American Journal of Epidemiology Copyright O 1999 by The Johns Hopkins University School of Hygiene and Public Health ADrightsreserved Vol. 150, No. 5 Printed In USA. Mutagenic Drinking Water and Risk of Male Esophageal Cancer: A Population-based Case-Control Study Xuguang Tao,1-2 Huigang Zhu,1 and Genevieve M. Matanoski2 Drinking mutagenic downstream water from the Huangpu River was hypothesized to have increased the risk for male esophageal cancer in Shanghai, China. The authors conducted a population-based case-control study of a total of 71 esophageal cancer deaths and 1,122 controls collected during a 5-year follow-up period, 1984-1988, from four male cohorts born before January 1, 1944, living in four communities consuming water with different mutagenicities in the Shanghai area. The controls represented a 1% random sample of the defined living cohorts selected at the end of each of the 5 years of follow-up. Logistic regression showed an odds ratio of 2.77 (95% confidence interval: 1.52, 5.03) for drinking mutagenic downstream water from the river versus drinking nonmutagenic upstream water after controlling for possible confounders including age, disease history (hepatitis, cirrhosis, schistosomiasis, digestive tract ulcer), hazardous occupational history, pesticide exposure, lifestyle factors (cigarette smoking, tea intake, and alcohol intake), dietary habits (intake of pickled vegetables, maize, peanuts, and cured meat), education, poverty, urban environment, and water chlorination. Am J Epidemiol 1999; 150:443-52. case-control studies; chlorine compounds; esophageal neoplasms; mutagenicity tests; pesticides; risk; smoking; water pollution Drinking mutagenic downstream water from the Huangpu River was hypothesized to have increased the risk for male esophageal cancer in Shanghai, China. Shanghai has one of the busiest harbors and is one of the largest industrial and commercial cities in China. The total metropolitan area is 6,100 km2 with about 12 million people. It is composed of a centralized city area of 12 districts and 10 counties at its outskirts. Until 1987, about 98 percent of the drinking water consumed in the city was taken from the Huangpu River, and 90 percent of it was from the downstream part of the river, where the water was heavily polluted. This river is 88 km in length, with an average runoff of 300 m3 per second. In 1987, all water intakes at the downstream section were moved to Linjiang at the midstream part of the river to improve water quality (1). In Shanghai, chlorination, in conjunction with coagulation, sedimentation, and filtration, has historically been used against bacteria contamination for tap water. Concentrations of halogenated hydrocarbons, such as trihalomethanes and other chlorination by-products, were apparently elevated because of the action of chlorine on the high levels of natural organic substances present in the untreated, downstream water of the river. A study showed that the average chloroform concentration, a useful indicator for the presence of halogenated hydrocarbons, in the chlorinated downstream water was 45.6 jig/liter, over 40 times higher than the value of 1.1 (Xg/liter found in upstream water (2). Prior to the relocation of the water supply in 1987, Shanghai water supplies were also tested for mutagenicity, as both tap water and source water, using the Ames Salmonella typhimurium/mammalian microsome mutagenicity test on strains TA100 and TA98 with and without enzymatic activation. The test was considered positive if the number of reversions per liter of water was two or more times the blank and if the standard error was less than 20 percent of the mean (3). Between 1983 and 1985, chlorinated tap water samples from a downstream water plant, the largest tap water plant in Shanghai, were 100 percent positive. The raw water samples taken from the same river section were 96.9 percent positive. However, both raw and chlorinated tap water samples taken from an upstream water plant were negative (4). The results of micronucleus tests on peripheral erythrocytes of cru- Received for publication June 26, 1998, and accepted for publication January 22, 1999. Abbreviation: Cl, confidence interval. 1 Department of Environmental Health, School of Public Health, Shanghai Medical University, Shanghai, China. 2 Department of Epidemiology, School of Hygiene and Public Hearth, The Johns Hopkins University, Baltimore, MD. Reprint requests to Dr. Xuguang Tao, Department of Epidemiology, The Johns Hopkins School of Hygiene and Public Hearth, 111 Market Place, Suite 850, Baltimore, MD 21202-6709. 443 444 Tao et al. cian carp caught along the river and unscheduled DNA synthesis tests on water samples from different sections of the river also showed that downstream water samples had much stronger mutagenicity than did upstream samples (5, 6). The carcinogenic effects on humans from drinking mutagenic downstream water from the Huangpu River in the early years are unclear. Previous studies showed that drinking the mutagenic downstream water from the Huangpu River had a positive association with the risk for stomach and liver cancers in Shanghai males. The odds ratio was 2.02 (95 percent confidence interval (CI): 1.27, 3.21) for male stomach cancer and, for liver cancer, the odds ratio was 1.85 (95 percent CI: 1.45, 2.36) controlling for possible confounders (7, 8). Polluted water had been reported to be related to the risk for esophageal cancer elsewhere in China and other countries (9-13). Less attention has been paid to the possible association between