Drinking Water Mutagenicity and Urinary Tract Cancers: A Population-based Case-Control Study in Finland

Transcription

1 American Journal of Epidemiology Copyright O 1998 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 146, 7 Printed In USA. Drinking Water Mutagenicity and Urinary Tract Cancers: A Population-based Case-Control Study in Finland Koivusalo, 1 Timo Hakulinen, 1-2 Terttu Vartiainen, 34 Eero Pukkala, 1 Jouni J. K. Jaakkola, 296 and Jouko Tuomisto 3 The detection of mutagenic and carcinogenic chlorination by-products in chlorinated drinking water has raised concern in many countries over the potential hearth effects of long-term exposure to these products. The relation between estimated exposure to historical drinking water mutagenicity and cancer was studied in Finland by using a population-based case-control study comprising bladder cancer cases, kidney cancer cases, and controls. The cases were obtained from the nationwide Finnish Cancer Registry for the years The controls, frequency matched by age and sex, were randomly selected from the national population registry. Information on past drinking water sources and confounding factors was acquired through a questionnaire. Historical exposure to drinking water mutagenicity was estimated by using information on past residence, past water source, and historical data on water quality and treatment. Odds ratios were calculated for an increase of 3,000 net revertants per liter (net rev/liter) in average exposure from 1950 to 1987, adjusting for age, sex, socioeconomic status, and smoking in logistic regression models. A small, statistically significant, exposure-related excess risk was found for kidney cancer for men (odds ratio = 1.49, 95 percent confidence interval (Cl) ) for 3,000 net rev/liter exposure level. For women, the association remained nonsignificant, with a lower odds ratio of 1.08 (95 Cl ). For bladder cancer, the odds ratio for both men and women was 1.22 (95 Cl ) for 3,000 net rev/liter exposure. However, a higher odds ratio of 2.59 (95 Cl ) for 3,000 net rev/liter exposure was observed for male nonsmokers. Am J Epidemiol 1998; 148: bladder neoplasms; case-control studies; chlorine; kidney neoplasms; neoplasms; water supply Chlorination is of crucial importance for the microbiologic safety of drinking water. However, chlorination of water rich in organic material also produces a complex mixture of organochlorine by-products, including carcinogenic and/or mutagenic compounds (1, 2). The formation of chlorination by-products depends on raw water quality and chlorination practices. High levels of mutagenicity have been detected in Finland due to chlorination of surface waters rich in humic Received for publication July 22, 1997, and accepted for publication March 5, Abbreviations: Cl, confidence interval; MX, (3-chtoro-4-dichloromethyf)-5-hydroxy-2(5H)-furanone; net rev/liter, net revertants per liter;, odds ratio. 1 Finnish Cancer Registry, Uisankatu 21 B, Helsinki, Finland. 2 Department of Public Health, University of Helsinki, Helsinki, Finland. 3 Division of Environmental Health, National Public Hearth Institute, Kuopio, Finland. * Department of Environmental Sciences, University of Kuopio, Kuopio, Finland. 5 Department of Epidemiology, The Johns Hopkins University, Baltimore, MD. 8 Department of Environmental Health, Harvard School of Public Health, Boston, MA. Reprint requests to Dr. Koivusalo, Finnish Cancer Registry, Uisankatu 21 B, 000 Helsinki, Finland. material. The large differences in mutagenicity of Finnish drinking waters provide necessary variation in exposure needed to detect small environmental risks. Epidemiologic studies have suggested an association between chlorinated drinking water and cancers of the gastrointestinal and urinary tract, but the evidence of such a causal relation has remained insufficient (3-6). Trihalomethanes were a major concern in earlier studies of cancer using quantitative estimates of exposure to chlorination by-products (7-9). Among the current major concerns are the brominated trihalomethanes and haloacetates (3, 4, 10). The detection of a highly mutagenic compound (3-chloro-4- (dicwcnx)rnethyl>5-hydroxy-2(5h)-ftiranone), termed MX, has drawn attention to the potential carcinogenic impacts of mutagenic compounds in chlorinated drinking waters. MX has been detected widely in Finnish chlorinated drinking waters (11) and also in drinking water in Great Britain (12), Japan (13), and the United States (14). The concentration of MX correlates well with the degree of measured mutagenicities (15). MX is a very potent bacterial mutagen (16) and genotoxic compound in laboratory animals (). It has also been 704

