United States
          Environmental Protection
            Office of Research and
            Washington, DC 20460
September 1990
Waterborne Disease
Selected Reprints of
Articles on Epidemiology,
Investigation, and
Laboratory Analysis

                                        September 1990
    Waterborne  Disease Outbreaks

      Selected Reprints of Articles on
Epidemiology, Surveillance, Investigation,
          and Laboratory Analysis
                     Edited by
                  Gunther F. Craun
             Health Effects Research Laboratory
             Office of Research and Development
             U.S. Environmental Protection Agency
                 Cincinnati, OH 45268

The opinions presented in this document are those of the authors and do not necessarily reflect the
views of the Agency. Mention of trade names or commercial products does not constitute endorsement
or recommendation for use.

This volume is dedicated to the engineers, sanitar-
ians,  microbiologists, epidemiologists, and other
health officials who have labored throughout  this
century to improve the quality of our drinking water.
Because of their endeavors, we  in the United States
and most developed countries now enjoy a quality of
life in which waterborne disease outbreaks are no
longer a constant threat or danger. This status has
been achieved by the tireless efforts of many  and
often with little public recognition.

Although  the incidence of waterborne disease  has
been drastically reduced in  the United States, we
must not forget that infectious diseases still remain a
threat for billions of people in the developing world
where water supplies are not as well protected from
sewage and may be inadequately treated. The lessons
we have learned are important for reducing  the
worldwide incidence of waterborne diseases. We must
also continue  our vigilance because a  residual
number of outbreaks still occur  in the United States,
and this indicates  the  vulnerability of our water

We have learned that many of our water sources are
inadequately protected and treated  to prevent the
transmission of giardiasis. The waterborne trans-
mission of this disease was first reported in  the
United States in 1965  and  has increasingly been
reported  since 1971. Giardia is  now the most
commonly identified  pathogen  in  waterborne  out-
breaks. Contaminated surface water  is a significant
source of infection  for giardiasis, and almost half of
the outbreaks of waterborne giardiasis have occurred
in surface water systems where the  only treatment
was disinfection. In addition to Giardia, a number of
new etiologic agents have been recognized in recent
years and as laboratory analysis and investigation
procedures improve,  more agents are likely to be
identified. Some of the newly recognized waterborne
pathogens include Cryptosporidium, Campylobacter,
Yersinia, hemorrhagic  E. coli 0157:H7,  rotavirus,
and Norwalk agent. This volume also describes a
chronic diarrhea for which  an agent has yet to be
identified despite extensive laboratory analyses.

Only a fraction of the  waterborne outbreaks that
occur in the United  States are recognized, inves-
tigated, and reported, and in only half of the reported
waterborne outbreaks was an etiologic agent iden-
tified. We must improve the investigation and sur-
veillance of waterborne disease to identify the causes
of these outbreaks and the etiologic agents. The
identification of Giardia as an important waterborne
pathogen  stimulated the research which has
identified  the types of filtration, operating param-
eters, and disinfection levels necessary for removal
and inactivation of this pathogen. Water  treatment
research is now being conducted to obtain similar
information for Cryptosporidium.
The emphasis of this volume is on drinking water
systems and the prevention of infectious waterborne
diseases. We must improve surveillance,  reporting,
epidemiologic  investigations, and laboratory anal-
yses, and we cannot rest on past  accomplishments.
Continuing microbiological challenges are a remind-
er that our task to eliminate waterborne illness is yet


Preface  	   iii
Acknowledgments 	 vi

  I.   Introduction  	  1

      Epidemiology in Search of Infectious Diseases: Methods in Outbreak Investigation 	   2

 II.   Disease Surveillance 	  7

      Waterborne Disease Outbreaks in Colorado:
         Three Years' Surveillance and 18 Outbreaks 	    8
      A Three-State Study of Waterborne Disease Surveillance Techniques  	   12
      Routine Coliform Monitoring and Waterborne Disease Outbreaks  	   14
      A Follow-up Study of Gastrointestinal Diseases Related to
        Bacteriologically Substandard Drinking Water 	   18
      Risk Factors for Endemic Giardiasis  	   21
      Epidemiologic Surveillance for Endemic Giardia Lamblia Infection in Vermont 	   24
      Cryptosporidiosis and Surface Water	   31

 III.   Investigation of Waterborne Disease Outbreaks  	   35

      Community Health Effects of a Municipal Water Supply
        Hyperfluoridation Accident   	   36
      An Epidemic of Gastroenteritis Traced to a
        Contaminated Public Water Supply  	   39
      Chronic Diarrhea Associated with Drinking Untreated Water   	   47
      Epidemic Giardiasis Caused by a Contaminated Public Water Supply 	   54
      A Waterborne Outbreak of Cryptosporidiosis in Normal Hosts   	   59
      Large Community Outbreak of Cryptosporidiosis Due to
        Contamination of a Filtered Public Water Supply  	   62
      Waterborne Outbreak of Cryptosporidiosis 	   67
      What Happens When the Multiple Barrier Concept Is Neglected  	   68
      Chlorination to Prevent and Control Waterborne Disease   	   74

 IV.   Sampling and Analytical Methods 	   79

      Small Round Structured Viruses: The Norwalk Family of Agents  	   80
      DNA Hybridization Probe for Clinical Diagnosis ofEntamoeba histolytica 	   84
      Diagnosis of Giardia lamblia Infections by Detection of Parasite-Specific Antigens  	   90
      Identification of Cryptosporidium Oocysts by Monoclonal Antibody 	   95

A number of previously published articles have been reprinted in this volume to
supplement information contained in a separate publication, Methods for the Investigation
and Prevention of Waterborne Disease Outbreaks. These articles provide examples of
outbreak investigations and surveillance activities. They have been reprinted to illustrate
various aspects of epidemiologic methods and engineering principles and compiled as a
convenience to those investigators with limited access to library facilities. Articles have
also been included on new clinical procedures for viral and protozoan pathogens as these
are discussed only briefly in the methods volume. The contributions of the authors of the
journal articles are gratefully acknowledged.

Permission to reprint these articles has been provided by: Journal of Epidemiology and
Community Health,  American Journal  of Public Health, Journal of Environmental
Health, American Journal  of Epidemiology,  Annuals  of Internal Medicine,  The New
England Journal of Medicine, Lancet, Journal American Water Works Association, and
Journal of Clinical Microbiology.

This report has been reviewed by the U.S.  Environmental Protection Agency and approved
for publication.

I. Introduction

                                         Reproduced,  with  permission,  from  Journal
                                         of  Epidemiology and  Community Health.
Journal of Epidemiology and Community Health, 1989, 43, 311-314
 Review article: Research methods in epidemiology, III

 Epidemiology  in search of infectious diseases: methods
 in outbreak investigation

 From the Communicable Disease Surveillance  Centre (Welsh Unit), Cardiff Royal Infirmary, Newport Road,
 Cardiff CF2 1SZ, UK.
The value  of epidemiology in  the investigation of
incidents of infectious disease in the UK does not
seem to  be  always  readily  appreciated  within
microbiology' 2  or  environmental  health.3  Yet
because microbial disease is a result of the interaction
of the agent, the host  and environmental  factors4
neither  laboratory  diagnosis  nor  environmental
inspection are of themselves sufficient for successful
investigation and control. Simply isolating microbes
from the environment, even if appropriate samples are
available, cannot distinguish between a reservoir of an
agent and the outbreak source or cross contamination
from the   reservoir  or  source.  Microbiologically
plausible explanations of outbreaks may be wrong,
but it is often difficult to persuade investigators of the
need to  test such hypotheses.
  On the other hand, whilst epidemiology may have
evolved from the study of infectious disease epidemics,
academic epidemiology  seems in  the main  to have
passed infectious  diseases  by in the  UK.  This  is
regrettable as there is still the need to consolidate the
foundations of a section of public health medicine
which has  refused to disappear.5 This review, in
following   the   steps   taken   in   an  outbreak
investigation,6 draws attention to both the usefulness
of   epidemiology  and   to  regularly  recurring
methodological problems which may be encountered.

Describing the outbreak

The  classical "shoe leather" approach to infectious
disease  incidents  is the  detailed investigation of
individual  cases in  the light of the well established
microbiological  features  of the particular infection
and   a  confidence   in  the  possible  routes  of
transmission.7 The background incidence of infectious
diseases  in  the  UK is  usually  well known  and  is
sufficiently low not to require sophisticated statistical
methods of cluster  analysis8 to  recognise  a local
outbreak. However, with newly recognised infections
or diseases presumed to be caused by an unknown
infectious agent, surveys  of disease  incidence or
laboratory isolation rates may be necessary to confirm
that an outbreak has occurred. The methodological
problems encountered are  similar to those found in
investigating clusters of cancers.8
  The cases first reported in an outbreak are usually
only a small and sometimes biased sample which may
mislead,  rendering  control  measures  ineffective.
Identification  and   definition  of  the  exposed
population  is necessary  to  enable thorough  case
finding to be carried out.  Paradoxically, however, a
single case which does not fit the usual characteristics
of time,  place and  person may provide the most
valuable evidence of the source of infection.7
  Data from cases are characterised by time, place and
person.6 The epidemic curve, a plot of the number of
cases over time of onset of illness, is the most useful
and  immediate means  of  assessing  the type of
outbreak. In point source outbreaks in which all cases
are exposed at a given time, onset of symptoms of all
primary cases will  cluster within the  range  of the
incubation  period.  For each  organism there  is  a
characteristic range within which infecting dose9 and
portal of entry, as well as host factors10 (eg, age,
immunosuppression)   give   rise   to   individual
variability. In outbreaks propagated from person to
person the  occurrence of  cases will be spread over
several incubation periods with peaks at intervals of
the incubation period. In larger community outbreaks
of diseases spread by person to person the epidemic
curve is  usually smoother  and the peaks at the
generation time of the new  cases less obvious. Despite
their almost universal use in outbreak investigations,
there has been little formal evaluation of their use in
practice.  Sartwell  in 1950  established  that the
frequency  distribution  of incubation  periods  of
infectious diseases of both  long and short incubation

period followed a  log normal curve.11 lie suggested
(lie use ol the dispersion factor, (he anlilogarilhm ol'
Ihe  logarithmic  standard  deviation,  to   measure
variation in incubation periods. Such analyses may be
used  to  distinguish  between  point  source  and
continuing source  outbreaks, estimate Ihc data  ol' a
common  exposure,  determine  time  of  maximum
infecliousness, help explain pathogenic mechanisms,12
'dcnlily  Ihe  occurrence of secondary spread,  anil
evaluate control measures. However, Armenian and
Lilienfckl have claimed (he initial efforts of Sarlwell
"have not been followed by any systematic review of
the data on incubation  periods".13

Developing ii hypothesis

The  traditional  distinction  between  descriptive
(hypothesis generating) and  analytical  (hypothesis
testing) studies has been criticised by  Rothman,14 a
criticism which holds true in outbreak investigation.
At the  simplest level cpidemiological investigation
may be described  in two stages: Instly, generating a
hypothesis  of  causation  from   descriptive   data
analysed  by lime, place and person,  and  secondly,
lesting  thai hypothesis by a case-control or cohort
stuily. liul in almost all outbreaks competing a priori
hypotheses are voiced at  Ihe outset by public health
investigators, patients, employers, food  producers,
elc. Some ol these  hypotheses may be excluded by the
"descriptive" data. These descriptive dala in turn may
stimulate alternative hypotheses. Indeed, as in clinical
investigation of an individual patient,  this process of
hypothesis  generation  anil  refutation  goes  on
continuously as new cases are being interviewed, and
challenges the  rigid  dichotomy of descriptive versus
analytic.  l;or example,  in an outbreak ol'.Sulinmii'lla
eating, cases were mainly infants who were discovered
to have been fed on a single brand of milk formula.15
Comparison with a group of well  infants showed  a
significant difference in use  of this brand.  Was this
comparison  lesting  the hypothesis, or could such  a
design  only  firm  up the  hypothesis—the hypothesis
seemingly was  generated on the same set of cases as it
was "tested"?  In such outbreaks a  priori hypotheses
are   formulated  on  Ihe  prior  knowledge of the
epidemiology of salmonellas and, in this case, on the
observation of the age distribution of eases. The search
for a common milk formula led to Ihe infants followed
a prediction based  on the deduction lhal  Ihe  mosl
likely vehicle of infection in infants was formula feed.
   In some outbreaks  the descriptive epidemiology
suggests hypotheses for which there are no biological
explanations'6 '7 (eg, a cluster of three casesof aplasliu
anaemia in children"1)  or which go  against the
accepted microbiological theory (eg, an outbreak of
shingles'''1 2I). In the  practice of  infectious disease
control,  Ihe  framework  of proven  microbiological
cnuses and modes of transmission enables statistical
data  to  be  interpreted  confidently. Once  outside
known causal pathways statistical associations are
difficult to interpret and public health action, like Ihe
clinical diagnosis of an individual patient, usually is
taken on the basis of clinical experience and instinct.

Hypothesis testing

When   hypotheses   are   formally   tested   the
epidemiological    techniques   usually    used   are
retrospective  cohort and  case-control  studies.  In
certain outbreaks, eg, food poisoning outbreaks  in
institutions or  following  receptions, it  is  usually
possible  to identify retrospectively and follow up the
whole   population    to   identify   simultaneously
exposures and'disease slalus. (This itself may lead  to
biases which  are described below.) Comparison  of
attack rates in  exposed  and  unexposed (eg,  food
specific allack rales) or exposures in cases and controls
will often implicate the source. Paradoxically, when a
vehicle is universal, eg, lap water, it may be impossible
to demonstrate an association  with illness unless ihere
is a dose response (which is less likely if the vehicle of
infection is  heavily contaminated).  In addition  to
universal problems  of matching, confounding and
appropriate  analysis  (here are  regularly recurring
problems of design and implementation in outbreak

Sampling may not be required at all  as Ihe si/.e of the
population at risk or the number of cases is usually
small. I lowever (he corollary of this  is lhal Ihe power
of possible studies in outbreak investigations cannot
usually   be  controlled '  by   Ihe  investigators.
Nevertheless, the speed at which such studies can  be
done by experienced investigators, and  the relatively
low cosl of Ihe studies, as well  as the potential benefits
of demonstrating an association between infection and
a possible source, mean that studies of low power will
usually still be worth carrying out.
   Diagnostic suspicion  bias"2  in case selection is
always a problem when a possible vehicle  of infection
is widely publicised. For example, people who have
eaten a food believed to be the source of an  outbreak
of botulism may present with  neurological symptoms
 unrelated to botulism. When laboratory confirmation
of diagnosis is  not possible the only safeguard is to
choose  cases which  meet the strictest clinical case
definition (and  therefore who are at the most severe
end of the disease spectrum).
    In case control studies  Ihe selection  of controls
always poses a problem of practicality versus validity.
 Sampling frames commonly  used to select  controls

 Methods in outbreak investigation
 include general practitioner age/sex registers, hotel
 and reception guest lists, family members  of cases,
 neighbours of cases, and persons investigated by the
 laboratory but who were negative for the disease in
 question. When diagnostic access bias22 is considered
 potentially, important (eg, cases  of enteric infections
 are only recognised in the main  if they have general
 practitioners  who  regularly   recommend  faecal
 sampling), it is important to match for neighbourhood
 of  residence,  or  use  patients   investigated  by
 laboratories but found to be negative, since failure to
 do so may lead to a spurious association with, say, a
 restaurant in the neighbourhood of the "keen" general
 practitioner. However, using general practitioner lists
 may result in overmatching. Obtaining controls from
 general practitioner registers is laborious and slow,
 and in  our  experience  about  5-10%  of general
 practitioners are reluctant to cooperate.
   Increasingly, acquaintances nominated by cases are
 used as controls15 although the representativeness of
 such groups and  possible biases  they may introduce
 have not been evaluated systematically. It is possible,
 for example,  that if cases know  that a particular
 exposure  is   the  source  of the  outbreak their
 nomination    of   controls   will   be   influenced.
 Additionally, the case may know  the exposure  history
 of several acquaintances and  reveal this with their
 names to the interviewer, so possibly influencing the
 interviewer's choice of controls. A protocol to avoid
 this may be derived, although in  practice there is the
 danger  of  a conflict  of  priorities  when local
 investigators, eg,  environmental  health officers with
 their   own   local  responsibilities  for  outbreak
 investigation, are called upon to carry out the national
   Misclassification  of controls can  be  avoided by
 asking all the individuals in the cohort  or control
 group about  symptoms  over  the  appropriate time
 period so that those who may be unrecognised cases
 can  be excluded  or reclassified. The feasibility  of
 serotesting, swabbing or otherwise testing controls to
 exclude or reclassify asymptomatic infected persons
 should  always be considered.  However, few  well
 people  are willing to  give a sample of faeces,  and
 obtaining blood samples from well children is difficult
 and  may be  considered  by some clinicians to be
In all but the smallest outbreaks data are collected by
questionnaire, but often practicalities require groups
to be interviewed by a combination of telephone, mail,
face to face and proxy interviews. The sensitivity and
specificity of questionnaires is seldom considered. In
one study food consumption recalled by people 2-3

days after luncheon was compared with that observed
at the time and recorded on video tape.23 Only four of
32 attendees made  no  errors  in  reporting.  The
sensitivity   of  the  food  consumption   histories
questionnaire was 87.6% and the specificity 96.1%.
Thirteen percent of the respondents reported eating
"dummy" food items listed on the questionnaire but
not served at the luncheon. Such errors in recall can be
reduced by  providing background details of events,
good questionnaire design, and making use of other
sources  to   check  data  such  as diaries,  menus,
discussion with  relatives, etc. Validation of patient
histories is often not  possible in practice due to the
nature of the exposures, the lag between exposure and
investigation  and  the  urgency  of  investigation.
However, random errors in recall are unlikely to give
rise  to false  associations between illness and,  for
example, a  food item.  Rather they will reduce the
power of a study  to identify  the  true vehicle of
infection. For some types of data, such as history of
immunization, patient or parent recall are particularly
inaccurate  and  validation of vaccine history from
medical records  is required.24
   Loss of cases or controls from the study because of
refusal to be interviewed or failure to trace patients is
much less a problem in outbreak investigation since
the affected population is usually eager to know the
cause, and  response  rates should approach  100%.
Throughout the course of their illness cases will have
gone over in their minds possible exposures and their
recall may be biased by their own preconceptions or by
press and media speculations.
   Cases  will often  have been interviewed on many
occasions (by general  practitioners,  environmental
health officers,  community  physicians)  before an
analytical study  is  carried out, and  this may have
introduced   bias  from  suggestions  made   by
interviewers, as  well as  prompting a more detailed
recall. Family members may have suggested likely
sources of infection to the cases. Patients may have
anticipated the view of the interviewer if the suspected
cause of the outbreak has been learned from local
media reports and they  may  alter their  answers,
accordingly. These biases can often be anticipated and
can sometimes be assessed in analysis. For example, in
a  suspected waterborne  outbreak of giardiasis in
Reno, Nevada, considerable local publicity occurred
before  the   case-control study  was  conducted.25
However, because of the long incubation period of
giardia many new cases became ill several days after
the water supply was treated and so media attention
persisted and water  was still considered  suspect.
However, increased water consumption  was  only
associated with cases exposed before the water supply
was treated. Bias from media reports was likely to have
affected later cases also but no association was found.

   A major methodological problem is (he bias which
may result from  the interviewer knowing the disease
status of the person and having his own suspicion or
prejudice as to the source. Selective histories may be
taken  in which the patient with  salmonella is  only
•usked   in  detail  about  a  particular  food.  Blinded
interviewing is rarely possible since- resources are
limited (most outbreaks are investigated  by only one
.or two people), anil indeed exposure history is usually
obtained  at  the  same  lime  as  disease  history.
.Professionalism in the technique of interviewing, and
use  of a  structured  questionnaire  are  the  only
safeguards. Controls, who  anyway have no special
reason  to  recall  certain  exposures,  are  usually
interviewed several days  after cases,  which adds to
problems of accuracy and  detail of recall. To help
overcome  this  problem,  questionnaires may include
memory jogging points related to  notable events, etc.
It is important that the controls are interviewed as
carefully and thoroughly  as the cases were.


 Do epidemiological studies in outbreak investigations
have to be "quick and  dirty" or can they be "quick and
clean"? They must be quick. There is a need to develop
and evaluate epidemiological  methods in outbreak
investigation so  that  they are  both  practical  and
.sound. Hpidemiological standards for the practice of
the  newly  re-established  discipline  in public health
medicine  -communicable disease control - must be
developed \so that  attributed causes of outbreaks are
not  left to be a "matter of opinion".2'1

 1  Anonymous. Stanley Royd: the epidemiological lesson, lir
    Med J. 1986; 292: 644  5.
 2 Powell OKU. Salmonella cnlerilidis. Lancet 1989; i: 99.
 3 North R. .•/ failure <>t government. A report on ilie mA of
    /'(Kill puixiHiiiif; from i-jf^.v anil i'KK I'riiiliicls. Bradford:
   Toodguard Publications, 1989.
 "* Bracliman I'S.  The control of  infectious diseases.  In:
    Holland WW, Delels R, Knox Ci, eds. Oxfordlexllwok of
    pt/hlic liciillli, Vol 2. Oxford: Oxford University Press,
    1985: 8  18.
 5 Kviins AS.  The  eradication  of communicable diseases:
    myth or reality? Am J  KpidcinMl 1985; 122:  199 207.
''Gregg  Mil.  The   priiH'iplcs  of  an  epidemic  Held
 , investigation.  In: Holland WW, Delels R, Knox G, cds.
   Oxford lexlhook of pnhtie health. Vol  2. Oxford: Oxford
   University Press, '1985: 284 99.
7 Galbraith NS. The application of epidemiological methods
   in the investigation and control of an acute episode of
   infection.  In:  Holland WW, Delels  R, Knox  (i, eds.
   Oxford ti'Mhinik tifpnhlic health, Vol  4. Oxford: Oxford
   University Press, '1985: 321.
"Openshaw  S,  Craft  AW,   Charllon   M,  Birch  JM.
   Investigation  of  leukaemia  clusters  by  use  of a
   geographical analysis machine. Linnet 1988; i: 272 3.
'' Blaser MJ, Newman I.S. A review of human salmonellosis
   I. Infective dose. Rev Infect /)/v 1982; 4: 109ft I Oft.
10 Khoury MJ, Handcrs WO, Greenland  S, Adams MJ  On
   the  measurement   of susci-plihilily  in cpidcmiologic
   studies. Am J Lpidemiol 1989;  129: 18.1 90.
" Sartwell PI1'. The tlisliJlnilion of incubation pounds of
   infectious disease.  Am .1 llyx 1950; 51: .110  18.
12 Sarlwell PK. The incubation period and  the dynamics of
   infectious disease.  Am J lipiilemiol 1966; 83: 204  16.
13 Armenian  UK,  l.ilienfeld  AM.  Incubation  period  of
   disease. Kpidemiol Kev 198.1; 5: I  15.
14 Rolhnian  KJ.  Modern cpiile»ii<>lot;i'.  Boston, Toronto:
   Little Brown, 1986.
15 Rowe B, Begg NT, I lulcliinson DN, el i//. Salmonella
   ealing infections associated with consumption of infant
   dried milk. hineet  1987; ii: 900 .V
"'Scliwarl/ RS, Callen JP, Silva J. A cluster of llodgkin's
   disease in a small commnnily. Am J K/iitlemiul 1978; 108:
   19 26.
17 Ciordis L. Challenges to epidemiology in  the next decade.
   Am J Kpidemiol 1988; 128:  I 9.
111 Morgan  GJ,  Palmer  SR,  Onions  O,  Anderson  M,
   ("artwrighl RA, Benlley DP. A cluster of three cases of
   apliistie anaemia in children. Clin l. llucmiiiiil 1988; 10:
   29 32.
I>J Palmer SR, Caul  HO, Donald Ol-:, Kwantes W, Tillell II.
   An outbreak of shingles1.' l.aneel 1985; ii: 1108  II.
20 PeloTKA, (iilks CI-', Juel-Jensen UK. Clusters of shingles.
   Laiiert 1985; ii:  14.1.1.
21 Palmer SR,  'lillell  II.  Shingles clusters. Lancet  1986, i:
   27.1 4.
22 Siickelt 1)1.. Bias in analytic lesearch. ./ ('hrnnie /)/'.v 1979;
   M: 51  63.
-'Decker  MO,  Booth  AL,   Dewey   MJ, l-'ricker  RS.
   llulclieson  Rll,   SchalTncr  W.  Validity of  food
   consumption   histories  in  a  I'oodhorne oulbieak
   investigation. Am J Kpulemiiil 1986, 12-1, 859 6.1.
24 Orenslein  WA, Bernier Rll, Oondero  TJ, el ,il. l-'ield
   evaluation  of vaccine efficacy. Hull H'llO 1985; 63:
   1055 68.
25 NavinTRJuranek DO, ford M, Minedew D.I. l.ippy K.C,
   Pollard RA. C'nse control study of walei borne giaidiasis
   in  Reno, Nevada.  Am J Kpidemitit 1985; 122: 269 75.
26 I louse of Commons Agriculture Committee. Salmonella in
   I'XKS-  l''irxl Report, Vol  2  (Minnies of evidence  and
   Appendices).  London: IIMSO", 1989.

II.  Disease Surveillance



Reproduced,  with  permission,
Journal  of  Public Health
                               Waterborne Disease in Colorado:
                       Three  Years' Surveillance and  18 Outbreaks
                             LINDA EISNACH, MPA, AND RICHARD J. KARLIN, PE
    Abstract: The Colorado Department of Health conducted inten-
sive surveillance for waterborne diseases  during the three-year
period July 1,  1980-June 30,  1983.  Eighteen outbreaks of water-
borne illness were investigated. Outbreaks involved from 15 to 1,500
ill persons. Giardia lamblia was confirmed or suspected as the agent
in nine outbreaks, rotavirus in one, and no agent could be identified
in eight. Seventeen outbreaks occurred on surface-water systems;
none of these had adequate chemical pretreatment and filtration.
                Investigation of water systems exhibiting positive coliform results
                during the first year detected no outbreaks. Activities important to
                effective surveillance included educational outreach programs to
                local health agencies, physicians and the public, and the designation
                of one individual to whom all water-related public complaints and
                health department inquiries were directed. (Am J Public Health
                1985; 75:254-257.)
    There has been  a  steady  increase in  the  number of
reported  waterborne  disease  outbreaks  occurring in the
United States since the early  1960s. The  average annual
number reported  for 1971-72  was 32; from 1980-82 the
average was 41.1.1 The actual number of outbreaks that
occur is certainly greater than this.
    Over a three-year period prior to June  1980, six water-
borne outbreaks were reported in Colorado but about 20
additional clusters of gastrointestinal illness was suspected
to be waterborne. Energy development, tourism, and popu-
lation growth combine to place stress on water systems that
typically have relied on protected or remote surface sources
of supply with marginal treatment.
    Prior to'June  1980, the Colorado Department of Health
had a passive waterborne outbreak surveillance system, and
responsibility for follow-up of water-related complaints was
divided among two Sections of the Water  Quality Control
Division and the Communicable Disease Control Section.
    From June  1980 through June 1983, a field epidemiolo-
gist was  made responsible for improving the detection of
waterborne illness.* The project was housed in the Commu-
nicable Disease Control Section, Division of Disease Control
and Epidemiology, which  receives reports of communicable
disease cases from the state's 63 counties.
    In the first year, surveillance was expanded to include a
daily review of coliform results from the required testing of
community and non-community water systems. When posi-
tive samples were received, physicians, pharmacists, public
health  nurses, and schools  in areas  served by the affected
water  systems were promptly surveyed by telephone for
:    *US Environmental Protection Agency contract #68-03-2927, June 1980-
June 1982.
    Address reprint requests to Dr. Richard S. Hopkins, Colorado Depart-
ment of Health. Communicable Disease Control Section, Division of Disease
Control and Epidemiology, 4210 E. 11th Avenue, Denver. CO 80220. Authors
Shillam, Gaspard, and Eisnach are also affiliated with that Division; author
Karlin is with the Drinking Water Section. Water Quality Control  Division,
CDH. This paper, submitted to the Journal May 7, 1984, was revised and
accepted for publication October 15, 1984.
 © 1985 American Journal of Public Health 0090-0036/85 $1.50
                evidence of an outbreak of gastrointestinal disease, without
                waiting for check samples.
                    Throughout the project, engineers in the Water Quality
                Control Division supplied information from their inspections
                of water  systems that  indicated  a potential  hazard, and
                investigated water systems when a waterborne outbreak was
                suspected. In the second and third  years of the project,
                questionnaires were mailed to all persons with laboratory-
                confirmed giardiasis reported to state or local health depart-
                ments. Two  or  more laboratory-confirmed giardiasis case
                reports linked by time and place or two or more independent
                citizen reports from  the same  area of excessive illness
                suspected of being waterborne prompted an active surveil-
                lance effort as described above.
                    In May  1981, an education program directed toward
                county health officials was  instituted; this included formal
                presentations to I'O of the 13 largest  county health depart-
                ments. Public health nurses and sanitarians were encouraged
                to report to  one another and to the  project office unusual
                occurrences  of  disease  and problems discovered in local
                water systems.
                    Public  awareness  about waterborne  diseases was
                heightened through media coverage of outbreak  investiga-
                tions. Complaints from the  public about illness or poor
                quality drinking water were directed from the engineering
                sections  to the waterborne disease project.
                    The state health department's bi-weekly Colorado Dis-
                ease  Bulletin (circulation 4000) was used to inform private
                physicians and local health agencies of the project and to
                solicit reports  of potential waterborne  disease  problems.
                Outbreak investigations were described in  the Bulletin to
                maintain interest in the  project.
                     Community surveys conducted in the investigation of all
                potential  waterborne outbreaks  sought information about
                 various symptoms, onset dates,  duration, amount of water
                consumption prior to illness,  and water supply source. The
                 standard water question was: "How many 8-ounce glasses
                 of unheated tap water, or drinks made up with unheated tap
                 water, did you drink on an average day before (date of
                 outbreak)?'  These  surveys  were usually  conducted via
                 telephone, using modified  random digit dialing techniques
                 when a  water system's distribution  area approximated an
                 area  served by all or part  of one telephone  exchange.
                 Samples  were  randomized and  consisted  of at least 40
                 households.  In  several  investigations dealing with persons
                 who  had dispersed from a common site, mailed question-
                 naires were used.
                                                                                      AJPH March 1985, Vol. 75, No. 3

                                                                    WATERBORNE DISEASE OUTBREAKS IN COLORADO
    An outbreak was considered to be waterborne when a
positive dose-response relationship between amount of wa-
ter consumed and attack  rate could be demonstrated and
when there  was a significant difference in attack rate be-
tween persons drinking from the implicated supply and those
drinking no  water from the implicated supply. The project
protocol required a minimum of six stools from  ill persons
for  examination  and  ideatification  of pathogenic  agents
including  bacteria (Salmonella,  Shigella,  Campylobacter
and Yersinia), intestinal parasites, and viruses and the same
number from well  controls.  When  a viral outbreak was
indicated  by symptom  profile, the protocol also required
collection of paired sera from a minimum of six ill and six
well individuals. Viral specimens were frozen and sent to the
Health  Effects Research  Laboratory, US  Environmental
Protection Agency (EPA), Cincinnati, Ohio.  Bacteriological
and parasitological laboratory analyses were conducted at
the Colorado State Laboratory  by  standard  methods.  Be-
cause  of  late notification of outbreaks, not all indicated
specimens could be collected in all outbreaks.
    In suspected waterborne outbreaks, water samples were
collected  and examined for total coliforms  and  fecal coli-
forms.  In outbreaks of suspected giardiasis, 1,000 gallons
each of treated and untreated water were filtered through a
one micron filter and the filter examined for the presence of
Giardia cysts  by Dr.  Charles  Hibler  of  Colorado State


    Eighteen outbreaks of gastrointestinal  disease were
shown to be waterborne.  An additional 10 outbreaks were
strongly suspected of being waterborne. Twelve reports of
outbreaks were sufficiently suspicious to warrant significant
further investigation, but  most  were shown to  be  due to
person-to-person spread, not waterborne.
Etiology of Outbreaks
    Nine  of the 18 waterborne outbreaks are considered to
have been caused by Giardia lamblia, although confirmation
of the responsible agent is not equally good in all nine (Table
1).  One was associated  with   rotavirus.2  No   agent was
identified for the remaining eight outbreaks. In four of these
outbreaks (#1,  2, 3, 6) which resembled  Norwalk virus
gastroenteritis clinically, epidemiologic investigation began
at a time when no new cases were occurring and a decision
was made not to collect stool or serum samples.  In the four
others (#13, 16-18), the symptom profile suggested bacterial
or parasitic  disease but laboratory  tests on ill persons for
bacteria and parasites were negative.
    In one of the presumed giardiasis outbreaks (#9), initial
reports of numerous laboratory confirmed  giardiasis cases
dissuaded us from aggressive pursuit of stool specimens only
to find  that,  in fact, physicians' diagnoses were based on
clinical picture. In another (#7), all the symptomatic cases
had already been treated by the time we were able to collect
stool specimens. The only  stool  positive for Giardia lamblia
was from a well, untreated camper.
Source of Notification of Outbreaks
    Eleven of the 18 waterborne outbreaks were  brought to
the  attention of local health departments by citizens com-
plaining by  telephone of illness in themselves and  their
neighbors; four were brought  to state health department
attention directly by such  citizen phone calls.  Three out-
breaks were  picked up by  local health departments in the
course of routine active disease surveillance, as clusters of
illness were reported by  doctors' offices when the local
health department made its regular weekly telephone con-
tact. Two outbreaks were  reported by physicians, and two
outbreaks came  to our attention through a combination of
direct contacts by ill persons and ongoing review of positive
Giardia lamblia results in the state laboratory.
    No outbreaks were found through 12 months of daily
review of positive coliform test  results  from  state water
systems although an average of two unsafe coliform results
per week were  followed  up.  This activity was therefore
discontinued in subsequent years.
Epidemiologic Data
    In all 18 outbreaks, a statistically significant association
was  found  between self-reported water  consumption and
illness (Table 1). In 11 of the  18 outbreaks, a control group
on a  different water  supply  was interviewed and  had a
significantly lower  attack rate. Community  attack  rates
ranged from 10 per cent to 64.4 per cent. Outbreaks occurred
in  all seasons, but 11 of 18 occurred between June 1 and
September 30.
Type of Water System
    Community  water systems** accounted for 15 of the 18
confirmed and seven of the  10 suspected waterborne out-
breaks. One confirmed outbreak (#2) was related to a well
not intended as a drinking-water supply, and the remainder
(two confirmed,  three suspected) were on non-community
systems. There are 715 community water systems and 1,250
non-community  systems registered with the Colorado De-
partment of Health. The largest system with a community-
wide  problem served approximately  3,500 persons. One
outbreak (#1)  occurred on a supply serving approximately
50,000 people, but only a few hundred were affected by  the
    All but one of the outbreaks (#1) occurred on systems
using surface  water as the  source  of  supply. In  three
outbreaks (#11,  16, 17), the sources were shallow wells in
glacial alluvium  immediately  adjacent to a river;  pollen
grains and  other particles characteristic of surface water
were found when the "well"  water was  sampled through a
one micron filter.
Coliform Counts and Chlorine Residual
    Coliform counts were elevated at or near the time of the
outbreak in seven outbreaks,  normal in nine, and unknown
in  two. Chlorine  residuals  were inadequate in six, adequate
in  eight, and unknown in four outbreaks. In only one of the
nine documented or suspected Giardia  lamblia outbreaks
were there elevated coliforms and absent disinfectant at the
time of the outbreak. In one Giardia outbreak (#9), coliform
counts indicated  safe water on two  occasions during  the
outbreak but were elevated two weeks  after the outbreak
Deficiency Responsible for Outbreak
    One outbreak (#1) was caused by a contaminated well
and  another (#2) by a cross-connection.  The  remaining
outbreaks occurred on systems without  adequate chemical
pretreatment, filtration, and chlorination. In the nine water-
borne outbreaks attributed to Giardia lamblia, there was
either no filtration at  all (six outbreaks), a malfunctioning
filter and decreased chlorine contract time (one outbreak), or
   "Defined as 15 or more service connections and/or 25 or more residents
AJPH March 1985, Vol. 75, No. 3

TABLE 1—Epidemlologlc Features of 18 Waterborne Disease Outbreaks, Colorado, 1980-83


































# of

































. 11.4

Relative Risk1
(90% C.I.)

1.56(1.13, 2.16)

2.7(1.70, 5.38)



2.22(1.17, 4.20)

4.21 (1.09,


8.12 (2.86,
2.3 (.98, 5.42)


1 1 .02)

2.71 (1.59,4.63)


4.54 (.74, 27.7)

2.18 (.66, 7.2)

6.00 (.71, 51.1)


















1-2 days
3-5 days

2-3 days

3-4 days

3-1 4 days

2-3 days

7-10 days

7-14 days

4-30 days

7-14 days

5-30 days

7-30 days

2 days

5-30 days

3-20 days

3-30 days

1-5 days

1-10 days
















Evidence for Agent


Positive serology &
stools in cases
clinical picture
Positive stools in
cases, clinical
picture, water filter
showed large
particles passing

Clinical picture in
cases, positive
water filter
Positive stools in
cases, negative
stools in controls,
clinical picture,
positive filter
Clinical picture,

Positive stools

Positive stools in
cases, clinical
picture, positive
Stools syndrome


Stools syndrome,
positive filter
Stools filter




Agents Cause of
Ruled Out3 Outbreak
— Cross-connection
SS.C Laundry well used
for drinking
SS,C Chlorinator failure,
SS.C.EEC Chlorinator failure,
— Unfiltered surface

— Unfiltered surface
— Filter off line

SS.C Shortened chlorine
contact time,
filter inadequate

— Unfiltered pond
water backed
up into spring
SS No pretreatment
before filtration
No pretreatment
before filtration

C Unfiltered surface
source, heavy
S,C virus Unfiltered surface
source, heavy
unknown Unfiltered surface
— Filtration with
G.C.SS Filter
by heavy runoff
G Filter
heavy runoff,
G,C No filtration, heavy
    1. For persons drinking more than the median amount of water versus those drinking less than the median
    2. C - cramps,   D = diarrhea,   V = vomiting
    3. SS = Salmonella/Shigella.   C = campylobacter,   G = Giardia lamblia,   EEC = Enteropalhogenic E. coli
 filtration  without  any  chemical  pretreatment  (two  out-
 breaks). In the four suspected or confirmed viral outbreaks,
 effective  filtration was  also absent, but  the precipitating
 event was a Chlorinator failure (two outbreaks), bypass of a
 Chlorinator (one outbreak), or very high raw water turbidity
 leading to low free chlorine residuals (one outbreak). Out-
 breaks  #16,  17, and 18  were associated with very high raw
and  finished  water turbidity during  spring  runoff,  when
filters were overwhelmed or bypassed.
    Specific precipitating events were found for only two of
the suspected or confirmed giardiasis  outbreaks: backup of
beaver  pond water into a spring  house in #9 decreased
chlorine contact time  compared to the normal condition in
                                                                                              AJPH March 1985, Vol. 75, No. 3

                                                                       WATERBORNE DISEASE OUTBREAKS IN COLORADO

    Some ol'lhe lessons we have learned from Ihis surveil-
lance experience will be summarized briefly. Daily monitor-
ing and follow-up of colilbrm tests results failed to uncover
any gastroenteritis outbreaks in the communities involved.
Increased  communication  with county  health personnel,
sanitarians, public health nurses, and physicians apparently
paid oil', although we were still hampered by delay between
the occurrence  and reporting  of illness.  The presence of a
single person who reviewed all complaints and all communi-
cable disease reports and initiated investigations was impor-
    Resort areas may pose  special problems  in identifying
outbreaks.  Unsolicited  reports from  out-of-state  visitors
alerted  us  to four of the waterborne outbreaks eventually
attributed to Giardia. In the Winter Park outbreak (#11),
local  physicians had seen no increase ia  diagnosed cases of
giardiasis in residents,  but  letters  from visiting skiers con-
vinced us that  a problem existed.  A survey of out-of-state
visitors to the area showed an attack rate  of 15.1 per cent for
visitors drinking from (he town's main water system during
the first two weeks of February. Most  residents  are not
served by the town water but most  visitors stay in hotels and
condominium units served  by  that water. Long-term resi-
dents may  have lower attack rates  than residents even when
comparably exposed.4
    The decision  to classify  outbreaks in  this study  as
waterborne was based on demonstrations of a dose-response
relationship between water consumption and illness. As long
as over- and underestimation of water  consumption are not
associated with illness status, the magnitude of any observed
association between illness and amount of water consumed
should be an underestimate.
    Ill persons may selectively recall higher water consump-
tion than well persons.  Interviewer bias  is also possible. In
the Eagle-Vail outbreak (#4), however, there was no associ-
ation between reported water consumption and gastrointesti-
nal illness during the two weeks prior to  the outbreak, even
though  the respondents clearly thought these illnesses were
.also waterborne.2 In Aspen Highlands (#8), a relationship
was seen'between illness and amount of water consumption
in short-term but not long-term residents."
    Analysis of the data from these two outbreaks using the
method described by Marshall  et a IS shows that 30 to 40 per
cent  of  ill  persons  would have had to  overestimate  their
water consumption by two or more glasses per day (while no
well  persons  overestimated  their water consumption)  in
order to show a dose-response relationship significant at the
.05 level when none really existed.  Differential  misclassifica-
tion would have to be considerably more common  than this
to cause a spurious relationship at the levels seen in most of
the outbreaks reported  here.
    The attribution of outbreaks to particular agents was the
most  difficult  and  frustrating  part of these investigations.
Only  in one instance. Aspen Highlands,4 were these major
criteria  for agent clearly met:
    • The  agent should be found in ill  persons significantly
more often than in well persons from the affected community
or from an unaffected control community;
    • The  agent should be  found  in the epidemiologically
incriminated source, the water; or there should be an appro-
priately timed event that would have allowed  the presumed
.agent to pass through the treatment system.
    A positive water sample from a sampling filter is not  by
itself sufficient  evidence  that  a  waterborne outbreak  is
caused  by Giardia  lamhliu.  We  have  been  able  to  find
morphologically intact  organisms  in  finished water from
communities with  no current outbreak.*** Nonetheless we
have classified two outbreaks (#7, 9) as presumed giardiasis
when there was an appropriate clinical syndrome for giardia-
sis, evidence of walerborne spread of disease, and evidence
by filtration sampling for Giardia la/nblia in the water, even
in the absence of positive stools in ill persons.
    The patterns  seen  in  this waterborne disease surveil-
lance effort are consistent  with  the efficacy of the multiple
barrier concept in preventing watciborne disease. None of
the 17 surface-water outbreaks  occurred on the  many  sys-
tems  which have  coagulation, flocculation,  sedimentation,
filtration, and disinfection.
    Bacteriological standard violations were associated with
only one of the Giardia outbreaks  (#14). Giardia lamblia is
relatively  resistant to chlorine disinfection. The large num-
ber of coliform organisms relative to Giardia lamblia in
wastes has not allowed  the presence of coliforms to predict
the presence of Giardia.
    None ol'lhe Ginrd'ui outbreaks had turbidity violations.
Turbidity is a poor measure of the microbiological  quality of
water in areas (such as  Colorado)  where raw water turbidi-
ties are low. Systems with  raw  water turbidities below the
1.0 NTU standard  often discontinue chemical addition ahead
of filtration since  it  is  not needed  to  meet  the  standard.
Filtration without  chemical pretreatment has been shown to
be ineffective in removing  Giardia-tized particles.h
    Although beaver and other  animals have been implicat-
ed  as Giardia  carriers, ubiquitous  human and  domestic
animal activity  has led us to the conclusion that all surface-
water supplies are at risk  for Giardia lamblia.'Since 1977.
Colorado  has required that all surface water be completely
treated, including  coagulation, flocculation,  sedimentation,
filtration,  and chlorination steps. This  requirement is  sup-
ported by  the absence of reported outbreaks of all types
(Giardia and viral) where full treatment is employed. Such
systems are expensive  for small  communities.  Effective
inexpensive surface water treatment methods could substan-
tially reduce the risk of waterborne disease.
    "*L.apham SC. Hopkins RS. White M, Blair JR. Smadcs RH, Bissel RC.
Simpson (il..1 Giardiasis in travelers 10 Colorado. A prospective .study of 484
visitors. (Manuscript submitted Tor publication.)
 I Water-related Disease Outbreaks. 1982 Surveillance Summary. Atlanta:
  Centers for Disease Control, August 1983.
 2. Hopkins RS, Uaspard GB, Williams KP. Karlin RJ. Cukor G, Bladlow NR:
  A community walerborne gastroenteritis outbreak: evidence for rotavirus
  as the agent. Am J Public Health 1984: 74:263-26?.
 3. Rothman K I,  Boicc  JD' Hpidemiologic  Analysis with a Programmable
  Calculator. NIH Pub. No. 79-1649:36-37. Washington,  DC: Govt Printing
  Ofliec, 1979.
 4. Istre GR. Dunlop TS. Gaspare! GB. Hopkins RS: Waterborne giardiasis at
  a mountain resort: evidence tor acquired immunity. Am J Public Health
  1984: 74:602-604.
 5. Marshall JR.  Priore R. Graham S.  Brasure J: On the distortion of risk
  estimates in multiple exposure level case-control studies. Am J Epidemiol
  1981: I 1.3:404-473
 6. I.ogsdon GS. L.tppy t:C: 1 he role of filtration in preventing  waterborne
  disease. J Am Water Works Assoc 1982;  74:649-655.
   This  rescaich  snppi
Contract #68-0.3-2927.
•ted in pail by Hnviiomncnlal Protection Agency
AJPH March 1985, Vol. 75, No. 3



Reproduced,  with  permissions
Journal  of  Public Health
       A Three-state Study  of  Waterborne Disease Surveillance Techniques
      Abstract: For a two-year period, the states of Colorado, Ver-
  mont and Washington tested the effectiveness of ten surveillance
  methods for identifying waterborne disease. Nine were active sur-
  veillance methods, soliciting illness reports; one was passive, rely-
  ing on voluntary disease reporting. One waterborne disease out-
  break was identified through use of the nine active methods, while
  14 were reported through  the  passive  surveillance method. The
  presence of coliform bacteria during routine water testing was not
  related to illness in the community. (Am J Public Health  1985;

      Since the early 1960s, there has been a steady increase
  in the reported number of waterborne disease outbreaks in
  the  United States.' An average of 41  outbreaks were re-
  ported annually, 1979-83. The increased number of reported
  outbreaks may reflect  more complete reporting rather than
  an actual increased risk.
      To examine the methods for improving surveillance of
  waterborne  disease,  the  US  Environmental Protection
  Agency (EPA) funded  three two-year projects by the State
  Health Departments  of Colorado (1980-82),  Vermont
  (1980-82), and Washington (1981-83). Although each project
  differed somewhat in  approach, all were  designed to test
  alternative surveillance techniques and to improve reporting
  of what would have been previously unreported waterborne
  disease. Detailed individual reports are available from each


      Before the implementation of the  special surveillance
  projects,  Colorado,  Vermont,  and  Washington detected
  enteric diseases and disease outbreaks through passive sur-
  veillance systems, based upon voluntary reporting with little
  or no  effort  devoted  to soliciting specific illness  reports
  related to  contaminated drinking water.  Disease  reports
 from physicians, citizens, and health officials were investi-
 gated when there was evidence  of an outbreak. Reportable
 diseases which are caused by potentially waterborne patho-
 gens were reviewed and investigated when drinking water
 appeared to be a possible source of infection. Salmonellosis
 and shigellosis are reportable in Colorado, Vermont,  and
 Washington;  as of  1984,  Colorado  and  Vermont report
 giardiasis and campylobacteriosis.
    During the project period,  each of  the  three  states
 expanded waterborne disease surveillance activities. A full-
 time  project epidemiologist was employed in each state;
    Address reprint requests to Lucy Harter. Division of Health, Washington
Department of Social and Health Services. Mail Stop ET-I3, Olympia. WA
98504. Dr. Frost is affiliated with the same Department; Dr. Vogt and Ms.
Little are with the Vermont Department of Health,  Epidemiology Office,
Burlington; Dr. Hopkins and Mr. Gaspard are with the Colorado Department
of Health. Epidemiology and Disease Control Division. Denver; Mr. Lippy is
with the US  Environmental Protection Agency, Health  Effects Research
Laboratory, Cincinnati. This paper, submitted to the Journal December 21.
1984. was revised and accepted for publication May 13. 1985.

                  TABLE 1—Active Waterborne Disease Surveillance Methods In Colorado, Vermont, and Washington
Daily Review of Mandatory Collform Tests

Weekly Telephone Surveillance of Schools, Industries, Camps
One-Time Customer Surveys of 12 Water Systems with Water
Quality Problems
Twice Weekly Additional Coliform Tests of Selected Water Systems
Pathogen Testing of 60 Water Systems with Water Quality
Surveillance of Laboratory Confirmed Giardiasis Cases

Surveillance of Laboratory Confirmed Yersiniosis and
Campylobacteriosis Cases
Surveillance of Private Laboratories for Increased Enteric Work
Surveillance of Camps, Parks, Ski Programs, Recreational
Passive Surveillance Enhanced with Staff Epidemiologist and Public

Time Period (State)
6/81-5/82 (CO)
9/81-12/81 (VT)
12/80-9/81 (VT)

4/81-3/83 (WA)
2/82-1 2/82 (VT)

4/81-3/83 (WA)
4/81-3/83 (WA)
6/81-5/82 (CO)

4/81-3/83 (WA)

4/81-3/83 (WA)

4/81-3/82 (WA)
6/80-5/82 (CO)
12/80-12/82 (VT)
4/81-3/83 (WA)
Total Months
of Observation







 toring of laboratory confirmed  infections,  the  passive  sur-
 veillance system provides the most timely reports. There is
 an inherent delay in laboratory based surveillance. Stool
 sample collection, sample delivery to the laboratory,  and
 specimen  analysis  take one  to three weeks, whereas, a
 physician can become aware of illness in a community and
 report it within days of the onset.
     Under the Safe Drinking Water Act, coliform bacteria is
 the  criterion used  to judge the microbiological safety of
 drinking water. In this study, routine and  intensive surveil-
 lance  of water systems with  coliform violations  did  not
 reveal illness  among water users. Coliform  bacteria were
 elevated in only  10  of the  15  outbreaks  which occurred
 during the project period. The presence of coliforms failed to
 predict illness in the community. The absence  of coliforms
 did not assure safe drinking water. Although coliform bac-
 teria can be indicative of water contamination, it  appears
 less reliable as an indicator of health risk.
     Except in Colorado, the surveillance activities that were
undertaken in this study did not result  in  more  reported
outbreaks  of waterborne disease. The  methods  used in
Vermont and Washington were either too  insensitive to pick
up additional  outbreaks or the current passive surveillance
systems employed by the states are adequate. The special
surveillance methods described in this paper may  be more
useful for  other local  and state  public  health agencies if
current surveillance is inadequate.

1. CDC:  Water related disease outbreaks in the US 1980. MMWR 1982;
2. Hopkins RS, Shillam P, Gaspard  B, Eisnach L, Karlin RJ: Waterborne
  disease in Colorado: three years surveillance and 18 waterborne outbreaks.
  Am J Public Health 1985; 75:254-257.
3. Vogt RL, Little AA: Vermont Waterborne Disease Project. Final Report
  on EPA contact #68-03-2967. Vermont Department of Health, Burlington,
  VT 05402.
4. Harter LC. Frost FJ, Perkins-Jones K, Holman R: Washington State
  Waterborne Disease Surveillance Study. Final Report on EPA Contract
  #68-03-3059. Washington Department  of Social  and Health Services.
  Olympia, WA 98504.
                        AJPH November 1985, Vol. 75, No.11

                                             Reporduced,  with  permission,  from Journal
                                             of Environmental  Health.
 Routine  Coliform  Monitoring and

 Waterborne Disease  Outbreaks
                                                Odette Batik
                                                Gunther F. Craun
                                                Wesley O. Pipes

      To provide protection against the transmission of waterborne disease
    all public drinking water systems are required to monitor for coliform
    bacteria at a prescribed frequency. However, waterborne outbreaks have
    previously been documented in public water systems which have not ex-
    ceeded these coliform regulations. A comparison of coliform monitoring
    results in a sample of both community and non-community water
    systems showed no statistical differences in coliform results between
    systems in which an outbreak had occurred and systems in which no out-
    breaks had been reported. Although the results of this study must be
    interpreted cautiously, the authors feel that public health officials should
    reevaluate  the  importance of routine  coliform  monitoring in their
    surveillance program for the prevention of waterborne disease.
   All public water systems, which in-
 clude both community systems  ser-
 ving at least 25 year-round residents
 or  15 service connections and non-
 community systems serving transients
 or temporary residents, are required
 by  the U.S. Environmental Protec-
 tion Agency (EPA) to routinely moni-
 tor  for coliform bacteria(6). The fre-
 quency of  coliform monitoring in-
 creases as the size of the population
 served increases, with systems of 1000
 persons  or  less required  to  collect
 only one sample per month. It is not
 until a system serves a population of
 25,001 that 30 samples per month are
 required.  There is   no  requirement
 that samples be collected daily, and
 no rationale is provided in the EPA
 regulations to support the frequency
 of monitoring.
  Journal of Environmental Health, V. 45 (5)

    Odette  Batik, Columbia  University,
  School of Medicine, New York, NY 10027;
  Gunther F. Craun, U.S.  Environmental
  Protection  Agency, Cincinnati, Ohio
  45268; and Wesley O. Pipes, Department of
  Biological  Sciences,  Drexel  University,
  Philadelphia, Penn. 19104.

