RESEAAssessment of Water Use for Estimating Exposure to Tap WaterContaminantsGayle H. Shimokura,1 David A. Savitz,1 and Elaine Symanski9'Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC 27599 USA; 2School of PublicHealth, University of Texas Health Science Center, Houston, TX 77030 USA

Epidemiological studies examining the association between exposure to tap water contaminants(such as chlorination by-products) and disease outcomes (such as cancer and adverse reproduc-tive outcomes) have been limited by inaccurate exposure assessment. Failure to take into accountthe variation in beverage and tap water consumption and exposure to volatile contaminantsthrough inhalation and dermal absorption can introduce misclsification in assessing the associ-ation between exposure to tap water contaminants and health. To refine exposure assessment oftap water contaminants, we describe in detail the tap water consumption, showering, andbathing habits of pregnat w'omen and their male partners as aessed by a questionnaire and a3-day water diary. We found good agreement between questionr and 3-dy water diary val-ues for drinking water intake (Peaons r = 0.78) and for time spent showering (r = 0.68) andbathing (r = 0.78). Half of the paricipants consumed tap water on a regular basis with an overallmean * 1 standard deviation (SD) of 0.78 d0.51 /day. Our results fiuther suggest that ful-timeemployees, compared to women working part-time or less, have more heterogeneous consump-tion patterns over time. Seventy-nine percent of women and 94% of men took showers for anaverage of 11.6 ± 4.0 min and 10.4 ± 4.8 mm, respectively. Badts were taken more frequendy bywomen than men (21% vs. 3%) for an average of 22.9 ± 10.1 min and 21.3 ± 12.4 mi, respec-tively. Thus, these patterns of tap water use should be considered in the design and interpreta-tion of environmental epidemiology studies. Aiy words bathing, chlorination by-products, chlo-roform, exposure assessment, pregnat women, tap water, tap water consumption, showering,trihalomethanes. Environ HeathPerpect 106:55-59 (1998). [Online 15 January 1998]bttpq://enpnatl.niehs.nih.govldocs/1998J106p55-59shimokura/abstract.hsml

The main methodological difficulty withepidemiological studies that examine theassociation between tap water contami-nants, including chemical and microbialagents, and disease outcomes lies in accurateassessment of individual exposure to tapwater. Often, exposure to tap water is classi-fied by the source of water (i.e., ground orsurface supply), the type of water treatment,the concentration of the tap water contami-nant at the water utility serving the partici-pant's residence, or some combination ofthe three. Broad exposure definitions,which do not take into account intra- andinterindividual sources of variation either intap water concentrations of the contami-nant or in individual tap water consump-tion, can introduce exposure misclassifica-tion and thereby distort associationsbetween tap water exposure and adversehealth outcomes (1). In addition, recentstudies have shown that exposure to volatiletap water contaminants through inhalationor dermal absorption may be equal to orgreater than exposure through ingestion (2).Thus, the assessment of showering, bathing,and swimming habits can be importantwhen evaluating tap water exposure.

This report provides information fordeveloping exposure assessment methodsfor epidemiologic studies of adverse healtheffects associated with tap water exposure.

We describe in detail the tap water con-sumption, showering, and bathing habitsof pregnant women and their male part-ners as assessed by a questionnaire and a 3-day water diary. We also consider a rangeof possibly important behaviors that mayaffect tap water exposure, such as locationof tap water consumption, frequency ofshowers, and duration of baths. We exam-ine the pattern of variation in tap waterconsumption over time and across individ-uals and compare the questionnaire datawith the water diary data to evaluate theconsistency between methods of data col-lection.

Materials and MethodsSubject selection. From October 1994 toJanuary 1995, potentially eligible participantswere identified by reviewing the medicalrecords of the Obstetrics and GynecologyClinic of the University of North CarolinaHospitals, Chapel Hill, North Carolina.Initially, the study was restricted to womenwho were no more than 14 weeks pregnantat the time of medical record review, had atleast 10 years of schooling, lived with theirmale partners, and resided within the servicearea of the local water utility (the OrangeWater and Sewer Authority). To obtain asufficient number of participants, the inclu-sion criteria were expanded in December

1994 to include women who were up to 30weeks pregnant and who lived outside theaforementioned service area.

