United States
Environmental Protection
Agency
Office of Research and
Development
Washington DC 20460
EPA/630/R-96/003
September 1996
Report on the Peer
Review Workshop on
Revisions to the Exposure
Factors Handbook
RISK ASSESSMENT FORUM
-------
EPA/630/R-96/003
September 1996
REPORT ON THE PEER REVIEW WORKSHOP ON
REVISIONS TO THE EXPOSURE FACTORS HANDBOOK
Prepared by:
Eastern Research Group, Inc.
110 Hartwell Avenue
Lexington, MA 02173
EPA Contract No. 68-C1-0030
September 1995
Risk Assessment Forum
U.S. Environmental Protection Agency
Washington, DC
Printed on Recycled Paper
-------
NOTICE
Mention of trade names or commercial products does not constitute endorsement or
recommendation for use. Statements are the individual views of each workshop participant; none
of the statements in this report represents analyses or positions of the Risk Assessment Forum or
the U.S. Environmental Protection Agency (EPA).
This report was prepared by Eastern Research Group, Inc. (ERG), an EPA contractor, as
a general record of discussions during the Peer Review Workshop on Revisions to the Exposure
Factor Handbook. As requested by EPA, this report captures the main points and highlights of
discussions held during plenary sessions and includes brief summaries of the work group sessions.
The report is not a complete record of all details discussed, nor does it embellish, interpret, or
enlarge upon matters that were incomplete or unclear. In particular, each of the four work group
summaries was prepared at the workshop by individual work group chairs based on the work group
discussions held during the workshop. Thus, there may be slight differences between the four
groups' recommendations. ERG did not attempt to harmonize all the recommendations.
u
-------
CONTENTS
Page
Foreword iv
SECTION ONE—INTRODUCTION 1-1
Background 1-1
Peer Review Workshop 1-2
SECTION TWO—CHAIRPERSON'S SUMMARY OF THE WORKSHOP 2-1
P. Barry Ryan
SECTION THREE—WORK GROUP SUMMARIES . .- 3-1
Food and Beverage Consumption 3-1
Barbara Petersen
Nondietary and Dermal Exposure Factors 3-15
John Kissel
Human Activity Patterns 3-21
Steve Colome
Housing Characteristics and Indoor Environments 3-56
P J. (Bert) Hakkinen
SECTION FOUR—OVERVIEW 4-1
Reviewers' Preliminary Comments 4-1
Summary of Workshop Deliberations 4-8
Observers' Comments 4-15
APPENDIX A. REVIEWER LIST A-l
APPENDIX B. PREMEETING COMMENTS , B-l
APPENDIX C. WORKSHOP AGENDA C-l
APPENDIX D. WORK GROUP ASSIGNMENTS D-l
APPENDIX E. FINAL OBSERVER LIST E-l
m
-------
FOREWORD
This report includes information and materials from a peer review workshop organized by
the U.S. Environmental Protection Agency's (EPA's) Risk Assessment Forum (RAF) and the
National Center for Environmental Assessment. The meeting was held in Washington, DC, at the
Doubletree Hotel Park Terrace on July 25-26, 1995. The subject of the peer review was the
document entitled Exposure Factors Handbook (External Review Draft, EPA/600/P-95/002A, June
1995). A copy of this report was made available to the public through EPA's Office of Research
and Development publications office, CERI, U.S. EPA, 26 West Martin Luther King Drive,
Cincinnati, Ohio 45268 (703 487-4650). The expert technical reviewers were convened to
independently comment on the draft document and make recommendations that will enhance the
final Handbook.
Notice of the workshop was published in the Federal Register on July 13,1995 (60 FR 36142).
The notice invited members of the public to attend the workshop as observers and provided logistical
information to enable observers to preregjster. About 40 observers attended the workshop, including
representatives from federal government, industry, environmental andhealth organizations, the press,
trade organizations, and consulting firms.
A balanced group of expert peer reviewers were selected from academia, industry, and
government. Selected reviewers provided scientific and technical expertise in the following
disciplines: water ingestion, food ingestion, inhalation rates, soil ingestion, fish consumption, dermal
contact, human activity patterns, residence characteristics, and survey statistics.
In outlining the scope of the peer review, EPA emphasized that peer involvement is a key
component of the process of developing a useful Handbook. EPA explained that the intended
audience for the Handbook includes members of the risk assessment community within and outside
of the Agency involved in developing exposure assessments, scientists involved in studies for which
exposure data are collected, and scientists conducting research on exposure assessment. EPA
explained further that the comments and recommendations of outside experts will greatly benefit the
development of the final Handbook. EPA asked the expert reviewers to concentrate their review
on determining whether the data presented in the Handbook will be useful and support both point
estimates and probabilistic analyses of exposure. EPA will use the expert reviewers' comments and
recommendations drawn from this workshop in considering revisions to the draft Handbook.
The workshop report is organized as follows. The report opens with a brief introduction
concerning the purpose of the workshop and the background of the Handbook (section 1). This is
followed by the chairperson's summary (section 2) and then the four work group chairs' summaries
(section 3). The last section of the report provides highlights of peer reviewers' preliminary
comments, a summary of meeting deliberations, and observers' comments (section 4). Appendices
to the workshop report include a list of reviewers, the reviewer's premeeting comments, the agenda,
reviewer work group assignments, and a list of observers.
William Wood, Ph.D.
Executive Director
Risk Assessment Forum
iv
-------
SECTION ONE
INTRODUCTION
This report highlights issues and conclusions from a workshop convened to gather
information from expert reviewers on the U.S. Environmental Protection Agency's (EPA's) Exposure
Factors Handbook (the Handbook) (External Review Draft, EPA/600/P-95/002A) published in June
1995. This information will be used by EPA in further developing the Handbook. The workshop
was sponsored by the EPA's Risk Assessment Forum and the National Center for Environmental
Assessment (NCEA).
BACKGROUND
Seven years ago, in response to requests for guidance and information on how to select
values for exposure factors, the Exposure Assessment Group of EPA's then Office of Health and
Environmental Assessment issued the Exposure Factors Handbook. The Handbook addresses factors
frequently relied on in exposure assessments and provides a common set of statistically based values
(default values) suggested for use by EPA program and regional offices. The Handbook was
intended to encourage consistency in exposure assessments, while allowing risk assessors the
flexibility to tailor assessment approaches to specific situations.
The 1989 Exposure Factors Handbook is divided into two parts. Part I provides equations
and data on factors used in assessing exposure by ingestion, inhalation, and dermal routes. Part I
also provides values for other factors used for exposure calculations such as lifetime, body weight,
and activity patterns. Part II presents standard exposure scenarios and a discussion concerning
analysis of uncertainties. Standard exposure scenarios include, for instance, ingestion of
recreationally caught fish/shellfish from large water bodies and inhalation of vapors outside
residences. The scenarios provide basic equations for calculating exposures as well as default values
that can be used when site-specific data on factors are not unavailable. Both qualitative and
quantitative methods for assessing uncertainties associated with exposure assessment are presented.
1-1
-------
Although originally developed as a support document in connection with EPA's 1986
Guidelines for Estimating Exposure (FederalRegister 51:34042-34054) and 1988 Proposed Guidelines
for Exposure-Related Measurements (Federal Register 53:48830-48853), the Handbook quickly
became an extremely popular tool in conducting exposure assessments. Then in 1992, two events
prompted efforts to revise the Exposure Factors Handbook: (1) EPA's Risk Assessment Council
issued a memorandum on risk characterization that emphasized moving away from single-value risk
assessments (i.e., in favor of assessments that consider both central tendency and high-end
exposures); and (2) EPA published the revised Guidelines for Exposure Assessment (FederalRegister
57:22888-22938). Moreover, risk assessors were using and seeking updated exposure factors.
As a first step toward revision, EPA initiated a survey of Agency exposure assessors to
develop recommendations on what factors should be included in the updated Handbook. EPA's
two-day peer involvement workshop held in July 1993 represented another step in planning the
Handbook's revision. Subsequently, based on the results of the workshop and new data obtained
for various factors, the Handbook was revised. For example, because experts at the meeting held
diverse opinions on whether to include scenarios and default parameters, neither types of
information were provided in the draft revision of the Handbook. The draft Handbook (the subject
of this peer review workshop) presents a significant amount of new material over the original 1989
Handbook.
PEER REVIEW WORKSHOP
To involve outside scientific and technical experts in development of the Handbook, EPA's
Risk Assessment Forum and NCEA sponsored a two-day workshop, which was held on July 25-26,
1995, at the Doubletree Hotel Park Terrace, Washington, DC. The meeting gathered 25 experts (see
Appendix A for a list of expert reviewers) with the objective of ensuring that the Handbook is of
sufficient scientific quality to distribute as an EPA publication.
Prior to the workshop, EPA provided each reviewer with a copy of the draft Exposure Factors
Handbook. EPA asked workshop participants to review this material before the meeting and to
prepare premeeting comments with the following issues in mind:
1-2
-------
• EPA sought expert opinion on specific questions: Are the data presented in a way
that is useful to exposure assessors? For example, the data presented in the home
produced section have been broken out in various ways (e.g., by regions,
urbanization, race, age groups). Is this the best way to present the data? Also, are
the data presented in a way that will support both joint estimate and Monte Carlo
assessments?
• The studies included have been grouped into key studies and other relevant studies
based on the Agency's judgment about the adequacy of the data and their
applicability to the exposure factors being evaluated. EPA sought reviewer
comments on whether these groupings have been made appropriately.
• Recommendations are presented at the end of each section. These are based on the
Agency's interpretation of the key studies. EPA sought opinions on whether this is
the proper interpretation of the data and whether the limitations/uncertainties have
been appropriately emphasized/described.
• EPA would like to .develop a new chapter (or sections at the end of each chapter)
that highlights data gaps and future research needs. EPA sought suggestions for this
material.
Appendix B contains the reviewer's premeeting comments.
To begin the workshop, William Wood, Ph.D., Executive Director of the Risk Assessment
Forum, and Michael Callahan, Director of the National Center for Environmental Assessment's
Washington office, explained the need to revise the 1989 Exposure Factors Handbook and the process
for producing a final document. They emphasized that the revision will be based on the results of
the peer review workshop, discussions with EPA program offices, and a review by EPA's Science
Advisory Board. Next, they reviewed the charge to reviewers (i.e., the four issues presented above)
and emphasized the need for reviewers to address whether the data presented in the Handbook will
be useful in supporting exposure analyses (e.g., site-specific and national exposure assessments).
P. Barry Ryan, Ph.D., a professor at the Rollins School of Public Health at Emory University,
served as the chairperson of the workshop. In his introductory remarks, Dr. Ryan reviewed the
agenda for the workshop (see Appendix C), providing an explanation of the format for work group
sessions. Reviewers were divided into four work groups according to the following topic areas:
1-3
-------
" food and beverage consumption;
• nondietary and dermal exposure factors;
• human activity patterns; and
• housing characteristics and indoor environments.
(See Appendix D for reviewer work group assignments.) To help focus the groups' efforts on
addressing each question in the charge, Dr. Ryan reviewed the purpose and goals of the workshop.
He reminded reviewers to focus on identifying and elucidating issues relevant to the draft Handbook,
rather than attempting to reach a consensus on issues.
Dr. Ryan explained that whereas the 1989 Handbook provided guidance on exposure
scenarios, the draft Handbook provides, to the degree possible, guidance on the distribution of
exposure factors. Further, he noted that the broader purpose of the document is to present a
compilation of scientific data that will facilitate consistency among exposure assessment presentations
by providing recommended data for use in preliminary assessments.
1-4
-------
SECTION TWO
CHAIRPERSON'S SUMMARY OF THE WORKSHOP
P. Barry Ryan
Rollins School of Public Health
Emory University
Atlanta, Georgia
This section of the report paraphrases general comments provided by the four work group
chairs in their oral presentations on the draft Exposure Factors Handbook. Some of these comments
reiterate or elaborate on comments provided by expert reviewers in their premeeting submissions
(see section 4 for an overview of premeeting comments and Appendix B for the premeeting
comments themselves). Written summaries of work group discussions provided by the four chairs
are presented in the next section (section 3).
This section also summarizes comments from general discussions at the workshop concerning
the adequacy of information on uncertainty analysis provided in the draft Handbook (i.e., in chapter
8) and highlights areas of general agreement among peer reviewers.
WORK GROUP ON FOOD AND BEVERAGE CONSUMPTION
Although this work group's presentation focused on exposure via fish consumption, many of
the comments raised also are applicable to other areas of dietary exposure and exposure in general.
Overall, this work group agreed with the basic approach taken for revising the document, referring
to information included in the draft as sound.
Several comments raised by this work group concern the studies selected for the Handbook.
The panel proposed including more recent studies that provide guidance on a site-specific basis. In
particular, panel members suggested including new data based on geographic region or on the type
2-1
-------
of water from which fish are caught. This is an overarching theme, given that it applies by analogy
to media for other exposures.
Other comments regarded the use of certain data sets that are not necessarily population
based in place of those that are and the use of other available studies. Both of these issues arose
in other work groups as well.
This work group ended its presentation by posing two questions for consideration: Are there
differences in dietary exposures (and other exposures) that can be attributed to ethnic differences?
What is the relationship between short-term measurements of dietary intake and long-term exposures
through the diet?
The group could not cite studies in these areas and thus characterized the questions as
indicative of data gaps.
WORK GROUP ON NONDIETARY AND DERMAL EXPOSURE FACTORS
This work group echoed several comments raised by other groups. In particular, group
members emphasized the need for explaining short-term exposure measurements and long-term
exposures. They also advocated for the inclusion of longer term studies.
Additionally, the panel urged reorganization of the literature reviews by type of study.
Members suggested, for example, separating staged and unstaged studies. In general, studies should
be organized such that surveys or population-based studies would not be viewed as equivalent to
special studies on specific groups.
WORK GROUP ON HUMAN ACTIVITY PATTERNS
This work group's presentation focused primarily on general considerations rather than the
specific topic of human activity patterns. One of the panel's suggestions was to present data in
2-2
-------
chapter 5 in graphical form where possible. The group noted that pie charts, bar graphs, and similar
visual representations can make the presentation of data more immediate.
Panel members suggested beginning each section with an overview and including an index
at the back of the Handbook. Each overview could include a summary of the information in the
section, an explanation of the reason for including the information, and an outline showing the
organization of the information provided. This would make the document considerably more user-
friendly.
The work group also suggested that EPA conduct a more thorough literature search to
identify more appropriate studies. This sentiment was echoed by all the work groups. In the panel's
opinion, the studies presented discuss survey-related issues in an unsophisticated and inadequate
fashion. Moreover, panel members found some of the data presented in the time/activity surveys
to be obsolete.
WORK GROUP ON HOUSING CHARACTERISTICS AND INDOOR ENVIRONMENTS
Members of this work group concentrated their efforts on developing an outline for their
proposed reorganization of the chapter. In their view, the chapter has all the appropriate
information, but the pieces do not fit together well.
In regard to their charge, panel members advocated positioning reference residence exposure
in the larger exposure context. Additionally, they suggested positioning the entire document within
a single conceptual framework. The group developed such a framework for exposures experienced
in residences and suggested that a single framework could be developed similarly for all aspects of
exposure. In the panel's opinion, such a framework would provide a firm foundation for all
discussions of exposure in the Handbook.
2-3
-------
UNCERTAINTY ANALYSIS
Although none of the groups specifically addressed the treatment of uncertainty analysis in
the Handbook (chapter 8), the workshop chairperson solicited comments on this topic. All peer
reviewers agreed that the Handbook provides a useful introduction to uncertainty analysis, but that
more information on this topic should be included. Reviewers suggested that if information in the
chapter cannot be expanded upon, then the discussion on uncertainty analysis should be incorporated
in the introductory chapter (chapter 1).
CONCLUSIONS
Based on comments made during workshop discussions and on work group presentations,
peer reviewers generally agreed on the following:
» The revised Handbook will serve an important need.
• The Handbook should provide some method of evaluating the quality of the studies
included.
• Although certain studies in the Handbook are "key" and "relevant," some studies are
inappropriate or dated. Moreover, studies on specific populations should be
eliminated or included with a strong caution about their use.
• Presentation of data is important, but could be enhanced with a graphical format.
• Available literature should be more thoroughly reviewed (studies seemed to have
been selected without regard for their specific contribution to the exposure
assessment field).
2-4
-------
SECTION THREE
WORK GROUP SUMMARIES
Food and Beverage Consumption Work Group
Work Group Chain Barbara Petersen, Technical Assessment Systems, Inc.
Work Group Members: J. Mark Fly, University of Tennessee
Patricia Guenther, U.S. Department of Agriculture
Mary Kama, U.S. Department of Agriculture
Paul Price, ChemRisk
John Risher, The Agency for Toxic Substances and Disease
Registry
Frances Vecchio, U.S. Department of Agriculture
INTENDED USES AND AUDIENCE
Information provided in the Handbook should be appropriate to meet the needs of the
intended users of the data. Thus, the work group-identified potential users for the chapter on food
and beverage consumption:
• groups and individuals evaluating food additives/packaging (i.e., using the data as a
shortcut to look at potential exposure through foods, though not for regulatory
purposes);
• groups and individuals assessing indirect risk (e.g., from air pollution, sludge,
material leaching from a large area of environmental media, multiple exposures over
wide area of food crops);
• researchers; and
• state, federal, and local health departments.
The types of information that would be needed for such uses of the data include:
• data on the entire U.S. population and for subgroups;
3-1
-------
• point estimates as well as distributions (i.e., high-end 90th or 95th percentile); and
• data with a variable level of precision depending on the application (user might be
willing to make worse case assumptions instead of more accurate estimates).
Because some users of this information may have limited expertise, the chapter should
provide more guidance on how to use the data. This guidance should be provided not as default
values but as "reference" values, accompanied by explanations of appropriate use. In particular,
more guidance is needed oa how to use the fish consumption data. In the opinion of work group
members, the addition of scenarios to the chapter would not be particularly useful to the reader.
MISSING DATA
In general, the work group found the data to be outdated and often incomplete. Thus,
members recommended that the following data be added:
• data on all categories: USDA CSFH1989-91 (available on data tapes;
• fish consumption data: Michigan Survey of Fish Consumption (1992) (currently
available); and
• meat consumption data: USDA Ag. Econ. Ranching Survey; Home Slaughtering, of
Sheep and Beef Cattle (conducted on a national basis).
At the workshop, representatives of the U.S. Department of Agriculture (USDA) indicated
that some data reports are available for the 1989-91 surveys and that these could be provided to
EPA
GUIDANCE ON MULTIPLE ANALYSES OF THE SAME DATA
The work group recommended the following on this topic:
• Fully inform the user that the Handbook includes multiple analyses of the same data.
3-2
-------
Provide complete documentation for tables, including, for instance, the source of the
data and an indication of the form of the food (e.g., dry-weight basis, cooked,
uncooked). (The work group strongly recommends this because panels often had
difficulty matching text with accompanying tables.)
Provide guidance on which estimates to use for which purpose, and note limitations.
Include a column in each table that provides the source of the data and summarizes
the data from each study.
Include a chart that helps users identify which tables provide original data and which
provide data that are the results of reanalyses. (The work group developed a flow
chart for its own use in identifying the source of data used in the analyses reported
in the food consumption chapter.)
Ensure that the data presented in the table of water intakes for different activity
levels does not conflict with whatever data are provided (if any) in other parts of the
Handbook. (This group did not evaluate the water intake estimates, but did raise a
concern about the consistency of data between chapters.)
PRECISION
The document should not imply precision beyond what the study authors produced. The
group recognized, however, that EPA should not be asked to reassess the appropriate level of
original data. Therefore, after much discussion, the group recommended that EPA not report more
significant digits than the source data.
Moreover, an effort should be made to be consistent throughout the handbook. In
particular, consistency may be a problem when per capita estimates are derived for infrequently
consumed foods such as fish.
Additionally, the rules of rounding should apply to interpolated percentiles.
NUMBER OF SUBJECTS PER SUBGROUP
The group concluded that without a cell size of at least 30 observations, the cell should be
3-3
-------
left empty. Although the group did not recommend a specific number of observations, panelists felt
that when distributions are to be generated the cell size should be more than 30.
SUBGROUPS
•
Does the Handbook present data for subgroups likely to be used by exposure assessors?
» - For age and sex subgroups, the work group found the data to be adequate.
• For fish consumption, the work group recommended adding data on the following:
ethnic background subgroups;
— subgroups reiving on fishing for economic subsistence; .
— information by water bodies (river/stretch of the coast), marine vs.
freshwater, nature of the fisheries;
— Michigan study;
— creel survey data; and
— list of studies in Paul Price's comments.
• For meat consumption, the group recommended adding data on:
— home-produced meat (USDA ranching survey).
• For game, deer, wild foul intake, the group recommended adding:
— data available from USDA surveys (but few users);
— data on how frequently these foods are consumed; and
— data on total amounts consumed.
• For breast milk, the group noted the following:
— some of the results of the studies included in the Handbook conflict. Thus,
a comment should be added regarding these conflicts and guiding the user
in the appropriate selection of data;
— the Handbook should convey the level of uncertainty in these estimates, in
both the text and in the degree of precision in the estimates; and
3-4
-------
— estimates of intake are needed for specific age subgroupsO to 6 months, 6
to 12 months, greater than a year.
In general: The group attempted to provide guidance for instances where
observations are too few for a specific cell; thus, if too few people are in a cell:
— combine foods (this would not necessarily be easy, and may depend on
specific substance);
— combine age/sex groups;
— include information about how to obtain original data (it may be better for
researchers to go back to the original data).
FOOD DISAPPEARANCE DATA
The food disappearance data are of extremely limited utility and should only be included
when other sources of information are not available. Given the usefulness of data already included,
the group recommended dropping the food disappearance data entirely. (See also Attachment A.)
SURVEY METHODOLOGY
The work group recommended the following on this topic:
adding a checklist for every study, including reference period, sample size, and
methodology (e.g., year, 3 days);
if appropriate data are available, adding food frequency data for infrequently
consumed foods; and
providing guidance on the use of total population data; total population data may be
too general for specific analyses (site specific, state specific), but useful for screening
and preliminary analyses.
3-5
-------
PRESENTATION
The work group recommended the following on this topic:
» add a discussion about how to obtain upper percentile estimates when the
contaminant is present in more than one food (adding upper percentiles for
individual foods is not the correct approach; user needs to be advised to go back to
raw data);
• add a discussion on the use of per capita estimates and when per user estimates are
more appropriate; and
• provide estimates of precision (standard deviations/standard errors/confidence
intervals).
DERIVATION OF ESTIMATES FOR HOMEGROWN FOODS
On this topic the work group recommended adding a discussion on the limitations of
estimates given as well as the methods/assumptions used to derive estimates.
Additionally, the group expressed interest in reviewing the analysis currently being conducted
by Paul White. Some additional information from USDA is attached (see Attachments B and C).
3-6
-------
ATTACHMENT A
3-7
-------
I!»«<•• I'.
Food Consumption,
Prices, and
Expenditures,
1968-89
Ju.filh .Jon<>v, Putn.un
I. MIS-I Ali-JiuiK..-
3-8
-------
LJfae tuuy ti»« auto, UM data an mow mufti 1 M fndicminn Of tnudf QVff tint &?3 at ffifissarea^att of
absolute leveli. In other words, taia seriesi provide* aa indication of whether or not America!*, oa
avenge, an containing more or lew of various food* over tins*, it i» not a direa fflentire of aejstl
consumption nor at the quantity ingested. The disappearance data tor feed hswa proved aceertta enough
to permit Measurement! of tiw avenge level of food consurapdott ia the country M a *&&&, to show
year-to-year change! in consumption of the major iboda, to permit calculation of th* appn»te£s auteiest
content of the food supply, to establish long-term trends, and to perait statistical analyses of item of
prioea and tocomea oa conaumpttea of the principal foods.
The toed supply data series Is the oaly data. $et that it oontiifeat; that is, iuppiy uid total me
baltnce. It meaiiuea ntflizttion of bttie cpmaoditiei without gettog involved with Uteattfyifif all end uia
product! and the problem* of detonpoiiBf compound food* bade to oomnsdit? iqpedisata. It itteaitirai
food lupptie* for eaniumptiOA through in outJets, it-hone «ad awsy Iron bom It ia a loa^ eoot&noci
serisi, pubUihed nut in 19*1 aad extended tadc to 1909 tor meat eomBodldec, It te the oolydtta set
avmJUble Sat determining long-term trend! la mppjy and flonrufaption by major toed group*,
ooven the complete spectrum of primary foodirulft. Hcaes, it can *e wed to mecrat*
icterrditiooihip! between fbodi aad fior meawriflg total flood mpply and apparent ue*. It fc ptrtiailarly
usafal for ettimadag complete demand lyttaau that meaavre price and toesmz «itstldtiaa of demawi ia a
coiuitteat way.
Usually the Ibod supply ia t residual which make* the suppty-utttlation coanodtty tabls baJaaos. The
disappeartftce netbixS of calculation relegata to the tbod supply aH retidual UHB toi wakii data ate cot
available, such u mlKeUancou! nonfood vaet, stock change* at retail and coaauner levebi and tttopBng
and measurement errors accumulated In ib» cadnwduo wf «h»r eantpefieatt af the balsa** sheet For
example, an increasing proportion of the total chicken supply (eap©siaHy backs, aeda, aad gibtea) joei
Into pet fboda. But since such use has yet to be oifldaHy eittaated or entered is a
component of the supply-Htfllzation balance aneet, it I* inctaded ia Hood diitppcMmfie. Thus, thto report
probably oventatei chicken coonunptiop.. In coatraM, th« lack of rdiabto Mtfaaates of gan« naa nippttei
nteana that fith coosumption ia likely uptftfrmtfttli
Pood disappearance is often used aa a proay to estimate human counnspdoa. Uied in thM oosszr, the
food supply usually pfcwtta u upper bound on the amount of flood avaflsble Sot coassnptioB, Food
disappearance estimate! can overitate actual consumptioa beauw they indode; ipofis*! aad wtte
accumulated through the marketing lyitem and in the &COM. In general, 1bod dUtppoaa^ aatt liive
mori appropriately M iadleaton of trend! ia coasumptioii over tie* thtn u meciareasaa of absolute
levels of food eaten. Thii ii the caae so long ai changes in tood prodnai^a and mftriwtfej pnedei or
conaumer behavior over time do not alter the relative disparity besweia mod dlsappearanea and feed
actually eaten.
The food disappearance seriea may no longer be a reliable indicator of chmfi over tint* fat tegsstkm of
food ats and oils. While tood disappearance airly acevfaf^jr refleea aemdj ia (ata and ofla sow lor
human fbod, it probably does not accwatery mcaiure tresdJ in food eatea bcatsas the umie pontoa of
tood disappearance) tor ats tad oOt hat iacreaaed dariai the past two 4ece£ss «fth the growth ia awa^-
(rom-aome eating placet, eajMtiaty ott-feod plaeei. Fooj!-rvice ata&Mg?unemti tbit deepofiy ftxxB caa
gcaei*w*igiU§c«rttasMun» of wwte grease, rete^ A recast study by SRI,
Internationai indicate! that the qcantiry of uted flrying at diapeei of by reatauraatt and proowd by
renderen for use in animal fiMdf> pet foods* industrial operatioai, aad tor capon now tma^aay amonati to
about 6 pound* per capita, or nearly 10 percent of the 1989 dtuppetraiiei of food ati tM oili,
Food supply data are aggregate of bod 'obtained from aHMiirm Ralafl-^nliti e
-------
-------
ATTACHMENT B
3-11
-------
Untttd Stitw
Dlpwrnent of
Agriculture
Agriculture!
RtMirch
Strvlcf
Bolttvillf Ar«t
Btlttvitlt Agricultural
Rtswrch C*nt»r
Beltsville, Maryland
.20705
The Continuing Survey of Food Intakes by
Individuals and the Diet and Health
Knowledge Survey, 1989-91
BACKGROUND:
METHODS:
DATA TAPES:
The Continuing Survey of Food Intakes by Individuals (CSFII), conducted as three separate l-ycar
surveys in 1989, 1990, and 1991. was designed to measure what Americans eat and drink.
Information from the surveys is used to develop nutrition education programs, to assess dietary
changes associated with participation in food programs, to develop food fortification and
enrichment policies, to monitor the safety of the food supply, and to assess demand for
agricultural products and marketing facilities.
The Diet and Health Knowledge Survey (DHKS), conducted as a telephone follow-up to the
CSFII, is designed to improve our understanding of factors that affect food choices and 10 obtain
Information on people's knowledge and attitudes about the Dietary Guidelines for Americans.
Together, the CSFII and the DHKS provide the first opportunity on a national scale to link an
individual's knowledge and attitudes (from the DHKS) to his or her dietary behavior as indicated
by food intake infoonation collected from the same Individual in the CSFII.
Individuals who took part in the CSFII were asked to provide 3 consecutive days of dietary data.
The first day's data were collected in a personal in-home interview using a 1-day dietary recall,
The second and third days' data wens collected using a self-administered 2
-------
ATTACHMENT C
3-13
-------
United Stitt* Agricultural Bflltsvillft Arm Bultwille, Miryltnd
Dcpartmint of RoMarch B»ltivill« Human Nutrition 20706
Agriculture Strvlc* R«M«rch Cantflr
ESTIMATION OF USUAL INTAKS DISTRIBUTIONS
ARS is sponsoring cooperative research with statisticians at Iowa
State University to develop statistically defensible methods for
estimating the distributions of usual food and nutrient intakes
for populations and subpopulations. These distributions are
required to determine the proportions of the population who are
at risk for inadequate intake of essential nutrients or for
excessive intake of undesirable dietary constituents, such as
pesticide residues. This information is needed by regulatory and
policy decision makers- in both the nutrition and food safety
arenas.
Our approach la based on the assumption that individuals can more
accurately recall the types and amounts of foods they ate
yesterday than they can recall intake over any longer period of
time. When at least two days of dietary information are
available for individuals in the sample, it is possible to
develop a statistical method for estimation of long-term average
intake by removing the within-person variation in intake, rather
than by having the individuals come up with an estimate for
their long-term intake.
Many statistical procedures are based the assumptions that the
data under investigation are normally and identically distributed
and come from a simple random sample. Dietary survey data
typically do not meet chase assumptions/ and the method developed
at Jowa State does not require them. Also, nuisance effects
caused by .seasonal!ty, day of week effects, and sequence of
survey day can be removed from the data with this method.
Software implementing the Iowa State method for estimating
nutrient intake distributions is in the beta-testing stage.
Beta-testers include researchers at other Federal agencies and
universities.
Methods for estimating distributions of usual food intakes are
under development. This problem is more difficult because of the
high fraction of zero intakes in 1-day data. Research plans also
include the development of software for implementing the usual
food intake distribution methods.
April 1995
3-14
-------
Nondietary and Dermal Exposure Factors Work Group
Work Group Chair: John Kissel, University of Washington
Work Group Members: Dennis Druck, U.S. Army
Larry Gephart, Exxon Biomedical Sciences
Peter Robinson, Procter & Gamble Company
Brad Shurdut, DowElanco
INTRODUCTION
The Nondietary and Dermal Exposure Factors Work Group was given primary responsibility
for reviewing draft Handbook sections 2.2 (Water Ingestion) and 2.8 (Soil Ingestion) and chapter
4 (Dermal Exposure), and the work group shared responsibility for chapter 8 (Analysis of
Uncertainties).
WATER INGESTION (Handbook Section 2.2)
The revised Handbook cites eight studies of tap water ingestion. Good agreement among
the studies is apparent and the prior recommended mean of 1.4 liters tap water/day is a reasonable
interpretation of those studies. This finding is tempered somewhat by the fact that four of these
studies (Pennington, 1983; U.S. EPA, 1984; Ershow and Cantor, 1987; Roseberry and Burmaster,
1992) are analyses of the same data, the 1977-78 U.S. Department of Agriculture (USDA) Food
Consumption Survey (a fact that should be made more explicit in the commentary). Nevertheless,
similar results are produced by the other studies. In addition, the number of individuals in these
data sets is large compared to data sets available for many other exposure factors. The data are
sufficient to produce a probability density function (PDF) for tap water ingestion; the work group
recommends graphical presentation of such a result. Before a PDF is adopted or prepared, however,
more recent U.S. Food and Drug Administration (FDA) figures should be examined to confirm
similarities to the 1977-78 data.
3-15
-------
Available data, regarding regional variability suggest, that differences; are; not important.
Similarly, differences associated with pregnancy are small.
Some questions do remain regarding peak consumption rates associated with especially
strenuous activity. These rates can be very high, although how long they are; sustained is not clear:
Loss of volatiles before consumption due to heating is possible. For this reason,
distinguishing between tap,water that is heated before consumption and total tap water is: desirable.
Incidental ihgestion while swimming remains essentially unqualified.
SOIL INGESTION (Handbook Section 2.8)
General
The literature review provided in the Handbook stops at 1991. Because this topic is the
subject of much ongoing speculation in the literature, references through 1995 must be included and
reviewed (see premeeting comments of I. Kissel for list). Unfortunately, the recent literature on soil
presents increasingly complex analyses that have the effect of adding to rather than reducing doubts
about the adequacy of the available data.
Studies that represent hypotheses only and not actual data may be cited for completeness,
but should not be represented in a way that suggests that they have the same weight as empirical
studies.
Children
Recent improvements in the apparent consistency in results obtained with different tracers
reflect a change in the assumed ihgestion-to-excretion lag period from 12 to 23 hours. The validity
-------
using a constant lag period of any length has not been adequately demonstrated. Moreover, the
fundamental assumption that these tracers are not bioavailable has not been justified.
The central tendency (mean) of the ingestion rate of children based on six tracers (Calabrese
et al., 1995) is (to one significant figure) 100 mg/day. This value can be conditionally accepted as
a point estimate but requires further validation.
Recently Calabrese and Stanek fit their (child) data to a log normal distribution and
produced estimates of annual average soil ingestion rates that appear notably high relative to data
from all but one subject. The extrapolation of short-term (4-day) studies to generate a distribution
of annual ingestion rates is questionable. The work group is unwilling to accept the resulting
distribution at face value.
In qualitative terms, the actual distribution of soil ingestion in children is likely to be quite
skewed, with many persons at the low end and a few at the high end. Members of the work group,
however, have little confidence in current quantitative knowledge about the shape of the distribution
The summary table on p. 2-410 of the Handbook should clearly distinguish the Davis et al.
and Calabrese et al. studies from the other (nonbalance) studies. (It also should be given a table
number and proofread.) Both old and new interpretations by Calabrese should be included.
Additionally, explicit calculation of averages should be removed since methodologies were riot
equivalent.
Given that this is a particularly important pathway that often drives risk assessments, the
current understanding of the available data is especially unsatisfactory. Multiple steps should be
taken to alleviate this problem. These include:
• conduct independent reevaluation of the Davis et al. and Calabrese et al. data sets
with respect to the signal vs. noise question;
• fund longer term studies that will provide a better evaluation of the fluctuation of
excretion of relevant tracer compounds;
3-17
-------
• identify data sets in which both body burdens and environmental (soil and dust)
levels of tracers are known for purposes of dose reconstruction; and
• investigate the status of research by Calabrese, Bornschein, and others at Helena,
Montana, that addresses soil ingestion.
Ultimately estimates of soil ingestion should correspond to observed exposures. A recent
attempt at dose reconstruction based on arsenic exposure produced a median soil ingestion rate of
85 mg/day (Lee and Kissel, in press, Env. Geochem. Health). This value reflects assumptions that
dermal absorption and inhalation exposures were negligible. Such a result requires corroboration
by additional reconstructions.
Pica (Geophagia)
Current data are grossly inadequate. This is true for both the prevalence in the population
of geophagja (which is of greater interest than the more generic pica) and for estimates of the
related soil ingestion rate. Currently n=l for this condition. Lasztity et al. (J. Anal. Atom.
Spectrom. 4:737-742, 1989) should be checked as a possible second case (the work group has not
reviewed this reference). ,
Adults
The current adult soil ingestion estimate is based on quite limited data (n = 6) from which
it is not possible to justify generation of a PDF. The central tendency of the four tracers (i.e.,
aluminum, silicon, yttrium, and zirconium) designated as most reliable in Calabrese et al. (1990) is
28 mg/day (median of medians) to 39 mg/day (mean of means). Calabrese has disavowed his former
recommended tracers for children, but apparently has not revisited the adult tracers. Thus, the work
group was uncertain about how to interpret the adult data. A value of 50 mg/day for adult ingestion
is conditionally acceptable as a point estimate, but requires validation.
3-18
-------
DERMAL EXPOSURE (Handbook Chapter 4)
The emphasis in this chapter is on dermal contact with contaminated soil or water.
Acknowledgment should be made that dermal exposure also can occur as a result of contact with
surfaces such as floors, countertops, or carpets in the absence of soil or water phases.
Additional data are required concerning skin area actually exposed versus surface area of the
body. Estimation techniques should be expanded to describe subareas of body surface associated
with consumer product use. This discussion should be tied to chapter 6 (Consumer Products).
Because this also has a behavioral component, a cross-reference in section 5.3 should be included
(Activity Patterns).
Soil
Explicit mathematical formulations for dermal dose should be removed; however, relevant
factors should be enumerated. Potential problems also should be cited with the use of a percent
absorption fraction if loadings in the exposure scenario do not match loadings in the studies from
which absorption efficiency is taken.
Soil adherence literature should be reorganized to reflect the type of study (staged vs.
unstaged activity, direct vs. indirect measure of soil loading). Also, studies by Charney et al. (1980),
Duggan et al. (1985), Gallacher et al. (1984), and Sheppard andEvenden (1992) should be included.
Description of the Kissel et al. data should be expanded to assist users in interpretation.
Since relevant activity patterns currently are not well understood, a PDF cannot be justified.
A review of the Finley et al. paper, which presents a PDF, should be included. Also, the .error in
evaluation of the Que Hee et al. data should be noted. Moreover, equal weighting of dissimilar
studies to produce the proposed PDF should be questioned.
3-19
-------
Because adherence is a function of activity, the key question is how to apply data in the
absence of adequate activity pattern data. The work group recommends use of multiple ranges with
descriptions of representative activities. Other body-surface loadings can be estimated as fractions
of hand loading by range or activity (and are likely to be lower than hand loading in most cases).
The outcome for hand data would look something like the data in Table 1.
Table 1. Hypothetical Hand Data
Activity Range
Background
Low contact
Moderate contact
High contact
Nominal Hand Loading
(mg/cm2)
0.01
0.1
1
10
Representative
Activity
Post-bathing, preactivity
Soccer
Rugby, farming
Children playing in mud
UNCERTAINTY ANALYSIS (Handbook Chapter 8)
The work group developed the following general recommendations:
• increase uniformity of summary statistics in various sections of the Handbook;
» provide more graphical representations of data;
• distinguish more explicitly among empirical and non-empirical data sources, key
studies, and other studies;
• deemphasize "default" values wherever reasonable substitutes exist; and
• expand chapter 8 to be of practical use (this may warrant publication as a separate
document).
OTHER
The work group also suggests adding a glossary to the Handbook.
3-20
-------
Human Activity Patterns Work Group
Working Group Chair: Steve Colome, Integrated Environmental Services
Working Group Members: Ed'Avol, University of Southern California
Neil Klepeis, Information Systems and Services
John Robinson, University of Maryland
INTRODUCTION
The Human Activity Patterns Work Group was asked to review sections of chapters 3,5,6,
and 8 that involve issues of time use, microenvironmental occupancy, and activity patterns.
Exposure and risk assessment models require information and data on human activity
patterns. A substantial amount of information is included in the current draft of the Handbook, but
in the opinion of the work group that data will be lacking or limited for many exposure scenarios
encountered by an exposure or risk assessor.
The Handbook covers a new and developing field and often relies on one or two studies to
support an information need. New information is being produced rapidly in this field and some of
that information will be more directed and of higher quality than the earlier studies cited in the
Handbook. The panel recommends that the Handbook be considered one edition of a changing
volume and that revisions be considered at intervals of 2 to 5 years.
A number of general recommendations were common to all of the chapters reviewed by this
panel:
• It is important to recognize that the number of potential exposure scenarios is too
large to present every possible combination of time, location, and activity for all
major demographic groups. The current draft includes a large number of these
scenarios, but for any particular risk assessment the scenario needed may not exist
in the tables presented. The draft should acknowledge this limitation.
3-21
-------
The panel recommends that key combinations of location (residence, work/school,
outdoor) and activity (sleep, exercise, low activity) be presented graphically for
purposes of illustration. Also, raw time activity data from the human activity pattern
(HAP) studies should be included in the Handbook so that exposure and risk
analysts can address time activity patterns directly for the scenarios particular to their
assessments.
The work group found it difficult to find information fitting many individual
assessment scenarios. This difficulty might be unavoidable because the Handbook is
an evolving document related to an emerging field of investigation.
To increase the utility of the Handbook, the panel recommends starting each chapter
with an index that indicates exactly what information can be found in the chapter and
points out the role of that information in exposure assessment.
Literature cited in some of the sections did not take full advantage of related fields
and was too narrowly focused. For example, a rich literature is available on
inhalation rates from the fields of sports medicine, occupational health, and
pulmonary physiology. These fields were not explored in the chapter involving
inhalation rates, and reference to such information would help develop a better
understanding of the variability and uncertainty of this factor, which has an important
influence on the dose of inhaled contaminants.
The panel recommends that a full literature search of related fields be conducted for
the topics of inhalation rate, consumer product use, exposure assessments using
activity patterns, population mobility, lifetime, and body weight.
All of the chapters lack a general introduction that identifies the role of the
information provided in conducting exposure or risk assessments. Effectively
positioning the information in each chapter would enhance the utility of the data and
help to focus the authors of the Handbook in selecting the most useful tables and
information.
The panel recommends that each chapter open with a general introduction giving the
reader a context for the information'provided in the chapter and guidance on how
data in the chapter is used in exposure and risk assessment.
The Handbook would benefit from an introductory section to the document that
presents a conceptual framework for the risk model showing the role of exposure
assessment. The introduction should contain a flow chart of the interconnection of
exposure components, with cross-references to sections of the Handbook where the
particular information is addressed. Additionally, definitions could be established in
the introduction. In particular, exposure should be distinguished from dose.
3-22
-------
The panel recommends that an introductory chapter be written that establishes a
conceptual framework for the Handbook. See figure 1 for the type of general
diagram that would be helpful to orient the reader of the Handbook. The boxes in
the figure could be used to identify specific chapters and sections of the Handbook
that deal with the topics.
Many of the data tables in the Handbook are based on social surveys and thus they
are subject to several of the sources of limitation that affect all surveys. A major
problem with the Handbook, however, is that it tends to treat all surveys as equal,
when in fact they vary widely in sophistication and utility in terms of sample design,
field quality control, question framing, and presentation of results.
Because the description of survey methods for each of the surveys presented is
inadequate and unsophisticated, the panel recommends incorporating a full
description of survey methods into chapter 1 of the Handbook (see Attachment A).
INHALATION ROUTE (Handbook Chapter 3)
Introduction
This chapter presents a number of recent studies reporting on ventilation rates of children
and adults over a range of age distributions and exertion levels. One additional study reported on
the measured ventilation rates of California outdoor construction workers. Ten studies were
discussed, ranging in study population size and description from nine nonsedentary adult volunteers
aged 21 to 37 years on which direct measurements of ventilation were made, to several thousand
households completing the USDA 1977-78 Nationwide Food Consumption Survey, for which
ventilation rates were calculated based on metabolic relationships (i.e., oxygen consumption and
associated energy expenditure for activities of varying duration).
General Recommendations
The discussion and recommendations drawn from this chapter are based on a limited cited
data base of recently published work. A body of untapped work exists in the occupational and
physical therapy, sports medicine, and exercise physiology literature that could provide improved
estimates of ventilation rates for a range of exertion levels and life activities. The identification and
3-23
-------
3-24
-------
review of additional published work in this area also would address panel concerns about day-to-day
and subject-to-subject variability associated with the reported values and potentially expand and
enhance the demographic nature of the cited data base.
A second stated concern of the review panel was the representativeness of the cited sample.
The relatively small number of subjects participating in several of the reported studies (9 to 30
subjects) raised issues about applying these factors to larger populations of interest. The interesting
(but limited) report on nine California outdoor construction workers left the panel wondering how
applicable the activities and metabolic costs associated with outdoor construction in California are
to construction workers in other parts of the country in other weather regimes.
Accordingly, the panel's primary recommendation is to undertake a search of literature on
exercise physiology, occupational and physical therapy, and sports medicine to identify and include
additional published and peer-reviewed information about ventilation rates over a range of life
activities.
Several studies of varying size, scope, and focus were presented in the chapter; some of the
cited work is research being performed for the purpose of acute respiratory assessment, while other
studies are analyses of data initially collected for other purposes. The authors of the chapter made
some attempts to identify perceived limitations or advantages associated with the studies being
reviewed, but the evaluations made were inconsistent and occasionally superficial. The utility of the
Handbook will depend on a critical assessment of the data that are included and, in that sense,
endorsed as valid for subsequent use. Accordingly, the panel's second recommendation is to provide
a uniform, objective, and critical review of the studies presented in the Handbook, with some
judgment as to their value and applicability from the perspective of exposure assessment.
Specific Recommendations
The chapter on inhalation included a significant amount of information that will be of use
to Handbook users. The following specific recommendations, however, should be considered:
3-25
-------
table 3-22 in the Handbook is a valuable source of information and should be moved
to the front of the chapter to serve as a coherent summary and a guide to the
information presented;
the limitations and advantages presented in table 3-22 should be revised to provide
a standardized summary of information with a critical judgment of the relative value
of the data provided;
the former reference inhalation rate of 20 m3/day is too high; theoretical
considerations and field data support a reduction—but not to the precision of 13.3
m3/day; the data would seem to support values rounded at least to the nearest whole
number (a range would be even better here); and
the chapter should be corrected to differentiate between dose, which involves the
delivery of a chemical species beyond a portal of entry to a target organ, and
exposure, which involves the presence of a chemical species at a portal of entry.
OTHER FACTORS FOR EXPOSURE CALCULATION (Handbook Chapter 5)
Introduction
Chapter 5 of the Handbook (Other Factors for Exposure Calculation), is composed mostly
of summaries of HAP studies (56 pages plus 32 pages in the appendix), with smaller sections on
body weight (about 12 pages) and population mobility (15 pages plus 4 pages in the appendix) and
only one paragraph on lifetimes. The introduction to the chapter is only a few sentences long and
lacks elaboration on how the data presented in this chapter are to be used in exposure assessments.
In particular, thie data are not given a conceptual context. Several of the HAP studies (Robinson,
1965-75; Juster et al., 1975-81; Timmer et al., 1985) are outdated and not explicitly exposure
relevant. Most of the HAP data are presented in terms of mean time spent instead of frequency
distributions. In addition, little information on the percentage of time spent or the percentage of
respondents in each location or activity category is provided. Also, no clear distinction is made
between calculations involving doers (i.e., those actually experiencing a location or activity) and
overall calculations (i.e., doers plus nondoers).
3-26
-------
General Recommendations
The panel offers the following general recommendations:
The number of possible human activity pattern analyses is very large (locations x
activities x background activities x socioeconomic subgroups x geographic subgroups);
thus, the Handbook should stress that users will not always be able to find the
analyses they desire, even though chapter 5 and its appendix appear to be
comprehensive (i.e., thick).
Provide a general framework or context for the use of human activity patterns in
human exposure assessment with citations from the literature.
Since HAP studies are part of a relatively new field and they can be used in a variety
of ways, examples should be given from the literature of past human exposure
assessments that have used HAP studies.
Most of the data in the Handbook are presented as means of time spent instead of
frequency distributions and thus are not useful in probabilistic exposure assessments.
Although many of the studies do not report frequency distributions, the kinds of
frequency distributions that can be obtained from the raw data (e.g., locations,
activities, demographic breakdowns) should be summarized.
More clarification of the kinds of data presented from each study should be
provided: Are the results for all respondents or do they represent the doers only
(those who engaged in the microenvironments described)?
Specific Recommendations
The panel offers the following specific recommendations:
Make it clear that the HAP studies presented represent (1) what location, activity,
and demographic breakdowns were available, and (2) the authors' choice of which
breakdowns are appropriate to analyze.
3-27
-------
Clarify the kinds of data that are available from each HAP data base and their utility
in regard to human exposure assessment:
— 24-hour minute-by-minute diaries (for accurate determinations of time spent
in locations and activities); or
— followup questions (for occurrences of specific kinds of exposure that
respondents may overlook in their 24-hour diaries, e.g., going to a gas station
on the way to work).
Distinguish between the material presented by the authors in each study and the
information available in the HAP raw data.
Include a "usage table" that describes the kinds of data available in each study, the
specific exposure-related categories the study uses in its analyses, the other kinds of
analyses that may be possible with the raw data used in the study, and the study's
usefulness in connection with exposure assessments.
Since HAP studies represent avast data resource that cannot possibly be adequately
represented in the Handbook, consider making a version of the Handbook available
on CD-ROM or over the Internet that contains the raw data from various studies.
If the raw data are made available, then the Handbook needs to have a section
summarizing its variables, breakdowns, and codes.
Add the new 1992-1994 National Human Activity Pattern Survey (NHAPS) data to
the Handbook (see examples in Attachment B).
The Sexton/Ryan "study" should be moved into the introduction to the chapter.
Old and outdated studies—both test and table—should be deleted from the
Handbook (Robinson, Juster, and Tinner, pp. 5-14 to 5-30 and 5A15 to 5A21 in the
appendix; Tarshis, pp. 5-60 to 5-62; Sell, pp. 5-64 to 5-66; and James and Knuiman,
pp. 5-62), although they can be retained as references for those who may be
interested in them.
See the suggestion in John Robinson's premeeting comments to omit material.
Mention ways to improve future HAP data collection efforts:
— more specific exposure-related activities should be included (e.g., different
categories of food preparation [baking and frying vs. sandwiches/salads] and
cleaning [vacuuming, dusting, waxing the floor vs. general tidying]); and
— more specific exposure variables should be included (e.g., smoker present,
heat on, gas oven in use).
3-28
-------
Where appropriate, footnotes should be included at the bottom of each table
explaining whether the calculations are for doers or both doers and nondoers.
Some discussion should be provided on the concept of a microenvironment (a
specific combination of a location and an exposure-related activity) with appropriate
citations.
CONSUMER PRODUCTS (Handbook Chapter 6)
Introduction
This chapter attempts to present data on the specific usage of consumer products as it relates
to potential exposure. The data come from one survey of particular consumer products conducted
in 1987 by Westat. The only data presented appear to be in terms of minutes of exposure for those
exposed for this restricted range of consumer products.
General Recommendations
The panel offers the following general recommendations:,
• The title and range of potential exposures covered in this chapter need to be greatly
expanded. The Westat data refer only to solvents and neglect the many other sources
from consumer products, including "secondhand" exposure after the product has been
used. This would include exposure to secondhand tobacco smoke, exhaust from
gasoline engines, and usage of dishwashers by other household members. In addition,
the list of potential pollutants include tap water, benzene, pesticides, and paints
among many others.
