisson
(presented in study
in modified form)
with a assumed
uniformly
distributed in the
range 0.125 to 0.5
and p in the range
0.21 to 0.84
Probability of
symptomatic
response range 0.1
to 0.45.
Approximate beta
Poisson dose-
response, with a =
0.3126 and A/so =
23,600
Secondary
Transmission
Secondary
transmission
modeled via a
deterministic
time-
dependent
transmission
model
accounting for
the immune
status of the
population
Not considered
Sensitivity
Analysis
Sensitivity
analyses
performed for
several
variables;
variables
were set to
low, medium,
and high
values to
determine
whether their
variation
changed the
study findings
Not reported
August 2010
34
-------
U.S. Environmental Protection Agency
Study
Strachan et
al. (2002)
Teunis et al.
(1997)
Risk of Interest
Risk of £. co// O1 57
infection of humans
using pasture for
recreational
activities
Risk of infection by
Cryptosporidium or
Giardia in drinking
water
Microorganism(s)
Escherichia coli
O1 57
Cryptosporidium
parvum, Giardia
lamblia
Pathogen
Concentration and
Variability
Empirical data for
concentration in
cattle feces (range 0
to > 5 log™ CFU/g)
provided the
distribution of
pathogens in feces.
The number of days
prior to human
exposure that
animals were in
fields was a random
variable with a
uniform distribution.
Empirical data for
Giardia and
Cryptosporidium fit
with a negative
binomial distribution;
because seasonal
variations in both
parasites noted,
data sets broken into
winter and summer
sets and fit
separately.
Treatment efficiency
assumed beta-
binomially
distributed.
Ingested Volume
or Mass
Mass of soil
ingested during a
24-hour camp and
during an 8 hour
day were random
variables with
triangular
distributions
Log-normal
distribution of
drinking water
consumption,
mean equal to
0.153L/daywith
uncertainty factor
of 8.2.
Dose-Response
Dose-response
assumed to follow
a beta-binomial
dose-response
model; metrics for
assessing model fit
not provided.
Details of the
model not provided.
Exponential dose-
response relations
used for both
Cryptosporidium
and Giardia.
Readers referred to
cited studies for
parameters.
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Sensitivity
analysis
(called
importance
analysis) was
performed
and results
were
presented as
Spearman
Rank
correlations
Stochastic
risk
assessment
performed;
no explicit
sensitivity
analysis
reported
August 2010
35
-------
U.S. Environmental Protection Agency
Study
van
Heerden et
al. (2005)
Westrell et
al. (2004)
Risk of Interest
Risk of human
adenovirus infection
via drinking water or
recreational water
exposure.
Risk of infection
during treatment,
handling and soil
application of sludge
and wastewater, and
risk of infection via
consumption of
crops irrigated with
biosolids or via
recreation in waters
receiving wastes
from land-applied
wastes
Microorganism(s)
Adenovirus
Rotavirus,
adenovirus, EHEC,
Salmonella,
Cryptosporidium,
Giardia.
Pathogen
Concentration and
Variability
Adenovirus density
assumed Poisson-
distributed (in time,
not space) with the
distribution mean
determined from
frequency of positive
determinations
among drinking
water and surface
water samples.
Mean adenovirus
densities (in viruses
per 100 ml_) were
0.0014 and 0.00245
for two drinking
waters, 0.0546 for
river water, and
0.0097 for water
behind a dam.
Pathogen
concentrations in
untreated sewage
based on measured
concentrations;
distributional forms
and parameters not
reported.
Inactivation rates in
anaerobic digestion
drawn from the
literature, but not
reported. Die-off
after land application
assumed negligible
or outpaced by
regrowth.
Ingested Volume
or Mass
Drinking water
consumption rate
fixed at 2 L per
capita per day and
recreational water
consumption rate
fixed at 30 ml_ per
capita per day.
Not reported.
Dose-Response
The exponential
model used for
adenovirus dose-
response. The
model parameter
not explicitly
provided, though
based on the
citation provided in
the study, it can be
inferred to be that
for inhalation of
adenovirus
aerosols, r = 0.417.
Parameters and
models not
reported; other
studies were also
cited
Secondary
Transmission
Not considered
Not considered
Sensitivity
Analysis
Univariate
sensitivity
analyses
conducted to
assess the
impact of
consumption
rates, dose-
response
parameters
and recovery
rates on risk
estimates
Not reported
August 2010
36
-------
U.S. Environmental Protection Agency
Study
Wong et al.
(2009)
Risk of Interest
Risk of enteric virus
infection associated
with swimming at
coastal beaches
impacted by POTW
discharges.
Microorganism(s)
Adenovirus
Pathogen
Concentration and
Variability
Experimental
distribution for
adenovirus
occurrence based
on regression on
order statistics to
account for non-
detect observations.
Ingested Volume
or Mass
1 00 mL/day
Dose-Response
Exponential model,
r = 0.417 (based on
data for inhalation
of adenovirus
aerosols)
Secondary
Transmission
Not considered
Sensitivity
Analysis
Not reported
1 The Gompertz dose-response model is found in Table 4.
Note that several relevant studies were also published after this review was conducted but before it was finalized. Thus, it was not feasible to include Schoen and
Ashbolt (2010) or Seller et al. (2010a,b,c).
August 2010
37
-------
U.S. Environmental Protection Agency
IV NOVEL AND CUTTING-EDGE QMRA-RELATED TECHNIQUES
This section summarizes and describes techniques specific to exposure assessment, health effects
modeling, and risk characterization that are novel, new, or beyond the approaches that typically
have been employed in QMRA studies. For each of these topical areas, an overview of
"classical" approaches QMRA is provided, followed by a review of the novel techniques
identified in the state-of-the-science literature review. Where possible, results of papers
describing recommended practices or methods for selecting techniques are also provided.
I V.I EXPOSURE ASSESSMENT
Exposure modeling of animal-derived pathogens is a particularly difficult component of
developing QMRAs for animal-derived pathogens. Like most biological systems, there is
substantial variability in many facets of exposure to animal-derived wastes. Pathogen loads are
known to vary between individual animals, from farm to farm, and from region to region.
Manure handling practices, too, differ greatly among farms and regions and profoundly impact
persistence and abundance of pathogens that can potentially reach receiving waters. Pathogen
survival (or growth) varies between pathogens, with environmental media, and with other
conditions such as moisture content, temperature and pH. Uncertainty of estimates of pathogen
and indicator loads may be significant, particularly for pathogens like Cryptosporidium spp. that
are highly infectious, and for pathogens with frequent non-detects.
In this section, novel techniques for conducting exposure assessments or estimating model
parameters and their variability are described. Both deterministic and stochastic models are
reviewed, and the uses of Bayesian methods in exposure assessment are highlighted.
IV. 1.1 General Description
Pathogen loads from animal sources differ with animal, season, region, manure management
practices, and method of land application. The prevalence of pathogens differs significantly
among animals, with cattle producing a high proportion of E. coli O157:H7 and Cryptosporidium
loading; poultry and dairy cattle contributing significantly to Campylobacter loading; and swine
and poultry both contributing high loads of Salmonellae. Manure handling varies widely
between U.S. farms, with manure directly deposited on pasture, land-applied as solids, or land-
applied as slurries. Solid manure and manure slurries may undergo treatment before application,
though the degree of treatment for applied manure may not be even for all applied manure.
Ferguson et al. (2003) divide processes governing the relationship between watershed pathogens
and concentrations in surface waters into those most important in organism inactivation
(water/osmotic potential, temperature, sunlight, pH, and inorganic and organic nutrients) and
those most important in transport (adsorption/desorption effects, hydrological movement, and
mechanical or biological movement). These processes are illustrated in Figure 3, which shows
that wastes of animal-origin reach receiving waters via multiple pathways (e.g., in surface runoff,
in interflow, after adsorption/desorption to soils or vegetation) and the transport of pathogens and
fecal indicator organisms is dependent on processes that are highly variable.
August 2010 38
-------
U.S. Environmental Protection Agency
Barnyard runoff
Storage system
leakage
/ / / / s /
/ ,-',-' xVVV" /'/'/'' Precipitation
Sunligbt
Ultraviolet radiation
Water table
Air temperature
Freeze-tbaw cycle
Humidity
Soil type, '$ 'V D[Y|ng
water content, pH V*fe Dilution Vegetative
"---,, Interflow' -4^-- cover
""**--, Soil bacteria
^~"---_ Competition $oj| organ|c matte|. ae|.atjon
Riparian buffer
Figure 3. Factors Affecting the Viability of Pathogens and Indicators Along with Pathways
(SOURCE: Adapted from Rosen, 2000)
Atwill et al. (2002) describe the pathway followed by enteric microorganisms from fecal material
to a specified downslope location (in this case a receiving stream) as comprising the following
steps:
1. rainfall of sufficient intensity erodes the top layer of a fecal pat, releasing pathogenic
organisms onto the wetted soil surface;
2. rainfall intensity reaches infiltration capacity and pathogenic organisms are carried
downslope via sheet flow, preferential rill flow, or exfiltration in variable source areas;
3. pathogens are transported downslope or infiltrate into the subsurface; and
4. vegetative buffer strips intercept flows laden with pathogenic organisms and enhance
infiltration.
Other processes not discussed by Atwill and colleagues that may be significant in determining
overall loading and transport of pathogens from fecal material to receiving waters are treatment
practices (for agricultural wastes), adsorption to plants and other matter, inactivation via
exposure to UV radiation, desiccation or toxic materials, and predation.
A final aspect of exposure assessment that has recently received attention is the distribution of
pathogens and indicators in environmental matrices (e.g., Englehardt et al., 2009; Gale, 2005)
and consideration of method uncertainty and use of censored data within QMRA frameworks
(e.g., Petterson et al., 2007, 2009; Signer and Ashbolt, 2006).
IV.1.2 Cutting-Edge Exposure Assessment Techniques
Cutting edge techniques for exposure assessment found from the literature search include use of
Bayesian methods for leveraging data related to one component of the exposure pathway to
develop knowledge about conditions in another part of the pathway where data are scarce; use of
August 2010
39
-------
U.S. Environmental Protection Agency
highly resolved transport models, including the use of stochastic fate and transport models with
varying degrees of resolution; and the use of alternatives to the log-normal and Poisson
distribution for describing the temporal and spatial distribution of pathogens in environmental
settings. Each of these types of techniques for exposure assessment is discussed below.
Application ofBayesian techniques to exposure assessment
The use ofBayesian techniques in exposure assessment was demonstrated by Eisenberg et al.
(2008) in their risk analysis for exposure to amended biosolids (treated sludge) projects. The
authors used prior information about pathogen concentration in raw sludge and removal efficacy
of sludge treatment processes to predict post-removal pathogen concentrations in biosolids. The
data for post-treatment pathogen concentrations were then used to inform the likelihood of
pathogen occurrence. As described elsewhere in this report, transport of animal-derived fecal
pollution is highly variable and many of the processes comprising the transport are relatively
poorly characterized in the literature. Bayesian techniques such as those demonstrated by
Eisenberg and colleagues offer a means for using data from better-characterized processes to
improve estimates for processes associated with greater variability or for which there are fewer
data. Other studies have used Bayesian techniques in exposure analysis for quantifying
uncertainty associated with assumed distributional forms for Cryptosporidium density (Signer et
al., 2007) and for quantifying uncertainty related to microbial counts (Clough et al., 2005).
In microbial risk analyses, Bayesian techniques offer the opportunity to estimate model
parameters given a relatively small amount of data pertaining to a highly variable system. For
example, Clough et al. (2003) used Bayesian methods to estimate herd-level prevalence from
pathogen prevalence in fecal pats. Variabilities in the system included the fecal material
produced per animal per day (particularly fecal production for infected and non-infected
animals), farm-to-farm differences in infection rates and fecal production rates, seasonal
variations in infection prevalence and fecal production rates, and number of animals in herds.
Prior distributions used for number of infected animals included uniform and beta-binomial
distributions. These analyses indicate that the posterior distribution is relatively insensitive to
the choice of prior distribution and that the number of pat samples used influences the CI for
parameter estimates.
Ranta et al. (2005) also applied Bayesian techniques to the problem of estimating Salmonella
infection in cattle herds and animal populations. In surveillance of cattle for Salmonella,
individual animals are either infected or not, and are either tested or not. The rate of detection of
an infected cow, therefore, depends both on the rate of infection and the frequency of testing.
Testing generally occurs if an animal exhibits adverse health effect symptoms but may occur for
other reasons. The probability of an animal being tested given that it shows symptoms was
assigned a beta prior distribution while the probability that an animal is tested given no
symptoms was assigned a uniform prior distribution. Based on the model and knowledge of
testing results for Salmonella among Finnish cattle, the authors developed an estimate for the
overall prevalence of Salmonella infection among Finnish cattle and determined that assessment
of the status of Finnish cattle could be made with a "modest" sample size.
Petterson et al. (2001b) used Bayesian methods to estimate virus inactivation rates on salad
crops. Bayesian methods were chosen over other techniques such as bootstrap methods to
reduce computational complexity of estimating inactivation rate parameters. Prior distributions
for model parameters (some of which were log- or logistic-transformed) were normal and were
August 2010 40
-------
U.S. Environmental Protection Agency
tested to ensure choice of prior distribution did not unduly influence posterior distributions. As
for the previously discussed researchers, Petterson and colleagues found that Bayesian
techniques allowed use of basic assumptions for certain biological processes while replacing
"hidden" distributional assumptions for other processes with transparent analyses. Additionally,
the outputs of Bayesian analyses were deemed to allow a more meaningful investigation of
uncertainty. From their analyses, Petterson et al. (2001b) determined that a negative binomial
distribution of viruses on salad crops was a better characterization than a Poisson distribution and
estimated inactivation rates on multiple salad crops.
Bayesian techniques have also been used in the estimation of parameters for distributions
describing variability of pathogens in environmental waters (Petterson et al., 2007; 2009;
Pouillot et al., 2004). For example, Petterson et al. (2009) developed a point estimate for
uncertainty in counts of E. coli O157:H7 using data from experiments in which the pathogen was
spiked into water samples and detected via polymerase chain reaction (PCR). They then used
Bayesian techniques to develop a distribution describing the temporal variability of E. coli
O157:H7 at a raw water intake of a drinking water treatment plant. In the hierarchical model
used in that study, several distributions—including a gamma distribution, log gamma
distribution, and a constrained log-gamma distribution—were evaluated as potential meta-
distributions for E. coli O157:H7 counts. Estimates of pathogen density were found to be highly
dependent upon the selected model type, which indicates the need to use additional information
in selection and development for models describing temporal variability in density. Improved
models could be developed via inclusion of fate and transport models (Ramachandran, 2001),
distributions based on analysis of additional or alternative data, or expert judgment. In a similar
study of the treatment of distribution of Cryptosporidium oocysts in drinking water source
waters, Petterson et al. (2007) showed the importance of separate analyses of method uncertainty
and variability. Hierarchical Bayesian analysis was used to develop a distribution for oocyst
density. Estimates for the parameters of the distribution of oocysts were dependent upon
whether method uncertainty and oocyst variability were treated separately and on the
assumptions made of method uncertainty.
In the food microbiology and risk literature, several studies have been published documenting
use of Bayesian techniques for estimation of growth rate parameters in different matrices. These
techniques could be extended to circumstances in which extra-enteric growth relevant to
recreational water exposures may occur. Three studies (Albert et al., 2005; Crepet et al., 2009;
Pouillot et al., 2003) have used hierarchical Bayesian models for development of growth models
ofListeria monocytogenes in foods. These studies are an exemplary application of Bayesian
techniques because of uncertainty in growth parameters for the pathogen and because of the
variability in growth rate due to factors (e.g., temperature, physiological state of inocula,
variability among strains, etc). In the study by Pouillot et al. (2003), expert knowledge regarding
difference in growth rates among strains was required to develop parameter estimates.
Albert et al. (2005) used the growth uncertainty and variability parameters determined by
Pouillot et al. (2003) within a Monte Carlo simulation of L. monocytogenes growth in milk
during storage prior to transport. The model was developed for incorporation into a QMRA
model of risks associated with exposure to L. monocytogenes in milk. During storage, the
temperature of the milk varies. Other variable elements in the growth process are related to
timing (milking time, storage time, etc.), bacterial abundance in milk added to storage, and
differences among bacteria related to strain. A deterministic model of the L. monocytogenes
August 2010 41
-------
U.S. Environmental Protection Agency
growth process was run as an element of a Monte Carlo simulation in which input variables (23
total) were drawn from appropriate distributions. Despite what the authors refer to as a "simple"
process model, their findings were that the stochastic model could not succinctly characterize the
growth process.
Crepet et al. (2009) performed a Bayesian analysis similar to that of Pouillot et al. (2003) to
determine the variaiblity and uncertainty of L. monocytogenes growth parameters on fresh leafy
salads. As for the studies conducted for milk, the use of Bayesian techniques allowed use of data
corresponding to a range of environmental conditions in parameter estimation and accounted for
uncertainty and variability more comprehensively than previously published growth models.
Use of high-resolution fate and transport models as components in QMRAs
Highly-resolved hydrodynamic and transport models have been developed for analysis of the
stocks, flows, sources and sinks of pathogens in watersheds (a review of established watershed
transport models may be found in Coffey et al., 2007); in coastal waters (e.g., by Liu et al.,
2006); for mixing in a river reach downstream of a WWTP discharge (Seller et al., 2003); for
flow in a combined sewer overflow (CSO)-impacted river and bay (King County Department of
Natural Resources, 1999a); and in analysis of an impaired waterbody (Seller et al., 2006). For
effective use of highly-resolved models in QMRA, two challenges must be overcome: (1)
collection of data required for model development or calibration, and (2) experimental design of
modeling such that conditions corresponding to conditions that can reasonably be expected to
arise are considered. As for samples collected on a particular day, models predict microbial
occurrence for a set of input hydrodynamic, biological and physical conditions. Modeling efforts
should therefore be designed to account for variation in these conditions rather than producing a
"snapshot" related to a single set of conditions.
A dated but still informative example of a watershed scale fate and transport model is provided
by Walker and Stedinger (1999). Their Cryptosporidium source, fate, and transport model
included addition of oocycts to manure on fields, storage and inactivation of oocysts after
deposition, washoff of manure and oocysts into watercourse, oocyst routing in streams, oocyst
fate in a drinking water source reservoir, and occurrence of oocysts in undisinfected drinking
water. Calculations were based on the generalized watershed loading function model that, in
turn, used a soil conservation model for erosion from fields. A single set of input parameters
were used as input to a single year-long time period simulation. Their results indicated that dairy
oocyst loads were small compared with those of WWTP effluents and that there is a strong
seasonality in oocyst loading of the drinking water source. Using the QMRA framework, results
of a single run of a model such as this with a single (albeit representative) set of input parameters
are of limited value. Two approaches for using deterministic models such as that of Walker and
Stedinger (1999) or results from such models within a stochastic framework more conducive to
use in QMRA are described below.
A comparison and critique of two hydrologic and erosion models that have been adapted for
modeling transport of bacteria is provided by Benham et al. (2006). Models described in that
review for release of microorganisms from fecal material are a linear model (equation 1), an
exponential model (equation 2), a power law model (equation 3), and an empirical model
(equation 4).
(1)
August 2010 42
-------
U.S. Environmental Protection Agency
(2)
(3)
(4)
In equations 1 to 4, AA//?is number of microorganisms released during time period A^; Ns is
number of organisms in the manure top layer; AQ is the runoff yield for the time period A^; TVi/is
the total number of microorganisms in surface applied manure before a runoff event; Q is the
runoff depth during the rainfall event; PW is the density of water; nid is the dry mass of applied
manure; and ki, #2, £j, a, 6, and (3 are empirical parameters.
Ferguson et al. (2003) note that other factors including moisture content of the manure, whether
the manure originated as diarrhea, and slope may have a strong influence on the mobilization of
microorganisms from land-applied manure. For manure applied as a slurry, leaching losses of
microorganisms from the applied manure may be significant and the rate of leaching is
dependent on how well the soils are drained.
Downslope transport of microorganisms occurs in both surface and subsurface flows (Benham et
al., 2006; Ferguson et al., 2003) and tends to follow preferential flow paths. Benham and
colleagues hypothesize that larger amounts of bacteria are transported in rills than in sheetflows
and that accumulation of organisms occurs in microponds. Features of subsurface soils believed
to create preferential flow paths for microorganisms include groundcover (i.e., planted regions
appear to enhance infiltration of microorganisms), tilling practice, and earthworm burrows
(Ferguson et al., 2003).
Adsorption of organism to media (soils, vegetation) remains poorly understood and models to
predict it are rudimentary. As reported by Benham et al. (2006), two commonly used watershed
transport models employ a simple linear partitioning model (equation 5) to predict the fraction of
bacteria sorbed to soil particles.
S = KDC (5)
In equation 5, S is the sorbed bacteria density (CFU per g), C is the bacteria concentration in
suspension (CFU/mL), and KD is a partition coefficient. Data for estimating KD for pathogens of
interest is a data gap in the knowledge of organism fate and transport in watersheds, streams, and
coastal waters. Ferguson et al. (2003) suggest adhesion of bacteria to soils may play a minor role
in transport during periods of high rainfall intensity, when the majority of microorganism
transport occurs.
Atwill et al. (2002) evaluated the filter efficiency of vegetative buffer strips of differing soil
types, slopes, and vegetative cover. In that study, vegetative buffer strips were assembled in soil
boxes and subjected to artificial rainfall of varying intensities. Removal of C. parvum by the
buffer strips was between 1-log/m for a buffer constructed with sandy loam soil to 3-log/m for a
buffer constructed with a silty clay or loam soil. Pathogen removal was observed at moderate
rainfall rates and for slopes as high as 20%.
Several reports of the use of stochastic models for occurrence, persistence, and growth of
pathogens were found in the literature database. These reports, described in greater detail below,
August 2010 43
-------
U.S. Environmental Protection Agency
generally predict distribution of pathogen or indicator occurrence based on Monte Carlo
simulations of stochastic models or Monte Carlo simulations of deterministic models whose
parameters are chosen from distributions for each simulation. Montville and Schaffner (2005)
used a Monte Carlo analysis of pathogen growth during sprout production to evaluate monitoring
schemes focused on reducing the incidence of contaminated sprouts reaching consumers. In
contrast to the model of Albert et al. (2005) discussed above, their model was purely stochastic.
The model inputs included pathogen prevalence on seeds (which can be surprisingly high), decay
rates, growth rates, and detection probabilities for monitoring at different phases of the sprout
production process. The resulting model indicated that disinfection of seeds alone could not
ensure pathogen elimination and that monitoring of seeds for pathogens is an essential part of an
overall risk management strategy.
Alternatives to log-normal and Poisson distributions for describing spatial and temporal
variability
Most commonly, the distribution of indicators and pathogens at a particular location and time is
described by a log-normal distribution or Poisson-log-normal distribution. The spatial
distribution of microorganisms in a well mixed volume at a moment in time is described by a
Poisson or, less commonly, by a negative binomial distribution. Although quite useful and
relatively easy to manipulate, these distributions may not be appropriate for use in circumstances
such as analysis of time series with frequent non-detects; analysis of systems with non-
homogeneous distribution of microorganisms (e.g., due to gradients with distance source,
association of microorganisms with particles or detached biofilms; clumping of microorganisms
with each other or due to harboring of microorganisms in other organisms); or analysis of
distribution of organisms after processes such as water treatment and wastewater treatment
(Gale et al., 2002).
Englehardt et al. (2009) recently proposed and verified a theoretical distribution for describing
microbial counts in water. Their model is developed based on the concept that the number of
organisms present at a particular time and location is the result of some initial number of
organisms at some prior time undergoing series and parallel events — each changing the number
of organisms. The distribution is termed the discrete growth distribution (DGD); however, it is
important to note that growth does not refer to growth of the microorganisms. Rather, it refers to
a particular mathematical process. Reported advantages of the DGD over alternative models are
that it has a theoretical basis and, perhaps most importantly according to the author, when fit to
data it can predict the probability that zero organisms are observed. The Poisson distribution and
other discrete distributions share the ability to predict the occurrence of no organisms. The DGD
distribution is given by
v'
P(V) = -IL - ; 0<<7<1, r/>0 (6a)
V,
where v is a discrete number of organisms, q is parameter of the distribution (0 < q < 1), and 77 is
a parameter of the distribution related to the number of processes (called causes) leading to the
distribution observed. The denominator in the expression for the DGD is denoted Dq, ^ and can
be approximated by
August 2010 44
-------
U.S. Environmental Protection Agency
M
!)*(-Ing)]
(6b)
where Mis a parameter governing the accuracy of the approximation and F indicates the upper
incomplete gamma function. Englehardt et al. (2009) did not provide guidelines on the selection
of M To estimate distribution parameters for the DGD, equation 6a can be linearized via
double-log transformation, to the form
^ - log
'(0).
(6c)
A plot showing the DGD probability distribution with the Poisson distribution and negative
binomial distribution is presented in Figure 4. Not that the three distributions were generated for
illustration of their differences (i.e., not fit to actual data). All three distributions have the same
expected value and the parameters for the negative binomial distribution were chosen such that
the resulting distribution is similar to the DGD distribution (though no optimization was
performed). Figure 4 also shows that the DGD and negative binomial distributions are similar to
each other and very different from the Poisson distribution. Both the DGD and negative
binomial distributions are relatively flat at the lower range of indicator density. The DGD offers
the advantage of being mechanistic (i.e., based on a series of plausible events) and therefore
o
?
0)
CN
O
0)
Zi
ro
-Q
O
0)
<£>
O
A
o
o
° DGD
0 Negative binomial
o Poisson
o
VD
OD
<> CD
O
OD
on
O
O
o
°o
O D
°nn
O D
\
2
5 10 20 50 100
Number of organisms
Figure 4. DGD, Negative Binomial, and Poisson Probability Distribution Illustration
August 2010
45
-------
U.S. Environmental Protection Agency
appropriate for extrapolation outside the data range for which data are available. The negative
binomial offer simplicity in use, including convolution with other distributions for developing
analytical dose-response models.
As described above, alternatives to the log-normal and Poisson distributions for describing
spatial variability in indicator density are the negative binomial and DGD. These distributions
are both flatter than the Poisson distribution and are appropriate for describing the distribution of
indicators at a site with known heterogeneity. Other distributions have been suggested for
describing indicator and pathogen occurrence, though none of the studies reviewed for this report
employed them. These alternative distributions are (Haas et al., 1999) the Poisson log-normal,
the Poisson inverse Gaussian, and the Poisson Generalized Inverse Gaussian (Poisson-GIG).
IV.1.3 Summary: Cutting-Edge Techniques for Exposure Modeling
Characterization of variability and uncertainty of the complex systems giving rise to human
exposure is perhaps the most pressing need in QMRA exposure assessment. Techniques in
development for filling this need include the following:
• estimation of model parameters for deterministic and stochastic fate and transport models
using Bayesian techniques;
• incorporation of high-resolution fate and transport models (e.g., computational fluid
dynamic models, field- and drainage-scale overland fate and transport models) into
QMRA models—preferably within a stochastic (Monte Carlo) framework;
• explicit treatment of both variability and uncertainty within QMRA; and
• development of novel distributions that better characterize pathogen and indicator
occurrence and temporal and spatial variability.
The application of Bayesian techniques and the development and implementation of high-
resolution fate and transport models are accessible to QMRA practitioners. In cases where use of
these techniques is justified, analysis of complex systems that are recalcitrant to analysis with
conventional techniques may be evaluated and characterized.
IV.2 HEALTH EFFECTS MODELING
IV.2.1 Dose-Response Modeling
Dose-response modeling overview
Dose-response models relate the density of a pathogen, surrogate organism, or indicator
organism to the probability of a particular (adverse) response (endpoint). Endpoints for dose-
response models employed in previous QMRA studies have included infection (presence of
antibodies or other measureable changes such as the presence of substances in blood); illness
(exhibition of symptoms such as diarrhea); mortality (especially in studies of category A
biological agents4) (Bartrand et al., 2008); acute febrile response; and development of lesions in
lung tissue, skin, or their organs.
4
According to the U.S. Centers of Disease Control and Prevention (CDC), high-priority agents include
microorganisms that pose a risk to national security because they (1) can be easily disseminated or transmitted from
person-to-person; (2) result in high mortality rates and have the potential for major public health impact; (3) might
August 2010 46
-------
U.S. Environmental Protection Agency
The exponential and beta-Poisson are the dose-response models most often used in QMRA
performed for waterborne pathogens. Favorable properties of those two models are that they are
mechanistic (can be derived based on assumptions about the infection process) and that they
predict a non-zero probability of infection at low dose (Haas et al., 1999). The exponential dose-
response model, presented in equation 7, is derived based on the assumptions of a Poisson-
distributed dose and equal and independent probability, r, of each ingested organism initiating an
infectious focus.
p(R\d) = \-e-rd (7)
In equation 7, R denotes response, dis dose, and r is the probability that a single organism
initiates an infectious focus. The beta-Poisson dose-response model, shown in equation 8, is
based on the assumptions that the ingested dose is Poisson-distributed and the probability of a
response at a given dose is beta-distributed.
p(R\d) = \-,F,(a,a + p-d) (8)
In equation 8, \F\ denotes a confluent hypergeometric function and a and (3 are parameters of the
distribution. An approximation to the exact beta-Poisson model, presented in equation 9a
(Furumoto and Mickey, 1967), is frequently used. The approximate version is given by
(9a)
that may also be reparameterized as follows:
P(R\d) = \-
d
(9b)
where NSQ is dose corresponding to median response (e.g., ID50). The relation in equation 9b
provides a good approximation to the exact beta-Poisson model when |3»1 and oc«p.
Other dose-response models have been used in MRA studies, especially in literature pertaining to
risks associated with food. Representative examples are reviewed below, with mechanistic dose-
response models presented first, empirical models presented second and models developed using
Bayesian techniques.
Teunis et al. (1999) proposed a dose-response model for gastroenteritis illness given infection.
Assuming the duration of illness is gamma-distributed the authors derive the following dose-
response model for illness given infection
^(illness infection) = 1 ——l—— (10)
(l + i//A)
In equation 10, a is a parameter of the distribution (the shape factor for the gamma distribution
for disease duration) and y/A is treated as a single dose-dependent parameter. The authors
evaluated the following functional relationships for i/s. y/A = r/d; y/A = 77 /d; i//A = rj , finding
cause public panic and social disruption; and (4) require special action for public health preparedness. See
http://www.bt.cdc.gov/agent/agentlist-category.asp#b for further information.
August 2010 47
-------
U.S. Environmental Protection Agency
that the probability of illness given infection was not dose-dependent for Salmonella enterica
and Cryptosporidium, and that the probability of becoming ill when infected decreases with
increasing dose for Campylobacterjejuni.
Numerous empirical dose-response models have been used in prior QMRAs and related studies,
particularly in food risk literature. Representative examples are listed in Table 4; however, a
discussion of these dose-response models is beyond the scope of this report. Unless there are
compelling reasons to the contrary, the exponential and beta-Poisson dose-response models are
preferred over these alternatives because they are biologically-based and better suited to low-
dose extrapolation. Although this appears to be a majority position in the literature (as evidenced
by the use of the exponential and beta-Poisson or related dose-response models in the majority of
studies in Table 3), there is a minority but significant opinion to the contrary. Brookmeyer et al.
(2005) developed a mechanistic dose-response model (referred to as a "competing risks" model)
for inhalation of anthrax. The model relies on assumptions of a constant, uniform risk of anthrax
spores being cleared from the lung, and a constant, uniform risk of germination of anthrax
sporesin the lung. The resulting dose-response model is
(11)
where F(f) is the cumulative attack probability function (probability that a spore germinates at or
before time i), A: is rate of clearance of spores from the lung, and $is spore germination rate.
Table 4. Select Empirical Microbial Dose-Response Models
Model
Equation
Weibull-Gamma /'(
-------
U.S. Environmental Protection Agency
The two model parameters were estimated using physiological (clearance rates from lungs) and
microbiological data (spore germination rates) rather than quantal response data. When the
estimated parameters were used in equation 11, the model accurately predicted a median
incubation period for an anthrax outbreak. Note that the exponential dose-response model results
from equation 11 when t —» QO. The implication of this equivalence is that, in the absence of
dose-response data or a published dose-response model, there is the potential to estimate the
exponential dose-response mode parameter based on physiological and microbiological data
related to the pathogen of concern.
Hamilton et al. (2006) used a beta-binomial dose-response model to estimate the incidence of
gastroenteritis resulting from consumption of vegetables irrigated with reclaimed water. The
authors describe the beta-binomial model as an extension of the beta-Poisson model, although it
is unclear whether the dose is assumed Poisson-distributed in the model presented. The dose-
response model, as described in their study, is presented in equation 12. Details on the
derivation of the model are not provided in the study by Hamilton et al. (2006) or the study
(Cassin et al., 1998).cited as the source of the beta-binomial model.
P(R\d) = \- ^ \ (12)
(a + (3 + d-\ j
{ a )
Haas et al. (1999) present three empirical dose-response models (the log-logistic model,
equation 13; the log-probit model, equation 14; and the Weibull model, equation 15) for
consideration, although they indicate that these models should be used with caution because their
accuracy may be poor under conditions significantly different from those for which the models
were developed. Additional empirical models, including a Weibull-gamma model (equation 16)
were assessed by Holcomb et al. (1999), who also assessed a flexible exponential model that is
not described here because details on its use and derivation were not provided.
P = (13)
- q2 \n(d)]
(14)
(15)
(16)
y 2
In equations 13 to 16, q\, q^ /?, %, and s are parameters and $(y) = Fexp( )dx . In their
comparison of dose-response models for risk of infection in food consumption, Holcomb et al.
(1999) determined that the Weibull-gamma model (equation 16) yielded the best fit to data they
were studying. This finding is not surprising, since the Weibull-gamma has three parameters and
August 2010 49
-------
U.S. Environmental Protection Agency
all other models had two or fewer parameters. The authors did not test whether the 3-parameter
model provided a statistically-significant improvement in fit over models with fewer parameters.
Cutting-edge dose-response modeling techniques
Bayesian techniques
Bayesian methods are being increasingly used in microbial risk assessment to estimate dose-
response model parameters and their uncertainties. These methods, described briefly below,
provide opportunities for development of dose-response models for small data sets, or when
sufficient data are not available for use of frequentist (traditional) techniques, or to improve
dose-response model confidence using data that might otherwise not have been used in dose-
response model development.
In the simplest implementation of Bayesian inference for estimating parameters for dose-
response models, an assumption is made that dose-response characteristics for related exposures
belong to a distribution of responses whose precise form is unknown but whose distribution
arises from the biology, pathology, and other features of the pathogen-host combinations. In the
case of dose-response modeling, the parameter(s) of the dose-response model is/are assumed to
be a random variable drawn from the known or unknown distribution. Bayesian inference
involves making an assumption about the functional form of the distribution of dose-response
parameters (referred to as the "prior" or prior distribution) and treating observed responses as
conditional probabilities. Using Bayes' theorem (equation 17), if R is a set of observations of
responses (each corresponding to a different dose) and 9 is the set of values from which the
dose-response model parameters arise, the probability that a particular parameter describes the
dose-response observations, R, is given by
- (17)
The conditional probability P(R 9} (i.e., the probability that the observations, R, occur given a
specific set of parameters, 9) is equal to the likelihood of the observations. If host heterogeneity
in response is low and response at a particular dose may be assumed binomially distributed, the
likelihood of observing^? given
# doses
where nt is the number of subjects in dose group i,pi is the number of positive responses in dose
group /', and 7t(9, d) is the dose-response model (e.g., exponential or beta-Poisson) evaluated at
parameter(s) 6 and dose dt. If host heterogeneity were high, the likelihood function in equation 2
could be replaced with one based on an alternative distribution such as a beta distribution.
The mechanics of generating estimates for dose-response model parameters then become
• selection of the most appropriate dose-response model, n(0, d) (or comparison of
multiple models);
• selection of a prior distribution describing the variation in dose-response model
parameters, P(ff); and
August 2010 50
-------
U.S. Environmental Protection Agency
• evaluation of equations 17 (using equation 18) over the range of plausible values of the
model parameter(s).
The result of this procedure, termed the posterior distribution, gives the refined probability
distribution for the dose-response model parameter(s) consistent with available data. The
integral in the denominator of equation 17, in general, precludes analytical solutions and
estimation of the posterior distribution is usually made via MCMC methods (Gilks et al., 1996).
Meta-analysis (or hierarchical modeling) of multiple data sets corresponding to differing
conditions (e.g., experiments conducted using different strains of a pathogen, as explored by
Messner et al. (2001) may be conducted within the Bayesian framework. In Bayesian meta-
analyses, an assumption is made about the distributional form of the population from which
dose-response model parameters is drawn. The parameters of the distribution assumed for the
model parameters are called the hyperparameters and are estimated in the course of analysis.
The posterior distribution is calculated as described above using data from all experiments
included in the meta-analysis. The resulting posterior distribution may be used for evaluating
parameters for conditions different from those under which data were collected (e.g., to estimate
expected dose-response parameter and credible interval for an unknown strain) and as a
component in microbial risk assessment.
As with frequentist approaches, model fits from Bayesian inference must be assessed prior to
adoption. Alternative approaches for assessing dose-response models and application of
Bayesian techniques include the following:
• use of a Q-Q plot (graphical technique for assessing whether data fit a given distribution)
to verify the fit to the assumed distribution;
• use of a graphical plot of the data and the model to assess dose-response model fit as well
as the classic likelihood ratio test to compare the dose-response model with the best
(completely parameterized) model;
• Bayesian information criterion (BIC), deviance information criterion (DIC),or other
appropriate tests of model fit;
• cross-validation (subj ect to availability of data);
• assessment of the sensitivity of the posterior distribution to the choice of prior
distribution (relative insensitivity of the posterior distribution to choice of prior indicates
the prior is not biasing the posterior distribution); and
• comparison of Bayes factors (Kass and Raftery, 1995).
Studies in which dose-response models have been developed using Bayesian techniques are
described below and compared in Table 5. In Table 5, analysis type specifies whether individual
data sets were analyzed separately ("individual") or whether multiple data sets corresponding to
different experiments of factors were analyzed as part of a meta-analysis. When a meta-analysis
was performed, the factor(s) that differed between data sets are reported.
Messner et al. (2001) used Bayesian techniques to refine dose-response parameter estimates for
three isolates (IOWA, TAMU, and UCP) of Cryptosporidium parvum and to explore the
variation of the dose-response parameter among isolates and for unknown strains. Bayesian
methods were selected, in part, because of high variation in dose-response among strains. First,
August 2010 51
-------
U.S. Environmental Protection Agency
Table 5. Comparison of Bayesian Dose-Response Studies
Study
Messneret al.
(2001)
Messneret al.
(2001)
US EPA (2003,
2006)
Teunis et al.
(2002a)
Teunis et al.
(2004)
Teunis et al.
(2005)
Teunis et al.
(2008)
Engelhardt
(2004)
Englehardt
and Swartout
(2004)
Pathogen
Cryptosporidium
parvum
C. parvum
C. parvum
C. parvum
E. co/;0157:H7
Campylobacter
jejuni
E. co/;O157:H7
Rotavirus
C. parvum
Analysis Type
Individual models
for 3 isolates
Meta-analysis
Meta-analysis
Meta-analysis
Individual models
for teachers and
pupils
Meta-analysis
Meta-analysis
Individual
Meta-analysis
Factor Varied in
Meta-Analysis
Not applicable
Isolate
Isolate
Isolate
Not applicable
Outbreak and
feeding study
data
Outbreak data
Not applicable
Isolate
Dose-Response
Model
Exponential
Exponential
Exponential
Generalized
hypergeometric
Beta-Poisson with
variables transformed
Beta-Poisson infection
model (parameters a,
ft) with conditional
probability of illness
(parameters 77 and r)
Beta-distributed
probability that
pathogens initiate an
infectious focus and
negative binomial
distribution of dose
Exact beta-Poisson
Exact beta-Poisson
Prior Distribution for Dose-
Response Parameter(s)
Uniform log-transformed
Log-normal
Logit-normal, Logit-t, and beta
Uniform log-transformed for all
parameters
Uniform(0,1) for transformed variable
u = a/(a + j3)
Normal(0,10) for transformed variable
v = log10(« + /?)
All dose-response parameters (a, ft, 77,
and r) were transformed and
noninformative priors were used for all
transformed variables
A log-normal prior was used for C.
jejuni density in milk
Priors for transformed variables were
normal; hyperpriors for the mean and
standard deviation of transformed
variables were normal- and gamma-
distributed
Uniform (noninformative) priors for
both a and ft
Not described
August 2010
52
-------
U.S. Environmental Protection Agency
Study
Englehardt
and Swartout
(2006)
Chen et al.
(2006)
Pathogen
C. pan/urn
C. jejuni
Analysis Type
Individual
Meta-analysis
Factor Varied in
Meta-Analysis
Not applicable
Isolate
Dose-Response
Model
Modified beta-Poisson
for illness endpoint,
morbidity ratio
independent of dose
Approximate beta-
Poisson
Prior Distribution for Dose-
Response Parameter(s)
A joint, noninformative log-normal prior
used for the model parameters (a and
ft
Hyper-parameters (assumed to be the
mean and standard deviation of the
parameter distributions) assumed
normally-distributed with an extremely
wide range of parameter values;
transformations used to generate
hyperparameters not explicitly defined
August 2010
53
-------
U.S. Environmental Protection Agency
Messner and colleagues assumed all data were best fit by an exponential dose-response model
and analyzed data sets for individual strains separately using a noninformative prior (;r[ln(&)] =
constant) to develop estimates for the dose-response parameter and credible interval for each
strain. Second, they conducted a meta-analysis in which the prior for dose-response model
parameter, &, was assumed to be log-normal (among isolates) and hyperparameters for the prior
distribution were the log-mean and standard deviation of log-transformed parameter. In that
study, meta-analysis only slightly changed parameter estimates for individual strains, but
provided insights into the envelope of values within which the dose-response model parameter
for an unknown isolate of Cryptosporidiumparvum is expected to fall.
In developing a dose-response model for use in evaluating criteria for Cryptosporidium density
in finished drinking water, EPA used Bayesian techniques to account for variability in dose-
response model parameter among isolates and to assess hypotheses about the distribution of the
dose-response model parameter over the range of C. parvum isolates (U.S. EPA, 2003, 2006).
As described above (see Messner et al., 2001), the exponential model dose-response parameter
varied by two orders of magnitude among three isolates for which human graded dose data were
available. There were also concerns that data corresponding to one of the isolates were not
representative of human response because of the status of the isolate at the time it was
administered to volunteers. To explore the sensitivity of the dose-response model to selection of
data and meta-distribution choice for the dose-response model parameter, EPA generated and
compared four dose-response models. A MCMC technique was used to estimate the parameters
for the distribution of the exponential dose-response model parameter, r, and subsequently the
expected value and 95% credible interval for r. The resulting credible intervals were sensitive to
both model choice and data choice. Because all the models were considered equally plausible,
the model selected for use in regulation was a composite of the alternative models. Additional
means for comparing models would have been to assess the parsimony of the models or to
compare the fits of the models (e.g., with Bayes factors).
Teunis et al. (2002a) analyzed the same Cryptosporidium parvum data as Messner et al. (2001)
but based their Bayesian analysis on the two-parameter exact beta-Poisson dose-response
models, which accounted for variability within strains and between strains. Uniform log-
transformed priors were selected for all (four) dose-response model parameters. In contrast to
the model developed by Messner and colleagues (2001), the model developed by Teunis et al.
(2002a) accounts for heterogeneity at the pathogen level, and is more consistent with
experimentally-observed dose-response. Teunis and colleagues suggest this difference makes
their model a better candidate for estimating population-level risks.
Teunis et al. (2004) assessed outbreak data (number of teachers and pupils exposed, number of
teachers and pupils ill, and dose) for the development of an E. coli O157:H7 dose-response
model using Bayesian techniques. This study demonstrates the ability of Bayesian techniques to
develop a dose-response model using data insufficient for development of a model using
frequentist approaches. A beta-Poisson dose-response model was assumed because it accounts
for heterogeneity in the host-pathogen interaction and because it has desirable behavior at low
dose. The model parameters were transformed for analysis and the priors assigned to the
transformed variables were a uniform distribution for u = a/(a + /?) and a normal distribution for
v = logio(a + P). The authors note that use of only two data points is an extreme demonstration
of the ability (and necessity) of Bayesian models in dose-response model development with
limited data.
August 2010 54
-------
U.S. Environmental Protection Agency
Subsequent work by Teunis et al. (2008) accounted for heterogeneity in dose among individuals
in anE1. coli O157:H7 outbreak using data (estimated dose, number exposed and attack rate)
from nine previous outbreaks with different exposure vehicles (food, water, soil, and incidental
direct ingestion). This information was used to develop a dose-response model that is consistent
with variability in infection and illness arising from different conditions under which outbreaks
occur. These conditions include the vehicle by which hosts are exposed to pathogens,
differences in characteristics of the exposed population, genetic variations among pathogens
associated with different outbreaks, and condition of the pathogen in different outbreaks. In
contrast to models developed using data generated in controlled exposures, Bayesian models
based on outbreak data may capture the variability associated with typical and unintended human
exposures and infections. Teunis et al. (2008) proposed a novel dose-response model to account
for heterogeneity in exposure among individuals for each outbreak. Assuming the number of
pathogens in a given vehicle (mass of food, volume of water) is Poisson-distributed and that the
Poisson mean is gamma-distributed among different food samples, water samples, (etc.), the
dose associated with a particular outbreak may be expressed as Poisson-gamma mixture (i.e.,
negative binomially distributed). Under these assumptions, the probability of exposure is
^(CF,0)=l-(l + ^j (19)
where C is average density of the pathogen in a given source, Fis the ingested volume of the
source, and ^is a parameter of the distribution. Assuming the ability of pathogens to initiate
infection is beta-distributed (an underlying assumption of the beta-Poisson model), the
probability of infection of a negative-binomially distributed dose is given by
(20)
where iF\ denotes a hypergeometric function. Teunis et al. (2008) used fixed estimates of the
dispersion parameter, $ and estimated the dose-response parameters a and (3 via hierarchical
Bayesian analysis. Beta-Poisson model parameters were transformed as follows:
z = log(v)
where
a
U 'a + B (22)
Priors for w and z were normal, and hyper-priors for the mean and standard deviation of w and z
were normal and gamma, respectively. The dose-response model and credible interval for the
parameters resulting from the analysis were suggested to be an improvement over dose-response
models proposed in prior studies because it more accurately accounts for heterogeneities present
in non-controlled exposures.
Outbreak data were evaluated in a Bayesian framework along with feeding study data in a study
of Campylobacter jejuni dose-response and risk (Teunis et al., 2005). Two features of that study
that distinguish it from other Bayesian dose-response studies are use of feeding study and
August 2010 55
-------
U.S. Environmental Protection Agency
outbreak data for development of a single model and use of a four-parameter dose-response
function developed in a prior study (Teunis et al., 1999). Outbreak data were for two outbreaks
of campylobacteriosis associated with ingestion of contaminated milk. In the dose-response
model, the probability of illness at a given dose is the product of the probability of infection
(P(d\a,P) when the beta-Poisson model is used), and the conditional probability of illness given
infection (h(d\r,rj) when the model for conditional probability of illness proposed by Teunis et
al., 1999 is used). Here, r and 77 are models parameters:
(l + r}dy] (23)
Estimates of dose used for outbreak data were based on reported milk consumption (data were
binned as "none," "draught," "l/2 cup", "1 cup", and "2 cups") and density ofC.jejuni in milk,
which was considered a random variable with a log-normal prior distribution. Dose-response
relations for the combined outbreak and feeding study data were markedly different from those
developed using feeding study data alone. The authors noted that, as for most feeding studies,
those conducted for campylobacteriosis utilized immunocompetent adult volunteers, whereas the
outbreak data were mainly for children. This fact merits consideration of the modified dose-
response relation from the Bayesian analysis in QMRAs, because it includes effects on a
sensitive population (children) and consequently may provide a more conservative estimate of
population level risk.
Englehardt (2004) proposed use of Bayesian techniques for dose-response models developed to
overcome limitations of scarce data for low dose and the limited number of data sets for
describing exposure and infection scenarios. Exposure and infection scenarios can be expected
to be highly variable. A beta-Poisson dose-response relation was selected for evaluation and
noninformative (uniform) priors were used for the dose-response model parameters (a and /7). A
single data set for infection of human volunteers with rotavirus was used to demonstrate the
utility of the technique. The dose-response model, evaluated at an unspecified value of the
Bayesian parameter estimates (perhaps the median values), was compared with a modified set of
parameters estimated via traditional optimization techniques and an upper limit dose-response
curve corresponding to an exponential model with r = 1. The dose-response curve developed via
Bayesian techniques fell between the other curves. However, the CI for the traditional model
and the credible interval for the Bayesian model were not provided in that study, making
comprehensive comparison of the models difficult. The author advocates increased use of
Bayesian techniques for more complete inclusion of expert knowledge in risk assessment and
because the techniques are well suited for data-scarce applications such as microbial dose-
response modeling.
In their analysis of Cryptosporidium dose-response, Englehardt and Swartout (2004) addressed
the bias introduced into dose-response models developed based on quantal dose-response
experiments employing homogeneous host populations. Noting that none of the human
volunteers in three Cryptosporidium feeding studies were immunocompromised and that the
antibody positive (Ab+) population may have been over-represented, the authors conducted a
parametric bootstrap analysis to generate a set of dose-response model (beta-Poisson) parameters
that would better reflect risks to the general population. In their analysis, 10% of the population
was assumed to be sensitive to Cryptosporidium infection and 20% was assumed to Ab+.
Among sensitive individuals, a dose of 10 oocysts or higher was assumed to cause infection.
Dose-response models from the bootstrap analysis for three isolates of Cryptosporidium
August 2010 56
-------
U.S. Environmental Protection Agency
(analyzed separately) were not substantially different from estimates made using the original data
sets—perhaps indicating that the Ab+ population offset the sensitive population in the overall
population incidence of infection. Like Messner et al. (2001), Englehardt and Swartout (2004)
used Bayesian methods to explore variability in response between isolates ofCryptosporidium.
Rather than directly predicting the distribution for parameters of the beta-Poisson model, the
authors determined the single pathogen probability of initiating infection, r, in the Bayesian
meta-analysis. The likelihood function was based on the conditional probability of observing the
mean single-organism probabilities of infection as determined in the bootstrap analysis described
above. The prior used in the analysis is not explicitly stated.
Similar to the analysis of Teunis et al. (2005) for C.jejuni, Englehardt and Swartout (2008)
incorporated conditional probability of illness given infection into a Bayesian analysis of C.
parvum dose-response. Based on novel modeling and interpretation of the literature, the authors
developed a dose-response model, termed the generalized beta-Poisson model, in which the
proportion of infected persons becoming ill, 7, was independent of dose (one of the relationships
explored in a previous study of GI illness by Teunis et al., 1999):
Pm(d
7
1-
/3 »1 and a « /3
(24)
7 D~i^i (a>a + P,-d)] otherwise
where iFi denotes a confluent hypergeometric function. The modified and unmodified beta-
Poisson models were both fit to aggregated (pooled) data for five isolates ofCryptosporidium via
maximization of the log-likelihood ratio ("traditional techniques"). The improvement in fit the
3-parameter modified model provided over the 2-parameter beta-Poisson model was not
significant however, and the unmodified beta-Poisson was selected over the modified version for
further analyses. Parameters (a and /?) for the aggregate (pooled) data set for all 5 isolates were
estimated via Bayesian techniques and a joint, noninformative log-normal prior was used for the
model parameters.
Chen et al. (2006) explored the impact of Campylobacter jejuni isolate (14 total) and origin
(fresh vs. laboratory) on dose-response using Bayesian analysis. Dose-response parameters were
transformed as described in Teunis et al. (2008) and priors were chosen such that their influence
on the predicted parameters was deemed minimal. Significant differences in the envelope of
dose-response models for the fresh and lab isolates were noted, with the IDso for the fresh
isolates found to be less than that for the lab isolates. The researchers also reported a greater
spread (range of doses corresponding to each isolate's IDso) among the fresh isolate dose-
response curves. Lab isolate dose-response curves tended to have shallower slopes (higher intra-
isolate variability). The authors note that Bayesian analysis allowed generation of population-
level infection estimates based on a limited amount of data.
Physiologically-based biokinetic models
In their assessment of anthrax dose-response models, Gutting et al. (2008) outline the
components of a hypothetical physiologically-based biokinetic (PBBK) model of infection and
response to aerosols of Bacillus anthracis. In the model, the fate and transport of B. anthracis
spores and vegetative cells is tracked in regions of the respiratory system, in macrophages, in the
August 2010 57
-------
U.S. Environmental Protection Agency
blood and in lymph nodes. As done by Brookmeyer et al. (2005) in their development of a
competing risks model for inhalation anthrax, Gutting et al. (2008) estimate model parameters
for use in their biokinetic model using physiological and microbiological data not collected in
quantal dose-response studies or epidemiological investigation. Details of the techniques used
for parameter estimation or of the model were not provided.
An earlier mechanistic dose-response model was proposed by Coleman and Marks (2000). In
that study the important events occurring in the course of a Salmonella illness were determined
to be survival of ingested bacteria to the target, colonization, engulfment, intracellular survival,
migration and multiplication, damage, and subsequent GI illness. The authors suggest use of
stochastic models for each of these processes and present an alternative formulation based on a
predator-prey framework. They note the existence of physiological and biological processes that
do not conform to the assumptions underlying the beta-Poisson or exponential dose-response
model and that can occur during infection and the progression to illness. Examples of these
processes are clumping of pathogens in the ingested dose, quorum sensing, and the possibility
that organisms do not exhibit independent action.
Accounting for variations between strains
Accounting for strain heterogeneity in pathogens is another cutting-edge area of QMRA dose-
response research that is difficult because dose-response data for all strains may not be available
and because all pathogen strains may not have been identified. The simplest means to predict
response for exposure to a variety of strains is to ignore inter-strain variations by using a dose-
response model based on pooled data, using a dose-response model based on the most virulent
strain among the strains considered, or to select a dose-response model for a "representative
strain." In some cases, models of pooled data may not exhibit goodness of fit (Coleman and
Marks, 1998).
A more systematic technique for addressing strain-to-strain variation is described by Seller et al.
(2007). In that study, and drawing from the previous work of Coleman and Marks (1998, 2000),
Gompertz-log dose-response models (alternatively called the Weibull dose-response model)
were fit to data for all strains of Salmonella for which data were available. As part of this effort,
one of the model parameters was assumed to be a random variable drawn from a uniform
distribution whose minimum and maximum values corresponded to the minimum and maximum
values found in fitting the model to individual (single strain) dose-response data. This technique
appears amenable to use of Bayesian techniques, since distributions other than a uniform
distribution may be assessed for describing variations related to pathogen strain differences.
IV.2.2 Accounting for Susceptible Populations
Susceptibility may be defined as "a capacity characterizable by a set of intrinsic and extrinsic
factors that modify the effect of a specific exposure upon the risk/severity of outcomes in an
individual population" (Balbus et al., 2000). Parkin (2004) advises QMRA practitioners to
define susceptibility clearly during the problem formulation phase. The author describes
intrinsic factors—including age (Parkin et al., 2003), gender, prior disease, immune status
(Makri et al., 2004; Parkin, 2004), pregnancy (Lanciers et al., 1999), and diabetes (Currie et al.,
2000)—as well as extrinsic factors (e.g., residence, income, co-exposures, access to health care)
that may result in differing susceptibility among subpopulations. Because many factors may
give rise to differences in susceptibility, this suggests that, where techniques and data are
available, they should be considered in all elements of a QMRA. In problem formulation,
August 2010 58
-------
U.S. Environmental Protection Agency
susceptibility should be defined, susceptible groups should be identified, and susceptible
populations should be included as a distinct stakeholder group (Parkin, 2004). Sampling and
surveying plans should include collection of quantitative data relevant to susceptible
populations — particularly because available data related to susceptible groups are, at present,
limited.
Parkin et al. (2003) conducted a feasibility assessment of incorporating analyses of populations
with different susceptibilities in a QMRA of the risk of GI illnesses resulting from recreational
use of surface waters. In that study, qualitative and quantitative data related to differences
between subpopulations were drawn from a literature search for studies describing the infection
of children by enteric viruses. The authors found several studies documenting an increased risk
for children (as compared with adults) for infection (including a study providing odds ratios by
age group), but did not provide definitive information on the relative severity of the duration of
illness associated with exposure to recreational waters. The results of this feasibility assessment
led the authors to recommend that researchers should provide greater detail in reports and studies
of outbreaks, distinguishing between outcomes for subpopulations, and providing more details
related to risk factors for subpopulations.
To our knowledge, analysis of susceptibility is not well developed in QMRAs conducted to date.
Obvious elements of QMRA analysis in which susceptibility could be addressed are dose-
response modeling and exposure assessment. The dose-response models most commonly used at
present (the exponential and beta-Poisson models) are derived based on an assumption that host
response at a given dose is binomially-distributed. This assumption may be relaxed by assuming
the exposed population may be divided into distinct groups, each of which has a uniform
susceptibility. Short! ey and Wilkins (1965) suggested use of such a model in dose-response and
proposed a model of the form
P(R\d) = \-xe-rid -(\-x)e-^d (25)
for a population comprised of two distinct subgroups. In equation 25, r\ and r^ are the
probability that a single organism can initiate infection in subgroup 1 and subgroup 2,
respectively; x is the fraction of the population belonging to subgroup 1. Short! ey and Wilkins
(1965) found that the model provided a good description of responses of a group of mice given
interperitoneal exposure to anthrax spores. Use of dose-response models such as that of equation
25 is, however, hampered at present by (1) the lack of estimates for dose-response models for
susceptible groups for most agents of concern, and (2) difficulty in estimating the proportion of
the population belonging to each subgroup.
An alternative to use of models such as that described by equation 25 is to assume the responses
for a population exposed to a given dose are not binomially distributed. In this regard, Haas et
al. (1999) suggest the beta-binomial distribution (equation 26) as an alternative.
,
a a
2 - - — 2 - (26)
y'! (»-;•)! B(* — ^
V 0 }
In equation 26, j is the number of individuals exhibiting a response in a population of n exposed
persons, ;ris the expected probability of response at the dose level, and $is a parameter
describing the dispersion in response. Haas and colleagues demonstrated use of the beta-
August 2010 59
-------
U.S. Environmental Protection Agency
binomial host response model for an hypothetical data set that could not be fit with binomially-
distributed host response models and noted that assessing the goodness of fit of models relying
on beta-binomial host response is problematic.
Two studies of Cryptosporidium parvum infection attempted to account for sensitive sub-
populations in dose-response models. Pouillot et al. (2004) used two dose-response models for
Cryptosporidium parvum infection. For immunocompetent sub-populations, an exponential
dose-response model with r = 5.26xlO"3, 95% CI (2.88xlO"3, 10.9xlO"3) was developed using
data from a human feeding study with a single isolate. For the immunocompromised sub-
population, data from experiments with immunodeficient mice were used to develop an
exponential dose-response model with parameter r = 0.354, 95% CI (0.221, 0.612). For the
immunocompetent sub-population, the probability of illness given infection was assumed beta-
distributed, with/? = beta (9,11) based on experimental data. For the immunocompromised sub-
population, a worst case scenario estimate that 100% of infected persons became ill was used.
Englehardt and Swartout (2004) used a bootstrap procedure for simulating dose-response data
intended to reflect responses of the population as a whole, rather than only those of healthy adult
volunteers. Although the authors justified and demonstrated the effectiveness of the procedure,
they found that for drinking water scenario studied, the inclusion data for sensitive and resistant
sub-populations in their analysis had little effect on overall predicted risks.
To summarize, avenues for quantitative analysis of susceptibility in dose-response include the
following:
• adjustment of ingestion/ventilation rates to account for variability among subpopulations;
• use of transport models of sufficient spatial resolution to estimate pathogen
concentrations in different neighborhoods/vicinities;
• use of non-human dose-response data to develop alternative dose-response models for
sensitive sub-populations;
• use of bootstrap techniques for development of data sets inclusive of sensitive and
resistant sub-population responses, and
• integration of spatially-resolved socio-economic data into exposure models.
IV.2.3 Secondary Transmission
Secondary transmission refers to infection spreading from one infected person to another person.
Secondary cases (often reported as a secondary attack rate) generally refer to cases that occur
following exposure to a primary case. In some cases, direct person-to-person transmission
cannot be separated from contamination of the immediate environment and subsequent
transmission to another person (e.g., toddlers sharing toys versus direct physical contact during
play). In most cases, it is appropriate that the definition of secondary transmission include
infections that result from propagation of the specific exposure of interest, but not encompass
distant transmissions (separated by time and/or space) that may be more appropriately considered
to result as a function of person-to-environment-to-person transmission.
Temporal and spatial limitations should be specifically noted in the definition of secondary
transmission for a given pathogen. Full discussion of the range of scenarios that qualify as
secondary transmission should be included where appropriate. It is important to note that the
above definition of secondary transmission is limited to avoid overlap with pathogen occurrence
in the environment (person-environment-person)—even though people are part of the
August 2010 60
-------
U.S. Environmental Protection Agency
environment. However, the potential for re-introduction of the pathogen into the media could
also be included within the definition of secondary transmission.
Secondary transmission models may be deterministic compartmental dynamic models, stochastic
compartmental models, individual-based (individual history) models, or spatially-structured
models. Descriptions of these models are provided by Rothermich and Murphy (2004).
Dynamic MRA models (illustrated in Figure 5) can characterize secondary cases that occur
among contacts following exposure to a primary case, whereas static MRA models usually
consider secondary transmission to be negligible or include it as a non-fluctuating multiplicative
factor (e.g., secondary cases equal primary cases multiplied by 0.1, if 10% secondary
transmission assumed). The problem formulation documentation should indicate if/how
secondary transmission is included in the assessment. If it is not included, justification for this
decision should be provided.
A recent study of secondary transmission for waterborne diseases by Joh et al. (2009) explored
the common assumption of a threshold number of organisms is needed to initiate an infection as
a component of a transmission model for diseases transmitted indirectly to humans (i.e., via
environmental reservoirs whose pathogen level is linked to human and animal infection levels).
The objective of employing a threshold assumption was an attempt to match the observed
dynamics of diseases such as cholera with model predictions. Although the use of threshold
models in QMRA dose-response is a subject of ongoing debate, the study by Joh and colleagues
is significant in its treatment of environmental reservoirs and ability to predict sporadic
outbreaks of disease. As noted in the preceding exposure assessment section (IV. 1) of this
report, modeling of the temporal variability of pathogen loading in environmental systems
remains an area of research within QMRA.