2 Drinking Water Mutagenicity and Cancer 705 shown to be carcinogenic in rats (). Pharmacokinetic and toxicologic studies have indicated that the urinary tract might be the potential target for carcinogenic effects of MX (19, 20). In this case-control study, our aim was to assess the relation between the estimated historical exposure to drinking water mutagenicity and the risk of kidney and bladder cancers by using estimates of historical exposures to drinking water mutagenicity based on actual historical information about water treatment and individual-level information on past residences, past water sources, occupations, and lifestyle. MATERIALS AND METHODS We included in our study all of the 2,433 incident and histologically confirmed bladder and kidney cancer cases (International Classification of Diseases, Seventh Edition, codes 0 and 1) for whom a contact address was acquired or, in the case of those who were deceased, those for whom a contact address of a spouse or close relative (child/mother/father/ sister/brother) could be acquired. The contact addresses were obtained from the Central Population Registry after receiving a permission by the treating hospital. The cancer cases were diagnosed in and reported before December 1993 to the nationwide Finnish Cancer Registry. Randomly selected, living controls were frequency matched by sex and 5-year age group and were obtained from the nationwide Central Population Registry, which covers the total population in Finland. This ensured that the source population from which the controls were randomly selected was the same as that from which the cases were derived. A questionnaire was sent to 4,134 persons, of whom 1,721 were living cases, 699 were close relatives of cases who had died before January 1994, and 1,714 were controls. Informed consent was requested in the questionnaire. Altogether, 2,861 persons (69 percent) responded. However, sufficient information on residential history, smoking, and occupation was ascertained from only 2,812 persons (68 percent) (table 1). The criterion for information on exposure for at least 30 years from 1950 to 1988 further diminished the number of persons included in the study participants to 2,349 (57 percent) (table 1). In the final analysis, there were bladder cancer cases, kidney cancer cases, and controls (table 2). Information on previous water sources for and addresses of the participants was obtained through the questionnaire. Further information on water-pipe connections and past drinking water quality and treatment practices by waterworks was obtained from administrative registers and municipal waterworks. The level of mutagenicity was estimated by an equation giving the level of mutagenicity in net revertants per liter (net rev/liter) on the basis of information on raw water quality (e.g., permanganate consumption, ph, and color) and water treatment practices (21, 22). This equation was based on a previous survey in the study area using actual measured mutagenicities in waterworks (21, 22). The equation used was the following (21): Mutagenicity in TA Salmonella typhimurium In the equation, the mutagenicity is measured in net revertants per liter of water by using standard procedures (23). The TA 5. typhimurium is the Ames mutagenicity tester strain; A and k are constants; and c is a water quality parameter dependent on total organic carbon, chlorine dose, and ammonia concentration. The r refers to water quality parameters in raw water and the t to water-quality parameters after treatment. The parameter c can be calculated on the basis of routinely collected information on water quality and treatment practices, and thus, actual historical values could be used in the estimation of past mutagenicity. Exposure to mutagenicity was estimated for each TABLE 1. Number of cases by site and number of controls, those who responded, those who responded adequately, and those with 30 years of information on exposure, by gender, Finland, Group Al cancer cases and controls Ad who responded Those who responded adequately* Those wtth 30 years of Information on water sourcef Bladder Sufficient Information on residential history, smohng, and occupation, f Level of mutagenicity exposure estimable tor 30 years or more Kidney Controls , ,134 2,881 2,812 2^ Am J Epidemiol Vol. 148, 7, 1998

