Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
IEpidemiological Studies CANCER FREQUENCY AND CERTAIN ORGANIC CONSTITUENTS OF DRINKING WATER After comparing cancer rates among people in Louisiana whose drinking water comes from the Mississippi River with rates in populations served by water from other sources, Harris (1974) concluded that something in the river water led to increased cancer rates. He suggested that the lower Mississippi contained many chemical pollutants, some, or some combi- nation, of which were carcinogenic. The initial study was challenged by DeRouen and Diem (1975a, 1977), and several other hypotheses were put forth. At that time the Environ- mental Protection Agency (EPA) was monitoring a large number of chemicals in U.S. water supplies among them, chloroform and the other trihalomethanes (THM's). The National Cancer Institute (NCI) also studied the biological effects of chloroform in its large animal bioassay program and subsequently demonstrated an increase in liver and kidney tumors among animals exposed to high doses of chloroform (Page and Saffioti, 1976~. Given these several pieces of information, the EPA asked a number of research groups to detains whether there was indeed a relationship between cancer rates and chloroform and other THM's in water supplies. Even though the liver and kidney were the only sites at increased risk in the animal experiments, the various research groups attempted to evaluate the cancer risk at many other sites as well. The 5
6 DRINKING WATER AND H"LTH rationale for this is that exposure to a single carcinogen may result in cancer at different sites in different experimental animals and in humans. For example, vinyl chloride has produced nephroblastoma in rats but not in humans. There is also ample evidence of chemicals that produce cancer in some species but not in others; for example, 2-naphthylamine is a potent bladder carcinogen in humans, dogs, and monkeys but is . . . . . inactive ln rats and mice. Most of the EPA-requested studies used indirect evidence of the presence of THM in water supplies. Populations receiving surface water, which often contains organic materials and is sometimes treated with chlorine, were contrasted with those receiving groundwater. Because groundwater usually contains less organic matter and is less often chlorinated in the purification process than surface water it is less likely to contain THM's. Two of the studies used direct measurements of chloroform and other THM concentrations, which had been obtained by the EPA in two recent analytical surveys in a regression equation for U.S. county mortality rates with allowance being made for various demographic factors by including further linear regression terms. The above epidemiological studies (published and unpublished), which comprise to our knowledge the sum total of epidemiological work on this subject, were reviewed by the subcommittee. None of the studies reviewed was adequately able to take into account many well-established risk factors for cancer rates at different sites. For example, for bladder cancer one needs to control for occupation' cigarette smoking, use of alcohol and drugs, nonaqueous sources of THM, coffee consumption, socioeconomic status, and ethnicity. The studies also had to assume that present exposures to THM reflected lifetime exposures of the popula- tions studied: this is not only because THM concentrations in water have only recently been measurable, but also because the necessary information on migration patterns with associated THM concentrations is not readily available. The subcommittee has summarized the various studies, made a critical assessment, pointed out where it believes the evidence is deficient, and discussed the potential for further research on this subject. The probabilities of false associations when many comparisons are made were also considered when final conclusions were made.
Epidemiological Studies 7 Summary and Conclusions The studies that the subcommittee reviewed were divided into two groups: those in which nonspecific measures of exposure to putative carcinogens in water (e.g., the use of surface water versus groundwater) were examined and those in which water quality was characterized by measurements of THM concentrations. The subcommittee gave greater weight to the conclusions of the latter group of studies because crude measures of exposure, which lead to comparisons of cancer between surface-water users and groundwater users, must be of limited value. They do not permit the quantitation of exposure to contaminants in water consumed, which is needed to determine dose-response relation- ships between THM concentrations and cancer frequencies and to estimate the ejects of reducing THM concentrations. The conclusions drawn in the second group of studies, in which many cancer sites were examined, suggest that higher concentrations of THM's in drinking water may be associated with an increased frequency of cancer of the bladder. The results do not establish causality, and the quantitative estimates of increased or decreased risk are extremely crude. The effects of certain potentially important confounding factors, such as cigarette smoking, have not been determined. The bladder is not one of the sites found to be at increased risk in experiments on animals exposed to chloroform. Tumors of the liver and kidney developed in laboratory animals. The positive association found for bladder cancer was small and had a large margin of error- not only statistical, but, much more importantly, because of the very nature of the studies. Hogan et al. ( 1978, unpublished) found regression coefficients of bladder cancer mortality (deaths/105/year) on chloroform concentration (,ug/liter) in drinking water to be approximately 0.003 for males and 0.0021 for females. Thus, if there is a causal relationship, an increase in chloroform concentration of 100 ~g/liter might lead to increases of 0.3 deaths/105/year from bladder cancer in males and 0.2 deaths/105/year in females. These compare with the U.S. National Mortality Rates (195~1969) for cancer of the bladder of 6.8 in males and 2.4 in females. Thus, a decrease of 100 ,ug/liter of chloroform in water may lead to a 4.4% decrease in bladder cancer death rates in males and 6.7% in females. These changes, which have been shown in case-control studies to be explicable by as little as one to two cigarettes per day differences in average cigarette consumption, would probably be too small to be distinguished from possible confounding ejects by any epidemiological study. Situations that could be more favorable for investigation should be sought so that further epidemiological work might be more rewarding.