3-29
-------
For many of these potential sources, more recent and generalizable data from the
1992-1994 MAPS (or NHAPS) data collection for EPA are readily available. The
base data are provided by Robinson and Blair (1995) and several sample pages from
this document are attached as examples of statistical information that could be
directly reported in the Handbook. A list of all MAPS sources covered by their
questions is shown in Attachment C (exhibits 1 and 2 and figures 2 and 3, and
percent exposed data are given in tables 3-6 and 10; frequency distribution data for
some of these questions are shown in table 7). An example of differences in
percentages of the population exposed is shown in table 13, and parallel data are
published in the report for the other pollution sources in tables 3-7. More detailed
breakouts of data for environmental tobacco smoke exposure are shown in table 2,
along with an analysis of differences by time of day in figure 1.
In line with the wide variety of products covered and the primary/secondary usage
split, the chapter needs to be retitled to "Exposure to Specific Potential Pollution
Sources."
Specific Recommendations
The panel offers the following specific recommendations:
The list of potential pollutants extends far beyond those covered in MAPS, or in the
Westat solvent study. Nor did MAPS cover specific brand names or products
involved. These are potentially covered in market research data collected by
Simmons, which are available through Pandian at the University of Nevada at Las
Vegas. These data sources need to be cited at least; presumably, however, there is
an EPA list of complete pollution sources that could be cited.
The Westat data that are cited are too narrow. They should include at least
percentages of the population who report using the product and their estimates for
number of uses per year. The need for single-day validity data should be noted, given
that estimate data typically involve overreports. The other measurement
procedures/limitations of the study should be reported, particularly the response rate
for the crucial mail-back portion of the study (not just the initial 73 percent
telephone response rate) and information on the fairly cumbersome questionnaire
respondents were asked to fill out.
3-30
-------
ANALYSIS OF UNCERTAINTIES (Handbook Chapter 8)
Although this is a well-written and general summary, it provides little direct guidance to the
risk assessor. The intent may be to force careful thinking with each assessment undertaken. If the
other sections had been written with the same clarity, the rest of the document would have been
easier to read.
3-31
-------
-------
ATTACHMENT A
GENERAL COMMENTS ABOUT SURVEY METHODS
Because many of the data tables in the Handbook are based on social surveys, they are
subject to several sources of limitation that affect all such surveys. Thus, a major problem with the
Handbook is that it tends to treat all surveys as equal, when in fact they vary widely in sophistication
and utility in terms of sample design, field quality control, question framing, and presentation of
results.
In general, a well-conducted survey of the public is expected to meet the following criteria:
1. A probabilistic sampling frame, in which all individuals have ah equal (or at least
known) chance of selection;
2. Sample sizes selected at random from the population that allow generalization to
that larger population. (While statisticians argue about that sample size, it is the case
that a random sample of 100 individuals has a sampling error of +/- 10 percent,
which can be tolerable for some estimation purposes—if criteria #1 and #3 are met.
Sample sizes below 20 or 30 individuals have 2 to 3 times that level of imprecision
and are usually considered to be quite unreliable, particularly if the sample
respondents are not chosen at random—as is usually the case.)
3. A high rate of response from those individuals chosen at random for the survey.
(This is usually not a problem for surveys conducted by the U.S. Census Bureau with
response rates above 90 percent, but it can be a serious problem for typical survey
organizations that tolerate response rates of 60 percent or less. Few "consumer
panel" surveys achieve response rates close to that level, if strict response rates are
calculated. Unfortunately, the possibility of biased samples of respondents are high
in such circumstances.)
4. Careful attention to the ways information and questions are framed to respondents.
(Different ways of framing questions have been found to produce differences of 20
to 60 percentage points in estimates, compared to the 3 to 5 point error ranges
associated with sampling error.)
Unfortunately, much less is known about these latter contributors to "nonsampling error" and so field
procedures to overcome them are much less subject to control.
3-33
-------
Some ways of asking behavioral questions are more generally accepted by survey practitioners
than others, however. In general, the easier the reporting task expected of the respondent, the
better. Thus, asking respondents to maintain accounts of what they are doing at the moment is
easier and more reliable/understandable than asking what they do "regularly" or "typically." This
approach also is preferable to long-term recall (e.g., "over the last six months"). Asking respondents
to recall what they did yesterday, however, has not been found to generate serious recall difficulties
(as is implied in several passages ,in the Handbook). A problem with "yesterday" behavior is that it
provides only a limited view of the behavior of individual respondents; however, it can produce quite
reliable data on what the population does on a particular day.
A major problem does arise, however, when these one-day data are used to model the long-
term consequences of exposure for individuals. An individual can be exposed to an average carbon
monoxide level per day at certain levels, but if the individual receives all of that dosage in a few
minutes of a single day, it can be lethal. These long-term consequences at the individual level need
to be considered. Thus, in general, the reader needs to take into account myriad factors before
treating these data as factual or as scientific and free of mundane or naturally occurring sources or
error. This should be done at the outset and in the context of each chapter, much as in the spirit
of the current text, but more targeted on the most important sources of error.
Along the same lines, possibly each chapter could end with a call for needed measurement
advances to produce the kind of statistical data that would be most appropriate for policy purposes.
A further problem arises from the lack of essential data for understanding the implications
of the-data that are presented. Thus, for drinking water or point application, what proportion of the
population are involved in the activity for a day or a year? The percentile data appear virtually
uninterpretible without such basic statistics, which should be readily available in the original source
(if not, the original authors should be chided for omitting it). Many of these parameters are now
available from our 1992-1994 MAPS study that should soon be published.
3-34
-------
ATTACHMENT B
EXAMPLES OF ANALYSES DONE WITH THE NEW NHAPS 1992-1994 DATA
3-35
-------
Time Activity Panel - 8/9/95
Residential-Indoors 68.73
Other Indoor 2.07
Bar/Restaurant i
School/Public Bldg.
Mall/Other Store 2.26
Residential-Outdoor 3.69 5 52
In Vehicle
Office/Factory r
Other Outdoor J'
Near Vehicle
1.7
2.19
Figure 1. The overall weighted percentage of time spent by the respondents in each location. The total amount of time is 1,440
min x 9,196 respondents = 13,242,240 minutes.
OS
Residential-Indoors
Residential-Outdoor
In Vehicle
Near Vehicle
Other Outdoor
Office/Factory
Mall/Other Store
School/Public Bldg.
Bar/Restaurant
Other Indoor
,
1,001.39
Weighted (N=9196) '
Unweighted (N=9386)
226.46
225.75
0 200 400 600 800 1,000
Mean 24-hour Cumulative Duration (minutes)
1,200
Figure 2. The overall weighted and unweighted mean 24-hour cumulative durations in each location (for doers only). In the
weighted analyses, 190 respondents with missing age or gender values were excluded.
INTRO.DOC
3-36
-------
Time Activity Panel - 8/9/95
so
60
40
20
72.47
Other Indoor
Bar/Restaurant
School/Public Bldg.
Mall/Other Store
Office/Factory
Residential -Indoors
Residential -Outdoor
In Vehicle Near Vehicle
o
Residential-Indoors In Vehicle Other Outdoor Mall/Other Store Bar/Restaurant
Residential-Outdoor Near Vehicle Office/Factory School/Public Bldg. Other Indoor
Location
Figure 3. The weighted percentage of time spent in each location for males vs. females (doers plus non-doers).
INTRO.DOC
3-37
-------
Time Activity Panel - 8/9/95
Bathing 9.38
Yard/Mainten. 11.28
Sports/Exercise 13.26
Housekeeping 12.42
Dishes/Clean Kitch 6.02
Food Preparation 11.8
Eating/Drinking 35.82
Figure 4. The overall weighted percentage of time spent by the respondents in each exposure activity — excluding time spent in the
No Exposure category. The total amount of time is 1,737,104 minutes = 1,440 mm x 9,196 respondents (13,242,240 min) minus
11,505,136 min (the 86.88% spent in the No Exposure category, REGACT = 0).
Food Preparation
Dishes/Clean. Kitch
Housekeeping
Bathing
Yard/Mainten.
Sports/Exercise
Eating/Drinking
I
50 100 150
Mean 24-hour Cumulative Duration (minutes)
200
Weighted D Unweighted
Figure 5. The overall weighted (N=9196) and unweighted (N=9386) mean 24-hour cumulative durations in each exposure activity.
In the weighted analyses, 190 repondents with missing age or gender values were excluded. See Section 4 for a discussion of the
weighting methodology.
INTRO.DOC
3-38
-------
ON
i
1
O
en
en
.2
"3
I
1
1
w
I
1
0>
.S
I
ex,
en
V
8P
I
QH
1
I
o
O
O
3-39
-------
-------
ATTACHMENT C
LIST OF ALL MAPS SOURCES
3-41
-------
Exhibit 1: Chemicals/Pollutants Associated with Survey Questions (Form A)
1. a) Gasoline Storage
b) Lawn Mowers
c) Pa1nt/Varn1shes
2. a) Mothballs
b) St1ck-ups
c) Deodorizers
d) Humidifiers
CHEMICAL
£S HQj & VOCs/Bensene PAHs RFPs
x
x
X
X
X
X
Lead Pesticides Chloroform Other
P-DCB
0
Naphtha
P-DCB
Lfmolene
3. a) Paints
b) Fried food
c) Open flames
d) Glues
e) Solvents
f) Pesticides
g) Floor wax
h) Gas equipment
1) Cleaning agent
j) Dust
... Spot remover
1) Nail polish
m) Perfumes
4. Smoking
5. Diary smoking
6. Gas station
7. Gas range/pilot
8. Shower/bath
9. Dishwasher
lO.Washing machine
11.Aerosol spray
12.Heating
13.Traffic/parking
Pll. Child floor
P12. Child outside
x x
x x
x x
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Cadmium
Cadmium
P6/Br
Dust
3^42
-------
Exhibit 2: Chemicals/Pollutants Associated with Survey Questions (Form B)
ygg Benzene Particles Pesticides Hetal$ Chloroform PAHs Other
l.a) Gasoline storage x x
b) Lawn mowers x x
c) Pa1nt/varn1shes x x
2. Solvents, etc. x x
3. Renovation x x x
4. Pesticide x
treatment x
5. Floors swept x x
6. Welcome oats x x x
7. Smoking x x x x x
11. Fish, eating x PC-B5
Al-5) Drinking/ Washing x x Other
trihalomethanes
A6) Humidifier Biological aerosols
(metals) bacteria.
etc.
A7-11) Bathing, other , x Tnms
3-43
-------
FIGURE 2
Implicit Model for EPA (MAPS Survey — Form A)
Background Factor
Outcome Variables
Biological
40. Age
46-7. Race
ob. Gender
gtatuf
41-4. Employment
Si. Education
Role
kidl. Children
51. Other Adults
41-4. Work hour
a. evenings
b. outdoors
Geographic
45. ZIP-work
57. ZIP-home
54-8. Housing
59-65. Structure
Stories,rooms,
carpet,basement,
garage
Life-style
49-50. Health
POTENTIAL POLLUTANT EXPOSURE FORM —
Air
Storage
1. Gas cans
2. Lawnmower
3. Paints
4. Mothball
6. Deodorz
Water Ingestion
Yesterday
8-9.
a.
10.
11.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
b.
25.
f.
29.
30.
b.
d.
e.
f.
g-
34.
35.
36.
37.
38.
65a.
Smoking 26.
home/away 27.
Others smoke 28.
Paints
Open flame
Glues
Solvents
Pesticides
Floor wax
Gas equipment
Cleaning agents
Excessive dust
Stain removers
Perfumes
Nail polish
Gas stop
Pump gas
Gas stove
Microwave
Aerosol spray
Furnace
Fuel •
Wood stove
Kerosene
Electric space
Fireplace
Heavy traffic
Road run/walk
Parking garage
Walk to car
Other outside
54-6.Kid<5
Shower/bath
Dishwasher
Wash machine
,
.
+Diarv Variables
SMOKING
HARD BREATHING
OUTDOORS
TRAVEL
COOKING
ETC.
Garage started car
7. Humidifier
13. Windows
open
32a. Doors open
3-44
-------
FIGURE 3
Implicit Model for EPA (MAPS Survey -- Form B
Background Factors
Outcome Variables
Biological
40. Age
46-7. Race
ob. Gender
Status
41-4. 1
41-
48.
Employment
Education
Income
Children
51. Other Adults
41-4. Work hour
a. evenings
b. outdoors
ooraphj
45. ZXP-work
57. ZIP-home
54-8. Housing
59-65. Structure
Stories,rooms,
carpet,basement,
garage
Life-stvle
49-50. Health
POTENTIAL POLLUTANT EXPOSURE FORM »
Air
Storage
1. Gas
2. Lawnmower
3. Paints
4. Solvent
water
Ingestion
Yesterday
8-9. Smoking
a. home/away
c. Other smoke
55. Car starts
ast 6 Months
Renovations
a. Paint
b. Floors
c. Addition
d. Carpets
e. Glues
f. Sleep elsewhere
6. Pesticides
c. Personal
7. Vacuum floors
8. Humidifier
56. Gas stove
57-8.Heat sources
54-6. Kid <5
15-7. Shower/bath
18. Dishwasher
20. Wash machine
19. Dish washing
9. Water source
10. Bottled water
11. Tap water
12. JUices
13. Soft drinks
IB. Pool swimming
D17. Soil
24. Fish
25. Black
•fDiarv Variables.
SMOKING
HARD BREATHING
OUTDOORS
TRAVEL
COOKING
ETC.
3-45
-------
TABLE 3: EXPOSURE TO DIFFERENT AIR POLLUTANTS "YESTERDAY"
Open paint
Fried/grilled food
Open flame
Glues/adhesives
Solvents/fumes
Pesticides
Floor wax
Gas equipment
Cleaning agents
Excessive dust
Stain remover
Colognes/fragrance
Nail polish
Aerosol spray
Mothballs
Air fresheners
Toilet deodorizer
ALL AGES
Median
% Exposed
6%
24
10
7
11
6
8
10
19
16
3
50
5
32
13
65
46
Minutes/(Times)
Exposed/Exposure
60 Minutes
17
20
17
20
10
10
60
10
120+
5
1-2 Times
NA
1.6
NA
NA
NA
3-46
-------
TABLE 4
GENERAL EXPOSURE TO VARIOUS HOME SOURCES OF AIR POLLUTION
% Reporting
Gasoline 20%
Lawnmowers 29
Paints 45
Solvents 28
In last 6 months:
Renovate home 34%
Indoor painting 27
New flooring 5
Added room 4
Carpeting (with glue) 9 (1 %)
Pesticides applied 43%
Indoors 16
Professionally 18
Personally 26
Vacuum floor (3+/week) 74 %
Use Welcome mat 89
Work with soils
38%
Eat seafood
57%
Ate blackened food
25%
Used microwave 54% (5 minutes)
Smoking allowed in home 32%
3-47
-------
TABLES
EXPOSURE TO GASOLINE AND OTHER PRODUCTS/EXHAUST "YESTERDAY"
At gas station
Pump gas
Others pump gas
Drive in heavy traffic
Walk/run near road
Indoor parking garage
Walk to car
Other time outdoors
Used gas oven
Used for heat
Heated home
with gas
with furnace
Used other heating
Electric
Coal
Wood stove
Kerosene
Electric space heater
Fireplace
Other heating
Window open
Outside door open
Smoked at all
Smoker in house
Smoking by self at home
Cigar
% Reoortine Time
21%
12%
6
25%
8
6
72
27
23
0.6
40
19
30
5
3
2
3
1
3
2
4
42
84
21
15
17
1
10 Minutes
NA
17
15
5
5
30
30
11 hours
NA
NA
.
6-9 times
NA
NA
NA
Started auto
in attached garage
22
1.4 Times
3-48
-------
TABLE 6
WATER-RELATED EXPOSURES
Took bath or shower 91 %
Shower 76
Bath 15
15+ minutes
(Bath or Shower) 31
Swimming 8%
Drank tap water 72%
3+ glasses 41
Drank juice mix 61
3+ glasses 28
Drank sodas 54
3+ cans 15
Use bottled water 43
Get public water 80
Get well water 16
Washed dishes 84
Dishwasher used at home 23
Washing machine used at home 43
Use humidifier 24
3-49
-------
•si
a J
(£4
d
tt
•5
So
CO
l
CM
1
S
g
Kl
S
CM
Si
s
S
s
8!
CM
S
s
s
CM
CM
K»
3E£
ii
o
CM
to
c8
i
3-50
§
3
•f
CM
1
-------
TABLE 10
Means and Percentages of Selected Diary Exposure Variables
LOCATIONS:
Travel
Vehicle Travel
Outdoors
Bar/Nightclub
Restaurant
Auto Repair
Dry Cleaners
ACTIVITIES:
Food preparation
Repair cars
Car repair shop
Play sports
Wash, shower etc.
*
EXPOSURE:
Others' smoke (ETS)
Overall
Mean
85 rain/day
77
110
6
20
4
0.4
23
2
1
21
15
167 min/day
Doing
87
82
59
3
21
2
0.3*
43
1
2
16
63
44
Mean
per Doer
98 min/day
93
187
176
97
222
112
54
111
33
129
24
Median
per Doer
70 min/day
69
120
145
60
60
10
40
60
10
95
20
382 min/day 320 min/day
3-51
-------
M
J!
ea
5!
\
I
3j
1
fi
i
fi
J
Jj
I
li
I
ll
i\
$
»
5
s
s
tf
2
*
**
f
s
n
=
1
•1
J
*
3
«*
V
n
*
••
r*
«n
M
n
«
•
*
^
i*
M
•4
PI
S
*
n
•
R
i
i
i
S
*<
M
*
f
m
0
3
S
*"
**
Wt
V
M
H
R
•
%J
S
:
m
ft
;
n
?
Wl
N
n
•4
A
8
W
«
1*
£
n
„
M
W
n
s
•4
•
•
S
3
S
^
n
i*
o
**
;
5
n
•
ft
S
«
f
V
•4
8
S
•
n
n
n
M
«
«
«
m
«
M
*4
«
3
«•
S
I
S
»
Ml
«
S
S
3
n
•4
n
:
3
*
S
4*
S
5
f»
n
s
2
5
i!
rft
|A
VI
m
n
n
s
*
n
««
n
s
*
*
M
M
n
«
*
#4
5
*
s
•
M
§"*
*
»•
M
t»
n
:
m
*4
«
S
c
1
e
*
«
- *
S
s
e
*"
N
S
i
,
"
#i
N
W
«
:
M
M
rt
5
t
I
v»
^
•
M
•
V
s
ft
ul
b
4
•
8
£
«>
net
«•
n
ft
;
w
e
»
^
•
M
*•
*
*
f»
m
M
S
^
4
s
1
(
e
X
«
r»
<*
«
<4
8
V
S
I
I
"8
1
«s
^
r»
rf«
;;
s
s
8
i
I
t
M
3
*
4
•ft
m
rf
«
n
s
3
r
MC-Grad
£
3-52
-------
TDtlet 1
Deoderlier |
al
|
i
•
s!
1
il
I
fj
02
!
ll
|
li
j
»i
ll
c£
li
t
l
s
H
"
g
°
•f
n
£
e
r
i
1
,
5
-
«
e
«
**
c
J
1
I
5
V
s
:
°
w
- ?
.
b
1
2
•;•'
\
(-
•
-«
;
e
3
"
tf
£
|
i
*
«j
^
Ml
- f"
"
S
w
J
i
^
in
5
S
N
f-
"
,
J
I
^
,
-
n
m
*
"
s
i
„
5
a
a
e
«
n
"
i*
t
i
„
3
2
M
tft
e
••
;
i
i
,
j
3
j
«
"
U
1
«
3
-
Z
<•
H
e
o
**
i
i
„
,
•4
S
fj
rt
3
\
„
3
2
»
0
H
V
I
m
l-
V
s
5
«
5
n
S
I
1
-
1
„
-
S
•
M
H
rt
"
|
•
i
„
2
s
(A
"
-
n
i
I
s
s
«
s
s
.
"
1
!
0
3
t-
n
»
•
H
n
I
i
S
„
5
M
«
S
.
"*
'I
!
m
J
-
S
M
e
"
|
I
„
;
3
S
wt
u
**
Jj
1
o
0
^
s
.
0
«*
p
i
s
3-53
-------
Table 2. Differences in the Average Duration ofMPE
by the Most Significant Predictors.
Average Maximum Duration of Exposure
to EPS
(Minutes per Day)
Unadjusted Adjusted
TOTAL
1579
178
178
EDUCATION
Grammar School
High School Incomplete
High School Graduate
Some College
College Graduate
Graduate School
Not Reported
DAY
Sunday
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
EMPLOYMENT
Working
Looking for Work
Laid Off from Work
Retired
Going to School
Keeping House
Something Else
MARITAL STATUS
Married
Living Together
Widowed
Separated
Divorced
Never Married
Refused
(48)
(132)
(488)
(471)
(267)
(154)
(18)
(226)
(305)
(322)
(269)
(30)
(201)
(226)
(1066)
(50)
(14)
(202)
(55)
(140)
(52)
(918)
(86)
(75)
(25)
(125)
(334)
(16)
Eta=.13
169
193
210
172
180
99
83
Eta=.09
172
149
167
182
271
204
200
Eta=.20
206
202
303
117
68
85
154
Eta=.I3
174
282
96
145
193
184
110
Beta=.ll **
201
209
200
170
176
109
173
Beta=.10 **
209
146
163
170
257
184
208
Beta=.06 (NS)
182
190
194
190
109
156
177
Beta=.09 (NS)
181
247
153
114
176
171
82
Table 2. Differences in the Average Duration of MPE by the Most Significant
Predictors (Cont'd)
n Average Maximum Duration of Exposure
toETS
(Minutes per Day)
Unadjusted Adjusted
NO. OF CIGARETTES Eta=.34 Beta=.32 ***
SMOKED YESTERDAY
None (1230) 141 144
1-9 (91) 211 193
10-19 (102) 279 286
20-29 (103) 337 316
30-39 (24) 443 427
40 or more (27) 565 568
Multiple R (Squared)
.46 (.21)
Note: ** - significant at p < .01; *** - significant at p < .001
3-54
-------
^^
n
Q
"5
o
LU
O
(0
c
.2
E
8
Q.
(0 LU
o
£ I
-------
Housing Characteristics and Indoor Environments Work Group
Work Group Chain P. J. (Bert) Hakkinen, Procter & Gamble Company
Work Group Members: James Axley, Yale University
Andrew Persily, National Institute of Standards and
Technologies
Thomas Phillips, California Air Resources Board
P. Barry Ryan, Emory University
John Talbott, U.S. Department of Energy
INTRODUCTION
The Housing Characteristics and Indoor Environments Work Group focused its review on
the draft Handbook's chapter 7 on reference residence, which was developed from input provided
by the 1993 work group (panel) assigned to help develop this new chapter. The 1995 work group
included five members of the 1993 work group. This summary report is organized as follows:
• review of 1993 workshop comments on what to include in the chapter;
• overview of the contents of the draft chapter reviewed by the 1995 work group;
• summary of the 1995 premeeting comments on the draft chapter;
• summary of discussions of the July 25-26,1995 housing characteristics and indoor
environments work group;
• the residential model proposed by the 1995 work group for this chapter;
• the 1995 work group's proposed outline for the chapter, based on premeeting
comments, the July 25-26 work group's discussions, and the proposed model;
• the 1995 work group's guidance on what to include in the proposed sections of the
revised chapter;
» data gaps and research needs noted by the 1995 work group, based on the July 25-26
discussions; and
• summary and recommendations.
3-56
-------
REVIEW OF 1993 WORKSHOP COMMENTS ON WHAT TO INCLUDE IN THE CHAPTER
The 1993 work group's opinion was that the reference residence chapter should focus on
inhalation, because of the relevance of this route to many indoor pollutants, and should cover the
residential factors judged to have the greatest potential impact on these types of exposure
assessments. The 1993 work group identified the following "high-priority" data and recommended
that they be covered in the chapter:
• For single-zone assessments:
— whole residence volume; and
— air exchange rates.
• For multizone assessments:
— . room/zone volumes; and
— room/zone air exchange rates.
• For "sink" terms for deposition and sorption:
— surface, areas for walls, floors, and ceilings; and
— composition of walls, floors, and ceilings.
• For water-related assessments:
— water usage for baths and showers;
— water usage for appliances; and
— water temperatures for appliances.
3-57
-------
OVERVIEW OF THE CONTENTS OF THE DRAFT CHAPTER REVIEWED BY THE 1995 WORK
GROUP
The input from the 1993 work group led to development by EPA and its contractor of the
draft chapter provided to the 1995 work group for review and comments. The draft chapter
reviewed by the 1995 work group was organized into the following sections:
7.1 Introduction
7.2 Indoor Volumes.
7.2.1 Volumes of Residences
7.2.2 Room Volumes and Surface Areas
7.3 Airflows
7.3.1 Background
7.3.2 Air Exchange ,
7.33 Interzonal Airflows
7.3.4 Variability Within Zones
7.4 Water Supply and Use
7.4.1 Background
7.4.2 Water Use
7.5 References for Chapter 7
SUMMARY OF THE 1995 PREMEETING COMMENTS ON THE DRAFT CHAPTER
The 1995 work group assigned to this chapter, and some members of the other 1995 work
groups, provided extensive premeeting written comments about the draft chapter. Overall, the
premeeting comments were favorable (e.g., "excellent material," "useful"). The premeeting
comments discussed on July 25 and 26 by the 1995 work group are as follows:
• Modify the Introduction to help readers understand why this type of information on
house and other residential volumes, air exchange rates, and water uses and volumes
is important and how it can be used. The introduction could cite key publications
to help readers understand use of the above information (e.g., McKone, T.E.
Household exposure models. Toxicol. Lett. 49:321-329 (1989); Wilkes, C.R. et al.
3-58
-------
Inhalation exposure model for volatile chemicals from indoor uses of water.
Atmosp. Environ. 26A:2227-2236 (1992); and perhaps publications by Barry Ryan,
Ken Sexton, and others).
Add information on:
— modeling approaches (e.g., microscopic and macroscopic modeling) and
computational tools; and
— sources of exposure (e.g., airborne, waterborne, dust and aerosol, and
transport between source types).
Make sure key new documents are included. For example, sections 7.2 and 7.3.2
state that no measurement surveys have been conducted to directly evaluate the
range and distribution of residential volumes and residential air exchange rates.
Some candidates for addition, however, include:
— Pandian, M. et al. Residential air exchange rates for use in indoor air and
exposure modeling studies. J. Exposure Analysis Environ. Epidemiol. 3:407-
416 (1993). (Residential air changeovers in different regions of the United
States, different seasons, and different levels within the homes.)
— Murray, D.M., and Burmaster, D.E. Residential air exchange rates in the
United States: Empirical and estimated parametric distributions by season
and climatic region. Submitted for publication in Risk Analysis. (Residential
air changeovers in different regions of the United States, different seasons,
and as a function of "heating degree days.")
— Murray, D.M. Residential total house and zone volumes in the United
States: Empirical and estimated parametric distributions by season and
climatic region. Submitted for publication in Risk Analysis. (House volumes
in the United States as a whole and for eight states. Results also presented
for house zone volumes.)
Add other products to those listed for wall coverings and floor surfaces.
Add a listing of types of airborne contaminants likely to be emitted by wall covering
and floor covering products.
Add a discussion of the roles of season and temperature in affecting air exchange.
Discuss the role of indoor versus outdoor influences and contributions.
Discuss the possible impact of chemical and physical transformation.
Discuss the possible impact of exposure from soil gas and ground water (e.g., via
basement).
3-59
-------
Discuss changes in residential parameters such as ceiling height as a function of the:,
year of construction of the residence.
Add information on various residential surfaces that might be useful for dermal
exposure assessments. This could include surface areas of various objects that might
be handled.
Expand indoor volumes section to become "Building Characteristics" (e.g., include
configurations; surface areas of walls, floors, and ceilings; and characteristics of
construction materials and furnishings).
Discuss need to treat some materials as porous solids (e.g., include thickness,
porosity, mass per unit volume, specific surface area).
Discuss suburban versus urban and rural residences.
Discuss possible relationships between house volumes and air exchange rates, and
what might happen in individual rooms and zones.
Discuss the best way(s) to present chapter 7's current and possible new information
(e.g., format, key studies versus other studies, nature of the distributions).
List the significant data gaps and research needs.
Since this chapter and chapter 6 on consumer products are new, consider asking
readers and others to help ensure that all potentially useful residential exposure data
are included in future revisions of the Handbook. Consider establishing an EPA, or
other, contact to receive data for possible inclusion in future revisions; perhaps
listings of new data could be sent to known users of the Handbook, or the listings
could be posted as an update file on the Internet (e.g., on a World Wide Web
homepage).
SUMMARY OF THE JULY 25-26 DISCUSSIONS OF THE 1995 HOUSING CHARACTERISTICS
AND INDOOR ENVIRONMENTS WORK GROUP
The 1995 work group discussed the following:
The chapter should focus on information useful for single zone modeling, but this
emphasis should be placed in the larger context of other modeling possibilities (e.g.,
multizone).
3-60
-------
The draft version of the chapter should be restructured. The group discussed
possible sections, figures, and tables that could be modified or dropped, and, if
dropped, how the key information could be captured elsewhere in the chapter. The
work group examined the existing draft very carefully to assess how the information
might be better organized and presented, what information should and should not
be included, and what would be best for possible future revisions of the Handbook.
The contents of the chapter need to go past inhalation as a route of exposure to
include information useful for dermal assessments.
The current references need to be updated to include new studies.
Significant data gaps and research needs should be addressed, either in a new chapter
or as a section at the end of the current chapter. The work group developed a listing
of these gaps and needs.
THE RESIDENTIAL MODEL PROPOSED BY THE 1995 WORK GROUP FOR THIS CHAPTER
The 1995 work group proposed the following for use in a revised figure 7-1, Elements of
Residential Exposure:
Air exchange, leakage
Sources:
— direct emission (S)
— transport from outdoors (air, water, soil)
— re-emission, re-suspension (R)
Sinks and loss mechanisms:
— deposition (D)
— transport out
— reaction (RJ
— reversible Sinks
Air In
Water In
Soil In
Concentration, *
S R I
t t
Exposure, E for Occupant(s)
D Rx
^ Out
3-61
-------
THE 1995 WORK GROUP'S PROPOSED OUTLINE FOR THE CHAPTER, BASED ON
PREMEETING COMMENTS, THE JULY 25-26 DISCUSSIONS, AND THE PROPOSED MODEL
(SHOWN ABOVE)
The 1995 work group proposed the following outline for the revised chapter:
7.1 Introduction
7.2 Building Characteristics
7.2.1 Volumes and Surface Areas of Residences
7.2.2 Volumes and Surface Areas of Rooms
7.23 Mechanical System Configurations
7.2.4 Building Materials and Furnishings
7.2.5 Basement and Crawl Spaces
7.3 Transport Rates
7.3.1 Airflow Rates
Air Exchange Data
Lawrence Berkeley Laboratory. Model
Mechanical Systems (Kitchen, Bathroom, and Newer Mechanical Ventilation
Systems)
7.3.2 Deposition and Filtration
7.33 Interzonal Airflow
7.3.4 Water Supply
73.5 Water Filtration
7.3.6 Soil Tracking
73.7 Soil Removal/Resuspension
73.8 Wind and Outdoor Temperature (for Predictive Models)
7.4 Sources
7.4.1 Airborne Sources (Outdoor Air Concentrations and Indoor Airborne
Sources)
7.4.2 Waterborne Sources
7.43 Soil and House Dust Sources
3-62
-------
7.5 Complications
7.5.1 Personal Versus Micro-environmental Exposures
7.5.2 Reversible Sinks
7.6 References
THE 1995 WORK GROUP'S GUIDANCE ON WHAT TO INCLUDE IN THE PROPOSED
SECTIONS OF THE REVISED CHAPTER
The 1995 work group offered the following guidance on revising the chapter:
7.1 Introduction
The initial two paragraphs should discuss the general framework for residential exposure
analysis (e.g., how data such as building characteristics, transport rates, and sources and reversible
sinks are used in a "residential model" to develop estimates of concentration^]), followed by use of
other information such as human activity patterns to develop the assessment of exposure(s).
The next two paragraphs should discuss residential modeling'approaches (can use information
from pages 7-1, 7-2, and 7-3 of the draft chapter). This would include a more complete
classification and discussion of macro- and micro-contaminant dispersal, flow, and integrated analyses
and should also include a presentation of a simple single zone "case study" along with citation to one
or more key reviews (e.g., McKone, T.E. Household exposure models. Toxicol. Lett. 49:321-329
[1989]). Also, these paragraphs should inform readers about multizone and microcomputational
analyses as well as the tools available for advanced/complex studies.
The last paragraphs in the section should outline the rest of the chapter, set the emphasis
on simple single zone analyses, and identify limitations. For example, these paragraphs could include
statements about the known accuracy of some of the data (e.g., those for transport rates).
3-63
-------
7.2 Building Characteristics
7.2.1 Volumes and Surface Areas of Residences. Use information from section 7.2.1 of the
draft chapter. Also, the current section 7.2.1 on this topic includes table 7-2,,
Residential Volumes in Relation to Household Size and Year of Construction. The
work group recommends that an attempt be made to include representative building
configurations in this table (e.g., single family detached and attached, multifamily
units, and mobile homes as noted in table 7-1).
7.2.2 Volumes and Surface Areas of Rooms. Use information from section 7.2.2 of the
draft chapter and provide some text on how to use information presented in table 7-
4, Examples of Products and Materials Associated with Floor and Wall Surfaces in
Residences (Tucker, 1991, citation). Some work group members recommended
removing table 7-3 (Room Volumes and Surface Areas from Energy Conservation
and Indoor Air Quality Research Homes) and table 7-4 from the chapter, and
attempting to replace them with tables containing better data.
7.2.3 Mechanical System Configurations. Identify as a data gap?
7.2.4 Building Materials and Furnishings. Identify as a data gap? Need information on
surface areas, compositions, and porosities of materials and furnishings.
7.2.5 Basement, Crawl Spaces, and Other Possible House Areas/Units to Consider
Transport Rates
73.1 Airflow Rates
» Air Exchange Data. Use some of the information from sections 7.3.1 and 7.3.2 of
the draft chapter, along with other possible studies. The work group recommends
that the format for presentation be carefully examined, with the data perhaps best
presented in both "visual" (figures) and tabular form. The work group also
recommends that "the net be broadened" in an attempt to ensure that all useful
studies are included. Also, the work group recommends that statements be included
to address small sample sizes in some of the data sets, the type of study (e.g., 12-
hours or 1 week in duration), and the need to assess the applicability of the data for
particular assessments (e.g., exposure assessors should try to use representative data
for a particular region of interest; also, can some of the data shown in table 7-5, such
as the 23.32 air changes per hour value, be considered as very extreme and beyond
what might reasonably be considered as possibly occurring?).
• Lawrence Berkeley Laboratory Infiltration Model.
• Mechanical System (Identify as a data gap? Need information on kitchen, bathroom,
and newer mechanical ventilation systems).
3-64
-------
The work group noted the usefulness of the ASHRAE (1993) "Fundamentals" handbook as
a publication readers of this section and chapter could be directed to for additional
information.
73.2 Deposition and Filtration
7.3.3 Interzonal Airflow. The current section 7.3.3 on this topic includes figure 7-2,
Residential Configurations. The work group recommends that an attempt be made
to include other representative building configurations (e.g., multifamily units and
mobile homes as noted in table 7-1).
7.3.4 Water Supply. Use information from section 7.4 of the draft chapter, try to include
some appliance use data, and point to other references such as the University of
Pittsburgh work performed by Julian Andelman and others.
7.3.5 Water Filtration
7.3.6 Soil Tracking. Include work by John Roberts and others.
7.3.7 Soil Removal/Resuspension. The work group recommended including work by David
Layton on surface areas for resuspension, and discussing "the effective surface area"
of furniture and other residential surfaces for helping to determine the amount of
dermal exposure that can occur.
73.8 Use of Wind and Outdoor Temperature Information (for Predictive Models)
7.4 Sources
This section could be immense in size; however, an attempt should be made to keep it to
about two pages while discussing types of sources and referring to key publications for further
information.
7.4.1 Airborne Sources. (Outdoor Air Concentrations and Indoor Airborne Sources.)
7.4.2 Waterborne Sources
7.43 Soil and House Dust Sources
7.5 Complications
This section could use information from section 7.3.4 of the draft chapter for the new 7.5.1.
7.5.1 Personal Versus Micro Environmental Exposures
7.5.2 Reversible Sinks
7.6 References
3-65
-------
DATA GAPS AND RESEARCH NEEDS NOTED BY THE 1995 WORK GROUP, BASED ON JULY
25-26 DISCUSSIONS
The 1995 work group identified the following data gaps and research needs (not listed in
order of recommended importance):
• source emissions;
• urban versus suburban versus rural housing characteristics;
• ceiling height as a function of the year of construction and location (urban versus
suburban versus rural) (Note: The work group noted that variations in ceiling height
affect data on house volumes, including the information presented in table 7-1, and
that a footnote could be added to that table to note this along with noting it as a
data gap and research need);
• single versus multifamily residences, including representative building plans;
• ensuring all useful air exchange data sets are covered;
» appliance characteristics (temperatures and volumes);
n building materials and furnishings;
• mechanical system configurations and rates;
• transport (e.g., soil tracking); and
• need for reality checks of exposure assessments, including factor values used, and the
need for validation of models.
SUMMARY AND RECOMMENDATIONS
The 1995 workgroup assigned to this chapter provided extensive comments that should make
chapter 7 easier to follow and use, with the aim of having the revised proposed chapter provide key
exposure factor information needed for assessments of indoor environments. It is recommended
that EPA and its contractor review the above summary for suggested changes to the current draft
of the chapter.
3-66
-------
SECTION FOUR
OVERVIEW
REVIEWERS'PRELIMINARY COMMENTS
Prior to the workshop, each expert reviewer was asked to read the draft Exposure Factors
Handbook and provide written comments. (Appendix B presents the reviewers' premeeting
comments.) Relying on their technical knowledge and best professional judgment, reviewers
responded with comments on:
» the usefulness of presented data in support of both joint estimates and Monte Carlo
exposure assessments;
• the Agency's grouping of key studies and other relevant studies based on judgments
about the adequacy of the data and their applicability to the exposure factors being
evaluated;
» the adequacy of interpretations of the data and the appropriateness of the
limitations/uncertainties emphasized/described in the Handbook's recommendations;
and
• the usefulness of developing a new chapter (or sections at the end of each chapter)
that highlights data gaps and future research needs.
In his introductory remarks, Dr. Ryan, the workshop chairperson, summarized several
recurrent observations from reviewers in their preliminary, general comments on the Handbook:
• Overall the document is not "user-friendly" because of its size, nongraphical data
presentations, inconsistent referencing, and inconsistent formatting of footnotes,
tables, and references.
• Terms are undefined or poorly defined (e.g., upper percentile, default values). Also,
the use of significant figures varies across factors.
• Terminology is inconsistent (e.g., geometric mean versus arithmetic standard
deviations).
4-1
-------
Studies and distributions are presented without explanation of their strengths and
weaknesses. Further research is needed on identifying the "best" surveys or studies.
No distinctions are apparent between primary and secondary studies or in defining
"key" studies and "relevant" studies. Because the details of survey design are lacking,
the document assumes all studies are equal.
The chapter on uncertainty analysis (i.e., chapter 8) is incomplete. It presents tools
but describes no methods or procedures. The chapter should be either expanded or
abandoned. Also, the discussion in this chapter might be more appropriately placed
at the beginning of the document.
Some old data are used for areas in which newer and better data are available.
A Handbook needs to present a condensed summary of ranges.
The document should be made available on the Internet via EPA's homepage and
on CD-ROM.
Overall, the comments raised a number of issues for consideration at the workshop.
Comments on Food and Beverage Consumption
Barbara Petersen, Ph.D., of Technical Assessment Systems, Inc., reviewed the premeeting
comments that focused on exposure factors dealing with food and beverage consumption. Dr.
Petersen identified several major themes in reviewers' comments:
• A list of resource people needs to be developed and included in the Handbook, as
well as a list of training programs.
» Although the Handbook focuses on five major types of food and beverages, people
consume many different types of food. It is important to estimate precise and
realistic consumption values for whole diets to avoid errors in food consumption
distribution rates.
• The results of the USDA National Survey for 1989 through 1991 are not provided
in the Handbook. At a minimum, references for this data should be provided.
• Better guidance is needed on selecting appropriate exposure data for specific types
of studies, including the advantages and limitations of these data.
4-2
-------
• Categories of food (e.g., cooking) and types of data (e.g., data on meal size and daily
intake needed to assess acute exposures) are missing.
Dr. Petersen presented an array of comments on the treatment of data in the Handbook.
For example, it was felt that some statistical interpretations are not justified (e.g., presenting
percents for small sample sizes and using data on one individual to develop a range); the inconsistent
use of significant figures throughout the Handbook can effect the precision of assessment results;
expanded footnotes are needed to facilitate moving between tables and text. Also, reviewers
expressed the opinion that equal consideration was given to both primary and secondary data, and
it was suggested that the Agency look at interpersonal variations in data where point estimates and
Monte Carlo simulations can and cannot be used. Other comments included the need for the
Agency to:
• conduct systematic peer review of new data;
• provide a second source of information;
• adequately present study results to illustrate intake rates;
• review study results to ensure that accurate information is presented (e.g., cooked vs.
uncooked; dry vs. wet weight);
• clarify how household food is assigned to individuals;
• update Eleanor Pao's USDA data;
• clarify the terminology and methods used in studies to describe different types of
food (e.g., sources of water); and
• provide more information on total infant intake (breast milk and formula).
Comments on Nondietary and Dermal Exposure Factors
John Kissel, Ph.D., of the University of Washington, presented a summary of the premeeting
comments on water ingestion, soil ingestion, and the dermal route. He identified the two major
issues concerning water ingestion as (1) the presence of volatile organic compounds (VOCs) in water
and (2) the effect of showering and the use of heat in the preparation of food and incidental
4-3
-------
ingestion of water while swimming and bathing. All of the studies on water ingestion were found
to be relatively consistent.
Dr. Kissel noted several comments regarding soil ingestion. For example, it is felt that the
Handbook provided a good review of the soil ingestion literature up to 1991. Data on adult
ingestion of soil as well as on the pica child, however, was found to be inadequate. Although
multiple studies are available on ingestion of soil by children, reviewers view the mean child
ingestion rate of 180 mg/day as conditional because of their low confidence in the studies. Issues
in regard to the studies include the validity/sufficiency of input/output studies, the methods used for
short-term versus long-term studies, and the lack of confidence in the Calabrese study's probability
density function of annual averages generated from 4-day data. Given the importance of. this
exposure pathway, reviewers suggested conducting the following and including the results in future
revisions of the Handbook:
• new longer term studies;
" interim dose reconstructions; and
• independent revaluation of existing data.
Issues in regard to the dermal route and its presentation in the Handbook include:
• The lack of data on individual behaviors that can affect a determination on what skin
surfaces are exposed limits the probability density function of related exposure"
factors. Historically, exposure assessors assumed that people wore shorts andt-shirts.
Data on more specific surface areas are needed.
• The lack of nonsoil and nonwater dermal exposures.
» The literature review should be reorganized by type of study (e.g., direct v. indirect).
• More data by Kissel should be presented to put the discussion in context.
• Although soil adherence is a function of activity, the data do not reflect this
association. For example, in an episodic dermal exposure to soil, information on the
time between contact and showering is needed.
4-4
-------
Current protocols are not consistent for different routes of exposure. For example,
percent absorbed is used to calculate dermal exposures to contaminants in soil,
whereas mass transfer coefficients are used to calculate dermal exposures to
contaminants in water. The protocol for dermal exposure should be the same for all
media.
Comments on Human Activity Patterns
Steven Colome, Ph.D., of Integrated Environmental Services, presented the following
summary of reviewers preliminary comments:
• No major conflicts exist among reviewers.
• The information assembled in the Handbook is useful.
• More information is needed to orient the exposure assessor on why this information
is required and on how the information can be applied.
• More effective and critical evaluation of the adequacy and quality of information is
needed.
• The relationship between time-activity data and data applications needs to be made
clearer.
• The Handbook as written reflects the multiple authors (e.g., uneven editing,
incomplete or selective referencing).
Many of the reviewers' specific comments on chapter 3 (Inhalation Route) were of an
editorial nature. In addition, however, reviewers noted that inconsistent definitions are used and in
some cases overly precise, but not fully justified, recommendations are made (e.g., 13.3 mgf/day).
Also, study evaluations were found to be not particularly useful without the overall summary of the
full body of information. Also, it was felt that the small size of some studies may not be
representative.
Reviewer's commented that the information presented in chapter 5 (Other Factors for
Exposure Calculations) was not well synthesized. Also, the 1992-1994 NHAPS study will replace
4-5
-------
many of the earlier, outdated studies. Reviewer's identified the need to distinguish data needs for
dose and exposure.
There was general agreement among reviewers that limited data are available on consumer
products (chapter 6). They also contended that the tables containing the Westat data are not
integrated and it is unclear how this information would be used in an exposure assessment. Thus,
some suggested that a critical evaluation of data quality and information should be conducted.
All reviewers commented on the lack of direction and lack of information in chapter 8
(Analysis of Uncertainties). Several reviewers suggested that the information might be better used
as part of the introduction to the Handbook.
Comments on Housing Characteristics and Indoor Environments
Based on the results of the 1993 peer involvement workshop, chapter 7 (Reference
Residences) was added to the Handbook. PJ. Hakkinen, Ph.D., of the Proctor and Gamble
Company, summarized reviewers comments on housing characteristics and indoor environments.
Overall, most reviewers found that the chapter contained useful material. It was suggested that the
introduction be modified to help readers understand why this type of information is important and
how it can be used. It also was suggested that information be added on:
• modeling approaches (e.g., microscopic and macroscopic modeling) and
computational tools;
• sources of exposure (e.g., airborne, waterborne, dust and aerosol, and transport
between source types); and
• key new documents.
Reviewers noted that sections 7.2 and 7.3.2 currently state that no measurement surveys have
been conducted to directly evaluate the range and distribution of residential volumes and residential
air exchange rates. Some pointed out, however, that a great deal of published information, as well
as information submitted for publication, from various studies could be added to this chapter. Also,
4-6
-------
it was suggested that the publications currently cited in chapter 7 on air exchange rates need to be
verified to ensure that they have been corrected for a known data coding problem.
Reviewers suggested that information on the following be added to the chapter:
• other wall covering and floor surface products;
• types of airborne contaminants likely to be emitted by wall covering and floor
covering products;
• the role of season and temperature in affecting air exchange;
• the role and contribution of indoor versus outdoor influences;
• the possible impact of chemical and physical transformation;
• the possible impact of exposure from soil gas and ground water (e.g., via the
basement);
• changes in residential parameters (e.g., ceiling height as a function of the age of the
residence); and
• sizes and other relevant information for various residential surfaces (e.g., for possible
use in dermal exposure assessments).
Moreover, reviewers suggested that the section on reference residence:
• expand the "Indoor Volumes" section to become "Building Characteristics"
(addressing configurations; surface areas of walls, floors, and ceilings; and
characteristics of construction materials and furnishings);
• discuss the need to treat some materials as porous solids (e.g., include thickness,
porosity, mass per unit volume, specific surface area);
• discuss suburban versus urban versus rural residences;
• discuss possible relationships between house volumes and air exchange rates, and
what might happen in individual rooms and zones;
• discuss the best way(s) to present chapter 7's current and possible new information
(i.e., format, key studies versus other studies, nature of distributions); and
• discuss the significant data gaps and research needs.
4-7
-------
Dr. Hakkinen explained that as with chapter 6 on consumer products, chapter 7 is a recent
addition to the Handbook. Based on this, it was suggested that the Agency:
• ask readers and others to help ensure that all potentially useful residential exposure
data are included in future revisions of the Handbook;
• establish a contact to receive data for possible inclusion in future revisions; and
• send updates of new data to known users of the Handbook, or post an update file
on an Internet World Wide Web homepage.
Reviewers in this work group concurred with other workshop reviewers that chapter 8
(Analysis of Uncertainties) needs extensive expansion and thought. Reviewers also suggested that
this chapter might benefit from some examples.
SUMMARY OF WORKSHOP DELIBERATIONS
The workshop provided a forum for the expert reviewers to discuss the scientific aspects,
thoroughness, and completeness of the draft Handbook. Workshop participants contributed useful
and substantive suggestions and recommendations for improving the Handbook. Section 3 of this
report provides summaries and recommendations as reported by the chairpersons of the four work
groups.
All workshop reviewers acknowledged the frequent use of the Handbook by diverse groups
within the public and private sectors and commended EPA's efforts in updating and expanding the
contents of the Handbook. Reviewers recommended that EPA provide an overview discussion on
how all of the exposure factors provided in the Handbook can be integrated (e.g., why is it important
to understand activity patterns and indoor environments?) and provide conceptual frameworks for
each chapter.
4-8
-------
Food and Beverage Consumption
Considerable discussion among reviewers focused on data that are missing from the
Handbook. Although data are available from the USDA Continuing Survey of Food Intake by
Individuals (CSFII) from 1989 to 1991 and the Michigan study on fish consumption from 1992, the
results have not been included in the Handbook. The results of a survey conducted by the USDA
on ranching activities (i.e., cattle and sheep slaughtered for home consumption) have never been
evaluated regarding the distribution and frequency of consumption. Reviewers also noted that the
bioavailability issue is missing from any discussions in the Handbook on homegrown foods.
Several reviewers commented on the multiple analysis and use of the same data in different
sections of the document. Reviewers recommended that a genealogy (flow chart) of the data be
provided so that readers can follow the chronology and relationship between different studies and
their uses. Reviewers also suggested that expanded footnotes be provided in tables that indicate
where the data came from, note any analyses conducted, and explain uses for the data. Reviewers
expressed the opinion that the tables should provide sufficient documentation to enable them to
stand alone (e.g., intake rates should specify whether the observation was on an as-eaten or a dry-
weight basis; household or individual basis). One reviewer asked whether multiple sources of data
are needed to determine consistency.
Considerable discussion focused on the level of precision throughout the Handbook.
Reviewers cautioned that the Handbook should not create false precision when using existing data-
Most reviewers considered the use of one significant figure sufficient for data on food consumption.
The use of rounding was also found applicable for interpolation and percentages. Reviewers also
discussed criteria for the number of observations or subjects/cell necessary to calculate central
tendency versus percentiles. They considered n = 30 acceptable for distributions, n S: 30 acceptable
for central tendency, and n > 30 acceptable for percentiles.
Reviewers emphasized the importance of identifying subpopulations and providing for an
appropriate level in disaggregation (e.g., data on fish consumption by ethnic group). They expressed
the opinion that food disappearance data may not be useful because values can be overestimated.