drinking the polluted water of Huangpu River and esophageal cancer in Shanghai, probably because of the declining incidence of this cancer (14, 15). This trend is different from those reported in many other countries where esophageal cancer rates have changed little or actually increased (16-23). A recent population-based casecontrol study identified cigarette smoking and heavy alcohol intake as major determinants of esophageal cancer risk, accounting for nearly half of all the cancer cases among men in urban Shanghai (24, 25). Other risk factors that might account for the other half of esophageal cases are still unclear. The purpose of this study was to explore the possible association between drinking mutagenic downstream water from the Huangpu River and male esophageal cancer deaths after controlling for possible confounders. MATERIALS AND METHODS Definition of cohorts A stratified random sampling method was used in the selection of study communities. To control for the effects of urbanization-related factors such as chlorinated tap water, air pollution, and population density, the community selection was stratified by urban or rural locations. In Shanghai, all urban areas used chlorinated tap water, and most of the rural areas used raw water before 1980. The sampling goal was to identify the following four communities: • two urban communities • used nonmutagenic, upstream, chlorinated tap water (community A) • used mutagenic, downstream, chlorinated tap water (community B) • two rural communities • used nonmutagenic, upstream, raw water (community C) • used mutagenic, downstream, raw water (community D). The following criteria were used in selecting the communities. 1) The total population base was big enough to have at least 100,000 person-years' observation (20,000 persons for 5 years) of males aged 40 years or over, with the expectation of collecting enough cases to have about 80 percent power to detect an odds ratio of 2 or above. Each of the communities should have about 5,000 males aged 40 years or over at the beginning of the follow-up. 2) The distance between the two communities in each stratum (communities A and B or communities C and D) should be far enough (more than 30 miles or 48 km) to reduce the possibility of subject exchange and to have big differences in the mutagenicity of their drinking water. 3) All four communities should have used their defined water sources for at least 40 years before the start of follow-up, and the two rural communities should have used raw water; those communities that changed their water source to tap water no more than 5 years before the beginning of follow-up were to be included, however, because many rural communities changed thenwater source to tap water in the late 1970s and early 1980s. 4) It was decided a priori that community A should be selected from the urban area that used chlorinated tap water from the water plant that located its water intake at the far end of the Huangpu River upstream and had negative Ames test results for all of its water samples during 1983-1985. It was further decided that community B should be selected from the urban area that used chlorinated tap water from the water plant that located its water Intake at the far end of the river downstream and had 100 percent positive Ames test results for all of its water samples collected during the same period. 5) All communities should be in historically stable residential areas with few newly developed blocks and in nonindustriallzed regions. Based on the above criteria, four groups of communities were listed as candidates. Each of the four study communities was randomly selected from the group of candidate communities that met the above criteria. Figure 1 shows the location of the four residential communities selected. Communities A and B were located in urban areas, and communities C and D were located in rural areas. All males born before January 1, 1944, and living in the four communities on January 1, 1984, comprised the study cohorts and were followed from January 1,1984, to December 31,1988. The cohort definition was based on data collected from the Residence Registration and Administration agencies of Am J Epidemiol Vol. 150, No. 5, 1999 Mutagenic Water and Esophageal Cancer 445 5 years of follow-up. Table 1 shows detailed information on cohort size, cases, and controls collected during the 5 years of follow-up on each cohort. Controls were not required to match cases with any characteristics. Jlangsu Provinc* Data collection Yangtze Rlw / t Suzhou Rlwr Shanghai Huangpu Rlvw East China Saa Zhoflang Province FIGURE 1. Locations of the four residential communities selected for the case-control study of mutagenic drinking water and male esophageal cancer, Shanghai, 1984-1988. A, urban community that used nonmutagenic, upstream, chlorinated tap water; B, urban community that used mutagenic, downstream, chlorinated tap water C, rural community that used nonmutagenic, upstream, raw water; D, rural community that used mutagenic, downstream, raw water. Solid black line, area boundary. the Public Security Bureau. By law, residents of all ages are registered at the local Residence Registration and Administration agencies at birth or upon entering a community as a resident The registration of a resident in a community is canceled immediately upon that resident's moving to a new community or at death. Demographic data of the populations in the communities are available from the local agencies at the end of each year. The mutagenicity data of the drinking water were based