Household Exposures to Drinking Water Disinfection By-Products: Whole Blood Trihalomethane Levels

www.nature.com/jea

Household exposures to drinking water disinfection by-products: whole blood trihalomethane levels

LORRAINE C. BACKER,a DAVID L. ASHLEY,b MICHAEL A. BONIN,b FREDERICK L. CARDINALI,b STEPHANIE M. KIESZAKa AND JOE V. WOOTENb aDivision of Environmental Hazards and Health Effects, National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, Georgia bDivision of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, Georgia

Exposure to drinking water disinfection by-products (DBPs), such as trihalomethanes (THMs), has been associated with bladder and colorectal cancerin humans. Exposure to DBPs has typically been determined by examining historical water treatment records and reconstructing study participants' water consumption histories. However, other exposure routes, such as dermal absorption and inhalation, may be important components of an individual's total exposure to drinking water DBPs. In this study, we examined individuals' exposure to THMs through drinking, showering, or bathing in tap water. Thirty-one adult volunteers showered with tap water for 10 min (n=11), bathed for10 min in a bathtub filled with tap water( n=10), or drank 1 l of tap water during a 10- min time period (n=10). Participants provided three 10 ml blood samples: one sample immediately before the exposure; one sample 10 min after the exposure ended; and one sample 30 min (for shower and tub exposure) or 1 h (for ingestion) after the exposure ended. A sample of the water (from the tap, from the bath, or from the shower) was collected for each participant. We analyzed water samples and whole blood for THMs (bromoform, bromodichloromethane, dibromochloromethane, and chloroform) using a purge-and-trap/gas chromatography/mass spectrometry method with detection limits in the parts-per- quadrillion range. The highest levels of THMs were found in the blood samples from people who took 10-min showers, whereas the lowest levels were found in the blood samples from people who drank 1 l of water in 10 min. The results from this study indicate that household activities such as bathing and showering are important routes for human exposure to THMs. Journal of Exposure Analysis and Environmental Epidemiology (2000) 10, 321±326.

Keywords: chlorination, disinfection, disinfection by-products, drinking water, household exposures, trihalomethanes.