8 DRINKING WATER AND H"LTH Studies Based on Indirect Measures of Water Quality MISSISSIPPI RIVER-LOUISIANA, FIRST STUDY The first papers reporting relationships between cancer frequency and water quality were those of the Environmental Defense Fund (Harris, 1974, unpublished; Page et al., 19764. These two papers were based on the published cancer mortality data by county in the United States (Mason et al., 19741. In each Louisiana parish the proportion of water that was obtained from the Mississippi River was related, in a regression analysis, to cancer mortality. The study was controlled for urban-rural status, income, and employment in certain potentially hazardous industries. Significant positive correlation coefficients were found in two of the four race-sex groups for genitourinary cancer, and in all four groups for gastrointestinal cancer. MISSISSIPPI RIVER-LOUISIANA, SECOND STUDY Using the same water quality data and county mortality data as Harris (1974, unpublished), DeRouen and Diem (1975b, unpublished) reported a major difference between northern and southern regions of Louisi- ana a potentially important confounding factor for the cancer-water association. This was also noted by Buncher (1975, unpublished). DeRouen and Diem's analyses of cancer mortality and water source were restricted to southern Louisiana where parishes using only Mississippi River water were located. DeRouen and Diem compared cancer mortality at 16 anatomic sites in four race-sex groups (52 site-sex- race groups in all) between persons who received river water and those who did not. They found seven positive associations (P < 0.05) of mortality with use of river water in one or more sex-race groups for seven sites or site groups. Similarly, significant negative associations were found for seven sites. The consistency of associations in both sex and racial groups suggested positive associations for cancer of the bladder, breast, (i.e., lower mortality in populations using river water) for cancer of the liver and corpus uteri. Cancer of the lung and cervix uteri had a significant positive association in one population and a significant negative association in anot. her. OHIO RIVER AREA, FIRST STUDY A similar approach was used on data from the Ohio River area. Buncher (1975, unpublished) studied the four-state area comprised of Ohio, West
Epidemiological Studies 9 Virginia, Kentucky, and Indiana; Kuzma et al. (1977) confined their attention to Ohio. In the Ohio study, white males and females in all Ohio counties were compared according to the predominance of surface water or groundwater in their supplies. Sites with statistically significant (P<0.05) excess mortality associated with surface water were the stomach in males and females and the bladder in males. In two additional site-sex groups, the liver and breast in females were positive at the 20% level. These rates were adjusted for age, county population, percent urbanization, median income, an index of manufacturing activity, and an index of agricultural, forestry, and fishing activity. OHIO RIVER AREA, SECOND STUDY Harris et al. (1977~ unpublished) reviewed the Ohio studies. Although they generally agreed with the findings of Kuzma et al. (1977), they failed to confirm the positive association for liver and breast cancer in females. In addition. they reported positive associations for the esophagus in males (not studied by Kuzma) and the pancreas in males. They studied the effect of water source on cancer mortality by regression analysis similar to that used in the first Ohio area study. but they used as a water variable the percer~tage of a county population that received surface water. UPSTATE NEW YORK Alavan~a et al. (1977 unpublished) conducted a case-control study of persons dying with gastrointestinal or genitourinary cancer in several counties in upstate New York. The study was first confined to women to decrease the potentially confounding factor of occupational exposures, and the cancer sites were selected on the basis of indications from prior reports. Consequently. this may be thought of as an hypothesis-testing study. Each cancer case was individually matched with a woman dying of a cause other than cancer. Matching variables were county, age, race, and birthplace. Statistical analysis procedures did not take the matching into account, but it is impossible to judge the effect of this omission. Water supplies were determined from addresses and were characterized as chlorinated or not chlorinated and as surface water or groundwater. A doubling of risk was found to be associated with chlorinated supplies in urban areas. A 50~c increase in risk was found to be associated with chlorinated groundwater when compared with r~onchlorinated water irk rural areas. There was no
1O DRINKING WATER AND HEALTH increase (risk ratio, 1.01) in rural areas supplied with chlorinated surface water in comparison with rural areas using nonchlorir~ated water. In 1978, Alavanja's study was expanded to include men. He reported: Males living in the chlorinated water areas of Erie, Rensselaer and Schenectady counties and females living in the chlorinated water areas of Erie and Schenectady counties are at a greater risk of gastrointestinal [GI] and urinary tract tUT] cancer mortality than are individuals living in nonchlonnated water areas. Moreover, this excess risk of GI and UT cancer mortality is not due to a disparity in the ages race or ethnic distribution of the population or to an A . . _ ~ ~ ' · ' · ~ _ urban/rural factor, or hazardous occupation, or to Inorganic carcinogens two, As, Be, Pb. Ni, NO3 ~ or a surface water/groundwater difference. In one county (Chautauqua) there was a significant deficit of both kinds of cancer mortality in chlorinated areas. Because of the large population of Erie County the data acquired there strongly influence the total risks reported. Urbanization is a potentially confounding variable in the study of ejects of chlorination. and parceling out shares of the excess risk between urbanization and chlorination does not seem possible. although some attempts have been made. For example, urban and rural nonchlorinated areas were compared with urban and rural chlorinated areas WASHINGTON COUNTY, MARYLAND 1 1 Kruse (1977. unpublished) studied cancer of the kidney and liver in a Population in Washington County' which includes the city of Hagers- town Maryland. This population has been investigated extensively as an epidemiological community models and the source of home drinking water had been determined in an earlier survey. This previously gathered information was used in a total population study designed to correlate water source characterized as chlorinated or not chlorinated. with presence or absence of cancers at specific sites. The approach involved estimating the effect of water source through a multivariate analysis including eight additional potentially confounding demographic and sociological variables. There was an increased incidence of liver cancer associated with the chlorinated water supply with a risk ratio of 1.5. but the increase was nest statistically- significant at the 20~ level. Similarly. a decrease in incidence of cancer of the kidney with chlorination (risk ratio. 0.96) was not significantly associated with water source. There were only 91 cancer cases in this study.