Post-Infection p
ir
S . E
Susceptible A A Exposed
(J! I
J -
Pathogen I
, Background 1
from , 1
„ . , , Pathogen 1
Reclaimed _ . 1
Concentration 1
Water I
a
C1 I .
r Carrier I r
Car
p5V01
. D I g
r" Diseased 1
a
2 1
rier 1
'
Figure 5. States and Flowpaths in a Dynamic Disease Transmission Model (SOURCE: Adapted
From Seller And Eisenberg, 2008)
August 2010
61
-------
U.S. Environmental Protection Agency
IV.3 RISK CHARACTERIZATION
Schaub (2004) states the major goals of risk characterization are to answer questions raised
during problem formulation, describe confidence in estimates, and describe limitations of the
QMRA process. The major elements that may be included in a risk characterization are listed in
Table 6. Risk estimates may be expressed as individual or population estimates (ILSI, 2000) as
per-event risks; as risk due to multiple exposures or as annual risk (Petterson et al., 2006); or as
risk of illness, such as daily adjusted life years (DALY). Choice of the metric used for
presenting risk data should be based on needs of risk managers and how estimates will answer
questions from the problem formulation stage.
IV.3.1 Sensitivity Analysis
Prior studies and reports (ILSI, 2000; Mokhtari and Frey, 2005b) list the following roles for
sensitivity analyses in model development and risk management:
• prioritization of potential critical control points (points, steps, or procedures at which a
control can be applied in risk management);
• identification of key sources of uncertainty and variability;
• identification of data gaps; and
• model refinement, verification, and validation.
There is often a temptation to perform sensitivity analysis as a final step in a QMRA and a
component of risk characterization. Delaying sensitivity analysis to this stage may result in
unnecessary modeling and data collection for model components to which the solution is not
particularly sensitive. Rather, the last two roles in the list above indicate that sensitivity analyses
should be performed iteratively during model development and data collection (ILSI, 2000).
Sensitivity analysis techniques are chosen based on the objectives of the sensitivity analysis.
These objectives might include (Mokhtari and Frey, 2005b) the following:
Table 6. Elements that May be Included in Risk Characterization (SOURCE: Adapted from ILSI,
2000)
Evaluate health consequences of exposure scenario
• Risk description (event)
• Risk estimation (magnitude, probability)
Characterize uncertainty/variability/confidence in estimates
Conduct sensitivity analysis
• Determine the most important variables and the information needs
Address items in the problem formulation
Evaluate various control measures and their effect on risk magnitude or profile
Conduct decision analysis
• Evaluate alternative risk management strategies
August 2010 62
-------
U.S. Environmental Protection Agency
• ranking the importance of model inputs (e.g., critical control points);
• identifying contributions of input values that contribute to high exposure and/or risk
scenarios;
• identifying and prioritizing key sources of uncertainty and variability;
• identifying critical limits; and
• evaluating the validity of the model.
Broadly, sensitivity analysis methods may be classified as mathematical, statistical, or graphical.
General descriptions and specific techniques for these three types of sensitivity analyses are
listed in Table 7. The choice of the QMRA model form depends on its use (screening, model
selection and validation, evaluation of variability and uncertainty in risk characterization, etc.).
Among complete QMRAs reviewed for this report (see Table 3), very few attempted sensitivity
analyses beyond simple nominal range sensitivity analysis (NRSA) and many studies did not
report sensitivity analysis results at all.
An important and cutting-edge facet of sensitivity analysis in QMRA is development of models
and sensitivity analysis techniques that allow independent analysis of model sensitivity to
uncertain factors and factors with variability. Several studies in which such analyses were
performed are highlighted in this section as examples of QMRAs employing state of the art
sensitivity analysis techniques.
Two studies (Mokhtari and Frey, 2005a; Pouillot et al., 2004) employed two-dimensional risk
analysis (nested analyses in which uncertain parameters are selected from distributions and used
as fixed inputs to stochastic risk models) to allow separate consideration of uncertainty and
variability. Pouillot et al. (2004) addressed sensitivity through estimation of risk using pre-
determined percentile estimates for model parameters. Mokhtari and Frey (2005a) addressed
sensitivity directly, using and comparing an ANOVA analysis and rank correlation analyses. In
their study, estimates and sensitivity rankings from two dimensional analyses (in which
uncertainty and variability are addressed separately) were compared with similar sensitivity
analyses for a one-dimensional model (in which uncertainty and variability were lumped). The
two-dimensional model with ANOVA sensitivity analysis yielded the results best suited for use
in risk management, while the correlation based methods were found to provide misleading
results when used in conjunction with the two-dimensional model. Regardless of the choice of
sensitivity model, the one dimensional model was insufficiently resolved to allow identification
of processes over which risk managers can exert control.
Petterson et al. (2007) conducted a model sensitivity study (comparison of outputs of models
developed under different assumptions) to evaluate the importance and best assumptions for
including method recovery in estimates of occurrence of protozoans in drinking source waters.
This study differs from those described above in that a specific uncertainty (method recovery)
was addressed and in that alternative models for describing Cryptosporidum oocyst occurrence
variability were developed. As with the above studies, however, uncertainty (related to method
recovery) was separated from variability, which was characterized via distributions whose
parameters and their uncertainty distributions were estimated using Bayesian techniques. These
analyses indicated that the risk estimates were highly dependent on assumptions made regarding
method recovery and that conservative assumptions about method recovery (low recovery rates)
should be employed in future risk analyses.
August 2010 63
-------
U.S. Environmental Protection Agency
Table 7. Sensitivity Analysis Methods and Techniques (SOURCE: Adapted from Frey and Patil,
2002; Frey et al., 2004)
Sensitivity
Analysis Type
General
Description
Techniques
Description
Mathematical
Quantification of
the variation in
model output with
the range of
variation of an
input. Typically
involves
systematic
variation of input
parameters,
evaluation of
model, and
assessment of
the influence of
the input
parameters on
the model output.
Nominal Range
Sensitivity
Analysis
(MRS A)
Variation of individual inputs over their range
while holding all other inputs at their nominal
values. Sensitivity is assessed via comparison
of model outputs corresponding to the range of
values. When model output is probability, the
difference in log odds ratio (ALOR) method may
be preferred.
Differential
sensitivity
analysis (DSA)
Variation of individual input in small range near
central tendency values. Sensitivity is assessed
based on variation in model output in the range
around the central tendency.
Automatic
Differentiation
(AD)
Difference in
log odds ratio
(ALOR)
Worst-case
determination
Break-point
analysis
This method is similar to DSA, except sensitivity
is assessed based on numerical partial
derivatives for the variation in model output with
changes in input parameters.
ALOR = in
Similar to NRSA, except sensitivity is assessed
via the ALOR, where
p(event with changes in input)
ot event with changes in input)
[ /?(event | w/out changes in input)
^ />(not event | w/out changs in input)
Similar to the ALOR approach, quantifies
sensitivity to a factor via a factor sensitivity ratio
given by
Nk (extreme)
FSk = log
Nk (average)
where k refers to the factor, N is the output (e.g.,
dose in the study conducted by Petterson et al.,
2006) and extreme and average refer to worst-
case and baseline values.
Search for values of inputs at which decision-
makers would be indifferent between two or
more risk management options.
August 2010
64
-------
U.S. Environmental Protection Agency
Sensitivity
Analysis Type
Statistical
Graphical
General
Description
Inputs to models
are assigned
probability
distributions and
sensitivity is
assessed via the
effect of variance
of the inputs on
model output.
Inputs may be
varied using
Monte Carlo
simulation, Latin
hypercube
sampling, or
other methods.
Techniques for
visualizing the
change in model
outputs with
changes in model
parameters.
Techniques
Regression
techniques
(sample
regression or
rank
regression)
Analysis of
variance
(ANOVA)
Sample
(Pearson)
correlation or
rank
(Spearman)
correlation
Classification
and regression
tree (CART)
Scatter plots
Conditional
sensitivity
analysis (CSA)
Description
Linear models (either based on known
relationships or analysis of scatter plots, etc.)
are developed for the dependence model output
on input variables. Regression is performed on
a sample of data generated from the model
(e.g., by Latin hypercube sampling, as
demonstrated by de Vos et al., 2006).
Sensitivity to input variables may be assessed
via comparison of standard errors of regression
coefficients or via application of stepwise
regression techniques.
ANOVA is used to determine whether there is a
statistical relationship between input variables
and model output; in contrast to regression
techniques, no functional form for the
relationship is assumed and data may be
qualitative or quantitative.
Sample correlation measures the strength of
linear association between input variables and
model outputs. Rank correlation is a measure
of the strength of the monotonic relationship
between two random variables.
Nonparametric technique that can select from
among a large number of variables those and
their interactions that are most important in
determining the whether an outcome variable
reaches a criterion value (Seller and Eisenberg,
2008). Output variables are divided into classes
(e.g., above and below a criterion) and a tree of
events leading to the output variable is
developed and analyzed.
Plots providing information on the relationship
between input variables and model outputs are
constructed.
Evaluating (usually graphically) the effect of
changes in a subset of model inputs while other
inputs are held at fixed values.
August 2010
65
-------
U.S. Environmental Protection Agency
V REFERENCES
Albert, I., Pouillot, R., and Denis, J.B. (2005) Stochastically modeling Listeria monocytogenes growth in farm tank
milk. Risk Analysis 25: 1171-1185.
An, Y.J., Yoon, C.G., Jung, K.W., and Ham, J.H. (2007) Estimating the microbial risk of E. coli in reclaimed
wastewater irrigation on paddy field. Environmental Monitoring and Assessment 129: 53-60.
Asano, T., Leong, L.Y.C., Rigby, M.G., and Sakaji, R.H. (1992) Evaluation of the California wastewater
reclamation criteria using enteric virus monitoring data. Water Science and Technology 26: 1513-1524.
Ashbolt, N.J., and Bruno, M. (2003) Application and refinement of the WHO risk framework for recreational waters
in Sydney Australia. Journal of Water and Health 1: 125-131.
Astrom, J., Petterson, S., Bergstedt, O., Pettersson, T.J., and Stenstrom, T.A. (2007) Evaluation of the microbial risk
reduction due to selective closure of the raw water intake before drinking water treatment. Journal of Water and
Health 5 Suppl 1: 81-97.
Atwill, E.R., Hou, L., Karle, B.M., Harter, T., Tate, K.W., and Dahlgren, R.A. (2002) Transport of Cryptosporidium
parvum oocysts through vegetated buffer strips and estimated filtration efficiency. Applied and Environmental
Microbiology 68: 5517-5527.
Balbus, J., Parkin, R., and Embrey, M. (2000) Susceptibility in microbial risk assessment: Definitions and research
needs. Environmental Health Perspectives 108: 901-905.
Bartrand, T.A., Weir, M.A., and Haas, C.N. (2008) Dose-Response models for inhalation of Bacillus anthracis
spores: Interspecies Comparisons. Risk Analysis 28: 1115-1124.
Bastos, R.K.X., Bevilacqua, P.O., Silva, C.A.B., and Silva, C.V. (2008) Wastewater irrigation of salad crops:
Further evidence for the evaluation of the WHO guidelines. Water Science and Technology 57: 1213-1219.
Benham, B.L., Baffaut, C., Zeckoski, R.W., Mankin, K.R., Pachepsky, Y.A., Sadeghi, A.M. et al. (2006) Modeling
bacteria fate and transport in watersheds to support TMDLs. Transactions oftheASABE 49: 987-1002.
Bollaerts, K.E., Messens, W., Delhalle, L., Aerts, M., Van der Stede, Y., Dewulf, J. et al. (2009) Development of a
quantitative microbial risk assessment for human salmonellosis through household consumption of fresh minced
pork meat in Belgium. Risk Analysis 29: 820-840.
Brookmeyer, R., Johnson, E., and Barry, S. (2005) Modelling the incubation period of anthrax. Statistics in
Medicine 24: 531-542.
Buchanan, R.L., Smith, J.L., and Long, W. (2000) Microbial risk assessment: Dose-response relations and risk
characterization. International Journal of Food Microbiology 58: 159-172.
Cassin, M.H., Lammerding, A.M., Todd, E.C.D., Ross, W., and McColl, R.S. (1998) Quantitative risk assessment
for Escherichia coli O157:H7 in ground beef hamburgers. International Journal of Food Microbiology 41: 21-44.
Charles, K., Roser, D., Ashbolt, N.J., Deere, D., and McGuinness, R. (2003) Buffer distances for on-site sewage
systems in Sydney's drinking water catchments. Water Science and Technology 47: 183-189.
Chen, L., Geys, H., Cawthraw, S., Havelaar, A., and Teunis, P. (2006) Dose response for infectivity of several
strains of Campy lob acterjejuni in chickens. Risk Analysis 26: 1613-1621.
August 2010 66
-------
U.S. Environmental Protection Agency
Clough, H.E., Clancy, D., O'Neill, P.D., and French, N.P. (2003) Bayesian methods for estimating pathogen
prevalence within groups of animals from faecal-pat sampling. Preventive Veterinary Medicine 58: 145-169.
Clough, H.E., Clancy, D., O'Neill, P.O., Robinson, S.E., and French, N.P. (2005) Quantifying uncertainty associated
with microbial count data: A Bayesian approach. Biometrics 61: 610-616.
Coffey, R., Cummins, E., Cormican, M, O Flaherty, V., and Kelly, S. (2007) Microbial exposure assessment of
waterborne patogens. Human and Ecological Risk Assessment 13: 1313-1351.
Coleman, M.E., and Marks, H.M. (1998) Topics in dose-response modeling. Journal of Food Protection 61: 1550-
1559.
Coleman, M.E., and Marks, H.M. (2000) Mechanistic modeling of salmonellosis. Quantitative Microbiology 2: 227-
247.
Crepet, A., Stahl, V., and Carlin, F. (2009) Development of a hierarchical Bayesian model to estimate the growth
parameters ofListeha monocytogenes in minimally processed fresh leafy salads. InternationalJournal of Food
Microbiology 131: 112-119.
Currie, B.J., Fisher, D.A., Howard, D.M., Burrow, J.N.C., Lo, D., Selva-nayagam, S. et al. (2000) Endemic
melioidosis in tropical northern Australia: A 10-year prospective study and review of the literature. Clinical
Infectious Diseases 31: 981-986.
de Vos, C.J., Saatkamp, H.W., Nielen, M., and Huirn, R.B.M. (2006) Sensitivity Analysis to evaluate the impact of
uncertain factors in a scenario tree model for classical swine fever introduction. Risk Analysis 26: 1311-1322.
Diallo, M.B., Anceno, A.J., Tawatsupa, B., Houpt, E.R., Wangsuphachart, V., and Shipin, O.V. (2008) Infection risk
assessment of diarrhea-related pathogens in a tropical canal network. The Science of the Total Environment 407:
223-232.
Dowd, S.E., Gerba, C.P., Pepper, I.L., and Pillai, S.D. (2000) Bioaerosol transport modeling and risk assessment in
relation to biosolid placement. Journal of Environmental Quality 29: 343-348.
Eisenberg, J.N.S., Seller, J.A., Scott, J., Eisenberg, D.M., and Colford, J.M.J. (2004) A dynamic model to assess
microbial health risks associated with beneficial uses of biosolids. Risk Analysis 24: 221-236.
Eisenberg, J.N.S., Moore, K., Seller, J.A., Eisenberg, D., and Colford, J.M.J. (2008) Microbial risk assessment
framework for exposure to amended sludge projects. Environmental Health Perspectives 116: 727-733.
Englehardt, J., and Swartout, J. (2004) Predictive population dose-response assessment for Cryptosporidium
parvum: Infection endpoint. Journal of Toxicology and Environmental Health Part A 67: 651-666.
Englehardt, J., Swartout, J., and Loewenstine, C. (2009) A new theoretical discrete growth distribution with
verification for microbial counts in water. Risk Analysis 29: 841-856.
Englehardt, J.D. (2004) Predictive Bayesian dose-response assessment for appraising absolute health risk from
available information. Human and Ecological Risk Assessment 10: 69-78.
Englehardt, J.D., and Swartout, J. (2006) Predictive Bayesian microbial dose-response assessment based on
suggested self-organization in primary illness response: Cryptosporidium parvum. Risk Analysis 26: 543-554.
Englehardt, J.D., and Swartout, J. (2008) Development and Evaluation of Novel Dose-Response Models for Use in
Microbial Risk Assessment. National Center for Environmental Assessment, Office of Research and Development
U.S. Environmental Protection Agency. Washington, DC.
August 2010 67
-------
U.S. Environmental Protection Agency
Ferguson, C., de Roda Husman, A.M., Altaville, N., Deere, D., and Ashbolt, N.J. (2003) Fate and transport of
surface water pathogens in watersheds. Critical Reviews in Environmental Science and Technology 33: 299-361.
Flemming, C., Seller, J.A., Simhon, A., and Seto, E.Y. (2009) Pathogen characterization of fresh and stored
biosolids and implications of a screening level microbial risk assessment. In Residuals andBiosolids 2009. Portland,
Oregon: Water Environment Federation.
Frey, H.C., and Patil, S.R. (2002) Identification and review of sensitivity analysis methods. Risk Analysis 22: 553-
578.
Frey, H.C., Mokhtari, A., and Zheng, J. (2004) Recommended Practice Regarding Selection, Application, and
Interpretation of Sensitivity Analysis Methods Applied to Food Safety Process Risk Models. Office of Risk
Assessment and Cost Benefit Analysis, U.S. Department of Agriculture, Washington, DC.
Furumoto, W.A., and Mickey, R. (1967) A mathematical model for the infectivity-dilution curve of tobacco mosaic
virus: Theoretical considerations. Virology 32: 216-223.
Gale, P., Pitchers, R., and Gray, P. (2002) The effect of drinking water treatment on the spatial heterogeneity of
micro-organisms: Implications for assessment of treatment efficiency and health risk. Water Research 36: 1640-
1648.
Gale, P. (2005) Land application of treated sewage sludge: Quantifying pathogen risks from consumption of crops.
Journal of Applied Microbiology 98: 380-396.
Gerba, C.P., Rose, J.B., Haas, C.N., and Crabtree, K.D. (1996) Waterborne rotavirus: A risk assessment. Water
Research 30: 2929-2940.
Giannoulis, N., Maipa, V., Konstantinou, I., Albanis, T., and Dimoliatis, I. (2005) Microbiological risk assessment
of Agios Georgios source supplies in Northwestern Greece based on faecal coliform determination and sanitary
inspection. Chemosphere 58: 1269-1276.
Gilks, W., Richardson, S., and Spiegelhalter, DJ. (eds) (1996) Markov Chain Monte Carlo in Practice. London,
UK: Chapman and Hall.
Gutting, B.W., Channel, S.R., Berger, A.E., Gearhart, J.M., Andrews, G.A., Sherwood, R.L., and Nichols, T.L.
(2008) Mathematically modeling inhalation anthrax. Microbe 3: 78-85.
Haas, C.N. (1983) Estimation of risk due to low doses of microorganisms: A comparison of alternative
methodologies. American Journal of Epidemiology 55: 573-582.
Haas, C.N., Rose, J.B., and Gerba, C.P. (eds.) (1999) Quantitative Microbial Risk AssessmentNew York, NY, John
Wiley and Sons.
Hamilton, A.J., Stagnitti, F., Premier, R., Boland, A.-M., and Hale, G. (2006) Quantitative microbial risk assessment
models for consumption of raw vegetables irrigated with reclaimed water. Applied and Environmental Microbiology
72: 3284-3290.
Holcomb, D.L., Smith, M.A., Ware, G.O., Hung, Y.-C., Bracket!, R.E., and Doyle, M.P. (1999) Comparison of six
dose-response models for use with food-borne pathogens. Risk Analysis 19: 1091-1100.
ILSI (International Life Sciences Institute) (1996) A conceptual framework to assess the risks of human disease
following exposure to pathogens. Risk Analysis 16: 841:848.
ILSI (2000) Revised Framework For Microbial Risk Assessment. ILSI Press, Washington, DC.
August 2010 68
-------
U.S. Environmental Protection Agency
Job, R.I., Wang, H., Weiss, H., and Weitz, J.S. (2009) Dynamics of indirectly transmitted diseases with
immunological threshold. Bulletin of Mathematical Biology 71: 845-862.
Jolis, D., Pitt, P., and Hirano, R. (1999) Risk assessment for Cryptosporidium parvum in reclaimed water. Water
Research 33: 3051-3055.
Julian, T.R., Canales, R.A., Leckie, J.O., and Boehm, A.B. (2009) A model of exposure to rotavirus from nondietary
ingestion iterated by simulated intermittent contacts. Risk Analysis 29: 617-632.
Kass, R.E., and Raftery, A.E. (1995) Bayes factors. Journal of the American Statistical Association 90: 773-795.
King County Department of Natural Resources (1999a) King County Combined Sewer Overflow Water Quality
Assessment for the Duwamish River and Elliott Bay. Volume 1: Overview and Interpretation. Wastewater
Treatment Division and Water and Land Resources Division Seattle, WA.
Lanciers, S., Despinasse, B., Mehta, D.I., and Blecker, U. (1999) Increased susceptibility to Helicobacterpylori
infection in pregnancy. Infectious Diseases in Obstetrics and Gynecology 7: 195-198.
Liu, L., Phanikumar, M.S., Molloy, S.L., Whitman, R.L., Shively, D.A., Nevers, M.D. et al. (2006) Modeling the
transport and inactivation of E. coli and enterococci in the near-shore region of Lake Michigan. Environmental
Science and Technology 40: 5022-5028.
Makri, A., Mondarres, R., and Parkin, R. (2004) Cryptosporidiosis susceptibility and risk: A case study. Risk
Analysis 24: 209-220.
Mara, D.D., Sleigh, P.A., Blumenthal, U.J., and Carr, R.M. (2007) Health risks in wastewater irrigation: Comparing
estimates from quantitative microbial risk analyses and epidemiological studies. Journal of Water and Health 5:39-
50.
Medema, G.J., Teunis, P.P.M., Gornik, V., Havelaar, A.H., andExner, M. (1995) Estimation of the Cryptosporidium
infection risk via drinking water. In Protozoan Parasites and Water. Cambridge, UK: Royal Society of Chemistry.
Messner, M.J., Chappell, C.L., and Okhuysen, P.C. (2001) Risk assessment for Cryptosporidium: A Hierarchical
bayesian analysis of human dose response data. Water Research 35: 3934-3940.
Mokhtari, A., and Frey, H.C. (2005a) Sensitivity analysis of a two-dimensional probabilistic risk assessment model
using analysis of variance. Risk Analysis 25: 1511-1529.
Mokhtari, A., and Frey, H.C. (2005b) Recommended practice regarding selection of sensitivity analysis methods
applied to microbial food safety process risk models. Human and Ecological Risk Assessment 11: 591-605.
Montville, R., and Schaffner, D. (2005) Monte Carlo simulation of pathogen behavior during the sprout production
process. Applied and Environmental Microbiology 71: 746-753.
Moon, H., Chen, J.J., Gaylor, D. W., and Kodell, R.L. (2004) A comparison of microbial dose-response models fitted
to human data. Regulatory Toxicology and Pharmacology 40: 177-184.
NRC (National Research Council (1983) Risk Assessment in the Federal Government, Managing the Process.
National Academy Press, Washington, DC.
NRC (2004). Indicators for Waterborne Pathogens. National Academies Press, Washington, DC.
Nauta, M., van der Fels-Klerx, I., and Havelaar, A. (2005) A poultry-processing model for quantitative
microbiological risk assessment. Risk Analysis 25: 85-98.
August 2010 69
-------
U.S. Environmental Protection Agency
Ottoson, I, and Stenstrom, T.A. (2003) Faecal contamination of greywater and associated microbial risks. Water
Research 37: 645-655.
Parkin, R., Seller, J.A., and Olivieri, A.W. (2003) Incorporating susceptible subpopulations in microbial risk
assessment: Pediatric exposures to enteroviruses in river water. Journal of Exposure Analysis and Environmental
Epidemiology 13: 161-168.
Parkin, R.T. (2004) Addressing susceptibility in microbial pathogen risk assessment. Human and Ecological Risk
Assessment 10: 135-149.
Petterson, S., Signer, R.S., Ashbolt, N.J., and Roser, DJ. (2006) QMRA Methodology. MicroRisk.
Petterson, S.R., Ashbolt, N.J., and Sharma, A. (200 la) Microbial risks from wastewater irrigation of salad crops: A
screening-level risk assessment. Water Environment Research 73: 667-672.
Petterson, S.R., and Ashbolt, N.J. (200 Ib) Viral risks associated with wastewater reuse: modeling virus persistence
on wastewater irrigated salad crops. Water Science and Technology 43: 23-26.
Petterson, S.R., Signer, R.S., and Ashbolt, N.J. (2007) Incorporating method recovery uncertainties in stochastic
estimates of raw water protozoan concentrations for QMRA. Journal of Water and Health 5: 51-65.
Petterson, S.R., Dumoutier, N., Loret, J.F., and Ashbolt, N.J. (2009) Quantitative Bayesian predictions of source
water concentration for QMRA from presence/absence data for E. coli O157:H7. Water Science and Technology 59:
2245-2252.
Pouillot, R., Albert, I., Cornu, M., and Denis, J.B. (2003) Estimation of uncertainty and variability in bacterial
growth using Bayesian inference: Application to Listeria monocytogenes. International Journal of Food
Microbiology 81: 87-104.
Pouillot, R., Beaudeau, P., Denis, J.B., and Derouin, F. (2004) A Quantitative Risk Assessment of Waterborne
Cryptosporidiosis in France Using Second-Order Monte Carlo Simulation. Risk Analysis 24: 1-17.
Ramachandran, G. (2001) Retrospective exposure assessment using Bayesian methods. Annals of Occupational
Hygiene 45: 651-667.
Ranta, J., Tuominen, P., and Maijala, R. (2005) Estimation of true Salmonella prevalence jointly in cattle herd and
animal populations using Bayesian hierarchical modeling. Risk Analysis 25: 23-37.
Regli, S., Rose, J.B., Haas, C.N., and Gerba, C.P. (1991) Modeling the risk from Giardia and viruses in drinking
water. Journal of the Amercian Water Works Association 83: 76-84.
Roberts, J.D., Sibergeld, E.K., and Graczyk, T. (2007) A probabilistic risk assessment of Cryptosporidium exposure
among Baltimore urban anglers. Journal of Toxicology and Environmental Health, Part A 70: 1568-1576.
Rose, J.B., Haas, C.N., and Regli, S. (1991) Risk assessment and control of waterborne giardiasis. American Journal
of Public Health 81: 709-713.
Rosen, B.H. (2000) Waterborne Pathogens in Agricultural Watersheds. Watershed Science Institute, National
Resouces Conservation Service, Ithaca, NY.
Rothermich, M., and Murphy, P. (2004) Developing Dynamic Infection Transmission Models for Microbial Risk
Assessment Applications. National Center for Environmental Assessment, Office of Research and Development,
U.S. Environmental Protection Agency, Cincinnati, OH.
August 2010 70
-------
U.S. Environmental Protection Agency
Ryu, H., and Abbaszadegan, M. (2008) Long-term study of Cryptosporidium and Giardia occurrence and
quantitative microbial risk assessment in surface waters of Arizona in the USA. Journal of Water and Health 6: 263-
273.
Schaub, S. (2004) A risk assessment framework for waterborne pathogens and requirements for producing a
complete protocol. Human and Ecological Risk Assessment 10: 151-158.
Schijven, J., Rijs Gerard, B.J., and de Roda Husman Ana, M. (2005) Quantitative risk assessment of FMD virus
transmission via water. Risk Analysis 25: 13-21.
Schijven, I, and de Roda Husman Ana, M. (2006) A survey of diving behavior and accidental ingestion among
dutch occupational and sport divers to assess the risk of infection with waterborne pathogenic microorganisms.
Environmental Health Perspectives 114: 712-717.
Schoen, M.E., and Ashbolt, N.J. (2010) Assessing pathogen risk to swimmers at non-sewage impacted recreational
beaches. Environmental Science and Technology 44: 2286-2291.
Seidu, R., Heistad, A., Amoah, P., Drechsel, P., Jenssen, P.O., and Stenstrom, T.A. (2008) Quantification of the
health risk associated with wastewater reuse in Accra, Ghana: a contribution toward local guidelines. Journal of
Water and Health 6: 461-471.
Seto, E.Y., Seller, J.A., and Colford, J.M., Jr. (2007) Strategies to reduce person-to-person transmission during
widespread Escherichia coli O157:H7 outbreak. Emerg Infect Dis 13: 860-866.
Shortley, G., and Wilkins, J.R. (1965) Independent-action and birth-death models in experimental microbiology.
Bacteriological Reviews 29: 102-141.
Shuval, H., Lampert, Y., and Fattal, B. (1997) Development of a risk assessment approach for evaluating wastewater
reuse standards for agriculture. Water Science and Technology 35: 15-20.
Signer, R.S., and Ashbolt, N.J. (2006) Pathogen monitoring offers questionable protection against drinking-water
risks: A QMRA (quantitative microbial risk analysis) approach to assess management strategies. Water Science and
Technology 54: 261-268.
Signer, R.S., Ashbolt, N.J., and Roser, D.J. (2007) Microbial risk implications of rainfall-induced runoff events
entering a reservoir used as a drinking water source. Journal of Water Supply: Research and Technology -AQUA
56:515-531.
Sinclair, R.G., Jones, E.L., and Gerba, C.P. (2009) Viruses in recreational water-borne disease outbreaks: A review.
Journal of Applied Microbiology 107: 1769-1780.
Smeets, P., vanDijk, J., Stanfield, G., Rietveld, L., and Medema, G. (2007) How can the UK statutory
Cryptosporidium monitoring be used for quantitative risk assessment of Cryptosporidium in drinking water? Journal
of Water and Health 5 Suppl 1: 107-118.
Smeets, P., Dullemont, Y., Van Gelder, P., Van Dijk, J., and Medema, G. (2008) Improved methods for modelling
drinking water treatment in quantitative microbial risk assessment; a case study of Campylobacter reduction by
filtration and ozonation. Journal of Water and Health 6: 301-314.
Seller, J.A., Olivieri, A., Crook, J., Parkin, R., Spear, R., Tchobanoglous, G., and Eisenberg, J.N.S. (2003) Risk-
based approach to evaluate the public health benefit of additional wastewater treatment. Environmental Science and
Technology 37: 1882-1891.
Seller, J.A., Olivieri, A.W., Eisenberg, J.N.S., Sakaji, R., and Danielson, R. (2004) Evaluation of Microbial Risk
Assessment Techniques and Applications. Water Environment Federation, Alexandria, VA.
August 2010 71
-------
U.S. Environmental Protection Agency
Seller, J.A., Eisenberg, J.N.S., DeGeorge, J.F., Cooper, R.C., Tchobanoglous, G., and Olivieri, A.W. (2006) A
public health evaluation of recreational water impairment. Journal of Water and Health 4: 1-19.
Seller, J.A., Seto, E.Y., and Olivieri, A.W. (2007) Application of Microbial Risk Assessment Techniques to
Estimate Risk Due to Exposure to Reclaimed Waters. WateReuse Foundation.
Seller, J.A., and Eisenberg, J.N.S. (2008) An evaluation of parsimony for microbial risk assessment models.
Environmetrics 19: 61-78.
Seller, J.A. (2009) Potential implications of person-to-person transmission of viral infection to US EPA's
Groundwater Rule. Journal of Water and Health 7: 208-223.
Seller, J.A., Bartrand, T., Ashbolt, N.J., Ravenscroft, J., and Wade, TJ. (2010a) Estimating the primary etiologic
agents in recreational freshwaters impacted by human sources of faecal contamination. Water Research:
doi:10.1016/j.watres.2010.1007.1064.
Seller, J.A., Embrey, M, Tuhela, L., Ichida, A., and Rosen, J. (2010b) Risk-based evaluation of Escherichia coli
monitoring data from undisinfected drinking water. Journal of Environmental Management 91: 2329-2335.
Seller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J., and Ashbolt, N.J. (2010c) Estimated human health risks
from exposure to recreational waters impacted by human and non-human sources of faecal contamination. Water
Research: doi: 10.1016/j.watres.2010.1006.1049.
Steyn, M., Jagals, P., and Genthe, B. (2004) Assessment of microbial infection risks posed by ingestion of water
during domestic water use and full-contact recreation in a mid-southern African region. Water Science and
Technology 50: 301-308.
Strachan, N.I, Dunn, G.M., and Ogden, ID. (2002) Quantitative risk assessment of human infection from
Escherichia coli O157 associated with recreational use of animal pasture. International Journal of Food
Microbiology 75: 39-51.
Teunis, P.P.M., Medema, M.G.J., Kruidenier, L., and Havelaar, M.A.H. (1997) Assessment of the risk of infection
by Cryptosporidium or Giardia in drinking water from a surface water source. Water Research 31: 1333-1346.
Teunis, P.F.M., Nagelkerke, N.J.D., and Haas, C.N. (1999) Dose reponse models for infectious gastroentritis. Risk
Analysis 19: 1251-1260.
Teunis, P.F.M., Chappell, C.L., and Okhuysen, P.C. (2002) Cryptosporidium dose response studies: Variation
between isolates. Risk Analysis 22: 175-183.
Teunis, P.F.M., Takumi, K., and Shinagawa, K. (2004) Dose response for infection by Escherichia coli O157:H7
from outbreak data. Risk Analysis 24: 401-407.
Teunis, P.F.M., van den Brandhof, W., Nauta, M.J., Wagenaar, J., van den Kerkhof, H., and Van Pelt, W. (2005) A
reconsideration of the Campylobacter dose-response relation. Epidemiology and Infection 133: 583-592.
Teunis, P.F.M., Ogden, I.D., and Strachan, N.J.C. (2008) Hierarchical dose response of E. coli O157:H7 from
human outbreaks incorporating heterogenity in exposure. Epidemiology and Infection 136: 761-770.
U.S. EPA (U.S. Environmental Protection Agency) (1986) Ambient Water Quality for Bacteria. EPA440/5-84-002
U.S. Environmental Protection Agency Washington, DC.
U.S. EPA (2003) Economic Analysis for the Long Term 2 Enhanced Surface Water Treatment Rule Proposal. U.S.
Environmental Protection Agency. Washington, DC.
August 2010 72
-------
U.S. Environmental Protection Agency
U.S. EPA (2006) National Primary Drinking Water Regulations: Long Term 2 Enhanced Surface Water Treatment
Rule (LT2ESWTR); Final Rule. U.S. Environmental Protection Agency. Washington, DC.
vanHeerden, J., Ehlers, M.M., Vivier, J.C., and Grabow, W.O.K. (2005) Risk assessment of adenoviruses detected
in treated drinking water and recreational water. Journal of Applied Microbiology 99: 926-933.
Walker, F.R. I, and Stedinger, J.R. (1999) Fate and transport model of Crypto sporidium. Journal of Environmental
Engineering 125: 325-333.
Westell, T., Schonning, C., Stenstrom, T.A., and Ashbolt, N.J. (2003) Integration of QMRA and HACCP for
management of pathogens in wastewater and sewage sludge treatment and reuse. In 4th International Symposium on
Wastewater Reclamation and Reuse, Nov 12-14. Mexico City: International Water Association.
Westell, T., Schonning, C., Stenstrom, T.A., and Ashbolt, N.J. (2004) QMRA (quantitative microbial risk
assessment) and HACCP (hazard analysis and critical control points) for management of pathogens in wastewater
and sewage sludge treatment and reuse. Water Science and Technology 50: 23-30.
WHO (1999) Principles and Guidelines for the Conduct of Microbiological Risk Assessment. CAC/GL-30. World
Health Organization Geneva, Switzerland.
Wong, M., Kumar, L., Jenkins, T.M., Xagoraraki, I., Phanikumar, M.S., and Rose, J.B. (2009) Evaluation of public
health risks at recreational beaches in Lake Michigan via detection of enteric viruses and a human-specific
bacteriological marker. Water Research 43: 1137-1149.
August 2010 73
-------
ANNEX 2
Development of a QMRA Model to Evaluate the Relative Impacts to
Human Health Risks from Animal-Impacted Recreational Waters
(June 2009)
For
Quantitative Microbial Risk Assessment to
Estimate Illness in Freshwater Impacted by
Agricultural Animal Sources of Fecal Contamination
U.S. Environmental Protection Agency
December 2010
-------
Development of a QMRA Model to
Evaluate the Relative Impacts to Human Health Risks from
Animal-Impacted Recreational Waters
Prepared by Clancy Environmental Consultants, Inc.
and Seller Environmental, LLC
EPA Contract Number EP-C-07-036
June 30, 2009
Office of Science and Technology
Office of Water
U.S. Environmental Protection Agency
Washington, DC 20460
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Mention of commercial products, trade names, or services in this document or in the references
and/or endnotes cited in this document does not convey, and should not be interpreted as
conveying official EPA approval, endorsement, or recommendation.
in
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Disclaimer iii
Tables vi
Figures vii
Executive Summary viii
Introduction 1
Overview of QMRA and context for animal impacted waters 3
Problem formulation 4
Analysis 5
Exposure assessment 5
Characterization of human health effects 7
Risk characterization 7
QMRA model development 8
Quantification of model inputs through literature review 13
Farm size, housing and manure handling 14
Cattle 14
Swine 15
Poultry 15
Model input: farm size and housing conditions 16
Quantity of fecal material 16
Cattle 16
Poultry 17
Swine 17
Model input: mass of fecal excretion 17
Characterization of fecal material 17
Salmonella 18
E. coli O157:H7 21
Campylobacter 22
Cryptosporidium 23
Dynamic prevalence 24
Model input: prevalence of infection 25
Excretion density 25
Salmonella 25
E. coli O157:H7 26
Campylobacter 26
Cryptosporidium 27
Excretion patterns and super-shedders 28
Model input: excretion density 28
IV
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Persistence in fecal deposits, manure and soils 29
Bacteria (E. coli O157:H7, Salmonella and Campylobacter) 29
Cryptosporidium 31
Model input: persistence of reference pathogens in feces and soil 32
Inactivation in the storage pond 33
Bacteria (E. coli O157:H7, Salmonella and Campylobacter) 33
Cryptosporidium 33
Model input: persistence of reference pathogens in the storage pond 34
Pathogen mobilization 34
Grazing land 35
Model input: probability of overland transport from grazing land 36
Manure amended soil 36
Model input: probability of overland transport from manure amended soil 36
Streambed sedimentation 37
Model input: proportion of mobilized load retained in stream sediments 37
Inactivation in stream sediments 38
Model input: persistence of reference pathogens in stream sediments 38
Resuspension 38
Model input: proportion of retained load mobilized during a re-suspension event 38
Consumption by recreational bathers 38
Model input: Consumption of water by recreational bathers 39
Human health effects: Dose response relationships 39
Risk Characterization 41
Model simulation results 41
Minimum storage time 45
Time between application and rainfall event 46
Sensitivity analysis 47
Sensitivity to prevalence 47
Sensitivity to excretion rate 48
Sensitivity to Inactivation rate 48
Sensitivity to storage pond catchment size 50
Conclusions 52
References 54
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Table 1. Summary of inputs to the QMRA model 13
Table 2. Summary of model inputs for farm size and housing conditions 16
Table 3. Summary of model input values for mass of fecal excretion 17
Table 4. Salmonella serotype prevalence in humans, broilers, cattle, and market hogs 19
Table 5. Overlap between Salmonella enterica serotype prevalence in humans and livestock 20
Table 6. Summary of model input values for infection prevalence (%) 25
Table 7. Summary of model input values for excretion density (Logi0. microorganisms/g) 29
Table 8. Exponential die-off constants for E. coli and fecal coliform 30
Table 9. Inactivation rates of Salmonella and E. coli 015 7 in fresh cow manure and slurry 31
Table 10. Inactivation rates of Salmonella and E. coli 0157 in fresh poultry manure and slurry 31
Table 11. Model inputs for inactivation rates of reference pathogens in fecal deposits and manure 33
Table 12. Inactivation rates and times for E. coli, Salmonella enterica and Campylobacter jejuni 33
Table 13. Model inputs for inactivation rates of reference pathogens in storage pond 34
Table 14. Model inputs for probability of passage from grazing land to surface waters 36
Table 15. Model inputs for probability of passage from manure amended land to surface waters 37
Table 16. Model input for the proportion of load transported to stream sediments 37
Table 17. Model inputs for inactivation rates of reference pathogens in stream sediments 38
Table 18. Model input for the proportion of retained load that is re-suspended 38
Table 19. Summary of dose response relationships used in QMRA model 40
VI
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Pi'Hires
JL 1 was, \ 9 I \t^ l3
Figure 1. EPA/ILSI QMRA framework applied to animal impacted waters 4
Figure 2. Conceptual model for exposure 6
Figure 3. Conceptual approach for the mathematical translation of exposure 9
Figure 4. Hypothetical historical rainfall data used to populate the history of the QMRA model 10
Figure 5. Influence diagram of QMRA model constructed in Analytica® 11
Figure 6. User Interface of the QMRA model for input of model variables in Analytica® 12
Figure 7. Egg layer farm size, 2007 (USDA NASS) 15
Figure 8. Broiler farm sizes, 2007 (USDANASS) 16
Figure 9. Salmonella enterica serotype prevalence in humans and livestock 20
Figure 10. Ingestion volumes during recreational activities 39
Figure 11. Downstream illness risks from the Cattle model for each reference pathogen 42
Figure 12. Downstream illness risks from the Swine model for each reference pathogen 43
Figure 13. Downstream illness risks from the Poultry model for each reference pathogen 44
Figure 14. Modeled pathogen loads to grazing land and manure storage 45
Figure 15. Impact of minimum storage time on load applied via manure application 46
Figure 16. Mobilized pathogen loads from cattle versus time between application and rainfall event 47
Figure 17. Sensitivity of predicted illness rates to the prevalence of infection in farm animals 47
Figure 18. Influence on the predicted illness rates of the number of supershedders among swine 48
Figure 19. Sensitivity of the risk model to cattle excretion rate for event 1 illness risks 48
Figure 20. Estimated Campylobacter load remaining on manure applied ground over time 49
Figure 21. Estimated Cryptosporidium load remaining on manure applied ground over time 50
Figure 22. Sensitivity of the Event 1 illness risks to storage pond catchment size for the cattle model 51
vn
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
An important goal of the Clean Water Act (CWA) is to protect and restore waters of the U.S. for
swimming. A key component in the CWA framework for protecting and restoring waters for
swimming is State adoption of Water Quality Standards (WQS) to protect swimmers from
illnesses associated with microbes in the water. In this regard, one of EPA's key roles is to
recommend recreational water quality criteria (under Section 304(a) of the CWA)) for adoption
by the States.
It has been over 20 years since EPA last issued recreational criteria. The science underpinning
this topic has advanced significantly during this time. EPA believes that new scientific and
technical advances need to be considered, if feasible, in the development of new or revised
304(a) criteria. To this end, EPA has been conducting research and assessing relevant scientific
and technical information to provide the scientific foundation for the development of new or
revised criteria which are scheduled to be published in 2012. The Agency would like to be able
to apply relationships from discrete epidemiology studies to the broad set of waters covered
under the Clean Water Act.
The Health and Ecological Criteria Division (HECD) within the Office of Water, in conjunction
with the Office of Research and Development (ORD) requested the development of a QMRA
model that has the ability to mathematically encapsulate relevant scenarios for freshwater
recreational waters impacted by agricultural animals (e.g., animal feeding sites and/or areas
where animal manures are applied). Although specific sites of interest have been identified,
EPA's goal in this respect was to identify and incorporate into a transparent and defensible
model, the salient aspects of hazard fate and transport, representative exposure scenarios, and
risk characterization for several important zoonotic pathogens.
This work was initiated by HECD in several phases.
o The first phase of this work (Work assignments B-01 and subsequently 1-08 Task 2)
focused on model and scenario development for cattle-impacted waters. That work was
developed as a key facilitating component to an EPA QMRA workshop in November
2008 at the ORD offices in Cincinnati, OH.
o The second phase of work (WA 1-08 Task 2 Amendment 3) requested an extension of the
initial cattle scenario to one (or more) other agricultural animals (e.g., swine, poultry,
etc.) and
o
Conduct of a sensitivity analysis of the various (cattle, swine, and poultry) models and
model parameters for the three exposure scenarios to determine which data and model
components are the most crucial with respect to the conduct of QMRA for animal
Vlll
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
impacted waters. The identified exposure scenarios included rainfall induced runoff
transporting pathogens directly to a stream via overland flow, rainfall induced storage
pond overflow, and re-suspension of pathogens stored within stream sediments.
This report summarizes the literature review, model development, QMRA simulations, and
sensitivity analysis efforts that were undertaken to achieve the goals described above. During
the course of this work (conducted for WA 1-08 Task 2 and subsequently Work Assignment 1-
08 Task 2 Amendment 3) the following was accomplished:
o Conduct of a literature review, development, implementation and parameterization of a
QMRA model in Analytica software, and development and parameterization of three
exposure scenarios for cattle-impacted waters.
o Planning of and participation in an EPA QMRA workshop in November 2008 at the
ORD offices in Cincinnati, OH where the cattle model was presented and demonstrated.
o Extension of the initial cattle model to two other agricultural animals (swine and poultry)
for the three exposure scenarios.
o Conduct of a literature review of swine and poultry manure data to parameterize the
newly extended QMRA model for Salmonella, Camplyobacter jejuni, E. coli O157:H7,
and Cryptosporidiumparvum.
o Conduct of a sensitivity analysis of the various models (cattle, swine, and poultry) and
associated model parameters for the three exposure scenarios to identify which data and
model components are the most crucial with respect to the conduct of QMRA for animal
impacted waters.
The salient findings from this work includes the following:
o Onsite collection and storage of fecal material is an important barrier for preventing
pathogen mobilization downstream. Operations that collect and store fecal material for
land application may present short term peaks of pathogen risk, immediately following
application. These peaks are estimated to be roughly equivalent to the ongoing risk
associated with open grazing operations.
o When manure is to be stored and then land applied, the storage barrier is only effective
for pathogen removal when a minimum storage time is ensured.
o Managing land application to avoid periods of high rainfall will reduce risk.
o Prevalence of infection in any given herd is likely to be constantly changing, and the
within-herd temporal variation of can be substantial.
IX
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
o Understanding the prevalence of human infectious pathogenic strains could be a critical
component for not overestimating the risk associated with animal-impacted waters.
o Quantification of pathogen excretion density is a significant source of uncertainty in the
overall model. In particular, the existence of super-shedders appears to drive the overall
pathogen load. This aspect requires further research, particularly if identification and
containment of super-shedders is possible.
o Environmental inactivation rates of pathogens are highly uncertain. Therefore, ensuring
pathogen reduction via uncontrolled environmental processes is not feasible unless
extended residence times can be guaranteed. Given the current state of knowledge,
Cryptosporidium oocysts should be assumed to persist for long time periods unless site
specific data indicate otherwise.
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
An important goal of the Clean Water Act is to protect and restore waters for swimming. A key
component in the CWA framework for protecting and restoring waters for swimming is State
adoption of Water Quality Standards (WQS) to protect swimmers from illnesses associated with
microbes in the water. One of EPA's key roles is to recommend recreational water quality
criteria (under Section 304(a) of the CWA)) for adoption by the States. These EPA
recommended criteria have been historically based on fecal matter in the water; in the 1960's the
Federal government recommended a certain level of fecal coliforms as the recreational criteria.
In 1986, EPA recommended certain levels of enterococci and E. coli as its new recreational
criteria. These organisms do not cause human illness themselves (that is, they are not human
pathogens); rather, they are merely indicators of fecal contamination and therefore indicators of
the potential presence of human pathogenic organisms.
It has been over 20 years since EPA last issued recreational criteria. The science related to this
topic has advanced significantly during this time. EPA believes that new scientific and technical
advances need to be considered, if feasible, in the development of new or revised 304(a) criteria.
To this end, EPA has been conducting research and assessing relevant scientific and technical
information to provide the scientific foundation for the development of new or revised criteria.
The enactment of the BEACH Act provided EPA with an opportunity to conduct new studies and
provided additional impetus to issue new or revised criteria for coastal recreational waters
(specifically, for Great Lakes and coastal marine waters) to replace or amend the 1986 EPA
recommended criteria. EPA believes that the new or revised criteria must be scientifically sound,
implementable for broad CWA purposes, and provide for improved public health protection over
the 1986 criteria.
As one component of the work introduced above, the Agency would like to extrapolate
relationships from discrete epidemiology studies to the broader set of waters covered under the
Clean Water Act. Additionally, once new or revised recreational AWQC are published, the
Agency would like to provide States guidance on using Quantitative Microbial Risk Assessment
(QMRA) in developing WQS specific to local conditions. The Health and Ecological Criteria
Division (HECD) within the Office of Water, in conjunction with the Office of Research and
Development (ORD) has requested the development of a QMRA model that has the ability to
encapsulate relevant scenarios for freshwater recreational waters impacted by agricultural
animals (e.g., animal feeding sites and/or areas where animal manures are applied). Although the
actual sites of interest have yet to be determined, EPA's goal was to identify and incorporate into
a transparent and defensible model, the salient aspects of hazard fate and transport,
representative exposure scenarios, and risk characterization for the zoonotic pathogens E. coli
O157:H7 and Cryptosporidium parvum. The exposure scenarios of investigated were rainfall
1
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
induced runoff transporting pathogens directly to a stream via overland flow, rainfall induced
storage pond overflow, and re-suspension of pathogens stored within stream sediments.
This work was initiated by HECD in several phases. The first phase of this work (Work
assignments B-01 and subsequently 1-08 Task 2) focused on model and scenario development
for cattle-impacted waters. That work was developed as a key facilitating component to an EPA
QMRA workshop in November 2008 at the ORD offices in Cincinnati, OH. Based on the
success of that work, EPA requested additional model development (Work Assignment 1-08
Task 2 Amendment 3) as follows:
o Extension of the initial cattle scenario to one (or more) other agricultural animals (e.g.,
swine, poultry, etc.). The extent to which the initial cattle model could be extended to
other animals was to be evaluated and a justification was to be provided for extension of
the initial scenario versus starting from scratch for the development of the additional
scenario(s). This component of the work was to include a modest literature review of
swine and poultry manure data (based on available resources) and was not intended to be
an exhaustive review of all available literature.
o
Conduct of a sensitivity analysis of the various (cattle, swine, and poultry) models and
model parameters for the three exposure scenarios to determine which data and model
components are the most crucial with respect to the conduct of QMRA for animal
impacted waters.
With respect to the context of the sensitivity analysis introduced above, this effort was to be
inclusive of the relationship of pathogens and indicators, including: 1) the fate and behavior of
pathogens from these animal sources; 2) the fate and behavior of the indicators in these
situations; and, 3) the potential divergence or uncoupling of these two groups in the environment
due to the various environmental sources and sinks (i.e., regrowth, establishment, resuspension).
This report summarizes the literature review, model development, QMRA simulations, and
sensitivity analysis efforts that were undertaken to achieve the goals described above.
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Overview of QMRA and context for animal impacted waters
Quantitative microbial risk assessment (QMRA) (also known as MRA and pathogen risk
assessment) is a process that evaluates the likelihood of adverse human health effects that can
occur following exposure to pathogenic microorganisms or to a medium in which pathogens
occur (1996; ILSI, 2000) To the extent possible, the QMRA process includes evaluation and
consideration of quantitative information, however, qualitative information is also employed as
appropriate (WHO, 1999). QMRA methodologies have been applied to evaluate and manage
pathogen risks for a range of scenarios including from food (Bollaerts et al., 2009; Nauta et al.,
2005; Seto et al., 2007),drinking water(Astrom et al., 2007; Medema et al., 1995; Regli et al.,
1991; Seller, 2009), recycled water (Asano et al., 1992; Westrell et al., 2003) and recreational
waters (Ashbolt and Bruno, 2003; Seller et al., 2003; Seller et al., 2006).
The principles, processes and methods for carrying out risk assessments for chemical agents
were formalized in 1983 by the National Research Council (NRC) resulting in a four step
process or framework (National Research Council, 1983). The steps outlined by the NRC include
hazard identification, dose-response assessment, exposure assessment, and risk characterization.
Many of the early MRAs employed the NRC conceptual framework to provide a structure from
which the assessments could be conducted (Haas, 1983; Regli et al., 1991; Rose et al., 1991). As
the field of microbial risk assessment developed, it became clear that there were some
complexities associated with modelling the infectious diseases that are unique to pathogens, such
as person-to-person transmission of infection and immunity. Therefore, the conceptual
framework for chemicals may not always be appropriate for the assessment of risk of human
infection following exposure to pathogens (ILSI, 1996).
To address this concern, the EPA Office of Water developed a conceptual framework to assess
the risks of human infection associated with pathogenic microorganisms. The EPA Office of
Water is also developing a framework for microbial risk assessment to support human health
protection for water-based media. The EPA/ILSI framework emphasizes the iterative nature of
the risk assessment process (Figure 1), and allows wide latitude for planning and conducting risk
assessments in diverse situations. This framework consists of three principal components:
problem formulation, analysis, and risk characterization. The analysis phase is further subdivided
into the characterization of exposure and human health effects.
The problem formulation stage is used to identify: (1) the purpose of the risk assessment, (2) the
critical issues to be addressed, and (3) how the results might be used to protect public health.
Once identified, initial descriptions of the exposure and potential health effects are described and
then a conceptual model is developed. This conceptual model is used as a starting point for the
analysis phase of the risk assessment and later as an interactive tool along with components
developed in the analysis phase to initiate the risk characterization.
3
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
In the analysis stage, information about both the exposure and the health effects is compiled and
summarized. This compilation of quantitative and qualitative data, expert opinion, and other
information results in exposure and host/pathogen profiles that explicitly identify the data to be
integrated into the risk characterization and the associated assumptions and uncertainties. These
two elements, while separate, must also be interactive to ensure that the results are compatible.
The final stage, risk characterization, results in a statement of the likelihood, types, and/or
magnitude of effects likely to be observed in the exposed population under the expected
exposure scenario, including all of the inherent assumptions and uncertainties. Often, the risk
characterization phase includes data integration through parameterization of a mathematical
model, numerical simulation and interpretation.
Salient aspects of each of these Framework components are briefly summarized below to provide
context for the animal impacted water modelling effort undertaken in this work.
MICROBIAL RISK ASSESSMENT
A
> r
Problem Formulation
m, (model development) jL
W Analysis ^
Characterization of exposure Characterization of health effects
| Characterise source material | | Dose-response assessment |
1
Characterise pathogen mobilisation events Probability of illness 1
Interception and inactivation ^. ^
I
i
Water consumption by recreational bathersl
i A
^ T
1 >!
Y Risk Characterization *
4
>
k
f
Figure 1. EPA/ILSI QMRA framework applied to animal impacted waters
Problem formulation
The objective of the model development for animal impacted waters was to facilitate the
evaluation of the impact of fecal pathogens from animal feeding sites and/or areas where animal
manure is applied on gastrointestinal illness among recreational bathers. The range of potential
animal-impacted water sites was constrained in this work to consider dairy cattle, poultry and
swine operations. Within this context, the identified hazards were enteric pathogens with the
potential for zoonotic transfer from these animals to humans.
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Given the scope and goals described above, a wide range of pathogens could be considered.
However it is current practice within QMRA to select reference pathogens to represent the broad
behaviour of the three microbial groups (bacteria, parasites, and viruses). Due to the significance
of bacterial pathogens in zoonotic transfer, and the potentially different behaviour of different
types of bacteria, three bacteria were selected as reference pathogens for this work:
Campylobacter jejuni, Salmonella and E. coll O157:H7; along with one parasitic protozoa:
Cryptosporidiumparvum. While enteric viruses are thought to be primarily species specific (and
hence not transmitted via zoonoses), recent evidence has demonstrated the cross-species
transmission of HEV (Tei et a/., 2003), and concern exists regarding the potential for large scale
waterborne transmission to humans (Myint and Gibbons, 2008). The evidence for this method of
transmission is still building, and little environmental data is available to describe the prevalence,
excretion, fate and transport of HEV. Thus, HEV was not included in this QMRA model
development.
Analysis
Exposure assessment
As illustrated in Figure 1, exposure assessment for the QMRA involved modelling the
occurrence of each reference pathogen in the environment from their source to the location of
potential human exposure via recreational waters. The conceptual model for identifying and
evaluating the hazard pathways is provided in Figure 2 which illustrates a generic animal feeding
operation, where animals may either be housed within a shed, in open pens or contained in open
paddocks.
Fecal material was assumed to be shed in one of these three environments. The fate of animal
manure on animal feeding operations is a critical determinant of potential downstream risks.
Manure handling varies significantly between operations depending on the type of facility
(small, medium or large operation; breeding, feeding, laying, dairy, or finishing facility),
geographic region, and intended use of manure (land application, storage, reuse for cattle feed,
etc.). In general, manure is collected and stored as solids (solids comprise at least 20% of mass)
or slurries (mixtures of feces, urine and potentially cleaning water, rainwater, and small
quantities of feed) and the fate of the fecal material depends upon the local manure management
operations. Fecal material shed on open pastures was assumed to accumulate naturally and be
available for overland transport. Fecal matter from sheds and open pens is usually collected and
stored and then potentially applied later to land within the same site or a nearby site.
For pathogens to present a potential health risk to recreational bathers, they must be mobilized
from the farm. The sole mechanism for this mobilization was assumed to be rainfall induced
runoff which could mobilize fecal material deposited on pasture land, fecal material 'spilled'
during manure handling, land applied manure and potentially overflow of onsite storage facilities
(pond) during large events.
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Of those pathogens mobilized, some would be intercepted and some inactivated, preventing their
complete transport to the recreational waters in an infectious state. In addition, some pathogens
may be bound to soils (either on the soil surface or, when injected or tilled into the soil, in the
vadose zone) and their transport to receiving waters may be retarded or prevented. In receiving
waters, pathogens may be associated with sediments, and remain stored within the streambed to
be resuspended during a future event.
Animal (
farming I
operation
Potential pathways for faecal material to be mobilised to local waterways
Open G raZJ ng - Direct deposition of faecal materia
Some por lion ol
overland flow may
not be intercepted
by the pond and may
flow directly to stream
Pens - animals not fully housed, Faecal material
on ground and may be mobilised by rainfall
recreationa
| J*r\swimmingsite
Figure 2. Conceptual model for exposure
The exposure assessment encompasses three separate pathogen mobilization events:
1. Rainfall induced runoff transporting pathogens directly to a stream via overland
flow
2. Rainfall induced storage pond overflow
3. Re-suspension of pathogens associated with stream sediments
Human exposure to pathogens was assumed to occur through ingestion of ambient waters via
recreational activities in waterbodies impacted by the events identified above.
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Characterization of human health effects
The probability of infection and subsequent illness, given the modelled exposure was estimated
based on dose response relationships published in the literature. Potentially important aspects of
disease transmission such as secondary transmission (i.e. person to person or person to
environment to person) and immunity were not accounted for in this work.
The primary objective of constructing the QMRA model was to undertake the process of
describing the system from source to exposure, review available data, to describe model inputs,
and evaluate the sensitivity of the model and scenarios to the different variables. Risk
characterization was conducted employing a static stochastic model.
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
QMRA
An important aspect of the model development was to translate the conceptual model for
exposure (Figure 2) into a mathematical approach for quantifying the occurrence, inactivation
and transport of pathogens from their source to exposure. As indicated previously, three
mobilization events were considered in the model development. Those events are described
below.
• Event 1: Direct overland transport to stream.
• Event 2: Overflow of the onsite storage pond. Pond overflow was assumed to occur
during or as a result of a rainfall event, and therefore this event also includes the impact
of direct overland transport (Event 1+ overflow of the pond).
• Event 3: Re-suspension of pathogens stored within stream sediments. An objective of the
modelling was to investigate the magnitude of sediment re-suspension events on the
downstream illness risk. To simulate this event, it was necessary to model the
accumulation of pathogens within stream sediments prior to the hypothetical
mobilisation. Of those pathogens mobilized to the stream during a runoff mobilisation
event (Event 1) a portion of those were assumed to settle and remain stored within
sediments.
An overview of the conceptual approach for the mathematical translation is provided in Figure 3.
The pathogen load excreted each day by farm animals was calculated as the number of infected
animals, multiplied by the concentration of pathogens in the feces of the infected animal and the
daily fecal mass. This approach assumes that only infected animals excrete pathogens, and that
infected and non-infected animals excrete the same mass of feces each day.
Fecal material was assumed to either remain on the ground or to be transferred to a storage
facility. Fecal material deposited directly on the ground was assumed to inactivate over time
following first order kinetics. A logic basis was selected for all rate equations within the model.
The total load of pathogens contained within deposited fecal material on grazing land was
calculated as the sum of all previous daily loads accounting for inactivation over time since
deposition. Fecal material collected for storage was assumed to inactivate over time following
first order kinetics. Fecal material was assumed to be added to the facility on a daily basis, with
the total load at time of application calculated as the sum of all previous daily additions, while
accounting for inactivation over time in the storage facility. Conditions within the storage facility
and within fecal deposits on land would be very different and different inactivation rates would
be expected.
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
SOURCE MAT
BARRIERS
ERIAL
Faecal Material pathogen load =
Number of infected anima s X.
Concentration of pathogens in faeces (microorganisms.g-1 ) X
Mass of faecal mater idlpei day (g-1)
Shed
1
' &
Storage facility
Sequencing batch reactor
Inactivation during storage
±
Land application
Inactivat on in soil
X^o
/ Proba
transpa
MOBILISATION to
LOCAL STREAM
Pen
/ \
'\
\
Open grazing
» i
f
Daily direct deposition
nactivation on land
rerland transport ^\ S"^^* ***\
»liiy of pathogen bc.ng \ f Overland transport X
rted from and to storage ) f P^b-hty of pathogen bemg \
j u( slieam J \ transported from land to storage 1
r -^^VENT 1 EVENT 1 ^S-"v~_ _^^
EVENT 1
Storage pond
Sequencing batch reactor
Light(k,)and dark (kj (natlivation
with the portion of trip population experipndng
light Inactivation = a
C,=oC010fc'l+ C0(1-o)10*:it
> f EVENT 2
EVENT 1
MOBILISATION to Recreational Wate
I EVENT 1
Deposition of solid associated pathogens
.EVENTS
A Re-suspension
rs
Figure 3. Conceptual approach for the mathematical translation of exposure
Overland transport of pathogens from grazing land and manure application sites is dependent
upon many factors including the intensity of rainfall, vegetation cover, slope and distance to
waterway. The QMRA model accounts for overland transport rates by assuming that every
pathogen contained on land had a certain probability (p) of being transported to waterway during
an overland flow event. This approach considers land transport as a single barrier in the same
conceptual way as has been presented for drinking water treatment processes (Teunis et al.,
1999a). There is the potential to describe the probability of passage (p) as a variable fraction
accounting for differences in where pathogens may be deposited, vegetation cover and slope. A
distribution for/? could be selected based on the magnitude of the rainfall event.
Application of manure to land was assumed to occur as a single event, with the total load of
pathogens in the storage facility transported to land.
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
The storage pond was assumed to intercept overland flow during small to moderate events.
Pathogens contained within the pond were assumed to inactivate following first order kinetics,
however based on reported information about the importance of sunlight in the inactivation of
pathogens in water (Sinton et a/., 2007), light and dark inactivation phases were modelled
separately. The total inactivation was assumed to be the sum of two subpopulations 1) those
exposed to sunlight and 2) those protected from sunlight. The size of the total population (0-1)
exposed to sunlight was described by the parameter a.
The QMRA model describes the impact of rainfall events on pathogen mobilization, and not only
by a single event, but also upon the accumulation of pathogens within the storage pond and
stream sediments based on a rainfall history. It was therefore necessary to include a dynamic
component to the model which allowed this historical impact to be quantified (Figure 4).