March/April, 1983
   The coliform group has been used
 as an indicator of pollution for some
 70 years because they are present in
 the normal intestinal flora of humans
 and warm-blooded  animals and  are
 eliminated in large numbers in fecal
 wastes. The absence of coliform  or-
 ganisms in a public water supply is
 generally interpreted as evidence of a
 safe drinking water free of  water-
 borne  pathogens(6). Implicitly, rou-
 tine coliform monitoring  of drinking
 water is to provide protection against
 transmission of waterborne disease.

   Outbreaks  of waterborne disease
 have occurred in water systems where
 the  maximum  contaminant  level
 (MCL)  for coliforms  has not  ex-
 ceeded EPA regulations(3,4,5). Case
 studies  have suggested that the cur-
 rent EPA  regulations for  coliform
 monitoring frequency, especially  for
 small systems, are inadequate to pro-
 tect  against transmission  of water-
 borne disease and that it  is inappro-
 priate to use coliforms for  monitoring
 surface water systems where disinfec-
 tion  is  the  only treatment(3). Anec-
 dotal information indicates it is also
possible to  detect coliforms in water
systems  that have not reported out-
   Journal of Environmental Health
   A study was conducted to compare
 coliform  monitoring results  between
 water systems experiencing a water-
 borne outbreak and systems in which
 an outbreak had not been reported
 and to determine if reported water-
 borne outbreaks could have been pre-
 dicted on the basis of violations of the
 coliform  MCL.
   The study consisted of two separate
 but  related investigations,  one of
 community water systems and one of
 non-community water  systems. In
 both cases only the data from existing
 reports of outbreaks and records of
 coliform  monitoring results were uti-

        Community Systems

   Waterborne-disease   outbreaks
 were reported in 47 community water
 supplies in  the United States during
 the period between 1971-1977. Water
 supply officials from states in which
 these outbreaks  occurred were con-
 tacted, and coliform records for three
 years (the year preceding, following,
 and during the time  the outbreak oc-
 curred) were requested. Records for
 an  identical  period  were also re-
 quested for comparison communities
in which an outbreak had not been

   Two groups of comparison com-
munities  were  selected.  The  first
group was selected with the aid of the
state water supply officials: a com-
munity of about the same size, with
the  same type of water source, and
that employed the same type of water
treatment procedures as the outbreak
community. To minimize differences
in outbreak reporting and water qual-
ity surveillance practices (e.g. number
of coliform   samples collected)  be-
tween the comparison and outbreak
community, the comparison commu-
nity was selected from the same gen-
eral geographic area (health  district)
as  the  outbreak community.  The
second group  of comparison commu-
nities was selected at random from an
inventory of  water  supplies  in each
county in which an outbreak occurred
after excluding systems  not  within
15% of the same size as the outbreak
system.  No statistical differences in
demographic   characteristics  were
found between the  two groups  of
comparison communities.

   These  specific   questions  were
examined: (a) Were positive coliform
results more likely to occur in the out-
break community  during the three
months  prior to the outbreak?  (b)
Over the total three-year period, was
the coliform standard exceeded more
often for those communities  experi-
encing outbreaks? (c) When  positive
results were obtained over the three-
year period, were there more coliform
organisms per sample in the outbreak

   Unfortunately, much  of the data
requested was not made available:  no
monitoring results existed for 10 wa-
ter systems in  communities experienc-
ing  an outbreak; four systems had
been monitored, but  the records had
been discarded; for  16  systems  the
records  requested  were  not made
available by the state health officials.
Records were provided for 17 com-
munity water systems experiencing  an
outbreak but  not all could be used
either because the recording methods
prevented  determination  of  all  re-
quired variables or because the rec-
ords for the comparison community
had  been discarded. Only 32% of the
originally selected data set remained
for analysis.  Thirteen pairs of com-
munities selected with the aid of state
officials  and  10 pairs from the ran-
dom selection  procedure had suffi-
cient information  for comparison.

Regardless of the manner in which
the comparison community  was se-
lected,  there  were  eight  instances
where the community experiencing an
outbreak and its comparison commu-
nity  had  similar coliform  results in
the three months prior  to the  out-
break: in four instances both commu-
nities had positive coliform results,
and in four instances both  had nega-
tive coliform results.
  Because of the small number of
pairs  of communities with different
coliform results, Cox's(2)  exact test
for discordant  pairs  was  used to
evaluate the first question  (Table 1).
The second and third questions were
evaluated by  paired comparison  pro-
cedures using Students' t test (Table
2).  The outbreak  and  comparison
communities  were  not  found  to
differ in regard to positive coliform
results within the three months prior
                                  to the outbreak (Table 1). It is reasur-
                                  ing,  however, that  when different
                                  coliform  results  were  obtained be-
                                  tween the community experiencing an
                                  outbreak and its comparison commu-
                                  nity, the trend was toward detection
                                  of coliforms in the outbreak commu-
                                  nity arid negative coliform tests for
                                  the  comparison  community during
                                  the three prior months. With respect
                                  to the number of times the coliform
                                  standard was exceeded per number of
                                  samples collected and mean number
                                  of  coliform organisms in the water
                                  system during the entire  three year
                                  period, no  statistical difference was
                                  observed between outbreak and com-
                                  parison communities  (Table 2).

                                       Non-Community Systems
                                   In this study, coliform  monitoring
                                  results for non-community water sys-
                                  tems  experiencing an outbreak were
                                 Table 1
                         Community Water Systems
      Comparison of Coliform Monitoring Results Three Months Prior to Outbreak
A.  Comparison Community Selected by State
    (matched on size, water source, water
    treatment, location)
                                              Comparison Community
          Outbreak Community
                           Positive Test
                           Negative Test

                           Total Pairs
                           P(1) = 0.188
Comparison Community Selected at Random
(matched on size, location)
Positive Negative
 Test     Test
   4      4
   1      4

Outbreak Community
Positive Test
Negative Test
Total Pairs
P(1) = 0.250
                                 Table 2
                         Community Water Systems
          Comparison of Coliform Monitoring Results During 3-Year Period

                                                   Mean Coliform Levels
                                                       Multiplied by
                             Number of Times          Number of Times
                         Standard was Exceeded per  Standard was Exceeded per
                         Number of Samples During  Number of Samples During
                      •'         3-year Period              3-year Period
Outbreak Communities vs.
Comparison Communities
Selected by States
(n = 1 5 pairs)
t = 0.27
P(1) = 0.395

t = 1.07
P(1) = 0.151

    Outbreak Communities vs.
     Comparison Communities
         Selected at Random
             (n = 1 5 pairs)
                            t = 1.72
                          P(1) = 0.054
           t =  1.26
         P(1) =  0.1 14
    Journal of Environmental Health
                                                         Vol. 45, No. 5

  compared to non-community systems
  not reporting an outbreak in Pennsyl-
  vania during 1973-1978. Pennsylvan-
  ia was chosen because it reported the
  largest number of outbreaks during
  this period, reflecting a dedication to
  identifying and investigating water-
  borne disease outbreaks. Thirty-nine
  waterborne disease outbreaks in non-
  community  systems were reported
  from Pennsylvania in  this  six year
    Data  on coliform monitoring  of
  non-community water  systems  were
  obtained from county offices. During
  1973-1978,  non-community  water
  systems   in  Pennsylvania were in-
  spected at least once a year by a sani-
  tarian and were sampled for coliform
  analysis if they were considered un-
  satisfactory with regard to protection
  and construction or if the disinfection
  device was not  functioning properly.
  Data were obtained on 34 outbreaks
  occurring  in  18 different counties.
  There were five instances  where two
  outbreaks of waterborne disease were
  reported for the same system. The 29
  non-community  systems  with  re-
  ported outbreaks included 12 restau-
  rants, 12 campgrounds, two recrea-
  tional areas, one church, one apart-
  ment complex,  and one  fire house.
  However,  no  coliform monitoring
  data  were available  (apparently no
  samples had been taken) for 13 of the
  29 establishments.  This  eliminated
  the church, the apartment complex,
  the fire house,  the two  recreational
  areas,  two campgrounds,  and six
 restaurants from the analysis.  Coli-
 form monitoring data were also ob-
 tained for the remaining  798  non-
 community water systems  for which
 no waterborne disease outbreaks had
 been reported in these  18 counties.
These  included  715 restaurants, 81
campgrounds, and  two recreational
  The Chi-square test for comparison
of two proportions in an independent
sample was used to  determine if coli-
forms had been found in a larger pro-
portion of the establishments with re-
ported outbreaks than of the  estab-
lishments  with   no  reported   out-
breaks. Discriminant  analysis  was
also employed in an attempt to iden-
tify which of several variables were
the best predictors  of outbreaks of
waterborne  disease  in non-commu-
nity systems.
  This analysis also showed it is poss-
ible to find coliforms in non-commu-
nity systems for which there are no re-
ported outbreaks and  to have  out-
breaks in non-community systems for
which there  are  no  positive coliform
monitoring results.  No difference in
positive coliform results was  found
between non-community systems ex-
periencing an outbreak and those not
experiencing an outbreak (Table 3).
  For  the discriminant analysis,  the
dependent  variable  considered  was
waterborne outbreaks (yes or no) and
the independent  variables were estab-
lishment  type,  type of water  treat-
ment, total number of coliform sam-
ples collected over the five year per-
iod, and the  average coliform density.
The only variable showing a statisti-
cal  association was  the total number
of coliform samples collected over the
five year period, inferring that non-
community  water supplies  with out-
breaks are sampled  more frequently
than those  without outbreaks. The
frequency of sampling is controlled
by a sanitarian who has responsibility
for inspecting   and sampling  non-
                                  Table 3
                        Non-Community Water Systems
       Comparison of Coliform Monitoring Results Three Months Prior to Outbreak

                                                     Coliform Results
                                                  Positive  Negative
                                                   Result   Result    Total
Non-Community System
Experiencing an Outbreak
Non-Community System



             P(1) = 0.288
March/April, 1983
                                         Journal of Environmental Health
  community water  supplies. Appar-
  ently, the sanitarians are able to iden-
  tify by sanitary surveys those water
  supplies which are most likely to have
  outbreaks and concentrate the sam-
  pling where  they  think a problem
  exists. However, the sampling of sys-
  tems with identified problems did not
  always  cause an  outbreak  to  be
  avoided. These data provide no evi-
  dence that coliform  monitoring re-
  sults provide  any  prediction of out-
  breaks of waterborne disease in non-
  community systems.

    These studies suggest that the data
  obtained  from  routine  coliform
  monitoring required by EPA regula-
  tions may be of little help  to public
  health  officials  in  preventing  the
  transmission of waterborne  disease in
  public water systems; however, these
  results must be cautiously interpreted
  for  several  reasons.  The  study of
  community water  systems  included
  only 32% of the communities which
  experienced an  outbreak during the
  study period. This provided a rather
  small sample which may not be repre-
  sentative  of  communities experi-
  encing  outbreaks.  There is also in-
  complete  reporting  of  waterborne
  outbreaks and additional outbreaks
  which were not investigated and re-
  ported  might have occurred during
  this period. This would result in unre-
  ported  outbreaks which were not in-
  cluded  in the  study,  and it is  also
  possible  that unreported outbreaks
  occurred in one or more of  the com-
  parison  communities. It appears un-
  likely that  comparison communities
  used  in  this  study  experienced  a
  waterborne outbreak that was neither
  investigated nor reported during the
  time selected, as outbreak and com-
  parison  communities were  matched
  on both  geographic  location  (same
  local health district) and time to mini-
  mize possible differential disease  sur-
  veillance. The water quality  data for
  both the community and non-com-
 munity studies were collected from a
 period  before the EPA regulations
 were applied and may represent in-
 complete  reporting and  inadequate
 monitoring conditions which  no long-
 er exist.  It is not known to what ex-
 tent these conditions could influence
 results, but it is felt that differential
monitoring or reporting did  not  oc
cur. These data also represent results

from a number of different  labora-
tories, but this should not have an in-
fluence on interpretation of  results,
as laboratory procedures are gener-
ally standardized for coliform moni-

  The fact that no differences in coli-
form  monitoring  results  could  be
found between the systems with and
without  reported outbreaks  in this
study might  be explained  by several
other factors, which were observed
during analysis.  Of  the community
systems  in the  study,  the  median
population was 1,950 and half of the
communities  in  the  study were  re-
quired to  sample only two  times a
month or less. Most  non-community
systems  were sampled only  once a
year.  In instances  involving sharp
upward  fluctuations in contamina-
tion immediately prior to the  time of
outbreak,  routine  sampling could
miss the period of contamination by
several days  to several months. This
situation has previously been docu-
mented  in several waterborne out-
breaks(3).  Almost half  of the out-
breaks in  community systems  were
caused by Giardia lamblia. Data from
waterborne outbreaks  of giardiasis
have shown that Giardia cysts can be
recovered from drinking water meet-
ing the coliform  standard, especially
for systems providing chlorination as
the only treatment, and that water-
borne giardiasis outbreaks can occur
in water  systems meeting the coliform

  There are several possible explana-
tions why the use of routine coliform
monitoring may be ineffective in pre-
venting  transmission of  waterborne
diseases. It is possible that the coli-
form test may neither be sensitive  nor
specific enough to be useful in  alert-
ing officials  of the necessity  to take
action to prevent the occurrence of an
outbreak. In addition, routine moni-
toring may not include enough sam-
ples to provide a good probability of
detecting the presence of coliforms or
an  outbreak occurs so quickly after
contamination that it is highly  un-
likely  that coliforms would  be  de-
tected before the outbreak by means
other  than continuous  monitoring.
The elapsed time between sample col-
lection and transmittal of results to
operating  personnel under  current
regulations, even if contamination is
detected, may also be too lengthy to
permit corrective action to be taken in

a  timely  manner.  A  related  factor
concerns the type of corrective action
generally  stimulated  by a  positive
coliform result. Too often the initial
reaction to a positive result is the col-
lection of a check sample, as it is as-
sumed the result is due to a sampling
or  laboratory  error.  Many  times
check samples continue  to  be taken
until a negative coliform result is ob-
tained,  and no positive action is ever
taken to determine how the contami-
nation  occurred or if there  was a
treatment or distribution deficiency.
  Although the results of this study
must be interpreted  cautiously and it
is recommended that additional stud-
ies  of  this type be  conducted, it is,
nevertheless,   felt   that  regulatory
agencies should carefully examine the
relative importance  they have   as-
signed to routine coliform monitoring
of tap water in their surveillance pro-
gram.  It should be  realized that the
primary reason for the establishment
of such surveillance programs is the
prevention of  waterborne disease  and
other kinds of surveillance  activities
may be more  effective and cost effi-
cient in this regard.
  The  American Water Works Asso-
ciation  Committee on the Status of
Waterborne Disease  in  the  United
States and Canada has just released a
report in which they recommend that
surveillance activities be structured to
help prevent  waterborne disease(l).
Their conclusions on  the usefulness
of  the  coliform test  in preventing
waterborne disease are of interest:

     "Routine  coliform  surveil-
   lance  of  treated  water  is
   mandated by the Safe Drinking
   Water   Act;   however,  the
   committee  feels  that   its
   importance  is  to  provide  a
   historical record  of operation
   rather  than to  prevent water-
   borne outbreaks."

     "The committee  feels that it
   is more important to know  the
   quality of the source water and
   potential sources of contamina-
   tion  so  that source protection
   and  treatment can be provided.
   A water supply surveillance pro-
   gram  should emphasize  fre-
   quent  engineering  evaluation
   and  sanitary surveys to identify
   and  correct potential deficien-
   cies.  Microbiological  resources
   are better applied to  assessing
    Journal of  Environmental Health
  raw  water quality,  identifying
  sources of contamination, and
  evaluating  the   efficiency   of
  treatment than the routine sur-
  veillance of water quality in the
  distribution system. In some in-
  stances,   alternative  microbi-
  ological or chemical indicators,
  such as chlorine residual analy-
  sis, may be more useful  in pre-
  venting outbreaks than coliform
  surveillance of the distribution


1. Committee Report (1981) Waterborne Di-
  sease in the United States and Canada, J.
  Amer. Water Works Assoc. 73:528-529.
2. Cox, D.R. (1978) Analysis of Binary Data,
  Methuen and Co. LTD, London, p. 56-58.
3. Craun, G.F. (1978) Impact of the coliform
  standard on the transmission of disease,
  Evaluation of the Microbiology Standards
  for Drinking Water,  C.W.  Hendricks,
  Ed., USEPA 570/9-78-006, USGPO,
  Washington, D.C.
4. Craun, G.F. (1979) Waterborne giardiasis
  in the United States: a review, Amer. J.
  Publ. Health 69:817.
5. Craun,  G.F.  (1979) Waterborne  disease
  outbreaks in the United States, J. Environ.
  Health 41:259-265.
6. U.S. Environmental  Protection Agency
  (1976) National Interim Primary Drinking
  Water Regulations, USEPA 570/9-76-003,
  USGPO, Washington, D.C.
                       Vol. 45, No. 5



Reproduced,  with  permission,  American
Journal  of  Public Health
                      A  Follow-up Study of Gastro-lntestinal Diseases
                Related to Bacteriologically Substandard  Drinking Water
     Abstract:  In a prospective follow-up study conducted in 52
  French alpine  villages, one weekly water sample was taken in each
  village provided with untreated ground water and analyzed as to the
  presence of four indicator bacteria: total plate count, total coliforms,
  thermotolerant (fecal) coliforms, and fecal streptococci. Cases of
  acute gastro-intestinal disease (AGIO) occurring among 29,272
  inhabitants were reported through  physicians, pharmacists,  and
                    primary school teachers. A loglinear model identified fecal strepto-
                    coccus (FS) as the best predictor; the presence of fecal coliforms
                    enhanced the effect of FS. The total bacteria count and the total
                    coliforms had no independent contributions.  A  threshold analysis
                    suggested that any level of indicator bacteria above  zero was
                    associated with an excess of AGID. (Am J Public Health 1987;
      Microbiology standards for drinking water are periodi-
  cally evaluated.' The indicator bacteria used in the European
  community to assess whether a 100 ml  water  sample is
  potable are fecal coliforms (FC) and fecal streptococci (FS).2
  Total coliforms are also used as indicator bacteria in many
  countries,  such as the United States.3 The dramatic decline
  of "classical" waterborne diseases  in developed countries
  has  lessened  the  concern  of local  authorities with these
  issues.  The frequency  of substandard water samples is
  relatively common in the alpine regions, but water-related
  disease outbreaks are rather rare.4 Alpine health authorities
  are highly  concerned, however, about the increasing costs
  and constraints of the water policies.
     This study had the  following objectives: to assess the
  risks related to the consumption of drinking water that does
  not meet current  bacteriology standards; to identify  the
  indicator bacteria  that best  predict  this risk; to  determine
  whether there  is an indicator bacteria threshold other than
  zero above which water should be declared "non-potable".


     The study design is presented in detail elsewhere.5 Two
 parallel  surveys were conducted: 1)  a weekly water sample
 was taken in each of 52 villages and  analyzed the same day
 in a single central laboratory; and 2) a day-to-day count of the
 acute gastro-intestinal diseases occurring among inhabitants
 of the study villages and identified through 119 physicians, 52
 pharmacists, and 118 primary school teachers (for children
 aged 7 to 10 years). Physicians and pharmacists were located
 in the villages or within a 10 km surrounding area. None of
 these sources was informed of the  findings  of the weekly
 water samples.
     Villages were chosen to meet several criteria: population
 between 100 and 3,500;  non-tourist  areas (to prevent wide
 variations in village population across seasons); public water
 systems  with untreated ground water; sufficiently far from
 large cities so that the cases of AGID seeking help could be
   Address reprint requests to D. Zmirou, MD, MPH, Centre Alpin de
Recherche Epidemiologique et de Prevention Sanitaire, Centre Hospitaller
Universitaire, Pavilion D. B.P. 217 X, 38043 Grenoble CEDEX, France.
Ferley and Charrel are also affiliated with the Centre; Collin is with the Medical
School,  Nancy, France; Berlin  is with the Department of Biostatistics,
Harvard School of Public Health. This paper, submitted to the Journal
November 5, 1985, was revised and accepted for publication  September 4,
  1987 American Journal of Public Health 0090-0036/87$!.50
                    assumed to be completely enumerated by all the participating
                        The limits of the public water system in each village were
                    known to the survey coordinators. Any case reported for an
                    individual not using the municipal water system was excluded
                    from the analysis. The village population was  restricted
                    accordingly to users of the public water system.
                        Although 64 weekly samples were available for most
                    villages, because  of technical problems,  only 58 samples
                    could be analyzed  for  some villages. Of the 52 villages
                    available for study, two were excluded prior to examination
                    of the  data because  of especially large numbers (>14) of
                    missing observations, and two  villages were excluded be-
                    cause the ground  water they provided was  subjected to
                    treatment half-way through the survey, leaving 48 villages in
                    the study group.
                        The bacteriologic  assays  conducted to ascertain the
                    presence of indicator bacteria were as follows: 1) "standard
                    plate count" (aerobic bacteria): 1 ml water cultivated on
                    standard medium (PCA Institut  Pasteur Production), with a
                    24-hours incubation at 37° ± 1°C; 2) "total coliforms": 100 ml
                    filtered on cellulose  membrane and cultivated on Tergitol
                    TTC during 24 hours at 37°  ± 1°C; 3) "fecal coliforms"
                    (thermotolerant coliforms):  same technique  as  for  total
                    coliforms but incubated at 42° ± 1°C for 24 hours; 4) "fecal
                    streptococci: 100 ml filtered through cellulose membrane and
                    cultivated 48 hours at 37° ± 1°C on  Slanetz and Bartley
                    medium (Institut Pasteur Production).
                        The 48 villages were followed for 64 weeks and contrib-
                    uted 3,072 "village-weeks".  There were 1,807  cases of
                    gastro-intestinal disease reported during the study period.
                    This relatively small number of cases  precluded analyzing
                    each village separately across weeks. We assumed that each
                    village-week constituted an independent  observation. To
                    avoid being  misled by errors  in this assumption (i.e., in  a
                    given village the water quality  was  liable to be correlated
                    across  weeks), we used a variable that indicated whether
                    cases had occurred the previous week in the same village.
                    This variable was included in the  multivariate model to adjust
                    for the experience of the previous week. The unit of obser-
                    vation was  the village-week: the cases occurring during  a
                    given week in a given village were related to the population
                    of the village (person-week) exposed to the water quality
                    indexed by the weekly sample.
                        Because of the sparseness of AGID and of the large
                    denominator (each week, 29,272 person-weeks were accrued
                    in the whole set of villages), the events were modeled using
                    a Poisson distribution.
                                                                                        AJPH May 1987, Vol. 77 No. 5

                                                     GASTROINTESTINAL DISEASE AND SUBSTANDARD DRINKING WATER
TABLE 1—Risk of Acute Gastro-intestinal Diseases According to the
         Fecal Conforms Exposure, in Absence of Fecal Streptococci,
         by Village Size
                                         Relative Risk*
  Village      Fecal           Person-
Population  Coliforms   Cases   weeks
 Point    95% Confidence
Estimate      Interval

237 188
777 252

1 14

[.14; .51]

[.67; 1.92]

   *RR = rate-ratio
    Log-linear analysis can be used with Poisson models and
accommodate categorical covariates6"8 yielding estimates of
relative risk. Further details on methods and data analysis are
discussed elsewhere.* The following covariates were includ-
ed  as potential predictors or confounders of the AGIO
incidence rate: FS count had four levels (0, 1-5, 6-10, 11 and
over, per 100 ml); FC was a binary variable (absent, present);
so were total coliforms, total plate count, and village  size
(under or over 400 inhabitants); and indicator of at least one
case the previous  week.
    The data were restricted to the cases occurring within a
three-day period centered on the water sampling day. This
restriction, which removed 60 per cent of the total cases from
the analysis was implemented to assess and control for a
possible misclassification of exposure. Cases far from the
sampling day, say the day before the next week or just after
the previous week, were liable to result from water pollution
as indexed  by the water sample of the  next week (or the
previous week).
    The confidence  intervals of the relative risks should be
computed using the covariances between various  pairs of
variables. Unfortunately, the Loglin package* used on a Vax
11/780 computer  does  not generate covariance matrices.
Therefore,  the  confidence intervals of  the relative  risks
presented below were approximated using the appropriate
crude (i.e., unadjusted) data tables and the usual formulas for
the variance of rate ratios.9

    The total plate count  and total coliforms had  no inde-
pendent predictive value. When the water met the standards
(no FC  nor FS),  the incidence  rate was 3.44 per 10,000
person-weeks when  at least one germ was observed on the
total plate count; it was 3.43  per 10,000 person-weeks when
no germ was present. The corresponding incidence rates in
the presence of at least  one coliform was 3.74 per 10,000
person-weeks, and 3.32  per 10,000 person-week  when no
coliform was observed.
    On the  other hand, when both  FS and FC were absent,
354 cases occurred, out of a total of 698 (51 per cent). Out of
these, 198 cases (53 per cent) occurred when the total plate
count was positive, the other half of the cases occurring while
no germ was found.
    Table 1 presents the raw data for the effect of FC, in
absence of FS. This effect is somewhat unclear, and  depends
on the village size category: in large  villages, FC are not
                           TABLE 2—Relative Risk (and 95% Cl) due to Increasing Concentration of
                                   Fecal Streptococci, by  Level of Fecal Coliforms and Village
                                                                                       Fecal Streptococci Count/100 ml
Population  Coliforms  0*
                                                 11 +


1 4.25 (2
1 1.40(1
1 2.72 (1
1 .89 (.7-1.2)
4.74 (1.6-9.6)
                              'Level 0 ol FS is the referent category; Relative Risk = 1

                           related to AGIO risk, whereas they seem protective in small
                               The incidence rates when no FS were found served as
                           baseline risk in further analysis.  The absolute risk is rather
                           small;  however, it is underestimated by a factor  of 2.5
                           because only the cases occurring within the three-day period
                           about the sampling day were considered. One can infer from
                           Table 2 that the risk due to a given level of FS contamination
                           is higher when FC are also present; the risk due to a given
                           joint FC and FS exposure is higher in small  villages than in
                           larger villages;  in larger villages, no risk seems associated
                           with FS alone (i.e., when no FC are present), whereas they
                           suffice to predict a significant risk (although  weak) in small
                               These data do not show a clear trend, but it is notable
                           that  the highest concentration of FS corresponds  to the
                           highest relative risk. The Mantel extension for the analysis of
                           trend10 was used on the  crude data (i.e., unadjusted for the
                           previous week experience) and yielded a significant positive
                           trend only in small villages, in presence of FC [ slope = 6.0
                           x 10~5 cases per person-week per unit of FS increase; xi2
                           (slope) = 22.6;  \22 = 4.6, testing departure from linearity].
                           This test  for trend was done  using the program written by
                           Rothman and  Boice for HP.41, applied  to  incidence rate
                               In almost all strata,  the lowest positive level of FS (1 to
                           5) is associated with  a  point  estimate of the relative risk
                           greater than 1. This class may be too large to  assess whether
                           there is a threshold. Another categorization can  be  adopted,
                           where all concentrations greater than 5 bacteria/100 ml were
                           merged (thus any inference from this highly heterogeneous
                           class would be misleading). Table 3 shows how the relative
                           risk  changes with  small  increases  of  FS  concentration,
                           according to whether some cases had been reported the week
                           before or not. The risk is systematically enhanced when cases
                           had occurred previously. As to the threshold issue, no cut-off
                           point other than 0/1 can  be clearly identified.
                           TABLE 3—Relative Risk (and 95% Cl) due to Increasing Concentrations
                                   of  Fecal  Streptococci  with  Special Focus  on Low  Levels
                                   (adjusting for the previous week experience)
                                                   Fecal Streptococci Count/100 ml
                           Cases Week Before   0
                                                                            5 +
   'Berlin J, Zmirou D: Poisson regression for the analysis of relative risk
(Unpublished manuscript), Harvard School of Public Health, 1985.
1 +
1 .79 (.6-1.1)
1 1.47(1.0-2.1)
1.56 (1.0-2.4)
2.45 (1.1-5.5)
2.44 (1.8-3.3)
AJPH May 1987, Vol. 77 No. 5

       The categorization of villages into "small" (<400) and
  "large" (>400) populations split the total set of villages into
  two groups of 24 villages each. Biologic  and socioeconomic
  considerations may  contribute to  the understanding of the
  effect of the village size. In small villages, economic activity
  (agriculture is more  oriented toward cattle breeding) and/or
  general hygiene and wealth may facilitate  the spread  of a
  contamination, and make more dramatic  the effect of a given
  waterborne hazard, through direct interpersonal contamina-
  tion, for  example. This  hypothesis is consistent with the
  multiplicative model of  risk  that  underlies  the  analysis.9
  Although some work has been done on these issues, there is
  still no unequivocal evidence as to the relationship between
  indicator   bacteria and  pathogens, in  drinking12 or  non-
  drinking13'14 water. In  1981, 44 per cent of the water-related
  disease outbreaks reported to the Centers for Disease Control
  (CDC) remained of unknown origin.15
       This  study showed  that the four classes of indicator
  bacteria had very  different predictive value with respect to
  disease. Total bacteria and total coliforms had no predictive
  value; however, half of the cases of gastro-intestinal disease
  occurred  when the water met the standards. Half of these
  were observed while only the total plate count was positive.
       Those who are primarily interested in identifying which
  indicator  germ best  predicts  the risk of disease should be
  concerned with FS and FC (as modifying the effect of FS).
  Geldreich  claims  that  the FC/FS  ratio  provides a useful
  indicator  as to the origin of the fecal pollution, human or
  animal.16 In our study, the average  FC/FS ratio was 4.0 and
  was different in large villages (5.2) and in  small villages (1.3).
  This tends to confirm that the origin of the contamination in
  small villages  was more  related to animals than  in larger
  villages. Those who want to set standards to decide whether
  a water is safe or not might also be interested in the total plate
  count. Finally the "background risk" (when all  indicator
  germs are  absent) amounts to about one-fourth of total cases;
  however,  this figure may well depend on the specific condi-
  tions of this study such as the frequency of the water samples
  taken in the villages (one per week).
      FS was the most predictive indicator and the validity of
  the  standard threshold  (0 bacteria per  100  ml)  cannot be
 disputed with these data. We must recognize that our findings
 depend upon the ecological context and may not hold, for
 instance, in warm  waters or  treated  waters. Surprisingly,
 most of the cases of gastro-intestinal disease were sporadic;
 throughout the 18 months of the survey,  only one outbreak
 was  observed (with about 50 cases), even though 42 per cent
of the total water samples did not  meet  the bacteriology
standards. Hence, this study shed special  light on endemic
water-related hazard, which  may well  be overlooked  by
health authorities. Although modest, this risk persists, and a
lot  remains to be done in  order to  provide  a safe and
wholesome drinking water  to the consumers.
    This study was supported by the French Ministry of Health, the Ministry
of Agriculture, and the Water Agency Rh6ne-M£diterranee-Corse. The Min-
istry of External Affairs also contributed to this work through a scholarship for
one of the authors. We acknowledge the invaluable help of N. Laird and J.
Ware (on the statistical analysis), and N. Gutensohn and G. Hutchison (on the
epidemiological interpretation), from the Harvard School of Public Health. The
support of J. Vial, president of the water  section of the Council of Public
Hygiene of France, is also acknowledged.
 1. Hendricks CW (ed). Evaluation of the Microbiology Standards for Drink-
   ing Water. USEPA 570/9-78-OOC; 1979.
 2. Official Journal of the European Communities—No L229/11—Council
   Directive No 80/778/EEC 15 July 1980 relating to the quality of water
   intended for human consumption.
 3. Federal Register: EPA-Water Programs. National Interim Primary Drink-
   ing Water Regulations, 1982.
 4. Collin JF, Foliguet  JM: A propos de quelques epidemics d'origine
   hydrique survenues en France. Microbiol Alim Nut 1983; 1:27-33.
 5. Ferley JP, Zmirou D, Collin JF, Charrel M: Etude longitudinale des risques
   li£s  a la  consommation d'eaux non conformes aux normes bacten-
   ologiques. Revue Epidem et Sante Publ 1986; 34:89-99.
 6. Bishop VMM, Fienberg SE,  Holland PW: Discrete Multivariate Analysis:
   theory and practice. Cambridge: MIT Press, 1975.
 7. Laird N.Oliver D: Covariance analysis of censored survival data using
   log-linear analysis techniques. J Am Stat Assoc 1981; 76:374, 231-240.
 8. Frome EL, Checkoway H: Use of Poisson regression models in estimating
   incidence rates  and ratios. Am J Epidemiol 1985; 121:309-323.
 9. Kleinbaum DG, Kupper LL, Morgenstern H: Epidemiologic Research.
   Principles and  Quantitative Methods. Lifetime Learning Publications
10. Mantel N; Chi-square tests with one degree of freedom; extensions of the
   Mantel-Haenszel procedure. J Am Stat Assoc 1963; 58:590-700.
11. Rothman KJ, Boice  JD: Epidemiologic  Analysis with a Programmable
   Calculator. Boston: Epidemiology Resources, Inc, 1982.
12. Petersen NJ, Hines  VD: The relation of summertime gastro-intestinal
   illness to the sanitary  quality of the water supplies in six rocky mountain
   communities. Am J Hyg 1960; 71:314-320.
13. Robertson WJ, Tobin AS: The  relationship  between three potential
   pathogens and  pollution indicator organisms in Nova Scotian coastal
   waters. Can J of Microbiol 1983; 29:1261-1269.
14. Cabelli VJ: Health effects criteria for marine recreational waters, USEPA,
15. Harris JR, Cohen ML, Lippy EC: Water-related disease outbreaks in the
   United States in 1981. J Infect Dis 1983;  148:759-762.
16. Geldreich EE: Fecal coliforms and fecal streptococcus density relation-
   ship in waste discharges and receiving wastes. Crit Rev Environ Control
   1976; 6:349-369.
                                                                                                  AJPH May 1987, Vol. 77 No. 5

                                                                    Reproduced,  with  permission, American
                                                                    Journal  of  Public Health
                              Risk Factors for Endemic Giardiasis
    Abstract: In a mail survey, 171 Hitchcock Clinic patients with
giardiasis were compared with an age- and  sex-matched control
group of 684 clinic patients with respect to potential risk factors.
Households with shallow well or surface water sources had an odds
ratio (OR) for giardiasis of 2.1 (95% confidence interval (95%CI)
1.3-3.2) compared with  households with drilled well or  municipal
water supply. Other observed risks include family member in day
care program (OR 2.2,  95%CI 1.3-3.7) and family member with
diagnosed giardiasis (OR 17, 95%CI 7.4-37). Previously reported
risks such as travel out of country (OR 3.2, 95%CI 1.5-7.2) and
camping (OR 1.7, 95%CI 0.9-3.2) were also observed. Virtually no
giardiasis risk was observed associated  with report of dog or
barnyard animal proximity. Control for confounding and adjustment
for recall and non-response bias does not materially alter the risk
estimates. We suggest that shallow well or surface household water
source is an important and previously unrecognized giardiasis risk
factor. (Am J Public Health 1987; 77:585-587.)

    Giardiasis is traditionally considered an epidemic dis-
ease, often occurring in waterborne community outbreaks. '~3
Recently, however, increasing  numbers of isolated case
reports in New Hampshire and Vermont4-5 suggest multiple
sources of endemic infection.
    Relatively  little has  been written about endemic or
sporadic giardiasis. Probable risk factors include mountain
camping,6 day care exposure,7 person to person contact,8'9
and travel abroad.10''' We evaluated these and other potential
endemic risk factors such as shallow well household water
sources and animal exposure.
    Shallow or dug well household  water  sources were of
particular interest to us. Often old and deteriorating, these
shallow water sources are frequently open to surface con-
tamination,  and  are common in northern New England  (20
per cent among our controls overall). Several animal species
have been implicated as potential giardiasis reservoirs.12"14
We were interested in the giardia risk conferred by household
animal contact and whether such contact interacted with  the
risk conferred by shallow  household water sources.


    We used the Dartmouth-Hitchcock Medical Center lab-
qptory log  to identify  190 giardiasis cases seen between
January 1, 1977 and June 1, 1984. Nineteen were Dartmouth
College students, a highly mobile population, and were not
considered  in the  analysis. The  endemic nature of  the
remaining 171 cases was  demonstrated by  their  failure to
cluster by date of diagnosis or zip code; contact with the New
Hampshire and Vermont Health Departments confirmed our
impression that  there had been  no epidemic outbreaks of
giardiasis in our region during the study period.
    Control patients were selected from among the 500,000
Hitchcock Clinic patients  in the clinic computer registry (a
cumulative data base maintained since 1969). Four age- and
sex-matched controls for  each case, excluding Dartmouth
College students, were randomly drawn from the  computer
registry. We did not have access to patient medical records,
precluding knoweldge of medical diagnosis or vital status.
   From the Department of Medicine, Dartmouth-Hitchcock Medical Cen-
ter, Hanover, NH 03756. Address reprint requests to Dr. Robert P. Smith at
that address. Dr. Chute is currently with the Department of Epidemiology,
Harvard School of Public Health, Boston. This paper, submitted to the Journal
March 29, 1986, was revised and accepted for publication October 20, 1986.
    We mailed a one-page questionnaire to cases and con-
trols asking about place of residence and work, domestic
water supply, child day care  utilization, animal contacts,
foreign travel,  household exposure to giardiasis, diarrhea
history, and outdoor activities. The questionnaire specified
that answers should refer to the year and month of diagnosis
for the cases, or the year and month of the case for which a
given control was matched. A second mailing was sent to all
nonresponders. Finally, we attempted to interview by tele-
phone all the cases and one-third of the controls who did not
respond to both mailings (see Table 1).
    Standard contingency tables and unconditional  logistic
regression15'16 were used to analyze the data. The matching
variables (age,  sex, year of diagnosis) were included in the
logistic models. Recall bias was minimized by also including
the year of  diagnosis.  Response  bias  was evaluated  by
weighting the telephone subset to simulate the entire group of
non-responding cases and controls in a regression model, and
comparing the weighted and unweighted risk estimates. We
report exposure frequencies, etiologic fractions,17 and crude
and adjusted  odds ratios.15


    Table  1  summarizes the  response patterns for both
mailings and the telephone follow-up. The matching  criteria
are distributed  similarly for  responding  cases and controls
(Table 2), while crude exposure rates differ (Table 3).
    Table 3 outlines crude and adjusted giardiasis odds ratio
(OR) estimates with their 95 per cent confidence intervals
(95%CI)  and  etiologic  fractions. Adjustment for matching
factors does  not materially  alter the point estimates.  We
confirm the influence in our area of previously established
risk factors, although the camping risk estimate is imprecise.
Shallow well or surface water sources,  compared with all
other water sources (drilled  well or municipal) confers the
hypothesized risk for giardiasis (OR 2.1, 95%CI 1.3-3.2).

TABLE 1—Response Rates
© 1987 American Journal of Public Health 0090-0036/87$ 1.50
1st Mailing
2nd Mailing
Phone subset


AJPH May 1987, Vol. 77, No. 5

  TABLE 2—Characteristics of Respondents
Mean Age (years)
Year of Diagnosis
     "Year of Diagnosis" for Controls refers to that of the matched case.
  Compared with municipal water sources, the OR estimate is
  even larger (OR 2.6, 95%CI 1.6-4.3). Adjustment for match-
  ing factors, travel history, age of well, or animal exposure
  does not substantially change the estimate. Animal exposure
  is  not clearly associated  with  disease, with the possible
  exception of cats. Similar results are found when the analysis
  is limited to households with shallow wells.


      This case-control study of non-epidemic giardiasis iden-
  tifies several risk factors of importance in rural northern New
  England. We confirm the importance of some established risk
  factors (foreign travel, day care center exposure, and house-
  hold case contact) and calculate a positive but unstable risk
  estimate for camping (Table 3). We also find a previously
  unrecognized giardiasis  risk for the household use of shallow
  water sources, while domestic and barnyard animal contact
  does not seem to be important.
      Shallow wells are a common household water source in
  northern New England, and their association with giardiasis
  may be an important public health consideration. Inadequate
  construction or maintenance may be responsible for surface
  or septic contamination. Among households with shallow
  wells, those with pets or livestock did not have an increased
 giardiasis risk, perhaps reflecting a lack  of importance of
 domestic animal reservoirs.13'14 We believe our questionnaire
 did not reliably ascertain exposure to wild animals since the
 reported prevalence was unexpectedly low for this region.

 TABLE 3—Risk Measures for Giardiasis
Risk  estimates  for  wild  animal proximity are  extremely
imprecise and are excluded from this report.
    The risk associated with household contact may operate
through contaminated water, but this may simply represent
person to person contact or common>exposure to a different
risk factor. Large urban day care centers are recognized as an
important source of giardiasis7; we observed increased risk
among those exposed to children attending small rural day
care centers.  Camping and  domestic animal  exposure  in
northern New England appear to be less important giardiasis
risk factors than in other parts of the United States.6'12 The
etiologic fraction (14 per cent) for cat exposure incorporates
an unstable odds ratio estimate and should be interpreted
with caution.
    Several potential biases are possible in this retrospective
case-control study.  Recall bias was evaluated using logistic
models that included year of diagnosis as a covariate. This
adjustment did not materially affect the risk estimates. The
Hitchcock Clinic records  from  which the controls were
selected  included patients  seen since 1969, although cases
occurred only since 1977. To evaluate the potential impact of
this difference in time frames, the analysis was repeated using
the subset of controls with patient billing activity since 1980
(a list through  1977  was unavailable). This  subset analysis
generated risk  estimates similar  to  those reported for the
entire sample.
    Response bias was controlled by weighted logistic re-
gression  in which the cases  and controls  interviewed by
telephone were weighted to represent all non-respondents  to
the mailings. Crude and weighted odds ratios appear in Table
4. The point estimates differ substantially only for "travel
abroad"  and "household contact",.but these are based on
small numbers of exposed cases and are therefore unstable.
Overall, we do not think nonresponse greatly distorted our
findings.  Detection bias was possible if physicians tested for
giardiasis more commonly among  returning travelers or
household  contacts of known cases. We  believe this may
have inflated our risk estimates for these exposures.
    Information bias may misclassify water source informa-
tion. We did  not  ask  if  municipal water sources  were
filtered—a water treatment known to reduce water supply
giardiasis risk.  However, in 1981 fewer than 20 per cent of the
community water sources in our region were filtered. Since
we do not know the addresses of controls at the time orcase
diagnosis, we are unable to establish the filtering status of
municipal water sources in-our data. However, failure to
control for this would tend to bias conservatively our esti-
mate of the giardiasis risk associated with shallow wells. Our
information regarding onsite water sources is probably ac-
% Exposure
Risk Factor
Foreign Travel
Day Care
Household Case
Shallow Well
Household Dog
Household Cat
Farm Animals

Crude OR
95% Cl
(1 .3-3.2)
Adjusted OR*
95% Cl
(1 .2-4.0)
   •Adjusted for matching factors: age, sex, year of diagnosis.
                            AJPH May 1987, Vol. 77, No. 5

                                                                                      RISK FACTORS FOR ENDEMIC GIARDIASIS
TABLE 4—Impact of Potential Response Bias
                             Mailing OR
                   Weighted OR
Shallow Well
Day Care
Household Contact
curate:  rural  homeowners are  aware of household water
source, since municipal or drilled water sources can substan-
tially improve the market value of a home.
     It is possible that some of our controls may have had
giardiasis.  Among  controls, 3  per cent indicated  a past
diagnosis of  the condition,  while  7  per  cent reported  a
prolonged episode of diarrhea. Because we could not verify
the giardiasis  status of controls, we did not exclude respon-
dents on the basis of self-reported disease. This misclassifica-
tion would tend to  bias observed  risks  toward the null; we
believe this conservative  approach justifiable in the absence
of systematic disease ascertainment on all controls.
     We found no evidence of important confounding by the
matching variables (age,  sex, and year of diagnosis) using
unconditional logistic regression (Table 3). Adjustment for
camping and foreign travel did not  alter the risk estimates for
shallow well household water source. We cannot exclude the
possibility of residual confounding by unknown factors. In
particular, socioeconomic factors that may influence the type
of water source might also influence other giardiasis risks.
     In summary, our study confirms the previously reported
endemic or sporadic giardiasis  risks of foreign travel, day
care exposure, and household case contacts. We suggest that
shallow well household water sources are an  important and
previously unreported risk for endemic giardiasis in northern
New England. The  18 per cent etiologic fraction attributable
to this exposure suggests that it has a significant  impact on
endemic giardiasis.
    We gratefully acknowledge the Hitchcock Foundation for funding the initial
phase of this project. Dr. Baron was partially supported by the Milbank Memorial
Fund (Milbank Scholar Program). We thank Dr. E.  Robert Greenberg for his
critica! review. This study was initially presented at the Society for Epidemiologic
Research national meeting in Chapel Hill, NC, June 1985.

 1.  Shaw PK, Brodsky RE, Lyman EO, el al: A community wide outbreak of
    Giardisis with evidence of transmission by a municipal water supply. Ann
    Intern Med 1977; 87:426-432.
 2.  Lippy EC: Tracing a giardiasis outbreak at Berlin, New Hampshire. J Am
    Water Works Assoc 1978; 70:512-520.
 3.  Lopez CE. Dykes AC, Juranek DD, Sinclair SP, Conn JM, Christie RW,
    Lippy EC, Schultz MG, Mires  MH: Waterbome Giardiasis: A com-
    munitywide outbreak of disease and a high rate of asymptomatic infection.
    Am J Epidemiol 1980; 112-495-507
 4.  Vermont Department of Health, Epidemiology Division: Vermont Disease
    Control Bulletin 1985; 2.
 5.  New Hampshire Division of Public  Health Services: Giardiasis Study
    Planned. Communicable Disease Rep Lab News 1983.
 6.  Wright RA, Spencer HC, Brodsky RE, Vernon TM: Giardiasis in Colo-
    rado: An epidemiologic study. Am J Epidemiol 1977; 105:330-336.
 7.  Pickering LK, Woodward WE. DuPont HL, Sullivan P: Occurrence of
    Giardia lambliain children in day care centers. J Pediatr 1984; 104:522-526.
 8.  Meyer EA,  Jarroll EL:  Reviews and commentary: Giardiasis. Am J
    Epidemiol 1980; 111:1-12.
 9.  Schmerin NJ, Jones TC, Klein IJ: Giardiasis associated with homosexual-
    lity. Ann Intern Med 1978; 88:801-803.
10.  FiumaraN: Giardiasis in travelers to the Soviet Union. NEngl JMed 1973:
11.  Brodsky RE, Spencer HC, Schultz MG: Giardiasis in American travelers
    to the Soviet Union. J Infect Dis 1974; 130:319-323.
12.  Davies RB, Hibler CF: Animal reservoirs and cross-species transmission
    of Giardia.  In: Jakubowski  W, and Hoff JC:  EPA  Symposium  on
    Waterbome  Transmission of Giardiasis. EPA Pub.  No. 600.9-79-001.
    Cincinnati: EPA, 1979.
13.  Kirkpatrick C, Green G: Susceptibility of domestic cats to infections with
    Giardia  lamlia cysts  from  human  sources.  J Clin Microbiol  1985;
14.  Hewlett E. Andrews J. Ruffier J, et al: Experimental infection of mongrel dogs
    with giarrdia lamblia cysts and trophozoites. J Infect Dis 1982; 145:89-93.
15.  Breslow N, Day N: Statistical Methods in  Cancer Research: Vol 1-The
    Analysis of Case-Control Studies. Lyon: Internationl Agency for Research
    on Cancer, 1980.
16.  SAS  Institute: SAS User's Guide. Cary. NC: SAS Institute. 1982.
17.  Kleinbaum D,  Kupper  L, Morgenstern H: Epidemiologic Research.
    Belmont. CA: Lifetime Learning Publications.  1982.
AJPH May 1987, Vol. 77, No. 5

                                 Reproduced,  with  permission, from  American
                                 Journal  of  Epidemiology
 AMERICAN JOURNAL OF EPIDEMIOLOOY                                                Vo1- 129. N°- 4
 Copyright © 1989 by The Johns Hopkins University School of Hygiene and Public Health            Printed in U.S.A.
 All rights reserved
                            INFECTION IN VERMONT


                     GUTHRIE BIRKHEAD1 -:1 AND RICHARD L. VOGT1

          Birkhead, G., and R. L. Vogt (Vermont Dept. of Health, Burlington, VT 05402).
        Epidemiologic surveillance for endemic Giardia lamblia infection in Vermont:
        the  roles of waterborne and person-to-person transmission. Am J Epidemiol
          The authors studied 1,211 laboratory-confirmed, non-outbreak-related cases
        of giardiasis in Vermont residents reported through Vermont's laboratory-based,
        active surveillance system between  1983 and 1986. Giardiasis was the most
        common reportable disease in the state, with an average annual incidence rate
        of 45.9 cases per 100,000 population per year. This rate is higher than that in
        other states reporting giardiasis incidence.  Morbidity from giardiasis was also
        significant in that 30% of cases reported symptoms lasting four or more weeks.
        Waterborne transmission  was  suggested to  be an  important cause of  non-
        outbreak-related cases because rates of infection were highest  in persons
        receiving nonfiltered municipal or nonmunicipal residential drinking water. Rates
        were also higher at higher elevations, where water supplies may be difficult to
        protect from contamination.  In  addition, the pattern of age-specific incidence
        rates and the high estimated incidence of infection in children attending day care
        suggested that person-to-person transmission also played a role in causing non-
        outbreak-related cases. Routine surveillance data  can  serve to indicate likely
        important routes of transmission of giardiasis in the community.

         Giardia; giardiasis; water supply
   Giardia lamblia has been recognized as a   demies in the  United  States  in terms of
 human pathogen since outbreaks  of giar-   both the number of outbreaks and the num-
 diasis were first described in the 1960s and   ber of persons affected  (3). In  addition,
 1970s (1, 2).  It is the most commonly de-   there  have been  numerous reports of the
 termined cause of waterborne disease epi-   high prevalence of giardiasis in  child day-
 	                                  care settings due to person-to-person trans-
   T>   • j c    ui-  *•   r. u     ™ ,,«,„   ,.    mission (4).  However, the  extent of en-
   Received for publication February 29, 1988, and in   ,    .              .
 final form August 23,1988.                        demic, or non-outbreak-related, giardiasis
   ' Epidemiology Division, Vermont Department  of   and the risk factors  for endemic transmis-
 Heaith, Burlington VT.                          sion in the United States are less well de-
   -Division of Field Services, Epidemiology Program      .,
 Office, Centers for Disease Control, Atlanta, GA.       scribed, partly because this infection has
   ' Current address: New York State Department  of   only recently been made reportable in many
Altany NY™"  Of Communicable  Disease Contro1'   states and few population-based data are
   Reprint requests to .Dr. Richard L. Vogt, Epide-   available.
miology Division, Vermont Department of Health,     Epidemiologic  surveillance systems are
Box 70, 60 Main  Street, Burlington, VT 05402.         one source of information  used to deter-
   The authors thank Dr. Robert A. Gunn, Division         ,      j    •    ,    ,,       ,, ,.
of Field Services, Centers for Disease Control, for his   mme the Predominant patterns of disease
comments on the earlier versions of the manuscript.    transmission in a community. Giardiasis


has been subject to epidemiologic surveil-
lance in Vermont since it was made a re-
portable disease on July 1, 1979. We report
on a review of Vermont giardiasis surveil-
lance data for 1983 to 1986 conducted to
better define the epidemiology of endemic
giardiasis in the state.