Participants were initially contacted bymail; if they did not refuse to participate atthat time, the interviewer called to recruitthem for the study and arrange a time for ahome interview. Both the woman and herpartner were asked to participate. Out of 79couples initially identified during the med-ical record review, 71 (90%) actually met theeligibility criteria. Of these, 19 declined bymail (27%), 16 declined by phone (23%),and 36 were interviewed (51% of those eligi-ble). Informed consent was obtained from allinterviewed participants. Of those 36 coupleswho were interviewed, 33 couples and onewoman (but not her partner) completed the3-day water diary. Only those who complet-ed both the interview and water diary wereincluded in the analysis.

Interview and water diary. During theinterview, participants provided informationabout daily water consumption and weeklyconsumption of cold water-based beveragesthat had been prepared in the home duringthe previous 3 months. Questions were askedabout how long beverages were typicallystored and whether the storage container wasopen or dosed. We distinguished among bev-erages that were made from filtered tap water,unfiltered tap water, and bottled water.Filtered tap water was defined as tap waterthat had been filtered using an activated car-bon point-of-use device. The interviewer alsogathered information about the frequencyand duration of baths and/or showers. Thelevel of ventilation while bathing wasaddressed by asking the participants aboutwhether they usually kept the bathroom dooropen or closed, bathroom window(s) open orclosed, and the ventilation fan on or offwhilebathing. The frequency and duration of otherwater-related activities such as swimming(including water aerobics and use of jacuzzis);bathing children or pets; washing dishes,

Address correspondence to D.A. Savitz, CB# 7400,McGavran-Greenberg, University of North Carolinaat Chapel Hill, Chapel Hill, NC 27599-7400 USA.This research was supported in part by CooperativeAgreement CR 820076 with the U.S. EnvironmentalProtection Agency. E.S. was supported in part byNational Institute of Environmental Health SciencesTraining Grant T32-ES07018. We also thankChristine Moe for reviewing the manuscript.Received 7 July 1997; accepted 8 October 1997.

Environmental Health Perspectives * Volume 106, Number 2, February 1998

11 lc es

55

Articles * Shimokura et al.

dothes, or cars by hand; mopping floors; andfrequent handwashing were also ascertained.

At the time of the interview, participantswere instructed on how to complete theprospective water diary. For each day of thestudy, participants were asked to record everybeverage they consumed, its volume andtemperature (cold or hot), where and when itwas consumed, and whether it was preparedwith tap or bottled water. Unfortunately, fil-tered tap water was not an option listed onthe diary record form. For eight participants,the type of water question was inadvertendyomitted altogether. Frequency and durationof exposure to tap water via showering andbathing, swimming, bathing children, wash-ing dishes or dothes by hand, or by someother activity were also recorded. The partic-ipants were allowed to choose which 3 daysthey recorded in their water diary, althoughthey were asked to commit to the daysagreed upon during the interview. They wereinstructed to choose at least one nonworkingor weekend day and to complete the diary ata time when they were in town and healthy.Eighty-two percent of households completedthe water diary within 1 week from the dateof interview.

With the exception of the comparisonbetween water diary records and interviewdata, missing information (320 incompletediary entries from 35 participants) wasimputed to avoid exduding the entire diaryof that participant from the other analyses.Where possible, the missing information wasimputed from the interview since similarquestions were asked in both the interviewand the diary. Otherwise, similar assump-tions as those by Ershow and Cantor (3)were made: milk, juices (except reconstitutedfrozen fruit juices), alcoholic beverages, andsoft drinks were assumed to be made frombotded water; water, powdered drinks, cof-fee, and tea were assumed to be made fromunfiltered tap water (3). Information onselected missing beverage volumes and loca-tions (comprising 0.3% of all beverageentries) was imputed using information fromadjacent diary record entries. Where bever-age temperature was missing, beverages wereassumed to be cold, except for coffee and tea,which were assumed to be hot. All iced teabeverages were considered to be hot bever-ages because it was assumed that they wereheated during preparation, although no dif-ferentiation was made between instant andbrewed tea.

Statsitical methods. For each day of thewater diary, the total volume of 1) all bever-ages (hereafter referred to as total water), 2)drinking water and beverages made with tapwater (hereafter referred to as tap water),and 3) cold drinking water and tap waterbeverages consumed at home (hereafter

referred to as cold tap water consumed athome) were calculated and averaged overthe participant's 3 days of water records.Consumption of each beverage type forfemale and male participants was summa-rized by calculating the mean and standarddeviation (SD), along with selected per-centiles, of the overall distribution. Forfemale participants, the distribution of dailyintake of cold tap water consumed at homewas also stratified by employment status.Male participants were not stratified byemployment status because nearly all (85%)male participants were employed full time.