4-9
-------
Reviewers discussed the limitations of studies used in the Handbook, including survey designs
(i.e., 1-year recall versus 3-day) and criteria for including distributions. They recommended that for
each study the survey methodology and its limitations be addressed. For example, one reviewer
noted that taking national data and applying it to a site-specific situation is problematic. Reviewers
recommended adding a discussion on how per capita estimates can be used when "per user" data
would be more appropriate. A discussion on how to compute high-end exposures in situations where
more than one contaminated commodity is present also was recommended. Reviewers did state that
national data could be used for screening purposes.
Several reviewers addressed recreational fish consumption (self caught) and found the basic
approach for freshwater anglers to be sound. Exposure assessors should not rely on defaults but
rather should look at available studies and present distributions for the relevant studies. Reviewers
made several recommendations for revisions to the Handbook regarding exposures from recreational
fish consumption:
• new studies need to be added;
• criteria for selecting key or recommended studies need to be presented;
• guidance on selection of relevant studies based on the type of water body, regional
variation, etc. is needed; and
• creel and mail survey results should be separated to avoid the potential for bias.
Regarding Native Americans, reviewers recommended that new studies be added and that
data on Native Americans in surveys of recreational anglers be evaluated. Reviewers also
recommended that subsistence anglers be discussed in the Handbook and criteria be developed for
determining the presence or absence of such populations at individual water bodies.
Reviewers in this work group also discussed the need for consistency within tables and
chapters and between chapters.
4-10
-------
Nondietary and Dermal Exposure Factors
Members of the nondietary and dermal exposure factors work group concurred that there
was good agreement between the results of prior studies conducted on water ingestion. Large data.
sets are available from which distributions can be recommended.
Reviewers expressed the opinion that the even the best data on soil ingestion by children
were difficult to interpret. Reviewers held little confidence in the results of the Calabrese study or
his interpretations, which were based on 4 days of data. They could not justify a distribution based
on the Calabrese data or other data, but felt a central tendency and upper confidence could be
recommended. The value of 100 mg/day was viewed as conditional by the reviewers. Reviewers
questioned the validity of available studies and the issue of values based on short-term exposures
versus exposures for longer periods.
-i
Similarly for a pica child, reviewers contended that the data sets were too small (n = 1 or
2) and the uncertainty too high to interpret, and they recommend a distribution based on the data.
Reviewers also considered the data on adult ingestion of soil to be inadequate.
Reviewers provided recommendations on how to improve the exposure factors associated
with soil ingestion:
• conduct longer term studies to validate the 4-day study;
• as an interim measure, initiate dose reconstruction efforts; and
• sponsor (i.e., EPA) an independent reevaluation of existing data.
The work group reviewers addressed how the dermal route discussion in the Handbook could
be improved:
• add nonwater and nonsoil routes for dermal exposures, and relate this discussion to
exposure to consumer products (e.g., carpets);
• add more specific skin surface areas;
4-11
-------
» remove equations in the text;
» reorganize the literature review by type of study;
• present more Kissel data to put the dermal route discussion in context;
• add behavioral data that would permit probability density functions; and
• generate activity ranges rather than calculating probability density functions (see the
Nondietary and Dermal Exposure Factors Work Group summary in section 3).
The reviewers also made several general recommendations for improving the Handbook,
including:
• add a glossary;
» separate data more explicitly by quality (high confidence versus low confidence); and
• reference data by source (i.e., original data versus manipulated data).
Human Activity Patterns
The human activity patterns work group reviewed chapters 3, 5, and 6 of the Handbook.
Reviewers expressed the opinion that data included in chapter 3 (Inhalation Route) are limited and
that the chapter would benefit from a broader literature search that sought information from the
sports medicine-and occupational fields. Reviewers discussed the role of variability (i.e., day-to-day
variability and variability in demographic groups) and the selection of a representative sample in
developing inhalation rates. They agreed that the theoretical and field data support a reduction in
the daily inhalation rate from 20 m3/day, but did not support the precision of the 13.3 m3/day rate.
Another area of discussion was the need to distinguish the difference between dose and exposure.
Reviewers found that the few distributions presented in chapter 5 (Other Factors for
Exposure Calculations) would be difficult to use in probabilistic estimates. They expressed the
opinion that the human activity patterns area remains a confusing field of study that lacks a strong
consensus within the scientific community. Because the data can be arranged in many ways and the
4-12
-------
Handbook could not present all the distributions for each situation listed in the tables, the reviewers
suggested that a structure for using human activity patterns data in exposure assessments and
applications for the data be established. In addition, reviewers suggested that the chapter could be
reduced in size.
Although the Westat results are from an established study, reviewers recommended that the
survey and measurement procedures be described in the text of chapter 6 (Consumer Products).
Reviewers also suggested that the chapter would benefit from the addition of citations obtained from
available literature. For example, information is needed on either the proportion of users of
consumer products or exposure to environmental tobacco smoke.
The human activity patterns work group made the following recommendations:
• information should be presented graphically wherever possible, given that
illustrations can present information more comprehensively than text;
• raw time-activity data from the HAP studies should be made available in the
Handbook to facilitate site or situation-specific exposure assessments;
• a short introduction or index should be added to the beginning of each chapter to
increase the useability of the Handbook (the introduction could describe the context
of the information presented in the chapter, explain what the information can be
used for, and note where the information can be found);
• a comprehensive literature search of related fields should be conducted on inhalation
rate, consumer product use, exposure assessments using activity patterns, population
mobility, lifetime, and body weight;
• a full description of survey methods should be incorporated into chapter 1; and
• a conceptual framework for the information provided in the Handbook should be
developed and presented in an introductory chapter.
Housing Characteristics and Indoor Environments
Considerable discussion took place among members of the housing characteristics and indoor
environments work group on restructuring chapter 7 (Reference Residence) and developing a revised
4-13
-------
outline (which is provided in the work group chairperson's summary in section 3 of this report). The
proposed outline includes all of the sections in the current draft Handbook except the one on
interzonal mixing.
Reviewers agreed that the focus of this chapter should be on single zone modeling; however,
they expressed the opinion that this model also should be placed in a larger context. Information
should be made available for Handbook users who want to know more about multizones. Reviewers
suggested that the chapter also move beyond inhalation to include some discussion on indoor
exposures due to soil tracking and dermal contact with surfaces. In addition, reviewers
recommended that a discussion on modeling approaches (computational tools) be added to the
introduction of the chapter and that the references be updated to include new studies.
Reviewers described a residential model with direct emissions; transport of air, water, and
soil from the outdoors; and reemission and resuspension as sources to the indoor environment. The
model would account for air exchanges and leakage, as well as sinks and loss mechanisms due to
deposition, transport out of the residence, reactions, and reversible sinks.
Reviewers identified future data needs and existing data gaps related to housing
characteristics and indoor environments. In their view, information is needed on:
• source emissions;
• ceiling height as a function of construction year;
» suburban versus rural versus urban housing characteristics;
• single versus multifamily housing characteristics, including representative building
plans;
• all useful air change per hour data sets;
• appliance characteristics (e.g., wash water temperature);
• building materials and furnishings;
• mechanical system configurations and rates;
4-14
-------
validated models; and
transport (e.g., soil tracking).
OBSERVERS' COMMENTS
The workshop agenda included an opportunity for observers to make public statements
during the morning plenary session on Tuesday, July 25, and the afternoon plenary session on
Wednesday, July 26. Observers were asked to sign up if they intended to make a statement. At the
discretion of each work group chair, observers also were provided an opportunity during work group
sessions to participate in discussions.
Only one observer, Annette Guiseppi-Elie, Ph.D., of Mobil Oil Corporation and Chair of the
American Industrial Health Council's Environmental Health Risk Assessment Subcommittee, made
a statement during the plenary sessions. Dr. Guiseppi-Elie commended EPA for expanding the
Handbook to include considerable new data and for.sponsoring the peer review workshop. She
stated that her initial observations are consistent with comments made previously by each of the four
work group chairs. She expressed disappointment that single point distributions were used (chapter
4) and that the equivocal presentation of data made it difficult to generate distributions. She said
the data presented in chapter 6 is incomplete and chapter 8 needs to be substantially expanded to
include a discussion of the advantages and disadvantages of different methods of conducting
uncertainty analyses. Dr. Guiseppi-Elie suggested that peer reviewers and the public be given
adequate time to digest and comment on the Handbook. She concluded by suggesting the following
revisions to the Handbook:
With the many intended uses of the Handbook, the document is primarily a
compilation of data. As such, the Agency should critically review relevant studies
and avoid the use of default or reference values. Exposure assessors should have
access to all available data, and the data should be presented in an appropriate
format. Nevertheless, the Handbook should explain that in some instances—for
example, screening analyses—reference values are useful.
4-15
-------
An extensive discussion on exposure assessment methodology is not needed. This
subject is covered in other documents and the Handbook is not intended to be a
guidance manual. For newer exposure assessment areas (e.g., consumer
products),some guidance in the form of examples might be appropriate.
The most up-to-date data should be provided in the Handbook.
Data should be presented in a user-friendly manner. .
Because limited data are available on exposure parameters for soil ingestion and this
pathway drives most risk assessments, the Agency should hold a smaller peer review
meeting with experts in this field to address the issues associated with ingestion of
soil by children and adults.
The revised Handbook should be made available to the public before it is finalized.
Chapter 8 should be either eliminated as it currently exists in the Handbook or
expanded, improved, and peer reviewed.
4-16
-------
APPENDIX A
REVIEWER LIST
A-l
-------
-------
SEPA
United States
Environmental Protection Agency
Peer Review Workshop on Revisions to
the Exposure Factors Handbook
Doubletree Hotel Park Terrace
Washington, DC
July 25-26, 1995
Reviewer List
Edward Avol
Department of Preventive Medicine
School of Medicine
University of Southern California
1540 Alcazar Street - CHP 236
Los Angeles, CA 90033
213-342-1090
Fax: 213-342-3272
James Axley
School of Architecture
Yale University
180 York Street
New Haven, CT 06520
203-432-2283
Fax: 203-432-7175
David Burtnaster
Alceon Corporation
4th Floor
2067 Massachusetts Avenue
Cambridge, MA 02140
617-864-4300
Fax: 617-864-9954
Steven Colome
Integrated Environmental Services
92715 University Tower
4735 Royce Road
Irvine, CA 92715
714-854-1167
Fax: 714-854-1840
Michael DiNovi
Chemistry Review Branch
U.S. Food & Drug Administration
200 C Street, SW (MC: FHS-247)
Washington, DC 20204
202-418-3003
Fax: 202-418-3030
Dennis Druck
Environmental Scientist
Center for Health Promotion
and Preventive Medicine
U.S. Army (MCHB-DE-HR)
Aberdeen Proving Ground, MD
21010-5422
410-671-5207
Fax: 410-671-5237
J. Mark Fly
Department of Forestry,
Wildlife, and Fisheries
University of Tennessee
274 Plant Sciences Building
Center Drive
Knoxville, TN 37916
615-974-7126
Fax: 615-974-4714
Larry Gephart
Exxon Biomedical Sciences, Inc.
Mettlers Road (CN-2350)
East Millstone, NJ 08875-2350
908-873-6319
Fax: 908-873-6009
Patricia Guenther
Beltsville Human Nutrition
Research Center
U.S. Department of Agriculture
USDA Center - Room 6C63
4700 River Road - Unit 83
Riverdale, MD 20737
301-734-5618
Fax: 301-734-5496
P.J. (Bert) Hakkinen
Paper Product Development and
Paper Technology Divisions
The Proctor and Gamble Company
6300 Center Hill Avenue
Winton Hill Technical Center
Cincinnati, OH 45224
513-634-2962
Fax: 513-634-3496
Mary Hama
Beltsville Human Nutrition
Research Center
U.S. Department of Agriculture
USDA Center - Room 6C63
4700 River Road - Unit 83
Riverdale, MD 20737
301-734-5617
Fax: 301-734-5496
i Printed on Recycled Paper
(over)
A-3
-------
Dennis Jones
Agency for Toxic Substances
and Disease Registry
1600 Clifton Road (MS: E29)
Atlanta, GA 30333
404-639-6300
Fax: 404-639-6315
John Kissel
Department of Environmental Health
School of Public Health
and Community Medicine
University of Washington (SC-34)
Box 357234
Seattle, WA 98195-7234
206-543-5111
Fax: 206-543-8123
Neil Klepeis
Information Systems and
Services, Inc.
Suite 311
4220 South Maryland Parkway
Las Vegas, NV 89119
702-734-6602
Fax: 702-734-7647
Andrew Persily
National Institute of
Standards and Technologies
Building 226 - Room A-313
Clopper Road
Gaithersburg, MD 20899
301-975-6418
Fax: 301-990-4192
Barbara Petersen
Technical Assessment
Systems, Inc.
1000 Potomac Street, NW
Washington, DC 20007
202-337-2625
Fax: 202-337-1744
Thomas Phillips
Research Division
California Air Resources Board
835 A Street
Davis, CA 95616
916-322-7145
Fax: 916-322-4357
Paul Price
ChemRisk
1685 Congress Street
Stroudwater Crossing
Portland, ME 04102
207-774-0012
Fax: 207-774-8263
John Risher
Division of Toxicology
The Agency for Toxic Substances
and Disease Registry
1600 Clifton Road (E29)
Atlanta, GA 30333
404-639-6304
Fax: 404-639-6315
John Robinson
University of Maryland
3131 Art Sociology Building
College Park, MD 20742
301-405-5734
Fax: 301-314-6892
Peter Robinson
The Proctor and Gamble Company
11810 East Miami River Road
Route 27
Ross, OH 45061
513-627-1474
Fax: 513-627-1908
P. Barry Ryan
Department of Environmental and
Occupational Health
Rollins School of Public Health
Emory University
Room 264
1518 Clifton Road, NE
Atlanta, GA 30322
404-727-3826
Fax: 404-727-8744
Val Schaeffer
U.S. Consumer Product Safety
Commission
4330 East West Highway
Betheseda, MD 20814
301-504-0025
Fax: 301-504-0124
Brad Shurdut
DowElanco
9330 Zionsville Road
Building 206/AZ
Indianapolis, IN 46268-1053
317-337-3806
Fax: 317-337-3214
John Talbott
U.S. Department of Energy
Room 5E-098
1000 Independence Avenue
(MS: EE-421000)
Washington, DC 20585
202-586-9455
Fax: 202-586-1628
Frances Vecchio
Beltsville Human Nutrition
Research Center
U.S. Department of Agriculture
USDA Center - Room 6C63
4700 River Road - Unit 83 *
Riverdale, MD 20737
301-734-5615
Fax: 301-734-5496
A-4
-------
APPENDIX B
PREMEETING COMMENTS
B-l
-------
-------
U.S. Environmental Protection Agency
Peer Review Workshop on Revisions to the
Exposure Factors Handbook
Washington, DC
July 25-26, 1995
PREMEETING COMMENTS
B-3
-------
-------
Work Group #1
Food and beverage consumption
B-5
-------
-------
Barbara Petersen
B-7
-------
-------
.Barbara Petersen - Page 1
COMMENTS ON THE EPA's Risk Assessment Forum
General comments:
The authors have attempted to include a wide variety of
different data sources. It appears that the authors had
little personal knowledge of the national food consumption
surveys and the other data presented in the tables.
Equal weight was given to secondary analyses of
original data (and even to secondary analyses of summary
data) without adequate explanation as to the reason for
reliance of secondary analyses. Prior to publication, the
data tables must be carefully evaluated by experienced
statisticians with specific knowledge of the original
surveys designs, data reporting, etc. This will ensure that
the data are not misused. The data need to be evaluated
rather than simply summarized.
It is also important that each table be revised to
address the following issues: (1) correct expression of the
precision of the results, e.g. appropriate rounding of
numbers, (2) presentation of the uncertainty of each data
point, and (3) expansion of the footnotes to adequately
explain data sources and manipulations as well as
characteristics of the food categories, collapsing of foods
into categories, etc.
In the time available, I have reviewed many of the
tables and .identified some issues. Although I have
summarized some of these below, I am concerned that I may
have missed some very important issues. Nonetheless, I
wanted to provide a general indication of the types of
issues that are of concern.
B-9
-------
Barbara Petersen - Page 2
Water Intake;
• Needs an additional reference to the new USDA Continuing
Survey of Food Intakes and the ability to estimate tap
intake from this survey. (It is possible to use the
factors developed by Ershow and Canter with the newer
USDA surveys with minor additional considerations)
Also needs a comment about the control of water intake.
It seems highly unlikely that individuals continue over
the long run to require 6-11 liters water/day (as noted
on page 2-43). Such amounts are more likely to reflect
short-term intakes before the body's homeostatic
mechanisms can initiate water conservation. A factor for
lengths of time individuals can be expected to maintain
such high levels should be included.
• In the past few years, the quantities of commercial
beverages and bottled water have increased and a source
of such information should be included.
Food Intake Studies — Section 2.3.2;
This section is seriously out-of-date. In addition,
there are several reanalyses of the same data that do not
appear to provide potential users with any unique .
information but do appear to introduce significant sources
for potential misuse of the data. The information that
seems to be required should be obtained by directly
analyzing the original survey and presenting those results
along with references so that users can obtain the raw data
for additional analyses.
USDA is now conducting a continue survey on an annual
basis, called the Continuing Survey of Food Intakes by
B-10
-------
Barbara Petersen - Page 3
.Individuals (CSFII). This data is much more recent than
the survey cited in this chapter. Data about food
consumption in the United States are now available (on CD-
ROM) through 1991-92. Three years can be combined to
provide sample sizes of greater than 10,000 individuals
(almost 30,000 days of food intake data).
The 1987-88 survey had methodological issues which
should be noted in this chapter if data are included from
that survey.
The tables that were developed by EPA's Office of
Pesticide Programs and are presented in detail should be
regarded as historical data. Similar tables should be
generated using current (1989-1992) food consumption data or
the user should be referred to these databases and provided
with methodology to permit the user to compute similar
estimates. Dietary patterns have changed substantially
since 1977-78.
Likewise, the reference by Pao et al. (1982) is based
on data which is almost 20 years old and should be updated
using the more recent food consumption information. It
should also be noted in the text that the categories used by
Pao et al. (1982) do not necessarily capture all of the
consumption of the item since foods in many mixed dishes
were not included. For example, broccoli consumption
includes only those dishes that are primarily broccoli and
would miss the broccoli consumed in mixed dishes. Nor are
all categories of foods included in the report.
The data in table 2-27 is useful for trend analyses but
the user will require similar estimates to be derived from
the 1989-92 CSFII. If this is not possible, a reference
should be included to direct the user to the relevant data
B-ll
-------
Barbara Petersen - Page 4
sources. These data are particularly useful because it
includes foods consumed as part of mixed dishes (which
represent a significant percentage of U.S. fruit and
vegetable consumption).
It appears that the data in Table 2-28, 2-29, 2-30 and
2-32 also do not include fruits and vegetables consumed as
part of a mixed .dish. If this is correct, a footnote should
be added.
On page 2-64, there is a presentation of an analysis by
EPA using USDA NFCS food categories which are presented in
Appendix 2-A. It is unclear whether the analysis was
conducted using the USDA household data or the individual
data. This needs to be carefully explained. If the
household data are used, there needs to be a justification
for why - when individual data are available. While this
appears to include foods that are in "mixed dishes" it is
very likely that this only includes those foods where the
fruit or vegetable is the major component and that it also
includes the quantities of components that are not intended
to be included, e.g. grams of beef in a beef-vegetable stew.
The EPA ORES approach is much more suitable for this type of
an analysis. It is quite simple to combine EPA ORES
individual fruits/vegetables into categories if that is the
additional information that is being obtained through this
approach.
There are significant differences in the estimates of
intake which are obtained using the USDA household data and
those from the individual intake and the user should be made
aware of these differences. Furthermore, it is not a simple
matter to apportion intake among members of the household -
it requires careful development of factors reflecting the
differing ages, activities, etc. This is particularly
B-12
-------
Barbara Petersen - Page 5
difficult when the results are then used to create
distributes of intake and there is a significant likelihood
of distorting the results.
The information presented in Tables 2-32 through 2-73
will certainly be misused without additional explanation of
how the data were generated and the degree of uncertainty;
the treatment of individuals within the households, the
handling of mixed dishes, etc. etc. This data has been
adjusted by body weight and, again, the handling of this key
statistic within a household and the assignment of food
intake/unit body weight needs to be explained. This
analysis should have been done with the individual data
rather than the household data (it may be that the data were
generated using the individual data and that this is
inadequately described in the accompanying text). In sum,
Tables 2-32 through 2-73 should either be revised and fully
explained or deleted from the handbook. (Note that there
are similar tables obtained by the same analysis in other
sections of this chapter and these comments apply to those
Tables as well.)
2.3.2.3.t
These analyses are based on the out-of-date 1977-78
USDA survey. Similar methodology could readily be applied
to the newer data and the user should be so instructed.
Page 2-110;
Canadian Department of National Health and Welfare
Canada Survey.
Newer data are now available for many Canadian
provinces.
B-13
-------
Barbara Petersen - Page 6
Page 2-111;
A footnote should be added indicating that these are on
a dry-weight basis. While this is clear in the text, it is
extremely easy for such a table to be misused and it is
worth the extra effort to associate this fact with the data.
Page 2-114 through 2-177;
The references for this information are entirely
inadequate. Handbook 8 contains multiple entries for each
of these foods and the procedure for extracting such
estimates needs to be documented and accompanied with the
associated uncertainty of the estimates. This table implies
precision to two digits to the right of the decimal point -
when, in fact, the precision is probably no more than +.5%
for most of the commodities listed. Presumably, this
information is to be used in combination with the FDA Total
Diet Study data - yet the data are presented for different
categories of foods.
Table 2-79;
It needs to be reiterated that the data presented in
Pao, et al. do not include all fruit and vegetable
consumption.
The "traditional definition" needs to be carefully
defined.
Page 2-121;
"Upper percentile per capita rates may be calculated
using the consumer only distribution data in Tables 2-32
through 2-73 and the survey size data presented in Section
B-14
-------
Barbara Petersen - Page 7
2.7. I do not believe this is correct. Before this
statement is included, survey statisticians who have worked
with the design and analysis of the USDA data (preferably at
the USDA) should be consulted to ensure that the results are
correctly presented and that the user is. given appropriate
guidance in correct use. The surveys have complex
statistical designs and care must be taken to avoid
misinterpreting the findings.
2.4.1 Intake Studies;
The same comments noted above apply here - the more
recent surveys need to be included.
Likewise, the comments above apply to the information
presented in Table 2-82 through 2-103.
Tables 2-103 through 2-105 reflect disappearance data.
Waste and cooking losses need to be added to these estimates
or the user fully informed as to the potential extent of
these differences. This is particularly important for
animal fats. The degree of precision implied in the
estimates (2 significant digits to the right of the decimal)
is unjustified.
Table 2-107. I do not understand the significance of
the footnotes, e.g. composition of household. Also, it is
not clear whether the amounts of consumption defined the
categories or vice versa.
Table 2-108 would be more appropriate if it were taken
from the nutrient data that accompanies the USDA survey
results, e.g. for the same categories of food as are
reflected as consumed in the survey and which are used for
many of the other tables in this chapter. The degree of
B-15
-------
Barbara Petersen - Page 8
precision in this table is, again, unjustified; estimates of
uncertainty should be included. These do not appear to
match the levels of "trim" that were developed by USDA for
use with the most recent surveys.
Page 2-165;
National Health and Nutrition Examination Survey III
(NHANES III) is apparently mentioned for the first time in
conjunction with food consumption information. A full
description of the survey needs to be added along with
references and a notation of the dates the information was
collected. Also, the response rate needs to be defined more
appropriately, e.g. I believe this only applies to the
response of those individuals who otherwise completed the
NHANES survey (clinical component, etc). The NHANES surveys
provide another useful source of information and results
from these surveys could be included in the handbook.
2.4.4. Recommendations;
Indicates that all results were based on USDA NFCS
data. This is generally true, but it should be noted that
data are included from other sources as well. Also, there
have been many significant modifications of the USDA data -
some with quite surprising impact on estimates of intake.
Also, I note in several places the statement, "the
recommended average intake rates" I do not believe that
USDA recommends an amount of intake of any specific food...
Is the writer, rather suggesting that the "recommended value
to use as an estimate of the average intake???
B-16
-------
Barbara Petersen - Page 9
Table 2-111;
Since contaminant residues will be quite different for
different sources of fat, I wonder what the utility of this
table would be. I would recommend deleting it.
Breast Milk;
I do not feel qualified to accurately comment on this
section. However, given my comments above and the
importance of this information to EPA assessments, I
strongly recommend that this chapter be submitted to a
formal peer review ... by individuals experienced in the
measurement of breast milk intake.
Fish and Shellfish;
On page 2-218, there is a presentation of an analysis
by EPA using USDA NFCS food categories which are presented
in Appendix 2-A. It is unclear whether the analysis was
conducted using the USDA household data or the individual
data. This needs to be carefully explained. If the
household data are used, there needs to be a justification
for why - when individual data are available. While this
appears to include foods that are in "mixed dishes," it is
very likely that this only includes those foods where the
fish/shellfish is the major component and that it also
includes the quantities of components that are not intended
to be included, e.g. grams of fish in a vegetable/fish dish,
etc. The EPA DRES approach is much more suitable for this
type of an analysis. It is quite simple to combine EPA DRES
individual fruits/vegetables into categories if that is the
additional information that is being obtained through this
approach.
B-17
-------
Barbara Petersen - Page 10
There are significant differences in the estimates of
intake which are obtained using the USDA household data and
those from the individual intake and the user should be made
aware of these differences. Furthermore/ it is not a simple
matter to apportion intake among members of the household —
it requires careful development of factors reflecting the
differing ages, activities, etc.
The information presented in Tables 2-141 through 2-145
will certainly be misused without additional explanation of
how the data were generated and the degree of uncertainty;
the treatment of individuals within the households, the
handling of mixed dishes, etc. etc. This data has been
adjusted by body weight and, again, the handling of this key
statistic within a household and the assignment of food
intake/unit body weight needs to be explained. This
analysis should have been done with the individual data
rather than the household data (it may be that the data were
generated using the individual data and that this is
inadequately described in the accompanying text).
Table 2-137 and 2-138;
It seems highly unlikely that the degree of precision
is justified given the conversion from Ib/year.
Table 2-151t
It seems highly unlikely that the degree of precision
is justified.
The reader should be given guidance as to the relevance
of commencement Bay to other U.S. waters or to selected
populations, etc.
B-18
-------
Barbara Petersen - Page 11
Page 2-237;
The study that is referenced in the first full
paragraph is unclear, e.g. "These values are much higher
than the values obtained in this study...?" Is the study
that is referenced U.S. EPA, 1993. According to this
paragraph the U.S. EPA 1993 study is a reanalysis of the
pierce data - is that correct? If so, the discrepancy needs
to be explained.
Table 2-15-51
The degree of precision expressed is unwarranted given
the methodology, e.g. the balsa fish wood model.
Page 2-277;
Are the results expressed on a raw or cooked basis?
Page 2—292. First paragraph:
The conclusion confuses the Pao report with the
original USDA data. Upper percentiles can certainly be
obtained from any of the USDA surveys.
Table 2-181;
Implies far more precision than can be justified by the
underlying data source.
Page 2-303;
Confuses the source of the information about homegrown
food usage. Although there is a detailed description of the
individual component of the USDA survey, that does not
B-19
-------
Barbara Petersen - Page 12
appear to be the source of information about homegrown
vegetables. USDA collects substantially more information
about the farm use of many commodities and EPA should make
arrangements to obtain and use that data.
Tables 2-185 through 2-249;
The estimates of the intake of homegrown food will
probably be one of the more important uses of the handbook.
The selection of categories and the handling of the data
need to be carefully evaluated. I have particular concern
about the generation of percentile for subgroups,
particularly age groups, which were derived from household
data. Table 2-186 is a good examp'le of the undue level of
precision that is applied in these estimates.
The division of the population into regions and then
into ages, seasons, etc. provides extremely small sample
sizes. For example, in Table 2-189, a total of 3 < 01 from
the south region. Yet these data were somehow used to
generate percentiles!H1 I question whether even the mean
is a useful value. Similarly in Table 2-207, percentiles
were generated for Age 06-11 based on a single
individual MM. (It should be noted that the intake
estimates are the same at all the percentiles — thus, a
user would erroneously be estimating intakes). It is also
extremely important that the user not combine intake of
different foods at the upper percentiles.
The values do not appear to be reasonable estimates of
intake from homegrown sources and should be carefully
evaluated. An independent validation using data from
another source is absolutely essential. At a minimum, these
tables need to be edited and all values that have inadequate
sample numbers be removed.
B-20
-------
Barbara Petersen - Page 13
There is a reference to "homegrown exposed fruits"
several places within the chapter. I did not find a list of
the fruits that are included in this category. It is also
important to ensure that the categories are consistently
used in all estimates for exposed fruits, etc.
Soil;
I will defer to other peer reviewers who have worked on
these types of estimates for their comments on this topic.
B-21
-------
-------
Paul Price
B-23
-------
-------
Comments on the June 1995 External Review Draft of
the Exposure Factors Handbook
The 1995 revision of the Exposure Factors Handbook (Handbook) represents a substantial
expansion of the basis on which exposure factors are derived. The Environmental Protection
Agency (EPA) needs to be commended for the level of effort expended in preparing this draft The
topics critical to performing exposure analyses have been addressed and, in general, the Agency
has done an reasonable job of summarizing the available literature on these key exposure factors.
The comments included in this document are divided into two sections. The first section presents
general comments on the overall document and provides suggestions for how the document may be
improved. The second section presents comments by chapter including general comments for the
chapter and specific recommendations for revisions of the Handbook. In cases where important
information was absent from a section, suggestions are made regarding additional references or the
need for additional research.
GENERAL COMMENTS ON THE DOCUMENT
1. Providing Support for Standard or "Default" Distributions for Exposure Factors
A major issue of discussion at the 1992 workshop on the Exposure Factors Handbook was the
extent to which the revised Handbook would provide' support for statistical techniques such as
Monte Carlo Analysis. Specifically, the discussion focused on whether the new Handbook should
provide recommended or default distributions, in addition to the recommended point estimates for
the typical and high percentile individuals. A consensus was not reached at the workshop on this
issue.
The current document reflects this ambiguity. The Handbook provides detailed information on
interindividual variation in factor values as presented in key studies; however, no default
distributions for any of the factors are specified. Typically, the information is presented as a table
of values (consumption rates, duration periods, etc.) for selected percentiles of the surveyed
population. In addition, the document discusses many of the papers by Dr. David Burmaster and
various coauthors, that present distributions of interindividual variation in exposure factors, as well
as the 1992 Guidelines for Performing Exposure Assessment, that emphasizes the importance of
characterizing interindividual variation in exposure. However as stated earlier, the current draft
does not specify a default distribution for any of the factors. For certain factors, such as tapwater
consumption, the Handbook does make a recommendation for using a particular distribution.
However, for other factors such as fish intake, no distribution is recommended.
Not providing a distribution on interindividual variation presents a fundamental contradiction for
EPA because the data required to document values for a typical individual (represented by the
average or median individual) and the high-end individual (represented by a value for an individual
in the top 10 percent of the population) will also be adequate to justify a distribution. For certain
factors EPA may find that there are insufficient data to recommend either a distribution or values
for point estimates for the typical or high-end individual. In such cases, the Agency should
acknowledge the lack of data and simply put forward the results of available studies. However,
the Agency cannot maintain that data are available to specify the point estimates, but not for
describing a distribution.
An appropriate way for the Exposure Factors Handbook to address the issue of distributions of
interpersonal variation would be to identify three data availability categories as a means to classify'
data on each of the various factors. The first category would represent those factors with
sufficient data to allow the development of both point estimates for the typical and high-end
individual and a sufficient number of other percentiles of the distribution to allow Monte Carlo and
B-25
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 2
other statistical techniques to be used. The second category would include those parameters that
had e'nough data that EPA was able to specify point estimates for the typical and high-end
individual. Parameters in this category might have some uncertainty; however the uncertainty
would fall within a sufficiently narrow range that the point estimate would still have meaning,
although data would be insufficient to allow the development of a distribution. Finally, a third
category would contain parameters that the Agency is unable to specify either point estimates or
distributions.
Guidance for developing distributions for factors can be found in Finley et al. (1994) and in the
draft versions of "Developing Distributions for Use in Probabilistic Exposure Assessment" by
Cullen and Frey (draft version expected late 1995).
2. Discussion of Scenario-Based Exposures
The Exposure Factors Handbook of 1989 and the revised draft version provide risk assessors
using scenario-based exposure estimate techniques with default or general guidance on values of
exposure parameters. However, neither document clearly identifies the role of the Handbook in
developing scenario-based exposure assessment techniques. To this end, it is recommended that
the introduction be expanded to clearly identify that the Handbook is intended to provide guidance
for scenario-based exposure assessments. In addition, the introduction should discuss the
strengths and limitations of scenario-based estimates of exposure.
Issues that should be discussed in an expanded introduction include: impacts of simplifying
assumptions in exposure scenarios, uncertainty in the applicability of exposure scenarios to
specific sites of environmental contamination, and the use of scenario-based exposure assessments
in models of interindividual dose rate variations. Although some of these issues are discussed in
the current chapter on uncertainty, they should be mentioned earlier in order to provide a context
for decisions concerning the development of specific factors.
3. Research Needs
While the purpose of the Exposure Factors Handbook is not to present new research findings, but
rather, to summarize the information in the published literature, in several instances relatively
simple analyses on existing data could be conducted to produce distributions of point estimate
exposure factors that will be greatly superior to the information currently available in the published
literature. EPA should give serious consideration to providing support to short-term data analysis
projects which would improve the basis for many exposure factors. Specific examples of such
research needs will be provided in the following sections.
4. ' Need to Differentiate Between Different Types of Studies
In many of the chapters, EPA apparently used a single format for summarizing the results of
published studies. This single format did not differentiate between studies that generated new data
(surveys) and studies which analyzed the existing data. This approach obscured the significance of
many studies. In the chapter specific section, recommendations are made for reorganizing the
relevant studies into various categories.
B-26
-------
ChemRisk-A Division of McLaren/Hart
July 13, 1995
Page 3
5. Additional References
Attached are several references to studies that warrant consideration for inclusion in the Handbook.
6. Discussion of Uncertainty in the Studies
Throughout the Handbook a paragraph is included at the end of each study that dutifully tolls a list
of potential limitations and biases. While such a list may be helpful for individuals looking up an
individual study in the Handbook to obtain a brief overview of the study and its findings, the list of
potential biases is presented without any indication of how the use of the study would be affected
by the individual biases. A more helpful approach would be to group similar types of studies and
discuss the overall implications of the biases common to that particular group.
A larger problem with the listing is that the information on uncertainty is not used in any objective
way in the selection of the recommended point estimate values. EPA should discuss how the
limitations of each study (biases, uncertainties, etc) were taken into consideration in the selection of
recommended values.
7. Significant Figures
All recommendations on point estimates should be limited to the appropriate number of significant
figures. For example four significant figures are given for the typical breast milk ingestion rate and
three figures for inhalation rate.
CHAPTER SPECIFIC COMMENTS
CHAPTER 1 INTRODUCTION
Specific Comments
• On page 1-2 EPA provides guidance for eliminating Part n of the 1989 Handbook. This
decision is a proper one; however, it should be recognized that by proposing specific
equations for scenarios, EPA is still defining a specific method for deriving lexicological
relevant doses for various exposure pathways. Although useful, these methods should not
preclude the use of more sophisticated techniques for modeling, when warranted.
• In the last paragraph on page 1-3, the text lists steps for performing exposure assessments.
The first step, determining pathways of exposure, should include steps 2 and 5. Defining a
pathway of exposure requires identification of the source media by which the contaminant
is transmitted to an exposured population as well as a characterization of the exposed
population.
• EPA states on page 1-6, that the averaging time period for chronic noncancer effects is the
actual period of exposure. This is not an appropriate assumption and is inconsistent with
the following sentence that states that the averaging time should express the dose that is
comparable to the dose response relationship of the effect being evaluated. Many RfDs are
based upon reproductive or systemic effects that occur as a result of exposure over a few
months or years. Where the periods of exposure are significantly longer than the duration
of exposure associated with the adverse effect, the use of an exposure duration will
B-27
-------
ChcmRisk-A Division of McLaren/Hart
July 13,1995
Page 4
underestimate the potential for risk. This issue is not important for estimates of chronic
exposure that do not consider temporal variations in exposure parameters; however, it can
be critical in the evaluation of time-varying sources of contamination.
• On page 1-8, the last sentence in Section 1.2 should be rewritten to read:
"since a different person could be exposed during each of seven sequential 10-year
periods."
CHAPTER 2 INGESTION ROUTE
2.1 Dose Equation for Ingestion
• No comments
2.2 ' Drinking Water Consumption
Specific Comments
• On page 2-15, EPA lists Rosemary and Burmaster (1992) as another study of tapwater
intake that produced raw data. As discussed above, studies which reanalyze existing data
should be evaluated differently than those which present raw data. Specifically, EPA
should indicate whether the new analysis is more useful (than the original study) for
characterizing the values for exposure factors.
• The EPA provides data on page 2-26, on the consumption of raw tapwater from USEPA
(1984d). In certain instances where the chemicals of concern include highly volatile
chemicals such as radon or -vinyl chloride, the potential for exposure from tapwater in
coffee, tea, or reconstituted soups and beverages may not be relevant to the exposure
assessment During the process of beverage or food preparation, such chemicals are likely
to be volatilized from tapwater. In these cases, information on the amount of water
consumed directly (and likely to have the highest potential of retaining the volatile
compounds) is most relevant
• In Section 2.2.2.5., High Activity Levels/High Climates, EPA reviews a series of studies
on the impact of ambient temperature on tapwater consumption rate. This section should be
expanded to consider the information on seasonal and geographic variation in tapwater
consumption rate in the general population as presented by Ershow and Cantor (1989) as
well as other studies based upon the Pennington data. These studies demonstrate mat
regional and seasonal differences have almost no effect on the distribution of tapwater
consumption rates for any age group in the general population. This suggests that the vast
majority of the U.S. population deals with high external temperatures by spending more
time in climate controlled areas or by reducing -activity, not by drinking more water.
The finding from Ershow and Cantor (1989) is not inconsistent with other individual
studies that documented increased needs for fluid intakes under a combination of high
levels of activity and high temperatures. However, it is important to note that high levels of
activity in high temperatures do not typically occur in the general population. Therefore,
only scenarios that specifically assume that the high levels of activity will occur during
B-28
-------
ChemRisk-A Division of McLaren/Hart
July 13, 1995
Page 5
times of high temperature should consider the potential for increased fluid intake. In all
other cases, there appears to be little need for deriving seasonal or geographic-specific
estimates of tapwater consumption for the general population. This point is not well made
in the current document.
• On page 2-41, EPA proposes an estimate for the typical and upper percentile drinking water
consumption rates for adults. In the last sentence, the Handbook also indicates that the
distribution generated by Roseberry and Burmaster (1992) may be used. Is the Agency
endorsing the results of this study as default distribution for tapwater consumption rate? If
so, then why have similar studies not been endorsed for other parameters? Assuming EPA
wishes to endorse a distribution for tapwater consumption rates, EPA should also
acknowledge that the interindividual variation in tapwater consumption rates could be
characterized using the percentiles presented by Ershow and Cantor (1989).
• On page 2-43, EPA presents its recommendations for high activity/hot climates. The
recommendations should be modified to reflect previous comments. It is not appropriate to
recommend that the value of 6-11 liters per day be used to characterize tapwater
consumption rate for the portion of the general population not specifically involved in high
levels of activities.
2.3 Consumption of Fruits and Vegetables
Specific Comments
• In Section 2.3, it would be helpful to provide the distribution of total root crops and total
above ground surface crops that are consumed. Such estimates would be relevant when
evaluating the different exposure pathways that may affect food crops.
2.4 Consumption of Meat, Poultry, and Dairy Products
Specific comments
• On page 2-152, information collected by the USDA's economic research service is
discussed. The USDA also collects information on cattle and other livestock beef that is
slaughtered for home consumption. This information can be used to derive consumption
rates for beef and other types of meat for ranchers and their families, an important
subpopulation for risk assessment The data is collected as part of the yearly agricultural
survey performed on the nation's farms and ranches, although it is not reported in the
USDA annual report on food consumption prices and expenditures. A copy of the survey
can be provided to EPA, if requested. Adjustments to account for home grown beef
consumption from this data would be very useful, because they are based upon annual
reports and would provide a measure of long-term beef consumption.
• The expansion of the Handbook to consider poultry products is commendable.
2.5 Breast Milk Intake
Based upon the data presented by EPA in the Handbook, the mean and upper percentile breast milk
and lipid uptake rates for infants calculated by EPA appear to be reasonable and based on the most
B-29
-------
ChemRisk-A Division of McLaren/Hart
July 13, 1995
Page 6
current science. The available data on this parameter appears to be sufficiently adequate to warrant
including the full distribution of consumption rates in the recommendations. Providing all the
relevant data will enable risk assessors to develop probability distributions for evaluating
uncertainty and to complete probabilistic exposure analyses. Of equal importance, EPA should
provide an interpretive discussion of the recommended values for the breast milk related factors,
including the limitations and uncertainties associated with each, for evaluating infant exposures to
chemicals that have accumulated in breast milk.
Specific Comments
• The Handbook would be greatly improved by providing more comprehensive information
on the distribution of breast milk intake rates for different breast feeding intervals as well as
distributions for time-weighted averages. In addition, the Handbook should identify
distributions for exposure duration, that is, the length of time that infants are breast-fed.
• Exposure duration data should cover not only cover the general population, but should also
be stratified by applicable demographic factors such as geographic location, race, and
socioeconomic status.
• A significant short coming of the draft guidelines is the lack of any discussion regarding the
use of recommended factors to estimate exposure to infants from breast feeding. Such a
discussion is critical to assure that risk assessors use the recommended exposure factors
appropriately and have a clear understanding of the limitations and uncertainties associated
with the data.
• On page 2-185, EPA again treats the Maxwell and Burmaster study (1993)as a study that
presents new data, even though the study is a reanalysis of Duey et al. (1991).
• On page 5-195, the recommendations for breast milk intake should not be given to four
• significant figures. Rather, the numbers should be rounded to 700 and 1000 milliliters per
day for the two point estimates.
2.6 Intake of Fish and Shellfish
General Comments
Overall, Section 2.6 has been greatly expanded over the 1989 Handbook. The expanded section
reflects the fact that fish and shellfish consumption has been the subject of many studies since the
late 1980s.
Unfortunately, the current section reads like a compendium with little synthesis or interpretation.
Because the body of research is now reasonably large, for any particular application there are likely
to be several potentially applicable studies to serve as basis for fish consumption rate estimates.
EPA should provide guidance on which studies are the most relevant for various scenarios. In
addition, judgment as to quality or applicability of studies for exposure assessment is confined to
classifying a study as either "key" or "relevant" and to including a brief advantages/limitations
paragraph in summarizing each study. The reason for classifying any one study as "key" vs.
"relevant" is not given. Li addition, the advantages/limitations comments are not directly
B-30
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 7
comparable across studies to assist in understanding the key vs. relevant classifications. This is a
major flaw in the current document
Specific comments
• A more useful format than the existing Table 2-180 for evaluating the studies for
applicability on a site-specific basis might be a matrix classifying studies by (1) population
type (anglers, general population); (2) survey type (creel, mail, telephone); (3) water body
type and size (marine/estuarine, freshwater, single waterbody, regional coverage); (4) recall
period (length, seasonal coverage); (5) available data (summary statistics only vs. full
distributions, groupings of rates by demographic categories) and some qualitative ratings
for key quality evaluation criteria; (6) design relevance; (7) sample size; (8) response rate;
and (9) representativeness).
• The term "fishermen" should be replaced with the more general term "anglers" throughout.
• The distinction between fish consumption measurement/estimation studies and reanalyses
of measurement/estimation studies published by others should be clearly stated. The latter
can be used for interpreting and applying the former. For example, the Price et al. (1994)
study should be discussed in the section with the Puffer et al. (1981) and Pierce et al.
(1981) studies, and the Ruffle et al. (1994) study should be discussed with the Javitz
(1980) and Rupp et al. (1980) studies. This approach will provide commentary on whether
the reanalyzed study or the original study provides the best basis for deriving parameter
values and distributions.
• Potentially useful information missing from the present draft is a compilation and summary
of the published information on fish/shellfish meal size. Meal sizes ranging from
approximately 120 to more than 250 grams have been used both in measurement/estimation
studies and subsequent risk assessments. While annualized fish consumption rates
sometimes negate the need for meal size estimates, there are often occasions where meal
size is a relevant exposure factor or interpretive tool.
• The discussion of other factors to consider in selecting and using fish/shellfish
consumption rates (e.g., edible parts/portion, preparation methods, losses due to cooking,
lipid normalization, dry to wet weight conversions) should be moved out of the
recommendations section (2.6.9) and into its own section.
• The issue of cooking loss should be discussed in the Handbook. While the degree of
cooking loss is chemical-specific, the frequency of use of cooking methods is not. The
frequency of cooking method is a key issue in evaluating cooking loss and should be
presented in the handbook. Data on relative frequencies of various cooking methods are
available in a number of studies (Connelly, et al. 1990,1992; Chemrisk, 1991).
• On page 2-201, the last paragraph on this page appears to be incorrect. Individuals at or
above the 90th percentile of fishing frequency are by definition frequent fisherman. As
such, they would be expected to contribute more than 10% of the overall fishing effort
from the total population of anglers. The same applies to the median intakes.
B-31
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
PageS
*
• On page 2-273, in EPA's discussion of the Pierce et al. (1981) study, the Agency confuses
the finding of Pierce et al. (1981) with those of USEPA (1993). Pierce did not attempt to
develop estimates offish consumption for individual anglers. EPA developed estimates in
the original Exposure Factors Handbook and in EPA (1993).
• The discussion on page 2-242 of Price et al. (1994) is difficult to follow and fails to
highlight the essential finding of the paper (i.e., that all creel surveys have the potential to
overestimate consumption rates). This finding is glossed over and the study is treated as a
survey of fish consumption rates. Instead, the results of Price et al. (1994) should be
included with the discussion of the Puffer et al. and Pierce et al. case studies and a
discussion of the importance of the bias inherent in the use of creel surveys in developing
recommendations for point estimates for fish consumption rates should be added.
• On page 2-291, EPA's recommendations for marine anglers is inappropriate. As
demonstrated in Price et al. (1994), the values reported by Puffer and Pierce do not
accurately reflect the distributions of consumption rates in the population of anglers.
Rather, they are biased towards reporting the results for anglers that have a high frequency
of using the surveyed bodies of water. As a result, the estimate of the mean or median, as
well as the upper percentiles of the distribution dose rate, overestimate the true values by
one to two orders of magnitude for the median and upper percentiles. Although, this point
is made on page 201, it is not carried through to the recommendations section. Similarly,
while Price et al. (1994) did not evaluate the results from the Santa Monica survey, the
same potential for bias also exists.
Therefore, EPA is encouraged to perform a reanalysis of the Santa Monica survey using the
methodology outlined in Price et al. (1994). A reanalysis will allow the development of
adjusted estimates of the median and 90th (or 95th) pereentile, that can be used as the basis
of the typical and upper pereentile fish consumption rate. The results of the three creel
studies (Puffer et al., Pierce et al., and Santa Monica) after adjustment would be
appropriate guidance for evaluating consumption rates for marine anglers fishing in highly
productive waters. In addition, because the methodology and the survey results produce a
full distribution offish consumption rates, the entire distribution can be recommended as a
distribution for probabilistic analysis.
• The basis for the selection of studies included in the recreational freshwater anglers is
unclear. Several studies which surveyed a larger number of anglers and had a higher
response rate were not included in the list At a minimum, EPA should provide a clear
rationale for including certain recreational angler studies and excluding others.
• On page 2-292, the basis for the Native American freshwater angler estimates do not
include other studies which indicate that other Native American populations consume lower
fish consumption rates. These studies (cited in attached reference list) should also be
discussed and evaluated as a potential consumption rate estimate for Native American
anglers.
• On page 2-293, EPA provides an equation for developing estimates of the consumption of
fat from fish meals. Fish consumption rates in most surveys do not take into account the
reduction in fish size that occurs during trimming or cooking. Therefore, the use of fish
consumption rates in conjunction with levels of fat that reflect as consumed can lead to poor
B-32
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 9
estimates of the total amount of fat consumed in fish meals. Since Table 2-181 presents
many species where the total percent of fat is reported based upon cooked or prepared fish,
there is a significant potential for error.
Section 2.7 Intake Rates for Various Home Produced Food Items
The homegrown intake studies in Section 2.7 are based on a U.S. Department of Agriculture
Nationwide Food Consumption Survey (NFCSs) from 1987-88. These data were used over
earlier NFCS studies because they are believed to be more reflective of current eating patterns in
the US. Using these data, EPA has developed a series of equations that result in homegrown
intake rates for fruits, vegetables, meats, poultry, and dairy products in different regions and for
specific ethnic groups.
Specific Comments
• It is unclear why EPA is assuming that the NFCS surveys should be the only source for
determining homegrown intake rates. In the 1989 Handbook, a variety of studies were
referenced to derive a homegrown percentage for intake rates. The NFCS derivation may
be more definitive, but a comparison to alternative derivations would also be useful. In
addition, the real usefulness of this information is somewhat questionable.
• For adults, the fish consumption mean is about 70 g/day, which is very high in comparison
to mean rates reported in Section 2.6.
• The conclusions to Sections 2.7, as well as 2.3 and 2.4 are quite disappointing. Instead, it
would be more helpful if EPA presented total intake rates in the text similar to the 1989
version. Without these conclusions, the reader must wade through the myriads of tables to
determine total intake of poultry or other food items.
2.8 Soil Ingestion and Pica
Overall, the 1995 Handbook bases its estimates upon a considerably more robust data set than was
available in 1989.
Specific Comments
• In deriving soil ingestion estimates for children, EPA focuses on mean values as measures
of central tendency. However, each of the five data sets that EPA cites in its presentation
of the range of soil ingestion rates is characterized by considerable variability. In such
cases, the median is the more appropriate measure of central tendency, as it is less likely to
be influenced by extreme values in the data set.
• EPA should be commended for its recognition of the extreme value presented by the pica
child in the Calabrese (1989) data set, and the variability of the titanium tracer in all of the
cited studies, as a justification for recommending a soil ingestion estimate of 100 mg/day.
• A conspicuous absence is Calabrese (1992), "What Proportion of Household Dust is
Derived from Outdoor soil?". This paper estimated the amount of outdoor soil in indoor
dust in the Calabrese et al. (1989) study. Based upon the results of this study, Calabrese et
B-33
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 10
al. (1989) recommend that the median outdoor soil ingestion rate be reduced by 35%,
stating that 'for the three most reliable tracers, the median soil ingestion estimates would be
reduced from 29 to 19 mg/day for Al, 55 to 36 nag/day for Ti, and 16 to 190 mg/day for
Zr."