  The Vermont Department of Health con-
ducts active laboratory surveillance  for re-
portable  diseases and has previously  re-
ported on the attributes of this system (5,
6). Briefly, all 18 licensed clinical labora-
tories in the state are required to contact
the Epidemiology Division weekly by tele-
phone or by mail whether or not they have
cases to report. Laboratories failing to re-
port regularly are contacted by the  health
department. Physicians are also required to
report cases.
  A public  health nurse interviews each
case using a standard form recording  de-
mographic information, the date of onset
and nature of symptoms, and possible risk
factors for infection.  The risk factors  in-
quired about are residential drinking water
source, child day-care attendance, drinking
of untreated water while away from home
(e.g.,  while hiking), and history of foreign
travel in  the month before illness. The
nurse offers free laboratory stool testing to
asymptomatic members of case households
and recommends that those with symptoms
be treated by a physician.
  For this study, a case of giardiasis was
defined as a Vermont resident with G. lam-
blia cysts  or trophozoites in the stool who
either had onset of symptoms of giardiasis
(diarrhea  or cramping) beginning  in  the
period January 1, 1983 through December
31, 1986,  or, if asymptomatic, submitted
the laboratory specimen during this time.
Laboratory reports were examined through
March 31, 1987, to ensure complete ascer-
tainment of 1986 cases. Reports were  not
included in the study if the case was part
of a  recognized community outbreak or
child day-care center outbreak (two or more
cases with the same probable source of in-
formation and not living in the same house-
hold), was a recently arrived refugee, or had
previously  reported  giardiasis. Symptom
and  risk factor information was reviewed
in detail for 1985 cases.
  Only symptomatic cases were included in
incidence calculations because we assumed
that a larger and more constant proportion
of symptomatic than  asymptomatic cases
come to medical  attention. We felt that
information on symptomatic cases would
then be more representative of all sympto-
matic cases and less subject  to biases  re-
lated to ease of attaining medical  care or
other factors. In addition, the date of infec-
tion for incidence calculations  could  be
known with certainty only for symptomatic
cases.  Finally,  including  asymptomatic
cases identified in the case-household  in-
vestigations  could  bias  incidence rates
downward  in groups   less  completely
screened, such as  adults (compared with
children) or persons  in areas of the state
served by  a  less vigorous public health
  Age-  and  sex-specific  population  esti-
mates for Vermont for 1983-1986 were used
to calculate rates (7). Incidence rates by
town were  calculated from  the  median
number of cases per town during the four-
year period.
  The rates of symptomatic G. lamblia in-
fection  were  calculated for persons  with
different sources of  residential drinking
water.  For this  analysis, cases  were  ex-
cluded if they reported other possible  risk
factors  for  infection  such as day-care  at-
tendance, contact with a day-care attendee,
drinking of untreated water  while away
from home, or a history  of recent foreign
travel. Cases with a household member who'
had  G.  lamblia infection beginning more
than three days before their own illness
were also excluded to eliminate cases that
were possibly due  to  person-to-person
transmission  in  the  home.  Residential
water sources were categorized as munici-
pal or private. Municipal systems were fur-
ther categorized by type of source (surface

or well) and by type of treatment for surface
sources (filtered or not filtered). The types
of water systems serving each town  and
estimates of the population served in 1985
were obtained from the municipal water
system  inventory  of the Environmental
Health Division, Vermont Department of
Health. Since we could  not determine the
type of water system supplying the homes
of giardiasis cases  in towns with multiple
municipal water systems, we assumed that
cases were on the system serving the largest
population in the town.  We calculated the
total population receiving drinking water
from nonmunicipal water systems by sub-
tracting the number of persons on munici-
pal systems from the total state population.
We determined the risk of G. lamblia infec-
tion by elevation of town of residence using
each town's elevation above sea level (8).
  The rate of infection was also calculated
for children attending day care.  An esti-
mate of the number of children under age
five years attending licensed  or registered
day-care centers in the  state in 1985  was
obtained from the  Department of Social
and Rehabilitative Services, State of Ver-

  A total  of 1,406 laboratory-confirmed
cases of giardiasis  in Vermont residents
were reported to the Vermont Department
of Health between  1983 and 1986 (table 1).
Of these, 1,211 (86 per cent) cases were not
          outbreak-related. These form the basis of
          the study.  Case investigation forms were
          completed  for  1,202 (99 per cent) cases.
          Symptomatic cases accounted for 977  (81
          per cent) and asymptomatic cases for 234
          (19 per cent) of the cases. A total  of 127
          outbreak-related  cases were  identified  as
          being part  of one community outbreak or
          more than 20 day-care outbreaks. Refugees,
          predominantly  from southeast  Asia,  ac-
          counted for the remainder of the cases.
            The mean  annual  incidence  rate  for
          symptomatic cases  was  45.9  cases per
          100,000 population per  year (table 1). Giar-
          diasis  was  the most common  reportable
          enteric disease in  Vermont during this
          time. The occurrence of symptomatic cases
          peaked during the months of July through
          November  (figure 1). This pattern was ev-
          ident  regardless  of residential  drinking
          water  source or whether  drinking  of un-
          treated water  outside  the home was re-
          ported (data not shown). The rate of symp-
          tomatic G.  lamblia seemed to be highest in
          towns located in the mountainous northern
          and central regions of the state (figure 2).
            The mean annual age-  and sex-specific
          incidence rates for symptomatic cases for
          1983 to 1986 are shown in figure  3. The
          highest rates are in children aged  one to
          four years.  In this age group, males had 1.5
          times the rate of symptomatic giardiasis of
          females (95 per  cent  confidence interval
          (CI) 1.1-2.0). The rate was also high for
          persons, especially women, aged 20 to  39
                                     TABLE 1
 Laboratory-confirmed cases of giardiasis reported to the Vermont Department of Health, 1983-1986: Vermont
                       Department of Health giardiasis surveillance data
Non -outbreak -related
Total reported
  * Population estimates by the Vermont Department of Health (7).

    Jon  Fob Mar Apr May Jun  Jul Aug  Sop  Oct  Nov  Dec
  FIGURE 1. Number of symptomatic G. lamblia in-
fections, by month, in Vermont, 1983-1986. Vermont
Department of Health giardiasis surveillance data.
years. Women in this age group were 40 per
cent more likely to have symptomatic giar-
diasis than men (relative risk (RR) =  1.4,
95 per cent CI 1.1-1.6).
  Diarrhea  was  the most commonly re-
ported symptom (96 per cent), followed by
cramps (64  per cent)  and weight loss (48
per cent). Thirty per cent of symptomatic
cases reported  having symptoms lasting
more than four weeks.  The most commonly
reported risk factor for infection was his-
tory of contact with  a person with diarrhea
in the last month (38 per cent), followed by
drinking of untreated water while  away
from home (13 per cent), attendance at day
care (12 per cent), contact with a day-care
attendee (8 per cent),  and foreign travel (6
per cent).
   Information on the  source of residential
water supply was  available for 243 (96 per
cent) symptomatic 1985 cases, of whom 118
received their drinking water from munic-
ipal water systems and 125 obtained it from
nonmunicipal  water sources.  The  health
department's   water  systems   inventory
showed  that  an  estimated 56 per  cent
(300,273) of  the  state's  population  was
served by municipal water systems in 1985.
Using these data, we  estimated that state
residents with nonmunicipal  residential
water systems have approximately a 30 per
cent excess risk of developing symptomatic
giardiasis compared with those  on munici-
pal water systems (RR = 1.3, 95 per cent
                       Rates per 10,000 per year
                         >5 to IO/IO,000/Year
                               Scale of Kllomatan
                               )    20   40
  FIGURE 2. Median yearly incidence rate of symp-
tomatic G. lamblia infection, by town, in Vermont,
1983-1986. Rales are given per 10,000 population per
year. Vermont Department of Health giardiasis sur-
veillance data.

CI 1.0-1.7). The rates of giardiasis varied
by type of water system (table 2). Nonmun-
icipal water system users had 2.2 times the
rate of giardiasis of those  on filtered, sur-
face  water  municipal  systems.  Municipal
surface water  systems without filtration
had almost twice  the  rate of giardiasis of
those with  filtration. Rates  were elevated
for users of municipal well water  systems.
The  risk of infection  also increased  with
increasing elevation of town of residence
(table 2).
  Among 92 cases less than five years of
age, 36 (39 per cent) attended day care. The
Department of Social and  Rehabilitative
Services estimated  that  12,000 positions
were available in licensed or registered day-
care centers in 1985, which meant that 29.4
per cent of  children in the state under age
five years were in  officially sanctioned day

                                     BIRKHEAD AND VOGT
                                                             E3  Male
                                                             ESS  Female
            <1     1-4    5-9   10-14  15-19  20-2930-3940-4950-5960-69  70+

                                          AGE GROUP (years)
   FIGURE 3.  Incidence rates of symptomatic G. lamblia infection, by age and by sex, in Vermont, 1983-1986.
 Vermont Department of Health giardiasis surveillance data.

                                            TABLE 2
 Risk factors for development of G. lamblia infection in Vermont, 1985: Vermont Department of Health giardiasis
                                        surveillance data
               Risk factor
Population at   Rate/100,000
    risk        population
                                                                                     95% CI*
Residential water supplyt
Surface water, filtered
Surface water, not filtered
Well water
Child day-care attendance^:
Elevation of town of residence (m)§
















  * RR, relative risk; CI, confidence interval.
  t Cases without other risk factors for giardiasis.
  $ Cases aged less than five years.
  § Case data for 1983-1986, missing elevation for five towns and missing town of residence for four cases.
Rates of G. lamblia infection in this section are calculated per 100,000 population per year.

care, assuming full enrollment. The relative
risk of day-care attendees  compared with
that of nonattendees is shown in table 2.

  Our review of surveillance data showed
that G. lamblia was the most common re-
portable enteric disease in Vermont during
the period 1983-1986. The incidence rate
of symptomatic giardiasis (45.7 per 100,000
population) in  Vermont is much greater
than rates  reported previously by  other
states. In Minnesota, the incidence rate was
9.8  per 100,000 population based on data
from the state laboratory for 1971 through
1975 (9). A survey of most laboratories in
Colorado in 1972 and 1973  found a giardi-
asis incidence rate of 11.6 per 100,000 pop-
ulation per year  (10). This had increased
by  1981-1982  to approximately  16  per
100,000 population per year based on state
laboratory reports and active case finding
among most primary-care  providers (11).
These  reports did  not  exclude asympto-
matic,  outbreak-related cases or cases in
  One explanation for the higher giardiasis
incidence rate in Vermont is that the active,
laboratory-based reporting  and  active case
investigations may detect a  greater propor-
tion of cases than do other state surveil-
lance systems. In addition, physician rec-
ognition  and diagnosis  of  giardiasis may
have improved since the previous  reports.
Finally, the incidence of  giardiasis may
truly be higher in Vermont than in  other
states, perhaps partly because of the rural
nature of Vermont and the  possibly higher
frequency of contact with sources of infec-
tion in the environment, particularly water.
The true giardiasis incidence in  Vermont is
probably higher than that reported here
because some state residents may leave the
state to be  diagnosed. This is  a threat to
the sensitivity  of most surveillance sys-
tems. We have no information on the  num-
ber of these cases, but we suspect that they
are probably few.
  Our data suggest that waterborne trans-
mission is  an important  source of non-
outbreak-related  G.  lamblia  infection in
Vermont. Persons receiving drinking water
from  nonmunicipal  residential  drinking
water systems, which often have an unpro-
tected water source and may not be filtered,
had the highest rate of symptomatic giar-
diasis of any group. Municipal water sys-
tems with filtration had the lowest rates of
giardiasis,  which is in  accord with the
knowledge  that filtration is necessary for
adequate treatment  of water  to remove
Giardia cysts (3, 12). Shallow, poorly con-
structed wells have been shown by  others
to be  a risk factor for giardiasis (13).  The
increased risk of  infection among users of
municipal well water in Vermont suggests
that some of these systems are not adequate
to prevent giardiasis.
  The increase in the rate of infection with
elevation  also  suggests  that waterborne
transmission is  occurring. Outbreaks of
giardiasis in mountainous areas have been
traced to drinking contaminated water (14).
Rates of illness are higher in these areas
where easily contaminated surface waters
may be the predominant source of drinking
water  and where low  water  temperature
may prolong the  survival of Giardia cysts
(14). A final point in favor of the possibility
that waterborne transmission is important
in endemic giardiasis is that more cases had
onset  in the summer months when outdoor
activities bring more  people  into contact
with possibly contaminated water. It is un-
clear why the peak in the number of cases
continued into the fall.
  Our data are limited by lack of knowledge
of the exact municipal  water source for
cases  in towns with multiple water systems
and by the crude nature of the estimates of
the population served. It is possible  that
other  factors associated with  having unfil-
tered  municipal or nonmunicipal residen-
tial drinking water, such as lower socioeco-
nomic status, poorer access to medical care,
or crowding, may have a role in the in-
creased risk of infection in this group. The
variable  sensitivity of the giardiasis  sur-
veillance system in different regions of the
state  may  also contribute to some  of the

 differences in rates. These factors are un-
 likely to significantly affect our results be-
 cause of the  consistency  of relative  risks
 associated  with different  types  of water
 supply and elevation and because the  more
 rural  parts of the  state,  where access  to
 medical  care  and reporting  might be ex-
 pected to be  the poorest, had the highest
 rates of giardiasis.
   Our data also indicate that person-to-
 person transmission may play a role in non-
 outbreak-related giardiasis. Children  aged
 one to four years had the highest rates  of
 symptomatic  infection of  any  age group.
 This pattern  has been described in other
 enteric infections (15-18) and may be due
 to poorly developed immunity or increased
 fecal-oral  contact among children.  Child
 day-care centers are a known site of person-
 to-person transmission among children (4).
 The higher rates observed for male children
 are unexplained, but a similar pattern has
 been observed by others with both Giardia
 and Campylobacter infections (18).
  The data we have presented indicate that
 G. lamblia may be a more common pathogen
 than has been previously  recognized  from
 reports of outbreak investigations. Surveil-
 lance data are limited in identifying caus-
 ative factors  by their descriptive nature,
 but where  data  are  available on the popu-
 lations at risk,  they can serve to confirm
 known risk factors and to estimate  their
importance in disease transmission in the

 1. Moore. GT, Cross WM, McGuire D,  et al. Epi-
   demic giardiasis at a ski resort. N Engl J Med
 2. Shaw  PK, Brodsky RE,  Lyman DO, et  al.  A
   community-wide outbreak of giardiasis with evi-
   dence of transmission by a municipal water sup-
    ply. Ann Intern Med 1977;87:426-,12.
 3.  Centers for Disease  Control. Water-related dis-
    ease outbreaks: annual summary, 1983. Atlanta,
    GA: Centers for Disease Control, September 1984.
 4.  Pickering LK, Woodward WE, DuPont HL, et al.
    Occurrence of Giardia lamblia in children in day-
    care centers. J  Pediatr 1984;104:522-6.
 5.  Vogt RL, Clark SW,  Kappel S. Evaluation of the
    state surveillance system using hospital discharge
    diagnoses, 1982-1983. Am J Epidemiol 1986;123:
 6.  Harter L, Frost F, Vogt RL, et al. A three-state
    study of waterborne disease surveillance  tech-
    niques. Am J Public Health 1985;75:1327-8.
 7.  Vermont Department of Health.  1985 estimates
    of the population and housing of Vermont coun-
    ties and towns. (Vermont public health statistics
    bulletin). Burlington, VT: Vermont Department
    of Health, March 1986.
 8.  Hennenberger JD, ed. 1982  Vermont year book.
    Chester, VT: The National Survey, 1982:135-507.
 9.  Weiss HB, Winegar DA, Levy BS, et al. Giardiasis
    in Minnesota 1971-1975. Minn Med 1977;60:815-
    Wright RA, Spencer HC, Brodsky RE, et al. Giar-
    diasis in Colorado: an epidemiologic study. Am J
    Epidemiol 1977; 105:330-6.
11.  Hopkins RS, Shillam P, Gaspard B, et al. Water-
    borne disease in Colorado: three years surveillance
    and 18  outbreaks. Am  J  Public Health 1985;
12.  Logsdon GS, Symons JM, Hoye LH. Water filtra-
    tion techniques for  removal of cysts and cyst
    models. In: Jakubowski W, Hoff JC, eds. Water-
    borne transmission of giardiasis. Cincinnati, OH:
    US Environmental Protection Agency, 1979:240-
    56. (Publication no. EPA-GOO 9-79-001).
13.  Chute CG, Smith RP, Baron JA. Risk factors for
    endemic  giardiasis. Am J  Public Health 1987;
14.  Meyer EA, Jarroll EL. Giardiasis. Am J Epidemiol
           15. Ryder RW, Merson MH, Gangarosa EJ. Salmo-
              nellosis in the United States, 1968-1974. J Infect
              Dis 1976;133:483-6.
           16. Blaser MJ, Pollard RA, Feldman RA. Shigella
              infections in the United States, 1974-1980. J In-
              fect Dis 1983;147:771-5.
           17. Tauxe RV, Pegiies DA, Hargrett-Bean N. Cam-
              pylobacter infections: the emerging national pat-
              tern. Am  J Public Health 1987;77: 12 19-21.
           18. Hopkins RS, Olmstead RN. Campylobacter jejuni
              infections in Colorado:  unexplained excess  of
              cases in  males. Public  Health Rep  1985; 100:



    Reproduced,  with  permission,
    from  American  Journal  of
    Public Health
                             Cryptosporidiosis and Surface Water

                             DEBRA J. GRABOWSKI, MS, RS, AND HARRY F. HULL, MD
    Abstract: In the period July through October, 1986, 78 labora-
tory-confirmed  cases of cryptosporidiosis were identified in New
Mexico. To determine possible risk factors for development of this
disease, we conducted a case-control study; 24 case-patients and 46
neighborhood controls were interviewed. Seventeen (71 per cent) of
the 24 case-patients were  females, seven (29%) were males; their
             ages ranged from 4 months to 44 years, median 3 years. There was
             a strong association between drinking surface water and illness: five
             of the 24 case-patients, but none of the 46 controls drank untreated
             surface water. Among children,  illness  was also associated with
             attending a day care center where other children were ill (odds ratio
             = 13.1). (Am J Public Health 1989; 79:39^2.)

    Ciyptosporidiitm, first recognized as a cause of diarrhea
in humans in 1976.' has become known as a common cause
of diarrhea worldwide.2'3 No effective therapy is known; the
diarrhea is often prolonged and accompanied by weight loss.
In immunocompromised persons,  cryptosporidiosis can
cause a life-threatening dehydration.
    Cryptosporidium can be  transmitted through contact
with animals, especially cattle and sheep,4  through person-
to-person  contact, especially  in  day care centers,5 and
through a contaminated community water source.6 In several
reported  outbreaks,  the  source  of the infection was
    In 1984, 24 cases of cryptosporidiosis were reported to
the New Mexico Office of Epidemiology, 13  of them were
associated with two day care centers. In  1985, 24 cases were
reported. In July 1986, a cluster of 20 cases of cryptospori-
diosis  was recognized in Bernalillo County  (the city  of
Albuquerque). Because these cases did not  appear to  be
related and the source of  infection  was unclear, a case-
control study to evaluate potential risk factors for illness was

    Since February 1984, all stool samples submitted to the
New Mexico Health and Environment Department's Scien-
tific Laboratory Division for ova and parasite testing have
been  examined  for  Cryptosporidium.  Stool  samples are
concentrated using a formalin-ethyl  acetate technique. A
heat-fixed fecal  smear is  stained with Kinyoun modified
acid-fast stain9 and counterstained with Loeffler methylene
blue.  The specimen is then examined under a light micro-
scope for the presence of Cryptosporidium oocysts.
    In August 1986, the Office of Epidemiology sent a letter
to primary care physicians in Bernalillo County, alerting
them  to the cluster of cases, reviewing the  symptoms and
             diagnosis  of cryptosporidiosis, and  requesting that  stool
             samples be sent if cryptosporidiosis was suspected.
                 A case-patient was defined  as  a  Bernalillo County
             resident with laboratory-confirmed cryptosporidiosis report-
             ed to the Office of Epidemiology between July 1 and October
             1, 1986. If more than one person with laboratory-confirmed
             cryptosporidiosis  was identified within a household or day
             care group, only the one with the earliest onset of symptoms
             was included in the study.
                 Controls were matched for age, sex, and neighborhood
             using a reverse telephone directory. The home  of the  case-
             patient was  identified in the directory. The homes listed on
             either  side of the  case-patient were contacted by telephone
             alternately until  suitable  controls were identified.  Case-
             patients 10 years  old or younger were matched within two
             years of age, and those over 10 years were matched within
             five years  of age. A control could not have had diarrhea in the
             four weeks before the onset of disease in the case-patient to
             which he/she was matched. Twenty-two of the 24 index  cases
             were matched with two controls each. Two index cases were
             matched with only one control.
                 A standard questionnaire, administered via telephone to
             both case-patients and controls, requested  information on
             size of household, day care attendance or contact, travel,
             surface water exposure, pets and domestic animal exposure
             during the four weeks before  onset  of illness  in the  case-
                 Two  surface  water samples were collected on October
             27, 1986 from a lake in southern New Mexico and a lake in
             Arizona where three of the case-patients had been exposed to
             surface water. These samples  were tested for Cryptospori-
             dium  and Giardia using a fluorescent  labeled  monoclonal
             antibody technique.10'11
                 Using the matched triplets  or  pairs,  we calculated
             Mantel-Haentzel  chi-square,  odds ratios  (OR),  and  95%
             confidence  intervals (CI) and fitted  a conditional logistic
             regression to determine which variables were independently
             associated with infection with  Cryptosporidium.
   Address reprint requests to Margaret M. Gallaher, MD, Office of Epi-
demiology, New Mexico Health and Environment Department, P.O. Box 968,
Santa Fe, NM 87504-0968. Ms. Nims and Dr. Hull are also with the NM Health
and Environment Department in Santa Fe. Ms. Herndon is with the Epidemi-
ology Program Office, Centers for Disease Control, Atlanta; Dr. Gallaher is on
assignment to New Mexico from the CDC Epidemiology Program Office. Dr.
Sterling is with the Department of Veterinary Sciences, University of Arizona,
Tucson; Ms. Grabowski is with the Albuquerque (NM) Environmental Health
Department. This paper, submitted to the Journal December 24, 1987, was
revised and accepted for publication April 22, 1988.
© 1989 American Journal of Public Health 0090-0036/89$!.50

                  Of the 56 laboratory-confirmed cases of cryptosporidi-
              osis in Bernalillo County reported between July  1 and
              October 1, 26 met the definition of a case-patient. Two (8
              percent) of the cases were lost to follow  up (a  1-year-old
              female  who attended a day care  center and  a 43-year-old
              female). Of the 30 cases excluded from the study, two were
              patients with  acquired  immune deficiency syndrome who
              died during the course of the study, 10  were  household
AJPH January 1989, Vol. 79, No. 1

                                          H Male

                                          :;::: Female
                                                              TABLE 1—Frequency of Reported Symptoms of Cryptosporidiosis, New
                                                                      Mexico, 1986(N = 24)
Abdominal Pain
Abdominal distention
                        15-19   20-2J   25-29  30-34   35-39   40-44

                     AGE GROUP (Years)
 FIGURE 1—Cryptosporidiosis, by Age Group and Sex, New Mexico, 1986

 contacts of case-patients, and 18 were day care contacts of
     The 24 remaining case-patients ranged in  age  from 4
 months to 44 years with a median age of 3 years. Figure 1
 shows the age and sex distribution of the index cases.
 Seventeen of the cases were 10 years of age or younger; 13
 of whom were females; 17 of all  24  cases were female.
 Thirteen (54 per cent) of the case-patients were White and 11
 (46 per cent) were Hispanic. Controls showed similar demo-
 graphic characteristics.
     Case-patients had onset of symptoms from May 28 to
 September 2,  1986  (Figure 2).  The duration of symptoms
 ranged from five to 60 days with a median of 21 days. The
 most frequent symptom (Table  1) was diarrhea followed by
 flatulence and abdominal pain.
     Matched set analysis identified two possible sources of
 infection: surface water (rivers, lakes and streams) and day
 care  centers (Table 2). Five case-patients drank untreated
 surface water, two in Mexico, and three in southern Colorado
 and New Mexico; none of the controls drank surface water.
 Case-patients reported more swimming in surface water in
                   6 —
            the four weeks prior to illness. The time between exposure to
            surface water (swimming or drinking) and  onset of illness
            ranged from four to 21  days with a median of seven days.
            Fourteen  (82  per cent) of  17 household  members  with
            exposures to surface water similar to the case-patients also
            became ill with diarrhea. These household members became
            ill within two to seven days of the onset of illness in the
                Case-patients were substantially more likely to attend a
            day care center where other children were ill with diarrhea
            than were their matched controls (Table 2). Attendance at
            day care centers where no other children were ill was not
            related to risk.
                The logistic regression model for multivariate analysis
            included all variables except drinking surface water, which
            did not vary.  Infection was associated  with swimming in
            surface water,  attending a day care center, camping, and
            having a pet which was less than six months of age or was ill,
            with odds ratios all greater than 3.0. Having a pet over six
            months of age  was associated with a  decreased risk of
            disease.  All of these  estimates  were unstable,  with  95%
            confidence limits including 1.0.
                Among the 24  case-patients,  eight were exposed to
            surface water,  eight were  day  care  center attendees  or
            employees, six were household contacts of a day care center
            attendee, and eight had young and/or ill pets. Four had none
            of these exposures.
                The surface water samples  from  New Mexico and
            Arizona where cases had been exposed did not yield Cryp-
            tospondium or Giardia on testing.
                Two of the 24 index cases were coinfected with Giardia.
                      28  1   5   9   13  17  21  25  29  3   7  11  15  19  23  27  31  4
                                                                                 12  16  20  24 28  1
                    MAY       JUNE            JULY                       AUQ
                FIGURE 2—Cryptosporidiosis, by Date of Onset, New Mexico, May 28-September 4, 1986
                                      AJPH January 1989, Vol. 79, No. 1

                                                                          CRYPTOSPORIDIOSIS AND SURFACE WATER
                TABLE 2—Exposures Associated with Cryptosporidlosis

Drink surface water

Swim in surface water

Attend a day care center
where other children
were ill

Household contact of
day care center

No. of Cases




of controls





95% Confidence
Odds Ratio Interval
00 2.44, 00

3.7 1.02,13.5

13.1 2.58, 66.4

3.0 0.91, 9.85

                   Numbers in parentheses represent cases with only one control.
One of them attended a day care center. The other had no
apparent risk factors for cryptosporidiosis or giardiasis.


    Person-to-person transmission  of cryptosporidiosis  is
well recognized, especially in day care centers.5'12 Lack of
hygiene and frequent mouthing of objects among diapered
babies and toddlers can lead to the rapid spread of this and
other enteric infections in day care centers and homes.13 This
study confirmed the association between attending a day care
center, especially where other children are ill, and crypto-
    Cryptosporidiosis  has also been found to be a cause of
traveler's diarrhea, particularly  in people traveling to less
developed countries.2'14 In this study, two (8 per cent) of the
index cases had recently traveled to Mexico; both of these
case-patients also drank untreated surface water in Mexico.
    Although puppies  and kittens  have been  shown to
excrete Cryptosporidium,* these pets  have  not  previously
been demonstrated  to transmit cryptosporidiosis to humans.
Contact with a young and/or ill pet was associated with an
increased risk of disease in this study, but the numbers are
too small to be convincing. Nevertheless, this observation is
in  keeping with what is known of  cryptosporidiosis  in
domestic animals: young animals contract cryptosporidiosis
more often than older animals, and the severity and duration
of illness  is inversely related to the age of the animal.1'15
    This study suggests for the first time that drinking surface
water may be a mode of transmission of cryptosporidiosis in
the United States. This is feasible since waterborne crypto-
sporidiosis has been previously demonstrated in a commu-
nity in  Texas due  to  a contaminated  artesian  well.6 The
plausibility of waterborne transmission is increased by the
time  relationship between exposure  to  surface water and
onset of illness. The median incubation period of seven days
in those case-patients exposed to surface water is consistent
with  what  is  known  of  the  incubation period  for
    Cryptosporidium sp. has been found in a broad variety of
both domestic and wild animals. Cattle and sheep appear to
be  important hosts, but mice, snakes, rabbits, and deer have
also been  found to excrete the organism.4'15 Much of the
recreational water in New Mexico and other western states
drains land which has  multiple uses including the grazing of
livestock.  Livestock could contaminate surface water with
Cryptosporidium by direct deposit of feces in the water or via
surface runoff washing animal feces into the water. Similarly,
contamination of the water by wild animal feces might also
play an important role. New Mexico received  heavier than
normal rainfall during the summer months of 1986, a factor
which may have promoted contamination of surface water.
    The  finding of an  elevated risk of cryptosporidiosis
among persons who had recreational contact with, but did not
drink surface water, supports waterborne transmission of the
disease.  Recreational surface  water contact, specifically
swimming, has been  shown  to be a risk factor  for other
enteric infections.17'18 Swimmers become infected by swal-
lowing small amounts  of contaminated water,  especially
following submersion. The infective dose of Cryptosporidium
is unknown, but is probably small, because of the organism's
ability to complete its  life  cycle in  a  single  host and to
reinitiate the developmental cycle in that host.4 A small
infective dose for cryptosporidiosis  would facilitate trans-
mission by water contact without frank water consumption.
     Our studies of surface water in New Mexico were limited
in scope, but failed to demonstrate the presence of Crypto-
sporidium. These studies were done two months after the
outbreak and probably represent a failure to detect intermit-
tent, low-level, or seasonal contamination. Additionally, the
samples  were taken from a large lake and it was difficult to
determine the exact site of exposure. Studies of surface water
conducted in Arizona showed that  there  was widespread
contamination of streams with Cryptosporidium during the
summer of 1986," but that most of the water sources did not
contain Cryptosporidium during the fall months.*
     Cryptosporidiosis occurs more  commonly in summer
and fall in these and other data.3'4 Increased outdoor activity
during the summer and early fall with drinking of or swim-
ming in  surface water  may  partially explain  the seasonal
     Coinfection with Giardia lamblia has been  found in 2 per
cent to 23 per cent of cryptosporidiosis cases  in  previously
described outbreaks19"21 and  8 per cent in this  study. The
coincidental occurrence of two parasitic infections suggests
a common mode of transmission. Person-to-person transmis-
sion of  both giardiasis and  cryptosporidiosis in day care
centers has been well described.  Waterborne transmission
of giardiasis from surface water is also well known.22'23 Our
    * Unpublished data, CR Sterling, University of Arizona, 1986.
 AJPH January 1989, Vol. 79, No.

 study suggests that waterborne transmission of cryptospori-
 diosis may explain coinfection with these two parasites, even
 though neither of the coinfections in this  study occurred in
 persons who had exposure  to surface water.
      The median duration of diarrhea  in the study population
 was longer than the three to  14 days described by others,' •'7-25
 perhaps because patients who had stool samples submitted for
 ova and parasite  examination had diarrhea that lasted longer
 than three to five days and did not respond to usual symptomatic
      The predominance of  young children  found here has
 been seen in other studies, and follows the same pattern seen
 in domestic animals.  In addition, young children are more
 likely to come to medical attention earlier in the  course of a
 diarrheal  illness than adults and more likely to be tested for
 the  cause of their diarrhea.  The explanation for the predom-
 inance of females seen in this group of children  is not clear
 and may reflect only incomplete case ascertainment.
      Further study of the role of surface water in transmission
 of cryptosporidiosis is clearly needed. Meanwhile, boiling of
 surface water  rather than chemical disinfection  may  be
 prudent, since Cryptosporidium is resistant to most common-
 ly  used  chemical  disinfectants including  chlorine  and
 iodine,15'24 but  oocysts are destroyed by heating  above 65°C
 for  30 minutes.24  Boiling  also  destroys  other  organisms,
 including  Giardia,  that may  be in  the water.  We  also
 recommend that physicians  consider cryptosporidiosis in the
 differential diagnosis of persons with diarrhea and a history
 of surface water  contact.
     The authors wish to thank Kathy Powers. RN, and Judy Knott, RN, for
 their assistance in this study. A portion of the data described in this paper was
 accepted for presentation at the annual EIS Conference. Centers for Disease
 Control, April 6-10. 1987. in Atlanta, Georgia.

  1.  Navin TR, Juranek DD: Cryptosporidiosis: Clinical, epidemiologic, and
     parasitologic review.  Rev Infect Dis 1984; 6:313-327.
  2.  Navin TR: Cryptosporidiosis in humans: Review of recent epidemiologic
     studies. EurJ Epidemiol 1985; 1:77-83.
  3.  Mata L: Cryptosporidium and other protozoa in diarrheal disease in less
     developed countries. Pediatr Infect Dis 1986; 5:8117-8130.
  4.  Current WL: Cryptosporidiosis. J Am Vet Med Assoc 1985; 187:1334-
 5.  Centers for Disease Control: Cryptosporidiosis among children attending
    daycare centers—Georgia,  Pennsylvania,  Michigan. California, New
    Mexico. MMWR  1984: 33:559-601.
 6.  D'Antonio RG, Winn RE, Taylor JP, el al: A waterborne outbreak of
    cryptosporidiosis  in normal hosts. Ann Intern Med 1985:103:886-888.
 7.  Hunt DA. Shannon R. Palmer SR, Jephcott AE: Cryptosporidiosis in an
    urban community. Br Med J 1984; 289:814-816.
 8.  Wolfson  JS. Richter JM, Waldron MA, Weber DJ,  McCarthy DM,
    Hopkins CC: Cryptosporidiosis in immunocompetent patients. N Engl J
    Med 1985; 312:1278-1282.
 9.  Ma P, Soave R: Three-step stool examination for cryptosporidiosis in 10
    homosexual men with protracted watery diarrhea. J Infect Dis 1983; 147:
10.  Garcia LS, Brewer TC, Bruckner DA: Fluorescent detection of Crypto-
    sporidium oocysts in human fecal specimens using monoclonal antibodies.
    J Clin Microbiol 1987; 25:119-121.
11.  Madore MS. Rose JB, Gerba CP, Arrowood MJ, Sterling CR: Occurrence
    of Cryptosporidium oocysts  in sewage effluents and selected surface
    waters. J Parasit 1987; 73:702-705.
12.  Pickering LK, Bartlett AV, Woodward WE: Acute infectious diarrhea
    among children in day care: Epidemiology and control.  Rev Infect Dis
    1986; 8:539-547.
13.  Pickering LK. Evans DG, DuPont HL, Vollet JJ, Evans DJ:  Diarrhea
    caused by Shigella, rotavirus and Giardia in day-care centers: Prospective
    study. J Pediatr 1981; 99:51-56.
14.  Sterling CR. Seegar K. Sinclair NA: Cryptosporidium as a causative agent
    of traveler's diarrhea. J Infect Dis 1986; 153:380-381.
15.  Tzipori S: Cryptosporidiosis in animals and humans. Microbiol Rev 1983;
16.  Jokipii L,  Jokipii  AMM: Timing of symptoms and oocyst excretion in
    human cryptosporidiosis. N Engl J Med 1986; 315:1643-1647.
17.  Centers for Disease Control: CDC Shigella surveillance. Atlanta: CDC,
    1976: 37:4.
18.  Makintubee S. Mallonee J. Istre G: Shigellosis outbreak associated with
    swimming. Am J Public Health 1987; 77:166-168.
19.  Skeels MR, Sokolow  R, Hubbard CV, Foster LR: Screening for coinfec-
    tion  with Crypwsporidium and Giardia in Oregon public  health clinic
    patients. Am J Public Health 1986; 76:270-273.
20.  Wolfson JS, Hopkins CC, Weber  DJ,  el al: An association  between
    Crypiosporidium and  Giardia in stool. N Engl J Med  1984;  310:788.
21.  Jokipii L, Pohjola S, Jokipii AMM: Cryptosporidium: A frequent finding
    in patients with gastrointestinal symptoms. Lancet 1983:  2:358-361.
22.  Dykes  AC, Juranek  DD, Lorenz RA,  el  al: Municipal  waterborne
    giardiasis:  An epidemiologic investigation. Beavers implicated as a pos-
    sible reservoir. Ann Intern Med 1980; 92:165-170.
23.  Lopez  CE, Dykes AC, Juranek DD,  el al:  Waterborne giardiasis: A
    communitywide outbreak  of disease and a high rate of asymptomatic
    infection. Am J Epidemiol 1980; 112:495-507.
24.  Soave R,  Armstrong D:  Cryptosporidium and  cryptosporidiosis. Rev
    Infect Dis  1986; 8:1012-1023.
25.  Current WL, Reese NC, Ernst JV, Bailey WS, Heyman MB, Weinstein
    WM: Human cryptosporidiosis in immunocompetent and immunodeficient
    persons: Studies of an outbreak and experimental transmission.  N Engl J
    Med 1983; 308:1252-1257.
                             AJPH January 1989, Vol. 79, No. 1

III.  Investigation of Waterborne Disease Outbreaks


   Reproduced, with permission,
   from American Journal  of
   Public  Health
                Community Health  Effects  of a Municipal  Water Supply
                                     Hyperfluoridation Accident
                                    AND STEVEN D. HELGERSON, MD, MPH
     Abstract: For 12 hours, excess hydrofluorosilicic  acid was
  diverted to a 127-home community  water supply. Fluoride  levels
  peaked at 51 parts per million (ppm). Water acidification caused
  copper to leach from the domestic plumbing; raising copper levels to
  25-41 ppm. Fifty-two (33 per cent) of those who drank hyper-fluorid-
  ated water developed mild gastroenteritis. Vomiting was uncommon
  and symptom onsets usually occurred >30 minutes after drinking
  water; suggesting that fluoride, rather than copper, caused illness.
  Skin contact with hyperfluoridated water caused itching and skin
  rashes. (Am J Public Health 1988; 78:711-713.)
     Hyperfluoridation accidents in public water supplies are
 uncommon1"5 and accidents in municipal water supplies have
 been reported only twice previously.4-5 This report describes
 the public health effects of a third municipal water supply
 hyperfluoridation accident that  occurred  in a  residential
 Connecticut community.
     The community has 127 homes. Its water source is a
 nearby treatment plant  which also supplies water to three
 distant larger metropolitan areas. The community is  on a
 "dead-end" portion of  the water distribution system. The
 street water mains are cement lined while feeder lines  from
 the street to each house are copper as is the piping within the
     At the treatment plant, hydrofluorosilicic acid (H2SiF6)
 is injected into the water supply.  At approximately 3:00 pm
 on March 11,  1986, an inadvertently opened valve began to
 divert hydrofluorosilicic  acid that  normally would have  been
 injected  into water supplying the community and distant
 metropolitan areas, solely  into the community's water sup-
 ply. Hyperfluoridated water would have reached the domes-
 tic taps at approximately  6:00 pm (beginning of exposure
 period, time = 0 hours). At +1  hours (7:00 pm) residents
 began notifying  water company  personnel  that  the water
 tasted abnormal and turned blue on contact with soap, and of
 itching and gastrointestinal symptoms. At +1 and +4 hours
 (7:00 and 10:00 pm), household tap water samples revealed
 fluoride and copper concentrations >40 times normal (fluo-
 ride 42-51 ppm [normal 1.0 ppm], copper 25-41 ppm [normal
 0.03 ppm]).
    At +10 hours (4:00 am), a sample of water from a water
 main had fluoride and copper concentrations of 50 ppm and
 0.03 ppm, respectively. The water mains were then flushed.
 Beginning at +12 hours (6:00 am), residents were told not  to
    From the Division of Field Services,  Epidemiology Program Office,
Centers for Disease Control (Dr. Petersen); the Preventable Diseases Division,
State of Connecticut Department of Health Services (Ms. Denis, Dr. Brown,
Dr.  Hadler); Office of the Director, Centers for Disease Control (Dr.
Helgerson); and the Yale University School of Medicine, Department of
Epidemiology and Public Health (Dr. Helgerson).
    Address reprint requests to James L. Hadler, MD, MPH, Epidemiology
Section, State of Connecticut Department of Health Services, 150 Washington
St., Hartford, CT 06106. This paper, submitted to the Journal August 31, 1987,
was  revised and accepted for publication December 14, 1987.
             drink or bathe in the water and to discard ice or beverages
             made with tap water.

             Epidemiologic Investigation
                 On March  15, 1986, a door-to-door survey of the 127
             community households was conducted. Data were obtained
             from at least one adult (>16 years in age) to determine each
             household member's water consumption, dermal exposure to
             water, and symptoms.
                 The exposure period was considered to have been +0
             hours to +12 hours. The quantity of ingested hyperfluorid-
             ated water was estimated  as the number of glasses of tap
             water consumed during the exposure period plus the number
             of glasses of beverages made  from tap  water  during that
             period. The latent interval was defined as the time from
             consumption of the last glass of hyperfluoridated  water to the
             illness onset. A person was considered dermally exposed to
             the hyperfluoridated water if that person bathed or showered
             at home during the exposure period.
                 The outbreak period was defined from 0 to +54 hours. A
             case of gastrointestinal irritation was  a resident with onset
             during this period of any of the following symptoms: diarrhea
             (£2 watery stools in  a  24-hr period),  abdominal cramping,
             severe nausea, or vomiting. A case of skin irritation was a
             resident with onset of unusual  itching during the outbreak
                 Statistical relations between water consumption or der-
             mal exposure and illness were determined by the method of
             related ratios to control for household.'-7 The dose-response
             relation of water consumption  and gastroenteritis was as-
             sessed by the Mantel  Chi square test  for  trend8; other
             differences were assessed by  the Chi  square test.  The
             effective dose of hyperfluoridated water  that would be
             expected to cause .gastroenteritis in 50 per cent of a popula-
             tion (ED50) was calculated  by the log probit method.9


                 Information concerning 321 persons was gathered from
             86 (68 per cent) of the 127 households.  Representatives from
             two households refused interviews and in 39 households no
             adult was available for interview on the survey day. Abnor-
             mal taste or color of water was reported in 62 per cent (53/86)
             of households.

             Gastrointestinal Illness

                 Gastrointestinal symptom histories  were obtained for
             312 persons, 55 (18 per cent) of whom were cases; symptoms
             included: abdominal cramping (66 per cent), nausea (62 per
             cent), headache (49 per cent), diarrhea (42 per cent), vomiting
             (13 per cent), diaphoresis (12 per cent), and fever (4 per cent).
             The onset of illness for the 46 persons with known time of
             symptom onset is depicted in Figure 1. The median duration
             of gastroenteritis symptoms was 5.5 hours with a range of 1
             to 60  hours. No person  sought  medical  evaluation  for
             gastroenteritis symptoms.
AJPH June 1988, Vol. 78, No. 6

    1 Exposure Period
                                                             TABLE 1—Cases of Skin Irritation by Bathing or Showering Status,
                                                                      Connecticut Community, March 11-12, 1986
    6P  10P  2 A
    March 11
                             6P  10P  2 A
                                         6A  10A  2P
                                           March 13
                6 A  10A   2P
                    March 12
                    Time  ot Symptom Onset

FIGURE 1—Gastroenteritis, by Time of Symptom Onset, Connecticut Commu
nity, March 11-13, 1986
    Water consumption and symptom histories were avail-
able for 301 persons. Of the 160 persons who drank water, 52
(33 per cent) had gastroenteritis compared to only two (1.4
per cent) of the 141 persons who did not drink water (relative
risk =  23; 95 per cent confidence intervals = 5.7,  92.4).
Information about latent interval was available for 37 persons
and, for those persons, the median latent interval was two
hours  (Figure  2). Only four persons (11 per  cent) had
symptom onsets  <30 minutes after drinking water.  Attack
rates were similar by gender, age, and location of residence.
    Attack rates  by  quantity of tap water consumed for 160
persons for whom  data were available  were:  29 per cent
(33/114), 1-2 glasses; 31 per cent (11/36),  3-4 glasses; and 80
per cent (8/10), >5  glasses. The  calculated ED50 was 2.7
glasses of water (95 per cent CI = 1.8, 4.1). This correspond-
ed to 33.8 mg of fluoride and 20.2 mg of copper if an average
glass of water was 250 cc and contained 50 ppm of fluoride
and 30 ppm of copper.

Skin Irritation

    Of the 300 persons whose skin irritation histories were
obtained, 30 (10 per cent) reported unusual itching, with a
duration of 2 to 62 hours. Of these 30  persons, 12 reported
skin rash compared  to only  two of 270 persons who did not
report itching. No person had a skin rash on the survey day.
                                   9   10   11  12  13  14
                    Latent Interval (Hours)
FIGURE 2	Interval from Water Consumption to Onset of Gastroenteritis,
Connecticut Community, March 11-13, 1986




Rate %



                                                                 °21 persons had unknown skin irritation histories.
    Persons dermally exposed to hyperfluoridated water in a
shower or bath (Table 1) were 2.7 times as likely to have
reported itching than unexposed persons (95 per cent CI = 4,
5.3). Persons who reported itching, but who had not bathed
or showered, had other water exposures such  as dishwash-
ing,  and their itching  was localized to the area of water
                                                                  This hyperfluoridation accident was caused by diversion
                                                              of excess  hydrofluorosilicic acid  into a community  water
                                                              supply. The resultant water acidification substantially  ele-
                                                              vated water copper levels by solubilizing copper compounds
                                                              coating the domestic copper service lines, as demonstrated
                                                              by elevated copper levels in tap water and normal levels in
                                                              water mam water.  In the previous municipal water supply
                                                              outbreak'caused by excess hydrofluorosilicic acid, high iron
                                                              levels resulted from contact with iron pipes.4
                                                                  Although one-third of those who drank the water became
                                                              ill, symptoms were mild and of short duration. The EDso of
                                                              2.7 glasses of water in this outbreak corresponded to symp-
                                                              tom-producing doses of copper and fluoride observed in other
                                                              studies. Copper ingestion causes vomiting usually within 10
                                                              minutes,10 and outbreaks have  occurred  in water supplies
                                                              containing from 4-430 ppm of copper10"13 and with as little as
                                                              a 5 mg ingested dose.10
                                                                  Low-dose fluoride ingestion causes nausea, vomiting,
                                                              abdominal cramping, and diarrhea,''2'4'14 and outbreaks have
                                                              occurred in  water  supplies with levels from  30  to > 1,000
                                                              ppm.1"5 Symptoms occur with a 5 mg ingested dose.15 Both
                                                              copper and fluoride  may have had additive effects in  this
                                                              outbreak; however, the infrequency of vomiting and symp-
                                                              toms occurrences <30 minutes after water ingestion suggest-
                                                              ed fluoride toxicity.
                                                                  Skin irritation symptoms with a hyperfluoridation acci-
                                                              dent have not been reported and, in this outbreak, could have
                                                              been caused  by fluoride, copper, or the water's acidity.
                                                              Although hydrofluorosilicic acid  produces burns in high
                                                              concentrations,16 its cutaneous effects in low concentrations
                                                              are  unknown. Dermally applied copper salts  can produce
                                                              itching and skin rash17 but the cutaneous effects of copper at
                                                              levels recorded in this outbreak are unclear.
                                                                 The authors thank Loraine Good for editorial assistance with manuscript
                                                              preparation and Dr. Robert Gunn for his advice during all phases of the
                                                                                          AJPH June 1988, Vol. 78, No. 6

                                                                                                                    PUBLIC HEALTH BRIEFS
   1. Centers for Disease Control: Acute fluoride poisoning—North Carolina.
   2. Hoffman R, Mann J, Calderone J, Trumbull J, Burkhart M: Acute fluoride
     poisoning  in  a New  Mexico  elementary school.  Pediatrics  1980;
   3. Craun GF: Waterborne outbreaks  in the United  States,  1971-78.  Am
     Water Works Assoc  1980 Annual Conference Proceedings. Denver: Am
     Water Works Assoc, 1980.
   4. Leland DE, Powell KE, Anderson RS:  A fluoride overfeed incident at
     Harbor Springs, Michigan. J Am Water Works Assoc 1980; 72:238-243.
   5. Centers for Disease  Control:  Fluoride intoxication in a dialysis  unit—
     Maryland.  MMWR 1980; 29:134-136.
   6. Cochran WG:  Sampling Techniques. New York: John Wiley, 1977.
   7. Mendenhall W, Ott L, Schaeffer RL: Elementary Survey Sampling. North
     Scituate, MA: Duxbury Press, 1979.
   8. Mantel N: Chi-square tests with one degree of freedom: Extensions of the
     Mantel-Haenzel procedure. J Am Stat Assoc 1963; 58:690-700.
   9. Gad SC, Weil CS: Statistics for lexicologists. In: Hayes AW (ed): Principles
    and Methods of Toxicology. Vol 2. New York: Raven Press, 1984: 273-320.
10.  Gosselin RE, Hodge HC, Smith RP, Gleason MN (eds): Clinical Toxicol-
    ogy of Commercial Products, 5th Ed. Baltimore: Williams and Williams,
11.  Wyllie J: Copper poisoning at a cocktail party. Am J Public Health 1957;
12.  Nicholas PO: Food-poisoning due to copper in the morning tea. Lancet
    1968; 2:40-42.
13.  Hopper SH. Adams, HS: Copper poisoning from vending machines. Public
    Health Rep 1958;  1:910-914.
14.  Semple AB, Parry WH, Phillips DE: Acute copper poisoning: an outbreak
    traced to contaminated  water  from a corroded geyser.  Lancet  1960;
15.  Spoerke DG, Bennett DL, Gullekson D: Toxicity related to acute low dose
    sodium  fluoride ingestions. J Fam Pract 1980; 10:139-140.
16.  Burke WJ, Hoegg VR, Phillips RE: Systemic fluoride poisoning resulting
    from a fluoride skin burn. JOM 1976; 15:39-41.
17.  Clayton GD, Clayton FE (eds): Patty's Industrial Hygiene and Toxicology.
    2nd Ed., Vol II. New York: Wiley and Sons, 1963.
AJPH June 1988, Vol. 78, No. 6

                  Reproduced, with  permission,  from American
                  Journal  of Epidemiology

AMERICAN JOURNAL OF EPIDEMIOLOGY                                                   Vnl  10:! No -1
Copyright© 1976 by The Johns Hopkins University School of Hygiene and Public Health             Printed iii'il.S.A.