Because location of consumption (homeor other location), temperature of beverage,and type of cold beverage (drinking water ortap water-based beverage) are potentiallyimportant determinants of tap water expo-sure, tap water consumption was stratifiedby these factors. Means of total water, tapwater, and cold tap water consumption werecompared across sex and employment status(full time, not full time) using analysis ofvariance. Of those female participants whoconsumed tap water at home, the percentageof participants who stored any tap waterbeverage, the average number of days storedfor each beverage type, and the percentageof containers that were dosed were analyzedseparately for drinking water and tap water-based beverages.

The frequency and duration of showersand baths were determined, as well as thefrequency of nonventilated showers in win-ter and summer. A nonventilated showerwas defined as a shower where the bathroomdoor and window (if present) were closedand the ventilation fan (if present) was off.

The pattern of variation over time (i.e.,from one day to the next) and across indi-viduals was examined using the 3-day diaryinformation on daily consumption of unfil-tered cold drinking water (iced tea exclud-ed). By employing a one-way random-effects analysis of variance model, these(natural log-transformed) data were used toestimate the between-person and within-person components of variance (becausesome participants reported no daily con-sumption of drinking water, 1 was added toeach value before logarithms were comput-ed). Variance components were estimatedseparately for male and female participantsand for female participants employed fulltime or employed part-time or less. Theintraclass correlation coefficient (ICC),defined as the percent of total variance dueto the between-person variance component,was estimated as well. Because of the inher-ent bias of this formula, the results usingthis formula were confirmed using an unbi-ased but more intricate ICC formula (4).

To compare the questionnaire data with

the water diary data (3-day averages),Pearson's and Spearman's correlation coeffi-cients were computed. Comparisons werealso made by evaluating the mean differencebetween values obtained from the two datacollection methods. Differences were deter-mined for each participant first and then theaverage and 95% confidence interval (CI) ofall the differences for the study populationwere calculated (5). We compared three val-ues, one for each route of exposure: volumeof drinking water consumed per day athome (tap water-based beverages wereexduded because volume estimates from theinterview data were not obtained), the totalnumber of minutes spent showering perday, and the total number of minutes spentbathing per day. All analyses were conduct-ed using the Statistical Analysis System ver-sion 6.10 (SAS Institute, Cary, NC).

ResultsParticipant characteristics. Thirty-fourpregnant women and 33 of their male part-ners completed both the water diary andthe interview. Overall, they tended to behighly educated nonsmokers in their twen-ties or thirties and having their first child(Table 1). Three-fourths of the participantswere white, with the remainder split evenlybetween African-American and Asian eth-nicities. Although the women's gestationalages ranged from 12 to 37 weeks, most(59%) were in their second trimester ofpregnancy. Men were much more likely tobe employed than women, though theywere equally likely to be students. Abouthalf of the participants consumed tap waterat home on a regular basis, and the majority(82%) was served by treated water supplies.

Daily water consumption. Consumptionof total water, tap water, and cold tap waterat home (Table 2) as reported by the ques-tionnaire indicates that participants con-sumed on average close to the NationalAcademy of Sciences' standard assumedintake of 2 liters water/day (6), but less than20% was cold tap water at home. Overall,men and pregnant women had similar waterconsumption habits, and only cold tap waterconsumption at home differed by employ-ment status.

Table 3 shows mean daily tap waterconsumption by location, temperature, andbeverage type (drinking water or tap water-based beverage) as reported by the question-naire. Women employed part-time or lessconsumed the most tap water overall, buttotal consumption levels among sex andemployment groups did not vary signifi-cantly (p>0.10). Women employed part-time or less averaged the most tap waterconsumption at home (due to consumingmore drinking water), whereas outside the

Volume 106, Number 2, February 1998 * Environmental Health Perspectives56

Articles * Tap water exposure assessment

Table 1. Selected lifestyle, physical, and demograph-ic characteristics of study participants

Demographic Female Malevariable (n= 34) (n= 33)aEducation (%)Some college 15 _bCollege graduate 24 _bSome graduate/professional 62 _bSmoking during pregnancy (%)Yes 0 9No 100 91Number of children at home (%)0 67 671+ 33 33Age distribution (%)20-29 41 3330-39 59 5840-47 0 9