Introduction One difficulty in demonstrating the association between exposure and health outcome is the long latency Chlorine is the most commonly used chemical for disinfect- period for cancer development and the subsequent need ing U.S. watersupplies (King and Marrett,1996); however, to reconstruct water consumption and exposure histories chlorine reacts with organic compounds in the water to overa long period of time. Forexample, King and produce halogenated hydrocarbon by-products, including Marrett (1996) found that people exposed to THM trihalomethanes (THMs; e.g., chloroform, bromoform, levels 50 g/l for 35 years or more had 1.63 times the bromodichloromethane, and dibromochloromethane). Ex- risk of bladder cancer compared with those exposed for posure to these disinfection by-products (DBPs) has been less than 10 years. McGeehin et al. (1993) found that associated with cancer, particularly bladder cancer in the number of years of exposure to chlorinated drinking humans (Cantoret al., 1987, 1998; Zierleret al., 1988; waterwas positively associated with riskforbladder McGeehin et al., 1993; King and Marrett, 1996; Freedman et cancer. The risk for bladder cancer increased for longer al., 1997). Because of the human health risk associated with exposure durations (i.e., the odds ratio, OR, was 1.8 for exposure to THMs that results from chlorinating drinking 30 years of exposure). water, the maximum contaminant level (MCL) for total Anotherdifficulty in examining the health effects THMs in finished (treated) drinking water is 80 g/l (U.S. produced by DBPs is determining how to define EPA, 1998). exposure. Investigators have used water quality indicators such as the source water (ground water vs. surface water), THM concentrations in the water supply, and the 1. Address all correspondence to: Dr. Lorraine C. Backer, Environmental historical chlorine dose for the water supply as indicators Epidemiologist, National Center for Environmental Health, 4770 Buford of exposure to DBPs. Other studies (e.g., Zierler et al., Highway NE, MS F-46, Atlanta, GA 30341. Tel.: +1-770-488-7603. Fax: +1-770-488-3506. E-mail: [email protected] 1988) used length of residence in a community using Received 4 February 2000; accepted 17 May 2000. chlorinated drinking water as the exposure indicator. But Backer et al. Household exposures to drinking water disinfection by-products these approaches cannot assign individual exposure Methods levels. King and Marrett (1996) assessed individual exposures using retrospectively collected data on lifetime Study Design and Population waterconsumption. The difficulties associated with using This study was conducted in Atlanta, Georgia, using a any of these exposure assessment methods include consent form and protocol approved by the Institutional variability in DBP concentrations in a given water Review Board of the Centers for Disease Control and system over time and problems with reconstructing water Prevention (CDC). We obtained informed consent from consumption history. In addition, these methods may not 31 adult volunteers (20 women, 11 men) who agreed to account for all relevant exposure routes. For example, participate in one of the following household activities: household wateruse (e.g., showering, bathing, flushing showering for 10 min with tap water (n=11), immersing toilets, orusing the dishwasher,washing machine, or themselves for10 min in a bathtub filled with tap water faucets) can result in significant indoor air concentra- (n=10), or drinking 1 l of water during a 10-min time tions of volatile organic compounds (VOCs), including period (n=10). Forthe showeringand bathing segments, DBPs and other contaminants (e.g., trichloroethylene) we asked participants to collect a sample of the water (Andelman, 1985; Giardino et al., 1992; Weisel and they used and measure with a thermometer the Chen, 1994; Wallace, 1997; Howard-Reed et al., 1999), temperature of the water they used. The water flow, and subsequent VOC exposure through inhalation will which was the same forall participants, was 7.2 l/min. vary for different individuals within a household (Wilkes Automatic ventilation fans were on in the bathrooms at et al., 1996). In addition, DBPs, such as chloroform, can all times during study activities. be absorbed through the skin (Jo et al., 1990a). Study participants were also requested to provide three Biological specimens can be used to determine the 10 ml blood samples associated with the tap water dose of DBPs resulting from exposure to tap water by exposure: one sample immediately before the exposure; different routes. Both blood and breath analyses should one sample 10 min afterthe exposurehad ended (the reflect the total internal dose from all routes of exposure exposure ended when they got out of the bath, turned off (i.e., inhalation, dermal absorption, and ingestion) and the shower, or finished drinking the water ); and one have been used to examine internal doses of VOCs in sample 30 min (forshowerand tub exposure) or1 h non-occupationally exposed individuals. Forexample, (foringestion) following the end of the exposure.Study Ashley et al. (1994) found detectable levels of benzene, participants did not drink water, shower, or bathe during ethylbenzene, chloroform, and other VOCs in blood the time between drawing of the second and third blood samples from non-occupationally exposed subjects se- samples. Study participants were not requested to lected from the Third National Health and Nutrition complete a questionnaire. Examination Survey. Weisel et al. (1992) and Jo et al. (1990a,b) have utilized breath analysis to examine Data Collection exposure to chloroform in people who showered with chlorinated tap water. In these studies, although the Blood Samples A certified phlebotomist collected blood amount of time between the end of the exposure and samples using Vacutainer1 tubes that had been processed sample collection varied among participants, breath to remove VOC contamination (Cardinali et al., 1995). The samples indicated that exposure to chloroform occurs blood samples were analyzed for THM (bromoform during showering. Lindstrom et al. (1994) found that [CHBr3], bromodichloromethane [CHCl2Br], dibromo- inhalation exposures to benzene (from gasoline-con- chloromethane [CHClBr2], and chloroform [CHCl3]) taminated water) during actual showering were approxi- levels using a headspace purge-and-trap/gas chromato- mately two to five times higher than corresponding graphy/mass spectrometry method with detection limits in bathroom exposures (i.e., benzene concentrations ranged the parts-per-quadrillion range. The mass spectrometry from 758 to 1670 g/m3 in the showerstall and 366 to portion of the method had been modified from the original 498 g/m3 in the bathroom during and immediately after VOC analysis procedure (see Ashley et al., 1992) to the shower). achieve lowerdetection limits. To increase the sensitivity of To more fully examine the association between the analysis forTHMs in blood, we operated the mass exposure to DBPs and cancer, researchers must account spectrometer in selective ion recording (SIR) mode instead forall of the relevant household activities (e.g., of in full scan. Extraction of VOCs from blood, trapping on bathing, showering, drinking water) in estimating an Tenax in a glass-lined steel tube, removal of water, and individual's lifetime dose of DBPs. In this study, we concentration on a liquid nitrogen trap were done with a examined whole blood levels of THMs in individuals Tekmar (Cincinnati, Ohio) 3000 purge-and-trap concen- exposed by bathing orshoweringin tap water,orby trator with an attached ALS 2016 automated sampler. We drinking tap water. separated the analytes with a J&W (Folsom, California) 30

