BACTERIA COLONIZING POINT-OF-ENTRY

             GRANULAR ACTIVATED  CARBON  FILTERS  AND

               THEIR RELATIONSHIP TO HUMAN HEALTH
                               BY
                      Rebecca  L.  Calderon*
          Department of Epidemiology and Public Health
                    Yale  School of Medicine
                  New Haven, Connecticut 06510
                         CR-811904-01-0
                         Project  Officer

                        Alfred P. Dufour
      Environmental Monitoring and Surveillance Laboratory
         United States  Environmental  Protection Agency
                     Cincinnati,  Ohio   45268
* Current address:  Office of Health Research (RD-683)
                    USEPA
                    401 M. St, S.W.
                    Washington, D.C.   20401

-------
                            ABSTRACT

     An epidemiological study on the health effects associated
with high densities of heterotrophic bacteria in drinking water
was conducted by the Department of Epidemiology and Public
Health, Yale School of Medicine.  The study consisted of Yale
families that lived in houses that received their water from the
South Central Connecticut Regional Water Authority.  The
households were divided into two group.  One group had granular
activated charcoal filters installed where the water line
entered the home.  Hence, all water in the home was filtered
through a GAC filter.  The second control group had no filter.
Water samples and health diaries were collected during monthly
visits to both control and filter homes.  All water samples were
analyzed for heterotrophic bacteria using R2A media.   Within a
month the filters had colonized with bacteria from the
distribution system and were discharging significantly higher
numbers of bacteria (1000 cfu/ml) into the household plumbing
than the control houses.   Information on upper respiratory
infections, lower respiratory infections and gastrointestinal
infections were collected.  Health data collected showed no
difference in acute symptomatology between the filter households
and the control households.  Bacterial diagnoses made by health
care professionals were not related to bacteria in drinking
water.

     This report was submitted in fulfillment of CR-        by
Yale University under the sponsorship of the U.S. Environmental
Protection Agency.  This report covers a period from September
1987 to October 1989,"and work was completed as of September
1990.

-------
                            CONTENTS
                                                           Page

Abstract	•	ii
Figures	iv
Tables   	   v
Acknowledgement  	  vi

Section 1.  Introduction 	   1
Section 2.  Conclusions  	
Section 3.  Materials and Methods  	
Section 4.  Results and Discussion 	
Section 5.  Bibliography 	
Section 6.  Figures and Tables 	
Appendices 	
     A.  Science Advisory Committee  . . .
     B.  Filter Specifications 	
     C.  Questionnaires and Health Diaries
     D.  Aerosol measurements  	

-------
                            FIGURES
Number                                                     Page

1.   Schematic diagram of point-of-entry granular activated
     filter (courtesy of Lu Cole)  	

2.   Heterotrophic bacterial densities in cold water tap
     for filters (non-working and working) and controls. .

3.   Heterotrophic bacterial densities in hot water taps
     for filters (non-working and working) and controls. .

4.   Heterotrophic bacterial densities in filter effluent
     and influent for non-working and working filters.  . .

5.   Heterotrophic bacterial densities and water temperatures
     in working filter effluent	

6.   Heterotrophic bacterial densities and free chlorine
     residual in filter households	
     Age and Gender of POE subjects.

-------
                             TABLES

Number                                                     Page

 1.  Six month geometric means of heterotrophic bacterial
     densities per ml	
 2.  Total number of water samples collected for each
     sample type	
 3.   Geometric mean and standard deviations of hetero-
     trophic bacterial from installation water samples .