10
30
Dynamic index
Figure 4. Hypothetical historical rainfall data used to populate the history of the QMRA model
The presented approach was then used to create the QMRA model using Analytica® Profession
Version 4.1 (Lumina Decision Systems, Inc.) which allows for the construction of influence
diagrams that can then be used for Monte Carlo simulation. The Analytica influence diagram
and the user-interface of the model are illustrated in Figure 5 and Figure 6, respectively.
10
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
QMRA Model: Animal impacted waters
( E
Interface Characterising faecal material
Shed Pen Open grazing/direct deposition
Proportion of time animals in shed
J
Number of
f Pathogen load to "~1
^_ shed J
Faecal storaj ie facility
[Pathogen load to storage |
1 "="> )
t «.
Remaining pathogen load from
Total remaining pathogen load
nfected
Proportion of time animals in Pen
J\
Number of excretion
animals in the pen
Proportion of timeanimals 'free range' / Animal /
animals on field
mass to pen") f Path°ge" ° ") fpaecal mass to field"]
s^ // Pathogen a :cumulation on grazing land
Inactivation rate in faecal ^ '
storage facility
Total faecal mass in storage
fina product
Land Application of treated faecal material
Cum
Over!
Direc
mulative oad on field *- * °
grazing land
and tr msport : Rajnfa|| dafa
t dep< sition to overland flow
Mobilised Probab ility
grazing land -*" transport
Proportion
of flow from Rainfall data
* grazing land (mm]
pond
Total Time since * Inactivation ^^-"""^ /^--A""""" / I
manure application load applied * remaining « app|jed , ^^^^""^ I \ /
applied (kgl (days] _^_ manure ^^^^"^ j \ / /
Overland tra
^3_ --"""
~^^_ 1 ^^^
sportrl-and application to pverland flow
Overiand ^ Probab lily of 4^^"
applicati
mobilised
Proportion
of flow from
-, to storage
\^ \
^~~~-~-^^
Sediments
Portion of oad lost to
Inactivation _^ -^" ^
ediments . Proportion of load ^^_--~~"~^
sediments mobilised during event 2 ^ ^__^-^
/ ^-^^^
Load to
Load
Proportion of load
nt, nr1 mobilised during event 3
Mass of direct faecal input to recreational wa
Mass of d rt
faecal nput
recreation
waters - (k
Storage pond:
inTto^T'ond Patho
from applied stora
/ \/
\ X i
Load remaining
=n load in/.
gepond from^ng^nd
; load moblilised rate in storage pond
Fast phase nactivation
L^^/ \ ^^^^^~ 7 ra6 m storage pond
"\ _^-~~~~~ / __/-^^" \ ___/-^-^^
^^^-^^J/
^ — - Direct downstrea
~~~~~~~ — L__^ I
i-^-^* /
im load Load from
N
Event ' ^^^1^^
ters - (kg] | IO|
ct Pathogen
al recreational
g] waters
Recreational water co
Dilution J Cone (event
Volume V
Event 2 load
•i +
Event 2
Load
cent rat ion
^cC~D < c-""^ 1
volume (ml) ^ Cone by event J-^ Exposure J
Dose-response ""~--^^
^ functions J
- J Pill|inf J
-^ 1000 ^ Analysis J
j^ exposures J
Figure 5. Influence diagram of QMRA model constructed in Analytica®
11
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
User Interface - Animal impacted recreational waters
Length of dynamic simulation (days) 60
Total number of Animals Edit Table
Rainfall data (mm)
Prevalence of infection
Edit Table
Faecal mass per animal per day (g) Edit Table
Number of excretion events per day 10
Proportion of time animals in shed I Edit Tabl I
Proportion of time animals in Pen I Edit Tabl I
Proportion of time animals 'free range' I Edit Tabl I
Storage facility for collected faecal material
Inactivation rate in faecal storage facility Edit Table
Land application of treated faecal material
Total amount of manure applied (kg) 1000
Time since manure application (days) 0
Inactivation rate in land applied manure I Edit Table
Investigate sensitivity to point prevalence? No ^
Use Dynamic prevalence? No ^
Excretion Density
Investigate sensitivity to pathogen density? No ^
Accumulation of faecal material on grazing land
Pathogen inactivatio rate on grazing land I Edit Table I
Storage pond
Proportion of flow from grazing land to storage pond 0.2
Proportion of flow from manure site to storage pond O.I
Proportion of storage pond load moblilised during event 2 11
Slow phase Inactivation rate in ... Edit Table
Deposition in stream sediments and resuspension
Portion of load lost to sediments during Event I I Edit Tabl I
Proportion of load mobilised during event 2 0
Proportion of load mobilised during event 3 11
Inactivation rate in sediments Edit Table
Recreational Water Body
Mixing volume of recreational water body (L) 10000000
Illnesses per 1000 exposures
Figure 6. User Interface of the QMRA model for input of model variables in Analytica®
12
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
s literature review
W.J V I I 1. \../ (i~* *« II II Iv5*/ I, %.*- V IB* I- \!/ 1. \s/ T I \»/ T ?
The scientific literature was reviewed with the objective of informing the selection of
quantitative values for each of the model inputs (Table 1) for cattle, swine, and poultry impacted
waters. A broad range of studies of relevance to the QMRA model were identified, however in
many cases the studies had goals that were unrelated or tangentially related to our focus of
waterborne human health risks. Special care was taken to consider the context of original
datasets and the representativeness of those datasets to the input variables of interest in the
QMRA model.
Table 1. Summary of inputs to the QMRA model
Input Comments
Number of animals The expected number of animals housed at a feeding operation for
cattle, swine and poultry.
Proportion of time in shed, pen and grazing Account for different housing conditions between each animal type
Fecal excretion rate (kg.day"1) The mass of feces excreted per day for each animal
Prevalence of infection Needed to predict the number of animals infected with each
reference pathogen on a farm for each animal type.
Density of pathogens in feces of infected For each animal type, for each reference pathogen
animals (microorganisms.g"1)
Inactivation rate in fecal deposits The rate is expected to be different for each reference pathogen and
may vary for each animal due to size of deposits and composition.
Inactivation rate in stored manure The rate is expected to be different for each reference pathogen and
may vary for each animal.
Inactivation rate in land applied manure The rate is expected to be different for each reference pathogen and
may vary for each animal.
Overland transport from grazing land Probability of passage from grazing land to surface waters expected
to be different for each pathogen and animal.
Overland transport from manure amended Probability of passage from manure amended soil to surface waters
soil expected to be lower than from grazing land.
Inactivation rate in storage pond Two population model to account for a protected sub-population
Proportion of mobilized load transported to Proportion will vary for each pathogen, and vary between runoff
sediments events.
Inactivation rate in stream sediments Two population model to account for a protected sub-population.
Proportion of retained load mobilized Re-suspension event will depend on individual pathogen and
during a re-suspension event stream flow hydrodynamics
Dilution factor What is the magnitude of the mixing zone of the load mobilized
from agricultural land.
Exposure volume Volume of water consumed by recreational bathers
13
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Few of the model inputs were expected to be represented by a single value, but rather were likely
to be variable, changing with time and space both on an individual farm and between farms.
Variable model inputs can be described by a probability density function (PDF) which allows the
variable to take one of a range of values each with defined probability of occurrence.
Analytica® has the ability to describe any of the model inputs as either a point value or a
distribution. The choice of whether to use a single value or a distribution was driven by two
factors:
1. The nature of the input variable itself, and whether incorporating variability was
deemed informative to the overall objective of the modelling exercise; and
2. Whether the available data was suitable for describing variability.
Uncertainty in quantifying model inputs from sparse data is difficult to separate from variability.
However it is conceptually important to distinguish between what is considered to be actual
variation in the model inputs and what is poor precision due to a lack of knowledge. In this
study, uncertainty was investigated through the sensitivity analysis (described later). A
summary of the data that were obtained from the literature review is provided below for the
model variables summarized in Table 1.
Cattle
About 2/3 of cattle and calves are raised on pasture and ranges, where manure collection is
generally not practiced and an estimated 85% of beef cattle in the United States are fed in
feedlots (Eghball and Power, 1998), primarily in the Central and Southern Great Plains. Thus, in
their lifespan most beef cattle likely deposit manure both in pastures and in feedlots. Manure
from grazing cattle is deposited directly on fields and pathogens and indicators in the manure are
available for liberation from the manure matrix and down-slope transport. As with pigs raised on
feedlots, manure from cattle on feedlots may be scraped and composted as a solid, or may be
stored along with urine and other liquids as a slurry. Solids and slurries are nearly always land-
applied on the farm where they are generated or on a nearby site.
Dairy cattle practices are more varied, due in part to the wider distribution of farms throughout
the United States, in part to variation in practices with season, and due to the trend toward
organic or other non-conventional farming practices. The two most common manure
management strategies practiced on dairy farms are direct deposition on pasture lands and indoor
collection and storage as solids or slurries (Hubbard and Lowrance, 1998). In both cases,
manure is land-applied in the vicinity of its production. Increasing practice of organic dairy
farming has produced trends toward smaller herd sizes (average of 82 compared to 153 milk
cows per farm for organic and conventional farms, respectively) and increasing use of pasture for
feeding cows (63% for organic dairy farms compared to 18% for conventional dairy farms)
(Greene et al., 2009). Manure handling practices at dairy operations may vary with season, with
14
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
farmers housing cows indoors during the winter and collecting manure as liquid and/or solids
during that period and more frequent use of pasture during other seasons.
Swine
In the US, more than 60% of swine are produced on large1 farms with more than 5,000 head. On
these large farms the majority of manure is collected as slurries and land applied; with sites for
land application nearly always in the vicinity of the production facility. Small and medium sized
operations are more likely to include pasture production (estimated at approximately 5%) or
open feedlot production (estimated at approximately 35%). Open feedlots are scraped every two
weeks, with manure stored in piles and subsequently land-applied via spreader.
Poultry
Poultry farms operate in the US for production of both eggs and broilers, with the size of
operations illustrated in Figure 7 and Figure 8, respectively. Broiler operations produce only
solid wastes (mixture of bedding and manure), which is almost entirely applied to land, on an
annual basis following stacked storage. Laying hen operations produce both solid and liquid
manure, with the liquid waste applied to land 2-3 times per year. Some solid poultry waste is
used as livestock feed or in the production of biofuel, however these uses are considered to be
small in comparison to land application. An increasing trend in the production of organic and
free range chickens and eggs is expected, and will most likely have the greatest influence on
small to medium operations.
100000000
= 10000000
1000000
100000
1 to 49 50 to 99 100 to 400 to 3200 to 10,000 20,000 50,000100,000
399 3199 9999 to to to or more
19,999 49,999 99,999
Farm Size
Figure 7. Egg layer farm size, 2007 (USDA NASS)
1 The USEPA considers swine operations with at least 2500 swine weighing 55 pounds or more or 10,000 swine
each weighing less than 55 pounds to be Large Concentrated Animal feeding Operations (CAFOs). Operations with
750 - 2500 swine each weighing 55 pounds or more or 3000 swine, each weighing less than 55 pounds AND a man
made ditch or pipe carrying manure or wastewater or animals in contact with surface water in the confinement are
considered medium CAFOs
15
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
10000
Ito49 50 to 99 100 to 399 400 to 3199 3200 to 9999 10,000 to
19,999
Farm Size
Figure 8. Broiler farm sizes, 2007 (USDA NASS)
Model input: farm size and housing conditions
While there is variability in the size of animal feeding operations across the United States, in the
context of the QMRA model the farm size was considered to be a fixed point input. The model is
constructed to investigate the downstream impacts of a single farming operation. The number of
animals housed can be changed in the user interface of the model to compare for example the
impact of a large or small farm, but it is not considered to be a random variable for the Monte
Carlo simulation. A typical farm size was selected for each of the animal types (Table 2).
A similar approach was adopted for selecting the proportion of time in shed, pen and open
grazing. While the review of practices across the United States indicated that there is variability
between farms, any given farm was assumed to operate under one set of conditions. The most
typical conditions were selected as the model input and are summarized in Table 2.
Table 2. Summary of model inputs for farm size and housing conditions
Farm size
Proportion of time: shed
pen
open grazing
Cattle
500
0.1 (milking)
0
0.9
Poultry
20,000
1
0
0
Swine
5,000
1
0
0
Quantity of fecal material
Representative estimates of animal excretion rates from the published literature are summarized
below.
Cattle
The average wet weight of feces excreted by cattle per day has been reported as 23kg (Dorner et
al., 2004). The number of excretion events per day has been estimated at 12 (Larsen et al., 1994;
16
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Thelin and Gifford, 1983) and 8.1 (Stromberg, 1997). Individual cow pats have been measured
at 920g (Davies-Colley et al., 2004) and modelled assuming 1 kg per cow pat, and two cow pats
per excretion event (Ferguson et al., 2007).
Poultry
Manure production estimates for poultry have been reported as 0.11 kg.day"1 for laying hens and
0.04 kg.d"1 for broilers (Dorner et al., 2004); and 0.088 and 0.12 kg.day"1 for layers (reviewed by
Ferguson et al. (2009)).
Swine
Manure production estimates for swine included 15 kg.day"1 for sows and gilts for breeding, 2
kg.day"1 for nursing and weaner pigs, and 5 kg.day"1 for growing and finisher pigs (Dorner et al.
(2004)); and 5.1 and 6.2 kg.day"1 for pigs in the US and Australia respectively (reviewed by
Ferguson et al. (2009)).
Model input: mass of fecal excretion
Manure production rates vary with the health of the animal, type of feed, animal age, and
potentially other factors. No data were identified that quantified the difference between
excretion rates of infected and non-infected animals, and hence only average or typical rates
were available for the model input. This is expected to be representative for long term endemic
excretion by asymptomatic animals, but may not be representative of ill animals; a small portion
of the total infected, but likely to be responsible for the bulk of the pathogen load (Chase-
Topping etal, 2008).
While there is variability between animals in the amount of fecal material excreted each day, the
literature values were interpreted as reported averages. A point estimate was considered suitable
for describing this average since adequate data were not available to describe the between animal
variability. Conservative values, within the range of the reported averages, were selected and are
summarised in Table 3.
Table 3. Summary of model input values for mass of fecal excretion
Cattle Poultry Swine
Average fecal mass per 24 0.11 5
animal per day (kg/day)
Point prevalence studies reviewed from the literature typically involved sampling a large number
of animals, often across a number of farms, and reporting the overall proportion of those animals
that tested positive for the target organism as the 'prevalence'. In the context of the current study
it is worth considering the averaging effect of the adopted approaches and that localized effects
can be lost. In addition, analytical methods for identifying pathogens from fecal samples are
imperfect, often with poor rates of recovery. Therefore, fecal sampling may actually
underestimate the true prevalence (Van Hoorebeke et al., 2009).
17
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Salmonella
Serotypes of Salmonella enterica differ widely in the doses required to initiate infection in
humans and the severity of associated illness (Coleman and Marks, 2000; Coleman et al., 2004;
Seller et al., 2007). Thus, salmonellosis risk is related to both the prevalence of Salmonella
among animal and the serotypes prevalent in the animals. Though not conclusive, the relative
risk posed by Salmonella enterica serotypes in animals may be inferred from comparison of the
prevalent serotypes in different animal hosts and humans. The Centers for Disease Control
(CDC, 2006) have identified the serotypes from human Salmonella enterica isolates for the
period 1996-2006 and the United States Department of Agriculture Food Safety and Inspection
Service (USDA FSIS, 2009) has identified the serotypes for Salmonella isolates identified in
broilers, market hogs, steer and heifers, and cows and bulls for the period 1998 - 2007. The 24
most common serotypes of non-typhoid Salmonella from human isolates are summarized in
Table 4 and Figure 9. Prevalence of serotypes for broilers, steers/heifers, cows/bulls, and market
hogs are also presented in Table 4 and Figure 9. Inspection of these data indicate that the
prevalence of serotypes within a given host changes significantly from year to year, though for
humans, the serotypes Typhimurium (including the Copenhagen variant) and Enteriditis were
consistently among the top three serotypes isolated. It is also noteworthy that there is overlap
between the most common human and animal Salmonella serotypes (Figure 9), with all animals
exhibiting relatively high prevalence of human-infecting serotypes Typhimurium, Newport,
Saint-Paul, Infantis, Anatum, and Mbandaka and all of these hosts but swine are subject to
infection with Montevideo. Over the period examined here, the percent of samples positive for
broilers was highest (12.2%), followed by market hogs (3.3%), cows and bulls (1.3%) and steer
and heifers (0.3%).
The overlap between serotypes prevalent in humans and those present in livestock can be used to
develop a lower bound on the potential loading of human-infectious Salmonella from livestock.
The 24 serotypes most commonly isolated from humans account for 79.5% of all positive
isolates. The prevalence of the 24 most common human serotypes among livestock ranges from
52.5% to 59.8% of isolates (Table 5). Caution should be used in interpreting serotype prevalence
data in risk estimation because assuming the overlap in serotype prevalence between human and
livestock is an indication of relative risk disregards the role of exposure in risk. Sometimes lack
of overlap between prevalent human and animal serotypes indicates no serious human health
effects for those serotypes, but other times lack of overlap may indicate lack of exposure.
Because Salmonella enterica infections are sporadic (Callaway et al., 2008) and serotype
prevalence may change dramatically from year to year (USDA FSIS, 2009), there exists the
possibility for an animal-associated outbreak (among humans) for a relatively uncommon or an
unknown serotype.
18
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Table 4. Salmonella serotype prevalence in humans, broilers, cattle, and market hogs
Serotype
Typhimurium (w. Copenhagen)
Enteriditis
Newport
Heidelberg
Javiana
Montevideo
Muenchen
Oranienburg
Saintpaul
Infantis
Thompson
Braenderup
Agona
I, 4, [5], 12:i-
Hadar
Mississippi
Typhi
Paratyphi B var L(+) tartrate (+)
Poona
Berta
Stanley
Anatum
Bareilly
Mbandaka
Other or not identified
Human
21.64
17.80
8.43
5.24
3.46
2.42
2.04
1.74
1.62
1.54
1.51
1.49
1.49
1.20
1.12
1.04
1.02
1.02
0.79
0.64
0.58
0.57
0.52
0.52
20.54
Broiler
10.64
6.76
17.44
2.42
0.94
1.18
2.21
1.24
0.34
0.75
56.08
Steer/
Heifer
2.30
5.75
3.45
5.75
1.15
2.30
4.60
2.30
2.30
2.30
2.30
4.60
1.15
59.77
Cow/Bull
9.82
0.73
13.45
1.09
8.36
1.09
0.36
3.64
0.36
5.82
0.36
2.55
52.36
Market
Hog
13.96
3.15
0.30
4.52
7.43
1.37
1.31
9.45
0.36
58.16
19
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
1C
S 20
o.
n
VI
•B 15 -
V)
£
M
| 10
ra
w
™ 5
I
n -
JL
^
111
bJ[ l
t
• Human
D Broiler
nstcer/Hcifcr
D Cow/Bull
D Market Hog
1 J
i i i i i i
1 n m\h
^^^/^^^^^
^
^
Figure 9. Salmonella enterica serotype prevalence in humans and livestock
Table 5. Overlap between Salmonella enterica serotype prevalence in humans and livestock
Host
Human
Broilers
Steers/Heifers
Cows/Bulls
Market hogs
Percent Positive Isolates with Serotype in
the 24 Most Prevalent Serotypes among
Human Isolates
79.5
43.9
40.2
47.6
41.8
Prior research suggested a method to account for the variation in dose-response characteristics
across Salmonella enteric serotypes (Seller et al., 2007). In these QMRA analyses, a Gompertz
dose-response model was assumed with one of the dose-response model parameters treated as a
random variable. The prevalence data above suggests a potential modification to this approach,
in which the selection of the dose-response model parameter is dependent on the likelihood of
observing a serotype known to cause infection in humans frequently. Serotypes that appear in
animals, but seldom in humans, may be assumed less able to initiate infection than those
appearing regularly in both hosts.
20
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Dairy cattle: Point prevalence studies have been undertaken by identifying the frequency of
Salmonella positive fecal samples from cows. Reported point prevalence rates for Salmonella
spp. in cow feces include 3.4% ((Rodriguez et al., 2006); total samples n=2,496), 5.4%, (Wells et
al., 2001), 7.3% (Blau et al., 2005), 9.96% ((Callaway et al., 2005); total samples n=960), 31%
(Huston etal, 2002) and 0-93% ((Edrington etal., 2004); total samples n=720).
The frequency of at least one positive fecal sample for each dairy (dairy prevalence) has also
been reported. Those results include 21.1% (Wells et al, 2001), 30.9% (Blau et al., 2005), and
56% (Callaway et al., 2005). Pangloli et al. (2008) identified seasonal variability in the
prevalence of Salmonella in both animal and environmental samples on dairy farms. With some
exceptions, prevalence increased with the seasonal temperature. Callaway et al. (2006)
investigated the incidence of Salmonella in fecal samples collected from cattle feedlot pens.
Within pen prevalence varied from 0 to 6.67% with the overall average prevalence at 3.75%.
Poultry: A review of studies prior to 1998 undertaken by Byrd (1998) estimated the prevalence
of Salmonella among chicks to be 5-9%. The horizontal transmission between chicks co-housed
in hatcheries was observed to be highly efficient. Garber et al. (2003) identified the Salmonella
enterica serotype enteritidis in 7.1% of layer houses (n = 200, all U.S. facilities). Factors
associated with higher incidence of S. enterica were large flock size (>100 000), young age,
housing conditions (floor-reared as opposed to cage reared), and lack of cleaning and
disinfection of feeders and hoppers between flocks. Hayes et al. (2000) detected Salmonella spp.
in poultry litter and drag swabs from 48/71 (55.8%) broiler and roaster house facilities.
Hutchison et al. (2004) detected Salmonella spp in 17.9% of fecal samples (n=67) from
commercial farms.
Swine: Dorr et al. (2009) reported Salmonella spp. prevalence of 10.4% for 5 farms based on
the detection of Salmonella spp in fecal samples. The highest and lowest observed prevalences
(on individual farms) were 21.7% and 5%. Salmonella prevalence increased significantly with
age. Hutchison et al. (2004) detected Salmonella in 7.9% of fecal samples (n=126) from
commercial farms. Sanchez et al. (2007) undertook a systematic review and meta-analysis to
identify study-level variables that could explain the variation in apparent Salmonella spp.
prevalence estimates in swine. The median farm-level and animal-level prevalences were 59%
and 17%, respectively.
E.coli O157:H7
Cattle: E. coli O157:H7 is commonly excreted by dairy cattle in the United States. According to
Edrington et al. (2004) "Research has suggested that almost all dairy farms will have cattle
testing positive for E.coli O157:H7 if screened often enough((Hancock et al., 1997))". Elder and
co-workers (2000) reviewed the literature regarding prevalence rates of E. coli O157:H7 or E.
coli O157:nonmotile shedding in cattle and noted a widespread prevalence of E. coli O157:H7
with 63% of feedlots observing at least one positive sample. Overall cattle prevalence though
appeared to be low at only 1.8% of fecal samples (Animal and Plant Inspection Service, 1995).
21
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Method improvements employing enrichment followed by immunomagnetic separation (IMS) in
the late 1990s had an important impact on analytical recovery rates, and led to an increase in
reported prevalence rates. This improvement in method is demonstrated by Donkersgoed et al.
(1999) who collected 1,247 fecal samples from 293 feedlots over 12 consecutive months; and
analysed each sample by both direct culture and enrichment with IMS. Direct culture alone
identified E. coli O157:H7 in 2.6% (33/1427) of samples while enrichment with IMS identified
7.5% (93/1247) as positive, n that study prevalence of E. coli O157:H7 was highest in summer
(19.7%) followed by spring (4.9%), autumn (4.7%) and winter (0.7%).
Point prevalence of E. coli O157:H7 in fecal samples analysed with enrichment and IMS have
been reported as 23% (Smith et al., 2001), 0-35% (Edrington et al., 2004); n=720), 5.9%
(Barkocy-Gallagher et al., 2003), 13% (LeJeune et al., 2004); n=4790), and 13% (Fegan et al.,
2004). Callaway et al. (2006) investigated the incidence of E. coli O157:H7 in fecal samples
collected from cattle feedlot pens. Within pen prevalence varied from 5 to 20% with the overall
average prevalence at 11.6%.
Poultry: E. coli O157:H7 appears to have a very low prevalence in poultry. Of those studies
identified in this review no detects were observed (Chapman et al. (1997) n=1000; Hutchison et
al. (2004) n=67 fresh manure).
Swine: E. coli O157:H7 has been isolated from swine in ranges from 11.9 % (n = 126) by
Hutchison et al. (2004) to 0.4% (n=1000) by Chapman et al. (1997). In the Chapman et al.
study, E. coli O157 was isolated by an immunomagnetic separation technique and culture of
magnetic beads on cefixime teelurite sorbitol MacConkey agar. Isolates that gave positive results
were confirmed as E. coli by biochemical tests and as serogroup O157 or serotype O157:H7.
Campylobacter
Like Salmonella, the ability of Campylobacter isolates to infect humans varies among species
and isolates and the prevalence of strains differs in animals and humans. Though other species
may play smaller roles in human health effects, Campylobacter jejuni and Campylobacter coli
are the most important human-disease-causing species of Campylobacter commensal in livestock
(Wesley et al., 2000). Among all livestock hosts, the prevalence of Campylobacter and likely
the relative prevalence of different Campylobacter species varies between farms and regions,
with age of animal, with season, and probability with other factors (El-Shibiny et al., 2005;
Weijtens et al., 1999; Wesley et al., 2000) and estimating prevalence of individual species is
difficult given available data. Furthermore, the dose-response characteristics C. jejuni appear to
differ among fresh cultures and laboratory cultures (Chen et al., 2006). Given this lack of
species-specific prevalence data and the absence of a general dose-response model for human
infection with C. coli, a reasonable approach is to assume that Campylobacter spp. dose-
response parameter for campylobacteriosis (illness endpoint) fall within the range of related
illness rates provided by Teunis et al. (2005).
22
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Cattle: Molecular typing studies have shown that cattle and sheep are colonized with and excrete
strains of C. jejuni which are capable of causing disease in the local community (Stanley and
Jones, 2003); and outbreaks of disease from Campylobacter spp. have been associated with
contaminated water from agricultural runoff (Hrudey and Hrudey, 2004; Vogt et al., 1982).
Poultry: Avian species are thought to be the natural hosts of thermophilic Campylobacters
because they have a core temperature of 42°C, which is the optimum growth temperature of C
.jejuni. According to the review undertaken by Stanley and Jones (2003), many studies have
shown that once Campylobacter has been introduced into a shed, birds rapidly excrete high
numbers in their feces and the organism spreads rapidly so that 100% of birds may be colonized
within a few days.
Point prevalence of Campylobacter in poultry studies have been reviewed by Dorner et al.
(2004)who reported a range of prevalence from 3.1% to 100% across 14 studies.
Swine: Campylobacter coli appears to be the predominant species found in swine. In a study of
two Dutch piggeries (multiplier farms as opposed to fattening farms) with similar facilities and
handling practices (Weijtens et al., 1997), Campylobacter spp. were detected in feces of 9/10
sows (5 from each of two different farms) prior to delivery and in 10/10 sows post delivery. At 1
week after delivery Campylobacter was found in 29/60 piglets. At 8 weeks after delivery it was
found in 56/60 piglets. Dorner et al. (2004) reviewed the point prevalence of Campylobacter in
Sows and Gilts 45.9% positive (w=315); and 79.7% positive (w=59).
Cryptosporidium
Cattle: Atwill et al., (2003) reviewed reported prevalence and excretion rates for
Crytosporidium with an aim to estimate environmental loading rates and noted that "There is a
wide range of reported prevalences of fecal shedding of C. parvum for adult beef and dairy
cattle. Numerous investigators have reported mean prevalences of fecal shedding from -20 to
-70% in groups of clinically healthy adult cattle (Lorenzo et al., 1993; Quilez et al., 1996; Scott
et al., 1995), yet several large cross-sectional epidemiologic surveys have observed prevalences
of only 2% or less in asymptomatic adult cattle populations (Atwill et al., 1999a; Hoar et al.,
2001; Wade etal., 2000)."
Sources of this variation were identified as:
• Different investigators using diagnostic assays of differing sensitivity and specificity
(Faubert and Litvinsky, 2000; Payer et al., 2000; Pereira et al., 1999)
• Different populations (beef versus dairy), age distributions and management practices.
For example, they found in two different studies that calving duration for beef herds was
associated with a three- to sixfold difference in the proportion of cattle shedding C.
parvum (Atwill et al., 1999b; Hoar et al., 2001), making interstudy comparisons of the
shedding prevalence for beef cattle potentially confounded if not adjusted for calving
duration.
23
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Atwill and co-workers (2003) reported the overall apparent prevalence of adult beef cattle testing
positive for C. parvum was 7.1% (17 of 240), with 8.3 and 5.8% of cattle shedding oocysts
during the pre- and postcalving periods, respectively. Starkey and co-workers (2005)
investigated the prevalence of Crytosporidium oocysts in the feces of dairy cows from a New
York watershed. Of the 9914 fecal samples collected, 747 (7.5%) were found to contain C.
parvum.
Payer et al. (2006) investigated the prevalence of Cryptosporidium species in 571 1-2-year-old
heifers on 14 dairy farms in seven states on the East Coast of the United States and reported
11.9% of 571 heifers were infected with Cryptosporidium., 0.7% with Cryptosporidium parvum.
Of 68 PCR-positive specimens 1, 4, 10, 24, and 29 calves were infected with Cryptosporidium
suis, Cryptosporidium parvum, Cryptosporidium deer-like genotype, Cryptosporidium bovis, and
Cryptosporidium andersoni, respectively. These findings demonstrate a lower prevalence of
infection in 1-2-year-old dairy cattle than in younger cattle as well as an increase in the diversity
of species present. Consequently, the risk of humans acquiring infection with C. parvum from
exposure to feces from yearling and older cattle appears much lower than from exposure to pre-
weaned calves.
Poultry: Ferguson et al. (2009) reviewed two studies that reported prevalence rates for
Cryptosporidium in poultry: Ley et al. (1988) reported prevalence rates in the US at 6% (n=17)
and 27% (n=33) in broiler and layer chickens; Medema et al. (2001) reported prevalence of 27%
(n=16) in The Netherlands.
Swine: Ferguson et al. (2009) reviewed 6 studies that investigated the prevalence of
Cryptosporidium in swine, rates ranged from 0 - 100%.
Dynamic prevalence
As indicated by the summaries presented above, the reported rate of prevalence was highly
variable for all reference pathogens. Several detailed prevalence investigations have been
conducted which evaluated the dynamics of herd prevalence. Chapagain et al. (2008) modelled
the dynamics of Salmonella Cerro infection in a US dairy herd. The data collected for that study
included tracking an outbreak of Salmonella Cerro in a single milking herd. In March, 2004
only one (n=102) cow was shedding Salmonella at this time and the isolates from this sample
were identified as S. Enteric Typhimurium (var. Copenhagen). Six months later, 43.5% of the
herd was reported to be infected with Salmonella enterica Cerro. Within 6 weeks, the fecal
prevalence rate of S. Cerro dramatically increased to 75% and persisted at or near this level for
~6 months. By August, 2005 the number of cows shedding Salmonella had dropped to 9% and
the results of a subsequent sampling in September indicated that 29% of the cows were shedding
this organism.
Van Kessel et al. (2007a) described the course of a Salmonella outbreak and subsequent endemic
infection on a dairy farm in Pennsylvania. Shedding of Salmonella Cerro was reported to be
24
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
greater than 60% of the cows in the herd throughout the fall of 2004 and the spring of 2005. With
this high level of persistent shedding, most of the environmental samples tested positive for
Salmonella during this time frame. The animal level prevalence of Salmonella fell dramatically
between March (67%) and August 2005 (8%) before rising to 88% by December 2005.
The dynamic nature of within herd prevalence observed for Salmonella, at least in part appears to
explain the variability in reported point prevalence rates. It is logical to assume that other
organisms may exhibit similar behaviour within a herd - with differences driven by the relative
virulence and infectivity of the different organisms in the host.
Model input: prevalence of infection
Given the high level of variability in the reported results, selection of prevalence rates for each
pathogen and animal was difficult, and most likely not representative of the all realistic
situations. While a large number of studies have been reviewed, it was not considered
reasonable to simply fit a distribution to these data for determining a variable input for the
QMRA. Investigating the significance of a changing the prevalence rate on the model output is
considered to be a more important attribute of the QMRA model, with a view to modelling
dynamic prevalence rates within a single herd in the future. As a starting point for the
simulations, point estimates were selected that reflected the general patterns reported in the
literature. The values employed in the modelling effort are summarised in Table 6.
Table 6. Summary of model input values for infection prevalence (%)
Salmonella
E. coli O157
Campylobacter
Crypto sporidium
Cattle
10
20
40
30
Poultry
10
0
80
10
Swine
10
15
40
10
1 ' S
The vast majority of studies aimed at quantifying the prevalence of infection have sought to
identify the presence/absence of the target organism in fecal material. Not only the presence of
pathogens however, but the magnitude of their concentration is essential for quantitative risk
assessment calculations. Far fewer studies have been undertaken with the aim of quantifying
pathogen concentration in feces of infected animals.
Salmonella
Cattle: Very limited data were identified describing the density of Salmonella in the feces of
infected cattle. In their modelling study Chapagain et al. (2008) assumed the excretion density
of infected cows to be 109 (units not specified).
9/1 f\
Poultry: Byrd (1998) reported on challenging day-old chicks with 10 , 10 or 10 Salmonella
typhimurium by gavage. The concentration of Salmonella in litter ranged from 10205 to 10455
25
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
cfu/g litter (n = 10), with the highest concentrations resulting from chicks inoculated with 104
and 106 cfu. Hutchison et al. (2004) reported the mean and maximum concentration of
Salmonella in fresh chicken manure to be 102'34 and 104'34 cfu/g (n = 12).
Swine: Hutchison et al. (2004) reported the geometric mean and maximum concentration of
9 7R ARQ
Salmonella in fresh swine manure to be 10 and 10 cfu/g (n = 10).
E.coli O157:H7
Cattle: A great deal more information was available for E. coli O157:H7 in comparison to
9 c
Salmonella. Reported concentrations of E. coli O157:H7 in cow feces include: 2x10 to 8.7x10
cfu/g (Shere et al, 2002); 4 to >l.lx!06- per 10 g of feces (Widiasih et al., 2004); and <100 to >
36xl06cfu/g (Chase-Topping et al., 2007). Providing insight into the nature of the variability,
Fegan et al. (2004) reported the concentration of E. coli O157:H7 in cattle feces to vary from
undetectable <3 MPN to 105 MPN/g with 67% of samples < 10 MPN/g and 8% in the range of
103-105 MPN/g. Matthews et al. (2006) undertook bacterial counts on 440 fecal samples
positive for E. coli O157. Approximately one quarter of samples had concentrations >100 cfu/g
reaching a maximum of around 106 cfu/g. The authors reported an outlier of 3.6x107 cfu/g that
had been excluded from the statistical analysis.
The most comprehensive dataset identified regarding the variability in shedding was undertaken
by Robinson et al. (2004) in which two groups of naturally infected calves were intensively
sampled for periods of 5 and 15 days, respectively. In that study the reported median level of
shedding was 103 up to a maximum up 106 cfu/g (Robinson et al., 2004). Recovery of the
method was not taken into account with quantitative estimates, however the limit of detection
was estimated at 100 cfu/g feces. The authors argued that the highly variable excretion densities
may have been be due to actual variability in shedding, clumping of organisms within the fecal
material, or variability in method recovery.
Poultry: No studies were identified that described the excretion density of E. coli O157:H7 in
infected birds. Prevalence ofE. coli O157:H7 in poultry appears to be very low.
Swine: Cornick and Helgerson (2004) challenged three month old pigs with graded doses of E.
coli O157:H7. Pigs were housed indoors on concrete floors or decks in Iowa. Shortly after
inoculation fecal E. coli O157:H7 density ranged between 103 and 107 cfu/g. After two weeks
and then two months, the shedding density ranged from 50 to 1000 cfu/g, and non-detect to 104
cfu/g, respectively. Hutchison et al. (2004) reported the mean and maximum concentrations of
O CQ C OO
E. coli O157:H7 in fresh swine manure at 10 ' and 10 ' cfu/g respectively (n = 15).
Campylobacter
Cattle: Stanley et al. (1998) reviewed the excretion patterns of Campylobacter in sheep and
cattle and concluded that young cattle are exposed to Campylobacter infection within the first
few days of life, and calves can excrete very high numbers (108 per g feces); while adult cattle
26
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
shed Campylobacter intermittently throughout their lives at excretion densities of around 10 -
103per g feces.
9 & T Q
Poultry: Cox et al. (2002) reported concentrations of Campylobacter in breeders of 10 - 10
cfu/g feces and broilers of 103'5-106'5 cfu/g feces. Studies reviewed by El-Shibiny et al. (2005)
reported a range of 106- 109 cfu/g feces.
Swine: Average density of Campylobacter in sow feces at two separate farms one week before
delivery were 105'0±L1 and 103'6±0'4 cfu/g feces (Weijtens etal, 1997).
Cryptosporidium
Cattle: For adult cattle, the total load has been has been estimated to be 3900 - 9200 oocysts per
adult animal per day (Atwill et al, 2003). In contrast Davies and co-workers (2005a) reported
that the mean number of oocysts excreted by apparently healthy grazing adult cattle was 331 per
g feces (dw), which relates to a total load of around 107 oocysts per animal per day. The
prevalence and shedding intensity of Cryptosporidium oocysts among animals in the Sydney
drinking water catchment has also been investigated by Cox et al. (2005). In that study, the
concentration of Cryptosporidium across cows, sheep and horses ranged from 0 to >6897 oocysts
per g feces (ww). This expected mean excretion density on the order of hundreds of oocysts per
g feces has also been reported in Canada (Heitman et al., 2002) and Uganda (Nizeyi et al., 2004).
Calves however, may excrete higher numbers of oocysts during infection in comparison to older
cows with reported averages of 1.3x 105 (Starkey et al., 2005) up to 107 oocysts per gram of
feces (Blewett 1989). Nydam et al. (2001) investigated the shedding patterns of calves naturally
infected with Cryptosporidium and reported that an infected 6 day old calf would produce
3.89xl010 oocysts until 12 days old. This is a much higher loading, particularly given the high
expected prevalence of symptomatic Cryptosporidiosis infection among calves. In addition, the
results of genotyping studies suggest that the risk of human infection from oocysts excreted by
pre-weaned calves may be considerably higher than from yearling and older cattle(Santin et al.,
2004).
Poultry: Among the two Cryptosporidium species most often isolated from poultry, C.
meleagridis and C. bayeli (de Graaf et al., 1999; Thompson et al., 2008), only C. meleagridis is
known to infect humans (Ramirez et al., 2004). Turn ova et al. (2002) found that shedding rates
for broilers experimentally infected with C. meleagridis ranged from 0 oocysts to approximately
20,000 oocysts per mL of liquid chicken feces (data were presented graphically). Shedding
began approximately 2 days after experimental infection, peaked after about 7 days and lasted
for more than 15 days post infection. Ferguson et al. (2009) note a Cryptosporidium spp.
shedding rate of 2100 oocysts/g feces for layers in the Netherlands.
27
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Swine: Natural Cryptosporidium parvum infection by bovine and porcine genotypes of C.
parvum has been observed in pigs (Guselle et al., 2003). This porcine genotype has been
identified in a human infection (Xiao et al., 2002). Reported shedding rates of Cryptosporidium
parvum from pigs ranged from 0 oocysts at 18 days post-weaning to a maximum geometric mean
for 33 pigs (94% infection rate) of 1596 oocysts/g feces at 53 days post-weaning. Dorner et al.
(2004) fit the same data with a gamma distributions to describe shedding intensities observed
during the course of the study. In their review of pathogen loads from animal sources, Ferguson
et al. (2009) note Cryptosporidium spp. shedding rates from prior studies include 14.3 oocysts/g
for pigs in an Australian study and 472 oocysts/g for 6 - 8 week old pigs.
Excretion patterns and super-shedders
All studies that investigated the concentration of pathogens in the fecal material of infected
animals reported a wide variability in the measured concentration. In a review by Chase-
Topping et al. (2008) the authors argue that for E. coli O157 excretion density could not be
described by a single distribution as most (75%) fecal samples were positive for bacterium
9 "7
containing <10 cfu/g of feces, while some animals excreted up to >10 cfu/g of feces. The
implications could be substantial, since in one study the high shedders (defined as >104 cfu/g
feces) made up 9% of a sample of slaughter cattle but were responsible for >96% of all E. coli
O157 bacteria shed.
Chase-Topping et al. (2008) recommend the following definition of a 'super-shedder': "An
individual who for a period yields many more infectious organisms of a particular type than most
other individuals of the same host species. Typically, many more infectious units are released
from a super-shedder. The term is most useful when there is a clear biological basis for the
distinction between super -shedders and non-super-shedders (such as host genetic differences,
host immune suppression, type differences in the infectious organism, or the presence or absence
of co-infections)."
Based on the brief review presented in this paper, the pattern of heterogeneity in excretion
density appears to exist for all reference pathogens.
Model input: excretion density
Two separate excretion densities were selected for each pathogen to represent endemic excretion
rates and 'super-shedder' excretion rates (Table 7). The initial simulations were undertaken with
zero super-shedders in the catchment (i.e. only endemic excretion), but then additional
simulations were undertaken including 1 to 10 super-shedders on top of the baseline endemic
level. The literature consistently reported variability in shedding rates within each group, and
therefore a variable shedding density was assumed. Sufficient data were not available to define
the shape of the distribution, therefore, the triangular distribution was used to describe the
estimated range of likely values. Distinguishing between different animal types was not
possible.
28
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Table 7. Summary of model input values for excretion density (Logi0. microorganisms/g)
Salmonella
E. coli O157
Campylobacter
Cryptosporidium
Endemic
Triangular(2, 3, 4)
Triangular (1, 2, 3)
Triangular (2, 3,4)
Triangular (1, 2, 3)
Super-shedder
Triangular (5, 6, 7)
Triangular (5, 6, 7)
Triangular (6, 7, 8)
Triangular (5, 6, 7)
B a cter i a (£. coli 015 7: H 7, Salmonella and Campylobacter)
Muirhead et al. (2005) reported that there was an initial increase in the number of bacteria in
defecated cow feces for the first 2 weeks, and by the 5th week the bacteria were back to their
initial levels. A study undertaken by Van Kessel et al. (2007b) concurred with these results
reporting that cowpats may remain a substantial source of E. coli for at least 30 days after
deposition and maybe much longer, mostly because of a substantial growth in E. coli (up to 1.5
logic) during the first 4-8 days. Temperature appeared to be the leading factor affecting the
magnitude of the initial growth of the E. coli population in freshly deposited bovine feces. The
range of temperatures between 20 and 35°C appeared to be the most favourable for the post-
deposit growth.
Regrowth of bacteria including Salmonella has been previously reported (Gibbs et al., 1997;
Zaleski et al., 2005) and has been associated with increasing soil moisture after rainfall events
(Pepper et al, 1993). Gagliardia and Karns (2000) examined the impact of intermittent wetting
and drying on E. coli O157:H7 survival and transport through soil columns. The researchers
expected the level of E. coli O157:H7 in leachate to significantly decrease during dry periods,
however no decrease was observed indicating that only limited soil moisture was required to
support bacterial survival and possible growth.
E. coli O157:H7 has been shown to survive in sewage sludge for greater than 2 months (Avery et
al., 2005). Campylobacter spp. were thought to survive poorly in digested sludge applied to land
(Jones et al., 1990), but subsequent work identified that 8% of swab samples taken from sludge
put to land were positive for C. jejuni (Jones, 2001), identifying the need for further work to
evaluate the persistence of Campylobacter in sludge amended soil.
Rosen (2000) reported that pathogenic organisms including Salmonella typhimurium were
observed to survive less than half as long in aerated manure than in non-aerated manure, largely
because temperatures in aerated manure were significantly higher than those observed in the
non-aerated manure .
Microcosm studies (Astrom et al., 2006) indicate that Campylobacter from piggery effluent is
much less persistent than either Salmonellae or indicator bacteria. Campylobacter coli
29
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
inoculated into microcosms with high soil moisture, moss, and lichen covers was not detected
within 3 hours post inoculation, even in microcosms kept in the dark and at relatively low
temperature (conditions believed optimal for Campylobacter survival). The authors note that
other studies conducted in wastewaters and surface waters have shown greater Campylobacter
persistence than observed in their study. Guan and Holley (2003) reviewed the persistence of
Campylobacter in a range of matrices, and reported persistence in water for 8-120 days at 4-8°C
but <2 days at 20 - 30°C; in soil for 20 days (4-6 °C) and 10 days (20-30 °C); in human feces for
12-21 days (4 -6 °C); in cattle manure for 3 days (20-37 °C); and in cattle manure slurry for 3
days (4-37 °C).
Dieoff constants for E. coli and fecal coliform reported by Van Kessel et al. (2007b) are
summarized in
Table 8. The average air temperature during the observation period was 25.8°C. Rate constants
in laboratory incubated samples were substantially smaller than the rate constants in field
incubated samples, however the difference between open and shade groups from the field was
not significant.
Table 8. Exponential die-off constants for E. coli and fecal coliform
Data from Van Kessel et al. (2007b)
Field
Open Shade 21.1°C
E. coli
Fecal coliform
0.205±0.070
0.225±0.017
0.230 ±0.012
0.169 ±0.030
0.08 ±0.02
0.071 ±0.018
Laboratory
26.7°C
0.125 ±0.044
0.103 ±0.039
32.2°C
0.166 ±0.028
0.125 ±0.016
Cattle Manure: Guan and Holley (2003) reviewed Salmonella survival and reported persistence
for 27-60 days (temperature range 4-37°C) in liquid manure (slurry); for 48 days in solid manure
at 4-5°C and 48 days in solid manure at 20-37°C. Rosen (2000) reported a 0.246-0.986 cfu/d
reduction in Salmonella in cattle manure slurry. Inactivation rates for E. coli O157 and
Salmonella in fresh cow manure and slurry reported by Himathongkham et al., (1999) are
summarized in Table 9.
30
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Table 9. Inactivation rates of Salmonella and E. coli O157 in fresh cow manure and slurry
Data from Himathongkham et al. (1999)
Inactivation rate k Log10 inactivation/day
Fresh manure Slurry
Temp
4°C
20 °C
37°C
Location in Pile
Middle and bottom
Top
Middle and bottom
Top
Middle and bottom
Top
Salmonella
0.049
0.079
0.107
0.040
0.578
0.120
E. coli O157
0.054
0.111
0.074
0.046
0.279
0.112
Salmonella
0.061
0.079
0.422
E. coli
0.04
0.068
0.315
O157
Poultry Manure: Hutchison et al. (2005a) reported Salmonella inactivation rates of k = 0.565
logic cfu/d in farmyard manure; E. coli inactivation rates of k = 0.725 logic cfu/d; and
Campylobacter of k = 0.395 cfu/d. Inactivation rates for E. coli O157 and Salmonella in fresh
poultry manure and slurry reported by Himathongkham et al., (1999) are summarized in Table
10.
Swine Manure: Hutchison et al. (2005a) reported Salmonella inactivation rates of k = 0.552
logic cfu/d in farmyard manure; and E. coli inactivation rates of & = 0.55 logic cfu/d.
Table 10. Inactivation rates of Salmonella and E. coli O157 in fresh poultry manure and slurry
Data from Himathongkham et al. (2000).
Temp Location in Pile
Inactivation rate k LoglO inactivation.day-1
Fresh manure Slurry
4°C
20 °C
37°C
Middle and bottom
Top
Middle and bottom
Top
Middle and bottom
Top
Salmonella
0.086
0.117
0.627
0.631
1.634
1.634
E. coli O157
0.061
0.070
0.689
0.688
1.683
1.683
Salmonella
0.022
0.151
0.571
E. coli O157
0.0064
0.145
1.527
Cryptosporidium
For Cryptosporidium, the mechanism(s) of inactivation in fecal material and soils are not well
understood. However temperature has been shown to be an important factor determining the
31
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
inactivation rate of oocysts, with a rapid increase in inactivation as temperatures exceeded 35°C
(Davies et al., 2005b; Jenkins et al., 1999). Other factors considered important included
soil/biotic effects since much higher rates of inactivation were observed in fecal material in
comparison to water (Olson et al., 1999). By implication, pathogens may be expected to be
inactivated more quickly in microbially rich matrices such as cow pats. Other factors including
ammonia (microbial production of ammonia in stored animal wastes) can also contribute to
inactivation of oocysts in stored fecal material (Jenkins etal, 1998).
Davies et al. (2005b) investigated Cryptosporidium oocysts persistence in closed soil
microcosms over time, using fluorescence in situ hybridization (FISH) as an estimate of oocyst
'viability'. Time for one logic inactiviation (calculated from reported inactivation rates) ranged
from 13-24 days at 35°C (Log 10 k=0.077 to 0.042 day'1) to 45-75 days at 20°C (Log 10 k= 0.022
to 0.013 day"1) depending on soil type.
Model input: persistenceof reference pathogens in feces and soil
Manure handling has a profound influence on pathogen persistence, with greater persistence
observed, in general, in liquid and slurry manure storage than in solid manure storage. Input
assumptions for the model are summarised in the following sections and in Table 11.
Cow pats: The inactivation rate for bacterial pathogens in cow pats was assumed to be zero for
the first 4 days, and then follow open field conditions of 0.2 day"1 exponential decay rate (0.087
day"1 on a Logic scale). Inactivation rate for Cryptosporidium oocysts was assumed to be similar
to inactivation rates at 20°C, and hence a triangular distribution of (0.05, 0.04, 0.03) was
selected. This equates to T(0.02,0.015,0.013) on a Logic scale. Higher rates can be expected at
higher temperatures, and hence this assumption is likely to be conservative.
Cattle manure: The inactivation rate for bacterial pathogens in cow manure was based on
results reported by Himathongkham et al. (1999) for 20 °C, in the middle or bottom of the pile,
rounded down. No additional data were available for Cryptosporidium, so inactivation was
assumed to be the same as in cow pats.
Poultry manure: The inactivation rate for bacterial pathogens in poultry manure was based on
results reported by Himathongkham et al. (2000) for 20 °C, in the middle or bottom of the pile,
rounded down. Campylobacter was assumed to inactivate at a rate equal to the value reported by
Hutchison et al. (2005a). No additional data were available for Cryptosporidium, so inactivation
was assumed to be the same as in cow pats.
Poultry slurry: The inactivation rate for bacterial pathogens in poultry slurry was based on
results reported by Himathongkham et al. (2000) for 20 °C, in slurry, rounded down.
Campylobacter was assumed to inactivate at a rate equal to the least persistent of E. coli O157
and Salmonella. No additional data was available for Cryptosporidium, so inactivation was
assumed to be the same as in cow pats.
32
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Swine manure slurry: The inactivation rate for bacterial pathogens in swine manure was based
on the only available values reported by Hutchison et al. (2005a) of 0.5 Logic inactivation.day"1.
No additional data were available for Cryptosporidium, so inactivation was assumed to be the
same as in cow pats.
Manure amended soil: In the absence of any additional data, inactivation rates in manure
amended soil were assumed to be the same as in the manure alone.
Table 11. Model inputs for inactivation rates of reference pathogens in fecal deposits and manure
Inactivation rate k LoglO inactivation per day
Salmonella E. coli O157:H7 Campylobacter Cryptosporidium
Cow pats
Cattle manure
Poultry manure
Poultry slurry
Swine manure slurry
zero inactivation for
first 4 days then
0.087 day-1
0.10
0.60
0.15
0.50
zero
first
0.087
0.07
0.60
0.14
0.50
inactivation for
4 days then
day-1
zero inactivation for
first 4 days then
0.087 day-1
0.10
0.395
0.15
0.50
T(0.02,0.015,0.013)
B a cter i a (£. coli 015 7: H1, Salmonella and Campylobacter]
Sinton et al., 2007 investigated the inactivation rates of E. coli., Salmonella enterica and
Campylobacter jejuni under light and dark conditions. Temperature was maintained at 14°C.
Inactivation was consistently slower in 'dark' conditions in comparison to sunlight. The
reported inactivation rates are included in Table 12.
Table 12. Inactivation rates and times for E. coli, Salmonella enterica and Campylobacter jejuni
Data from Sinton et al. (2007)
Indicator/Pathogen
E. coli
Salmonella enterica
Campylobacter j ejuni
Tgo
(hours)
548
67.4
82.6
Dark
k
(Log 10
decay rate/d)
0.04
0.36
0.30
Sunlight
Winter
(T90 hours)
17.3
26.8
1.58
k (Log 10
decay
rate/d)
1.4
0.9
15
Summer
(T90 hours)
3.85
4.81
0.8
k (Log 10
decay
rate/d)
6.3
5.0
30
Cryptosporidium
Ives et al. (2007) undertook bench-scale survival studies with Cryptosporidium parvum. C.
parvum inactivation rates ranged from 0.0088 Logio/day at 5°C to 0.20 Logio/day at 30°C.
33
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Temperature, surface water or groundwater type, and the interaction of these factors had
statistically significant effects on the survival of C. parvum.
Model input: persistence of reference pathogens in the storage pond
Given the results and relevance of the work undertaken by Sinton et al. (2007), the QMRA
model was constructed to allow for differentiating between light and dark inactivation. The
inactivation rates as reported by Sinton et al. (2007) were used in the model, using winter
sunlight values to be conservative and a sub-population size of 10% (chosen arbitrarily) to
differentiate between those organisms protected from sunlight at depth in the pond (Table 13).
Cryptosporidium is less sensitive to inactivation by sunlight in comparison to bacterial pathogens
(Mendez-Hermida et al., 2005); however whatever the key mechanism of inactivation is for
Cryptosporidium a similar sub-population phenomena was considered likely, with a small
portion of organisms finding protection. Using the results of Ives et al. (2007) that highlight the
importance of temperature on Cryptosporidium survival, the slow phase was assumed
represented by the low temperature inactivation rate and fast phase for high temperature
inactivation. The sub-population was assumed to be 10% (chosen arbitrarily). The inactivation
rate for 30 °C of 0.2 Logic per day appears to be high, and including the two population model
allows for a more conservative consideration of persistence. These values should be tested and
refined in the future.
Table 13. Model inputs for inactivation rates of reference pathogens in storage pond
Inactivation rate k LoglO inactivation per day
Salmonella E. coli O157:H7 Campylobacter Cryptosporidium
Fast (Light) 0.9
inactivation
Slow (dark) 0.36
inactivation
1.4
0.04
15
0.3
0.2
0.0088
Persistent sub- 10
population size (%)
10
10
10
Overland flow across agricultural land occurs when rainfall intensity exceeds the infiltration
capacity, or when the soil becomes saturated. Once overland flow occurs, there is potential for
mobilization of pathogens to surface waters. The ease with which contaminants are mobilized is
influenced by their partitioning status; free unattached cells are more easily incorporated into
mass flow in comparison to those microorganisms attached to soil or manure particles. For the
contaminant to reach surface waters, these forces of entrainment must be maintained for the
overland distance to the waterway.
34
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Grazing land
For all animals, the mobilization of pathogens from fresh fecal deposits depends on numerous
factors including nearness to the waterway. For this work, it was assumed that for any specific
pathogen, the animal source did not impact the relative mobilization. Davies-Colley et al. (2004)
documented the water quality impact of a herd of 246 dairy cows crossing a stream. The cows
defecated approximately 50 times more per metre of stream crossing than elsewhere on the
raceway. In this example, access to the stream combined with the increased frequency of
defecation would lead to a high probability of mobilization.
Muirhead and coworkers have undertaken several studies investigating the mobilization of E.
coli from fresh cowpats with the following conclusions:
• The number of E. coli in the cowpat runoff was highly variable and was strongly
correlated with the number of E. coli in the cowpat (Muirhead et al., 2006c);
• E. coli mobilized from cowpats were transported as single cells, and only a small
percentage (approximately 8%) were attached to particles; implying that particle transport
is not necessary for pathogen mobilization (Muirhead et al., 2005); and
• These results implied that in runoff generated by saturation-excess conditions, bacteria
are rapidly transported across the surface and have little opportunity to interact with the
soil matrix. The removal of E. coli from overland flow under saturation excess
conditions was also investigated, and reported to be limited (< 50%). Instead, most
bacteria remained entrained within the overland flow down the length of the plots (5m
long) (Muirhead et al., 2006b).
On soil blocks from Sydney's catchment, Davies et al. (2004) investigated the dispersion and
initial transport of Cryptosporidium oocyst from fecal pats under simulated rainfall events. The
oocyst load in runoff was significantly affected by the vegetation status, the slope of the soil, and
the event characteristics in terms of rainfall intensity. These same factors significantly affected
the concentrations of oocysts retarded on the surface soil a short distance (10 or 30cm) downhill.
Devegetated or heavily grazed soils represented a higher risk than vegetated soils. The freshly
crusted cow pats containing 107 oocysts transported from 1002 oocysts on vegetated loam soil to
1045 on unvegetated loam soil over a distance of 1m. In addition, the bovine manure matrix has
been reported to enhance the attachment of Cryptosporidium oocysts to soil (Kuczynska et al.,
2005b).
Atwill et al. (2002) suggested that vegetated buffers constructed with sandy loam or soils with
higher bulk densities were less effective at removing oocysts (1-2 logic reduction per m) than
buffers constructed with silty clay, loam, or solids with lower bulk densities (2 to 3 logio
reduction per m). Their study suggested that on slopes of <20% a length of 3m should function
to remove 3 logic (99.9%) of C. parvum oocysts from agricultural runoff generated during events
involving mild and moderate precipitation.
35
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Ferguson et al. (2007) studied the dispersion and transport of Cryptosporidium parvum oocysts,
Escherichia coli and PRD1 bacteriophage seeded into artificial bovine fecal pats during
simulated rainfall events. Transportation efficiency increased with decreasing size of the
microorganism studied; Cryptosporidium oocysts were the least mobile followed by E. coli and
then PRD1 phage. Rainfall events mobilized 0.5 to 0.9% of the Cryptosporidium oocysts, 1.3-
1.4% of E. coli bacteria, and 0.03-0.6% of PRDlbacteriophages from the fresh fecal pats and
transported them a distance of 10m across the bare soil sub-plots.
Model input: probability of overland transport from grazing land
Several studies were identified that investigated the mobilisation of pathogens from direct fecal
deposits on agricultural land. While trends have been identified, quantification of the probability
of passage of the reference pathogens from the grazing land to surface waters is still not possible,
and realistically is going to be site specific. To consider what could be quantified however,
animal access to waterways can reasonably lead to an assumption ofp=l, and vegetated buffers
may allow the probability to be estimated atp < 0.001 (3 Logio).
As a starting point in the model, the probability of transport to waterway was based on results by
Ferguson et al. 2007, with a uniform distribution selected to represent the reported ranges.
Assuming no direct access of animals with the stream or pond, this is likely to be a conservative
estimate since they are estimates for a distance of 10m (Table 14).
Table 14. Model inputs for probability of passage from grazing land to surface waters
Probability of passage (p)
Salmonella
Uniform
(0.013,0.014)
E. coli O157:H7
Uniform
(0.013,0.014)
Campylobacter
Uniform
(0.013,0.014)
Cryptosporidium
Uniform
(0.005,0.009)
Manure amended soil
In comparison to the mobilisation of pathogens from grazing land, similar principals apply for
the conditions that lead to overland transport form manure attenuated soil. Importantly however,
the pathogen transport rates would be expected to be lower from manure amended soil.
Amending soil leads to higher infiltration, increased surface retention as demonstrated for human
biosolids application (Joshua et al., 1998; Moffet et al., 2005) and increased vegetative cover. In
addition, sites selected for manure application tend to be of lower relief.
Model input: probability of overland transport from manureamended soil
Because the objective for this study was a screening-level analysis of animal-derived pathogen
risk in recreational waters, a coarse, conservative model (with a tendency to overpredict
pathogen loading) was implemented for the transport of pathogens from manure amended soils;
models with higher resolution were not implemented because manure management practices are
diverse and give rise to widely differing pathogen loading rates. As a conservative starting point,
the same probability of transport rates applied for grazing land were applied to describe transport
36
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
from manure amended land (Table 15). This is certainly an overestimate of pathogen transport
rates given the known reduction in pathogen runoff when best practices for land application are
used (Goss and Richards, 2008; Vinten et al., 2004).
Table 15. Model inputs for probability of passage from manure amended land to surface waters
Probability of passage (p)
Salmonella
Uniform
(0.013,0.014)
E. coli O157:H7
Uniform
(0.013,0.014)
Campylobacter
Uniform
(0.013,0.014)
Cryptosporidium
Uniform
(0.005,0.009)
During low flow conditions bacterial and protozoan pathogens can settle to a stream or river bed
and be retained. Searcy et al. (2006) investigated the transfer of Cryptosporidium oocysts from
surface water to the sediment beds of streams and rivers using controlled laboratory flume
experiments. The association of C. parvum with other suspended sediments increased both the
oocysts' effective settling velocity and the rate at which oocysts were transferred to the sediment
bed.
Model input: proportion of mobilized load retained in stream sediments
In the absence of real quantitative estimates of sediment transport rates, and given the results of
Searcy et al. (2006), the proportion of organisms associated with solids was used as a first pass to
estimate that portion that may be retained (Table 16). For bacteria, this proportion was assumed
to be 8% (given 8% solids associated (Muirhead et al., 2006a)) and for Cryptosporidium
representative values were not found for attachment of Cryptosporidium to stream sediments. In
experiments with soils and manure amended soils (Kuczynska et al., 2005a), Cryptosporidium
was observed to adsorb readily to sandy loam and clay loam soils (72% and 93% attachment,
respectively) when introduced to the soils in a suspension in dilution water and even more
readily (97.4% and 97.7 %) when oocysts were introduced to the soils in a dilute suspension of
bovine manure. In a study of urban stormwater Clostridium perfringens partitioning following
high rainfall events ranged from 20 to 60% (Characklis et al., 2005). Based on the very high
tendency of Cryptosporidium to attach to soils, Clostridium perfringens is expected to be a
conservative surrogate for Cryptosporidium mobilization from stream sediments in this case.
Therefore 20% was selected as a reasonable conservative estimate (here, conservative indicates
C. perfringens is expected to over-predict Cryptosporidium mobilization to the water column) for
Cryptosporidium.
Table 16. Model input for the proportion of load transported to stream sediments
Salmonella E. coli O157:H7 Campylobacter Cryptosporidium
Proportion of load O08 O08 O08 O2
37
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
No studies were identified that specifically addressed the inactivation rate of pathogens within
stream sediments. The studies identified to address the persistence of pathogens in the storage
pond were considered equally relevant to the freshwater stream context.
Model input: persistence of reference pathogens in stream sediments
The same values as applied to estimate persistence in the storage pond, were applied to the
persistence in stream sediments (Table 17).