322 Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(4) Household exposures to drinking water disinfection by-products Backer et al. m DB-624 column, with a 1.8 m film thickness mounted determined water concentrations using isotope dilution in a Hewlett-Packard (Avondale, Pennsylvania) Model mass spectrometry methods. 5890 series II gas chromatograph. The effluent from the gas chromatograph was directed through a heated interface into Data Analysis We compared the mean concentrations of the mass spectrometer ion source. Mass spectral analysis THMs in the blood of people exposed to tap water, both was done with a Micromass (Beverly, Massachusetts) before and after exposure, and by the three different routes, Ultima high-resolution mass spectrometer that operated at using Statistical Analysis System software (SAS Institute, 10,000 resolving power (5% valley definition) in the SIR Cary, North Carolina). voltage mode. Masses were calibrated vs. a mixture of perfluorokerosene (low boiling) with a 50/50 mixture of toluene and benzene. The mass spectrometry analysis, Results and discussion which included methyl tert-butyl ether(MTBE),was The whole blood levels of bromodichloromethane, dibro- divided into five groups: MTBE/MTBE-d10, 4.40±6.20 min, masses: 73.065, 74.071, and 82.122 amu; mochloromethane, and chloroform are presented in Figures 13 1±3, respectively. There were measurable levels of THMs CHCl3 / CHCl3, 6.20±7.50 min, masses 82.946, 84.943, 13 in all the waterand whole blood samples that wereanalyzed. 83.949, and 85.946 amu; CHCl2Br/ CHCl2Br, 8.30± 10.10 min, masses 82.946, 84.943, 83.949, and 85.946 The highest median levels of THMs were found in the blood 13 samples from people who took 10-min showers, whereas amu; CHClBr2/ CHClBr2, 11.10±12.10 min, masses 126.895, 128.892, 127.898, and 129.896 amu; and the lowest median levels were found in the blood samples 13 from people who drank 1 l of water in 10 min. For example, CHBr3 / CHBr3, 13.00±14.20 min, 172.843, 174.840, 173.846, and 175.844 amu. Peak areas were determined the median levels of bromodichloromethane in blood using the OPUSquan software of Micromass. We deter- samples from people who took a shower were 3.3 pg/ml mined blood concentrations of bromoform, bromodichlor- (baseline), 19.4 pg/ml (10 min afterexposureended), and omethane, dibromochloromethane, and chloroform by using 10.3 pg/ml (30 min afterexposureended). The median isotope dilution mass spectrometry methods. levels of bromodichloromethane in blood samples from people who took a bath were 2.3 pg/ml (baseline), 17.0 Water Samples To provide a measure of VOC exposure, we pg/ml (10 min afterexposureended), and 9.9 pg/ml (30 collected tap watersamples at the time of exposure.The min afterexposureended). Forthose who drank1 l of watersamples werecollected in borosilicate glass vials water, the median levels of bromodichloromethane in the which contained 125 l of a phosphate buffer/sodium blood samples were 2.6 pg/ml (baseline), 3.8 pg/ml (10 thiosulfate solution. The solution was made using J.T. Baker (Phillipsburg, NJ) HPLC grade water and 1.0 M sodium dihydrogen phosphate, 1.0 M sodium hydrogen 30 phosphate and 0.08 M sodium thiosulfate. This solution T=+10** T=+10** raised the pH of the tap water to approximately 6.5 and l) 25 /m neutralized any free chlorine present in the sample. We g