 4.   Geometric bacterial means and standard deviations
     for water samples by group and sample type ....

 5.   Number of households, average number of people per
     household and average length of time (months)  in
     study	

 6.   Categories of symptoms and reported symptoms for
     each category	

 7.   Study group symptom illness rate (per person year)
     by illness category	
 8.  Highly credible symptom rate (per person year)  for
     each study group by category of symptoms	
 9.  Percentage of illness categories with fever by
     study group	
10.  Percentage of highly credible illness categories
     with fever by study group	

-------
                        ACKNOWLEDGEMENTS

The project benefited from the help of several organizations and
individuals.  We wish to acknowledge the following:

From U.S.E.P.A. for their expert advice:
     Frank Bell, Office of Drinking Water
     Alfred Dufour, EMSL and project officer

From the Water Quality Association for supplemental fundings and
     expertise on the design of the filter for this study:
     Lucius Cole, Technical Director

From the South Central Connecticut Regional Water Authority for
     the chemical and physical analysis of water samples and
     providing information on the water distribution system:
     Alan Hess, Director
     Daryl Smith,
     Noel Grant, Laboratory Director

From the students and staff of the research team for their
tireless efforts in seeing that this study was completed:
     Elaine Okawa
     Chris Ellis
     Sean Buckley
     Mark Clark
     Meg Harvey
     Suzanne Lea
     Mary Nelligan

-------
                          SECTION  1

                        INTRODUCTION

    The recent environmental awareness of the population has
lead to an increase in the use of home water treatment
devices.  These devices can be used to treat a specific
problem such as water hardening or removal of iron or be
used to improve the aesthetics of the water such as taste,
odor and color.  The most commonly used devices contain
granular activated carbon (GAC).   Filters with GAG remove
common tastes and odors, some turbidity, chlorine, and many
organic contaminants.  The large surface area of activated
carbon provides an excellent adsorbent for a wide range of
chemicals and provides and excellent surface for biomass
attachment and growth of bacteria.  Of special concern is
the health hazard that may be created if pathogenic bacteria
adhere to and grow on carbon surfaces.  While the microbial
quality of the influent water meets current microbial stan-
dards, many of water's autochthonous bacteria have been
implicated in nosocomial infections.  Studies by Geldrich
and coworkers (1985) reported densities of heterotrophic
bacteria exceeding 1000 per ml in the effluent of GAC water
filters.  While little may be known about infective doses
for the autoucthnous bacteria, their diverse nature and
ability to colonize GAC still raises health concerns.
    A previous study (Calderon and Mood, 1988), examined the
possible health effects associated with bacteria colonizing
point-of-use (POU) GAC filters.  Since these filters are
installed in the kitchen and are used to provide water for
drinking and cooking, the health outcome examined was acute
gastrointestinal symptomatology.   In that study, no signi-
ficant health effect was associated with people drinking
water from GAC-POU filters.   The filters had average bac-
terial densities of 1000 per ml in their filter effluent.
In a this study, conducted by the Department of Epidemiology
and Public Health, Yale School of Medicine, whole house GAC
filters were studied to look at the effect of total body
exposure to higher heterotrophic bacterial densities in
filter effluents.  This study was funded by a cooperative
contract with the Environmental Monitoring, Surveillance
Laboratory, U.S. Environmental Protection Agency and by a
grant from the Water Research Council.

-------
Report for project: bacteria associated with POE filters
                               SECTION  2
                              CONCLUSIONS

      1.  Whole house granular activated carbon filters are
          colonized by heterotorphic bacteria from the water
          supply distribution system.  Once colonized, the
          filters discharge significantly higher densities of
          heterotrophic bacteria than unfiltered tap water.

      2.  Households with filters discharging higher numbers
          of bacteria in their effluent did not report a
          higher incidence of acute symptomatology.  Subjects
          in the study were asked to report on upper respira-
          tory, lower respiratory, gastrointestinal and other
          (rash, infected wounds, body aches, etc.).

      3.  Of the bacterial infections reported to health care
          providers, none were related to the bacteria colon-
          izing the GAC filters.

-------
Report for project: bacteria associated with POE filters
                               SECTION  3
                         MATERIALS AND  METHODS

      A. SCIENCE ADVISORY COMMITTEE

          A science advisory board was formed to cover the various
      disciplines needed to carry out this project.   The SAB
      contained members from the Yale study team, USEPA's Office
      of Research and Development as well as the Office of Drink-
      ing Water, members from the Water Quality Association, the
      Water Research Council and the local Regional  Water Auth-
      ority.  The members of the SAB are listed in Appendix A.