Table 17. Model inputs for inactivation rates of reference pathogens in stream sediments
Inactivation rate k LoglO inactivation per day
Salmonella E. coli O157:H7 Campylobacter Cryptosporidium
Fast (Light) inactivation 0.9 1.4
Slow (dark) inactivation 0.36 0.04
Persistent sub- population 10 10
size (%)
15 0.2
0.3 0.0088
10 10
Re-suspension of microbes from stream sediments can be a source of increased pathogen
concentrations during high flow events. While research has been undertaken to quantify re-
suspension (Jamieson et al., 2005; Muirhead et al., 2004); no real data relevant to the current
modelling exercise was identified. Jamieson et al. (2005) reported that bacterial re-suspension
was primarily limited to the rising limb of storm hydrographs implying that a finite supply of
sediment-associated bacteria are available for re-suspension during individual storm events.
Model input: proportion of retained load mobilized during a re-suspension event
In the absence of any quantitative data from the literature, and given the results of Jamieson et al.
(2005), the impact of resuspending the total sediment load was investigated in a conservative
approach to evaluate the potential impacts of reasonable worst case conditions (Table 18).
Table 18. Model input for the proportion of retained load that is re-suspended
Salmonella E. coli O157:H7 Campylobacter Cryptosporidium
Proportion of load 11 11
In a study undertaken by Dufour et al. (2006), fifty-three recreational swimmers participated
using a community swimming pool disinfected with cyanuric acid stabilized chlorine. The
swimmers were asked to actively swim for at least 45 minutes and to collect their urine for the
next 24 hours. Results of the study indicated that non-adults ingest slightly more water than
38
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
adults do during swimming activity (Figure 10). The predicted median volume of water
swallowed by all participants combined was approximately 19mL.
100
-J
"8 10
1
T
I 1
0.1 -
n
n
• Adults
•:'.j Adults • below detectable lira ills
* Chikten
A Chik±en- below detectable limits
1 £ 5 10 20 30 40 50 60 70 80 90 95 98 99
Percent of Observations Less Than Correponding Value
Figure 10. Ingestion volumes during recreational activities
Data from Dufour et al. (2006)
Model input: Consumption of water by recreational bathers
A normal distribution was fitted to the natural log transformed combined adult and child dataset:
Ln mean of 2.92 with Ln sd of 1.43 this yields a mean of 18.6 ml.
Human health effects: Dose response relationships
During dose-response analysis, data from human clinical studies, epidemiological studies, animal
studies, and/or outbreaks are used to develop a mathematical relationship between the intensity
of exposure or amount of intake and the subsequent occurrence of disease or infection. Dose-
response models are mathematical functions that take as input the dose to which individuals or
populations are exposed and yield a probability (bounded by 0 and 1) of the particular adverse
health effect (Haas et al., 1999).
Dose-response models are generally derived using statistical estimation techniques, and the form
of the relationship between exposure and response is determined by (1) assumptions related to
the biological processes leading to infection, and (2) the "shape" of the relationship found in the
data between exposure and the health outcome of interest. The mathematical form of the dose-
response model may vary with pathogen or strain, route of administration, distribution of host
statuses, and other factors. For this study, dose response models were obtained from the
scientific literature and applied. The dose-Response models applied in the QMRA model are
summarised in Table 19.
39
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Table 19. Summary of dose response relationships used in QMRA model
Salmonella
E. coli O157:H7
Dose Response
Model
Gompertz
function for
Bareilly strain
Hypergeometric
beta-Poisson
model*
Parameter
values
log(a) =
11.68; b=0.82
Median value
ofMarkov
chain Monte
Reference
(Coleman and
Marks, 1998,
2000; Coleman et
al, 2004; Soller et
al, 2007)
(Tennis et al.,
2008)
Probability of
illness given
infection
1.0 (Note that
dose response
is for illness
not infection)
0.825
Reference
(Havelaar et al.,
2003)
E. coli O157:H7
Campylobacter
Cryptosporidium
Hypergeometric
beta-Poisson
model*
Beta-Poisson
approximation
Exponential
Median value
ofMarkov
chain Monte
Carlo sample
alpha=0.37
and beta
=37.6
alpha=0.145
beta=7.59
r = uniform
distribution
(0.04-0.16)
(Teunis et al, 0.825
2008)
(Haas et al., 1999; 0.2
Medema et al.,
1996; Teunis et
al., 1996)
(U.S. EPA, 2006) 0.71
(Havelaar et al.,
2003)
Point estimate
from results
presented by
Teunis et al.
(1999b)
(Havelaar and
Melse, 2003)
Beta-Poisson approximation at high doses and at low dose (<0.1) with the exponential model
r=a/a+b
40
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
The QMRA model was run with the described inputs using a Monte Carlo simulation approach
for 10,000 iterations to construct the cumulative density function (CDF) for downstream illness
risk following Event 1, Event 2, and Event 3. As indicated in the previous sections of the report,
all of the inputs to the QMRA were highly uncertain. The variability in the input values
extracted from the literature was considerable. In addition, uncertainty exists regarding whether
the relatively simple model structure was a reasonable representation of types of realistic
exposure scenarios that were investigated. Therefore, the output from these simulations is best
used to compare and understand the relative risks associated with the various scenarios and to
prioritize the types of data that could be collected in the future to most significantly enhance
interpretation of future analyses.
Despite limited predictive capabilities, the simplified model structure was more than adequate
for investigating the interactions of model parameters and to identifying the most important
sources of uncertainty - or rather the sources of uncertainty that have the largest impact upon the
illness predictions. This approach to model exploration is termed sensitivity analysis and is the
study of how the uncertainty in the output of a model can be apportioned to different sources of
uncertainty in the model input.
The model output represented in terms of CDFs of predicted illnesses per 1000 exposures for
each reference pathogen and animal type are illustrated in Figure 11 (cattle), Figure 12 (swine)
and Figure 13 (poultry). In all simulations the predicted rates of gastrointestinal illness were
consistently higher for events 1 and 2 compared to event 3. Thus, mobilization of fresh fecal
material by overland flow was the primary driver of downstream risk. Intercepted and stored
fecal material was less of a concern due to the inactivation of pathogens, even for
Cryptosporidium which was modelled with a persistent inactivation rate. Event 2 risks were
slightly higher than Event 1. The difference between event 1 and 2 was the impact of the storage
pond overflow (in addition to direct overland flow alone) on the overall downstream risk, which
was surprisingly small. The difference was greater for Cryptosporidium in comparison to the
bacterial pathogens due to the persistence of oocysts, however the high order of magnitude of
daily pathogen excretion still overshadowed the build up of pathogens in the storage pond.
41
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Cattle model: Campylobacter
jr
15
Cattle model: £ co//O157:H7
£
Cattle model: Cryptosporidium
r/7
0.001
Event
1 Direct runoff
o.oi o.i i 10 100
Illnesses per 1000 exposures
— 2 Runoff and Pond Overflow 3 Sediment resuspension
Cattle model: Salmonella
= //
/
z_
^^^ _
/
/
//
//
/ //
'
^
'/
Figure 11. Downstream illness risks from the Cattle model for each reference pathogen
42
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Swine model: Campylobacter
Swine model: £ co//O157:H7
Swine model: Cryptosporidium
Event
1 Direct runoff
Illnesses per 1000 exposures
— 2 Runoff and Pond Overflow 3 Sediment resuspension
Figure 12. Downstream illness risks from the Swine model for each reference pathogen
43
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
ility Cumulative Probability Cumulative Probability
>PP PPPPPPPPP PPPPPPPPP
1
Cumulative Prc
p o o o o
3 M NJ Li) fc Ul
Poultry me
— •
0.0<
/
/
zz_
/
^
/
~#L.
=5=^
(del: Camp
)01 0.0
/-— .
/r
#
//
/
flobacter
/
/
____^
/
/
/
/>
^_
=£^
^--
/>
//
~2_
//
//
-^^
^
/
01 0.01 0.1 1 10 100 1000
*^-
Pnultr
y model: Salmonella
D O.0001 0.001 O.O1 0.1 1 10 100 1000
Poultry model: Cryptosporidium
\
^^
~^^_^'
^
'^^^:-
^^
^^^
"^ ^^"
, <^^
/
/ /
T_ //
7
^-^*~
"^ — ^^
/
?zL
/^
-- —
0 0.0001 0.001 0.01 0.1 1 10 100 1000
Illnesses per 1000 exposures
Event
— 1 Direct runoff — 2 Runoff and Pond Overflow 3 Sediment resuspens on
Figure 13. Downstream illness risks from the Poultry model for each reference pathogen
Note for Figure 13: no results for E. coli O157:H7 as the prevalence was assumed to be 0.
Comparison between Campylobacter and Cryptosporidium illness risks for poultry (Figure 13)
indicates that the risk from Cryptosporidium was predicted to be higher than for Campylobacter.
This was due to the relatively high prevalence of Cryptosporidium (-25%) in the few reported
studies, and the high persistence in the environment. These results however highlight the
importance of accounting for the human infectious component of the animal infections. While
poultry do have Cryptosporidium infections, the majority appear to be C. meleagridis which is
unlikely to cause infection in healthy humans. In contrast, poultry are frequently infected with
Campylobacter jejuni and Campylobacter coli which are most frequently responsible for
Campylobacter infection in humans. Failure to consider which strains of each organisms are
most likely to cause infection in humans can lead to an overestimation of risk, and unnecessarily
conservative illness estimates. In this case, despite the relatively high prediction for
Cryptosporidium risk from poultry, epidemiologic evidence allows this to be discounted with a
greater focus on Campylobacter.
44
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Comparison of the predicted risks for Campylobacter and Cryptosporidium between the swine
and cattle operations show that they are of a similar magnitude. This was surprising due to the
significant differences in the onsite management of fecal material on each type of feeding
operation. While cattle were assumed to openly graze 90% of the time, swine operations were
assumed to be entirely housed with fecal material collected and stored before managed
agricultural application. A comparison between the modelled load of pathogens to grazing land
and storage is illustrated in Figure 14.
Pathogen load on grazing
Pathogen load to manure storage
Reference pathogens
| Salmonella |E.coli0157:H7 FJCampvlobacter Q Cryptosporidium
Reference pathogens
| Salmonella | E.coli 01S7:H7 | Campylobacter Q Cryptosporidium
Figure 14. Modeled pathogen loads to grazing land and manure storage
The fact that the risks associated with each of these operations were predicted to be roughly
equivalent does not seem reasonable. There are two factors that contribute to this result: the
importance of the minimum storage time for manure prior to application; and the impact of time
between application and the rainfall mobilising event.
Minimum storage time
While the manure was stored prior to application, the design of the storage facility assumed that
fecal material was added every day. At the time of application, all manure from the storage
facility was assumed to be applied to land. While the mean storage time may have been some
months, the modelled load of pathogens in the manure was still high - leading to high predicted
downstream risk. For the same reasons as mentioned previously, the magnitude of the daily
input of pathogens to the storage facility overshadowed the impact of the inactivation on the final
load. Therefore the pathogen load to agricultural land with manure application was driven by
the minimum storage time, rather than the overall mean storage time (Figure 15).
Storage time has a profound influence on pathogen loads in land-applied manures
(Himathongkham et al., 1999; Himathongkham et al., 2000; Hutchison et al., 2005b; Hutchison
et al., 2005d; Kasorndorkbua et al., 2005; Kelley et al., 1994; Meals and Braun, 2006; Nicholson
et al., 2005), regardless of the manure origin, type (liquid v. solid), and storage system. Yet,
while pathogen reduction is a consideration in farmer choice of timing and rate of manure
application and of storage facility volume, other considerations including nutrient requirements
45
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
of soils, generation of noxious odours, rainfall, availability of labour, construction cost, and
available land also play roles in design and operation of manure application schemes. Some of
these considerations may be at odds with long storage times for pathogen reduction, as in the
case of odour, which tends to be lowest for fresh manure (Dougherty et al., 1998). Because
storage systems must meet multiple objectives, minimum storage times and timing of land
application are generally best management practices (BMPs) though it is likely that for some
individual NPDES permits numerical limits on minimum storage times may have been
stipulated.
T3 108
I io6-
0
'C i n4
00 10
cd
0
1 100
a
1 i "
0
Reference pa
Q Salmone
No withholding period
:hogens
//a | £ coft 0 1 57 :H7 Q
minimum storage: 30 days minimum storage: 90 days
Campylobacte^^ Cryptosporidium
Figure 15. Impact of minimum storage time on load applied via manure application
Time between application and rainfall event
Unlike the continuous contribution of grazing animals openly defecating on land, manure
application is an infrequent occurrence; perhaps 2-3 times per year. The pathogen load available
for transport to surface waters can be expected to decline over time following application due to
environmental inactivation. For the purposes of conservatively estimating the downstream risks
associated with land application, the rainfall mobilizing event was assumed to occur immediately
following application. Figure 16 illustrates the modelled load of pathogens mobilized with a
rainfall event versus time since manure was applied. In comparing the risks associated with
swine and cattle operations it is necessary to consider the duration of the potential risk. While
open grazing poses a potential risk throughout the year, risk associated with manure application
exhibits short term peaks that follow the time of application. The magnitude of these peaks
appear to be, at worst, equivalent to the ongoing risk associated with open grazing.
46
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
a 100000
o
o.
o.
•a
-a
•a
01
o
•a
10000
1000
100
10
10 20 30 40
Time since application of manure (days)
50
60
Reference pathogens
Salmonella — E.coli O157:H7 Campylobacter Cryptosporidium
Figure 16. Mobilized pathogen loads from cattle versus time between application and rainfall event
Sensitivity analysis
Sensitivity to prevalence
Pathogens are only assumed excreted by infected animals and therefore the prevalence of animal
infection with human infectious pathogens was the basis for the QMRA model. A broad range
of prevalence rates were identified in the literature ranging from 1% to 100% of animals on a
given farm. The sensitivity of the predicted illness risk to the prevalence rate was explored and
the results for E. coli O157:H7 illness is illustrated in Figure 17. The broad range of prevalence
rates had a relatively limited impact on the predicted illness rates (2 Logic) largely due to the
high order of magnitude in excretion rate that overshadowed the calculations. In contrast
however, the number of supershedders in the catchment had a strong influence on the predicted
illness rates (Illustrated in Figure 18 for the swine model). Once a single super shedder was
assumed present, this high level excretion of pathogens dominated the risk model, and the risks
rapidly approached their maximum. Increasing the number of supershedders above 1 had less of
an impact.
Figure 17. Sensitivity of predicted illness rates to the prevalence of infection in farm animals
47
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
Point prevalence
— 0.01 --0.1 1
fel ft 0.001
0
1
Number of supershedders
Figure 18. Influence on the predicted illness rates of the number of supershedders among swine
Sensitivity to excretion rate
The sensitivity of the model to the assumed endemic excretion rate for cattle is illustrated in
Figure 19. Reported excretion rates from the literature cover several orders of magnitude (<10-
109 organisms per g), indicating that uncertainty in this input is an important driver in uncertainty
in the predicted illness risk.
100
g 10
s
o l
I lo.i.
ft 0 001
0) „
a 0
Cattle ^^ — ___— — — ~~
^**
^^
***
**^~'~
»•
_,...
^f
~^"
^^'
^. •
,.••*"**"**
_^^-^
-11111*--1^
.^
«•**"*
^^^
^^_^ ^ • "*
^,. •*
...
•••*'
.•****
•*
^ . — •"•"
.••**""
*
Reference j
1 &&
"
)athogens
nonelia
- • - ^nmpylobnctpr
— • E.coliOl51'Rl
Cryptosporidium
T. 1 2 3 4 5 6 7 8 9 10 11
S Logio pathogen density in cattle feces
1
Figure 19. Sensitivity of the risk model to cattle excretion rate for event 1 illness risks
Sensitivity to I nactivation rate
Pathogen inactivation under uncontrolled environmental conditions is a key barrier between
animal feeding operations and humans engaging in recreational activities. In the QMRA model,
inactivation of pathogens is predicted on open grazing land (within cow pats); in manure piles; in
48
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
manure applied to land; in the storage pond; and in stream sediments. Despite a large body of
research investigating the kinetics of pathogen inactivation in a variety of media, and under
laboratory and field conditions, these inputs are still extremely uncertain. While general patterns
can be described (such as pathogen inactivation increases with temperature), the actual
inactivation rates are not well known. The inactivation rates in the model are at best a starting
point to explore model sensitivity and to further motivate research into better defining these
quantitative relationships.
To illustrate the impact of the uncertainty in the inactivation rates, the inactivation of
Campylobacter and Cryptosporidium in manure applied to land is illustrated in Figure 20 and
Figure 21, respectively. Both figures illustrate that the QMRA model is very sensitive to the
assumed inactivation rate. Some references report Campylobacter to be inactivated readily in the
environment. When this upper value is used, the load on manure applied land is reduced quickly
and as soon as 5-7 days following application, the concentration of Campylobacter is predicted
to be negligible. Some references however have indicated that under some circumstances
Campylobacter can persist in the environment, represented by the 'low' inactivation rate. Under
this assumption, even after 3 months in the field, a reasonable concentration of Campylobacter
remains. While good reduction of bacterial pathogens could be assumed in the environment, this
uncertainty should be reduced by identifying conditions that support bacterial persistence before
this reduction can be relied upon as a critical barrier for the protection of public health.
Results of the sensitivity analysis for the inactivation of Cryptosporidium oocysts indicate that
even the faster reported inactivation rates support the assumption that oocysts persistence in the
environment, and that environmental inactivation is a poor barrier for protection of human health
unless extended residence times can be assured.
I
BI
'
I
I
0.01
0.0001
20
30 40 50
Time (days)
60
70
80
90
Figure 20. Estimated Campylobacter load remaining on manure applied ground over time
49
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
41
S ir>8
£ 10
•-
I 106
C5
a 4
•a in
.5
'3
2 1UU
a
J 1
"8
'C
§< 001
t
g 0.0001
O i
75 (
03
5
^^^
) 1
**>-
^^v
0 2
^*"5;^>^
^^s*^^
^^">
0 30 40 5
Time (days)
*«-.
^
0 6
1 Best estimate
^^**>»J^
^^^^IHigh
^S;**>>>
0 70 80 9
D
Figure 21. Estimated Cryptosporidium load remaining on manure applied ground over time
Sensitivity to storage pond catchment size
Onsite interception of runoff can be used to mitigate the downstream risks following rainfall
events. The sensitivity of the predicted illness rates following an overland flow event (Event 1)
to the assumed storage pond catchment size is illustrated for the cattle model in Figure 22. The
results indicate that the downstream risk from Event 1 is eliminated when 100% of the catchment
is intercepted by the storage pond. However, when 90% of the catchment was assumed to be
intercepted (only 10% outside the storage pond catchment) by the storage pond, the downstream
risks were predicted to be similar (within 1 Logic) to when only 10% or 0.1% interception was
assumed. These results indicate that for the storage pond to effectively reduce downstream risks,
it must intercept all the runoff, as even a small portion of flow can contain a significant load of
pathogens.
50
-------
U.S. EPA Office of Water
QMRA Model for Animal-Impacted Waters
!>
Salmonella E. coli O157:H7 Campylobacter Cryptosporidium
Reference pathogens Storage pond catchment
no.i% i—i 10% i—i 90% i—iioo%
Figure 22. Sensitivity of the Event 1 illness risks to storage pond catchment size for the cattle model
51
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
During the course of the work that was conducted for WA 1-08 Task 2 (including the first phase
of this work conducted under 1-08 Task 2 and subsequently Work Assignment 1-08 Task 2
Amendment 3) the following was accomplished:
o Conduct of a literature review, development, implementation and parameterization of a
QMRA model in Analytica software, and development and parameterization of three
exposure scenarios for cattle-impacted waters.
o Planning of and participation in an EPA QMRA workshop in November 2008 at the
ORD offices in Cincinnati, OH where the cattle model was presented and demonstrated.
o Extension of the initial cattle model to two other agricultural animals (swine and poultry)
for the three exposure scenarios.
o Conduct of a literature review of swine and poultry manure data to parameterize the
newly extended QMRA model for Salmonella, Camplyobacter jejuni, E. coll O157:H7,
and Cryptosporidiumparvum.
o Conduct of a sensitivity analysis of the various models (cattle, swine, and poultry) and
associated model parameters for the three exposure scenarios to identify which data and
model components are the most crucial with respect to the conduct of QMRA for animal
impacted waters.
The construction, parameterization, and evaluation of the QMRA model to describe the potential
impacts of an animal feeding site and/or areas where animal manures are applied on
gastrointestinal illness risk among recreational water users has allowed for the important
environmental variables to be evaluated and numerous issues to be identified and discussed. The
salient findings from this work includes the following:
o Onsite collection and storage of fecal material is an important barrier for preventing
pathogen mobilization downstream. Operations that collect and store fecal material for
land application may present short term peaks of pathogen risk, immediately following
application. These peaks are estimated to be roughly equivalent to the ongoing risk
associated with open grazing operations.
o When manure is to be stored and then land applied, the storage barrier is only effective
for pathogen removal when a minimum storage time is ensured.
o Managing land application to avoid periods of high rainfall will reduce risk.
52
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
o Prevalence of infection in any given herd is likely to be constantly changing, and the
within-herd temporal variation of can be substantial.
o Understanding the prevalence of human infectious pathogenic strains could be a critical
component for not overestimating the risk associated with animal-impacted waters.
o Quantifying pathogen excretion density is a significant source of uncertainty in the
overall model. In particular, the existence of super-shedders appears to drive the overall
pathogen load. This aspect requires further research, particularly if identification and
containment of super-shedders is possible.
o Environmental inactivation rates of pathogens are highly uncertain. Therefore, ensuring
pathogen reduction via uncontrolled environmental processes is not feasible unless
extended residence times can be guaranteed. Given the current state of knowledge,
Cryptosporidium oocysts should be assumed to persist for long time periods unless site
specific data indicate otherwise.
53
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Animal and Plant Inspection Service (1995) Escherichia coli O157:H7 Shedding by Feedlot Cattle. U.S. Dept. of Agriculture
National Animal Health Monitoring System Report N182.595
Asano, T., Leong, L.Y.C., Rigby, M.G., and Sakaji, R.H. (1992) Evaluation of the California wastewater reclamation criteria
using enteric virus monitoring data. Water Science and Technology 26: 1513-1524.
Ashbolt, N.J., and Bruno, M. (2003) Application and refinement of the WHO risk framework for recreational waters in Sydney
Australia. Journal of Water and Health 1: 125-131.
Astrom, J., Petterson, S., Bergstedt, O., Pettersson, T.J., and Stenstrom, T.A. (2007) Evaluation of the microbial risk reduction
due to selective closure of the raw water intake before drinking water treatment. J Water Health 5 Suppl 1: 81-97.
Astrom, J., Carlander, A., Sahlen, K., and Stenstrom, T.A. (2006) Fecal inidcator and pathogen reduction in vegetation
microcosms. Water, Air, and Soil Pollution 176: 375-387.
Atwill, E.R., Johnson, E., Klingborg, D.J., Veserat, G.M., Markegard, G., Jensen, W.A., Pratt, D.W., Delmas, R.E., George,
H.A., Forero, L.C., Philips, R.L., Barry, S.J., McDougald, N.K., Gildersleeve, R.R., and Frost, W.E. (1999a) Age, geographic,
and temporal distribution of fecal shedding of Cryptosporidium parvum oocysts in cow-calf herds. American Journal of
Veterinary Research 60: 420-425.
Atwill, E.R., Johnson, E.M., and Pereira, M.d.G.C. (1999b) Association of herd composition, stocking rate, and duration of
calving season with fecal shedding of Cryptosporidium parvum oocysts in beef herds. Journal of the American Veterinary
Medical Association 215: 1833-1838.
Atwill, E.R., Hou, L., Karle, B.M., Harter, T., Tate, K.W., and Dahlgren, RA. (2002) Transport of Cryptosporidium parvum
oocysts through vegetated buffer strips and estimated filtration efficiency. Applied and Environmental Microbiology 68: 5517-
5527.
Atwill, E.R., Hoar, B., das Gracas Cabral Pereira, M., Tate, K.W., Rulofson, F., and Nader, G. (2003) Improved quantitative
estimates of low environmental loading and sporadic periparturient shedding of Cryptosporidium parvum in adult beef. Applied
and Environmental Microbiology 69: 4604-4610.
Avery, L.M., Killham, K., and Jones, D.L. (2005) Survival of E. coli O157:H7 in organic wastes destined for land application.
Journal of Applied Microbiology 98: 814-822.
Barkocy-Gallagher, G.A., Arthur, T.M., Rivera-betancourt, M., Nou, X., Shackelford, S.D., Wheeler, T.L., and Koohmaraie, M.
(2003) Seasonal Prevalence of Shiga Toxin-Producing Escherichia coli, Including O157:H7 and Non-0157 Serotypes, and
Salmonella in Commercial Beef Processing Plants^. Journal of Food Protection 66: 1978-1986.
Blau, D.M., McCluskey, B.J., Ladely, S.R., Dargatz, D.A., Fedorka-Cray, P.J., Ferris, K.E., and Headrick, M.L. (2005)
Salmonella in Dairy Operations in the United States: Prevalence and Antimicrobial Drug Susceptibility. Journal of Food
Protection 68: 696-702.
Bollaerts, K.E., Messens, W., Delhalle, L., Aerts, M., Van der Stede, Y., Dewulf, J., Quoilin, S., Maes, D., Mintiens, K., and
Grijspeerdt, K. (2009) Development of a quantitative microbial risk assessment for human salmonellosis through household
consumption of fresh minced pork meat in Belgium. Risk Anal 29: 820-840.
Byrd (1998) Horizontal transmission of Salmonella typhimurium in broiler chicks Journal of Applied Poultry Research 7: 75-80.
Callaway, T.R., Keen, J.E., Edrington, T.S., Baumgard, L.H., Spicer, L., Fonda, E.S., Griswold, K.E., Overton, T.R.,
VanAmburgh, M.E., Anderson, R.C., Genovese, K.J., Poole, T.L., Harvey, R.B., and Nisbet, D.J. (2005) Fecal Prevalence and
Diversity of Salmonella Species in Lactating Dairy Cattle in Four States* Journal of Dairy Science 88: 3603-3608.
54
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Callaway, T.R., Edrington, T.S., Brabban, A.D., Keen, J.E., Anderson, R.C., Rossman, M.L., Engler, M.J., Genovese, K.J.,
Gwartney, B.L., Reagan, J.O., Poole, T.L., Harvey, R.B., Kutter, E.M., and Nisbet, D.J. (2006) Fecal prevalence of Escherichia
coli O157, Salmonella, Listeria, and bacteriophage infecting E. coli O157:H7 in feedlot cattle in the Southern Plains region of the
United States. . Foodborne Pathogens and Disease 3: 234-244.
Callaway, T.R., Edrington, T.S., Anderson, R.C., Byrd, J.A., and Nisbet, D.J. (2008) Gastrointestinal microbial ecology and the
safety of our food supply as related to Salmonella Journal of Animal Science 86: E163-E172.
CDC (2006). Salmonella annual summary. URL www.cdc.gov/ncidod/dbmd/phlisdata/salmonella.htm. (2009).
Chapagain, P.P., Kessel, J.S.V., Kams, J.S., Wolfgang, D.R., Hovingh, E., Nelen, K.A., Schukken, Y.H., and Grohn, Y.T. (2008)
A mathematical model of the dynamics of Salmonella
Cerro infection in a US dairy herd. Epidemiol. Infect. 136: 263-272.
Chapman, P.A., Siddons, C.A., Cerdan Malo, A.T., and Harkin, MA. (1997) A 1-year study of Escherichia coli O157 in cattle,
sheep, pigs and poultry. Epidemiology and Infection 119: 245-250.
Characklis, G.W., Dilts, M.J., III, O.D.S., Likirdopulos, C.A., Krometis, L.-A.H., and Sobsey, M.D. (2005) Microbial
partitioning to settleable particles in stormwater. Water Research 39: 1773-1782.
Chase-Topping, M., Gaily, D., Low, C., Matthews, L., and Woolhouse, M. (2008) Super-shedding and the link between human
infection and livestock carriage of Escherichia coli O157. Nature 6: 904-912.
Chase-Topping, M.E., McKendrick, I.J., Pearce, M.C., MacDonald, P., Matthews, L., Halliday, J., Allison, L., Fenlon, D., Low,
J.C., Gunn, G., and Woolhouse, M.E.J. (2007) Risk Factors for the Presence of High-Level Shedders of Escherichia coli O157 on
Scottish Farms Journal of Clinical Microbiology 45: 1594-1603.
Chen, L., Geys, H., Cawthraw, S., Havelaar, A., and Teunis, P. (2006) Dose response for infectivity of several strains of
Campylobacterjejuni in chickens. Risk Analysis 26: 1613-1621.
Coleman, M., and Marks, H. (1998) Topics in dose-response modeling. J FoodProt 61: 1550-1559.
Coleman, M., and Marks, H. (2000) Mechanistic Modeling of Salmonellosis. Quantitative Microbiol 2: 227-247.
Coleman, M.E., Marks, H.M., Golden, N. J., and Latimer, H.K. (2004) Discerning strain effects in microbial dose-response data. J
Toxicol Environ Health-Part a-Current Issues 67: 667-685.
Comick, N.A., and Helgerson, A.F. (2004) Transmission and Infectious Dose of Escherichia coli O157:H7 in Swine. Applied
and Environmental Microbiology 70: 5331-5335.
Cox, N.A., Stem, N.J., Musgrove, M.T., Bailey, J.S., Craven, S.E., Cray, P.F., Buhr, R.J., and Hiett, K.L. (2002) Prevalence and
level of Campylobacter in commercial broiler breeders (parents) and broilers. Journal of Applied Poultry Research 11: 187-.
Cox, P., Griffith, M., Angles, M., Deere, D., and Ferguson, C. (2005) Concentrations of Pathogens and Indicators in Animal
Feces in the
Sydney Watershed. Applied and Environmental Microbiology 71: 5929-5934.
Davies-Colley, R.J., Nagels, J.W., Smith, R.A., Young, R.G., and Phillips, CJ. (2004) Water quality impact of a dairy cowherd
crossing a stream. New Zealand Journal of Marine and Freshwater Research 38: 569-576.
Davies, C., Kaucner, C., Altavilla, N., Ashbolt, N., Hijnen, W., Medema, G., Deere, D., Krogh, M., and Ferguson, C. (2005a)
Fate and transport of surface water pathogens in watersheds. Report 91078F. Denver: American Water Works Research
Foundation.
Davies, C.M., Ferguson, C.M., Kaucner, C., Krogh, M., Altavilla, N., Deere, D.A., and Ashbolt, NJ. (2004) Dispersion and
transport of Cryptosporidium oocysts from fecal pats under simulated rainfall events. Applied and Environmental Microbiology
70: 1151-1159.
55
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Davies, C.M., Altavilla, N., Krogh, M., Ferguson, C.M., Deere, D.A., and Ashbolt, NJ. (2005b) Environmental inactivation of
Cryptosporidium oocysts in catchment soils. Journal of Applied Microbiology 98: 308-317.
de Graaf, D.C., Vanopdenbosch, E., M., O.-M.L., Abbassi, H., and Peeters, J., E, (1999) A review of the importance of
crypto sporidio sis in farm animals. InternationalJournal ofParasitology 29: 1269-1287.
Donkersgoed, J.V., Graham, T., and Gannon, V. (1999) The prevalence of verotoxins, Escherichia coli 0157:H7, and Salmonella
in the feces and rumen of cattle at processing. Canadian Veterinary Journal 40: 332-338.
Dorner, S.M., Huck, P.M., and Slawson, R.M. (2004) Estimating Potential Environmental Loadings of Cryptosporidium spp. and
Campylobacter spp. from Livestock in the Grand River Watershed, Ontario, Canada. Environmental Science and Technology 38:
3370-3380.
Dorr, P.M., Tadesse, D.A., Zewde, B.M., Fry, P., Thakur, S., and Gebreyes, W.A. (2009) Longitudinal study of Salmonella
dispersion and the role of environmental contamination in commercial swine production systems. Applied and Environmental
Microbiology 75: 1478-1486.
Dougherty, M., Geohring, L.D., and Wright, P. (1998) Liquid Manure Application Systems Design Manual. Northeast Regional
Agricultural Engineering Service Ithaca, NY NRAES-89
Dufour, A.P., Evans, O., Behymer, T.D., and Canui, R. (2006) Water ingestion during swimming activities in a pool: A pilot
study. Journal of Water and Health 4: 425-430.
E. Tumova, E., Skfivan, M., Marounek, M., Pavlasek, I., and Ledvinka, Z. (2002) Performance and oocyst shedding in broiler
chickens orally infected with Cryptosporidium baileyi and Cryptosporidium meleagridis Avian Diseases 46.
Edrington, T.S., Hume, M.E., Looper, M.L., Schultz, C.L., Fitzgerald, A.C., Callaway, T.R., Genovese, K.J., Bischoff, K.M.,
McReynolds, J.L., Anderson, R.C., and Nisbet, D.J. (2004) Variation in the faecal shedding of Salmonella and E. coli O157:H7
in lactating dairy cattle and examination of Salmonella genotypes using pulsed-field gel electrophoresis. Letters in Applied
Microbiology 2004 38: 366-372.
Eghball, B., and Power, J.F. (1998) Management of Manure From Beef Cattle inFeedlots and From Minor Classes of Livestock.
Agricultural Uses of Municipal, Animal, and Industrial Byproducts USDA ARS Washington DC Conservation Res. report
number 44
El-Shibiny, A., Connerton, P.L., and Connerton, I.F. (2005) Enumeration and Diversity of Campylobacters and Bacteriophages
Isolated during the Rearing Cycles of Free-Range and Organic Chickens. Applied and Environmental Microbiology 71: 1259-
1266.
Elder, R.O., Keen, J.E., Siragusa, G.R., Barkocy-Gallagher, G.A., Koohmaraie, M., and Laegreid, a.W.W. (2000) Correlation of
enterohemorrhagic Escherichia coli O157 prevalence in feces, hides, and carcasses of beef cattle during processing. Proceedings
of the National Academy of Science USA 97: 2999-3003.
Faubert, G.M., and Litvinsky, Y. (2000) Natural transmission of Cryptosporidium parvum between dams and calves on a dairy
farm. Journal ofParasitology 86: 495-500.
Payer, R., Morgan, U., and Upton, S.J. (2000) Epidemiology of Cryptosporidium: transmission, detection and identification
[Review]. International Journal for Parasitology 30: 1305-1322.
Payer, R., Santin, M., Trout, J.M., and Greiner, E. (2006) Prevalence of species and genotypes of Cryptosporidium found in 1-2-
year-old dairy cattle in the eastern United States. Veterinary Parisitology 135: 105-112.
Fegan, N., Vanderlinde, P., Higgs, G., and Desmarchelier, P. (2004) The prevalence and concentration of Escherichia coli O157
in faeces of cattle from different production systems at slaughter. Journal of Applied Microbiology 97: 362-370.
Ferguson, C., Charles, K., and Deere, D. (2009) Quantification of Microbial Sources in Drinking-Water Catchments. Critical
Reviews in Environmental Science and Technology 39: 1-40.
56
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Ferguson, C.M., Davies, C.M., Kaucner, C., Krogh, M., Rodehutskors, J., Deere, D.A., and Ashbolt, NJ. (2007) Field scale
quantification of microbial transport from bovine faeces under simulated rainfall events. Journal of Water and Health 5: 83-95.
Gagliardi, J.V., and Karns, J.S. (2000) Leaching of Escherichia coli O157:H7 in diverse soils under various agricultural
management practices. Applied and Environmental Microbiology 66: 877-883.
Garber, L.P., Smeltzer, M., Fedorka-Cray, P., Ladely, S., and Ferris, K. (2003) Salmonella enterica serotype enteritidis in table
egg layer house environments and in mice in U.S. layer houses and associated risk factors. Avian Diseases 47: 134-142.
Gibbs, R.A., Hu, C.J., Ho, G.E., and Unkovich, I. (1997) Regrowth of faecal coliforms and salmonellae in stored biosolids and
soil amended with biosolids. Water Science and Technology 35: 269-275.
Goss, M., and Richards, C. (2008) Development of a risk-based index for source water protection planning, which supports the
reduction of pathogens from agricultural activity entering water resources. Journal of Environmental Management 87: 623-632.
Greene, C., Dimitri, C., Lin, B.-H., Oberholtzer, L., and Smith, T. (2009) Emerging Issues in the U.S. Organic Industry. USDA
Economic Research Service Washington DC Economic Information Bulletin number 55
Guan, T.Y., and Holley, R.A. (2003) Pathogen survival in swine manure environments and transmission of human enteric illness-
-A review. Journal of Environmental Quality 32: 383-392.
Guselle, N.J., Applebee, A.J., and Olson, M.E. (2003) Biology of Cryptosporidium parvum in pigs: from weaning to market.
Veterinary Parasitology 113: 7.
Haas, C.N. (1983) Estimation of risk due to low doses of microorganisms: A comparison of alternative methodologies. Am J
Epidemiol 55: 573-582.
Haas, C.N., Rose, J.B., and Gerba, C.P. (1999) Quantitative microbial risk assessment. J.W. Wiley, Inc.
Hancock, D.D., Rice, D.H., Thomas, LA., Dargatz, D.A., and Besser, T.E. (1997) Epidemiology of Escherichia coli O157 in
Feedlot Cattle. Journal of Food Protection 60: 462-465.
Havelaar, A., and Melse, J.M. (2003) Quantifying public health risk in the WHO Guidelines for Drinking-Water Quality: a
burden of disease approach. PJVMraport 734301022. Bilthoven: RIVM.
Havelaar, A.H., van Duynhoven, Y.T.H.P., Nauta, M.J., Bouwknegt, M., Heuvelink, A.E., de Wit, G.A., Nieuwenhuizen,
M.G.M., and van de Kar, N.C.A. (2003) Disease burden in the Netherlands due to infections -with Shiga-toxin producing
Escherichia coli O157. RIVM rapport 284550008. Bilthoven: RIVM.
Hayes, J.R., Carr, L.E., Mallinson, E.T., Douglass, L.W., and Joseph, S.W. (2000) Characterization of the contribution of water
activity and moisture content to the population distribution of Salmonella spp. in commercial poultry houses. Poultry Science 79:
1557-1561.
Heitman, T.L., Frederick, L.M., Viste, J.R., Guselle, N.J., Morgan, U.M., Thompson, R.C.A., and Olson, M.E. (2002) Prevalence
of Giardia and Cryptosporidium and characterization of Cryptosporidium spp. isolated from wildlife, human, and agricultural
sources in the North Saskatchewan River Basin in Alberta, Canada. Canadian Journal of Microbiology 48: 530-541.
Himathongkham, S., Bahari, S., Riemann, H., and Cliver, D.O. (1999) Survival of Escherichia coli O157:H7 and Salmonella
typhimurium in cowmanure and cowmanure slurry. FEMS Microbiology Letters 178: 251-257.
Himathongkham, S., Riemann, H., Bahari, S., Nuanualsuwan, S., Kass, P., and Cliver, D.O. (2000) Survival of Salmonella
typhimurium and Escherichia coil O157:H7 in poultry manure and manure slurry at sublethal temperatures. Avian Diseases 44:
853-860.
Hoar, B., Atwill, E.R., Elmi, C., and Farver, T.B. (2001) An examination of risk factors associated with beef cattle shedding
pathogens of potential zoonotic concern. Epidemiology and Infection 127: 147-15 5.
Hrudey, S.E., and Hrudey, EJ. (2004) Safe Drinking Water. Lessons from Recent Outbreaks in Affluent Nations. London: IWA
Publishing.
57
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Hubbard, R.K., and Lowrance, R.R. (1998) Management of Dairy Cattle Manure. Agricultural Uses of Municipal, Animal, and
Industrial Byproducts USDA ARS Washington DC Conservation Res. report number 44
Huston, C.L., Wittum, T.E., Love, B.C., and Keen, J.E. (2002) Prevalence of fecal shedding of Salmonella spp in dairy herds.
Journal of the American Veterinary Medical Association 220: 645-649.
Hutchins, S.R., Walters, L.D., Avery, S.M., Synge, B.A., and Moore, A. (2004) Levels of zoonotic agents in British livestock
manures. Letters in Applied Microbiology 39: 207-214.
Hutchison, M.L., Walters, L.D., Avery, S.M., and Moore, A. (2005a) Decline of zoonotic agents in livestock waste and bedding
heaps. Journal of Applied Microbiology 99: 354-362.
Hutchison, M.L., Walters, L.D., Avery, S.M., Munro, F., and Moore, A. (2005b) Analyses of livestock production, waste storage,
and pathogen levels and prevalences in farm manures. Applied and Environmental Microbiology 71: 1231-1236.
Hutchison, M.L., Walters, L.D., Moore, A., and Avery, S.M. (2005d) Declines of zoonotic agents in liquid livestock wastes
stored in batches on-farm. Journal of Applied Microbiology 99: 58-65.
Hutchison, S.R., Walters, L.D., Avery, S.M., Synge, B.A., and Moore, A. (2004) Levels of zoonotic agents in British livestock
manures. Letters in Applied Microbiology 39: 207-214.
ILSI (1996) A conceptual framework to assess the risks of human disease following exposure to pathogens. Risk Anal 16:
841:848.
ILSI (2000) Revised framework for microbial risk assessment. ILSI Press Washington, DC.
Ives, R.L., Kamarainen, A.M., John, D.E., and Rose, J.B. (2007) Use of Cell Culture To Assess Cryptosporidium parvum
Survival Rates in Natural Groundwaters and Surface Waters. Applied and Environmental Microbiology 73: 5968-5970.
Jamieson, R.C., Joy, D.M., Lee, H., Kostaschuk, R., and Gordon, R.J. (2005) Resuspension of Sediment-Associated Escherichia
coli in a Natural Stream. Journal of Environmental Quality 34: 403-407.
Jenkins, M.B., Bowman, D.D., and Chiorse, W.C. (1998) Inactivation of Cryptosporidium parvum oocysts by ammonia. Applied
and Environmental Microbiology 64: 784-788.
Jenkins, M.B., Walker, M.J., Bowman, D.D., Anthony, L.C., and Ghiorse, W.C. (1999) Use of a sentinel system for field
measurements of Cryptosporidium parvum oocyst inactivation in soil and animal waste. Applied and Environmental
Microbiology 65: 1998-2005.
Jones, K., Betaieb, M., and Telford, D.R. (1990) Seasonal variation of thermophilic campylobacters in sewage sludge. Journal of
Applied Bacteriology 69: 185-189.
Jones, K. (2001) Campylobacters in water, sewage and the environment. Journal of Applied Microbiology 90: S68-S79.
Joshua, W.D., Michalk, D.L., Curtis, I.H., Salt, M., and Osbome, G.J. (1998) The potential for contamination of soil and surface
waters from sewage sludge (biosolids) in a sheep grazing study, Australia Geoderma 84: 135-156.
Kasomdorkbua, C., Opriessnig, T., Huang, F.F., Guenette, O.K., Thomas, P.J., Meng, X.-J., and Halbur, P.G. (2005) Infectious
Swine Hepatitis E Virus Is Present in Pig Manure Storage Facilities on United States Farms, but Evidence of Water
Contamination Is Lacking. Appl. Envir. Microbiol. 71: 7831-7837.
Kelley, T.R., Pancorbo, O.C., Merka, W.C., Thompson, S.A., Cabrera, M.L., and Bamhart, H.M. (1994) Fate of selected bacterial
pathogens and indicators in fractionated poultry litter during storage Journal of Applied Poultry Research 3.
Kuczynska, E., Daniel R. Shelton, D.R., and Pachepsky, Y. (2005a) Effect of bovine manure on Cryptosporidium parvum oocyst
attachment to soil. Applied and Environmental Microbiology 71.
Kuczynska, E., Shelton, D.R., and Pachepsky, Y. (2005b) Effect of Bovine Manure on Cryptosporidium parvum Oocyst
Attachment to Soil Applied and Environmental Microbiology 71: 6394-6397.
58
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Larsen, R.E., Miner, J.C., Buckhouse, J.C., and Moore, J.A. (1994) Water-quality benefits of having cattle manure deposited
away from streams. Bioresource and Technology 48: 113-118.
LeJeune, J.T., Besser, T.E., Rice, D.H., Berg, J.L., Stilbom, R.P., and Hancock, D.D. (2004) Longitudinal study of fecal shedding
of Escherichia coli O157:H7 in feedlot cattle: predominance and persistence of specific clonal types despite massive cattle
population turnover. Applied and Environmental Microbiology 70: 377-384.
Ley, D.H., Levy, M.H., Hunter, L., Corbett, W., and H.J., B. (1988) Cryptosporidia- positive rates of avian necropsy accessions
determined by examination of auramine O-stained fecal smears. Avian Dis 32: 108-113.
Lorenzo, M.J., Ares-Mazas, E., and Villacorta Martinez de Maturana, I. (1993) Detection of oocysts and IgG antibodies to
Cryptosporidiumparvum in asymptomatic adult cattle. Veterinary Parasitology 47: 9-15.
Matthews, L., Low, L.C., Gaily, D.L., Pearce, M.C., Mellor, D.J., Heesterbeek, J.A.P., Chase-Topping, M., Naylor, S.W., Shaw,
D.J., S.W.J.Reid, G.J.Gunn, and Woolhouse, M.E.J. (2006) Heterogeneous shedding of Escherichia coli O157 in cattle and its
implicatons for control. The National Academy of Sciences of the USW103: 547-552.
Meals, D.W., and Braun, D.C. (2006) Demonstration of methods to reduce E. coli runoff from dairy manure application sites.
Journal of Environmental Quality 35: 1088-1100.
Medema, G.J., Teunis, P.F.M., Gornik, V., Havelaar, A.H., and Exner, M. (1995) Estimation of the Cryptosporidium infection
risk via drinking water. In Protozoan Parasites and Water. Cambridge, UK: Royal Society of Chemistry.
Medema, G.J., Teunis, P.F.M., Havelaar, A.H., and Haas, C.N. (1996) Assessment of the dose-response relationship of
Campylobacterjejuni. International Journal of Food Microbiology 30: 101-111.
Medema, G.J., Ketelaars, H.A.M., Hoogenboezem, W., Rijs, G.B.J., and Schijven, J.F. (2001) Cryptosporidium and Giardia'.
Occurrence in sewage, manure and surface water. Association of River Waterworks - RIWA, The Netherlands
Mendez-Hermida, F., Castro-Hermida, J.A., Ares-Mazas, E., Kehoe, S.C., and McGuigan, K.G. (2005) Effect of batch-process
solar disinfection on survival of Cryptosporidium parvum oocysts in drinking water. Applied and Environmental Microbiology
71: 1653-1654.
Moffet, C.A., Zartman, R.E., Wester, D.B., and Sosebee, R.E. (2005) Surface Biosolids Application: Effects on Infiltration,
Erosion, and Soil Organic Carbon in Chihuahuan Desert Grasslands and Shrublands. Journal of Environmental Quality 34: 299-
311.
Muirhead, R.W., Davies-Colley, R.J., Donnison, A.M., and Nagels, J.W. (2004) Faecal bacteria yields in artificial flood events:
quantifying in-stream stores Water Research 38: 1215-1224.
Muirhead, R.W., Collins, R.P., and Breme, PJ. (2005) Erosion and subsequent transport state of Escherichia coli from cowpats.
Applied and Environmental Microbiology 71: 2875-2879.
Muirhead, R.W., Collins, R.P., and Bremer, PJ. (2006a) The association of E. coli and soil particles in overland flow Water
Science and Technology 54: 153-159.
Muirhead, R.W., Collins, R.P., and Bremer, PJ. (2006b) Interaction of Escherichia coli and soil particles in runoff. Applied and
Environmental Microbiology 72: 3406-3411.
Muirhead, R.W., Collins, R.P., and Bremer, PJ. (2006c) Numbers and transported state of Escherichia coli in runoff direct from
fresh cowpats under simulated rainfall. Letters in Applied Microbiology 42: 83-87.
Myint, K.S., and Gibbons, R.V. (2008) Hepatitis E: A neglected threat. Trans. R. Soc. Trap. Med. Hyg. 102
211-212.
National Research Council (1983) Risk assessment in the federal government, managing the process. National Academy Press
Washington, D.C.
59
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Nauta, M., van der Fels-Klerx, I., and Havelaar, A. (2005) A poultry-processing model for quantitative microbiological risk
assessment. Risk Anal 25: 85-98.
Nicholson, FA., Groves, S.J., and Chambers, B.J. (2005) Pathogen survival during livestock manure storage and following land
application. Bioresource Technology 96: 135-143.
Nizeyi, I, Cranfield, M., and Graczyk, T. (2004) Cattle near the Bwindi Impenetrable National Park, Uganda, as a reservoir of
Cryptosporidium parvum and Giardia duodenalis for local community and free-ranging gorillas Parasitology Research 88: 380-
385.
Nydam, D.V., Wade, S.E., Schaaf, S.L., and Mohammed, H.O. (2001) Number of Crypto sporidium parvum oocysts or Giardia
spp cysts shed by dairy calves after natural infection. American Journal of Veterinary Research 62: 1612-1615.
Olson, M.E., Goh, J., Phillips, M., Guselle, N., and McAllister, T.A. (1999) Giardia cyst and Cryptosporidium oocyst survival in
water, soil, and cattle feces. Journal of Environmental Quality28: 1991-1996.
Pangloli, P., Dje, Y., Ahmed, O., Doane, C.A., Oliver, S.P., and Draughon, F.A. (2008) Seasonal Incidence and Molecular
Characterization of Salmonella from Dairy Cows, Calves and Farm Environment. Foodborne Pathogens and Disease 5.
Pepper, I.L., Josephson, K.L., Bailey, R.L., Burr, M.D., and Gerba, C.P. (1993) Survival of indicator organisms in Sonoran desert
soil amended with sewage sludge. Journal of Environmental Science & Health Part A: Environmental Science and Engineering
28: 1287-1302.
Pereira, M.C., Atwill, E.R., and Jones, T. (1999) Comparison of sensitivity of immunofluorescent microscopy to that of a
combination of immunofluorescent microscopy and immunomagnetic separation for detection of Cryptosporidium parvum
oocysts in adult bovine feces. Applied and Environmental Microbiology 65: 3236-3239.
Quilez, J., Sanchez-Acedo, C., del Cacho, E., and Causape, A.C. (1996) Prevalence of Cryptosporidium and Giardia infections in
cattle in Aragon (Northeastern Spain). Veterinary Parasitology 66: 139-146.
Ramirez, N.E., Ward, L.A., and Sreevatsan, S. (2004) A review of the biology and epidemiology of cryptosporidiosis in humans
and animals. Microbes and Infection 6.
Regli, S., Rose, J.B., Haas, C.N., and Gerba, C.P. (1991) Modeling the risk from Giardia and viruses in drinking water. J
Am.Water.Works.Assoc. 83: 76.
Robinson, S.E., Wright, E.J., Hart, C.A., Bennett, M., and French, N.P. (2004) Intermittent and persistent shedding of
Escherichia coli O157 in cohorts of naturally infected calves. Journal of Applied Microbiology 97: 1047-1053.
Rodriguez, A., Pangloli, P., Richards, H.A., Mount, J.R., and Draughon, F.A. (2006) Prevalence of Salmonella in Diverse
Environmental Farm Samples. Journal of Food Production 69: 2576-2580.
Rose, J.B., Haas, C.N., and Regli, S. (1991) Risk assessment and control of waterborne giardiasis. Am J Public Health 81: 709-
713.
Rosen, B.H. (2000) Waterborne Pathogens in Agricultural Watershes. Watershed Science Institute, National Resouces
Conservation Service Ithaca, NY
Sanchez, J., Dohoo, I.R., Christensen, J., and Rajic, A. (2007) Factors influencing the prevalence of Salmonella spp. in swine
farms: A meta-analysis approach
Preventive Veterinary Medicine
81.
Santin, M., Trout, J., Xiao, L., Zhou, L., Greiner, E., and Payer, R. (2004) Prevalence and age-related variation of
Cryptosporidium species and genotypes in dairy calves. Veterinary Parisitology 122: 103-117.
Scott, C.A., Smith, H.V., Mtambo, M.A., and Gibbs, H.A. (1995) An epidemiological study of Cryptosporidium parvum in two
herds of adult beef cattle. Veterinary Parasitology 57: 277-288.
60
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Searcy, K.E., Packman, A.I., Atwill, E.R., and Harter, T. (2006) Deposition of Cryptosporidium Oocysts in Streambeds. Applied
and Environmental Microbiology 72: 1810-1816.
Seto, E.Y., Seller, J.A., and Colford, J.M., Jr. (2007) Strategies to reduce person-to-person transmission during widespread
Escherichia coli O157:H7 outbreak. Emerg InfectDis 13: 860-866.
Shere, J.A., Kaspar, C.W., Bartlett, K.J., Linden, S.E., Norell, B., Francey, S., and Schaefer, D.M. (2002) Shedding of
Escherichia coli O157:H7 in dairy cattle housed in a confined environment following waterborne inoculation. . Applied and
Environmental Microbiology
68: 1947-1954.
Sinton, L., Hall, C., and Braithwaite, R. (2007) Sunlight inactivation of Campylobacter jejuni and Salmonella enterica, compared
to Escherichia coli, in seawater and river water . . Journal of Water and Health 5: 357-365.
Smith, D., Blackford, M., Younts, S., Moxley, R., Gray, I, Hungerford, L., Milton, T., and Klopfenstein, T. (2001) Ecological
relationships between the prevalence of cattle shedding Escherichia coli O157:H7 and characteristics of the cattle or conditions of
the feedlot pen. Journal of Food Protection 64: 1899-1903.
Seller, J.A., Olivieri, A., Crook, I, Parkin, R., Spear, R., Tchobanoglous, G., and Eisenberg, J.N.S. (2003) Risk-based approach
to evaluate the public health benefit of additional wastewater treatment. Environ Sci Technol37\ 1882-1891.
Seller, J.A., Eisenberg, I, DeGeorge, I, Cooper, R., Tchobanoglous, G., and Olivieri, A. (2006) A public health evaluation of
recreational water impairment. J Water Health 4: 1-19.
Seller, J.A., Seto, E.Y., and Olivieri, A.W. (2007) Application of microbial risk assessment techniques to estimate risk due to
exposure to reclaimed waters. WateReuse Foundation, Final project report WRF-04-011
Seller, JA. (2009) Potential implications of person-to-person transmission of viral infection to US EPA's Groundwater Rule. J
.WaterHealth 7: 208-223.
Stanley, K., and Jones, K. (2003) Cattle and Sheep farms as reservoirs of Campylobacter. Journal of Applied Microbiology 94:
104S-113S.
Stanley, K.N., Wallace, J.S., Currie, J., Diggle, P., and Jones, K. (1998) The seasonal variation of thermophilic campylobacters in
cattle Journal of Applied Microbiology 85: 472-480.
Starkey, S.R., Wade, S.E., Schaaf, S., and Mohammed, H.O. (2005) Incidence of Cryptosporidium parvum in dairy cattle
population in a New York City Watershed. Veterinary Parisitology 131: 197-205.
Stromberg, B.E. (1997) Environmental factors influencing transmission
Veterinary Parisitology 72: 247-264.
Tei, S., Kitajima, N., Takahashi, K., and Mishiro, S. (2003) Zoonotic transmission of hepatitis E virus from deer to human
beings. .Lancet362\ 371-373.
Teunis, P.F., van der Heijden, O.G., van der Giessen, J.W.B., and Havelaar, A.H. (1996) The dose-response relation in human
volunteers for gastro-intestinal pathogens. EIVMReport 284550002.
Teunis, P.F., Ogden, I.D., and Strachan, N.J. (2008) Hierarchical dose response of E. coli O157:H7 from human outbreaks
incorporating heterogeneity in exposure. Epidemiol Infect 136: 761-770.
Teunis, P.F.M., Ever, E.G., and Slob, W. (1999a) Analysis of variable fractions resulting from microbial counts. Quantitative
Microbiology 1: 63-88.
Teunis, P.F.M., Nagelkerke, N.J.D., and Haas, C.N. (1999b) Dose response models for infectious gastroenteritis. Risk Analysis
19: 1251-1260.
Teunis, P.F.M., van den Brandhof, W., Nauta, M.J., Wagenaar, J., van den Kerkhof, H., and Van Pelt, W. (2005) A
reconsideration of the Campylobacter dose-response relation. Epidemiology and Infection 133: 583-592.
61
-------
U.S. EPA Office of Water QMRA Model for Animal-Impacted Waters
Thelin, R., and Gifford, G.F. (1983) Fecal coliform release patterns from fecal material of cattle. Journal of Environmental
Quality 12: 57-63.
Thompson, R.C.A., Palmer, C.S., and Ryan O'Handley, R. (2008) The public health and clinical significance of Giardia and
Cryptosporidium in domestic animals. The Veterinary Journal 117: 18-25.
U.S. EPA (2006) National primary drinking water regulations: Long term 2 enhanced surface water treatment rule (LT2ESWTR);
final rule. 40CFR Parts 9, 141 and 142, volume 71, Number 654, January 5, 2006.
USDA FSIS (2009). Serotypes Profile of Salmonella Isolates from Meat and Poultry Products. January 1998 through December
2007. URL www.fsis.usda.gov/Science/Serotvpes_Profile_Salmonella_Isolates/index.asp. (2009).
Van Hoorebeke, S., Van Immerseel, F., De Vylder, J., Ducatelle, R., Haesebrouck, F., Pasmans, F., de Kruif, A., and Dewulf, J.
(2009) Faecal Sampling Underestimates the Actual Prevalence of Salmonella in Laying Hen Flocks. Zoonoses Public Health.
Van Kessel, J.S., Karns, J.S., Wolfgang, D.R., Hovingh, E., and Schukken, Y.H. (2007a) Longitudinal Study of a Clonal,
Subclinical Outbreak of Salmonella enterica subsp. enterica Serovar Cerro in a U.S. Dairy Herd. Foodborne Pathogens and
Diseased. 449-461.
Van Kessel, J.S., Pachepsky, Y.A., Shelton, D.R., and Karns, J.S. (2007b) Survival of Escherichia coli in cowpats in pasture and
in laboratory conditions. Journal of Applied Microbiology 103: 1122-1127.
Vinten, A.J.A., Douglas, J.T., Lewis, D.R., Aitken, M.N., and Fenlon, D.R. (2004) Relative risk from surface water pollution by
E. coli derived from faeces of grazing animals compared to slurry application. Soil Use and Managment 20: 13-22.
Vogt, R.L., Sours, H.E., and Barrett, T. (1982) Campylobacter enteritis associated with contaminated water. Annals of Internal
Medicine 96: 292-296.
Wade, S.E., Mohammed, H.O., and Schaaf, S.L. (2000) Prevalence of Giardia sp, Cryptosporidium parvum and Cryptosporidium
muris (C. Andersoni) in 109 dairy herds in five counties of southeastern New York. Veterinary Parasitology 93: 1-11.
Weijtens, M.J.B.M., van der Plas, J., Bijker, P.G.H., Urlings, H.A.P., Koster, D., van Logtestijn, J.G., and Huis in't Veld, J.H.J.
(1997) The transmission of campylobacter in piggeries; an epidemiological study. Journal of Applied Microbiology 83: 693-698.
Weijtens, M.J.B.M., Reinders, R.D., Urlings, H.A.P., and Van der Plas, J. (1999) Campylobacter infections in fattening pigs;
excretion pattern and genetic diversity. Journal of Applied Microbiology 86: 63-70.
Wells, S.J., Fedorka-Cray, P.J., Dargatz, D.A., Ferris, K., and Green, A. (2001) Fecal shedding of Salmonella spp. by dairy cows
on farm and at cull cow markets, international journal of food protection 64: 3-11.
Wesley, I.V., Wells, S.J., Harmon, K.M., Green, A., Schroeder-Tucker, L., Glover, M., and Siddique, I. (2000) Fecal shedding of
Campylobacter and Arcobacter spp. in dairy cattle. Applied and Environmental Microbiology 66: 1994-2000.
Westrell, T., Scheming, C., Stenstrom, T.A., and Ashbolt, N.J. (2003) Integration of QMRA and HACCP for management of
pathogens in wastewater and sewage sludge treatment and reuse. In 4th International Symposium on Wastewater Reclamation
and Reuse, Nov 12-14. Mexico City: International Water Association, p. (in press).
WHO (1999) Principles and Guidelines for the Conduct of Microbiological Risk Assessment. CAC/GL-30. World Health
Organization Geneva, Switzerland WHO/SDE/WSH/99.1
Widiasih, D.A., Ido, N., Omoe, K., Sugii, S., and Shinagawa, K. (2004) Duration and magnitude of aecal shedding of Shiga
toxin-producing Escherichia coli from naturally infected cattle. Epidemiol Infect 132: 67-75.
Xiao, L.H., Bern, C., Arrowood, M., Slulaiman, L, Zhou, I.L., Kawai, V., Vivar, A., Lai, A.A., and Oilman, R.H. (2002)
Identification of the Cryptosporidium pig genotype in a human patient. Journal of Infectious Diseases 185: 1846-1848.
Zaleski, K.J., Josephson, K.L., Gerba, C.P., and Pepper, I.L. (2005) Potential regrowth and recolonization of salmonellae and
indicators in biosolids and biosolid-amended soil. Applied and Environmental Microbiology 71: 3701-3708.
62
-------
ANNEX 3
Distribution and Prevalence of Selected Zoonotic Pathogens in
U.S. Domestic Livestock
For
Quantitative Microbial Risk Assessment to
Estimate Illness in Freshwater Impacted by
Agricultural Animal Sources of Fecal Contamination
U.S. Environmental Protection Agency
December 2010
-------
DISTRIBUTION AND PREVALENCE OF SELECTED
ZOONOTIC PATHOGENS IN U.S. DOMESTIC LIVESTOCK
U.S. Environmental Protection Agency
Office of Water
Health and Ecological Criteria Division
August 2010
-------
U.S. Environmental Protection Agency
Disclaimer
Mention of commercial products, trade names, or services in this document or in the references
and/or footnotes cited in this document does not convey, and should not be interpreted as
conveying official EPA approval, endorsement, or recommendation.