3 706 Koivusalo et al. TABLE 2. Number of cases and controls according to demographic, socioeconomic, smoking, and mutagenicity categories in the population-based case-control study In Finland, Alive Deceased Category Bladder Kidney Cortrob 0 Gender Age (years) < Socioeconomic status Upper white collar Lower white collar Agricultural Professional blue collar Blue-collar worker Not classifiable* Smoking Never Past 1-19 cigarettes/day 220 cigarettes/day Mutagenicity No mutagenicity net rev/iiterf 1,000-2,499 net rev/liter 2,500 net rev/liter * In the group "not classifiable'' were included housewives (other than farmers), long-term pensioners, and those who did not report their occupational status, t Net rev/liter, net revertantsiiter. person for 5-year periods for Information on past water sources was acquired from the questionnaire and combined with information from the waterworks on water supply connections, raw water quality, and water treatment practices in past years to calculate the individual estimates of historical exposure to drinking water mutagenicity. Mutagenicity was studied both as a continuous variable for the average exposure from 1950 to 1987 (odds ratios calculated in accordance with our previous studies per 3,000 net rev/liter) and as a categorical variable according to approximate tertiles of estimated mutagenicity levels for cases and controls (no exposure and 1-999, 1,000-2,499, and >2,500 net rev/liter). In addition, timedependent variables for exposure were used, such as estimated cumulative exposure and years of exposure to estimated level of 3,000 net rev/hter or more. The latter variable was created to provide an estimate comparable with those used in studies in which exposure has been measured in years within chlorinated surface water supply. Information on confounding factors was gathered through the questionnaire, including data on smoking, lifetime occupations and occupational exposures, current and previous intake of water, intake of coffee and other beverages, medication during the previous 10 Am J Epidemiol Vol. 148, 7, 1998

4 Drinking Water Mutagenicity and Cancer 707 years, weight, height, and past urinary tract illnesses (infections and urolithiasis). Occupational exposures were inquired about, with specific reference to chemical, leather, dyeing, printing, and graphic industries. Smoking status was divided into four categories (nonsmoker, past smoker, and present smoker of 1-19 or 20 or more cigarettes per day). Social class was assessed on the basis of lifetime occupational career and grouped according to the standard social class categorization of Statistics Finland. A specific categorical variable was constructed for urbanization measured by living for more than 5 years in one of the three major cities (Helsinki, Tampere, or Turku). Consumption of beverages was assessed according to the reported amount of daily beverages in the late 1980s, reported mealtime beverage, and consumption of coffee and tea in the 1950s, 1960s, and 1970s. Statistical analysis for both cancer sites was conducted by gender using logistic regression models (24). The continuous variable for exposure explained the variation in the data better than did the categorical variable and was thus used in the majority of the final analyses. To ensure that the estimated average exposure measures quantitative exposure in a comparable way, only individuals with at least 30 years of existing exposure information from 1950 to 1987 were included in the final analysis. Odds ratios were estimated for a 3,000 net rev/liter increase in average exposure from 1950 to 1987, adjusting for age, sex, socioeconomic status, and smoking (24). Population attributable risks were calculated for the continuous variable of estimated exposure to mutagenicity in a way analogous to that for exposures with linear response (25). Model-based relative risks (u), u = exp(fem), where m is the estimated mutagenicity level and b the regression coefficient due to mutagenicity, were calculated for each control, taking into account the distribution of estimated mutagenicities among controls. The population-attributable risk of estimated historical mutagenicity can then be expressed by u, the population mean of these relative risks, using the formula (u \)lu. Since some relevant confounding factors in this study were assumed to be associated with occupation, living in the three main cities (urbanicity), other urinary tract diseases, and smoking, additional stratified analyses were conducted without these variables. Socioeconomic status, age, and smoking were used as