Ethnicity (%)White 76 74Other 24 26

Gestational age (%)First trimester 6 -

Second trimester 59Third trimester 35 -

Working full time (%)Yes 53 85No 47 15

Student (%)Yes 26 27No 74 73

Principal type of waterconsumed at home (%)CNo water consumption 9 21Tap water 53 49Bottled water 24 9Filtered water 15 21

Source of household water (%)Municipal water supply 82 82Well or spring water 18 18

&One male did not complete the water diary.bPaternal educational level was not obtained.CAs reported in the interview.

home, men consumed the most tap water(due to consuming more hot tap water).Differences in consumption patterns by sexwere less pronounced relative to differencesin consumption by extent of employment(full time vs. not full time), suggesting thatemployment status is a stronger determi-nant of tap water consumption patternsthan sex.

Slightly over half (53%) of the womendrank tap water at home. Of those, 28%stored the water for an average interval (± 1SD) of 1.4 ± 0.8 days. In contrast, ofwomenwho consumed at least one tap water-basedbeverage (59%; n = 20), all stored the bever-age an average interval (± 1 SD) of 5.1 ± 3.1days. Most (89%) beverage containers weredosed during storage.

The large proportion of total variancedue to between-person variance in womenworking part-time or less (ICC = 0.81, with

Table 2. Summary of participants' daily water intake (liters)

Percentile distributionNumber Mean ± SD 10 25 50 75 90

Daily intake of total waterWomen 34 1.86 ± 0.73 1.17 1.45 1.75 2.08 2.33Men 33 1.68 ± 0.70 0.70 1.34 1.59 2.08 2.39

Daily intake of tap water,Women 34 0.78 ± 0.51 0.20 0.43 0.62 1.12 1.39Men 33 0.78 ± 0.51 0.25 0.34 0.81 1.10 1.23

Daily intake of cold tap water at homebAll women 34 0.37 ± 0.40 0 0.02 0.26 0.55 0.97Employed full time 18 0.28 ± 0.30 0 0.04 0.15 0.53 0.60Employed part-time 16 0.47 ± 0.48 0 0.01 0.42 0.73 1.04or less

All men 33 0.29 ± 0.35 0 0 0.15 0.51 0.69'Filtered tap water excluded.bFiftered tap water and iced tea excluded.

Table 3. Average daily tap water consumption (liters) by location, temperature and beverage type

Womenemployed Women employed

All women full time part-time or less All men(n= 34) (n= 18) (n= 16) (n= 33)

Consumption at homeCold drinking water 0.25 0.18 0.34 0.19Cold tap water-based beverages 0.11 0.10 0.13 0.10Total home consumption of cold tap water 0.37 0.28 0.47 0.29Hottap water 0.15 0.11 0.20 0.14Total home consumption of tap water 0.52 0.38 0.67 0.43Consumption outside the homeCold drinking water 0.15 0.18 0.11 0.16Cold tap water-based beverages 0.04 0.06 0.02 0.03Total cold tap water consumed outside the home 0.19 0.23 0.14 0.18Hottap water 0.08 0.12 0.02 0.17Total consumption outside the home 0.26 0.36 0.16 0.35

Total cold tap water 0.56 0.51 0.61 0.47Total hot tap water 0.23 0.23 0.22 0.31Total tap water consumption 0.78 0.74 0.83 0.78

a total variance of 8.75) suggests that almostall the variability in unfiltered cold tap waterconsumption habits is due to individual dif-ferences, with very little day-to-day variabili-ty. On the other hand, full-time employedwomen and men had much lower between-person variance (ICC = 0.42, with a total vari-ance of 7.38, and ICC = 0.49, with total vari-ance of 8.80, respectively), indicating greaterheterogeneity in consumption patterns overtime, possibly between workdays and non-workdays. Overall, among all women, 61%of total variance was due to between-personvariance (CC = 0.61, with a total variance of7.83). A comparison of results between theapplied ICC formula with a second unbiasedformula resulted in negligible differences.

Showering, bathing, and other water-related activities. Nearly all participantsreported showering (90%); of those, all butnine showered at least daily for an average ofabout 10 min, with women tending to show-er longer than men (Table 4). More men thanwomen showered, and more showers wereventilated in summer than in winter. Women

took more baths than men; of those whoreported bathing (22%, n = 15), almost halfreported daily bathing. Baths lasted on aver-age twice as long as showers.