                            PUBLIC WATER  SUPPLY


         Weissman, J. B., G. F. Craun, D. N.  Lawrence, R. A. Pollard. M. S. Saslaw
      and E.  J. Gangarosa  (CDC, Atlanta, GA  30333). An epidemic of gastroen-
      teritis traced  to  a contaminated public water supply. Am J Epidemiol 103:
      391-398, 1976.
         Between  January  1 and March  15,  1974, approximately 1200 cases  of
      acute gastrointestinal illness  occurred  in  Richmond Heights, Florida, a resi-
      dential  community of  6500 in south Dade County.  Over  one-third of all fam-
      ilies in the area had at least one member affected. The findings of 10 culture-
      proven  cases of shigellosis among those  who became ill and clinical signs
      and symptoms in the  others suggest that  most of the other cases that were
      not  cultured  may  have  been  shigellosis also.  Epidemiologic  investigation
      showed that consumption of tap water was  associated with  illness in the ini-
      tial  cases of  affected families.  Evaluation of the  Richmond Heights public
      water supply  disclosed numerous inadequacies in both design and operation.
      One of the  wells providing water to the  community was continuously con-
      taminated with excessive  levels of fecal coliforms from a nearby septic tank,
      and a breakdown in  chlorination on January 14-15 caused approximately 1
      million  gallons of inadequately chlorinated water from the  contaminated
      well to be distributed  to the community 48 hours before  the epidemic began.
      Correction of deficiencies in  the water plant was undertaken by the utility
      company; the residents of Richmond  Heights were  instructed to boil their
      drinking water or to  use commercially bottled water pending  completion  of
      corrective measures.  A full scale study is planned for all  similar public water
      supplies in Dade County.

         disease outbreaks;  gastroenteritis; Shigella infections; water  pollution

              INTRODUCTION                 Miami  in south Dade  County,  indicated

  In  March 1974, routine  investigation of   that UP to 1000 clinical cases of  an illness
10  cases  of shigellosis   in  Richmond   characterized by diarrhea  and abdominal
Heights,  Florida,  a residential suburb of   Pain  had  occurred in  residents between
	                                  January 1 and March  15, 1974. Subsequent
  Received for publication May 21, 1975. and in final   studieg developed evidence that  illness re-
form October 1,  1975.                                          *              .
  •Bacterial Diseases Division, Bureau of Epidemi-   suited from  massive contamination ot the
ology.  Center for Disease Control, Atlanta, GA non:!.'!.    area's public water  supply. This  report
  • Water Supply Research Laboratory, Environmen-   describes  the results  on these  investiga-
tal Protection Agency, Cincinnati, OH.              .
  3 Field Services Division, Bureau of Epidemiology,   lions.
Center for Disease Control, Atlanta.
  'Dade  County Department  of  Public  Health,            EPIDEMIOLOGIC METHODS

MThe'auU,o« gratefully acknowledge the assistance     To evaluate the true extent of  disease in
of Robert  Quick, Margaret Pearson. R.N., and .John   Richmond  Heights,   we reviewed county
Black, D.V.M., Dade County Department of Public   records from 1970-1974 for cases of shigel-
Health, Prof. David Hertzig, Department of Mathe-   ,        ,    ,       ,,     j   ,  i
matics, University of Miami, and the staffs of the   losis and subsequently undertook a survey
Perrine and North Miami Health Units.              of 75  households  in  the  area  (approxi-

mately 5 per cent of all households) chosen
by a two-stage cluster sampling technique
  A survey form was developed for use by
public  health  nurses  in  the field. The
name, age, sex, and school or occupation of
everyone in the household was listed. The
respondent was asked about  illness, de-
fined as diarrhea and abdominal pain since
January 1, 1974, in each person. For every-
one who had  been ill, the date of onset,
duration of symptoms, symptom complex,
and  details  of medical  treatment were
obtained. Each family member, ill or well,
was  asked  to estimate  his daily  water
consumption,  indicating whether he drank
tap or bottled water.
   Each  respondent  furnished  additional
information:  whether  a private well was
present on the property, and if so, if it was
used only  for irrigation or as a source of
potable water, and whether there_4nad been
recent septic  tank malfunction. A variety
of questions were asked to uncover possible
common exposures, such  as the source of
 the household's meats and vegetables, res-
 tuarants frequented,  church and commu-
nity  activities, and  contacts with  people
who  had  similar illnesses.  Respondents
with ill  family members  were  also asked
how  they  thought  illness was introduced
into the home.
   Since the  endemic level of diarrheal
diseases in Richmond Heights  was  un-
known and could not be determined relia-
bly due to time  lapse and unusual inter-
vening  events, we selected a nearby com-
munity to serve as a control. After review-
ing a variety of socioeconomic characteris-
tics of approximately 2 dozen communities
in  Dade County, we  chose Bunche Park
about 15 miles (24 km) north of Richmond
Heights,  because its social,  racial, and
economic  profile  most  closely  approx-
imated that of Richmond Heights. A sur-
vey similar to the one performed in Rich-
mond Heights was conducted in Bunche
                              Florida Water and Utilities, Inc.,  the
                            company operating the Richmond Utilities
                            Company, provided records of water usage,
                            termination  of services to consumers be-
                            cause of non-payment of bills, blueprints of
                            the  water distribution system,  and daily
                            operational records. The Dade County De-
                            partment of Public Health  provided  rec-
                            ords of coliform counts on routine water
                            samples obtained  by health department

                                     LABORATORY METHODS
                              Rectal swabs and stool  specimens were
                            processed by the Epidemiologic Services
                            Laboratory  Branch,  Center for  Disease
                            Control  (CDC). Shigella  sonnei  isolates
                            from Dade County were forwarded to CDC
                            for  biochemical confirmation,  antibiotic
                            sensitivity  testing,  and  colicin   typing.
                            Water samples collected during the inves-
                            tigation  were  analyzed  by the  State's
                            Miami Regional Laboratory, at CDC  and
                            at the Water Supply Research Laboratory,
                            Environmental  Protection Agency  (EPA).
                            Water samples were analyzed for total and
                            fecal  coliforms  by the membrane  filter
                            technique; standard plate counts were per-
                            formed on several samples  to provide an
                            estimate of total  bacterial  population.
                            Fluorescein  was used in dye studies, and
                            fluorescence  was  measured  spectro-

                            Epidemiologic  investigations in Dade
                              Review of county records  of  culture-
                            proven cases of Shigella sonnei infection for
                            the  five-year period  1970-1974 disclosed
                            that a statistically significant excess num-
                            ber of isolates were reported in the first 10
                            weeks of 1974 as compared with the four
                            preceding years (p < .05) (table 1). With-
                            out the 10 cases reported from Richmond
                            Heights, however, the total in the first 10
                            weeks of 1974 was within 2 SD  of the mean

                 TARI.F, 1
Coses of shigellnais, Dadp County, Florida 1970-1974
in weeks
   Mean, 1970-1973
for the  same time period during the four
preceding years. Of the 29 reported cases,
those in Richmond Heights occurred signif-
icantly  earlier  than cases  elsewhere  in
Dade County. Seven of 10 cases in Rich-
mond  Heights occurred in  the first five
weeks of the 10-week  period,  compared
with  none of eight cases in communities
adjacent to Richmond Heights (p •=  .004)
or with  three of 16 cases elsewhere in Dade
County (p = .006).
   Shigella sonnei isolates from Richmond
Heights differed in their antibiotic resist-
ance  and colicin-typing patterns from  all
other isolates in Dade  County during the
period January-March 1974. All those from
Richmond Heights patients had an untyp-
able colicin pattern  and were sensitive to
all antibiotics tested. Four patients  from
elsewhere in Dade County  had  untypable
isolates that were antibiotic resistant. Four
additional colicin  patterns  (none were the
pattern of an established  colicin  "type")
were  observed  among six  of  the Dade
County isolates,  but none  was  an isolate
from a patient in Richmond Heights. There
were no shigella isolates from Bunche Park
during  the time period  studied.
   Interviewers .succeeded in contacting 73
of 75  households in Richmond Heights and
70 of 75  in Bunche Park.  In  Richmond
Heights, data were  obtained for 353 per-
sons,  and in Bunche Park for 292. Twenty-
seven of 72 households (37.5 per  cent) in
Richmond Heights had  members who were
ill, compared with 11 of 70 (15.7 per cent)
in Bunche Park (x2 = 7.52, p < .01).
  Sixty-seven of 353 people (19.0 per cent)
in Richmond Heights were ill, compared
with 16 of 292 (5.5 per cent)  in Bunche
Park (p <  .01, t test). Thus, the incidence
of diarrheal disease  in Richmond Heights
was significantly  greater than in a control
community during the time period exam-
ined.  Extrapolation  of these data  to  the
community as a whole would indicate that
1256 (±450)  cases of diarrheal illness  oc-
curred in Richmond  Heights in the first 10
weeks  of  1974  (95  per  cent  confidence
  Analysis of the distribution  of cases in
(he Richmond  Heights survey  population
disclosed  a mean age of patients of 17.2
years,  close to the mean of  20.9 years of
household  members who  were  not  ill;  at-
tack rates were  distributed  with  relative
uniformity throughout all age groups (table
2). The age distribution of cases was also
slightly different  from that of shigellosis in
the United Slates (2).
  Cases of diarrheal disease  in the survey
population  began  to occur in epidemic
numbers  on January  17, 1974,  peaking
about  one month later and rapidly falling
off afterward (figure 1). The lime course of
cases  in the survey population paralleled
that  of the culture-proven cases  of S.
ftonnei in  Richmond Heights. Illness in
both survey populations was characterized
by fever,  nausea, vomiting, and diarrhea;
overall illness  in Richmond Heights  was
characterized by a greater number of spe-
cific symptoms than in either Bunche Park
or among the culture-proven cases (table
3). Symptoms of patients in the Richmond
Heights survey population differed signifi-
cantly from those in Bunche Park patients
in that  there  was  a significantly higher
incidence  of tenesmus, lever,  and vomit-
ing, but  did not differ significantly from
the  culture-proven  cases. In none of  the
symptoms noted was the incidence higher
for patients in  the  Bunche  Park survey

                                        TABLE 2
 Age-specific illness in Dade County survey populations and in households of patients with culture-positive
                                    shigelloais, 1974

70 +

Survey populations
Richmond Heights


(N, .)
Bunche Park


                                 3  10  17
     23 30 I 6  S3  20 27  3  10  17  24
      DEC-      JAN.         FEB
      1973               1974
  FIGURE 1. Gastroenteritis by week of onset. Rich-
 mond Heights survey population, January-March

population than in the Richmond Heights
survey  group.
  There  were no  significant correlations
between illness in the Richmond Heights
survey  population and attendance at spe-
cific schools, churches, or clubs, particular
sources of meats  or vegetables,  specific
                                                              TABLE 3
                                             Symptom complex in survey populations, by per cent,
                                                          Dade County,  1974


Bloody diarrhea
Survey populations
restaurants,  or the presence  of a private
well; 23.9 per cent of the survey households
had private  wells,  generally  for use  with
sprinkler systems  and not  as a source of
potable  water.  Sixty-nine   per  cent  of
households  with  illness knew of  other
households in  which illness had occurred;
only 9.3  per cent  of households without
illness knew of families with  illness (x2 =
25.1, p < .0001). Many people volunteered
that their drinking water had been inter-
mittently  discolored,  "fishy" smelling or
tasting,  and  had  contained particulate
matter. In the  sample survey population as
a whole, illness was  not associated  with
water consumption. To eliminate the am-

biguity caused by households with multi-
ple  cases where some cases  may have
-occurred  by  person-to-person  spread  as
well as by a common source such as drink-
ing  water, the  analysis  of water-specific
attack rates  was recalculated,  this time
comparing water consumption of the initial
cases only with water consumption by the
non-ill members  of the same households.
Eighteen of 49 (36.7 per cent)  who drank
more than four glasses of water daily be-
came  ill; this incidence  was significantly
greater than  15 of 82 (19.5 per cent) who
drank four glasses of water or less daily (x2
 = 4.60, p < .05). If seven households who
drank only bottled water are  also removed
from analysis,  the association of illness
with water consumption increases further
in statistical  significance  (x2 =  5.44, p <
.03). Similar  analyses relating to illness in
Bunche  Park did not  identify water  as
being associated  with the  low level  of
illness there.
         Wafer system evaluation
   At the time of the outbreak, Richmond
Heights was served by a single public water
system operated for the Richmond Utilities
Company by Florida Water and Utilities.
Inc.  (figure 2).  The average  quantity of
water pumped for the time in question was
slightly more than 3790 kl per day. Water
was  pumped  from  several wells on  the
premises of the utility company and came
from  a  relatively shallow aquifer located
less than 15.2  meters below  the surface.
The  principal well  in use was  a well, 6,
drilled  to  a  depth  of 7.3 meters with  a
casing extending clown to 6.1 meters. On
January 31, a new and deeper well, 7, was
put into service and, together with well 6,
supplied  the  needs  of the  Richmond
Heights residents.
  Water from both wells was mixed, chlo-
rinated and distributed. A 379 kl elevated
storage tank  was connected to the system
in such a  manner  that  during  periods of
high usage, water was pumped from the
wells, and  after passing the point of chlo-
rine  addition,  was  distributed  directly
throughout the community.  As the de-
mand in the distribution system slackened,
the water surging from the wells under the
             [P~] PUMP   ® CHECK VALVE

            GAS CHLOR1NATOR
                      FIGURE 2. Richmond Heights water treatment plant.

influence of the pumps was intermittently
diverted into the storage tank. Then, dur-
ing periods of low  usage, the demand was
satisfied entirely by water in storage. Only
when the water level in the elevated tank
dropped to a predetermined  level did the
pumps automatically reactivate.
  During  the investigation,  bacteriologic
testing of untreated water from wells 6 and
7 revealed consistently elevated total coli-
form counts (>50 per ml) in samples from
well 6. Fluorescein dye introduced into the
flush toilet at a neighboring nursery school,
which  had a septic tank located approxi-
mately 38 meters from well 6, was identi-
fied in water from (he well after only nine
hours. Due to the  porosity of the soil, the
cone of inlluencc of the pumps had been
estimated to have a radius of perhaps up lo
400 meters.
  The method of chlorination in use  con-
sisted  of the addition  of chlorine  from a
68  kg  cylinder through  a  booster pump.
The average amount of chlorine used per
day during January.  1974, was  11.3 kg.
This was successful in achieving recorded
chlorine levels at  the pumping station  of
between 1.3 ppm and 3 ppm. Similar read-
ings were recorded for preceding and subse-
quent  months. However, review of pump-
ing station records revealed that on Janu-
ary  15, the plant operator arrived to find
the chlorinator booster pump malfunction-
ing and only 2.3 kg of chlorine used during
the preceding 24 hours. The chlorine resid-
ual recorded at the plant effluent  on that
day was 0. Thus,  it was  estimated  that
approximately 3790 kl of unchlorinated or,
at  best, inadequately  chlorinated water
had been distributed to the community 48
hours before the epidemic began.

  With bacteriologic evidence of continu-
ing fecal contamination of  the Richmond
Heights raw water supply and  epidemio-
logic evidence linking gastroenteritis in the
community with consumption of water, the
                            Dade County Department of Public Health
                            issued  a "boil-water emergency" alert on
                            March 29,  1974. Announcements in news-
                            papers  and on  radio  and  television in-
                            structed residents to boil drinking water.
                            Public  water fountains  in  schools  and
                            churches in the areas were closed. Person-
                            nel at  the  health department offices  were
                            made available to answer public inquiries.
                            The health department also made recom-
                            mendations to assure  the  safety of the
                            water  supply  after several  water system
                            deficiencies were designated. These in-
                            cluded the recommendations that coliform
                            counts in raw waters to be  treated exclu-
                            sively by disinfection should not exceed 20
                            fecal or 100 total coliforms  per 100 ml as
                            measured  by  monthly  arithmetic means
                            (in accordance with EPA guidelines), that
                            provisions  be made to insure continuous
                            chlorination of the raw water source, and
                            that, a chlorine contact  time of at least 30
                            minutes be assured.

                               Our  investigation indicated  that  over
                            1200 individuals  in  Richmond Heights,
                            about  20 per cent of the area's population,
                            became ill in  January  and February  1974.
                            A substantial body of evidence supports
                            the association of water consumption from
                            the  public water supply with  diarrheal
                            disease, and we  have  concluded that the
                            infectious  agent  responsible for this epi-
                            demic was waterborne.
                               No samples of  raw water from well 6, the
                            well supplying most of the community's
                            water  in January, were  obtained during
                            that time. However, multiple samples ob-
                            tained on  several days during our investi-
                            gation in March  were invariably contami-
                            nated with excessive numbers of fecal coli-
                               Moreover, the  recovery of fluorescein dye
                            from well  6 water shortly  after introduc-
                            tion of the dye into  a  nearby sewerage
                            system suggests  that well 6 was continu-
                            ously  contaminated  by sewage.  Conse-

quently, any  interruption in disinfection
would inevitably have resulted in the dis-
tribution  of contaminated water to the
community. On January  14, two days be-
fore the epidemic began,  a malfunction in
the chlorinator  booster pump resulted in
the distribution of inadequately treated
water for  an entire day. Numerous poten-
tial troUblespots, none of which was spe-
cifically implicated in this outbreak, in-
cluded pumping raw water when chlorine
cylinders  were being changed, inadequate
chlorine contact times,  substantial  pres-
sure drops in certain areas of  the distribu-
tion system, and evidence of substantial
water loss through  leaks or illegal connec-
tions; these factors afforded the possibility
of contamination of the system in its distri-
bution as well as its source.
  Ten cases of Shigella  sonnei infection
provided the initial stimulus for this inves-
tigation.  Culturing  the survey population
would  have been  both  impractical  and
unrewarding,  given the elapsed time be-
tween their illness and our investigation;
symptomatology was  used  instead as the
primary  means  of  differential diagnosis.
We cannot prove, therefore, that the  clini-
cal cases were shigellosis;  however, the
concordance of symptoms in the Richmond
Heights survey  population with the symp-
toms in the culture-proven cases (rather
than  with non-shigellosis cases in Bunche
Park) and the high  incidence of tenesmus,
thought to be a  relatively specific indicator
of  shigella infection, suggest  that  most
community illness  was shigellosis rather
than  another acute intestinal illness. Nu-
merous reports  in the literature of sewage
poisoning from contaminated water have
depicted   a  syndrome  characterized  by
shorter duration and milder  symptoms
than  we  observed here (3-5). The possi-
bility that some of the diarrheal illnesses
may  have been caused by etiologic agents
other than shigella cannot be ruled out.
   One year prior to this investigation  CDC,
EPA, and  Bade  County Health Depart-
ment workers  investigated another  epi-
demic  of  waterborne  disease  in Dade
County. In February and March 1973, 210
confirmed cases of typhoid fever occurred
among 1400 residents at the South Dade
Migrant  Farm  Labor  Camp,  in  Home-
stead, Florida  (6). Epidemiologic  investi-
gation at  that time  disclosed a  highly
significant correlation between water con-
sumption and illness. Bacteriologic exami-
nation of the  camp's well water supply
showed  the well  to  have  been intermit-
tently contaminated. The water was chlo-
rinated  before distribution;  however,  an
engineering evaluation of the system's op-
eration  revealed that the  single chlorine
cylinder  had run  out of chlorine and un-
treated water was distributed prior to the
  Deficiencies in the water system  that, led
to the Dade County waterborne  typhoid
epidemic of 1973  were duplicated in the
Dade County  waterborne gastroenteritis
epidemic of 1974: contamination of a raw
water source without, a reliable system for
preventing its  distribution in the  absence
of adequate disinfection, and  an  unsatis-
factory  surveillance  system  to pinpoint
long-standing  potential  hazards before
their translation into real disease.
  The morbidity which may occur from the
waterborne transmission of enteric infec-
tion is difficult to overestimate. Less than
 10  years ago,  contamination of a public
water system  in California  caused  over
 16,000 cases  of salmonellosis  in  a three-
week period (7). This and other outbreaks
 (4-6, 8,  9)  underscore that contaminated
public water is by no  means  a  problem
exclusively of underdeveloped  or economi-
cally disadvantagecl  societies but  rather a
significant problem of contemporary urban

]. Serfling R, Sherman  I: Attribute  Sampling
   Methods  for  Local Health  Departments. PHS
   Publication No 1230, 1965, pp 46-53
2. Center  for Disease Control: Shigella Surveillance

   Report No 36,  First and  Second  Quarters  1974.
   issued February 1975
3.  Lobel HO, Bisno AL, Goldfield M,  et al: A water-
   borne epidemic of gastroenteritis with secondary
   person-to-person  spread.  Am   J  Epidemiol
   89:384-392, 1969
4.  Werner  SB, Jones PH,  McCormack WM,  et  al:
   Gastroenteritis following ingest ion  of sewage-pol-
   luted water: An outbreak at a logging camp on the
   Olympic  Peninsula. Am J Epidemiol  89:277-285,
5.  Gastroenteritis—Illinois. Center for Disease  Con-
   trol: Morbidity  and  Mortality  Weekly Report
   21(23):198-199, June 10, 1972
6. Feldman KK, Maine WB, Nilzkin JI,, et al: Epide-
   miology of Salmiinclla typhi infection in a  migrant
   labor camp in Dade County, Florida. .J Infect  Ois
   130:334  342,  1974
7. A wat.erborne epidemic  of salmonellosis in River-
   side, California, 196.r>. A  collaborative report. Am J
   Epidemiol 93:3:! 48, 1971
8. Gorman AE, Wolman A: Waterborne outbreaks in
   the United States and  Canada and their signifi-
   cance. J Am Water Works Assn 31:225-37;), 1969
9. Weibel SR. Uixon  Fll, Weidner RB, et  al: Water-
   borne disease outbreaks 1946-1960. J  Am Water
   Works Assn 56:947-958,  1964

                       Reproduced,  with  permission,  from Annals
                       of  Internal  Medicine

Chronic Diarrhea Associated with Drinking
Untreated Water
Julie Parsonnet, MD; Susan C. Trock, DVM; Cheryl A. Bopp, MS; Cynthia J. Wood, BA;
David G. Addiss, MD; Frank Alai, MPH; Leo Gorelkin, MD; Nancy Hargrett-Bean, PhD;
Robert A. Gunn,  MD, MPH; and Robert V. Tauxe, MD, MPH
   Purpose: To determine the cause of an outbreak of chron-
 ic diarrhea and to define the clinical profile of the illness.
   Design: A case series of patients with chronic diarrhea
 and case-control and cohort studies to determine the vehicle
 and cause of the illness.
   Setting: Rural Henderson County, Illinois.
   Patients: Seventy-two patients who had onset of chronic
 diarrheal illness between May and August  1987. Controls
 were  local residents and eating companions who did not
 have  diarrheal  illness. A cohort study included  80 truck
 drivers from a local firm.
   Methods and Measurements: Nonbloody diarrhea  was
 characterized by extreme frequency  (median, 12 stools/d),
 marked urgency, fecal incontinence, and weight loss  (mean,
 4.5 kg). The median incubation period was 10 days.  Nine
 patients were hospitalized;  none died. Diarrhea persisted in
 87%  of patients after 6 months. Antimicrobial therapy pro-
 duced no clinical improvement. No bacterial, mycobacterial,
 viral, or parasitic agents known to be enteropathogenic were
 detected in stools or implicated water. Three of five small-bow-
 el biopsies showed mild inflammatory changes. Mild  inflam-
 mation was also seen in two of nine colonic biopsies. Case-con-
 trol studies implicated a local restaurant (P = 0.0001,  odds
 ratio  =19) and  subsequently  the untreated  well  water
 served in the restaurant (P = 0.04, odds ratio = 9.3) as the
 vehicle of transmission.
   Conclusions: This is the first outbreak of chronic diarrhea
 linked to drinking untreated water. The causative agent and
 pathophysiologic mechanism of the illness remain elusive.
In 1984 and  1985, outbreaks of a distinctive chronic
diarrheal illness occurred in Brainerd, Minnesota and
San Antonio,  Texas (1, 2). The illnesses were charac-
terized by dramatic, urgent, watery diarrhea persisting
for many months in  the absence of systemic symp-
toms.  Although  epidemiologic evidence strongly im-
plicated  unpasteurized  milk  as the  vehicle  in  the
Brainerd outbreak, the San Antonio case-patients were
not exposed to unpasteurized milk products. Extensive
clinical and microbiologic  studies revealed no likely
pathogens in  either episode,  and the causative agent
for "Brainerd diarrhea" remains unknown.
   Physicians  throughout the United States have re-
ported sporadic cases of chronic diarrheal illness clini-
cally similar to those seen in the Minnesota and Texas
outbreaks (3), and other smaller clusters have been
recognized (1). Medical evaluation of these patients is
often frustrating. Although some chronic diarrhea cas-
es have been linked with two newly denned  clinicopa-
thologic entities, collagenous colitis (4-7) and micro-
scopic colitis  (8, 9),  the vehicles of transmission and
causative agents are obscure.
   Henderson  County, Illinois is a corn and soybean
farming  community  of 9000 people located on  the
eastern bank of the Mississippi River. In July 1987,  a
physician reported 12 cases of a distinct chronic diar-
rheal illness to the Henderson County Health Depart-
ment. By the middle of August, the health department
had accumulated information on approximately 50 pa-
tients with chronic diarrhea.  Patients had been evalu-
ated by physicians at several local hospitals; no causa-
tive agent was identified, but a local restaurant was
suspected as the  source. This large outbreak provided
an opportunity to further  delineate  the clinical and
pathologic profile of "Brainerd diarrhea" and to define
a new  vehicle  of transmission, water.
Annals of Internal Medicine. 1989;110:985-991.

From the Centers for Disease  Control, Atlanta, Georgia,
and the Illinois Department of Public Health,  Springfield,
Illinois. For current author addresses, see end of text.

A case was defined as diarrhea (three or more loose stools
per day) of unknown cause lasting 4 or more weeks, with
onset after 1 May  1987, in a person who had been in Hen-
derson County after 1 April 1987. A case was considered to
be restaurant-associated if the case-patient had  visited the
implicated restaurant during the 6 weeks before onset of ill-
ness. Cases were identified by direct patient reports to the
Henderson  County Health Department, by calls to local
physicians and hospital infection control  administrators, and
by referral from known case-patients. The Illinois Depart-
ment of Public Health solicited case reports from region-
al and county health departments.  Requests were made
through the newspaper and television  media for persons
with chronic diarrhea to notify the  county health depart-
                                             15 June 1989 • Annals of Internal Medicine
                         Volume 110 • Number 12   985

Table 1. Symptoms of Chronic Diarrhea! Illness in 67 Case-
Patients in Henderson County, Illinois, 1987
Frequency, n(%)
 Fecal incontinence
    53  (79)
    43  (64)
    42  (63)
    30  (45)
    24  (36)
    16  (24)
     8  (12)
     5   (8)
 Clinical Studies

 Sixty-seven case-patients were interviewed regarding symp-
 toms and treatment of their diarrhea! illness. Available med-
 ical records, laboratory results, and pathology specimens
 from 51 case-patients were reviewed. Six months  after the
 initial investigation a follow-up questionnaire was sent to all
 case-patients to assess the duration of illness.
   Thirty-two case-patients submitted stool  specimens; all
 specimens  were submitted at least 4 weeks after the onset of
 illness.  Ten specimens  were  cultured  at the  Centers for
 Disease Control (CDC) for Salmonella species, Shigella spe-
 cies, Campylobacter jejuni/coli,  Yersinia  enterocolitica,
 pathogenic Vibrio species, Aeromonas species, Plesiomonas
 species, Bacillus cereus, and Staphylococcus aureus. Ten col-
 onies selected from the MacConkey agar were identified, ser-
 otyped (if identified as E. coif), and tested by enzyme-linked
 immunosorbent  assay for E. coli enterotoxins (heat-labile
 and heat-stable)  (10, 11). Shiga-like toxins I and  II, local-
 ized and  diffuse adherence,  and invasiveness genes were
 sought with specific DNA probes (12-15). Ten bacterial col-
 onies were also  selected from  blood agar plates and identi-
 fied. The National Veterinary Services Laboratories cultured
 ten case and eight control stools for mycobacterial species,
 including Mycobacterium paratuberculosis (16). Ten stools
 were cultured for yeast. Ova and parasite examination, in-
 cluding examination for Cryptosporidia organisms, Isospora
 belli, and Blastocystis hominis (ten stools), and electronmi-
 croscopy for viral agents (five stools) were also done. Twen-
 ty-two  case-patients and twenty-one asymptomatic controls
 were skin-tested for  Mycobacterium  avium-intracellulare,
 Mycobacterium   kansasii,  Mycobacterium  scrofulaceum,
 Mycobacterium  tuberculosis, Candida albicans,  and Tricho-
 phyton species. Pathologists at CDC and the Baylor Univer-
 sity Medical Center reviewed biopsy slides  obtained from
 local hospitals.

 Epidemiologic and Environmental Studies

 To determine risk factors for chronic diarrheal illness in the
 community, we conducted a case-control study (case-con-
 trol study  1) and a survey. Case-control study 1  included
 case-patients who  were residents of Henderson County
 (n = 23) or the neighboring town of Kirkwood in Warren
 County (n = 5)  and had consulted a  physician for their
 diarrheal illness. One  age- and neighborhood-matched con-
 trol, with no diarrheal illness lasting longer than 1 week in
 the previous 3 months, was systematically selected for each
 case-patient. Truck  drivers employed by an Iowa-based na-
 tional retail company  who frequently passed through  Hen-
 derson County and  had  reported cases of chronic diarrheal
 illness to the health department were surveyed.  In  both the
 case-control study and the survey, data gathered  included
 foods and beverages purchased  (particularly foods and dairy
 products purchased from farms or produce stands), restau-
 rants and food stores  visited, domestic  water supply used,
exposure to animals, history of previous gastrointestinal ill-
ness or  medication use,  and illness among household con-
  After these two studies implicated the restaurant, a sec-
ond  case-control  study and  a  cohort study were done. In
case-control study 2, 30 case-patients who named 65 meal-
companion controls were analyzed in a matched design with
a variable number of controls per case. The  cohort study
included all truck drivers  identified in the survey who had
eaten at the restaurant. Information asked included frequen-
cy, date and time of restaurant visits, specific food and bev-
erages consumed in the restaurant, seating and parking pref-
erences, restroom use, cigarette and alcohol use,  history of
gastrointestinal  illnesses, and use of medications.  All avail-
able  restaurant employees were interviewed.
  The restaurant was inspected by local and state sanitari-
ans,  and water quality reports from the restaurant and local
community were reviewed. Water from the implicated well
was analyzed for pesticides, herbicides, volatile organic com-
pounds, and heavy metals. The Environmental Protection
Agency cultured water for mycobacterial species (17). Viral
cultures of water, human enteroviral gene probe of a water
filter, and examination of water for  ova and parasites were
done at  the University of Arizona (18-20).
  The  statistical  analyses were done  using the  Wilcoxon
rank sum test, the Fisher exact two-tail test, the McNemar
test,  and conditional logistic regression.


Clinical Studies

Seventy-two case-patients were identified; 36 (50%)
case-patients were male. The mean and median age of
case-patients was 56 years (range, 9 to 83 years).  Ill-
ness was  characterized by  sudden  onset of  severe
watery diarrhea; in the initial week of illness, some
case-patients  described  more than 20  stools per day
(median,  12  stools/day). Urgency was  reported by
100% of case-patients  and was severe, often associated
with fecal incontinence (65%).  Weight  loss was re-
ported by 62% at 2 months (mean loss, 4.5 kg). Fever
was  uncommon. The frequency of symptoms  is shown
in Table 1.
  Six months  after the initial investigation, 62 (87%)
of 71  case-patients  who responded to the follow-up
questionnaire  continued to have diarrhea; 52 (73%)
reported having diarrhea every day (mean duration of
illness,  261 days). Frequency of bowel movements in
those with daily diarrhea  had decreased to a mean of
six bowel movements per  day. Complete resolution of
diarrhea occurred in 7 case-patients (mean time until
resolution, 183 days);  these  case-patients were young-
er than those who remained ill (median age,  37 com-
pared with 56, P =  0.01).
  A median incubation period of 10 days  (range, 2 to
40 days) was estimated from 15 case-patients who vis-
ited  the restaurant on  only 1 day. Of 150  case-patient
household members at  risk for secondary transmis-
sion, none developed diarrhea. One Henderson Coun-
ty case-patient frequently visited her adult daughter in
Iowa;  that daughter  developed chronic diarrhea  3
months after the mother's  diarrhea onset; however, be-
cause sporadic cases  of chronic diarrhea occur and
because contact was difficult to quantify, it was  not
possible to definitely  ascribe this  case to secondary
  Of the 72 case-patients,  60 (83%) consulted a phy-
sician. Nine persons, eight of them women, were hos-
pitalized; none died. Laboratory results are outlined in
986  15 June 1989  • Annals of Internal Medicine
      Volume 110 •  Number 12

Table 2.  Clinical Data for 67- Case-Patients in Henderson County, Illinois Compared with 78 Case-Patients in Brainerd,
Minnesota with Chronic Diarrhea*
Laboratory Test
Henderson County
n(%) (n)
Stool culture for Shigella,
Salmonella, Yersinia,
and Campylobacter
Stool for ova and parasites

Fecal leukocytes
Barium enema

Small-bowel series



Large-bowel biopsy

Small-bowel biopsy and
Leukocyte cell count




6 (9)

6 (9)


5 (7)








5 (6)

No pathogens (77) , Salmonella (2),
Campylobacter jejuni ( 1 )

No pathogens (79), Blastocystis hominis (2),
Giardia lamblia ( 1 )
None (4), few (16), moderate (1), many (1)
Normal (22), diverticuli (18), spasm (2),
possible colitis (2)
Normal (17), duodenal diverticulum (1),
rapid transit ( 1 ) , gastric bypass ( 1 )
Normal (29), proctitis (2), erythema (1),
edema (1), punctate hemorrhages (1)
Normal (5), nonspecific patchy erythema
(4), inflammation (3), punctate
hemorrhage (2)
Normal (4), nonspecific mild, chronic
inflammation (13), prominent lymphoid
aggregates (1)
Normal (8), increased histiocytes (1), mild
villous blunting (1), no Giardia (7)
Normal (35), low (1)
Normal (33), low (2), high (2)
  * Other tests performed on fewer than 10 case-patients (n) in Henderson County. All results were normal except where indicated: serum albumin (9;
 3 low), thyroid function tests (6), vasoactive inhibitory peptide (2), gastrin (3; 1 high), 5-hydroxyindole acetic acid (3), serum B12 (3), serum folate
 (3), serum carotene (3; 1 high), qualitative fecal fats (2), quantitative fecal fats (3; 1 high).
 Table 2; no consistent abnormalities to suggest inflam-
 matory bowel disease or malabsorption were observed.
 Therapy included use of numerous antimicrobial and
 antidiarrheal agents (Table 3); none cured the illness.
 Opioid analogs  provided subjective improvement in
 half of those receiving them.
   All bacterial and mycobacterial  stool cultures were
 negative for known pathogens,  and  no viral  agents
 were seen on electronmicroscopic examination of stool
 specimens. Giardia lamblia was identified for one case-
 patient in the first of several stool specimens submitted
 for parasite  examination. Diarrhea in this case-patient
 temporarily  improved with metronidazole therapy but
 worsened several days  later.  A  course  of quinacrine
 hydrochloride also produced temporary  improvement.
 Later examinations for stool parasites  and a  small-
 bowel aspirate did riot  show evidence of Giardia spe-
 cies although the case-patient continued  to have five to
 six stools per day. Two case-patients had B. hominis
 noted in isolated stool samples. Both had  continued
 symptoms despite later ova and parasite examinations
 that  revealed no B. hominis  organisms. Skin-test re-
 sponse to mycobacterial antigens was similar for case-
 patients and controls.
   Twelve patients had sigmoidoscopy or colonoscopy;
 no gross abnormalities were seen.  The  mean interval
 from date of diarrhea  onset to date of biopsy was 43
 days. Local  pathologists reported mild, nonspecific in-
 flammatory  changes in eight of the nine case-patients
 who  had either colonic or rectal  biopsies.  All  slides
 were reviewed at CDC  and the  Baylor  University
 Medical Center,  Dallas. The degree of inflammatory
 infiltrate was believed to fall within the normal range
 for colonic mucosa in all but two biopsies. Specifically,
 there was no evidence for either  microscopic colitis or
collagenous colitis.
  Five case-patients had small-bowel biopsies. No par-
asitic infestation in any specimen and no significant
abnormalities were noted by local pathologists. The
four biopsy specimens reviewed at Baylor University
Medical Center were believed to show a mild increase
in plasma cells and  eosinophils  within the lamina
propria. Similarly, the CDC pathologist reported in-
creased plasma cells and  eosinophils in three of five
biopsies. In all cases, changes  were concluded to  be
mild and nonspecific.

Epidemiologic and Environmental Studies

Attack  rates among Henderson County residents were
as follows: 0 in the group 0 to 4 years, 1.0/1000 in the
group 5 to 17 years, 3.2/1000  in the group 18 to  64
Table 3. Medications Used to Treat Chronic Diarrhea! Ill-
ness in 67 Case-Patients in Henderson County, Illinois, 1987

bismuth salicylate
2 (6)
1 (9)
  * Other medications used by more than one case-patient include van-
comycin (5), erythromycin (3), sulfasalazine (2), cimetidine (2), and
Donnagel (3).
                                              15 June 1989  • Annals of Internal Medicine • Volume 110 • Number 12   987

8 io-
(8  o
o  8
         8 152229 5  121926 3 10172431|7 14
          May         Jun          Jul      Aug
                   Date of Onset
Figure 1. Number of cases of chronic diarrhea by week of onset, in
Henderson County, Illinois, May through August 1987. The date of
closure (7/31) of the implicated restaurant is indicated by the arrow.

years, and 9.2/1000 in group 65 years and older. The
earliest onset dates were in  early May, with a peak in
mid-July, and a precipitous drop in the beginning of
August (Figure 1). The last case-patient had an  onset
on 7 August, 1  week after  the restaurant was closed.
Thirty (42%) of the case-patients resided in Hender-
son County.
  Eating at the  restaurant in the 4 weeks before  onset
of illness was strongly associated with illness in case-
control study 1. Twenty-five (89%) of twenty-eight
case-patients visited the restaurant, and 6 (21%) of 28
neighborhood controls visited the restaurant (odds ra-
tio = 19.0, P < 0.001).
  Sixty-three (79%) of eighty truck drivers respond-
ed to the survey questionnaire and nine cases  (14%)
were identified. Three  truckers described diarrheal ill-
ness but did not supply enough information to catego-
rize  them as case-patients   and they were excluded
from further analysis. One trucker did not respond to
questions concerning restaurants. Nine (20%) of for-
ty-five truckers who ate  at the restaurant and none of
                                                       14 truckers who did  not eat  at the  restaurant  were
                                                       case-patients (relative risk = indeterminate, P = 0.1,
                                                       Fisher exact two-tail test).
                                                         The results of case-control study 2 (study of restau-
                                                       rant-associated risk factors for illness) are outlined in
                                                       Table 4. Case-patients were older than controls  (me-
                                                       dian age of 60.5 years for case-patients and 52 years
                                                       for controls, P —  0.02),  and visited the restaurant
                                                       more frequently than  controls (median visits in  June
                                                       and July, four for  case-patients and two for controls,
                                                       p = 0.5). The only exposure that increased the risk of
                                                       developing illness  was drinking water  at  the restau-
                                                       rant.  Twenty-nine  (97%) of  thirty case-patients re-
                                                       ported  drinking water  at the restaurant, compared
                                                       with 40 (71%) of 56 controls (P =  0.04; 9 controls
                                                       who could not recall whether they drank  water  were
                                                       excluded from analysis). Water remained a risk factor
                                                       for illness when controlled for number of visits to the
                                                       restaurant (P = 0.03) and age (P =  0.06). Drinking
                                                       coffee  prepared with  water heated to 65 °C did not
                                                       increase the risk of illness (odds ratio  = 1.0). The res-
                                                       taurant did not  serve unpasteurized milk products.
                                                         Total restaurant water  consumption  was estimated
                                                       by multiplying  the average number  of glasses  con-
                                                       sumed per visit times the estimated number of visits in
                                                       the 3-month period from May to  July. Case-patients
                                                       drank more water  at the restaurant than did matched
                                                       controls (median, 4.0  glasses compared with  1.8 glass-
                                                       es; P = 0.3). When the one case-patient and 16 con-
                                                       trols who drank no water were excluded from analysis,
                                                       this difference in total water consumption became less
                                                       pronounced (median, 3.0 glasses compared with 2.0
                                                       glasses). No trend  to  suggest a dose-response was ob-
                                                       served even when controlled for age.
                                                          Case-patients and  controls were  similar in  their
                                                       use of antibiotics (Table  4), antacids,  and histamine
                                                         Twenty-four  (53%) of the forty-five truckers who
                                                       had  visited the restaurant participated in the truck-
                                                       driver cohort study. The chronic diarrhea attack rate
                                                       was higher among the older age group (Table 5). A
                                                       high frequency  of  restaurant visits also increased risk
Table 4. Risk Factors for Diarrhea] Illness in the Restaurant Matched Case-Control Study, Henderson County, Illinois
Matched Risk Factor
Mean age
Median number of restaurant visits,
June to July
Drank water
Drank coffee
Previous antibiotic use
Drank water
Visit frequency
Drank water
Drank water
Drank coffee

60. 5y










95% Confidence

0.3 to 1.9

1.2 to 73.8
0.4 to 2.6
0.3 to 8.4

1.3 to 82.0

0.9 to 61.5

1.2 to 81. 2
0.3 to 3.7
P Value*



  * Statistics done using conditional logistic regression.
  t Not applicable.
988   15 June 1989 • Annals of Internal Medicine
                                          Volume 110 • Number 12

Table 5. Risk Factor for Chronic Diarrhea Cohort Study of 24 Truck Drivers in Henderson County, Illinois, 1987
< 40
> 40
Visits to the restaurant, n
< 15
> 15
Glasses of water drunk, J n
> 30





Number of












95% Confidence

1.0 to 46.0

1.2 to 54.5

0.4 to 23.1
0.7 to 31.9

P Valuef



  * Attack rate in case-patients per 100 persons exposed.
  t Fisher exact test, two-tailed.
  | Excludes one person who did not drink water.
  § Chi-square due to linear trend.

 of illness as did drinking a greater total  volume  of
 water (Table  5).
   The restaurant employed  11 people, 8  of whom
 were interviewed. Three employees met the case-pa-
 tient definition. Two additional employees, including
 the  cook-manager, described prolonged diarrheal ill-
 ness but did not meet the case-patient definition. The
 cook-manager had G.  lamblia  in a stool sample and
 reported  cure with metronidazole therapy. The other
 employee described having two to  three loose stools
 per  day only intermittently over the summer months.
   Henderson  County rests  on limestone  through
 which water drains from east to west into the Missis-
 sippi River. Although there are scattered  beef cattle,
 there  are no dairies within the county. Municipal wa-
 ter is  available within the county's towns,  but county
 residents who live in  more rural areas rely on either
 private shallow sandpoint wells or deep wells that pen-
 etrate into the limestone aquifer.
   The restaurant obtained untreated  water from a 4-
 inch drilled well, 125  feet deep, which also supplied a
 neighboring, vacant house. Routine water samples col-
 lected between September 1981 and May 1987 showed
 satisfactory results for total  coliforms (15 samples)
 and total nitrate (16  samples). In June 1987, three
 power failures and two mild earthquakes occurred at
 or near the well. The restaurant continued using water
 during these  power failures. On 30 July, i day before
 the  restaurant was closed, a  water sample collected
 from  the restaurant's  combination soft drink and wa-
 ter  dispenser showed a  high  total   coliform  count
 (>  16/100 mL, tube  dilution method). On 1  August,
 the  well was hyperchlorinated.
   The well had been inspected in April  1987, and the-
 owners were  put under notice  to correct several con-
 struction deficiencies.  At a May reinspection, deficien-
 cies remained that could have permitted  water  from
 contaminated sources to siphon back into the  well,
 particularly from within the vacated house and from
 abandoned, blind pipes that branched from the well
 into a surrounding field. In the 2 months after chlorin-
 ation, 2 of 20 water samples taken from the well or its
 attached plumbing showed a mild elevation  of fecal
 coliforms. One sample, collected on 9  September from
the vacant house, revealed a high fecal coliform count
(2300  organisms/100  mL of water).  Several nearby
wells of varying depths tested satisfactorily.
  Microscopic examination  of water  collected on  9
September revealed no parasites. A water filter was
nonspecifically positive by gene probe for a human en-
terovirus; however, viral cultures of the water grew no
organisms. No pesticides, herbicides, heavy metals,
or volatile organic agents were detected from well wa-
ter collected on 9  September. The restaurant owner
drilled a new well, replaced the restaurant's plumbing
system, and reopened on 8  October.  No  subsequent
illness  has been reported.


This is the first reported outbreak of chronic diarrhea
associated with drinking untreated well  water. The
dramatically severe symptoms in this and the two pre-
viously described  outbreaks  suggest a distinct illness
with a common pathophysiologic mechanism and cau-
sative  agent. Because  an  environmental  exposure ap-
pears  to  cause illness, a standard case definition  is
needed so that responsible vehicles can be identified
and transmission interrupted. We suggest  a case defi-
nition  for "Brainerd diarrhea"  that includes the fol-
lowing criteria:  nonbloody   diarrhea lasting longer
than  1 month with negative stool examination for
known bacterial and parasitic pathogens;  acute onset
of  illness; fecal urgency; absence of fever; eight or
more stools per day throughout the first  week of ill-
ness; normal stool fecal fat or D-xylose  test,  or both;
and normal or mild, chronic  inflammatory changes on
colon  biopsy. The proposed case definition,  though
untested, would  distinguish this chronic  diarrheal
syndrome from mild diarrheal illness that may be het-
erogeneous in cause and from inflammatory bowel dis-
eases,  malabsorptive diseases, and collagenous or mi-
croscopic colitis.
   The causative  agent of outbreak chronic  diarrhea
remains obscure.  Heating water  to 65 °C may have
prevented transmission of illness, however, suggesting
either  a heat-labile toxin or  a microorganism. In the
Minnesota outbreak, a possible case of person-to-per-
                                             15 June 1989 • Annals of Internal Medicine
                                                                                 Volume 110 • Number 12   989

 son transmission suggested an infectious pathogen. We
 were  unable  to confirm this finding in  Henderson
 County. Blastocystis hominis organisms were found in
 the stools of two case-patients; each of these case-pa-
 tients later had negative  stool  examinations  despite
 continuing symptoms. Furthermore,  the clinical pic-
 ture  ascribed to Blastocystis infection—recurrent ab-
 dominal cramps, fatigue, and anorexia with only mild
 diarrhea—is not consistent  with the outbreak illness
 (21,  22). We believe B. hominis infection to be an
 incidental finding  in these two  case-patients,  and no
 other known pathogens were identified.
   Microscopic  colitis and  collagenous colitis, two
 newly defined pathologic entities, both cause  watery,
 urgent, chronic diarrhea (4-8). Although these symp-
 toms are compatible  with  the  outbreak  illness, the
 mild  pathologic  abnormalities  noted on  Henderson
 County biopsy specimens  are not consistent with ei-
 ther  microscopic or collagenous colitis. In microscopic
 colitis, mucosal goblet cell  depletion, excess  neutro-
 phils, round cells within  the  lamina propria, and pene-
 tration of neutrophils into the surface epithelium are
 typically seen  (9).  In  collagenous colitis, the  sub-
 epithelial collagen layer is  thickened,  and lympho-
 cytes and plasma cells infiltrate the lamina propria (7,
 23).  Neither the acute  inflammation of microscopic
 colitis nor the collagen banding of collagenous colitis
 was evident in the Henderson County biopsies. How-
 ever, the time from onset of illness to colon biopsy was
 comparatively brief (mean, 43 days). Collagenous co-
 litis  and microscopic  colitis are generally diagnosed
 histopathologically after more than 1 year  of symp-
 toms. It is conceivable that  the  three clinical  entities
 represent different stages of the same disease. In some
 studies, progression from  microscopic  colitis  to col-
 lagenous colitis has been demonstrated (24, 25). Out-
 break diarrhea may be  the  forme fruste. Sequential
 colonic  biopsies of  Henderson  County patients  are
 planned to assess this possible association.
   Subtle changes  in  small-bowel mucosal  biopsies
 were  also present; however, clinical  studies by local
 physicians  do not  evince  malabsorption.  Three pa-
 tients from the Minnesota  outbreak had abnormal
 small-bowel motility, but again, abnormalities were in-
 constant and nonspecific  (26).  Understanding these
 mild changes would be enhanced by follow-up  studies.
   Although  recall of water  consumption  by  case-
 patients and controls was subject to considerable bias,
 the association between the restaurant's water and ill-
 ness was both strong and plausible. Unlike food items
 that had rapid turnover  in the restaurant, well-water
 usage continued  uninterrupted  throughout the  out-
 break period. Some  restaurant patrons reported that
 the water appeared  filmy  or foamy, or  smelled foul.
 Inspections  of the  restaurant and the well revealed
 plumbing deficiencies that could  have permitted water
 contamination;  a sample taken  from the water dis-
 penser 1 day before the restaurant closed was contami-
 nated with coliforms. Several points along the plumb-
 ing system (for example, filters, the water softener, or
 the combination water-soft-drink dispenser) could
 have  been a source for continuous or sporadic water
contamination. Furthermore, the restaurant continued
to  use water during power failures, enhancing the
opportunity of contamination by back-siphonage. The
unlikely  possibility remains however, that water di-
rectly from the limestone aquifer could  have been re-
   A substantial proportion of sporadic, chronic  diar-
rheal illness may be caused by exposure to an as yet
unidentified environmental agent. Because no diagnos-
tic test for this illness exists, it is important that  both
primary-care physicians and specialists  recognize the
syndrome  clinically  and  inquire about  case clusters
and possible  exposures to untreated  water or unpas-
teurized  milk.  Improved  national  awareness,  early
case recognition, and continued clinical and epidemic-
logic investigation  of both sporadic and outbreak-
related chronic diarrhea  are fundamental  to further
definition of this debilitating disease.
Acknowledgments: The authors thank Ms. Joan Ray and Ms. Judy Kir-
berg of the Henderson County Health Department for their assistance in
the  epidemiologic investigation; Drs. Doris D. Vendrell, Joan B.  Rose,
Gary C. duMoulin, Roger I. M. Glass, and Mr. Jerald Jarnagin for their
laboratory support; and Drs. Mitchell L. Cohen, Paul A. Blake, and
John S. Fordtran for their reviews of the manuscript.
Use of trade names is for identification only and does not imply endorse-
ment by the Public Health Service or by the U.S. Department of Health
and Human Services.
Requests for Reprints: Julie Personnel, MD, EDB/DBD/CID, Room 1-
5428, Centers for Disease Control, Atlanta, GA 30333.
Current Author Addresses: Drs. Parsonnet, Addiss, Hargrett-Bean, and
Tauxe, and Ms. Bopp: Enteric Diseases Branch, Division of Bacterial
Diseases, Center for Infectious Diseases, Centers for Disease Control,
Atlanta, GA 30333.
Drs. Trock and Gunn: Division of Field Services, Epidemiology Pro-
gram Office, Centers for Disease Control, Atlanta, GA 30333.
Ms. Wood and Mr. Alai: Illinois Department of Public Health, 535 West
Jefferson Street, Springfield, IL 62761.
Dr.  Gorelkin: Division of Host Factors, Center for Infectious Diseases,
Centers for  Disease Control, Atlanta, GA 30333.
 1. Osterholm MT, MacDonald KL, White KE, et al. An outbreak of a
   newly recognized chronic diarrhea syndrome associated with raw
   milk consumption. JAMA. 1986;2S6:484-90.
 2. Martin DL, Hoberman LJ. A point source outbreak of chronic diar-
   rhea in Texas: no known exposure to raw milk [Letter]. JAMA.
 3. Read NW,  Krejs GJ, Read MG, Santa Ana CA, Morawski SG,
   Fordtran JS. Chronic diarrhea of unknown origin. Gastroenterolo-
   gy. 1980;78:264-71.
 4. Rams H, Rogers AI, Ghandur-Mnaymneh L. Collagenous colitis.
   Ann Intern Med. 1987;106:108-13.
 5. Wang KK,  Perrault J, Carpenter HA,  Schroeder KW, Tremalne
   WJ. Collagenous colitis: a clinicopathologic correlation. Mayo Clin
   Proc. 1987;62:665-71.
 6. Kingham JG, Levison DA, Morson BC, Dawson AM. Collagenous
   colitis. Gut. 1986;27:570-7.
 7. Jessurun J, Yardley JH,  Giardiello FM, Hamilton SR, Bayless
   TM. Chronic colitis with thickening of the  subepithelial collagen
   layer. Hum Pathol. 1987;18:839-48.
 8. Kingham JG, Levison DA, Ball J, Dawson AM. Microscopic coli-
   tis—a cause of chronic watery diarrhoea. BrMedJ. 1982;285:1601-
 9. Bo-Linn GW, Vendrell DD, Lee E, Fordtran JS. An evaluation of
   the significance of microscopic colitis in  patients with chronic diar-
   rhea. / Clin Invest. 1985;75:1559-69.
10. Svennerholm AM,  Holmgren J. Identification  of Escherichia coli
   heat-labile enterotoxin by  means of a ganglioside immunosorbent
   assay (GM1-ELISA) procedure. CUTT Microbiol. 1978;l:19-23.
11. Lockwood DE, Robertson DC. Development of a competitive en-
   zyme-linked immunosorbent assay (ELISA) for Escherichia coli
   heat-stable  enterotoxin (STa). / Immunol Methods.  1984;75:295-
990  15 June 1989  • Annals of Internal Medicine • Volume 110 •  Number 12

12. Nataro JP, Baldini MM, Kaper JB, Black RE, Bravo N, Levine
   MM. Detection of an adherence factor of enteropathogenic Esche-
   richia coli with a DNA probe. JInfect Dis. 1985;152:560-5.
13. Moseley SL, Oausen CR, Smith AL. A new bacterial adhesin asso-
   ciated with enteritis in  infants [Abstract]. In: 25th Interscience
   Conference on Antimicrobial Agents and Chemotherapy. Washing-
   ton D.C.: American Society for Microbiology; 1985:1128.
14. Newland JW, Neill RJ. DNA probes for Shiga-like toxins I and II
   and for toxin-converting bacteriophages.  / Clin Microbiol.  1988;
IS. Venkatesan M, Buysse JM, Vandendries E, Kopecko DJ. Develop-
   ment and testing of invasion-associated DNA probes for detection of
   Shigella spp. and enteroinvasive Escherichia coli. J Clin Microbiol.
16. National Veterinary Service Laboratories. Laboratory Methods in
    Veterinary Mycobacteriology for the Isolation and Identification of
   Mycobacteria. Ames, Iowa: U.S. Department of Agriculture; 1985.
17. du Moulin GC, Stottmeier KD. Use of cetylpyridinium chloride in
    the decontamination of water for culture of mycobacteria. Appl En-
    viron Microbiol. 1978;36:771-3.
18.  Gerba CP, Margolin A, Trumper B, Rose JB, Zhang C. Low cost
    rapid methods for enterovirus detection in water. In: Water Quality
    Technology  Conference. Advances in Water Analysis  and Treat-
    ment. Denver: American Water Works Association; 1987:417-24.
19.  Rose JB, Madore MS, Riggs JL, Gerba CP. Detection of crypto-
    sporidia and giardia in environmental waters. In: Water Quality
    Technology Conference.  Advances in Water Analysis and Treat-
    ment. Denver: American Water Works Association; 1987:1025-41.
20.  Musial CE, Arrowood MA, Sterling CR, Gerba CP. Detection of
    Cryptosporidia in water using polypropylene cartridge filters. Appl
    Environ Microbiol. 1987;53:687-92.
21.  Shcehan DJ, Raucher BG, McKitrick JC. Association of Blastocys-
    tis hominis with signs  and symptoms of human disease. / Clin Mi-
    crobiol.  1986;24:548-50.
22.  Markell EK, Udkow MP. Blastocystis hominis: pathogen or fellow
    traveler? Am J Trop Med Hyg.  1986;3S:1023-6.
23.  Giardiello FM, Bayless  TM, Jessurun  J, Hamilton SR, Yardley
    JH. Collagenous colitis:  physiologic and histopathologic studies in
    seven patients. Ann Intern Med. 1987;106:46-9.
24.  Teglbjaerg PS, Thaysen  EH, Jensen HH. Development of collage-
    nous colitis in  sequential  biopsy specimens.  Gastroenterology.
25.  Jessurun J, Yardley JH, Lee EL, Vendrell DD, Schiller LR, Ford-
    tran JS. Microscopic and collagenous colitis: different names for the
    same condition? [Letter]. Oastroenterology.  1986;91:1583-4.
26.  Kellow J, Phillips S, Miller L, Osterholm M, MacDonald KL. Ab-
    normalities of motility and absorption  in an outbreak of chronic
    diarrhea [Abstract]. Gastroenterology. 1985;88:1442.
                                                        15 June 1989 • Annals of Internal Medicine  •  Volume 110  •  Number 12    991



Jtl6£L.Lull  reprint
Reproduced,  with  permission,
from  American Journal  of
Public  Health
     Epidemic Giardiasis Caused by  a Contaminated  Public Water Supply
    Abstract: In the period November 1, 1985 to January 31, 1986,
703 cases of giardiasis were reported in Pittsfield, Massachusetts
(population 50,265). The community obtained its  water from two
main reservoirs (A and B) and an auxiliary reservoir (C). Potable
water was chlorinated but not filtered. The incidence of illness
peaked approximately two weeks after the city began obtaining a
major portion of its water from reservoir C, which had not been used
for three years. The attack rate of giardiasis for residents of areas
supplied by reservoir C was 14.3/1000, compared  with 7.0/1000 in
areas that received no water from reservoir C. A case-control study
showed that persons with giardiasis were more likely to be older and
             to have drunk more municipal water than household controls. A
             community telephone survey indicated that over 3,800 people could
             have had diarrhea that might have been caused by Giardia. and 95 per
             cent of households were either using alternate  sources of drinking
             water or boiling municipal water. Environmental studies identified
             Giardia cysts in the water of reservoir C. Cysts were also detected
             in the two other reservoirs supplying the city, but at lower concen-
             trations. This investigation highlights the risk of giardiasis associated
             with unfiltered surface water systems.  (Am J Public Health 1988:
    Waterborne giardiasis is becoming an important public
health problem in the United States: during the period from
1965-84, 90 outbreaks and 23,776 cases of giardiasis were
reported. Sixty-nine per cent of outbreaks and 74 per cent of
cases related to contamination of public water supplies, the
majority of which were surface  water systems with inade-
quate  filtration or chlorination.' We  describe one of the
largest waterborne outbreaks of giardiasis ever reported in
the United States and highlight the  risk of giardiasis in
communities that are supplied by unfiltered surface water
    The outbreak occurred in the western Massachusetts
city of Pittsfield (population, 50,265, 1985 census).  The
problem was first recognized during the first week of Decem-
ber J 985, when the health department of Pittsfield received 70
reports of laboratory-confirmed giardiasis. Only 53  positive
stool  samples for Giardia  had been submitted during the
entire one-year prior to November 1, 1985. Interviews of the
70 case-patients revealed that they had negative stool cul-
tures for bacterial enteric  pathogens, no common personal
contact, and they were widely  distributed  throughout the
city. This information, coupled  with the  knowledge that a
new reservoir had been added to the community water supply
in November, led city and state health officials to suspect a
waterborne outbreak of giardiasis.
Water Supply
    Prior to  November 1985, the community  obtained  its
drinking water from two surface reservoirs, A and B. Water
was chlorinated but not filtered. However, during November
1985, alterations were made to the water system to facilitate
the construction of a filtration system  for water originating
   Address reprint requests to Division of Field Services. Epidemiology
Program Office. Centers for Disease Control, Atlanta, GA 30333. Dr. Kent is
affiliated with the Division of Field Services at CDC; Dr. Greenspan is with the
Division of Field Services at CDC, currently assigned to  the Division of
Communicable Disease Control. Massachusetts Department of Public Health.
Boston; Ms. Herndon is with the Division of Surveillance and Epidemiologic
Studies. EPO. CDC; Drs. Mofenson. Harris and Eng are with the Division of
Communicable Disease Control, Massachusetts Department of Public Health,
Boston: Dr. Waskin is with the Division of Parasitic Diseases. Center for
Infectious Diseases. CDC. This paper, submited to the Journal June  16. 1987,
was revised and accepted for publication August  17, 1987.
Editor's Note:  See also related editorial p 123 this issue.
             from reservoir A. Specifically, on November 5, an auxiliary
             reservoir C was brought on line for the first time in over three
             years to replace water from reservoir A which was to be
             phased  out while the filtration system was being installed.
             This was expected to be a temporary yet necessary step in
             order to meet the usual demand for water, as installation of
             filtration equipment for reservoir A required decreasing  the
             flow of reservoir A water into the city's water system during
             the construction period.  On November 14, as construction
             proceeded, the flow from reservoir C was increased and  the
             flow from reservoir A was correspondingly increased. On this
             day, the water department received complaints of turbid
             water from many areas of the city; the increased turbidity was
             attributed to  reversals in water flow which  suspended sedi-
             ment in the water mains. After approximately two days,  the
             turbidity resolved spontaneously. Reservoir C supplied wa-
             ter to the city until December 21, when it was replaced by
             reservoir A.