categorical variables in the control for confounding. RESULTS In bladder cancer, the odds ratios for both genders remained small and statistically nonsignificant after control for confounding. The odds ratio was 1.22 (95 percent confidence interval (CI) ) for both genders combined for a 3,000 net rev/liter increase in average exposure level for those with at least 30 years of estimable exposure history and was slightly lower for the same increase when all persons were included in the analysis (table 3). The inclusion of variables for beverage use, urbanicity, medication, or urinary tract illnesses did not materially change the estimates. For kidney cancer, the odds ratio for both genders TABLE 3. Odds ratios estimated for 3,000 net rev/liter* increase in average exposure to mutagenicity for all respondents and those with 30 years or more of estimable exposure after adjustment for age, smoking, socioeconomic status, and, in the analysis of both genders, also for sex, Finland, Cancer sites by subgrcxj) Cases/controte * 95 CI* Bladder cancer Any estimable exposure and men 230 years of estimable exposure and men 878/1, / /748 / 0/ / Kidney cancer Any estimable exposure and men 30 years of estimable exposure and men 829/1, / /748 / 3/ / Netrev/Gter,netrevertants/Trter;, odds ratio; CI, confidence interval Am J Epidemiol Vol. 148, 7, 1998

5 708 Koivusalo et al. included was 1.31 (95 percent CI -1.74) for 3,000 net rev/liter increase in average exposure after adjustment for age, smoking, and socioeconomic status for those for whom there was 30 years or more of estimable exposure history and slightly lower when no requirement for estimable exposure was made. The odds ratios by gender were 1.49 (95 percent CI ) for men and 1.08 (95 percent CI ) for women for a 3,000 net rev/liter increase in estimated average exposure and at least 30 years of estimable exposure history (table 3). The inclusion of variables for beverage use, weight, medication, urbanicity, or urinary tract illnesses in the model did not change the estimates to any larger extent for men. However, for women, a statistically significant interaction between urinary tract illnesses and exposure was observed. In the stratified analysis, the odds ratio for women was 1.68 (95 percent CI ) for a 3,000 net rev/liter increase in average exposure after adjustment for age, socioeconomic status, and smoking for those with no urinary tract diseases and 30 years or more of estimable exposure history (table 4). The use of a categorical variable for the estimated historical drinking water mutagenicity tertiles resulted in estimates similar to those of analyses using continuous variables (table 5). When odds ratios were based on lifetime years of exposure, higher risk estimates were observed than in earlier results based on average exposure (table 6). In both cancers, the addition of a variable for any known occupational exposure did not materially change the estimated risks. When those with occupational exposure were omitted from the analysis, slightly higher odds ratios could be seen in both cancers, and odds ratios of 1.34 (95 percent CI -1.79) for bladder cancer and 1.42 (95 percent CI ) for kidney cancer per 3,000 net rev/liter increase in average exposure increase were observed for both genders combined for those with 30 years or more of estimable exposure history. In the analysis by specific groups, a substantially higher and statistically significant odds ratio of 2.59 (95 percent CI ) for a 3,000 net rev/liter increase in average exposure was observed for nonsmoking men with 30 years or more of estimable exposure history (table 4). The observed odds ratios did not change substantially for men in other "less confounded" categories. For women, the observed odds ratios became slightly higher when analyzed by less confounded categories, but remained statistically nonsignificant. The present and past intakes of drinking water were not statistically significant TABLE 4. Odds ratios by specific groups in stratified analyses for 3,000 net rev/liter* increase in average exposure to drinking water mutagenidtyt, Rnland, Cancer site by stratification criteria Bladder cancer Urbanicity No main cities Main cities Smoking Nonsmokers Current and past smokers Urinary tract illnesses No urinary tract illnesses Past urinary tract illnesses Cases/ controls 125/199 55/94 122/224 58/69 85/139 95/154 * CI« Cases/ controls 420/ /131 79/89 473/ / / CI Kidney cancer Urbanicity No main cities Main cities 235/199 82/ / / Smoking Nonsmokers Current and past smokers 236/224 81/ /9 299/ Urinary tract illnesses No urinary tract Hlnesses Past urinary tract idnesses 122/ / /499 85/ ^t.6O Net rev/liter, net revertants/dter;, odds ratio; CI, confidence interval. f AD analyses have been adjusted for age and socioeconomic status and analyses by urbanicity and reported urinary tract illnesses for smoking as well. Am J Epidemiol Vol. 148, 7, 1998