Swimming was reported for only five(7.5%) of the study participants during theinterview; of those, only one reportedswimming in her water diary. However,because this study was conducted in thewinter, assessments made at other times ofthe year are likely to yield very differentresults. The discrepancy between the ques-tionnaire and the water diary may be dueto the fact that swimming is not a dailyactivity and therefore may not have beenwell represented during the 3 days of thewater diary. The other water-related activi-ties were too diverse to be evaluated, butwashing dishes was the most commonlyreported activity, while occupational expo-sures such as watering gardens and milkingcows were the most time intensive.

Questionnaire and diary assessmentmethods. The questionnaire and the diaryshowed good agreement, especially for

Environmental Health Perspectives * Volume 106, Number 2, February 1998 57

Articles * Shimokura et al.

Table 4. Showering and bathing habits according tointerview

Female Male(n = 34) (n= 33)

Participants who shower 79 94at least daily (%)

Average duration of shower 11.6 ± 4.0 10.4 ± 4.8± SD (min)

Distribution of showerduration (%)5-9 min 17.9 37.510 min 35.7 37.511-30 min 46.4 25.0

Ventilated showers (%)During winter season 43 56During summer season 61 75

Participants who bathe 18 3at least daily (%)

Participants who bathe 21 3occasionally (%)Average duration of 22.9 ± 10.18 21.3 ± 12.4bath ± SD (min)

Distributon of bath duration.8-19 min 41.7 5020 min 16.7 021-0 min 41.7 50

SD, standard deviation.8One value missing.

drinking water intake (Table 5). ThePearson's and Spearman's correlation coeffi-cients ranged from 0.52 to 0.78, which arefairly large for exposures assessed in epi-demiologic studies (7) but they do indicatesome discrepancies between data collectionmethods. On average, greater beverage con-sumption and shower duration values, butsmaller bathing duration values, wereobtained from the interview relative to thediary. However, in each case, the differencebetween the two estimates was small.

DiscussionQestionnaire and diary assessment methods.In this study we applied two data collectiontools: the questionnaire, intended to reportexposures averaged over time, and theprospective water diary, designed to reflectcurrent day-to-day exposure. Questionnairessuffer from potential errors in recall and theinability to report finely detailed information.The quality of data gathered from diarymethods depends on the extent to which eachparticipant is willing to measure and recordcurrent habits and the representativeness ofthe days for which the diary is completed (8).

Conducting an interview is logisticallypreferable to completing diaries because datacollection is less burdensome and providesdata on exposures averaged over time. Asdemonstrated in this study, the interviewprovided essentially the same information asthe prospective water diary. Overall meandifferences between assessment methods for

Table 5. Comparison of interview-based questionnaire data with diary results

Drinking water Time spent Time spentintake at home showering bathing

(liters/day) (min/day) (min/day)Number of participants 42a68b 60bMean value (questionnaire) 0.75 10.5 2.1Mean value (diary) 0.40 9.8 3.3Pearson's correlation coefficientc 0.78 0.68 0.78Spearman's correlation coefficientc 0.75 0.63 0.52Mean difference between questionnaireand diary valuesd (Cl) 0.35 (0.21-0.48) 0.7 (-0.62.0) -1.2 (-2.4-0.0)

Cl, 95% confidence interval.'rwenty-five participants were not included in this analysis because they were not asked about the volume of their water glasses.bSeven participants had missing values.cAnalyses compared individual results obtained from the frequency questionnaire to 3-day averages computed from diary entries.dDifferences were first computed for each participant and then averaged over the study population.

consumption of drinking water (0.35 1/day),showering (0.7 min/day), and bathing (1.2min/day) are negligible, and the interviewand mean diary record intakes correlate rea-sonably well. While the 3-day water diary isnot necessarily the gold standard, it appearsthat individuals have the ability to estimatewith a fair amount of precision their owntrue mean values for water intake and fortime spent on water-related activities, at leastover relatively short periods in the recentpast. However, the frequency questionnairemay not perform as well if used to estimateconsumption patterns from a more tempo-rally remote time period.