             Epidemiologic Investigation
                 Daily reports of positive Giardia isolates  were obtained
             from the two clinical laboratories in the region. The investi-
             gators attempted to reach every reported  case-patient  by
             telephone; patients were asked to report demographic infor-
             mation  and the onset date of symptoms. Local physicians
             were sent letters encouraging them to obtain  stool  analyses
             from  patients  with  symptoms  clinically  consistent with
             giardiasis, and were urged to report positive cases to the local
             health department.
                 Laboratory analysis of stool samples  were performed
             locally  by the major  clinical laboratory  in the area. Stool
             samples were cultured for  Salmonella, Shigella, and Cam-
             pylobacter. Analysis  for ova and parasites was performed
             using standard direct saline and formalin-ethyl acetate meth-
                 A confirmed case of outbreak-associated giardiasis was
             defined as a Pittsfield resident with a stool sample positive for
             Giardia lamblia submitted after November 1, 1985. A suspect
             case of giardiasis was anyone interviewed in the community
             surveys who reported having had diarrhea  (>3 loose stools
             per 24-hour  period) of undetermined etiology of  >5 days
             duration with onset since November I,  1985.
                 Information obtained from the city water engineer con-
AJPH February 1988, Vol. 78, No. 2

cerning placement of values in the municipal water system
was used to construct a water distribution map depicting the
reservoir origin of drinking water and the distribution pattern
throughout the city between November 1, 1985 and January
31, 1986 (Figure  1). We estimated the population in each of
these regions using census tract data obtained in March 1985;
populations  in areas that were included  in more than one
census tract data were divided in proportion to the length of
city streets in each area. Attack rates were then determined
for each water distribution area by counting the number of
reported case-patients residing in each region divided by the
estimated population.

Case-Household Telephone Survey
    Questionnaires were administered by telephone to each
of the first 139 reported case-patients during the first week of
the investigation (December 13-19);  each of these case-
patients was asked to supply information about all household
members. Whenever possible, all household members were
interviewed individually. They were asked to describe their
principal source of drinking water and the number of glasses
of beverages containing both unboiled and boiled municipal
water  they  customarily drank  per  day, as well  as any
gastrointestinal symptoms they may have experienced.
     A case-control study was performed using the survey
data to identify risk factors for giardiasis. Cases were defined
as persons  with  laboratory-confirmed giardiasis;  controls
were defined as persons who did not have  positive stool
samples  and had no diarrhea.  Cases and  controls were
matched within households, and conditional logistic regres-
sion3 was used to determine whether  age  and water con-
sumption were independently associated with giardiasis.

Case Telephone Survey
     Questionnaires similar to those used  in the case-house-
hold survey were administered to the next 100 reported
case-patients.  Additional questions were  included for these
 100  persons specifically describing  their consumption  of
municipal water at home and away from home. Information
about household members was not obtained.
FIGURE 1—Water Distribution by Reservoir Source, Pittsfield, Massachusetts,
November 1 to December 21, 1985
Community Telephone Survey
    A community telephone survey of 220 households ran-
domly selected from the telephone directory was conducted
on January  16,  1986. The  purpose of the survey was  to
determine whether laboratory reports of positive stool sam-
ples were true indicators of the occurrence of disease in the
community.  By this time, physicians may have been treating
an increasing number of symptomatic patients empirically
without having obtained stool samples because of the wide-
spread publicity  associating diarrhea with giardiasis.  Each
household member was to be interviewed concerning water
consumption,  gastrointestinal symptoms,  diagnosis,  and
treatment. Heads  of households were also asked about
compliance with  control measures. The systematic sampling
technique was performed using methods described by Kish,4
and 95% confidence intervals (CI) of estimated numbers  of
cases were determined using  a  method  described  by
Environmental Investigation
    We reviewed all available water quality testing results
indicating turbidity, coliform, and chlorine levels since Jan-
uary 1,  1985. Personnel from city and state  environmental
engineering agencies inspected the water distribution system
for cross-connections with sewage lines.
    Chlorinated  water from reservoir C immediately distal to
the chlorination  point  was sampled for  Giardia cysts on
December 12; untreated water in all three reservoirs  was
sampled on December 16; additional samples were obtained
from three sites in the distal portion of the distribution system
on four  occasions  in late  December and early January.
Sampling for Giardia cysts was  performed  using volume-
sampling orlon yarn filters supplied by the US Environmental
Protection Agency (EPA).6 Analysis of water samples for
Giardia was also  performed by the EPA  using  an im-
munofluorescent antibody technique.7 Giardia cysts extract-
ed from the  water samples  obtained from each of the three
reservoirs were tested for infectivity by inoculation in ger-
bils.8 Giardia trophozoites obtained from these animals were
then examined at the University  of Minnesota by scanning
electron  microscopy for morphological  characteristics  in
order to determine whether they were from human or animal
    Inspectors surveyed the  watershed areas of the  three
reservoirs for signs of  sewage  contamination and beaver
activity in  December  1985; beavers obtained during the
survey were examined for  Giardia. During the summer  of
1986, muskrats were obtained from the watershed areas of all
three reservoirs  and examined for infection with Giardia.
    Eleven outbreak-related human fecal samples were sent
to the University of Minnesota  for use in a cross-species
transmission experiment. Beavers and muskrats were admin-
istered Giardia cysts from these  samples  and examined for
infection. Giardia trophozoites obtained from these animals
were also studied for morphological characteristics.

    From November 1, 1985 to January 31,1986,703 persons
were  identified  with  positive stool samples for Giardia.
During this  same period, approximately 2,600 persons sub-
mitted stool samples, yielding a  Giardia-positive rate of 27
per cent. Onset dates of illness were obtained from 604 (86 per
cent) of the 703 confirmed case-patients, and residential
address was confirmed  for 635 persons (90  per cent). The
greatest number of persons reported onset of symptoms  on
                         AJPH February 1988, Vol. 78, No. 2

                                                                                    WATERBORNE GIARDIASIS EPIDEMIC
                                                              Reservoir C
                                                                                         Boil water order
7    13    19    25
                                                           13    19   25
T"   i  	i~
 12    18    24
                        •604 stool-positive cases. Date of onset unknown for 99 cases.
                  FIGURE 2—Laboratory-confirmed Cases of Giardiasls, by Date of Onset, Pittsfield, Massachusetts, November 1,1985
                  to January 31, 1986
 November 28, two weeks after the flow from reservoir C was
 increased and the episode of turbid water occurred (Figure 2).
 When attack rates of laboratory-confirmed giardiasis were
 calculated by place of residence, the highest attack rates were
 found for areas of the city served by reservoir C (Table 1).

 Case-Household and Case Telephone Surveys
     In the case-household survey of 139 households, infor-
 mation was obtained about all 437 household members. Eight
 additional persons with positive stool samples, who had not
 yet been reported, were identified, yielding 147 confirmed
 case-patients. Of the next 100 case-patients who were inter-
 viewed in the case telephone survey, seven had been inter-
 viewed before in the case-household survey; thus, a total of
 240 confirmed case-patients were interviewed in detail con-
 cerning their symptoms. Among these 240, 236 (98 per cent)
 had diarrhea. The mean duration of diarrhea was 11.3 days,
TABLE 1—Attack Rates by Residential Water Source, Laboratory-con-
         firmed Cases of Giardiasls, Pittsfield, Massachusetts, Novem-
         ber 1,1985 to January 31,1986
Water Source
Reservoir A
Reservoir B
Reservoir C
Mixed (A, B, and C)
Residential water
from reservoir C
Attack Rate
(per 1000)
Attack Rate
(per 1000)
                                 with a range of 1-34 days. Additional gastrointestinal symp-
                                 toms were frequent (Table 2).
                                     Among the 139 households interviewed, we identified
                                 147 persons with laboratory-confirmed giardiasis (cases) and
                                 239 persons who  had neither diarrhea nor positive stool
                                 samples  for  Giardia  (controls).  Seventy-three  per cent
                                 (108/147) of cases reported drinking two or more glasses of
                                 municipal water per day, compared to 36 per cent (86/239) of
                                 controls (OR = 4.9  95% CI = 3.1-8.0). Eighty-eight per cent
                                 (130/147) of cases were age 20 years or older, compared with
                                 52 per cent (124/239)  of controls (OR  = 7.1  95% CI =
                                 3.9-13.3). Conditional logistic regression analysis, used to
                                 match cases and controls by household and to control for age
                                 and water consumption, indicated that drinking two or more
                                 glasses of unboiled municipal water per day and being age 20
                                 years or older were independently associated with giardiasis
                                 [OR (water) = 9.2  95% CI = 4.2-19.9; OR (age) = 3.3  95%
                                 CI =  1.7-6.2].
                                 TABLE 2—Frequency of Symptoms, Laboratory-confirmed Cases of
                                         Giardiasls, Case-Household and Case-Telephone Surveys,
                                         Pittsfield, Massachusetts, November-December 1985
                                                                     Persons with Symptom (%)
                                                              Diarrhea (any duration)
                                                                    (a 5 days)
                                                              Abdominal cramps
                                                              Loss of appetite
                                                              Weight loss
                                                              Fever (undocumented)
                                                              Bloody diarrhea
                                                                            236 (98)
                                                                            190 (79)
                                                                            214 (89)
                                                                            207 (86)
                                                                            197 (82)
                                                                            180 (75)
                                                                            178 (74)
                                                                            165 (69)
                                                                             37 (15)
                                                                              7  (3)
AJPH February 1988, Vol. 78, No. 2

    Among 100 case-patients interviewed  concerning the
number of glasses of unboiled municipal water consumed at
home or away from home, 74 reported that all or most of their
water consumption occurred at home. They reported drink-
ing a mean of 3.2 glasses per day at home, and 0.9 glasses per
day away from home.
Community Telephone Survey
    Of the  220 households selected, 200 were contacted; we
obtained information  on  all  527  persons  living in  these
households. Fourteen persons (2.6 per cent) had  positive
stool  samples for Giardia since November  1  (confirmed
cases); all of these persons had  had  diarrhea for a 5  days
before receiving treatment. Twenty-six persons (5 per cent)
had diarrhea of undetermined etiology with > 5 days duration
(suspect  cases).  Projections from this survey indicated that
between  November  1, 1985  and January  15,  1986,  1,335
people (95% CI = 598-2072) had confirmed giardiasis and
2,480 additional persons (95% CI = 1426-3534) were suspect
cases.  The onset date of illness in  suspect cases resembled
that  of the confirmed  cases, suggesting that many of these
suspect cases could have been  outbreak-associated.  Rela-
tively few cases were reported with onset in January, which
was also consistent with reports from laboratory-confirmed
    Of the 200  households interviewed,  189 (95 per cent)
reported continuous compliance with a boil-water order that
had been issued on December 12; 78 (39 per cent) were boiling
municipal water, and 111 (56 per cent) were either  using
commercial bottled water or were obtaining water from local
springs or wells. All respondents indicated awareness of the
outbreak and the boil-water order.
Environmental Investigation
    Daily records of free residual chlorine levels in water
entering  the distribution  system immediately distal to the
chlorination point indicated that chlorine levels in water from
reservoir C were between 0 and 0.5  mg/1 during the entire
month of November due to a malfunction in the chlorinating
machinery which was not successfully repaired until Novem-
ber 30. During December, chlorine levels in the distribution
system water from reservoir C averaged 2 mg/1 with approx-
imately  15 minutes minimum contact time before water
reached the first households. Chlorine levels in water enter-
ing the distribution system from reservoirs  A and B imme-
diately distal to the chlorination point during both months
averaged 3 mg/1 and 2 mg/1, respectively,  with  15-minute
minimum contact  times.   Water temperature in all  three
reservoirs was approximately 40°F, measured turbidity levels
were within acceptable limits, and pH ranged from 7.0 to 7.4.
    During October 1985, all bacterial cultures of water
samples  from various points in  the  distribution  system
contained levels of coliform bacteria below the  acceptable
limit of 4 colony forming  units (CFU)/100  ml; no bacteria
were detectable in 71 of 80 samples. However, on November
19, five days after the episode of turbid water (observed but
not measured), five of 17 samples had > 5 CFU/100 ml. All
of these five samples were taken from areas of the city that
were at least partially supplied by  water from reservoir C.
Sampling was performed daily for the next seven days; after
two days, all samples were  negative  for coliform bacteria
except for a single site located in the region served exclu-
sively by reservoir C, which continued yielding elevated
coliform  counts  (8^41  CFU per  100 ml) until December 1,
when the defective chlorinator was repaired.
    Microscopic examination of water samples revealed that
chlorinated water from reservoir C obtained immediately
distal to the chlorination point contained 80 Giardia cysts/100
gallons; untreated water obtained directly from reservoir C
contained 28 Giardia cysts/100 gallons. The samples obtained
on December 16 directly from untreated water in  reservoirs
A and  B contained seven cysts and nine cysts/100 gallons,
respectively.  Subsequent water samples taken from distal
points  in the distribution system in December and January
yielded no Giardia cysts. G. lamblia cysts and trophozoites
were detected in the laboratory animals inoculated with the
extract from untreated reservoir C water; animals receiving
portions of water samples obtained from reservoirs A and B
were negative.  Morphological studies suggested that the
trophozoites obtained from reservoir  C did not resemble
those  of beaver origin,   but instead  more resembled
trophozoites  found  in  humans.  Giardia  cysts  that  were
isolated from the outbreak-related human fecal samples and
inoculated in beavers and muskrats produced infection  in
both types of animals and also resembled Giardia typically of
human origin.
    The original source of contamination for reservoir C was
not identified.  Inspectors found no  evidence of sewage
contamination in any of the water mains or in any  of the
watershed  areas of the three reservoirs.  However, they
encountered signs of human  activity (graffiti and empty
beverage containers) near the shoreline of reservoir C  in
officially restricted areas;  they also noted  many signs  of
recent  beaver activity (fresh decorticated wood  cuts) near
reservoir C, and near a small pond that ultimately drained
into reservoir B, but no evidence of beavers near reservoir A.
Nine beavers were found during the December 1985 survey
of the three watershed areas; one of three beavers found near
reservoir C yielded  stool  samples positive for Giardia,
compared to none of six beavers found in the watershed area
of reservoir B. All of seven muskrats obtained during summer
1986 from  watershed areas of all three reservoirs were
infected with Giardia. Thus, the water and animal studies
suggested that beavers and muskrats in the area contributed
to contamination of the water supply and may have originally
been infected from  a  human source, but  they did  not
determine whether the reservoir water had also been directly
contaminated by human sewage.
    Control measures for the outbreak included the follow-
    • A boil-water order was issued on December 12, 1985.
  Pittsfield  residents  were  asked to either boil  municipal
  water or use alternative sources of drinking water.
    • Reservoir C water was removed from the  municipal
  water supply on December 21,  1985.
    • Water in storage  tanks and in the distribution system
  was  hyperchlorinated and flushed in  order to kill  any
  remaining Giardia cysts that may have settled in the water
  mains. On December 21, free residual chlorine levels were
  increased to  3  mg/1, and minimum contact times with
  chlorine were increased to one hour. The entire distribution
  system was then flushed over a nine-day period.
    • The boil-water order was removed after the following
  conditions were met: a)  laboratory reports showed no
  evidence  of significant increase in new illness after two
  incubation periods  (four weeks) following completion  of
  the hyperchlorination and flushing; b) results of the second
  community survey over the same two incubation periods
  were consistent with the data from confirmed cases, show-
  ing no evidence of continuing infection;  c) adequate chlo-
  rination and  contact times  were maintained; d)  water
                         AJPH February 1988, Vol. 78, No. 2

                                                                                      WATERBORNE GIARDIASIS EPIDEMIC
  sampling during January in the distribution system revealed
  no Giardia cysts.
     • Continued  weekly reporting of laboratory-confirmed
  cases of giardiasis was required for at least one month after
  removing the boil-water order.
     The boil-water order was rescinded on January 24,1986,
 six weeks after it was issued. As of April 1, 1986, only one
 case of giardiasis had been reported with onset of symptoms
 after late January.
     It is difficult to maintain appropriate concentrations of
chlorine that prevent waterborne giardiasis; thus, communi-
ties should not rely upon chlorination alone to protect public
water supplies. Proposed  additions or alterations to munic-
ipal water supplies which lack filtration should be thoroughly
evaluated  to ensure that contamination levels do not exceed
the  capacity  of the  chlorination system. Untreated water
should be examined  for  the  presence of fecal coliforms,
turbidity and, if possible,  Giardia cysts. This outbreak was
facilitated by a  malfunction in  the  chlorinating machinery
which left very low levels of chlorine in water from reservoir
C, but even if chlorine levels had been maintained at 2 mg/1
for  15 minutes as intended, viable Giardia cysts might still
have been present;  water  at 40°F and  pH 7.0 probably
requires more than 30 minutes  of contact with 2  mg/1  of
chlorine to adequately kill Gardia cysts.9
     The dramatically  increased  incidence of illness noted
during  the last  few  days  of November suggests  that  a
relatively high dose of Giardia cysts was present in drinking
water during a  short period  of time. Assuming  a mean
incubation period of two weeks,  as is commonly described,
the time course of infection implicates the episode of turbid
water on November 14,1985 as being potentially responsible
for the increased number of cases.  Giardia cysts that had
settled  to the  bottom of water mains  could have been
resuspended by the increased turbulence in water flow. The
result could have been a bolus of higher concentrations of
Giardia  cysts causing more infections.  Additionally, the
increased  turbidity probably reduced the effect of chlorine
which was already  at inadequate  levels.   The increased
concentrations of bacteria in water samples during the week
following the episode of turbid water could have occurred as
a result of inadequate chlorination.
     The validity of calculating  attack  rates  by place  of
residence to implicate a specific water  source depends upon
the assumption that most water consumption occurs at home;
our survey supports this assumption. John Snow, the famous
British  epidemiologist, used this technique effectively  to
implicate the Southwark  and  Vauxhall  Company as the
source of contaminated water  in  his classic cholera investi-
gations of  1853.10
     This is the largest reported  outbreak to  date of labora-
tory-confirmed cases  of waterborne giardiasis (703 cases).
Previous investigations of waterborne giardiasis outbreaks
have described a larger number  of possible  cases and com-
munity-wide attack rates of clinical illness ranging from 3.8
per  cent to 10.6 per  cent.n~15 It is likely that those  who
obtained a laboratory diagnosis of their infection in Pittsfield
represented only a portion of the infected  population:  the
survey indicates 1.9  additional symptomatic suspect cases
for every one laboratory-confirmed case. Additional persons
were likely to have had asymptomatic infection. Estimates of
numbers of infected persons in outbreaks depend upon the
sensitivity and specificity of the case definition  and on the
accuracy and validity of telephone surveys.
     Community health officials were encouraged by the high
degree of reported compliance with  their recommendations
to avoid drinking unboiled municipal water. However, few
people actually boiled tap water; most people  were using
alternate sources of drinking water although all were aware
of the boil-water order.
     It is ironic that this  outbreak occurred as a result of the
attempt to install a filtration system that would probably have
prevented the  outbreak. Many small  communities lack fil-
tered water supplies;  they are costly and require technical
expertise to maintain. In Massachusetts, among  124 surface
water supplies, 76 had either no filtration system or partial
filtration systems in 1985  (personal  communication, John
Higgins, Massachusetts Department of Environmental Qual-
ity Engineering).  Giardiasis  outbreaks may  become even
more frequent in the future, as increasing population pressure
on watershed areas increase  the risk of Giardia  contamina-
tion  for many communities that depend on unfiltered surface
water supplies.
   The authors acknowledge Louis Bolduc and the Pittsfield Health Depart-
ment staff, Dr. George Douglas of Berkshire Medical Center, Tom Keefe of the
Massachusetts Fish and Game Division, Walter Jakubowski of the Environ-
mental Protection Agency, and Dr. Stanley Erlandsen of the University of
Minnesota for their assistance in this investigation.

 1. Craun GF: Waterborne giardiasis in the United States 1965-1984. Lancet
   1986; 2:513-514.
 2. Young KH, Bullock  SL, Melvin DM, Spruill CL: Ethyl acetate as a
   substitute for diethyl ether in the formalin-ether sedimentation technique.
   J Clin Microbiol 1979; 10:852-853.
 3. Breslow NE, Day NE: In: Statistical Methods in Cancer Research,
   Volume I—The Analysis of Case-Control Studies. Lyon: International
   Agency for Research  on Cancer, 1980; 248-281.
 4. Kish L: In: Survey Sampling. New York: John Wiley and Sons, 1965;
 5. Mendenhall W, Ott L, Scheaffer R: In: Elementary Survey Sampling.
   Belmont, California: Duxbury Pres, 1979; 137-139.
 6. Jakubowski W, Chang SL, Erickson TH, et al: Large volume sampling of
   water supplies for micro-organisms. J Am Water Works Assoc 1978;
 7. Sauch JF: Use of immunofluorescence and phase-contrast microscopy for
   detection and  identification of Giardia cysts in water samples. Appl
   Environ Microbiol 1985; 50:1434-1438.
 8. Belosevic M, Faubert GM, MacLean  JD,  Law C, Croll NA: Giardia
   lamblia infections in Mongolian gerbils: an animal model. J Infect Dis 1983;
 9. Jarroll EL, Bingham AK, Meyer EA: Effect of chlorine on Giardia lamblia
   cyst viability. Appl Environ Microbiol 1981; 41:483-487.
10. Snow J: Snow on Cholera. London: Oxford University Press, 1936.
11. Veazie L, Brownlee I, Sears JH: An outbreak of gastroenteritis associated
   with Giardia lamblia. In: Waterborne Transmission of Giardiasis. Cincin-
   nati: US Environmental Protection Agency, 1979; 174-191.
12. Shaw PK, et al: A community-wide outbreak of giardiasis with docu-
   mented transmission by  municipal water.  Ann Intern  Med 1977;
13. Dykes AC, et al: Municipal waterborne giardiasis:  an epidemiologic
   investigation. Ann Intern Med 1980; 92:165-170.
14. Lopez CE, et al: Waterborne giardiasis: a community-wide  outbreak of
   disease and high rate of asymptomatic infection.  Am  J Epidemiol 1980;
15. Navin TR, et al: Case-control study of waterborne giardiasis in Reno,
   Nevada. Am J Epidemiol 1985; 122:269-275.
AJPH February 1988, Vol. 78, No. 2

                                                         Reproduced,  with  permission,  from Annals
                                                         of  Internal  Medicine
 A Waterborne Outbreak of Cryptosporidiosis in Normal  Hosts

                 RICHARD  G.  D'ANTONIO,  M.D.;  RICHARD  E. WIIMN, M.D.;  JEFFERY P.  TAYLOR, M.P.H.; TRACY  L.
                 GUSTAFSON, M.D.;  WILLIAM L.  CURRENT,  Ph.D.;  MAMIE  M.  RHODES, B.S.;  G.  WILLIAM GARY, Jr.,
                 Dr.P.H.; and  ROBERT A.  ZAJAC, M.D.;  Lackland Air  Force Base and Austin, Texas;  Greenfield,
                 Indiana; and  Atlanta,  Georgia
 In July 1984, an outbreak of gastroenteritis occurred in a
 suburban community in Texas. A random telephone
 survey of 100 of 1791 households in the community
 identified an attack rate of 34%. The outbreak was traced
 to contamination of the community water supply, an
 artesian well. Fecal conforms were identified in untreated
 drinking water from the well during July. Stool
 examinations and serologic tests identified
 Cryptosporidium as the etiologic agent. Cryptosporidium
 should be added to the list of waterborne organisms
 capable of causing outbreaks of gastroenteritis.

 CRYPTOSPORIDIUM has been recognized recently as an
 important cause of diarrhea in immunocompromised (1)
 and immunocompetent  persons  (2, 3). Most Cryptospori-
 dium infections are reported to  occur in animal  handlers
 and  patients  with the  acquired immunodeficiency  syn-
 drome. In patients with the  syndrome, this  parasite caus-
 es a  prolonged and  debilitating diarrheal illness (1);  in
 immunocompetent adults, it produces a self-limited diar-
 rheal illness  lasting an average of 5  days (2,  3).  Two
 recent reports described outbreaks of cryptosporidiosis in
 children attending day-care centers (4, 5), and in surveys
 of patients  with diarrhea, prevalence rates as  high as 4%
 have been reported in Australia and Costa Rica (6, 7).
 Infection with this organism may  be more common  than
 previously suspected.
   In July 1984, a waterborne outbreak of diarrhea oc-
 curred  among residents of Braun Station, a  suburb of San
 Antonio, Texas. The outbreak was caused by Cryptospo-

  Stool specimens from Braun Station residents and outpatients
 treated  at the  Wilford Hall Medical  Center,  Lackland Air
 Force Base, in July 1984 were examined for ova and parasites at
 the parasitology laboratory of the medical center or at the Bu-
 reau of  Laboratories,  Texas Department of Health. Fresh and
 formalin-preserved stool  specimens  were  concentrated  by  a
 modification of Ritchie's formalin-ether concentration method
 (8) using ethyl acetate instead of ether. Concentrated  speci-
 mens were examined by light microscopy with iodine stain. To
 identify  Cryptosporidium  oocysts, air-dried, heat-fixed smears
 of fecal  concentrates were stained by  Truant's fluorescent pro-
 cedure (9) and a modified Kinyoun's acid-fast stain (10). Fecal
 films  of specimens preserved in polyvinyl alcohol  were stained
> From the Infectious Diseases Service, Wilford Hall Medical Center, Lackland
Air Force Base, Texas; Bureau of Epidemiology, Texas Department of Health,
Austin, Texas; Animal Health Discovery Research, Eli Lilly Research Laborato-
ry, Greenfield, Indiana; and Viral Gastroenteritis Laboratory, Division  of Viral
Diseases, Centers for Disease Control, Atlanta, Georgia.

886    Annals of Internal Medicine. 1985;103(6pt l):886-888.
by Wheatley's  trichrome method  (11). Stool specimens were
examined for bacterial pathogens including Shigella, Salmonel-
la, Campylobacter, and Yersinia by routine microbiologic tech-
niques. Isolation of viruses was not attempted. Total and fecal
coliform counts in water samples were made by standard proce-
dures at the San Antonio Metropolitan Health District Labora-
tory. Serum specimens from Braun Station residents were tested
for antibody to Norwalk virus by radioimmunoassay (12), for
antibody to Cryptosporidium by an indirect immunofluorescent
technique (13).
  A telephone  survey of community households was done dur-
ing 25 through 27 July 1984. Pages in the residential telephone
directory for San Antonio were selected from a table of random
numbers. The  first  number listed  on each page with a prefix
corresponding to northwest San Antonio was selected.
  Persons contacted  by telephone were asked to verify the
street address,  provide  the number  and names of household
members, their ages and sex, occurrence of any illnesses in May
and July, and week of onset of symptoms. The first person in
each household who was ill with diarrhea or vomiting, or if no
one was ill, the person answering the telephone, was asked addi-
tional questions about symptoms;  duration of illness; dates of
overnight trips; pets;  contact  with farm  animals;  children en-
rolled in day-care centers; recent swimming in pools, lakes, or
rivers; place of employment; recent consumption of raw seafood
or raw milk; source of drinking water; and types of fluids ingest-
ed daily. A total of 100 Braun  Station households and 50 house-
holds outside Braun  Station  were surveyed. In addition, all
Braun Station residents with stool  specimens positive for Cryp-
tosporidium were interviewed with this questionnaire. Statisti-
cal analyses were done with  the chi-square test and Mantel-
Haenszel chi-square test (14).
  Braun Station is a suburban community of approxi-
mately 5900 persons located 20 miles northwest of San
Antonio, Texas. Potable water was supplied to all house-
holds from the same artesian well. Water was not filtered
before use but it was chlorinated with gas shortly before
distribution.  We  surveyed  100 of 1791  households  in
Braun Station for a total of 346  persons. Our investiga-
tion showed two distinct outbreaks of gastroenteritis in
the community between 1 May and 30 July 1984.
  The first outbreak in May was  identified retrospective-
ly. We found that  251  persons (72%)  in 85 households
had  an acute gastrointestinal  illness. Only 2.5% of the
120  persons  living in the 50 households outside Braun
Station had diarrhea or vomiting in  May  (p < 0.00001,
chi-square test).  The dates of onset of symptoms for the
first ill persons in 74 households are  shown in Figure 1.
Dates  of onset were unknown for 11 persons. Only six
serum specimens had been collected in May at the time of

Figure 1. Dates of onset of symp-
toms in the first ill family members
in  100  surveyed  households in
Braun Station, Texas, May and July
                                       1  3  5  7  9 11 13 15 17 19 21 23 25 27 29
                                                                         24 26 26 30 2 4  6  6  10 12 14 16 18 20 22 24 26 26 30
                                                MAY                JUNE                  JULY
                                                       DATE OF ONSET OF SYMPTOMS
the outbreak, but these as well as convalescent sera were
tested for Norwalk virus antibodies. Four specimens had
fourfold rises in titer to Norwalk virus and the remaining
two had seroconverted  from undetectable to low  liters.
Because no sera were available from well persons, further
serologic testing was not done on them. Symptoms had
been similar to those in  other outbreaks of Norwalk gas-
  The second outbreak was investigated prospectively. In
July 1984, 117 persons  (34%) in 60 households  had a
gastrointestinal illness.  Only 2.5% of  persons  residing
outside  Braun Station were ill in July (p = 0.00001, chi-
square  test).  No  geographic  clustering of households
with ill  persons was evident, and the ill persons ranged in
age from  1 to 72 years  with a mean of 32  years. Symp-
toms began during the week of 1 July in 53% and during
the week of 8 July in 20%.  Dates of onset of symptoms
for the first ill persons are presented in Figure 1.
  The symptoms  of persons  in  the  60 Braun Station
households are presented in Table 1. Most had diarrhea
and abdominal cramps that lasted from 1 to 25 days, with
a mean  of 6 days. No one was immunocompromised.
  No association was evident between pet ownership, ex-
posure to  farm animals, or swimming in the community
pool and occurrence  of diarrhea in July.  There was, how-
ever, an inverse association between  the occurrence of
diarrheal illness in July  and taking a vacation between 1
and  11 July. Twenty-one percent of persons who took a
vacation in this period developed a diarrheal illness com-
pared with 72% of those who remained  in Braun Station
(p = 0.00006,  chi-square test). Analysis for each day
during this interval indicated that absence from the com-
munity  on July 2 was  most protective  (p = 0.01, chi-
square test). A linear relation was seen between the pro-
portion of total daily fluid intake of tap water or tap
water  equivalents (reconstituted  frozen juices,  instant
iced tea, and powdered drinks)  and the occurrence of
diarrhea in July, the greater the proportion of daily fluid
intake of tapwater, the higher the attack rates of diarrhea
(p = 0.002, Mantel-Haenszel chi-square test  for trend)
(data not presented).
  Cryptosporidium oocysts were identified in the stools
of 47 of 79 residents who became ill in July. Forty of
these forty-seven persons had diarrhea, 12 had vomiting,
12 had fever, and 2 were asymptomatic. One was hospi-
talized. Only 12 of 194 stool specimens submitted by out-
patients  who did not reside in Braun Station but who
were treated for gastroenteritis at the Wilford Hall Medi-
cal Center contained Cryptosporidium oocysts. Four of
these persons had had person-to-person contact or drank
tap  water  in Braun  Station during July. Twenty-three
stool specimens that contained Cryptosporidium oocysts
were cultured for bacterial pathogens. Salmonella enteri-
ditis was identified in 1 and Giardia lamblia  cysts were
identified in  5. Only  3  of the 32 Braun Station families
with cryptosporidiosis had children in a day-care center.
  Forty-three persons with a stool positive for Crypto-
sporidium  submitted  at least one additional stool. Nine-
teen passed oocysts for at  least 5  days and 9  passed oo-
cysts for at least 20 days. One person continued to pass
oocysts 30 days after his first positive specimen was sub-
mitted. Paired serum specimens  were obtained from 12
residents with positive stool. None showed detectable li-
ters to Cryptosporidium in acute specimens, but 9 devel-
oped liters of 1:40 or greater in convalescenl specimens.
  Unlrealed waler samples from wells in this part of
Texas are nol usually tested. Pretreatment water samples,
however, were examined beginning in mid-June because
                                                                     D'Antonioet al, • Water-borne Cryptosporidiosis

Table 1. Symptoms and Duration of Illness in a Cryptosporidiosis
Outbreak in July 1984 at Braun Station, Texas

                                                111 Persons

Symptoms, %
  Diarrhea                                          100
  Abdominal cramps                                  76
  Nausea                                            45
  Vomiting                                          19
  Fever                                              14
  Headache                                          29
  Muscle aches                                       13
Duration of illness, d
  Mean                                               6
  Range                                           1-25
of community complaints.  Although chlorinated water
samples collected from residences remained negative for
coliforms, nine samples of untreated well water collected
between 2 and 11 July contained fecal coliforms. Samples
collected  on  5  and 6 July contained 2600  and  1300
counts/100 mL, respectively. An order to  boil water be-
fore drinking  was issued on 11  July and remained in ef-
fect until the community was provided with an alternate
water supply  later  that month. Subsequent water tests
showed that dye introduced into the community's sewage
system appeared in the well water.  Attempts to identify
the exact  site of surface water or sewage contamination
were unsuccessful.

   Two outbreaks of gastroenteritis occurred in a single
suburban  community in Texas in the summer of 1984,
and evidence shows that both were caused by contamina-
tion  (probably  intermittent) of the community's water
supply. In both outbreaks, the incidence of diarrheal ill-
ness  in Braun Station residents was at  least 12 times as
high as that in neighboring communities with a different
water supply.
   In the July outbreak, the wide age range  of ill persons,
the  lack of geographic clustering, the strong protection
afforded to persons on vacation during the first week in
July, and analysis of tap water consumption indicate that
the  outbreak was transmitted  through the community
water supply. Coliform tests of unchlorinated water con-
firm  that  the  water supply was contaminated between 2
and  11 July.  Limited laboratory data  indicate that  the
major pathogen in the first outbreak was Norwalk virus,
whereas the major pathogen in the  second  outbreak was
Cryptosporidium, indicating intermittent contamination
of the water supply.
   The typical clinical manifestations produced  by Cryp-
tosporidium in immunocompetent adults are not well de-
fined. Preliminary evidence from outbreaks in children at
day-care centers shows that it can produce a short-lived
diarrheal illness accompanied by  very little vomiting or
fever (4,  5).  The symptoms of children in  the  July  out-
break were generally milder than those of young adults.
Diarrhea was the most prominent  symptom, as  well as
the longest lasting, in adults. Significant volume depletion
occurred in  two persons, one of whom was hospitalized
for fluid therapy. Because of the prolonged nature of the
diarrhea, weight loss occurred in 41% of the ill adults. A
proper estimate of the incidence of asymptomatic cryp-
tosporidiosis could not be calculated, but asymptomatic
Cryptosporidiosis clearly  occurred in some family mem-
bers of households with  proven cases. In some persons
the fecal  shedding of the  organism was prolonged. Sixty-
nine percent of persons tested had stool specimens posi-
tive for Cryptosporidium after their  clinical symptoms
had resolved.
   Our report describes the  largest common-source out-
break  of Cryptosporidiosis  to  date.  The epidemiologic
data show that the source was fecally contaminated wa-
ter. The  results also suggest that  Cryptosporidium, like
Giardia,  may be resistant to the usual levels of chlorina-
tion of potable water supplies.  Cryptosporidium should
be added to the list of waterborne microorganisms capa-
ble of causing gastroenteritis.

ACKNOWLEDGMENTS: The authors thank James Perdue, of the Texas
Department of Health, for assisting in the telephone survey; Kathleen Can-
non and Judy Schwitzer for editorial and secretarial support; the staff of the
San Antonio Metropolitan Health District for laboratory support and assist-
ance in  the telephone survey; and Alan Rebischke for technical assistance.
> Requests  for reprints should be addressed to  Richard E. Winn, M.D.;
SGHMMI,  Wilford Hall Medical Center; Lackland Air Force Base, TX
 1. SOAVE R, DANNER RL, HONIG CL, et al. Cryptosporidiosis in homo-
   sexual men. Ann Intern Med. 1984;100:504-11.
   WEINSTEIN WM.  Human Cryptosporidiosis in immunocompetent and
   immunodeficient persons: studies of an outbreak and experimental trans-
   mission. N EnglJ Med. 1983;308:1252-7.
   MCCARTHY DM. An association between Cryptosporidium and giardia
   in stools  [Letter]. NEnglJMed.  1984;310:788.
 4. ALPERT G, BELL LM,  KIRKPATRICK CE, et al. Cryptosporidiosis in a
   day-care center [Letter]. N EnglJ Med. 1984;311:860-1.
 5. CENTERS FOR DISEASE CONTROL. Cryptosporidiosis among children
   attending day-care centers—Georgia, Pennsylvania, Michigan, Califor-
   nia, New Mexico. MMWR. 1984;33:599-601.
   is in hospital patients with gastroenteritis. Am J Trop Med Hyg.
 7. MATA L, BOLANOS H, PIZARRO D, VIVES M. Cryptosporidiosis in chil-
   dren from some highland Costa Rican rural and urban areas. Am J Trop
   Med Hyg. 1984;33:24-9.
 8. RITCHIE LS. An ether sedimentation technique for routine stool exami-
   nations. Bull US Army Med Dept. 1948;8:326.
 9. TRUANT JP, BRETT  WA, THOMAS W,  Fluorescence microscopy of tu-
   bercle bacilli stained  with auramine and rhodamine. Henry Ford Hosp
   Med Bull. 1962;10:287.
10. PAIK G, SUGG MT. Reagents, stains, and miscellaneous test procedures.
   In: LENNETTE, EH, SPAULDING EH, TRUANT JP, eds. Manual of Clini-
   cal Microbiology. 2nd ed, Washington D.C.: American Society for Mi-
   crobiology;  1974:930-50.
11. WHEATLEY WB. A rapid staining procedure for intestinal amoebae and
   flagellates. Am J Clin Pathol. 1951;21:990-2.
12. GREENBERG HB, WYATT RG,  VALDESUSO J, et al. Solid-phase microti-
   ter radioimmunoassay for detection of the Norwalk strain of acute non-
   bacterial, epidemic gastroenteritis virus and its antibodies. / Med Virol.
13. CAMPBELL PN, CURRENT WL. Demonstration of serum antibodies to
   Cryptosporidium sp.  in normal and immunodeficient humans with con-
   firmed infections. J Clin Microbiol. 1983;18:165-9.
14. SCHLESSELMAN JJ, Case-Control Studies. New York: Oxford University
   Press; 1982.
888    December 1985 • Annals of Internal Medicine • Volume 103 • Number 6 (Part 1)


                                THE NEW ENGLAND JOURNAL OF MEDICINE
                                        May 25, 1989
                                                Reproduced,  with permission,  from the
                                                New  England  Journal  of Medicine
                               A FILTERED PUBLIC WATER SUPPLY

           THOMAS W. McKiNLEY,  M.P.H., GARY S. LOGSDON, D.Sc., P.E., JOAN B. ROSE, PH.D.,
                                 AND DENNIS D. JURANEK, D.V.M., M.Sc.
Abstract  Between January 12 and February 7,1987, an
outbreak of gastroenteritis affected an estimated 13,000
people in a county of 64,900 residents in western Georgia.
Cryptosporidium oocysts were identified in the stools of 58
of 147 patients with gastroenteritis (39 percent) tested dur-
ing the outbreak. Studies for bacterial,  viral, and other
parasitic pathogens failed  to implicate any other agent. In
a random telephone survey, 299 of 489 household  mem-
bers exposed to the public water supply  (61 percent) re-
ported gastrointestinal illness, as compared with 64 of 322
(20 percent) who were not exposed (relative risk, 3.1; 95
percent confidence interval, 2.4 to 3.9). The prevalence of
IgG to cryptosporidium was significantly higher among ex-
posed respondents to the survey who had become ill than
among nonresident controls.

     HUMAN cryptosporidiosis was first described in
      1976 in a three-year-old child with no obvious
immunodeficiency.1 It was subsequently  recognized
as a cause of severe diarrhea in patients with the  ac-
quired immunodeficiency syndrome and found to be a
common cause of sporadic and epidemic gastroenter-
itis in immunocompetent  persons.2"13 Transmission
from calves  to humans, from person to person, and by
a common source have all been described.4'8'9'12 Out-
breaks have occurred in day-care centers,10'11 among
international travelers,3'13 and in a community sup-
plied with unfiltered drinking water from a contami-
nated municipal well.9
  We report the contamination of a filtered  public
  From the Centers for Disease Control (E.B.H., T.D.M., T.R.O., P.P.P.,
M.L.C., E.G.L., E.S.H., D.D.J.) and the Georgia Department of Human Re-
sources (T.W.M.), Atlanta; the Environmental Protection Agency, Cincinnati
(G.S.L.); the University of Arizona, Tucson, Ariz. (J.B.R.); the Uniformed
Services University of the Health Sciences, Bethesda, Md. (B.L.P.U.); and the
Georgia Department of Natural Resources (D.M.W.) and the Morehouse School
of Medicine (M.A.W.), Atlanta. Address reprint requests to Dr. Juranek at the
Centers for Disease Control, Division of Parasitic Diseases (F-12), Atlanta, GA
  Cryptosporidium oocysts were identified in samples of
treated public water with  use of a monoclonal-antibody
test. Although the sand-filtered and  chlorinated  water
system met all regulatory-agency quality standards, sub-
optimal flocculation  and filtration probably allowed the
parasite to pass  into the drinking-water supply.  Low-
level cryptosporidium infection in cattle in the watershed
and  a sewage overflow  were considered as possible
contributors to the contamination  of  the surface-water
  We conclude that current standards for the treatment
of public water supplies may not prevent the contamina-
tion of drinking water by cryptosporidium, with consequent
outbreaks of  cryptosporidiosis.  (N Engl  J  Med  1989;

water supply by cryptosporidium that occurred even
though the treated water met federal and state stand-
ards for drinking water and that resulted in the largest
outbreak of cryptosporidiosis described to date.

  In mid-January 1987  a college physician informed
health authorities of a dramatic increase in gastro-
enteritis among students at West Georgia College in
Carrollton, Georgia. Carrollton is the principal city in
Carroll County (population, 64,900), located in west-
ern Georgia.
  An initial investigation by state and federal epide-
miologists  indicated  that the outbreak involved the
entire community. The laboratory director at Carroll-
ton's hospital reported that four stool specimens ob-
tained from patients with  acute gastroenteritis were
positive for cryptosporidium. The first positive report
had been on January  12.
  An initial case-control study of patients presenting
with gastroenteritis to the  hospital emergency room
and a comparison of gastroenteritis rates at nursing

homes connected to the public water system with rates
at homes independent of the system implicated the
water supply as a possible source of the outbreak. An
advisory to boil water was issued on January 30.

Surveillance and Laboratory Studies

  To define the course of the epidemic, we reviewed emergency-
room logs at the Carrollton hospital and recorded the number of
visits of patients who presented each week with acute gastrointesti-
nal illness from December 1986 through March 1987.
  From January 20 through March  10, stool specimens for ova and
parasite examination were obtained from 147 patients with gastro-
enteritis who presented to outpatient facilities. Seventy-six of these
patients completed questionnaires on their symptoms at the time
of the stool collection. Two aliquots of each stool were preserved
separately  in 10 percent formalin and in polyvinyl alcohol. The
formalin-preserved stool was examined  unstained,  after  iodine
staining, and (after concentration by formalin—ethyl acetate sedi-
mentation) after modified acid-fast and auramine—rhodamine stain-
ing.14 Stool preserved with polyvinyl alcohol was stained with tri-
chrome and examined  by light microscopy. All  stool samples
were examined at the Parasitic Disease Laboratory  of the Cen-
ters for Disease Control (CDC).
  To evaluate the role of bacterial or viral pathogens, we reviewed
the results of all bacterial stool cultures performed at the hospital
laboratory during January, and we collected stool samples from five
acutely ill  college students for examination by immune electron
microscopy for viral particles. In addition, paired serum samples
obtained six weeks apart — during the acute phase of the illness and
during convalescence — from 10 college students were tested for
seroconversion to rotavirus by IgG and IgA enzyme-linked immu-
nosorbent assay  (ELISA). Seven of these paired serum samples
were also analyzed for the presence of antibody to Norwalk virus
with use of the avidin—biotin test. Viral studies were done at the
Gastroenteritis Laboratory of the CDC.

Telephone Survey

  From January 31 to February 2, the Carroll County Health De-
partment surveyed by telephone 304  households  systematically
sampled from the county telephone directory.  Carrollton house-
holds were oversampled to ensure an adequate sample of persons
who had been exposed to the public water supply.
  Adult respondents (over 18 years of age) were asked about the
source of their home water, their age, their sex, the place of employ-
ment of all household members, and whether any household mem-
ber had been ill with abdominal pain or diarrhea since January 1.
They were also asked about their consumption  of specific foods,
their use of restaurants, the amount of tap water they consumed,
and their exposure to children in day-care centers and to farm ani-
mals. Data were obtained on all 304 respondents and 507 additional
household members. County engineers determined the water supply
for given work sites and schools. People whose home,  school, or
work site was supplied with public water were considered  to have
been exposed to the public water supply.
  From March 7 through 12, we were able to obtain serum samples
from 86 of the 159 telephone-survey respondents who were living in
Carrollton and had been exposed to the public water supply. These
samples and randomly selected, banked control serum samples from
20 CDC employees who had not traveled outside the United States
were tested for the presence of IgG and IgM to cryptosporidium by
ELISA15 at the Uniformed Services University of the Health Sci-
ences in Bethesda, Maryland.

Investigation of the Water System

  During the outbreak,  federal and local engineers evaluated the
water-treatment plant in  accordance with published procedures.16
Samples of raw and treated water from various points in the water
system were examined for cryptosporidium and giardia by  passing
379 to 3785 liters (100 to 1000 gal) of water through a l-/nm poly-
propylene filter.  The eluted sediment was examined directly for
giardia, and a fluorescein-tagged monoclonal antibody (Meridian
Diagnostics) was used to detect cryptosporidium in the sediment at
the Parasitic Disease Laboratory of the CDC, according to methods
described by Jakubauski," and at the Department of Microbiology
and Immunology of the University of Arizona, according to meth-
ods described elsewhere.18'19
  To evaluate the means of contamination of the water supply, we
reviewed data on sewage and water maintenance and on rainfall,
and examined fresh stool samples for cryptosporidium from 67  of
the 226 cattle (including 3 calves) pastured in the watershed.

Surveillance and Laboratory Studies
   The number of visits of patients with gastroenteritis
to the hospital emergency room increased in the sec-
ond week of  January, peaked from the  third week
of January through  the first week of February, and
then declined  to preoutbreak levels by tne last week of
February  (Fig.  1).
   Gryptosporidium was identified in the stools of 58
of the 147 patients with gastroenteritis (39 percent)
tested during the outbreak. Giardia was  detected in
stool from only one subject, a two-year-old child who
also had cryptosporidiosis.
   Of the 70 stool samples cultured for bacteria at the
hospital laboratory during January,  only 2 were posi-
tive: 1 for shigella and the other for campylobacter.
The results of all viral studies in the 10 college stu-
dents who became ill were negative.
   Information on signs and symptoms was obtained
from  30  outpatients  with cryptosporidium-positive
stools. These  patients ranged in age from less than  1
year to 46 years (mean age, 20). Diarrhea (defined as
three or more loose bowel movements per day) was
reported by 87 percent, stomach pain by  80 percent,
nausea by 67 percent, vomiting by 33 percent, fever by
30 percent, and muscle aches  by 20 percent.

Telephone Survey
   Of the  811 household members surveyed by tele-
phone, 363 had been ill with  diarrhea or abdominal
pain since January 1.  After adjustment for the over-
sampling of the city of Carrollton,  where the attack
rate was 54 percent,  the overall estimated attack rate
for the county was 40  percent.
   Of the 489  household members who were exposed
to the public water supply, 299 (61 percent) had been
ill, as compared with 64 of the 322 unexposed subjects
(20 percent; relative risk, 3.1; 95 percent confidence
interval,  2.4  to 3.9;  Taylor series approximation).20
When we used logistic regression models to control for
the exposure  to the public water supply,  we  found
no strong or  significant association between illness
and any other risk factor studied (i.e., exposure to day-
care centers, farm  animals, or particular restaurants;
household size; or consumption of chicken or beef).
   Among the 489 exposed  household members, the
attack rate  was 67 percent  (170 of  254)  in  female
and 55 percent (128 of 232) in male household mem-
bers (the sex of 3 subjects was not recorded) (relative
risk, 1.2; 95 percent confidence interval,  1.0  to 1.4).