on the results of Ames tests on water samples collected from 1983 to 1985. The percentages of positive tests for the water sample from communities A, B, C, and D during 1983-1985 were 0, 100, 0, and 96.9 percent (4). Cases and controls During the 5-year follow-up period of 1984-1988, all 71 esophageal cancer deaths in the study cohort were identified from death certificates collected from the local Department of Vital Statistics, with a total of 112,176 observed person-years from four male cohorts born before 1944. To get a representative sample of the cohort population, we selected a 1 percent random sample of the total living cohort (n = 1,122) excluding cancer cases, stratified by community, at the end of each of the Am J Epidemiol Vol. 150, No. 5, 1999 Water supply system data were obtained from the local Environmental Protection Agency and the Shanghai Water Supply Company. The type of drinking water for all subjects was determined according to their residential addresses. Water sources for each address were verified by the Shanghai Water Supply Company. Demographic data, necessary for sampling controls, were collected from the Vital Statistics Department of the Anti-epidemic Station and from the Residence Registration and Administration agencies of the Public Security Bureau. Personal exposure data were collected by questionnaire at home visits. Data for the 71 cases were collected and verified by interviewing the immediate adult family member (spouse, parent, son, daughter, brother, or sister) or other relative who had lived with the case. For the 1,122 controls, individual exposure data were collected directly from the subject or from the immediate adult relative who had lived with the subject, if the subject was not available at the time of the home visit. The interview included questions on birth date, duration of residence in the current community, medical history, occupational history, lifestyle factors, dietary habits, education, and the food expense budget. Data analysis Crude incidences of male esophageal cancer death in the four studied cohorts by type of drinking water were calculated from the number of cases and the observed person-years among the defined male cohorts. Incidence ratios and 95 percent confidence intervals were calculated for downstream cohorts versus upstream cohorts by tap water or raw water (26). Stratified by urban tap water or rural raw water, a Mantel-Haenszel incidence ratio (27) for drinking mutagenic downstream water versus nonmutagenic upstream water from the Huangpu River was also calculated. An unconditional logistic regression was used to estimate the odds ratio and confidence interval for drinking mutagenic water against male esophageal cancer, controlling for possible confounders including age, medical history (hepatitis, cirrhosis, schistosomiasis, and digestive tract ulcer), occupational history (with emphasis on selected hazardous occupations), pesticide exposure, lifestyle factors (cigarette smoking, tea intake, and alcohol intake), dietary habits (intake of pickled vegetables, maize, peanuts, and 446 Tao et al. TABLE 1. Male esophageal cancer cases and controls collected during 5 years of follow-up In four cohorts born before January 1,1944, with different drinking water sources, Shanghai, 1984-1988 Year Two urban cohorts Cohort A (upstream tap water) Cohort size No. of controls No. of cases Cohort B (downstream tap water) Cohort size No. of controls No. of cases Two rural cohorts Cohort C (upstream raw water) Cohort size No. of controls No. of cases Cohort D (downstream raw water) Cohort size No. of controls No. of cases Total Cohort size No. of controls No. of cases 1984 1985 1986 1987 1988 Total 6,413 64 4 6.386 64 2 6,291 63 6,250 63 2 6,167 62 4 31,507 316 16 6,531 65 5 6,491 65 6 6,451 65 4 6,406 64 6 6,349 64 3 32,228 323 24 5,901 59 0 5,842 58 4 5,811 58 3 5,744 57 1 5,621 56 4 28,919 288 12 4,178 42 5 4,038 40 5 3,918 39 6 3,724 37 2 3,664 37 1 19,522 195 19 23,023 230 14 22,757 227 17 22,471 225 17 22,124 221 11 21,801 219 12 112,176 1,122 71 cured meat), education, and monthly food expenses. Subjects with hazardous occupations were defined as those ever occupationally exposed to radiation (alpha, beta, gamma, X-ray), arsenic and its compounds, asbestos, ethylene chloride, chromium and its compounds, cadmium and its compounds, chloromethyl ether, formaldehyde, nickel and its compounds, coal tar, asphalt, and acrylonitrile. Because of the small number in each occupational subset, a combined indicator, ever exposed to any of the above hazardous occupations, was used in the analysis. Likelihood ratio tests and a backward method were used to finalize the regression model (28). The overall goal was to obtain the best fitting model while minimizing the number of parameters. As the first step, all independent variables were included in the initial model. The next step was to fit a reduced model containing only those variables found to be significant (p < 0.05) and to compare it with the full model containing all the variables. The likelihood ratio test comparing these two models was obtained using a G statistic. G was defined as the difference of a -2 log likelihood between the two models. RESULTS The distribution of duration of residence in the selected communities for cases and controls is shown in 4 table 2. All subjects were at least 40 years old at the time of the study by cohort definition; 91.5 percent of the cases and 93.5 percent of the controls had been living in their current communities for 20 years or more. Cases tended