Other limitations of our data for assess-ing the reliability between diary and ques-tionnaire records were that the questionnairefocused only on home consumption of coldtap water and that the respondents averagedintake levels over working and nonworkingdays, when consumption habits are expectedto differ. Furthermore, the diary did notprovide enough information to evaluateeffects of weekday or consecutive days.Finally, no direct validation of reported bev-erage volume, water type, or shower orbathing duration was done for either datacollection method.

Our study sample came from an educat-ed and health-conscious population becauseonly women who received prenatal care,who had at least a high-school education,and who volunteered to participate wereincluded in the study. Their water con-sumption habits differ from the general pop-ulation because they can probably affordalternatives to tap water and may tend toswitch from tap water to bottled or filteredwater in order to protect their fetus' health.The percentage of individuals who reportedbottled or filtered water as the principal typeof water consumed at home was higher forwomen (39%) than for men (30%). Ourresults are somewhat higher than anotherstudy, which reported that 17% of healthypostpartum women consumed mainly bot-tled water during the first 3 months of their

pregnancy (9). On the other hand, anadvantage of this educated convenience sam-ple is that they should be more compliantabout completing and returning the waterdiary than a random sample of the generalpopulation would be. Consequently, dataquality and correlations between the diaryand questionnaire may be less favorable inother populations.

Daily water consumption. The distrib-ution of daily consumption of total water,tap water, and cold tap water consumed athome is relevant to a wide range of healthconcerns with drinking water, includingwaterborne exposure to agents such as fluo-ride, arsenic, toxic metals, and pesticides,and microbial pathogens such as hepatitisA, cryptosporidia, and Norwalk virus. It isalso useful for developing a standard intakedistribution so that an individual's waterintake level can be compared to others ofthe same sex and age (3).

On average, we found that pregnantwomen and their male partners had similarwater consumption intake and that morecold than hot beverages were consumed.Although minor differences in consumptionby sex were noted, bigger differences wereobserved by extent of employment (full timevs. not full time). These data indicate thatemployment status had a greater influencethan sex in shaping individual water con-sumption habits. It may be that sex waslargely a surrogate for working status inanother study in which all the men andnone of the women were employed outsidethe home (JO).

Although the majority (67%) of all tapwater beverages consumed by pregnantwomen was consumed at home, a significantamount (33%) was consumed either atwork, a restaurant, or at a friend's or rela-tive's house. Therefore, exposure assessmentstudies that only take into account waterconsumption at home or that only take intoaccount levels of tap water contaminants atthe residences of study participants would bedisregarding approximately one-third of

Volume 106, Number 2, February 1998 * Environmental Health Perspectives58

Articles * Tap water exposure assessment

their participants' total exposure due to tapwater consumption.

To our knowledge, the only other studyof tap water and total water consumption inpregnant women relied on data from theU.S. Department of Agriculture's 1977-1978 Nationwide Food ConsumptionSurvey (11). While our results are slightlyhigher for total water consumption (1.9I/day vs. 1.6 I/day), we observed lower tapwater intake levels than those reported previ-ously (0.78 I/day vs. 1.1 I/day). The smallsample size in our study makes it difficult tomake any definitive statements; however,this difference may reflect a decline in tapwater consumption because the current U.S.population is drinking more bottled andcanned beverages than ever before (12).

A substantial proportion of the variationin consumption levels was due to differencesbetween individuals for women workingpart-time or less (ICC = 0.81), but less so forfull-time employed women (ICC = 0.42)and men (ICC = 0.49). These findings sug-gest that use of a standard water intake levelin studies evaluating health effects associatedwith drinking water contaminants would notbe appropriate because it would likely resultin attenuated exposure-response relation-ships and substantial exposure misdassifica-tion. These results also indicate that data col-lection methods for tap water exposureshould employ repeated measures over time.Such measures of tap water exposures couldbe obtained through multiple records ofdaily water intake or by one-time question-naires that collect information on tap waterexposures integrated over time by therespondent. Similarly, although not the focusof this study, the sources of variation inwaterborne contaminant levels would alsohave to be considered when assessing expo-sure to drinking water contaminants.

Assessment ofbehavioral indicators oftapwater exposure. For assessing total exposureto tap water, it is not appropriate to combineall tap water exposures into one summaryestimate because each type of tap water expo-sure represents a slighdy different combina-tion of inhalation, ingestion, and dermalabsorption and likely targets different organs

(X). We were not able to relate behaviors tobiological measures of exposure, so we areunable to directly address contributions fromdifferent sources.