6 Drinking Water Mutagenicity and Cancer 709 TABLE 5. Odds ratios and 95 confidence intervals for all those with at least 30 years of known exposure according to the approximate tertiles* of exposure categories compared with the nonexposedf, Finland, Exposure category Nonexposed Low (1-999 net rev/liter}:) Medium (1,000-2,499 net rev/liter) High ( 2,500 net rev/liter) * (/)= 0 cases/ 293 controls) CI$ Bladder cancer (n = 552 cases/ 621 controls) Cl {n = 3 cases/ 293 controls) Cl Kidney cancer (n = 386 cases/ 621 controls) Cl * The tertiles of exposure have been calculated amongst those for whom the estimated average exposure to mutagenicity was 1 net revertants/iiter (net rev/liter) or more. t Age, soctoeconomic status, and smoking have been adjusted for in the analysis, i, odds ratio; Cl, confidence interval; net rev/liter, net revertants/lrter. TABLE 6. Odds ratios and 95 confidence intervals according to years of drinking substantially mutagenic water (3,000 net rev/liter* or more) for all those with at least 30 years of estimable exposure^ Finland, Duration ot exposure (years) (n= 0 cases/ 293 controls) * 95 Cl* Bladder cancer (n = 552 cases/ 621 controls) 95 Cl (n = 3 cases/ 293 controls) 95 Cl Kidney cancer (n = : 386 cases/ 621 controls) 95 Cl < * net rev/liter, net revertants/liter;, odds ratio; Cl, confidence interval. t Odds ratios have been estimated for a categorical variable and for a continuous variable for 30 years of estimated exposure. Age, soctoeconomic status, and smoking have been adjusted for in the analysis. $ Continuous variable for exposure-years, odds ratio calculated for 30 years of estimated exposure. risk factors and did not have a statistically significant interaction with estimated levels of mutagenicity. The population attributable risks were estimated on the basis of main results and odds ratios of 1.31 (95 percent Cl -1.74) in kidney cancer and 1.22 (95 percent Cl ) in bladder cancer. These odds ratios would indicate respective attributable proportions of 9 percent for kidney cancer and 6 percent for bladder cancer. When the 95 percent confidence intervals are taken into account, the lowest attributable proportion in both cancers would be zero, and the highest would be percent for kidney cancer and 14 percent for bladder cancer. DISCUSSION Our results have shown a small excess risk with exposure-response of kidney and bladder cancers associated with estimated historical exposure to drinking water mutagenicity. The observed excess risks are supported by the results from earlier ecologic and cohort studies in Finland that focused on cancer in 56 towns during earlier time periods (26, 27). Several previous epidemiologic studies suggest a small excess risk in bladder cancer due to exposure to chlorination by-products (5, 8, 9, 28, 29), and the observed associations in bladder cancer are in line with a metaanalysis conducted on chlorination by-products and cancer (5). The association between estimated historical mutagenicity and cancer was strongest and most consistent for kidney cancer in men. A recent Canadian study reported an exposure-response relation between exposure to trihalomethanes and bladder cancer (8). To our knowledge, our study is the first to report an exposureresponse relation between the quantitative level of drinking water chlorination by-products and kidney cancer. Previous studies have used qualitative or lessquantitative methods for exposure assessment, with negative or inconclusive results (5, 30, 31). For bladder cancer, higher risks have been observed in previous studies on chlorination by-products after several years of exposure (28, 29). The higher odds ratios for Am J Epidemiol Vol. 148, 7, 1998