to have longer times of residence than controls, since cases on average were older than controls, even though controls on average had 6 months' more follow-up time than cases, which was the result of selecting controls at the end of each year of follow-up. The crude incidences of male esophageal cancer death in four cohorts for the 5 years of follow-up are TABLE 2. Years of living In the current community for male esophageal cancer cases and controls, selected from four cohorts born before January 1,1944, Shanghai, 1984-1988 Years of Uvtng Cases Controls No. % No. £70 0 6 10 7 15 7 13 13 0.0 8.5 14.1 9.9 21.1 9.9 18.3 18.3 14 59 117 195 340 201 121 75 1.2 5.3 10.4 17.4 30.3 17.9 10.8 6.7 Total 71 100.0 1,122 100.0 community <10 10-19 20-29 30-39 40-49 50-59 60-69 % Am J Epidemiol Vol. 150, No. 5, 1999 Mutagenic Water and Esophageal Cancer shown in table 3. The crude incidences for the two downstream cohorts are higher than those for the two upstream cohorts. Crude incidence ratios for drinking downstream water versus upstream water are 1.47 (95 percent CI: 0.78, 2.76) for rural chlorinated tap water drinking cohorts and 2.35 (95 percent CI: 1.14, 4.83) for rural raw water drinking cohorts, without controlling for confounding factors. Stratified by urban tap water or rural raw water, the Mantel-Haenszel incidence ratio for drinking mutagenic downstream water versus nonmutagenic upstream water from the Huangpu River is 1.80 (95 percent CI: 1.12, 2.88). Because the distributions of some characteristics of the four cohorts were found to be dramatically different, the above crude incidence ratios may not represent the real risk. The distributions of possible risk or confounding factors in the four cohorts are shown in table 4. Age distributions are well balanced in the four cohorts. The distributions of hepatitis, liver cirrhosis, and digestive duct ulcer histories show no significant differences either. However, the two cohorts using tap water in urban areas have slightly higher rates of heavy smokers (20 or more cigarettes a day) and higher rates of heavy alcohol drinkers (one and more times a week) than the two cohorts drinking raw water in rural areas. The two upstream cohorts had significantly higher schistosomiasis prevalence rates than did the two downstream cohorts, because the area upstream of the river was a schistosomiasis epidemic area. The two upstream cohorts also had higher rates of smokers and tea drinkers but a lower rate of alcohol users. In addition, the two rural cohorts had higher rates of pickled vegetable intake, cured meat intake, pesticide exposure, poverty, and poor education. Cohort C, the upstream rural cohort, was very special, having the highest rates of schistosomiasis prevalence, pesticide exposure, and poor education and the lowest rate of hazardous occupation. The upstream cohort, using chlorinated tap water from upstream, had significantly higher intake rates of maize products and peanut products. Some of these factors were possible confounders, since they were associated with exposures in this study and are reported to have associations with esophageal cancer in previous studies. However, it is important to point out that all of these rates are group-based ecologic data; to connect them to incidence data directly may lead to ecologic fallacy. Individual-based multivariate analysis is needed to adjust for these potential confounders. The odds ratios and confidence intervals of the independent variables in the initial and final logistic regression models are shown in table 2. In the first model that included all the independent variables, no significant associations were found between the risk of esophageal cancer death and the history of hepatitis, cirrhosis, schistosomiasis, and digestive tract ulcer; intakes of alcohol, tea, peanut products, and cured meat; and a monthly food expense less than the median of controls (p > 0.05). After removing these 10 nonsignificant variables, we established the final model using procedures described in Materials and Methods. The difference of a - 2 log likelihood between the initial model and the final model was 12.378, which was much smaller than 18.307, the chi-square value at a = 0.05 and df = 10. This means that, after exclusion of those nonsignificant variables, the final model fit the data as well as the initial model. The result shows that, adjusted for the other independent variables in the model, the odds ratio for drinking mutagenic downstream river water compared with drinking nonmutagenic upstream river TABLE 3. Incidences and Incidence ratios of male esophageal cancer In four cohorts born before January 1,1944, wtth different mutageniclty of drinking water sources, Shanghai, 1984-1988 Cohort Two urban cohorts Cohort A (upstream tap water) Cohort B (downstream tap water) Two rural cohorts Cohort C (upstream raw water) Cohort D (downstream raw water) Total Drinking water Ames test positive, 19831985 0 100 0 96.9 Personyears (no.) Incidence (1/100,000 personyears) Incidence ratio 24 31,507 32,228 50.78 74.47 1.00 1.47* (0.78, 2.76)t 12 19 28,919 19,522 41.50 97.33 1.00 2.35* (1.14,4.83) 112,176 63.29 1.80* (1.12,2.88) Cases (no.) 16 * Crude incidence ratio. t Numbers in parentheses, 95% confidence interval. % Mantel-Haenszel incidence ratio stratified by water type, that is, tap water or raw water. Am J Epidemiol Vol. 150, No. 5, 1999 447 448 Tao et al. TABLE 4. Characteristics among population-based controls from four male cohorts bom before January 1,1944 (Shanghai, 1984-1988) Rural raw water Urban tap water