Temperature and storage time can affectthe type of tap water exposure, especially forvolatile tap water contaminants such as tri-halomethanes (THMs). For instance, colddrinking water tends to have the highestconcentration ofTHMs because it is usuallyconsumed straight from the tap, and if it isstored, it is stored briefly (average of 1.4days). Relative to initial tap water levels, hotbeverages will have lower THM levelsbecause of volatilization by heating as willcold beverages stored for extended periodsin open containers. We found that all coldbeverages prepared at home with tap waterwere stored for at least 1 day, and most werestored for longer periods (average of 5.1days) before consumption, thereby permit-ting equilibrium between the air and water.However, because most of the storage con-tainers were closed (which limits the amountof THM that escapes into the householdair), such beverages would still represent apotential source of THM exposure (P.Singer, personal communication). Non-volatile chlorination by-products and otherwater contaminants would, of course,remain in the water and be ingested regard-less of temperature or storage conditions.

Showering and bathing have been foundto be significant sources of THM exposure(2) although the importance of these expo-sure routes to other water contaminants arenot well defined. We found that most peo-ple take showers daily and that their showerslast an average (± 1 SD) of 10 ± 5 min, sothat dermal and inhalation exposure fromthis activity is substantial on a populationlevel. Although bathing is not a daily activityfor the general population, it can represent asignificant source of exposure among per-sons who bathe regularly. Because nearlyone-fifth of the women in our study bathedat least daily and an additional 21% bathedoccasionally, exposure through this activityshould not be ignored, particularly amongpregnant women.

ConclusionStudies of the potential health problemsassociated with tap water have been limitedby crude estimates of tap water exposure.The incorporation of a comprehensiveassessment of behaviors related to con-sumption and other uses of water, coupledwith improved estimates of levels of tapwater contaminants in drinking water,should enhance our ability to evaluate asso-ciations between water contaminants andthe potential for associated health risks.

REFERENCES

1. Lynch CF, Woolson RF, Cantor KP. Chlorinated drink-ing water and bladder cancer: effect of misclassifica-tion on risk estimates. Arch Environ Health44:252-259 (1989).

2. Weisel CP, Jo WK. Ingestion, inhalation, and dermalexposures to chloroform and trichloroethane from tapwater. Environ Health Perspect 104: 48-51 (1996).

3. Ershow AG, Cantor KP. Total water and tap waterintake in the United States: population-based esti-mates of quantities and sources. Life SciencesResearch Office Monograph. Bethesda,MD:Federation of American Societies forExperimental Biology, 1989.

4. Winer BJ. Statistical Principles in ExperimentalDesign. 3rd ed. New York:McGraw-Hill, 1991.

5. Bland JM, Altman DG. Statistical methods for assess-ing agreement between two methods of clinical mea-surement. Lancet 1(8976):307-310 (1986).

6. Assembly of Life Sciences, Safe Drinking WaterCommittee. Drinking Water and Health. Contract no.68-01-3139. Washington, DC: National Academy ofSciences, 1977.

7. Willett WC. Nutritional Epidemiology. NewYork:Oxford University, 1990.

8. Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA,Litin LB, Willett WC. Reproducibility and validity of anexpanded self-administered semiquantititative foodfrequency questionnaire among male health profes-sionals. Am J Epidemiol 135:1114-1126 (1992).

9. Hertz-Picciotto I, Swan SH, Neutra RR, Samuels SJ.Spontaneous abortions in relation to consumption oftap water: an application of methods from survivalanalysis to a pregnancy follow-up study. Am JEpidemiol 130:79-93 (1989).

10. Sharrett AR, Carter AP, Orheim RM, Feinleib M. Dailyintake of lead, cadmium, copper, and zinc from drink-ing water: the Seattle study of trace metal exposure.Environ Res 28:456-475 (1982).

11. Ershow AG, Brown LM, Cantor KP. Intake of tapwater and total water by pregnant and lactatingwomen. Am J Public Health 81:328-334 (1991).

12. U.S. Department of Agriculture. Nationwide FoodConsumption Survey: Continuing Survey of FoodIntakes by Individuals, Women 19-50 Years and TheirChildren 15 Years, 1 Day. NFCS, CSFII Report 86-1.Hyattsville, MD:Nutrition Monitoring Division, HumanNutrition Information Service, 1987.

Environmental Health Perspectives * Volume 106, Number 2, February 1998 59