                                                                                           May 25, 1989
                           i— Advisory to Boil Water
Week Ending
Figure 1. Numbers of Visits of Patients with Gastroenteritis to the
Hospital Emergency Room in Carroll County, Georgia, According
               to the Week of the Visit.
Among the exposed respondents,  the  higher attack
rate in female members remained significant when
we controlled for  age  and water consumption. The
attack rates according to age group among all 489
exposed household members ranged from 52 to 72 per-
cent and  were highest among  those 20 to 29 years
of  age, but  these differences were not statistically
significant (P>0.05 by the  chi-square  test for inde-
   The telephone survey indicated that approximately
50 percent of Carroll County residents, or 32,450 peo-
ple, were exposed  to the public water supply at home
or at work. The risk of illness attributable to this expo-
sure was 41 percent (61 percent of exposed members
who became ill minus 20 percent of unexposed mem-
bers who became ill). We therefore estimate that ap-
proximately 13,000 persons had an illness attributable
to public-water exposure during the outbreak.
  For  the 86  exposed  respondents to the telephone
survey who provided serum samples, the prevalence
of  detectable  IgG to  cryptosporidium was  higher
among the 68 respondents who had become ill than
among the 18  who had not become ill (76 percent vs.
56 percent; P = 0.08 by  the chi-square test), and  it
was significantly higher among respondents who had
become ill than among the  20  CDC employees (76
percent vs. 35 percent; P = 0.001  by the chi-square
test). The seroprevalence of  cryptosporidium-specific
IgM with IgG was 12 percent in respondents who
became ill, 6 percent in those who had not become ill,
and 5 percent in the CDC controls  (P>0.50, by Fish-
er's exact test  for each  comparison).

Investigation of the Water System
  Roughly 7900 of the 19,000 households in the coun-
ty receive water from  the Carrollton water system.
The remainder have wells or are connected to separate
municipal water systems. The Carrollton water sys-
tem draws water from a small river that runs through
surrounding pasture land. The water is treated by the
                      addition of alum, lime, and chlorine, rapid mixing,
                      mechanical flocculation (a process to promote the ag-
                      gregation of particulates),  sedimentation, and rapid
                      sand filtration.
                        Samples obtained from the municipal water system
                      on January 5, 22, and 23 were tested for coliform bac-
                      teria by standard methods21 and were all negative. All
                      routine measurements of treated water were within
                      the limits set by the Environmental Protection Agency
                      and the State of Georgia for turbidity, coliform bacte-
                      ria, and residual free chlorine. Nevertheless, on Janu-
                      ary 28, samples of treated water contained particles as
                      large as  100 pm  (cryptosporidium oocysts are 4  to 6
                      fjim in size).
                        Cryptosporidium oocysts were ultimately identified
                      in samples of treated water taken from the water-
                      treatment plant on January 28, February 4, and Feb-
                      ruary 5 and from four dead-end water mains, includ-
                      ing one at the college, between February 4 and 6. Two
                      samples  of raw  water  obtained  from streams that
                      drain into the river above  the treatment plant were
                      positive for cryptosporidium, but five samples taken
                      from the river itself, upstream  from the treatment
                      plant, were all negative. Three samples of treated wa-
                      ter taken as controls from a nearby town with a sepa-
                      rate water source were also negative.
                        No outbreaks  of  diarrheal illness were reported
                      among cattle pastured in  the county. Cryptosporid-
                      ium  was found in low numbers in stool samples ob-
                      tained from 3 of the 56 cattle tested from pastures
                      along the river upstream from the water plant (2 of the
                      positive stool samples were from calves), but it  was
                      absent in the stool samples  from  11 cattle tested from
                      pastures along the streams  in which cryptosporidium
                      was found. There were no major rainfalls in early Jan-
                      uary, but more than 25 cm (10 in) of rain and snow fell
                      from January 15  to 22.
                        Although no serious breaks in the water or sewage
                      systems were reported during January, a sewage over-
                      flow caused by a blocked major sewer line was discov-
                      ered in mid-February in a wooded area  above the
                      water-treatment plant. When dye was released at the
                      overflow point, it reached  the river and the water-
                      treatment plant in approximately six hours. We could
                      not determine when the sewage pipe had  begun to
                      overflow. An analysis of the surface-water supply for
                      coliform bacteria after the  sewage  spill  had  been
                      cleaned up found no indication of ongoing discharge
                      of raw sewage.
                        We attributed  the passage of particulates through
                      the treatment plant to three factors. First, mechanical
                      agitators  that were  scheduled  for replacement  had
                      been removed from the flocculation basins in Decem-
                      ber in anticipation of the arrival of new agitators. This
                      reduced the efficiency of the flocculation step and im-
                      paired particulate removal. Second,  the efficiency of
                      filtration was impaired by  the equipment and proce-
                      dures used to control the flow of water through the
                      filters and to monitor turbidity. Finally, the filters
                      were sometimes  restarted  without first being back-

Vol. 320  No. 21
washed.  This could have discharged dirt, flocculent
particles, and microorganisms from the filter beds into
the  treated  water.  The  Environmental Protection
Agency has set a turbidity limit for water entering the
distribution system of 1 nephelometric turbidity unit
(NTU) or less, measured at least once daily. The Car-
rollton water system met this regulation; however, tur-
bidity measurements of water obtained from  individ-
ual filters on February 3 showed that three properly
backwashed and restarted  filters were producing wa-
ter with  a turbidity of 0.07 to 0.18 NTU,  whereas
three filters restarted without being backwashed pro-
duced water with a turbidity ranging from 0.2 to 3.2
NTU in  the first three hours after they were restarted.
During the first week of January the  number of filters
restarted without undergoing backwashing increased
to 38 from a usual weekly average of 22, because water
use increased after the holidays.
   The water-treatment process was improved in Feb-
ruary  so that turbidity of continuously monitored
treated water was 0.2 NTU or less — a value consis-
tent with the removal of particles larger than approxi-
mately 1 /u,m. A sample of treated water taken on
February 11  showed no oocysts. The consistently low
turbidity readings resulted in the lifting of the adviso-
ry to boil water on  March 2. Clinical and laboratory
surveillance through the end of May showed no recur-
rence of  cryptosporidiosis in Carroll County.

   To our knowledge, this is the first reported contami-
nation of a filtered water system by cryptosporidium.
The sudden onset of widespread gastroenteritis affect-
ing persons of all ages is typical of a waterborne out-
break. There was a strong association of illness with
exposure to the public water supply among the county
residents surveyed  by  telephone  and a high prev-
alence of cryptosporidium-specific  IgG in  exposed
respondents to the survey who had  been ill.  Crypto-
sporidium oocysts were identified in  the  stools of pa-
tients during the outbreak as well as in the  treated
public water. Neither laboratory examination of pa-
tients nor evaluation of the water system yielded any
evidence that other enteric pathogens contributed to
the outbreak.
  Although there was adequate chlorine  in the water
to inactivate most known bacteria and viruses,22 cryp-
tosporidium oocysts are highly resistant to chlorine.23
The efficacy of boiling water to inactivate cryptospo-
ridium is uncertain, but temperatures above 100°C kill
all other  protozoa that have been tested, and inactiva-
tion of cryptosporidium oocysts by exposure to a moist
heat (55°C) for 20  minutes has been demonstrated.
Although we  have  no information on the residents'
compliance with the advisory to boil water, it appears,
in conjunction with the improvements in water-treat-
ment procedures, to have controlled  the  outbreak.
  Exposed persons were defined as  county residents
with access to the public water supply at home, school,
or work.  Since Carrollton is the commercial center of
                                   the county, it is likely that a number of persons who
                                   were considered to be unexposed in our study were
                                   actually exposed to the  public water supply during
                                   recreational or commercial visits to Carrollton. This
                                   may explain the illness rate of 20 percent among those
                                   considered unexposed in the study.
                                     The higher attack  rate among female residents in
                                   the survey remains unexplained. Surveys in settings
                                   with  no outbreak have found a  higher rate of self-
                                   reported diarrheal illness in women than in men (Ho
                                   MS: unpublished data), but our data do not allow us
                                   to determine whether the sex-based difference we ob-
                                   served was due to a difference in reporting behavior or
                                   susceptibility to symptomatic infection.
                                     Cryptosporidiosis is an illness of variable severity.
                                   Many patients reported  mild, nonspecific symptoms
                                   during the outbreak in Carroll County. Of those with
                                   cryptosporidiosis confirmed by stool culture, 13 per-
                                   cent did not have diarrhea, as defined by three or more
                                   loose stools per day; therefore, the diagnosis of crypto-
                                   sporidiosis should not be excluded on the basis of the
                                   absence of diarrhea.
                                     The  prevalence  of cryptosporidium-specific  IgG
                                   was higher among exposed survey respondents who
                                   had not been ill than among CDC employees (56 per-
                                   cent  vs. 35 percent)  and among a group of Peace
                                   Corps volunteers tested before placement  overseas
                                   (of 75 subjects,  24, or 32 percent, had IgG to crypto-
                                   sporidium).24 Although  not statistically significant,
                                   these differences suggest that  before  the outbreak,
                                   the background exposure to cryptosporidium among
                                   residents of Carroll County was higher than the typi-
                                   cal background exposure among U.S. residents,  or
                                   that  many residents had  asymptomatic infection
                                   during the outbreak. Since the stool samples examined
                                   for the  presence of ova and parasites from the Car-
                                   rollton  area were not routinely examined for  cryp-
                                   tosporidium before November 1986, previous  cases
                                   of  cryptosporidiosis  were not likely to have  been
                                     Infected cattle in the watershed could cause long-
                                   term low-level  contamination of the surface water;
                                   however, we could not conclude that the outbreak was
                                   caused directly by cryptosporidiosis in cattle, since the
                                   level of infection in cattle was low and the distribution
                                   of cattle that tested positive did not match the distri-
                                   bution of positive water  samples.
                                     Regardless  of the source of contamination of the
                                   untreated surface-water  supply,  we believe that the
                                   changes in water treatment around the holiday season
                                   allowed oocysts to pass into the treated water in suffi-
                                   cient numbers to cause widespread illness.  In addi-
                                   tion,  the sewage spill above the treatment plant and
                                   the increase in rainfall just after the onset of the out-
                                   break may have increased the load of cryptosporidium
                                   reaching the plant, but this remains speculative.
                                     Contamination of rivers and streams by cryptospo-
                                   ridium  has  been reported in several states.25 The re-
                                   sults of our investigation demonstrate that cryptospo-
                                   ridium  can contaminate filtered public water systems

                                        THE NEW ENGLAND JOURNAL OF MEDICINE
                                                                                                                           May 25, 1989
— even when  the water quality is within regulatory
limits for coliform bacteria, chlorine, and turbidity —
causing large epidemics of gastroenteritis in otherwise
healthy persons. Further study is needed to assess  the
prevalence and the  effect on public health of crypto-
sporidium  in surface-water supplies and  to evaluate
methods of preventing future contamination of public
drinking water by this hardy  parasite.

   We are indebted to Mr. Olan Ledford for assistance in coordinat-
ing and organizing work involving the Carroll County  Health De-
partment; to Dr. Mary Miles for bringing the outbreak to our atten-
tion and for assistance in  collecting specimens and epidemiologic
data; to Dr. Charles Gerba and the laboratory staff at the Universi-
ty of Arizona for the application of the most recent water-sampling
technology; to Dr. Henry Mathews, Dr. Govinda Visvesvara,  and
Ms. Jennifer Dickerson,  CDC, for examining the stool and water
specimens; to Dr.  Roger Glass and Dr. Mei-Shang Ho, CDC, for
viral studies; to Dr. Bernard Nahlen, CDC, for assisting the investi-
gation; to Mr. Lewis Mason and  Mr. Jim Baxley, Carrollton City
and County  Water Utilities, for assistance in evaluating the water
system; to Mr. Ted Mikalsen, Georgia Department of Natural Re-
sources, for testing streams in the watershed for coliform bacteria; to
Ms. Cheri Toni for help in case  finding and surveillance; to  Ms.
Shirley Johnson for administrative assistance; to Ms. Robbie Green
for notifying  us of the sewage overflow; to Dr.  Keith Sikes,  Dr.
Jeffrey Sacks, and Mr. David Smith,  Georgia Department of Hu-
man Resources, for assisting the investigation; to the staffs  of the
college, the hospital, and  the Carroll County Health Department for
their assistance; and to Ms. Voughn  Trader for assistance  in the
preparation of  the manuscript.
1  Nime FA, Burek JD, Page DL, Holscher MA, Yardley JH. Acute enteroco-
   litis in a human being infected with the protozoan Cryptosporidium. Gasfro-
   enterology 1976; 70:592-8.
2.  Holley  HP Jr, Dover C. Cryptosporidium: a common cause of parasit-
   ic diarrhea in otherwise healthy individuals. J  Infect Dis 1986; 153:365-
   Jokipii  L,  Jokipii AMM. Timing of symptoms and oocyst excretion in
   human  cryptosporidiosis. N  Engl J Med 1986; 315:1643-7.
   Current WL, Reese  NC, Emst JV, Bailey WS, Heyman MB, Weinstein
   WM. Human cryptosporidiosis in immunocompetent and immunodeficient
   persons: studies of an outbreak and experimental transmission. N Engl J
   Med 1983; 308:1252-7.
  5.  Navin TR, Juranek DD. Cryptosporidiosis: clinical, epidemiologic, and
     parasitologic review. Rev Infect Dis 1984; 6:313-27.
  6.  Soave R, Armstrong D. Cryptosporidium and cryptosporidiosis. Rev Infect
     Dis 1986; 8:1012-23.
  7.  Payer R, Ungar BL. Cryptosporidium spp. and cryptosporidiosis. Microbiol
     Rev 1986; 50:458-83.
  8.  Human cryptosporidiosis — Alabama. MMWR 1982; 31:252-4.
  9.  D'Antonio RG, Winn RE,  Taylor JP,  et al. A waterborne outbreak of
     cryptosporidiosis in normal hosts. Ann Intern Med 1985; 103:886-8.
 10.  Alpert G, Bell LM, Kirkpatrick CE, et al. Outbreak of cryptosporidiosis in a
     day-care center. Pediatrics 1986; 77:152-7.
 11.  Cryptosporidiosis among children attending day-care centers — Georgia,
     Pennsylvania, Michigan, California, New Mexico. MMWR 1984; 33:599-
 12.  Koch KL, Phillips DJ, Aber RC, Current WL. Cryptosporidiosis in hospital
     personnel: evidence for  person-to-person transmission. Ann Intern Med
     1985; 102:593-6.
 13.  Jokipii LrPohjola S, Jokipii AM. Cryptosporidiosis associated with travel-
     ing and giardiasis. Gastroenterology 1985; 89:838-42.
 14.  Ma P, Soave R. Three-step  stool examination for cryptosporidiosis in 10
     homosexual men with  protracted watery diarrhea. J Infect  Dis  1983;
 15.  Ungar BL, Soave R, Payer R, Nash TE. Enzyme immunoassay detection of
     immunoglobulin M and G antibodies to Cryptosporidium in immunocompe-
     tent and immunocompromised persons. J Infect Dis 1986; 153:570-8.
 16.  Logsdon GS.  Evaluating treatment plants for paniculate contaminant re-
     moval. J Am Water Works Assoc 1987;  79(9):82-92.
 17.  Jakubauski W. Detection ofGiardia cysts in drinking water: state of the art.
     In:  Erlandsen SL, Mayer EA, eds. Giardia and giardiasis: biology, patho-
     genesis and epidemiology. New York: Plenum Press, 1984:263-85.
 18.  Rose JB. Detection of Cryptosporidium from wastewater and  freshwater
     environments. Water Sci Technol 1986;  18:233.
 19.  Musial CE, Arrowood MJ, Sterling CR, Gerba CP. Detection of Cryptospo-
     ridium in water by using polypropylene cartridge filters. Appl Environ Mi-
     crobiol 1987; 53:687-92.
20.  Kleinbaum DO, Kupper LL, Morgenstem H. Epidemiologic research: prin-
     ciples and quantitative methods. Belmont, Calif.: Lifetime Learning Publi-
     cations, 1982:296-9.
21.  American Public Health Association, American Water Works Association,
     Water Pollution  Control  Federation, Georgia Department of Natural Re-
     sources. Standard methods for the examination of water and wastewater.
     16th ed. New York: American Public Health Association, 1985.
22.  Laubusch EJ. Chlorination and other disinfection processes.  In:  American
     Water Works  Association. Water quality and treatment: a handbook of
     public water supplies. 3rd ed. New York: McGraw-Hill, 1971:201.
23.  Campbell I, Tzipori AS, Hutchison G, Angus KW. Effect of disinfectants
     on survival of Cryptosporidium oocysts.  Vet Rec 1982; 11:414-5.
24.  Ungar BLP, Mulligan M, Nutman TB. Serologic evidence of Cryptosporid-
     ium infection in  US volunteers before and during Peace Corps  service in
     Africa. Arch Intern Med  1989; 149:894-7.
25.   Rose JB. Occurrence and significance of Cryptosporidium in water. J Am
     Water Works Assoc 1988; 80(2):53-8.

              Reproduced,  with  permission,  from  the  Lancet
                                                                                         THE LANCET,DECEMBER24/31,1988
                                                                   of free chlorine following booster chlorination procedures are not
                                                                   oocysticidal.5 The waterborne route should be considered when
                                                                   clusters of cryptosporidiosis associated with potable water occur,
                                                                   even in the absence of bacterial or viral contamination.
  SIR,—Despite the fact that Cryptosporidium spp have animal
reservoir  hosts  and  are  transmitted by  the faecal-oral  route,
Cryptosporidium, unlike Giardia  duodenalis, has  not yet been
acknowledged  widely  as  a significant  cause of waterborne
gastroenteritis. In  1985, D'Antonio  et  al1  described  the  first
waterborne  outbreak,  in   Texas.  Rush   et  al2   reported
Cryptosporidiian spp  oocysts in raw drinking water samples and
postulated waterborne spread to account  for a peak  in cases of
cryptosporidiosis in  the  Sheffield  area  in 1986.  Hayes et al3
described a waterborne outbreak in Georgia, USA, affecting 13 000
people exposed to a sand-filtered, chlorinated public water supply.
  During  April, 1988,  several cases of cryptosporidiosis were
reported  to  the community  medicine  specialist  in  Ayrshire.
Extensive  epidemiological  investigations  failed to implicate  a
common source of food or milk or consistent history of animal
contact. Mapping of residences of the affected cases revealed that
many of the households concerned received the same public potable
water supply. 27 patients either resident in the area  served by that
supply  or who  consumed that  water  were  diagnosed  as
Cryptosporidium spp positive (age range 4 months to 93 years), of
whom 12 were admitted to hospital (11 required intravenous fluid
  Following notification of the possible involvement of the water
supply and the introduction of booster chlorination procedures, the
water supply system was thoroughly examined. Twice during the
outbreak Esclierichia  coli, coliforms, and faecal streptococci were
isolatedj from a service tank at the treatment works, but all other
samples proved negative bacteriologically. Cryptosporidium spp
oocysts were detected by Wright-Giemsa, modified Ziehl-Neelsen,
saffranin/methylene-blue,   phenol-auramine,  and  monoclonal
antibody staining in the treated, chlorinated water supply system.
Oocysts were present in sludge samples from the treatment works,
in the final  water leaving one of  the service tanks, and in  a
break-pressure tank delivering chlorinated final water to the supply
(range 0-13 to 1000 oocysts/1). Reinspection of the break-pressure
tank revealed an old 50 cm fireclay pipe discharging into it; the pipe
collecting run-off from the surrounding area. Water samples from
the  pipe  and  a nearby  stream  showed evidence  of faecal
contamination (244 E coli, 403 coliforms, 58 faecal streptococci per
100ml, and 107 Ecoli, 152 coliforms, 34 faecal streptococci, per 100
ml, respectively).  Samples from the  stream  and soil and grass
adjacent to the fireclay pipe were positive for oocysts (0-13 oocyst/1
and  32 oocysts/g, respectively).  Oocysts  were also detected in
samples of the treated supply in the absence of faecal bacterial
indicators. Following the isolation, drainage and disinfection of the
contaminated storage  tanks, extensive   mains  flushing,  and
emergency booster  chlorination,  no  new cases  of waterborne
cryptosporidiosis have  been reported, and  bacteriological and
parasitological analyses of subsequent water samples have proved
  Oocysts appear to have been introduced via the fireclay pipe into
the break-pressure tank which contained final water for distribution
rather than by failure of the water treatment process to retain them.
There was evidence that cattle slurry had been sprayed on land in
the vicinity of the fireclay pipe before the outbreak. Cryptosporidiian
spp oocysts can remain viable for at least 12 months at 4°C5 and
oocyst  contamination  of  final  waters   in  the  absence   of
bacteriological contamination has been reported.1 Attainable levels
Scottish Parasite Diagnostic Laboratory,
Department of Bacteriology,
Stobhill General Hospital,
Glasgow G21 3UW

Community Medicine Department,
Ayrshire Central Hospital, Irvine

Area Laboratory,
Crosshouse Hospital, Kilmamock

Srrathclyde Water Department,

Environmental Health Department,
Cunninghame District Council, Irvine

Communicable Diseases (Scotland) Unit,
Ruchill Hospital, Glasgow

Scottish Home and Health Department,




]. C. M.  SHARP

 I. D'Antonio RG,  Winn  RE, Taylor JP,  et al. A  waterborne  outbreak of
     cryptosporidiosis in normal hosts. Aim Intern Med 1985; 103:886-88.
 2. Rush BA, Chapman PA, Ineson RW. Crvplosporidiitin and drinking water. Lancet
     1987; ii: 632.
 3. Hayes EB, Matte TD, O'Brien TR, et al. Contamination of a conventionally treated
     filtered public water supply by Cryptesporiditttn associated with a large community
     outbreak of cryptosporidiosis. N Eiigl J Mcd (in press).
 4. Anon. The bacteriological examination of drinking water supplies. London: HM
     Stationery Office, 1983.
 5. Smith AL, Smith HV, Girdwood RWA, Carrington E. The effect of free chlorine on
     the viability  of Cryptosporidium  spp  oocysts.  WRc Publ  PRU 2023-M.
     Medmeacham, Bucks: Water Research Centre, 1988.

                                                           Reprinted from Journal AWWA,  Vol. 73,
                                                           No.  1,  by permission.  Copyright   1981,
                                                           American Water Works Association.
         What Happens  When the Multiple Barrier
         Concept  is  Neglected
  Waterborne disease:  occurrence  is on
  the upswing
Edwin C. Lippy
Three of the waterborne disease outbreaks in 1979 typified problems common to
many water systems. Investigation of the outbreaks identified as causative factors
deficiencies in water treatment practices and plant operation and a breakdown in
surveillance activities designed to protect public health. Preliminary information
indicates that a total of 45 outbreaks occurred during 1979, increasing the annual
average of outbreaks to  36  for the  four-year  period from 1976  to  1979. This
represents a 50 percent increase over the 1971-75 average of 24. The five-year
averages have steadily increased from an annual average of ten during 1951-55.
Before that period the trend was declining.
  The disturbing trend toward more fre-
quent occurrence of outbreaks of water-
borne disease should serve as a warning
to all who share in the responsibility for
the delivery of a safe and potable water.
The  "multiple  barrier"  concept   that
relies  on placing  protective  systems
between the water consumer and actual
as well as potential sources of contami-
nation  should  be  emphasized,  with
appropriate  consideration  for  natural
features (distance, dilution, geologic fac-
tors),  man-made  facilities  (adequate
waste treatment, water treatment, opera-
tional considerations), and conscientious
surveillance  by   regulatory   agencies

(monitoring,  inspection,  certification).
Investigation of three of  the outbreaks
that occurred in 1979 illustrates what can
happen  when  these  barriers  are  ne-
glected. The outbreaks occurred in Brad-
ford, Pa.; a state park near Lake Havasu
City, Ariz., and  a  church  camp  near
Brevard, N.C.

Bradford, Pa.
  Bradford  is located in northwestern
Pennsylvania  about  210  km (130  mi)
northeast of Pittsburgh and some 97 km
(60 mi) south of Buffalo, N.Y. It is situ-
ated  in the midst of the  Allegheny  oil
region and  is an industrial  center for
producing and refining crude oil.
  The water system serves a population
of  approximately  18 000.  Three  im-
poundments located within a few miles
of the city  serve as sources of supply.
The impoundments and watersheds are
wholly owned and protected by the city,
and recreational use is prohibited. Chlo-
rination  is the only treatment provided;
gas injection facilities  are located imme-
diately downstream from each impound-
ment. Transmission mains convey water
from the impoundments to the distribu-
tion system, where  the supplies con-
verge. A number of service connections
draw off each transmission main prior to
their convergence.
  An open finished water  reservoir of
13-ML (3.5-mil  gal)  capacity  provides
storage  and regulates pressure in the
distribution system. Chlorination is not
provided at this reservoir. Data for each
raw water  impoundment  are  given in
Table 1.
  The outbreak begins. A waterborne out-
break was suspected  when fifteen resi-
dents reported within one week to the

                       E.C. LIPPY  57

Bradford Hospital for treatment of diar-
rheal  illness. When tested, ten of  the
fifteen were  stool-positive for Giardia
cysts.  Newspaper reports implicated  the
water supply and advised residents to
boil their water. The advice was based
upon  reporters'  discussions with  state
and city  health officials and the Center
for Disease Control (CDC) in Atlanta,  Ga.
The suspicion that the illness was related
to the water supply probably arose from
recent water quality problems.
  Heavy rain in the Bradford area in July
and  August  contributed  to  turbidity
problems and resulted in numerous citi-
zens' complaints of discolored and mud-
dy water. This was compounded by pub-
lic notification of failure to comply with
the coliform maximum contaminant lev-
el (MCL) in August and September. The
city was not required to report violations
of the turbidity MCL because it had been
granted a waiver by the  Environmental
Protection Agency (EPA). Records indi-
cated that the reservoir at Marilla Brook
had turbidity values in excess of 10 NTU
on numerous occasions.
  The bacteriological record  appeared
acceptable until August  when  three of
twenty samples collected were  reported
to be  positive for coliform bacteria (Ta-
ble 2). Closer inspection of the record,
however, showed that extensive  time
had elapsed between sample collection
and analysis, casting doubt on the valid-
ity of al! the results prior to the use of
local laboratories in September.
  Determining the cause. The  Meadville
regional   office   of  the  Pennsylvania
Department of Environmental Resources
invited EPA to assist in the investigation
of the water system to isolate Giardia
cysts from the water, identify the source
of contamination, and propose correc-
tive measures. Also, a team of epidemio-
logical investigators was dispatched by
the Pennsylvania Department of Health
and was  soon joined by a team  from  the
Center for Disease Control.
  Within ten days after the investigation
was begun, cysts were isolated from  the
water system. A beaver trapped from a
large lodge in the Gilbert Run reservoir
was sacrificed for necropsy and a sample
from the colon showed an abundance of
Giardia cysts. At this point in the inves-
tigation the  number  of confirmed cases
of giardiasis at  the  Bradford  Hospital
had increased to 180.
  Precautionary  measures  taken at  the
outset of the investigation included a
recommendation  that consumers boil
water for one  minute  and that plant
operators increase chlorination to a con-
centration of sufficient strength to  de-
stroy  Giardia cysts.  Chlorine contact
times  were  calculated for each source
based on the flow in each main and  the
distance from the point of chlorine injec-
tion to the first service connection. These
times  were  confirmed  by  increasing

                                   TABLE 1
              Characteristics of reservoir facilities in Bradford, Pa., 1979

Reservoir 2 (Gilbert
Reservoir 3 (Marilla
Reservoir 5 (Hazelton


sq mi





mil gal













                                   TABLE 2
                   Coliform sample results for Bradford, Pa., 1979


Number of

Number of
Number of Samples and Time
Between Collection and Labora










           'Began using local laboratory
           tlncludes six check samples
           ^Includes two check samples
chlorine feed and clocking the amount of
time it took for the increase to appear at
the first  service connection. The  com-
puted and observed times were in rea-
sonable agreement, and  chlorine doses
were  adjusted according to  the  most
current information available for  de-
struction of  Giardia Cysts.
  The  chlorine contact time at the Gil-
bert  Run reservoir was most critical at
about 20 min, which required a chlorine
dosage that would give a residual of 2.5
mg/L  at the  first  service  connection.
Contact time at the Marilla Brook reser-
voir was 30  min, requiring a residual of
2 mg/L. The Hazelton Mill reservoir pro-
vided a contact time of more than 90 min,
which permitted the use of  a more rea-
sonable chlorine dose. Feeding the entire
system from Hazelton  Mill was consid-
ered as a possibility but was rejected
because inadequate pressure would pro-
hibit service to all customers.  However,
after beaver lodges were found in  both
the Gilbert Run and Marilla Brooks res-
ervoirs, the  output from these sources
was reduced and the deficit sustained by
Hazelton Mill. Common problems noted
at all  of   the  chlorination   facilities
included antiquated equipment (a single
chlorine  cylinder was  connected to the
chlorinator,  causing lapses in feed when
empty cylinders were replaced  so that if
a 15-min change time at each facility
occurred on the same day 250 kL [66 000
gal] of nonchlorinated water  would be
delivered  to the  system);  inadequate
chlorinator  capacity;  and an  incorrect
procedure for determining chlorine dose
to achieve a residual in the distribution
system.  Chlorine  residuals had  been
determined by collecting samples from a
utility tap located a short distance from
the injection point and adjusting the feed
rate to attain 1 mg/L. This resulted in no
residual in the system.
  The city provided new chlorine feed
equipment, which  included  manifolded
cylinders  with  automatic  switchover
devices to improve the reliability of chlo-
rination. Chlorination was the  only pro-
tective barrier that could be employed in
this situation, so uninterrupted, contin-
uous  disinfection  was  very important.
Even   though   consumers  had  been
instructed to boil water, this could not be
considered an effective barrier because it
was not known how diligently the boil
water order was being followed in  the
  The Pennsylvania Game Commission
estimated that six to eight beavers inhab-
ited the lodge in Gilbert Run reservoir. It
was suggested that they be trapped and
moved out  of the watershed; however,
the commisson  was concerned that  the
beavers would not survive  in new sur-
roundings that late in the year. Holding
the beavers  in captivity for release in the
spring was then  considered.- Suitable
holding facilities were not available and
other  problems associated  with this

                       JOURNAL AWWA

                  Hillside Spray
                  Irrigation System )L~  ~jk-	} (.
                                  Colorado River
      Figure 1. Water and sewage facilities for the Sand Point marina and campground
 alternative  made  it unworkable.  Even-
 tually all the beaver in the Gilbert Run
 and Marilla Brook reservoirs and a few
 from  the watershed of  Hazelton Mill
 reservoir were  trapped and sacrificed.
 Two beavers  in  Gilbert  Run, one  in
 Marilla Brook, and one in Hazelton Mill
 tested positive for Giardia cysts.
   The number of giardiasis cases  deter-
 mined by positive  stool examination at
 the  Bradford  Hospital  eventually  in-
 creased to 407. However, some local phy-
 sicians  began  prescribing  medication
 without the benefit of stool examination.
 Sales of drugs normally prescribed for
 giardiasis were recorded by city pharma-
 cies as numbering 3487 during the six-
 week period beginning with  the start of
 the  investigation.  The epidemiological
 study  conducted by CDC  and  the state
 health department included a random
 household survey to identify occurrence
 and distribution of illness and collection
 of stool specimens to determine infectiv-
 ity rate in residents of Bradford as well
 as  a control community. The control
 community  had  little interaction with
 Bradford and  used groundwater as  a
 water supply source.  The survey results
 showed that 5.4 percent of the  Bradford
 residents had giardiasis-like illness (diar-
 rhea for > 10 days) with only 0.6 percent
 of the control community experiencing
 the  same  symptoms. Examination  of
 stool specimens showed an infection rate
 of 16 percent for Bradford respondents

(seventeen of 106 stools positive) and all
40 stools negative for the control commu-
nity.1 The stool infectivity rate applied to
the  Bradford  population  amounted  to
2900 people  affected.
  The   water  quality  problems  that
occurred before the outbreak generated
numerous complaints, and  public  sup-
port was  building for improvement  of
the water system. The additional incon-
venience of boiling water combined with
the noticeable odor created by increased
chlorination fostered  additional  public
pressure for action. The city directed a
consulting engineering firm  to  begin a
feasibility study for the provision of ade-
quate treatment only  eleven days after
investigation of the outbreak began.
  Rebuilding  protective  barriers. The Brad-
ford water system is typical of many US
systems  faced with  making improve-
ments to comply with  the National Inter-
im Primary Drinking Water Regulations.
In Pennsylvania alone there are 113  sim-
ilar systems that rely on surface water as
a source but do not have the  facilities to
reduce turbidity. From an energy conser-
vation  aspect,  the  Bradford system  is
very efficient,  relying on  raw  water
sources situated at sufficient elevation to
maintain pressure throughout the distri-
bution system without pumping. Howev-
er,  nearly two thirds  of the system is
unmetered, which is reflected in an aver-
age per capita demand of 1365 L/day (350
gpd) and in  the size and cost of sewage
 collection and treatment facilities.
  Breakdown occurred in  all compo-
 nents of the  protective systems in the
 multiple barrier concept. With informa-
 tion accumulating from waterborne out-
 breaks of giardiasis in California,1 Wash-
 ington,2-3  Oregon,1  and   New  Hamp-
 shire,--'3 and  other studies"  implicating
 beaver  as  an intermediate host of the
 cyst that is transmitted to man, protec-
 tive public  health  measures should  be
 incorporated  into the  multiple  barrier
 concept for certain water  systems. Sys-
 tems using surface  water as  a source of
 supply,  with  disinfection  as  the  only
 means of treatment, should consider con-
 trol  of  the beaver population  in  the
 watershed.  Control  measures  may in-
 clude  periodic  surveys  to  determine
 where lodges are being established, trap-
 ping during periods of the year when the
 beaver can be  safely moved, and forestry
 practices that replace  deciduous trees
 with coniferous species, thereby limiting
 the food supply along waterways and
 around  reservoirs  to discourage beaver
 habitation.  These  precautionary  mea-
 sures could be associated  with natural
 and surveillance barriers, with both util-
 ity  and  regulatory agencies sharing  the
 responsibility  for implementation.
  Other breakdowns that occurred were
 inadequate chlorination (man-made bar-
 rier)  and   ineffective  microbiological
 monitoring  (surveillance)  which  cast
 doubt on the validity of coliform results.
 The turbidity values prior to September
 were also questionable  because the for-
 mazin standard was leaking. It is ironic
that a waiver from compliance with the
turbidity MCL was granted  by EPA in
this  situation: such  a  waiver  is predi-
cated on qualifiers including demonstra-
tion that turbidity does not interfere with
 disinfection, prevent maintenance of an
 effective residual throughout the distri-
bution system, or interfere with microbi-
 ological determinations.

State park—Lake  Havasu City, Ariz.
  Cattail Cove state park is  located on
the eastern bank of the Colorado River
north of Parker Dam and south of Lake
Havasu  City. A recreational area located
within the park-is divided into two facil-
 ities, one operated  by the  state and an
 adjacent facility operated by concession-
 aires (Sand  Point  marina and  camp-
ground). Sand  Point, the larger of the two
facilities, includes a store, comfort sta-
tion, marina, and 170 spaces for campers
 and recreational vehicles.
  The water system serving  the recrea-
 tional area is simple in design and oper-
 ation. The source is a shallow well con-
 structed in a diked area adjacent to the
 Colorado River, using  bank storage or
infiltrated river  water. After  chlorina-
 tion, water is pumped to a hillside stor-
 age tank overlooking the  two  facilities.
The  Cattail Cove and Sand Point distri-

                         E.C. LIPPY  59

bution systems are fed by separate mains
off a wye connection near the tank (Fig-
ure 1), Each facility has a sewage system
consisting of a  package plant  with
effluent disposal by irrigation.
  Cause lor concern. Disease control offi-
cials  of  the Arizona  Department  of
Health Services  requested  assistance
when they suspected a waterborne out-
break related either  to swimming  or
water supply at the recreational  area.
Reports of illness began in June and
continued through September. Most of
the  reports   originated  in  California,
which caused some problems in follow-
up of cases. Of 74 cases investigated, four
were stool positive for  Giardia  cysts.
Information  gathered  by  questionnaire
suggested  that drinking water at Sand
Point was associated with illness.
  Coliform sample results from the Sand
Point water  system  showed two of two
samples positive  (TNTC and confluent
growth) in June  with followup  check
samples  negative  thirteen  days later.
(Two consecutive negative check  sam-
ples were initiated on only one sampling
point.) One of four samples collected in
July  was  unsatisfactory   (confluent
growth) with no followup check sample,
and one negative sample was collected in
August. Samples from the state facility
distribution  system were all  negative.
Samples   from  the  swimming  areas
showed coliform  and  fecal  coliform
results in excess of the single sample
recommendation (200 fecal/1000 total) in
August and September.  After the Labor
Day weekend beaches were posted with
the warning that  water quality did not
meet standards  and that swimming was
therefore not recommended. While the
preliminary epidemiological information
implicated Sand Point  drinking water,
bacteriological data for  the swimming
areas also  cast  suspicion  on  water-
related recreational activities. Suspicion
increased when it was learned that the
sewage plant serving  Sand 'Point over-
flowed into the marina  during a power
failure in  July.  Overflow facilities had
not  been incorporated  in  the  plant
  Erratic  evolution plagues system. The
water and wastewater systems at Sand
Point evolved over a period of six years
under four different owners. An impor-
tant  feature  unique to  the Sand Point
facility was the design  of the sewage
disposal system. The system was initially
designed to dispose of the total sewage
plant effluent through the use of subsur-
face  bubbler irrigators. The  design was
based on  150 trailer or camper spaces
utilizing  380  L/day/space  (100  gpd/
space)  and 57 L/day/bubbler (15  gpd/
bubbler),  which required 1000 subsur-
face  bubblers to irrigate an equal num-
ber of trees  and shrubs. A surprising
feature of the design was the use of the
same type and  color  of pipe for the

effluent irrigation system as for the pota-
ble water distribution system.
  At some stage in the process of system
development and changes  in ownership,
the installation of bubblers was halted
and spray  irrigation was  added on an
on-site hillside to aid in the disposal of
effluent. Tree and shrub irrigation in the
camping  areas was  then  accomplished
by  drip  irrigation  using potable  water
with  aboveground  drip  fittings  sub-
merged in pits formed around the bases
of  trees  and  shrubs.   As the facility
changed  ownership and was expanded
to  accommodate   additional  trailer
spaces, records or sketches were not kept
to  indicate where  plumbing  changes
were made, and the extent  of the bubbler
system was not known.  (The  changes
were made without the required approv-
al by the health department.)
  Conditions were optimized  for  cre-
ating a direct cross connection between
the water  system and the subsurface
bubble-irrigation system.  Each  drip fit-
ting submerged  in  a tree or shrub pit
provided an entry point  for contami-
nated water if subatmospheric pressure
occurred in the water system.
  To  complicate the situation further,
the location chosen for spray irrigation
was on the side  of a hill directly above
the transmission main between the well
and storage tank and the feed line to the
campground. The feed lines to the camp-
ground were located in a gully at the
base of the hill.  Because the vegetation
and soil on the hillside could not absorb
all  the effluent  being  applied, effluent
ran off into the gully. To prevent drain-
age into  the marina  and beaches, small
catchment areas were fashioned to retain
the runoff  in pools. These  pools were
located almost directly on  top of the line
carrying  potable water. This presented
another unusual  situation.
  When the well was developed, it was
constructed in  fill material  placed to
extend the  bank  into the river and form
an infiltration gallery. Sufficient fill was
excavated to set the well at a depth of 4.5
m (15 ft) and the material packed around
the screen was to be selected and placed
by specifications referenced on the engi-
neering plan.  It is  assumed  that  this
material  was intended  to be carefully
selected,  washed, and graded sand  and
gravel. However, local material was used
in backfilling the gallery.
  A raw water quality problem subse-
quently developed which is thought to be
related to the use of material excavated
from a local site that  contains manga-
nese.   After  chlorination,  the  water
pumped  from the  well  turned black in
the transmission main to the storage
tank, indicating the oxidation of manga-
nese.  The  transmission  main,  rising
about 45  m (150 ft) and running a length
of approximately 300 m (1000 ft), served
as an inclined settling tube that permit-
ted  an accumulation  of black  water
which had to be bled off every three or
four days and before weekends through
a discharge port at the well and into the
swimming area. This process discolored
the river  for about 30  min before the
black water was diffused.
  The unique situation thus  presented
was the possibility that while the trans-
mission main to  the  storage  tank was
being  bled,  siphonage  of sewage  or
drainage from the irrigated area could
occur by a venturi effect through a crack
or bad joint in the main, causing contam-
ination of the beaches or affecting the
water being infiltrated into the well.
  Standard bacteriological samples were
collected as a result of the concern with
problems  of the water and sewage sys-
tems and  the swimming areas. A larger
volume sample of raw  water  was col-
lected  from the well, using a filtering
device containing three membrane pads
in series with pore sizes  of 5, 2, and 0.45
/mi.  Nineteen litres, or  190  times the
volume of  a  standard  bacteriological
sample, were filtered through  the pads
for analysis of coliform and fecal coli-
form bacteria. (The pads were halved at
the laboratory and processed separately
for the analysis.) The large-volume sam-
pling technique was used to detect low-
level and, possibly, intermittent contami-
  Sample results showed no coliforms in
the large-volume  sample of raw  water
and no occurrence of contamination dur-
ing bleed-off of the transmission main
from the well to  the storage tank. (The
latter finding was also confirmed by  a
dye  test.) Positive  coliform and fecal
coliform results were obtained on sam-
ples from the Sand Point distribution
system in an area that served campers
who had reported illnesses.
  Discussion of  the  results  with  the
departments of health services  and state
parks led to the agreement to close Sand
Point marina and campground. By plac-
ing dye in the effluent pumping well of
the sewage treatment plant, the northern
regional  office of the  Department  of
Health Services in Flagstaff proved that
a cross connection existed between the
water system and the subsurface bubbler
irrigation  system. After dye  was  ob-
served coming from the camp faucets,
excavation located the cross connection.
Extensive corrections were required to
improve the water and sewage systems.
Improvements also were required in sur-
veillance  activities,  including prompt
followup  on positive coliform samples,
timely reporting of results, and a review
of the state regulations permitting instal-
lation  of wastewater irrigation systems.
Particular emphasis was placed on the
necessity  of color coding  or use of dis-
similar materials for water and waste-
water piping.
  Determination  of the occurrence and

                       JOURNAL AWWA

 distribution of illness associated  with
 this  outbreak was difficult  and time-
 consuming.  Epidemiologists  from  the
 Department of  Health Services were
 hampered in their investigative efforts
 by a number of factors,  including  (1)
 difficulty in locating visitors because of
 incomplete  or  erroneous  identification
 entered on the camp register; (2) commu-
 nication across state boundaries; (3) con-
 tact  of visitors  who  might have been
 exposed five or six months before and
 would not remember;  (4) selection of an
 appropriate sample population; (5) defin-
 ing an incubation period where exposure
 to multiple pathogens could  have oc-
 curred; (6) complications related to the
 possibility that  the agent  of hepatitis
 might have been involved; and (7) limited
 investigative resources.
   Investigators eventually chose to inter-
 view a sample of  the population that
 visited one of the two  facilities on Labor
 Day  weekend. A case was  defined as a
 person who experienced a  diarrheal ill-
 ness  during or within a three-week peri-
 od after  the visit.  Results  based on  a
 telephone survey  of 44 of 179 groups
 visiting Sand  Point marina and camp-
 ground and 14 of 42 groups who regis-
 tered at the state facility showed that  57
 (35.2 percent) of 162 campers from Sand
 Point were  ill compared  to only 6 (8.0
 percent) of  the 75 state park campers.
 The  attack rate was significantly higher
 for Sand  Point campers who consumed
 campsite  water than for those who did
 not,  and  the incidence of illness  was
 greatest in the vicinity of the cross con-
 nection.  Approximately  6800  persons
 camped at  Sand  Point  from  June  1
 through September. If  water contamina-
 tion  occurred  in this  period, it is  esti-
 mated that as many as  1900 persons may
 have become ill in  this outbreak.7
   Hazards  lurk  in noncommunity  systems.
 The  water supply serving  Sand  Point
 marina and campground is classified as a
 noncommunity system under the Nation-
 al Interim Primary Drinking Water Regu-
 lations. Current estimates place the total
 number of  noncommunity systems  at
 250000. Generally, this type  of system
 has not received  the same  degree  of
 attention  from regulatory  agencies as
 have  the community water systems. The
 59 000 community  water  systems have
 been   given  priority  consideration  in
 implementation of  the  Safe Drinking
 Water Act, which is not reassuring since
 they  also  suffer from  lack  of effective
 surveillance, as exemplified by the situa-
 tion in Bradford, Pa. If such  problems as
 those found  at Sand Point  marina and
 campground are  common,  regulatory
 surveillance  will  require  substantial
•changes to control  a potentially serious
 public health situation.
  There was clearly  a breakdown in
 regulatory  surveillance for this  system.
 Design  criteria for  reuse systems were

                                       Storage Tank
                 Toxaway Creek
               Figure 2. Water and sewage facilities at The Wilds camp
 not  adequate, making review  of  plans
 and  specifications  ineffective. Proce-
 dures for inspections  during construc-
 tion and for changes after the system was
 in operation  also were inadequate, per-
 mitting hazardous conditions to develop.
 Microbiological monitoring was not ef-
 fective in followup procedures  for posi-
 tive samples.
  The problems discussed that pertain to
 investigation of the outbreak do not fully
 describe the hazards observed in this
 situation.  More than 100 potential  cross
 connections existed with the drip irriga-
 tion system;  drains  for surface water
 could be mistaken for sewage hookups,
 allowing contamination of bathing areas;
 spray from the effluent irrigation system
 could drift over the camping area; valves
 on  the water system were installed by
 bringing plumbing runs to  the  soil sur-
 face without  using vaults or boxes; and
 potable  water lines at the sewage dump
 station (for spaces without sewage hook-
 ups) were not properly protected.
  Perhaps a more important  concern is
 that noncommunity systems get the least
 regulatory attention and need the  most.
 These water systems are frequently  oper-
 ated and maintained by untrained people
 and  are  beset with design, construction,
 and operational problems.

 Church camp-Brevard, N.C.
  A church camp named "The Wilds" is
located  in Pisgah  National  Forest in
 southwestern North  Carolina. Religious
groups use the camp for retreats held
throughout the week during the summer
months and over weekends during  other
seasons. Camp facilities include dormi-
tories,  assembly  hall,   cafeteria,   and
extensive recreational facilities.
  The water system consists of two  wells
 that supply untreated water to a storage
 tank from which it is pumped  to  the
 distribution system. The major  water-
 using facilities discharge wastewater to a
 package sewage treatment plant. Septic
 tanks serve the few buildings inaccessi-
 ble to the collection system (Figure 2).
  Large number of illnesses closes camp.
 An epidemiologist with the North Caro-
 lina Department of Human  Resources
 visited  the  camp  the  first weekend  in
 November, after receiving reports of ill-
 ness. It was determined that illnesses had
 been occuring in  groups visiting  the
 camp since August. A survey was con-
 ducted  during the investigator's visit,
 with questionnaire results  showing that
 146  of  186  visitors were affected with
 vomiting and diarrhea within 24 hours of
 arrival.  Stool and acute serum specimens
 were collected and arrangements made
 for  collection  of  convalescent  serum
 after a four-week interval. The camp was
 closed while the investigation evaluated
 food preparation and  handling, general
 hygienic practices, and water  supply and
 waste disposal facilities.
  EPA participated in the  investigation
 and  evaluated water supply  and waste
 disposal facilities. A large-volume water
 sample  (1900 L [500 gal]) was collected
 for viral assay and 19-L (5-gal) samples
 for coliform and fecal coliform analyses
were collected for  each well using the
filtering technique described  in connec-
tion  with  the  Arizona  investigation.
 (Two standard bacteriological samples
 collected by the county sanitarian from
the  cafeteria kitchen the last  two week-
 ends in October were  positive for coli-
 forms.) Collection of 19-L (5-gal) bacteri-
ological samples from each well on mem-
brane filters showed coliform and fecal
coliform contamination of well 1  but no

                        E.G. LIPPY   61

coliform colonies  present on the mem-
brane filter for well 2.
  Sources of  contamination that could
conceivably affect  the wells included the
septic tanks and the effluent line from
the sewage treatment plant. Well 1 was
located about  15 m (50 ft) from the plant
and the  effluent line passed within 7.5 m
(25 ft)  of  the well  on its  path to  a
discharge point in  Toxaway  Creek (Fig-
ure 2). The effluent line crossed a drain-
age ditch and  was  found to be damaged
and leaking sewage  in September. The
area is  underlain  by  limestone  which
can become fractured, making it possible
for the well to bo  infiltrated by sewage
effluent.  (It  was reported that an earth
tremor occurred in the area in late July.)
In addition, a short distance upstream
from the location of the gutter crossing, a
septic tank had  been damaged. A truck
had been driven over the site, rupturing
the fiberglass  tank. The tile  field, which
was located at the base of a steep slope,
was eroded from  surface drainage, and
deep gullies had formed that drained into
the gutter. It was  therefore  possible for
sewage  to enter the gutter from the rup-
tured septic tank or the gullied tile field.
This, coupled  with a water line crossing
the drainage  ditch and  passing  within
immediate  proximity  of the  well, was
another  possibility  for  contamination.
Sewage flowing down the gutter could
follow the path  of the water line in the
excavated and back-filled trench.
   Both possibilities were tested by dam-
ming the gutter downstream from  the
water line and effluent line crossings and
dosing the pool with fluorescein dye. The
well was sampled unsuccessfully  for flu-
orescein for three  weeks. While fluores-
cein was apparently not charged into the
well, it is conceivable that solution chan-
nels  in  the   limestone  strata became
clogged  and  were  sealed,  preventing
entry of the dye. Possibly other  factors
intervened. One septic tank that could
have remotely influenced well 2 was also
dye-tested with negative results.
  The camp was  permitted  to  reopen
pending  improvements which included
installation  of chlorination  equipment
on both wells, flushing and  disinfection
of the well system, and repairs to sewage
disposal facilities. Well 1 was kept out of
service during the  opening weekend and
no illness was noted among visitors.
  Epidemiological   survey  data, rein-
forced by  laboratory  tests   of medical
specimens, determined  that  the agent
responsible for this outbreak was a "Nor-
walk"  type virus. Serological   studies
conducted  by the Center  for Disease
Control  showed that thirteen of fourteen
symptomatic cases were seropositive for
antibody  titers  against  the Norwalk
agent.   Immune   electron   microscopy
studies of stools from 35 ill volunteers
failed to defect  virus.8 Detection  of
Norwalk agent in  stool specimens col-

lected during outbreaks has been very
difficult. Volunteer studies under labora-
tory conditions  have shown that  about
one half of those exposed to  Norwalk
agent excrete the particles in  sufficient
concentration to be detected by immune
electron microscopy.  Particles are not
detected prior to onset of symptoms and
are found in less than 20 percent of stools
collected 72 hours or longer after onset of
  Untreated groundwater supplies  spell trou-
ble.  This   outbreak is  discussed  here
because it  typifies  many   of  those
occuring in systems using groundwater
as a source of supply without treatment.
A review of the causes  of waterborne
outbreaks shows that for the period from
1971 through  1978, 107 of 224 reported
outbreaks were caused by use of ground-
water  that was not chlorinated or not
adequately treated. The  outbreaks were
responsible for a total of 17 391 illnesses.
An etiologic agent was identified in few-
er than half of the outbreaks, the remain-
der  being  characterized  as acute gas-
trointestinal illnesses that were presum-
ably caused  by bacterial pathogens or
enteric  virus  for which isolation and
identification techniques were not avail-
able. It is not unusual for laboratories to
culture specimens for only two bacterial
pathogens, Salmonella and ShigeJJa, and
if they are not  present, the  causative
agent is never determined.
  There has been an improvement in the
capability of some laboratories to detect
other bacterial and viral pathogens that
can be cultured. However, nonculturable
viruses, which can be  detected in stool
specimens  only  by electron microscopy
or through the use of stool material from
infected individuals reacted with  acute
and convalescent serum to produce spe-
cific rises in antibody titer, have become
increasingly associated with waterborne
outbreaks in groundwater systems. It is
typical of these outbreaks that investiga-
tors detect sources of sewage contamina-
tion in close  proximity to wells  but
usually  cannot  demonstrate  either
through dye-testing  or by sampling for
microbiological  indicators that  contami-
nation occurred. Viruses identified with
these outbreaks are extremely small in
size (in the range of 30-70 nm), which
probably  enables  them  to   penetrate
water-bearing aquifers whereas coliform
organisms are trapped or die. Many fac-
tors interfere with dye-testing in ground-
water  situations,  and the  inability to
demonstrate  a positive test is not proof
that the "likely event" did not  occur.