August 2010
-------
Table of Contents
DISCLAIMER i
TABLE OF CONTENTS n
LIST OF TABLES m
LIST OF FIGURES m
ACRONYMS iv
EXECUTIVE SUMMARY 1
1. INTRODUCTION 5
1.1. Purpose 5
1.2. Background 5
1.3. Approach 7
2. DISTRIBUTION OF LIVESTOCK IN THE UNITED STATES 9
2.1. Methods 9
2.2. Cattle Distribution 10
2.2.1. All cattle 10
2.2.2. Milk cows 10
2.2.3. Cattle on feed 10
2.3. Swine Distribution 11
2.4. Chicken Distribution 11
2.5. Summary 11
3. PATHOGEN OCCURRENCE IN LIVESTOCK 17
3.1. Key Zoonotic Hosts 18
3.2. Occurrence and Abundance of Reference Pathogens in U.S. Cattle, Swine, and Chicken 20
3.3. Large-Scale Studies of Pathogen Occurrence in U.S. Livestock 22
3.3.1. Large-scale studies of Salmonella prevalence 29
3.3.2. Large-scale studies of Campylobacter prevalence 30
3.3.3. Large-scale studies of E. coli O157:H7 prevalence 30
3.3.4. Large -scale studie s of Cryptosporidium and Giardia prevalence 30
3.4. Summary 31
4. FARM FACTORS AND THE OCCURRENCE OF PATHOGENS IN LIVESTOCK MANURES 32
4.1. Farm Factors with Regional Implications 33
4.1.1. Farm size 34
4.1.2. Operation type 35
4.1.3. Longitudinal (life stage) studies 38
4.1.4. Seasonality 40
4.2. Farm Factors without Regional Implications 42
4.2.1. Water disinfection and hygiene 43
4.2.2. Mixed production 43
4.2.3. BMPs and manure management 44
4.3. Summary 50
5. REFERENCES 52
APPENDIX A. LITERATURE SEARCH STRATEGY AND RESULTS A-l
APPENDIX B. OCCURRENCE DATA B-l
APPENDIX C. ABUNDANCE DATA C-l
APPENDIX D. FARM FACTORS DATA D-l
August 2010 ii
-------
U.S. Environmental Protection Agency
List of Tables
Table 1. Cryptosporidium species and associated maj or and minor hosts 18
Table 2. Significant Giardia species and genotypes and associated hosts 19
Table 3. Significant Campylobacter species and genotypes and associated hosts 20
Table 4. Frequently-encountered Salmonella serotypes for select hosts 20
Table 5. Prevalence (occurrence) of human infectious species of pathogens in livestock
manures 22
Table 6. Abundance of human infectious pathogens in livestock manures 23
Table 7. Review of large-scale studies of Salmonella prevalence on livestock operations 24
Table 8. Large-scale studies of Campylobacter prevalence among U.S. livestock
operations 26
Table 9. Review of large-scale studies of E. coli O157 prevalence on livestock
operations 27
Table 10. Review of large-scale studies of Cryptosporidium prevalence on livestock
operations 28
Table 11. Review of a large-scale study of Giardia prevalence on livestock operations 29
Table 12. Cryptosporidium shedding 39
Table 13. Effect of manure management options on the number of microorganisms
contained in manure 46
Table 14. Typical reductions of pathogens during manure treatment processes 47
Table 15. Typical reductions of viruses during animal waste treatment processes 47
Table 16. Dairy farm oocyst stormwater density and loading by age class 50
List of Figures
Figure 1. Cattle density in the conterminous United States, 2007 12
Figure 2. Milk cow density in the United States, 2007 13
Figure 3. Density of cattle on feed in the United States, 2007 14
Figure 4. Swine density in the United States, 2007 15
Figure 5. Chicken density in the United States (layers and broilers combined), 2007 16
Figure 6. Farm sources of zoonotic pathogens and pathways to receiving waters 45
August 2010
in
-------
U.S. Environmental Protection Agency
AIAO
APHIS
AWQC
BEACH Act
BMP
CPU
CPSP
CWA
EPA
GI
MMS
MPN
NAAS
PCR
POTW
QMRA
QPCR
SE
STEC
U.K.
U.S.
USDA
WHO
WQS
Acronyms
all-in-all-out management
Animal and Plant Health Inspection Service
ambient water quality criteria
Beaches Environmental Assessment and Coastal Health Act
best management practice
colony forming unit
Critical Path Science Plan
Clean Water Act
U.S. Environmental Protection Agency
gastrointestinal
manure management systems
most probable number
National Agricultural Statistics Service
polymerase chain reaction
publicly owned (wastewater/sewage) treatment works
quantitative microbial risk assessment
quantitative polymerase chain reaction
Salmonella Enteriditis
Shiga toxin producing Escherichia coli
United Kingdom
United States
U.S. Department of Agriculture
World Health Organization (United Nations)
water quality standard(s)
August 2010
IV
-------
U.S. Environmental Protection Agency
Executive Summary
This report describes the distribution,
prevalence,1 and abundance2 of key
waterborne zoonotic pathogens in
domestic cattle, swine, and poultry
(livestock) in the United States. It
evaluates if the prevalence and/or
abundance of these pathogens from
livestock varies systematically due to
geography, farm management, or manure
handling practices. The methods
employed are summarized in the text box
to the right.
The major finding is that U.S. farm
management conditions are more
important determinants of pathogen
occurrence and abundance in livestock
than location within a specific region.
The conditions with the greatest influence
are farm size, whether the farm is a
feedlot, seasonality, and the age of the
animals on the farm.
As depicted in Exhibit 1, exposure to
zoonotic pathogens during recreational
activities depends upon the occurrence of
the pathogens in manure and subsequent
manure and farm management practices.
Risks to users of recreational waters can
be reduced significantly via management
of surface water runoff from stored or
land-applied manures and through
treatment of manures prior to land
application.
The zoonotic pathogens3 evaluated are
the bacterial pathogens Salmonella,
Campylobacter, andE. coll O157:H7,
and the protozoan parasites
Cryptosporidium and Giardia. All of
Methods
This work was conducted into four phases, (1)
selection and justification of key zoonotic pathogens
whose net risk may be used to represent risks
posed by livestock wastes, (2) review of literature to
develop a datasets of distribution and prevalence of
the key zoonotic pathogens in the United States, (3)
review of peer-reviewed literature on the impact of
farm management practices on key pathogen
prevalence and persistence, and (4) evaluation of
the datasets.
The literature review includes collection of national-
scale geospatial data, studies on the prevalence of
pathogens in regions of the United States, and
studies exploring factors that can affect pathogen
distribution and prevalence, including best
management practices (BMPs) and manure
handling practices.
Pathogen
production
(prevalence &
abundance)
Animal management £T Manure
Farm type^^^J management
Treatment
Runoff routing
••^^k Land application
Surface water ^^^
loading
BMPs
Hydrology
Exposure during
recreation
Exhibit 1. Pathogen Flow from Farms to Recreation Sites
1 Defined as the population proportion (%) of animals shedding a particular pathogen at a specific point in time.
2 Defined as the density of organisms in manures of shedding individuals (# of organisms per g wet weight of solid
feces or per L of manure slurry).
3 From "Review of Zoonotic Pathogens in Ambient Waters" (EPA 822-R-09-002).
August 2010
1
-------
U.S. Environmental Protection Agency
Zoonotic Pathogen Species and Human Infections
In cattle, all of the key zoonotic pathogens discussed in this report occur frequently. The species and
serotypes found in cattle feces are similar to those posing hazards to humans. Swine Salmonella,
Campylobacter, Giardia, and E. coli O157:H7 tend to be the same species/serotypes as the most
commonly implicated species/serotypes in human infections. Most swine Cryptosporidia appear to be
host-adapted and pose a reduced hazard to humans. For chickens, the only key zoonotic pathogens
that occur with frequency and are similar to the pathogens implicated in human infections are
Salmonella and Campylobacter. Both are highly prevalent among U.S. flocks.
these pathogens are found in the feces of cattle, swine, and chicken, except Giardia that does
not occur in chicken feces.
The particular pathogen species and types prevalent in livestock feces are not necessarily the
species and types posing the greatest risk of human infection and subsequent illness (see text box
above). The species and types of the key pathogens in cattle are the most similar to those
infecting humans. For chickens, Salmonella and Campylobacter are the only key pathogens
occurring as human-infectious species.
Risks to users of recreational waters impacted by animal wastes depend on the following:
1) whether human-infectious pathogens are present in animal waste,
2) the abundance of the pathogens in those wastes,
3) the survival of those pathogens during manure management and overland transport (to a
stream), and
4) the in-stream transport of those pathogens to a recreation site.
This report focuses on the first three of these factors with particular emphasis on identifying
specific geographical or farm management conditions that influence the relative prevalence or
abundance of the key zoonotic pathogens.
Pathogen Occurrence in Livestock
Prevalence and abundance of zoonotic pathogens
Representative ranges for the prevalence and abundance of the key zoonotic pathogens are
summarized in Exhibit 2 below. These ranges were chosen from values reported in the literature
because they are from studies with large scales and long durations and because the authors
believe they reflect typical conditions in the United States. Despite high variability in pathogen
abundance in all livestock wastes, the following general conclusions can be made:
• E. coli O157:H7 abundance is greatest in cattle.
• Campylobacter abundance is similar in all livestock evaluated and nearly ubiquitous in
chickens and swine.
• Salmonella and Cryptosporidium abundance are greatest in cattle and swine.
• Giardia abundance is greatest in swine.
August 2010
-------
U.S. Environmental Protection Agency
Exhibit 2. Pathogen occurrence and abundance ranges for cattle, swine, and chicken manures
Fecal Source
Prevalence
Abundance
Cattle
Swine
Chickens
Cattle
Swine
Chickens
^
?:
IO
^
O
^
o
o
Uj
9.7-28
0.1-12
NA
1031-084
ND-107
NA*
^
§
(0
"5
•2
Q.
|
5-38
46-98
57-69
1012-1073
1020-1057
1028_1065
JO
0)
c
0
^
(0
V)
5-18
7.9-15
0-95
103-058
1028-1049
10-10-1045
s
.3
'C
o
8-
1
1
o
0.6-23
0-45
NA
1023-1039
1017_1036
NA
•2
•5
I
••S
0.2-37
3.3-18
NA
10°-1049
10°-1068
NA
Conditions and Factors that can Influence the Prevalence or Abundance of the Key
Zoonotic Pathogens
Geography and pathogen occurrence
Studies of national scope indicate some U.S. regional differences in the occurrence of pathogens
in animal wastes. In some cases, regional differences may be the result of regional differences in
agricultural practice (e.g., tendency toward different herd size or feed type) and not intrinsic
differences in the occurrence or survival of pathogens. General findings on regional differences
include the following:
• There is a higher prevalence of Salmonella in layers (chickens) and dairy cattle in the
Midwest (including Great Lakes states) than in other U.S. regions.
• Regional clustering of Salmonella-positive feedlot cattle was not observed. However,
cattle on feed longer appear associated with higher Salmonella shedding prevalence.
• Campylobacter are uniformly distributed among dairy cattle in the United States.
• E. coli O157 occurs at low levels in dairy manure across the United States. One study
noted prevalence is highest in the southwestern states (7.6% of fecal samples) and lowest
the northeast (1.6% in the northeast). Dairies in the southwestern United States typically
have larger herds than other U.S. regions.
• Samples from pens of feedlot cattle receiving barley were 2.7x more likely to have
positive samples for E. coli O157 than for pens without cattle fed barley. Because
feeding barley to cattle is effectively a regional practice, this finding might partially
explain high regional prevalence of E. coli 0157 in the southwest.
• In a 13-month study of feedlot beef cattle, E. coli O157 occurs at low levels in beef cattle
manure across the United States, but prevalence is highest in the southwestern states
(13% of fecal samples from California, New Mexico, and Texas). Only the top 11
feedlot cattle states were included in that study.
August 2010
-------
U.S. Environmental Protection Agency
• In both dairy and feedlot beef cattle operations, prevalence was consistently higher
during summer months and for larger herd sizes.
• Beef feedlot cattle with the shortest time on feed had the highest E. coli O157 prevalence.
This may be the result of stress during travel and greater susceptibility of animals newly
arrived at feedlots.
Farm management conditions
Numerous farm management conditions can influence the relative prevalence or abundance of
the key zoonotic pathogens. Important findings include the following:
• The prevalence of most of the key zoonotic pathogens increases with larger farm size
(number of animals) for cattle and chicken operations.
• Campylobacter and Salmonella prevalences are higher among feedlot cattle than among
pasture or range cattle. This difference is likely (primarily) due to diet and housing.
• Cattle Cryptosporidium infection is more related to animal age than housing or feed;
very young calves shed Cryptosporidium with higher prevalence and abundance than
older cattle.
• E. coli O157:H7 shedding in cattle is highly variable. However, the prevalence is higher
in spring/summer than fall/winter.
• Chicken Campylobacter prevalence is higher in the summertime than other seasons, and
may be the result of pathogen aerosol transport via ventilation systems.
• Differences in pathogen prevalence between conventional and organic farming operations
can be related to intrinsic differences in these farming practices—such as farm size
(organic farms tend to be smaller), animal housing, and age of the animals on the
operation. For example, Campylobacter prevalence increases with chicken age and
organic chickens are typically older when slaughtered.
Manure management practices
• Pathogen loads to streams may be substantially reduced via manure management and use
ofBMPs.
• Effective BMPs include routing stormwater away from manures and farm areas with high
potential for pathogens, and use of filter strips and other means for slowing overland
flow.
• Treatment alternatives may be used to reduce manure pathogen density by as much as 5-
logs, with higher reductions associated with higher costs and maintenance requirements.
August 2010
-------
U.S. Environmental Protection Agency
1. Introduction
1.1. Purpose
This report summarizes available information on the prevalence, distribution, and abundance of
select waterborne zoonotic pathogens in domestic livestock cattle, swine, and chicken in the
United States. It is, in large part, a continuation of the U.S. Environmental Protection Agency
(hereafter EPA or the Agency) report Review of Zoonotic Pathogens in Ambient Waters
(USEPA, 2009). In that report, the following six key ("reference") waterborne zoonotic
pathogens were identified based on their relevance in the United States and their potential to be
associated with outbreaks in ambient recreational waters and/or drinking water: pathogenic E.
co//', Campy lobacter, Salmonella, Leptospira, Cryptosporidium, and Giardia. The first four
microorganisms are bacteria while the last two are parasitic protozoa. With the exception of
Leptospira (see Text Box 1), the remaining five zoonotic waterborne pathogens are discussed
throughout this report.
1.2. Background
A central goal of the Clean Water Act (CWA) is to protect and restore waters of the United
States for swimming and other recreational activities. A key component in the CWA framework
for protecting and restoring recreational waters is for EPA to recommend ambient water quality
criteria (AWQC) to provide public health protection from illnesses—historically gastroenteritis
or gastrointestinal (GI) illness—associated with exposure to fecal contamination during
recreational water contact and for subsequent adoption by the States as water quality standards
(WQS). Water quality criteria, WQS, guidelines, or their equivalent, as they relate to microbial
waterborne illness, are generally specified throughout the world in terms of densities of fecal
indicator organisms because fecal matter can be a major source of pathogens in ambient water
and because it is not practical or feasible to monitor for the full spectrum of all pathogens that
may occur in water (NRC, 2004). For decades, these fecal indicator organisms have served as
surrogates for potential pathogens and subsequent health risks in recreational and drinking
waters.
The EPA currently recommends recreational AWQC under CWA Section 304(a) and utilizes the
fecal indicator bacteria enterococci and/or E. coli (USEPA, 1986), which are non-pathogenic and
present in both human and animal feces. This approach effectively assumes that animal fecal
material is as hazardous as human fecal material and does not allow the exclusion or
"discounting" of disease risk associated with animal fecal contamination.
According to the World Health Organization (WHO, 2004), zoonoses are "those diseases and
infections which are naturally transmitted between vertebrate animals and man." Examples of
zoonotic infections have been recognized among all the major groups of infectious agents:
bacteria, protozoa, viruses, helminths, and prions—though the latter three groups are not
discussed in this paper. Some zoonoses may infect only one type of animal other than humans,
while others may infect several types of animals as well as humans (Moe, 2004). Fenton and
Pederson (2005) reported that most pathogens can infect several host species; for example, >60%
of human pathogens and >90% of domesticated animal pathogens infect multiple hosts.
August 2010
-------
U.S. Environmental Protection Agency
Text Box 1
Selection of Key Waterborne Zoonotic Pathogens
There are many zoonotic pathogens and many waterborne pathogens; however, there is a much more
limited subset of pathogens that are both. In the EPA report, Review of Zoonotic Pathogens in Ambient
Waters (USEPA, 2009), a total of 70 pathogens from warm-blooded animals were evaluated for their
potential to be both waterborne and zoonotic using the following four criteria (partially adapted from Bolin
etal.,2004):
1. The pathogen must spend part of its lifecycle within one or more warm-blooded animal species.
2. Within the lifecycle of the pathogen, it is probable or conceivable that some life stage will enter water.
3. Transmission of the pathogen from animal source to human must be through a water related route.
There are zoonotic pathogens for which waterborne exposure has not been detected as a significant
route of cross-species transmission. This does not exclude the possibility that these zoonotic
pathogens could be transmitted via water.
4. The pathogen must cause infection or illness in humans. There are animal pathogens that have
waterborne transmission between animals, yet are not known to cause illness in humans.
Of the 70 pathogens evaluated, 20 met the above criteria. Notably, several well-known waterborne
pathogens were excluded from analysis because they are not zoonotic. See Appendix A, Table A-1 of
USEPA (2009) for a summary of waterborne pathogens that meet the above criteria and selected
pathogens that meet some, but not all, of the criteria. Six of the 20 waterborne, zoonotic pathogens from
warm-blooded animals were selected for further discussion based on their relevance in the United States.
Five (£. co/;, Campylobacter, Leptospira, Cryptosporidium, and Giardia) were selected based on their
potential for outbreaks in ambient (untreated) recreational water and one (Salmonella) was included
based on outbreaks in drinking water. Notably, this list correlates very well with the top five waterborne
pathogens for recreational and drinking waters identified previously by Craun et al. (2004). Of the 6 key
pathogens, Leptospira is excluded from further discussion in this report. This is because the source of
infection in humans is usually either direct or indirect contact with the urine (not feces) of an infected
animal (USEPA, 2009), and because there are no peer-reviewed, livestock-associated dose-response
data for this pathogen. Notably, other studies reviewing risks associated with zoonotic organisms in
livestock feces also do not include Leptospira among the main pathogens of concern (Bicudo and Goyal,
2003; Goss and Richards, 2008; Venglovsky et al., 2009). The key pathogens selected for analysis in
this study are the same as those suggested for emphasis in research and regulation in another review of
zoonotic pathogens (USEPA, 2005). On the basis of surveillance of water and foodborne outbreaks in
the United States, the authors suggested that priority for standard methods and recreational and drinking
water guidelines should be given to Salmonella spp., Campylobacter jejuni, E. coli O157:H7,
Cryptosporidium, Giardia, and selected viral agents indicative of viral contamination.
Interested readers may refer to USEPA (2009) for detailed descriptions of strain variation and zoonotic
potential, routes of exposure, health implications, incidence, and interactions and survival in the aquatic
environment, for each of these key pathogens.
The presence of fecal indicator organisms in recreational waters generally indicate point- and/or
nonpoint sources of human and/or animal fecal wastes from agricultural animals (e.g., cows,
pigs, and chickens), domestic animals (e.g., dogs and cats), and/or wildlife (NRC, 2004).
Excreted feces and other animal waste products (e.g., urine) are the predominant sources of
waterborne zoonoses (WHO, 2004). Zoonotic pathogens in the feces of an animal or human
reservoir can be transported to a particular waterbody where their stability in that environment
will ultimately influence their infectivity and disease risk to exposed humans. There is evidence
August 2010
-------
U.S. Environmental Protection Agency
that zoonotic pathogens may change in infectivity, virulence, and the severity of health outcomes
they cause in humans depending on their previous host environment. There is also evidence that
some of these host-factor changes can influence subsequent infection cycles in exposed hosts.
The key mechanisms of phenotypic change in pathogens are genetic diversity (coinfection and
quasispecies), cryptic genes, mutators, and epigenetic effects, and are reviewed in USEPA
(2009).
Understanding which pathogens could be present depending on the source of fecal contamination
might allow the Agency to better estimate human health risks from identified sources of fecal
matter. To this end, the information in this report could be used to support quantitative microbial
risk assessment (QMRA) in two ways, (1) to estimate potential risks from warm-blooded animal
feces in ambient (untreated) recreational waters; and (2) as a complement to recreational water
epidemiological studies in support of the development of new or revised recreational water
quality criteria and/or implementation. Risks of illness from recreation in waters receiving
livestock fecal pollution arise from the presence of human-infectious pathogens in livestock
wastes, the abundance of the pathogens in the fecal wastes, the survival of pathogens during on-
farm treatment of the wastes or during transport to streams, and the transport of pathogens from
their points of entry (often diffuse) to surface waters or recreation sites. This report focuses on
the on-farm portion of this risk scenario—particular emphasis is placed on the impact of the
animal source and farm operations on the prevalence and abundance of key zoonotic pathogens
in manures of swine, cattle, and chicken.
EPA has conducted a significant amount of research since the Beaches Environmental
Assessment and Coastal Health (BEACH) Act of 2000 was enacted and plans to issue new or
revised criteria for coastal recreational waters (under the CWA, defined as Great Lakes and
coastal marine waters) by 2012. EPA believes that the new or revised criteria must be
scientifically sound, implementable for broad CWA purposes, and provide for improved public
health protection over the current (1986) criteria.
1.3. Approach
Preparation of this report was divided into (1) a literature review to develop a dataset of available
studies on the U.S. distribution and prevalence of five key waterborne zoonotic pathogens; and
(2) evaluation of the dataset. The literature search included the identification, collection, and
summarization of available data and studies (e.g., peer-reviewed literature, EPA or other State
agency reports) on the distribution and prevalence—and the factors affecting distribution and
BMPs or manure handling practices—of these key zoonotic pathogens in U.S. domesticated
livestock cattle, swine, and chicken. The review and analyses focused on ambient water
transmission, to include animal feeding operations (AFOs), concentrated feeding operations
(CAFOs), and small farms. The report also identifies and discusses seasonal and regional
variability of pathogen occurrence and abundance. These data and information were used along
with available national-scale geospatial data to support extensive mapping and analysis of the
spatial distribution of zoonotic pathogen sources and their potential impacts on water quality in
the United States.
August 2010
-------
U.S. Environmental Protection Agency
Following this introduction, Section 2 provides an overview of the distribution of livestock in the
United States. Section 3 summarizes information on pathogen occurrence in U.S. livestock,
while Section 4 discusses important factors affecting pathogen occurrence. Section 5 includes all
references cited within this report. Appendix A summarizes the approach and results of the
literature review. Appendices B and C summarize individual studies of zoonotic pathogen
occurrence and abundance, while Appendix D summarizes studies of farm factors and their
impact on zoonotic pathogen occurrence and abundance.
August 2010 8
-------
U.S. Environmental Protection Agency
2. Distribution of Livestock in the United States
In this section, the distribution of livestock cattle, swine, and chicken in the conterminous United
States is presented. Distribution, together with data on pathogen prevalence and abundance in
livestock manures (discussed in Sections 3 and 4), suggests regions of the United States where
pathogen occurrence in recreational waters differ from "typical" values and where the risks
associated with fecal indicator densities differ from those associated with waters impacted by
human fecal pollution. Different livestock species are associated with different suites of
pathogens. Those pathogens may differ in their ability to infect humans and also differ in the
abundances in which they occur in manures. Thus, while livestock distribution alone does not
allow estimation of risks associated with swimming in recreational waters receiving wastes from
livestock operations, it contributes to the comparison of U.S. regions with respect to their risks.
2.1. Methods
Livestock data were acquired from the 2007 Agricultural Census (NASS, 2010). This data set
provides data on all livestock species of interest to this report. It is a single and consistent data
set for the entire conterminous United States and contains data available at a county level scale
for visualization of regional differences in livestock. The Agricultural Census (the census)
provides the number of animals for each livestock type as well as related data not used in this
report.
Cattle data are presented as all cattle (adults and calves) and are also subdivided into milk cows
and cattle on feed. Cattle on feed are those being fed some ration of grain, silage, hay and/or
protein supplement in preparation for slaughter. As described in Sections 3 and 4, these
distinctions are important because cattle on different operations have different prevalence and
shedding rates of the key pathogens.
Swine data available from the census includes data for hogs (swine weighing over 120 pounds)
and pigs (swine weighing under 120 pounds), and also hogs/pigs used for breeding. Because the
data do not indicate a significant difference between these two groups, the combined hog and pig
data were extracted from the census and used in this section, while the data for breeding
hogs/pigs was not used.
For chicken, the census provides layer and broiler data separately. Broilers are raised for meat
while layers are raised for egg production. The data for broilers and layers were combined for
this report because no significant difference between prevalence of pathogens is apparent
between both types of operations.
To allow meaningful comparison of livestock numbers between counties and regions, the
number of animals for each livestock type was normalized by county area, and livestock
numbers were presented as densities (number of animals/mi ). Data from numerous counties
was withheld from the livestock census to prevent disclosure of data on individual operations
that are identified in the maps shown below. Because the proportion of counties with suppressed
data was relatively minimal, this did not preclude interpretation of maps and identification of
August 2010
-------
U.S. Environmental Protection Agency
trends in livestock density for the conterminous United States. However, the absence of these
counties does preclude state-level or finer spatial scale analyses of livestock density.
Maps illustrating density by county for each of the livestock types (all cattle, milk cows, cattle on
feed, all swine, and all chickens) were developed using ArcGIS 9.2. The maps are graduated in
color (light to dark representing low to high density), categorized by quartile, and the counties
withholding data are shown in yellow. Alternative schemes for presenting data were explored,
however categorization by quartiles most suitably affords easy discrimination of regions of high
and low density on maps of the entire United States.
2.2. Cattle Distribution
2.2.1. All cattle
The distribution of all cattle (including milk, beef and cattle on feed) in the conterminous United
States is shown in Figure 1. Several large regions of the United States show relatively high
densities. These include the central United States (from North Dakota through Texas); the Great
Lakes states; less urbanized portions of Pennsylvania, New York, and Vermont; portions of
Virginia, Maryland, Tennessee, and Kentucky; and counties in California and Florida. A
majority of these regions lie within temperate climate zones and drain to inland waters, with a
large portion of the high-density counties located within the Mississippi River basin. The
potential for runoff impacting coastal waters is relatively high for Wisconsin and western
Michigan coastal streams draining to Lake Michigan, for Florida and East Texas streams
draining to the Gulf of Mexico, and for California streams draining to the Pacific Ocean.
2.2.2. Milk cows
Figure 2 shows milk cow density in the conterminous United States. The distribution of U.S.
milk cows differs substantially from that of all cattle, with clear high-density regions extending
from the Great Lakes through the Northeast states, through the mid South, in a cluster in
southwestern Missouri, and along the Pacific coast. The regions with highest milk cow densities
drain to the Great Lakes and to inland waterways including the waters of the Chesapeake Bay
system, the Ohio and Tennessee Rivers, inland waters in the Pacific Northwest and California,
and the upper Mississippi River.
2.2.3. Cattle on feed
The density of cattle on feed is shown in Figure 3. The distribution of feedlot cattle differs
substantially from overall cattle distribution and from distribution of milk cows. There is a clear
high-density region of cattle on feed in the upper Midwest (a majority of Nebraska, most of
Iowa, southern Minnesota and Wisconsin, eastern South Dakota, much of Kansas, and eastern
Colorado), and there are regions of medium density from the Great Lakes states through the mid-
Atlantic states. Data was withheld by the census for many counties in Texas, Virginia, and
Tennessee. It can be assumed that Texas also has high feedlot cattle density. The majority of
high-density regions drain to the Mississippi River basin, while many of the medium-density
counties drain to the Great Lakes.
August 2010 10
-------
U.S. Environmental Protection Agency
2.3. Swine Distribution
Because few data were available to differentiate between swine operations, all swine density data
are presented in a single map (Figure 4). There are clear regional trends in swine density, with
high-density regions in the Midwest (the highest densities seen in Iowa), the Great Lakes States
(particularly in Illinois, Indiana, southern Michigan, and western Ohio), eastern North Carolina,
and in south-central Pennsylvania. These regions drain primarily to the Mississippi River basin,
the Ohio River, the Great Lakes, and mid-Atlantic coastal streams.
2.4. Chicken Distribution
Among the different livestock species, chicken density (shown in Figure 5) appears the most
uniform. Although the census provided bird counts for layers and broilers, these data were
combined because insufficient evidence was found that would differentiate these types of
operations with respect to their anticipated pathogen productions. Chicken production is
common across the entire eastern United States, with particularly high bird densities in the
Southeast (excluding Florida) and Arkansas, the upper Midwest (Wisconsin, Minnesota, and
Iowa), and inland counties of the Pacific states (California, Oregon, Washington). The high-
density regions drain to coastal (Great Lakes and Atlantic) waters, and many U.S. inland waters.
2.5. Summary
Cattle and swine tend to be clustered in identifiable regions of the United States, whereas
chickens are more disperse. Regions of particularly high milk cow density are the Great Lakes,
Northeast, and inland counties of California and Washington. Cattle on feed have different
prevalences of many of the key pathogens than milk cows. High densities of cattle on feed are
noted in the Midwest (Texas to Minnesota). Swine are the most heterogeneously distributed of
these livestock types, with high-density regions in the upper Midwest and North Carolina.
Chicken are more evenly distributed than other livestock types, with production occurring
throughout the eastern United States and inland counties of the Pacific states. These livestock
distributions may be used, along with pathogen prevalence data, to identify regions where risks
from recreation may differ from those either in waters primarily impacted by publicly-owned
treatment works (POTW) discharge (wastewater/sewage) or waters impacted by mixed sources
of fecal pollution.
August 2010 11
-------
U.S. Environmental Protection Agency
Figure 1
Density of cattle in the conterminous United States
2007
375
750
1 500 Miles
I
Density of Cattle by County
(all cattle/ area in sq.mi.)
Conterminous U.S. 2007
Quarfrfes:
| | 0-10
^26-56
I 57-607
1 '.Withheld to avoid disclosing
I data for individual farms (n=4:
I State boundaries
] Majorwaterbodies
Dafa source. United States Department of Agriculture
The Census of Agriculture
Nation at Agricultural Statistics Service
Figure 1. Cattle density in the conterminous United States, 2007
August 2010
12
-------
U.S. Environmental Protection Agency
Figure 2 o
Density of milk cows in the conterminous United States L.
2007
375
750
N
''V£
1 500 Miles
=>iyj-31tf¥^ r^-r^^^
~-3r zLe J-Jt^r ••>" • -. I i.i.' . ••:
- • hr TIT .<£+*,
Density of Milk Cows by County
(milk cows / area in sq.mi.)
Conterminous U.S. 2007
Quarfn'es.
| | 1-2
| | 3-8
I I 9-193
1 vVithheld to avoid disclosing
J datafor individual farms (n=133:
I State boundaries
Major waterbodies
Dais source. U^fed States Department of Agriculture
The Census of Agriculture
National Agricultural Statistics Service
Figure 2. Milk cow density in the United States, 2007
August 2010
13
-------
U.S. Environmental Protection Agency
Figure 3 o
Density of cattle on feed in the conterminous United States L
2007
375
750
1.500 Miles
_i I
Density of Cattle on Feed by County
(cattle on feed / area in sq.mi.)
Conterminous U.S. 2007
Qua/tifes.
< 1
1 -2
3-10
11-570
1 vYithheldto avoid disclosing
J data for individual farms (n=179;
I State boundaries
"J Majorwaterbodies
Dafs source. United States Department of Agriculture
The Census of Agriculture
National Agricultural Statistics Service
Figure 3. Density of cattle on feed in the United States, 2007
August 2010
14
-------
U.S. Environmental Protection Agency
Figure 4
Density of swine in the conterminous United States
2007
375
750
1 500 Miles
|
Density of Swine by County
(swine.1 area in sq.mi.)
Conterminous U.S. 2007
4-45
46-2802
~~| '.".'it h he I d to av oi d d iscl o sin g
I 1 data for in dividual farms (n=1Q5;
Estate boundaries
Majorvvaterbodies
Data source. United States Department of Agriculture
The Census of Agriculture
National Agricultural Statistics Ser/.-ce
Figure 4. Swine density in the United States, 2007
August 2010
15
-------
U.S. Environmental Protection Agency
Figure 5
Density of chickens in the conterminous United States
(layers and broilers combined)
2007
375
750
1.500 Miles
I
Density of Chickens by County
(chickens / area in sq.mi.)
Conterminous U.S. 2007
Qualities:
| | 3-12
1 13 - 74,423
1 Withheld to avoid disclosing
I data for individual farms (n=228:
I State boundaries
Major watersodies
Data source. United States Department of Agriculture
The Census of Agriculture
National Agricultural Statistics Service
Figure 5. Chicken density in the United States (layers and broilers combined), 2007
August 2010
16
-------
U.S. Environmental Protection Agency
3. Pathogen Occurrence in Livestock
The livestock distribution data in Section 2 provide an indication of where gross pathogen loads
may occur in the United States, but additional data are required to develop a more nuanced and
defined understanding of the distribution of specific pathogens. This section presents and
describes those additional data and, where sufficient, the data are then related to the regional
distribution of pathogens. It is important to emphasize that the risks associated with particular
livestock species are specific for each host-pathogen combination. Further, pathogen occurrence
and shedding rates, host-specificity, and manure handling practices all play roles in the
generation and transport of pathogens with the potential to infect humans. These factors and
processes differ among host-pathogen combinations.
To characterize the occurrence and distribution of pathogens among livestock wastes, relevant
studies were collected and data extracted characterize the occurrence of the five key zoonotic
pathogens listed in Section 1. Numerous study types and data sets contributing to these
characterizations were identified and grouped as follows:
• review papers describing the hosts for the individual pathogens and the extent to which
host adaption may have occurred;
• studies reporting prevalence (either herd- or animal-level) of the pathogens in manures of
cattle, swine, and chickens;
• studies reporting the abundance (number of organisms per mass of feces);
• studies reporting the national distribution of the reference pathogens; and
• studies relating farm factors to the occurrence of pathogens.
The following sections provide summaries of each of these groupings and refer to expansive
tables in the appendices (B, C, and D) in which all studies identified during the literature review
are presented.
In these reviewed studies, the authors use the terms pathogen prevalence and pathogen
abundance, albeit with different meanings. In general, pathogen prevalence is the fraction of
manure samples, herds, individual animals, or other measurable quantities associated with
pathogens. The three measures of prevalence with the greatest relevance and that are the most
frequently used in this report are (1) herd-level prevalence (the percentage of herds having at
least one positive sample for a given pathogen); (2) sample-level prevalence (the percentage of
fecal samples positive for a given pathogen); or (3) individual animal prevalence (the percentage
of tested animals positive, either based on samples of fresh feces or other measures of infection).
Note that throughout this report, the terms prevalence and occurrence are used interchangeably.
Abundance refers to the count of pathogens or indicator in a known mass or volume of a given
environmental medium. Because it is the most useful measure of pathogen density, this report
focuses on fecal abundance of pathogens (units of organisms/g for solid feces or organisms/L for
slurries). In most cases, studies presented pathogen densities as number of organisms per wet
weight of feces, though several studies presented fecal abundance data as organisms per dry
weight of feces.
August 2010 17
-------
U.S. Environmental Protection Agency
3.1. Key Zoonotic Hosts
The five reference zoonotic pathogens occur in different hosts, with the hosts frequently
associated with specific species, serotypes, or genotypes of the pathogens. This section
associates the species and genotypes of the reference pathogens with hosts—both to identify the
most important pathogen species/serotypes/genotypes in human infections and to identify the
hosts likely to produce human-infectious pathogens. For all of these pathogens, and for
Salmonella in particular, the prevalence of species or serotypes among specific host populations
appears to shift over time, so the data presented in this section must be interpreted cautiously.
Among the more than 16 species of Cryptosporidium identified to date, C. parvum and C.
hominis are believed to cause the majority of human infections among immunocompetent hosts.
As noted in Table 1 below, other animals are considered major hosts for both species. Humans
are considered minor hosts for other Cryptosporidium species, including C. muris, C.
meleagridis, C. foils, and C. canis. Among livestock species, cattle prevalence of
Cryptosporidium species aligns closely with species infecting humans, whereas swine
cryptosporidia are more seldom isolated in human infections (Xiao et al., 2006). Chicken
cryptosporidia do not appear to overlap with species causing human infections, with the
exception of C. meleagridis that has been implicated in infections of immunocompromised
persons (Hunter and Nichols, 2002).
Giardia taxonomy remains unsettled and the species ofGiardia causing the majority of human
illnesses is called G. lamblia, G. duodenalis, and G. intestinalis by different researchers (Adam,
2001; Thompson, 2004). Thompson (2004) notes that Giardia, isolates from humans fall into one
of two major genotype assemblages and that some Giardia genotypic groupings are confined to
specific animal hosts. Table 2 (Adam, 2001) presents the most important Giardia species and
genotypes and their associated hosts. Based on this listing, cattle and pigs appear to have the
potential for shedding human-infectious Giardia, though chicken are not a significant source of
human-infectious Giardia cysts.
Table 1. Cryptosporidium species and associated major and minor hosts (SOURCE: adapted from
Xiao etal., 2004, 2006)
Reference
Pathogen
Cryptosporidium
Species or
Serotype
C. muris
C. andersoni
C. parvum
C. hominis
C. felis
C. canis
C. meleagridis
C. baileyi
C. galli
C. bovis
C. suis
Major Host
Rodents, Bactrian camels
Cattle, Bactrian camels
Cattle, sheep, goats, humans
Humans, monkeys
Cats
Dogs
Turkeys
Chicken, turkeys
Finches, chicken, capercalles,
grosbeaks
Cattle
Pigs
Minor Host
Humans, rock hyrax,
mountain goats
Sheep
Deer, mice, pigs
Dugongs, sheep
Humans, cattle
Humans
Parrots, humans
Cockatiels, quails, ostriches,
duck
—
—
—
August 2010
18
-------
U.S. Environmental Protection Agency
Table 2. Significant Giardia species and genotypes and associated hosts (SOURCE: adapted from
Adam, 2001)
Reference
Pathogen
Giardia
Species
G. agilis
G. muris
G. lamblia
G. ardae
G. psittaci
G. microti
Genotype
Genotype A-1
Genotype A-2
Genotype B
Hosts
Amphibians
Rodents
Human, cat, beaver, lemur, sheep, calf, dog,
chinchilla, alpaca, horse, pig, cow
Human, beaver
Human, beaver, guinea pig, dog, monkey, cow,
sheep, alpaca, goat, pig, rat
Herons
Psittacine birds
Voles and muskrats
Campylobacter species may be grouped as those playing a major role in human infection, those
playing a minor role, and those unlikely to cause human infection. Ketley (1997) designated C.
jejuni and C. coli as the species playing a major role in human infections (80% to 90% of
Campylobacter infections) and notes that other species have the potential for initiating human
infections. Campylobacter species in humans, livestock, and other hosts are summarized in
Table 3. C. jejuni and C. coli are prevalent among cattle, pigs, and chickens. The non-human
reservoirs of C. upsaliensis are not fully known, though dogs are known to shed that species and
have been implicated in transmission to humans (Bourke et al., 1998). C. hyointestinalis has
been identified in pig and cattle feces and hamster intestines (Gebhart et al., 1985) and is a
suspected cause of human enteric disease (Edmonds et al., 1987).
Salmonella serotypes prevalent among different hosts appear to be fluid; over the past 20 years
the predominant serotype for both swine and chickens has changed (Foley et al., 2008).
Nonetheless, examining the predominant serotypes for different hosts provides insight into the
potential for animal species to shed zoonotic Salmonellae. Baumler et al. (1998) associated
humans, livestock, and selected other host species with Salmonella serotypes frequently
encountered (Table 4) and noted the age groups that are most susceptible to infection and illness.
Given the fluidity in serotype frequency for each of the hosts, the authors believe that the
associations presented in Table 4 demonstrated in that all three livestock species of interest
(cattle, swine, and chicken) have the potential to shed human-infectious Salmonellae. Callaway
et al. (2008) caution against making inferences on the zoonotic potential of Salmonellae based
only on serotype prevalence.
As discussed in Section 3.2 below, E. coli O157:H7 are frequently detected in cattle feces, less
prevalent in swine feces, and seldom reported in chicken feces. Assessing the potential for cattle
and other wildlife to generate virulent E. coli O157 is difficult given the apparent ability of
Shiga-toxin-negative E. coli 0157 to acquire the stx virulence gene in a variety of hosts and
settings (Wetzel and LeJeune, 2007), and the potential for differences in virulence between
isolates from humans and other sources, though these differences were not observed in a recent
study by (Lenahan et al., 2009). Given the variability of E. coli O157:H7, even among those
isolates originating from the same source, we adopt a conservative approach and assume thatE1.
coli O157:H7 from any source pose the same hazard to humans.
August 2010
19
-------
U.S. Environmental Protection Agency
Table 3. Significant Campylobacter species and genotypes and associated hosts
Reference Pathogen
Campylobacter
Role in Human Infections
Major
Minor
Unlikely
Species
C. jejuni
C. coli
C. upsaliensis
C. hyointestinalis
C. lari
C. pylori
C. fetus
Non-Human Hosts
Cattle, pigs, chickens
Cattle, pigs, chickens
Dogs
Pigs, cattle, hamsters
Sea gulls, chickens
Not known
Cattle, sheep
Table 4. Frequently-encountered Salmonella serotypes for select hosts (SOURCE: Baumler et al.,
1998)
Reference
Pathogen
Salmonella
Host
Species
Humans
Cattle
Chicken
Pigs
Sheep
Horses
Wild rodents
Disease
Salmonella enteritis
Typhoid fever
Paratyphoid fever
Salmonellosis
Pullorum disease
Fowl typhoid
Avian paratyphoid
Pig paratyphoid
Salmonellosis
Chronic
paratyphoid
Salmonellosis
Salmonellosis
Murine typhoid
S. Enterica Subspecies/
Serotypes most Frequently
Encountered
Typhimurium, Enteritidis
Typhi
Sendai; Paratyphi A, B and C
Typhimurium, Dublin
Pullorum
Gallinarum
Enteriditis, Typhimurium
Cholerasuis
Typhimurium
Typhisuis
Abortusovis
Typhimurium
Abortusequi
Typhimurium
Typhimurium, Enteritidis
Most Susceptible
Age Groups
Children (<4 years)
Children and adults
Children and adults
Calves and adults
Newly-hatched birds
Growing stock &
adults
Newly-hatched birds
Weaned and adult
pigs
Weaned pigs (>4
months)
Not specified
Lambs, adult sheep
Lambs
Foals and adults
Foals
Not specified
3.2. Occurrence and Abundance of Reference Pathogens in U.S. Cattle, Swine, and Chicken
The major determinants of whether swimmers are exposed to pathogens of livestock origin
include the following:
• whether livestock are shedding the pathogen (prevalence/occurrence),
• the rate of shedding (abundance), and
• the attenuation of pathogens between their introduction in the watershed and their arrival
at the swimmer.
This section summarizes occurrence and abundance data for each of the livestock species for
each of the reference pathogens. The data characterizes host-pathogen occurrence and
abundance and includes data collected as a component of preparing the report "State-of-the-
Science Review of Quantitative Microbial Risk Assessment: Estimating Risk of Illness in
August 2010
20
-------
U.S. Environmental Protection Agency
Recreational Waters" (August 2010) and data identified during the literature review (see
Appendix A). For all host-pathogen combinations, both occurrence and abundance are widely
variable. Occurrence varies between herds/flocks and within herds/flocks. For some host-
pathogen combinations infection is typically chronic (e.g., Campylobacter in chickens) whereas
for others infection is epidemic or more prevalent within a particular age cohort (e.g.,
Cryptosporidium in calves). Likewise, shedding (abundance) typically varies during the course
of infection for an individual animal, with some animals shedding asymptomatically, some
animals shedding at mean rates, and some animals acting as "super-shedders" (animals shedding
at above-average rates).
The occurrence of the reference pathogens in livestock manure (% of samples positive) is
summarized in Table 5 and a full listing of occurrence data extracted from the literature is
presented in Appendix C. In Table 6, the maximum and minimum reported prevalences of
pathogens in feces are presented for each of the livestock animals and pathogens. Notable
important findings from Table 5 are
• Campylobacter is quite prevalent (typically >50%) in both pig and chicken manure and is
detected often, but less frequently in cattle manure;
• although E. coli O157:H7 shedding was observed in pigs, the most important source of E.
coli O157:H7 is cattle;
• Salmonella occurs in all of the livestock species and is highly variable among chickens;
and
• chickens are not significant sources for Cryptosporidium and Giardia (Cryptosporidium
species typical of chicken infection pose a low risk of human infection and Giardia have
not been observed in chicken feces).
Reported abundances of pathogens in feces of livestock are summarized in Table 6 and all data
are provided in Appendix D. The studies reviewed employed various sampling strategies and
where possible, the authors attempted to provide consistent abundance data in Table 6. The basis
refers to the weight basis for the density ([W]et vs. [D]ry). The sample type is either composite
(C) or direct (D), and the chicken manure type is either fresh (F) or litter (L). Notable
observations from Table 6 include the following:
• fecal shedding abundances exhibit wide variability, with extreme variability observed for
cattle shedding of E. coli O157:H7 and Campylobacter, for pig shedding of Giardia, and
for chicken shedding of Salmonella;
• cattle have the highest manure production rate among the livestock species and therefore
may produce very high pathogen loads (number of pathogens excreted by an animal over
a period of time); and
• no data were available for estimating E. coli O157:H7 shedding rates in chickens.
August 2010 21
-------
U.S. Environmental Protection Agency
Table 5. Prevalence (occurrence) of human infectious species of pathogens in livestock manures
Pathogen
E. co//
O157:H7
Campylobacter
spp.
Salmonella
en ten 'ca
Cryptosporidiu
Giardia spp.
Cattle (Beef & Dairy)
0
c
15
Q.
observed
E
^
—
ii
3.3
5
5
0.6
0.2
o
15
Q_
observed
E
E
X
ro
28
38
18
23
46
Reference(s)
Berry et al.
(2007)
Wesley et al.
(2000); Hoar
etal. (2001)
Hutchison et
al. (2004);
Fossler et al.
(2005)
Sturdee et al.
(2003); Atwill
et al. (2006);
USDA
(1994); Payer
et al. (2000);
Wade et al.
(2000)
Pigs
0
c
15
Q.
observed
E
^
—
ii
0.1
46
7.9
0
3.3
0
15
Q_
observed
E
E
X
ro
12
98
15
45
18
Reference(s)
Cornick and
Helgerson (2004);
Hutchison et al.
(2004);
Dorneretal. (2004)
Hutchison et al.
(2004); Dorr etal.
(2009)
Heitman et al.
(2002); Xiao et al.
(2006)
Heitman et al.
(2002); Xiao et al.
(2006)
Chicken
0
c
15
Q.
observed
E
^
—
ii
0
57.1
0
6
0
15
Q_
observed
E
E
X
ro
0
68.5
95
27
Reference(s)
Chapman et al.
(1997); USDA
(2001)
Cox et al.
(2002) and El-
Shibiny(2005)
Byrd (1998);
Martin et al.
(1998)
Ley etal. (1988)
None reported
3.3. Large-Scale Studies of Pathogen Occurrence in U.S. Livestock
Relatively few national-scale studies of U.S. livestock pathogen prevalence are available in the
open literature. Among the studies identified, most were either of Salmonella, Campylobacter,
or E. coll O157:H7 occurrence. A few focused on Cryptosporidium or Giardia occurrence.
Because these studies provide the most direct evidence of regional differences in pathogen
occurrence and properties, they were reviewed and are summarized in Table 7 (for Salmonella),
Table 8 (for Campylobacter), Table 9 (for E. coli O157:H7), Table 10 (for Cryptosporidium),
and Table 11 (for Giardia).
August 2010
22
-------
U.S. Environmental Protection Agency
Table 6. Abundance of human infectious pathogens in livestock manures
Pathogen
E. co/;O157:H7
Campylobacter
Salmonella
Cryptosporidium
Giardia
Cattle6
Log10 range3
a
3.f
1.2
3'
2.3
0.0
b
8.4
7.3
5.8
3.9
4.9
&
in
'if>
CO
00
W
W
W
W
W
o
CD
Q.
><
-i—*
CD
Q.
CO
W
C
D
C
C
D
Reference
Hutchison et
al. (2004)
Moriarty et
al. (2008)
Hutchison et
al. (2004)
Atwill et al.
(2006)
Wade et al.
(2000)
Pigs6
Log10
range
a
ND9
2.0
2.81
1.71
0
b
7
5.7
4.9
3.6
6.8k
CD
Q.
><
-i—*
CD
Q.
CO
W
D
D
C
C
D
Referernce
Cornick and
Helgerson
(2004)
Weitjens et
al. (1999)
Hutchison et
al. (2004)
Hutchison et
al. (2004)
Maddox-
Hyttel et al.
(2006)
Chicken
Log10
range
a
b
NRh
2.8
-1.0
6.5
4.5
NAf
NA
CO
'co
CO
GO
W
D
CD
Q.
><
-I—*
CD
Q.
CO
W
D
Dj
-o
CD
Q.
><
-i—*
CD
3
C
CO
F
F
Reference
Cox et al.
(2002)
Kraft et al.
(1969)
Notes:
a Units are Iog10(cfu L"1 or oocysts L"1 or cysts L"1). "a" denotes the minimum observed value and "b" denotes the maximum observed value
b Basis refers to weight basis for manure. D denotes dry weight and W denotes wet weight
c Sample type is either composite (C) or direct (D)
d Poultry manure type is litter (L) or fresh (F)
6 All cattle and swine fecal abundances reported are for solid, fresh fecal samples (not slurries or treated manure)
f Not applicable
9 Not detected
h None reported
' Geometric mean among samples (minimum abundance among positive samples not provided in the original study)
j Samples taken at random from the top of the litter pile; because the droppings were fresh, it is presumed they were derived from a single bird
k Estimated from data presented graphically
August 2010
23
-------
U.S. Environmental Protection Agency
Table 7. Review of large-scale studies of Salmonella prevalence on livestock operations
Study
Livestock
Species
Regional Variations
Serotype Prevalence Trends
Ebel et al.
(1992)
Spent
laying hens
Prevalences of Salmonella enteriditis
(SE) in positive layer houses
Northern U.S.: 45%
Southeastern U.S.: 3%
Central/Western U.S.: 17%
No significant regional differences in
occurrence of Salmonella reported; serotype
prevalence among positive flocks not
assessed
Foley et al.
(2008)
Swine and
chickens
Not considered
Chickens:
Prior to 2000, serotype S. heidelberg
replaced serotype SE as the predominant
serotype among U.S. chickens. The decline in
SE prevalence may be the result of targeted
programs or in development of natural
resistance among chicken flocks. Salmonella
(all serotypes) prevalence is highly variable,
varying with time and with operation/chicken
type.
Swine:
From 1986 to 1995, S. cholerasuiswas the
most common serovar among U.S. swine
isolates. In 1995, Derby was identified as the
most common serotype, and since 1996,
Typhimurium has been most common. As
with chickens, prevalence of Salmonella
infection is highly variable from year-to-year
and herd-to-herd; estimates of overall
prevalence were 1.4-3.1% and 3.4-33%.
Fossleret al.
(2005a)
Dairy cows
Salmonella shedding was more
prevalent in midwestern states (Ml,
MN, Wl) than in NY
Not considered
Garber et al.
(2003)
Layers
Overall, SE was isolated from 7.1%
of U.S. layer houses, regional
prevalence estimates were:
Southeast: 0%;
Central region: 9.0% (standard error
= 7.2);
West: 4.4% (standard error = 2.5);
and
Great Lakes region: 17.2% (standard
error = 13.7)
Only SE studied; regional differences in
prevalence may differ for other serotypes
Kabagambe
et al. (2000)
Dairy cows
Herds from the south (defined as
study states CA, NM, TX, FL, TN)
had 5.7* odds of Salmonella
shedding (from at least one animal)
than herds from the north (study
states OR, WA, ID, MN, IN, IO, Ml,
Wl, NO, OH, NY, VT, and PA).
Prevalences (% of herds with at least
one shedder) were:
South: 66.7%
North: 15.7%
Not considered
August 2010
24
-------
U.S. Environmental Protection Agency
Study
Livestock
Species
Regional Variations
Serotype Prevalence Trends
USDA(2000)
Layer
houses
Environmental samples of manure,
egg belts, elevators, and walkways
in layer houses used to estimate the
prevalence of SE. Regional
prevalences (based on at least one
positive sample per layer house
studied) were:
Great Lakes: 17.2%
Southeast: 0%
Central: 9.0%
West: 4.4%
All farms: 7.1%
Only SE was studied; non-detects of SE in
the south may reflect a difference in serotype
prevalence from other regions rather than a
complete absence of Salmonella
USDA
(2009a)
Swine
operations
Regional differences in swine
Salmonella prevalence were not
reported; overall prevalence was
7.2% of fecal samples positive and
52.6% of sites with at least one fecal
sample positive
Top four serotypes from Salmonella isolates:
Derby: 29.6% of isolates and 23.0% of sites
Typhimurium (Copenhagen): 22.6% of
isolates and 15.6% of sites
Agona: 10.8% of isolates and 9.6% of sites
Anatum: 7.5% of isolates and 5.2% of sites
The isolate prevalence among swine has not
changed since 1995
USDA
(2009b)
Dairy cows
Overall percentage of Salmonella-
positive dairy operations has
increased with each APHIS study
since 1996, from 21.1% in 1996 to
39.7% in 2007.
Percentage of Sa/mone//a-positive
cows increased from 5.4% in 1996 to
13.7% in 2007; percentage of
operations with any Salmonella-
positive samples by herd size:
From 1996 to 2007, the most common
serotypes from dairy cattle isolates shifted
from S. Montevideo and Meleagridis to Cerro
and Kentucky
Region
West
East
All
operations
Herd size
1 to
499
30.0
42.9
41.5
500 or
more
36.4
79.5
61.0
USDA
(2009c)
Beef cow-
calf
operations
Number of Sa/mone//a-positive cow-
calf operations did not vary by region
or farm size. Number of positive
operations declined from 11.2% in
1997 to 9.2% in 2007/8. Number of
positive sampled cows declined from
1.4% in 1997 to 0.5% in 2007/8.
Serotype Montevideo identified most among
isolates (17.6%), followed by I 6, 7:k:- (8.8%)
and Braenderup, Meleagridis, Newport and I
August 2010
25
-------
U.S. Environmental Protection Agency
Study
Livestock
Species
Regional Variations
Serotype Prevalence Trends
Wells et al.
(2001)
Dairy cows
Regional herd-level prevalences (a
positive herd had at least one
positive fecal sample):
Northwest: 20.0%
Midwest: 15.4%
Northeast: 9.5%
South: 45%
Authors noted farm-size differences
in shedding rates that may manifest
as regional differences if sizes of
farms are different in different
regions.
48 different serotypes were identified in the
studied herds, with the top 5 most prevalent
being Montevideo (21.5% of isolates), Cerro
(13.3%), Kentucky (8.5%), Menhaden 7.7%),
and Meleagridis (6.1%).
Distribution significantly different from prior
reported national distributions of serotypes in
which serotype Typhimurium was the most
commonly identified serotype. Regional
differences in the abundance of individual
serotypes not reported.
Table 8. Large-scale studies of Campylobacter prevalence among U.S. livestock operations
Study
Livestock
Species
Regional Variations
Species Prevalence Trends
Harvey et
(2004)
al.
Dairy cattle
Prevalences of Campylobacter relatively low and
regional differences not significant, regional
prevalences were:
Northeast: 2.9%
Desert southwest: 5.2%
Pacific west: 5.0%
70% of Campylobacter
determined to be C. jejuni;
other studies report a range of
species prevalence among
dairy cattle isolates
USDA
(2009b)
Dairy cattle
Campylobacter found to be present on most dairy
operations; in 2007, at least one healthy cow in
92.6% of sampled operations (n = 121) was
shedding Campylobacter. Percentage of cows
positive for Campylobacter decreased from 51.3%
in 2002 to 33.7% in 2007.
In 2007, number of C. coll
isolates was very small
compared with the number of
C. jejuni isolates
Wesley et
(2000)
al.
Dairy cattle
No significant differences in herd-level and animal-
level prevalence of Campytobacferforthe north
(ID, IL, IO, Ml, MN, NY, OR, PA) and south (CA,
FL, TN, TX) regions of the United States
80.6% of herds positive for C.
jejuni and 19.4% of herds
positive for C. coli.
30.3% of cows positive for C.
jejuni and 2.5% positive for C.
coli.
August 2010
26
-------
U.S. Environmental Protection Agency
Table 9. Review of large-scale studies of E. coli O157 prevalence on livestock operations
Study
Livestock
Species
Overall Prevalence and Regional Trends
Observations and Notes
USDA
(2003)
Dairy cows
Overall:
Prevalence for culture-positive E. coli O157 was
4.3%; 38.5% of operations had one or more
positive cows.
Prevalence highest in summer (June - 8.2%) and
lowest in spring (April - 1.5%).
Regional:
Highest prevalence of positive cows found in West
region (7.6%), Midwest (3.5%), Southeast (3.1%),
and Northeast (1.6%).
Large operations (>500 cows) were more likely to
have positive samples than medium operations
(100-499 cows) or small operations (<100 cows).
The majority of large dairies are in the West region.
Samples collected from
March to September 2002;
total of 3,733 samples for
culture and ID of E. coli O157,
sfx 1, six 2, and antigens.
Samples collected from 5
operations from each of the
21 participating states:
• West region: CA, CO, ID,
NM, TX, WA
• Midwest region: IL, IN, IA,
Ml, MN, MO, OH, Wl
• Northeast region: NY, PA,
VT
• Southeast region: FL, KY,
TN, VA
USDA
(2001)
Beef cattle
(feed lot)
Overall:
Ranged from 3.3 % in samples taken in the winter
(Feb) to 19.9% in fall samples (Sept). No
geographic trends for STEC prevalence. All
feedlots had at least one positive sample during
the study.
Regional:
Prevalence of culture-positive samples per region:
8.4% in Middle Region (CO, KS, and OK), 11.5% in
Northern Region (ID, IA, NE, SD, WA), 13% in
Southern Region (CA, NM, TX)
73 feedlots/422 pens in 11
leading cattle feeding states
sampled for STEC from Oct
'99 to Sept '00. Total of
10,415 samples.
Samples from pens for cattle
that had been on feed the
shortest (13.9%) were more
likely to be positive than
samples from pens for cattle
that had been on feed the
longest (8.6%).
USDA
(1998)
Dairy cows
Overall:
24.2% of operations and 30.9% of markets had at
least one culture-positive E. coli O157 sample.
Prevalence on farm - 0.9% of samples positive.
Prevalence for cows to be culled with 7 days -
2.8% of samples positive.
Prevalence for culled dairy cows at markets - 1.8%
of samples positive.
Regional:
Authors did not comment on regional differences.
Fecal samples collected from
91 dairy operations and 97
cull dairy cow markets in 19
states during a one-time
sampling event.
Samples collected from Feb-
July1996. Seasonal pattern
of shedding was observed,
samples more likely to be
positive after May 1 than
before May 1.
No significant differences
found between cows on farm
and cows going to slaughter.
Prevalence was higher for
herds with 100 or more cows
(39.1% of herds had at least
one positive sample) than for
herds with fewer cows (8.9 %
of herds had at least one
positive sample), however
seasonality may have been a
factor.
August 2010
27
-------
U.S. Environmental Protection Agency
USDA
(1995)
Beef cattle
(feed lot)
Overall:
Prevalence 1.61% of collected samples.
Regional:
Prevalence of positive feedlots per region: 59.4%
in Middle Region (CO, NE, KS, OK), 58.3% in
Northern Region (ID, IA, IL, MN, SD, WA), 71.9%
in Southern Region (CA, AZ, TX)
Pens in feedlots from 13
leading cattle feeding states
sampled for E. coli O157:H7
in fall of 1994. Total of
11,881 samples.
Samples from pens for cattle
that had been on feed the
shortest (47.1%) were more
likely to be positive than
samples from pens for cattle
that had been on feed the
longest (16.8%).
Table 10. Review of large-scale studies of Cryptosporidium prevalence on livestock operations
Study
Livestock
Species
Overall Prevalence and Regional Trends
Observations and Notes
USDA
(1994)
Beef calves
from beef
cow/calf
operations
(Fresh fecal
samples
included both
diarrheic
calves <3
months and
nondiarrheic
calves <6
months)
Overall:
Prevalence of positive calves was 20.1% for
diarrheic calves and 11.2% for nondiarrheic
calves.
Prevalence of positive operations submitting
samples from diarrheic calves was 39.1%.
Prevalence of positive operations submitting
samples from nondiarrheic calves was 41.8%.
Prevalence was related to and decreased with
age of calves (23.1% for 1-30 days old; 9.2%
for>121 days old).
Regional:
No discussion regional differences in
prevalence.
Study included 391 samples from
diarrheic calves from 69
operations and 1,053 samples
from nondiarrheic calves from 141
operations.
Average age of diarrheic calves
testing positive was 41.1 days.
Average age of nondiarrheic
calves testing positive was 75.8
days.
Shedding was common in calves
of beef herds whether the calves
had diarrhea or not.
USDA
(1993)
Dairy calves
(preweaned)
Overall:
Prevalence across U.S. 22% of calves and
>90% of farms.
Prevalence increased slightly with herd size but
still high prevalence on (about 80%) on smaller
farms (<100 cows).
Prevalence higher in summer months than in
other months.
Prevalence was highest in heifers 1-3 weeks
old (>50%). Prevalence drops to <15% for
calves over 5 weeks old.
Regional:
Prevalence higher in western herds; authors
note that these are also the largest operations.
Study included 1,103 farms in 28
states, with 7,369 samples
collected.
States included:
• West: WA, OR, CA, ID, CO
• Midwest: NE, IA, MN, Wl,
Ml, IL, IN, OH
• Northeast: ME, VT, NH, NY,
PA, CT, MA, Rl
• Southeast: VA, NC, TN, GA,
AL, FL, MD
August 2010
28
-------
U.S. Environmental Protection Agency
Table 11. Review of a large-scale study of Giardia prevalence on livestock operations
Study
Livestock
Species
Overall Prevalence and Regional Trends
Observations and Notes
USDA
(1994)
Beef calves
from beef
cow/calf
operations
(Fresh fecal
samples
included both
diarrheic
calves <3
months and
nondiarrheic
calves <6
months)
Overall:
Prevalence of positive calves was 26.9% for
diarrheic calves and 45.9% for nondiarrheic
calves.
Prevalence of positive operations submitting
samples from diarrheic calves was 63.8%.
Prevalence of positive operations submitting
samples from nondiarrheic calves was 90.8%.
Prevalence peaked in calves 61-90 days old
(59.6%) and decreased with age of calves
(29.9% for calves >121 days old).
Regional:
No discussion of regional differences in
prevalence.
Study included 391 samples from
diarrheic calves from 69
operations and 1,053 samples
from nondiarrheic calves from
141 operations.
Average age of diarrheic calves
testing positive was 47.1 days.
Average age of nondiarrheic
calves testing positive was 79.1
days.
Shedding common in calves,
especially older calves, of beef
herds whether the calves had
diarrhea or not.
These large-scale studies were "snapshots" of prevalence of the pathogens. Prevalence was
either as herd prevalence (the fraction of herds for which one or more samples were positive for
the pathogen of interest) or individual level (the number of positive samples at a farm or among
all samples in a study). Because large-scale studies were snapshots, data from these studies must
be interpreted in light of known seasonal and non-seasonal (e.g., epidemic) variations in
pathogen prevalence and prevalence of specific pathogen species, serotypes, (etc.), and in light
of market and other conditions that differentiate the regions from each other. For example, farm
size is known to influence prevalence of both Salmonella and Campylobacter in most livestock
operations. Regions with a greater proportion of large operations may have an attendant high
prevalence of shedding of these pathogens. In those cases, care should be taken to associate at
least some of the cause for shedding prevalence difference among regions with farm size rather
than other regional factors such as climate.
3.3.1. Large-scale studies of Salmonella prevalence
More large-scale studies have focused on Salmonella prevalence than any of the other reference
pathogens (Table 7). Studies of layers and dairy cows indicate a higher prevalence of
Salmonella shedding in the Midwest and very low levels of shedding in the Southeast. Notably,
at least two large-scale studies considered only Salmonella enteriditis serotype enteriditis and so
may have missed the presence of unexpected Salmonella serotypes among animals in the
Southeast. A single study (Kabagambe et al., 2000) observed higher shedding in states broadly
classified as "south" than those classified as north, though Florida was the only southeastern
state included in that study. Regional differences in shedding prevalences were not observed for
swine of beef calf-cow operations.