7 710 Koivusalo et al. those with 30 years or more of exposure to 3,000 net rev/liter could indicate that our exposure period was too short to study appropriately the relation with bladder cancer and led to potential underestimation of the actual risk. On the other hand, the completeness and the quality of the exposure information prior to 1950 are not as good as in the subsequent years. The exposures were estimated for the time period because most of the Finnish waterworks began chlorination in the 1940s. Moreover, exposure estimation for the years before World War II would have necessitated inclusion of former Finnish areas that are currently part of Russia. The data were also analyzed by stratified groups to study whether the results would differ when "less" and "more" confounded groups were analyzed in two groups defined by selected expected confounding factors. In the estimation of small excess risks due to common exposures, residual confounding needs to be considered, and in the analysis by different groups, the aim was to study the direction toward which our results could be confounded. The changes in the odds ratio for women after exclusion of those with urinary tract infections could indicate that the lack of association in women may be due to confounding or effect modification. For bladder cancer, the observed odds ratios remained statistically nonsignificant. In earlier studies (8, 28), higher odds ratios were observed for nonsmokers. However, higher odds ratios have also been reported for smokers (29). The target sites of kidney and bladder are supported by the pharmacokinetic and toxicologic studies of the highly mutagenic compound MX, which has recently been proven to be carcinogenic in rats (). However, since chlorination by-products form a complex mixture, the observed increased risk linked to the estimated historical levels of mutagenicity may also be due to other compounds produced by the same mechanism as drinking water mutagenicity and not necessarily directly to the specific mutagenic compounds or the total level of mutagenicity. While the quantity of mutagenicity correlates well with the quantity of MX, the observed concentrations of MX in drinking water have been very small in comparison with some other products, such as trichloroacetates or brominated trihalomethanes, which in recent analyses have been of more concern (32). The level of mutagenicity results predominantly from the water treatment practices, and the use of disinfectant (chlorine, chlorine dioxide, and chloramine) influences the level of mutagenicity. Thus, the correlation between the level of mutagenicity and other chlorination by-products is dependent upon the variation of the water treatment practices used. Chlorination by-products were not measured routinely before the 1970s, and thus, all assessments of past exposure are based on estimation. However, in this study, the estimation of past mutagenicities was based on actual historical measurements. In a previous survey conducted between 1985 and 1987 in 97 waterworks within the study area, the estimated mutagenicities correlated well with the measured mutagenicities. The correlation coefficients were between the measured and the modeled mutagenicities (21). Confounding by other water-related carcinogenic exposures, such as arsene in nonmutagenic well water, is possible but should be minor and, if present, would, in principle, lead toward underestimation of the risk. The use of bottled water or other water sources was inquired about on the questionnaire and taken into account in the analyses as exposure to nonmutagenic water. In some studies, the quantity of daily water use has been employed. However, in this study, we found limited variation in water use, with over 80 percent of the respondents reporting the use of less than 1.5 liters of water per day, calculated on the basis of information on all beverages consumed. Owing to the complete and accurate populationbased cancer registration, the validity and reliability of cancer data are good in Finland (33). Nevertheless, the problem of nonresponse remains and needs to be considered, as observed risk ratios are small. It is prudent to assume that controls in rural areas would reply more often and be able to give a more thorough history of addresses, whereas those of the highest social class would answer less frequently than others and would more often live in urban areas with mutagenic drinking water. However, when percentages according to urban versus rural status were calculated for those who responded and those who did not, no major differences could be observed when these proportions among cases in both cancers and controls were compared by gender. To take into account the bias due to differences in response rates by socioeconomic status, we included a variable for socioeconomic status in the model. However, because of the small magnitude of the observed risks, residual confounding may still be considered to have a role in explaining the observed excess risks. In the analysis, the number of participants in the study was reduced due to nonresponse and requirements for data availability on exposure, and thus, a risk of further bias could have been introduced. It was assumed that there was a possibility of more complete information on those with rural residential status and little exposure in comparison with those who had moved to cities or between urban areas because substantial migration from rural to urban areas has taken Am J Epidemiol Vol. 148, 7, 1998