Epidemiological Studies 11 LOS ANGELES COME In Los Angeles County, Mah et al. (1977, unpublished) correlated cancer mortality rates and incidence rates with chloroform content in drinking water. They divided the county into nine subareas categorized as having low, medium, or high chloroform content of the water. Eight cancer sites were studied: esophagus, stomach, colon, rectum, liver, lung, kidney, and bladder. Tabular and graphic inspection of data and limited analyses suggested no association of cancer rates with chloroform content of water, and the investigators believed no more detailed analysis was justified in view of certain limitations in the basic material, viz., absence of precise data on chloroform content, extensive use of bottled and other water transported from other sources by this population, and large population movements into and out of Los Angeles County in the last several decades. OHIO RIVER, THIRD STUDY Salg (1977, unpublished) studied populations in areas served by water from the Ohio River, including all counties within the boundaries of the Ohio River Valley Basin as determined from water drainage maps. These counties lie within a seven-state area: Illinois, Indiana, Kentucky, Ohio, Pennsylvania, Tennessee, and West Virginia. The exposure variables, indicating possible intake of pollutants in the water supply, were the percent of county population served by surface water and the percent served by prechlorinated (defined as chlorination prior to filtration) water. The outcome variable was cancer mortality for 346 counties, specific for 19 anatomic sites of cancer. The main variables were investigated in a multivariate analysis using as possible confounding factors nine variables representing demographic and socioeconomic characteristics of the counties. There were positive associations (P < 0.2) for 13 of the 19 site categories in one or more of four sex-race population groups. Similarly significant negative associations were found for seven site categories in one or more populations. In two or more population groups there were positive associations for cancer of the esophagus, rectum, breast, larynx, and for Hodgkin's disease. Salg interpreted her findings to indicate the need for further study of carcinoma of the large intestine, rectum, and bladder. Large intestine cancers were found to be significantly increased in white males. Bladder cancer incidence was significantly elevated in white males and sig- nificantly decreased in nonwhite females.
12 DRINKING WATER AND HEALTH PITTSBURGH STUDY Carlson and Andelman (1977, unpublished) conducted a study in the Pittsburgh region. They associated site, race, and sex-specific, age- adjusted cancer incidence rates by 1969-1971 census tract with drinking water, which was characterized by source of raw water (surface, ground, and river) and by water treatment plant. Their paper is principally an investigation of statistical methodology. Their general conclusion is that significant associations, both positive and negative, are found between water quality and cancer incidence. The data are not summarized in a way that permits identification of cancers of specific sites. NEW JERSEY STUDY Vasilenko and Magno (1975, unpublished) studied all 21 counties in New Jersey to determine the relation between water source and age- adjusted cancer mortality from lung' stomach, and urinary tract cancer of white males between 1950 and 1969. Water quality was estimated from the ratio of the number of households served by public systems and private water companies to the number served by individual wells. Confounding variables included in the analysis were urban-rural characteristics, income, education, migration, occupation, industrializa- tion, and concentration of sulfur dioxide in the air. In a multiple regression analysis, the water variable was positively associated with mortality from respiratory cancer and stomach cancer. A nonsignificant negative association was found for bladder cancer. SUMMARY Nine of the ten studies described above showed a number of associa- tions, some of which were statistically significant, between indirectly characterized water quality and cancer rates (incidence or mortality). One, the Los Angeles County study, reported no associations, but this study appeared to have greater limitations than any of the others. Cancer rates at several sites were positively associated with water quality in one or another study, but no site consistently predominated. The bladder, stomach, large intestine, and rectum, which were cancer sites identified in a number of geographic areas, warrant further study. The effects of certain important demographic variables were consid- ered in a number of studies of this group, but other confounding factors, e.g., cigarette smoking, were not.