No studies evaluated regional differences in the prevalence of Salmonella isolate serotypes.
However, multiple studies noted significant shifting in the dominant serotypes among layer
operations, dairy operations, and beef calf-cow operations for the period between 1996/7 and
2007/8.
August 2010
29
-------
U.S. Environmental Protection Agency
33.2. Large-scale studies of Campylobacter prevalence
Large scale studies of Campylobacter prevalence were conducted only for dairy cattle. These
study results (Table 8) are difficult to interpret. All studies indicate a relatively uniform
prevalence of Campylobacter across the conterminous United States and a preponderance of C.
jejuni isolates. However, one study indicates low prevalence (Harvey et al., 2004) whereas other
studies (Wesley et al., 2000; USD A, 2009b) indicate high prevalence among operations and
among individual animals. It is not possible to generalize the findings of these studies, except to
note that they do not indicate regional differences in either Campylobacter prevalence or the
prevalence of individual Campylobacter species.
3.3.3. Large-scale studies of E. coli O157:H7 prevalence
Large-scale studies of E. coli O157:H7 prevalence were conducted for dairy and beef cattle.
These study results (Table 9) indicate consistently low albeit widespread prevalence of E. coli
O157:H7 across the conterminous United States for dairy operations and for cattle on feed. The
results of these studies indicate slightly higher prevalences for feedlots in southwestern states
(i.e., California, Texas, Arizona, New Mexico). Authors of a study relating these national results
to farm factors (USDA-APHIS-VS, 1997) believe the presence of barley in the diet of cattle on
feed in these states may be a region-based contributing factor. Additionally, in feedlots,
prevalence was higher in cattle that had been on the feedlot a short time compared to cattle that
had been on the feedlot for a longer time. In one study of dairy operations (USDA, 2003), again
there was a slightly higher prevalence in the western states (California, Colorado, Idaho, New
Mexico, Texas, Washington). This finding may be related to the size of the operation; the largest
dairy operations are in this region. In all of these large-scale studies, higher prevalence occurred
during the summer/fall months. And, when evaluated, prevalence was higher for larger herds
than it was for smaller herds (USDA, 1997, 1998).
3.3.4. Large-scale studies of Cryptosporidium and Giardia prevalence
The USDA conducted a large-scale study of Cryptosporidium in dairy calves (USDA, 1993) and
a large-scale study of Cryptosporidium and Giardia in beef calves (USDA, 1994). These study
results (Tables 10 and 11) indicate widespread prevalence of both Cryptosporidium and Giardia
in U.S. dairy and beef calves. The results of these studies indicate a higher prevalence of
Cryptosporidium in dairy operations in western states (i.e., California, Oregon, Washington,
Idaho, and Colorado). This finding may be related to the size of the operation; the largest dairy
operations are in this region. Prevalence of Cryptosporidium peaked in very young calves (<30
days old) while Giardia prevalence peaked in slightly older calves (31-60 days old). In the study
of dairy calves, higher prevalence occurred in the summer months. In both studies, parasites
were found in calves with and without diarrhea, indicating that many producers may be unaware
of the extent of infection in their herds. Large-scale studies of Cryptosporidium and Giardia
among beef and dairy cow/calf operations in the United States indicate that prevalence of these
parasites is widespread, and that there may be slight regional differences in prevalence with the
west having the highest Cryptosporidium prevalence of the regions studied.
August 2010 30
-------
U.S. Environmental Protection Agency
3.4. Summary
With the exception of the absence ofGiardia in chickens, all of the key zoonotic pathogens
evaluated in this report (Salmonella, Campylobacter, E. coli O157:H7, and Cryptosporidium) are
all found in cattle, swine, and chicken—though the particular species and types prevalent in the
livestock are not necessarily the species and types posing the greatest risk of human infection
and subsequent illness. All of the key zoonotic pathogens occur frequently among cattle and the
species and serotypes found in cattle feces are generally similar to those posing hazards to
humans. Swine Salmonella, Campyiobacter, Giardia, and E. coli O157:H7 tend to be the same
species/serotypes as those most commonly implicated in human infections, whereas most swine
Cryptosporidia appear to be host-adapted and pose a reduced hazard to humans. Among
chickens, the only key zoonotic pathogens that occur with some frequency and are similar to the
pathogens implicated in human infections are Salmonella and Campylobacter—both of which
are highly prevalent among U.S. flocks.
Relatively few studies reporting regional differences in pathogen prevalence for specific host-
pathogen combinations were identified. Large-scale studies of both layer and dairy cattle
operations indicate higher prevalence of Salmonella among Midwest (including Great Lakes
States) operations than in other U.S. regions. In particular, the serotype SE was frequently
encountered in Midwest layer operations but has not been reported in operations in the south.
However, these observations must be interpreted cautiously. Differences among regions
detected in these studies may relate less to intrinsic (climatic) differences among regions than to
farm management decisions and practices, such as the types of units used for housing animals
and the typical farm sizes in different regions. Additionally, studies based only on SE (rather
than all Salmonella serotypes) may not accurately indicate the presence of human-infectious
Salmonella. Large-scale studies of Campylobacter among dairy herds indicate Campylobacter
are widespread in the United States (no regional differences) and the dominant species among
positive manure samples is C.jejuni.
Large-scale studies of E. coli O157 among the largest beef and dairy operations in the United
States indicate that a low prevalence of E. coli 0157 is widespread, and that there may be slight
regional differences in prevalence with the west/southwest having the highest prevalence of the
regions studied.
August 2010 31
-------
U.S. Environmental Protection Agency
4. Farm Factors and the Occurrence of Pathogens in Livestock Manures
Livestock distributions and the relatively few large-scale studies on pathogen occurrence are
insufficient to allow an assessment of how livestock pathogens might differ between U.S.
regions. However, additional data related to farm factors (farm type, animal management, farm
size, etc.) can be used to evaluate whether pathogen occurrence and abundance from a particular
operation could be expected to fall at the top, bottom, or middle of the ranges reported in Section
3.2. This section divides farm factors into those with potential regional implications and those
without regional implications. An example of a farm factor with regional implications is farm
size. Farm sizes differ by region, with large farms often concentrated in specific regions (e.g.,
swine operations in North Carolina) and small farms in other regions. Farm factors without
regional implications are generally animal rearing and manure management choices implemented
on farms. Individual farmers choose how to raise animals and manage manure based on
operation of their farms. Although the latter choices play a role in the pathogen occurrence and
discharges from the individual farms, there appears to be no systematic or consistent reason (e.g.,
besides "tradition") for these choices to be clustered or have regional differences. However,
because there are no data to allow an evaluation of these types of factors, they are not discussed
in this report.
This section presents synopses of studies relating farm factors to the prevalence and abundance
of the key pathogens in manures. First addressed are farm factors with regional implications.
We identify those factors as farm type (e.g., cattle feed lot vs. dairy operation vs. pasture
operation) and farm size. Next discussed are results from longitudinal studies comparing
prevalence of infection among different age cohorts of animals or among animals before and
after changes in animal management. Also discussed are studies on seasonality of pathogen
prevalence. This is followed by an evaluation of farm factors without regional implications and
is limited to those farm factors for which we could obtain data for multiple pathogens and
livestock species. The section concludes with a brief review of manure management practices.
The discussion of manure management differs from the other portions of this section because it
relates to the probability that manure-borne pathogens reach receiving waters—not the
probability that the pathogens are excreted in manures.
It is important to note that the literature review for this section identified many studies
comparing organic and conventional farming practices. Most of the reported differences in
pathogen prevalences among conventional and organic farms can be explained based on intrinsic
differences between these farming practices, such as farm size (organic farms tend to be smaller),
animal housing, and age of the animals on the operation. For example, Campylobacter
prevalence increases with chicken age and organic chickens are typically older when
slaughtered. Therefore, in reviewing studies reporting pathogen prevalence differences among
conventional and organic operations, we sought to identify the underlying cause rather than
ascribing the observed pathogen prevalence to a generic organic vs. conventional difference.
The results presented cannot be used for comparing risks and benefits of organic farming. Such
a comparison must include many factors not relevant for discussion in this report. A total of 21
articles were reviewed to compare and describe the Salmonella, E. coli O157:H7, and
Campylobacter prevalence and variability reported for organic and conventional farming
August 2010 32
-------
U.S. Environmental Protection Agency
practices for cattle, swine, and chicken. Detailed reviews of those studies are provided in
Appendix D.
Differences in pathogen prevalence and abundance between CAFOs and non-CAFO operations
can be attributed to multiple factors such as herd size, animal density, selection of feed, and
others. These factors are likely more important than the distinction of whether or not a farm is a
CAFO. As noted in the USEPA (USEPA, 2005), "The U.S. Environmental Protection Agency
defines a concentrated animal feeding operation (CAFO) as an animal feeding facility that
houses more than 1,000 animal units (AU), has 300 to 1000 AU but meets certain conditions, or
is designated a CAFO by the state. The number of animal units is based on an equivalent
number of beef cattle. Therefore, 1,000 AU equals 1,000 beef cattle, 700 mature dairy cattle,
2,500 swine, 5,000 ducks, 10,000 sheep, 55,000 turkeys, or between 30,000 and 100,000 laying
hens or broilers depending on the animal waste management system employed." This definition
is a useful distinction with respect to the volume of manure produced by the operation and the
manure management techniques likely employed. However, for the purposes of this report,
designation as a CAFO is used only to provide information on whether a farm is above a certain
size threshold.
4.1. Farm Factors with Regional Implications
Regional differences in pathogen prevalence may relate to intrinsic differences between the
regions (precipitation, temperature, solar radiation, microbial ecology, and other features related
to the survival of pathogens in regions) or may relate to practices that differ by region. In
addition to regional pathogen prevalence and animal density data, the authors identified studies
from the literature that relate pathogen prevalence (either herd-level or animal level prevalence)
to farm practices and features. Some of the farm factors and features may be different among
U.S. regions and have the potential to cause different pathogens prevalences in different regions.
Examples of the factors and features that could give rise to regional differences in pathogen
prevalence are the following:
• operation type;
o cattle (dairy, beef cattle pasture, beef cattle feedlot)
o chicken (broilers, layers, broiler-breeder)
o swine (farrow to finisher operations or birth to market; farrow to feeder operations
or birth to about 15 days; feeder to finisher operations)
o or for any livestock, use of organic or conventional farming practices (as self-
defined in studies); and
• farm size.
Because different operations house animals of different ages and because, in many cases,
prevalence of pathogens varies with animal age, this section includes a discussion on
longitudinal studies of pathogen prevalence in livestock cohorts. The results of these studies
provide additional data for understanding and interpreting pathogen prevalence differences
reported for operations of different types. Likewise, seasonality provides insights into some
findings of the farm factors studies.
August 2010 33
-------
U.S. Environmental Protection Agency
4.1.1. Farm size
Regional difference in the proportion of large farms might give rise to regional differences in
pathogen prevalences in livestock manures. As described below, there is a general association
between larger farms and increased prevalence of pathogens among herds (herds with at least
one positive sample) and within herds (prevalence from sample to sample). However, this
association is not universal and studies reporting farm-size related differences cannot directly
identify the mechanisms by which larger farms are prone to higher pathogen prevalences.
Cattle
An association between large dairy herd sizes and Salmonella prevalence in feces has been
observed in numerous studies (Kabagambe et al., 2000; Wells et al., 2001; Huston et al., 2002;
Warnick et al., 2003; Cummings et al., 2009). Although the mechanism(s) underlying the
association of Salmonella infection with large herd sizes is unclear, it may relate to intensive
management practices, introduction of more cows, and stress due to crowding, transportation,
and animal mixing (Huston et al., 2002). Huston and colleagues studied dairy cattle from 105
Ohio dairy farms and found that the odds of a herd being infected by Salmonella increased by
about 5% per each 25 additional cows on the farm. In a large study of Salmonella at dairy
operations, Wells et al. (2001) observed prevalence of Salmonella in fresh feces for farms with
fewer than 100 head to be 0.7% during late winter (February to April) and 0% during early
summer (May to July). For farms with more than 100 head, the prevalences were 3.3% and
14.0%, respectively. Fossler and colleagues conducted the largest study of Salmonella shedding
in dairy cows (Fossler et al., 2005a) and the only one evaluating herd level characteristics.
Salmonella shedding was more likely on farms with at least 100 cows. A single study (Fossler et
al., 2005b) was identified that did not report an association between dairy herd size and
prevalence of Salmonella in feces. We find that the weight of evidence points toward increasing
prevalence of Salmonella shedding with increased cattle herd size for both dairy and beef
operations.
Several studies associating cattle herd sizes with Campylobacter shedding at both the herd and
animal level were identified. Hoar et al. (2001) found an association between herd size,
measured as number of females on the farm, and Campylobacter fecal shedding prevalence on
beef cattle farms. Ellis-Iversen et al. (2009) observed that larger cattle herd size was associated
with increased fecal shedding of Campylobacter on dairy and beef cattle operations in England
and Wales. Sato et al. (2004) found that Campylobacter prevalence was significantly higher in
smaller farms than in large farms. Wesley et al. (2000) could not associate increased dairy farm
Campylobacter shedding prevalence with herd size at the herd level, but did find that increased
fecal shedding at the animal level was associated with larger farms. Together, these studies point
to increased Campylobacter shedding prevalence with increasing farm size.
Two large-scale studies of E. coli O157 occurrence among the largest U.S. dairy operations
found that prevalence was related to farm/herd size. One study (USDA, 2003) found that large
operations (>500 cows) were more likely to have positive E. coli O157 samples than medium
operations (100 to 499 cows) or small operations (<100 cows). Another similar study (USDA,
1998) found that prevalence was higher for herds with 100 or more cows (39.1% of herds had at
August 2010 34
-------
U.S. Environmental Protection Agency
least one positive sample) than for herds with fewer cows (8.9 % of herds had at least one
positive sample). When comparing prevalence of Shiga toxin producing E. coli (STEC) at
conventional and organic dairy farms, Cho et al. (2006) found significantly higher prevalence in
organic farms when data from all farms were combined, but found no difference in prevalence
between organic and conventional farms when only farms with 100 or fewer head were included
in the analysis. Ellis-Iversen et al. (2009) associated increased cattle E. coli O157 prevalence
with farm size.
Hoar et al. (2001) did not find an association between herd size (measured as number of females
on the farm) and Giardia fecal shedding prevalence on beef cattle farms. However, in a
nationwide study of dairy calves, the USDA (USDA, 1993) found that almost all dairy farms
with over 100 milks cows tested positive for Cryptosporidium, while farms with <100 cows had
slightly lower prevalence (-80%).
Swine
A single study was identified assessing the influence of swine herd size on shedding prevalence.
Kurd et al. (2002) conducted a longitudinal study of pigs before and after transport to abattoirs
(slaughter houses) to determine the role transportation and housing conditions at the abattoir play
on acquisition of Salmonella infections among pigs. All farms selected for the study were small
to moderate sized operations (approximately 193 sows) located in Midwestern states. Although
linear regression of infection rates demonstrated no effect of herd size on infection rates, the
statistical power of this finding was low because of small sample size.
Chicken
Ebel et al. (1992) describe a general trend of greater SE prevalence at larger farms, citing
differences in prevalence among Canadian and northern U.S. farms as evidence of this trend. In
a nationwide study of layers (Garber et al., 2003), approximately 4% of houses with fewer than
100,000 layers were environmentally positive for SE, whereas 16.5% of houses with 100,000 or
more layers were environmentally positive for SE. Young et al. (2009) summarized the
prevalence of zoonotic and potentially zoonotic bacteria in organic and conventional chicken
using systematic review and meta-analysis methodology. Those results indicate that the
prevalence of Salmonella spp. was higher in conventional laying hen flocks than in organic
flocks. This finding could however be confounded by larger flock sizes.
A limited number of studies (Berndtson et al., 1996; Newell and Fearnley, 2003) indicate that
chicken Campylobacter prevalence increases with farm size. Mechanisms responsible for this
finding may be greater numbers of animals feeding and watering from common sources.
4.1.2. Operation type
Operation types differ from each other with respect to typical herd/flock sizes, animal densities,
feed, hygienic conditions, manure handling, and other features. As shown in Section 2, the
geographic distributions of different types of operations are uneven in the United States. In
particular, feedlot cattle are far more prevalent in the Midwest states than in other regions. The
August 2010 35
-------
U.S. Environmental Protection Agency
following section compares operations where different types of animals are kept, different age
classes of animals are kept, or animals are housed and/or fed in a specific manner.
Cattle
Cattle operations include beef cattle calving operations, dairy cattle calving operations, dairy
operations, beef cattle fed on pasture, and beef cattle in feedlots. There are also differences
between organic and conventional operations. With respect to shedding prevalence of
pathogens, the primary differences in these operations are the following:
• the presence or absence of calves;
• the feed provided to cattle;
• whether cattle are on feedlots (in pens), on pasture, or in irrigated pasture;
• manure handling practices;
• use of antibiotics; and
• animal density.
In a study of the impact of transport on shedding of Salmonella and Campylobacter, Beach et al.
(2002) sampled rectums of feedlot cattle and adult pasture cattle before and after transport to
facilities for slaughter. There were marked differences in shedding prevalence among feedlot
and pasture cattle: 64% (n = 100) of feedlot samples and 6.3% (n = 96) of pasture cattle samples
were positive for Campylobacter. Similarly, Riley et al. (2008) determined that feedlot steers
had greater odds of Campylobacter detection (odds ratio 8.5; 95% confidence interval [3.7,
19.5]).
Midwestern states were more likely to have cattle shedding Salmonella than cattle from New
York. Cattle that had been treated with antibiotics within 14 days were less likely to be shedding
Salmonella (Fossler et al., 2005a,b,c; Wilhelm et al., 2009). In the United States, there was a
significantly higher (p = 0.0001) seroprevalence of Salmonella from anti-microbial free herds
(54%) than from conventional indoor reared herds (39%), although there were some
geographical variations in Salmonella. Wisconsin had the highest prevalence at 59%, followed
by North Carolina at 34%, and Ohio at 34% (Gebreyes et al., 2008).
Bae et al. (2005) studied the prevalence of Campylobacter in cattle at different farm types.
Prevalence of C. jejuni and C. coli excretion differed by farm type. The highest C. jejuni
prevalence was observed at beef cow-calf operations (47.1%) and the lowest at calf rearing
operations (23.8%). The highest C. coli prevalence was at calf rearing operations (20.0%) and
the lowest was at beef calf-cow operations (0.6%). Comparing feedlot and pasture beef cattle,
Krueger et al. (Krueger et al., 2008) found higher Campylobacter shedding prevalence among
pasture cattle than cattle fed concentrate. Among cattle shedding Campylobacter, abundance in
feces were a factor of 10* less for animals fed concentrate than pasture-fed animals. Riley et al.
(2008) determined shedding prevalences of steer fed on pasture or in feedlots. Feedlot cattle had
greater odds (odds ratio 8.5; confidence interval [3.7, 19.5]) of shedding Campylobacter than
steers grazing wheat. In a study that evaluated conventional and organic dairy herds, the
prevalence of Campylobacter spp. in organic and conventional farms was 26.7% and 29.1%, and
the prevalence was not statistically different between the two types of farms (Sato et al., 2004).
August 2010 36
-------
U.S. Environmental Protection Agency
Different feedlot operations use different feed mixes. In a large U.S. study of farm factors
affecting the prevalence of E. coli 0157 on beef cattle feedlots (USDA, 1997), samples from
cattle pens receiving barley in the ration were 2.75x more likely to be positive for E. coli O157
than samples from cattle pens receiving no barley. The prevalence of E. coli O157:H7 was
reported as 14.8% for organically raised cattle, 14.2% for naturally raised1 cattle, and 11.2% for
conventionally raised cattle (Reinstein et al., 2009). There were no statistically significant
differences in herd-level prevalence of STEC in the United States but researchers found greater
individual prevalence on organic dairy farms (Cho et al., 2006). No differences in STEC
prevalence were found between organic and conventional farms in Switzerland (Kuhnert et al.,
2005) or the Netherlands (Franz et al., 2007).
Swine
Davies et al. (1997) compared Salmonella prevalence in swine operation feces for multiple-site
production (different phases of production raised on separate sites) and all-in-all-out (AIAO)
management (all animals are removed from a location before introducing a new group) of both
nursery and finisher phases of production to prevalence in traditional operations at which
multiple age groups were on the farm simultaneously. Contrary to expectations, Davies and
colleagues found that Salmonella prevalence was not lower in multiple-site production systems
using all-in-all-out management of finishing pigs compared with conventional farrow-to-finish
systems. As with many other studies, very high variability in both the Salmonella prevalence
and serotype prevalence among positive samples was noted. Conflicting results were reported in
studies that examined the prevalence of Salmonella spp. in swine on farms and at slaughter in the
United States, Denmark and Germany. Studies conducted in the United States showed higher
Salmonella prevalence in organic farms (Gebreyes et al., 2008) whereas international studies
showed contrary results.
Chicken
Campylobacter in chicken has received relatively more attention on the topic of organic vs.
conventional farming practices than most of the other animal/pathogen combinations. Heuer et
al. (2001) evaluated three rearing systems (organic, conventional, and extensive indoor) and
reported that Campylobacter spp. were isolated from 100% of organic broiler flocks, 36.7% of
conventional broiler flocks, and 49.2% of extensive indoor broiler flocks. The proportion of
Campylobacter positive flocks was significantly higher for organic flocks compared with the
others. Furthermore, they found that no single factor related to organic broiler production can be
pointed out as the sole determinant of high Campylobacter prevalence. Rather, the prevalence
results reported reflect the combined effect exerted by factors that are inextricably related to each
broiler rearing system.
The above results are similar to those reported by (1) Luangtongkum et al. (2006) who indicate
that the prevalence of Campylobacter on conventional broiler farms was slightly lower (44 to
80%) than organic farms (70 to 100%) and the prevalence on conventional turkey farms was
1 The study does not explicitly define "naturally raised cattle." In general, natural rearing programs disallow use of
antimicrobials, ionophores, and hormones, which necessitates special animal management, including handling of
sick animals (SDES, 2007).
August 2010 37
-------
U.S. Environmental Protection Agency
similar (63-98%) to organic farms (6-100%); (2) Young et al. (2009) who report in a review
article that the prevalence of Campylobacter was higher in organic broiler chickens than those
raised conventionally; and (3) Van Overbeke (2006) who found that Campylobacter infections
were significantly higher in organic flocks (however, organic flocks were slaughtered at 12
weeks, compared to 6 weeks for conventional flocks).
4.1.3. Longitudinal (life stage) studies
For each animal-pathogen combination, there may be variation in the within-herd prevalence and
abundance of the pathogen in feces as animals age, are moved between facilities, are provided
different diets, or are in contact with other animals of the same or different species. Numerous
studies have explored longitudinal variation in infection rates from birth to slaughter and, as
described below, noted consistent trends in pathogen shedding prevalence consistent with age
groups. In this section, studies that have identified trends in shedding prevalence for specific
livestock host-pathogen combinations are summarized.
Cattle
Besser et al. (2005) conducted a longitudinal study of Campylobacter jejuni prevalence in fresh
feces at U.S. cattle feedlots. In that study, prevalence of C. jejuni increased markedly from cattle
arrival. In samples from pens with newly-arrived cattle C. jejuni was detected in 1.6% of fecal
samples. In samples from pens with animals within two weeks of slaughter prevalence was
62.2%. Campylobacter appears to be persistent in at least a portion of feedlot cattle (Inglis et al.,
2004). Among those cattle, species prevalence of Campylobacter varied between lots and from
season to season. In their 4-month controlled study of Campylobacter in feedlot cattle, Inglis et
al. (2004) observed C. lanienae to be the most prevalent species, followed by C. jejuni. No C.
coli were detected. In a study comparing conventional and organic dairy herds in Wisconsin,
Sato et al. (2004) found that Campylobacter prevalence was significantly higher in calves than in
cows.
E. coli O157:H7 appears to differ between calves and adult cattle and between cattle before and
after their arrival on feedlots. E. coli O157:H7 infection peaks in young cattle between 3 and 18
months of age, and declines thereafter (Ellis-Iversen et al., 2009a). In a large study of feedlot
beef cattle, LeJeune et al. (2004) observed a general trend of increasing prevalence of E. coli
O157:H7 among animals with their duration in the feedlot. However, during periods of highest
E. coli O157:H7 detection, prevalence in individual pens was sporadic. A reduced risk for the
presence of STEC was found for older than younger cows (Kuhnert et al., 2005).
Calves (<6 months) and particularly young calves (<2 months) are especially prone to
Cryptosporidium infection and high rates of fecal shedding. Wade et al. (2000) found
Cryptosporidium shedding prevalence was strongly dependent on cattle age, with much higher C.
parvum prevalence among calves than older cattle (Table 12). Age-related differences in
Cryptosporidium prevalence were confirmed in a study of runoff water from cattle operations.
Miller et al. (2008) measured Cryptosporidium densities in runoff from various locations on
dairy farms and estimated Cryptosporidium loading from areas housing cattle in different age
classes. Runoff from areas housing calves <2 months of age had much higher runoff water
August 2010 38
-------
U.S. Environmental Protection Agency
Table 12. Cryptosporidium shedding
prevalence among cattle of different
age groups (SOURCE: Wade et al., 2000)
Parasite
C. pan/urn
C. muris
Prevalence (%)
CO
O>
O)
CO
^
<
0.9
1.1
CO
-i—*
O
E
CD
V
2.4
0.5
CO
-I—*
o
b
•t
CD
0
1.7
CO
.C
C
.
CM
A
0
1.5
density of oocysts and overall loading than that from areas housing cattle 3 to 6 months of age.
Runoff from areas housing dry, milking, and calving cows was even lower than from areas
housing calves 3 to6 months of age.
Swine
Dorr et al. (2009) observed significant increases of Salmonella prevalence in pig feces with age.
At late nursery age, prevalence of fecal shedding of Salmonellae ranged from 5.0 to 21.7%,
whereas at slaughter infection rate varied from 25 to 63%. The authors note that other studies
have not reported monotonically-increasing shedding with age and suggest that environmental
contamination, particularly during transport, may account for increasing prevalence with age.
Alter et al. (2005) evaluated how Campylobacter varied over rearing time and found that no
Campylobacter were detectable in the feces of piglets at the day of birth. The Campylobacter
incidence rose within days to 32.8%. After transfer to the nursery unit the prevalence increased
to 56.6%. Approximately two-thirds of the pigs remained C. coli shedders in the fattening unit.
The detection rate before transportation was 79.1%.
Chicken
Stern et al. (2001) studied 32 flocks belonging to 4 major U.S. chicken producers. Low flock-
and sample-level prevalences of Campylobacter were observed for young flocks and high flock-
and sample-level prevalences were observed for flocks pre-slaughter. In most infected flocks,
increasing prevalence among samples in successive sampling events indicated rapid spread of
infection among the chickens. For many infected flocks 100% of fecal samples were positive for
Campylobacter within a short period after the initiation of infection.
In a study of Campylobacter changes with broiler chicken age, Bull et al. (2006) examined fecal
droppings from 10 houses in the U.K. over a production cycle for meat chickens.
Flock-level prevalence of Campylobacter increased from 10% (n = 10) at 18 days to 40%
between 28 and 33 days and to 60% at depletion. Once colonized with Campylobacter., flocks'
fecal droppings tended to have consistent and high densities of Campylobacter. Five flocks were
colonized by C. jejuni exclusively, one with C. coli exclusively, and one with both species.
Similar increases in flock-level prevalence of Campylobacter in broiler chickens were also
August 2010
39
-------
U.S. Environmental Protection Agency
observed by Arsenault et al. (2007) in a study of broiler houses in Quebec, Canada, and by
Gregory et al. (1997) in a study of broiler flocks in northeastern Georgia. Other investigators
have reported that the risk of flock infection increased with the age of the broilers (Berndtson et
al., 1996; Evans and Sayers, 2000). Newell and Fearnley (2003) found that the prevalence of
flock positivity is dependent on flock size and the type of production system. Flock positivity is
generally higher in organic and free-range flocks compared to intensively reared flocks
presumably due to the level of environmental exposure as well as the increased age of the birds
at slaughter.
4.1.4. Seasonality
Seasonal effects include seasonal life-cycle effects (e.g., calving periods), seasonal temperature
fluctuations (with related fluctuations in survival of microorganisms and changes in microbial
ecology), and changes in precipitation (particularly in climates with distinct dry and wet
seasons). As described below, seasonal variations in occurrence are frequently associated with
the lifecycles of the livestock. Occurrence of pathogens may differ among different livestock
age cohorts (e.g., high Cryptosporidium shedding among calves) or may have a trend with
animal age (e.g., increasing prevalence of Campylobacter with chicken age). These differences
are not intrinsically regional—a farm in the northern United States experiences the same life-
cycle-driven seasonality as a farm in the South. Other seasonal factors are less important than
life-cycle seasonality, but could cause regional differences in pathogen shedding. Among these,
the seasonal changes most directly related to potential regional difference in pathogen occurrence
are temperature fluctuations.
In the reviews of studies addressing seasonality below, attempts were made to ascertain the
underlying cause (temperature, precipitation, life-cycle, etc.) of the changes in prevalence,
though in many cases researchers did not explore or could not determine the underlying causes.
Cattle
Several studies report seasonal variations in Salmonella prevalence within cattle herds, with fecal
shedding higher in the summer months than winter and early spring (Wells et al., 2001;
Edrington et al., 2004; Fossler et al., 2005b). Edrington et al. (2004) also noted that the most
prevalent serotypes of Salmonella changed with season, though the changes may have been
related to epidemiological factors rather than consistent seasonal differences. In contrast to these
studies, Huston et al. (2002) did not find a seasonal pattern in shedding rates among cattle on a
large number of Ohio dairy farms, and Kunze et al. (2008) did not find seasonal variations in
Salmonella prevalence in feces from feedlots in the U.S. southern high plains.
Prevalence of E. coli O157 in cattle appears to be seasonal. In two large studies of E. coli O157
shedding in dairy operations, the USDA found a higher prevalence of E. coli 0157 in the
summer months (USDA, 1998, 2003). Another large USDA study of beef cattle feedlots found
the highest prevalence of E. coli O157 during September (USDA, 2001). Edrington et al. (2004)
observed a generally high prevalence of E. coli O157:H7 shedding among dairy cows on large
commercial dairy operations in the southwest United States as well as seasonality in prevalence
of shedding. In that study of relatively large farms, no positive samples for E. coli O157:H7
August 2010 40
-------
U.S. Environmental Protection Agency
were collected during winter sampling events. Summer E. coli O157:H7 was higher than winter
prevalence, but highly variable. Similarly, in a 26-month study of feedlot cattle, Berry et al.
(2007) observed higher summer and fall (22.7% in fall) prevalence of E. coli O157:H7 in manure
samples from feedlot pens than in winter samples (9.7% of samples), while Chapman et al.
(1997) observed higher spring and summer E. coli O157:H7 prevalences than winter prevalences
for U.K. cattle herds.
Miller et al. (2008) noted seasonality in Cryptosporidium oocyst runoff from coastal California
dairy operations, which was attributed to rainfall patterns. Oocyst runoff was high during the
wet season and was highest early in the wet season and during early portions of storms. In these
conditions, fecal matter is easily mobilized and density of oocysts in fecal material is relatively
high. Trotz-Williams et al. (2007) observed higher fecal shedding prevalences of C. parvum for
calves born and raised in summer months than in winter. A nationwide USDA study (USDA,
1993) found prevalence of Cryptosporidium in dairy calves to be higher during the summer
months than in other months. In a U.K. study of cattle, Hutchison et al. (2005) observed
significantly higher Cryptosporidium prevalence in June and December than in other months,
though overall seasonal trends were not observed. In contrast, Wade et al. (2000) did not find
seasonality in Cryptosporidium oocyst shedding from dairy cattle, but did find a distinct pattern
of decreasing C. parvum shedding prevalence with increasing animal age.
Sato et al. (2004) observed significantly higher cattle prevalence of Campylobacter shedding in
March than September. This observation was consistent with Wesley et al. (2000) who noted a
seasonal trend in dairy cattle shedding of Campylobacter. In their review of both dairy and beef
cattle studies, Stanley and Jones (2003) note that in both the northern and southern hemispheres
there is a bimodal, seasonal trend in Campylobacter shedding prevalence among cattle, with
northern hemisphere peaks in the spring and autumn and southern hemisphere peaks in
prevalence during climatic periods similar to the summer and autumn. Although these
observations point toward a likely climatic component to the seasonality, no clear explanation of
the role of temperature or climate was offered.
Chicken
Campylobacter prevalence in chickens appears to be seasonal (Stern et al., 2001; Newell and
Fearnley, 2003), with summer rates generally higher than winter rates, and with peak rates
varying with latitude. Campylobacter prevalence in chicken (layer) feces was observed to vary
significantly with season (Doyle, 1984). In a 42-week study, the authors observed two peak
periods of high prevalence in layer feces—September/October and April/May. During low
prevalence months, prevalence was found to be generally less than 8% of fecal samples, whereas
in high prevalence months, rate of infection was as high as 25%. However, the authors could not
relate the seasonal variation in Campylobacter prevalence to any particular factor or
phenomenon. Newell and Fearnley (2003) suggest use of ventilation during summer months as a
possible cause of higher summertime infection rates. Studies by Berndtson et al. (1996) and
Evans and Sayers (200) found that if a broiler flock is Campylobacter infected, a large
proportion of the birds within the flock is infected; and that there is a seasonal variation in the
prevalence of Campylobacter positive broiler flocks (i.e., significantly higher in the period from
May to October than the period from November to April)
August 2010 41
-------
U.S. Environmental Protection Agency
4.2. Farm Factors without Regional Implications
Farm features and practices beyond those listed in Section 4.1 are important determinants of
pathogen prevalences, but their use cannot be related to regions or states. The authors of the
studies we reviewed have attempted to relate many of these non-regional farm factors to manure
pathogen prevalence. Note that measures for reducing the carriage of pathogens in livestock are
different from those designed to limit the transport of pathogens occurring in livestock to surface
waters (Sobsey et al., 2006).
Farm practices may be modified to reduce prevalence of pathogens in livestock. Although these
practices cannot be related to spatial distribution of pathogens (no data were found on the
prevalence of farm practices in the United States), farm practices and their relationship with
pathogen prevalence are reviewed below. This section provides a survey of the relevant
literature but is not exhaustive. Doyle and Erickson (2006) and Sobsey et al. (2006) suggest the
following farm practices for reducing the carriage of pathogens in livestock:
• mono-species farms;
• genetic selection of animals resistant to colonization;
• breeding treatments (e.g., antibiotic treatment of semen, antimicrobial egg dips);
• sanitation and hygiene for farm and transportation environments;
• choice of bedding material;
• maintenance of dry litter;
• housing design;
• elimination of pathogens from water;
• elimination of pathogens from feed;
• feed withdrawal (prior to shipping and during molting);
• animal diet modifications;
• feed and water additives;
• vaccination; and
• prophylactic antibiotic treatment.
A relatively small list of these types of farming practices is reviewed in this report. The
following factors were selected because they are considered the most important and because
sufficient data were found in the literature to adequately comment on their association with
manure pathogen prevalence:
• watering practices (type of watering, use of disinfected water)
• practice of mixed farming; and
• use of BMPs for manure management.
Among these factors, manure management is clearly the most important. However, a
comprehensive review of the relationship between manure management options and pathogen
occurrence is beyond the scope of this report, which primarily deals with the occurrence of key
pathogens in livestock waste—not the treatment, fate, and transport of these zoonotic pathogens.
August 2010 42
-------
U.S. Environmental Protection Agency
Nonetheless, this report provides reviews of several studies illustrating the relationship between
manure management BMPs and pathogen loading of receiving waters.
4.2.1. Water disinfection and hygiene
Water source and use of disinfected water do not provide consistent reductions in pathogen
prevalence. Besser et al. (2005) did not observe a significant difference in feedlot cattle
Campylobacter jejuni prevalence among cattle provided chlorinated water and cattle provided
unchlorinated water. Wesley et al. (2000) determined that chlorination of drinking water was not
a protective factor for Campylobacter jejuni infection in dairy herds. However, Ellis-Iversen et
al. (2009) found an association between emptying water troughs at cattle operations and
prevalence of Campylobacter excretion. Note that this study did not explore the benefits of
disinfection.
Arsenault et al. (2007) did not observe a difference in prevalence in Salmonella infection
between chicken flocks using disinfected and undisinfected water, while LeJeune et al. (2004)
did not find a difference in fecal prevalence of E. coli O157:H7 between cattle provided
chlorinated water in their troughs and those provided unchlorinated water.
The manner in which water is offered to animals also appears to play a role in the prevalence of
pathogens in the animals. Bahnson et al. (2006) observed that pigs from herds with at least some
bowl drinkers had 8x the odds of testing positive for Salmonella than did pigs from herds with
only nipple drinkers. Berndtson et al. (1996) did not find an association between chicken
drinking facilities (bell, cup, or trough) and Campylobacter prevalence on Swedish broiler
chicken operations. Moreover, Newell and Fearnley (2003) indicate that the water source is a
low-risk factor for Campylobacter infection and water contamination usually follows rather than
precedes colonization of a flock suggesting that this is a consequence of the tracking up through
the water lines of organisms excreted from the birds.
Together, these studies show that drinking water is just one of many exposure routes for
livestock to acquire infections and the reduction in transmission of infectious agents via drinking
water might not produce a measurable reduction in prevalence of infection among herds and
flocks.
4.2.2. Mixed production
Mixed production provides a route for infection of one livestock species by another, either via
direct contact, aerosol transmission, transmission on equipment, boots, and other materials used
across operations, and possibly other routes. Mixed production may also entail use of practices
different from those used on farms where single species are raised. Despite these potential
transmission routes, the studies on mixed production did not demonstrate higher pathogen
prevalences on mixed production farms than on farms with single species and, in fact, one study
reported lower pathogen prevalence on a mixed production farm than on comparable operations
with single species.
August 2010 43
-------
U.S. Environmental Protection Agency
Boes et al. (2005) did not find a significant herd-level or animal level difference in prevalence of
Campylobacter jejuni or C. coli in swine herds that were on farms with or without mixed
production (commercial cattle or chicken production). Further, the ratio of species of
Campylobacter present in swine herds (high proportion of C. coif) was consistent across herds
with and without mixed production. This finding, along with high prevalence of C. jejuni
infection rates in both chickens and cattle, led the authors to suggest transmission among mixed
livestock populations is not an important factor in swine infection.
Wright et al. (2008) studied Campylobacter prevalence, overall and by species, in turkey and
swine farms in eastern North Carolina. Although high Campylobacter prevalence was observed
among swine and turkey populations, the prevalence of C. coli (by far the most prevalent
Campylobacter species among the swine herds studied) in swine was not a good predictor of
Campylobacter infection prevalence among turkeys on the same farm. Further, adult turkeys
were far more likely to be colonized by C. jejuni than C. coli. These findings led the authors to
hypothesize that even though turkeys and swine grown in proximity to each other may both be
colonized by thermophilic campylobacters, the C. jejuni and C. coli in the animals are likely
host-associated. Arsenault et al. (2007) indicate that risk factors for Campylobacter spp. include
vertical transmission; contamination from previous flock; and exposure to potential sources of
the bacterium, such as other animals on the farm, insects, rodents, environment, litter, and
drinking water. Odds of Salmonella colonization were 2.6x greater for chicken flocks without
permanently locked houses.
Ellis-Iversen et al. (2009) observed lower prevalence of E. coli O157:H7 infection among cattle
on farms where chicken were also present. They speculated that this lower prevalence among
cattle may be related to the adoption of practices for chicken rearing that had a mutually
beneficial effect on cattle health.
4.2.3. BMPs and manure management
Manure management (the choice of manure treatment, storage, and ultimate disposition) is a
critically important farm factor, both because it determines the pathogen loads ultimately
reaching waters used for recreation, and because it is a class of activities over which farmers can
maintain control. While it is outside the scope of this report to provide detailed quantitative
BMP and manure treatment system performance data, a brief review is provided of the role of
manure management in determining the risks of livestock fecal pollution as well as a list of
commonly used manure management techniques with their anticipated ranges in pathogen
reduction. The data and descriptions presented below may be used as a starting point for
collection of data on pathogen reduction in treatment and BMPs or for developing or refining a
QMRA framework for risks from livestock pathogens. Manure management and use of BMPs
for attenuation of pathogens in runoff to receiving waters do not relate to regional differences in
pathogen loads to streams. Although manure management and installation of BMPs are expected
to have profound effects on manure loads to receiving waters, agriculture decision making
usually takes place at the farm level (Garcia et al., 2008) and is unrelated to differences among
regions of the United States.
August 2010 44
-------
U.S. Environmental Protection Agency
A model showing pathways by which zoonotic pathogens may reach either surface waters or
groundwater is presented in Figure 6 (Goss and Richards, 2008). The sources may be in barns,
yards, or grazed fields. As described above, the prevalence and abundance of shedding depends
on the animal and farm features. Pathogens shed in barns, yards, and grazed fields may be
transported directly to receiving waters via rainwater, may be given some degree of treatment,
and/or may be applied to arable fields. Manure management determines the proportion of
pathogens present in manure that reaches streams and groundwater. The presence of tile drains
can dramatically change the connection between farm and receiving water and result in high
indicator and pathogen loads to receiving waters even for relatively small operations (USEPA,
2005).
Manure management systems are primarily designed to match nutrient content of wastes to
nutrient needs of crops (Eghball and Power, 1994; Van Horn et al., 1994; Moore et al., 1995;
Garcia et al., 2008). Nonetheless, substantial changes occur in the pathogen and fecal indicator
densities of wastes during storage and treatment. Wastes may be added intermittently to
treatment processes (e.g., compost heaps or lagoons), resulting in an accumulation of wastes with
a distribution of residence times, indicator levels, and pathogen levels.
Farm
Soil
Surface
water
Ground-
water
Animal management
Field crop management
Figure 6. Farm sources of zoonotic pathogens and pathways to receiving
waters (SOURCE: adapted from Goss and Richards, 2008)
August 2010
45
-------
U.S. Environmental Protection Agency
Overview of manure treatment systems
Bicudo and Goyal (2003) reviewed studies of manure management systems and divided the
manure management process into components that determine the number of microorganisms in
manures (e.g., choice of feed) and components relating to the ultimate discharge of those
microorganisms into surface or subsurface waters. The processes and farm factors they
described differ from the others listed in Section 4. Specifically, the other factors in Section 4
relate to the occurrence of pathogens in the fresh manures of livestock whereas the factors and
processes explored by Bicudo and Goyal (2003) relate to the pathogen loads in manures intended
for land application or other processes (after some level of treatment). The pathogen and
indicator reductions for the options they identified and assessed (Table 13) are variable and differ
between implementations at different farms and with season and other factors at a given farm. In
general, high, rapid reductions in microorganisms have attendant high-energy requirements or
use of chemicals that may be expensive or whose handling may be difficult. The adoption of
manure management strategies is likely related to the labor and capital costs associated with the
processes. Manure management strategies may be classified as passive or active on the basis of
the operational requirements of the system (USEPA, 2005). Lagoons, storage prior to disposal,
Table 13. Effect of manure management options on the number of microorganisms contained in
manure (SOURCE: Bicudo and Goyal, 2003)
Type of
Option
Animal
housing and
feed
Physical
treatment
Chemical
treatment
Biological
treatment
Management
Option
Diet changes
Production
systems
Vegetative filter
strips
Thermal
treatment
Electrolytic
treatment
Chemical
addition
Ozonation
Anaerobic lagoon
Anaerobic
digester
Aeration
Thermophilic
aerobic
stabilization
Aerobic
sequencing
batch reactor
Effect
Reduction of acid-resistant
£. co//
Reduction in Salmonella
numbers
Reduction in Salmonella
numbers in swine manure
Reduction of bacterial
indicator and
Cryptosporidium oocyst
numbers
Reduction of viruses
Reduction of bacterial
indicators
Reduction of bacterial
indicators
Reduction of bacterial
indicators
Reduction of bacterial and
viral indicators
Reduction of bacteria and
viruses
Reduction of bacteria and
viruses
Reduction of bacteria and
protozoa
Reduction of bacterial
indicators
Notes
Hay diet seems to have a significant effect on
the reduction in E. coll numbers in cattle
manure
Conflicting results from swine nutrition
experiments
Prevalence of Salmonella was lower on slotted
floors compared with all other floor types
Wide variations; reductions from 20 to 90%
have been reported
Temperatures between 60°C and 70°C appear
to be high enough to inactivate several viruses
in pig slurry
Energy costs may be a concern, as the system
consumes 26 kW per m3 of pig slurry
Does not attract much interest because of high
concentrations needed and toxic nature of
chemicals
High doses (up to 3 g/L) required to achieve a
99.9% reduction in £. coll
Reduction of enteric microbes and coliphages
varies between 90 and 99%
Rapid inactivation of pathogens is achieved
under thermophilic conditions
Temperature and time of treatment play a
significant role
Most microorganisms are inactivated within 24
hours when the temperature reaches 50°C
Reductions between 90 and 99.9% can be
achieved with treatment times ranging from 5
to 10 days
August 2010
46
-------
U.S. Environmental Protection Agency
vegetated buffer strips, constructed wetlands, separation of different ages of animals, and land
application are examples of passive systems and composting, anaerobic digesters, aerobic
digesters, and actively operated lagoons are examples of active systems.
Goss and Richards (2008) provide general ranges of bacteria reductions that may be expected in
manure treatment processes and the stresses (inactivation or removal mechanism) for each of the
processes (Table 14). The ranges proposed by Goss and Richards are not directly comparable to
those presented by Bicudo and Goyal (2003) because the classification of treatment processes
differs for the two studies. Further, given the between- and within-operation variabilities in the
treatment processes, the ranges presented in those two studies are expert judgments and, in the
case of Goss and Richards' estimates, are not specific to classes of pathogens. Table 15 (from
Sobsey et al., 2006) summarizes typical viral pathogen reductions for various physical, chemical,
and biological treatment processes. Though we are not considering manure-borne viruses to be
as significant a risk as bacterial and protozoan pathogens in this report, these data provide a
Table 14. Typical reductions of pathogens during manure
treatment processes (SOURCE: Goss and Richards, 2008)
Treatment
Lagoon
Constructed wetland
Deep stack (composting)
^. t. mesophilic
Digestion therm0philic
Composting
Air drying
Heat drying
Pasteurization
Alkaline processes
Log
Reduction
1-3
2-3
1-?
1-2
5
1-5
1-2
4-5
5
3-5
Stress
Time
Time, filtration
NH3, heat
Time, heat
Heat, time
Desiccation
Heat, desiccation
Time, heat
Heat, NH3
Table 15. Typical reductions of viruses during animal waste treatment processes (SOURCE: Sobsey
etal.,2006)
Treatment Process
Estimated Virus Reduction (Iog10)
Physical processes
Heat/thermal processes
Mesophilic
Thermophilic
Freezing
Drying or desiccation
Gamma irradiation
Typically 1-2
Typically >4
Variable; depends on temperature, type of waste and pathogen
Typically >4 at <1% moisture; typically <1 at >5% moisture
Typically >3
Chemical processes
HighpH(>11)
Low pH (< 2 to < 5)
Ammonia
Inactivation at high pH (e.g., alkaline/lime stabilization; typically >3)
Inactivation at low pH; acidification; typically <2
Inactivation at higher pH (> 8.5) where NH3 predominates
Biological processes
Aerobic, mesophilic
Aerobic, thermophilic (composting)
Silage treatment (mesophilic)
Land application
Typically 1-2
Typically >4
Variable
Highly variable and largely unknown; potentially high
August 2010
47
-------
U.S. Environmental Protection Agency
means for evaluating that assumption and are of relevance to other sources of fecal pollution
including land-applied biosolids.
Composting is a relatively simple and flexible means of manure treatment and reduction of
pathogen densities in solid manures, manures mixed with bedding or other materials, or litter.
Composting may be conducted in heaps or reactors (Heinonen-Tanski et al., 2006) and the
degree of pathogen inactivation depends upon the temperature maintained in the compost and the
availability of oxygen in the heap. Heinonen-Tanski and colleagues report large variations in
measured reductions of indicator bacteria in composted manures for composting conducted over
different durations and with different techniques. Larney and Hao (2007) compared straw-
bedded and wood-bedded windrow composting of feedlot cattle manures. Both types of compost
performed similarly and produced declines in E. coli of >99.95% during the first 7 days of
composting, despite relatively low temperatures in the compost (34 to 42°C). High inactivation
despite low temperatures lead Larney and Hao to conclude that desiccation played a significant
role in E. coli reduction.
Based on a review of the literature, Larney and Hao noted that Giardia and Cryptosporidium
reductions in composts are more gradual than those observed for E. coli and also more dependent
on achieving a high temperature in the compost heap. The authors suggest that composting for
15 days at temperatures > 55°C is adequate for inactivation of Giardia and Cryptosporidium in
feedlot wastes. Shepherd et al. (2007) explored within-stack differences in inactivation of non
stx E. coli O157:H7 inoculated into dairy manure composts comprised of dairy manure, feed
waste, sawdust, calf feces and fresh hay. Heaps were kept on a concrete slab during the
experiments and were not turned. Commensal E. coli populations in the heaps declined -4.6 to
4.9 logs during the first 3 days of composting and were not detectable via direct plating after
seven days, though they were detectable via enrichment for up to 14 days. A 6-log reduction in
E. coli O157:H7 was observed in the center and at the top of the stack after 3 days and a 4-log
reduction was observed at the bottom of the stack. After the third day surface samples from the
heap had consistently higher E. coli O157:H7 abundance than samples from elsewhere in the
stack. Inactivation ofE. coli O157:H7 and commensal E. coli were correlated.
The following generalizations may be drawn concerning composting:
• Removal of pathogens and indicators is highly variable and dependent on
o duration,
o temperature achieved in the heap, and
o mixing of the heap.
• Composting heaps may be comprised of different materials. The make-up of the compost
heap may be less important than management of the heap to maintain high temperature
and mixing.
• Inactivation is variable within heaps, with the highest temperature and removal achieved
at the heap interior.
• Vegetative bacterial removals in composting are similar and greater than those of Giardia
and Cryptosporidium.
August 2010 48
-------
U.S. Environmental Protection Agency
Land application and passive treatment systems
With respect to land-applied manure, Bicudo and Goyal (2003) identified that the application
technique and timing (with respect to rainfall) influenced microbial water quality in both surface
runoff and in subsurface drainage. Management options in land application of manures include
(Midwest Plan Service, 1993)
• timing of land application;
• land application technique;
• degree to which land-applied manures are tilled/incorporated into soil;
• selection of ground cover for lands receiving manures;
• choice of land application site; and
• mass or volume applied per land area.
Wastes must be stored prior to land application and waste holding facilities are designed to meet
storage requirements dictated either by regulations or by practical considerations like the staging
of land-application with other activities. The combination of storage time prior to land
application, timing and nature of land application determine the risks land-applied manures pose
to recreational waters.
Lewis et al. (2005) measured storm-flow fecal coliform densities in key locations on dairy farms
to determine the routes by which fecal indicators reach streams and to identify the farm
management practices with the greatest potential for reducing loads. Although between storms
loads from various portions of the dairy facilities were highly variable, the relative loading by the
sources, averaged over all storms, may be assumed mode consistent. Relative contributions from
various components of the dairy operations' manure management systems (MMS) > lots >
stockpiles > drains > runoff > pastures > gutters and the relative average fecal coliform density
in waters from these locations were MMS > stockpiles > lots > drains > runoff > pasture >
gutters. These results indicate that the greatest potential for reducing loads to streams lies in
management of runoff from manure management systems and stockpiles, whereas pastures and
gutter flows present less opportunity for transport of fecal bacteria loads to streams.
Miller et al. (2008) estimated Cryptosporidium oocyst loading to surface waters from dairy farms
similar to those studied by Lewis et al. (2005). Attempts were made to relate loading to both
infrastructure and non-infrastructure factors including age class of cattle and implementation of
specific BMPs. In this study the BMPs were vegetative buffer strips (grassy channels for
directing and slowing lot runoff) and the use of straw mulch. Vegetative buffer strips and use of
straw mulch both were associated with significant reductions in Cryptosporidium loads from
dairy operations. Each additional 10% of straw mulch coverage placed on dairy cattle high-use
areas was associated with a reduction in stormwater oocyst load by a factor of 0.76 and each
meter increase in vegetative buffer length was associated with a reduction in the stormwater
oocyst load by a factor of 0.98. Age class of cattle in a given farm area was a strong determinant
of the oocysts load in runoff from the area. Mean oocyst density and load from areas housing
calves under two months of age were 2000 oocysts/L; mean oocyst density from areas housing
older cattle was around 6 oocysts/L (Table 16). Collectively, the findings of Miller et al. (2008)
August 2010 49
-------
U.S. Environmental Protection Agency
Table 16. Dairy farm oocyst stormwater density and loading by age class (SOURCE: Miller et al.
2008)
Age Class
<2 months
3-6 months
Dry, milking, and
calving cows
Cryptosporidium
Mean Runoff
Density
2000 oocysts/L
50 oocysts/L
~ 6 oocysts/L
Mean Loading
Rate
1400 oocysts/s
320 oocysts/s
1-20 oocysts/s
Giardia
Mean Runoff
Density
989 cysts/L
38.9 cysts/L
86.9 cysts/L
Mean Loading
Rate
7908 cysts/s
1 09 cysts/s
450 cysts/s
support the targeted use of infrastructure BMPs on farm areas with greater potential for oocyst
stormwater loading (areas housing calves <2 months of age).
Miller et al. (2007) also studied the efficacy of BMPs in reducing Giardia cyst loads to receiving
waters. In addition to BMPs related to isolation of manures from young animals, the authors
examined structural BMPs, including use of vegetative buffer strips, application of straw mulch,
seed application, scraping of manure, and cattle exclusion. Among these BMPs, only vegetative
buffer strips were found to reduce Giardia load to receiving streams and the density of Giardia
in receiving streams. Other farm factors found to influence Giardia loads coming from pens
were animal age and cumulative annual precipitation. An association between event rainfall and
cyst loading was also observed. Cyst loading increased with prior 24-hours' precipitation up to a
net rainfall of 30mm. Above 30mm cyst loading was observed to decline with increasing
rainfall.
Meals and Braun (2006) compared the impacts of manure storage, timing of land application
ground cover, and tilling of manures into soils to determine which BMP reduced runoff from
land-applied manures most. Runoff density of E. coli from land-applied aged manure declined
significantly with manure age—runoff from land-applied 90-day-old manure was 99.6% less
than that from non-stored manures. Delay to rainfall (from the time of land application) also
reduced E. coli in runoff. Manures applied 1 day before rain resulted in twice the E. coli density
in runoff water than manures applied 3 days before a rain event. Neither vegetation height nor
incorporation of manures into soils alone produced a significant reduction in runoff E. coli
density, though incorporation accompanied with either storage or 3-day lag between application
and rain resulted in significant reductions in runoff E. coli density. This study indicates that
manure storage results in the greatest reduction in pathogen density compared with timing of
land application, height of vegetation, or tilling the applied manure into the soil. These results
are expected to be general to all bacterial pathogens.
4.3. Summary
Animal management factors with the potential to influence the regional and local prevalence of
the key pathogens include operation type, farm size, and whether the operation uses practices
typical of organic operations. A general association was found between increased prevalence
and larger farm size (number of animals) for cattle and chicken operations—though for at least
one host-pathogen combination (Giardia in beef cattle), this trend was not observed. Only one
study attempted to relate swine operation size to prevalence of Salmonella. In that study, no
association was found; however, the sample size was small and the results may not be
August 2010
50
-------
U.S. Environmental Protection Agency
representative of general trends. Studies indicate little difference in pathogen prevalence among
chicken operation types (layer vs. broiler) and swine operation types (all-in-all-out multisite
systems vs. conventional systems) and significant differences among cattle on feedlots and other
cattle. Prevalence of both Campylobacter and Salmonella were markedly higher among feedlot
cattle than among animals prior to their arrival at feedlots. Cryptosporidium infection among
cattle appears to be more related to animal age than housing and feed, and no studies were found
associating Cryptosporidium prevalence with cattle operation type. Differences in pathogen
prevalences between farms using organic and conventional practices could typically be explained
based on differences such as farm size (organic farms tend to be smaller), animal housing, and
age range of animals on the operation (organic chickens are typically older when slaughtered).
Although numerous animal management factors that are individual farmer choices were
identified and evaluated, they are not believed to contribute to regional differences in zoonotic
pathogen occurrence. However, these farm practices do relate to local pathogen prevalences and
points for controlling zoonotic pathogen sources. Examples of animal management practices
with the potential to influence the prevalence of zoonotic pathogens in manures include the
following: mixed-production practices; genetic selection of animals resistant to colonization;
breeding treatments (antibiotic treatment of semen, antimicrobial egg dips); sanitation for farm
and transportation environments; choice of bedding material; maintenance of dry litter;
elimination of pathogens from water; elimination of pathogens from feed; feed withdrawal (prior
to shipping and during molting); feed and water additives; and vaccination. A brief survey of
studies related to these animal management factors indicates that mixed production facilities are
not generally associated with higher prevalence of pathogens (via cross-infection), and that
drinking water chlorination may be of limited value for limiting infections among herds (though
the vessel used for delivering drinking water may be important).
The exposure of humans to zoonotic pathogens during surface water recreation relates both to the
source of the pathogens (as quantified by the occurrence and prevalence of the pathogens in fresh
livestock manures) and to the manure handling practiced on individual farms. Manure handling
entails collection, storage, treatment, and eventual use of collected manures. Each of these steps
affords opportunities for reducing pathogen densities in source materials or preventing transport
of the pathogens to receiving waters. Best management practices, including infrastructure for
isolating runoff waters from manure stocks or for promoting the removal of pathogens from
runoff waters (e.g., vegetative buffers), are effective for reducing pathogen loading to streams.
High reductions in pathogen loads may be achieved via treatment of the animal wastes.
Treatment systems may be physical, chemical, or biological, and the level of treatment varies
widely among alternative systems. In general, higher pathogen removal rates are accompanied
by higher costs (energy, chemical, or complexity of systems). Regardless of the treatment
process, removal varies widely between systems (whose designs and operations may vary) and
for systems over time. Land application of manures is typically designed based on nutrient
considerations, though how and when manures are land-applied can have profound impacts on
the runoff of manure indicators and pathogens to groundwater and streams.
August 2010 51
-------
U.S. Environmental Protection Agency
5. References
Adam, R.D. (2001) Biology of Giardia lamblia. Clinical Microbiology Reviews 14: 447-475.
Alter, T., Gaull, F., Kasimir, S., Gurtler, M., Mielke, H., Linnebur, M., and Fehlhaber, K. (2005)
Prevalences and transmission routes of Campylobacter spp. strains within multiple pig
farms. Veterinary Microbiology 108: 251-261.
Arsenault, J., Letellier, A., Quessy, S., Normand, V., and Boulianne, M. (2007) Prevalence and
risk factors for Salmonella spp. and Campylobacter spp. caecal colonization in broiler
chicken and turkey flocks slaughtered in Quebec, Canada. Preventive Veterinary
Medicine 81: 250-264.
Atwill, E.R., Hoar, B., das Gra9as Cabral Pereira, M., Tate, K.W., Rulofson, F., and Nader, G.
(2003) Improved quantitative estimates of low environmental loading and sporadic
periparturient shedding of Cryptosporidium parvum in adult beef cattle. Applied and
Environmental Microbiology 69: 4604-4610.
Atwill, E.R., Pereira, M.D.G.C., Alonso, L.H., Elmi, C., Epperson, W.B., Smith, R. et al. (2006)
Environmental load of Cryptosporidium parvum oocysts from cattle manure in feedlots
from the central and western United States. Journal of Environmental Quality 35: 200-
206.
Bae, W., Kaya, K.N., Hancock, D.D., Call, D.R., Park, Y.H., and Besser, I.E. (2005) Prevalence
and antimicrobial resistance of thermophilic Campylobacter spp. from cattle farms in
Washington state. Applied and Environmental Microbiology 71: 169-174.
Bahnson, P.B., Fedorka-Cray, P., Ladely, S., and Mateus-Pinilla, N.E. (2006) Herd-level risk
factors for Salmonella enterica subsp. enterica in U.S. market pigs. Preventive Veterinary
Medicine 76: 249-262.
Barwick, R.S., Mohammed, H.O., White, M.E., and Bryant, R.B. (2003) Factors associated with
the likelihood of Giardia spp. and Cryptosporidium spp. in soil on dairy farms. Journal of
Dairy Science 86: 784-791.
Baumler, A.J., Tsolis, R.M., Ficht, T.A., and Adams, G. (1998) Evolution of host adaption in
Salmonella enterica. Infection and Immunity 66: 4579-4587.
Beach, J.C., Murano, E.A., and Acuff, G.R. (2002) Prevalence of Salmonella and Campylobacter
in beef cattle from transport to slaughter. Journal of Food Protection 65: 1687-1693.
Berndtson, E., Danielsson-Tham, M.L., and Engvall, A. (1996) Campylobacter incidence on a
chicken farm and the spread of Campylobacter during the slaughter process. Internal!on
Journal of Food Microbiology 32: 35-47.
Berry, E.D., Woodbury, B.L., Nienaber, J.A., Eigenberg, R.A., Thurston, J.A., and Wells, J.E.
(2007) Incidence and persistence of zoonotic bacterial and protozoan pathogens in a beef
cattle feedlot runoff control-vegetative treatment system. Journal of Environmental
Quality 36: 1873-1882.
Besser, I.E., LeJeune, J.T., Rice, D.H., Berg, J., Stilborn, R.P., Kaya, K. et al. (2005) Increasing
prevalence of Campylobacter jejuni in feedlot cattle through the feeding period. Applied
and Environmental Microbiology 71: 5752-5758.
Bicudo, J., and Goyal, S. (2003) Pathogens and manure management systems: a review.
Environmental Technology 24: 115-130.
Boes, J., Alban, L., Bagger, J., Mogelmose, V., Baggesen, D.L., and Olsen, J.E. (2005) Survival
of Escherichia coli and Salmonella Typhimurium in slurry applied to clay soil on a
Danish swine farm. Preventive Veterinary Medicine 69: 213-228.
August 2010 52
-------
U.S. Environmental Protection Agency
Booher, S.L., Cornick, N.A., Moon, H.W. (2002) Persistence of Escherichia coli O157:H7 in
experimentally infected swine. Veterinary Microbiology 89: 69-81.
Bourke,B., Chan, V.L., Sherman, P. (1998) Campylobacter upsaliensis: Waiting in the Wings.
Clinical Microbiology Reviews 11(3):440-449
Bolin, C., Brown, C., and Rose, J. (2004) Chapter 2. Emerging zoonotic diseases and water. In
Waterborne Zoonoses: Identification, Causes and Control. Cotruvo, J.A., Dufour, A.,
Rees, G., Bartram, J., Carr, R., Cliver, D.O. et al. (eds). London, U.K.: IWA Publishing.