      B. MICROBIOLOGY

      Sample Collection

          Four water samples were collected monthly  from each
      filter household and two were collected from each control
      household.  A cold water tap sample from the kitchen sink
      was collected first.   The cold water was allowed to run 10
      seconds and then a 100 milliliter sample was collected in a
      sterile bottle containing sodium thiosulfate.   The second
      sample was a hot water sample from the kitchen sink.  The
      hot water tap was turned on and allowed to run until the
      maximum hot temperature was reached (indicated by steam
      rising or by touch).   A cold and hot water sample were
      collected from filter and control households.
          A filter effluent sample and a filter influent sample
      were collected from filter households.  The filter effluent
      sample was collected first by allowing 200 mis to discharge
      and then a 100 ml sample was collected into a  sterile bottle
      containing sodium thiosulfate.   Finally the filter influent
      sample was collected by allowing 200 mis to discharge and
      then collecting 100 mis into a sterile bottle  containing
      sodium thiosulfate.

      Bacterial Analysis

          The total heterotrophic plate count was measured in all
      water samples using R2A (Reasoner, 1985).   All samples
      collected before 7:00 pm were plated within two hours of
      collection.   Occasionally samples collected after 7:00 pm
      were refrigerated and plated the next morning.   Originally
      0.1 and 0.5  ml aliquots of  samples were plated in triplicate
      onto predried (35° C  for two hours) R2A.  As the study
      progressed 10 fold dilutions in sodium phosphate buffer (pH
      7.2)  were performed as indicated.
          The cold water tap,  filter influent and filter effluent

-------
Report for project: bacteria associated with POE filters

      samples were incubated at room temperature (22°C)  for one
      week.  The hot water tap samples were incubated at 35°C  for
      one week.  A previous laboratory experiment indicated sligh-
      tly better recovery of bacteria in hot water samples at the
      elevated temperature of incubation.  The extended time of
      incubation was used as Reasoner and workers (1985) reported
      increased recoveries of bacteria with extended incubation.

      Chemical Analysis

          The primary chemical analysis done by our study group
      was free and total chlorine determination.  A 25 ml sample
      was collected from the filter effluent and influent taps in
      glass screw cap tubes with teflon caps.  A Hach DPD test kit
      was used to determine the free and total chlorine concen-
      trations of each test.  After three months of testing the
      total chlorine test was dropped and only the free was done.
      One sample from each day of collection was repeated to
      verify that days collections.
          Approximately seven households with filters were chosen
      for extended organic and inorganic chemical testing.  The
      samples were collected by the Yale researchers.  The anal-
      ysis was done by the South Cental Connecticut Regional Water
      Authority (RWA).  All sampling and analysis procedures were
      done according to Standard Methods (1985).  The RWA is a
      state certified laboratory for the analysis of drinking
      water.

      C.  FILTERS

      Design

          The filter was designed by representatives from the
      Water Research Council and the Water Quality Association.
      The design was pressure tested by the Water Quality Assoc-
      iation.  The design chosen was generic in that it was a
      single tank containing two cubic feet of carbon (Figure 1).
      The two spigots, before and after, were added for sampling,
      the pressure gauges ( before and after) to monitor reduced
      flow, and a water meter to measure volume passed through the
      filter.  The carbon specified was 12 by 40 mesh with an
      iodine number of 1000.  The filter specifications are listed
      in  Appendix B.
          A competitive bid process was set up using a list of GAG
      filter manufacturers provided by the Water Quality Assoc-
      iation.  A brief description of the study and the filter
      design was sent to the manufacturers requesting a bid for
      filters,  their installation, maintenance and potential
      removal at the end of the study.  A manufacturer was chosen
      on  the quality of the services offered rather than the

-------
Report for project: bacteria associated with POE filters

      price.