• Several investigators have shown that mouthing behavior declines after the age of three
(Hawley, 1985; LaGoy, 1987; Sedman, 1989). This would lead to reduction in soil
ingestion among children aged 3 to 6. There is no mention of these studies in the soil
ingestion chapter.
• It is unreasonable to base the recommended abnormal soil ingestion estimate of 10-14 g/day
upon data provided by a single child (the pica child in the Calabrese study). Additional
abnormal soil ingestion data should be reviewed in order to derive a more accurate estimate
for pica children.
CHAPTER 3 INHALATION ROUTE
General Comments
This section presents a useful segregation of available studies into those which evaluated short-
term inhalation rates and those which evaluated chronic inhalation rates. EPA has correctly
determined estimates of long-term inhalation rates using the stoichiometric approach to estimating
oxygen needs presented in Layton (1993).
Specific Comments
• On page 3-46, EPA inappropriately recommends using a value of 13.3 cubic meters per
day based upon a simple arithmetic mean of the recent approaches presented in Layton
(1993). Instead, EPA should determine which of the three approaches used by Layton will
provide the most accurate estimate and use that approach to derive a recommended value.
In addition, providing three significant figures for a "typical" inhalation rate*is
inappropriate.
• The recommendations for the upper end of inhalation rates and the inhalation rates for
children are better estimated by using the relationship between body weight and inhalation
' rate developed by Layton to calculate the distribution of inhalation rates in the human
population (see Finley et al., 1994). The result of this approach is a series of age-specific
distributions for inhalation rates which can be used to select both the typical and high
percentile inhalation rate.
• To be consistent, the value for the ventilation rate during heavy exercise given on page 3-2,
paragraph 1 should be in the units of liters/minute.
Footnote C in Section 3.2.2, Table 3-12 incorrectly refers to Tables 3-8 and 3-9. The
tables that should be referred to are Tables 3-11 and 3-10.
In Section 3.2.2, Table 3-13, the "Age" column and the "Adult Male" category are
incorrectly footnoted. The correct footnote is k, not d. In this same table under column
B-34
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 11
labeled "moderate", the value corresponding to the adult female is either incorrectly
footnoted or the value should not be in the table.
• The footnote "h" in Table 3-14 shows moderate activity for males as "moving." The text,
however, states that the males are "mowing." These should be consistent.
• The units in Table 3-15 for VR should read m3/hr, not MP/hr.
• The footnotes a and b in Table 3-16 should be switched.
• It appears as if the conclusions on page 3-33 to the Shamoo et al., study are omitted from
the text. Since the purpose of this study was to demonstrate the accuracy of self estimating
ventilation rates, a discussion regarding the accuracy of self estimating ventilation should
be provided.
• On page 3-36, paragraph 1, the text directs the reader to Table 3-18 for a presentation of
inhalation rates by age, gender, and activity level. However, this is not the data that is
shown in Table 3-18. It appears as if a table is missing from the main body of text If this
is the case, this table should be included.
• The statement made on page 3-45, paragraph 2, comparing the activity of the 13-17 year
old age group to that of the older adults is subjective and should be deleted.
• No backup information is provided on page 3-46 of the Handbook regarding the
justification of the 20 m3/day value from the EPA Ambient Water Quality Criteria
Document The studies used to derive the 20 m3/day inhalation rate for the EPA Water
Quality document should be summarized in either sections 3.2.2 or 3.2.4, with a reference
provided in the text
• The reference to Table 3-11 on page 3-47 should be deleted from the summary table.
• No upper percentile value is provided for infants or children on page 3-47.
• The inhalation rate values for the outdoor worker/athlete for the slow and medium
categories given in Section 3.2.4, page 3-48, paragraph 2 (1.1 m3/hr and 1.5 m3/hr,
respectively) are low when compared with the values for these categories in Tables 3-7 and
3-9. In Table 3-7, the inhalation rate for outdoor workers at a slow and medium pace are
1.26 m3/hr and 1.50 m3/hr, respectively. Similarly, the average of all subjects at a slow
and medium pace as summarized in Table 3-9 are 1.44 m3/hr and 1.86 m3/hr, respectively.
If the lower values are given as exposure factors, further explanation of their derivation
should be provided.
CHAPTER 4 DERMAL ROUTE
Specific Comments
• The study by Phillips et al. (1993) indicated that due to the strong correlation (0.986)
between surface area and body weight the use of surface area to body weight ratios in
exposure assessments is more appropriate than treating each as an independent variable in
B-35
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 12
the equation. This point is important for Monte Carlo analyses of risk and less important
for point estimates. The Handbook should make note of this fact in its discussion of the
study advantages and disadvantages.
• . The revised Handbook gives little attention to the issue of fraction of surface area exposed.
In drawing upon EPA (1992), the Handbook does note that clothing cannot always be
assumed to protect against dermal exposures to contaminants carried on fine dust or in
liquid suspension (e.g., some pesticides). With respect to soil exposures, the Handbook
continues to assume that clothing limits exposure to contaminants in soil.
• The allowance for modifications to the estimates based on climate considerations appear to
be reasonable. Assumptions of 5% for winter, 10% for spring and fall, and 25% for
summer are appropriate defaults in the absence of site-specific information.
• The Handbook should indicate that the fraction of skin exposed is highly dependent on site-
or exposure scenario-specific assumptions and that the proposed values should not be take
as absolute values.
• The section on dermal adherence of soil should include the paper by Finley et al.(1994).
• Due to the comprehensive nature of the study by Kissel et al. (1995), and the fact that it
provides data on actual field conditions, the revised Handbook recommends that this study
serve as the basis for dermal adherence assumptions. While dermal adherence factors for
most activities and body parts were well within the range originally identified by EPA
(1992) (0.2 - 1.0 mg/cm2), the activity termed "kids-m-mud" resulted in adherence values
between 35 and 58 mg/cnA
Risks from dermal exposure are directly proportional to the dermal adherence factor; thus,
risk estimates resulting from the use of the Kissel et al. (1995) study would be greater by at
least a factor of 35 up to a factor of almost 300 than those estimated with the EPA (1992)
range for activities of this type. While the reported values appear to be plausible for the
scenario described. The use of their values in the equation provided in this section may not
result in plausible estimates of dermal exposure. Particularly when a fraction of
contaminant approach is used to characterize the dose rate (see page 4-3). In many
instances, assumptions about the fraction of contaminant that is absorbed is based upon a
scenario of a thin layer of dirt which comes in direct contact with the stratum corneum, in
the absence of an intervening barrier of water. In such situations, nonpolar compounds can
directly defuse from the organic portion of the soil into the stratum corneum and be
available for absorption. This scenario is not appropriate for thick layers of mud. Under
these circumstances, the vast majority of the material would not come in direct contact with
the skin and would not be available for absorption. In addition, where sufficient water is
present a significant barrier may prevent the transport of lipophilic compounds from soils
particles to the stratum corneum. The current document should provide a warning on the
use of such high values for dermal adherence.
B-36
-------
ChetnRisk-A Division of McLaren/Hart
July 13, 1995
Page 13
CHAPTER 5 OTHER FACTORS FOR EXPOSURE CALCULATIONS
5.1. Lifetime
Specific Comments
• The current text inappropriately suggests that gender- and race-related differences in
lifespan may be incorporated in the estimate of the LADD. This may not be appropriate.
Lifetime, is used as a.factor in the derivation of the lifetime average daily dose (LADD). In
this equation, the factor serves as a metric for extrapolating the impact of duration in
toxicological testing across species. As such, the impact of variation in lifespan in different
subpopulations or in individuals is unclear. To the extent that the variation in a
subpopulation reflects genetic-specific differences in lifespan and not differences in
socioeconomic behavior, an argument could be made for use of slightly different lifespans
for different subpopulations (such as men, women, racial groups). However, differences
in lifespan that are due to other factors such as elevated death rates from accidents or other
non-health related factors should not be considered. In addition, the actual age when an
individual dies should not be considered in deriving the LADD. The Exposure Factors
Handbook should include this discussion to avoid errors when placing distributions of
lifetimes in Monte Carlo models of LADDs.
5.2 Body Weight Studies
No comments
5.3. Activity Patterns
Specific comments
• The factor for occupational mobility is inappropriately grouped in the activity pattern
section. It should be placed in a separate section, similar to population mobility;
Additional information on occupational tenure is available from studies by the Department
of Labor. Specific references can be provided, if requested. In addition, a simulation of
occupational tenure was developed in Price et al. (1991).
5.4. Population Mobility
General Comments
• Section 5.4 is insufficient in that EPA has failed to provide any analysis or evaluation of the
studies presented. To improve the usefulness of this section, EPA should include insight
on how the study results should be used in calculating duration for residentially-related
exposures.
Specific Comments
• Section 5.4 would benefit from an additional discussion of how residential duration, as
determined by population mobility, influences various scenarios and how information on
changing residence, changing counties, and changing states, can be use in different
B-37
-------
ChemJRisk-A Division of McLaren/Hart
July 13,1995
Page 14
exposure scenarios. For example, where a source of contamination is related to a specific
home or a specific location, any change in housing location can be assumed to remove the
exposure. As a result, duration is determined by the residential occupancy period. In
contrast, where a source of contamination affects an entire county (such as a wide-spread
air pollution, or the use of a local body of water for fishing) moves that do not result in the
individual leaving a community or area will not result in the ending of exposure.
• Population mobility information from the IRS is available for any county in the U.S. and
should be identified as an additional source of information.
• The use of residential occupancy in determining doses presents a number of complexities
which warrant discussion in Section 5.4. For example, if the purpose of an exposure
assessment is to characterize future doses to a population currently living at an affected site
then a distribution of residential duration should only reflect the future residency starting
from the present. If the scenario assumes that an individual will be exposed during his or
her entire duration of residence the distribution should be based on the period of total
residency. A third distribution should be used if the assessor was interested in
characterizing historical exposures to a population currently living at a site. This
. distribution should be based on the number of years in the past individuals have lived in
their current houses.
• This section also fails to provide any guidance for selecting either point estimates or
distributions of duration.
CHAPTER 6 CONSUMER PRODUCTS
No comments
CHAPTER 7 REFERENCE RESIDENCE
General Comments
Referencing of research papers and reports in Chapter 7 is not consistent with other sections of the
Handbook.
Specific Comments
• Latex and oil based paints should be added as common wall coverings on page 7-8. In
addition, carpeting, waxes, and acrylic floor finish should be'added as common products
used for floor surfaces.
• ' A column should be added to Table 7-4 describing the types of airborne contaminants likely
to be emitted by each of the common products associated with wall and floor coverings.
This information may aide the analyst in determining the potential sources and source
contribution of indoor air pollution.
• The last sentence in paragraph 2 on page 7-10 regarding the strength of stack forces during
the warming season versus the cooling season generalizes that the cooling season indoor-
outdoor temperature differences are not pronounced. This is dependent on what part of the
B-38
-------
ChenxRisk-A Division of McLaren/Hart .
July 13, 1995
Page 15
country is being studied. However, where there is a pronounced indoor-outdoor
temperature difference (i.e., in the higher latitudes), it is likely that the stack effect will be
minimized due to the restricted air flow caused by closed windows perhaps affixed with
additional storm windows and closed doors.
• It should be noted that the exception to diluting indoor air pollution with outdoor occurs
when the outdoor air is the source of the indoor air pollution. If this is the case, increased
mixing of indoor air with outdoor air will not necessarily result in the dilution of
contamination in indoor air (page 7-11, paragraph 1).
••• Values for air exchange rates given in the "Earlier Studies" subsection on page 7-12 are
reported as the geometric mean of the sample population. This indicates that the data in that
population were lognormally distributed. Air exchange rates given in Table 7-5 are
expressed as both the arithmetic and geometric mean of the sample population. It should be
made clear to the reader which value, the arithmetic or the geometric mean, should be used
in assessing the potential air exchange rates within a household.
• An extra open parenthesis is found just prior to the word "volume" on page 7-17,
paragraph 1. This parenthesis should be deleted.
• The variable, V, representing the household volume is expressed as both lower and upper
case. This should be made consistent throughout Section 7.3.3.
• In the example given on page 7-18, the resulting air exchange rate shown as 27.0 m3
should be 27.0 m3 h-i.
CHAPTER 8 ANALYSIS OF UNCERTAINTIES
General Comments
This chapter addresses an important topic that is the focus of much recent research. While the
existing sections provide a discussion of several important aspects, they do not address many of
the issues included in the March 1995 memorandum on risk communication issued by Carol
Browner and the recent literature on methods and applications of uncertainty analysis within
exposure assessment. Recent volumes of Risk Analysis and the Journal of Exposure Analysis and
Environmental Epidemiology provide relevant material as well as the 1990 text Uncertainty by
Morgan and Henrion (1990).
EPA should revise and expand Chapter 8. A suggested outline for the chapter is attached.
Specific Comments
• The @Risk software package (Palisades Corp.) identified on page 8-6, can perform
sensitivity analysis within the probabilistic analysis setting. In addition, analyses that
incorporate dependencies among parameters can be constructed relatively easily either
directly or by age- or gender-stratification. In addition, Monte Carlo analysis offers an
extremely powerful tool for evaluation of nonlinear relationships between factors and dose
rates. Scatterplots of matched pairs of inputs for selected factors and the dose results for
B-39
-------
CfaemRisk-A Division of McLaren/Hart
July 13,1995
Page 16
individual iterations of a Monte Carlo model provide a very useful description of the
interrelationship between any given factor and the dose estimates.
• Monte Carlo analysis can also incorporate information on interdependence in exposure
factors. In addition, intelligent design of distribution selection can allow the effective
modeling of the parameter interaction (see for example Phillips et al. (1993)).
B-40
-------
ChemRisk-A Division of McLaren/Hart
July 13, 1995
Page 17
Additional References for Use in the Preliminary Draft Exposure Factors
Handbook
2.4 CONSUMPTION OF MEAT, POULTRY AND DAIRY PRODUCTS
Texas Agricultural Statistics Service. 1994. Agricultural Survey: January 1, 1994. Texas
Agricultural Statistics Service, Austin, TX.
2.6 INTAKE OF FISH AND SHELLFISH
ATSDR. 1995. Final Report: Exposure to PCBs from Hazardous Waste Among Mohawk
Women and Infants at Akwesasne. Prepared by the Bureau of Environmental and Occupational
Epidemiology, New York State Department of Health and Health Research, Inc. for the U.S.
Department of Health and Human Services, Atlanta, Georgia. PB95-159935. January.
ATSDR. 1995. Final Report: Health Study to Assess the Human Health Effects of Mercury
Exposure to Fish Consumed from the Everglades. Prepared by the Division of Environmental
Epidemiology, Florida Department of Health and Rehabilitative Services, Tallahassee, Florida for
the U.S. Department of Health and Human Services, Atlanta, Georgia. PB95-167276, January.
Connelly, N.A. and T.L. Brown. 1995. Use of angler diaries to examine biases associated with
12-month recall on mail questionnaires. Trans. Am. Fish. Soc. 124:413-422.
Connelly, N.A., B.A. Knuth, and C.A. Bisogni. 1992. Effects of the Health Advisory Changes
on Fishing Habits and Fish Consumption in New York Sport Fisheries. Human Dimension
Research Unit, Department of Natural Resources, New York State College of Agriculture and Life
Sciences, Fernow Hall, Cornell University, Ithaca, NY. Report for the New York Sea Grant
Institute Project No. R/FHD-2-PD. September.
Degner, R.L:, C.M. Adams, S.D. Moss, and S.K. Mack. 1994. Per Capita Fish and Shellfish
Consumption in Florida. Prepared by Florida Agricultural Market Research Center a part of the
Food and Resource Economics Department Institute of Food and Agricultural Sciences, University
of Florida, Gainesville, FL for the Florida Department of Environmental Protection. August 31.
Ebert, E.S., S.H. Su, TJ. Barry, M.N. Gray, and N.W. Harrington. 1995. Estimated rates of
fish consumption by anglers participating hi the Connecticut Housatonic River creel survey. N.
Am. J. Fish. Management (In press)
Finley, B., D. Proctor, P. Scott, N. Harrington, D. Paustenbach, and P. Price. 1994.
Recommended distributions for exposure factors frequently used in health risk assessment. Risk
Anal. 14(4):533-553.
Fitzgerald, E., S.A. Hwang, K.A. Briz, B. Bush, K. Cook, and P. Worswick. 1995. Fish PCB
concentrations and consumption patterns among Mohawk women at Akwesasne. J. Exp. Anal.
Environ. Epid. 5(1): 1-19.
B-41
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995 • • ' • '
Page 18
West, P.C., J.M. Fly, R. Marans, F. LarMn, and D. Rosenblatt. 1993. 1991-92 Michigan Span
Anglers Fish Consumption Study. Prepared by the University of Michigan, School of Natural
Resources for the Michigan Department of Natural Resources. Technical Report No. 6. May.
2.8 SOIL INGESTION AND PICA
Calabrese, EJ. and EJ. Stanek. 1992. What proportion of household dust is derived from
outdoor soH? /. Soil. Contain. l(3):253-263.
Hawley, J.K. 1985. Assessment of health risk from exposure to contaminated soil. Risk Anal.
5(4):289-302.
LaGoy, P.K. 1987. Estimated soil ingestion rates for use in risk assessment. Risk Anal.
7(3):355-359.
Sedman, RJM. 1989. The development of applied action levels for soil contact: A scenario for
the exposure of humans to soil in a residential setting. Environ. Health Perspect. 79:291-313.
3.2 INHALATION RATE
Finley, B., D. Proctor, P, Scott, N. Harrington, D. Paustenbach, and P. Price. 1994.
Recommended distributions for exposure factors frequently used in health risk assessment. Risk
/lnaL14(4):533-553.
4.3 DERMAL ADHERENCE OF SOIL
Finley, B.L., P.K. Scott, and D.A. MayhatL 1994. Development of a standard soil-to-skin
adherence probability density function for use in monte carlo analyses of dermal exposure. Risk
Anal. 14(4):555-57L
5.3 ACTIVITY PATTERN
Price, P.S., J. Sample, and R. Stricter. 1991. PSEM-A model of long-term exposures to
emissions from point sources. In: Proceedings of the 84th Annual Meeting and Exhibition of the
Air & Waste Management Association. Vancouver, British Columbia.
5.4 POPULATION MOBILITY
Price, P.S., J. Sample, and R. Stricter. 1992. Determination of less-than-lifetime exposures to
point sources emissions. Risk Anal. 12(3):367-382.
B42
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 19
Suggested Outline for Chapter 8
I. Introduction
A. Uncertainty in exposure assessment is a key issue
1. Uncertainty is a critical component in the exposure assessment process
2. Uncertainty must be characterized in a form mat enables decision makers to evaluate the
uncertainty in risk findings.
B. Management of uncertainty in different types of exposure assessments (e.g., screening vs.
refined assessments).
II. Presentation of a taxonomy of uncertainty in exposure assessment
A. Definitions for uncertainty and variation
B. Uncertainty in data related to exposure assessment (Section ffl)
C. Uncertainty in exposure factors used in exposure scenarios (Section IV)
D-. Uncertainty in estimates of dose rates derived from scenarios (Section V)
III. Data collection and evaluation
A. Characterization of the quality and/or limitations of data used in residential exposure
assessment
1. Sources of uncertainty
a. Random and systematic errors
b. Measurement errors
c. Analytical limitations
d. Limitations of survey design
e. Dependence and correlation
2. Uncertainty in characterizations of variation
3. Statistical evaluation of data
a. Parametric techniques
(1) Common distributions
(2) Summary statistics
(3) Distribution fitting
(4) Probability plots
b. Non-parametric techniques
IV. Key issues in evaluation of exposure factors
A, Is the factor dominated by variation or uncertainty
B. Time-scale of data and implications for the exposure factor
C. Certainty of factor values at the upper and lower limits of the reported values
D. Choice of parametric or empirical descriptions of uncertainty
E. Uncertainty in distributions of variability
F. Applicability of factors to populations that differ from the sampled population
V. Uncertainty in modeling dose rates
A. Types of models used in exposure assessment
1. Screening assessments
,2. Refined assessment
B. Key issues
1. Differences between modeling uncertainty and addressing uncertainty in data
2. Model structure
B-43
-------
ChemRisk-A Division of McLaren/Hart
July 13,1995
Page 20
3. Errors due to extrapolation of models • .
4. Disagreement between models
C. Techniques for managing modeling uncertainty
1. Model validation
a. Complete data
b. Incomplete data
c. Use of multiple models
2. Attribution of uncertainty in models
a. Sensitivity analysis
b. Scatter plots
c. Linear regression
d. Probabilistic techniques
VI. Propagation of uncertainty in models
A. Available techniques
1. Analytical methods
2. Simulation methods
a. Factorial design
b. Discrete probability distribution (DPD) arithmetic
c. Monte Carlo
d. Latin Hypercube
e. Fuzzy arithmetic
B. Software
C. Examples
VH. Presentation of uncertainty
A. Summary Statistics
1. Variance/Standard Deviation
2. Range
3. Interquartile Range
B. Graphical techniques
1. PDFs
2. CDFs
3. Box plots
4. Two dimensional plots of uncertainty and variation
C. The role of uncertainty in risk and exposure descriptors
B-44
-------
J. Mark Fly
B-45
-------
-------
Fly
These comments are concerned primarily with section 2.6
"Intake of Fish and Shellfish" of Chapter 2 "Ingestion Route" since
these topics are most closely tied to my background and experience.
The comments, however, may be generalizable to other sections of the
handbook. The comments correspond with the issues outlined by Dr.
Wood in his review directive.
1. Are the data presented in a way that is useful to exposure
assessors? Obviously it will be very useful for exposure assessors
to have available in one document a summary of the existing
literature and data on fish and shellfish consumption in the U.S.
This will save a lot of time and effort by eliminating the need for
assessors to start their efforts with a literature review and
possibly additional data analysis. The data are also presented in
a reasonable fashion. If I place myself in the imagined role of an
exposure assessor, however, I wonder how I would ever use these data
to make specific recommendations or to set standards for pollution
or fish consumption that could reasonably be defended in the public
arena or in -court and to not appear quite arbitrary. For these
reasons, Chapter 8 (Analysis of Uncertainties) may be the most
important chapter of all and appears to serve its purpose quite well
although I am no expert on uncertainty and exposure or risk
assessment. If expertise on uncertainty is rather limited
across the country, then it may be helpful to local and state
government, federal agencies, non-governmental organizations and
industry to develop a list of resource people that could be called
upon for consultation. Training programs may be needed to expand
this expertise in the U.S. If it does not already exist, there
appears to be a need'for a professional organization with accepted
standards of practice for dealing with uncertainty in exposure
assessment. The purpose of these standard procedures would be to
reduce the opportunity for administrative or legal challenges to
decisions made related to exposure rates. This would be similar to
standard procedures of medical practice or for playground safety,
for example. Such procedures would help remove the feelings of
neurotic helplessness and consequent indecisiveness that most
decision makers in this arena must face.
Regardless of how the data are presented or the
recommendations made, the handbook should emphasize that the primary
B-47
-------
Fly
consideration needs to be on those segments of the population who
are most at risk. Examples would be anglers rather than the general
population, fish consumers as opposed to non-consumers, and those
who studies indicate consume considerably more fish than the
average person, such as the elderly. Native and African Americans,
and people who fish for subsistence.
2. Have the division of key studies and other relevant studies
been made appropriately? Given the limited number of fish
consumption studies available, one can understand how the division
between key studies and. other relevant studies was made. My
greatest concern about the key studies for recreational freshwater
fish is that four of the five studies used a one year recall period.
Current research in survey methodology on autobiographical memory
raises considerable question about the validity of one year recall
data. The general tendency is for respondents to overestimate their
participation rates, particularly in recreation related activities.
People like to believe that they do more of their favorite
activities than they really do. However, we cannot assume that
recall of fish consumption would err in the same direction. For
these reasons, the National Hunting and Fishing Survey has dropped
their one year recall approach and instituted alternative methods.
The survey is conducted every five years by the U.S. Fish and
Wildlife Service and'the U.S. Census Bureau. Of the key studies,
that leaves only one study (West et al., 1989) with a shorter recall
period (7 days).
3. Do the recommendations represent proper interpretation of
the data? On page 2-285 under recommendations for Chapter 2.6
Intake of Fish and Shellfish, the report states: "Recommendations
for consumption rates were classified into the following categories:
General Population - Per Capita;" etc. What is actually presented
in the report, however, does not appear to be recommendations as
much as purely results from the key and related studies. Perhaps
these results could be interpreted by exposure assessors as
guidelines. Clearly, the report avoids making recommendations
because of the confounding factors affecting any particular
situation, such as locale and differences in rate and type of
consumption by sub-populations.
B-48
-------
Fly
Should the report go beyond merely summarizing study results
and make suggestions on how these results might be used? For
example, the report might describe hypothetical scenarios and
indicate what recommendations would be made based on available data.
If we could map out the key scenarios that might occur across the
country and provide recommendations for each scenario, then we would
be closer to actually having recommendations of practical use.
The greatest problem, however, is that the mean values seem to
be so wide ranging. By merely presenting these results, along with
their limitations, the assessors are likely to throw up their hands
in frustration.
4. Are there suggestions for data gaps and future research
needs? There is a need for more data on the consumption of
recreational freshwater fish using a much shorter recall period than
one year. It would be helpful to convene a panel of experts to
develop survey methodology guidelines for determining fish
consumption that would avoid as many of the limitations as possible
that have been noted in the handbook. It would be more difficult
for future studies that followed the guidelines to be discredited
in terms of their application to exposure assessment.
B49
-------
-------
Peter Robinson
B-51
-------
-------
Peter Robinson, Corp. PfcRS, HSD * (513) 627-0474 BS7/14/95 ©4:04 PM
Preliminary comments on EPA Exposure Factors Handbook
Section 1: Introduction
• Some discussion and a listing of software available; (Doth within EPA and elsewhere) to help in
, exposure assessment would be very valuable, eitier here in the introduction or at the end of
the document. Example software may include:
• CONSEXPO (RIVM)
• THERdbASE (EPA)
• Ordering of sections: perhaps it wouldbe preferable to have Breast Milk Intake following
shellfish and home-produced sections?
Section 2, 2.1: Dose Equation for Ingestion
Section 2.3: Consumption of Fruits and Vegetables
• p. 2.49.5 lines from end: (e.g. some items... 199:2)).
Section 2.4: Consumption of Meat...
Section 2.5: Breast Milk Intake
• Would exposure to infant formula also be a separate issue for the handbook?
Section 2.6: Intake of Fish and Shellfish
Section 2.7: Intake Rates for Various Homeproduced Food Items
Section 3: Analysis of Uncertainty
• p. 8.5: 4 lines from end: This is true only in linear systems (think of averages of fractions
compared with ratio of averages of numerator and denominator)
• p. 8.6, 8.7: The criteria for the selection of models are reasonable, but they should be applied
to the recommended models in the document itself. I'm not sure that this is done. For
example, in the dermal absorption section (Seclon 4), the "nonsteady state approach" is
recommended for estimating the dermally absorbed dose for organics without any discussion
of its validation, and verification status, of how well it represents the situation being addressed,
and without any discussion of plausible alternatives. I am aware of some discussion within the
scientific community of the applicability of this particular model in this case. (I present more
detailed comments on Section 4 separately).
• The American Industrial Health Council Exposure Factors Sourcebook is an invaluable
resource that emphasises parameter distributions, and should be mentioned here.
B-53
-------
Peter Robinson, Corp. P&RS.HSD «(513) 627-0474 BS07/14/95 04:05PM,
Preliminary comments on EPA Exposure Factors Handbook
Dermal Route
• p. 4.2: The "nonsteady-state approach" should be explained. In particular, ot is not
_ mentioned what kinds of information and parameters are required to apply this model in a
realistic manner. I think a referral to the EPA dermal exposure document is not sufficient to
give the reader an idea of what is involved in applying this model. Some discussion of the
current validation status of this model should also be made. Perhaps a discussion of
alternative approaches that may be more useful when certain kinds of data are (or are not)
available would also be most useful.
• Section 4.1 should, I think, be expanded. For e>ample, it would greatly benefit from some
discussion of factors that may affect dermal abso ption. These should at least be mentioned
as "watch-outs" that may affect dermal exposure. Examples may include:
• vehicle effects
• compromised skin
• skin hydratfon
• absorption of compounds from thin film on the skin (the film thickness may be an
important parameter)
• role of the stratum corneum (and other skin components)
• Many of these factors may have a much more important effect on dermaf exposure than some
of the nuances of the "nonsteady-state approach".
B-54
-------
Patricia Guenther
Mary Hama
Frances Vecchio
B-55
-------
-------
Guenther, Hama, and Vecchio
In response to the request for review of the draft Exposure Factors Handbook, we
submit the following comments, organized into two sections: (1) General Comments,
which pertain to large sections of the Handbook and which address the four questions
listed in Dr. Wood's memo dated July 5, 1995, and (2) Specific Comments, which are
presented by section and page number. We focused our comments on sections
relevant to USDA's NFCS data.
GENERAL COMMENTS
DATA PRESENTATION
o We are pleased that Section II has been dropped from the Handbook.
o An overall description of the major USDA surveys should be provided in one
place.
Data from USDA's Nationwide Food Consumption Survey (NFCS) 1977-78 and NFCS
1987-88 and related data sources (e.g. Pao et al. in "Foods Commonly Eaten by
Individuals"; ORES) are described in several sections of the Handbook. Perhaps a
more efficient organization would be to have one section or chapter that includes a
complete description of each survey or data source, including the descriptive
information now found in the text, and the differences among them. The reader could
be referred to that section/chapter as needed. This would eliminate any discrepancies
among repeated survey descriptions, response rates, and so on. (For example, the
NFCS 1987-88 sample size of 4,300 households on page 2-303, paragraph 1 differs
from the value of 4,500 households mentioned in an earlier description of the survey.)
This survey description section/chapter should include a description of the two distinct
components of both the NFCS 1977-78 and NFCS 1987-88: (1) household food use
B-57
-------
Guenther, Hama, and Vecchio
during a 7-day period -about 4,500 households with a response rate of 37 percent,
and (2) individual food intakes by household members for a 3-day period-8,468
individuals, with an estimated response rate of 68 percent of participating households;
10,172 individuals completed a Day 1 recall, an estimated response rate of 81 percent
of participating households. This section/chapter should also include a description of
the data presented by Pao et al., namely the estimated distributions of mean daily
intakes for individuals who consumed the specified food at least once during any given
3-day period. Other survey-based data (e.g. the ORES data) should be described here
also.
o We recommend the use of "household members" throughout, rather than
"family members."
o "Response to questionnaire" in the tables for home-grown food and caught fish
refers to activities undertaken the previous year on a household basis rather
than on an individual basis (tables 2-141 through 2-145, table 2-184, and tables
2-185 through 2-231).
o It should be mentioned that the body weights of individuals were self-reported
(not "actual").
o "Complex foods" should be defined or a few examples given when the term is
introduced.
o Detailed descriptions on how the survey data were used should be provided.
Detailed information should be provided on what data were used in the Handbook and
how they were used, including the assumptions made. When the MFCS data are
referenced, it must be made clear whether the reference is to the household food use
data or the individual intake data. This is especially necessary when the MFCS 1987-
B-58
-------
Guenther, Hama, and Vecchio
88 household data were used to estimate individual intakes. In reality, household food
is not evenly distributed among household members so assumptions made for these
calculations should be clearly stated.
For example, the source of the data in tables 2-141 through 2-145 is unclear,
particularly since the household food use component of the NFCS 1987-88 is not
referenced prior to this section in the present draft. It should be clarified that the data
were collected during a week-long period and then converted to, and reported as,
g/kg/day. As another example, the derivation of daily intake of home-grown foods
(described on pages 2-303 through 2-309) is not a simple activity. We recommend the
inclusion of a short outline of the data variables used and the steps taken in the
calculations. For example, how was the serving size (q,) for an individual within the
age and sex category derived?
Every time food data are described it must be made clear if complex foods were
disaggregated or not. This has not been done consistently. Inclusion of the definitions
of food groups in the Appendix might also be useful.
o Provide an overall description of the data strengths and limitations for use in
exposure assessment.
The report adequately pulls together the food intake resources that are available for
exposure assessment. However, the general strengths and limitations of these data
for use in exposure assessment should be made clearer to the readers, particularly if
specific intake levels are included in the Handbook with the intent that a researcher
can use the data to link to substance concentration data. The following types of
information might be useful to include:
a. The information required from food intake data depends on the substance being
measured (i.e., what the food intake data are being linked to) and the degree of
B-59
-------
Guenther, Hama, and Vecchio
accuracy required by the researcher. For example, for some substances an
accurate exposure estimate may require knowing the specific type of fruit, the
source of the fruit (e.g., home-grown or commercial), the degree of processing
and cooking, the storage conditions, etc., while for other substances some of
this information may not be needed. For some substances, the accuracy of a
specific exposure estimate may be decreased if default recipes are used rather
than coding the specific ingredients reported, or if ingredients from complex
foods are not categorized into their respective food groups, while these factors
may have less of an effect on other substances.
b. To estimate chronic exposure, the distribution of long-term food intakes is
desired.
c. Error [both variable (random) error and bias (nonrandom) error] is introduced
into food intake estimates through nonobservation (i.e. coverage error,
nonresponse error, and sampling error) and observation error (i.e. during data
collection and data processing).
d. Similar sources of error exist in substance concentration data.
e. The food intake-substance linkage should take into account that both food
intake data and substance concentration data are better represented by a
distribution of values, rather than a mean value; i.e., the distribution of
exposures can be better estimated by convolving the two distributions.
f. Error is often introduced into the linkage process when assumptions are made
about the data sources, or when one or both of the sources are modified so
that they are compatible. The food intake data may have been measured at a
greater level of detail, regarding ingredients or preparation methods, than that
of the substance data, so that the researcher is forced to ignore some
information in the food intake database. Alternatively, the food intake data may
be less detailed than that of the substance data, so that the researcher may
need to choose between items in the substance database or use a combination
of items.
g. The error in the final exposure distribution will reflect errors in both the food
B-60
-------
Guenther, Hama, and Vecchio
intake and substance distributions, as well as from the process of linking the
data.
h. Factors that increase the variance of the linked food intake-substance data
beyond the true variance are likely to bias estimates of upper centiles towards
higher values than the true values, and factors that decrease the variance of
the data may bias estimates of upper centiles downward. For example, food
intake data collected for only several days will likely result in an overestimate of
the prevalence of high intakes unless adjusted statistically to correct for intra-
individuai variability.
i. Some researchers believe that risk is affected not only by lifetime substance
exposure, but also by the combination of foods eaten or the pattern of food
intake over time. These issues cannot be addressed easily, but should be
considered in conducting exposure assessment.
If this type of information is considered beyond the scope of this Handbook, the
readers could be provided with a reference instead, such as:
Life Sciences Research Office, Federation of American Societies for
Experimental Biology. 1988. S. A. Anderson (ed.) Estimation of Exposure to
Substances in the Food Supply.
The specific discussions of the strengths and limitations of the NFCS 1977-78 and
MFCS 1987-88 are repeated several times throughout the Handbook. Again, a more
efficient organization might be to provide one section/chapter on the general data
strengths and (imitations, and refer the reader to this as needed.
Where applicable, notations should be included on the tables to indicate the degree of
reliability of the data. For example, each table should alert readers of values based on
samples of less than specific cell counts. Cells with unacceptably low cell counts
should be suppressed. Otherwise, values with small cell sizes are misleading.
B-61
-------
. Guenther, Hama, and Vecchio
DESIGNATION OF KEY STUDIES
We agree with the designation of USDA's Nationwide Food Consumption Surveys as
key studies in this Handbook. Individuals or organizations interested in doing their own
analyses should be referred to the primary data. (References for the NFCS 1987-88
Datasets are listed under "References for Chapter 2.") More recent food intake data
are available from USDA's Continuing Survey of Food Intakes by Individuals (CSFII)
1989-91:
U. S. Department of Agriculture. (1994) Dataset: 1991 Continuing Survey of
Food Intakes by Individuals and 1991 Diet and Health Knowledge Survey. U.S.
Dept. of Commerce, National Technical Information Service, 5285 Port Royal
Rd., Springfield, VA 22161. Accession No. PB94-500063.
U. S. Department of Agriculture. (1993) Dataset: 1990 Continuing Survey of
Food Intakes by Individuals and 1990 Diet and Health Knowledge Survey. U.S.
Dept. of Commerce, National Technical Information Service, 5285 Port Royal
Rd., Springfield, VA 22161. Accession No. PB93-504843.
U. S. Department of Agriculture. (1993) Dataset: 1989 Continuing Survey of
Food Intakes by Individuals and 1989 Diet and Health Knowledge Survey. U.S.
Dept. of Commerce, National Technical Information Service, 5285 Port Royal
Rd., Springfield, VA 22161. Accession No. PB93-500411.
Weights for the combined CSFII 1989-1991 sample are on the CSFII 1991 data tape.
PRESENTATION OF KEY STUDIES
As described under "Data Presentation," the strengths and limitations of the data
sources should be spelled out, preferably before the Recommendations section.
B-62
-------
Guenther, Hama, and Vecchio
RESEARCH NEEDS
Distributions of 1-day food intakes would be an additional analysis useful for assessing
acute exposure.
Research is needed to develop improved statistical methods for conducting risk
analyses. To estimate chronic exposure, the distribution of long-term food intakes or
other behaviors is desired. However, it is not possible to observe long-term behavior
directly with an acceptable degree of accuracy. Thus, research must be conducted to
develop scientifically sound statistical methods for estimating long-term distributions
from short-term observations.
SPECIFIC COMMENTS
SECTION 1:1: BACKGROUND
Page 1-7, paragraph 3, last sentence: This combination of low body weight and high
consumption is likely in the case of some foods.
SECTION 2.2: DRINKING WATER CONSUMPTION
"Water" and related terms (e.g. "drinking water," "tapwater," "source-specific drinking
water") are used inconsistently throughout this section. Although these terms are used
differently by different authors, it would be helpful to the reader if EPA would use these
terms consistently, with clarification on the specific terms and meanings used by
different authors. Definitions could be as follows:
B-63
-------
Guenther, Hama, and Vecchio
Tapwater: Water from the tap, whether filtered or not.
Other sources of water: Bottled, spring, etc.
Drinking water: Water that is drunk alone; it may be tapwater or
other sources of water.
Source-specific Drinking water and water added to
water foods, such as in reconstituting juices, coffee, and
soups; it may be tapwater or other sources of
water.
For example, the following uses of these terms seem inconsistent:
Page 2-2, paragraph 4: "These rates include drinking water consumed in the
form of juices and other beverages containing tapwater (e.g. coffee)." —
"Drinking water" as used includes some food sources of added water. Does
this include reconstituted juices only? Are beverages containing bottled water
included?
Page 2-3, paragraph 2: "However, for the purposes of exposure assessments
involving contaminated drinking water, intake rates based on total tapwater are
more representative of source-specific tapwater intake." - Are non-tapwater
sources of water included in "drinking water?"
Page 2-4, paragraph 1: "Tapwater used in cooking foods" - Does this phrase
refer to water used to reconstitute foods, water used for boiling, etc., or both?
Page 2-10, paragraph 3: Describe how drinking water intake was estimated.
Page 2-37, paragraph 2: The Recommendations section is lacking recommendations.
B-64
-------
Guenther, Hama, and Vecchio
SECTION 2.3 CONSUMPTION OF FRUITS AND VEGETABLES
Page 2-44, paragraph 1: There is no botanical definition of a "vegetable." Delete "not
the botanical definition."
Page 2-47: Are white potatoes correctly listed under raw vegetables?
2.6 INTAKE OF FISH AND SHELLFISH
Page 2-219 Footnote (a) in Table 2-140: The correct reference is USDA Nationwide
Food Consumption Survey 1987-88.
Pages 2-279 through 2-282, table bottom: Should SW be SE (standard error)?
Page 2-279: The last piece of the Source (".pd < (95th) < 194 gpd") seems to be
misplaced.
2.7 INTAKE RATES FOR VARIOUS HOME-PRODUCED FOOD ITEMS
Page 2-303, paragraph 1: Does the sample include only those households that
provided only 1 day of diary data?
Page 2-303 through 2-309: AH assumptions should be described. For example, the
estimates assume that regardless of the sex or age, each household member used
home-grown foods in proportion to the number of meals eaten from the household. A
standard serving size for all individuals within any sex/age category was used.
Page 2-305: How the serving size (q,) for an individual within an age and sex category
B-65
-------
Guenther, Hama, and Vecchio
was derived should be described. Also, an explanation is needed of the values
reported for infant intakes of asparagus and onions, or these values should be
suppressed.
?
Page 2-306: The source reference should be such that it can stand alone. We
recommend citing the survey as follows: USDA Nationwide Food Consumption Survey
1987-88 or, more appropriately, citing the publication.
Page 2-309, paragraph 1: This statement is incorrect; the intake of home-grown dairy
products is not highest for individuals in the South. Suggest instead: "Results of the
regional analyses indicate that intake rates of home-grown fruits, vegetables, and meat
are generally higher for individuals in the Midwest and South than in the Northeast
regions of the United States. Intake rate of home-grown dairy products was also
higher in the Midwest than in the Northeast."
SECTION 2.9 REFERENCES FOR CHAPTER 2
Pages 2-421 and 2-422: The following are corrections to USDA references:
The references listed as USDA (1966) and USDA (1972) are for the same
publication. The correct.citation is:
USDA. (1972) U.S. Department of Agriculture. Food Consumption: Households
in the United States, Seasons and Year, 1965-66, HFCS Rept. No. 12.
USDA. (1979-1992) U.S. Department of Agriculture. Composition of
Foods...Raw, Processed, Prepared. Agriculture Handbook No. 8-1 through 8-
21.
B-66
-------
Guenther, Hama, and Vecchio
USDA. (1983) U.S. Department of Agriculture. Food Consumption:
Households in the United States, Seasons and Year, 1977-78, NFCS Rept. No.
H-6.
U. S. Department of Agriculture. (1991) Dataset: Nationwide Food
Consumption Survey 1987-88 Household Use of Food. U.S. Dept. of
Commerce, National Technical Information Service, 5285 Port Royal Rd.,
Springfield, VA 22161. Accession No. PB92-500016.
A citation is also needed for the NFCS 1987-88 Individual Intake database:
U. S. Department of Agriculture. (1990) Dataset: Nationwide Food
Consumption Survey 1987-88 Individual Intake. U.S. Dept. of Commerce,
National Technical Information Service, 5285 Port Royal Rd., Springfield, VA
22161. Accession No. PB90-504044.
USDA. (1992a) (References should be listed by author.) Lutz, S.M;
Smallwcod, D.M.; Blaylock, J.R.; Hama, M.Y.
USDA. (1992b) and (1993a) refer to the same publication.
USDA. (1993a) (This citation is listed twice; the second reference should be
USDA (1993b).)
APPENDIX 2A
Cover: The title is incorrect; it should read:
Food Codes and Definitions Used in Analysis of the NFCS 1987-88 Data
B-67
-------
Guenther, Hama, and Vecchio
SECTION 8.1 TYPES OF UNCERTAINTY
Page 8-5, paragraph 2 and page 8-6, paragraphs 2 & 3: Suggest adding "Dietary
intake data, for example, are not normally distributed and have heterogenous
variances."
B-68
-------
Work Group #2
Nondietary and dermal
exposure factors
B-69
-------
-------
John Kissel
B-71
-------
-------
J.C. Kissel
EFH Review
Page I
FOODINGESTION
2.6 Intake of Fish and Shellfish
Recent literature - Toy et al. reported a study of fish consumption among native
Americans at the 2nd International Congress on Health Effects of Hazardous Waste in
Atlanta in June 1995. A copy of the abstract is appended to this review.
ACTIVITY PATTERNS
5.3 Activity Patterns
Data gap - No links between physical activity that might lead to soil contact, clothing
worn, and subsequent bathing are available in the existing data. Exposure event duration
hi the dermal soil contact pathway is therefore undefined. Current EPA cooperative
agreement CR 824065-01-0 includes some relevant information gathering.
Comment - Unless the underlying source of figures obtained from Tarshis (1981) can be
identified and validated, that source should be dropped (p. 5-60).
Recent literature - Zartarian et al. (J. Expos. Assess. & Environ. Epid., 5(l):21-34,1995)
presented a videotaping study that raises doubts about the accuracy of questionnaire data.
NONDIETARY AND DERMAL
2.8 Soil Ingestion and Pica
Recent Literature - The reference list in the handbook seems to stop in 1991 on this topic.
The literature is dense and contradictory, but the pathway is too important to to be treated
casually. Calabrese et al. (Human & Exp. Tox., 10:245-249,1991) is referenced in the
text, but cited incorrectly in the reference list. Papers that are not referenced, but should
be, include: two 1991 Calabrese and Stanek papers (Reg. Tox. & Pharm., 13:263-277,
278-292) which provided an alternative (but temporary) interpretation that differs from
the 1989 paper; Calabrese and Stanek (Reg. Tox. & Pharm., 15:83-85,1992) which deals
with relative contribution of outdoor soil vs. indoor dust; two 1994 papers that deal with
detection limits (Stanek and Calabrese, J. Soil Contain., 3(2): 183-189; 3(3):265-270); yet
B-73
-------
J.C. Kissel
EFH Review
Page 2
another interpretation of the Amherst mass balance studies (Calabrese and Stanek,
Envkon. Health Prespec., 103(5):454-457,1995) in which mean ingestion rates for
children are estimated to range from 97 to 208 mg/day for all six tracers and from 97 to
136 mg/day for the three tracers deemed most reliable; a paper (Stanek and Calabrese,
Envkon. Health Perspec., 103(3):276-285,1995) that fits ingestion estimates to a
lognormal distribution which produces very large values in the upper tail: a review paper
Sheppard (Envkon. Monitoring. & Assess, 34:27-44,1995) that includes arguments
based on dermal loadings and hand-to-mouth contact and soil residues on edible plants;
and another review paper (Sedman and Mamood, J. AWMA, 44:141-144,1994). In a
paper that has not yet appeared, Lee and Kissel (Environ. Geochem. & Health, in press)
back calculated soil ingestion rates necessary to explain observed urinary As
concentrations in 2-6 year old children (n=73) living in the vicinity of a smelter using
assumptions regarding background exposures and neglecing dermal and inhalation
exposures. The resulting estimated median soil ingestion rate was 85 mg/day (mean, 261
mg/day). Use of the Davis et al. data (median, 31 mg/day) resulted in signficarit
underprediction of observed urinaiy levels.
4.1 Equation for Dermal Pose
Comment - The units of equation 4-1 do not make sense as written because of confusion
over EV and EF.
DA^t tjga • EV PS*.. ED M • EP ffi • SA ^
ADD[*g>day** BW [kg] • AT [days]
Both EV and EF are defined as event frequency on p. 4-2, but assigned different units.
EF is exposure frequency [days/yr]. EV (the true even? frequency) should have units of
[events/day], not [events/yr]. Redefinition of EF as (effectively) EV-EF for the soil case
is confusing. EF [days/yr] (water case) should be distinguished from EF [events/yr] (soil
case) by calling the latter EF or EVF.
Comment - Recommendation of the "absorption fraction" approach for dermal absorption
from soil should be abandoned. Percent absorbed is a function of soil loading (Duff and
Kissel, J. Tox. & Environ. Health, in press). Computational requirements associated with
extrapolation of the fraction absorbed from experimental to actual conditions are not
simpler than computation of apparent permeability coefficients (which can be generated
B-74
-------
J.C. Kissel
EFH Review
Page 3
from existing data and are therefore no less available than percent absorbed data). Use of
percent absorption has frequently led (erroneously) to direct transfer of laboratory data to
field conditions. Consistent treatment of aqueous and soil media would reduce confusion
and be an improvement. Also, in the dermal case, event duration is not constant and
bioavailability is likely to vary with time of exposure. Careless use of the term
bioavailability in a manner that suggests it is a constant should be avoided.
4.2 Surface area
Data gap - None of the existing surface area models distinguish face and neck from total
head area. Situations arise in which the more limited surface area is the one of interest.
4.3 Dermal Adherence of Soil
General comments - Some reorganization of the literature is required. Studies should be
more clearly distinguished with respect to nature of activity (real or staged) and measure
of soil loading (direct or indirect).
Roels et al. data. - Roels et al. reported lead mass and lead concentration in soil, not soil
mass. The average 159 mg figure was generated by Sedman (1989) by dividing lead
recovered by lead concentration in soil, not by Roels et al. It refers to boys only, and
reflects equal weighting of four study populations of somewhat different sizes. The
corresponding average that can beproduced from the girls' data is 88 mg, and the overall
average is 123 mg. Those figures are not corrected for lead recovery efficiency. Roels et
al. did not report any measure of efficiency. Que Hee et al. (1985) reported that the
absolute efficiency of a single dilute HNOs rinse, using a method that involved the entire
hand, was 45 percent. Roels et al.'s rinse protocol (rinsing -with, not rinsing in 500 ml
dilute HNOs) could not be expected to clean the entire hand. The overall lead recovery
efficiency used to calculate total soil load on the Roels et al. subjects should therefore be
less than 45 percent. A reasonable interpretation of the entire Roels et al. data set might
be as follows:
123 mg 1 1 nn
-- =
This result is the same as that generated by Sedman using the boys only data, an overall
recovery of 60 percent, and a slightly lower hand area. A more recent intepretation of the
B-75
-------
J.C. Kissel
EFH Review
Page 4
Reels et aL data by Finley et al. (Risk Analysis 14(4):555-569, 1994) produced mean and
median values that appear too low and cannot be reproduced using stated assumptions.
Que Hee et al. data. - The protocol did not include shaking of hands (and was limited to
one hand). The data were obtained in experiments in which a subject (described as a
small adult) pressed a hand onto a petri dish containing house dust and then inverted and
reinverted both hand and dish over weighing paper. Mass adhering to the palm was
determined as net mass loss from the dish and paper. Prior interpretation of this data by
Sedman employed an incorrect contact area and is too low. As a result Sedman's lumped
estimate (0.5 mg/cm2) of the Roels et al., Lepow et al., and Que Hee et al. data is about
half what it should be.
Driver et al. data -, (Discussion of particle size effects) Preferential adherence of finer
soil fractions has also been shown by Duggan et al. (Sci. Tot. Environ., 44:65-79, 1985)
and Sheppard and Evenden (J. Environ. QuaL, 23:604-613, 1994) and Kissel et al.
(unpublished). Only the Que Hee et al. data do not show this effect. Those data
represent house dust rather than soil, were limited to a total of six points, and included no
replicates. Selection of finer particles is very likely. The key here is not that some
persons will be exposed to fine soils and will experience greater mass loadings. Size
distributions in real soils are heterogeneous. The important point is that adhering
particles are likely to have different properties (such as greater surface area to volume
ratios) than bulk soils. This has implications for both contaminant concentration and
desorption kinetics.