(p 20 T=+30** examined residual chlorine and pH for all samples. n

tio T=+30**

We analyzed the watersamples forTHM levels using tra 15 n T=0 e solid phase microextraction (SPME) /gas chromatography/ c n T=0 T=+10* mass spectrometry with method detection limits in the o 10 C T=+60

d T=0 parts-per-trillion range. Extraction of the THMs from the o lo 5 watersample was done using a Supelco (Bellefonte, PA) B 0 100 m carboxen/PDMS SPME fiber attached to a Varian Drinking Bathing Showering (Walnut Creek, CA) 8200 automated sampler. THMs were absorbed onto the SPME fiber at room temperature for 10 min and then desorbed into the GC inlet at 2008C for3 min. Figure 1. Bromodichloromethane levels (pg/ml whole blood) in the blood of study participants who bathed or showered in, or drank 1 We separated the analytes with a Supelco 10 m VOCOL liter of, tap water and provided blood samples before the exposure column with a 1.20 m film thickness mounted in a (T=0), 10 min after the exposure ended (T=10), and 30 (for Hewlett-Packard (Avondale, PA) Model 5890 gas chro- bathing or showering) or 60 (for drinking) min after exposure ended. The mean concentration of bromodichloromethane in tap matograph. Mass spectral analysis was done with a water was 6 g/l. For the box plots: the box extents indicate the 25th Hewlett-Packard 5972 mass selective detector. The mass and 75th percentiles of the column, the line inside the box marks the selective detectorwas operated in the selected ion value of the 50th percentile, capped bars indicate the 10th and 90th percentiles, and circles mark the 5th and 95th percentiles. *Sig- monitoring mode with dwell times varying from 50 to 100 nificantly different from baseline, P<0.01; **Significantly different ms perion fora total scan rateof 1.69 cycles/s. We also from baseline, P<0.001.

Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(4) 323 Backer et al. Household exposures to drinking water disinfection by-products

250 decreased significantly, but were still higher than baseline T=+10** T=+10* levels. Bromoform was not found above the detection limit l) 200

/m in any blood orwatersamples in this study.

g T=+30*

(p We found a dramatic difference between the whole blood n 150 T=+30** tio levels resulting from exposure by ingestion and those T=+10* tra

n T=+60* resulting from showering and bathing (a combination of e T=0 c 100 n

o inhalation, dermal contact, and possibly ingestion). The T=0 T=0 C

d blood levels of THMs increased by an average of 16.1 pg/ o 50 lo

B ml in subjects who showered, an average of 14.7 pg/ml in subjects who bathed, and an average of 1.2 pg/ml in Drinking Bathing Showering 0 subjects who drank 1 l of water. Thus, drinking 1 l of water increased blood levels by less than 10% of the increase Figure 2. Chloroform levels (pg/ml whole blood) in the blood of resulting from bathing or showering for 10 min. The study participants who bathed or showered in, or drank 1 liter of, tap increases in blood THMs from showering or bathing were water and provided blood samples before the exposure (T=0), 10 min after the exposure ended (T=10), and 30 (for bathing or significantly greater than the increases from drinking 1 l of showering) or 60 (for drinking) min after exposure ended. The mean water( P<0.001 for bromodichloromethane and dibromo- concentration of chloroform in tap water was 28.0 g/l. For the box chloromethane; P=0.017 forchloroform). plots: the box extents indicate the 25th and 75th percentiles of the column, the line inside the box marks the value of the 50th percentile, In addition to finding the differences in blood THM capped bars indicate the 10th and 90th percentiles, and circles mark levels among the exposure groups, we found that the the 5th and 95th percentiles. *Significantly different from baseline, increases in blood THM levels in people who showered or P<0.01; **Significantly different from baseline, P<0.001. bathed fell roughly into two groups; one group had a lower range of increases in blood levels while the other group was clustered at a higher blood level of bromodichloromethane min afterdrinking all the water), and 2.8 pg/ml (1 h after after showering or bathing (Figure 4). The mean increases drinking the water). We obtained similar relative findings in bromodichloromethane blood levels for the two clusters for dibromochloromethane and chloroform. observed after participants bathed (8.2, standard deviation The highest blood levels of each THM were found in the [SD]=2.97, n=5 forthe lowercluster;and 21.2 pg/ml; samples collected 10 min afterexposureended. By the SD=4.26, n=5 forthe highercluster) orshowered (12.1 second post-exposure time (30 min for showering or pg/ml, SD=1.87, n=6 forthe lowercluster;and 21.0 pg/ bathing; 60 min fordrinkingwater), the blood levels had ml, SD=2.07, n=5 forthe highercluster) weresignificantly different (P<0.001). The same individuals who had higher increases of bromodichloromethane also had higher increases of dibromochloromethane and chloroform after 8 T=+10** T=+10** these exposures. The mean increases for total THMs for the clusters were also statistically significantly different for the l) 6 /m groups that bathed (P=0.05) orshowered ( P<0.01). g T=+30** (p n T=+30** tio 4 T=+10* T=0 tra 30