      Installation and Maintenance

          The GAC filters were installed in homes immediately
      after the Regional Water Authority's water meter.   The
      installation was carried out by a Connecticut licensed
      plumber.  The initial cold and hot kitchen tap water samples
      were taken prior to the filter being installed. After the
      filter had been installed and the filter flushed until the
      water no longer had carbon fines, a filter influent and
      effluent sample was taken.
          The maintenance of the filters originally consisted of
      monitoring the before and after water pressures to be sure
      there would be no loss of flow due to particulate blocking
      of the filter.  Due to problems following installation
      chlorine removal was tested on a monthly basis.  Repairs
      were performed by a licensed plumber and consisted of re-
      pairing parts stressed by high pressure of the Regional
      Water Authority's distribution system.

      D. STUDY SITE AND POPULATION

      South Central Connecticut Regional Water Supply

          The South Central Connecticut Regional Water Authority
      (RWA)  has both ground and surface water supplies.   To have a
      homogenous supply as possible, the study participants were
      limited to those that lived only in the areas supplied by
      three surface supplies that comixed in the distribution
      system.  The study area was also chosen as the RWA has a
      known coliform problem that occurs occasionally within the
      distribution system.

      Study Population

          All families studied were recruited from the members of
      the Yale Community.  This population was chosen because 1)
      most lived in the specified water supply area; 2)  they had
      access to free medical care and; 3)  this is a relatively
      non-transitory population and therefore would minimize
      participants dropping out of the study.  Approximately 1000
      study participants were randomly selected from the Yale
      Directory.  A initial letter of invitation was sent,  briefly
      describing the study, what their participation would entail
      and a return form that asked them to answer a few questions
      (Appendix C).  From the pool of returned questionnaires,
      people were selected for the study if 1)  they indicated they
      wished to participate 2)  the source of their household water

-------
Report for project: bacteria associated with POE filters

      supply was the RWA and; 3)  there were more than two people
      in the household.  From this group only those that owned
      their own homes were eligible for the filter group.  Those
      that did not own their own home were immediately assigned
      status as a control household.   The remainder that owned
      their own home were broken into groups according to how many
      people were in their household.  Each group of households
      containing two through six people were randomly assigned to
      either a control or filter study group.  The designated
      filter households were then contacted and asked to parti-
      cipate in the study.  This initial conversation described
      that filter and asked if such could be installed in the
      house.  Some households had no basement or accessible point
      to install the water filter.  These people were then asked
      if they were willing to be controls and were replaced from
      the control list.

      Questionnaires and Human Investigation Committee

          The fist questionnaire was a short screening question-
      naire on how residents of the house felt about taste, odor
      and color of their water.  Additional questions were asked
      about number of people in the household, source of the
      household's water supply, ownership of house and source of
      medical care.  This questionnaire was accompanied by an
      information letter stating the purpose and requirements of
      the study (Appendix C).
          At the first visit all subjects, filter and control,
      were asked to read and sign a consent form (Appendix C).
      They were given a form to fill out on the age,  sex and
      allergy status of all participating members of the house-
      hold.  Each member of the household was assigned a number
      and a copy of the form given back to the household.  In this
      study health was measured in terms of acute symptomatology.
      The symptomatology was measured using a revised diary based
      on those used previously in EPA studies and a standardized
      questionnaire used to assess pulmonary health.   The diary
      method is a variation on self-administered questionnaires.
      Self-administered questionnaires were preferred for many
      reason.  The first is that this avoids the problems of
      interviewer error.  Secondly, since information on all
      household members was collected,  the self-administered
      questionnaires have been shown to collect information more
      accurately (Cannel and Fowler,  1963).   The diary consisted
      of a list of respiratory, gastrointestinal and dermato-
      logical symptoms.  The severity of the symptoms was measured
      in terms of effect on subjects  daily living such as staying
      home, staying in bed and seeking medical care.  If they
      sought medical care they were asked to call us immediately.
      The subjects were requested to  ask their health care pro-

-------
Report for project: bacteria associated with POE filters

      viders to take cultures when the cause was thought to be
      bacterial.
          This diary was exchanged for a new one at the monthly
      visit and the symptoms, if any, were reviewed by the re-
      searcher with an adult member of the house.  This supple-
      mental interviewing is recommended for self-administered
      questionnaires (Moser and Kalton, 1972).   The researcher
      confirmed the doctor's visit and diagnoses.