8 Drinking Water Mutagenicity and Cancer 711 place in Finland since the 1950s. There were no major differences between the proportions of those of the highest socioeconomic status and those involved with agriculture between all those who responded adequately and those with 30 years or more of estimable exposure history. However, when all cases and controls were used, the odds ratios were lower and had broader confidence intervals in comparison with analyses based on only those with 30 years or more of estimable exposure history. Information on smoking, demographic factors, socioeconomic status, occupational exposure, urban status, and history of urinary tract illnesses was used in order to take into account the potential confounding, effect modification, or different baseline risk by groups of persons. Age, gender, smoking, and socioeconomic status were included in the statistical model. Additional analyses were conducted by gender, smoking status, occupational exposure, urbanicity, and urinary tract illness history to study the potential impact of these factors on the odds ratios. Recall bias was addressed in several ways in the study to minimize the impact of different reporting between cases and controls. The respondents were not aware of the specific research hypothesis under study, and exposure assessment was based on a combination of information on residence history and average exposure among those with at least 30 years of exposure. The use of residential water source might have led to misclassification of exposure status. This would be most substantial for those with a private well or groundwater source in a city using predominantly surface water and most probably lead toward underestimation of the risk. Our results shed additional light on health effects of long-term exposure to chlorination by-products. They also confirm findings of earlier epidemiologic studies with an observed exposure-response relation for men between estimated level of historical exposure to drinking water mutagenicity and the risk of urinary tract cancers. However, any public health implications of our results should be interpreted with caution. Although the large exposed populations render even small, long-term excess risks important, concern over small excess risks due to chlorination by-products should never lead to compromise in microbiologic quality of drinking water. ACKNOWLEDGMENTS Financial support was received from the Swedish Cancer Foundation, the Finnish Cancer Institute, the Cancer Society of Finland, and the Duodecim Society. The authors acknowledge Professor Lyly Teppo, Marja- Liisa Weckman, and Esko Voutilainen of the Finnish Cancer Registry and the contact persons in municipal waterworks for devoting their time to the study. REFERENCES 1. Bellar TA, Lichtenberg JJ, Kroner RC. The occurrence of organohalides in chlorinated drinking waters. J Am Water Works Assoc 1974;66: Cheh AM, Skochdopole J, Koski P, et al. Nonvolatile mutagens in drinking water: production by chlorination and destruction by sulfite. Science 1980;207: International Agency for Research on Cancer. Chlorinated drinking water; chlorination by-products; some other halogenated compounds, cobalt and cobalt compounds. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. Vol. 52. Lyon, France: International Agency for Research on Cancer, 1991: Bull RJ, Birnbaum L, Cantor KP, et al. Symposium overview. Water chlorination: essential process or cancer hazard? Fundam Appl Toxicol 1995;28: Morris RD, Audet A-M, Angelillo IF, et al. Chlorination, chlorination by-products and cancer: a meta-analysis. Am J Public Health 1992;82: Craun G. Epidemiology studies of water disinfectants and disinfection by-products. In: Craun G, ed. Safety of water disinfection: balancing chemical and microbiological risks. Washington, DC: ILSI Press, 1993: Lawrence CE, Taylor PR, Trock BJ, et al. Trihalomethanes in drinking water and human colorectal cancer. J Natl Cancer Inst 1984;72: King WD, Marrett