Overall summary
  A study  of three  waterborne disease
outbreaks that occurred in 1979 shows
that problems contributing to the occur-
rence of outbreaks can be identified and
related  to  a  breakdown of protective
systems in the multiple barrier concept.
Designers,- utility personnel, and regula-
tory agencies share a responsibility for
maintaining the integrity of the  protec-
tive systems  necessary to ensure  the
delivery of a safe and potable water.

  The author acknowledges the coopera-
tion of  the  following persons: Isabel
Guerrero, CDC, Atlanta, Ga., James Erb
and John Mead, both of  the bureau of
water  quality control, Pennsylvania  De-
partment of Environmental  Resources,
Meadville, Pa.; Patrick Nuzzo, Bradford,
Pa. Karen Starko, bureau of disease con-
trol, Arizona Department  of Health Ser-
vices and  William Shafer, both of  the
bureau of water quality control, Arizona
Department of Health Services, Phoenix,
Ariz.; Robert Anderson, Mojave County
Health Department, Lake Havasu City,
Ariz.; Douglas Collup, Lake Havasu State
Park, Ariz.; Charles Finch, Department
of Human Resources, Raleigh, N.C.; Jack
McGuiness, Transylvania County Health
Department,  Brevard,  N.C.; and David
Cleary, The Wilds Camp, N.C.
  The  author  also acknowledges fellow
employees  S.C. Waltrip, F.W.  Schaefer,
T.H. Ericksen, and S.C. Underwood.

1. Waterborne  Giardiasis—California, Col-
    orado, Oregon, Pennsylvania. Morbidity
    and Mortality Weekly Rep., 29:11. Center
    for Disease Control, PHS, DHEW, Atlan-
    ta,  Ga. (Mar. 21, 1980).
2. Waterborne Giardiasis Outbreaks—Wash-
    ington, New Hampshire. Morbidity and
    Mortality Weekly Rep., 26:21  Center for
    Disease Control,  PHS, DHEW,  Atlanta,
    Ga (May 27, 1977).
    A Waterborne Outbreak of Giardiasis in
    Camas, Wash. Jour. AWWA, 70:1:35 (Jan.
4. LIPPY, E.G. Tracing a Giardiasis Outbreak
    at Berlin, New Hampshire. Jour. AWWA,
    70:9:512. (Sep. 1978).
5. LOPEZ, CARLOS ET AL. Waterborne Giardia-
    sis—A Community-Wide Outbreak of
    Disease and a High Rate of Asymptomat-
    ic  Infection.  Am. Jour. Epidemiol. (in
 6. DAVIES, R.B. & HIBLER, C. P. Animal Reser-
    voirs and Cross Species Transmission of
    Giardia. Waterborne Transmission of
    Giardiasis Symposium, Natl. Tech. Infor-
    mation Service,  Springfield, Va., EPA-
    600/9-79-001 (June 1979).
 7. STARKO, K. ET AL. Diarrhea in Campers (in
 8. FINCH, C. ET AL. Recurrent Gastroenteritis
    at  a Camp in North Carolina (in prepara-
    J.S. Viral  Enteritis-Seminars  in Infec-
    tious Disease. Stratton  Intercontinental
    Medical Book Corp. (1978).
 Edwin C. Lippy (Active Member, AWWA) is a
 sanitary  engineer with  the  acute  disease
 branch, health effects  research laboratory,
 US EPA, Cincinnati, Ohio 45268.

                       JOURNAL AWWA

                                                         Reprinted  from  Journal,  AWWA,  Vol.  78,
                                                         No.  1,  by  permission.  Copyright   1986.
                                                         American Water  Works Association
Chlorination  to  Prevent and  Control
Waterborne  Diseases
Edwin C. Lippy
An analysis of statistics compiled on outbreaks of waterborne disease from 1946 to 1980 and
from 1965 to 1982 showed that disinfection was not provided where it was needed and was not
properly operated where it was provided.  Chlorine concentration, contact  time, pH,
temperature, and interfering substances all must be evaluated to determine disinfection
requirements. Use of reliable chlorinatioh facilities and proper application of current tech-
nology can control and prevent waterborne diseases.
  Two  recent publications  identified
 shortcomings in disinfection contribut-
 ing to waterborne disease outbreaks.
 Statistics compiled for the period from
 1946 to 1980 showed glaring deficiencies
 in disinfection practices.1 The primary
 problems were that disinfection was not
 provided where it was needed and it was
 not properly operated where it was
 provided, both of which caused or con-
 tributed to nearly half of the outbreaks
 of waterborne disease  in public water
 systems. The same shortcomings were
 identified when the 63 giardiasis out-
 breaks occurring in the period 1965-82
 were  analyzed.2 Two-thirds, or 42, of
 these outbreaks were related to a failure
 to provide adequate or reliable treatment,
 with chlorination deficiencies identified
 as a primary cause.
  This article reviews  the factors that
 influence chlorine's effectiveness  in in-
 activating bacteria, viruses, and proto-
 zoans for the purpose of preventing and
 controlling waterborne  disease out-
 breaks.  The factors reviewed are: (1)
 chlorine concentration, (2) contact time,
 (3) pH,  (4) temperature, and (5)  inter-
 fering substances. Each of these must be
 evaluated to determine disinfection
 capability under current operating con-
 ditions and to determine the adj ustments
 necessary for coping with the challenges
 of pathogens.

 Chlorine concentration and contact time
  The product of chlorine concentration
 C (in milligrams   per millilitre) and
 contact  time T (in  minutes) produces a
 CT relationship that is used  in deter-

mining the requirements for inactivation
of bacteria, viruses, and protozoans.
Because CT relationships are developed
through laboratory research under care-
fully controlled conditions, applying the
data to a  real  world situation requires
judgment. The CTvalues listed in Table
1 for viruses were developed by White3
and also can be applied for inactivation
of bacterial agents. The viral CTvalues
are based on inactivation of Coxsackie
A2  viruses, which are more resistant
than poliovirus or any pathogenic vege-
tative bacteria implicated as causative
agents in waterborne disease outbreaks.
Values for protozoan CTwere interpreted
from  research conducted  by White,3
Jarroll et  al,4 and Rice et al.5 A rule of
thumb can be derived from  the  CT
relationships at pH values of 7-8, which
are common to  most water systems.
Systems using groundwater as a source
of supply should apply a minimum CTof
15-30 to provide for the inactivation of
viruses and bacteria, which are generally
the contaminants of concern; i.e., the
size of viral particles  permits easier
penetration of strata overlying aquifers
whereas protozoan  cysts are trapped
and retained (protozoan cysts are about
1000 times larger than virus particles),
and bacteria are capable of reproducing
in the environment  whereas protozoan
cysts are not.  Systems using surface
supplies, especially those located in areas
where source water temperatures can be
as low as  41°F (5°C),  should apply a
minimum CT of 100-150 for protozoan
inactivation. At temperatures lower than
41°F (5°C), the CT requirements will
increase. Figure 1  illustrates the CT
requirements included in Table 1.
  Once the CT value is known, the ac-
tual contact time is determined through
evaluation  of the treatment facility, and
the required chlorine concentration is
determined mathematically.  Because
several factors must  be considered,
determination of the actual contact time
should be  done carefully. A common
mistake is using hydraulic displacement
calculations to compute contact time.
For example, a 1-mil gal  (3.8-ML) rec-
tangular storage tank, with inflow and
outflow at  opposite ends, provides  2.4 h
of contact  time when calculations con-
sider the average daily plant output of 10
CT relationships for 99-percent inactivation of pathogenic agents
Viral CT
Protozoan CT

                                                       F.nwiN r T IPPV AQ

    Disinfection problems
    caused or contributed
      to more than half
        of the outbreaks
    ofwaterborne disease.
 mgd (38 ML/A), but the contact time
 actually provided may  be one fourth of
 the calculated  value.6  With an  actual
 contact time of 36 min, a chlorine con-
 centration based on the 2.4-h contact
 time would be insufficient. This differ-
 ence in the calculated  displacement or
 theoretical retention (contact) time and
 the actual retention (contact) time stems
 from  short-circuiting and dead  spaces
 common to both rectangular and circular
 storage tanks. The use of  the average
 daily plant output can also lead to errors.
 Plant output varies with customer de-
 mands. Peak daily demands for some
 systems can  be twice the average daily
 plant output. If this is the case,  the
 contact time is underestimated by an
 additional 50 percent. If the chlorinator
 is not paced to respond to changes in
 plant output, fluctuation in demand must
 be taken into consideration when eval-
 uating contact time and chlorine dosage
   The preferred method of determining
 contact time  is to conduct a tracer test.
 One simple  and quick method is to
 increase chlorine  feed  and determine
 chlorine residual concentrations in  the
 basin effluent.  The initial increase in
 concentration signals the flow-through
 time for a fraction of the plant output,
 and subsequent samples will reflect a
steady rise in chlorine residual until the
incremental dosage has been accounted
for. The  mixed-flow and dead  space
components of basin hydraulics can also
be estimated through application of the
tracer methods discussed by Hudson.7
Another simple and quick method for
tracer tests is to cease fluoride feed and
read the decrease in fluoride concen-
tration over time. The initial decrease in
fluoride concentration signals the flow-
through time  for a fraction of plant
output and represents the contact time
provided   for a portion  of  the  basin
volume. The advantage of using chlorine
and fluoride as tracers is that there is no
need to introduce other dyes, salts, or
chemicals into the water system. Use of
other tracers may require special per-
mission because of concern about toxic-
ity. Decreasing  the concentration  of
fluoride or increasing the concentration
of chlorine for a short time should not
cause objections to conducting the test.
The elapsed time between the increase
or decrease of chemical dosage and the
initial corresponding change in the basin
effluent represents the contact time and
should be used in the CT relationship to
calculate the concentration of chlorine
that will be required to achieve a micro-
biocidal residual. The CT relationships
are based  on maintaining the chlorine
            o  ViralcratSO'F 10°C
            •  ViralCrat32-41°F 0-5°C
            D  Protozoan CTal 59° F 15° C
            •  Protozoan CTal 41° F 5°
                20       40
   Figure 1. CT relationships for 99-percent inactivation of pathogenic agents
concentration for the entire contact time;
therefore, the dosage must be adjusted
accordingly to compensate for chlorine
demand and dissipation.
  Tracer tests are normally not required
where transmission  or distribution
mains are depended on for contact time.
The mains are pressurized and, under
full-flow conditions, the hydraulic char-
acteristics approximate plug flow. With
plug flow  the  theoretical and  actual
contact times are equivalent. However,
they can be easily checked by increasing
the chlorine dosage and measuring for
the increase in residual at an appropriate
point downstream. Table 2 is a ready
reference for computing contact time in
transmission and  distribution mains.
By dividing the pipe capacity (adjusted
for available length) by the plant output
or flow, the contact time can be determ-
ined. For example, at an output of 1 mgd
(3.8 ML/d) with 1000 ft (300 m) of 24-in.
(600-mm) pipe, contact time = 34 min.

pH and temperature
  The effectiveness of  chlorine as a
disinfectant is influenced by the pH and
temperature of the water being dosed.
The pH  determines the  amount  of
hypochlorous acid (HOC1) and hypochlo-
rite ion (OC1~) in solution. HOC1 is the
predominant form at lower pHs (6-7.5)
and  OC1"  is the predominant form  at
higher pH values of (8-10). At pH 7 and
32°F (0°C), the dissociation of chlorine
produces 87.04 percent HOC1 and 12.96
percent OC1"; at pH 9, it produces 6.29
percent HOC1 and 93.71 percent OC1~.8
It is important to know the dissociated
forms of chlorine present at various pH
values because HOC1 is the form that
possesses  the destructive power as a
microbiocide. According to White, "the
germicidal efficiency of HOC1 is due to
the  relative ease  with which  it can
penetrate cell walls. This penetration is
comparable to that of water, and can be
attributed to both  its modest size (low
molecular  weight) and its  electrical
neutrality (absence of an  electrical
charge)."  White states  that  the hypo-
chlorite ion "is a relatively poor disin-
fectant because of its inability to diffuse
through the cell wall of microorganisms
due to the negative electrical charge."
Research  has shown that HOC1 is 80
times more effective than OC1" in the
inactivation of Escherichia coli and about

                       JOURNAL AWWA

                                                                                   HOCl is the form
                                                                                        of chlorine
                                                                                    that possesses the
                                                                                   destructive power
                                                                                   as a microbiocide.
 150  times more effective for cysts of
 Entamoeba histolytica.3
  The key point to remember in the
 evaluation of chlorination is that the CT
 relationship increases with pH. Contact
 time, or the T portion of the CTrelation-
 ship, is controlled by structural facilities
 such as contact basins, storage tanks, or
 transmission mains whose size cannot
 be readily adjusted to compensate for a
 needed increase in CT. Therefore, the
 only immediate alternative is to increase
 chlorine concentration. If pH is such
 that OC1" is the predominant form, it
 may be necessary to lower the pH so that
 HOCl is the predominant form. Adjust-
 ment of pH should be considered when
 the  chlorine requirement becomes  ex-
 cessive and, perhaps, objectionable.
  In water, chlorine hydrolyzes rapidly—
 occurring in tenths of a second at 65°F
 (18°C) and in a few seconds at 32°F (0°C).
 Although temperature has little effect
 on chlorine's solubility, chlorine's reac-
 tion with microorganisms is affected.
 CTrelationships are inversely related to
 water temperature, and as it decreases,
 the  CT increases (Table 1).  Therefore,
 temperature must be considered when
 selecting  a bactericidal, viricidal, or
 cysticidal chlorine dosage.

  Chemical interference. When chlorine
 is introduced  into water, it  not only
 reacts to form microbiocides but  also
 reacts with other substances to form
 organic and inorganic compounds that
 have little or no disinfection capability.
 The reaction  of chlorine with organic
 and inorganic substances constitutes
 chemical interference with disinfection
 through consuming a part of, or all of,
 the applied chlorine—leaving little or no
 chlorine  to react  as a  disinfectant.
 Organic and inorganic forms of nitrogen
 are among the compounds that signifi-
 cantly interfere with the chlorination of
 surface water. Hydrogen sulfide, iron,
 and manganese are commonly found in
 groundwaters and interfere with disin-
 fection by combining with chlorine to
 form sulfates, ferric hydroxide,  and
 manganese  dioxide, respectively. The
 significance of this interference is  ex-
 emplified  by the requirement of 8.32
 mg/L of chlorine to oxidize 1 mg/L of
 hydrogen  sulfide.3  In fact, the ability of
 chlorine to oxidize hydrogen sulfide, iron,

and manganese is used to advantage as a
treatment technique and is employed at
some water plants for removal of these
  Physical interference.  This  type  of
interference occurs when microorgan-
isms become entrapped within the matrix
of particulate  matter and are  shielded
from  the effects of disinfection.  Re-
search9-14 indicates that the protective
effect of shielding by inorganic particu-
lates  is  minimal and that inorganic
particulates are of little consequence as
physical interfering substances. Organic
particulates such as  those  associated
with sewage effluent solids do, however,
afford protection and permit survival of
microorganisms after long chlorine con-
tact periods (60 min). A recent study by
LeChevallier et al15 on the survival of
coliform  organisms in the presence of
turbidity indicated that after  1 h of
contact time there was a 99.5-percent
reduction in the number of organisms at
a concentration of 6 ntu or less and an
approximately 20-percent reduction at a
concentration  of 8-13 ntu. The study
showed that turbidity control, to at least
the 1 and 5 ntu concentrations as  re-
quired by US Environmental Protection
Agency regulations, is justified. Although
more research is needed to specifically
define the degree of physical interference
resulting from turbidity and particulates,
common sense dictates that where sew-
age discharges affect raw water quality,
turbidity should be reduced by appro-
priate treatment (e.g., mixing, coagula-
tion, flocculation, settling, filtration) for
chlorination to be effective. If chlorina-
tion  is the only  treatment, it must be
adjusted to compensate for the physical
interference of turbidity.

Example of optimizing chlorination
  An example is provided to illustrate
the  procedure  that  can be used in
optimizing chlorination to cope with a
pathogen challenge.  Assume  that a
community uses an impoundment on a
mountain stream as a source of supply.
There are no sources of human contam-
ination  in the watershed. Chlorination
is provided at the impoundment, with
contact  time occurring in 5000 ft  (1500
m) of 24-in. (600-mm) transmission main.
The average daily demand is 2 mgd (7.6
ML/d), which is  increased by 50 percent
during peak periods. The water temper-
ature is 41°F (5°C), pH is 7, and turbidity
is 1 ntu. Research data indicate the CTis
100 to offset a particular pathogen chal-
lenge. The chlorine dosage requirement
is calculated for  peak output, and it has
been assumed that the chlorinator feeds
in proportion to flow.

    Peak output:  2 mgd X 150% = 3 mgd
   Contact time:  (2350 gal*/100 ft)
              x (5000 ft/3 mgd)
              = 56 min
Chlorine demand:  test with chlorine comparator
              kit yields a demand of 0.5 mg/L
              at 56 min
  •From Table 2
Fluid capacity of pipes
Pipe Diameter— d
gal/ 100 ft

                                                            EDWIN C.LIPPY  51

       properly applied,
            could have
          the outbreaks.
  Chlorine dosage
    requirement: (CT/T) + demand
              (100/56) + 0.5
              =1.8 + 0.5
              = 2.3 mg/L

  The calculations account for all factors
 affecting chlorination, except for extrap-
 olating the CT relationship from labora-
 tory research data to applied use.  An
 increased dosage of 0.5-1.0 mg/L would
 not he unreasonable. For this example, a
 dosage of 3 mg/L might be chosen to
 account for unknown variables  in  ex-
 trapolation. Increasing  the dosage to 3
 mg/L has the effect of adjusting the  CT
 relationship by 40 percent (i.e., 1.8 + 0.7
 X 56 = 140). The free chlorine residual
 that is produced, 2.5 mg/L after 56 min
 contact time, should be confirmed by
 monitoring. Microbiological monitoring
 for the pathogenic agent  may be per-
 formed to confirm the efficacy of chlo-
 rination and account for extrapolation.
 If sampling or analytical methods  are
 not available for pathogen monitoring,
 chlorination efficacy could be gauged by
 using methods that determine inactiva-
 tion of the total bacterial population.
  The above example assumes a flow-
 proportioning chlorinator that automat-
 ically  adjusts  dosage  in  response  to
 changes in plant output. If a constant-
 feed chlorinator is used, several adjust-
 ments of dosage are required throughout
 the day. For example, between the hours
 of  10:00 p.m. and  6:00  a.m. customer
 demands  decrease,  and plant output
 responds  accordingly.  If  demand  de-
 creases to 50 percent of the average,  the
 chlorine residual is going to increase to
 objectionable concentrations unless  the
 dosage  is adjusted.  With a  50-percent
 decrease in average output, or 1 mgd (3.8
 ML/d), and the chlorinator adjusted to
 maintain the 3-mg/L dosage calculated
 for 3 mgd (11 ML/d), the chlorine residual
 will increase dramatically during low-
 demand hours  to  8.5   mg/L. In this
 example, the resulting chlorine residual
 would likely cause widespread consumer
 complaints about tastes and odors.
  Equally as important as providing  the
 correct dosage is assuring that reliable
 chlorination is in operation 24 hours a
 day. Reliability demands dual cylinder
 feed, auxiliary power,  standby  equip-
 ment and spare parts, and maintenance
 of records documenting operation. Dual
 cylinder feed is necessary  to ensure

continuous dosage so that if one cylinder
becomes depleted, the fully-charged one
automatically  comes on-line.  A  dual
cylinder hookup also is needed to ensure
continuous feed during cylinder replace-
ment. Changing an empty cylinder for a
full cylinder takes about 30 min, and if a
dual hookup is not provided,  unchlo-
rinated water will enter the distribution
network. This means that for a plant
output  of 1 mgd (3.8 ML/d), 21 000 gal
(79 493  L) of unchlorinated water would
be distributed. Auxiliary power is neces-
sary to  sustain operation during power
outages. A standby chlorinator and spare
parts are needed for uninterrupted opera-
tion during maintenance and  repairs.
Records to substantiate  reliability  in
operation, including pounds of chemical
used, plant output on a daily basis, and
chlorine residual measurements, are also
required. Current technology to ensure
reliability in chlorination utilizes loop-
controlled feed with remote sensing and
automatic adjustment of dosage  to a
predetermined residual. The residual is
continuously analyzed and recorded on a
circular or strip chart, and an alarm
system  can be installed to warn of low
residuals. The continuous record sub-
stantiates chlorination, or lack of it, and
is a very important factor  in verification
of proper disinfection.
  The factors that influence the  effec-
tiveness of chlorination were evaluated
and  the required  CT successfully ap-
plied, halting transmission of waterborne
disease. If adequate chlorination facilities
had been in place and current technology
properly applied to cope with the chal-
lenge of'pathogenic agents, the outbreaks
could have been prevented.

  1. LIPPY, E.C. & WALTRIP, S.C. Waterborne
   Disease Outbreaks, 1946-1980: A Thirty-
   Five  Year Perspective.  Jour. AWWA,
   76:2:60 (Feb.  1984).
  2. LIPPY, E.C. & LOGSDON, G.S. Where Does
   Waterborne Giardiasis Occur, and Why?
   Proc. 1984 Specialty Conf.  on Envir.
   Engrg. ASCE, New York (1984).
  3. WHITE, G.C.  Handbook of Chlorination,
   Van  Nostrand Reinhold  Co., New York
   E.A. Effect of Chlorine on Giardia lamb-
   lia Cyst Viability. Appl. & Environ.
   Microbiol., 41:483 (1981).
  5. RICE, E.W.; HOFF, J.C.; &  SCHAEFER, F.W.
   Inactivation  of Giardia  Cysts by Chlo-

   rine. Appl. & Environ. Microbiol, 41:250
 6. LIPPY, E.C. Tracing a Giardiasis Outbreak
   at Berlin, N.H. Jour. AWWA, 70:9:512
   (Sept. 1978).
 7. HUDSON, H.E. JR. Water Clarification
   Processes: Practical Design and Evalua-
   tion. Van Nostrand Reinhold Co., New
   York (1981).
 8. WHITE, G.C. Disinfection of Wastewater
   and Water for Reuse.  Van  Nostrand
   Reinhold Co., New York (1978).
 9. BOARDMAN, G.D. Protection of Waterborne
   Viruses by Virtue of Their Affiliation
   With Paniculate Matter. Doctoral disser-
   tation, Univ. of Maine (1976).
10. HOFF.J.C. The Relationship of Turbidity
   to Disinfection of Potable Water. Evalua-
   tion of the Microbiology Standards for
   Drinking Water(C.W. Hendricks, editor).
   USEPA-570/9-78-006 (1978).
11. BOYCE, D.S.; SPROUL, O.J.; & BUCK, C.E.
   The Effect of Bentonite Clay on Ozone
   Disinfection of Bacteria and Viruses in
   Water. Water Res., 759 (1981).
12. WALSH, D.S.; BUCK, C.E.; & SPROUL, OJ.
   Ozone Inactivation of Floe  Associated
   Viruses and Bacteria. Jour. Envir. Engrg.
   Div.—ASCE, 106:761 (1980).
13. STAGG, C.H.; WALLIS, C.; & WARD, C.J.
   Inactivation of Clay-Associated Bacterio-
   phage MS-2 by Chlorine. Appl. & Environ.
   Microbiol., 33:385 (1977).
14. SCARPING, P.W. Effect of Particulates on.
   Disinfection of Enteroviruses in Water
   by Chlorine Dioxide. USEPA 600/2-79-
   054 (1979).
   SEIDLER,  R.J.  Effect of  Turbidity on
   Chlorination Efficiency  and Bacterial
   Persistence in Drinking Water. Appl. &
   Environ. Microbiol., 42:159(1981).
                                                                JOURNAL AWWA

IV. Sampling and Analytical Methods

                                      Reprinted by the
                                  PUBLIC HEALTH SERVICE
                                CENTERS FOR DISEASE CONTROL
                                    VIRUSES AND THE GUT
                       (Proceedings of the Ninth BSG -SK&F International Workshop 1988)
                                   October 2-4,1988,pp. 87-90
Roger I. Glass, Stephan S.  Monroe, Sarah Stine, Paul Madore,
David Lewis, David Cubitt, Gary Grohmann and Charles Ashley
Viral Gastroenteritis Unit, Centers for Disease Control, Atlanta, Georgia, USA; Division of Infectious
Diseases, University of Rochester, Rochester, Neic  York, USA; Public Health Laboratory Service, Taunton,
United Kingdom; PHLS, Central Middlesex Hospital, London, United Kingdom; Westmead Hospital, Sydney,
Australia; PHLS, Bristol, United Kingdom
SRSVs of the Norwalk family have remained
among the most difficult viral agents to detect,
characterise and understand.  Members of this
family  share  an  asymmetrical  structured
morphology when visualised directly in the
electron microscope, but they are antigenically
distinct  when  examined   using  immune
electromicroscopy (IEM). None of these agents
has been successfully cultivated, few have been
adequately  purified  so that either polyclonal
antisera  or  monoclonal  antibodies can  be
prepared, few are visible with ease by EM, and'
nearly all  assays  developed  to  detect the
antigen of these agents require human sera and
stool as basic reagents. Since exposure to these
agents is common, and since many of  these
agents may be antigenically related,1 antibody
responses in humans may not be restricted to
the infecting particle. Consequently, relatively
few  of the  convalescent  human  antisera
obtained, even from volunteers, can be used as
specific reference reagents.
  The  literature  on  the Norwalk  family of
agents is filled with a colourful variety of names
of particles found in a single epidemic. These
agents, of  which nearly a dozen  have been
identified, are often  unsuitable as references
because samples  from the key outbreaks have
been  exhausted   in  characterising  the new
particle against other reference agents.  Only
where volunteer  studies have  been conducted
to amplify the volume of these reagents has the
particle been further characterised  and assays
set up for  its detection.2"1


  In  recent  years,  there  have  been  more
reviews written about these novel agents of
viral gastroenteritis than  substantive advances
in  our understanding  of  these  agents.^"
Consequently, rather than review the history of
each  of  these  particles  again, I  present an
approach  used at CDC  to  assess  where we
would like to go in the future in  combining
epidemiological and laboratory studies. In the
past, epidemiological investigations have been
key to identifying novel  diarrhoea agents and
establishing their  role as pathogens in  man.
Improvements  in  the  investigation of future
outbreaks—paying special attention to  more
voluminous specimen collection,  more  basic
virology  and extended  laboratory collabor-
ations— may help improve our understanding
of the epidemiology of disease caused by  these
agents  and   provide   the   basic  diagnostic
reagents  required  to  achieve major  break-
throughs in understanding these viruses.

Epidemiological investigations as a
key to discovery of new agents
In 1986, we reviewed the activities of the viral
gastroenteritis group at CDC to asess the role of
epidemic  investigations  in  discovering new
viral agents and in establishing their causal role
in gastroenteritis. Two of the SRSVs recognised
to be causes of diarrhoea—the Norwalk agent12
and the  Snow Mountain agent13—were each
identified during the investigation of outbreaks
of diarrhoea of undetermined aetiology by CDC
epidemiologists. Human specimens from each
of these  outbreaks were subsequently used to
establish the identity and causative role of the
virus  using  techniques  of  immune electron-
microscopy and  to prepare inocula for both
primate and human volunteer challenges. Had
either of  these agents been seen in a single
individual, establishing their role as the  cause
of  diarrhoea  would  have   been  impossible.
However,  the coincidence of finding particles
in  many  of  the  outbreak  victims  but not in
controls, corrfbined with the determination of
seroconversion to the particle by IEM among
victims  but not controls,  helped link the
agent in the chain of causality. Specimens from
volunteer studies with the Norwalk agent were
used by Greenberg and Kapikian to develop an
RIA to detect antigen and a blocking assay  to
 measure antibody litre.1 Use of this assay with
 specimens from the CDC collection of unsolved

outbreaks  led  to  the  recognition that  the
Norwalk agent was a prime cause of about 25%
of the outbreaks of non-bacterial gastroenteritis
in the USA. This RIA assay was adapted by
Gary  for use with biotin-avidin, leading to  a
sensitive and non-radioactive method to detect
antigen and seroconversion  to  the  Norwalk
agent.3'14 These assays have remained the key
diagnostic tests in our laboratory. At the same
time,  the infrequency  of detecting SRSVs by
direct  EM led over  the  years to the near
abandonment of the use of electronmicroscopic
methods   in    our    laboratory.    Further
seroepidemiological studies with the  Norwalk
agent  established  that Bangladeshi  children
and American  adults had a high prevalence of
antibody, indicating that  the Norwalk  virus
could be a cause of sporadic diarrhoea as well.l3
  Investigations by  other groups have led to
the identification of a variety of other particles,
including  the  Hawaii agent,   Montgomery
County agent and Marin County agent in the
USA;   the  Taunton,  Moorcroft  and  Barnett
agents  in  the  UK;  the Paramatta agent in
Australia; the Sapporo and Otofuke agents in
Japan  as  well  as  a  group of other similar
particles differentiated  by IEM in Japan.16 The
wide  antigenic diversity of  this group,  the
inability to compare  agents  and  to refine
nomenclature, and  the difficulty in obtaining
quantities of the virus from outbreak specimens
or passage in  animals or cell  culture  have
slowed progress in the field.

Review of outbreak investigations
We have reviewed results from the  last  100
outbreaks  investigated by CDC staff where
stools and paired  sera have been collected and
assayed using the  biotin-avidin test for  the
Norwalk  agent  (Fig. 1). Our  intent was to
confirm   previous   observations   on   the
prevalence of Norwalk-associated outbreaks in
this collection while looking at outbreaks that
were still without an  aetiological diagnosis  to
see if the causative agent could be found. An
outbreak was  defined  as being associated with
the Norwalk agent if,  from six  or more paired
sera  collected from victims of the  outbreak,
50%   or  more demonstrated  seroconversion
defined as a four-fold  rise in litre in the biotin-
avidin assay.  We found  thai the  20-30%
positive rate for Norwalk-associated outbreaks
observed  in  earlier   sludies17'18  has  been

maintained  recently.  Frequency  distribution
was  not made up  of  two clear and distinct
peaks   of  Norwalk-positive  and  Norwalk-
negative outbreaks. Rather,  rates  of sero-
conversion  represented  a  continuum  with
definite positive and negative outbreaks as well
as a  clear middle  ground  distinguished  by
outbreaks  having   a   low  percentage  of
seroconversions    (10-49%)    in    which
convalescent litres  rose only  to intermediate
values  (e.g.  800 or 1600).  The intermediate
group might represent infection with  a virus
that  antigenically   cross-reacted   with  the
Norwalk agent, that produced an  anamnestic
response to a related antigen, or was ill-defined
or  non-specific due to an insensitive assay
          10  20  30  40  50  60  70  80  90100
            Seroconversions to Norwalk (%)

Figure 1:  Serological  analysis  of  outbreaks  of
Norwalk virus infection (n = 100), 1985-1988.
  Selected outbreaks in the intermediate group
(Group II) have been examined carefully and in
collaboration with  other investigators in  an
effort to identify other viral agents. Panels of
specimens from  10-20  outbreaks  have been
shared with other investigators including Dolin
and Madore  (Rochester)  to  look for Snow
Mountain  agent,  Blacklow and  Herrmann
(Worcester)  to  assay for astrovirus,  Cubitt
(London) to look for SRSVs using an antigen
detection system still being developed,  and
Lewis  (Taunton) to  look by IEM for  three
specific  Norwalk-like particles  such as  the
Taunton agent. Some of our initial results  are
presented below.
Taunton agent
Paired sera from  10 outbreaks from  the inter-
mediate group were tested  using an immune
capture assay to look for a serological  rise to the
Taunton agent.14 Paired sera from one outbreak
in a nursing home in Florida proved positive for
the Taunton  particle, so 13 additional paired
sera from  this outbreak were examined.  Of
these,  11 showed a significant and diagnostic
rise to the Taunton agent, identifying this to be
the first such  outbreak identified in  the USA.
These  paired  sera  represent   a   reference
resource to look further for the Taunton particle
in future outbreaks.

Snow Mountain agent
In January 1988, we investigated an outbreak of
diarrhoea aboard a British cruise ship that had
come into US territorial waters. A small particle
had been seen in an outbreak on a previous
cruise. More  than  20%  of  the  passengers
developed  diarrhoea in the two weeks before
the investigation. Paired sera and stools were
negative  for  the  Norwalk agent  by  both
antibody and antigen detection assays.  The
sera did respond  to the Snow Mountain agent
with   four   of   19   passengers   showing
seroconversion and  11 of  19 demonstrating
high  acute  titres. One faecal  specimen  had
a visible particle which was negative in an assay
for Norwalk or Snow Mountain antigen. This
outbreak appears to be due to a small particle
antigenically  related  to  the Snow  Mountain

In February 1988, an outbreak of gastroenteritis
occurred among children in an day care centre
in Sydney,  Australia. Caliciviruses were seen in
small quantity from  several stool  specimens
taken from the children in the outbreak. These
stool specimens and paired sera from patients
were negative for the  Norwalk  antigen by
biotin-avidin assays. They were examined in an
enzyme  immunoassay  for  Calicivirus2"  and
found to be  positive,  making  this the  first
outbreak  involving  Calicivirus  identified  in

Detection  of SRSV antigen

Faecal specimens from 23 outbreaks in the CDC
collection were tested for SRSV antigen using a
biotin-avidin assay developed by Cubitt.  This

assay recognises a variety of SRSVs, including
the   Norwalk  agent,   although   the  .exact
specificities  of  the  capture  and  detection
antibodies are unknown. The assay was able to
determine between  outbreaks  where  many
specimens were positive and other  outbreaks
where all specimens were negative, a clustering
which could not occur by chance alone. While
the assay remains to be further  characterised,
incorporation    of    biotinylated     human
convalescent  sera  from  other  non-Norwalk
outbreaks  into assays   similar  to the biotin-
avidin test for Norwalk antigen may provide a
novel way to detect other SRSVs.

The new challenge  in epidemic

Our experience suggests that the investigation
of new outbreaks  of undiagnosed  diarrhoea
may  be  particularly  valuable for  identifying
new agents  and  collecting  reference  reagents
when the outbreaks are recurrent, ongoing, or
located  in  enclosed  environments   such  as
aboard cruise ships,  in nursing homes, or  in
institutional  settings.  Investigations  in these
settings permit collection of large volume stool.s
repeatedly from patients who are acutely ill and
where both acute and convalescent sera can be
drawn   in  large  volume.  Convalescent  sera
which   have  particularly  high  titre  have
encouraged us to go back to patients and collect
a one unit bleed  to  use  for  future reference
work. On the basis of our experience with both
the Snow Mountain agent and Norwalk agent
we believe  that  outbreak investigations  can
provide adequate enriched specimens  suitable
for virus  purification,  preparation  of inocula
for challenge, experiments with cultivation  and
attempts at blind  cloning.  A major advance in
understanding the virology of these organisms
will be essential to advance our understanding
of the true importance  of  these agents. In  the
interim,  enzyme   immunoassays  appear  to
provide  a method to screen outbreak specimens
for the  Norwalk  agent,  caliciviruses, Snow
Mountain  agent and other  unspecified  SRSV

 1  Cubitt VVD, Blacklow XR, Herrmann JE,  \owak \'A.
   \akata  S,  Chiba  S  (1987)  Antigenic relationships
   between human caliciviruses and Xorwalk virus. J Infect
   Dis 156: 806-814.
 2. Dolin R, Levy AG, Wyatt RG, Thornhill TS, Gardner JD
   (1975) Viral gastroenteritis induced by the Hawaii agent:
   jejunal histopathology and seroresponse. Am J Med  59:
 3. Greenberg HB, Wyatt RC, Valdesuso J ct al. (1978) Solid-
   phase microtiter radioimmunoassay for detection of the
   Xorwalk strain of acute nonbactenal, epidemic gastro-
   enteritis virus and its antibodies. J Med Virol 2: 97-108.
 4. Dolin  R, Roessner KD, Treanor JT c'l al. (1986)  Radio-
   immunoassay for detection of the Snow Mountain agent.
   J Med Virol 19: 11-18.
 5. Blacklow \R, Gukor G (1981)  Viral gastroenteritis. N
   Engl J Med 304: 397-406.
 6. Caul EO. Small round  human fecal viruses. In: Pattison
   JR (ed) Parvoviruses and Human  Disease, pp.140-153.
   Boca Raton: CRC Press.
 7. Ci'ba Foundation.  Novel diarrhoea viruses. Chichester:
   John Wiley, 1987. (Symposium no. 128).
 8. Cukor  G,' Blacklow NR (1984) Human  viral gastro-
   enteritis. Microbiol Rev 48: 157-179.
 9. Dolin R, Treanor JJ,  Madore HP (1987) Novel agents of
   viral enteritis in humans. J Infect Dis 155: 365-376.
10. Kapikian AZ,  Chanock RM (1985)  Norwalk group of
   viruses.  In: Fields BN (ed) Virology, pp.1495-1517. New-
   York:  Raven Press.
11. Madeley  CR  (1987)  Viruses  associated with  acute
   diarrhoeal disease.  In: Zuckerman AJ, Banatvala  JE,
    Pattison  JR (ed)  Principles and  Practice ol  Clinical
    Virology, pp. 159-196.  New York: John Wiley.
12. Adler I, Zickl R (1969) Winter vomiting disease. J Infect
   Dis 119: 668-673.
13. Morens DM, Zweighaft RM, Vernon TM el al. (1979) A
   waterborne outbreak of gastroenteritis with secondary
   person-to-person spread: association with a viral agent.
   Lancet I: 964-966.
14. Gary GW, Kaplan JE, Stine SE, Anderson LJ  (1985)
   Detection of Norwalk virus antibodies and antigen with a
   biotin-avidin immunoassay. j Clin Microbiol 22: 274-278.
15. Black RE,  Greenberg HB, Kapikian AZ f(  al.  (1982)
   Acquisition of serum antibody to  Norwalk  virus and
   rotavirus and relation to diarrhea in a longitudinal study
   of young children in rural Bangladesh. J Infect Dis 145:
16. Ando T, Sekine S, Hayaski Y ct al. (1987) Role of small
   round viruses in foodborne gastroenteritis occurring in
   winter:   Outbreaks  in-  metropolitan  Tokyo   and
   epidemiologic research. (JaPl-  Jap J Food Microbiol 4:
17. Greenberg HB, Valdesuso J, Yolken RH cl al. (1979) Role
   of Norwalk virus in outbreaks of nonbacterial gastro-
   enteritis. J Infect Dis 139: 564-568.
18. Kaplan  JE,   Gary  GW, Baron  RC  c'l   al.  (1982)
   Epidemiology ol Norwalk gastroenteritis and the role of
   Norwalk  virus  in  outbreaks  of  acute nonbacterial
   gastroenteritis. Ann Intern Med %: 756-761.
19. Lewis DC, Lightfoot NF, Pether JVS (1988) Solid phase
   immune  electron  microscopy  with human  immuno-
   globulm M for serotyping of Norwalk-like viruses. J Clin
   Microbii.l 26' 938-942.
20. Nakata S, Estes MK, Chiba S (1988) Detection of human
   calicivirus antigen  and antibody  by  enyzme-linked
   immunosorbenl assays. ) Clin Microbiol 26: 2001-2005.

Copyright © 1989, American Society for Microbiology
                                                           Reproduced,  with  permission,  from  the
                                                           Journal  of  Clinical  Microbiology
                                                                                                   Vol. 27, No.
              DNA Hybridization Probe  for Clinical  Diagnosis of

                                     Entamoeba  histolytica

                                             AND DYANN WIRTH1*
  Department of Tropical Public Health, Harvard School of Public Health,1 and Department of Pathology, Brigham and
  Women's Hospital,2 Boston, Massachusetts 02115; Division of Infectious Diseases, University of California, San Diego,
Medical Center,  San Diego,  California 920933; and Laboratorio de Parasitologia, -Colonia Napoles, Mexico City, Mexico4'

                                  Received 17 August 1988/Accepted 15 December 1988

          As an alternative  to microscopic  identification of Entamoeba histolytica  parasites isolated from stool, a
        sensitive and species-specific DNA hybridization probe was made for rapid diagnosis of E. histolytica parasites
        in clinical samples directly applied to nylon membranes. The DNA hybridization probe was made by screening
        a genomic library of a virulent HM-1:IMSS strain of E. histolytica to detect recombinant plasmids containing
        highly repeated parasite DNA sequences. Four plasmid clones that reacted  across Entamoeba species coded for
        highly repeated rRNA genes of E. histolytica. Four other plasmid clones were/?, histolytica specific in that they
        bound to four axenized  and nine xenic  strains of E.  histolytica but did not recognize closely related E.
        histolytica-\ike Laredo, Entamoeba moshkovskii,  or Entamoeba invadens parasites. The diagnostic  clones
        detected as few  as eight  cultured amoebae and did not distinguish between pathogenic and nonpathogenic
        zymodemes of E. histolytica. The diagnostic clones were sequenced and  contained 145-base-pair  sequences
        which appear to be tandemly repeated  in the genome. No stable transcript which  is homologous  to the
        diagnostic DNA was  detected. In a study  of stool samples from Mexico City shown by microscopy to contain
        E. histolytica, Entamoeba coli, Giardia lamblia, Endolimax nana, Trichuris trichiuria, and Chilomastix mesntti
        parasites, the DNA hybridization probe demonstrated a sensitivity of 1.0 and a specificity of 0.93. We conclude
        that the DNA hybridization probe can be used for rapid and accurate diagnosis of E. histolytica parasites.
  Entamoeba  histolytica is an enteric  protozoan parasite
which causes  amoebic dysentery in humans (12, 26). The
most common infection with E. histolytica is asymptomatic,
in which the amoeboid forms of the parasite, called tropho-
zoites, reside  in the lumen of the colon and transform to
cysts,  which are passed in the feces. Such asymptomatic
infections with E. histolytica are frequent among homosex-
ual men in Western countries (1). However, E. histolytica is
a major cause of morbidity  and  mortality in  developing
countries such as Mexico and India, because the trophozo-
ites may invade the colonic mucosa and cause dysentery or
liver abscesses or both.
  Presently, infections with E. histolytica are diagnosed by
light microscopy identification of trophozoites  or of cysts
separated from  stool by dilution,  filtration, and flotation
during centrifugation (5). These methods are time consuming
and  require extensive experience  to assure accuracy of
morphological identification. Past or present invasive amoe-
biasis may be suggested by the identification of antiamoebic
antibodies in patient serum (24).
  In this paper, recombinant  DNA  methodologies previ-
ously used to create sensitive and specific diagnostic probes
for Plasmodium falciparum and Onchocerca volvulus para-
sites (3, 21) are applied to E. histolytica. We show that the
DNA hybridization  probe binds to  highly repeated  and
species-specific DNA sequences in E. histolytica and that
the probe can be used to identify parasites directly in patient
stool samples.
  * Corresponding author.
  Entamoeba strains. Trophozoites of E. histolytica strains
(Table 1), E. histolytica-like Laredo, and Entamoeba mosh-
kovskii were obtained from L. Diamond, National Institutes
of Health, Bethesda, Md., and grown axenically in TYI-S-33
medium (8). Entamoeba invadens trophozoites were ob-
tained from the American Type Culture Collection, Gaith-
ersburg, Md., and cultured axenically. Clinical isolates of E.
histolytica (Table 1) obtained from stools or liver abscess
fluids submitted to the Microbiology  Laboratory of the
University of California,  San Diego,  Medical  Center  were
cultured in Robinson medium (18) and subsequently trans-
ferred to TYSGM medium containing  Escherichia coli 0111
(7). For zymodeme identification, parasites were lysed and
electrophoresed on thin-layer starch gels, and the relative
mobilities of glucose-phosphate isomerase, NADP+ oxido-
reductase, phosphoglucomutase, and  hexokinase were de-
termined (20). Other zymodeme-characterized strains (Table
1) were from P. Sargeaunt, London School of Hygiene and
Tropical Medicine, London, England.
  Preparation of E. histolytica nucleic  acids. For DNA, tro-
phozoites of E. histolytica growing in axenic cultures  were
washed in phosphate-buffered saline, lysed in 10 volumes of
lysis buffer (1% N-lauroyl sarcosyl [sarcosyl], 8 M urea, 0.16
M sodium phosphate buffer  [pH 6.8]),  and  extracted with
phenol-chloroform. DNA in the lysis solution was bound to
a hydroxylapatite column, which was washed in 8 M urea
and  0.16 M phosphate and  then  in 0.19 M phosphate to
remove carbohydrates  and RNA, respectively.  DNA was
eluted with 0.48 M phosphate, dialyzed exhaustively  in 10
mM  Tris and  1 mM EDTA, pH 8,  and concentrated by
ethanol precipitation. RNA was prepared by lysing tropho-
zoites in 5.7  M  guanidinium isothiocyanate, 1% sarcosyl,
and 5% (3-mercaptoethanol and centrifuging them through a

                                     J. CLIN. MICROBIOL.
                             TABLE 1. Strains of Entamoeba species and hybridization results
                                                                                                Hybridization with:
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. histolytica
E. moshkovskii
E. invadens

II ((P)
II (P)
II (P)
II (P)
II (P)

Diamond + +
Diamond + +
Diamond + +
Diamond + +
Reed + +
Sargeaunt + +
Sargeaunt + +
Sargeaunt + +
Sargeaunt + +
Reed + +
Reed + +
Sargeaunt + +
Diamond - +
Diamond - +
ATCC - +
  " Zymodemes as defined by Sargeaunt et al. (20). P, Pathogenic; NP, nonpathogenic.
  * Hybridization with pEH2, an E. histolytica-specific probe, or pEH6, an rRNA gene probe which hybridizes across Entamoeba species.
cesium chloride cushion (14). RNA was radiolabeled for
Southern hybridization by a kinase reaction (14).
  Screening of  an E. histolytica  genomic DNA library for
highly repeated  DNA sequences. Genomic DNA of the HM-
1:IMSS strain of E. histolytica was partially cut with the
restriction enzyme Sau3A to produce fragments ranging in
mplecular size  from 300 to 5,000 base pairs (bp) on 1%
agarose gels. The genomic DNA fragments were ligated into
the plasmid pUC18, which was cut with BamHl and treated
with calf intestinal phosphatase, and were transformed into
JM109 cells. Some 400 colonies were lifted onto nitrocellu-
lose and hybridized with nick-translated (14) genomic DNA
of E. histolytica.
  Characterization  of  recombinant DNA  clones. Twelve
recombinant plasmid clones (pEHl through pEH12) were
selected, purified on cesium chloride gradients (14), radiola-
beled, and hybridized to (i) dot blots of Entamoeba tropho-
zoites (as prepared below), (ii) Southern blots of E. histolyt-
ica DNA which was restriction cut, separated on 1  or 2%
agarose gels (1 p,g per lane), and transferred to nitrocellulose
or Genescreen  Plus nylon membranes (Dupont,  NEN Re-
search Products, Boston,  Mass.), or (iii) Northern (RNA)
blots of E.  histolytica RNA which was separated on dena-
turing gels  containing 1% agarose and 6% formaldehyde or
1% agarose and 1.2% methyl mercury (14) and transferred to
nitrocellulose. For nitrocellulose, hybridizations  were per-
formed at 42°C  overnight in 50% formamide, 10 x  Denhardt
solution, 5x  SSC  (Ix  SSC is 0.15 M Nad plus 0.015 M
sodium citrate), and 500 ^g of herring sperm DNA per ml
(22), followed  by three  30-min  washes in  0.5% sodium
dodecyl sulfate  (SDS) and  0.1% SSC. For Genescreen Plus,
hybridizations were performed at 42°C overnight in 50%
formamide, 1 M NaCl, 1% SDS, 10% dextran, and herring
sperm DNA, followed by two 30-min washes in 2x SSC at
22°C, 2x SSC and 1% SDS at 65°C, and 0.1 x  SSC at 22°C.
The size of the insert DNA in recombinant DNA clones was
determined  by  cutting  with  restriction enzymes and  by
agarose gel electrophoresis.
  DNA sequence analysis. pEH5 and pEHll, two E. histolyt-
j'ca-specific recombinant DNA clones containing  145-bp in-
serts in the opposite orientation  within the polylinker, and
pEH3, pEH4, pEH6, and pEHIO, four recombinant  clones
coding for  rRNA  gene  sequences, were subcloned  into
M13mpl8 and  M13mpl9.  Single-stranded DNA was  pre-
pared from recombinant phage, and the DNA sequence was
determined with the dideoxy chain termination method of
Sanger et al. (19). Both pEH5 and pEHll were sequenced
completely in both directions, while partial sequences of the
rRNA clones were compared with homologous sequences
for the small-subunit rRNA sequence of Dictyostelium dis-
coideum  (15) to locate the clones on the rRNA gene. To
determine whether there were any significant DNA sequence
homologies between the 145-bp E. histolytica-specific diag-
nostic sequence and  other  known DNA  sequences,  the
entire GenBank was searched by using a Lipman-Pearson
FASTN algorithm with a k-tuple of 2.
  Dot blots. E. histolytica DNA was diluted in 10 mM Tris
and 1 mM EDTA,  pH 8, denatured in 0.25 N NaOH, and
spotted onto nylon membranes by using a 96-spot minifold
apparatus (Schleicher  & Schuell, Inc., Keene, N.H.). For
determination of copy number, inserts of pEH5 and pEH6
were electroeluted  from agarose gels, diluted, and spotted
onto nylon membranes and hybridized in parallel with tro-
phozoites. Trophozoites of axenized amoebae were washed
in phosphate-buffered  saline, counted with a hemacytome-
ter, diluted, lysed  in  10 mM Tris, 100 mM EDTA, 0.1%
Triton X-100, and 100  |o,g of proteinase K per ml for 60 min
at 50°C  (3), denatured with NaOH, and spotted.  Xenic
cultures of trophozoites were separated from most bacteria
by means of a Percoll gradient (17), extensively washed in
phosphate-buffered saline, frozen in 10 mM Tris and 100 mM
EDTA, and shipped from San Diego to Boston. Frozen xenic
cultures were thawed, diluted, and lysed before application
to nylon  membranes.
  Determination of the best conditions for detecting E.  his-
tolytica parasites in stool. Optimal conditions for preparing
parasite DNA from stool samples were determined  by trial
and  error experiments in which  cultured E.  histolytica
trophozoites were mixed with uninfected stools, dot blotted,
and hybridized with the E. histolytica-speciftc. clone pEH12.
Samples containing 1,000 to 5,000 E. histolytica trophozoites
and 200 (jil of stool were diluted into 800 |o,l of 10 mM Tris
buffer and then  incubated for 1 h at  37°C with either 0 to 500
mM  EDTA, 0 to 1% SDS, or 0.5 N NaOH. Subsequently,
stool bacteria and debris were removed by centrifugation for
1 min in a microcentrifuge. Then, 100 ^,1 of each supernatant
was  spotted in duplicate onto nylon membranes, denatured
with 0.5  N NaOH if this had  not  been done already, and

VOL. 27,1989
                                                        DNA HYBRIDIZATION PROBE FOR E. HISTOLYTICA     673
hybridized with  radiolabeled pEH12.  Alternatively,  the
1,400-bp insert  of pEH12 was cut out  of the plasmid,
separated on an agarose  gel, and radiolabeled  to  reduce
background hybridization.
  Because it is not possible to make E.  histolytica cysts in
vitro, E. invadens cysts prepared by removing glucose from
the medium and incubating for 4 days (25) were used to
determine the best way to release DNA from cysts. Contam-
inating E. invade ns trophozoites were removed from the cyst
preparations by lysis in  1%  Nonidet P-40,  which did  not
disrupt the cysts. E. invadens cysts (1,000 to 5,000/ml) were
washed three times by centrifugation in  phosphate-buffered
saline and then incubated for 1 h at 37°C either in 1% SDS,
5.7 M guanidinium isothiocyanate, and 0.5 N NaOH in 0.14
M NaCl or in 100 ng of proteinase K per ml. Alternatively,
cysts were frozen in dry ice and ethanol and thawed in a 37°C
water bath three times, boiled for 5 min  in 10% SDS, or
sonicated for 1 min with a probe sonicator. Cyst walls were
pelleted with a microcentrifuge, and samples of the super-
natants were spotted onto nylon filters and hybridized with
radiolabeled genomic E.  invadens  DNA, because  the E.
histolytica-specific probes did not bind to them.
  Preparation of patient stool samples. Stool samples (123)
came from patients referred by their  physicians  to  the
Laboratorio de Parasitologia  in Mexico City for identifica-
tion of intestinal parasites. Cysts  and  eggs were concen-
trated by flotation from 10 to 50 g of stool (5), and parasites
were identified by morphology by microscopic examination.
Parallel samples contained 1 g of untreated stool directly
frozen or cysts concentrated by flotation. The concentrated
cysts were divided into two samples: one frozen directly  and
the other fixed with 10% Formalin before freezing.
  To each patient sample, 0.5 ml of 10 mM Tris and 500 mM
EDTA, pH 8, were added upon thawing  to prevent degrada-
tion of DNA, and samples were frozen in ethanol-dry  ice  and
thawed in water at 37°C for three cycles to rupture cysts.
One hundred microliters of each sample was directly applied
to the minifold after a spin (14,000  rpm, 1 min)  to remove
unlysed  bacteria and  stool debris, and the DNA  on  the
membranes was subsequently denatured with 0.5 N NaOH
and 1.5 M NaCl. To prevent background hybridization of
vector sequences to stool, the 1,400-bp  insert of pEH12, a
recombinant clone specific for E. histolytica DNA, was cut
out with EcoRI and Hindlll, electrophoresed in low-melting-
point agarose, and radiolabeled by random-oligomer priming
(10). After all the DNA hybridizations of the stool samples
were completed  and positive and negative samples were
identified, the microscopic diagnoses were sent from Mexico
City. The sensitivity of the hybridization probe was calcu-
lated by dividing the number of samples identified as positive
with both microscopy and DNA hybridization by the number
positive with microscopy alone. The specificity of the probe
was calculated by dividing the number of samples identified
as negative with both microscopy and the DNA probe by the
number negative with microscopy alone.