      Water Use Survey

          To help quantify the various exposure routes to drinking
      water in the home, a water use questionnaire was given to a
      subset of filter and control households (Appendix C).  The
      questionnaire was given twice during the course of the
      study.  The first questionnaire was given during January and
      February to cover winter months.  The second questionnaire
      was given to the same group in June and July. Approximately
      fifteen households in the control group and fifteen house-
      holds in the filter group were issued two questionnaires per
      household.   The households were instructed that at least one
      of the two people filling out the questionnaire had to be an
      adult.  The subjects were asked to fill the questionnaire
      out for two different days.  Subjects were asked about type
      of activity and the length of time the water ran during that
      activity.

      Human Investigation Committee

          All questionnaires, health diaries, consent forms,
      information letters, went to the Yale University Human
      Investigation Committee (HIC) for clearance and approval.
      No questionnaire was used until an HIC approval was re-
      ceived.

      E. MEDICAL CLINIC VISITS AND CLINICAL LABORATORY PROCEDURES

          The health survey approach to establishing an assoc-
      iation between the microbial quality of GAG filtered water
      and illness in water consumers was supplemented by examining
      the direct relationship between bacteria isolated from
      patients visiting a health care provider and bacteria colon-
      izing a GAC filter used by the study participant.  The
      participants were requested at the beginning of the study to
      notify the study coordinator if they consulted a physician
      so that a follow-up investigation could be conducted to
      determine how the illness was diagnosed and if a clinical
      specimen was obtained.  In the event that a bacterial iso-
      late was obtained from a clinical specimen, it was to be
      characterized biochemically and identified to the species

-------
Report for project: bacteria associated with POE filters

      level if possible.  Subsequent to the identification of the
      clinical bacterial isolate, the GAG in the filter serving
      the patient would be removed from the filter housing and
      examined for the presence of a bacterial isolate identical
      to the patient isolate.  Although the finding of a bacterial
      strain in a GAC filter that is similar or identical to one
      isolated from a patient using the filter is not unequivocal
      evidence of cause and effect it does present a fairly strong
      indication that the filter may have been the source of the
      causative organism.

      F.  PRIMARY DATA HANDLING, PROCESSING AND ANALYSIS

          Two classes of data were generated in this study, one
      from the field (health diaries and water use questionnaires)
      and laboratory data (bacterial analysis and chlorine deter-
      minations) .  A system of data verification and cross check-
      ing was instituted for all forms and data.  All data (field
      and laboratory) were entered into computers for data anal-
      ysis.  A system of data entry error checking was instituted
      by duplicating entry on a subset of the data and by com-
      paring various computer listings of data with orginal re-
      cords .
          For each individual water sample raw laboratory bact-
      erial data were converted to cfu per ml.  The triplicate
      values were then entered into a computer and converted to
      logarithms to generate geometric means.  Chlorine data were
      handled similarly.
          Two software programs were used to analyze the data.
      The majority of data entry and error checking were done with
      STATPAC GOLD (Walnick, 1985).  The remainder of the analysis
      and data management was done with SYSTAT ( ).
                                   8

-------
                            SECTION  4

                             RESULTS

A.  CHEMICAL ANALYSIS

     At the second visit to filter households (visit after
installation),  a subset of those households began to chronically
complain about chlorine taste and odor.  After four months of
data collection, there was a clear dichotomy in bacterial
colonization of filters.  Some filters clearly were colonized in
that the filter effluent had bacterial densities 100 to 1000
times higher  than the filter influent.  The other filters
showed little difference if any between the filter influent and
effluent bacterial densities.  Routine testing of water samples
from filter influent and effluent for chlorine residual was
instituted.  At the recommendation of the Science Advisory
Committee, a filter was considered to be not working if it did
not remove 85% of the chlorine in water.  Based on that number
it was found that 46 of the filters did not work four months
after installation.  Bacterial densities in working filters were
significantly higher than non-working filters effluents as well
as in kitchen hot and cold water tap samples (Table 1).  The
non-working filters were rebuilt and for the remainder of the
study period, filter efficiency was checked by monthly chlorine
determinations on filter influent and effluent.
     Extended chemical analyses on a subset of filters were
performed by the Regional Water Authority.  The results of those
analyses are in Appendix D.  For the majority of inorganic
measurements there was very little difference between filter
influent and effluent.  Except for filters that were not working
most filters removed total organic carbon.