Epidemiological Studies 13 Studies Based on Trihalomethane Concentrations Three studies (Cantor et al., 1977, unpublished; Hogan et al., 1978, unpublished; McCabe, 1975, unpublished) have been conducted using measured THM concentrations. All three studies used cancer mortality data rather than incidence data. Although incidence data are usually preferable, the studies could not use these data as they are not generally available for the areas covered by the THM surveys. Cantor et al. studied the relation between age-standardized cancer mortality for 1968 to 1971 of white men and women in U.S. counties that were categorized as urban on the basis that more than 50~o of the county population lived in urban areas in 1970. County THM concentrations were estimated from data obtained in two surveys conducted by the EPA (U.S. Environmental Protection Agency, 1975~. The National Organics Reconnaissance Survey (NORS) sampled finished water in 80 water treatment plants across the country, and a survey conducted by EPA's Region V office covered 83 plants in Illinois, Indiana, Michigan Minnesota, Ohio, and Wisconsin. The analysis took into account the median school years completed by county inhabitants over age 25, foreign stock composition of county, county population, ratio of 1970 to 1950 county population, percent of county that is urban, percent of the county work force engaged in all manufacturing industries (U.S. Bureau of the Census, 1970), and major geographic region of the United States. An attempt was made by multivariate regression to explain the variability among counties of mortality rates for each site of cancer with sex-specific mortality rates greater than 1.5/100,000/yr. The residual mortality rates, which were "unexplained" by these other variables, were then correlated directly with measured THM levels (logarithms of the concentrations) for the 76 counties in which 50S'o or more of the population was served by the sampled water. To calculate the correlation coefficients, the data were weighted by the square root of the sex-specific person-years at risk in the population served by the sampled water supply as estimated by the product of percent of population served and population at risk. The statistical techniques used by the authors, particularly the weighting factors, are not standard and probably tend to decrease statistical . . ~ slgnl~cance. The results of Cantor's analysis are shown in Tables II-1 and II-2. Among males, a significant (P ~ 104703 positive correlation for the 76 counties was found between nonchloroform trihalomethane (NCTHM) concentration and bladder cancer. When the percent of county popula- tion served by the sampled water supply is increased to reduce the
14 .5 a' _1 Ct o Cal U. et - (,, Ct a' I_ ._ _ _ ~ ~ Cal ~ _ . _ ~ Cal o 3 ~ C U) ~ ~ o .O V ~ C,_ o o ·- I_ Cal a' o . 3 - o,a ~ 5 ~ .= ha A: o . - G o o ID - U, EM - .= o 3 lo :~ Cal ._ Cal a, o o - ~ ,~ ~ 2 ~ .O o ~ - ~ oo ~ u, .O ~ ~: ae v, c~ a, ~ ~ - : ~o ~ ~ ~o o ~ a, a' 0 c, - o .3 ~3 :^ . ~ :., C~ ~ ~ _ ~ __ _ _ ~ a~ 0 ~ ~ ~ ~ - ~ ~ \O oo t- \ _ ~ o V) _ _o oo o ~ _ _ o . . . . . .. . . . . . . . o o, o o o o~o o, o o o o, o o, ~_'_ _ _ o ~ oo ~ ~ ~ eh ~, _ ~ _ ~ o ~ o j o o o,o o o o o o _ _ _ _ _ _ o oo ~ o ~ o oo ~ _ _ ~ _ - , . . . . . . . . . . o o o ~ o o, o _ o o _ _ _ _ _ ~ ~ ~ o ~ ~ ~ ~ o o X ~ _ ~ ~ _ ~ _ ~(, . . . . . . . . . . . o o o o o o o o o o o ~o, Co "o ~o o o _ _ o o ~ Z Z ·~ on o o s _ ~ ·s ~o ~ m Z :C ~o C ~ V, ~ ~ _ _ _ 00 O - ~ O oo _ . . . . . . . . O O O O O 0~ 0 0~ '_ ~ _ _ CJ~ =' ~ - - ~ ~ O _ ~ o oo ~ - O C~ 7 ° ° ° 7 V' a' =: o _ Ct s .= _ _ _ ~ _ _ ~ ~ t ~ ~ _ 0 ~ - tn 0 ~#2 °°°7 7 x . =: o 2 ., ~o o - o o ao D o 3 - o .3 .= C~ - - ~: 3 C~ o ._ - .8 V, u C~ ~o 3 o C~ a' ~o - - ~ 3 _ -. ~ ~ _ 7 CL ~ f~ ~ t0) _ O D ~ ~ 5