Characteristics Cohort A (upstream) CohortB (downstream) CohortC (upstream) CohortD (downstream) Total Ch t-square p value Numerical distributior l(no.) Person-years No. of controls 31,507 316 32,228 323 28,919 288 19,522 195 112,176 1,122 Percentage distribution (%) Age (years) 40M9 50-59 60-69 270 History of disease Hepatitis Liver cirrhosis Schistosomiasis Digestive tract ulcer Smoking (cigarette(s)/day) 1-4 5-19 220 Tea intake 21 cup/day* Alcohol intake 21 time/week Pickled vegetables intake 23 months/year Maize products intake frequently Peanut products intake <1 time/week 21 time/week Cured meat intake 23 months/year Hazardous occupationst Ever occupatjonally exposed to pesticide Money for food/month < control median Education < primary school graduate 37.4 28.8 19.9 13.9 32.8 26.0 21.4 19.8 34.7 29.9 19.1 16.3 36.9 28.7 19.0 15.4 35.2 28.3 20.0 16.5 0.73 7.0 0.9 16.5 11.1 7.4 0.6 0.0 13.3 4.5 1.4 25.0 10.8 6.2 0.5 2.6 6.2 6.3 0.9 11.5 10.8 0.48 0.70 0.00 0.09 15.8 32.7 21.1 65.5 22.8 1.9 83.5 10.2 30.1 26.6 45.2 34.1 4.6 32.3 22.9 43.4 18.4 52.1 21.9 47.2 24.7 11.8 39.0 25.1 27.7 30.3 48.2 27.1 15.3 35.8 22.7 49.6 27.1 22.4 43.9 0.00 0.00 0.00 0.00 0.00 89.6 4.7 3.8 8.2 3.5 45.9 44.3 67.5 9.3 2.8 10.8 3.1 42.4 44.8 75.3 2.8 8.0 0.7 69.8 56.6 88.2 50.8 2.6 5.1 8.2 12.3 59.5 49.7 72.8 5.2 4.8 7.0 21.9 50.0 55.8 0.00 0.02 0.00 0.00 0.00 0.00 • One cup is approximately 240 ml. t Including occupations exposed to radiation (alpha, beta, gamma, X-ray), arsenic and its compounds, asbestos, ethylene chloride, chromium and its compounds, cadmium and its compounds, chloromethyl ether, formaldehyde, nickel and its compounds, coal tar, asphalt, and acrylonitrile. water was 2.77 (95 percent CI: 1.52, 5.03). The odds ratios for other risk factors or confounders having a significant positive association (p < 0.05) with esophageal cancer death were 2.63 (95 percent CI: 1.20, 5.78) for drinking chlorinated tap water in urban areas, 4.50 (95 percent CI: 2.09, 9.69) for age 50-59 years, 5.91 (95 percent CI: 2.61, 13.36) forage 70 years and older, 2.53 (95 percent CI: 1.16, 5.50) for smoking 5-19 cigarettes per day, 3.25 (95 percent CI: 1.47, 7.18) for smoking 20 cigarettes or more per day, 3.19 (95 percent CI: 1.54, 6.59) for pickled vegetable intake 3 months or more a year, 2.02 (95 percent CI: 1.15, 3.56) for maize products intake 3 months or more a year, 2.47 (95 percent CI: 1.01, 6.01) for hazardous occupations, 3.64 (95 percent CI: 1.75, 7.57) for pesticide exposure, and 5.19 (95 percent CI: 2.17, 12.38) for education lower than primary school graduation. DISCUSSION Esophageal cancer remains a deadly disease with a poor prognosis in China. The worldwide variation in incidence suggests that environmental exposures contribute to the development of esophageal cancer. This study found that males from communities drinking heavily polluted, highly mutagenic, downstream water from the Huangpu River had higher incidences of esophageal cancer than did males from communities drinking relatively clean, nonmutagenic, upstream water. This difference did not disappear after controlling for possible confounding factors such as age, disease history, occupational history, lifestyle factors, dietary habits, education, water chlorination, and urban location. This suggests that drinking mutagenic water appears to be a risk factor for the disease. This study is consistent with many previous studies, which Am J Epidemiol Vol. 150, No. 5, 1999 Mutagenic Water and Esophageal Cancer showed that possible risk factors of esophageal cancer are age, smoking, alcohol, pickled vegetables, corn and wheat products, pesticides, and drinking polluted water (9-12, 24, 25, 29-37). However, the risk from alcohol intake was significant in this study (odds ratio = 1.54, 95 percent CI: 0.86, 2.76). The frequency indicator used for recording alcohol intake in the investigation might not be as sensitive as a quality indicator. This may partly explain why alcohol intake was elevated but not significant in the study. However, the subjects who used alcohol more frequently would most likely have consumed more alcohol. If poor data on alcohol led to residual confounding, it probably was 449 not very great. Another issue related to alcohol intake was that the two downstream cohorts had higher alcohol intake rates than did the two upstream cohorts, which suggested the possibility of correlation between alcohol and water mutagenicity. Including or excluding the alcohol intake and some other variables in the model (table 5), however, did not change the odds ratio of drinking mutagenic water very much. The test result of -2 log likelihood showed that excluding those variables did not cause significant change of the goodnessof-fit of the model either. Thus, although the risk estimate for alcohol intake may not be accurate, for controlling purposes, it may not bias the water muta- TABLE 5. Odds ratios for independent variables In the first and the final unconditional multlvarlate logistic regression models based on 71 male esophageal cancer cases and 1,122 controls, Shanghai, 1984-1988 Variables Drinking water mutagenicity, mutagenic vs. nonmutagenic Drinking water type, tap water vs. raw water Age (years) 50-59 vs. 40-49 60-69 vs. 45-49 £70 vs. 40-49 History of disease Hepatitis (yes vs. no) LJver cirrhosis (yes vs. no) Schlstosomiasis (yes vs. no) Digestive tract ulcer (yes vs. no) Smoking (clgarette(s)/day) 1-4 vs. <1 5-19 vs. <1 £20 vs. <1 Tea intake (cups/day), £1 vs. <1$ Alcohol intake (times/week), £1 vs. <1 Pickled vegetables intake (months/year), £3 