LD. Case-control study of bladder cancer and chlorination by-products in treated water (Ontario, Canada). Cancer Causes Control 1996;7: McGeehin MA, Reif JS, Becher JC, et al. Case-control study of bladder cancer and water disinfection methods in Colorado. Am J Epidemiol 1993;138: Dunnick JK, Melnick RL. Assessment of the carcinogenic potential of chlorinated water experimental studies of chlorine, chloramine and trihalomethanes. J Natl Cancer Inst 1993; 85: Vartiainen T, Liimatainen A. High levels of mutagenic activity in chlorinated drinking water in Finland. Mutat Res 1986; 169: Horth H, Fielding M, Gibson T, et al. Technical report PRD 2038-M. Swindon, United Kingdom: Water Research Centre, Suzuki N, Nakanishi J. The determination of strong mutagen, 3-chloro-4-(dichloTomeniyl)-5-hydroxy-2(5H)-furanone in drinking water in Japan. Chemosphere 1990;21: Meier JR, Knohl RB, Coleman WE, et al. Studies on the potent bacterial mutagen, 3-chloro-4-(dichloromethyl)-5- hydroxy-2(5h>furanone: aqueous stability, XAD recovery and analytical determination in drinking water and in chlorinated humic acid solutions. Mutat Res 1987; 9: Kronberg L, Vartiainen T. Ames mutagenicity and concentration of the strong mutagen 3-chloro-4-(dichloromethyl)-5- hydroxy-2(5h)-furanone and its geometric isomer E-2-chloro- 3-{dichloromethyl)-4-oxo-buteinoic acid in chlorine treated tap water. Mutat Res 1988;206: Holmbom B, Voss RH, Mortimer RD, et al. Fractionation, isolation, and characterization of Ames mutagenic compounds in kraft chlorination effluents. Environ Sci Technol 1984;: Jansson K, Maki-Paakkanen J, Vaittinen S-L, et al. Cytogenetic effects of 3-chloro-4-(dichloromethyl)-5-hydroxy- 2(5H)-furanone (MX) in rat peripheral lymphocytes in vitro and in vivo. Mutat Res 1993;299:25-8. Am J Epidemiol Vol. 148, 7, 1998

9 712 Koivusalo et al.. Komulainen H, Kosma V-M, Vaittinen S-L, et al. Cancer by the drinking water mutagen 3-chloro-4-(dichloromethyl)-5- hydroxy-2(5h)-furanone (MX) in the rat. J Natl Cancer Inst 1997;89: Komulainen H, Vaittinen S-L, Vartiainen T, et al. Pharmacokinetics of 3-chloro-4-(dichloromethyl)-5-hydroxy-2(5H> furanone (MX), a drinking water mutagen, after a single dose in the rat Pharmacol Toxicol 1992,70: Haataja L, Vartiainen T, Lampelo S, et al. Binding of the strong mutagen 3-chloro-4-(dichloromethyl)-5-hydroxy- 2(5H)-furanone (MX) to bovine serum albumin. Toxicol Lett 1991;59: Vartiainen T, Liimatainen A, Kauranen P, et al. Relations between drinking water mutagenicity and water quality parameters. Chemosphere 1988;: Koivusalo M, Jaakkola JJK, Vartiainen T. Drinking water mutagenicity in past exposure assessment of the studies on drinking water and cancer application and evaluation in Finland. Environ Res 1994;64: Maron DM, Ames BN. Revised method for the Salmonella mutagenicity test Mutat Res 1983;113: Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol. 1. The analysis of case-control studies. (IARC scientific publication no. 32). Lyon, France: International Agency for Research on Cancer, Lubin JH, Boice JD. Estimating Rn-induced lung cancer in the United States. Health Phys 1989;57: Koivusalo M, Jaakkola JJK, Vartiainen T, et al. Drinking water mutagenicity and gastrointestinal and urinary tract cancers: an ecological study in Finland. Am J Public Health 1994;84: Koivusalo M, Pukkala E, Vartiainen T, et al. Drinking water chlorination and cancer a historical cohort study in Finland. Cancer Causes Control 1997;8: Cantor KP, Hoover R, Hartge P, et al. Bladder cancer, drinking water source, and tap water consumption: a case-control study. J Natl Cancer Inst 1987;79: Freedman M, Cantor KP, Lee NL, et al. Bladder cancer and drinking water a population-based case-control study in Washington County, Maryland (United States). Cancer Causes Control 1997,8: Wilkins JR, Comstock GW. Source of drinking water at home and site-specific cancer incidence in Washington County, Maryland. Am J Epidemiol 1981;114: Doyle TJ, Zheng W, Cerhan JR, et al. The association of drinking water source and chlorination by-products with cancer incidence among postmenopausal women in Iowa: a prospective cohort study. Am J Public Health 1997;87: Bull RJ. Toxicology of disinfectants and disinfection byproducts. In: Craun G, ed. Safety of drinking water disinfection. Balancing chemical and biological risks. Washington, DC: ILSI Press, 1993: Teppo L, Pukkala E, Lehtonen M. Data quality and control of a population-based cancer registry. Experience in Finland. Acta Oncol 1994;33: Am J Epidemiol Vol. 148, 7, 1998