15 ._ U. Ct a' - C~S U' an - C~ ~ AL .= _ at ~ ._ ~ o > ._ a.) .° C: :~ AL o 3 m U' ~ ,~ V Cal o o o 8 ·- ~ ~d As-, o ~ V a,) 1 ~ _ ~ ·= ._ A: o . - Ct - o :` I: o - U, a, . 3 Is - - rO - ._ C) Cal o I: o ._ - C~ ~o o C) ~ C~0 g ·= _ o~ O ~ ~ .O oo i~ _ 00 C~ o . o_ _ X :3 o~ 8 U) C~ ~ c) o ._ o_ _ 1 o o~ o V~ _ ~ ~ o So .g e~ ~ ~ 5 o ~ ° a,,~ · °e .- ~ U' ~ ~ _ _ ~ _ _ _ __ l- - ] ~ ~ ~ _ ~ ~ ', _ ~ ~ ~_ t_ _ _ o ~ C~ o ~ o o oo o o~ o ~o . . . . . . . . . . . . . .. . o o o o, o o~ o o, o o, o ~ o oo o~ _ _% _ _ _ _ _ ~o ~ _ ~ ~ ~ U~ ~ ~ r~ ~ ~ ~ ~ o ~ - - ~ ~ o o ~ o oo _ ~o . . . . . . . . . . . . . .. T°~°°-°°°°-=I °~=I o~oo Ol _ _ _ _ _ _ __ _ ~ ~ ~ X ~ _ o o ~ ~ ~ ~ o~ o _ ~ _ V) ~ _ ~ ~ ~ ~ _ ~ o ~ ~ . . . . . . . . . . . . . .. . o o, o o o o, o o o o, o o o oo o, _ _ _ _ _ _ __ _ ~ _ ~ ~ ~ _ ~ ~ ~ ~ _ ~ ~ X 0 ~ ~ - c~ r~ 0 ~ r~ ~ ~ - 0 ~ 0 . . . . . . . . . . . . . .. . o o o o o o o o o o o o, o oo o :r ~ ~ ~ ~0 X ~0 ° X :r o ~o ~Z Z U) "c ~c =. c D o 3 o Cd .= ._ ao - C~ - - _ ~ .- s ~ ~ ~o ._ Ct _ ~ ~ o s U' o ~ "o ~ _ o U, s ~ s o `. ~ V) _ ~ Z ~ . ~ ~ (~ o C~ e~ - s t> a o s 3 _ ,cn _ X ~ ~ £ q~ U~ ° ~ ,~ o o ~ C ' - ° .= D ~ B0 ~ C _
16 o 3 ~2 g .~ .ou cn - g ._ .~ - ca ~n o U) a, ~ 3 C o s: C: ·_ c: .- o ~ =: C: .= ~ _ V, o - LU {: ~: ~ o E~ U' ~a - . 3 . 3 o .= a' C~ ~ . ~ V' ~: o z o ·Eb o z . o o _ . . . . o o o o ~ ~o ~ . . . o o o o _ o ~ ~ _ o o 8888888 o o o o o o o _ ~ ~ _ _ ~ o o o o ~ o ~ oo . . . . . . . o o o o o o o ~ _ ~ o ~ ~ - o ~ _ ~ ~ o o _ o o o o o o o o o o o o o . . . . . . . o o o o o o o 1 1 1 1 _ ~ c~ ~ ~n ~ ~ o o ~o o ~ ~ o 0 0 0 0 0 oo ~ ~ ~ oo _ ~ _ ~ ~ _ 0888888 oo 0 0 0 0 0 1 _ ~ 0 oo ~ ~ ~ 0 ~ _ 0 0 0 _ . . . . . . . O 0 0 0 0 0 0 r~ ~ ~ ~ O _ ~ ~ ~ o~ ~ ~ _ 0 0 0 0 0 0 0 0 0 0 0 oo . . . . . . . O 0 0 0 0 0 0 1 1 1 1 . =, E o o cn ~ ~ oo r~ - - CS - x E o ~s
Epidemiological Studies 17 misclassification of exposure, correlations with bladder and brain cancer tend to increase. Among women, positive correlations were found between total THM's and lung cancer and between NCTHM and bladder cancer. The lung cancer correlation coefficients did not show a dose-response relationship with the proportion of the population receiving the measured water. But the correlations with bladder cancer did: from 85% to 1007 reached conventional significance (P = 0.02~. A preliminary analysis showed a fairly strong association between halo- methane levels and colon cancer rates, but control for composition of the population by ethnicity removed this association. The authors computed correlation coefficients separately for three geographic regions (North, South, and Mountain Pacific) for the 51 counties in which 65% or more of the population received measured water. There were positive correlations between NCTHM and bladder cancer in males in all regions combined (r = 0.30, P = 0.03), and in females in each region separately as well as combined (combined r = 0.33, P = 0.024. For men in the North, a significant correlation was observed between NCTHM and bladder cancer (r = 0.52, P = 0.02~. Hogan et al. (1978, unpublished) conducted a similar study, in which they related earlier cancer mortality data to chloroform levels in finished water as determined by the same surveys. They used the National Cancer Institute's 20-year age-adjusted county cancer mortality rates for white men and women (Mason et al., 1974) and county chloroform concentrations in drinking water estimated from the EPA's NORS and Region V surveys (U.S. Environmental Protection Agency, 19754. Regression equations were fitted to the mortality rates using, as independent variables, chloroform concentrations, 1960 county popula- tion, county population density, percent of county that is urban, percent of county population that is nonwhite, percent of county population that is foreign born, median number of school years completed by county residents over age 25, median family income of county, and percent of county work force engaged in manufacturing. Weighting was by total population exposed (both sexes and all races combined). The results for sites where any association approached statistical significance are shown in Table II-3. The data are consistent with an increase in cancer rates of the rectum, the bladder, and possibly the large intestine with increased chloroform concentration. McCabe showed that age-adjusted total cancer mortality rates correlated positively with estimated chloroform concentrations in 80 cities. Since no allowance appears to have been made for any of the confounding factors, it is inappropriate to draw conclusions from this study to compare with the Cantor and Hogan studies.