vs. <3 Maize products intake (months/year), £3 vs. <3 Peanut products intake <1 time/week vs. no £1 time/week vs. no Cured meat intake (months/year), £3 vs. <3 Hazardous occupation,§ ever vs. never Exposure to pesticide, ever vs. never Monthly food expenses < control median, yes vs. no Education < primary school graduate, yes vs. no Case/control nos. in two groups* Odds ratio Initial model - 2 ln(Bke)ihood) = 413.713 Final model - 2 In(likelihood) = 426.091 43/518 vs. 28/604 2.70 (1.38, 5.28)t 2.77(1.52,5.03) 40/639 vs. 31/483 2.64(1.17,5.99) 2.63(1.20,5.78) 11/317 vs. 3/396 32/224 vs. 3/396 25/185 vs. 3/396 0.91 (0.42, 1.96) 4.99 (2.29, 10.88) 6.77(2.95, 15.51) 0.93 (0.44, 1.98) 4.50 (2.09, 9.69) 5.91 (2.61, 13.36) 5/71 1/10 7/129 7/121 vs. 66/1,051 vs. 70/1,112 vs. 64/993 vs. 64/1,001 2.46 (0.85, 7.07) 2.77(0.29,26.13) 0.98 (0.37, 2.56) 0.80(0.32, 1.97) vs. vs. vs. vs. vs. 11/294 11/294 11/294 35/565 42/818 1.55(0.51,4.68) 2.51 (1.11,5.68) 3.20(1.35,7.59) 0.70(0.39, 1.25) 1.54(0.86,2.76) 1.42 (0.48, 4.20) 2.53(1.16,5.50) 3.25(1.47,7.18) 27/251 vs. 44/871 3.11 (1.46,6.63) 3.19(1.54,6.59) 31/492 vs. 40/630 1.81 (1.00,3.28) 2.02(1.15,3.56) 7/58 vs. 10/247 54/817 vs. 10/247 1.92(0.88,4.18) 1.56 (0.49, 4.94) 6/54 vs. 65/1,068 8/79 vs. 63/1,043 22/246 vs. 49/876 2.63 (0.94, 7.32) 2.64(1.07, 6.55) 3.99(1.89, 8.43) 43/561 vs. 28/561 1.29(0.74,2.24) 64/626 vs. 7/496 5.51 (2.28, 13.31) 6/172 29/401 25/255 36/557 29/304 2.47(1.01,6.01) 3.64(1.75,7.57) 5.19(2.17, 12.38) * This column gives the distribution of case and control numbers in two comparison groups. However, the odds ratios and confidence intervals in the table are the results of multivariate logistic regression and thus do not match the crude odds ratio and confidence interval calculated based on the numbers from this column. t Numbers in parentheses, 95% confidence interval. t One cup is approximately 240 ml. § Including occupations exposed to radiation (alpha, beta, gamma, X-ray), arsenic and its compounds, asbestos, ethylene chloride, chromium and its compounds, cadmium and its compounds, chloromethyl ether, formaldehyde, nickel and its compounds, coal tar, asphalt, and acrylonitrile. Am J Epidemiol Vol. 150, No. 5, 1999 450 Tao et al. genicity analysis result. Another interesting thing in this study is that drinking chlorinated tap water in urban areas was found to be significantly associated with the risk for esophageal cancer in multivariate analysis (odds ratio = 2.63, 95 percent CI: 1.20, 5.78). This finding was inconsistent with the univariate incidence comparison outcomes, in which the difference between urban and rural cohorts was very small. One of the important reasons might be the skewed distribution of other confounders. For instance, the two rural cohorts had much higher rates of pickled vegetable intake and pesticide exposure than did the two urban cohorts. These two factors are significant risk factors in the study (table 5). So the existing higher risk attributed to urban chlorinated water than to rural raw water found in multivariate analysis was masked by the high rates of pickled vegetable intake and pesticide exposure that occurred in rural cohorts when doing a univariate comparison. It should be pointed out that drinking chlorinated tap water could not be separated from urban residence, since all chlorinated tap water users in the study lived in urban communities. The risk found among chlorinated tap water drinkers is attributed to both chlorinated tap water and other urban factors, such as population density and air pollution. The possible association of chlorinated drinking water with cancer has been widely investigated. Although one study showed an odds ratio of 2.12 (95 percent CI: 1.10, 4.08) associated with drinking chlorinated tap water and the risk of esophageal cancer, other studies did not show any significant risk, with odds ratios ranging from 0.97 to 1.76 (13, 38-42). In this study, a new control sampling method, stepwise population-based proportional sampling, was used to collect controls. Usually there are three control sampling strategies in nested case-control studies (43, 44). The first strategy is that, at the end of follow-up, controls are selected from subjects who do not develop the disease of interest (cumulative incidence sample) or are alive and disease free (cumulative survival sample). The problem with this strategy is that a biased estimation of risk could result from competing causes of death and withdrawal, especially when the followup period is long and the mortality for other diseases is high. The second strategy is that, at the beginning of follow-up, controls are selected from subjects who are alive and disease free (case base sample). The problem with this strategy is that biased underestimation could result from the failure to exclude subjects who later become cases or die from other diseases. The third strategy is that, during the follow-up period, controls are selected from subjects who are alive and disease free at the time the case is diagnosed (density sampling). This method is considered to yield an unbiased estimate of risk. It is easy to conduct with matched controls but difficult for population-based proportional sampling. The control sampling strategy used in the present study was that controls were selected in a stepwise fashion from persons who were alive and cancer free at the end of each subperiod (1 year) during the total 5-year follow-up period. There were three advantages to this method. 