Yang et al. data - The first sentence under section 4.3.4 on p. 4-28 says that the Yang data
was not included, but the data appears in the summary (Table 4-14). The in vitro
estimate of mass required to produce a monolayer was apparently determined visually
(and presumably without aid of microscope). Nine mg/cm2 appears too high hi light of
electronmicrographs (also of sub-150 |im soil) and calculations presented by Duff and
Kissel (J. Tox. & Environ. Health, in press). In addition, my interpretation of the paper is
that the in vivo tests were done at the same loading for consistency, not that a second and
corroborating measurement of monolayer mass loading was generated.
Kissel et al. data - Three components should be distinguished more clearly. Hand press
experiments were similar to Driver et al.'s work, but include evidence of positive effect of
moisture on adherence and post-adherence soil fractionation. Greenhouse experiments
B-76
-------
J.C. Kissel
EFH Review
Page 5
demonstrated that coverage is uneven and very incomplete on surfaces other than hands.
Field studies demonstrate that loadings vary substantially with activity (so activity pattern
data is needed), that average loadings on hands exceed average loadings on other body
parts within given activity, but that hand loadings are not conservative predictors across
activities. Measures of central tendency in figures in Table 4-12 should be identified.
Other data - Papers by Charney et al. (Pediatrics, 65(2):226-231, 1980), Gallacher et al.
(Arch. Dis. Child., 59:40-44,1984) (mentioned but not discussed or referenced), Duggan
et al. (Sci. Tot. Environ., 44:65-79,1985), and Sheppard and Evenden (J. Environ. Qual.,
23:604-613,1994) can also be used to generate soil adherence estimates.
Recent literature - Finley et al. (Risk Analysis 14(4):555-569, 1994) have proposed a
probability density function for soil adherence based on Monte Carlo sampling of six
distributions generated from data from the prior literature. The published version
includes a very significant misinterpretation of the Que Hee et al. data. Many additional
questions are raised by an implausible claim of universal applicability, failure to support
conclusions with appropriate statistical tests, failure to justify equal weighting of
dissimilar data sets, understatement of uncertainty by inclusion of point estimates, and
use of arguments regarding monolayer loadings that show no familiarity with relevant
loading ranges.
4.4 Recommendations
Comments on Table 4-14 - The Lepow et al. entry should read > 0.5 mg/cm2 since
recovery was undoubtedly less than 100 percent, but was not taken into account because
it wasn't quantified. Schaum's interpretation of the Roel's et al. data (1.5 mg/cm2) is
presented, but not explained in the text. The number of subjects and number of replicates
should be added. Add notation that all figures except Kissel et al. (and Yang et al. if
retained) represent hand data only. It is reasonable to assume that average loadings on
non-hand surfaces are less than hand loadings.
Comment - Final notation (p. 4-35) that more research is needed to deal with
interpretation of specific acitivity loading data could include mention that CR 824065-01-
0 is addressing this question.
B-77
-------
MONDAY, JUNE 5,1995
1:30-3iOOpwi
BREAKOUT SESSION 8
AFish-Consumption Survey of thelulalip and Squaxin Island Tribes
KA. Toy, G.D. Gawne-Mittelstaedt,M.P.A., TulalipTribes DepartmentofEnvironment,
Marysville, N. Polissar, Ph.D, andS. Liao, PhD., Statistics and Epidemiology Research
Corporation, Seattle, Washington
The Environmental Protection Agency (EPA) has adopted criteria for toxic pollutants to protect
human health. These criteria are based on a fish-consumption rate of 6.5 grams per day. This default value
was obtained through a 1973 nationwide survey and did not recognize regional or cultural consumption
patterns. To protect the health of all populations, criteria must be based on sound scientific rationale.
This survey was conducted to determine the fish-consumption rates of two Puget Sound tribes.
Interviews were conducted between February and May of 1994. A total of 263 tribal members, age 18 years
and older, were surveyed. Data were also collected for 77 children from birth to 5 years of age. Information
wasobminedforspecies consumed, fish paits consumed, preparationmethods,sourceoffish,andchildren's
consumption rates. Consumption rates were estimated by age, sex, income, and species groups. Species
groups (anadromous, bottom fish, pelagic, and shellfish) were defined by life history and distribution in the
watercolumn. Fish consumed were primarily from PugetSound. The mean consumption for both tribes was
found to be 10-12 times higher than EPA's default value.
B-78
Monday, June 5 — Page 35
-------
INTERNATIONAL CONGRESS
ON HAZARDOUS WASTE:
IMPACT ON HUMAN AND
ECOLOGICAL HEALTH
ABSTRACTS
June 5-8,1995
Marriott Marquis Hotel
Atlanta, Georgia
»«11"0t,.
U.S. Department of Health and Human Services
Public Health Service
Agency for Toxic Substances and Disease Registry
B-79
-------
-------
David E. Burmaster
B-81
-------
-------
A.155.03 REFH Burmaster
13 July 1995 ,
Comments by:
David E. Burmaster
Alceon Corporation
PO Box 382669
Harvard Square Station
Cambridge, MA 02238-2669
tel: 617-864-4300x222
fax: 617-864-9954
email: deb@Alceon.com
Overall Comments:
First, a disclaimer. I have not had a chance to read all the sections on nondietary
and dermal exposure factors as assigned to our work group, much less the whole
document. Notwithstanding that limitation, I have these preliminary comments.
1. LogNormal Distributions
In my experience, LogNormal distributions appear again and again in exposure
and risk assessments. As a practical matter, most risk assessors do not
understand the power and ubiquity of LogNormal distributions, nor do most risk
assessors understand (i) how to fit LogNormal distributions to data or (ii) how to
manipulate LogNormal random variables in equations.
Dee Hull and I anticipated this need. Last fall, with the Exposure Factors
Handbook in mind, we drafted two essays to fill this perceived gap (copies
attached). The first attachment discusses the three common parameterizations
for LogNormal distributions, and the second one shows how anyone with a
spreadsheet can make LogNormal probability plots, Dee and I hope that US EPA
will reprint these essays as appendices to the main Exposure Factors Handbook.
2. Visualization
Wow! this report is dense, dense, dense with digits in black type. Thumbing
through the report now, I do not see a single graph, plot, or picture -- just oceans
of black type! (Oops, I just found one pie-chart and one histogram.)
Alceon Corporation • PO Box 2669 • Harvard Square Station • Cambridge, MA 02238-2669 • Tel: 617-864-4300
B-83
-------
A.155.03 REFH Burmaster
\.
I strongly urge the Agency and its contractor to add graphs and plots of many of
the data sets and results for two reasons: (i) analysis and (ii) communications.
There are now hundreds of books and reprints that stress the need to visualize
data to understand them and communicate them to botri technical and lay
audiences. As appropriate, I can supply many references in this direction.
If an external risk assessor submitted a report like this to one of the Agency's
Regional Offices concerning, say, for a Superfund site, I dare say the Agency
would reject it as impenetrable --too dense to read, a classic example of poor
risk communications.
3. Parametric Distributions
Looking through the report, 1 see many summary statistics reported by various
researchers ~ e.g., arithmetic means, standard deviations, geometric means,
selected percentiles, minima, and maxima - but I see very few parametric
distributions fit to the data. At first blush, I believe there are more well-fit
parametric distributions than mentioned in the report. For the ones known,
certainly, it is essential to show graphs of the fits, including plots of the residuals
of the fits. I plan to study this issue further in preparation for the meeting in
Washington.
4. Constraints and Dependencies
In the section on food, I see no mention of constraints or dependencies among
the distributions conveyed as tables of digits. For example, given the Agency's
fondness for choosing near-maximum values for many if not all exposure factors
simultaneously, I do not think it is possible for a person to eat all the foods listed
at the 95th percentile of each foods intake rate. For example, I strongly doubt that
a person who eats bread at the 95th percentile of dietary rate also eats rice,
beans, and corn at the 95th percentile of dietary rate as well. After all, there is a
constraint operating on the intakes -- the total calorific intake and its distribution
across the population.
Alceon Corporation • PO Box 2669 • Harvard Square Station • Cambridge, MA 02238-2669 • Tel: 617-864-4300
B-84
-------
A.155.03 REFH Burmaster
5. Variability vs Uncertainty
At first reading, I do not see sufficient distinction between variability and
uncertainty in the report.
As a practical matter, most risk assessors agree that all the variables in an
exposure or risk assessment contain both (i) variability [representing knowledge
of heterogeneity in a well-characterized population, which is usually not reducible
through further measurement or study] and/or (ii) uncertainty [representing
ignorance about a poorly characterized phenomenon or model, which is
sometimes reducible through further measurement or study]. Thus, variability
describes the diversity found in nature, while uncertainty describes our states of
knowledge or ignorance.
I will bring more material on these issues to the meeting in Washington.
6. Computational Issues
So far, I have found no material in the report that discusses -- or gives reference
to --the essential topic of using these distributions in calculations. Having lots of
measurements and summary statistics -- especially with several data sets
reported for a particular phenomenon -- leaves open the question of how to
combine values to estimate: (i) the full distribution (the most useful result), (ii) the
average (much less useful) or (iii) any particular percentile of the distribution (also
much less useful). At the Workshop, I will raise this question.
I attach a copy manuscript that demonstrates that the average risk is usually not
equal to the function of the average value of the input variables.
7. Section 8 - Analysis of Uncertainties
This section is completely inadequate.
It also perpetuates the false statement that Monte Carlo simulations cannot deal
with input variables that have correlations or dependencies among them. The
Alceon Corporation • PO Box 2669 • Harvard Square Station • Cambridge, MA 02238-2669 • Tel: 617-864-4300
B-85
-------
A.155.03 REFH
Burmaster
reverse is true - Monte Carlo simulation is often the only way to work with input
variables that have correlations or dependencies among themselves.
That's all for now See you at the Workshop.
Atcoon Corporation • PO Box 2669 • Harvard Square Station • Cambridge, MA 02238-2669 • Tel: 617-864-4300
B-86
-------
Draft 4
f
A Tutorial on LogNormal Probability Plots
David E. Burmaster, Ph.D. Delores A. Hull, M.S.
Alceon Corporation Alceon Corporation
PO Box 382669 PO Box 382669
Cambridge, MA 02238-2669 Cambridge, MA 02238-2669
deb@Alexandria.LCS.MIT.edu 617-864-4300
1.0 Introduction
This presentation supplements a companion piece titled "A Tutorial on the
LogNormal Distribution" (Burmaster & Hull, 1994).
Statisticians have designed "probability plots" for many kinds of probability
distributions, e.g., normal, lognormal, and exponential distributions, but no
probability plots exist for some distributions, e.g. gamma distributions. For a
general discussion of probability plots, see, e.g., Chapter 1 in Goodness-of-Fit
Techniques (D'Agostino & Stephens, 1986).
LogNormal probability plots have many, many uses in probabilistic risk
assessments precisely because LogNormal distributions occur naturally and are
ubiquitous in probabilistic risk assessments. Figure 1 shows a typical LogNormal
probability plot.
By definition, a probability plot is any 2D graph (with special or transformed axes)
on which values realized from the corresponding probability distribution plot in a
straight line (Benjamin & Cornell, 1970). For example, a set of values that are
randomly sampled from an exponential distribution will plot in a straight line on an
exponential probability plot (or in an almost straight line, given the randomness of
the sample). As another example, data measured from many physical, chemical,
or biological processes follow LogNormal distributions in theory and in practice
(Hattis & Burmaster, 1994).
In this presentation, we teach the reader how to create a LogNormal probability
plot using only a spreadsheet program. As a practical matter, we think all risk
assessors need to know how to plot their own probability plot for three reasons.
First, it teaches important skills. Second, it allows the risk assessor to extend the
25 November 1994 Alceon
B-87
-------
Draft 4
technique to develop and plot data on related graphs, e.g., a CubeRoot
probability plot. Third, it gives the risk assessor a way to correct a flaw in many
commercial statistics programs (e.g., Systat, 1992) that reverse (or transpose)
the axes.
In this presentation, we do not consider making a LogNormal probability plot for a
•set of values or data that include censored or truncated entries, e.g., chemical
concentrations reported as BDL (below the detection limit), although such plots
are sometimes easily accomplished if only a few values are truncated or
censored (see, e.g., Travis & Land, 1990).
2.0 The Functions p(z) and zfp)
2.1 The Function p(z)
Most introductory books on probability or statistics introduce the "standard" or
"unit" Normal distribution with a mean u. = 0 and a standard deviation a= 1. Here,
we write the unit Normal distribution as N(0,1).
For this section, let us assume thsit the random variable Z is distributed as a unit
Normal distribution: Z - N(0,1). The probability density function (PDF) for this
random variable is (Feller, 1968 & 1971; Stuart & Ord, 1987 & 1991):
1 z2
f(z) - -=-.exp[--2-] Eqn1
for -oo ^ z £ +00. This is the familiar bell-shaped curve.
The cumulative distribution function (CDF) for this unit Normal distribution is often
written (Feller, 1968 & 1971; Stuart & Ord, 1987 & 1991):
p(z)
f(x) dx Eqn2
with x as the dummy variable of integration. Figure 2 shows a plot of Eqn 2.
Almost every introductory text on probability and statistics includes a table of this
25 November 1994 Alceon
B-88
-------
Draft 4
integral (Benjamin & Cornell, 1970). The function (z) ranges from a minimum of
0 at z = -oo to a maximum of 1 at z = +00. Some easily memorized values are <£(-
2) = 0.023, O(-1 ) = 0.1 59, O(0) = 0.50, "1(p).
This new function, z(p) -- the inverse of p(z) -. allows us to compute the variable
z associated with each percentile of a unit Normal distribution. With this inverse
function, we want to recover the value z = -1 as corresponding to the 16th
percentile, z = 0 as corresponding to the 50th percentile (the median), and z = +1
as corresponding to the 84th percentile.
Happily, the function z(p) is well defined because the function O(z) has a well
defined inverse function (Feller, 1968 & 1971 ; Stuart & Ord, 1987 & 1991). Figure
3 shows a plot of the inverse function, <3>-1 (p) for most of the domain 0 < p ^ 1 . As
expected, over this domain, the inverse function O'1(p) has a range from -co to
. Note that the inverse function O'1(p) is an odd function:
-1(P) = -0-1 (-p) Eqn 3
3.0 Computing the Function z(p)
To make a LogNormal probability plot, the goal of this Tutorial, we need values
for the function z(p) evaluated at each of the sampled or measured values. There
are generally two ways to do this.
25 November 1994 , Alceon
B-89
-------
Draft 4
First, from standard tables. It Is easy but tedious to read standard tables p(z)
backwards, i.e., to read values for z(p) from tables of p(z) (e.g., Benjamin &
Cornell, 1970).
Second, by computation. Many commercial spreadsheet products and many
other commercial software packages calculate the function z(p). For example, in
Microsoft Excel™ 5.0 for the Macintosh and for Windows (Microsoft, 1994), the
built-in function called NORMSlNV(probability) computes z(p) for -«> < p < +<». In
Mathematica™ (Wolfram, 1991), the user may define a function z(p) in terms of
functions built into the software:
z[pj := Sqrt[2] lnverseErf[2 p -1] Eqn 4
With the basic mathematical formulae available in standard mathematical
handbooks (e.g, Abramowitz & Stegun, 1964), the analyst can evaluate the
function z(p) by knowing the right built-in function or by writing a short subroutine.
Also, Kenneth Bogen (1993) has published a fast intermediate-precision
approximation for z(p).
4.0 Plotting a LogNormal Probability Plot
h
In this Tutorial, we use the symbols xi, xg,.... xn,..., XN to denote a set of N
values sampled (or realized or measured) from a random variable X:. We want to
see if these xn values come from a LogNormal distribution. Even though X is a
random variable, each of the N realizations from it, denoted xn (for n = 1,.... N),
is a point value.
We recommend a 6-step process to make a LogNormal probability plot to
visualize a set of N values xi, xa,..., XN.
Step 1: Sort the N values from the smallest to the largest, so that x'i < xa ^.... <
XN. This presentation allows for some ties among the N values. In the rest of this
algorithm for LogNormal probability plots, we assume that the N values are
sorted from the smallest to the largest.
25 November 1994 Alceon
B-90
-------
Draft 4
Step 2: Check to see if each of the values xn > 0 for n = 1 N. If some values
are zero or negative, Stop, because a 2-parameter Log Normal distribution cannot
fit the data. If all xn are positive, Go to Step 3, because a 2-parameter LogNormal
distribution may fit the data.
Step 3: Take the natural logarithms of the xn values for n = 1,..., N. Work "in
logarithmic space" with the ln[xn] values in all of the remaining steps in this fitting
process. Go to Step 4. [EndNote 1}
Step 4: For each of the N data points, compute an empirical cumulative
probability as:
Pn - " 'N°'5 for n = 1,2,.... N. Eqn 5
This simple formula works well in most cases, but the statistical literature
contains discussions of other formulae for computing the empirical cumulative
probability for use in probability plots.
Step 5: Compute z(pn) for n = 1, 2,..., N. [EndNote 2]
Step 6: Plot the points with coordinates {z(pn), ln[xn]} for n = 1, 2,..., N on a
LogNormal probability plot with z(pn) on the abscissa and ln[xn] on the ordinate. If
the N points plot in a curved line on these axes, Stop, because a 2-parameter
LogNormal distribution cannot fit the data [EndNote 3]. If the N points plot in an
approximately straight line on these axes, Continue, because a 2-parameter
LogNormal distribution will fit the data. Include this graph in your final report.
Some authors (e.g., D'Agostino & Stephens, 1986) and some commercial
software packages (e.g., Systat, 1992) transpose the axes by plotting ln[xn] on
the abscissa and z(pn) on the ordinate.
Next Steps: Complete the additional steps discussed in the companion piece
titled "A Tutorial on the LogNormal Distribution."
25 November 1994 Alceon
B-91
-------
Draft 4
5.0 Discussion
LogNormal probability plots are a powerful technique because they allow the
analyst to see all the data in comparison to a full LogNormal distribution. Data
points falling in a straight line on a LogNormal probability plot imply that a
LogNormal distribution will fit the data with high fidelity (e.g., Figure 1). In such a
situation, the analyst may estimate the two parameters of the best-fit LogNormal
distribution by using ordinary least squares to fit a straight line to the data and to
compute the regression coefficients.
With a LogNormal probability plot, the analyst can see the nature and the quality
of the fit over the whole distribution, and she or he can use any systematic
departures from a fit to investigate other models for the data (D'Agostino et al,
1990). For example, Figure 4 in Brainard and Burmaster (1992) shows how a
systematic curvature of data points plotted on a LogNormal probability plot led to
a new understanding of the distribution of women's body weights;
Traditional GoF tests do not let the analyst visualize the data. With a traditional
GoF test, one or two errant data points may lead to a conclusion that a
LogNormal distribution does not fit the data, but a LogNormal probability plot may
show that the fit is excellent over the range of primary interest.
gndNotes
1. Some authors (e.g., Hattis & Burmaster, 1994) use common logarithms (to the base 10) in making
LogNormal probability plots. This convention is internally consistent, but any parameters estimated
by linear regression on such a plot require conversion if the rest of the analysis uses Napierian
logarithms.
2. Given that z(p) is an odd function, z(pi) = -Z(PN) when pn = "'N' for n -1,2..... N.
3. If the points tend to follow a smooth, nonlinear curve on a LogNormal probability plot, D'Agostino &
Stephens (1986) suggest other types of probability plots to consider. For example, the data may
plot in a straight line on a Normal probability plot, a CubeRoot probability plot, or another
Power-Transformed probability plot.
Acknowledgments
Alceon Corporation supported this work.
25 November 1994 Alceon
B-92
-------
Draft 4
References
Abramowitz & Stegun, 1964
Abramowitz, M. and I.A. Stegun, Eds, 1964, Handbook of Mathematical Functions with Formulas,
Graphs, and Mathematical Tables, National Bureau of Standards, Applied Mathematics Series
Number 55, Issued June 1964, Tenth Printing with corrections in December 1972, US Government
Printing Office, Washington, DC
Benjamin & Cornell, 1970
Benjamin, J.R. and C.A. Cornell, 1970, Probability, Statistics, and Decision for Civil Engineers,
McGraw Hill, New York, NY
Bogen, 1993
Bogen, K.T., 1993, An Intermediate-Precision Approximation of the Inverse Cumulative Normal
Distribution, Communications in Statistics, Simulation and Computation, Volume 23, Number 3, pp
797 - 801
Brainard & Burmaster, 1992
Brainard, J. and D.E. Burmaster, Bivariate Distributions for Height and Weight of Men and Women
in the United States, Risk Analysis, 1992, Vol. 12, No. 2, pp 267 - 275
Burmaster & Hull, 1994
Burmaster, D.E. and D.A. Hull, 1994, A Tutorial on the LogNormal Distribution, Alceon Corporation,
Cambridge, MA
D'Agostinoetal, 1990
D'Agostino, R.B., A. Belanger, and R.B. D'Agostino, Jr., 1990, A Suggestion for Using Powerful and
Informative Tests of Normality, American Statistician, Volume 44, Number 4, pp 316 - 321
D'Agostino & Stephens, 1986
D'Agostino, R.B. and MA Stephens, 1986, Goodness-of-Fit Techniques, Marcel Dekker, New
York, NY
Feller, 1968 & 1971
Feller, W., 1968 and 1971, An Introduction to Probability Theory and Its Applications, Volumes I
and II, John Wiley, New York, NY
Hattis & Burmaster, 1994
Hattis, D.B. and D.E. Burmaster, 1994, Assessment of Variability and Uncertainty Distributions for
Practical Risk Assessments, Risk Analysis, Volume 14, Number 5, pp 713 - 730
Microsoft, 1994
Microsoft Corporation, 1994, Microsoft Excel 5 Worksheet Function Reference, Microsoft Press,
Redmond. WA
Stuart & Ord, 1987 & 1991
Stuart, A. and J.K. Ord, 1987 & 1991, Kendall's Advanced Theory of Statistics, Fifth Editions of
Volumes 1 and 2, Oxford University Press, New York, NY
Systat, 1992
Systat, Inc., 1992, Users Manual, Evanston, IL
Travis & Land, 1990
Travis, C.C. and M.L. Land, 1990, Estimating the Mean of Data Sets with NonDectable Values,
Environmental Science & Technology, Volume 24, Number 7, pp 961 - 962
Wolfram, 1991
Wolfram, S., 1991, Mathematica™, A System for Doing Mathematics by Computer, Second Edition,
-Wesley, Redwood City, CA
25 November 1994 . Alceon
B-93
-------
ln[xn]
Figure 1
A LogNormal Probability Plot
for 51 Random Samples from
X - exp[N(2, 2)]
B-94
-------
P(z)
0.8
0.6
0.4
0.2
0
-4 -2
Figure 2
A Plot of p(z)
0.2 0.4 0.6 0.8
Figure 3
A Plot of z(p)
B-95
-------
-------
Draft 5
A Tutorial on the LogNormal Distribution
David E. Burmaster, Ph.D. Delores A. Hull, M.S.
Alceon Corporation Alceon Corporation
PO Box 382669 PO Box 382669
Cambridge, MA 02238-2669 Cambridge, MA 02238-2669
deb@Alexandria.LCS.MIT.edu 617-864-4300
1.0 Introduction
The lognormal distribution (with two parameters) has a central role in human and
ecological risk assessment for at least three reasons. First, many physical,
chemical, biological, and statistical processes tend to create random variables
that follow LogNormal distributions (Hattis & Burmaster, 1994). For example, the
physical mixing and dilution of one material (say, a miscible or soluble
contaminant) into another material (say, surface water in a bay) tends to create
non equilibrium concentrations which are LogNormal in character (Ott, 1990).
Second, the mathematical process of multiplying a series of random variables will
produce a new random variable (the product) which is LogNormal in character,
regardless of the distributions from which the input variables arise (Benjamin &
Cornell, 1970). Finally, LogNormal distributions are self-replicating under
multiplication and division, i.e., products and quotients of lognormally distributed
random variables are themselves distributed lognormally (Crow & Shimizu,
1988), a result often exploited in back-of-the-envelope calculations.
2.0 Concepts and Notations for Random Variables
In this appendix, we use the symbol^to denote a positive random variable, i.e.,
a variable in an equation that can take any value greater than zero. Here, the
double underscores indicate that ^ is a random variable. The relative frequency
of values sampled (or "realized") from the distribution is governed by a
mathematical function called a probability distribution. We use random variables
described by probability distributions to represent the variability and/or the
uncertainty inherent in a quantity.
15 November 1994 Alceon
B-97
-------
Draft 5
3.0 Symbolic Approach
In this Tutorial, we do not manipulate the probability density function (PDF) or the
cumulative distribution function (CDF) for any distributions (Feller, 1968 & 1971;
Stuart & Ord, 1987 & 1991). Instead, we demonstrate an alternative symbolism,
complete with its own algebra, that makes the concepts and the calculations
easier to understand (Springer, 1979). This abstract symbolism is, of course, not
what a computer does in a numerical simulation with Monte Carlo or Latin
Hypercube sampling. Computer algorithms are beyond the scope of this Tutorial
(see, e.g., Knuth, 1981).
4.0 The Two-Parameter Lognormal Distribution
The 2-parameter lognormal distribution takes its name from the fundamental
property that the logarithm of the random variable is distributed according to a
Normal or Gaussian distribution (Evans et al, 1993):
InjXj ~ N(n, c) Eqn 1
where ln[»] denotes the natural or Napierian logarithm function (base e) and
N(«, •) denotes a Normal or Gaussian distribution with two parameters, the mean
H and the standard deviation a (with a > 0). In Eqn 1, X is a lognormal random
variable, and InfX] is a normal random variable. In Eqn 1, jo. is the mean and a is
the standard deviation of the distribution for the normal random variable In^X], not
the lognormal random variable^- Although sometimes confusing, jj. is also the
median of the normal random variable ln[X] because jj. is the median of N(|i, a).
Many people say that Eqn 1 represents the lognormal random variable^ "in
logarithmic space." As can be seen in Eqn 1, the random variable ln()Q is
distributed normally, but the random variable^ is distributed lognormally.
Figure 1 shows graphs for both the PDF and the CDF for an illustrative Normal
distribution, N(n, o) = N(2,1). In Figure 1, the three dotted vertical lines show the
values of the distribution at x = \i and x = jj±a. As for every Normal distribution,
some 68 percent of the area under the PDF occurs between x = p.-o and x = \I+G.
The information coded in Eqn 1 is identical to the information coded in Eqn 2:
15 November 1994 , Alceon
B-98
-------
Draft 5
X ~ exp[N(n,o)] Eqn 2
where exp[»] denotes the exponential function and N(«, •) again denotes the same
Normal or Gaussian distribution with the same two parameters, mean \i and the
standard deviation a (with a > 0) as above. In Eqn 2,X\s a lognormal random
variable. As earlier, ji is the mean and a is the standard deviation of the normal
random variable ln[XJ, not the lognormal random variable J(. Many people say
that Eqn 2 represents the lognormal random variable^ "in linear space." When
working with Eqn 2 as the representation for a lognormal random variable .X,
many people refer to N(|o., CT) as the "underlying Normal distribution" or "the
Normal distribution in logarithmic space" as a way to remember its origins.
Figure 2 shows graphs for both the PDF and the CDF for the LogNormal
distribution, exp[N(ji, CT)] = exp[N(2, 1)], i.e., the LogNormal distribution for which
the Normal distribution in Figure 1 is the underlying Normal distribution. In Figure
2, the dotted vertical lines show the values of the LogNormal distribution at x =
exp[n] and x = exp[n±a]. As for every LogNormal distribution, some 68 percent of
the area under the PDF occurs between x = exp[|i-a] and x =
Of course, these two alternate representations for a lognormal random variable --
Eqn 1 and Eqn 2 -- contain identical information. For a particular lognormal
distribution, the normal or Gaussian distributions N(p., a) in Eqn 1 and Eqn 2 have
numerically identical parameters. The graphs in Figures 1 and 2, then, show two
ways to visualize a particular LogNormal distribution, exp[N(2, 1 )]. Figure 1
shows the Normal distribution ("in logarithmic space") underlying the LogNormal
distribution ("in linear space") in Figure 2.
5.0 Percentiles of Random Variables ln[XJ and)<
The two random variables InfXI and X are related intimately to each, other by a
common transformation -- either ln[«] or exp[»] ~ depending on the direction of the
transformation. In either the direction, the transformation is 1 :1 and monotonic, so
the percentiles are closely related by the same transforms. For example, the 95th
percentile for X is the exponential of the 95th percentile for ln[Xj, and, in the other
15 November1994 Alceon
B-99
-------
Draft 5
direction, the 95th percentile of ln[XJ is the natural logarithm of the 95th percentile
of X.
{XJo.95 == exp[ {ln[XJ}o.95 ] Eqn 3
Similarly, the median (or 50tn percentile) oiXls the exponential of the median of
InQC], and, in the other direction, the median of ln[XJ is the natural logarithm of the
median of X:
=
{Xjo.50 := exp[{ln[XJ}o.5o] Eqn 4
For example, if the 95tn percentile of InQCj is 4 (i.e., in logarithmic space), then
the 95tn percentile oiXls exp(4) or 54.60 (i.e., when the distribution is converted
to linear space).
More generally, for a Normal distribution, the (1 00 • p)th percentile (0 < p < 1 )
occurs at a z(p), where z(p) is the inverse of the cumulative distribution function
of the standard (or unit) normal distribution. Values for the function z(p) are
widely available in most text books on statistics as tables of the cumulative
distribution function for the standard (or unit) normal distribution (e.g., Benjamin &
Cornell, 1970). For example, here are three values frequently used and easily
remembered; z(0.16) = -1 , z(0.50) = 0, and z(0.84) = +1 .
The (100 • p)th percentile for the underlying Normal distribution may be
calculated as:
{ln[XJ}p = {N(n,a)}p Eqn 5
By extension, the (100 • p)th percentile for the LogNormal distribution may be
calculated as:
(XJp = {exp[ N(ji, a) ]}p Eqn 6
exp[{N(|a, a)}p]
exp[ IL + (z(p) • a) ]
15 November 1994 Alceon
B-100
-------
Draft 5
This last result is particularly pleasing.
Figures 1 and 2 graph a particular Normal distribution, N{2, 1), underlying a
particular LogNormal distribution, exp[N(2, 1)]. The median (or 50th percentile
where z = 0) in Figure 1 is n = 2, and the median in Figure 2 is exp[2] = 7.39. We
know that z(0.16) = -1, so, by Eqns 5 and 6, the 16th percentile of the underlying
Normal distribution occurs at u,-a = 1 and the 16th percentile of the LogNormal
distribution occurs at exp[u.-a] = exp[1] = 2.72. We also know that z(0.84) = +1,
so, by Eqns 5 and 6, the 84th percentile of the underlying Normal distribution
occurs at u,+c = 3 and the 84th percentile of the LogNormal distribution occurs at
exp[u.+a] = exp[3] = 20.09. In addition, we know that z(0.95) = 1.645, so, again by
Eqns 5 and 6, the 95th percentile of the underlying Normal distribution occurs at
u,+(1.645»a) = 3.645 and the 95th percentile of the LogNormal distribution occurs
at exp[u.+(1.645«o)] = exp[3.645] = 38.28. Thus, Figures 1 and 2 show two
alternative ways to visualize the same LogNormal distribution.
6.0 Arithmetic Central Moments of Random Variables ln[X} and )(
The first two arithmetic central moments for the Normal random variable ln[Xj are
straightforward:
AMean[ln[Xl] = AMean[ N(n, c) ] Eqn 7
u,
AStdDev[ InlXJ ] = AStdDev[ N(n, a) ] Eqn 8
= a
Here, the notation AMean[»] refers to the arithmetic mean of a random variable,
more property the expected value calculated by the expectation operator, £[•].
The notation AStdDev[«] refers to the arithmetic standard deviation of the random
variable.
The first two central moments for the LogNormal random variable^ are more
complicated and not easily derived. They are:
AMean[>(] = AMean[ exp[ N(n, a) ] ] Eqn 9
15 November1994 Alceon
B-101
-------
Draft 5
AStdDev[ exp[ N(n, o) ] ]
•= exp[ [i ] • V exp[ a2 ] • ( exp[ a2 ] - 1 ) Eqn 1 0
For the LogNormal distribution shown in Figure 2, the arithmetic mean is 12.18
and the arithmetic standard deviation is 1 5.97.
7.0 Geometric Moments of Random Variable _X
The first two geometric moments of a positive random variable >£ are defined as:
GMean[^] = exp[ AMean[ ln[Vj ] ] Eqn 11
GStdDev[V] = exp[ AStdDev[ ln[^ ] ] Eqn 12
where GMean[«] denotes the geometric mean of a positive random variable and
GStdDev[»] denotes the geometric standard deviation of a positive random
variable.
When applied to Eqn 2, these formulae yield:
GMean[^] = exp[n] Eqn 13
GStdDevf^] . = exp[o] Eqn 14
Thus, for LogNormal distributions, the median of ^ equals the geometric mean of
2£. Note that the arithmetic mean of a LogNormal distribution is always greater
than the geometric mean of the distribution.
8.0 Different Ways to Parameterize the LogNprmal Distribution
Fundamentally, it takes two and only two parameters to describe a particular
LogNormal distribution. There are an infinite number of ways to pick the two
values. First, the analyst could pick two parameters in "logarithmic space," two
parameters in "linear space," or one in each. Second, the two parameters chosen
15 November 1994 Alceon
B-102
-------
Drafts
could be two arithmetic moments, two geometric moments, two percentiles, or
one of each of two types. With some effort, it is generally possible to convert one
representation of a particular lognormal distribution to another representation for
the same distribution. After all, the particular lognormal distribution remains the
same, only the parameterization changes from one representation to another. We
have seen many different parameterizations in the literature, and we have seen
some authors use several different parameterizations in the same article. Given
the infinite number of representations for just one lognormal distribution, the
possibilities for confusion and misunderstanding and mistakes are boundless.
In this Tutorial, we emphasize the central importance of (i and
-------
Draft 5
£ ~ 10A[N(mo,aio)] Eqn 2'
where logioM denotes the common logarithm function (base 10), 10A[»] indicates
the number 10 raised to a power, and N(»,») denotes a Normal or Gaussian
distribution with two parameters, the mean mo and the standard deviation a-io-
The information coded in Eqn 1' is identical to the information coded in Eqn 2'. In
Eqns 1' and 2', we have used subscripts on the parameters to indicate the use of
common logarithms.
The fact that some authors use common logarithms (instead of Napierian
logarithms) introduces another dimension of confusion. Without giving the full
derivations, there are some convenient formulae to convert from the
parameterization in common logarithms to Napierian logarithms:
p. = ln[10]*|o.io Eqn 15
a = ln[10]«aio Eqn 16
GMean[X] = 10A[niol Eqn 17
GStdDev[^] = 10A[aioJ Eqn 18
With these conversions in place, the reader may now convert among the four
most common but different parameterizations of a particular lognormal
distribution.
9.0 A Constant Times a LogNormal Distribution
In many hilman or ecological risk assessments done in a probabilistic framework,
the risk assessor must multiply a lognormai distribution^ by a constant c, say, for
example, to convert from one set of units to another. To begin, we set c1 = ln[c].
Then
c»>< ~ c • exp[ N(|i, a) ] Eqn 19
exp[c']«exp[N(|i, a)]
15 November 1994 Alceon
B-104
-------
Draft 5
exp[ c' + Nfoi, a) ]
exp[ N(|i + c1, a) ]
Thus, in this symbolism, the multiplication of a LogNormal distribution by a
constant shifts the mean jj, of the underlying Normal distribution by c' = ln[c], but
the operation does not change the standard deviation
-------
Draft 5
where all inputs are positive random variables, Xj (for i = 1,..., I) and Yj (for j = 1,
.... J).
In the special case in which all theXj and Yj are independent LogNormal random
variables, R is also a lognormal random variable:
R ~ exp[N(u.R, OR)] Eqn23
with
Eqn24
-------
Draft 5
distribution to a data set with censored or truncated values, e.g., chemical
concentrations reported as BDL (below the detection limit), although such fits are
sometimes easily accomplished.
Step 1: Check to see if each of the values xn > 0 for n = 1 N. If some values
are zero or negative, Stop, because a 2-parameter LogNormal distribution cannot
fit the data. If all xn are positive, Go to Step 2, because a 2-parameter LogNormal
distribution may fit the data.
Step 2: Take the natural logarithms of the xn values for n = 1,.... N. Work "in
logarithmic space" with the ln[xn] values in all of the remaining steps in this fitting
process. Go to Step 3.
Step 3: Plot a histogram of the ln[xn] values. If the histogram of the ln[xn] values
is asymmetric by having a long tail to the left or the right, Stop, because a 2-
parameter LogNormal distribution cannot fit the data. If the histogram of the ln[xn]
values is symmetric, Go to Step 4, because a 2-parameter LogNormal
distribution may fit the data.
Step 4: Plot a LogNormal probability plot with z(p) on the abscissa and ln[xn] on
the ordinate. If the N points plot in a curved line on these axes, Stop, because a
2-parameter LogNormal distribution cannot fit the data. If the N points plot in an
approximately straight line on these axes, Go to Step 5, because a 2-parameter
LogNormal distribution will fit the data. Include this graph in your final report.
Some authors (e.g., D'Agostino & Stephens, 1986) and some commercial
software packages (e.g., Systat,. 1992) transpose the axes by plotting ln[xn] on
the abscissa and z(p) on the ordinate.
Step 5: Using ordinary least-squares regression, fit a straight line to the data
plotted on the LogNormal probability plot with z(p) on the abscissa and ln[xn] on
the ordinate. The line will have this functional form, with z as the independent
variable in the regression:
line = a + (b • z) Eqn 26
15 November 1994 Alceon
B-107
-------
Draft 5
where a is the intercept of the fitted line when 2 = 0 and b is the slope of the fitted
line. Include this graph in your final report, along with all the goodness of fit
statistics for the regression. Then, p. = a is a good estimate for the parameter [i
in Eqns 1 and 2 and a = b is a good estimate for a in Eqns 1 and 2. Usually the
regression package will report confidence internals for a and b. Go to Step 6. In
this Step 5, a regression line fit to the transposed LogNormal probability plot with
In[xn] on the abscissa and z(p) on the ordinate will not give correct estimates for
jiand a because the regression does not have the proper independent variable.
Step 6: Calculate the values of these two estimators to obtain alternate estimates
of parameters \L and CT:
Eqn27
£(ln[xr,]-ln[x])2 _
j——- tqn 28
Then, ji = ln[x] is an alternate good estimate for the parameter |i in Eqns 1 and
2 and a = s is an alternate good estimate for a in Eqns 1 and 2. If the alternative
estimates of ji from Steps 5 and 6 are numerically close to each other, AND if
the alternative estimates for d from Steps 5 and 6 are numerically close to each
other, go to Step 7.
Step 7: Do one or more goodness of fit (GoF) tests (Madansky, 1988; D'Agostino
& Stephens, 1986) on the ln[xn] values to see if they do or do not fit a Normal
distribution. Even though these methods do not visualize the data and are not as
robust as the probability plot above, discuss the results of these tests in your final
report. Go to Step 8.
Step 8: Discuss the adequacy of the fit compared to the use of the LogNormal
distribution in a narrative in your final report. Note any outliers, problems, or
issues. State the conditions and circumstances in which the results apply; also
state the conditions and circumstances in which the results do not apply. Discuss
alternative fits and conduct numerical experiments to see if use of an alternative
fit would lead to a different decision in the real world.
15 November1994 Alceon
-------
Draft 5
Discussion:
After the initial exploratory data analysis and data visualization, we recommend
an 8-step process for fitting a LogNormal distribution to data. First, we
recommend that the analyst work with the ln[xn] values to fit the parameters ji
and a of the underlying Normal distribution -- precisely because working with the
untransformed xn values is numerically unstable in most cases. Second, we
recommend that the analyst complete all 8 steps in entirety - precisely because
we have seen egregious mistakes when an analyst ignores a particular step.
Third, visualize! visualize!! visualize!!! in each step in the procedure. These 8
steps form the framework of many publications in the refereed literature (e.g.,
Roseberry & Burmaster, 1992; Murray & Burmaster, 1992)
Although we have found that these 8 steps work well for many univariate data
sets and for the marginal distributions of many multivariate data sets, the
methods will not work to fit a multivariate distribution to multivariate data that may
include non negligible correlations and/or dependencies. Finally, although this
recommended 8-step process rests on powerful and recognized statistical
techniques with long pedigrees -- I.e., probability plots, the method of moments,
and the method of maximum likelihood - there are other powerful and accepted
techniques not included - e.g., maximum entropy methods (Kapur & Kesavan,
1992) and model-free curve estimation (Tarter & Lock, 1993).
12.0 Numerical Simulations with LogNormal Variables
When a person is first starting a numerical simulation with LogNormal random
variables, we recommend a two-step process.
First, generate or simulate values for InfX] by drawing values from the underlying
Normal distribution N(ji, c) in logarithmic space. Second, exponentiate those
values for ln[X] to obtain values for^Xfrom the LogNormal distribution
exp[N(|o., c)] in linear space.
This two-step process basically reverses the 8-step fitting process just presented
in Section 11.0 above. For example, when using a commercial software product
15 November 1994 Alceon
B-109
-------
Draft 5
such as Crystal Ball™ (Decisioneering, 1992) or@Risk™ (Palisade, 1992) in
conjunction with a spreadsheet on a desktop computer, the analyst would
simulate the underlying Normal distribution, N(|i,
-------
Draft 5
Acknowledgments and Dedication
Alceon Corporation supported this work.
We dedicate this Tutorial in memory of Jerome Bert Wiesner.
15 November 1994 Alceon
B-lll
-------
Draft 5
References
Benjamin & Cornell, 1970
Benjamin, J.R. and C.A. Cornell, 1970, Probability, Statistics, and Decision for Civil
Engineers, McGraw Hill, New York, NY
Bogen, 1992
Bogen, K.T., 1992, RiskQ: An Interactive Approach to Probability, Uncertainty, and
Statistics for Use with Mathematica, Reference Manual, UCRL-MA-110232 Lawrence
Livermore National Laboratory, University of California, Livermore, CA, July 1992
Brainard & Burmaster, 1992
Brainard, J. and D.E. Burmaster, Bivariate Distributions for Height and Weight of Men and
Women in the United States, Risk Analysis, 1992, Vol. 12, No. 2, pp 267-275
Cleveland, 1993
Cleveland, W.S., 1993, Visualizing Data, AT&T Bell Laboratories, Hobart Press, Summit,
NJ
Cleveland, 1994
Cleveland, W.S., 1994, The Elements of Graphing Data, AT&T Bell Laboratories, Hobart
Press, Summit, NJ
Crow & Shimizu, 1988
Crow, E.L. and K. Shimizu, Eds., 1988, Lognormal Distributions, Theory and Applications,
Marcel Dekker, New York, NY
D'Agostino & Stephens, 1986
D'Agostino, R.B. and M.A. Stephens, 1986, Goodness-of-Fit Techniques, Marcel Dekker,
New York, NY
Decisioneering, 1992
Decisioneering, Inc., 1992,, Users Manual for Crystal Ball, Denver, CO
Evans et al, 1993
Evans, M., N. Hastings, and B. Peacock, 1993, Statistical Distributions, Second Edition,
John Wiley & Sons, New York, MY
Feller, 1968 & 1971
Feller, W., 1968 and 1971, An Introduction to Probability Theory and Its Applications,
Volumes I and II, John Wiley, New York, NY
Hattis & Burmaster, 1994
Hattis, D.B. and D.E. Burmaster, 1994, Some Thoughts on Choosing Distributions for
Practical Risk Assessments, Risk Analysis, in press
Kapur & Kesavan, 1992
Kapur, J.N. and H.K. Kesavan, 1992, Entropy Optimization: Principles with Applications,
Academic Press, Harcourt Brace Jovanovich, Boston, MA
Knuth, 1981
Knuth, D.E., 1981, The Art of Computer Programming, Seminumerical Algorithms,
Volume 2, Second Edition, Addison-Wesley, Reading, MA
Lumina, 1993
Lumina Decision Systems, 1993, Users Manual for DEMOS™, Los Altos, CA
15 November 1994 Alceon
B-112
-------
Draft 5
Madansky, 1988
Madansky, A., 1988, Prescriptions for Working Statisticians, Springer-Veriag, New York,
NY
Murray & Burmaster, 1993
Murray, D.M. and D.E. Burmaster, 1993, Review of RiskQ: An Interactive Approach to
Probability, Uncertainty, and Statistics for Use with Mathematica, Risk Analysis, Volume
13, Number 4, pp 479 - 482
Murray & Burmaster, 1992
Murray, D.M., and D.E. Burmaster, 1992, Estimated Distributions for Total Body Surface
Area of Men and Women in the United States, Journal of Exposure Analysis and
Environmental Epidemiology Volume 2, Number 4, pp 451 - 461
Ott, 1990
Ott, W.R., 1990, A Physical Explanation of the Lognormality of Pollutant Concentrations,
Journal of the Air and Waste Management Association, Volume 40, pp 1378 et seq.
Palisade, 1992
Palisade Corporation., 1992, Users Manual for@Risk, Newfield, NY
Roseberry & Burmaster, 1992
Roseberry, A.M., and D.E. Burmaster, 1992, Lognormal Distributions for Water Intake by
Children and Adults, Risk Analysis, Volume 12, Number 1, pp 99 -104
Springer, 1979
Springer, M.D., 1979, The Algebra of Random Variables, John Wiley & Sons, New York,
NY
Stuart & Ord, 1987 & 1991
Stuart, A. and J.K: Ord, 1987 &.1991, Kendall's Advanced Theory of Statistics, Fifth
Editions of Volumes 1 and 2, Oxford University Press, New York, .NY
Systat, 1992
Systat, Inc., 1992, Users Manual, Evanston, IL
Tarter & Lock, 1993
Tarter, M.E. and M.D. Lock, 1993, Model-Free Curve Estimation, Chapman & Hall, New
York, NY
Tukey, 1977
Tukey, J.W., 1977, Exploratory Data Analysis, Addison-Wesley, Reading, MA
15 November 1994 Alceon
B-113
-------
° [o+tljdxa
CO
o
10
o
o
[o+Ti]dxa
[Tl]dxa
[o-Tl]dxa
c\j
O)
-------
Larry Gephart
B-115
-------
-------
Larry Gcphart
General Comments
EPA should be commended for efforts the agency has undertaken to update the Exposure
Factors Handbook. The new draft Handbook contains significant new information on exposure
factors. Also, both new and old data are presented in a manner which allows for better use of the
data in exposure assessments.
To assist users of the handbook, it is recommended that for all exposure factors, central
tendency values (e.g., medians), defined upper percentile values (e.g., 90th percentiles), and where
possible, distributions should be presented. This would allow exposure assessors to tailor the use
of exposure factor data in specific risk assessments.
It is recommended that EPA consider removing the policy concept of "default" from the
Exposure Factors Handbook. For most exposure factors, data are sufficient to define a central
tendency and an upper percentile value. However, reference to "default" values still receives
favorable attention in the draft revised handbook. For example, current data support changing the
default child soil ingestion rate of 200 mg/day to a central tendency value of 50-100 mg/day, the
default adult soil ingestion rate of 100 mg/day to a central tendency value of 25-50 mg/day, the
default adult daily inhalation rate of 20 m3/d to a central tendency value of 13 m3/d, and the adult
life expectancy from 70 years to 75 years.
Section Specific Comments
Section 2.2 - Drinking Water Consumption
1. Are data presented in a way that is useful to exposure assessors?
In general, data are presented in way that will facilitate their use in exposure
assessments. For example, data from table 2-7 will allow for preparation of exposure
distributions. Data from other tables will allow for exposure assessments for certain unique
s.
situations (e.g., as a function of physical activity, sex, or geographical area).
B-117
-------
Larry Gcpharl
It is recommended that section 2-2 should not begin with a statement concerning the
drinking water consumption value currently used by EPA (pg 2-3). The term "default" value
should be deleted, or, at least defined.
Drinking water intake rate and tapwater intake rate are used to define the same
exposure factor. This is potentially confusing to the reader. It is recommended that the latter
term be used, since it better meat!; the definition provided.
For purposes of risk assessment, it is useful to maintain separate figures for tapwater
intake, which is defined to include food and beverages reconstituted with tapwater, and for
total fluid intake, which is defined to include consumption of commercial products. However.
the final recommendations do not maintain this separation.
It is not clear what is meant by "upper percentile tap water intake" on pg 2-3 (i.e., is
this a 95, 99 percentile value?). Similarly, on pg 2-3, it is not clear what is meant by the statement
"the data tend to support EPA's use of 2 L/day for upper percentile tapwater intake."
2. Have the key studies been identified?
The key studies have been identified and emphasis has been placed on making
recommendations from the key studies (Cantor et al., Ershow and Cantor).
3. Are the interpretations of the studies and recommendations appropriate?
The final recommendations are appropriate. However, some minor re-wording is
needed. As clearly supported in the studies by Ershow and Cantor and Cantor et al., and
confirmed by other studies, the average drinking water consumption rate is 1.4-1.5 L/day.
Therefore, as stated on pg 2-41, a value of 1.4 L/day is appropriate to recommended as the
average drinking water rate for adults. However, it is not clear why later in the same
paragraph, values of 2 L/day and 2.27 L/day, which are the 82th and 90th percentile values
from the study by Ershow and Cantor, are recommended for "chronic" and "acute" exposure
assessments, respectively. The methods used by Ershow and Cantor and Cantor et al. do not
B-118
-------
Larry Gcphart
indicate a systemic bias towards lower intake rates. Rather, the data collected by various
investigators indicates consistency around 1.4 L/day and 2.0 L/day values as central and upper
percentile estimates, respectively. The implication is that only the upper percentile values
should be used in exposure assessments, and use of central tendency values should be
discouraged. This is inconsistent with the goals stated in the preface to the handbook. Perhaps
including the statement "for use in chronic exposure assessments" to the sentence describing
the average value would help clarify this issue.
4. What are the significant data gaps?
There is apparent agreement amongst studies on total fluid intake rates. However, as
described in the June 1994 Exposure Factors Handbook Workshop, more data are needs on the
portions of total water consumption: 1) ingested directly from the tap; 2) ingested after heating
or after treatment; 3) used in commercial beverages and 4) used as an ingredient in home
prepared beverages. Consumption rates for specific sub-populations (e.g., infants, athletes,
pregnant women) are lacking. Data are also needed for incidental water ingestion which occurs
during swimming.
Section 2.8 • Soil Ingestion and Pica
1. Are data presented in way that is useful to exposure assessors?
In general, the technical summaries of the studies effectively bring out the strengths
and weakness of the various study designs. However, as described below, there are a number
of technical issues on individual studies that need of resolution. A summary table providing the
final conclusions of the various authors from the published papers would be useful. Most
importantly, given the extreme differences in quality of the studies and how much confidence
there is in the quantitative estimates of soil ingestion rates derived from the individual studies,
it is not clear if data from multiple studies should be used to derive a mean composite estimate
of soil ingestion for children (pg 2-410). Is such a composite is deemed useful, as described
below, the information in the table contains a number of errors which should be corrected.
B-119
-------
Larry Gcphart
There are also a number of publications available which are not cited in the text or reference
section which should be added.