n T=+60

e T=0 c

n T=0 o 2 25 C l) d o /m lo g

B 20

0 (p Drinking Bathing Showering n tio a

tr 15 n e c n o

C 10 Figure 3. Dibromochloromethane levels (pg/ml whole blood) in the d o blood of study participants who bathed or showered in, or drank 1 lo B liter of, tap water and provided blood samples before the exposure 5 (T=0), 10 min after the exposure ended (T=10), and 30 (for bathing or showering) or 60 (for drinking) min after exposure ended. The mean concentration of dibromochloromethane in tap 0 Shower Bath water was 1.1 g/l. For the box plots: the box extents indicate the 25th and 75th percentiles of the column, the line inside the box marks the value of the 50th percentile, capped bars indicate the 10th and Figure 4. Increase in bromodichloromethane levels (pg/ml whole 90th percentiles, and circles mark the 5th and 95th percentiles. blood; ppt) in the blood of study participants 10 min after they *Significantly different from baseline, P <0.01; **Significantly bathed or showered for 10 min in tap water. The mean concentration different from baseline, P<0.001. of bromodichloromethane in tap water was 6.0 g/l.

324 Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(4) Household exposures to drinking water disinfection by-products Backer et al.

The reason why the data fall into two groups is not clear. Table 1. Means and ranges of trihalomethane concentrations in tap water The clustering could not be explained by sex (we did not used by study volunteers who participated in different household activities. collect information about race or smoking status). It is possible that this clustering is the result of individual THM Mean concentration (g/l)SD variations in the ability to metabolize THMs. For example, ShowerBath Drinking brominated THMs are substrates for glutathione S-transfer- Bromoform

Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(4) 325 Backer et al. Household exposures to drinking water disinfection by-products average temperature of the water used for showering was occupationally exposed U.S. population and in groups with suspected 378C (range 34±438C) and the average temperature of the exposure. Clin Chem 1994: 40 (7): 1401±1401. Cantor K.P., Hoover R., Hartge P., et al. Bladder cancer, drinking water waterused forbathing was 38 8C (range 36±448C). source, and tap water consumption: a case±control study. J Natl The most consistent human health effect associated with Cancer Inst 1987: 79: 1269±1279. exposure to THMs is bladder cancer. The studies done by CantorK.P., Lynch C.F., Hildesheim M.E., Dosemeci M., Lubin J., Cantoret al. (1987, 1998), Zierleret al. (1988), McGeehin Alavanja M., and Craun G. Drinking water source and chlorination et al. (1993), King and Marrett (1996) and Freedman et al. byproducts: I. Risk of bladder cancer. Epidemiology 1998: 9: 21±28. Cardinali F.L., McCraw J.M., Ashley D.L., et al. Treatment of vacutainers (1997) produced strikingly similar results. However, foruse in the analysis of volatile organic compounds in human blood despite the consistency, the reported risk for bladder cancer at the low parts-per-trillion level. J Chromatogr Sci 1995: 33: 557± from lifetime exposure to high levels of THMs reported by 560. these studies was nearly universally less than 2.0. Exposure Freedman D.M., Cantor K.P., Lee N.L., Chen L.-S., Hei H.-H., Ruhl was typically assessed by examining historical water C.E., and Wang S.S. Bladder cancer and drinking water: a population- treatment plant records or by reconstructing a person's based case±control study in Washington County, Maryland (United States). Cancer Causes Control 1997: 8: 738±744. waterconsumption history. In ourstudy, we found that the Giardino N.J., Gumerman E., Esmen N.A., and Andelman J.B. Shower dermal and inhalation exposures associated with showering volatilization exposures in homes using tap water contaminated with and bathing resulted in much higher blood levels of THMs trichloroethylene. J Expos Anal Environ Epidemiol 1992: (suppl 1): than drinking 1 l of water. If blood THM levels are a risk 393±412. Howard-Reed C., Corsi R.L., and Moya J. Mass transfer of volatile organic factorforsubsequent human health effects, such as bladder compounds from drinking water to indoor air: the role of residential cancer, then water consumption may not be the only dishwashers. Env Sci Technol 1999: 33: 2266±2272. important route of exposure. Accounting for an individual's Jo W.K., Weisel C.P., and Lioy P.J. Routes of chloroform exposure and participation in household activities involving tap water, as body burden from showering with chlorinated tap water. Risk Anal well as his tap waterconsumption, may allow a more 1990: 10 (4): 575±580. Jo W.K., Weisel C.P., and Lioy P.J. Chloroform exposure and the health accurate individual exposure assessment. Our study has risk associated with multiple uses of chlorinated tap water. Risk Anal demonstrated that public health messages indicating that an 1990: 10 (4): 581±584. individual can significantly reduce his or her exposure to King W.D., and Marrett L.D. Case±control study of bladder cancer and THMs by not drinking tap water may lead to a false sense of chlorination by-products in treated water (Ontario, Canada). Cancer security regarding the risk for adverse health outcomes Causes Control 1996: 7: 596±604. associated with THM exposure. Landi S., Hanley N.M., Warren S.H., Pegram R.A., and DeMarini D.M. Induction of genetic damage in human lymphocytes and mutations in Salmonella by trihalomethanes: role of red blood cells and GSTT1-1 polymorphism. Mutagenesis 1999: 14 (5): 479±482. Acknowledgments Lindstrom A.B., Highsmith V.R., Buckley T.J., Pate W.J., and Michael L.C. Gasoline-contaminated ground water as a source of residential The authors thank Daniel Huff and Charles Dodson for their benzene exposure: a case study. J Expos Anal Environ Epidemiol 1994: 4 (2): 183±195. assistance in collecting blood samples, Judy Bass and Maria McGeehin M., Reif J., BecherJ., and Mangione E. A case±controlstudy Mirabelli for their assistance in conducting the field study, of bladdercancerand waterdisinfection methods in Colorado. Am J and the volunteers from National Center for Environmental Epidemiol 1993: 138: 492±501. Health fortheirparticipation in this study. U.S. Environmental Protection Agency. National primary drinking water regulations: disinfectants and DBPs: final stage I rule. December (effective date, February 1999). Federal Register vol. 63, no. 241, 1998. References Wallace L.A. Human exposure and body burden for chloroform and other trihalomethanes. Crit Rev Environ Sci Technol 1997: 27 (2): 113±194. Andelman J.B. Inhalation exposure in the home to volatile organic Weisel C.P., and Chen W.J. Exposure to chlorination by-products from hot contaminants of drinking water. Sci Total Environ 1985: 47: 443±460. wateruses. Risk Anal 1994: 14 (1): 101±106. Ashley D.L., and Prah J.D. Time dependence of blood concentrations Weisel C.P., Jo W.K., and Lioy P.J. Utilization of breath analysis for during and after exposure to a mixture of volatile organic compounds. exposure and dose estimates of chloroform. J Expos Anal Environ Arch Environ Health 1997: 52 (1): 26±33. Epidemiol 1992: (suppl 1): 55±69. Ashley D.L., Bonin M.A., Cardinali F.L., McCraw J.M., Holler J.S., Wilkes C.R., Small M.J., Davidson C.I., and Andelman J.B. Modeling the Needham L.L., and Patterson D.G. Determining volatile organic effects of waterusage and co-behavioron inhalation exposuresto compounds in human blood from a large sample population by using contaminants volatilized from household water. J Expos Anal Environ purge and trap gas chromatography/mass spectrometry. Anal Chem Epidemiol 1996: 6 (4): 393±412. 1992: 64: 1021±1029. Zierler S., Feingold L., Danley R.A., and Craun G. Bladder cancer in Ashley D.L., Bonin M.A., Cardinali F.L., McCraw J.M., and Wooten J.V. Massachusetts related to chlorinated and chloraminated drinking water: Blood concentrations of volatile organic compounds in a non- a case±control study. Arch Environ Health 1988: 43: 195±200.

326 Journal of Exposure Analysis and Environmental Epidemiology (2000) 10(4)