1) It could reduce the influence of competing causes of death and withdrawal, since in the new method the subperiod is short although the entire follow-up period is long. 2) It resembles density sampling, since cases and controls are collected relatively concurrently compared with sampling controls at the beginning or the end of the 5-year follow-up. 3) It is easy to draw population-based proportional control samples. Some limitations to this study should be considered. This study was done only on male cohorts aged 40 years or more from only four communities, so the result may not be truly representative of the risk for the 12 million persons in Shanghai. However, the study can help to test the hypothesis that drinking mutagenic downstream water from the Huangpu River could have some association with male esophageal cancer. If the hypothesis is correct, the information would be important to everyone who has consumed that mutagenic water. Bias might exist when using deceased persons as cases and living populations as controls. Information about cases came from an adult firstdegree family member or relative who had lived with the case, while information about controls was primarily self-reported. However, since some factors, such as dietary intake, were usually similar among family members, and since some personal behaviors such as smoking were easily remembered by family members, the response bias to these questions should be limited. The water supply information was not collected through the interview but from the water supply company, thus eliminating recall or family member bias. A limiting issue is that drinking water sources were determined according to residential addresses. Misclassification might occur if a subject worked far from his residential area and had a different quality of drinking water there. Additionally, 8.5 percent of the cases and 6.5 percent of the controls had lived less than 20 years in their current community or in neighboring communities with similar water quality. This means that their drinking water would be a mixture of mutagenic and nonmutagenic water but that it was classified in our study as either mutagenic or nonmutagenic water. However, this type of misclassification could only underestimate the risk rather than contribute to the elevated risk found in the study. Drinking bottled water or using home charcoal water cleaners has been Am J Epidemiol Vol. 150, No. 5, 1999 Mutagenic Water and Esophageal Cancer popular since the mid 1980s in Shanghai. However, considering the latency period of esophageal cancer, the change should not affect the results of this study. The variability of water mutagenicity within each cohort was very small for the two urban tap water cohorts, since subjects within each cohort had as their water source the same water supply plant. More variability might exist among subjects in the two rural cohorts. However, the variability within a rural cohort should be much less than the difference between the two rural cohorts, since they are located at two extreme ends of the river. In spite of these limitations, this study has identified a significant association between drinking mutagenic downstream water from the Huangpu River and the risk for male esophageal cancer after controlling for possible confounders. This finding, although not conclusive, may lead to a better understanding of the etiology of male esophageal cancer in that area and the association of drinking water mutagenicity with esophageal cancer. ACKNOWLEDGMENTS This work was supported by the Shanghai Environmental Protection Agency, Shanghai, China, and was finished during the tenure of a Research Training Fellowship awarded to Xuguang Tao (IARC/R.2170, 1993-1994) by the International Agency of Research on Cancer, Lyon, France. The authors thank Dr. S. Z. Yu and Dr. C. J. Hong for their help in research design; Drs. Q. Y. Zhao, H. Jiang, J. R. Wang, J. P. Bao, F. Y. Wang, Y. Lu, B. Q. Zhang, and J. Cheng for help in data collection; G. D. Wu, X. F. You, W. L. Zhai, C. Li, and Z. Q. Sun for their additional help in data collection; and L. Schwartz, D. Lantry, and L. Swartz for help in manuscript preparation. REFERENCES 1. Song RY, Zhang YJ. The status and prospect of water supply in Shanghai. Rev J Int Water Supply Assoc 1989;8:4-63. 2. Shanghai Environmental Protection Institute, Shanghai Water Supply Company, Shanghai Medical University. General report on the investigation of the status of micro oiganic contaminants in the water protection area at the upstream of Huangpu River and study on pollution control planning to the Shanghai Environmental Protection Agency. (In Chinese). Shanghai: Shanghai Medical University Press, 1991. 3. Maron DM, Ames BN. Revised methods for the Salmonella mutagenicity test. Mutat Res 1983;113:173—215. 4. Zhu HG, Jiang SH, Yang MD. Study on the mutagenicity of raw and tap water of Huangpu River in Shanghai. (In Chinese). Shanghai: Shanghai Medical University Press, 1985. 5. Zhu HG, Xu SQ, Yang MD. Report of the study on mutagenicity of water from Linjiang water intake: evaluation of the first project for upstream water channeling. (In Chinese). Am J Epidemiol Vol. 150, No. 5, 1999 451 Shanghai: Shanghai Medical University Press, 1989. 6. Xu SQ, Wei X, Zhu HG. A new in situ monitoring system for water quality: a study on nuclear anomalies in the peripheral blood erythrocytes of the crucian carp. Chin Environ Sci 1990; 1:84-91. 7. Tao X, Zhu HG, Yu SZ, et al. Pilot study on the relationship between male stomach and liver cancer death and mutagenicity of drinking water in the Huangpu River area. Public Health Rev 1991/1992;19:219-27. 