Li the background section, I suggest changing "toxics" to chemicals and "dirt" to soil.
\
In the study by Calabrese et al. (1990), based on percentage recoveries, the authors
clearly indicated a higher level of confidence in the data for the tracer substances Al, Si, Y and
Zr. Similarly, in the study by Calabrese et al. (1989), the authors indicate higher confidence in
the data for Al, Si, and Y. These differences should be noted in the study summaries and data
for these tracers, plus data for the tracer Zr (see text below) should be included in the summary
table on pg. 2-410.
There are a number of important papers on soil ingestion detection limits not included
section 2.8. For example, in a paper by Stanek and Calabrese (Reg. Tox Pharmacol. 13, 263-
277, 1991; Stanek and Calabrese Reg. Tox. Pharmacol. 13, 263-177, 1991), the authors
indicate that the studies by Binder et al (1986) and Van Wijnen et al (1990), which did not
employ a mass balance approach and did not therefore adequately account for intake of tracer
materials though the diet or medicines, do not provide quantitative estimates of soil ingestion.
In addition, with the exception of data for the single tracer Zr from the study by Calabrese et
al. (1989), data from the other studies do not provide quantifiable estimates of soil ingestion.
More recent reports by these investigators, which are not cited in this section, indicate that Al,
Si,- and Y are may be the most reliable tracers for soil ingestion in children (Calabrese and
Stanek Env. Hth. Persp. 103 (5), 454-4457, 1995; Stanek and Calabrese Env Hlth Persp.
103:276-285, 1995; Stank and Calabrese J. Soil Cont. 3(2), 183-189, 1994). While this
subject is controversial and still under review, it raises the question concerning the validity of
producing a composite soil ingestion rate using data from multiple studies of differing design
(pg 2-410).
A summary of a paper on methods to distinguish outdoor soil ingestion from indoor
dust ingestion in a soil pica child by Calabrese and Stanek (1992) should be included in this
section (Reg Toxicol. Pharmcol 15, 83-85,1992).
B-120
-------
L.irry Gcphart
In the table on page 2-410, there are a number of errors. Both data points from the
study by Clausing et al. are not corrected for control (hospitalized children) values. The
corrected value for Al should be 176 mg/day (232 mg/day - 56 mg/day). A footnote should be
included noting the unavailability of control data for the AIR data. Similarly, the corrected
values from study by Van Wijnen et al range from 69-120 mg/day instead of 162-313 mg/day.
The mean soil ingestion rate for Al in the study by Davis et al. was 39 mg/day rather than 3
mg/day. As described above, for the study by Calabrese et. al (1989), the data for Zr should
included, since data for this tracer may in fact be the most reliable. Using the corrected values
from the studies by Clausing et al. and Van Wijnen et al, and higher confidence data points
from the study by Calabrese, the mean composite value should be recalculated. To decrease
the impact of extreme values, considerations should be given to calculating a median instead of
or in addition to a mean composite value. Notation to the method used to account for ingestion
of tracers in food should be added. Finally, using all the data cited in the table, it is not
possible to calculate a separate mean for soil ingestion versus soil/dust ingestion, as indicated.
In the studies by Binder et al and Van Wijnen et al, no distinction was made for intake of
housedust and soil. Therefore, only the data on soil and dust combined should be included in
the table and the composite value should be designated as representing intake of soil plus
household dust. Data for Zr should be included in calculating the composite value. A revised
sample table is presented below.
B-121
-------
Larry Gcphart
Estimated Daily Soil and Dust Ingestion Rate in' Children3
(mg/kg)
Soil Tracer Substance
Al
121b
175b
52d
30d
69-120*-'
Si
112d
49a
AIR
136b
129C
Yi
lld
Zr
lld
Reference
Binder
Clausing
Davis
Calabrese
Van Wijnen
* all values represent medians except where noted
b value is corrected for ingestion of food using a hospital control group
c value is not corrected for ingestion of food due to detection limit considerations
d value is corrected for ingestion of food using mass balance methodology
e values are geometric means since medians were not available
Concerning adult soil ingestion, in the summary of the study by Calabrese et al. (1990)
it should be mentioned that due to recovery considerations, Al, SI, Y, and Zr were considered
the most reliable tracers. For this reason, the summary table on pg. 2-411 should include
information for Zr rather than Ti. Also, it is recommended that this table list median rather
than mean values, or list both mean and median values. The mean values listed in the study
summary on pg 2-402 were taken from week 1 data alone (table 7 of the study) and are not
representative of the whole data set. The more appropriate values, as listed by the authors, are
taken from table 8. For example, for Al, Si, Y, and Zr, the correct values are 77, 5, 53, and 22
for Al, Si, Y, and Zr, respectively, instead of 110, 30, 63, and 134 mg/day. Finally, the value
of 480 mg/d as an "upper percentile" taken from the Hawley paper should be deleted from the
summary table and from the discussion following the table. This value was obtained from an
exposure reconstruction rather than from an actual study. The recommended revised table
appears as below.
B-122
-------
Larry Gcphurt
Estimated Soil and Dust Ingestion Rate In Adults"
(mg/day)
Mean
Median
Soil Tracer Substance
Al
77
57
Si
5
1
Y
53
65
Zr
22
-4
mean of mean = 39 mg/day
median of median = 28 mg/day
1 Data are from Calabrese et al. (1990)
2. Have the key studies been identified?
The text correctly indicates that higher consideration should be given to placing less
emphasis on studies with serious design limitation (e.g., Binder et al. and Clausing et al.) and
more emphasis on the studies which have fewer study design weaknesses (Calabrese et al,
Davis et al., Van Wijnen et al.). However, due to design limitations, consideration should be
given to not including the studies by Binder et al. and Clausing et al. in the key study section.
Similarly, in the section on adult soil ingestion, the papers by Hawley et al. (1985) and
Krablin (1989) do not provide actual quantitative information on soil ingestion rates. Rather,
they were exposure reconstructions which attempted to estimate soil ingestion. Therefore, it is
recommended that these data should not be included as key studies but as "other information
on soil intake among adult." The only actual quantitative study is by Calabrese et al (1990).
3. Are the interpretations of the studies and recommendations appropriate?
On pg 2-411, the discussion is inappropriately worded in a way to support the current
EPA "default" values of 200 and 100 mg/day for children and adults, respectively. For the soil
intake rate in children, the text appropriately indicates that more weight should be given to data
from studies which were corrected for dietary intake of tracer substances. These studies were
conducted subsequent to when the 200 mg/day value was recommended. However, as
B-123
-------
Larry Gcphart
described above, the values in the summary table for the Wijnen study are not corrected for
background. Once the corrected values are included and presented as medians, the data clearly
indicate that the average rate lies in the range of 11-112 mg/day and the 100 me/day value is
towards the upper end of this range. Therefore, the statement concerning the 200 mg/day value
being a "conservative mean average" should be deleted.
For adults, the-only quantitative data available indicates that the average soil intake
• value lies in the range of 25-50 mg/day. (see table above). The statement on the bottom of pg
2-411 starting with "This set of values is consistent with the 50 mg/day range often used by
program offices" should be deleted. No information is cited to verify the statement and such
recommendations may change over time. In fact, an adult soil intake rate of 100 mg/day is the
often recommended value (e.g.,m EPA RAGS I, 540.1-89/002). If 50 mg/day is chosen, the
text should indicate that this value if towards the upper end of the average range based on
limited available data.
4. What are the significant data gaps?
There is only one published study on soil ingestion rates for adults and only 6 data
points were included in the study. Clearly, more data are needed on adult soil ingestion. Much
of the published data on soil mgestion rates for children are of questionable reliability.
Therefore, much more data are needed for this critical exposure factor. Data on the frequency
of pica and ingestion rates for children exhibiting pica are needed.
Section 3 - Inhalation Route
1. Are data presented in a way that is useful to exposure assessors?
Much of the data are presented in a format useful for exposure assessors. It is
apparent that many of the original references did not provide all of the data required for a
complete exposure assessment. For example, the Linn et al (1993) study presented
statistics on the hourly inhalation rate for different activity levels but failed to include the
time spent at each activity. Additionally, arithmetic means of breathing rate are
presented, not median values. In some instances, (e.g., Shamoo et al., 1991) data are
B-124
-------
Larry Gcphart
presented in such a way that distributions can be made of ventilation rate which may then
be coupled to the appropriate time at each activity level.
2. Have the key studies been identified?
Yes
3. Are the interpretations of the studies and recommendations appropriate?
All of the key studies have been summarized in adequate detail to give the
exposure assessor the necessary information. The advantages and disadvantages of each
study have been clearly expressed. Based on the information provided in the Handbook, it
is clear that there is only one study that represents the general U.S. population: Layton,
1993. All other studies were limited to the Los Angeles area and may be biased. For
long-term exposures Layton presents arithmetic mean£ of daily inhalation rates for different
cohorts (age, sex) and considers both active and inactive periods. Other studies (Spier et
. al, 1992) present daily inhalation rates for active periods only, which will significantly
overestimate daily inhalation rate.
The recommended average daily inhalation rate for adults for continuous exposures
where activity patterns are not known of 13.3 m3/day (based on the Layton studies) is an
appropriate recommendation. However, the recommended upper percentile value of 20
m3/day is not supported by the key studies. This value is an upper percentile for active
periods (supported in Layton, 1993) but is not a representative upper percentile which
considers both active and inactive periods. The recommended upper percentile should not
be higher than 17 m3/day, which is the maximum reported value from Layton (1993) in all
three approaches.
Breakdown of time spent at each activity are relatively consistent among the
studies. The recommended time spent at each activity seem appropriate.
B-125
-------
Larry Gcplum
4. What are the significant data gaps.
Only one study is representative of the general U.S. population. Additional
information on general U.S. trends to support Layton 1993 would be beneficial. Also,
complete statistics on breathing rates within a cohort (to create distributions) would assist
in performing cohort specific exposure assessments. ,
Section 4 - Dermal Route
1. Are data presented in a way that is useful to exposure assessors?
Much of the data, particularly the tables, are presented in a way that will facilitate the
use of the data by exposure assessors.
2. Have the key studies been identified?
Yes, the new, studies by Kissel et al. on soil adherence, and the available studies on
surface area have been identified.
3. Are the interpretations of the studies and recommendations appropriate?
Overall, the recommendation to use data summarized in table 4-4 on surface area is
appropriate. This table presents mean and 90th percentile values for specific body parts which
allows the risk assessors flexibility in performing assessments. The data set which forms the
basis of this table, while dated, is perhaps the best available. In addition, other assessments do
not reveal marked inconsistencies.
The recommendation on pg 4-23 concerning default total body area should be
revisited. It is stated that the total adult body surface area can vary from 17,000 cm2 to 23,000
cm2 (with reference to table 4-4). Based on this range, a value of 20,000 cm2 (central estimate)
is recommended for use in exposure assessments. However, the values presented in table 4-4
B-126
-------
Larry Gcpliiirt
actually range from 14,500 cm2 (minimum for women) to 23,000 cm2 (maximum for men),
with a central estimate of 18,750 cm2. Also, the mean of the values for men and women
combined is 18,150 cm2. Therefore, 18,000 cm2 should be recommended as the central
tendency value for use in exposure assessments, in lieu of using age or sex specific values.
On soil adherence, the recommendation to use the high quality data from the study by
Kissel et al. (1995) is appropriate. The data presented in table 4-12 will allow the risk assessor
to perform activity specific assessments with consideration for exposure to specific body
areas. This approach is very superior to the approach recommended in the previous handbook.
which assumed a constant upper bound soil adherence rate for all body areas.
4. What are the significant data gaps?
As described in the previous exposure factors workshop, many of the serious data
gaps lie in the area of skin exposed under various exposure scenario (soil contact, use of
various commercial products, seasonal impacts, etc.).
Section 5.3 - Activity Patterns
1. Are data presented in such a way that are useful for exposure assessors?
Data on some of the important time activity patterns (e.g., residence time, shower
duration) are presented in a way that will facilitate their use in exposure assessments, including
preparation of exposure distributions. Data on other factors, for example, occupational tenure,
are presented in a way that will help exposure assessors prepare assessments for selected
occupational, ethnic, and age groups. However, the data on occupational tenure are not
presented in a way which allows for preparation of exposure distributions. All of the data
collected by Robinson et al. (tables 5-28 to 5-30) do not allow for preparation of distributions.
B-127
-------
Larry Gcphart
2. Have the key studies been identified?
The two comprehensive compilations of time activity patterns (Robinson et al., 1991
and CARB 1991) are cited. Also, the key studies for occupational tenure (Carey, 1987/1988),
residence time (Isralei, 1992 and Cappel, 1992), and shower duration (James and Knuiman,
1987) are cited. However, none of these studies are clearly identified as "key" studies.
3. Are the interpretation of the studies and recommendations appropriate?
The recommendations on a number of important time activity patterns are not clearly
stated. For example, for residence time, summaries of the 3 primary studies (Nelson and
Isralei and Nelson, 1992; Johnson and Cappel, 1992; and U.S. Bureau of census, 1993) and
some of the limitations of the studies are provided. However, it is recommended that the text
include a summary paragraph which provides guidance on preference for which data set to use,
based on technical considerations. An example is given below.
In the study by the U.S. Bureau of Census, the assumption of even distributions within
ranges for which the data were collected, severely limits the usefulness of the study. The
approaches for estimating a distribution of the average total residence time used by Isralei and
Nelson (1992) and Johnson and Cappel (1992) were fundamentally different in two significant
areas. First, the Israeli and Nelson study took survey data to determine the desired distribution
for households. Johnson and Cappel centered their model around individuals. Second, their
data sources and data manipulation differed greatly. Israeli and Nelson took current residence
time data and performed rigorous probability calculations to determine a moving rate and then
a total residence time, whereas Johnson and Cappel utilized available data on mobility to do
simple calculations for probabilities of moving and then ran these probabilities through as
simulation to arrive at their final distribution. Because of the above differences and the fact
that Johnson and Cappel used the more current data of the two, the data by Johnson and
Cappel (tables 5-49 and 5-50) are recommended for use in exposure assessments (e.g., 50th
and 90th percentile values of 9 and 26 years, respectively).
B-128
-------
Larry Gcphnrt
4. What are the significant data gaps?
Much of the data available on activity patterns do not allow for preparation of
exposure distributions.
Section 8 - Analysis of Uncertainties
1.- Are data presented in a way that is useful to exposure assessors?
This section differs from the other sections in that it presents methods rather than data.
It's specified goal is to discuss "methods that can be used to evaluate and present the
uncertainty associated with exposure estimates". However, it deals more with characterizing
"types of uncertainty" than with methods to evaluate and present uncertainty. In some cases,
the descriptions of the types of uncertainty are vague or confusing. In most cases, the
description of methods for evaluating uncertainty is not made clear.
A recommendation is to specifically highlight the methods, either by bold type, by
numbering, or by putting them into a section of their own. My reading of this section uncovers
(by careful highlighting) the following 18 specific methods recommended:
1. Classify uncertainty into one or more types, e.g., scenario, parameter or
model.
2. Identify sources of uncertainty for each type, e.g., professional judgment
for scenario selection.
3. Describe rationale for professional judgment.
4. Characterize uncertainty as high, medium or low.
5. Do sensitivity analysis to set credible upper limit.
B-129
-------
Larry Gcphart
6. Avoid surrogate data.
7. Use bounding estimates.
8. Use a "best" estimate.
9. Use a probabilistic distribution based on data for a parameter that
"profoundly influences" the exposure estimates.
10. Use expert judgment to generate subjective probability representation.
11. Do a sensitivity analysis by using upper and lower limits.
12. Use "analytical uncertainty propagation".
13. Use probabilistic uncertainty analysis.
14. Use statistical methods.
15. Describe rationale for selection of models.
16: Use different models to establish a range of modeled estimates.
17. Confirm modeled computer code output.
18. Compare performance of model to actual observed data similar to
scenario.
It is obvious that the above items are not presented in a way that is useful to an
exposure assessor. Aside from being hidden within the text, the methods are not described
well, nor are there any examples which would more clearly illustrate their use.
B-130
-------
Larry Gcpluirt
2. Are the data presented in a way that will support both point estimate and Monte Carlo
assessments?
No. On the contrary, the recommendations in Section 8 argue against Monte Carlo
assessment by providing a laundry list of disadvantages with limited advantages (general
applicability, no restriction on form of input distributions, and straightforward computations).
The list of "do not use" reasons includes:
- only use when there are credible distribution data for most key variables
- don't use if you only need average exposure values
- don't use if you only need a bounding estimate
- sensitivity analysis is difficult to do and doesn't work.
- assumption of independent distributions is a problem
It appears that the writer wants to discourage Monte Carlo analysis and has a great deal of
uncertainty about its usefulness. This is especially apparent in the statements regarding
sensitivity analysis. The need to rerun the entire calculation several hundreds or thousands of
times is not a disadvantage since the software and hardware available to do this is practical
and quick (seconds to minutes). Secondly, a check on the shape of the resultant exposure
distribution against the shapes of the input distributions is a quick way to pinpoint potential
sensitive or "driving" distributions.
The alternatives to Monte Carlo analysis are to use "analytical uncertainty propagation"
and "classical statistics". The document does not describe the first with any degree of clarity,
and the second, of course, is desired but often not possible due to lack of data. The result is
that the reader is left with one choice: use sensitivity analysis. Worse yet, the recommended
method for doing a sensitivity analysis is incomplete: only use the upper and lower bound.
More complete use of sensitivity analysis is to alter the parameters by a constant percentage to
test the sensitivity of the mathematical model.
B-131
B-133
-------
Larry Gcphart
3. Are there data gaps?
-------
DEPARTMENT OF THE A HMY
U S ARMY CENTER POR HEALTH PROMOTION AND PREVENTIVE MEDICINE (PROVISIONAL)
ABERDEEN PROVING GROUND. MARYLAND 21010-6422
. o»
July 14, 1995
Health Risk Assessment and Risk
Communication Program
SUBJECT: Review of the Draft Exposure Factors Handbook
Ms. Helen Murray
Eastern Research Group, Incorporated (ERG)
110 Hart we 11 Avenue
Lexington, Massachusetts 02173-3198
Dear Ms. Murray:
Thank you for the opportunity to review the Draft
Exposure Factors Handbook. Dr. Jsick M. Heller and
Mr. Dennis E. Druck of the Health Risk Assessment and
Risk Communication Program reviewed the subject
handbook with special emphasis on the sections
pertaining to water and soil ingesition and dermal
contact . The presentation of the data and
recommendations is organized in a manner which should
be useful to exposure assessors. Overall, the handbook
is well done and provides information that should
improve the exposure assessment process.
Our only recommendation is ths.t the Background
Section of the Introduction induces an expanded
discussion of the importance of using site- specific
exposure factors in lieu of default values when such
information is available. Our point of contact is
Dr. Heller at commercial (410) 671-2953.
. Sincerely/
Arthur P. Lee, P.E.
Major, U.S. Army
Program Manager, Health Risk
Assessment and Risk
Communication
Copies Furnished:
Headquarters, Office of The Surgeon General
Commander, U.S. Army Medical Command
Readiness thru Health
B-135
-------
-------
Brad Shurdut
B-137
-------
-------
Shurdut Comments
Pagel
Comment for Draft Exposure Factors Handbook
Non-Dietary and Dermal Exposures
I. General:
The additions to the Drinking Water Ingestion, Soil
Ingestion, and Dermal Exposure sections have significantly
bolstered the utility of the book. The presentation of
material in this book does provide the reader with a fairly
objective listing of pertinent studies from which exposure
factors have been derived and forces the user to use his
discretion as to the factors which he deems most appropriate.
Although previous input has suggested that the development of
standard scenarios will not be pursued, the use of a brief
and simple example which either.precedes or follows the
textual discussion of the exposure parameters would certainly
clarify the utility of the data presented.
Should the authors or sponsors of this book promote the use
of some results more than others to ensure the consistency or
validity of exposure assessments? Stronger suggestion rather
than a recommendation may achieve this end. I think the
utility of this book is two-fold: (1) to present a
compilation of credible scientific data and studies to be
used to facilitate and enhance the development of exposure
assessments and (2) to improve the consistency of the data
used in exposure assessment by suggesting 'recommended' data
to be used for assessments. One of the greater utilities of
exposure and risk assessments is the use of the work product
within the regulatory framework of the EPA. In my
interactions with the EPA, especially in the area of
pesticide human exposure assessment, one of the larger
problems faced by myself and others in industry is the
inconsistent use of exposure data and values by the assessor
and the EPA scientist. It seems to me that the Exposure
Factors Handbook presents an opportunity to sift through the
B-139
-------
Shurdut Comments
Page 2
pertinent studies and highly recommend the data sets and
methodologies that should be used in assessments in concert
with other EPA generated guidelines. This can be more easily
achieved by a table at the end of each section or chapter
summarizing the recommended values or point estimates for
describing each variable.
Although this format provides a rich source of information
for the exposure assessor from which he can choose
appropriate factors for his assessment, the data (and data
tables) may confuse rather assist the assessor. For those
data sets that have gained greater levels of acceptance
within either the. scientific or regulatory communities, a
notation should be provided. This may in fact be done by the
classification of studies as either 'key' or 'other relevant
studies'.
XI.Water Ingestion;
Usefulness of Data Presentation:
The 'Drinking Water Consumption' chapter would be easier to
follow if the chapter sections were presented in a manner
which paralleled the variables within the dose equation. For
instance, the section entitled 'Key General Population
Studies' could be re-titled as 'Ingestion Rate (IR): Key
Studies' . For those variables, such as body weight and
exposure duration, which are discussed in other parts of the
book can be referenced to applicable sections. In addition,
those variables, such cis concentration (C) , diet fraction
(DF) , and averaging time (AT) , which are either not
applicable for drinking water consumption or specific to a
chemical and/or event can be briefly touched upon before the
primary analysis of water consumption rates.
The 'Ingestion Rate' discussion could be organized into sub-
categories. For instance, there are several factors of
B-140
-------
Shurdut Comments
Page 3
interest that potentially affect the rate of intake. These
may include, but are not limited to, geographic regions
living and activity level. Therefore, the data may be
presented by category rather than exclusively by author. The
effect of activity pattern on consumption rates, for example,
is discussed in the Ershow results and then in the Mcnall and
Schlegel results. Consolidation would more efficiently
direct the exposure assessor to a section of the chapter and
preclude the need to comb the entire chapter. Another
suggestion may be to first present the capsule summaries of
each study included in the chapter and then to group the
tables at the end of each section.
A summary of the factors presented would be useful to include
at the end of the drinking water section. This would boil
down the studies into a 'quick' reference form. Each factor
and/or study describing the parameters could be accompanied
by a 'high1, 'medium1, or 'low1 ranking based on the strength
and accuracy of the study reporting the data.
Means and standard deviations should b£ included in drinking
water tables. This would facilitate the use of point
estimates if desired by the assessor (Tables 2.2 and 2.3),
Data contained in the National Food Consumption Survey was
collected in 1977-78 'which was used in the Ershow analysis.
Not only is this database slightly outdated which may effect
the consumption values, but the data contained therein refers
to commodities regardless of the mode of preparation. As a
result, the water content of commodities may significantly
change following the consumer^ preparation.
Water consumption is a function of the ingestion of water
over the course of a day from many sources. Most of the
water consumption data are aggregate results. When concerned
about source-specific exposure it is necessary to estimate
B-141
-------
Shurdut Comments
Page 4
potential water consumed at one location versus another- not
total water ingested from a variety of sources over the
course of a day (restaurants, place of employment, etc.). A
percentage breakdown of the source of ingested water needs to
be split out of the general discussion and presented in a
distinct section describing the dietary fraction (DF)
variable.
Another source of water for ingestion, although not easily
quantified, results from swimming in potentially contaminated
bodies of water and during the taking of showers. Although
these routes of exposure 'can be considered minimal,, they
should be acknowledged.
The use of arithmetic and geometric means are not clearly
identified. As on page 2-41, use of a 'mean' heading does
not sufficiently describe the measure of central tendency
used.
Gaps/Future Research, needs:
EPA is currently considering the use of more recent
consumption databases for DRES (Dietary Risk Analysis).
There, is a need to use more up-to-date consumption data than
the NFCS upon which many of the studies are 'based.
Consumption data is underestimated by approximately 15-20%
compared to the more recent databases. In addition, dietary
patterns may be quite different today than those reported
within studies from 1976. Today there is an increased
emphasis on eating healthy which may also affect consumption
patterns.
There is a need to additionally refine some of the ingestion
data. Source specific and location-specific data would be
useful to assess water ingestion exposures resulting from
distinct sources (home vs. other sources) and the source of
B-142
-------
Shurdut Comments
PageS
tap water whether the source be from a well or reservoir,
etc.
III. DERMAL ROUTE OF EXPOSURE
The organization of this chapter does present and develop the
data well. Since only body surface area studies are
presented in the general dermal exposure section of this
chapter, the reader should initially be directed to other
applicable sections of the book to find some of the other
exposure factors, i.e. exposure/activity frequency, event
duration. In addition, the section dealing with dermal
adherence of soil has not been broken down into 'key1 and
'other1 studies. Since some of the studies contain empirical
data while others like the Sedman study presents mere re-
calculations of other previously reported data, clearly some
results are more key than others. The studies by Lepow and
Driver may arguably be more reliable than the- -other studies
provided in this section, but less reliable than the Kissel
study contained in the 'New Soil Adherence Research1 section.
Other factors which may be required for dermal exposure
assessment are measures of the frequency of dermal.contact
with a surface and the size of the surface dermally
contacted. The data to answer these questions are most likely
presented in the 'Reference House1 of •Activity Patterns1
sections within the book. The user should be directed to
these areas in the sections briefly describing each exposure
parameter in the dermal exposure assessment equation. In
addition, ranges for transfer factors reflecting the amount
of a material capable of being dislodged from a surface need
to be developed. Studies using pesticides have shown that
generally less than 1% of the material applied to a carpet
matrix is actually removed from the carpet onto the skin.
These factors have been demonstrated using both dislodgeable
residue techniques with dosimeters and biological monitoring
B-143
-------
Shurdut Comments
Page 6
performed in concert with human activity on the treated
surfaces.
Since the surface area data is of primary importance to the
determination of dermal exposure, the data must be coupled
with the estimated area of skin exposed while conducting
various tasks. A section may be added that summarizes these
results which further supplements Table 4-11. Several
publications have presented this data (USEPA, 1992; Hawley,
1985).
The authors are correct in saying that contrary to initial
perceptions, clothing does not eliminate dermal contact with
a chemical. However, depending on the chemical potentially
exposed to, clothing is generally an effective barrier
against chemical penetration and greatly reduces the amount
of pesticide contacting the skin. Studies conducted to
evaluate exposures to pesticide workers demonstrate that
generally less than 10% of a pesticide contacting the outer
surface of clothing penetrates through the clothing with a
majority of the results being much less than 10%. Secondary
dermal exposure may also be a consideration. Some chemicals
may be trapped in clothing materials which are subsequently
transferred to the skin over time especially as the clothing
becomes wet with perspiration.
The soil adherence studies presented in the Dermal Exposure
section, except for the Kissel study, are very limited. It
is generally regarded that the amount of soil found adhering
to exposed skin regions are highly dependent on the type of
activity performed. Many of the studies do not specifically
look at this variability when performing different tasks.
However, the Kissel study does attempt to describe this
variability which occurs in the real world by documenting
soil-skin adherence for several tasks and for all exposed
skin regions.
B-144
-------
Shurdut Comments
Page 7
Recommendations:
The recommendations presented are appropriate for this
section.
Data Gaps:
There is a need to include additional data regarding site-
specific absorption differences for different regions of the
body (variability of skin permeability). Maibach has
discussed this issue and documented this regional variation
in percutaneous penetration. Penetration indices for regions
of the body have been developed by comparing penetration of a
challenge compound to the penetration of a chemical through
the forearm. The penetration indices were specifically
derived from hydrocortisone skin penetration data and from
absorption results using the pesticides, malathion and
parathion (Guy and Maibach, 1984).
Another need is for a description of regions of the body that
may be potentially exposed during the conduct of various
activities. In addition, the size of an area contacted by an
individual performing a certain task directly influences his
potential exposure. Therefore, dermal exposure is not only a
function of the surface area of the body part being exposed
but the number of times that the area will be exposed to the
contaminated media (i.e. exposure when walking barefoot on
contaminated grass or soil is dependent on the surface area
of the receptor (the foot) contacting the grass and the
either the amount of the total lawn walked upon or number of
times contacted).
There exists a need to develop additional soil adherence
numbers for certain tasks or activities conducted in various
types of soils (sand, loam, or silt with varying moisture
contents). Furthermore, there is a need to investigate the
B-145
-------
Shurdut Comments
PageS
potential relationship between, residence time on the skin and
absorption, and the relationship between loading levels
(greater than a monolayer) on the skin and subsequent dermal
absorption. The general relationship between dermal
adherence, exposure to solvents and particulates in or on the
soil, and subsequent dose needs to be more fully investigated
before recommendations for the use of this data can be
encouraged.
In order to measure potential exposure and extrapolate
dermally absorbed dose additional factors should be
considered in the analysis:
•Permeability differences between.skin of child/adult ,
•Permeability differences of hydrated vs. dry skin.
IV. Soil Ingestion and Picas
The organization of the chapter is good. However, since 'a
majority of the studies deal with ingestion rates discrete
time period, the Hawley study describes results for each
season. Since seasonality could be considered a significant
factor for the amount of soil ingested, this data set should
be distinguished.
The soil_ ingestion studies (Clausing) using trace analysis
from outdoor soil which is subsequently correlated with fecal
levels of the trace material may potentially overestimate
soil ingestion quantities. These studies do not delineate
between potential exposure to trace elements from either
indoor or dietary sources. It is possible and likely that
the trace elements are present within the home in dusts,
etc., especially following tracking in of soil into the home,
which subsequently leads to potential exposures following
contact with contaminated surfaces. Furthermore, the control
population consisting of hospitalized children, are quite
B-146
-------
Shurdut Comments
Page 9
different than the studied population in terms of types of
activities performed over the course of a day and the limited
indoor environment to which he is exposed. As a result, both
the Calabrese and Davis studies yield more reliable results
than the preceding studies.
Data Gaps:
All the studies are lacking in the apparent documentation of
a child's activity and the amount of soil ingested. Soil
ingestion would intuitively be activity-dependent and
characterization by level of activity would be more useful
for the exposure assessor. By doing this, it would also
facilitate the use of the data to re-create exposure
scenarios in different regions and under different climatic
conditions.
Little work has been presented as to a child's 'mouthing
behavior1. The number of times that a child sucks his thumb,
touches his food (especially a sticky lollipop) with his
hands can have a strong impact on the final amount of soil
ingested. Data as to activity patterns, 'mouthing behavior1",
removal efficiency of soil by saliva, and soil loading on the
hands (Kissel and Lepow studies) may present an equally
predictive method for estimating soil ingestion.
Analysis of Uncertainties:
Good general overview of uncertainty analysis. It does
briefly summarize the uncertainties that one must be aware of
when conducting exposure assessments. It also presents the
limitations of potentially using point estimates in contrast
to probabilistic estimates using Monte Carlo simulations.
B-147
-------
-------
Work Group #3
Human activity patterns
B-149
-------
-------
Steven Colome
B-151
-------
-------
Colome comments
Page - 1
Review of:
EXPOSURE FACTORS HANDBOOK
External Review Draft
June 1995
EPA/600/P-95/002A
The National Center for Environmental Assessment published the first edition of the
Exposure Factors Handbook in 1989. Availability of newer data on human exposures and
further development of approaches to the craft of risk assessment led to the drafting of
this second edition. Sections have been added on the use of consumer products and the
reference house. Content has been updated throughout the text.
This reviewer did not participate in the 1993 workshop and comes to this task with a
fresh perspective. My review concentrated on sections of the draft that relate to activity
patterns including portions of Chapters 3, 5,6, and 8. These were the subject of my
assigned review. In addition, I scanned the section of Chapter 7 dealing with air
exchange rates.
Overall, the document presents information that may be useful to the risk assessor. The
task of compiling the exposure literature relevant for risk assessment was accomplished
in the Handbook. Of greater use to the risk assessor, there was some screening of the
literature for pertinence to the task of risk assessment. Individual studies have been
evaluated for strengths and weaknesses relative to risk assessment objectives and not
necessarily to the stated objectives of the investigators and the original purposes of the
studies. A substantial amount of raw and summary data are provided in the Chapters and
Appendices that may be used by the risk assessor for background information or
modeling.
Still, the Handbook was not particularly easy reading, and reviewing the assigned
sections of the document left me with the feeling that something was still missing. I tried
to approach this document as a risk assessor faced with a specific project. This is the
person who might turn first to this document for information.
The chapters I reviewed were generally weak on interpretation and evaluation of the
available studies. Many of the individual study evaluations read like the study-by-study
evaluations of the Criteria Documents. Those documents have legal status that requires
attempts at full and balanced evaluation. It seems so me that the Handbook could take
more latitude by skipping weaker studies, using what can be gleaned from available
studies, and directly identifying the data gaps and weaknesses. To assist in interpretation
and qualitative estimation of uncertainty, the Handbook could address biases and the their
potential magnitude. This directness was generally avoided and the authors opted to
point out limitations without discussion of the potential influence of those limitations on
B-153
-------
Colome comments
Page-2
the results of risk assessments. I would also have liked to see an attempt at more
integration of information across studies. How are the studies similar and how do they
differ? Can we learn something from the differences?
Detailed comments follow:
CHAPTERS
Page 3-1,1st paragraph. The word should either be paniculate matter or particles. You
should not use "particulates" as a noun.
Page 3-1, equation 3-1. It may be useful to add a conversion from ppm to jj,g/m3.
Gaseous measurements may be expressed in ppm.
Page 3-2, paragraph 1. The reference to "Heart" watches should be dropped. There are a
variety of techniques for monitoring heart rates and the apparent endorsement of a
product should be avoided.
Page 3-6, Table 3-2. The footnote should correspond more directly to the referenced
headings. This table is difficult to read and needs additional editing.
Page 3-9, paragraph 2. It is unclear from the presentation whether the time/activity survey
of 2126 Californians was conducted by Layton or by another investigator. The
discussion is deficient in evaluation of the quality of the exertion distributions, which are
based on recall. Since this study is heavily relied upon to provide summary information
on breathing rates, this omission is significant. It is also stated that this study is
representative of the general US population even though the participants are all drawn
from California. In other sections, regional studies are often said to be nonrepresentative
because subjects are drawn only from the region of the survey. I would tend to be more
accepting of regional studies when they provide the only available information. The
Handbook, however, needs to have consistency across chapters in its evaluation criteria.
Page 3-13, last paragraph. This is an example of the type of evaluative comment made
about a study. Chapter 8 indicates that this information can be used by the risk assessor
to establish qualitative uncertainty estimates for data drawn from the studies. However, I
am at a loss as to how a risk assessor can constructively use the information presented in
this paragraph hi order to assist a decijjion-maker to interpret uncertainly in a quantitative
risk assessment. More to the point, we need to know the quality of information for the
parameter presented. Are the results likely to be biased. If so, what is the direction and
potential magnitude of any bias. What are the likely results from a study being
nonrepresentative? Should we proceed to use the data distributions for lack of more
general data? The evaluative paragraphs are too general and inconsistent now to be of
much use.
B-154
-------
Colome comments
Page - 3
Page 3-2, paragraph 1.1 suspect the heart rates were regressed with log VR and not
"lognormal" VR. The distribution of VR may be approximately lognormal but the
manipulation of data takes the log of the value. This error appeared earlier in reference to
the Linn or Layton papers but I could not find it while looking back.
Page 3-22, Table 3-11. The number of students studied (EL=17; HS=19) should appear
on this table. It would be a good practice to provide numbers of subjects on all tables
since this helps in estimating variability and may be necessary for certain models. Also,
tables have a way of being reproduced absent the accompanying text.
•
Page 3-27, paragraph 2. Why does this California study fail to represent the US
population while the Layton study, also conducted in California, is interpreted as
providing generalizable information. The criteria used to evaluate studies needs to be
consistent.
Page 3-28, last paragraph. Why is the lack of heat stress information used as a limitation
of this study alone? I do not believe that any of the cited studies included consideration of
heat stress. There are a large number of demographic groups and health considerations
for which it would be useful to have additional information. The need for risk assessment
is to set priorities on this information and focus on factors that have the highest individual
or aggregate population risk.
Page 3-30, paragraph 2. Lognormal is again used incorrectly. Also, regression lines are
not "fed to" unless we have developed a new strain of equation-eating bacteria. Also,
discuss whether classification by the categories of essential vs. nonessential activities was
productive. It appears from Table 3-16 to be insignificant. If 'essentialness' of an
activity is not an important classifier, then it should be dropped from consideration in risk
assessments. Let the Handbook be a guide to where risk assessments may constructively
apply our limited efforts. We are still at a stage of development in this field of risk
assessment where information needs to be filtered.
Page 3-33, last paragraph. It is meaningless to discuss the "distribution of the data set" for
a sample of 9 people. The useful comparison is whether this small data set falls generally
in line with observations from larger studies. If the study merits presentation in this
handbook you should be able to discuss the added understanding derived from the effort.
Studies for special populations such as people with certain diseases, occupations or health
status, may be useful when the sample size is small. The values derived from these
special studies can be compared with those derived from population studies to determine
whether parameter estimates need adjustment for the population subgroups. I do not see
this type of evaluation here.
Page 3-37, Table 3-18. This table is difficult to interpret and will therefore be difficult to
use. How would a risk assessor apply the percentage of correct assessments of ventilation
range to a questionnaire of self-assessed activity? I suggest that the table be made more
B-155
-------
Colome comments
Page - 4
self-explanatory or else that it be dropped and the relevant information summarized in the
text.
Page 3-36. Something is lacking hi the transition between the first and second paragraphs
under the section on "US EPA". This needs an appropriate segue between ventilation
rates and percent tune spent indoors.
Page 3-46, first paragraph. The recommendation of 13.3 m3/day is far too exact for our
level of understanding. By using a false sense of precision, risk assessments often portray
greater certainty than exists. I recommend presenting a range of values and listing
characteristics such as gender that influence inhalation rates. Since the recommended
value differs substantially from the previously used ICRP value of 20,1 think it is
essential that the rationale be more tightly constructed.
Page 3-46, paragraph 2. It is stated that 20 m3/day represents "an upper percentile
estimate". What is meant by this? I suspect this means the upper 1% value. However,
this would be arbitrary and the phrasing could mean upper 1, 5,10 or 25% value. In
general, the recommendations made for this section are not convincing and have not been
presented in a manner that would be of great use hi a risk assessment.
Page 3A-5. If Table 3A-5 is essential it should be retyped for clearer presentation.
CHAPTERS
Pages 5(16-20). It may be useful to collapse this table since many of the categories of
activities are not needed hi exposure studies. How, for example, would a risk assessor
use family time or free time spent hi social life? These extra categories are present
because the original studies were not conducted to assist the risk assessment enterprise;
instead they had a clear purpose to address social behavior. Your task is something like
putting a round peg hi a square hole - you'll need to trim the edges to make it fit.
Page 5-23, paragraph 2. The "limitation" that tune-use relevant to exposure questions is
missing is not a design limitation. As stated above, these studies were conducted for
other purposes. It is useful to keep this hi mind while evaluating exposure applications of
data from these studies. It would be useful here to point out the need and utility of
conducting time-use studies with exposure-related objectives.
Page 5-24, last paragraph. What evidence is there that the activity patterns of children
have changed significantly since 1981? Unless these is some direct evidence of this, the
age of the study should not be listed as a limitation. It would be valid to state that there
are questions regarding the current application of exposure models due to possible
changes hi the activity patterns of children over the past 15 years. Even with a general
comment like that it would be useful to think through which behaviors might have
changed that would alter exposure patterns. This could serve the further use of this
B-156
-------
Colome comments
Page - 5
Handbook, which is to point the way to additional studies that will assist in reducing
uncertainty in risk assessments.
Page 5-(40 to 45). This study is useful hi part because it makes direct comparisons
between groups (i.e., Californians and the US as a whole) on exposure-relevant activities.
Page 5-45, paragraph 3. The CARS study and the national studies were not conducted
independently. John Robinson was a common factor and principal in both efforts.
Because of this, the study designs and methods were similar.
Page 5-62.1 would classify the paper by Sexton and Ryan as a review and concept-setting
piece, and not as a study. The Handbook treats this like a data presentation study. The
summary of this paper more appropriately belongs in the introduction to this chapter
where it can provide some orientation and direction for the detailed studies that follow.
CHAPTER 6
Pages 6-3 to 10. The tables on these pages present the mean value for minutes spent hi
use of various consumer products to the hundredth of a minute. Since the standard
deviations are typically express in tens of minutes to hours, the precision of the means is
silly and implies greater accuracy than exists. The means should be rounded to the
nearest minute. Also, you should more critically evaluate the quality of the use data since
they are based on one-year recall of product use. That information may be so inaccurate
as to be virtually worthless. The issue of data value should be more directly addressed in
this section.
Page 6-21. The recommendation section here is very good and should be a model for
Chapters 3 and 5. The summary addresses which information is needed in order to
estimate exposure. It proceeds to indicate which of these data are available in the chapter
and what data must be gathered from other sources or otherwise estimated.
CHAPTER?
I also reviewed the section on building ventilation presented in this chapter. Very little
guidance is given in this section to the risk assessor on how they might use the
information provided on building ventilation. Further, there is no reference to the joint
distribution of volume and ventilation and how a risk assessor might merge the
distributions for these two variables. Ventilation and building volume are the two most
critical factors affecting building exposures once the source strength hi an enclosed space
is known.
This section also does not spend much effort describing the role of season and
temperature in affecting air exchange. For example, while colder temperatures hi winter
will lead to tightening of the structure to minimize heat loss, thereby reducing ventilation
B-157
-------
Colome comments
Page-6
rates, the higher driving force associated with the temperature difference between the
inside and outside will tend to increase ventilation rates. The same is true for the air
conditioning periods of summer. It is hi the transitions that great differences can be
observed between two adjacent homes while one uses cross ventilation to cool the
residence, a neighbor may have tightened up the home to efficiently use air conditioning.
These differences have a significant influence on ventilation and may have a significant
influence on exposure. I get no sense of these relationships in reading this section.
This section should also cite the M. Pandian et al., paper published in J. Exposure
Assessment and Environ. Epidemiology in 1994. This manuscript used the VERSAR
ventilation data base. Following publication, it was found that the VERSAR data base
was incorrect for the western states due to errors in the coded data provided by
Brookhaven and misinterpretation of the data flags. These errors were corrected in errata
published by the same j ournal.
The Koontz & Rector manuscript from 1993 would have used the incorrect VERSAR
data base while the Koontz & Rector 1995 manuscript is not listed in the references and I
do not know for certain whether this version uses the correct data. Based on the values
presented in Table 7-5,1 suspect these values are corrected. It would be useful to cite the
Pandian manuscript, however, especially since it was published as an aid to conducting
exposure and risk assessments.
CHAPTERS
This is a well written and general summary. However, it provides little direct guidance to
the risk assessor. This may be purposive in order to force careful thinking with each
assessment encountered. If the other sections had been written with the same clarity, the
rest of the document would have been easier to read.
B-158
-------
Edward Avol
B-159
-------
-------
Comments on Chapters 3, 5, 6, and 8 (Activity Patterns Panel):
CHAPTER 3, INHALATION ROUTE
p.3-1, sec 3.1, equation definitions - There is an inconsistency in the definition and professed
use of Equation 3-1; In the text, the claim is made that ADD is to be used for non-
carcinogenic non-chronic effects, but in the equation definitions, AT is defined for
carcinogenic as well as non-carcinogenic effects. Since the LADD is used for carcinogenic-
related calculations, either correct Eqn 3-1 so that AT is replaced by ED, or add another
equation for LADD with AT appropriately defined.
p.3-1, sec 3.1, para 2, second to last sentence - If exposure duration is defined with respect to
a particular location, ADD would need to be a summation of multiple location-based
exposures, would it not?
p.3-1, sec 3.2.1, first sentence - health risk is also a function of the chemical species, not just
concentration, duration, and inhalation rate; for example, exposure to hexavalent chromium at
a given concentration, duration, and inhalation rate should result in a different health risk
assignment than exposure to sodium chloride at the same concentration, duration, and
inhalation rate.
p.3-2, sec3.2.1, para2 - Most discussions of ventilation rates and minute volumes are framed
in units of liters per minute; this practice is begun in the paragraph, then discontinued, but it
would be more valuable to users of the handbook to present the discussion in commonly used
units (liters per minute) than to have readers continually have to back calculate from cubic
meters per hour. Also, the reference for the Ozone Criteria Document should be EPA, not
CARB.
p.3-4 - general editing comment - having the tables interspersed with the text made it difficult
to read and follow...is it possible to put tables and figures at the end of each chapter, as is
done with references?
p.3-3, sec3.2.2, Layton discussion - the text discusses the three approaches but only addresses
the potential advantages and limitations of the third approach; in any event, it is a little
difficult to follow the discussion - would it be possible to add a summary table comparing the
three approaches with regard to limitation, advantages, and results?
p. 3-42, Table 3-22 - Layton reference should be 1993, not 1992.
p.3-45, Summary table of inhalation rates -
(a) The presentation, in the paragraph leading to this table, begjns,"...for purposes of this
recommendation,...", and then the summary table is presented. The presented values are
misleading to potential readers in that only after reading the text following the summary table
doe it become clear that the summary values in the table are not the recommended values for
use; it would be more direct and of more use to the reader to summarize recommended
values, and then explain them, if need be (i.e., re-work this summary table and use the
B-161
-------
surrounding paragraph of text as justification/explanation of what the table says; otherwise,
there is a risk that a reader may just use the values in the table, without ever reading that the
Spier value for upper percentile inhalation rate was too high for continuous exposure
assessment estimates. Perhaps a good compromise would be to add another line to each of
these summary tables, called "recommended value"...
(b) The title is misleading, and should include the word "adult" or "age 13-65+yrs" or some
other identifier, since the table on p3-47 is also summary of long-term exposure data (but for
children less than 12yrs).
p.3-46, Summary table of short-term exposure inhalation rates - title problem; no specification
of appropriate age range for use, but it appears in the section marked adult (yet CARB studies
cited included children)...? Again, another line with "recommended values" would be most
helpful.
Chapter 3 Issue Review:
1. The data is presented in a way that could be useful to assessors, if the recommendations
above are incorporated In terms of the best way to present the data, my personal reference
might have been to more clearly identify key studies, supporting studies, limitations,
advantages, and recommended values as numbered sub-headings so the reader/user could
quickly turn to the critical passage for technical support (since it is my expectation that all
potential users will not work their way through the text to find the qualifiers and
considerations that could have been more clearly identified.
2. The studies seem to have been appropriately grouped and fairly presented; clearer
identification, as described in the previous comment above, would help.
3. As Issue Review Comment #1 above suggests, the recommendations could be more clearly
presented in the tables; possible specific suggestions are presented above in comments on
summary tables.
4. A useful chapter here would include shortfalls in applicable activity pattern information for
ventilation rates of workers in a variety of field occupations (indoor and outdoor),
reproducibility of ventilation rates (perhaps the ranges we see reported are reflective of real-
world variability and not shortcomings of experimental design).
***
CHAPTER 5, OTHER FACTORS FOR EXPOSURE CALCULATIONS
p 5-1, sec 5.1, middle of para - should read "...Black females (75.6 years)."
p 5-1, secS.l , last sentence - Given concern for environmental justice and the observation
that minorities may be in locations of exposure to toxics, what is known from Census about
life expectancy of Asians? Hispanics?...in California, for example, these sectors of the
population are on their way to becoming larger in number than the white sector.
p.5-10, sec5.2.2 - Burmaster et al article submitted 2/1/94 for publication - is this a published
B-162
-------
reference now? .. - •
p.5-13, sec 5.2.2, para 2 - heights and weights here are presented in inches and pounds, but
everywhere else in this section, in metric units.
p5-13, sec 5.2.3 - A summary table of recommended values would be useful for readers.
p.5-29, Table 5-17 - Saturday time duration sum is 2440 instead of 1440 - should social
entertainment be 114 instead of 1114 minutes?
p. 5-41, first sentence - It is unclear what a "tomorrow" approach is; a sentence description
would clarify the discussion.
p. 5-51, para 1, line 9 - "One child was randomly selected from an English-speaking
household" suggests that all the rest were from non-English speaking households! Please
rephrase so that it is clear mat only English-speaking households were eligible for survey and
in any selected home, a child was randomly chosen (regardless of age) to participate in the
study.
p. 5-52, Table 5-31, note c - wording is incorrect; column totals may differ from 1440 due to
rounding error.
p.5-66 - There should be some sort of conclusion or recommendation here, after covering so
much time/activity data Recommendations for use?
p. 5-82 - Again, there needs to be some closing thought here - a recommendation or
something...It just sort of stops...
Chapter 5 Issue Review:
1) A great deal of information was presented in this chapter, but it was not really synthesized
into a usable body of data. When so much data is provided, it is often difficult to
recommend what values should be used, but at least, recommendations of study data sets, for
explicitly listed reasons, could be made. In this section, there did not seem to be any.
The presentation could have been better focused; in the face of so many tables and so
much overlapping data, it was difficult to see where any filtering, editing, or judgement about
the quality of the inherent data had taken place; a summary recommendation, similar to that
proposed in the Chapter 3 review above, would help here.
2) Rationale for why one set of study data was chosen to be emphasized over another similar
data set might be useful, but ultimately, a clear summary of what was persuasive for each
study considered would help the reader.
3) (See comments in Chapter 5 Issue Review, #1 above).
4) Suggestions for the discussion of data gaps in the activity pattern area include the
following: how have changes in the economy (shifting to service driven work force with less
B-163
-------
industrial base) and work force (male/female) affected distributions of exposure (the presented
work is generally from the '70s and '80s); have mobility patterns changed (do grown children
stay and live with their parents longer, resulting in longer and different exposures)? How
does one quantify the residential history pattern of off-spring, as opposed to homeowner?
***
CHAPTER 6, CONSUMER PRODUCTS
p.6-7, Table 6-3 - The amount of home solvent product used annually, in mean ounces per
year, presumably is determined on sales information and not based on the percentage of
active ingredient in each of the home products. If it is reasonable to suggest that some home
products have a greater percentage of ingredients of exposure interest, would it be useful to
have a table of percentage of active ingredients (the text prefaced the tables by specifying
methylene chloride or its substitutes)? Would this change apparent perspectives on potential
exposures?
p.6-22, References - Is any other infonnation available, or is Westat the only source of data?
Chapter 6 Issue Review:
1. The data is presented in a useful manner. As suggested in the first comment above for
Chapter 6, some re-ordering of potential exposures, based on active ingredients, might also be
helpful (although it may admittedly be out of the scope of typically available information).
2. Only three studies, all performed by Westat for EPA in 1987, were presented. These may
be the best information available, or the only information available. That being the case,
these are certainly relevant and appropriate for inclusion.