B. MICROBIOLOGY

     Approximately 6000 water samples were analyzed during the
study period.  Since 44 filters did not initially work, that
data were treated as a separate group.  Table 2 shows the
distribution of samples collected for the three study groups.
An initial set of samples was taken after the filters had been
installed.  Those installation samples are summarized in Table
3.  There were no significant differences between installation
samples in the working filter group and the non-working group.
     The overall geometric means and standard deviations for the
three groups are in Table 4.  The geometric means for working
filter cold and hot water tap samples are significantly higher
than the non-working filters and controls.  The filter effluent
bacterial densities of the working filters are significantly
higher than the non-working filters.  The geometric means of
bacterial densities for the three groups are plotted by month.
The graph of the cold water taps (Figure 2)  shows that the

-------
bacterial densities for the controls and the non-working filters
are very similar.  The filter cold water bacterial densities are
approximately one to two logs higher.  The hot water tap samples
(Figure 3) are similar.  Again the control and non-working
samples look very similar to each other and the working filter
water samples have higher bacterial densities.  The effluent
samples are compared to the influent samples in Figure 4.   The
working filter effluents were one to two logs higher than the
non-working filter effluents.  The working filter effluents were
highly correlated with water temperature as measured in the
influent  (Figure 5).
     The influent samples showed a steady increase in bacterial
density over the course of the study (Figure 6).  This may have
been related to water temperature but was probably due to a
steady decline of the free chlorine residual in the distribution
system over the course of the study.  Informal discussion with
the RWA indicated that they had slowly been lowering the
chlorine residual to approximately 1 ppm at the water treatment
plant.

C. DEMOGRAPHICS

     The numbers of households, average number of people per
household and average length in study are summarized in Table 5.
A few extra households were solicited for the control group as
it was anticipated that the drop out rate would be slightly
higher.  An especially important overall characteristic of both
groups was their willingness to continue in the study and,
hence, their relative stability as a subject population.  The
vast majority stayed until completion of the agreed upon 12
month period of enrollment.  The study had originally planned to
go for only 13 months but was extended to 21 months.  These
extra months were to cover for the months several of the filters
were not working. There was no difference between the average
number of people per households.
     The distributions of all the members of the study by age
and gender are in Figure 7.  The distribution between males and
females was very similar between the filter and control group.
The distribution in age was quite different between the two
groups.  The filter group had more subjects under the age of 10
years.  The control group had more subjects over 21,
particularly over 55 years of age.

D.  Illnesses observed

     For simplicity of reporting, the symptoms were divided into
four categories.  These categories are summarized in Table 6.
Upper respiratory was defined as anything from the neck above
including respiratory allergies.  Lower respiratory was defined
as anything below the neck.  The gastrointestinal illnesses were
obvious and the remainder was categorized as "other".  Many of
the illness episodes often had symptoms that were in more than