World Health Organization (WHO).
Bull, S.A., Allen, V.M., Domingue, G., Jorgensen, F., Frost, J.A., Ure, R. et al. (2006) Sources
of Campylobacter spp. colonizing housed broiler flocks during rearing. Applied and
Environmental Microbiology 72: 645-652.
Byrd, J.A., Corner, D.E., Deloach, J.R., Nisbet, D.J., and Stanker, L.H. (1998) Horizontal
transmission of Salmonella typhimurium in broiler chicks. Journal of Applied Poultry
Research 7: 75-80.
Callaway, T.R., Edrington, T.S., Anderson, R.C., Byrd, J.A., and Nisbet, DJ. (2008)
Gactrointestinal microbial ecology and the safety of our food as related to Salmonella.
Journal of Animal Science 86: E163-E172.
Callaway, T.R., Keen, J.E., Edrington, T.S., Baumgard, L.H., Spicer, L., Fonda, E.S. et al.
(2005) Fecal prevalence and diversity of Salmonella species in lactating dairy cattle in
four states. Journal of Dairy Science 88: 3603-3608.
Chapman, P.A., Siddons, C.A., Cerdan Malo, A.T., and Harkin, M.A. (1997) A 1-year study of
Escherichia coli O157 in cattle, sheep, pigs and poultry. Epidemiology and Infection 119:
245-250.
Cho, S., Diez-Gonzalez, F., Fossler, C.P., Wells, S.J., Hedberg, C.W., Kaneene, J.B. et al. (2006)
Prevalence of shiga toxin-encoding bacteria and shiga toxin-producing Escherichia coli
isolates from dairy farms and county fairs. Veterinary Microbiology 118: 289-298.
Cornick, N. A., and Helgerson, A.F. (2004) Transmission and infectious dose of Escherichia coli
O157:H7 in swine. Applied and Environmental Microbiology 70: 5331-5335.
Cox, N.A., Stern, N.J., Musgrove, M.T., Bailey, J.S., Craven, S.E., Cray, P.F. et al. (2002)
Prevalence and level of Campylobacter in commercial broiler breeders (parents) and
broilers. Journal of Applied Poultry Research 11: 187.
Craun, G., Calderon, R., and Craun, M. (2004) Chapter 8. Waterborne outbreaks caused by
zoonotic pathogens in the USA. In: Waterborne Zoonoses: Identification, Causes and
Control. Cotruvo, J.A., Dufour, A., Rees, G., Bartram, J., Carr, R., Cliver, D.O. et al.
(eds). London, U.K.: IWA Publishing. World Health Organization (WHO).
Cummings, K.J., Warnick, L.D., Alexander, K.A., Cripps, C.J., Grohn, Y.T., McDonough, P.L.
et al. (2009) The incidence of salmonellosis among dairy herds in the northeastern United
States. Journal of Dairy Science 92: 3766-3774.
Davies, P.R., Morrow, W.E.M., Jones, F.T., Deen, J., Fedorka-Cray, P., and Harris, IT. (1997)
Prevalence of Salmonella in finishing swine raised in different production systems in
North Carolina, USA. Epidemiology and Infection 119: 237-244.
Devane, M.L., Nicol, C., Ball, A., Klena, J.D., Scholes, P., Hudson, J.A. et al. (2005) The
occurrence of Campylobacter subtypes in environmental reservoirs and potential
transmission routes. Journal of Applied Microbiology 98: 980-990.
Donnison, A.M. and Ross, C.M. (1999) Animal and human faecal pollution in New Zealand
rivers. Zealand Journal of Marine and Freshwater Research 33: 119-128.
August 2010 53
-------
U.S. Environmental Protection Agency
Dorner, S.M., Huck, P.M., and Slawson, R.M. (2004) Estimating potential environmental
loadings of Cryptosporidium spp. and Campylobacter spp. from livestock in the Grand
River Watershed, Ontario, Canada. Environmental Science and Technology 38: 3370-
3380.
Dorr, P.M., Tadesse, D.A., Zewde, B.M., Fry, P., Thakur, S., and Gebreyes, W.A. (2009)
Longitudinal study of Salmonella dispersion and the role of environmental contamination
in commercial swine production systems. Applied and Environmental Microbiology 75:
1478-1486.
Doyle, M., and Erickson, M. (2006) Reducing the carriage of foodborne pathogens in livestock
and poultry. Poultry Science 85: 960-973.
Doyle, M.P. (1984) Association of Campylobacter jejuni with laying hens and eggs. Applied and
Environmental Microbiology 47: 533-536.
Ebel, E.D., David, M.J., and Mason, J. (1992) Occurrence of Salmonella enteriditis in the U.S.
commerical egg industry: Report on a national spent hen survey. Avian Diseases 36: 646-
654.
Edmonds, P., Patton, C.M., Griffin, P.W., Barrett, T.J., Schmid, G.P., Baker, C.N., Lambert,
M.A., and Brenner, DJ. (1987) Campylobacter hyointestinalis associated with human
gastrointestinal disease in the United States. Journal of Clinical Microbiology 25(4):
685-691
Edrington, T.S., Hume, M.E., Looper, M.L., Schultz, C.L., Fitzgerald, A.C., Callaway, T.R. et al.
(2004) Variation in the faecal shedding of Salmonella andE. coli O157:H7 in lactating
dairy cattle and examination of Salmonella genotypes using pulsed-field gel
electrophoresis. Letter in Applied Microbiology 38: 366-372.
Eghball, B., and Power, J.F. (1994) Beef cattle feedlot manure management. Journal of Soil and
Water Conservation 49: 113-122.
El-Shibiny, A., Connerton, P.L., and Connerton, IF. (2005) Enumeration and diversity of
Campylobacters and bacteriophages isolated during the rearing cycles of free-range and
organic chickens. Applied and Environmental Microbiology 71: 1259-1266.
Ellis-Iversen, J., Cook, A., Smith, R., Pritchard, G., and Nielen, M. (2009a) Temporal patterns
and risk factors for Escherichia coli O157 and Campylobacter spp. in young cattle.
Journal of Food Protection 72: 490-496.
Englen, M., Hill, A., Dargatz, D., Ladely, S., and Fedorka-Cray, P. (2007) Prevalence and
antimicrobial resistance of Campylobacter in U.S. dairy cattle. Journal of Applied
Microbiology 102: 1570-1577.
Evans, S.J., and Sayers, A.R. (2000) A longitudinal study of Campylobacter infection of broiler
flocks in Great Britain. Preventitive Veterinary Medicine 46: 209-223.
Payer, R., Trout, J.M., Graczyk, T.K., and Lewis, E.J. (2000) Prevalence of Cryptosporidium,
Giardia and Eimeria infections in post-weaned and adult cattle on three Maryland farms.
Veterinary Parasitology 93: 103-112.
Feder, I, Wallace, F.M., Gray, J.T., Fratamico, P., Fedorka-Cray, P.J., Pearce, R.A., Call, J.E.,
Perrine, R., Luchansky, J.B. (2003) Isolation of Escherichia coli O157:H7 from intact
colon fecal samples of swine. Emerging Infectious Diseases 9(3): 380-383.
Fenton, A., and Pederson, A.B. (2005) Community epidemiology framework for classifying
disease threats. Emerging Infectious Diseases 11: 1815-1821.
August 2010 54
-------
U.S. Environmental Protection Agency
Foley, S.L., Lynne, A.M., andNayak, R. (2008) Salmonella challenges: Prevalence in swine and
poultry and potential pathogenicity of such isolates. Journal of Animal Science 86: E149-
162.
Fossler, C.P., Wells, S.J., Kaneene, J.B., Ruegg, P.L., Warnick, L.D., Eberly, L.E. et al. (2005a)
Cattle and environmental sample-level factors associated with the presence of Salmonella
in a multi-state study of conventional and organic dairy farms. Preventive Veterinary
Medicine 67: 39-53.
Fossler, C.P., Wells, S.J., Kaneene, J.B., Ruegg, P.L., Warnick, L.D., Bender, J.B. et al. (2005b)
Herd-level factors associated with isolation of Salmonella in a multi-state study of
conventional and organic dairy farms. I. Salmonella shedding in cows. Preventive
Veterinary Medicine 70: 257-277.
Fossler, C.P., Wells, S.J., Kaneene, J.B., Ruegg, P.L., Warnick, L.D., Bender, J.B. et al. (2005c)
Herd-level factors associated with isolation of Salmonella in a multi-state study of
conventional and organic dairy farms. II. Salmonella shedding in calves. Preventive
Veterinary Medicine 70: 279-291.
Franz, E., Klerks, M.M., De Vos, O.J., Termorshuizen, A.J., and van Bruggen, A.H. (2007)
Prevalence of Shiga toxin-producing Escherichia coli stxl, stx2, eaeA, and rfbE genes
and survival of E. coli O157:H7 in manure from organic and low-input conventional
dairy farms. Applied and Environmental Microbiology 73: 2180-2190.
Garber, L.P., Smeltzer, M., Fedorka-Cray, P., Ladely, S., and Ferris, K. (2003) Salmonella
enterica serotype enteritidis in table egg layer house environments and in mice in U.S.
layer houses and associated risk factors. Avian Diseases 47: 134-142.
Garcia, A.M., Veith, T.L., Kleinman, P.J.A., Rotz, C.A., and Saporito, L.S. (2008) Assessing
manure management strategies through small-plot research and wholefarm modeling.
Journal of Soil and Water Conservation 63: 204-211.
Gebhart, C.J., Edmonds, P., Ward, G.E., Kurtz, H.J., and Brenner, D.J. (1985) "Campylobacter
hyointestinalis" sp. nov.: a new species of Campylobacter found in the intestines of pigs
and other animals. Journal of Clinical Microbiology 21(5): 715-720.
Gebreyes, W.A., Bahnson, P.B., Funk, J.A., McKean, J., and Patchanee, P. (2008)
Seroprevalence of Trichinella, Toxoplasma, and Salmonella in antimicrobial-free and
conventional swine production systems. Foodborne Pathogenic Diseases 5: 199-203.
Goss, M., and Richards, C. (2008) Development of a risk-based index for source water
protection planning, which supports the reduction of pathogens from agricultural activity
entering water resources. Journal of Environmental Management 87: 623-632.
Gregory, E., Barnhart, H., Dreesen, D., Stern, N., and Corn, J. (1997) Epidemiological study of
Campylobacter spp. in broilers: source, time of colonization, and prevalence. Avian
Diseases 41: 890-898.
Hakkinen, M., and Hanninen, M.-L. (2009) Shedding of Campylobacter spp. in Finnish cattle on
dairy farms. Journal of Applied Microbiology Early view (accepted for publication).
Haley, B.J., Cole, D.J., and Lipp, E.K. (2009) Distribution, diversity, and seasonality of
waterborne Salmonellae in a rural watershed. Applied and Environmental Microbiology
75: 1248-1255.
Harvey, R., Droleskey, R., Sheffield, C., Edrington, T., Callaway, T., Anderson, R. et al. (2004)
Campylobacter prevalence in lactating dairy cows in the United States. Journal of Food
Protection 67: 1476-1479.
August 2010 55
-------
U.S. Environmental Protection Agency
Hayes, J.R., Carr, L.E., Mallinson, E.T., Douglass, L.W., and Joseph, S.W. (2000)
Characterization of the contribution of water activity and moisture content to the
population distribution of Salmonella spp. in commercial poultry houses. Poultry Science
79: 1557-1561.
Heinonen-Tanski, H., Mohaibes, M., Karinen, P., and Koivunen, J. (2006) Methods to reduce
pathogen microorganisms in manure. Livestock Science 102:248-255.
Heitman, T.L., Frederick, L.M., Viste, J.R., Guselle, N.J., Morgan, U.M., Thompson, R.C.A.,
and Olson, M.E. (2002) Prevalence ofGiardia and Cryptosporidium and characterization
of Cryptosporidium spp. isolated from wildlife, human, and agricultural sources in the
North Saskatchewan River Basin in Alberta, Canada. Canadian Journal of Microbiology
48: 530-541.
Heuer, O., Pedersen, K., Andersen, J., and Madsen, M. (2001) Prevalence and antimicrobial
susceptibility of thermophilic Campylobacter in organic and conventional broiler flocks.
Letters in Applied Microbiology 33: 269-274.
Hoar, B., Atwill, E., Elmi, C., and Farver, T. (2001) An examination of risk factors associated
with beef cattle shedding pathogens of potential zoonotic concern. Epidemiology and
Infection 127: 147-155.
Hunter, P.R., and Nichols, G. (2002) Epidemiology and clinical features of Cryptosporidium
infection in immunocompromised patients. Clinical Microbiology Reviews 15(1): 145-
154.
Humphrey, T.J., and Becket, P. (1987) Campylobacter jejuni in dairy cows and raw milk.
Epidemiology and Infection 98: 263-269.
Kurd, H.S., McKean, J.D., Griffith, R.W., Wesley, I.V., and Rostagno, M.H. (2002) Salmonella
enterica infections in market swine with and without transport and holding. Applied and
Environmental Microbiology 68: 2376-2381.
Huston, C.L., Wittum, I.E., Love, B.C., and Keen, J.E. (2002) Prevalence of fecal shedding of
Salmonella spp. in dairy herds. Journal of the American Veterinary Medicine Association
220: 645-649.
Hutchinson, M.L., Walters, L.D., Avery, S.M., Synge, B.A., and Moore, A. (2004) Levels of
zoonotic agents in British livestock manures. Letters in Applied Microbiology 39: 207-
214.
Hutchison, M.L., Walters, L.D., Avery, S.M., Munro, F., and Moore, A. (2005) Analyses of
livestock production, waste storage, and pathogen levels and prevalences in farm
manures. Applied and Environmental Microbiology 71: 1231-1236.
Inglis, G., Kalischuk, L., and Busz, H. (2004) Chronic shedding of Campylobacter species in
beef cattle. Journal of Applied Microbiology 97: 410-420.
Johnson, J.Y.M., Thomas, J.E., Graham, T.A., Townshend, I, Byrne, J., Selinger, L.B., and
Gannon, V.P.J. (2003) Prevalence of Escherichia coli O157: H7 and Salmonella spp. in
surface waters of southern Alberta and its relation to manure sources. Canadian Journal
of Microbiology 49: 326-335.
Kabagambe, E.K., Wells, S.J., Garber, L.P., Salman, M.D., Wagner, B., and Fedorka-Cray, P.
(2000) Risk factors for fecal shedding of Salmonella in 91 U.S. dairy herds in 1996.
Preventive Veterinary Medicine 43: 177-194.
Ketley, J.M. (1997) Pathogenesis of enteric infection by Campylobacter. Microbiology 143: 5-
21.
August 2010 56
-------
U.S. Environmental Protection Agency
Kraft, D.J., LOlechowski-Gerhardt, C., Berkowitz, J., and Finstein, M.S. (1969) Salmonella in
wastes produced at commercial poultry farms. Applied Microbiology 18: 703-707.
Krueger, N., Anderson, R., Krueger, W., Home, W., Wesley, I, Callaway, T. et al. (2008)
Prevalence and concentration of Campylobacter in rumen contents and feces in pasture
and feedlot-fed cattle. Foodborne Pathogenic Diseases 5: 571-577.
Kudva, IT., Blanch, K., and Hovde, CJ. (1998) Analysis of Escherichia coli O157:H7 survival
in ovine or bovine manure and manure slurry. Applied and Environmental Microbiology
64: 3166-3174.
Kuhnert, P., Dubosson, C.R., Roesch, M., Homfeld, E., Doherr, M.G., and Blum, J.W. (2005)
Prevalence and risk-factor analysis of Shiga toxigenic Escherichia coli in faecal samples
of organically and conventionally farmed dairy cattle. Veterinary Microbiology 109: 37-
45.
Kunze, D.J., Loneragan, G.H., Platt, T.M., Miller, M.F., Besser, T.E., Koohmaraie, M. et al.
(2008) Salmonella enterica burden in harvest-ready cattle populations from the Southern
High Plains of the United States. Applied and Environmental Microbiology 74: 345-351.
Kwan, P.S.L., Birtles, A., Bolton, F.J., French, N.P., Robinson, S.E., Newbold, L.S. et al. (2008)
Longitudinal study of the molecular epidemiology of Campylobacter jejuni infection in
cattle on dairy farms. Applied and Environmental Microbiology 74: 3626-3633.
Larney, F.J., and Hao, X. (2007) A review of composting as a management alternative for beef
cattle feedlot manure in Southern Alberta, Canada. Bioresource Technology 98:3221-
3227.
LeJeune, J.T., Besser, I.E., Rice, D.H., Berg, J.L., Stilborn, R.P., and Hancock, D.D. (2004)
Longitudinal study of fecal shedding of Escherichia coli O157:H7 in feedlot cattle:
Predominance and persistence of specific clonal types despite massive cattle population
turnover. Applied and Environmental Microbiology 70: 377-384.
Lenahan, M., O'Brien, S.B., Byrne, C., Ryan, M., Kennedy, C.-A., McNamara, E.B. et al. (2009)
Molecular characterization of Irish E. coli O157:H7 isolates of human, bovine, and
porcine origin. Journal of Applied Microbiology 107(4): 1340-1349.
Lewis, D.J., Atwill, E.R., Lennox, M.S., Hou, L., Karle, B., and Tate, K.W. (2005) Linking on-
farm dairy management practices to storm-flow fecal coliform loading for California
coastal watersheds. Environmental Monitoring and Assessment 107: 407-425.
Ley, D.H., Levy, M.G., Hunter, L., Corbett, W., and Barnes, HJ. (1988) Cryptosporidia-positive
rates of avian necropsy accessions determined by examination of auramine O-stained
fecal smears. Avian Diseases 32: 108-113.
Luangtongkum, T., Morishita, T.Y., Ison, A.J., Huang, S., McDermott, P.F., and Zhang, Q.
(2006) Effect of conventional and organic production practices on the prevalence and
antimicrobial resistance of Campylobacter spp. in poultry. Applied and Environmental
Microbiology 72: 3600-3607.
Maddox-Hyttel, C., Langkjaer, R.B., Enemark, H.L., and Hakan, V. (2006) Cryptosporidium and
Giardia in different age groups of Danish cattle and pigs—Occurrence and management
associated risk factors. Veterinary Parasitology 141: 48-59.
Martin, S.A., McCann, M.A., and Waltman, W.D. (1998) Microbiological survey of Georgia
poultry litter. Journal of Applied Poultry Research 7: 90-98.
McLaughlin, M.R., Brooks, J.P., and Adeli, A. (2009) Characterization of selected nutrients and
bacteria from anaerobic swine manure lagoons on sow, nursery, and finisher farms in the
Mid-South USA. Journal of Environmental Quality 38: 2422-2430.
August 2010 57
-------
U.S. Environmental Protection Agency
Meals, D.W., and Braun, D.C. (2006) Demonstration of Methods to Reduce E. coli runoff from
dairy manure application sies. Journal of Environmental Quality 35:1088-1100.
Midwest Plan Service (1993). Livestock Waste Facilities Handbook, 3rd edition. MWPS-18,
Midwest Plan Service, Ames, IA.
Miller, W.A., Lewis, D.J., Lennox, M., Pereira, M.G.C., Tate, K.W., Conrad, P.A., and Atwill,
E.R. (2007) Climate and on-farm risk factors associated with Giardia duodenalis in storm
runoff from California coastal dairies. Applied and Environmental Microbiology
73(21):6972-6979.
Miller, W.A., Lewis, D.J., Pereira, M.D.G.C., Lennox, M.S., Conrad, P.A., Tate, K.W., and
Atwill, E.R. (2008) Farm factors associated with the reducing Cryptosporidium loading
in storm runoff from dairies. Journal of Environmental Quality 37: 1875-1882.
Minihan, D., Whyte, P., O'Mahony, M., Fanning, S., McGill, K., and Collins, J.D. (2004)
Campylobacter spp. in Irish feedlot cattle: A longitudinal study involving pre-harvest and
harvest phases of the food chain. Journal of Veterinary Medicine, Series B 51: 28-33.
Moe, C. (2004) Chapter 3. What are the criteria for determining whether a disease is zoonotic
and water related? In Waterborne Zoonoses: Identification, Causes and Control. Cotruvo,
J.A., Dufour, A., Rees, G., Bartram, J., Carr, R., Cliver, D.O. et al. (eds). London, U.K.:
IWA Publishing. World Health Organization (WHO).
Moore, P.A., Jr., Daniel, T.C., Sharpley, A.N., and Wood, C.W. (1995) Poultry manure
management: Environmentally sound options. Journal of Soil and Water Conservation
50: 321-327.
Moriarty, E.M., Sinton, L.W., Mackenzie, M.L., Kearki, N., and Wood, D.R. (2008) A survey of
enteric bacteria and protozoans in fresh bovine faeces on New Zealand dairy farms.
Journal of Applied Microbiology 105: 2015-2025.
NASS (National Agricultural Statistics Sevice). (2010). The Census of Agriculture. URL
http://www.agcensus.usda.gov/index.asp.
NRC (National Research Council) (2004). Indicators for Waterborne Pathogens. The National
Academies Press. Washington, DC.
Newell, D.G., and Fearnley, C. (2003) Sources of Campylobacter colonization in broiler
chickens. Applied and Environmental Microbiology 69: 4343-4351.
Olson, M.E., Thorlakson, C.L., Deselliers, L., Morck, D.W., and McAllister, T.A. (1997)
Giardia and Cryptosporium in Canadian farm animals. Veterinary Parasitology 68: 375-
381.
Ralston, B.J., McAllister, T.A., and Olson, M.E. (2003) Prevalence of infection pattern of
naturally acquired giardiasis and cryptosporidiosis in range beef calves and their dams.
Veterinary Parasitology 114: 113-122.
Reinstein, S., Fox, J.T., Shi, X., Alam, M.J., Renter, D.G., andNagaraja, T.G. (2009) Prevalence
of Escherichia coli O157:H7 in organically and naturally raised beef cattle. Applied and
Environmental Microbiology 75: 5421-5423.
Rice, D.H., Hancock, D.D., and Besser, T.E. (1995) Verotoxigenic E. coli O157 colonisation of
wild deer and range cattle. The Veterinary Record 137: 524.
Riley, D., Loneragan, G., Phillips, W., Gray, J., and Fedorka-Cray, P. (2008) Fecal shedding of
foodborne pathogens by Florida-born heifers and steers in U.S. beef production segments.
Journal of Food Protection 71: 807-810.
Sato, K., Bartlett, P.C., Kaneene, J.B., and Downes, F.P. (2004) Comparison of prevalence and
antimicrobial susceptibilities of Campylobacter spp. isolates from organic and
August 2010 58
-------
U.S. Environmental Protection Agency
conventional dairy herds in Wisconsin. Applied and Environmental Microbiology 70:
1442-1447.
Schoeni, J.L., and Doyle, M.P. (1994) Variable colonization of chickens perorally inoculated
with Escherichia coli 0157:H7 and subsequent contamination of eggs. Applied and
Environmental Microbiology 60(8): 2958-2962.
SDES (South Dakota Extension Service). (2007) SDES Extension Extra, available online at
http://agbiopubs. sdstate.edu/articles/ExEx 11020.pdf
Shepherd, M.W. Jr., Liang, P., Jiang, X., Doyle, M.P., and Erickson, M.C. (2007) Fate of
Escherichia coli O157:H7 during on-farm dairy manure-based composting. Journal of
Food Protection 70(12):2708-2716.
Sischo, W.M., Atwill, E.R., Lanyon, L.E., and George, J. (2000) Cryptosporidia on dairy farms
and the role these farms may have in contminating surface water supplies in the
northeastern United States. Preventive Veterinary Medicine 43: 253-267.
Sobsey, M.D., Khatib, L.A., Hill, V.R., Alocilja, E., and Filial, S. (2006) Pathogens in animal
wastes and the impacts of waste management practices on their survival, transport and
fate. In Animal Agriculture and the Environment: National Center for Manure and
Animal Waste Management White Papers. Rice, J.M., Caldwell, D.F., and Humenik, F.J.
(eds). St. Joseph, MI: ASABE.
Stanley, K., and Jones, K. (2003) Cattle and sheep farms as reservoirs for Campylobacter.
Journal of Applied Microbiology 94: 104S-113S.
Stanley, K.N., Wallace, J.S., Currie, J.E., Diggle, P.J., and Jones, K. (1998) The seasonal
variation of thermophilic campylobacters in beef cattle, dairy cattle and calves. Journal of
Applied Microbiology 85: 472-480.
Stern, N.J., Fedorka-Cray, P., Bailey, J.S., Cox, N.A., Craven, S.E., Hiett, K.L. et al. (2001)
Distribution of Campylobacter spp. in selected U.S. poultry production and processing
operations Journal of Food Protection 64: 1705-1710.
Sturdee, A.P., Bodley-Tickell, A.T., Archer, A., and Chalmers, R.M. (2003) Long-term study of
Cryptosporidium prevalence on a lowland farm in the United Kingdom. Veterinary
Parasitology 116: 97-113.
Thompson, R.C.A. (2004) The zoonotic significance and molecular epidemiology ofGiardia and
giardiasis. Veterinary Parasitology 126: 15-35.
Trotz-Williams, L.A., Martin, W.S., Leslie, K.E., Duffiled, T., Nydam, D.V., and Peregrine, A.S.
(2007) Calf-level risk factors for neonatal diarrhea and shedding of Cryptosporidium
parvum in Ontario dairy calves. Preventive Veterinary Medicine 82: 12-28.
Trotz-Williams, L.A., Martin, S.W., Leslie, K.E., Duffield, T., Nydam, D.V., and Peregrine, A.S.
(2008) Association between management practices in within-herd prevalence of
Cryptosporidium parvum shedding on dairy farms in southern Ontario. Preventive
Veterinary Medicine 83:11 -23.
USDA (U.S. Department of Agriculture). (1993) Centers for Epidemiology and Animal Health.
Cryptosporidium is common in dairy calves. Info sheet #N119.293. Fort Collins, CO:
USDA: APHIS.
USDA. (1994) Centers for Epidemiology and Animal Health. Cryptosporidium and Giardia in
beef calves. Info sheet #N137.0194. Fort Collins, CO: USDA: APHIS.
USDA. (1995) Centers for Epidemiology and Animal Health. Escherichia coli O157:H7
shedding in feedlot cattle. Info sheet May 1995. Fort Collins, CO: USDA: APHIS.
August 2010 59
-------
U.S. Environmental Protection Agency
USD A. (1997) Centers for Epidemiology and Animal Health. Factors associated with
Escherichia coll 0157 in feces of feedlot cattle. Info sheet #N242.1097. Fort Collins,
CO: USD A: APHIS.
USDA. (1998) Centers for Epidemiology and Animal Health. E. coli O157 and Salmonella -
status on U.S. dairy operations. Info sheet N286.598. Fort Collins, CO: USDA: APHIS.
USDA. (2000) Salmonella enterica serotype Enteritidis in table egg layers in the U.S. Fort
Collins: USDA:APHIS:VS, CEAH, National Animal Health Monitoring System,
#N333.1000.
USDA. (2001) Centers for Epidemiology and Animal Health. Escherichia coli O157 in United
States feedlots. Info sheet #N345.1001. Fort Collins, CO: USDA: APHIS.
USDA. (2003) Centers for Epidemiology and Animal Health. Escherichia coli O157 on U.S.
dairy operations. Info sheet #N345.1001. Fort Collins, CO: USDA: APHIS.
USDA. (2009a) Salmonella on U.S. swine sites - Prevalence and antimicrobial susceptibility.
Fort Collins, CO: USDA: APHIS.
USDA. (2009b) Salmonella and Campylobacter on U.S. dairy operations, 1996 - 2007. Fort
Collins, CO: USDA: APHIS.
USDA. (2009c) Salmonella on U.S. beef cow-calf operations, 2007-08. Fort Collins, CO:
USDA: APHIS.
USEPA (U.S. Environmental Protection Agency). (1986) Ambient Water Quality Criteria for
Bacteria. Washington DC: Office of Emergency and Remedial Response, EPA 440/5-84-
002.
USEPA. (2005) Detecting and Mitigating the Environmental Impact of Fecal Pathogens
Originating from Confined Animal Feeding Operations: Review. National Risk
Management Research Laboratory, Office of Research and Development, Cincinnati, OH
EPA/600/R-06/021.
USEPA. (2009) Review of Zoonotic Pathogens in Ambient Waters. Washington DC: USEPA,
EPA 822-R-09-002.
Van Horn, H.H., Wilkie, A.C., Powers, W.J., and Nordstedt, R.A. (1994) Components of dairy
manure management systems. Journal of Dairy Science 77: 2008-2030.
Van Overbeke, I, Duchateau, L., De Zutter, L., Albers, G., and Ducatelle, R. (2006) A
comparison survey of organic and conventional broiler chickens for infectious agents
affecting health and food safety. Avian Diseases 50: 196-200.
Venglovsky, J., Sasakova, N., and Placha, I. (2009) Pathogens and antibiotic residues in animal
manures and hygienic and ecological risks related to subsequent land application.
Bioresource Technology 100: 5386-5391.
Wade, S.E., Mohammed, H.O., and Schaaf, S.L. (2000) Prevalence ofGiardia spp.,
Cryptosporidium parvum, and Cryptosporidium muris (C. andersonf) in 109 dairy herds
in five counties of southeastern New York. Veterinary Parasitology 93: 1-11.
Warnick, L.D., Kaneene, J.B., Ruegg, P.L., Wells, S.J., Fossler, C.P., Halbert, L.W., and
Campbell, A.M. (2003) Evaluation of herd sampling for Salmonella isolation on midwest
and northeast U.S. dairy farms. Preventive Veterinary Medicine 60: 195-206.
Weijtens, M.J.B.M., Reinders, R.D., Urlings, H.A.P., and Van der Plas, J. (1999) Campylobacter
infections in fattening pigs; excretion pattern and genetic diversity. Journal of Applied
Microbiology 86: 63-70.
August 2010 60
-------
U.S. Environmental Protection Agency
Weijtens, M.J.B.M., van der Plas, I, Bijker, P.G.H., Urlings, H.A.P., Koster, D., van Logtestijn,
J.G., and Huis in't Veld, J.H.J. (1997) The transmission of campylobacter in piggeries; an
epidemiological study. Journal of Applied Microbiology 83: 693-698.
Wells, S.J., Fedorka-Cray, P., Dargatz, D.A., Ferris, K., and Green, A. (2001) Fecal shedding of
Salmonella spp. by dairy cows on farm and at cull cow markets. Journal of Food
Protection 64: 3-11.
Wesley, I.V., Wells, S.J., Harmon, K.M., Green, A., Schroeder-Tucker, L., Glover, M., and
Siddique, I. (2000) Fecal shedding of Campylobacter and Arcobacter spp. in Dairy
Cattle. Applied and Environmental Microbiology 66: 1994-2000.
Wetzel, A.N., and LeJeune, J.T. (2007) Isolation of Escherichia coll O157:H7 strains that do not
produce Shiga toxin from bovine, avian and environmental sources. Letters in Applied
Microbiology 45: 504-507.
Whyte, P., Collins, J.D., McGill, K., Monahan, C., and O'Mahony, H. (2001) The effect of
transportation stress on excretion rates of campylobacters in market-age broilers. Poultry
Science 80: 817-820.
Wilhelm, B., Rajic, A., Waddell, L., Parker, S., Harris, J., Roberts, K. et al. (2009) Prevalence of
zoonotic or potentially zoonotic bacteria, antimicrobial resistance, and somatic cell
counts in organic dairy production: current knowledge and research gaps. Foodborne
Pathogenic Diseases 6: 525-539.
WHO (World Health Organization). (2004) Waterborne Zoonoses: Identification, Causes and
Control. London, U.K.: IWA Publishing.
Wright, S., Carver, D., Siletzky, R., Romine, S., Morrow, W., and Kathariou, S. (2008)
Longitudinal study of prevalence of Campylobacter jejuni and Campylobacter coll from
turkeys and swine grown in close proximity. Journal of Food Protection 71: 1791-1796.
Xiao, L., Herd, R.P., and Bowman, G.L. (1994) Prevalence of Cryptosporidium and Giardia
infection on two Ohio pig farms with different management systems. Veterinary
Parasitology 52: 331-336.
Xiao, L., Payer, R., Ryan, U., and Upston, S. (2004) Cryptosporidium taxonomy: recent
advances and implcations for public health. Clinical Microbiology Reviews 17: 12-91.
Xiao, L., Moore, J.E., Ukoh, U., Gatei, W., Lowery, C.J., Murphy, T.M. et al. (2006) Prevalence
and identity of Cryptosporidium spp. in pig slurry. Applied and Environmental
Microbiology 72: 4461-4463.
Young, I, Rajic, A., Wilhelm, B., Waddell, L., Parker, S., and McEwen, S. (2009) Comparison
of the prevalence of bacterial enteropathogens, potentially zoonotic bacteria and bacterial
resistance to antimicrobials in organic and conventional poultry, swine and beef
production: a systematic review and meta-analysis. Epidemiology and Infection 137:
1217-1232.
August 2010 61
-------
U.S. Environmental Protection Agency
Appendix A. Literature Search Strategy and Results
Literature Search Overview
Literature searches were conducted to obtain data on livestock and pathogen geographic
distribution, data on the range of pathogen prevalences and abundances in livestock manures,
data on the relationship between farm factors and pathogen prevalence and abundance, and
manure management impacts on pathogen loads to streams. All searches were conducted using
the following databases:
> Highwire (full text searchable)
> Pubmed (full text searchable)
> Web of Science (keywords and titles searchable)
> American Society of Agricultural and Biological Engineers archives (keywords, titles and
abstracts searchable) and
> USDA APHIS online resources
Additional papers were acquired based on reviews of references cited in articles obtained from
database searches. Approximately one-third of papers used in the study were identified in
references obtained from studies from the database searches.
Geospatial data were searched both in databases and from online sources. Initially, state agency
(environmental protection agencies, departments of natural resources or state geospatial data
clearinghouses) databases were searched individually, but this strategy was abandoned in favor
of seeking a single source for consistent livestock data for the United States. The sole source
providing a consistent data set was the USDA agriculture survey.
Search Terms Used in Database Searches
A relatively simple search strategy was used to obtain studies related to occurrence and
abundance of pathogens in manures. Searches were conducted with manure as the primary
search term, cattle, swine or (poultry or chicken*) as a secondary search term and each of the key
pathogens or "pathogen*" as a tertiary search term. These searches typically returned fewer than
500 studies whose titles and abstracts were scanned to determine the relevant studies. Highwire
and Pubmed tended to return more studies since full text searches were used in those databases.
Studies related to microbial source tracking, fate of pathogens, and microbial methods were not
considered relevant. Abstracts of relevant studies were reviewed and highly relevant studies
were selected from the lists for inclusion in the report. In many cases studies were not selected
for inclusion in the report because they described farms not in the United States (though select
non-US studies were used to fill data gaps), because the emphasis was molecular biology, or
because the emphasis was human health epidemiology.
Specific searches were conducted for several types of studies, including the following:
> Studies documenting the impact of herd or flock size on pathogen occurrence,
> Studies comparing pathogen occurrence in conventional and organic operations; and
August 2010 A-l
-------
U.S. Environmental Protection Agency
> factors relating manure management to pathogen runoff.
Search terms used for herd and flock size studies were livestock type (primary search term)
"herd/flock size" OR "farm size" (secondary search term) and "pathogen*" (tertiary search
term). Searches for data on conventional and organic operations used the primary search term
"organic" the secondary search term "livestock type" and the tertiary search term "manure." The
search terms used to identify manure management factors were primary search term "manure
management" and secondary search term either pathogen* or [key pathogen name].
Summary of Literature Survey Results
The literature survey resulted in collection of 176 highly-relevant studies, of which over 120 are
cited in this document.
August 2010 A-2
-------
U.S. Environmental Protection Agency
Appendix B. Occurrence Data
Table B-1. Summary of studies reporting prevalence of Cryptosporidium in cattle, swine, poultry, other domestic animals, wildlife, and
environmental samples
Study
Moriarty et
al. (2008)
Atwill et al.
(2006)
Xiao et al.
(2006)
Hutchinson
et al. (2005)
Hutchinson
et al. (2005)
Atwill et al.
(2003)
Animal/Source
Dairy cattle
Beef cattle
Swine
Cattle
Swine
Beef cows, pre-
parturition and
post-parturition
Media
Freshly-
collected
manure
Feedlot
manure
samples
Slurry
Manure
collected as
composite
samples from
fresh and
stored stocks
Manure
collected as
composite
samples from
fresh and
stored stocks
Manure
Species
spp.
C. parvum
C. suis, pig
genotype II,
C. muris
C. parvum
C. parvum
C. parvum
Prevalence
Overall prevalence among 4 farms 5.2% (n=155);
oocysts detected at 2 farms
Overall prevalence of C. parvum in samples 0.2%
(n=5274); highest point prevalence (one farm, one
sampling event) of 1.7% (n=239)
Oocysts detected in 45% (n=56) of pig slurries
collected from 33 farms in Ireland. C. suis, pig
genotype II and C. muris were detected in 62%,
42%, and 2% of positive samples, respectively.
Prevalence among samples of fresh manure
5.4%; prevalence among samples of stored
manure 2.8%
Prevalence among samples of fresh manure
13.5%; prevalence among samples of stored
manure 5.2%
Overall prevalence for three herds 7.1%.
Prevalence by herd ranged from 6.25% to 8.75%.
Prevalence among pre-parturient cows and post-
parturient cows 8.3% and 5.8%, respectively.
Observations and Notes
An alternate estimate of 0.99-1.08%
point prevalence was developed
based on statistical analyses and
consideration of false negative rate.
Cryptosporidium spp. and individual
species prevalences varied with type
of slurry sample (liquid vs. solid).
Cryptosporidium prevalence was
significantly higher in June and
December than in other months,
though overall seasonal trends not
observed.
Data do not indicate a significant
difference in shedding during the pre-
and post-parturition periods. This
finding indicates the potential for inter-
herd transmission, particularly to
calves. Calves potentially shed very
high numbers of Cryptosporidium
oocysts.
August 2010
B-1
-------
U.S. Environmental Protection Agency
Study
Sturdee et
al. (2003)
Heitman,
(2002)
Heitman et
al. (2002)
Payer et al.
(2000)
Sischo et al.
2000
Wade et al.
(2000)
Animal/Source
Cattle
Dairy cattle, both
young and adult
Hogs
Dairy cattle, post-
weaned and adult
Dairy cattle and
calves
Dairy cattle, all
ages
Media
Manure
collected as
rectal
samples or
recently
deposited
feces
Manure (from
pasture)
Manure
(collected
from a single
hog operation
from pasture)
Feces (rectal
samples)
Fecal manure
(rectal
sampling)
Water
samples
Manure
Species
C. parvum
spp.
C. parvum,
C. muris
C. parvum,
C. andersoni
C. parvum
C. parvum,
C. muris
Prevalence
Description
Bull beef
Dairy cow
Home-bred
calf
Bought-in calf
Prevalence
over 6 years
3.6%
3.5%
52%
23.2%
Highest annual
prevalence
8.7%
8.8%
66.7%
48%
19.6% on two farms (n=92); prevalences on the
two farms were 8% and 28%, respectively
0%(n=40)
On a single farm, C. andersoni detected in 12.5%
of fecal samples (n=24)
On a second farm, C. parvum detected in 9.5% of
samples (n=42). Prevalence among cows was
10.5% and prevalence among heifers was 9.0%.
91% of the dairy farms; 15% of calves 0-3 weeks
of age; 90% of stream samples
Prevalence (%)
Parasite
C. pavum
C. muri
-------
U.S. Environmental Protection Agency
Study
Animal/Source
Media
Species
Prevalence
Observations and Notes
Atwill et al.
(1997)
Feral (wild) pigs
Fecal
samples
C. parvum
5.4% (12/221)
No association found between oocyst
shedding and sex of pig, body
condition, and presence of cattle in the
area. However, there was an
association between oocyst shedding
and pig age and density population:
younger pigs (<8 months) and high-
density area (>2 feral pigs/km2)
significantly more likely to shed
oocysts.
USDA
(1994)
Beef calves from
beef cow/calf
operations
Fresh fecal
samples
included both
diarrheic
calves <3
months and
nondiarrheic
calves <6
months
spp.
Prevalence of positive calves was 20.1% for
diarrheic calves and 11.2% for nondiarrheic
calves.
Prevalence of positive operations submitting
samples from diarrheic calves was 39.1%.
Prevalence of positive operations submitting
samples from nondiarrheic calves was 41.8%.
Prevalence was related to and decreased with
age of calves (23.1% for 1-30 days old; 9.2% for
>121 days old).
Study included 391 samples from
diarrheic calves from 69 operations
and 1,053 samples from nondiarrheic
calves from 141 operations.
Average age of diarrheic calves testing
positive was 41.1 days. Average age
of nondiarrheic calves testing positive
was 75.8 days.
Shedding was common in calves of
beef herds whether the calves had
diarrhea or not.
USDA
(1993)
Dairy calves
(preweaned)
Fresh fecal
samples
spp.
Overall prevalence across U.S. was 22% of
calves and >90% of farms.
Prevalence increased slightly with herd size but
still high prevalence on (about 80%) on smaller
farms (<100 cows).
Prevalence higher in western herds; these are
also the largest operations.
Prevalence higher in summer months than in
other months.
Prevalence was highest in heifers 1-3 weeks old
(>50%). Prevalence drops to <15% for calves
over 5 weeks old.
Study included 1,103 farms in 28
states, with 7,369 samples collected.
States included:
• West: WA, OR, CA, ID, CO
• Midwest: NE, IA, MN, Wl, Ml, IL,
IN, OH
• Northeast: ME, VT, NH, NY, PA,
CT, MA, Rl
• Southeast: VA, NC, TN, GA, AL,
FL, MD
August 2010
B-3
-------
U.S. Environmental Protection Agency
Table B-2. Summary of studies reporting prevalence of Giardia in cattle, swine, chicken, other domestic animals, wildlife, and
environmental samples
Study
Hutchison et
al. (2004)
Hutchison et
al. (2004)
Ralston et
al. (2003)
Heitman et
al. (2002)
Heitman et
al. (2002)
Payer et al.
(2000)
Payer et al.
(2000)
Wade et al.
(2000)
Olson et al.
(1997)
Animal/Source
Cattle (assumed to
include samples
from both dairy
and beef cattle
operations)
Swine
Range beef calves
and their dams
Dairy cattle
Beef cattle
Beef cattle 7-9
months old
Dairy cattle
Dairy cattle
Beef cattle
Media
Feces (farm
yard
manures) and
slurries
Feces (farm
yard
manures) and
slurries
Feces
Feces
Feces
Feces
Feces
Feces
Feces
Species
intestinalis
intestinalis
spp.
spp.
spp.
duodenalis
duodenalis
spp.
spp.
Prevalence
3.6%
2.4%
100% of calves shed Giardia cysts at some point
during the study
Cow shedding prevalences 10% prior to calving,
up to 15% one week post calving, and to 0%
within 23 weeks after calving
10-20%
15%
37.3%
0-17.3%
Giardia prevalence varied among animals with
different age groups.
Age group
All ages
< 6 months
6-24 months
> 24 months
Prevalence
8.9%
20.1 %
3.5%
0.2%
Giardia prevalence varied with age group.
Among animals < 6 months prevalence was 30%
and among animals > 6 months of age the
prevalence was 1 1 %
Observations and Notes
Samples taken from operations
throughout U.K.
Samples taken from operations
throughout U.K.
Study of Canadian dairy farms
Study of Canadian beef cattle farms
Study conducted in three Maryland
cattle farms (one beef and two dairy
operations).
Study conducted in three Maryland
cattle farms (one beef and two dairy
operations); dairy cattle include
replacement heifers and milk cows
No seasonal variation in Giardia
prevalence was observed.
August 2010
B-4
-------
U.S. Environmental Protection Agency
Study
USDA
(1994)
Xiao et al.
(1994)
Animal/Source
Beef calves from
beef cow/calf
operations
Swine
Media
Fresh fecal
samples
included both
diarrheic
calves <3
months and
nondiarrheic
calves <6
months
Feces (via
rectal swab
samples)
Species
spp.
spp.
Prevalence
Prevalence of positive calves was 26.9% for
diarrheic calves and 45.9% for nondiarrheic
calves.
Prevalence of positive operations submitting
samples from diarrheic calves was 63.8%.
Prevalence of positive operations submitting
samples from nondiarrheic calves was 90.8%.
Prevalence peaked in calves 61-90 days old
(59.6%) and decreased with age of calves (29.9%
for calves >121 days old).
On two farms, 0-17% of litters of pigs positive for
Giardia and 3-25% of weanlings positive for
Giardia
Observations and Notes
Study included 391 samples from
diarrheic calves from 69 operations
and 1,053 samples from nondiarrheic
calves from 141 operations.
Average age of diarrheic calves testing
positive was 47.1 days. Average age
of nondiarrheic calves testing positive
was 79.1 days.
Shedding was common in calves,
especially older calves, of beef herds
whether the calves had diarrhea or
not.
Two Ohio farms with different animal
housing types studied.
August 2010
B-5
-------
U.S. Environmental Protection Agency
Table B-3. Summary of studies reporting prevalence of Campylobacter in cattle, swine, chicken, other domestic animals, and
environmental samples
Study
Hakkinen and
Hanninen
(2009)
Mclaughlin et
al. (2009)
Kwan et al.
(2008)
Moriarty et al.
(2008)
Berry et al.
(2007)
Animal/Source
Dairy cattle
Swine (sows,
nursery swine, and
finishers)
Dairy cattle
Dairy cattle
Beef cattle (feedlot)
Species
C. jejuni
spp.
C. jejuni
C. jejuni,
C. co//
spp.
Media
Fecal samples,
water samples
from troughs, and
milk samples
Lagoon waters
Manure
Freshly-excreted
feces
Manure
(composite
samples)
Prevalence
C. jejuni detected in 169 out of 340
fecal samples, 1 out of 3 farm
drinking water supplies (one
sampling event), and in no milk
samples. C. co// detected in 3.2%
of fecal samples and C.
hyointestinalis detected in 15.3% of
fecal samples.
1 00% of sow slurry in lagoons (n =
7 farms, 102 samples); 100% of
nursery slurries (n = 10 lagoons
and 60 samples); and 100% of
finisher slurry in lagoons (n = 10
lagoons and 60 samples) positive
for Campylobacter
Overall prevalence 35.9% (5
farms); ranged from 26-50.8% for
low and high months; C. jejuni
prevalence among isolates 68%
96% of Campylobacter spp.
positive samples positive for C.
jejuni.
1 0% of Campylobacter spp.
positive samples positive for C.
CO//.
7% of Campylobacter spp. positive
samples positive for C. jejuni and
C. co//
Ranged from 2.2 % in samples
taken in the spring to 14.9% of
samples in the summer.
Observations and Notes
Finland; infection prevalence differed
significantly between farms and, for
farms with relatively low prevalence, by
season. For two farms, the lowest C.
jejuni prevalence (% of herd infected)
coincided with indoor housing of cattle;
higher C. jejuni prevalence coincided
with grazing periods.
Farms located in mid-south United
States
Seasonal variation in Campylobacter
spp. prevalence differed among the five
U.K. dairy farms studied, though overall
prevalence (annual average) among
farms was similar; genotype diversity for
C. jejuni isolates differed significantly
between farms
New Zealand farms, 4 regions
Samples collected every four weeks for
26 months; manure samples were
composite samples collected from
feed lots
August 2010
B-6
-------
U.S. Environmental Protection Agency
Study
Englen et al.
(2007)
Bull etal. (2006)
Luangtongkum
et al. (2006)
Bae etal.
(2005)
Besser et al.
(2005)
Devane et al.
(2005)
Animal/Source
Dairy cattle from 96
operations in 21
states
Broiler chickens at
U.K. farms
Broiler chickens from
conventional and
organic operations
Dairy cattle, beef
cattle, feedlot cattle,
cattle in calf-cow
operations
Beef cattle (feedlot)
Dairy cattle
Species
C. jejuni,
C. co//
C. jejuni,
C. co//
spp.
C. jejuni,
C. co//,
C. fetus,
C.
hyointestinalis
C. jejuni
spp,
C. co//,
C. jejuni
Media
Feces samples
Droppings (feces)
taken from floors
Intestinal tracts of
birds a slaughter
Fecal samples
(rectal or recently
deposited)
Manure from
feed lot (10
different fresh
samples per
sampling date, 10
sampling dates)
Feces
Prevalence
97.9% of operations and 51
samples positive for
Campylobacter
2% of
Flock level prevalence of
Campylobacter increased from
1 0% of flocks (n = 1 0) at 1 8 days to
40% of flocks between 28 and 33
days and to 60% at depletion
Prevalence on conventional broiler
farms slightly lower (44-80%) than
organic farms (70-100%)
C. jejuni and C. co// prevalence
differed with operation type;
prevalences for each farm type and
species were
Operation type
Beef cow-calf
Calf rearer
Dairy
Feedlot
Total
% positive
C.
jejuni
47.1
23.8
31.2
31.6
34.1
C.
CO//
0.6
20.0
5.8
13.3
7.7
For newly-arrived animals,
prevalence was 1.6% (n=10); after
final sampling (2 weeks prior to
slaughter) prevalence was 62.2%
(n=10)
spp.: 98% (n =91)
C. jejuni: 1 00% of Campylobacter
spp. positive samples.
C. co//: 10% of Campylobacter spp.
positive samples.
Observations and Notes
Study also assessed antimicrobial
resistance of isolates
Among flocks colonized at depletion,
71% colonized exclusively by C. jejuni,
14% colonized by C. co// exclusively,
and 14% were colonized by both
Higher prevalence in organic operations
can be explained, in part, by the higher
age at which organic chickens are
typically slaughtered
Samples collected from 15 farms in
Washington state; findings based on a
relatively large (n = 686) sample size
Study conducted in a large commercial
feedlot (>50,000 head)
Some serotypes prevalent in dairy cattle
feces also prevalent in human feces in
samples collected in the same region
August 2010
B-7
-------
U.S. Environmental Protection Agency
Study
Devane et al.
(2005)
Devane et al.
(2005)
El-Shibiny et al.
(2005b)
Dorner et al.
(2004)
Dorner et al.
(2004)
Dorner et al.
(2004)
Animal/Source
Beef cattle feces
Surface waters,
stream source
Chickens (free-range
and organic)
Sows or gilts
Growers or finishing
pigs
Nursing or weaner
pigs
Species
spp,
C. co//,
C. jejuni
spp,
C. co//,
C. jejuni
C. jejuni and C.
CO//
spp.
spp.
spp.
Media
Feces
Water
Animals (rate of
infection)
Animals (i.e.,
infected
proportion)
Animals (i.e.,
infected
proportion)
Animals (i.e.,
infected
proportion)
Prevalence
spp.: 84% (n =87)
C. jejuni: 1 00% of Campylobacter
spp. positive samples
C. co//: 19% of Campylobacter spp.
positive samples
spp.: 55% (n = 293) (result reflects
less than 100%
recovery/sensitivity)
C. jejuni: 1 00% of Campylobacter
spp. positive samples
C. co//: 7.4% of Campylobacter
spp. positive samples
Free-range birds: Campylobacter
isolated from 68.5% (n=54) of birds
during the rearing cycle; first
incidence of Campylobacter
colonization observed at an age of
31 days
Organic birds: Campylobacter was
isolated from 90% (n=42) of birds
during the rearing cycle; first
incidence of Campylobacter
colonization observed at 8 days
45.9% positive (n = 315); and
79.7% positive (n = 59)
91.9% positive (n = 595); 100%
positive (n = 24); and 98.1%
positive (n=160)
63.6% positive (n = 93) and 79.3%
positive (n = 294)
Observations and Notes
Some serotypes prevalent in beef cattle
feces also prevalent in human feces in
samples collected in the same region
High prevalence of Campylobacter spp.
positive samples in surface waters
consistent with expectations
Studies performed on organic and free
range chicken farms in the U.K.
Practices on these farms differ
significantly from those on more
conventional farms. C. jejuni more
prevalent in chicks between 0 and 5
weeks of age. After 5 weeks of age C.
co// detected more frequently. An
estimated 80% of U.K. chicken meat
believed to be contaminated with
Campylobacter (27% C. co// and 73% C.
jejuni)
Authors suggest representing
prevalence with a beta distribution. For
the two studies, beta distribution
parameters were (a, fl) = (146,172) and
(a, $ = (48,13)).
For the three studies, beta distribution
parameters were (a, /J) = (548,49) and
(a, fl) = (25,1) and (a, fl) = (158,4)
For the two studies, the beta distribution
parameters were (a, fl) = (60,35) and
(a, f!) = (234,62)
August 2010
B-8
-------
U.S. Environmental Protection Agency
Study
Dorner et al.
(2004)
Dorner et al.
(2004)
Inglis et al.
(2004)
Minihan et al.
(2004)
Cox et al.
(2002)
Animal/Source
Chicken, broilers
Chicken, layers
Feed lot cattle
Beef cattle (heifers)
Chicken (breeders
and broilers)
Species
spp.
spp.
C. jejuni,
C. lanienae,
C.
hyointestinalis,
C. co//
C. jejuni and C.
CO//
spp.
Media
Animals (infected
proportion)
Animals (infected
proportion)
Abundance in
fresh feces from
feed lot cattle
housed
individually;
longitudinal study
duration of 4
months
Rectal samples,
fecal samples,
water trough
samples, feed
samples, and
samples from
carcasses
Feces
Prevalence
8 studies were reported; the three
with the highest number of
observations were 3.1% positive
(n=1 9,700,000), 42.5% positive
(n=89,110), and 27% positive
(n=12,233)
66.1% positive (n=280) and 42.9%
positive (n=105)
Campylobacter jejuni detected
13.4% of fecal samples (range 8.2-
16.7% over 4 sampling events); C.
laneinae detected in 55.5% of
samples (range 46.7-63.3%); C.
co// not detected in any sample
On arrival at the feeding lot, rectal
fecal samples from cattle indicated
12% infection rate; after four
months at the feedlot, the infection
rate was 76%. Among rectal fecal
samples positive for
Campylobacter spp., C. jejuni
accounted for 68. 5% of isolates, C.
co// accounted for 29.9% of isolates
and C. lari accounted for 1 .6% of
isolates.
Overall: 57.1% of fecal samples
Within flock low: 0% of fecal
samples
Within flock high: 100% of fecal
samples
Observations and Notes
For the three studies with the largest
number of observations, beta
distribution parameters (a, j3 ) = (606
001 , 1 9 094 001 ), (a,/3) = (37 873, 51
240), and (a, ft ) = (3305, 8930)
For the two studies, the beta distribution
parameters were (a, (J) = (186, 96), (a,
$ = (46,61)
Irish feedlots; infection prevalence
differed between pens at a feeding lot
and with residence of cattle at the
feeding lot; prevalence increased from
12% at the introduction of cattle to the
feeding lot to 76% after cattle had been
at the feeding lot for 4 months.
Campylobacter spp. more prevalence in
rectal swabs than from feces on the pen
floors, trough water, feed, or dust.
Environmental occurrence of
Campylobacter increased with the
duration of cattle at the feedlot. C. co//
were more prevalent than C. jejuni
among environmental samples.
Based on a total of 350 observations for
14 different flocks located in diverse
U.S. locations; transmission from
breeder to offspring through the egg
noted in the study discussion
August 2010
B-9
-------
U.S. Environmental Protection Agency
Study
Hoar et al.
(2001)
Whyte et al.
(2001)
Wesley et al.
(2000)
Donnison and
Ross (1999)
Animal/Source
Beef cattle
Chicken, broilers
Dairy cattle
Sentinel freshwater
mussel (Hydridella
menziesi)
Species
spp,
C. jejuni,
C.coli
spp.
C. jejuni,
C. co//
C. jejuni,
C. co//
Media
Fecal samples,
17 herds
Animals (infected
proportion, based
on fecal samples)
Fecal samples
(assumed to be
fresh) from 13
farms in 23 U.S.
states
Mussel slurries;
river samples;
wastewater
samples
Prevalence
Campylobacter spp. detected in 20
out of 401 fecal samples (5%)
C. jejuni detected in15 out of 20
samples culture positive for
Campylobacter spp. (75%)
C. co// detected in 3 of 20 samples
culture positive for Campylobacter
spp. (15%)
69% positive (n=70); 10 farms
tested with n=7 fecal samples per
farm. At 7 farms there were either
6 or 7 positive samples out of 7. At
three farms Campylobacter was
not detected in any fecal samples.
C. jejuni isolated from 37.7% (n =
2085) of fecal samples
C. co// detected in 1.8% (n=2085)
of fecal samples
Recoveries from mussels by site:
MP = C. jejuni 67% (2/3)
DF = C. jejuni 75% (3/4)
SW1 = C. jejuni and C. co// 75%
(3/4)
SW2 = C.jejuni 33% (1/3)
Recoveries from untreated
wastewater:
Sheep = C. jejuni and C. co// 80%
(8/10)
Cattle = C. jejuni and C. co// 91 %
(10/11)
Sewage/human = C. co// 86% (6/7)
Observations and Notes
Proportion of female animals was
positively associated with occurrence of
Campylobacter spp. in individual herds.
Farms tended to be 100% infected or
not infected at all; transport and holding
in processing facilities appears to have
limited or no influence on occurrence
C. jejuni prevalence higher among
lactating cows than non-lactating cows
(46.9% vs. 39.8%) and on farms with
more than 1 00 head than smaller farms
(45.2% vs. 37.7%)
Analyzed mussels recovered from
river/stream systems with inputs of fecal
pollution from known sources
(MP=meat-processing wastewater;
DF=non-point inputs from dairy farms;
SW1 and 2=two treated sewage plants)
August 2010
B-10
-------
U.S. Environmental Protection Agency
Study
Weijtens et al.
(1997)
Humphrey and
Becket(1987)
Animal/Source
Swine (sows and
piglets)
Dairy cattle
Species
C. co//
C. jejuni
Media
Feces
Fecal samples,
rectal swabs, 12
herds
Prevalence
All Campylobacter isolated were C.
CO//
Sows: 9/10 infected before
delivery, 10/10 infected after
delivery
Piglets: 8/10 infected at 1 week
post-delivery; 10/10 infected at 4
and 8 weeks post delivery
Overall, 24.5% (n = 668) of cows
had positive C. jejuni fecal
samples; herd prevalence ranged
from 0% to 36%
Observations and Notes
Conducted on two Dutch farms;
between 1 and 4 weeks post-delivery,
piglets rapidly acquired infections and
began shedding. Fecal samples from
more than half of the piglets positive for
Campylobacter at the first fecal sample
(1 week post-delivery).
Prevalence did not appear significantly
different for cows whose water source
was chlorinated vs. unchlorinated
August 2010
B-ll
-------
U.S. Environmental Protection Agency
Table B-4. Summary of studies reporting prevalence of Salmonella in cattle, swine, chicken, other domestic animals, and environmental
samples
Study
Cummings et
al. (2009)
Dorr et al.
(2009)
Dorr et al.
(2009)
Haley et al.
(2009)
Kunze et al.
(2008)
Callaway et
al. (2005)
Fossleret al.
(2005a)
Animal/Source
Dairy cattle
Pigs (nursery)
Pigs (finisher)
Various
Feedlot beef
cattle
Dairy cattle
Dairy cattle
Media
Rectal fecal
samples
Fecal samples
Fecal samples
Receiving waters
(streams)
Feces from feedlots
Dairy cattle feces
Dairy cattle and calf
feces
Prevalence
Herd-level prevalence 11%; prevalence among all
individual samples 22.5%
1 0.4% positive for Salmonella (5 farms). Highest
and lowest observed prevalences (on individual
farms) were 21.7% and 5%. Serotype prevalences
were Typhimurium (48%); Derby (33%); Muenchen
(7.3%); London (4.1%); and Mbandaka (3.1%).
15.4% positive (based on fecal samples) for 9
farms. Highest and lowest observed prevalences
(on individual farms) were non-detect (0) and
33.9%.
Salmonella spp. were found in 79.2% of
environmental waters collected (n=72). Studies
were conducted in north central Georgia. Among
positive samples 13 serotypes were identified. The
prevalence of subspecies and serotypes was: S.
enterica subsp. arizonae 41%; Muenchen 14%;
Rubislaw 13%; Mikawasima 6%; Braenderup 6%;
Saint Paul 5%; other serotypes were Bareilly,
Liverpool, I 4,[5]:b, Gaminara, Montevideo,
Anatum, I47:z4z23, not typed.
Salmonella enterica recovered from 30.3% of fecal
samples (n=182)
Salmonella found in 56% of herds studied; overall
prevalence (among all fecal samples) of 9.07%
Prevalence of Salmonella spp. for all data
(collected in 5 states) ranged from 2.7% (n=5220)
in the winter to 6.4% (n = 6417) in the summer.
Prevalence 5.2% on organic farms and 4.8% on
conventional farms; the difference between
prevalence by farm type was not statistically
significant.
Observations and Notes
Large study in geographic extent (NE United
States) and number of samples analyzed (n =
2565 dairy cattle)
Salmonella prevalence increased significantly
with age
Salmonella prevalence increased significantly
with age
The strongest determinants of Salmonella
occurrence were antecedent rainfall and
temperature (highest Salmonella abundance
in August). Occurrence of serotypes highly
variable.
No apparent seasonal variation in prevalence
reported
Study conducted on herds from 4 states, with
state selection weighted toward western
states
Factors associated with increased prevalence
of Salmonella were:
• season
• health status of cattle
• Midwest farm location
• herd size >1 00 head
August 2010
B-12
-------
U.S. Environmental Protection Agency
Study
Edrington et
al. (2004)
Hutchison et
al. (2004)
Hutchison et
al. (2004)
Garber et al.
(2003)
Johnson et al.
(2003)
USDA (2003)
Warnick et al.
(2003)
Animal/Source
Dairy cattle
Pig feces
Chicken feces
Chickens
Various
Dairy cows
Dairy cattle
Media
Fecal samples from
large dairy
operations
Commercial farms
Fecal samples from
commercial farms
Hen houses (layers)
Streams
Rectal swabs
Fresh feces
Prevalence
Salmonella spp. detected in 0-93% of samples
taken at each sampling event (4 farms, two
summer samples and one winter sample)
Salmonella spp. detected in 7.9% of fecal samples
(n = 126)
Salmonella spp. detected in 17.9% of fecal
samples (n = 67)
Salmonella enterica serotype enteritidis found in
7.1% of layer houses (n=200, all U.S. facilities)
Salmonella detected in 14/468 (3%) of water
samples taken over two years
Overall prevalence for culture-positive E. coli O157
was 4.3%; 38.5% of operations had one or more
positive cows.
Prevalence highest in summer (June - 8.2%) and
lowest in spring (April - 1.5%). Highest prevalence
of positive cows found in West region (7.6%),
Midwest (3.5%), Southeast (3.1%), and Northeast
(1.6%). Large operations (>500 cows) more likely
to have positive samples than medium operations
(100-499 cows) or small operations (<100 cows).
The majority of large dairies are in the West region.
Within-herd prevalence highly-variable and ranged
from 0%to 100% of animals; overall, Salmonella
isolated from 9.3% of 4049 fecal samples
Observations and Notes
While a general seasonal trend in Salmonella
shedding was reported (higher summer
shedding than winter shedding), variability
was extremely high.