18 DRINKING WATER AND HEALTH All studies were seriously limited by the absence of data on past exposures, which are the only ones that are directly relevant to cancer that has already been diagnosed. Similarly, all studies were deficient in identifying populations that were stable in the areas where the water quality was studied. None of these papers discusses the nature of the correlation or regression coefficients obtained: in particular, whether the positive results found were due to one or two extreme observations or to a more relevant trend over the range of THM values. Similarly, none discusses the adequacy of using a simple linear regression equation to allow for certain confounding variables. Without such discussion interpretation must be especially circumspect. Results of these studies demonstrate the problems of establishing relationships between health statistics and environmental variables, and lend emphasis to the caution with which they should be interpreted. PROSPECTS FOR FURTHER EPIDEMIOLOGICAL STUDY Adequate exploratory and hypothesis-generating work has been done. Studies raising suspicion of higher cancer rates among persons whose major water supplies came from surface waters with generally higher THM concentrations compared to rates among those whose major water sources were groundwaters have been followed by attempts to relate the cancer rates to actual THM concentrations. Further studies of this kind are unlikely to lead to more useful information, although an examination of age-specific data might prove fruitful. Consequently, future studies should be more specific and should examine possible confounding factors in detail. When developing actual exposure data, i.e., THM concentrations in water consumed, investigators should gather information concerning duration of exposure to a particular water source. Further investigation of the validity and reliability of chemical analyses of water constituents would also be in order. Such epidemiological studies without experimen- tal intervention usually cannot uncover small effects. Moreover, with so many confounding factors, it would be difficult to ascribe an effect to any factor with certainty. For discussion, designs for further studies may be conveniently divided into two major types: case-control studies and cohort studies, . . . . . . He us sing intervention stuc lies. Case-Control Studies These studies can start with cases (incidence, prevalence) or deaths. For certain investigations the information from
Epidemiological Studies 19 deaths may be adequate. However, the degree of detail necessary for investigating carcinogens in water makes it highly desirable that information be collected from the subjects themselves, rather than from near relatives or other proxies. Some populations are exposed to varying amounts of chloroform, e.g., users of certain cough medicines and toothpastes that contain high concentrations of chloroform. Such individuals should be identified among cases and controls. The site most warranting a case-control study is the bladder, possibly followed by the colon and rectum. Confounding factors include occupation, cigarette smoking, use of alcohol and other drugs (including artificial sweeteners), nonaqueous sources of THM, coffee consumption, socioeconomic status, and ethnicity. Cohort Studies Cohort studies, attempting to follow "exposed'' and unexposed populations over time, may also be possible. Some natural experiments have beers created by using different methods of water purification. Thus, a population from a community using activated charcoal filtration that is adequately and frequently regenerated might be compared over time with a population from a community not using that process. Due attention must be given to latency the time difference between initial exposure and the appearance of disease. This applies to both retrospective and prospective studies. Studies should be undertaken only after there is a clear understanding of the magnitudes of detectable differences. Because the effects antici- pated from the usual concentrations of THM in drinking water are not expected to be great, large populations must be studied to demonstrate associations. Similarly, if there is no association, large populations must be studied to exclude the existence of the small effects that have been postulated. Some useful data may be available from foreign countries and the help of international agencies (World Health Organization-International Agency for Research on Cancer, North Atlantic Treaty Organization, U.S.-Japan Scientific Exchange, U.S.-USSR Scientific Exchange) should be sought. References Alavanja, M., I. Goldstein, and M. Susser. 1977. Report of Case Control Study of Cancer Deaths in Four Selected New York Counties in Relation to Drinking Water Chlorination. Report of EPA Contract 76-224. Unpublished. 16 pp. Alavanja, M. 1978. Gastrointestinal and Urinary Tract Cancer Mortality and Drinking Water Chlorination. Unpublished. 16 pp.