                               10

-------
one category of symptoms.  These were reported as combinations.
The months for the non-working filter groups were deleted from
the analysis.
     The reported illness experiences of each study group are
summarized in Table 7.  The overall frequency of reported
illnesses as well as the frequencies of the categories was not
significantly different for the two groups.  The vast majority
of illnesses were principally upper respiratory.  The most
common combination was gastrointestinal with upper respiratory
symptoms.  Another interesting combination was upper respiratory
and other.  A viral disease known as "foot and mouth" disease
that consisted of flu like illnesses as well as a rash accounted
for most of those.
     A highly credible illness was defined as an illness that
required a change in the individuals normal routine such as
stayed at home, stayed in bed or sought medical care.  The
categories were reanalyzed and the highly credible symptom rate
is in Table 8.  Approximately 31% of the filter individuals'
illness episodes were highly credible as compared to 32% of the
controls.  There were no significant differences between the two
groups for the highly credible illnesses.  The bacterial
diagnoses reported were strep throat and one case of
Campylobacter enteritis that was traced to a foodborne source.
The remainder of the diagnoses were attributed to viruses.
     Information concerning fever in relation to illness is
summarized in Table 9.  The filter group reported a fever with
4.6% of their illness episodes compared with 2.6% in the control
group.  The higher reported fever rate may have been due to the
higher percentage rate of children in the filter group.  In
general fever was reported evenly across all categories of
illness.  A higher percentage of highly credible illnesses
reported a fever in both filters and controls.  The filter group
reported a slightly higher rate than did the controls
particularly for high fevers.

E.  Water Use Survey
    (to go here)
                               11

-------
                           SECTION 5.
                           DISCUSSION

     The most common short term disease complex affecting the
population of the United States is respiratory illness followed
by acute enteric illnesses.  The role of potable water in the
transmission of enteric illness as been reviewed (ref).  The
role of drinking water in the transmission of respiratory
illnesses has been addressed primarily in the hospital setting
where it is a concern regarding immunocompromised patients.  The
emergence of Leaionella (ref) and non-avium Mycobacterium
(DuMoulin, 1989) as important nosocomial infections transmitted
through water as raised health questions regarding other
commonly found heterotrophic bacteria in drinking water and
their possible relationship to health effects in normal
populations.  This has become an important issue with the
increase in use of home water treatment devices.  The GAC
devices do colonize and elute elevated numbers of heterotrophic
bacteria.  This paper is the second of two reports that looked
at elevated levels of heterotrophic bacteria (bacteria colonized
GAC filters) in drinking water.  The first study examined the
gastrointestinal exposure route (point-of-use GAC filters).
This second paper address whole body exposure to elevated levels
of heterotrophic bacteria in drinking water with emphasis on
respiratory exposure.  The elevated levels of heterotrophic
bacteria arising from whole-house GAC filters.
     Work on bacterial colonization of whole house GAC filters
has not been as thorough as the point-of-use GAC filters.
Before this study was done, it was not known what levels of
colonization could be expected and what variables would affect
that colonization.  As this study showed, these filters do
colonize and elute significantly higher numbers of heterotrophic
bacteria than that found in the distribution system.  This level
of bacteria was shown to fluctuate up and down with water
temperature.  If there had been more time and resources the
following should have been addressed: 1) Ability of these
filters to flush to level of influent; 2) Types of bacteria
colonizing the filters and relationship to those found in
distribution and; 3)  The virulence characteristics of the
bacteria that colonize the filters.
                               12

-------
Table 1.  Six month geometric means of heterotorphic bacterial densities per ml.
Type
Kitchen Cold Tap
Kitchen Hot Tap
Filter Influent
Filter Effluent
Total
Table 2. Total number of
Type
Kitchen Cold Tap
Kitchen Hot Tap
Filter influent
Filter Effluent
Total
STUDY
Working Filters
416
134
7
399
3567
G
Non-Working
15
14
10
4
1738
water samples collected for each
STUD
Working Filters
893
890
891
893
3567
Y
Non-Working
434
436
435
433
1738
R 0 U P
Filters Controls
9
17
NA
NA
2318
sample type.
GROUP
Filters Controls
1167
1151
NA
NA
2318

-------
Table 3.  Geometric mean and standard deviations of heterotrophic  bacterial
          densities from installation water samples.
Sample
Kitchen
Kitchen
Filter
STUDY GROUP
Working Filter Non- work ing Filter
Mean Std Mean Std
Cold Tap 9 6
Hot Tap 15 7
Influent 3 3
Filter Effluent 3 4
8 7
7 6
6 5
3 3
Table 4.  Geometric bacterial means and standard deviations for water  samples by
          group and sample type.
Sample
Type
Working Filter
 Mean    Std
Non-working Filter
  Hean     Std
Controls
  Mean   Std
Cold Water Tap  2080      8