3. This chapter had a recommendation section that could be of great value to the intended
reader. However, given that only a few studies were available to draw from (and all were
performed by one agency), it would seem especially appropriate to summarize the limitations
and uncertainties of the reported work in the final section.
4. The dearth of reportable studies in this area in and of itself are a dat gap for future
research. In addition, it would be useful to learn more about what percent of the ingredients
in the products being tracked are of potential health concern (and order exposure by that
criteria); it would be useful to learn more about the exposure pattern of the actual end user -
it may be that a given product is only used for 10 minutes per event, but how often is the
same person (such as in an occupational setting for a janitor, or aircraft maintenance, or home
cleaning woman) exposed?
***
CHAPTER 8, ANALYSIS OF UNCERTAINTIES
p.8-3, sec 8.1.2, para 2, Iine3 - should read "...such as consumer product preference surveys
B-164
-------
or..."
p. 8-8, sec 8.2. - this section is a valuable introduction and overview of how the entire
document should be interpreted and used; it ought to be an overview comment at the front
end of the book, so that potential users read this section and keep this perspective in mind as
they seek information in the document.
Chapter 8 Issue Review:
1. The types of uncertainties in analyses are presented, one after the other, without example
or much discussion; in that sense, this chapter is much more abstract and different from the
previous chapters. Better sub-section identification of discussion points (such as 8.1.2.1
Sensitivity analysis, 8.1.2.2 analytical uncertainty propagation, 8.1.2.3 probabilistic analysis,
8.1.2.4 classical statistical methods...
2. Key and relevant studies don't apply in the same sense here as in earlier chapters; still,
several studies are appropriately cited for obtaining additional information.
3. Uncertainties have been fairly discussed, but no specific or general recommendations are
given at the end of discussion.
4. (no suggestions to offer on this point)
B-165
-------
-------
John Robinson
B-167
-------
-------
GENERAL COMMENTS ABOUT SURVEY METHODS
Many of the data tables in the handbook are based on social surveys. As such they are subject
to several sources of limitation that affect all surveys- A major problem with the handbook is that
it tends to treat all surveys as equal, w*-—» in fact they vary widely in sophistication and utility
in terms of sample design, field qualit ••»!. question framing and presentation of results.
In general, a, well-conducted survey of the public is expected to meet the following criteria:
1) A probabilistic sampling frame, in which ALL individuals have an e
-------
7-17-1995 9:35AM FROM COPYWORLD/MULTIACCES 5108499701 P. 2
Along the can* vein, it would not seem difficult to end each chapter with a call for needed
seasurement advances to produce the kind of statistical data that would be most appropriate for
policy purpose*.
X further problem arises from the lack of essential data to understand the implications of
what are presented. Thus if we look at drinking wa«er or paint application, what proportion of the
population are involved in the activity for a day or a yoar. Tho percentile data appear virtual?
uninterpretible without such basic statistics that should bo easily available in the original source
(if not the original authors should be chided for onitting it). Many of these parameters are now
available from our 1992-94 MAPS study that should soon be published — copies to bo sent with with
the hard copy of these coraoents.
Additional background information needed in the introduction to tho book include:
1) More detail for the initiate on tarw* like "default values", exposure scenarios" and
"site-specific situations" (page* 1-2)
2) The dif £erenee between Part I and Part K estimates (best noted for all tables throughout
the handbook)
3} Greater discussion of the page 1-5 model and ho« the page 1-8 examples can be interpreted
and considered for policy terms. Would it be possible to provide a graphic display of this model and
how it works for a wll-aeaaured exposure phenomenon,' or for the best understood example.
SOME SPECIFIC COMMENTS ON CHAPTER 3:
The above comments on sample and field weaUneosoc apply very clearly to the various studies
described in Chapter 3. Hot only are these based on small samples, but more importantly on highly
unrepresentative groups, like athletes amd construction worker*, or people living in California. The
applicability to any other population is almost absurd, perhaps even to construction workers in other
partii of the country. (This does not mean that tho data are totally wothless, only that tbeir
limitations should bo clearly noted. If we could chow the proportion of the population in
construction, or other at risk, populations that could also b* useful in understanding the import
of the data. These data could also be used with diary data for these groups to show how divergent
a group they-ore in terms of their activity patterns).
On the other hand, a good deal of irrelevant data are reported, such as on the estimate data
of Shasoo, appear in the text. If it is important, the reader needs to know why. Why show Tables 3-
45,3-47 which have only one entry, where are the "healthy adults" in Table 3-13?
SUGGESTIONS OK CHAPTER 3:
It is probably too early to include the 1992-94 MAPS (NHRPS) data in this chapter, bu-: at
least it should be cited as being available soon, along with (we hope) user-friendly instructions.
Con someone call Bill Kelson (919-541-3184) to get permission to use and cite the report at our
Boeting?
I would arguo that the data and studies on pages 5-14 to 5-30 can be eliminated, given the
more recent and risk-focused data from Robinson ami Thomas (1991). At most, these earlier data could
be cited in the Appondix or referenced in the earlier handbook. The Robinson-Thomas could be moved
into the nain text as a substitute for the excluded material. Note that i am suggesting killing my
own data, which are nov both obsolete and too imprecise for exposure assessment.
Th» Carny data on occupational tenure should be in another separate section. MORS TO COME BY ;)PM!
B-170
-------
More Specifiacally:
-Why not show scwe data in pounds, as well as kilos for us wrtrie retards? p5-4, 5-5.
-what is "Gaussian" p5-13, and bow do you get there from regression?
-Data in Tables 3-1^1,5-lS do not add to 1440 minutes, another reason to drop then
-pS-23 and elsewhere refer to "respondents that" rather than "respondents who"
-Is voluntary nobility useful when involuntary mobility has the sane exposure indications?
-p5-45 limitations do not seem that serious given the final sentence in the paragraph,
-p5-55 limitations are not serious ,
-Tar shis data are trota very poor and ancient surveys; MAPS data are available and far
superior -- also contain esti»ates to compare with Table 5-39
-Table 5-40 data are ancient -• put in Appendix at best (can be inferred from Robinson-Thomas
•Drop Sell data on kids time, as CARS data ar* far cuporior
-Section 5.4.2 Drop HAR data — ?saarpling frame & awful response rate: Census data are
definitive^? 5-79 line 14 demographic (drop^sj
-Who says 100% of people move, ever? My neighbor has lived here for almost 60 years. Data are
confusing at beat.
-Tables 5A-2 t» 3A4 can be dropped, in line with text ctropo
COMMENTS OH SECTION 6:
-p 6-1 discussion confuses sampling and data collection and needs to describe the specific
(difficult) questions asked of respondents, p 6-2 line 4 middle paragraph says "will" iaplyina new
data.
-Biggest question in Table 6.1 -- What % use the product at all and how nan? times for then?
-Why only 208 painters & again what % of those contacted?
-Sonehow it seeas that these details of the HCSTA? study do not give an integrated picture
and could be readily retabulated to give far nore useful data
COMMENTS OH SECTION 7:
-Don't the detailed Table 7-3 data refer to less than 20% of housing units in 'the us?
-Why are the California data in Table 7-6 so ouch higher (93,67,70)?
-Why can't fuller data b
-------
COMHSHTS ON SECTION S:
Thin nhould bo given in tho Introduction, with « brief
COMMENTS ON SECTIOH 2; , '
The data on page 2-14 XRS convergent, despite disparate sources. Mention should be mad* of
this at the outset, San* for the rich data (apparentli) in Tables 2-26 to 2-31.
In contrast, th* fish data fii&efe All over the max>.
COMMENTS OR SECTION 3:
Data fron activity pattern studioo are neoded here. SEE EARLIER COMMENTS.
«*• ?3-13 line 13, except rather than expect
COJWEHTS OK SECTIOH 4;
Whole fcodr data more appropriate for showers and baths than for soil /pesticide exposure.
HT GENERAL:
For each section, somo coBwmtc at the beginning on data needs and hotr adequate the current
are would help greatly, along with thoir relation to models and government policy guidelines, would
be most helpful to non-insiders. , , • . •
B-172
-------
Neil Klepeis
B-173
-------
-------
Neil Klepeis-7; 14/95
REVIEW OF THE EPA's
EXPOSURE FACTORS HANDBOOK
-Activity Patterns-
Contents:
,1. Comments on the Human Activity Patterns (HAP's) material in the Introduction and Chapter 5
of the handbook
2. Suggested new material on general exposure assessments that make use of HAP studies
3. Examples of HAP analyses from the recent national study by EPA (9,386 respondents
nationwide)
Note: The Exposure Factors Handbook is referred to as the handbook below.
Comments on the Introduction, pp. 1-1 to 1-10:
• Much of the introduction is devoted to a discussion of dose, which should be clearly
distinguished from a discussion of exposure.
• A crucial parameter in the dose equation is exposure duration, which can only be obtained
from human activity pattern studies (HAP's)
• Likewise, contaminant concentrations can only be obtained from microenvironmental
monitoring/measurement studies
• Some more discussion is needed in the introduction on how to use exposure durations and
contaminant concentrations to estimate population exposures
• The handbook should contain more background material on exposure — including a complete
definition of exposure as it is distinguished from dose, how exposure fits into the complete risk
model, which exposure factors are most crucial in making accurate population exposure
assessments, definitions of terms and techniques used in exposure assessment, and descriptions
of different exposure monitoring {e.g. the Total Exposure Assessment Methodology (TEAM)
studies by EPA) and modeling efforts (see references 12-17 below).
Review of the Exposure Factors Handbook....
B-175
-------
Neil Klepeis - 7/14/95
For example: '
• The complete risk model can be viewed as a sequence of dependent events: Pollutant Sources
-> Movement of Pollutants -> Exposure to Pollutants -> Dose -> Health Effects
• In this model total human exposure (THE) is defined as when a person is present in some
location at some time and the concentration of a pollutant is present at the same location at the
same time. In this way a pollutant concentration can come into contact with a person via the
air (lungs, skin), water (gut, skin), soil (gut, skin), or food (gut) pathways at any given instant.
The emphasis in THE assessment is on human beings and the sources of chemical toxins in
their immediate surroundings (environmental tobacco smoke, household goods/services, etc.)
• There can be multiple routes of exposure for different chemical pollutants, e.g., chloroform via
both air and water
• Predictions of dose require knowledge of metabolism, absorption, etc., which can be based on
body weight, inhalation rate, etc.
• Predictions of exposure require the study of factors leading up to the exposure event (Pollutant
Sources and Movement of Pollutants): chemicals emitted or present, emission rates, air
exchange rates, deposition rates, chemical reactions, reaction rates, etc.
• Modeling human exposure to air pollution requires the concentrations of pollutants at specific
locations (from monitoring/measurement studies), and the times that people spend there (from
activity patterns)
References on total human exposure concepts:
1. Ott, W., (1985), 'Total Human Exposure: An emerging science focuses on humans as
receptors of environmental pollution", Feature Article, Environmental Science and
Technology, Vol. 19, pp. 880-885.
2. Ott, W., (1990) "Total Human Exposure: Basic Concepts, EPA Field Studies, and Future
Research Needs," Journal of Air & Waste Management Association.. Vol. 40, No. 7, pp.
966-975.
4. M. Fugas, (1975) "Assessment of Total Exposure to Air Pollution," in Proceedings of the
International Conference on Environmental Sensing and Assessment,, Las Vegas, NV,
Paper No. 38-5, Vol. 2, IEEE #75-CH1004-l ICESA.
5. N. Duan, (1982) "Microenvironment Types: A Model for Human Exposure to Air Pollution,"
Environment International, Vol. 8, pp. 305-309.
6. Ott, W., (1982), "Concepts of Human Exposure to Air Pollution", Environment
International, Vol. 7, pp. 179-196.
Review of-the Exposure Factory Handbook.
B-176
-------
Neil Klepeis - 7/14/95
Comments on Chapter 5, pp. 5-1 to 5-86:
• Quantities such as body weight, inhalation rates, etc. are relevant for dose or health risk
assessments, but not for exposure assessments. Quantities directly related to exposure
assessment are microenvironment duration (related to life expectancy), averaging time, and use
of consumer products. Exposure is simply the confluence of a pollutant concentration and a
person in time and space, whereas dose is the amount that enters the person's system, i.e.,
blood stream. These ideas should be clarified in the introduction to Chapter 5.
• The introduction to Section 5.3 should be expanded to include more discussion of the use of
human activity patterns in total human exposure assessment (see detailed suggestions below).
• In addition to those studies described in Section 5.3, the handbook should include several
analyses of the California Air Resource Board's 1987-88 California Activity Pattern study on
exposure to environmental tobacco smoke (ETS), and by time-of-day (see references below).
• Human exposure is highly correlated with time-of-day, and recent HAP studies are very well
suited for analyses by time-of-day since the data is collected in 24-hour diaries with minute
resolution (see reference 10 below)
• Analysis of the recent national human activity pattern study by EPA should be included in the
handbook pending its completion (reference 11)
• In a section on future work, the handbook should include suggestions for improved human
activity pattern studies including:
=> better exposure-relevant activity categories in 24-hour diaries
=> inclusion of only those follow-up questions that have been shown to have a high
response rate in the past
• There does not appear to much data on the fraction of time spent in microenvironments, which
is useful to determine their relative significance to the entire population
• Usefulness to Exposure Assessors: Most of the data presented is in terms of mean
microenvironment durations or total minutes of time spent in different microenvironments
(mins/day). These.data will be useful for point estimates of exposure. However, to conduct
probabilistic exposure assessments that produce frequency distributions of exposure, it is
necessary to have as input into the model either: (1) frequency distributions of the time spent in
microenvironments, or (2) the raw data. It is impractical for reports on HAP studies to include
all the desired frequency distributions for all possible exposure assessments. Thus, more
Review of the Exposure Factors Handbook....
B-177
-------
Neil Klepeis - 7/14/95
emphasis in the handbook should be placed on how to use the raw HAP data in probabilistic
exposure assessments, than on comprehensive data listings. See suggestions below.
Other Activity Pattern Analyses:
3. Jenkins, P. L., Phillips, T. J., Mulberg E. J., and Hui, S.P., (1992) "Activity Patterns of
Califomians: Use of and Proximity to Indoor Pollutant Sources", Atmospheric
Environment, Vol 26A, No. 12, pp. 2141-2148,
4. J. Wiley, J. Robinson, T. Piazza, K. Garrett, K. Cirksena, U. Cheng and G. Martin,
(1991) "Activity Patterns of California Residents", Final Report Under Contract No. A6-
177-33, California Air Resources Board, Sacramento, CA.
5. J. Wiley, J. Robinson, T. Piazza, L. Stork and K. Pladsen, (1991) "Study of Children's
Activity Patterns", Final Report Under Contract No. A733-149, California Air Resources
Board, Sacramento, CA.
6. Robinson, J.P. and Blaire, J., (1995) "Estimating Exposure to Pollutants Through
Human Activity Pattern Data: The National. Microenvironmental Activity Pattern
Survey", Annual Report, Survey Research Center, University of Maryland.
7. J. P. Robinson, P. Switzer, W.R. Ott, (1994) "Smoking Activities and Exposure to
Environemental Tobacco Smoke (ETS) in California: A Multivariate Analysis", Report
No. 1 for the California Activity Pattern Survey, Department of Statistics, Stanford
University, Stanford, CA.
8. J. P. Robinson, P. Switzer, W.R. Ott, (1994) "Exposure to Environmental Tobacco
Smoke (ETS) Among Smokers amd Nonsmokers", Report No. 2 for the California
Activity Pattern Survey, Department of Statistics, Stanford University, Stanford, CA.
9. J. P. Robinson, P. Switzer, W.R. Ott,, (1994) "Microenvironmental Factors Related to
Califomians' Potential Exposures to Environmental Tobacco Smoke (ETS)", Report No.
3 for the California Activity Pattern Survey, Department of Statistics, Stanford
University, Stanford, CA.
10. W.R. Ott, P. Switzer, J. P. Robinson, (1994) "Exposures of Califomians to
Environmental Tobacco Smoke (ETS) by Time-of-Day: A Computer Methodology for
Analyzing Activity Pattern Data", Report No. 4 for the California Activity Pattern
Survey, Department of Statistics, Stanford University, Stanford, CA.
11. Klepeis, N., and Tsang, A. (1995) "Analysis of the National Human Activity Pattern
Study from a Viewpoint of Human Exposure Assessment", EPA Report in Preparation,
EMSL, Las Vegas, NV.
Additional Material/Clarifications to Include:
The handbook should contain a guide to conducting exposure assessments -with human
activity pattern (HAP) studies including examples of studies that have been done (see list of
references below). Some or all of the following ideas should be considered for expanded discussion
in the handbook:
Review of the Exposure Factors Handbook...
B-178
-------
Neil Klepeis - 7/14/95
By providing microenvironment durations, HAP studies are useful to compare relative
potential exposures between segments of the population without ever knowing the exposure
magnitudes. It is necessary to assume that the mean pollutant concentrations in each
microenvironment are approximately the same across different subgroups (region, age, gender,
etc.), i.e., the exposure mechanisms (source strengths, air exchange rates, deposition rates,
etc.) do not change appreciably for different socio-economic or geographical groups.
HAP's are probably most useful for comparisons of relative potential exposures from air
pollutants since these exposures are approximately proportional to the duration of time spent in
a microenvironment
Dermal, ingestion, etc. exposures require more complicated assessments (surface area, volume
eaten/applied, concentrations of toxins, etc.); and actual exposures may vary greatly between
subgroups due to unspecified factors even though the exposure durations are comparable.
HAP's may not be useful to model these exposures unless they also collect data (or are
combined with data from other studies) on the amount of material that is being ingested or
coming into contact with skin during the appropriate microenvironments
HAP's can be used for a complete population exposure assessment (giving either point
estimates or frequency distributions of exposure) by combining measurements of the magnitude
of air exposures in microenvironments for specific segments of the population with the amount
of time people spend being exposed — as obtained from HAP studies.
For a complete and accurate weighting of microenvironmental exposures by the amount of time
spent in each microenvironment, the population should be divided into subgroups that have
been shown to have different exposure magnitudes. If deterministic models, i.e., the mass
balance equation (see references below), are being used, then different parameters need to be
determined for each different subgroup.
Point estimates of population exposure to air pollutants can be made by multiplying the mean
microenvironmental exposure experienced by each subgroup by the fraction of time spent in
the microenvironment by that subgroup (i.e. weighting each microenvironmental exposure by
the fraction of time spent there), and summing over each of these contributions to obtain the
overall exposure.
In a probabilistic exposure assessment, frequency distributions of microenvironmental
exposure magnitude and exposure duration are Monte-Carlo sampled to predict population
exposure. When deterministic submodels (air exchange, source strength, etc.) are used to
Review of the Exposure Factors Handbook...
B-179
-------
Neil Klepeis - 7/14/95
predict exposure magnitudes, some model parameters may be correlated and a joint frequency
distribution should be calculated (as discussed in Chapter 8 of the handbook).
Some references for past or ongoing population exposure assessments:
12. Ott, W., (1984) "Exposure Estimates Based on Computer Generated Activity Patterns,"
Journal of Toxicology: Clinical Toxicology, Vol. 21,, pp. 97-128.
13. Ott W., J. Thomas, D. Mage, and L.Wallace, (1988) "Validation of the Simulation of
Human Activity and Pollutant Exposure (SHAPE) Model Using Paired Days from the
Denver, CO, Carbon Monoxide Field Study," Atmospheric Environment, Vol. 22, No.
10, pp. 2101-2113.
14. Ott W., Mage, D., and Thomas, J., (1992) "Comparison of Microenvironmental CO
Concentrations hi Two Cities for Human Exposure Modeling," Journal of Exposure
Analysis and Environmental Epidemiology, Vol. 2, No. 2,, pp. 249-267.
15. Lurmann, F. W. and Korc, M. E. (1994) "Characterization of Human Exposure to
Ozone and PM-10 in the San Francisco Bay Area", Final Report STI-93150-1416 FR,
for the BAAQMD, San Francisco, CA.
16. Behar, J.V., Thomas, J., and Pandian, M.D., "Estimation of the Exposure to Benzene of
Selected Populations in the State of Texas Using the Benzene Exposure Assessment
Model (BEAM)", EPA 600/X-93/002, Environmental Monitoring Systems Laboratory,
U.S. Environmental Protection Agency, Las Vegas, NV, January 1993.
17. Klepeis N. E., Ott W., and Switzer P., (1994) "A Total Human Exposure Model
(THEM) for Respirable Suspended Particles (RSP)", National Technical Information
Service (NTIS) No. PB94-197415, Presented at the 87th annual meeting of the
A&WMA meeting in Cincinnatti, OH.
List of references on deterministic submodels for predicting microenvironmental exposures to
environmental tobacco smoke (ETS):
18. Switzer, P., and Ott, W. (1992) "Derivation of an Indoor Air Averaging Time Model
from the Mass Balance Equation for the Case of Independent Source Inputs and Fixed
Air Exchange Rates," Journal of Exposure Analysis and Environmental Epidemiology,
Vol. 2, Suppl. 2, pp. 113-135.
19. Ott, W., Langan, L., and Switzer, P. (1992) "A Time Series Model for Cigarette
Smoking Activity Patterns: Model Validation for Carbon Monoxide and Respirable
Particles in an Chamber and an Automobile," Journal of Exposure Analysis and
Environmental Epidemiology, Vol 2, Suppl. 2, pp. 175-200.
20. Klepeis, N., Ott, W., Switzer, P, (1995) "Modeling the Time Series of Carbon Monoxide
and Respirable Suspended Particles from Multiple Smokers: Validation in Two Public
Smoking Lounges", presented at the 88th Annual Meeting and Exhibition of the
A&WMA, San Antonio, TX, June 1995.
21. Ott, W., Klepeis, N., and Switzer, P., (1995) "Modeling Environmental Tobacco Smoke
in the Home Using Transfer Functions", presented at the 88th Annual Meeting of the
A&WMA, San Antonio, TX, June 1995.
Review of the Exposure Factors Handbook..
B-180
-------
Neil Klepeis - 7/14/95
The following is a list of ideas that are important when analyzing HAP's. They might be
included in the introduction or in a subsection of Chapter 5.
• In using HAP studies to estimate relative exposures via the air pathway, the significance of a
microenvironment is determined by the amount of time spent experiencing them, i.e. their
duration. These significances are best compared for similar microenvironments since sources
(exposure magnitudes) vary from micro, to micro.
• Exposure magnitudes (obtained from monitoring/measurement studies) must be used in
conjunction with exposure durations to obtain accurate population exposures.
• The proportion of respondents in microenvironments, proportion of time spent in
microenvironments, frequency of occurrence of microenvironments, and mean durations of
microenvironments are used to approximate the relative significance of microenvironmental
exposure (assuming exposure depends mostly on duration) and to compare exposures between
subgroups (gender, age, race, region, etc.).
• . Analysis over all respondents (the doers — those experiencing each microenvironment -- plus
non-doers) indicates the significance of each microenvironment to the population as a whole.
Analysis of only the doers indicates the significance of each microenvironment to the pool of
respondents that are being exposed (see Table 1). See examples using the recent national study
presented below.
• When analyzing HAP studies, it is usually appropriate to weight each subgroup according to
the proportion of respondents in the "true" population, e.g., to compensate for oversampling.
Examples from the Recent EPA National Human Activity Pattern Survey:
Figures 1 to 4 contain some of the recent results of the national human activity pattern
study by EPA (reference 11): time-of-day analysis by location, percentage of time spent in each
location x activity microenvironment (over all respondents), percentage of respondents
experiencing a given microenvironment on the diary .day, and the 24-hour mean duration of
microenvironments (doers only). The most significant microenvironments over the 24-hour diary
days of the entire population (besides those involving non-exposure activities like sleeping) are
Eating/Drinking, Food Preparation, Housekeeping, and Bathing — all in the Residential-Indoor
location (Figure 2). These microenvironments are also among those that have the highest number
Review of .the Exposure Factors Handbook 7
B-181
-------
Neil Klepeis - 7/14/95
of respondents experiencing them on the diary day (Figure 3). Since its 24-hour duration (Figure
4) is one of the smallest (25 min), we can see that the overall significance of the Residential-
Indoor-Bathing microenvironment is due more to the number of respondents engaging in it than to
the amount of time it takes up. In addition, Housekeeping has about the same overall percentage of
time as Food Preparation, but it has a smaller fraction of respondents experiencing it. Thus, its
relative significance arises from its larger mean 24-hour duration.
Review of the Exposure Factors Handbook..
B-182
-------
1
B
"?
I
1
•3
o
I
1
I
I
H
"3
I
+
S5
i
i
Q
Io
it
fie
a* i
•s
ss
«
s
I
I
I
I
B-183
-------
I
vt
'S
§
o
T3
MS
llz
zi
8Z
2?
S3
ill
'1
a g
i3
3?
83
35
3
-38
•i
Near Vehicle
iz
I1
ll
Z8
?a
-2
28
i-Z
2§
8,
*2
a?.
2"
8z
22
ia
I-
13
38
i3
22
3a
"1
38
53
38
Sa
"3
32
a 3 s g "S
a-3
33-1
o
-§
A
I
O
izi
I
S"
• a
13
3i
h
S^
13
,^
i3
3i
h
a g s
1
III.
8z
- a 2
3 S
13
3i
la
38
83
Hgi
ii3
83
S i
wa
^
S
!iz
I1
iz
Z8
Pa
3a?
fi
J*
38
P3
]g?
38
§3
I
Si
3
3 S 3
B-184
-------
I
I
l-g
I
o
£ ffl c
Q-OC
•o'SS
O CD S
.?.ci
• « a
c «
« |
tsc 2 °>
?•= te c
2
dq gi^ 552
•~6i-Su5 c
""Sin
sd«
SH
s
joopui Jeqjo'OOl
juBJneisea/JBg'06
PI3 ouqnd/iooqos'08
AjojOBj/ooijiO'09
joopjnQ
OOpU|-|B|lU9plS9a'0 1
1
•a
a-
§,
'1
}uads
*
B-185
-------
I
tn
I
1
,fc
jnBjnBjsa-a/reg'06
•ioopuiMqio'001
•53
1
I
B-186
-------
$
ta
•8
1
fj a §
\ \\\\\\y
\ \ \\\\\Y^
\\\\\\\\
\\ \\AAAA
.s
gas
rig
03 _, W
I
I
*
B-187
-------
-------
Val Schaeffer
B-189
-------
-------
U.S. CONSUMER PRODUCT SAFETY COMMISSION
WASHINGTON, B.C. 20207
August 21, 1995
Helen Murray
Eastern Research Group
HOHartwell Ave.
Lexington, MA 02173-3198
Re: Comments on June 1995 Draft of the Exposure Factors Handbook
Dear Ms. Murray:
Thank you for the opportunity to comment on the updated draft of the EPA Exposure Factors
Handbook (EFH). Since neither Lori Saltzman or I could participate in the peer review workshop held
on July 25 and 26, 1995,1 am submitting post-meeting comments as we discussed. My review is
restricted to Chapters 3, 5, 6, and 8 since Lori was assigned to the workshop's Activity Patterns panel.
In general, Chapters 3 and 5 provide comprehensive and exhaustive reviews of multiple studies but the
information needs to be better organized and presented. On the other hand, Chapters 6 and 8 provide
information in a succinct and organized manner, but treat the subject matter in a cursory fashion.
Specific comments by chapter are as follows.
Chapter 3 - Inhalation Route
This chapter is a superior review of ten studies that determine inhalation rates by a number of
methods in a variety of populations under a range of activity levels. Eight of the ten studies were
reported since the last edition (1989) of the EFH and represent a sizable new data base. The studies
grouped as "key studies" and the recommended inhalation rate values are appropriately chosen. The
limitations of the various studies are adequately stated. The summary table (3-22) is particularly
helpful. Unfortunately, the data is not presented in a user-friendly way. The most useful data are the
recommended long-term and short-term inhalation rates for children and adults found at the end of the
chapter on pages 3-40 to 3-48. This means the interested reader has to go through 40 pages of study
descriptions and 21 tables of often conflicting data before reaching the critical information. I would
suggest that the recommended inhalation rates be presented early in the chapter followed by a briefer
discussion of studies that contributed to the values and an explanation of why the revised
B-191
-------
Val Schaeffer/CPSC
recommended breathing rates are superior to those advocated in the 1989 EFH. The more detailed
data tables and study descriptions can be incorporated in an appendix.
Chapter 5 - Other Factors for Exposure Calculations
Many of the comments to Chapter 3 also apply to Chapter 5. This chapter covers lifetime-,
body weight, activity patterns, and population mobility. The first two topics are relatively
straightforward and well treated. The available data on human activity patterns is more varied and
complex. While the key studies are identified, there was little attempt to group the data and present it
in a way that would be useful. Eleven studies are described and 35 data tables are introduced but no
clear guidance is provided as to. the most appropriate data to use. Recommended values or data sets
for activity patterns by age, sex, race, employment status, weekday/weekend, and season need to be
clearly stated early hi section 5.3. The same comments apply to the five studies and nine data tables
presented on population mobility.
Chapter 6 - Consumer Products
This chapter is restricted to usage data presented from three national surveys of selected
consumer product categories performed by Westat in the mid 1980's for the EPA Office of Pollution,
Prevention, and Toxics (OPPT). Abt Associates conducted a follow-up 1991 consumer use survey,
under contract with CPSC, for three of the product categories; paint strippers, aerosol spray paints, and
adhesive removers. The telephone survey of nearly 5000 respondents was modeled after the 1987
Westat usage survey of household solvent products using the random digit-dialing technique. The
pertinent data tables from the CPSC survey are enclosed. They can be used to update tables 6-1
through 6-4 for the three product categories. The source document is entitled Methylene Chloride
Consumer Use Study - Final Report, Abt Associates, December 1991. It can be obtained from the
CPSC Directorate for Economic Analysis by contacting Charles Smith (301-504-0962 x!325) or Bill
Zamula (301-504-0962 x!331). The document title and CPSC contacts were previously provided to
the Office of Health and Environmental Assessment (OHEA) at the 1993 EFH workshop.
The Westat and Abt surveys also supplied usage information on the location (outside, garage,
living room, etc.) and the indoor ventilation conditions (windows, doors open/closed) in which the
product categories were used. This information is necessary hi order to characterize exposure and
should be either presented in the chapter or its availability acknowledged in the study descriptions.
Table 6-3 on the amounts of various household products used would be more useful if the data were
»
presented as ounces per use rather than ounces per year.
It is disappointing that OHEA chose not to pursue the recommendations of the Activity
Patterns panel of the 1993 workshop for the presentation of consumer product-related exposure factors.
This included providing data on chemical composition and chemical emission factors as well as usage
B-192
-------
Val Schaeffer/CPSC
information for a limited number of reasonably well-studied consumer product categories. There is an
ever-increasing amount of formulation and emission rate information being gathered by various EPA
offices. This includes OPPT through its use cluster projects, the Indoor Air Division through its
indoor air source ranking data base, the Air Pollution Prevention and Control Division through its
product emissions testing, and the Office of Air Quality Planning and Standards through its study of
volatile organic compound emissions from consumer and commercial products. The EFH could be
used as a vehicle to provide some of this exposure-critical data in a structured and organized manner.
Paints might be a good product candidate to try this approach since data has been gathered on this
category in the above EPA efforts.
Another recommendation of the 1993 workshop panel was to provide some general guidance
on how the different consumer product-related factors would be used to assess consumer exposure.
The current chapter does not reflect this advice and EPA is encouraged to include this. At a
minimum, other documents (e.g. Standard Scenarios for Estimating Exposure to Chemical Substances
During Use of Consumer Products, EPA Contract No. 68-02-3968, 1986) should be cited as
references.
Chapter 8 - Analysis of Uncertainties
This chapter is a satisfactory introductory description of the terms and general principles
involved in uncertainty analysis. If feasible, it should be augmented by providing a framework and
some approaches to conducting an assessment of uncertainty. The chapter needs to more clearly
distinguish between characterizing exposure variability., that is the heterogeneity hi exposure received
by a population of individuals and characterizing exposure uncertainty, which is the lack of knowledge
of the true value of a particular exposure estimate. It should state early-on that the two topics should
be treated separately, and variability is sometimes erroneously included in the analysis of uncertainty.
The discussion of the Monte Carlo technique could be more positive. While there are
certainly cautions that must be exercised when using this method, it is a reasonable way to characterize
exposure variability and uncertainty provided sufficient information is known about the frequency and
probability distributions of the different exposure parameters and the dependencies among them. The
Monte Carlo technique is not cumbersome in terms of assessing sensitivity since modern day software
can run large numbers of simulations quickly and the method does not assume parameter
independence. Finally, the quantitative alternatives to Monte Carlo analysis are usually more difficult
to compute and generally more problematic.
B-193
-------
Val Schaeffer/CPSC
If there are questions regarding the comments or if further assistance is needed, I can be
reached at 301-504-0994 x!390/fax 301-504-0025.
Val Schaeffer, Ph.D.
Directorate for Epidemiology and Health Sciences
Enclosures:
cc: William Wood, EPA Risk Assessment Forum (letter only)
B-194
-------
Table 2-4: Number of Times of Use of Paint Removers/Strippers Within the Last 12
Months - Recent Usersa
Mean
Standard deviation
Minimum
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=316)
3.54
7.32
1.00
1.00
1.00
1.00
1.00
2.00
3.00
6.00
12.00
50.00
70.00
1986 Study
(Unweighted N=761)
3.68
9.10
0.03
0.03
0.23
0.69
4.0b
2.00
3.00
6.00
11.80
44156
100.00
aRecent users are those who have used, the product in the last year and purchased the product
in the past two years.
^Values are inconsistent with other values in this column.
B-195
-------
Table 2-5: Minutes Spent Using Paint Removers/Strippers Last Time Used
Users3
Recent
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=390)
144.59
175.54
2.00
5.00
15.00
20.00
45.00
120.00
180.00
360,00
480.00
720.00
1440.00
1986 Study
(Unweighted N=752)
125.57
286.59
0.02
0.38
5.00
5.00
20.00
60.00
120.00
240.00
420.00
1200.00
4320.00
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-196
-------
Table 2-6: Minutes Spent in the Room After Last Use of Paint Remover/Stripper
Recent Usersa
Including those who did not spend any time in room after use
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=309)
12.96"
85.07
0.00
0.00
0.00
0.00
0.00
0.00
0.00
10.00
60.00
180.00
1440.00
1986 Study
(Unweighted N=748)
31.38
103.07
0.00
0.00
0.00
0.00
0.00
0.00
20.00
60.00
180.00
541.20
1440.00
Including only those who spent time in the room
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile •
99th Percentile
Maximum Value
(Unweighted N=39)
93.88
211.71
1.00
1.00
1.00
3.00
10.00
60.00
120.00
180.00
420.00
1440.00
1440.00
(Unweighted N=340)
NA
NA
1.00
1.00
1.00
3.10
10.00
30.00
60.00
180.00
240.00
826.20
1440.00
'Statistically significant at the .05 level
"Statistically significant at the .01 level
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-197
-------
Table 2-7: Amount of Paint Remover/Stripper Used — Recent Users2
Fluid Ounces of Paint Remover/Stripper used in the past year
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=307)
142.05"
321.73
15.00
15.00
16.00
16.00
32.00
64.00
128.00
256.00
384.00
1920.00
3200.00
1986 Study
(Unweighted N=737)
63.73
144.33
0.64
1.50
4.00
8.00
16.00
32.00
64.00
128.00
256.00
512.00
2560.00
Fluid Ounces per use of Paint Removers/Strippers
•
Mean
Standard Deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
(Unweighted N=307)
64.84"
157.50
.35
2.67
8.00
10.67
16.00
32.00
64.00
128.00
192.00
320.00
2560.00
(Unweighted N= 735)
29.84
50.28
0.23
0.651
1.60
2.67
7.15
16.00
32.00
64.00
128.00
256.00
512.00
'Statistically significant at the .05 level
"Statistically significant at the .01 level
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-198
-------
Table 3-4 Number of Times of Use of Spray Paint Within the Last 12 Months - Recent
Usersa
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=775)
8.23"
31.98
1.00
1.00
1.00
1.00
1.00
2.00
4.00
11.00
20.00
104.00
365.00
1986 Study
(Unweighted N= 1178)
4.22
15.59
1.00
1.00 .
1.00
1.00
1.00
2.00
4.00
6.10
12.00
31.05
365.00
'Statistically significant at the .05 level.
"Statistically significant at the .01 level.
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-199
-------
Table 3-5: Minutes Spent Using Spray Paint Last Time Used — Recent Users3
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentiie
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=786)
40.87
71.71
1.00
1.00
3.00
5.00
10.00
20.00
45.00
90.00
120.00
360.00
960.00
1986 Study
(Unweighted N=NA)
39.54
87.79
2.00b
0.17b
2.00
5.00
10.00
20.00
45.00
60.00
120.00
300.00
1800.00
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
^Values are inconsistent with other values in this column.
B-200
-------
Table 3-6: Minutes Spent in the Room After Last Use of Spray Paint — Recent Users2
Including those who did not spend any time in room after use
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=791)
3.55"
22.03
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
120.00
300.00
1986 Study
(Unweighted N= 1158)
12.70
62.80
0.00
0.00
0.00
0.00
0.00
0.00
1.00
30.00
60.00
260.50
1440.00
Including only those who spent time in the room
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
(Unweighted N=35)
65.06
70.02
1.00
1.00
1.00
10.00
15.00
30.00
60.00
120.00
120.00
300.00
300.00
(Unweighted N=305)
NA
NA
1.00
1.00
1.00
2.00
5.00
15.00
60.00
120.00
222.00
480.00
1444.00
'Statistically significant at the .05 level. '
"Statistically significant at the .01 level.
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-201
-------
Table 3-7: Amount of Spray Paint: Used — Recent Users2
Fluid Ounces of Spray Paint used in the past year
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=778)
83.92"
175.32
13.00
13.00
13.00
13.00
13.00
26.00
.65.00
156.00
260.00
1170.00
1664.00
1986 Study
(Unweighted N= 1121)
30.75
52.84
0.02
0.75
2.01
3.25
7.00
13.00
32.00
65.00
104.00
240.00
1053.00
Fluid Ounces per use of Spray Paint
Mean
Standard Deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
(Unweighted N=778)
19.04"
25.34
0.36
0.36
3.47
6.50
9.75
13.00
21.67
36.11
52.00
104.00
312.00
(Unweighted N= 11 18)
13.80
24.40
0.01
0.19
0.80
1.50
3.50
8.00
16.00
26.00
39.00
96.00
526.50
"Statistically significant at the .05 level.
"Statistically significant at the .01 level.
aRecent users ace those who have used the product in the last year and purchased the product
in the past two years.
B-202
-------
Table 4-4: Number of Times of Use of Adhesive Removers Within the Last 12 Months
Recent Users3
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current .Study
(Unweighted N=58)
1.66"
1.67
1.00
1.00
1.00
1.00
1.00
1.00
2.00
3.00
5.00
12.00
12.00
1986 Study
(Unweighted N=167)
4.22
12.30
1.00
1.00
1.00
1.00
1.00
1.00
3.00
6.00
16.80
100.00
100.00
'Statistically significant at the .05 level.
"Statistically significant at the .01 level.
aRecent users are those who have used the product hi the last year and purchased the product
in the past two years.
B-203
-------
Table 4-5: Minutes Spent Using Adhesive Removers Last Time Used — Recent Usersa
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=52)
172.87
304.50
5.00
5.00 ;
10.00
15.00
29.50
120.00
240.00
480.00
1440.00
1440.00
1440.00
1986 Study
(Unweighted N= 1 68)
121.20
171,63
0.03 1
0.03
1.45
3.00 i
15,00
60.00
120.00
246.00
480,00
960.00
960.00
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-204
-------
Table 4-6 Minutes Spent in the Room After Last Use of Adhesive Remover — Recent
Usersa
Including those who did not spend any tune hi room after use
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=51)
13.79"
67.40
0.00
0.00
0.00
0.00
0.00
0.00
0;00
0.00
120.00
420.00
420.00
1986 Study
(Unweighted N= 166)
94.12
157.69
0.00
0.00
0.00
0.00
1.75
20.00
120.00
360.00
480.00
720.00
720.00
Including only those who spent time in the room
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
(Unweighted N=5)
143.37
169.31
5.00
5.00
5.00
5.00
20.00
120.00
420.00
420.00
420.00
420.00
1440.00
(Unweighted N= 131)
119.3
NA
1.00
1.00
1.60
4.00
10.00
60.00
120.00
420.00
504.00
720.00
720.00
'Statistically significant at the .05 level.
"Statistically significant at the .01 level.
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-205
-------
Table 4-7: Amount of Adhesive Remover Used — Recent Usersa
Fluid Ounces of Adhesive Remover used in the past year
Mean
Standard deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum Value
Current Study
(Unweighted N=51)
96.95*
213.20
13.00
13.00
13.00
16.00
16.00
32.00
96.00
128.00
384.00
1280.00
1280.00
1986 Study
(Unweighted N=155)
34.46
96.60
0.25 .
0.29
1.22
2.80
6.00
10.88
32.00
64.00
138.70
665.60
1024.00
Fluid Ounces per use of Adhesive Removers
Mean
Standard Deviation
Minimum Value
1st Percentile
5th Percentile
10th Percentile
25th Percentile
Median Value
75th Percentile
90th Percentile
95th Percentile
99th Percentile
Maximum -Value
(Unweighted N=51)
81.84*
210.44
5.20
5.20
6.50
10.67
16.00
26.00
64.00
128.00
192.00
1280.00
1280.00
(Unweighted N= 153)
22.04
85,44
0.04
0.06
0.33
0.67
3.00
8.00
16.00
32.00
64.00 .
574.72
1024.00-
'Statistically significant at the .05 level.
"Statistically significant at the .01 level.
aRecent users are those who have used the product in the last year and purchased the product
in the past two years.
B-206
-------
Work Group #4
Housing characteristics
and
indoor environments
B-207
-------
-------
Bert Hakkinen
B-209
-------
-------
_ Hakkinen 1
Procter&Gamble
The Procter & Gamble Company
Winton Hill Technical Center
6300 Center Hill Avenue. Cincinnati, Ohio 45224-1795
July 14, 1995
Ms. Helen Murray
Eastern Research Group
110 Hartwell Avenue
Lexington, MA 02173-3198
Fax #: 617-674-2906
Comments About June 1995 External Review Draft ofEPA/600/P-95/002A "Exposure
Factors Handbook" Update
This contains my comments as a reviewer of the above document. As assigned, I have focused my
review and comments on the housing characteristics and indoor environments portions of this
document, but have also commented on other sections. As requested by EPA, my comments have
kept the following issues in mind:
• Are the data presented in a way that is useful to exposure assessors?
• Are the data presented in the best way?
• Are the data presented in a way that will support both point estimate and Monte Carlo
assessments?
• Have the studies been appropriately grouped into key studies and other relevant studies?
• Are the recommendations at the end of the sections based on a proper.interpretation of the key
studies, and have the limitations/uncertainties been appropriately emphasized/described?
• What can be suggested about data gaps and research needs for each section?
As noted below in the comments about Sections 6 and 7, a broad recommendation would be to ask
readers to help ensure that all potentially useful exposure data are included in future revisions of
the Exposure Factors Handbook. A similar message could be sent to various trade associations,
key academicians, and others. Could someone in EPA be designated'in the handbook as a possible
contact to receive data for possible inclusion in future revisions? If done in this way, perhaps
periodic updates listing new data for the various sections could be sent to known users of the
Exposure Factors Handbook, or even posted as an update file accessible via EPA's Internet World
Wide Web "home page." These updates could include a statement that the data are, as yet,
unreviewed for final inclusion in upcoming editions of the handbook.
B-211
-------
Procter&Gamble
Hakkinen
Building on the above comments, I also feel that serious consideration be given to making the entire
handbook available on-line via EPA's Internet World Wide Web home page, and/or putting it on a
searchable CD (similar to how encyclopedias are now available and searchable as CDs). The
current about 1,000 pages is not very user friendly to store, search, and transport, and the.
technology exists to make it much more user friendly and accessible to exposure assessors. If
accessible via an Internet "home page," that means could also be used by EPA to receive comments
and information for consideration: and possible inclusion in future revisions.
Another general comment is that the American Industrial Health Council's Exposure Factors
Sourcebook should be carefully reviewed to see if any of its contents should be added to the revised
Exposure Factors Handbook. Also, the various ways data are presented in the AIHC document are
user friendly, and perhaps could be added to the revised Exposure Factors Handbook, e.g., the
AIHC document has very nice figures showing adult body weight distributions based on
information in tables from the original Exposure Factors Handbook.
The following additional comments are organized in the same order as the contents of the
handbook.
Section 1. Introduction
Page 1-2. After reviewing this document, I recommend that thought be given to returning to a two
part handbook, with the second part containing standard (or commonly applied) scenarios. A key
reason for this recommendation is the largely increased size of the handbook which makes it
intimidating and more difficult to find relevant information. A second part containing standard
scenarios would enable users to quickly find and understand the types of exposure assessments that
are commonly performed, and the variety of data and assumptions needed for the assessments.
Adequate cross-referencing and warnings about the need to consider the uniqueness of site-specific
situations would help encourage exposure assessors to apply data and assumptions focused on then-
particular exposure assessment needs without the need for separate guidance and support
documents. A "key words" master index would also help readers use the document more easily.
Section 2. Ingestion Route
No comments.
Sections. Inhalation Route
The following are potentially usefiil publications on respiratory volume as a function of the month
ofpregnancy:
B-212
-------
Procter&Gambte
Hakkinen 3
• Spatling, L. et al. The Variability of Cardiopulmonary Adaptation to Pregnancy at Rest and
During Exercise. BRITISH JOURNAL OF OBSTETRICS AND GYNAECOLOGY 99,
Supplement 8: 1-40 (1992). Has respiratory minute volume and other related information
as a function of the month of pregnancy. .
• Clapp, J. F. et al. Maternal Adaptations to Early Human Pregnancy. AMERICAN
JOURNAL OF OBSTETRICS AND GYNECOLOGY 159: 1456-60 (1988). Has data
similar to the Spatling publication.
• Pernoll, M. L. et al. Ventilation During Rest and Exercise in Pregnancy and Postpartum.
RESPIRATION PHYSIOLOGY 25: 295-310 (1975). Has data similar to the above two
publications.
Section 4. Dermal Route
For the sake of accuracy, the following publication may be worth noting in this section:
Slone, T. H.
Letter to the Editor on "Body Surface Area Misconceptions."
RISK ANALYSIS. 13:375-377(1993).
"Clearly the skin's surface is heterogeneous; it consist of numerous desquamating.
scales, sweat pores, follicular orifices, and follicles with hairs... One would obtain
different values depending on whether one is examining nearly flat nonfolllicular areas
or the three-dimensional follicles... There is insufficient evidence that surface area has
ever been measured accurately..."
The following is a potentially useful publication on how much of a skin-applied treatment is needed
to cover the surface area of an adult:
Sherertz, E. F.
Pharmacology. I. Topical Therapy in Dermatology.
JOURNAL OF THE AMERICAN ACADEMY OF DERMATOLOGY 21: 108-114 .
(1989).
"Approximately 30 grams of topical medication is needed to cover the body surface of an
adult in a thin layer." (If the medication is has a specific gravity of I gram per cm? and
the adult total body surface area is 18,000 cm^, the film thickness can be calculated to
be 0.0017 cm.)
B-213
-------
Procter&Gamble
Hakkinen 4'
The following is a potentially useful publication on the capacity of human skin to hold a liquid
product:
Rutiedge, L. C.
Some Corrections to the Record on Insect Repellents and Attractants.
JOURNAL OF THE AMERICAN MOSQUITO CONTROL ASSOCIATION 4: 414-425
(1988).
Most persons applying a liquid repellent qd_ libitum -will apply it at a rate of about 2
mg/cm?. Although it is possible to apply more than this intentionally, a limit is
eventually imposed by the inception of runoff from the skin.. For most repellents, this
limit is about 4 mg/cm^.
Sections. Other Factors for Exposure Calculations
See Section 6 comments for possible additions to Section 5.
Section 6. Consumer Products
Key Comments:
All three Westat studies forming the basis of this section were based on the recall of the subjects. ,.
As noted on Page 6-2, "Participants were asked to recall product usage data from the previous 12
months. This may degrade the response accuracy of the participants." Did this happen? Not
currently shown in Section 6 is evidence suggesting that this did occur in at least with the Westat
1987b study.
A 50 person subset of the original 193 person phone survey participated in a four-week diary study
of eight of the 14 cleaning tasks originally studied. The key finding of the diary study was that
much less time per day was spent on performing six of the eight tasks when the diary study data
were used. For example, wiping-off counters with a light-duty liquid decreased from a 50th
percentile value of 54.75 hours in the phone recall survey to just 18.45 hours in the diary study (=
33.4% of the recall study value).
While the text of Westat 1987b suggests seasonal differences in product usage as playing the key
role in the above observed differences, the diary data could at least be included for perspective, and
to serve at least as a lower, perhaps more accurate set of data. If similar follow-up diary studies
were performed as part of Westat 1987a and Westat 1987c; I would make the same comments and
suggestions. The Hakkinen and Hakkinen et al., publications noted below contain discussions
B-214
-------
Pnxter&Gambte
Hakkinen 5
about phone recall and diary studies that might be useful to note in the paragraphs discussing the
Westat studies in the Exposure Factors Handbook.
Other Comments:
I will bring copies of the following potentially useful publications to the workshop. To make this
section as user friendly as possible, I suggest considering adding a master table listing all the
various types of consumer products (over 70 if the following publications are included), and
showing which publications contain potentially useful information for a particular product type.
Also, some published tables, such as those in the Cosmetic, Toiletry, and Fragrance Association
document noted below could be added in their entirety:
• Cosmetic, Toiletry, and Fragrance Association, Inc., Summary of the Results of Surveys of the
Amount and Frequency, of Use of Cosmetic Products by Women. Contains usage amount
*
and frequency of use data for lotions, creams, mascara, sunscreen, hair sprays, shampoos,
toothpastes, underarm deodorants, etc. The frequency of use data are from several sources,
and are shown as average and upper 90th percentile values.
• Curry, K. K. et al. (P&G-sponsored). Personal Exposures... During Use of Nail Lacquers..."
JOURNAL OF EXPOSURE ANALYSIS AND ENVIRONMENTAL EPIDEMIOLOGY 4:
443-456 (1994). Contains consumer use data for nail lacquer products, i.e., nail polishes,
basecoats, and topcoats.
• European Centre for Ecotoxicology and Toxicology of Chemicals. Technical Report No. 58.
Assessment of Non-Occupational Exposure to Chemicals (1994). Contains usage amount and
frequency of use data for lotions, shampoos, toothpastes, mouthwashes, etc. (in all, 15
"cosmetic" product types). The frequency of use data are shown as "normal use" and
"extensive use" levels. Table 3 summarizes the data, and could be used "as is" in the
Exposure Factors Handbook as European data for direct use and comparison to U.S. values.