20 DRINKING WATER AND H"LTH Buncher, C.R. 1975. Cincinnati Drinking Water-An Epidemiologic Study of Cancer Rates. Division of Epidemiology and Biostatistics, University of Cincinnati Medical Center. Report to the Board of Health, City of Cincinnati. 123 pp. Cantor, K.P., R. Hoover, T.J. Mason, and L.J. McCabe. 1977. Association of Halometh- anes in Dnnking Water with Cancer Mortality. Environmental Epidemiology Branch, National Cancer Institute, Bethesda, Md. Unpublished. 24 pp. Carlson, W.S., and J.B. Andelman. 1977. Envirorunental Influences on Cancer Morbidity the Pittsburgh Region. Final Report prepared for Environmental Protection Agency Health Effects Research Laboratory, Cincinnati, Ohio, by Environmental Health Section, University of Pittsburgh. 21 pp. DeRouen, T.A., and J.E. Diem. 1975a. The New Orleans drinking water controversy. A statistical perspective. Am. J. Public Health 65:106~1062. DeRouen, T.A., and J.E. Diem. 1975b. Ethnic, Geographic Differences in Cancer Mortality in Louisiana. Unpublished. 18 pp. DeRouen, T.A., and J.E. Diem. 1977. Relationships between cancer mortality in Louisiana dnnking-water source, and other possible causative agents. Pp. 331-345 in H.H. Hiatt, J.D. Watson, and J.A. Winsten, eds., Origins of Human Cancer. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. Harris, R.H. 1974. The Implications of Cancer-Causing Substances in Mississippi River Water. A report submitted to James A. Moreau, Councilman-at-Large, New Orleans, La. Unpublished. 35 pp. Harris, R.H., T. Page, and N.A. Reiches. 1977. Carcinogenic hazards of organic chemicals in- drinking water. Evaluation and comparison of Louisiana and Ohio methods to determine risk. Environmental Defense Fund, Washington, D.C. Unpublished. 20 pp. Hogan, M. D., P-Y. Chi. T.J. Mitchell, and D.G. Hoel. 1978. Association Between Chloroform Levels in Finished Dnnking Water Supplies and Vanous Site-Specific Cancer Mortality Rates. Unpublished. National Institute of Environmental Health Sciences, Research Tnangle Park, N.C. 21 pp. Kruse C.W. 1977. Chlorination of Public Water Supplies and Cancer. Preliminary Report. Washington County, Maryland Experience. School of Hygiene and Public Health, The Johns Hopkins University, Baltimore, Md. Unpublished. 15 pp. Kuzma, R.J., C.M. Kuzma, and C.R. Buncher. 1977. Ohio drinking water source and cancer rates. Am. J. Public Health 67:725-729. Mah, R.A., G.H. Spivey, and E. Sloss. 1977. Cancer and Chlorinated Drinking Water. University of Califorrna. Los Angeles. Unpublished. 99 pp. Mason, T.H., F.W. McKay, R. Hoover, W.J. Blot, and J.F. Fraumeni. 1974. Atlas of Cancer Mortality for U.S. Counties: 195~1969. U.S. Department of Health, Education, and Welfare, National Institutes of Health, Washington, D.C. 103 pp. McCabe, L.J. 1975. Association Between Halogenated Methanes in Drinking Water and Mortality (NORS Data). Water Quality Division, Environmental Protection Agency. Unpublished. 4 pp. Page, N.P., and U. Saflioti. 1976. Report on Carcinogenesis Bioassay of Chloroform. U.S. National Cancer Institute, Bethesda, Md. Page, T., R.H. Harris, and S.S. Epstein. 1976. Drinking water and cancer mortality in Louisiana. Science 193:55-51. Salg, J. 1977. Cancer mortality rates and drinking water in 346 counties of the Ohio River Valley Basin. Final report EPA contract no. PO-5-03-4528. Department of Epidemiolo- gy, University of North Carolina. 136 pp.
Epidemiological Studies 21 U.S. Environmental Protection Agency. 1975. Preliminary assessment of suspected carcinogens in drinking water. Report to Congress. U.S. Environmental Protection Agency, Washington, D.C. 52 pp. U.S. Bureau of the Census. 1970. Census of Population. Government Printing Office, Washington, D.C. Vasilenko, P., and L. Magno. 1975. Factors Relating to the Incidence of Cancer Mortality in New Jersey. Pnuceton University, Princeton, N.J. Unpublished. 32 pp. WATER HARDNESS AND CARDIOVASCULAR DISEASE The Present Status of Knowledge The following statement regarding water "hardness" appears in Drinking Water and Health (National Academy of Sciences, 1977), a report of the National Academy of Sciences-National Research Council Safe Drink- ing Water Committee (SDWC). Hardness may be defined as the sum of polyvalent cations present in water. The most common such cations are calcium and magnesium. Hardness is usually expressed in terms of the equivalent of calcium carbonate (CaC03~. There are no distinctly defined levels of what constitutes a hard or soft water supply. Generally water with less than 75 mg/liter (ppm) of CaCO3 is considered soft and above this concentration as hard. The use of 75 mg/liter as the concentration separating soft and hard water was derived from practical considerations of effects, such as boiler scaling, rather than from observations on human health. This clas- sification of waters as hard or soft is an oversimplification of the concept of hardness since the major contributors to hardness, calcium and magnesium, vary in relative proportions. Moreover, the classification does not describe quantitatively the possible presence of a wide range of polyvalent cations, including strontium and barium, which are some- times found as trace elements in waters. Concentrations of vanadium, lithium, chromium, and manganese are higher in hard water. Cadmium, lead, copper, and zinc tend to occur in soft water at the tap, largely as a result of the greater corrosiveness (aggressiveness) of soft water in the distribution system. However, untreated hard water sometimes contains high levels of these elements. Therefore, the pattern is variable. The chemical composition of water that is softened by municipal treatment or in the home will depend upon the softening method that is used. Ion-exchange methods raise the concentration of sodium in water, an effect that may have adverse consequences for health, at least for some individuals.