Hot Water Tap     212      9

Filter Influent   13     13*

Filter Effluent      2917
                 11       9       77

                      13      1015  13

                     13      13*NA

                          25      13
                                                                     NA
* pooled samples

-------
Table 5.  Number of households, average number of people per household  and average  length of time  (months)  in
study.
Variable                                  Filters          Controls
Households                                  80                87

Average number of people
     per household                           3                 3

Average length in study (months)            17                14
Table 6.  Categories of symptoms and reported symptoms for each  symptom.
Upper                      Lower
Respiratory                Respiratory     Gastrointestinal    Other

HeadacheCoughVomitingBody aches
Runny nose                Phlegm         Nausea              Neckpain
Sore throat               Chestightness    Diarrhea           Rash •
Headcold                  Chest cold                          Infected Wound
Sneezing                  Shortness of breath
Itchy eyes                Bronchitis
Respiratory Allergies     Bronchiolitis
                          Pneumonia
                          Asthma

-------
Table 7.  Study group symptom illness rate (per person year) by illness
          cateaorv.
                                            STUDY    GROUP
Category                                 Filters          Controls
Total illness episodes                     1061             962
Rate per person-year                        4.4             3.7

Upper Respiratory                           3.7             3.0
Lower Respiratory                           1.5             1.7
Gastrointestinal                            1.2             1.1
Other                                       0.8             0.9

Combinations
 2 categories*                              1.5             1.1
 3 categories*                              0.4             0.4
 4 categories*                             <0.1             0.3

*Any combination of symptoms from upper respiratory, lower respiratory,
 gastrointestinal or other.
Table 8.  Highly credible symptom rate (per person year) for each study
          group by category of symptoms.
                                            STUDY    GROUP
Category                                 Filters          Controls
Total illness episodes                      327             305
Rate per person-year                        1.3             1.2

Upper Respiratory                           1.0             1.0
Lower Respiratory                           0.5             0.5
Gastrointestinal                            0.4             0.6
Other                                       0.4             0.4
Combinations
2 categories*
3 categories*
4 categories*

0.6
0.2
<0.1

0.4
0.2
<0.1
*Any combination of symptoms from upper respiratory, lower respiratory,
 gastrointestinal or other.

-------
Table 9.  Percentage of illness categories with fever by study group.
Category
Upper Respiratory Total
Low Fever < < 38.1 °C>
High Fever (> 38.0 °C>
Lower Respiratory Total
Low Fever
High Fever
Gastrointestinal Total
Low Fever
High Fever
Other Total
Low Fever
High Fever
STUDY
Filters
(N) %
(892) 100
(31) 3.5
(84) 9.3
(365) 100
(21) 3.0
(40) 11.0
(281) 100
(8) 2.8
(30) 10.7
(206) 100
(10) 4.8
(27) 13.1
GROUP
Controls
(N) %
(767)
(24)
(40)
(325)
(11)
(29)
(278)
(9)
(20)
(227)
(9)
(20)
100
3.1
5.2
100
3.4
8.9
100
3.2
8.3
100
4.0
8.8
Table 10.   Percentage of highly credible illness  categories  with  fever by
           study group.
Category
Upper Respiratory Total
Low Fever « 38.1 °C)
High Fever (> 38.0 °C)
Lower Respiratory Total
Low Fever
High Fever
Gastrointestinal Total
Low Fever
High Fever
Other Total
Low Fever
High Fever
STUDY
Filters
(N) %
(255)
(14)
(56)
(119)
(5)
(27)
(105)
(4)
(22)
(98)
(5)
(19)
100
5.5
22.0
. 100
4.2
22.7
100
3.8
21.0
100
5.1
19.4
GROUP
Controls
(N) %
(246)
(17)
(36)
(121)
(6)
(22)
(119)
(5)
(22)
(105)
(7)
(19)
100
6.9
14.6
100
5.0
18.2
100
1.7
18.5
100
6.7
18.1

-------