8. Tao X, Zhu HG, Yu SZ, et al. Effects of drinking water from the downstream of the Huangpu River on the risk of male stomach and liver cancer death. Public Health Rev 1991/1992; 19:229-36. 9. Hou J. Multiple stepwise regression analysis of etiological factors of esophageal cancer in Cixian County. (In Chinese). Chung Hua Chung Liu Tsa Chih 1989; 11:25-7. 10. Yu Y, Taylor PR, Li JY. Retrospective cohort study of risk-factors for esophageal cancer in Linxian, People's Republic of China. Cancer Causes Control 1993;4:195-202. 11. Lu JB, Yang WX, Dong WZ, et al. A prospective study of esophageal cytological atypia in Linxian County. Int J Cancer 1988;41:805-8. 12. Amer MH, El-Yazigi A, Hannan MA, et al. Water contamination and esophageal cancer at Gassim Region, Saudi Arabia. Gastroenterology 1990;98:1141-7. 13. Alvanja M, Goldstein I, Susser M. A case-control study of gastrointestinal and urinary tract cancer mortality and drinking water chlorination. In: Jolley RL, Gorchev H, Hamilton DH Jr, eds. Water chlorination: environmental impact and health effects. 2nd ed. Ann Arbor, MI: Ann Arbor Science Publishers, 1978:395-409. 14. Jin F, Devesa SS, Zheng W, et al. Cancer incidence trends in urban Shanghai, 1972-1989. Int J Cancer 1993;53:764-70. 15. Zheng W, Jin F, Devesa SS, et al. Declining incidence is greater for esophageal than gastric cancer in Shanghai, People's Republic of China. Br J Cancer 1993;68:978-82. 16. Menck HR, Garfinkel L, Dodd GD. Preliminary report of the National Cancer Data Base. CA Cancer J Clin 1991;41:7-18. 17. Kurihara M, Aoki K, Hisamichi S. Cancer mortality statistics in the world, 1950-1985. Nagoya, Japan: University of Nagoya Press, 1989. 18. Blot WJ, Devesa SS, Kneller RW, et al. Rising incidence of adenocarcinoma of the esophagus and gastric cardia. JAMA 1991;265:1287-9. 19. Moyana TN, Janoski M. Recent trends in the epidemiology of esophageal cancer. Ann Clin Lab Sci 1996;26:480-6. 20. Reed PI, Johnston BJ. The changing incidence of oesophageal cancer. Endoscopy 1993;25(suppl):606-8. 21. Armstrong RW, Borman B. Trends in incidence rates of adenocarcinoma of the oesophagus and gastric cardia in New Zealand 1978-1992. Int J Epidemiol 1996;25:941-7. 22. Thomas RJS, Lade S, Giles GG, et al. Incidence trends in oesophageal and proximal gastric carcinoma in Victoria. Aust N Z J S u r g 1996;66:271-5. 23. Bolt WJ. Esophageal cancer trends and risk factors. Oncology 1994;2:403-10. 24. Gao YT, McLaughlin JK, Blot WJ, et al. Risk factors for esophageal cancer in Shanghai, China. I. Role of cigarette smoking and alcohol drinking. Int J Cancer 1994;58:192-6. 25. Gao YT, McLaughlin JK, Gridley G, et al. Risk factors for esophageal cancer in Shanghai, China. D. Role of diet and nutrients. Int J Cancer 1994;58:197-202. 26. Greenland S, Rothman K. Introduction to categorical statistics. In: Rothman K, Greenland S, eds. Modern epidemiology. 2nd ed. Philadelphia: Lippincott-Raven Publishers, 1998:231-52. 27. Greenland S, Rothman K. Introduction to stratified analysis. In: Rothman K, Greenland S, eds. Modem epidemiology. 2nd ed. Philadelphia: Lippincott-Raven Publishers, 1998:253-79. 28. Hosmer DW, Lemeshow S. The multiple logistic regression model. In: Hosmer DW, Lemeshow S, eds. Applied logistic regression. New York: John Wiley & Sons, 1989:25-37. 29. Garidou A, Tzonou A, Lipworth L, et al. Life-style factors and 452 30. 31. 32. 33. 34. 35. 36. 37. Tao et al. medical conditions in relation to esophageal cancer by histologic type in a low-risk population. Int J Cancer 1996;68:295-9. Ghadirian P, Ekoe JM, Thouez JP. Food habit and esophageal cancer an overview. Cancer Detect Prev 1992;16:163-8. Zheng W, Blot WJ, Shu XO, et al. Risk factors for oral and pharyngeal cancer, with emphasis on diet Cancer Epidemiol Biomarkers Prev 1992;l:441-8. Steinmetz KA, Potter JD. Vegetables, fruit, and cancer. I. Epidemiology. Cancer Causes Control 1991;2:325-57. Cheng KK, Day NE. Nutrition and esophageal cancer. Cancer Causes Control 1996;7:33-40. Cheng KK. The etiology of esophageal cancer in Chinese. Semin Oncol 1994;21:411-15. Tzonou A, Lipworth L, Garidou A, et al. Diet and risk of esophageal cancer by histologic type in a low-risk population. Int J Cancer 1996;68:300-4. Lu RF, Xu DD. A study of association between cancer incidence and diet in Shanghai. (In Chinese). Tumori 1987;7: 68-70. Sathiakumar N, Delzell E, Austin H, et al. A follow-up study 38. 39. 40. 41. 42. 43. 44. of agricultural and other chemical production workers. Am J Ind Med 1992;21:321-30. Morris RD, Audet AM, Angeliio IF, et al. Chlorination, chlorination by-products, and cancer a meta-analysis. Am J Public Health 1992;82:955-63. Mughal FH. Chlorination of drinking water and cancer: a review. J Environ Pathol Toxicol Oncol 1992; 11:287-92. Flaten TP. Chlorination of drinking water and cancer incidence in Norway. Int J Epidemiol 1992;21:6-15. Cantor KP. Water chlorination, mutagenicity, and cancer epidemiology. (Editorial). Am J Public Health 1994;84:1211-13. McGeehin MA, Reif JS, Becher JC, et al. Case-control study of bladder cancer and water disinfection methods in Colorado. Am J Epidemiol 1993;138:492-501. Rodrigues L, Kirkwood BR. Case-control designs in the study of common diseases: updates on the demise of the rare disease assumption and the choice of sampling scheme for controls. Int J Epidemiol 199O;19:205-13. Flanders WD, Louv WC. The exposure odds ratio in nested case-control studies with competing risks. Am J Epidemiol 1986;124:684-92. Am J Epidemiol Vol. 150, No. 5, 1999
© Copyright 2026 Paperzz