22 DRINKING WATER AND H"LTH Current knowledge of the relationship of water hardness to cardiovas- cular disease is derived largely from "ecological" epidemiological studies, in which individual exposures or risk factors have generally not been considered. These studies were reviewed extensively by the National Academy of Sciences-National Research Council Panel on the Geochemistry of Water ire Relation to Cardiovascular Disease Notational Academy of Sciences, 19791. The status of knowledge in this field may be summarized as follows: 1. In general, when studies encompass large geographical areas, hard water is correlated with low cardiovascular disease rates. This correlation breaks down when smaller areas are considered or when the studied populations are grouped by altitude or the proximity of a seacoast. Some noncardiovascular ailments are also associated occasionally with soft water areas, raising the possibility that soft water may merely be a concomitant of some more basic risk factor. 2. Most studies have reported correlation coefficients and not risk estimates as a function of exposure. A few contain sufficient data for risk estimation: upper estimates of risk ratios for soft compared to hard water average approximately 1.25 for cardiovascular disease and 1.2 for arteriosclerotic, stroke, and hypertensive diseases CNational Academy of Sciences! 19791. 3. From autopsy studies in Canada and the United Kingdom, investigators have reported low magnesium levels in various tissues (heart, diaphragm, and pectoral muscle) of persons who died from myocardial infarction compared to those who died from accidental causes. They observed similar deficits in persons from soft water areas as compared to hard water areas. However, conflicting data have been reported. The SDWC was aware of detailed studies in progress in Canada and in the United Kingdom, but was unable to evaluate their significance since the results of these studies only became available late in the committee's deliberations. Recommendations for Alteration of Current Practices Given the current status of knowledge regarding water hardness and the incidence of cardiovascular disease, it is not appropriate at this time to recommend a national policy to modify he hardness or softness of public water supplies. The data do not indicate clearly which (if any) additions to soft water would benefit human health. However, certain
Epidemiological Studies 23 factors need to be considered carefully before individual decisions can be made. Municipal authorities who are either softening their water supplies or are considering doing so should carefully assess the necessity for this treatment. If they decide that it is essential, they should select a method, such as lime softening, that does not elevate sodium levels. Ion-exchange resins are often used to soften water, particularly in home water supplies. The ion-exchange process softens water by replacing calcium or magnesium with sodium. One atom of divalent calcium Or magnesium is replaced by two atoms of sodium. In terms of actual weights of the elements, 100 mg of calcium would be replaced by about 115 mg of sodium. Thus, softening can result in increases in the concentration of sodium in the softened water. Avoiding such an increase might be prudent given the state of knowledge on effects of sodium consumption. The increase in the sodium concentrations in the water caused by ion-exchange softening is a direct function of the hardness level of the untreated water and the operating efficiency of the softening device. When a physician advises that an increase in sodium intake is undesirable, a "three-line system" should be used if soft water is desired for laundering and bathing. The line that supplies the drinking water taps bypasses the softener thus preventing the consumption of water with an increased sodium level. Recommendations for Future Research Priorities Further studies should be conducted on the relationship between water hardness and cardiovascular disease. Repetition of the types of studies that have generally been conducted to date is unlikely to yield the information that is necessary for the determination of a nationally applicable policy. Epidemiological evidence should be gathered from studies of individu- al exposures to different water supplies, from studies of communities that have experienced changes in their water supplies, such as those resulting from discontinuation of softening and from changes in the composition of their drinking water. Studies to assess mortality rates should be controlled for such intercommunity differences as ethnicity and smoking. Some of these opportunities for studies could, with relatively little expenditure, be turned into very useful controlled prospective epidemio- logical studies if sufficient time is available for careful planning of control populations and maintenance of migration records. The results of clinical research from two types of studies may provide useful information on the relationship of water hardness and cardiovas
24 DRINKING WATER AND H"LTH cular disease. Tissue samples obtained during autopsy should be analyzed for a variety of elements, including magnesium, sodium, silicon, lead, and copper, and results examined for association with specific causes of death. Controlled trials of magnesium supplements (as fortified food or tablets) with groups of individuals at high risk of cardiovascular disease could also provide useful information. References National Academy of Sciences. 1977. Drinking Water and Health. Safe Drinking Water Committee, National Academy of Sciences, Washington, D.C. 939 pp. blational Academy of Sciences. 1979. Geochemists of Water in Relation to Cardiovascu- lar Disease. Panel on the Geochemistry of Water in Relation to Cardiovascular Disease, National Academy of Sciences, Washington, D.C. 98 pp.