  Specificity of the DNA hybridization probe for E. histolytica.
When a genomic library of E. histolytica was screened for
highly repeated DNA sequences, two classes of recombinant
plasmid clones were obtained.  In the first class were four
clones (pEH2, pEH5, pEHll, and pEH12) which appeared
specific for E.  histolytica DNA sequences. These  clones
bound to all the axenized strains of E. histolytica tested but
did not bind to closely related E. histolytica-like Lnredo, E.
                                                                                                 E.m.  EJ.
  FIG. 1. Specificity of binding of recombinant DNA clones to dot
blotted Entamoeba species. (A) Nick-translated pEH2, one of four
diagnostic hybridization probes, bound to trophozoites of HM-
1:IHSS (lane H), 200-NIH (lane 2), and CDC:0784:4 (lane C) strains
of E.  histolytica (E.h) but did not  bind to trophozoites of E.
histolytica-\ike Laredo (lane L), E.  moshkovskii (E.m.), or E.
invadens (E.i.). (B) In contrast, pEH6, a cloned portion of the
small-subunit rRNA gene of E. histolytica,  bound to all Entamoeba
species tested.
moshkovskii, or E. invadens parasites (Fig. 1A and Table 1).
The clones did not bind to human, Giardia, or bacterial DNA
(data not shown). These E. /iwfo/vf/co-specific clones recog-
nized  all E.  histolytica trophozoites tested, including three
pathogenic (II, XIV, and XIX) and two nonpathogenic (I and
III) zymodemes of E. histolytica growing in xenic cultures
(Fig. 2 and Table 1).
  In contrast, a second class of recombinant plasmid clones
(pEH3, pEH4,   pEH6,  and pEHIO)  recognized repeated
DNA  sequences which were conserved across Entamoeba
species (Fig. IB and Table 1).  These nonspecific clones
appeared  to code for portions of the rRNA gene of E.
                   It  of  parasites

               800  400  200  100   50   25   12
                         •  ••  •

        E.  Inv

  FIG. 2. Representative hybridization of the diagnostic recombi-
nant DNA clone  pEH2 to zymodeme-characterized Entamoeba
species.  Radiolabeled pEH2 insert bound  to axenically cultured
HM-1:IHSS  (HM-1) and xenic cultures of isolates 1519, 100, and
1721 (Table  1) with similar affinities,  suggesting about the same
target copy number in each. As a control. pEH2 did not bind to E.
invadens (E. inv) spotted in parallel.

                                      J. CLIN. MICROBIOL.
  FIG. 3. Southern blots of £. histolytica DNA hybridized with the
species-specific clone pEH5. DNAs of HM-1:IHSS (lane H), CDC:
0784:4 (lane C), and 200-NIH (lane 2) were cut with Taql (A) or
Sa«3A (B). The ladder produced by incomplete cutting with Taql
shows that  the sequence recognized is a  tandem  repeat.  With
Sa«3A, there  was a 1.2-kilobase  (Kb)-long restriction  fragment
length polymorphism, which distinguished 200-NIH amoebae from
the other two E. histolytica strains.
histolytica, because they hybridized to Northern blots of E.
histolytica  RNA and  to  bands on  Southern  blots of  E.
histolytica  DNA, which were also recognized by  radiola-
beled E. histolytica RNA (data not shown; 4). Partial DNA
sequencing of the clone pEH6 showed homology with the
small-subunit rRNA gene of D. discoideum (16) beginning at
about 300 bp from the 5' end and continuing to the 3' end of
the gene (data not shown).
  Molecular analysis of the E. histolytica-speciRc recombinant
clones. The four E.  histolytica-specific recombinant DNA
clones showed nearly identical patterns on Southern blots of
restriction-enzyme-cut E. histolytica DNA and  so recog-
nized the same highly repeated sequences within the  parasite
genome (Fig. 3). The  sequences  recognized were  tandem
145-bp long repeats, as shown by  the ladders formed when
the genomic DNA was partially digested with Taql (Fig. 3A).
A degeneracy in the repeat or a modification of a subset of
the repeat  sequences was  suggested by the failure to cut
completely with SaulA (Fig. 3B).
  When parasite  DNA was cut with 5a//3A and hybridized
with the diagnostic clone, there was a 1.2-kilobase  band in
the DNA from the 200-NIH strain that was not present in the
DNA of  the HM-1:IHSS or CDC:0784:4  strain (Fig.  3B).
This restriction fragment length polymorphism may have
potential  value  in developing a method to distinguish strains
of E. histolytica,  as has been shown for restriction fragment
length polymorphisms defined by an rRNA gene probe (4).
  pEH5 and pEHll each  contained single 145-bp inserts,
which were in  opposite directions within the  polylinker of
the plasmids. The DNA sequences  of these inserts were
identical, AT-rich, and contained no significant open reading
                                                                    5' GAATTATTCA AAATGGTCGT CGTCTAGGCC 30

                                                                      AAAATATTTT TTGACCAATT TACACCGTTG 60

                                                                      ATTITCGATT TCCTAAGAAC CTCACCATTT 90

                                                                      TTAATGAAAAGTACTAAATACAAAGTACAA 120

                                                                      TAATTTCTAACTGGGAAAATCGATC3'     145
                                                              FIG.  4. DNA sequence of 145-bp tandemly repeated E. histolyt-
                                                            ica-specific sequence. The DNA sequences of the inserts of pEH5
                                                            and pEHll were identical. Analysis of the sequence revealed no
                                                            significant open reading frames.
frames (Fig. 4). There were no DNA sequences found in a
search of the entire GenBank with significant homology to
the 145-bp E. histolytica diagnostic insert. pEH2 and pEH12
contained 1,800- and 1,400-bp inserts, which were composed
of tandem repeats of 145 bp as determined by partial digests
with Taql of gel-purified inserts (data not shown). The E.
histolytica-speciftc  recombinant  clones  did  not  bind to
Northern blots of parasite RNA, suggesting that the repeated
DNA sequences were not transcribed or that mRNAs were
too few or too unstable to be detected.
  Sensitivity  of the diagnostic clones. The  E. histolytica-
specific recombinant clones were very sensitive, detecting as
few as eight cultured E. histolytica trophozoites in dot blots
after an overnight exposure (Fig. 5A). Using  dilutions of the
145-bp insert spotted  in parallel to measure amoeba DNA
content, we estimated that each parasite contains 50 to 100 fg
of the  145-bp repeat (Fig. 5A). If we assume that each cell
            A        *  of  Parasites

          500   250  125 62    31   16    8   4
           40   20   10    5   2.5   1.2   0.6   0.3

                     pg ol pEHS Insert DNA

           B        # of  Parasites

          500   250 125  62    31   16    8   4
           40   20   10    5   2.5   1.2   0.6   0.3

                     pg of pEH6 Insert ONA
  FIG. 5. Estimation of copy number of inserts of pEH5 and pEH6
in HM-1:IHSS trophozoites.  Cultured parasites  were diluted and
spotted in parallel with dilutions of purified inserts of pEH5, which
contains a 145-bp species-specific DNA sequence, and pEH6, which
contains 1,600 bp of the small-subunit rRNA gene of £. histolytica.
Radiolabeled pEH5 (A) and pEH6 (B) inserts both detected as few
as eight parasites, and there appeared to be about 50 to 100 fg of
each target DNA per parasite.

VOL. 27, 1989
                                                        DNA HYBRIDIZATION PROBE FOR E. HISTOLYTICA

                   *    11  •

                   *    45  •

                   S.    32

                        68  •

  FIG. 6. Representative  hybridization  of the diagnostic  DNA
probe with stool samples. Clinical samples were frozen and thawed,
centrifuged, spotted in triplicate, and hybridized with a radiolabeled
pEH12 insert. On the left is the number of each stool sample.
Numbers 11, 45, and 68 are positive by our criterion.
contains about 500 fg of DNA (9), the 145-bp repeat com-
prises at least 10% of the amoeba DNA. There appeared to
be 50 to 100 fg of the pEH6 insert DNA per trophozoite (Fig.
5B), suggesting that the rRNA genes may also compose 10%
of the parasite genome.
  Conditions for detecting  E. histolytica parasites in stool.
Cultured E. histolytica trophozoites were mixed with unin-
fected stools, dot blotted,  and probed with the diagnostic
clone pEH12 to determine  by trial and error conditions for
preserving target DNA and eliminating background hybrid-
ization. We found that  the addition of 500 mM EDTA to
stool samples was necessary to prevent the rapid degrada-
tion of parasite DNA by stool DNases. A brief microcentri-
fuge spin to remove bacteria and paniculate  debris greatly
increased the amount of sample that could be applied to the
filters. Background  hybridization  was  reduced when  the
1,400-bp insert DNA of the diagnostic sequence pEH12 was
cut out and separated by agarose gel electrophoresis from
the  pUC vector before radiolabeling.  Comparison  of  E.
histolytica trophozoites spotted with  or without stool sug-
gested that the addition  of stool decreases the sensitivity of
the probe by about one  order of magnitude.
  The best method of releasing DNA from the E. invadens
cysts appeared to be by freezing and thawing three times,
which mechanically ruptured  the cyst  walls. Sonication also
released parasite DNA from  the cysts but was  labor inten-
sive and  potentially hazardous because of production  of
aerosols during the procedure. In contrast, treatment  of
cysts with SDS, 0.5 N NaOH, proteinase K, or guanidinium
did not  release the parasite DNA. We decided therefore to
dilute each of the patient  samples  in four parts of  buffer
containing 10 mM Tris and 500 mM EDTA, freeze and thaw
the samples three times, and then pellet away bacteria and
debris before applying the samples to the nylon filters.
  Detection of E. histolytica parasites in clinical samples. The
E. histolytica-specific probe pEH12 was tested against clin-
ical samples from Mexico City in which parasites had been
identified by  microscopy.  To perform a  blind  trial,  the
investigators performing the hybridizations did not know the
results of the microscopy until after the hybridizations were
performed and scored. Stool  or purified cysts and ova were
frozen and thawed to break  open cysts in the  presence of
EDTA to prevent DNase activity,  dot  blotted onto nylon
filters, and  hybridized  with   radiolabeled  insert DNA  of
pEH12  (Fig. 6). The DNA  hybridization  probe correctly
identified 25 of 25 stools containing E. histolytica parasites,
for a sensitivity of 1.0. The probe found seven stools to
contain E. histolytica parasites found by  microscopy to
contain other parasites, not E. histolytica, for a specificity of
0.93.  These so-called false-positives did not appear to be
systematic as they included 4 of 36 Giardia lamblia, 2 of 20
Endolimax nana, 1 of 14 Entamoeba coli, 0 of 8 Chilomastix
mesnili, and 0 of 3 Trichuris trichiuria. Ninety-one samples
were  found  not to contain E. histolytica parasites by the
DNA hybridization probe. Finally,  the probe recognized
equally well  parasites from unfractionated stool  or from
purified cyst preparations.


  DNA hybridization probes for diagnosis of E. histolytica
were  made by identifying and cloning highly repeated and
species-specific amoeba DNA  sequences. In a blind  study,
these diagnostic probes accurately identified E. histolytica
parasites in patient samples from Mexico City.
  Two types of DNA sequences are highly repeated in E.
histolytica. First, there is the tandemly repeated, slightly
degenerate, AT-rich 145-bp long sequence, which is species
specific.  This sequence does not appear to be transcribed
into stable RNA and so is similar to other species-specific
and highly repeated DNA sequences used for diagnosis of
malaria and onchocerciasis (3, 21). The function of these
parasite  sequences and  similar  species-specific repeated
DNA sequences in the mammalian genome is not known
(13).  Second, the  rRNA genes  of  E.  histolytica are also
highly repeated, presumably to amplify the amount of rRNA
produced by the parasites. Because major portions  of the
rRNA genes are conserved across species (15), the complete
rRNA genes were not useful as diagnostic probes. However,
portions of the rRNA genes that appear to be species specific
have recently been identified (4).
  The DNA hybridization probe developed here recognized
all E. histolytica parasites tested but did not recognize E.
histolytica-\ike  Laredo parasites. These Laredo parasites
likely represent a different species  from E. histolytica be-
cause they are nonpathogenic, grow at 25 instead of 37°C, do
not bind  monoclonal antibodies which bind to E. histolytica
(6), and fail  to hybridize with portions of the E. histolytica
rRNA gene (4).
  The DNA hybridization probes did  not  distinguish be-
tween E. histolytica zymodemes. The potential importance
of zymodeme analysis is that it may distinguish potentially
harmful from harmless parasites (20). A recent report sug-
gests that pathogenic and nonpathogenic zymodemes may be
distinguished by monoclonal antibody binding to cultured
trophozoites  (23).  However, whether  the zymodeme of a
parasite is fixed by its genotype or is a phenotypic trait with
differential expression dependent upon the environment of
the parasite  is presently a subject of controversy (16).
  When  compared with microscopy of cysts isolated from
stool by flotation, which is the gold standard for clinical
diagnosis of E. histolytica,  the DNA hybridization probe
performed well. The probe  correctly  identified 25 of 25
specimens containing E.  histolytica and called positive  an-
other 7 of 98 specimens called negative by microscopy. The
latter positives likely represent errors of the probe but may
represent errors  of the microscopist  (samples were  not
available for  review by microscopy). Fortunately,  these
errors did not appear to be systematic in that other common
stool parasites such as G. lamblia, E. nana, T. trichiuria, C.
mesnili,  or E. coli were not overrepresented. Similar false-
positives have been  reported  for  an ELISA that  uses  a

                                        J. CLIN. MICROBIOL.
monoclonal antibody to detect E. histolytica directly from
stool (6). The ELISA correctly identified all samples con-
taining E.  histolytica parasites  seen by microscopy but
showed an apparent false-positive rate of 13 of 46 positives
(6). Previous ELISAs detected trophozoites but not cysts in
stool (2, 11).
  The advantage of either the DNA hybridization probe  or
the ELISA over  microscopy is the brevity  of specimen
preparation: stool is directly spotted onto nylon filters or is
solubilized and placed in microdilution wells,  respectively.
In addition, with the DNA probe as many as 96 specimens
per sheet times 10 to 20 sheets can be hybridized in parallel.
Because  DNA hybridization  is  a simple  and  potentially
inexpensive method for diagnosing E. histolytica, this tech-
nique might be used to survey large numbers of persons  to
estimate accurately the prevalence of E. histolytica infection
and to evaluate  the effects  of control measures to limit
infection with amoebae. Finally, methods to use  the poly-
merase chain reaction to amplify target DNA and increase
the sensitivity of the amoeba probe and to use nonradioac-
tive detection  systems  to replace 32P-labeled probes are
presently being developed in our lab.

  This work was initiated with a grant from the Edna McConnell
Clarke Foundation. Further support came from a grant from the
John  D. and Catherine T.  MacArthur Foundation. J.S. is a Bur-
roughs-Wellcome Life Science Research Foundation Fellow. D.W.
is a Burroughs-Wellcome Scholar in Molecular Parasitology. R.A.-
S. was supported by a predoctoral training grant from CONACYT
(Mexico). S.R. is a Lucille Markey Foundation Fellow.

                    LITERATURE CITED
  1. AUason-Jones, E., A. Midel, P. Sargeaunt, and P. Williams.
    1986. Entamoeba histolytica as a commensal intestinal parasite
    in homosexual men. N. Engl. J. Med. 315:353-356.
  2. Anand, P., B. Malaviya, P. Das, M. A. Mateen, C. M. Habibul-
    lah, and S. R. Das. 1985.  Multilayer-enzyme linked immunoab-
    sorbent assay (MI-ELISA) for detection of Entamoeba histolyt-
    ica trophozoite coproantigen. Immunol. Invest.  14:443-453.
  3. Barker, R. H., Jr., H.  Suebsaeng, W. Rooney, G. C. Alecrim,
    H. V. Dourado, and D.  F. Wirth. 1986. Specific DNA probe for
    diagnosis of Plasmodium falciparum malaria.  Science 231:
  4. Bhattacharya, S., A. Bhattacharya, and L. S. Diamond. 1988.
    Comparison of repeated DNA from strains of Entamoeba his-
    tolytica and  other Entamoeba. Mol. Biochem.  Parasitol. 27:
  5. Biagi, F., and J.  Portilla. 1957.  Comparison of methods of
    examining stools for parasites. Am. J.  Trop.  Med.  Hyg. 6:
  6. del Moro, R., A. Oliva, P. Herion, R. Capin, and L. Ortiz-Ortiz.
    1987. Diagnosis of Entamoeba histolytica in feces by ELISA. J.
    Clin. Lab. Anal. 1:322-325.
  7. Diamond, L. S. 1982. A new liquid medium for xenic culture of
    Entamoeba histolytica. J. Parasitol. 68:958-959.
 8. Diamond, L. S., D. R. Harlow, and C. C. Cunnick. 1978. A new
   medium for the axenic cultivation of Entamoeba histolytica and
   other Entamoeba. Trans. R. Soc. Trop. Med. Hyg. 73:431-432.
 9. Edman, U., I. Meza, and N. Agabian. 1987. Genomic and cDNA
   actin sequences from a virulent strain of Entamoeba histolytica.
   Proc. Natl. Acad. Sci. USA 84:3024-3028.
10. Feinberg, A. P., and B. Vogelstein.  1983.  A technique for
   radiolabeling DNA restriction endonuclease fragments to high
   specific activity. Anal. Biochem. 132:6-13.
11. Grundy, M. S.  1982. Preliminary observations using a multilayer
   ELISA method  for the detection of  Entamoeba histolytica
   trophozoite  antigens in stool samples. Trans. R. Soc.  Trop.
   Med. Hyg. 76:396-400.
12. Guerrant, R. L.  1986. Amebiasis: introduction, current status,
   and research questions. Rev. Infect. Dis. 8:218-227.
13. Lewin, B. 1985. Genes, 2nd ed.( p. 381-392. John Wiley & Sons,
   Inc., New York.
14. Maniatis, T., E.  F. Fritsch, and J. Sambrook. 1982. Molecular
   cloning: a laboratory manual. Cold Spring Harbor Laboratory,
   Cold Spring Harbor, N.Y.
15. McCarroll, R., G. J. Olsen, Y. D. Stahl,  C. R. Woese, and M. L.
   Sogin. 1983. Nucleotide sequence of Dictyostelium discoideum
   small-subunit ribosomal ribonucleic acid inferred from the gene
   sequence: evolutionary  implications.  Biochemistry 22:5858-
16. Mirelman, D.,  R. Bracha, A. Chayen, A. Aust-Kettis, and L. S.
   Diamond. 1986. Entamoeba histolytica: effect of growth condi-
   tions and bacterial associates on isoenzyme patterns and viru-
   lence. Exp. Parasitol. 62:142-148.
17. Reed, S. L., J. G. Curd, I. Gigli, F. D. Gillin, and A. I. Braude.
   1986. Activation of complement by pathogenic and nonpatho-
   genic Entamoeba histolytica. J. Immunol. 136:2265-2270.
18. Robinson, G. 1968. The laboratory diagnosis of human parasitic
   amoebae. Trans. R. Soc. Trop. Med. Hyg. 62:285-294.
19. Sanger, F., S. Nicklen, and A. R. Coulson. 1977. DNA sequenc-
   ing  with chain-terminating inhibitors. Proc.  Natl. Acad. Sci.
   USA 74:5463-5467.
20. Sargeaunt, P. G., J. E. Williams, and J. D.  Greene. 1978. The
   differentiation  of invasive and non-invasive E. histolytica  by
   isoenzyme electrophoresis.  Trans. R. Soc. Trop. Med. Hyg.
21. Shah, J. S., M. Karam, W. F. Piessens, and D. F. Wirth. 1987.
   Characterization of an  Onc/iocerca-specific DNA clone from
   Onchocerca volvulus. Am. J. Trop. Med. Hyg. 37:376-384.
22. Southern, E. M. 1975. Detection of specific sequences among
   DNA fragments separated by gel electrophoresis. J. Mol. Biol.
23. Strachan, W. D., W. D. Spice, P. L. Chiodini, A. H. Moody, and
   J. P. Ackers. 1988. Immunological differentiation of pathogenic
   and non-pathogenic isolates of Entamoeba histolytica. Lancet
24. Trissl, D. 1982. Immunology of Entamoeba histolytica in human
   and animal hosts. Rev. Infect. Dis. 4:1154-1184.
25. Vazquezdelara-Cisneros, L. G., and A.  Arroyo-Begovich. 1984.
   Induction of encystation of Entamoeba  invadens by removal of
   glucose from the culture medium. J. Parasitol. 70:629-633.
26. Walsh, J. A. 1986. Problems in recognition and diagnosis of
   amebiasis: estimation of the global magnitude  of morbidity and
   mortality. Rev. Infect. Dis. 8:228-238.

                                                      Reproduced, with permission,  from the
                                                      Journal  of Clinical  Microbiology
Copyright © 1989, American Society for Microbiology
                                                                                                   Vol. 27, No. 3
           Diagnosis  of Giardia  lamblia  Infections  by Detection  of

                                   Parasite-Specific  Antigens

  Infectious Disease Section, Veterans Administration Medical Center, Denver, Colorado 802201; Division of Infectious
        Diseases, Department of Medicine, University of Colorado School of Medicine, Denver, Colorado 80262  ;
           Department of Pediatrics, Tulane University School of Medicine, New Orleans, Louisiana 701123;
            Program in Infectious Diseases and Clinical Microbiology, University of Texas Medical School,
                  Houston, Texas 770304; and Centers for Disease Control, Atlanta,  Georgia 30333s

                                Received 28 September 1988/Accepted 30 November 1988

          Antigen detection methods may facilitate diagnosis of Giardia lamblia in stool specimens. As determined by
        sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis and immunoblotting, G. lamblia cysts and
        trophozoites share several antigens, especially in the 65-kilodalton and 30- to 34-kilodalton regions. By using
        blind  methods, we compared results obtained  by counterimmunoelectrophoresis using cyst-immune rabbit
        serum and by enzyme-linked immunosorbent assay (ELISA) using trophozoite-immune rabbit serum with
        results obtained by microscopic examination of a preserved, concentrated,  and permanently stained stool
        specimen. Results were similar when these three methods were used to examine 118 stool specimens from
        clinical microbiology  laboratories (53 specimens with G. lamblia) and specimens from 239 day-care-center
        toddlers (39  specimens with G. lamblia). Compared with microscopy, we found, for counterimmunoelectro-
        phoresis and ELISA, respectively: sensitivity, 88 versus 94%; specificity, 97 versus 95%; positive predictive
        value, 86 versus 76%; negative predictive value, 98 versus 97%; and concordance, 89%. The false-positive rate
        by EUSA was 24% (10 of 42) in day-care-center toddlers but only 3% (1 of 32) in healthy adults (P < 0.04)
        as corroborated by  microscopy.  This discrepancy  suggests  that the ELISA may be  more  sensitive than
        microscopy,  which is considered the reference standard, and that results may be dependent, in part, on the
        epidemiology of the infection in the study subjects.
  Reliable  information on  the  epidemiology  and natural
history of Giardia lamblia infections has been limited be-
cause the sensitivity of microscopic diagnosis of the parasite
in concentrated stool specimens varies from 46 to 95% (1, 6,
12,13). Examination of multiple samples may be required for
diagnosis, particularly from persons with early (9) or chronic
infections (3). Other diagnostic methods,  some of which
require sampling of the small bowel, are more expensive,
uncomfortable, and invasive (6,10,13). Stool antigen detec-
tion systems for the diagnosis of G. lamblia infections have
been reported to be highly sensitive and specific (2, 7,14, 23,
24). These methods may permit  large numbers  of stool
samples to be tested rapidly and may  reduce technician time
and bias among observers.
  Standard diagnostic methods are designed to detect the
specific  stages of the  life cycle of G. lamblia, cysts and
trophozoites. Microscopic examination of stool  samples
most often detects cysts (4, 21), whereas duodenal aspiration
and biopsy primarily identify trophozoites (6).  In contrast,
antigen detection systems, although also designed to detect
the distinct stages of the life cycle, show similar diagnostic
accuracies. We identified the  antigens detected by two
antigen detection  methods to understand the  relationship
between cyst and  trophozoite antigens and  to better define
and standardize the assay systems. We performed the first
blinded comparison of three noninvasive methods  for diag-
nosis  of G. lamblia infections from  stool  samples:  micro-
  * Corresponding author.
  t Present address: Infectious Disease Section 111F. Veterans
Administration Medical Center, 1 Veterans Dr., Minneapolis, MN
scopic examination, detection of cyst antigens by counter-
immunoelectrophoresis (CIE), and detection of trophozoite
antigens by enzyme-linked immunosorbent assay (ELISA).
  (This work was presented in part at the 27th Interscience
Conference on  Antimicrobial Agents and  Chemotherapy,
New York, N.Y., 4 to 7 October 1987.)

  Antigen preparation. G. lamblia  Portland-1 trophozoites,
graciously provided by  Ernest Meyer,  were grown and
prepared as  previously described  (8).  Purified cysts were
prepared from human fecal specimens, as  previously de-
scribed (2), and were kindly provided, in part, by Charles
Hibler. The cyst purity was assessed by microscopy (2), and
the preparations were sonicated, centrifuged at 1,000 x g for
30 min, and preserved at -70°C.
  Antigen analysis. Rabbit sera used in the antigen detection
systems were compared by ELISA as previously described
(8), except that the solid-phase antigens were whole frozen
G. lamblia Portland-1 trophozoites (104 organisms per well)
or purified G. lamblia cysts (103 cysts per well). The end-
point was defined as the highest dilution of rabbit serum that
resulted in an optical density (OD) greater than 0.200 at 405
nm. The  specificities of the antibodies detected were deter-
mined by adsorption of immune serum four times for 45 min
each time at 37°C on a circular rotator with each of the
following Formalin-fixed organisms (number of organisms
per adsorption):  G. lamblia Portland-1 trophozoites  (6 X
106); Trichomonas vaginalis cultured from  a symptomatic
patient (6 x 106); Cryptosporidium  oocysts and  freshly
excysted sporozoites (107), provided by Charles Sterling and

432    JANOFF  ET AL.
                                     J. CLIN. MICROBIOL.
Michael Arrowood; Campylobacter jejuni (109); enterotoxi-
genic Escherichia coli (109); and Candida albicans (109).
  Whole-cell preparations of purified G. lamblia  cysts and
trophozoites were resolved by sodium dodecyl sulfate-poly-
acrylamide gel  electrophoresis  (SDS-PAGE) as previously
described (8). Immunoblots of trophozoite antigens were
prepared with  serum  from New  Zealand White rabbits
immunized with G. lamblia cysts or trophozoites and used in
the CIE and ELISA systems, respectively (2, 8).
  Stool specimens. In Denver, Colo., in 1985, we obtained (i)
single unpreserved stool samples from each of 239 randomly
selected toddlers in day-care centers (T.  Novotny, R.  S.
Hopkins, P. Shillam, and E. N. Janoff, Abstr. Annu. Conf.
Epidemic Intelligence Service, Centers for Disease Control,
1986) and (ii) 118 samples  submitted to the clinical microbi-
ology laboratories of the University of Colorado, Veterans
Administration Medical Center, and Denver General Hospi-
tal from 86 persons with diarrhea and 32 healthy adults with
no history of gastrointestinal symptoms or known G. lamblia
exposure within the past 2 months. Fresh stool specimens
were preserved in 10% Formalin or in poly vinyl  alcohol, and
a third fraction was stored  within 12 h of collection from test
subjects without preservative at —70°C until processed.
  Microscopy. The entire 22-mm2  cover slip from each
Formalin-ether (or Formalin-ethyl acetate) concentrate was
examined at a magnification of x 10, as was each trichrome
stain of polyvinyl alcohol-preserved stool specimen exam-
ined under an oil immersion lens. Each concentrated speci-
men was stained for Cryptosporidium oocysts and sporozo-
ites with the modified  Kinyoun acid-fast  (cold)  technique
(11), and the entire slide was scanned  for  5  min at a
magnification of x 40. Positive specimens were confirmed by
examination under an oil immersion lens. Identification of G.
lamblia by microscopic examination of preserved concen-
trate or of a permanently stained smear was considered the
reference standard for  the antigen  detection systems. The
numbers of G. lamblia organisms detected were  graded  as
rare to few, moderate,  or many in 31 of 39 positive speci-
mens from day-care-center toddlers.
  Antigen detection. The CIE assay was performed as pre-
viously described (2), with serum from cyst-immune rabbits.
For the ELISA  system, New Zealand  White  rabbits were
immunized with axenically grown  G.  lamblia Portland-1
trophozoites. A  crude serum immunoglobulin fraction was
precipitated with ammonium sulfate and  dialyzed,  and a
portion was  conjugated  to horseradish peroxidase (22).
Wells of microdilution plates (Immulon II; Dynatech Labo-
ratories, Inc., Alexandria, Va.)  were coated with unlabeled
rabbit anti-G. lamblia antibody in 100 mM carbonate buffer
(pH 9.0) overnight at 4°C. Freshly  thawed fecal specimens
were diluted  1:2 with 50% fetal calf serum-0.01  M phos-
phate-buffered saline (pH 7.2) plus  2 mM phenylmethylsul-
fonyl fluoride (Sigma Chemical Co., St. Louis,  Mo.) and
0.1% thimerosal, vortexed, and allowed to settle for 30 min.
The supernatant was incubated for 1 h  at 37°C in duplicate
wells.  After  wells were washed  three times  by using a
multichannel pipette, peroxidase-conjugated rabbit anti-G.
lamblia antibody was applied, incubated for 1 h at 37°C, and
developed  with  o-phenylenediamine in 53 mM  citrate-94
mM phosphate buffer (pH 5.0) for 15 min. The reaction was
stopped with 2.5 N H2SO4, and the A490 was measured on a
microELISA plate reader  (MR580; Dynatech).  Positive and
negative controls were included on each plate. Specimens
which produced an OD of less  than 0.100 were considered
negative, those with an OD between 0.100 and 0.250 were
considered equivocal,  and those which  produced an OD
above 0.250 were considered positive on the basis of results
previously  obtained from specimens characterized by mi-
croscopy (C. V. Knisley, P. G. Englekirk, L. K. Pickering,
S. West, and E.  N. Janoff, Am. J. Clin.  Pathol., in press).
Preliminary studies showed  that an OD of 0.250 resulted
from the addition of 250 trophozoites (1 ng of protein) to a
test well. We decided in advance to consider any specimens
which produced equivocal results upon CIE or ELISA to be
negative for the purpose of analysis.
  We evaluated each method on the basis of the number of
specimens  tested by that method; all  357 specimens were
tested by microscopic examination, 327 (92%) were tested
by CIE, and 312 (87%) were tested by ELISA. We compared
the three assays  by using results  from the 282 (79%) speci-
mens tested by all three methods. To investigate the repro-
ducibility of test results, nine stool samples were tested in a
blind manner in duplicate by CIE, as were  six by ELISA.
ODs by ELISA were compared for 20  samples tested twice
on separate days to establish the  test-retest reliability.
  Statistics. Results were evaluated by the  chi-square and
Fisher exact tests as indicated.


  Antigen analysis.  The trophozoite-immune rabbit  serum
used in the ELISA system showed a liter of 1:25,600 with the
trophozoites and a titer of 1:400 with  the cysts as the
solid-phase capture antigens. The cyst-immune rabbit serum
used in the CIE system showed  a titer  of 1:200 with the
trophozoites and a titer of  1:3,200 with  the  cysts as the
solid-phase  capture antigens. Preimmune  rabbit  serum
showed a titer of <1:100, the  lowest  dilution tested, with
both  antigens. Trophozoite-immune  serum  showed the
greatest decrease in titer following adsorption with G. lam-
blia  trophozoites,  from 1:25,600 to 1:3,200. Titers after
adsorption with other protozoal, bacterial, and fungal organ-
isms remained 1:12,800.
  Eight distinct bands, with major bands at 88, 60, 55,48, 33,
and 31 kilodaltons (kDa), were resolved from the soluble G.
lamblia cyst preparation (Fig. 1). Silver staining of solubi-
lized trophozoites also produced bands of 189, 88, 75, 68, 55,
47, and  34 kDa (13). Immunoblotting of whole trophozoites
with cyst-immune and  trophozoite-immune  rabbit sera re-
vealed common trophozoite antigens of 185,  65, and 57 kDa
and produced the strongest reactivity with heavy bands of 33
and 30 kDa (Fig. 2).
  Antigen detection in stool samples. The sensitivities and
specificities of the two antigen detection methods (CIE and
ELISA) were similar to results obtained by microscopy with
stool  specimens from  day-care-center toddlers  and with
specimens  submitted to clinical laboratories (Table 1). The
numbers of organisms identified by microscopy in 31 of 39
day-care-center samples examined in which  G. lamblia was
identified were rare to  few  in 48%, moderate in 16%, and
many in 35%. Both CIE and ELISA showed negative results
for two specimens  in which G.  lamblia  was detected  by
microscopy. Two stool  samples taken from patients 2 weeks
after they started therapy for documented G. lamblia infec-
tions were negative by microscopy; one sample was positive
by CIE, as was the one sample tested by ELISA. Thirteen
stool  samples with other parasites, including Entamoeba
histolytica (n = 2), Cryptosporidium spp.  (n = 7), Blastocys-
tis hominis (n = 3), and others (Entamoeba coli, Entamoeba
nana, Trichuris trichuris) showed negative  results by both
methods. Nine samples tested in a blind manner in duplicate
by CIE gave  concordant results (seven  positive and two

VOL. 27, 1989
                                                               ANTIGEN DETECTION FOR GIARDIA LAMBLIA     433

                              92-  —'-

                              66-  1
  FIG. 1. Silver-stained 12% polyacrylamide SDS-PAGE prepara-
tions of solubilized G. lamblia. Lanes:  a, cysts from an infected
patient; b, axenically grown Portland-1 strain trophozoites.
 negative), as did five of six samples tested by ELISA (three
 of four positive and two negative).  The test-retest reliability
 of ODs produced with 20 samples tested on two occasions by
 ELISA was high (correlation coefficient  = 0.953), and the
 mean values did not differ by the paired / test. The presence
 or absence of symptoms, such as diarrhea  and abdominal
 pain, in  the patients from whom  the specimens were ob-
 tained did not affect the  sensitivities or specificities of the
 two methods  (data not  shown).  The concordance  rates
 between the two antigen detection methods was 92% (196 of
 214) for  samples without G. lamblia and 81% (55 of 68)  for
 samples  with G. lamblia as corroborated  by microscopy.
  The predictive  value  of a  positive result was lower in
 samples  from day-care-center toddlers than from  routine
  FIG. 2. Immunoblots of G. lamblia Portland-1 trophozoites re-
solved  on  12% polyacrylamide SDS-PAGE gels.  Nitrocellulose
strips  were reacted with serum from rabbits immunized with G.
lamblia cysts (lane a), G. lamblia trophozoites (lane b), and preim-
mune serum (lane c).

samples from clinical microbiology laboratories. When mi-
croscopy results  were considered the  reference standard,
false-positive rates were higher among day-care-center tod-
dlers, among whom rates  of exposure to and carriage of G.
lamblia are often higher (1,16; Novotny et.  al., Abstr. Annu.
Conf.  Epidemic  Intelligence Service, 1986)  than among
healthy adults with no known G. lamblia exposure, who are
less likely to be infected (Table 2).
  Of  the samples tested  by  CIE, 5%  (16 of  327) showed
equivocal results  (grey  zones);  13%  of  those tested  by
ELISA (41 of 312) showed equivocal results. As part of our
protocol prior to testing, any  equivocal result was defined as
a negative test result for the purposes of analysis.  In the
clinical setting, these results  would have been called equiv-
ocal and a second specimen would have been requested. No
G. lamblia was identified by microscopy in 81% (13 of 16)
             TABLE 1. Comparison of antigen detection methods with microscopy for diagnosis of G. lamblia infections
Source of specimens
(diagnostic test)"
No. tested
No. (%) with
G. lamblia
Positive predictive
value (%)
Negative predictive
value (%)
 Day-care-center toddlers
  MICRO                          239         39 (16)          100          100
  CIE                             212         35 (17)          88           97
  ELISA                           218         42(19)          94           95

 Clinical microbiology laboratory
  MICRO                          118         53(45)          100          100
  CIE                             115         48(42)          96           97
  ELISA                            94         42(45)          90           91
  " MICRO. Microscopic examination of a single preserved, concentrated, and permanently stained specimen; CIE, CIE assay for cyst antigen; ELISA, ELISA
for trophozoite antigen.
  * Microscopic examination is considered the reference standard for this analysis.

434    JANOFF ET AL.
                                                    J. CLIN. MICROBIOL.
   TABLE 2.  Comparison of false-positive rates using antigen
   detection methods in two epidemiologically distinct groups
                               No. false-positive/total no. (%) by":
Day-care-center toddlers          5/35 (14)
Healthy adults                   0/32* (0)
10/42 (24)
 1/32* (3)
  ° For the purposes of this analysis, results are considered as false-positives
if the sample produced a positive result by ELISA or CIE but no organisms
were identified by microscopic examination of a single concentrated and
permanently stained sample from each subject.
  b P < 0.05, Fisher exact test.
and 90% (37 of 41) of these samples, which were equivocal
by  CIE and ELISA, respectively. Exclusion of these sam-
ples from  the  analysis  did not  significantly  change  the
evaluation of the tests.


  This  study establishes that both the CIE and  ELISA
systems are reliable diagnostic tests for the diagnosis of G.
lamblia infections on the basis of established criteria (18).
We performed a blinded comparison with a reference stan-
dard of diagnosis on stool samples from patients  with  a
spectrum of clinical manifestations.  We described the clini-
cal setting and exact methods for the study, the precision of
the tests, and variation among observers. The advantage of
antigen detection systems is that they provide rapid, sensi-
tive, and reproducible results for large numbers of samples;
they may be less labor-intensive, less expensive, and subject
to  less test-to-test variation than diagnosis of G.  lamblia
infections  by microscopy. These assays serve as a useful
adjunct to  microscopic examination, which  is directed to
detect a spectrum of enteric infections.
  CIE and ELISA antigen detection systems showed results
similar to those of microscopic examination in a large,
prospective, blinded  comparison for the diagnosis of G.
lamblia in stool  specimens.  Although  cysts  are  usually
identified five times more often than  trophozoites in stool by
microscopy (4, 21),  these assays,  which used rabbit sera
hyperimmune to two distinct phases of the life cycle of G.
lamblia, gave comparable results. The recognition of com-
mon antigens by these sera,  especially those bands of 65
kDa, as first described by Rosoff and Stibbs (17) and Stibbs
et al. (19), and those between 30 and 34 kDa, which are also
recognized by humans infected with G. lamblia (8, 20), may
in part explain this similarity. These  cross-reactive antigens,
which  are  recognized by the cyst-immune serum, are prob-
ably derived from trophozoites within the cysts. This cross-
reactivity  was not  described  by another group, perhaps
because of differences in immunization technique or in the
method of antigen preparation (5). The CIE assay may well
detect other protein  or nonprotein cyst wall antigens (5, 25).
  Antigen  detection may be more sensitive, rather than less
specific, than the reference standard, microscopic examina-
tion (14). The specificity of microscopic diagnosis in these
university-affiliated and state laboratories is high in Denver,
probably  because  of the  frequency of  diagnosis  of  the
parasite and the long-term interest in G. lamblia in Colorado.
We demonstrated the specificity of the trophozoite-immune
serum in  vitro by  adsorption studies and by the  lack of
cross-reactivity with other protozoa in clinical samples. The
false-positive rate was higher among day-care-center tod-
dlers,  among whom carriage rates of G. lamblia are high (1,
16; Novotny et al., Abstr.  Annu.  Conf. Epidemic Intelli-
gence Service,  1986) and who may excrete few organisms (3,
9, 15), than among adults, who are at low risk of infection.
False-negative  samples were uncommon  with both antigen
detection methods. The true sensitivities of these tests could
be determined  best during an outbreak by testing acutely ill
persons, with multiple specimens being taken from the same
subjects (14), and by using seeding studies.
  We conclude that  antigen detection systems provide an
acceptable and convenient adjunct to microscopic examina-
tion for noninvasive  method of diagnosis of G. lamblia in
stool specimens.  Further identification of parasite-specific
antigens may  be useful to  standardize  and improve  the
clinical  utility of diagnostic methods and to define the natural
history  of these infections.


  We thank Paul G. Englekirk (University of Texas Medical School
at Houston), Kurt Albrecht and Larry Briggs (Colorado Department
of Health), Elizabeth Lindsay, Sandy Barber,  Sally Dembeck, Stan
Merritt,  and Jan  Monahan (University of Colorado  Clinical Micro-
biology Laboratory), and Ben Miyahara and John Paisley (Denver
General  Hospital Microbiology Laboratory) for technical assistance
and support and Charles Sterling and Michael Arrowood (University
of Arizona School of Veterinary Science) for providing Ciyptoxpo-
ridium oocysts.
  This research was supported in part by the Veterans Administra-
tion Research Service and by Public Health Service grant HD13021
from the National Institutes of Health.

                   LITERATURE CITED
 1.  Black, R. E., A. C. Dykes, S. P. Sinclair, and J. G. Wells. 1977.
    Giardiasis in day-care centers:  evidence  of person-to-person
    transmission. Pediatrics 60:486-491.
 2.  Craft, J. C., and J. D. Nelson. 1982. Diagnosis of giardiasis by
    counterimmunoelectrophoresis of feces.  J.  Infect.  Dis. 145:
 3.  Danciger, M., and M. Lopez. 1975. Number of Giardia in feces
    of infected children. Am. J. Trop. Med. Hyg. 24:237-242.
 4.  Garcia,  L. S., T. C. Brewer, and D. A. Bruckner. 1979. A
    comparison of the formalin-ether concentration and trichrome-
    stained  smear methods for the  recovery and identification of
    intestinal protozoa. Am. J.  Med. Technol. 45:932-935.
 5.  Gillin, F. D., D. S. Reiner, M. J. Gault, H. Douglas, S. Das, A.
    Wunderlich,  and J. F. Sauch. 1987. Encystation and expression
    of cyst antigens by Giardia lamblia  in  vitro.  Science 235:
 6.  Gordts, B.,  W.  Hemelhof, K.  Van Tilborgh,  P.  Retore, S.
    Cadranal, and J. P. Butzler. 1985. Evaluation of a new method
    for routine in vitro  cultivation of Giardia  lamblia from human
    duodenal fluid. J. Clin. Microbiol. 22:702-704.
 7.  Green, E. L., M. A. Miles, and  D. C. Warhurst.  1985. Immun-
    odiagnostic detection of Giardia antigen  in  faeces by a rapid
    visual enzyme-linked immunosorbent assay.  Lancet ii:691-693.
 8.  Janoff,  E. N., P. D. Smith, and M.  J.  Blaser. 1988. Acute
    antibody responses to Giardia lamblia are depressed in patients
    with the acquired immunodeficiency syndrome. J. Infect. Dis.
 9.  Jokipii, A. M. M., and L. Jokipii. 1977. Prepatency of giardiasis.
    Lancet  i:1095-1097.
10.  Kamath, K. R., and R. Murugasu. 1974. A comparative study of
    four methods for detecting Giardia lamblia in  children with
    diarrheal disease and malabsorption. Gastroenterology 66:16-
11.  Ma, P., and R. Soave. 1983. Three-step  stool  examination for
    cryptosporidiosis in 10 homosexual men with protracted watery
    diarrhea. J. Infect.  Dis. 147:824-828.
12. Mahalanabis, D., T. W. Simpson, M. L.  Chakraborty, C. Gan-
    guli, A. K. Bhattacharjee,  and K. L. Mukherjee. 1979. Malab-
    sorption of water-miscible vitamin A in children  with giardiasis

VOL. 27. 1989
    and ascariasis. Am. J. Clin. Nutr. 32:313-318.
13.  Naik, S. R., N. R. Rau, and V. K. Vinayak. 1978. A comparative
    evaluation of three stool samples, jejunal aspirate and jejunal
    mucosal impression smears in the diagnosis of giardiasis. Ann.
    Trop. Med. Parasitol. 72:491^92.
14.  Nash,  T. E., D.  A. Herrington, and M.  M.  Levine.  1987.
    Usefulness  of an enzyme-linked immunosorhent  assay for de-
    tection  of Giardia antigens in  feces. J.  Clin. Microbiol. 25:
15.  Olveda, R.  K., J. S. Andrews, Jr., and E. L. Hewlett.  1982.
    Murine giardiasis: localization of trophozoites and small bowel
    histopathology during the course of infection. Am.  J. Trop.
    Med. Hyg. 31:60-66.
16.  Pickering,  L. K., W. E.  Woodward, H. L.  DuPont, and P.
    Sullivan. 1984. Occurrence of Giardia lamblia in children in
    day-care centers. J. Pediatr. 104:522-526.
17.  Rosoff,  J. D., and H. H. Stibbs. 1986.  Isolation and identification
    of a Giardia lamblia specific stool antigen (GSA 65)  useful in
    coprodiagnosis of giardiasis. J. Clin. Microbiol. 23:905-910.
18.  Sackett, D.  L., R. B. Haynes, and B. Tugwell.  1985. Clinical
    epidemiology. Little, Brown & Co.,  Boston.
19.  Stibbs,  H. H., M. Samadpour, and J. F. Manning. 1988. Enzyme
    immunoassay for detection of Giardia lamblia cyst antigens in
    Formalin-fixed and unfixed human stool. J. Clin. Microbiol.
20. Taylor, G.  T., and W.  M. Wenman. 1987. Human immune
    response  to Giardia lamblia infection.  J.  Infect.  Dis. 155:
21. Thornton, S. A., A. H. West, H. L. DuPont, and L. K. Pickering.
    1983.  Comparison of methods  for  identification  of Giardia
    lamblia. Am. J. Clin. Pathol. 80:858-860.
22. Tijssen, P., and E. Kurstak. 1984. Highly efficient and simple
    methods for the preparation of peroxidase and active peroxi-
    dase-antibody  conjugates  for  enzyme-immunoassays.  Anal.
    Biochem. 136:451-457.
23. Ungar, B. L. P., R. H. Yolken, T. E. Nash, and T. C. Quinn.
    1984. Enzyme-linked immunosorbent assay for the detection of
    Giardia lamblia in fecal specimens. J. Infect. Dis. 149:90-97.
24. Vinayak, V. K., F.  K. Kum, R. Chandna, K. Venkateswarlu, and
    S. Mehta. 1985. Detection of Giardia lamblia antigen in feces by
    counterimmunoelectrophoresis. Pediatr. Infect. Dis. 4:383-386.
25. Ward, H. D., J. Alroy, B. I. Lev, G. T. Keusch, and  M. E. A.
    Pereira. 1985. Identification of chitin as a structural component
    of Giardia cysts. Infect. Immun. 49:629-634.

        THE LANCET, JANUARY 3, 1987
             Reproduced,  with
             from  the  Lancet.
          SIR,—The protozoan parasite cryptosporidium can cause self-
        limiting gastroenteritis in otherwise normal individuals or a chronic
        life-threatening  diarrhoea  in  immunologically  compromised
        individuals, especially those with AIDS.1 Laboratory diagnosis has
        relied on  the  demonstration  of oocysts  in faeces by  modified
        Ziehl-Neelsen or auramine-phenol staining.2 One of us (D. P. C.)
        has also described an indirect fluorescent antibody test using human
        sera containing anti-cryptosporidium  IgA antibodies (HIgA) for
        confirmatory purposes.1-1 A monoclonal antibody for diagnostic use
        has been reported3A but few data on sensitivity and specificity have
        been presented. We report here a rapid and accurate method using a
        monoclonal antibody in an indirect immunofluorescent antibody
        (IFA) test, which seems to be highly  specific and more sensitive
        than  conventional techniques. It  allows  the  demonstration of
oocysts in hitherto unsuitable samples and their differentiation from
bodies that resemble cryptosporidium morphologically.
   Balb/c mice were immunised with an enriched oocyst suspension
(from the faeces  of an infected calf) which had been gamma-
irradiated and  sonicated before  intraperitoneal  inoculation at
about 10" oocysts per inoculum four times over 24 days. Spleen cells
taken on the 27th  day after the first inoculation were fused with a
drug-sensitive  tumour  cell  line  (JK-Ag8653),  and  antibody-
secreting hybridomas  were identified by an indirect IFA test. A
single cell line (MAB Cl) secreted antibody of the IgM isotype. At
first this  was used in an indirect assay, but we have since conjugated
the antibody directly with fluorescein isothiocyanate (FITC) and
developed a technique with a total assay time of less than 30 min.
   Faecal samples were homogenised and concentrated2 and tested
in parallel with conventional staining and HIgA techniques (table).
MAB  Cl  proved to be specific:  it reacted with all samples
containing cryptosporidium oocysts identified by staining and it did
not react with morphologically similar bodies or related parasites. It
also  seems  more  sensitive  than  conventional   tests,   since
cryptosporidium oocysts in  follow-up  samples from confirmed
cases that did not stain by conventional techniques were detectable
with MAB Cl. The monoclonal antibody was superior to the HIgA
because of reagent reproducibility and higher specificity (the HIgA
contained antibodies to other organisms).
   The   epidemiology   of  cryptosporidiosis   is   incompletely
understood, and this reagent may permit the specific identification
of the oocysts and the exclusion of other morphologically similar
bodies in environmental samples that are not suitable for testing by
conventional staining.

  We thank Dr D. Blcwitt and Dr K. Angus, Moredun Research Institute,
Edinburgh; Dr A. Balfour, Public Health Laboratory, Leeds; and Dr V.
McDonald, Houghton Poulry Research Centre for donations of antigens.
                                Division 01" Microbiological Reagents
                                  and Quality Control,
                                Central Public Health Laboratory,
                                London NW9 5HT

                                Public Healdi Laboratory, Rhy!

                                Division of Microbiological Reagents
                                  and Quality Control,
                                Central Public Health Laboratory

                                Animal Department,
                                Central Public Health Laboratory

                                Division of Microbiological Reagents
                                  and Quality Control,
                                Central Public Health Laboratorv
                                             J. MCLAUGHLIN
                                             D. P. CASEMORE

                                             T. G. HARRISON

                                             P. J. GERSON

                                             D. SAMUEL
                                             A. G. TAYLOR

                                           species a "new" human
                                 I  Casemore DP, Sands RL, Curry A. Crv/Jmspi
                                     pathogen. J C/in Pal/ml 1985,38: 1321-36.
                                 2. Casemore DP, Armstrong M, Sands RL. Laboratory diagnosis of cryptosporidiosis. J
                                     C/m Palhal 1985; 38: 1337-41.
                                 3. Sterling CR, Arrowood MJ. Detection of GYv/Ti^yonWntm sp infections using a direct
                                     immunofluorescem assay Palmlr hilea Du 1986,5:5139-42.
                                 4. Sterling CR, Seeger K, Sinclair NA. Ov/>fi)x/>onJiw;/ as a causative agent of traveller's
                                     diarrhoea. 7 Infal Dis 1986; 153: 380-81.

Assessment of sensitivity (specimens from individuals with cryptosporidiosis)
Human faeces (known positive or follow-up samples)
Enriched oocyst suspensions
Calf faeces
Lamb faeces
Guineapig faeces
Mouse faeces (infected by material from a calf)
Assessment of specificity
Human faeces (non-cryptosporidiosis)
Horse faeces (contains CLB and AFB)
Water filtrates (contains CLB and AFB)
Eimena tenella
Toxoplasma gondii (trophozoites)
No of specimens


Number of specimens positive by;








       *Higher numbers of oocysts were detected with MAB Cl than with the MZN and AP techniques in 16 samples.
       t7 additional follow-up samples were negative with all methods.
       MZN = modified Ziehl-Neelsen; AP = auramine-phenol; HIgA = human anti-cryptosporidium IgA antibody; MAB Cl = ami-cryptosporidium monoclonal antibody; AFB = acid-fast
       bodies; CLB = cryptosporidium-like bodies which were similar but morphologically distinct; NT = not tested.
*U.S.  GOVERNMENT PRINTING OFFICE:  1990—726-115/D09667