Although farms sampled were relatively close
in proximity, serotype prevalence was very
different from farm to farm.
Wastes taken from farms throughout Great
Britain and results believed representative of
overall prevalence in the region
Wastes taken from farms throughout Great
Britain and results are believed
representative of overall prevalence in the
region
Factors associated with higher incidence of
S. enterica serotype enteritidis were the
following:
• flock size >100,000
• flocks 0-16 weeks post-molting
• young flocks
• floor-reared (rather than cage reared)
• location in Great Lakes region
• no cleaning and disinfection of feeders
and hoppers between flocks
The highest prevalence was in storm drain
waters; the lowest prevalence was in samples
taken in an urbanized area
Samples collected from March to September
2002; total of 3,733 samples for culture and
ID of £. coli O1 57, sfx 1, six 2, and antigens.
Samples collected from 5 operations from
each of the 21 participating states:
• West region: CA, CO, ID, NM, TX, WA
• Midwest region: IL, IN, IA, Ml, MN, MO,
OH.WI
• Northeast region: NY, PA, VT
• Southeast region: FL, KY, TN, VA
Large, multi-state study; over the course of
the study, Salmonella isolated from at least
one fecal sample from every farm sampled
August 2010
B-13
-------
U.S. Environmental Protection Agency
Study
Huston et al.
(2002)
USDA(2001)
Wells et al.
(2001)
Hayes et al.
(2000)
Donnison and
Ross (1999)
Byrd(1998)
Animal/Source
Dairy cattle
Beef cattle
(feed lot)
Dairy cattle
Chicken litter and
drag swabs
Freshwater
mussel
(Hydridella
menziesi)
Chicks
Media
Feces
Fresh manure from
feedlot pens
Fresh feces
Broiler and roaster
houses
Mussel slurries;
river samples;
wastewater
samples
Chickens
Prevalence
Overall herd-level prevalence 31.4% and individual
cow prevalence 17.7% among cows from infected
herds
Ranged from 3.3 % in samples taken in the winter
(Feb) to 19.9% in fall samples (Sept). No
geographic trends for STEC prevalence. All
feedlots had at least one positive sample during the
study.
Prevalence of culture-positive samples per region:
8.4% in Middle Region (CO, KS, OK); 11.5% in
Northern Region (ID, IA, NE, SD, WA); 13% in
Southern Region (CA, NM, TX)
Overall prevalence of Salmonella 10% (n = 6595)
Salmonella spp. detected in 48/71 (55.8%) of
facilities
Recoveries from mussels by site:
MP = 67% (2/3) S. typhimurium
DF = 0% (0/4)
SW1 = 0% (0/4)
SW2 = 0% (0/3)
Recoveries from untreated wastewater:
Sheep = 50% (7/10)
Beef =50% (11/22)
Sewage (human) = 44% (4/9)
Incidence of Salmonella infection among chicks
leaving hatcheries was estimated at 5-9%. Within
three weeks of entering growing houses,
prevalence rose to between 72 and 95%.
Observations and Notes
The only factors associated with higher
shedding prevalences were farm size; larger
farms were associated with higher
Salmonella prevalences.
73 feedlots/422 pens in 11 leading cattle
feeding states sampled for STEC from Oct
'99 to Sept '00. Total of 10,415 samples.
Samples from pens for cattle that had been
on feed the shortest (13.9%) were more likely
to be positive than samples from pens for
cattle that had been on feed the longest
(8.6%).
Prevalence varied with season (higher
shedding prevalence in summer than winter)
and with herd size (higher shedding
prevalence in herds with more than 100 cows
than in herds with fewer than 100 cows.
Litter water content found to be a poor
predictor of Salmonella occurrence.
Analyzed mussels recovered from
river/stream systems with inputs of fecal
pollution from known sources (MP = meat-
processing wastewater; DF = non-point
inputs from dairy farms; SW1 and 2 = two
treated sewage plants).
Based on studies conducted prior to 1998.
Horizontal transmission between chicks co-
housed in hatcheries was shown to be highly
efficient. Fecal shedding (and transmission)
was dependent on doses ingested by chicks
when infection was acquired.
August 2010
B-14
-------
U.S. Environmental Protection Agency
Study
Animal/Source
Media
Prevalence
Observations and Notes
USDA(1998)
Dairy cows
Rectal swabs of
dairy cows, cows to
be culled within 7
days, and cows at
cull dairy cow
markets
24.2% of operations and 30.9% of markets had at
least one culture-positive E. coli O157 sample.
Prevalence on farm - 0.9% of samples positive.
Prevalence for cows to be culled with 7 days -
2.8% of samples positive.
Prevalence for culled dairy cows at markets - 1.8%
of samples positive.
Fecal samples collected from 91 dairy
operations and 97 cull dairy cow markets in
19 states during a one-time sampling event.
Samples collected from Feb^July 1996.
Seasonal pattern of shedding was observed,
samples more likely to be positive after May 1
than before May 1.
No significant differences found between
cows on farm and cows going to slaughter.
Prevalence was higher for herds with 100 or
more cows (39.1% of herds had at least one
positive sample) than for herds with fewer
cows (8.9 % of herds had at least one
positive sample), however seasonality may
have been a factor.
USDA(1995)
Beef cattle
(feed lot)
Fresh manure from
feedlot pens
Overall prevalence was 1.61% of collected
samples.
Prevalence of positive feedlots per region: 59.4%
in Middle Region (CO, NE, KS, OK); 58.3% in
Northern Region (ID, IA, IL, MN, SD, WA); 71.9%
in Southern Region (CA, AZ, TX)
Pens in feedlots from 13 leading cattle
feeding states sampled for E. coli O157:H7 in
fall of 1994. Total of 11,881 samples.
Samples from pens for cattle that had been
on feed the shortest (47.1%) more likely to be
positive than samples from pens for cattle
that had been on feed the longest (16.8%).
Ebel et al.
(1992)
Spent layer hens
Cecal contents
Overall prevalence (% of layer houses) of
Salmonella was 24% and overall prevalence of S.
enteriditis was 3%; no differences in Salmonella
prevalence associated with U.S. regions. S.
enteriditis more prevalence in the northern U.S.
(45%) than in the southeastern or western/central
U.S.
Authors speculate that high S. enteriditis
prevalence in the northern U.S. may relate to
the tendency toward larger flocks in that
region
August 2010
B-15
-------
U.S. Environmental Protection Agency
Table B-5. Summary of studies reporting prevalence of E. coli O157:H7 in cattle, swine, chicken, other domestic animals, wildlife, and
environmental samples
Study
Berry et al.
(2007)
Cornick and
Helgerson
(2004)
Cornick and
Helgerson
(2004)
Edrington et al.
(2004)
Hutchison et al.
(2004)
LeJeune et al.
(2004)
Feder et al.
(2003)
Johnson et al.
(2003)
Booher et al.
(2002)
Animal/Source
Beef cattle
(feed lot)
Swine
Swine
Dairy cattle
Swine
Feedlot cattle
Swine
Various
Swine
Media
Manure (composite
samples)
Animals (based on
multiple studies)
Feces
Feces
Samples of fresh
manure
Fresh manure from
feedlot pens
Pig colon
Streams in a mixed use
watershed
Fecal samples
Prevalence
Ranged from 9.7 % in samples taken in
the winter to 22.6% of samples in the
fall.
0.1-6% of animals
Shortly after inoculation fecal £ coli
density ranged between 103 and 107
CFU/g. Two weeks after inoculation
fecal £ coli O157:H7 density ranged
from 50 to 1000 CFU/g. Two months
after inoculation fecal £ coli density
ranged from non-detectto 104 CFU/g.
Variable prevalence observed from
farm to farm and season to season.
Prevalence ranged from 0-35% (of
fecal samples collected from a given
farm on a sampling event). No positive
samples were found on any farm
among winter samples.
11.9 % (n = 126) of samples were
positive for £. coli O157
13.3%(n=4790)
2% (6 out of 305)
£ coli O157:H7 detected in 13 out of
1483 (0.9%) of water samples taken
from 10 locations over 2 years
High dose experiments: recovery after
2 weeks for STEC strains varied from
75 (6/8) to 100% (8/8) and for all other
three strains at 12.5% (1/8); however,
Observations and Notes
Samples collected every four weeks for 26 months.
Manure samples were composite samples
collected from feedlots.
Range based on studies performed in Europe and
North America.
Three month old pigs challenged with graded
doses of £ coli O1 57: H7. Pigs housed indoors on
concrete floors or decks. Experiments conducted
in Iowa.
Prevalences from the same farms on successive
summers were highly variable.
Wastes taken from farms throughout Great Britain
and results are believed representative of overall
prevalence in the region.
There was no apparent influence of trough water
chlorination on fecal shedding prevalence.
No inferences could be made of £ co//O157:H7
isolation rates with respect to the season, or swine
or herd prevalence. PCR confirmed two genotypes:
isolates harboring the eaeA, stxl, and stx2 genes
and isolates harboring the eaeA, stxl, and hly933
genes; ribotyping did not discriminate among
isolates within the £ coli O157:H7 serotype.
Highest prevalence (9.1% of samples) observed for
a stream reach within an urbanized area; £ coli
O157:H7 not detected at many sites, including
those draining high-intensity livestock regions and
in storm drains
High (inoculation) and low (feeding) dose
experiments using 3-month old pigs and a mixture
of five £ coli strains, including two STEC O157:H7
strains, two enterotoxigenic £ coli (ETEC) strains,
August 2010
B-16
-------
U.S. Environmental Protection Agency
Study
Donnison and
Ross (1999)
Chapman et al.
(1997)
Chapman et al.
(1997)
Chapman et al.
(1997)
Chapman et al.
(1997)
Rice et al.
(1995)
Animal/Source
Sentinel
freshwater
mussel
(Hydridella
menziesi)
Chicken
Swine
Dairy cattle
Beef cattle
Cattle/deer
Media
Mussel slurries; river
samples; wastewater
samples
Fecal samples, taken
immediately after
slaughter
Fecal samples, taken
immediately after
slaughter
Fecal samples, taken
immediately after
slaughter
Fecal samples, taken
immediately after
slaughter
Fresh fecal samples
collected from cattle
and deer
Prevalence
at 2 months, most of the recovered
strains were STEC that varied from
37. 5 (3/8) to 50% (4/8).
Low dose experiments: recoveries
varied from 50% (all strains but one
STEC) to 67%. No STEC strain was
recovered at 2 months. Although one
STEC strain was not recovered from
the low dose group at 2 months, it was
recovered from the cecum (but not
elsewhere in the alimentary tract) of 2/6
pigs at necropsy.
Mussel slurries (shucked and
homogenized for analysis) and river
samples collected
0/1 000(0%) chickens
4/1 000 (0.4%) of pigs
16. 1%(n=1661) of culled dairy cattle
13.4%(n=1840)of beef cattle
1.85% (2/108) deer and 2.6% (5/191)
cow samples tested positive
Observations and Notes
and one enteropathogenic £ coli (EPEC) strain.
High dose: 3-month old pigs inoculated with the
mixture at 1010 CFU per strain). Low dose: 3-month
old piqs fed both STEC O157:H7 strains at a dose
of 10 CFU per strain and other strains at a dose of
1010CFU per strain.
STEC strains persisted in the alimentary tracts of
some pigs at 2 months post-inoculation for high
and low dose mixtures; when all strains were given
at 1010 CFU (high dose), STEC strains persisted in
greater numbers and in more pigs than did the
other £ coli strains.
3x102-5x10b CFU/100g slurry (mussel); 90-2x1 0J
CFU/mL (water samples)
Studies conducted in the U.K.
Studies conducted in the U.K.
Studies conducted in the U.K.; highest prevalence
observed in late spring and early summer.
Studies conducted in the U.K.; highest prevalence
observed in late spring and early summer.
Fecal samples measured for £ coli O157
August 2010
B-17
-------
U.S. Environmental Protection Agency
Study
Animal/Source
Media
Prevalence
Observations and Notes
Schoeni and
Doyle (1994)
Chickens
Fresh fecal samples
collected after 1 hour of
defecation
83-100% of chicks administered
2.6x101 to 2.6x105 £ co//O157:H7
colonized at some time during the 12
weeks of examination; £ coli recovered
from cecal tissue of two of six chickens
(33%)
£ coli also isolated from the shells of
eggs but not from the yolks and whites
at a rate of 13.9% (14/101)
£. coli inoculated orally: 2.6x10nto 2.6x1 Ob per
chick; £ coli colonization persisted up to 4 months
when inoculated up to 105 CFU/chicken and up to
10-11 months when inoculated with 108
CFU/chicken
August 2010
B-18
-------
U.S. Environmental Protection Agency
Appendix C. Abundance Data
Table C-1. Summary of studies reporting abundance (concentrations) of Cryptosporidium in feces and related media of domestic
animals
Study
Moriarty et al.
(2008a)
Berry et al. (2007a)
Atwill et al. (2006b)
Hutchison et al.
(2005b)
Hutchinson et al.
(2004)
Hutchison et al.
(2004)
Atwill et al. (2003)
Source/Media
Dairy cattle
manure
Beef cattle feces
from feedlot
Beef cattle feces
from feedlot
Fresh and stored
pig manure
Cattle manure
Fresh and stored
chicken manure
Beef cow (> 24
months) feces
Species
spp.
spp.
C. parvum
C. parvum
C. parvum
C. parvum
C. parvum
Description
Samples taken from freshly-
deposited manure
Manure from beef feedlots sampled
(composite samples) each 4 weeks
during a 26 month study
Manure from 22 feedlots in 7
western and central states sampled
from August 2000 to January 2002
Composite samples from fresh and
stored manure collected between
April 2000 and December 2002
Manure samples collected from
throughout U.K.
Composite samples from fresh and
stored manure collected between
April 2000 and December 2002
Manure samples from preparturient
and postparturient beef cows on
three California farms were sampled
and C. parvum was enumerated via
a sensitive method.
Abundance
Among positive samples,
Cryptosporidium density
ranged from 1-25 oocysts/g
feces
Average: 14 oocysts/g
Range: 0.5 oocysts/g manure
to 1510 oocysts/g manure
Among samples positive for C.
parvum, the geometric mean
was 447 oocysts/g manure
(range 203-7702 oocysts/g)
Geometric mean densities 58
for fresh manure and 33 for
stored manure
For fresh manure, geometric
mean density was 19
oocysts/g (n = 44); maximum
density 3500
For stored manure, geometric
mean density was 10
oocysts/g (n=12); maximum
density 480
No C. parvum were identified
in any chicken manure
samples
For samples positive for C.
parvum, the arithmetic mean
oocyst density was 3.38
oocysts/g feces and the
standard deviation was 2.64
oocysts/g feces
Notes
Prevalence low in the herds
studied
Cryptosporidum identified in
58% of composite manure
samples collected over a 26-
month study
C. parvum detected in only
0.2% of samples;
abundance data fit with a
negative binomial
distribution
No significant difference in
prevalence or shedding of C.
parvum between pre-
parturient and post-
parturient cows
August 2010
C-1
-------
U.S. Environmental Protection Agency
Study
Sturdee et al. (2003)
Heitman et al.
(2002)
Wade et al. (2000)
Wade et al. (2000)
Source/Media
Cattle feces
Manure from
dairy cattle
Dairy cattle feces
Dairy cattle feces
Species
C. parvum
C. parvum
C. muris
C. parvum
Description
Rectal and recently-deposited fecal
samples collected at a farm with
beef and dairy cattle and calf rearing
operations
Manure from two dairy operations
collected from pasture
Fecal samples collected rectally
from dairy cattle at 109 farms in
southeastern New York; data
stratified by cattle age
Fecal samples collected rectally
from dairy cattle at 109 farms in
southeastern New York; data
stratified by cattle age
Abundance
Description
Bull beef
Dairy cow
Calf, Home-
bred
Calf,
brought-in
Mean of
positive
samples
(oocysts/g)
1371
1778
107,025
24,448
Mean densities in manure from
the two farms were 18.8 and
490 oocysts/g (considering
only positive samples)
Mean: 24,413 oocysts/g feces
Range: 1-100,000 oocysts/g
feces
Mean: 21,090 oocysts/g feces
Range: 1-79,040 oocysts/g
feces
Notes
Highest observed density
was 280,000 oocysts/g
feces for a home-bred calf
C. muris not detected in any
fecal samples
C. muris recovered from
animals with a wide range of
ages
C. parvum recovered only
from calves less than 30
days of age
August 2010
C-2
-------
U.S. Environmental Protection Agency
Table C-2. Summary of studies reporting abundance (concentrations) of Giardia in feces and related media of domestic animals
Study
Hutchison et al.
(2004)
Hutchison et al.
(2004)
Ralston et al. (2003)
Heitman et al.
(2002)
Heitman et al.
(2002)
Wade et al. (2000)
Source/Media
Cattle farmyard
manures and
slurries
Swine farmyard
manures and
slurries
Range beef calf
and dam
manures
Dairy cattle
manure
Pig manure
Dairy cattle
manure
Species
G. intestinalis
G. intestinalis
spp.
spp.
spp.
spp.
Description
Results are for samples
collected throughout Great
Britain
Results are for samples
collected throughout Great
Britain
Fecal samples were collected
from calves and dams from
range operations in Canada
Fecal samples collected from
farms in Canada
Fecal samples collected from
farms in Canada
Fecal samples collected from
212 farms in southeastern New
York
Abundance
Geometric mean and
maximum cyst densities 10
and 5000 cysts/g, respectively
Geometric mean and
maximum cyst densities 68
and 160,000 cysts/g,
respectively
Giardia abundance in feces
varied with animal age group.
Density ranged from 0 at 1
week of age to a maximum of
2230 cysts/g (range 0-574,933
cysts/g of feces) of feces at 5
weeks of age. The geometric
mean decreased after week 5
to a low of 2 cysts/g at 25-27
weeks
of age
Mean cyst range 1.5-29.9
cysts/g
Mean cyst density 16.1 cysts/g
1-85,217 cysts, mean of 3039
cysts/g feces
Notes
August 2010
C-3
-------
U.S. Environmental Protection Agency
Table C-3. Summary of studies reporting abundance (concentrations) of Campylobacter'm feces and related media of domestic animals
Study
Mclaughlin et al.
(2009)
Moriarty et al.
(2008)
Bull etal. (2006)
El-Shibiny et al.
(2005b)
Hutchison etal.
(2005)
Hutchison et al.
(2005)
Hutchison et al.
(2005)
Dorner et al. (2004)
Dorner et al. (2004)
Source/Media
Swine lagoons
Dairy cattle
Chicken breeder
and broiler flocks
Chicken (range of
values from
published
studies)
Cattle
Swine
Chicken
Chicken (broiler)
feces
Nursing or
weaner pigs
Species
spp.
spp.
spp. (study
included
speciation)
spp.
spp.
spp.
spp.
spp.
spp.
Description
Samples taken from swine lagoons
receiving wastes from sows,
nursery pigs, or finishing pigs.
Samples taken from 4 farms
considered to span conditions in
New Zealand
Samples taken from environmental
areas (puddles, air) and from fresh
fecal deposits on house floors
Estimates based on multiple
published studies
Composite samples of manure
from pens collected
Composite samples of manure
from pens collected
Composite samples of manure
from pens were collected
Abundance data from multiple
studies were pooled and fit to a
gamma distribution
Abundance data from a single
study (Weijtens et al., 1999) fitted
to a gamma distribution
Abundance
Sows lagoon: mean density of
5100CFU/100mL
Nursery lagoon: mean density
of3500CFU/100ml_
Finisher lagoon: mean density
of 1900CFU/100mL
For all seasons: median 430
CFU/g, range 1 5-1. 8x107
CFU/g
Densities varied within a
relatively small range among
flocks (for flocks colonized by
Campylobacters) and did not
change significantly once a
flock was colonized; the range
of observed densities was
4.0x104-5.0x106 organisms/g
feces
10b-10a CFU/g excreta
320 CFU/g for fresh feces
530 CFU/g for stored feces
310 CFU/g for fresh feces
1600 CFU/g for stored feces
260 CFU/g for fresh feces
590 CFU/g for stored feces
Gamma-distributed
abundance, distribution
parameters (a, j3) = (27.78,
0.2558)
Gamma-distributed
abundance, distribution
parameters (a, ft) = (4.419,
0.6319)
Notes
For each type of waste, 1 0
lagoons were sampled; all
farms located in the mid-
south U.S.
Prevalence of C. jejuni and
C. co// reported, but not
related to abundance in
manure; Campylobacter
abundance bi-modally
distributed among samples
Species prevalence among
samples positive for
Campylobacter differed
between flocks; 5 flocks
colonized by C. jejuni
exclusively, 1 flock with C.
co// exclusively, and 1 flock
with both
August 2010
C-4
-------
U.S. Environmental Protection Agency
Study
Dorner et al. (2004)
Inglisetal. (2004)
Cox et al. (2002)
Whyteetal. (2001)
Source/Media
Sows and gilts
Feedlot cattle
feces
Chicken feces
Chicken feces
Species
spp.
C.jejuni, C.
lanienae, C.
hyointestinalis,
C. co//
spp.
spp.
Description
Abundance data from two studies
(Weijtens et al., 1997; Weijtens et
al., 1999) fitted to a gamma
distribution
Abundance in fresh feces from
feedlot cattle housed individually
Results from composite of samples
taken from 35 commercial broiler
farms; results segregated by age
of chicken (breeders vs. broilers)
Fecal samples from sacrificed
chickens from 10 Irish farms were
enumerated for Campylobacter.
Although samples analyzed before,
during, and after transport to a
processing facility, the only values
quoted here are for before
transport. Studies conducted in
Ireland.
Abundance
Gamma-distributed
abundance, distribution
parameters (a, /3 )= (4.207,
0.8859)
C.jejuni density range 0.01-
1.03 log-iocells/g (via RT-
PCR); C. lanienae density
ranged from 1.47-4.74 log™
cells/g; C. co// not detected in
any sample; C. jejuni detected
1.4% of fecal samples (range
8.2-16.7%)
Breeders: 2.8-3.9 log™
CFU/g feces
Broilers: 3.5-6.5 log™ CFU/g
feces
6.11+0.37 log-io CFU/g feces
for 5 farms
6.61+0.38 log-io CFU/g feces
for 5 additional farms
Notes
Longitudinal study duration
of 4 months
Campylobacter less
prevalent in broilers
(offspring) than breeders,
but shedding (colonization)
was higher in broilers than
breeders
August 2010
C-5
-------
U.S. Environmental Protection Agency
Study
Weijtens et al.
(1999)
Stanley et al. (1998)
Stanley et al. (1998)
Weijtens et al.
(1997)
Source/Media
Fattening pig
feces from 10
weeks of age to
25 weeks
Beef cattle feces
Dairy cattle feces
Sow feces at one
week prior to
delivery
Species
spp.
spp.
spp.
spp.
Description
For each sampling event, 6 feces
samples collected per pig. Pigs
monitored from birth and housed
with 16 pigs each on an
experimental farm.
Fresh beef cattle sampled at
slaughter
Fresh dairy cattle manure samples
collected in pens of 4 dairy herds in
the United Kingdom
Sow feces were sampled and
bacteria were enumerated 1 week
prior to delivery
Abundance
At 13 weeks: mean fecal
Campylobacter density
4.1+0.7 log™ CFU/g (n =8
pigs, average of 6 fecal
samples per sampling event
per pig)
At 19 weeks: mean fecal
Campylobacter density
3.3+1. Olog-io CFU/g (n =8
pigs, average of 6 fecal
samples per sampling event
per pig)
At 25 weeks: mean fecal
Campylobacter density
2.0+0.1 log™ CFU/g (n =8
pigs, average of 6 fecal
samples per sampling event
per pig)
61 OMPN/g feces
Adult cows: 69.9 MPN/g feces
(SD3)
Calves: 33,000 MPN/g (SD
170)
5.0+1.1 log-io CFU/g (farm 1,
n=5) and 3.6+0.4 log-io CFU/g
(farm 2, n=5)
Notes
The abundance (and
prevalence) of
Campylobacter varied
weekly and between fecal
samples on a given
sampling event. Several
pigs had periods of non-
detectable fecal
Campylobacter between
periods of high fecal
Campylobacter abundance.
Abundance highest shortly
after colonization and
generally decreased with
age.
Two peak periods
(seasonal) of shedding
were noted
Prevalence data for sows
and piglets were also
collected at 1 week, 4
weeks and 8 weeks post-
delivery
August 2010
C-6
-------
U.S. Environmental Protection Agency
Table C-4. Summary of studies reporting abundance (concentrations) of Salmonella in feces and related media of domestic animals
Study
Haley et al.
(2009)
Mclaughlin
et al. (2009)
Kunze et al.
(2008)
Boes et al.
(2005a)
Hutchison et
al. (2004)
Hutchison et
al. (2004)
Source/Media
Stream waters
Anaerobic primary
lagoon effluent
Feed lot cattle feces
Swine manure
slurry from 62
herds
Fresh pig manure
Fresh chicken
manure
Serotype
spp.
spp.
All
Typhimurium
spp.
spp.
Description
Water samples from a
mixed use (livestock, on-
site septic system, small
community) watershed
were sampled for
Salmonella
37 lagoons sampled;
lagoons received waste
from sow, nursery, or
finisher operations
Fecal samples taken from
feed lots
Samples were drawn from
swine manure slurry and
from soil after application
of swine manure slurry
Multiple commercial farms
Multiple commercial farms
Abundance
Geometric mean of Salmonella in waters did not
vary greatly among sampled sites. The highest
and lowest mean densities were 0.746
MPN/100 ml_ and 0.496 MPN/100 ml_
Mean densities for sow, nursery and finisher
operations were 28, 34, and 6.2 CFU/100mL,
respectively
Among samples positive for S. enterica, mean
fecal density was 0.75 log-io / g
Salmonellae detected in all slurry samples.
Average Salmonella Typhimurium density was
0.2 CFU/g (note: not log-io CFU); maximum
density was estimated to be 2500 CFU/g for a
sub-clinically-infected herd; observed
abundance among 112 slurry samples was 33%
of samples with < 0.1 MPN, 13% of samples
between 0.1 and 1 MPN, 28% between 1 and
10 MPN, 12% between 10 and 110 MPN, and
14%> 100 MPN.
Geometric mean of 600 CFU/g (n = 10)
Maximum observation of 78,000 CFU/g
Geometric mean of 220 CFU/g (n = 12)
Maximum observation of 22,000 CFU/g
Notes
All lagoons sampled
were located in the mid-
south of U.S.
Danish farms; authors
proposed a polynomial
survival model for
Salmonella in soil
Wastes taken from farms
throughout Great Britain
and results believed to
be representative of
overall prevalence in the
region
Wastes taken from farms
throughout Great Britain
and results believed to
be representative of
overall prevalence in the
region
August 2010
C-7
-------
U.S. Environmental Protection Agency
Study
Hutchison et
al. (2004)
Byrd(1998)
Source/Media
Fresh cattle
manure
Cecal material and
chicken litter from
hatcheries
Serotype
spp.
Typhimurium
Description
Multiple U.K. commercial
farms
Day-old chicks were
challenged with 100, 104,
or 106 Salmonella
typhimurium by gavage.
Litter and cecal contents
were monitored for 17
days.
Abundance
Geometric mean of 2100 CFU/g (n=62)
Maximum observation of 580,000 CFU/g
Pens containing chicks inoculated with 100
Salmonellae: 2.05 to 3.03 log™ CFU/g litter
(n=70)
Pens containing chicks inoculated with 10
Salmonella: 2.39 to 4.55 log™ CFU/g litter
(n=10)
Pens containing chicks inoculated with 10
Salmonella: 3.65 to 4.42 log™ CFU/g litter
(n=10)
Notes
Wastes taken from farms
throughout Great Britain
and results believed to
be representative of
overall prevalence in the
region; Salmonella
density higher in stored
manure than fresh
manure
Cecal colonization rate
and Salmonella count in
cecal contents varied
according to challenge
dose. The number of
chicks inoculated (5%,
10%, 25% and 50% of
chicks in a pen) did not
influence the overall
incidence of infection in
the pen.
August 2010
C-8
-------
U.S. Environmental Protection Agency
Table C-5. Summary of studies reporting abundance (concentrations) of E. coli O157:H7 in feces and related media of domestic animals
Study
Hutchison et
al. (2004)
Booher et al.
(2002)
Kudva et al.
(1998)
Kudva et al.
(1998)
Kudva et al.
(1998)
Schoeni and
Doyle (1994)
Animal/Source
Swine
Swine
Sheep
Cattle
Cattle
Chickens
Media
Manure
Fecal
samples
Manure pit
Manure
Manure
slurry
Fresh fecal
samples
Description
Samples collected from multiple
commercial farms in the U.K.
Fecal samples taken on 3
sucessive days during each of the
following three post-inoculation
periods: days 2-4; days 14-16;
days 58-60
Composite samples (from manure
pits receiving waste from multiple
animals) collected and
enumerated for £ coli O157:H7.
Experiments conducted in Idaho
Composite samples (from manure
pits receiving waste from multiple
animals) collected and
enumerated for E. coli O157:H7;
experiments conducted in Idaho
Untreated slurries and treated
slurries (the retentate post-storage
and separation) sampled and
enumerated for £ coli O157:H7;
experiments conducted in Idaho
Chickens inoculated orally with £
coli varying from 2.6x1 01-2. 6x1 05
CFU/chicken
Abundance
Geometric mean of 3900
CFU£ co//O157/g(n=15);
highest observed density
750,000 CFU£ coli O1 57/g
High dose experiments:
5to103CFU/gfeces
Low dose experiments:
5to104CFU/gfeces
(Note: values are
approximate because they
were obtained from a visual
inspection of figures in the
paper)
1.15x10BCFU/g feces from
a composite sample
Two samples yielded
2.04x1 07CFU/g feces and
4.35x108CFU/g feces
Two samples of untreated
slurry yielded 1.02x106
CFU/mLand2.36x106
CFU/mL.
A single sample of treated
slurry yielded 2.35x106
CFU/mL
Short-term experiment:
highest level of inoculation
(2.6x105) £ coli detected in
feces averaging 4.6x102
CFU/gram of feces
Long-term experiment:
chickens inoculated with
1.3x108 CFU/chicken
average recovery picked at
4monthsat3.2x106CFU
Notes
High dose: pigs inoculated with a
mixture of 5 £ coli strains at 1010
CFU per strain
Low dose: pigs fed 2 STEC O157:H7
strains at a dose of 107 CFU per
strain and the other 3 strains at a
dose of 1010 CFU per strain (low
dose)
Prior to shedding, sheep
experimentally inoculated with £ coli
O157:H7; some of the animals
contributing to the manure pit were
not infected
Prior to shedding, cattle
experimentally infected with £ coli
O157:H7; some of the animals
contributing to the manure pit were
not infected
Prior to shedding, cattle were
experimentally infected with £ coli
O157:H7; some of the animals
contributing to the manure pit were
not infected
August 2010
C-9
-------
U.S. Environmental Protection Agency
Appendix D. Farm Factors Data
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Alter et al. (2005)
Campylobacter
(all isolates
identified as C.
CO//)
Swine farms
including
slaughter pig
producing
operations (farrow
to finishing)
operations for
finishing only
A single "ecological"
farm with free range
access
Factors evaluated with respect to their
association with Campylobacter in
feces from:
• individual animals
• animal age (new-
born/weaner, nursery pig,
fattening, at slaughter)
• organic vs. conventional
farming
No Campylobacter detectable in feces of piglets at
the day of birth.
Campylobacter incidence rose within days to 32.8%;
after transfer to the nursery unit prevalence
increased to 56.6%.
Approximately two-thirds of pigs remained C. coli
shedders in the fattening unit; detection rate before
transportation was 79.1%.
Based on a single organic operation, infection with
and shedding of C. coli at the organic operation
appeared to occur at an earlier age than in
conventional operations.
On conventional farms, the best predictor of
prevalence in a given growth stage is infection
prevalence in prior growth stage.
Greatest increase in prevalence occurred during
piglet weaning.
Prevalence of infection in piglets was not related to
prevalence of infection in mothers (sows).
A single farm had a much lower infection rate at
slaughter than all other farms (5.3%); this difference
could not be attributed to difference in rearing
system or hygienic conditions.
August 2010
D-l
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Arsenault et al
(2007)
Salmonella spp.
Campylobacter
spp.
Broiler chicken and
turkey farms in
Quebec, Canada
Factors evaluated with respect to
presence of Campylobacter and
Salmonella in pooled cecal contents of
approximately 30 birds per flock
evaluated
The most important factors were:
• age at slaughter
• # birds in chicken house
• cleaning practices
• vermin control
• distance to nearest manure heap
• permanent locking of chicken
house
Prevalence of Sa/mone//a-positive flocks was 50%
for chickens and 54% for turkeys
Odds of Salmonella colonization were 2.6* greater
for chicken flocks that failed to lock the chicken
house permanently.
No other factors were associated with a significant
change in odds of Salmonella in chicken houses.
In turkeys, odds of Salmonella colonization were
4.8-7.7* times greater for flocks that failed to be
raised by <2 producers with no other visitors allowed
onto the premises, or origin from a hatchery.
Prevalence of Campy/obacter-positive flocks was
35% for chickens and 46% for turkeys
Odds of colonization were 4.1* higher for chicken
flocks raised on farms with professional rodent
control and 5.2* higher for flocks with manure heap
>200 m from the chicken house, and also increased
with the number of birds raised per year on the farm
and with the age at slaughter.
For turkeys, odds of Campylobacter flock
colonization were 3.2* times greater in flocks having
a manure heap at >200 m from chicken house and
4.2* greater in flocks drinking unchlorinated water.
Baeetal. (2005)
Campylobacter
(all isolates
speciated)
Cattle farms,
including calf
rearing, dairy, beef,
and feed lot
operations
Factors evaluated with respect to their
association with the prevalence of
Campylobacters in fresh feces;
including:
• farm type
• calf rearing
• dairy
• beef
• feed lot
Prevalence of C. jejuni and C. coll excretion differed
by farm type.
Highest C. jejuni prevalence was observed at beef
cow-calf operations (47.1%) and the lowest at calf
rearer operations (23.8%).
Highest C. coli prevalence was at calf rearer
operations (20.0%) and the lowest was at beef calf-
cow operations (0.6%).
August 2010
D-2
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Barwick et al. (2003)
Cryptosporidium
spp. and Giardia
spp.
Dairy farms
Factors evaluated with respect to
occurrence of Cryptosporidium or
Giardia in dairy farm soils;
evaluations made via multivariate
logistic regression
The most important factors studied
were:
• land use in the immediate vicinity
of sample site
• soil pH
• herd prevalence of
Cryptosporidium or Giardia
Cryptosporidium
Risk factors associated with occurrence of
Cryptosporidium in individual soil samples were:
• land use at the soil sample site (farming fields
were 4* times more likely to have oocysts in soil
than non-farming areas)
• soil pH (acidic soils were associated with a
higher likelihood of Cryptosporidium detection)
Giardia
Risk factors positively associated with Giardia
detection in soil were
• cattle access to soils
• herd-level prevalence of Giardia infection
• presence of grass cover and soil moisture
(linear association with Giardia detection
frequency, with higher prevalence in soils with
higher moisture)
Beach et al. (2002)
Campylobacter
spp. and
Salmonella spp.
U.S. cattle feedlots
and pasture
operations
Factors evaluated with respect to their
association with presence of
Campylobacter and Salmonella in
rectal samples of individual cattle.
Factors related to pre- and post-
transport (to slaughter) infection rates
were evaluated. The only factor related
to pre-transport prevalence was animal
origin
farm type (feedlot vs. pasture).
Campylobacter and Salmonella prevalences in pre-
transport cattle are provided in the table below
Pathogen
Campylobacter
Salmonella
Pre-transport prevalence
Feedlot
64%
3%
Pasture
6.3%
1%
Besseretal. (2005)
C. jejuni
U.S. cattle feedlots
Factors evaluated with respect to
association with C. jejuni in fresh feces
taken from feedlots
duration of animal at feedlot
water chlorination
Prevalence of C. jejuni increased markedly from
cattle arrival; in samples from pens with newly-
arrived cattle, C. jejuni detected in 1.6% of fecal
samples; in samples from pens with animals within
two weeks of slaughter prevalence was 62.2%.
Water chlorination did not result in a significant
difference in the prevalence of C. jejuni in fecal
samples.
August 2010
D-3
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Boes et al. (2005)
C.jejuni, C. coli
Danish swine herds
from farms with and
without mixed
livestock production
Factors evaluated with respect to their
association with C. jejuni and C. coli in
individual fecal samples, and mixed
livestock on farm
No significant herd-level or animal level difference in
prevalence of C. jejuni or C. coli in herds on farms
with or without mixed production.
The ratio of C. coli to C.jejuni in swine feces was
consistent across herds with and without mixed
production.
Bull etal. (2006)
C. jejuni, C. coli
Housed broiler
chicken operations
in the U.K.
Factors evaluated with respect to their
association with occurrence of C.jejuni
and C. coli in floor fecal droppings.
Factors evaluated were:
• time (bird age and time since
arrival in house)
• environmental conditions
(presence of Campylobacter
in the broiler house
environment)
Once colonized with Campylobacter, flocks' fecal
droppings tended to have consistent and high
densities of Campylobacter.
Flock level prevalence of Campylobacter increased
from 10% of flocks (n=10) at 18 days to 40% of
flocks between 28 and 33 days and to 60% at
depletion.
Among colonized flocks, five colonized by C.jejuni
exclusively, one colonized with C. coli exclusively,
and one colonized with both species.
Cho et al. (2006)
STEC/
Dairy cattle, farm
environment and
county fairs in
Minnesota
Organic and conventional farms, calf
pens, county fairs
Note that organic agriculture is a
production system that seeks to
promote and enhance the health of
agroecosystems by using few inputs,
avoiding synthetic substances and
promoting animal welfare (Codex
Alimentarius Commission, 1999).
Shiga toxic bacteria (STB) prevalence greater in
organic farms compared to conventional farms
especially at the individual sample level, but was not
statistically significant when restricting the analysis
to only herds with <100 cows or in a herd-level
analysis.
In samples collected from conventional farms, 2.3%
of fecal samples were STB-positive and 65% of
farms had at least one positive animal; 6.6% of fecal
samples from organic farms were STB-positive and
87.5% of farms had at least one positive animal.
STB detected from 17.4% of samples and 58.3% of
manure piles at county fairs.
Organic farms smaller than conventional herds with
a mean of 37 and 132 milking cows per herd on
organic and conventional farms, respectively
The percent of STB-positive samples for each farm
(within-herd prevalence for each farm) ranged from 0
to 26% (median 5.4%) on organic farms and from 0
to 13.9% (median 1.1%) on conventional farms. _
August 2010
D-4
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Cummings et al.
(2009)
Salmonella
Dairy cattle from
herds in the NE
United States
Factors evaluated with respect to their
association with prevalence of
Salmonella in rectal fecal samples;
factors evaluated were:
• herd size
• housing type
• vaccination status
• history of Salmonella infection
Herd size only significant predictor of Salmonella
prevalence
Ebeletal. (1992)
Salmonella spp.
and S. enteriditis
Spent hens from
farms throughout
the United States
Regional differences in pooled cecal
contents for layers were evaluated
Free range and
organic chicken
farms in the U.K.
Salmonella recovered from 24% of pooled samples;
S. enteriditis was recovered from % of pooled
samples.
Layer house prevalences of S. enteriditis among
northern, southeastern, and western/central layer
houses were 45%, 3% and 17%, respectively.
El-Shibiny el al.
(2005)
C. co// and C.
jejuni
Rearing cycle of free range (56 days)
and organic chickens (73 days)
evaluated for C. coll and C. jejuni
Campylobacter isolated from 68.5% of the organic
birds and 90% of the free-range birds over the
rearing cycles.
Following initial colonization campylobacters were
detected throughout the rearing period in both
flocks, with the exception of a single bird at day 31
from the free-range flock.
Organic flock was colonized by campylobacters
susceptible to the majority of the antibiotics tested.
Authors noted that the most intensively reared
broilers are killed at approximately 35 days old,
which is the time at which we have observed the
succession of C. jejuni by C. coll as the dominant
species in the free-range and organic flocks.
August 2010
D-5
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Edrington et al.
(2004)
Salmonella and
£ co//O157: H7
Dairy farms in the
Midwest and NE
United States
Factors evaluated with respect to their
association with fecal shedding of
pathogens, including:
• season
• year-to-year variation
Salmonella prevalence varied widely within farms,
from year-to-year and from season to season.
Although there was a general trend toward higher
summertime Salmonella prevalences, this trend is
small in the context of the overall variability in
prevalence.
£ co//O157:H7 prevalence was also highly variable,
though seasonal effects were more easily observed;
no £ co// O157:H7 positive samples were observed
on any farm during any winter sampling event.
Ellis-lversen et al.
(2009)
C. jejuni, C. co//,
£. co//O157
25 dairy and 10 beef
cattle operations
distributed through
England, Wales,
and Scotland
Farm factors evaluated were:
• contact with other herds
• housing
• herd size
• visit
• number of suckling calves on
farm
• water trough hygiene
• operation type
• presence of ringworm-
infected animals
Campylobacter
Only herd size, water trough hygiene and number of
suckling calves related to Campylobacter
prevalence. Higher herd size associated with
increased prevalence of Campylobacter, whereas
more frequent water trough emptying and presence
of suckling calves associated with decreased
Campylobacter prevalence.
£ co//O157
Larger herds associated with increased £ co// O157
prevalence and presence of chicken and suckling
calves associated with reduced £ co// O157
prevalence.
Fossleret al.
(2005a)
Salmonella
Multiple cattle and
farm types
evaluated, including
organic and
conventional
Factors evaluated were:
• size
• season
• state
• treatment with antibiotics
• cattle type (sick,
periparturient, to be culled,
preweaned calf, healthy)
• organic vs. conventional farm
Season, state, farm size and cattle status associated
with Salmonella shedding .
Farm type (organic vs. conventional) not associated
with shedding.
Midwestern states more likely to have cattle
shedding Salmonella than cattle from NY
Salmonella shedding more likely on farms with at
least 100 cows.
Cattle that had been treated with antibiotics were
within 14 days less likely to shed Salmonella.
There were too few large organic farms to evaluate
the role of large herd sizes on Salmonella shedding
in organic herds.
August 2010
D-6
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Fossler et al.
(2005b)
Salmonella
Multiple cattle and
farm types
evaluated, including
organic and
conventional
Paper discusses herd level factors that
modify the Salmonella shedding
factors described in Fossler 2005a
(above); many herd level
characteristics were considered.
Examples include presence of
chickens, turkeys, pigs, and wild
geese; maternity pens present; type of
bedding; protein feeds stored in
enclosed buildings (refer to Appendix
A of that report for complete listing)
Herd size, season, and state forced into the models;
note that this is the largest study of Salmonella
shedding in dairy cows and the only study evaluating
herd level characteristics.
Herd levels factors in the model were:
• lack of use of tiestall or stanchion facilities to
house lactating cows
• not storing all purchased concentrate or protein
feeds in an enclosed building
• not using monensin in weaned calf or bred
heifer diets
• access of lactating or dry cows to surface water
• disposal of manure in liquid on owned or rented
land
• cows eating or grazing of roughage from fields
where manure was applied in solid or liquid form
and not plowed under during the same growing
season
Herd size not associated with Salmonella shedding.
Season and State (location) associated with
Salmonella shedding.
Farm type not associated with Salmonella shedding
(authors noted a lack of information in the literature
on this topic).
August 2010
D-7
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Fossler et al.
(2005c)
Salmonella
Multiple cattle and
farm types
evaluated, including
organic and
conventional
Paper reviews herd level factors for
Salmonella shedding in calves; many
herd level characteristics considered
(refer to Appendix A of that paper for
complete listing)
Herd size, season, and state were forced into the
models; however, herd size not associated with
Salmonella shedding
Management practices and characteristics that were
not associated with Salmonella shedding in calves at
p<0.20 after adjustment for effects of herd size,
season, state of origin and the multiple sampling
occasions per herd included:
• farm type (organic vs. conventional)
• percent of cows born off the farm
• type of maternity facility
• use of a chlorinated water source for dairy cattle
• type of coccidiostats used
• amount of colostrums fed to calves
• washing of calf milk buckets between feedings
• placement of sick cattle in a pen separate from
other lactating cows
• average herd milk somatic cell count
The following were associated with an increased
odds of Salmonella shedding in calves:
• presence of Salmonella positive cow on the
operation
• lack of routine feeding of milk replacer
containing antibiotics to preweaned calves
• use of maternity housing as a hospital area for
sick cows more than once a month
Franz et al. (2007)
STEC and £. coll
O157:H7 genes
Organic (ORG) and
low- input
conventional (LIC)
dairy farms in the
Netherlands
Organic and low- input conventional
(LIC) dairy farms; note, the majority of
the management practices such as
feeding regimen and housing
conditions remains unclear and under
debate
Prevalence of a gene specific for O157 was 52%
overall, and was higher at organic farms (61%) than
at LIC farms (36%), but the difference was not
significan.
Relatively more LIC farms were positive for all STEC
virulence genes.
The four manures that best supported £ coll
O157:H7 (all organic) were derived from farms with
exclusively Frisian Holstein cows, while two of the
four farms from which the manure supported the
worst survival of £ coll O157:H7 (two ORG and two
LIC) harbored another breed (both ORG) next to
Frisian Holsteins.
August 2010
D-8
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Garber et al. (2003)
SE
Layer operations-
environmental
samples
Factors explored with respect to their
association with the presence of SE on
environmental surfaces of layer
operations included:
• geographic region
o Southeast
o Central
o West
o Great Lakes
• molting/age
• floor rearing (vs. cage rearing)
• rodent presence (measured via
trapping)
• cleaning and disinfecting between
flocks
• manure handling (flush vs. high;
rise vs. deep pit)
• presence of SE in feed
• age of layer house
• floor area per bird
Overall, SE was isolated from 7.1% of layer houses;
regional prevalence estimates were:
• 0% in the southeast
• 9.0% (standard error = 7.2) in the central region
• 4.4% (standard error = 2.5) in the west, and
17.2% (standard error = 13.7) in the Great
Lakes region
Approximately 4% of houses with fewer than
100,000 layers were environmentally positive for SE,
whereas 16.5% of houses with 100,000 or more
layers were environmentally positive for SE.
Molted flocks were more likely to be associated with
SE than unmolted flocks of the same age.
No association with SE was found for
presence of SE in feed age of layer house
floor area per bird.
Gebreyes et al.
(2008)
Salmonella
Anti-microbial free
and conventional
swine systems
Locations (Wl, NC, OH)
Anti-microbial free and conventional
production systems
Note: swine raised in outdoor
production units have full or partial
outdoor access on dirt with open
access to soil, vegetation and wild
fauna.
Significantly higher (p=0.0001) seroprevalence of
Salmonella from anti-microbial free herds (54%) than
conventional herds (39%).
Salmonella more common in anti-microbial free,
outdoor niche market than conventional indoor
reared herds, although there was some
geographical variation in Salmonella (Wl highest at
59%, followed by NC at 34%, and Ohio at 34%).
Harvey et al. (2004)
C. jejuni, C. coll
Dairy
Prevalence in fecal samples evaluated
with respect to association with region
of the U.S. (northeast, southwest, or
Pacific west) predominant
Campylobacter species
Low prevalence of Campylobacter observed overall
(5.2% for the desert southwest, 2.9% for the
northeast, and 5.0% for the Pacific west) and on-
farm prevalence ranged from 0-10%.
No difference in Campylobacter prevalence
observed between regions.
August 2010
D-9
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Heueretal. (2001)
Campylobacter
spp.
Conventional and
organic chicken
(broiler) flocks
3 rearing systems evaluated:
• organic
• conventional
• extensive indoor production farms
Campylobacters isolated from 100% of organic
broiler flocks, from 36.7% of conventional broiler
flocks and from 49.2% of extensive indoor broiler
flocks.
Proportion of Campy/obacter-positive flocks
significantly higher for organic flocks compared with
conventional flocks (p<0.001) and extensive indoor
flocks (p<0.001).
Organic broiler flocks constitute a strong potential for
introduction of Campylobacterto the processing line
upon arrival at slaughter
No single factor related to organic broiler production
can be pointed out as the sole determinant of high
Campylobacter prevalence; rather, prevalence
results reported reflect the combined effect exerted
by factors that are inextricably related to each broiler
rearing system.
Hoaretal. (2001)
Campylobacter,
Giardia,
Cryptosporidium
parvum
Beef cattle from 18
counties in CA
Factors evaluated were:
• herd size (as number of females
on the farm)
• increased weaning age
• scouring calves
• purchase of replacement calves
Campylobacter
Only herd size (number of females in herd) was
associated with increasing prevalence of
Campylobacter.
Giardia
No factors associated with an increased prevalence
of Giardia.
Crvptosporidium
Prevalence associated with length and timing of the
calving season, scouring calves, and mean herd
proportion of cows.
Huston et al. (2002)
Salmonella spp.
Conventional dairy
farms in OH
Factors evaluated with respect to their
association with fecal shedding of
Salmonellae; the most significant
factors evaluated were:
• herd size
• season
• housing
• use of straw bedding
Salmonella prevalence significantly associated with
herd size, use of free stalls for lactating and non-
lactating cows, and use of straw bedding for non-
lactating cows. No seasonal shedding pattern
observed.
August 2010
D-10
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Kabagambe et al.
(2000)
Salmonella spp.
Dairy farms,
nationwide
Factors evaluated with respect to herd-
level prevalence of Salmonella in
feces; the most significant factors
evaluated were:
• herd size
• region of the country (U.S.)
• manure disposal method
• manure management
• drinking water hygiene and
disinfection
• feeding brewers yeast products
Herd size and region have a significant impact on
Salmonella prevalence; cows from large herds (>
100 cows) had 5.8x greater odds of shedding
Salmonella than cows from smaller farms.
Cows from the south (defined as a large region
including CA, and NM) had 5.7* greater odds of
shedding Salmonella than cows from the north
(defined as a large region including Oregon and
Washington).
Kuhnert et al (2005)
STEC
Conventional and
organic dairy farms
in Switzerland
250 risk factor parameters evaluated
including
• management data (farm size,
hosing condition, etc)
• current milk production, others
specified in Roesch et al. (2004)
Note, several parameters known to be
different between these two types of
farming, ranging from feeding, therapy,
animal husbandry, and to processing
of the meat.
In general, no significant differences between the
two farm types concerning prevalence or risk for
carrying STEC or O157:H7 observed (cows tested
30 days postpartum).
Overall prevalence level based on PCR was 58%.
STEC detected in all farms and O157:H7 were
present in 25% of organic farms and 17% of
conventional farms.
STEC detected in 58% and O157:H7 were
evidenced in 4.6% of individual feces
Risk-factors mainly related to the potential of cross-
contamination of feeds and cross-infection of cows,
and age of the animals.
A reduced risk for the presence of STEC found for
older than younger cows
Increased risk for carrying STEC was associated
with elevated milk concentrations of lactose or urea,
with farms that had an Unifeed trailer (used for
mixing feed) or a paddock.
For O157, chain or lateral fixation versus bar
fixation, and final milking manually or automatically
were factors associated with an increased risk for
the presence of O157:H7
Note, previous studies produced conflicting results
regarding impact of diet on EC, some showed hay
diet resulted in reduction of EC, others showed
exact opposite results.
August 2010
D-ll
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Le Jeune et al.
(2004)
£. co//O157:H7
Beef cattle feed lots
Factors explored with respect to their
association with prevalence (%
positive samples) in feedlot pen
manure, including
cattle duration in feedlot
supply of chlorinated water for troughs
There was a general trend toward increasing £. coll
O157:H7 prevalence with duration of cattle in
feedlot; however, periods of high prevalence were
sporadic and unrelated to season or length of time
cattle were in the feedlot.
No significant difference in prevalence among cattle
provided chlorinated drinking water and those not
provided chlorinated drinking water.
Luangtongkum et al.
(2006)
Campylobacter
Conventional and
organic chicken
farms
Conventional and organic broiler and
turkey farms; antimicrobial resistance
to widely used antibiotics evluated
Campylobacter species highly prevalent in both the
conventional and organic chicken operations.
Broiler rates for Campylobacter prevalence 65%
(conventional) and 89% (organic) (significantly
different); turkey rates 83% (conventional) and 87%
(organic)
Prevalence on conventional broiler farms slightly
lower (44-80%) than organic farms (70-100%).
Prevalence on conventional turkey farms similar
(63-98%) to organic farms (6-100%)
The high prevalence of Campylobacter strains in
organically raised broilers in part seems to be
associated with the increased age of the birds at
slaughter.
Study also indicates the influence of conventional
and organic chicken production practices on
antimicrobial resistance of Campylobacter on
chicken farms.
Miller et al. (2008)
Cryptosporidium
Dairy farms in
coastal CA
Factors evaluated with respect to their
impact on density of oocysts in runoff
water and oocyst loading from various
farm locations, including:
• use of structural BMPs (vegetative
buffer strips, straw mulch)
• animal age class
Both straw mulch and vegetative buffer strips
provided significant reductions in oocysts loads.
For vegetative buffer strips, load reduction found to
be a function of the vegetative buffer strip length.
For straw mulch, reductions related to the areal
extent of mulch cover and the age class of cattle in
lots where mulch was applied.
August 2010
D-12
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Miller et al. (2007)
Giardia
duodenalis
Dairy lots and high-
use cattle areas
Factors evaluated with respect to their
association with the presence of G.
duodenalis in runoff water:
• type of dairy cattle (cow, milking or
dry cow calving (calf, 2.1-6.0
months of age; calf, 0.1-2.0
months)
• length of vegetated buffer (m)
• number of cattle in lot
• 24-hour precipitation (mm)
• cumulative annual precipitation
(mm)
• lot area
• slope
Increased concentrations and instantaneous loads
of G. duodenalis associated with:
• young calves
• absence of vegetative buffer strips
• presence areas of high cattle use
The following did not have an association with
concentrations or instantaneous loads of G.
duodenalis:
• percent slope
• area of lot
• cattle density
• 24-hour precipitation
• additional BMPs, such as straw mulching,
seeding, removal of manure via scraping, and
winter exclusion of cattle
G. duodenalis concentration in runoff increased
monotonically with precipitation up to a threshold
precipitation depth, after which concentration
decreased slowly with increasing precipitation depth.
Newell et al. (2003)
Campylobacter
spp.
NA (review article)
Review article evaluating sources of
Campylobacter m broiler chickens
Prevalence of flock positivity is dependent on flock
size and the type of production system.
Flock positivity is generally higher (up to 100%) in
organic and free-range flocks compared to
intensively reared flocks, which presumably reflects
the level of environmental exposure of birds as well
as the increased age of the birds at slaughter.
Most reviewed studies found that water sources
were a low-risk factor for positivity and that water
contamination usually follows, rather than precedes,
colonization of a flock.
August 2010
D-13
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Reinstein et al.
(2009)
£. co//O157:H7
Organically,
naturally, and
conventionally
raised cattle
Prevalence of £. co//O157 in organic-,
natural-, and conventionally-raised
cattle, comparison of antibiotic
susceptibility profiles
Prevalences of £. co// O157:H7 were 14.8% for
organically-raised cattle, 14.2% for naturally-raised
cattle, and 11.2% for conventionally raised cattle.
Study did not include a statistical comparison of the
prevalence data because of a number of differences,
particularly in diet, among the three production
systems.
Organically- and naturally-raised cattle are either
required to graze a pasture or fed a forage-based
diet.
Cattle fed a forage diet have both higher levels and
longer durations of fecal shedding of E. coli
O157:H7 than cattle fed a grain diet (Van Baale et
al. 2004).
No major difference in antibiotic susceptibility
patterns among the isolates observed.
Sato et al. (2004)
Campylobacter
spp.
Organic and
conventional dairy
herds
Prevalence and antimicrobial
susceptibilities of Campylobacter spp.
isolates from bovine feces compared
between organic and conventional
dairy herds
Prevalence of Campylobacter spp. in organic and
conventional farms was 26.7 and 29.1%, and the
prevalence was not statistically different between the
two types of farms
Campylobacter prevalence was significantly higher
in March than in September, higher in calves than in
cows, and higher in smaller farms than in large
farms.
No evidence that restriction of antimicrobial use on
dairy farms was associated with prevalence of
resistance to ciprofloxacin, gentamicin,
erythromycin, and tetracycline.
Trotz-Williams
(2008)
Cryptosporidium
parvum
Dairy farms
(Ontario, Canada)
Prevalence of Cryptosporidium parvum
in feces evaluated with respect to
associations with:
• number of calves
• number of milking cows
• perinatal management
• management of pre-weaned
calves
• calf feeding and medications
• other factors
30% of calves <1 month old shed C. parvum in the
study and 77% of farms in the study had at least one
shedding calf.
Predictors associated with increased in-farm
prevalence of shedding were the use of calf scour
prophylaxis in cows and calves and feeding of milk
replacerto young calves; predictors not associated
with C. parvum shedding were number of calves,
number of milking cows, and others.
August 2010
D-14
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Trotz-Williams et al.
(2007)
Cryptosporidium
parvum
Dairy farms
(Ontario, Canada)
Prevalence of Cryptosporidium parvum
in feces evaluated with respect to
associations with:
• calf age
• calf birth season
• calf birth in multi-cow calving pen
• calf feeding regimen
• calf and dam housing
• other factors
Calves born in summer months were more likely to
shed C. parvum than those born in summer months.
Age of calves at sampling and the time calves
remained with their dams after birth were both
associated with prevalence of fecal shedding of C.
parvum.
USDA(1997)
£ co//O157
Beef cattle feed lots
Factors evaluated with respect to their
association with £ co// O157 in feces
from:
• Length of time cattle are on feed
• Feeding barley
Refer to USDA (1995) for sampling
design details.
Pens for cattle that had been on feed <20 days were
3.4* more likely to have a positive £ co// O157
sample. Possible reasons include stress from
transportation to the feedlot or animal mixing with
the feedlot populations.
Pens receiving some portion of barley were 2.75x
more likely to have a positive £. co// O157 sample
than pens receiving no barley. Possible explanation
is the way barley is digested in cattle. Feeding
barley is regional practice. States in this study
feeding barley with positive feedlots were AZ, CA,
ID, TX, and WA.
Pens with cattle weighing >700 Ibs upon entry to the
feedlot were less likely to have a positive £ co//
O157 sample. Possible reasons include larger cattle
may handle transportation stress and a new
environment better than smaller/younger cattle.
Pens with at least 85% heifers (females) were less
likely to have positive £ co// O157 samples; authors
do not have an explanation.
Other factors not found to be associated with
positive samples:
• ionophore use
• feeding antibiotics, coccidiostats, probiotics,
urea, and other food additives
• animal density within pens
• previous health status of cattle
August 2010
D-15
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Van Overbeke et al.
(2006)
Campylobacter
spp. and
Salmonella
enterica
Conventional and
organic broiler farms
Conventional vs. organic farms in
Belgium
No significant differences could be found in
prevalence of Salmonella between organic and
conventional broilers; in contrast, Campylobacter
infections were significantly higher in organic flocks
(however, organic flocks were slaughtered at 12
weeks compared to 6 weeks for conventional flocks
and age is a known important factor).
In organic broilers, the following Salmonella
serotypes were found: Virchow, Hadar, and
Livingstone; serotypes Mbandaka and Virchow
found on the conventional farms
C. co//and C.jejuni found in both production
systems.
No statistically significant differences found between
organic and conventional meat; these results are
similar to other previous related studies.
Warnick et al. (2003)
Salmonella spp.
Dairy farms in the
Midwest and
northeast U.S.
Factors evaluated with respect to their
impact on shedding prevalence,
including:
• cattle group (lactating, cows to be
culled, cows to calf within 2
weeks, calves)
• antibiotic treatment within two
weeks
• sick within previous week
Within-herd Salmonella prevalence and serotype
prevalence varied widely over short time periods.
In general, cows nearing calving exhibited greater
Salmonella shedding than other groups, though high
variability was observed in shedding from all groups.
Wesley et al. (2000)
Campylobacter
spp.
Dairy farms
throughout the U.S.
Factors evaluated with respect to herd
prevalence and individual sample
prevalence of Campylobacter in feces,
including:
• region
• farm size
• chlorination of drinking water
Region not associated with herd- or individual
animal prevalence of Campylobacter.
Farm size not associated with herd-level prevalence
of Campylobacter shedding, but was with increased
shedding prevalence on the animal level.
Other factors associated with increased shedding
included manure management practices.
Drinking water chlorination not associated with a
decrease in shedding prevalence.
August 2010
D-16
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Wilhem et al. (2009)
Campylobacter,
E. co//,
Salmonella spp.,
Staphyloccus
aureus
Organic and
conventional dairy
farms
Review article with stated objective is
to identify, evaluate, and summarize all
primary research investigating the
prevalence of zoonotic bacteria in
organic dairy production or comparing
organic and conventional dairy
production
Bacterial outcomes reported in 17 studies.
Campylobacter, E. co//, and Salmonella spp. were
reported in 2, 7, and 4 studies, respectively.
Contradictory findings reported for differences in
bacterial outcomes between dairy production types
(organic vs. conventional); these findings may result
from geographic differences in organic regulations,
baseline prevalences, laboratory methods used, or
methods of analysis. Specifically, no significant
difference in Campvlobacter prevalence in fecal
samples was found between organic and
conventional farms.
No significant differences in herd-level prevalence of
(STEP in the U.S. Greater individual prevalence on
organic dairy farms.
No differences in STEC prevalence was found
between organic and conventional farms in
Switzerland or the Netherlands.
No significant difference was observed in the
prevalence of Salmonella spp. in fecal samples,
either at the farm or individual levels.
August 2010
D-17
-------
U.S. Environmental Protection Agency
Study
Organism(s)
Farm Type(s)
Factor(s) and Notes
Major Findings
Young et al (2009)
Campylobacter,
E. co//,
Salmonella spp.,
Staphyloccus
aureus
Organic and
conventional farms.
Chicken, swine and
beef farms.
Review article summarizing published
prevalences of zoonotic and potentially
zoonotic bacteria in organic and
conventional chicken, swine and beef
production using systematic review
and meta-analysis methodology
In 37 studies, specific bacterial and AMR outcomes
were compared between organic and conventional
chicken, swine or beef production.
The prevalence of Campylobacter was higher in
organic broiler chickens at slaughter (in 3/5 studies,
in the others no difference was noted)
Campylobacter isolates from conventional retail
chicken were more likely to be ciprofloxacin-
resistant.
Bacteria isolated from conventional animal
production exhibited a higher prevalence of
resistance to antimicrobials; however, the recovery
of some resistant strains was also identified in
organic animal production.
Limited or inconsistent research was identified in
studies examining the prevalence of zoonotic and
potentially zoonotic bacteria in other food-animal
species.
In four studies, researchers investigated Salmonella
spp. in broiler chickens on farms or at slaughter in
the U.S., Belgium, and Italy, and found very few or
no positive samples in both organic and
conventional populations.
Conflicting results were reported in six studies that
examined the prevalence of Salmonella spp. in
swine on farms and at slaughter in the U.S.,
Denmark, and Germany.
Studies conducted in the U.S. showed higher
Salmonella prevalence in organic farms;
international studies showed contrary results.
August 2010
D-18
------- |