Also, this publication contains task usage amount and task frequency data for various types
of laundry detergents, hand dish washing liquids, automatic dish washing products, and
fabric conditioners (in all, ten "laundry and cleaning"product types). Table 4 summarizes
the data, and could be used "as is" in the Exposure Factors Handbook as European data for
direct use and comparison to U.S. values.
Various other types of products are also discussed at least briefly in the above document.
B-215
-------
Hakkinen
• Hakkinen, P. J. et §1. (P&G). Exposure Assessments of Consumer Products: Human Body
Weights and Total Body Surface Areas to Use, and Sources of Data for Specific Products.
VETERINARY AND HUMAN TOXICOLOGY 33: 61-65 (1991). This review discusses
sources of exposure-related data for specific product types needed for exposure assessments.
The review also contains a discussion of the importance of statistical characterization of the
consumer data, and the importance of examining these data for correlative interactions.
• Hakkinen, P. J. (P&G). Cleaning and Laundry Products, Human Exposure Assessments.
HANDBOOK OF HAZARDOUS MATERIALS 145-151 (1993). Includes some exposure
information for assessing consumer exposures to cleaning and laundry products, along with
discussion of the topics covered in the 1991 P&G publication.
• International Sanitary Supply Association. Cleaning Time Estimator. Contains estimates of
times required to conduct various cleaning tasks, e.g., cleaning a shower stall, sinks, toilet,
stairways^ windows, etc. In all, over 50 estimated task durations are noted. Note that this
information could also be a possible addition to the "Activity Patterns" portion of Section 5
of the Exposure Factors Handbook.
• Vermeire, T. G. et aL Estimation of Consumer Exposure to Chemicals: Application of
Simple Models. THE SCIENCE OF THE TOTAL ENVIRONMENT 136: 155-176 (1993).
Includes some exposure information for assessing consumer exposures to detergents,
deodorants/antiperspirants, spray cleaners, etc.
• Wooley, J. etal. Release of .Ethanolto the Atmosphere During Use of Consumer Cleaning
Products. JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION 40:
1114-1120 (1990). Includes some exposure information for assessing consumer exposures
to liquid hand dish washing and laundry products.
The current section only contains information from the Westat data. A broad recommendation
would be to ask readers to help ensure that all potentially useful consumer product data are
included in future revisions of the Exposure Factors Handbook. A similar message could be sent
to various trade associations. Could someone in EPA be designated in the handbook as a possible
contact to receive data for possible inclusion in future revisions? If done in this way, perhaps
periodic updates listing new data could be sent to known users of the Exposure Factors Handbook,
or even posted as an update file in EPA's Internet World Wide Web "home page."
B-216
-------
Pnxter&Gambte
Hafcfcinen
Section 7. Reference Residence
The introduction to this section could use citation of one or more key publications readers could
consult to get an overview of how residential inhalation exposure assessments are performed, why
specific residential factors are needed, how the residential factors potentially relate to each other,
and the potential relative importance of residential inhalation exposure to other exposures, e.g.,
drinking or ingestion exposures to volatile contaminants in tap water. A very good publication
that covers the above is:
McKone, T. E. Household Exposure Models. TOXICOLOGY LETTERS 49: 321-329 (1989).
"There are 4 types of input data required by the indoor model: (1) house and room volumes, (2)
residence times for air in each household volume, (3) voter use by category, and (4) amount of
time individuals spend in the shower, bathroom,, and remaining house..." This publication's
Table JV also provides information from two other publications on ranges of water use per
person per day for toilets, showers, baths, laundry, dishwater, kitchen and sinks, and cleaning.
Other portions of the text provide information on house and room volumes, and shower stall
volumes.
Another overall perspective and information on various residential parameters, e.g., water
consumption, is available in:
Wilkes, C. R. et aJL Inhalation Exposure Model for Volatile Chemicals from Indoor Uses of
Water. ATMOSPHERIC ENVIRONMENT 26A: 2227-2236 (1992).
Section 7.2 starts by stating that no measurement surveys have been conducted to directly evaluate
the range and distribution of residential volumes. Likewise, Section 7.3.2 states that no
measurement surveys have been conducted to directly evaluate the range and distribution of
residential air exchange rates. A key comment about these statements is that some of the
following publications address these key needs. As discussed below, a great deal of potentially
very useful published and submitted for publication information from various studies could be
added to this section. These studies include: .
• Finley, B. L. et al. Evaluating the Adequacy of Maximum Contaminant Levels as Health-
Protective Cleanup Goals: An Analysis Based on Monte Carlo Techniques.
REGULATORY TOXICOLOGY AND PHARMACOLOGY 18: 438-455 (1993).
B-217
-------
Procter&.GambJe
Hakkinen 8
Portions of Table 2 and Table 3 may be worth using "as is "for the revised Exposure Factors
Handbook, including cited (when all were available or assumed) distribution type, mean,
standard deviation, minimum and maximum values for shower exposure time, shower and
house water use rates, shower, bathroom, and house air exchange rates, shower, bathroom,
and house exposure times, and transfer efficiencies from water to shower air and household
air (shown are transfer efficiencies based on perchloroethylene) .
McKone, T. E. and Bogen, K. T. Uncertainties in Health-Risk Assessment: An Integrated
Case Study based on Tetrachloroethylene in California Groundwater. REGULATORY
TOXICOLOGY AND PHARMACOLOGY 15: 86-103 (1992). Like the Finley et al
publication noted above, this contains a useful table (Table 1) containing distribution type,
arithmetic mean, and standard deviations for shower duration, shower and total house water
use per person, exposure times in the bathroom and house, bathroom and house ventilation
rates, and transfer efficiency from water to shower air and water to household air (shown
are transfer efficiencies estimated for tetrachloroethylene). This publication also notes
assumptions for representative shower, bathroom, and house volumes and air changeovers
for these locations.
Pandian, M. et al. Residential Air Exchange rates for Use in Indoor Air and Exposure .
Modeling Studies. JOURNAL OF EXPOSURE .ANALYSIS AND ENVIRONMENTAL
EPIDEMIOLOGY 3:407-416 (1993). Includes data from numerous studies and generates
frequency distributions and summary statistics for residential air changeovers in different
regions of the United States, different seasons, and different levels within the homes. The
summary statistics (Table 1) and cumulative frequency plots (Figures 1-3) should be
considered for addition "as is" to the Exposure Factors Handbook.
Murray, D. M. and Burmaster, D. E. Residential Air Exchange Rates in the United States:
Empirical and Estimated Parametric Distributions by Season and Climatic Region. Submitted
for publication in RISK ANALYSIS. Includes data from several key studies and generates
frequency distributions and summary statistics for residential air changeovers in different,
regions of the United States, different seasons, and as a Junction of "heating degree days."
In all, 25 frequency distributions are provided. The summary statistics (Tables I, II, and III)
should be considered for addition "as is" to the Exposure Factors Handbook.
Murray, D. M. Residential Total House and Zone Volumes in the United States: Empirical
and Estimated Parametric Distributions! by Season and Climatic Region. Submitted for
B-218
-------
Procter&Gamble
Hakkinen £
publication in RISK ANALYSIS. Includes data from several key studies and generates
frequency distributions and summary statistics for house volumes in the United States as a
whole and for eight states. Similar results are also presented for zone volumes for different
areas of the house. Also noteworthy is that the possible correlation between house volume
and air changeovers per hour was found to be very weak. The summary statistics (Tables I
and II) should be considered for addition "as is" to the Exposure Factors Handbook.
»„
. • Anonymous. What is the average house in Northeastern U.S? A University of Maine 100-
home study cited in SCIENCE NEWS (October 15, 1988, Page 254). This study found the
average home to be "A home with 2,000 square feet of floor space, eight-foot ceilings, 250
gallons of water use per day and a total venting of indoor air about once every 1.2 hours."
Calculating the house volume from some of these data gives a volume of 458,300 liters.
• Brambley, M. R. and Gorfien, M. Radon and Lung Cancer: Incremental Risks Associated
with Residential Weatherization. ENERGY 6: 589-605 (1986). "A number of studies have
involved measurement of air infiltration rates for both conventional and energy-efficient
homes. In a survey by Diamond and Grimsrud of 312 recently constructed homes
throughout the U.S. and Canada, the mean measured infiltration rate during the months of
November through March was 0.63 air changeovers per hour (ACPH). In a re-evaluation
of existing data, Nero estimates average infiltration rates in the U.S. to be 0.7-0.8 ACPH.
Although rare, air-exchange rates as great as 4.0 and lower than 0.25 ACPH have been
measured. Generally, an infiltration rate of 0.5 ACPH is considered by McNall as
representative of recently, well-sealed homes."
The U.S. Department of Energy has in years past apparently received various appliance-type
exposure information from the Association of Home Appliance Manufacturers (20 North Wacker
Drive, Chicago, Illinois 60606). This potentially very useful information included water volume
per appliance load, frequencies of appliance use, etc.
Like the previous section on consumer products, the current section is new. A broad
recommendation would be to ask readers to help ensure that all potentially useful residential
exposure data are included in future revisions of the Exposure Factors Handbook. A similar
message could be sent to various trade associations and key academicians. Could someone in
EPA be designated in the handbook as a possible contact to receive data for possible inclusion in
future revisions? If done in this way, perhaps periodic updates listing new data could be sent to
B-219
-------
Procter&Gambte
Hakkinen 10
known users of the Exposure Factors Handbook, or even posted as an update file in EPA's Internet
World Wide Web "home page."
Section 8. Analysis of Uncertainties
Two related reviews on human exposure assessment uncertainties that could be noted in this
section are:
*
Whitmyre, G. K. et al. Human Exposure Assessment I: Understanding the Uncertainties.
TOXICOLOGY AND INDUSTRIAL HEALTH 8: 297-320 (1992).
Whitmyre, G. K, et al. Human Exposure Assessment II: Quantifying and Reducing the
Uncertainties. TOXICOLOGY AND INDUSTRIAL HEALTH 8: 321-342 (1992).
Other Comments:
Finally, since one of the charges to reviewers was to identify data gaps and research needs for each
section, it should be noted that the following two recent publications in particular have discussed
ways to improve the science of exposure assessment. I don't necessarily recommend noting all of
the following in the revised Exposure Factors Handbook; however, some of the areas for
improvement and research needs are related to exposure factors and the use of exposure factors,
and seem worthy of highlighting in the appropriate section(s).
Whitmyre, G. K. etaJL Human Exposure Assessment II: Quantifying and Reducing the
Uncertainties. TOXICOLOGY AND INDUSTRIAL HEALTH 8: 321-342 (1992).
Potential improvements to human exposure assessment that were discussed in the above
publication included:
(I) Use of more appropriate exposure default values...;
(2) Incorporation of time-activity data...;
(3) The use of reasonable exposure scenarios...;
(4) The use of stochastic approaches...;
(5) Use of bivariate analysis...;
(6) Use of less than lifetime exposure...; and
B-220
-------
Procter&Gambte
Hakkinen 11
(7) Incorporation of physiological considerations relevant to absorbed dose
estimation...
The above publication also discussed other ways to improve the exposure
assessment process, and identified the following key research needs (see Pages
339-340 of original publication for full text):
(1) Exposure Parameters. Collecting statistical distribution data on parameters
for -which data are incomplete or absent.
(2) Methods for Calculating Joint Probabilities. More information on the inter-
relationships of exposure parameters is needed.
(3) Pharmacokinetic Modeling. Pharmacokinetic parameter data, such as blood
flow rates and tissue volumes, need to be developed on key chemicals of interest.
PBPK model uncertainties should be examined in more depth using Monte Carlo
and other stochastic methods. New models need to be validated. Chemical-
specific factors such as partitioning ratios and metabolic rate constants need to
be developed.
(4) Indirect Pathways. More research is needed in this area.
(5) Personal Monitoring and Human Activity Patterns. Total human exposure
monitors that measure personal exposures in a reproducible way need to be
developed for a variety of chemicals. More effort is needed in developing and
improving human activity models and databases. Further understanding of
microenvironments is needed, as is the need forfurtherstudies to define the
relative contributions of various routes, pathways, and microenvironments to
exposures to many types of compounds for various subpopulations and regions.
Paustenbach, D. J. The Practice of Health Risk Assessment in the United States (1975-1995):
How the U.S. and Other Countries Can Benefit from that Experience. HUMAN AND
ECOLOGICAL RISK ASSESSMENT 1: 29-79 (1995).
The above publication by Paustenbach presented several "lessons learned" in the
United States about how to improve exposure assessments. They include:
B-221
-------
Procter&Gamble
Hakkirien 12
(1) Don't put too much emphasis on risk estimates for the maximally exposed
individual (MEI);
(2) Evaluate the uptake (absorbed dose) for both the 50% and 95% persons;
(3) Do not repeatedly use conservative or worst-case assumptions. Incorporate
Monte Carlo techniques whenever possible;
(4) Ensure a proper statistical analysis of environmental data;
(5) Conduct sensitivity analysis to understand fragility of dose estimates;
(6) Understand the role of environmental fate when estimating exposure;
(7) Validate the reasonableness of the exposure estimates;
(8) Consider using biological monitoring to confirm exposure estimates; and
(9) Consider all indirect pathways of exposure.
Please feel free to contact me if you have any questions about my comments. Thank you again for
asking me to participate in the review of this document.
Very truly yours,
The Procter & Gamble Company
P.J.(Bert)Hakkinen,Ph.D.
Senior Scientist - Toxicology and
Risk Assessment
Paper Technology Division and Paper
Products Division
Phone: 513-634-2962
Fax:513-634-3496
PJHefh795.doc
B-222
-------
James Axley
B-223
-------
-------
EPA Exposures Factors Handbook Workshop - 7/95 James Axley <• Yale
Review Comments
Housing Characteristics and Indoor Environments
(Chapter 7. Reference Residence)
submitted by
James W. Axley
Yale School of Architecture
General Comments
The authors of Chapter? have taken on a very difficult challenge and have done a very
admirable job establishing a reasonable framework and first draft for the description of a
Reference Residence fo'r exposure analysis. They have been extremely careful to make
sure the data is presented in a form that is useful to exposure assessors; to discuss studies
that support the analytical approaches presented; and to present recommendations for use of
the data.
The general suggestions put forward below are largely proposals to expand the scope of the
chapter rather than offer corrections — adding, it is hoped, to the excellent material presently
included in this chapter.
• Orgamzaft'on:The general organization of the Chapter—7.1 Introduction; 7.2 Indoor
Volumes; 73 Airflows; 7.4 Water Supply and Use - could be expanded to include
sections on modeling approaches, sources, and analysis and the Indoor Volumes
section could be generalized to "Building Characteristics" to allow inclusion of a more
complete characterization of residential buildings as:
7.1 Introduction
7.2 Modeling Approaches
To include a general discussion of the approaches to modeling making the
distinction between microscopic and macroscopic modeling and establishing
the classes of data needed for analysis.
73 Building Characteristics (formerly Indoor Volumes)
To include building configurations; room volumes, wall, floor and ceiling
surface areas; construction material characteristics, furnishing characteristics
7.4 Airflows
7.5 Water Supply and Use
7.6 Sources
7.6.1 Airborne Sources
7.6.2 Waterborne Sources
7.6.3 Dust and Aerosol Sources (e.g. tracking of soil into homes)
7.6.4 Transport Between Source Types
7.7 Analysis
B-225
-------
EPA Exposures Factors Handbook Workshop • 7/95 _ James Axley - Yale
To include general formulations of single-zone and two- zone formulations
of the contaminant dispersal problem with steady-state solutions and
dynamic solutions for representative cases and an introduction to
computational tools for multizone analysis.
• Suburban Bias: The discussion of building configurations and room volumes, wall,
floor and ceiling surface areas appears to be biased toward suburban residences. There
is a clear need to consider urban residential environments and, possibly, rural
residences as well.
• Geomet Bias: As noted in the specific comments below, much of the discussion is
related to research completed at Geomet. While this Geomet research has been
consistently of the highest quality, it would be best to tie the discussion to a broader
array of studies. Furthermore, at this point in time, the studies have not been explicitly
classified into "key" and "other relevant" as mandated by EPA.
• Data Gaps and Future Research Needs: Some are noted below in the specific
suggestions and many could be enumerated, but at this time it would be best to address
this issue during the workshop.
Specific Comments and Suggestions
Page 7-2
Rgure 7-1 should include chemical and physical transformation (i.e., in addition to indoor
sources, reversible sinks, and decomposition and deposition). Examples include gaseous
chemical reactions such as Os + NO -> NC>2 + O2 and physical transformations such as
condensation or coagulation of aerosols.
The use of an assumed ceiling height of 8 feet to estimate residential volumes and surface
areas may introduce significant error. Historically, it is likely that the trend of ceiling
heights in detached single family homes and many urban attached single-family homes
ranged from below 8' before circa 1850 to above 10' by the great depression then,
influenced strongly by the code minimum of 7' 6" remained close to 8' until recent building
trends have revived higher ceiling heights. Urban neighborhoods not only tend to be
dominated by these older residences, a significant portion of these residences have been
transformed to multi-unit residential configurations. As a result, one might expect the
tabulated volume and, importantly, surface estimates (i.e., from a lead paint exposure point
of view) to significantly underestimate urban residential dwellings.
B-226
-------
EPA. Exposures Factors Handbook Workshop - 7/95 James Axley - Yaie
Page 7-5
It is not immediately clear whether the estimated volumes reported in Table 7-1 are whole-
building volumes or residential unit volumes. The magnitudes indicate, however, that the
volumes are residential unit volumes. This should be clarified.
Page 7-8 Table 7-4
Due to past beliefs regarding the nature of "surface" emissions "surface" materials have
been characterized by their surface area alone. More contemporary, and physically
consistent, views of the nature of "surface" emissions and sorption phenomena now
recognize the importance of treating most of these materials as porous solids and providing
more complete physical characterization of them (e.g., thickness, porosity, mass per unit
volume, specific surface area, etc.). For example, the ubiquitous gypsum board is very
porous, from the perspective of gas molecules, and has an extremely high specific surface
area ( > 500 m2/g) - characteristics far more significant than the surface covered.
Additional research is needed to provide more complete and more relevant characterization
of building materials in general and especially those building materials used in the
construction of room surfaces.
Page 7-10 52
A small technical point: The second sentence of the second paragraph of this page should
be altered to read:
"The forces causing the airflows are due to temperature differences, the actions of wind,
and mechanical ventilation systems."
Page 7-10 J3
This paragraph presents a macroscopic view of air circulation in buildings. From a
microscopic point of view, the circulation in a building with "free communication between
floors or stories" may be (is likely to be) far more complex than that described. It would
not be unreasonable to expect a complex overlay of recirculation loops at each level with
smaller flow loops or eddy-like flow structures here and there throughout a building. A
revision of this paragraph should be considered.
Page 7-11
Regarding the use of PFT air exchange data. It is an established fact that the application of
a constant injection tracer technique based on average tracer concentration measurements,
such as the conventional PFT technique, may be expected to consistently underestimate air
exchange rates. The problem results from the fact that air exchange rates vary with time
while the theory upon which these methods are based presumed constant air exchange
rates. The underestimation error that may result may be expected to be larger with longer
averaging times. Furthermore, the PFT database while large contains, I suspect, many
measurements taken by nonexperts that may, therefore, be suspect
B-227
-------
EPA Exposures Factors Handbook Workshop • 7/95 James Axley - Yale
This problem should be noted and ideally the uncertainty associated with it quantified. It is
significant that the Grot & Clark study and the Grimsrud studies noted on page 7-12 report
means significantly greater than that extracted from the PFT database for, I believe, these
two earlier studies were not based on data collected using the conventional PFT method.
Page 7-14
The background discussion is biased in two respects regarding characteristic residential
configurations is, I believe, biased toward single-family residences found primarily in
suburban areas. This should be noted and, if possible, urban and, possibly, rural (i.e.,
both farm and upscale residential dwellings that tend, to be more complex or exceptional in
configuration) configurations should be discussed.
Page 7-14
The heuristic relationship between internal airflows and house volume and air exchange is
novel and, as such, interesting, but it does not in any way represent a consensus view of
researchers in the field. In fact, I suspect few researchers in the field even know of this
approach. From my perspective it suffers from the following flaws:
• It is based on the tacit assumption that whole-house air exchange rate is distributed
to zones in proportion to volume. This is not likely to be the case. Among other
factors, rooms associated with entries and exits might be expected to experience
proportionately greater air exchange, room exposure must be expected to be
important, and occupant behavior will result in significant day-time, night-time, and
seasonal differences. : .
• As noted interzonal airflows are presumed to be "symmetrical" between two
zones. Air exchange is due to wind, buoyancy, and/or mechanical devices that, by
their nature, must be expected to result in a net transport from zone to zone. Thus
"symmetrical" flows must be expected to be the exception rather than the rule.
• Again the PFT database has been assembled from a variety of sources. Some data
has most certainly been collected by investigators not familiar with the many pitfalls
of multi-zone tracer gas measurements. Multi-zone tracer gas measurement is in
many (most) cases an ill-conditioned problem — i.e., sensitivity to measurement
error can be pathologically extreme - passive sampler-PFT techniques must be
expected to especially susceptible in this regard. (In this regard, multi-zone PFT
airflow data including negative values should be rejected out-of-hand.)
The heuristic method proposed certainly has the advantage of simplicity, but correctness
must be held as more important and this method should not be put forward as standard
practice.
Page 7-18
The first paragraph of section 7.3.4 Variability Within Zones refers to the very
interesting and rigorous work of Baughman et al., but does not properly establish the
context of the research discussed. From Baughman et al.'s Abstract:
B-228
-------
EPA Exposures Factors Handbook Workshop - 7/95 James Axley - Yale
"... This experimental study characterizes quantitatively the rate at which smoke
from a cigarette disperses within an unoccupied, 31 m3, low air-exchange rate room
[0.03-0.08 AGH] under natural convection flow conditions. Sidestream smoke ...
was simulated with ... SFe .. .[at] 41 locations within the room ... Duplicate runs
were conducted under three conditions: nearly isothermal surfaces, convection from
a 500 watt heater; and convection from incoming solar radiation. Characteristic
mixing times ranged from 7-10 minutes for the solar radiation case to 80-100
minutes for the nearly isothermal case." (Baughman, A.V., A.J. Gadgil, and W.W.
Nazaroff, Mixing of a Point Source Pollutant by Natural Convection Flow within a
Room. Indoor Air, 1994.1994(4): pp. 114-122.)
Importantly, these studies were conducted at extremely low air exchange rates - not at all
characteristic of airflows found in residences.
Furthermore, the statement in section 7.3.4 that "Similar finding might be expected for a
continuously emitting area source ..." is technically off the mark. Even molecular diffusion
from a point source will differ substantially from that from a planar source.
More to the point, are the consistent findings that microenvironmental monitors consistently
underestimate dose/exposure when compared to personal environmental monitors. Charles
Rodes provides a very useful review of these findings and establishes a reasoned position
relative to their importance to exposure modeling:
" In a less-than-ideal mixed situation, contaminant concentration gradients may be
large in close proximity to the source, even though the general area concentration at
some distance away may change insignificantly— Thus, the application of
integrated exposure models, using activity pattern information and compartmental
average concentration data, may give results that are unacceptably inaccurate and
produce [exposure] estimates that are often biased low." (Rodes, C.E., R.M.
Kamens, and R.W. Wiener, The Significance and Characteristics of the Personal
Activity Cloud on Exposure Assessment Measurements for Indoor Contaminants.
Indoor Air: International Journal of Indoor Air Quality and Climate, 1991. Vol. 2:
p. pp. 123-145.)
In this paper, Rodes discusses the importance of the so-called "proximity effect" and his
investigations of the "personal cloud" that appears to be central to this effect.
The second and third paragraphs of this section address this issue more appropriately, but
the research reported is limited to two studies. This section should be revised using a
broader collection of studies. The work reviewed by Rodes should go a long way toward
achieving this obj ective.
B-229
-------
-------
P. Barry Ryan
B-231
-------
-------
Comments of P. Barry Ryan- Exposure Factors Hand book
Comments on Chapter 7- Reference Residence
The Reference Residence is an important concept in modeling of exposures experienced in
indoor environments. As a large fraction of total time is spent in such locations, a proper
understanding of the environment is warranted. This Chapter presents field data and suggestion for
parameters to use in modeling of residential exposure.
The chapter is, by and large, a useful one. The compendium of data is unique and the
reference list at the end very valuable. I have several specific comments given below.
Introduction- The Introduction does not touch upon any soil gas contamination processes,
e.g., radon which show variable impact depending upon the characteristics of the residence- basement
condition, tightness, etc.
Tables 7-1 - 7-3 represent extremely useful data for indoor air quality modelers. They
(particularly 7-1 and 7-2) would be even more useful if the variability were described either by
presenting percentiles of distributions or even standard deviations.
The arguments on page 7-8 relating size of test homes to national average is very strained.
The test homes should be viewed as such without a lot of effort designed to that they are, somehow,
represented of trends in all homes.
Again on page 7-8 the discussion of surface-to-volume ratio (SAO is important. This
parameter is crucial in modeling deposition of particles and gases. The discussion is couched in
extremely confusing language related to S/V ratios for floors and walls. I was confused into trying to
figure out what the volume of a wall or floor might be. Specifically state the relationship- the ratio of
wall surface area to total volume and similarly for total wall surface area to total volume. Then it
becomes clear.
On page 7-10, there is an unqualified assertion the I/O temperature differences are smaller in
cooling seasons than in heating seasons. This is not categorically true; in Arizona, Nevada, and
doubtless many other locations, such differences may indeed be greater in cooling seasons. There is
no need to be so assertive.
hi Table 7-5, perhaps as a footnote, the nature of the distributions should be described. Are
the distributions essentially normal, lognormal, etc.? Skewness, boundedness etc. are all useful
parameters for modelers.
On page 7-12 the text states theat"... Statistical techniques were applied to compensate for
some of these imbalances (is seasonal and geographical coverage) ..." What techniques were applied
B-233
-------
Comments of P. Barry Ryan- Exposure Factors Hand book
and how? This should be discussed.
Page 7-12ff discusses air exchange rates. The listings for air exchange rates, a critical
parameter in any modeling exercise, report this parameter as mean ± geometric standard deviation
ACH, This is odd. First, it is not clear whether the deviation is geometric or arithmetic. If it is
arithmetic, the units are fine but there is an apples and oranges mix of parameters. If it is geometric,
the variability is multiplicative and, thus, has no units. Given the expected skewness in theses data
(NB max for west is 23.82) the reported values could be either. Also, given what I know about the
technique, I would be highly suspect of any measurement above about 8.0 ACH. For air exchange
rates higher than this, the measured tracer level is probably below detection limits.
On page 7-17, equations are given relating interzonal flows, Qz with air exchange rate, hi
that this appears to be a statistical regression model, it would be very useful to know the qulaity of
the fit. For example, what is the expected error in this fit, the R2, etc. Further, a discussion of
conditions resulting in failure of this model would be beneficial. Also, again with knowledge of the
technique used to gather the basic data, caution should be exercised in using this regression model.
An improper assumption regarding the location of zones can result in non-physical interzonal flows.
These may or may not be included in the data set used to derive these relationships.
Section 7.3.4 reads differently than the other sections. It is more a description of a research
experiment without really putting it into perspective for the exposure factors handbook.
Table 7-6 makes use of several data sets. The idea of using the mean or median of these
investigations is flawed as it gives each study identical weight. As I am not familiar with each
investigation, I cannot estimate the effect. I can speculate that some studies may have been geared
toward specific populations which may use one or more of these categories to a greater or lesser
extent than others. Although these may be the best data available, extreme caution should be urged
in their use.
In the overview documents we received, it was stated that a set of recommendations would
be presented at the end of each section. No such recommendations were found at the end of Section
7.
Comments on Chapter 8- Analysis of Uncertainties
Chapter 8 is a brief compilation of terms and ideas from the references (in particular,
USEPA, 1992). This in and of if self is quite useful and will supply the potential user with some
B-234
-------
Comments of P. Barry Ryan- Exposure Factors Hand book
knowledge of the nomenclature associated with the study of uncertainty. What is not supplied,
however, is a mechanism for implementing this information. Much of the section is devoted to
expanding on the three brief definitions given at the beginning regarding uncertainty in scenarios,
parameters, and models. Some approaches to investigating such uncertainties are suggested but
without sufficient detail to afford the unschooled practitioner to make any headway, hi this, there is a
severe failing. The final subsection discusses methods of presenting the data. Such issues are like
apple pie and motherhood; who can object to a goal of clear presentation? On the other hand, there is
little information on how to present. Detailed examples are found below.
In summary, I found this chapter to be sorely lacking. It is an excellent introduction to a
chapter on uncertainty estimates. If this handbook is to serve in the manner needed, this section
needs to be greatly expanded. I would urge the presentation of a series of worked examples ranging
from quite simple exposure assessments and their related uncertainties to more complex systems.
Throughout the discussion, analogy should be made to each of the paradigms discussed- uncertainty
in scenarios, parameters, and models- to tie in with the preliminary discussion. These examples
should be well-chosen in that they will be used as guidelines by users of this Handbook. If this is not
done, I suggest shortening the chapter and prompting the reader to assess the literature independently
by providing a more complete bibliography.
Specific Comments
The definitions of uncertainty characterization and uncertainty assessment are, I believe, not
at variance with commonly accepted nomenclature, but are not in agreement with it either. These are
new terms.
At the top of page 8-3, there is a discussion of "incomplete analysis" with an example given
focusing on overlooking an important pathway in an exposure. Why isn't this model
misspecification?
Parameter uncertainty is most easily understood by Monte Carlo simulation, especially for
random errors and sampling errors. Systematic errors are harder to model if they are unknown. Such
procedures as bounding estimates, expert judgement, and the like come with essentially unknown
(and unknowable) errors. This is not discussed clearly in the section.
Sensitivity analyses are useful but repeated simulation followed by analysis of variance also
supplies a useful, and often powerful, technique in lieu of simple "high-low" simulation. The
hierarchy of sensitivity analysis, analytical uncertainty propagation, Probabilistic uncertainty
B-235
-------
Comments of P. Barry Ryan- Exposure Factors Hand book
analysis, and classical statistical methods is not a lock-step with regard to complexity and data needs,
especially in a simulation model. Data needs are minimal as more simulations can be run to generate
more data and, thereby establish the importance of parameters.
At the bottom of page 8-5, a suggestion is made that is "average" values are needed, they
can be computed accurately by using average values for each parameters. What type of average are
we looking for here? Suppose distributions are highly skewed (as is likely in exposure assessment)?
Suppose parameters are correlated? Further possibilities exist which can make this assumption total
invalid. This deterministic approach to analysis is poor. Further, given the availability of commercial
software, the statements regarding the difficulty of repeating the simulations is not warranted. Also,
the last two sentences stating ".. Monte Carlo analysis assumes that the distributions of each
variable are independent." is just wrong. There is nothing in Monte Carl analysis that makes this
assumptions. Most new software is quite capable of including either Pearson or Spearman
correlation among distributions using either standard techniques for linear algebra on normal
distributions or the techniques of Iman and Conover on other distributions.
The discussion of skewness and data sparseness at the end of section 8.1.2 is incomplete.
There needs to be an assessment of the effect of < LOD values on the distributions estimated as well
as effects on other parts of the distribution if stratification is effected to increase the "tail" proportion
of samples. You can't get something for nothing.
The discussion of Model Uncertainty (Section 8.1.3) borders on philosophical. It discusses
the way things should be done without a lot of practicality. All of these things are desirable, but how
does one approach a real problem?
Section 8.2 on the Presentation of Uncertainty Analysis Results is the first Presentation of
this type of material I have seen. It is a good idea but needs expansion.
In summary, Chapter 8 is an embryonic development of methods of uncertainty analysis. I
believe it to be an essential component of the Exposure Factors Handbook but one that needs
extensive expansion and thought.
B-236
-------
Andrew Persily
B-237
-------
-------
Andrew Persily
July 7,1995
Table 7-3
These four houses have certainly been the subject of much interesting analysis, but there are others
out there was well. They are not necessarily representative of anything, and I wonder about the
implication that they are "the ones" to use. Why include these four houses and no others? At least
include some references to other house layouts. I wouldn't be surprised if the National
Association of Home Buildings1 (NAHB) could help you oti this?
Table 7-4.
Similar to comment on Table 7-3,1 am concerned about the presentation of this information as
uniquely representative of what's out there in the residential sector.
Top of page 7-1
My concerns regarding the measurement uncertainty associated with the interzone flows
determined with the PFT technique are at least an order of magnitude greater than the single zone
air change rates. The PFT interzone data is rarely presented with any uncertainty estimates, and
non-physical results (negative airflow rates ) are common. The so-called "heuristic relationship"
developed by Koontz and Rector may be a very good analysis of a questionable dataset, but its
presentation does not reflect any questioning of its appropriateness or reliability. It is not a
generality accepted approach.
, 1st paragraph of section 7.3.2
I am not' Comfortable with the reference to outdoor contaminant concentrations being zero. This is
a very inappropriate assumption in many situations. Outdoor concentrations are in fact higher than
indoors quite often.
Section 7.3.4
While mixing within spaces and the variation in contaminant concentrations are clearly critical to
exposure, I am not sure I see how the two studies cited here will help the user. They are both
interesting, quality work, but what does one do with them? How does the analysis account for
imperfect mixing? I am not advocating the use of mixing factors; in fact, it might be worth
including a discussion of mixing factors since the reader is probably familiar with them and would
v
benefit from a discussion of the fundamental problems with the concept So what constructive
information can you provide on imperfect mixing? Not much. You can tell them about
computational fluid dynamics, but I'm not sure what else will help.
B-239
-------
Andrew Persily
July?, 1995
2nd paragraph on page 7-12
In addition to the caveats on the PFT data based on its representativeness, there are also important
questions regarding measurement bias with this technique. See the article by Max Sherman in
Building Environment (Vol. 21,135-144,1986) in which is discusses the negative bias in using
this technique to conduct long-term measurements.
B-240
-------
'"
UNITED STATES DEPARTMENT OF COMMERCE
National Institute of Standards and Technology
Gaithersburg, Maryland 20899-0001
Building 226, Room A 313
20 My 1995
Helen Murray
Eastern Research Group, Inc.
110 Hartwell Avenue
Lexington, MA 02173-3198
Dear Ms. Murray,
After a little more thought, I have two additional comments on the Exposure Factors Handbook.
They are as follows:
Section 7.3
Why not discuss the use of models to predict whole building air change rates and interzone
airflow rates? A widely-accepted single-zone model to predict whole building rates (sometimes
referred to as the LBL model) is presented in the ASHRAE Fundamentals Handbook. Several
multi-zone mass balance models also exist, which can be used to predict airflow rates in multi-
zone building airflow systems. Examples of such models include CONTAM, BREEZE and
COMIS.
Section 7.3.2
There are more recent surveys of the airtightness of U.S. homes. See for example the paper by
Sherman and Dickeroff in the 1994 proceedings of the 15th conference of the Air Infiltration
and Ventilation Centre (AIVC), or contact Max Sherman (through Joan Daisey) directly for
even more recent information than in that paper.
If you have any questions, please contact me at (301) 975-6418 or at apersily@nist.gov.
Sincerely,
Andrew Persily
Group Leader, Indoor Air Quality and Ventilation
Building Environment Division
Building and Fire Research Laboratory
B-241
IM1SF
-------
-------
Thomas Phillips
B-243
-------
-------
•July 14, 1995 -~
Helen Murray
Eastern Research Group
110 Hartwell Aye.
Lexington, MA 02173-3.198
Dear Ms. Murray:
Subject: COMMENTS ON EXPOSURE FACTORS HANDBOOK, JUNE 1995 EXTERNAL REVIEW
DRAFT
Brief comments from my initial review of the Handbook are summarized below.
The focus 1s 4n issues listed and the comments are on the Reference
Residence as Well as other subject areas, as requested by Dr. Wood in his
June 29 letter!to reviewers.
1. USEFUL PRESENTATION OF DATA
la. In general, this draft has improved the usefulness of data by
presenting more frequency distributions. However, some of the sections do
not take full advantage of such data. For example, the distributions for
housing volume (Ch. 7.2), air exchange rate (th. 7.3), and time spent in
locations (Ch, 5.3) are not presented. Is it presumed that exposure
assessors will :obtain the original data sets and reports and be able to
analyze them in a short amount of time? Or ahe these distributions
presented in another EPA report?
'It would seem most useful and convenient to have the distributions all 1n
one package, at least for the critical parameters where extensive data are
available, as 1n the examples given above. It would also seem consistent
with EPA's effort to promote consistency in calculating exposure and dose
because it minimizes major sources of inconsistency such as the use of
different percentiles or the incorrect analyses of data.
B-245
-------
Ib. .For AER data (Ch. 7.3). 1t would be useful to emphasize data from
I 1 I """""**~^—mm
pojpqla.|1on-base'd 'samples and from studies with QA/QC programs In place. The
amalgamation of *T1 samples 1n the U.S. treats all data as equal In quality,
which 1s reallv nbt the case. The panel in the previous review round agreed
on this concept, as I recall.
Ic. For house.voilume data (Ch. 7 ,.21. 1t would be useful to Include data for
s|_ab_T on-grade .honies. This type of home 1s the predominant type in
California and "some neighboring states.
Id. ASa general! concept for indoor pollution. It would be worth mentioning
arid adding to figure 7-1 that 1) pollutants fjrom groundwater and soil gas
can enter the.Honfe though the slab or crawl sibace foundation, especially if
thert 1s build^ngideoressuriration, and 21 chemical reactions can not only
remoVe indoor^pollutants but they may add or create pollutants that could be
equally or moyq harmful than the original pollutants emitted. See the work
by Weschler ancj others for examples of carpet pollutants changing over time.
le. If building materials data are presented, include examples to show how
they would beiused. The data from Tucker (Table 7-9) show what materials
might be of Interest, but it is not clear how the data would be used,
especially if each building is a very specific case. In addition, the
example for gyrfsum board seems to be low by a factor of 2-3 for typical new
construction iij California. What is the source of these data?
If. Include estimates of the effect of different mechanical ventilation
systems on AEft^ in homes (Ch. 7.3). Breakdowns of this type would be useful
1s narrowing the data distribution somewhat and deciding which way to lean
1n estimating 4^R. For example, homes with forced air systems generally
have tighter shells (no ductwork and wall penetrations to Teak), homes with
B-246
-------
whole hjpuse o? exhaust ventilation, or heat-recovery ventilation will terad
to Mve; higher AEK's when the system operates!, and forced ailr systems or
large exhaust;ivstems can increase pollutant transport via building
deprfssurizatiqn.
Ig. Time spent ih locations or microenvironments should^bejemphasized mere,
and the time srierct in activities should be areatlv de-emphasized. Numerous
tables and ext4nsHve discussions are devoted to time spent iin various
activities (Chj 5c.3), but it is not clear how such sociological observations
relate to exposure assesors' data needs. It jis clear that time spent in
various;locations.; has direct input into exposure modeling, ais discussed in
NAS reports by Sexton and Ryan and by others.
2. G^UPING JCjF KEY VS. RELEVANT STUDIES
2a, Ithis hot .clfear whv a re-analvses of CARB's adult activiltv pattern
study, foit not the original study, is included. By the way,, the original
study includes * comparison to national data for time spent
-------
4. OAtA GAPSiANb IfUTURE RESEARCH NEEDS
4a. IDfaviouseiyiwte need better data on pollutant source strenqths and AERs.
I wouldj, add, to th^t-list the'removal, transport, and transformation
proce'ssi^s, e'spe'c'lially for particles, metals, and semi-VOC wHfch adhere to
and buijd up 1H building surfaces such as carpets. Most importantly, I
would a'tid personal! monitoring data and/or biological monitoring data as a
means! off sejfctirjg realistic benchmarks for an^ exposure assessment or
Yours truly,
Thomas J, Phillips
B-248
-------
APPENDIX C
WORKSHOP AGENDA
C-l
-------
-------
United States
Environmental Protection Agency
Peer Review Workshop on Revisions
to the Exposure Factors Handbook
Doubletree Hotel Park Terrace
Washington, DC
July 25-26, 1995
Final Agenda
Workshop Chair:
P. Barry Ryan
Rollins School of Public Health at Emory University
Atlanta, GA
TUESDAY, JULY 25, 1995
7:30AM
8:30AM
8:45AM
9:1 SAM
Registration/Check-in
Welcome
William Wood, Director, Risk Assessment Forum, U.S. Environmental Protection Agency (U.S. EPA)
Plenaiy Session: EPA Charge to the Peer Reviewers
Michael Callahan, National Center for Environmental Assessment, U.S. EPA
Summary of Premeeting Comments: Workshop Chair and Work Group Leaders
Workshop Chair:
Work Group 1:
Work Group 2:
Work Group 3:
Work Group 4:
P. Barry Ryan
Barbara Petersen
John Kissel
Steven Colome
Bert Hakkinen
10:30AM
11:OOAM
11:20AM
BREAK
Observer Questions and Brief Comments
Work Group Sessions
• Work Group 1:
12:OOPM
• Work Group 2:
• Work Group 3:
• Work Group 4:
LUNCH
Food and beverage consumption
Nondietary and dermal exposure factors
Human activity patterns
Housing characteristics and indoor environments
i Printed on Recycled Paper
C-3
(over)
-------
TUESDAY, JULY 25, 1995 (continued)
T:15PM Work Group Sessions - continued
4:OOPMT Plenary Session: Status Report by Work Group Leaders
5-.OOPM ADJOURN
WEDNESDAY, JULY 26, 1995
8:30AM Work Group Sessions - continued
12:OOPM LUNCH
1:15PM Plenary Session: Summary and General Discussion
2;45PM BREAK
3:OOPMr Observer Comments
4-30PM ADJOURN
C-4
-------
APPENDIX D
WORK GROUP ASSIGNMENTS
D-l
-------
-------
vvEPA
United States
Environmental Protection Agency
Peer Review Workshop on Revisions to
the Exposure Factors Handbook
Doubletree Hotel Park Terrace
Washington, DC
July 25-26, 1995
Work Group Assignments
Work Group #1
TERRACE BALLROOM
1st Floor
Work Group #2
CABINET ROOM
2nd Floor
Work Group #3
CHAIRMAN'S SUITE
2nd Floor
Work Group #4
DIRECTOR'S SUITE
2nd Floor
Food and beverage consumption
Work Group Leader: Barbara Petersen
EPA Resource Person: Jackie Moya
J. Mark Fly
Patricia Guenther
Mary Hama
Paul Price
John Risher
Frances Vecchio
Nondietary and dermal exposure factors
Work Group Leader: John Kissel
EPA Resource Person: John Schaum
Dennis Druck
Larry Gephait
Human activity patterns
Peter Robinson
Brad Shurdut
Work Group Leader: Steven Colome
EPA Resource Person: Karen Hammerstrom
Edward Avol
Neil Klepeis
John Robinson
Housing characteristics and indoor environments
Work Group Leader: Bert Hakkinen
EPA Resource Person: Kevin Garrahan
James Axley
Andrew Persily
Thomas Phillips
P. Barry Ryan
John Talbott
i Printed on Recycled Paper
D-3
-------
-------
APPENDIX E
FINAL OBSERVER LIST
E-l
-------
-------
&EPA
United States
Environmental Protection Agency
Peer Review Workshop on Revisions to
the Exposure Factors Handbook
Doubletree Hotel Park Terrace
Washington, DC
July 25-26, 1995
Final Observer List
Ronke Adenuga
Chemical Engineer
Exposure Assessment Division
Versar, Inc.
6850 Versar Center
Springfield, VA 22151
703-750-3000
Fax: 703-642-6954
Susan Artz
Analytical Contracts
BASF Corporation
Agricultural Research Center
P.O. Box 13528
Research Triangle Park, NC
27709-3528
919-248-6594
Fax: 919-248-6651
Leila Barraj
Executive Scientist
TAS, Inc.
The Flour Mill
1000 Potomac Street, NW
Washington, DC 20007
202-337-2625
Fax: 202-337-1744
Michael Callahan
National Center for
Environmental Assessment
Office of Research
and Development
U.S. Environmental
Protection Agency
401 M Street, SW (8603)
Washington, DC 20460
202-260-8909
Fax: 202-260-1722
Nancy Doerrer
Deputy Director
American Industrial
Health Council
Suite 760
2001 Pennsylvania Avenue, NW
Washington, DC 20006
202-833-2184
Fax: 202-833-2201
Cathy Fehrenbacher
Senior Industrial Hygienist
Office of Pollution
Prevention and Toxics
U.S. Environmental
Protection Agency
401 M Street, SW (7406)
Washington, DC 20460
202-260-0969
Fax: 202-260-0981
Kevin Garrahan
National Center for
Environmental Assessment
Office of Research
and Development
U.S. Environmental
Protection Agency
401 M Street, SW (8603)
Washington, DC 20460
202-260-2588
Fax: 202-260-1722
Mark Gibson
Staff Scientist
Karen & Associates
1707 K Street
Washington, DC 20036
202-463-0400
Fax: 202-463-0502
Laurie Gneiding
Project Manager/Risk Analyst
Environmental Liability
Management, Inc.
218 Wall Street
Research Park
Princeton, NJ 08540-1512
609-683-4848
Fax: 609-683-0129
Printed on Recycled Paper
E-3
-------
Annette Guiseppt-Elie
Manager, Environmental Health
Risk Assessment
Mobil Oil Corporation
P.O. Box 1029
Princeton, NJ 08543-1029
609-737-5636
Fax: 609-737-5737
Karen Hammerstrom
National Center for
Environmental Assessment
Office of Research
and Development
U.S. Environmental
Protection Agency
401 M Street, SW (8603)
Washington, DC 20460
202-260-8919
Fax: 202-260-1722
Karen Hentz
Senior Staff Scientist
Karch and Associates
1707 K Street
Washington, DC 20036
202-463-0400
Fax: 202-463-0502
Luis Hernandez
Senior Research Associate
Barrera Associates, Inc.
Suite 1120
733 15th Street, NW
Washington, DC 20005
202-638-6631
Fax: 202-638-4063
Patrick Kennedy
Supervisory Chemist
U.S. Environmental
Protection Agency
401 M Street, SW (7406)
Washington, DC 20460
202-260-3916
Fax: 202-260-0981
James Konz
Environmental Health Scientist
U.S. Environmental
Protection Agency (5204G)
401 M Street, SW
Washington, DC 20460
703-603-8841
Fax: 703-603-9103
Carolyn Leep
Associate Director, Risk Issues
Chemical Manufacturing
Association
2501 M Street, NW
Washington, DC 20037
202-887-1323
Fax: 202-778-4042
Ross MacDonald
Staff lexicologist
Shell Development Company
P.O. Box 1380
Houston, TX 77251-1380
713-544-6701
Fax: 713-544-8727
Robert McGaughy
Senior Scientist
Office of Health and
Environmental Assessment
U.S. Environmental
Protection Agency
401 M Street, SW (RD-689)
Washington, DC 20460
202-260-5889
Fax: 202-260-3803,
Jackie Moya
National Center for
Environmental Assessment
Office of Research
and Development
U.S. Environmental
Protection Agency
401 M Street, SW (8603)
Washington, DC 20460
202-260-2385
Fax: 202-260-1722
Rashmi Nair
Manager, Risk Assessment
Monsanto Company
A3ND
800 North Lindbergh Boulevard
St. Louis, MO 63167
314-694-8817
Fax: 314-694-8808
Stephen Olin
Deputy Director
International Life
Sciences Institute
Risk Science Institute
1126 Sixteenth Street, NW
Washington, DC 20036
202-659-3306
Fax: 202-659-3617
Pat Phibbs
Reporter, Environmental
Health Letter
Business Publishers, Inc.
951 Pershing Drive
Silver Spring, MD 20910-4464
301-587-6793
Fax: 301-587-1081
E-mail: 721lO,1536@compuserve.com
Linda Phillips
Environmental Scientist
Exposure Assessment Division
Versar, Inc.
6850 Versar Center
Springfield, VA 22151
703-750-3000
Fax: 703-642-5954
Paul Pinsky
Statistician
Office of Research
and Development
U.S. Environmental
Protection Agency
401 M Street, SW
Washington, DC 20460
202-260-1079
Resha Putzrath
Principal
Georgetown Risk Group
3223 N Street, NW
Washington, DC 20007
202-342-2110
Fax:202-337-8103
E-mail: rmputzrath@delphi.com
Susan Rieth
Manager
ENVIRON
4350 North Fairfax Drive
Arlington, VA 22203
703-516-2300
Fax: 703-516-2345
Sara Thurin Rollin
Reporter
Chemical Regulation Reporter
The Bureau of National
Affairs, Inc.
1231 25th Street, NW
Washington, DC 20037
202-452-4584
Fax: 202-452-4150
E-4
-------
John Schaum
National Center for
Environmental Assessment
Office of Research
and Development
U.S. Environmental
Protection Agency
401 M Street, SW (8603)
Washington, DC 20460
202-260-5988
Fax: 202-260-1722
Greg Schweer
Division Director
Exposure Assessment Division
Versar, Inc.
6850 Versar Center
Springfield, VA 22151
703-750-3000
Fax: 703-642-6954
Ken Sexton
University of Minnesota
Box 807
Minneapolis, MN
612-626-4200
Fax: 612-626-0650
Sanjay Thirunagari
Environmental Engineer Senior
Office of Waste
Resource Management
Waste Division
Virginia Department of
Environmental Quality
4900 Cox Road
Glen Allen, VA 23060
804-762-4193
Fax: 804-527-5233
Alberto Tohme
Senior Toxicologist
Safety & Environmental Resources
DuPont Environmental
Remediation Services
Suite 140
140 Cypress Station Drive
Houston, TX 77090
713-586-5812
Fax: 713-586-5650
Linda Triemer
Senior Staff Toxicologist
Environmental Sciences Division
Exxon Biomedical Sciences, Inc.
Mettlers Road (CN-2350)
East Millstone, NJ 08875-2350
908-873-6289
Fax: 908-873-6009
Eric Trinkle
Hydrologist
Delaware Department of
Natural Resources and
Environmental Control
P.O. Box 1401
89 Kings Highway
Dover, DE 19903
302-739-3689
Fax: 302-739-5060
Amy Wilkins
Environmental Scientist
National Center for
Environmental Assessment
U.S. Environmental
Protection Agency
401 M Street, SW (8603)
Washington, DC 20460
202-260-8909
Fax: 202-260-1722
Maggie Wilson
Environmental Scientist
Exposure Assessment Division
Versar, Inc.
6850 Versar Center
Springfield, VA 22151
703-750-3000
Fax:703-642-6954
Patricia Wood
Senior Project Manager
Exposure Assessment Division
Versar, Inc.
6850 Versar Center
Springfield, VA 22151
703-750-3000
Fax: 703-642-6954 .
William Wood
Director, Risk Assessment Forum
U.S. Environmental
Protection Agency
401 M Street, SW (8101)
Washington, DC 20460
202-260-1095
Fax: 202-260-3955
Li Yang
Senior Consultant
ARCO
515 South Flower Street
Los Angeles, CA 90071
213-486-0922
Fax: 213-486-2021
E-5
* U.S. GOVERNMENT PRINTING OFFICE: 1996 - 750-001/41040
------- |