Chemical contamination of water supplies.

Man-made organic chemicals have been found in drinking water for many years. Their numbers and varieties increase as our analytical capabilities improve. The identified chemicals comprise 10 to 20% of the total organic matter present. These are volatile or low molecular weight compounds which are easily identified. Many of them are carcinogenic or mutagenic. Chlorinated compounds have been found in untreated well water at levels up to 21,300 micrograms/L and are generally present at higher levels in chlorine-treated water than in untreated water. Aggregate risk studies for cancer are summarized. The most common sites are: bladder, stomach, colon, and rectum. Such studies cannot be linked to individual cases. However, they are useful for identifying exposed populations for epidemiologic studies. Five case-control studies were reviewed, and significant associations with water quality were found for: bladder cancer in two studies, colon cancer in three and rectal cancer in four. A large study by the National Cancer Institute found that there had been a change in the source of raw water for 50% of the persons in one area between the years 1955 and 1975. Such flaws in the data may preclude finding a causal relation between cancer and contaminants in drinking water. Large case-control and cohort studies are needed because of the low frequency of the marker diseases, bladder and rectal cancer. Cohort studies may be precluded by variations in the kinds of water contaminants. Definitive questions about these issues are posed for cooperative effort and resolution by water chemists, engineers, and epidemiologists.

Organic chemicals of anthropogenic sources are widely found in drinking water supplies. The problem is not a new one. In 1956, Middleton and Rosen (1) examined raw and finished water from five midwestern U.S. cities and reported the presence of benzene compounds, insecticides, kerosene, phenols, polycyclic hydrocarbon compounds, and synthetic detergents. With increasing sophistication of methods for chemical analysis, it is now possible to identify many organic chemicals in drinking water and to measure their concentrations in nanograms per liter. Summarizing reports from different countries, Kraybill (2) noted that, of 2221 organic chemicals found in water supplies on a worldwide basis, 765 were present in drinking water. Of these, 20 were recognized carcinogens, 23 were suspected carcinogens, 18 were carcinogenic promoters, and 56 were mutagens. Some of the recognized and suspected carcinogens identified in U.S. drinking water are shown in Table 1. Most of the chemicals reported in these surveys are volatile or low molecular weight organics which can be easily recovered from water (3), and these compounds are estimated to account for only 10 to 20% of total organic chemicals present in drinking water. Methods to analyze the nonvolatile organic fraction are still evolving but *Department of Epidemiology, University of North Carolina School of Public Health, Chapel Hill, NC 27514.  (4). most of the nonvolatile and high molecular weight compounds remain unidentified.
Organic chemicals enter water supply systems from a number of major sources. On a quantitative basis, the most important of these sources are industrial discharges, municipal waste water discharges, agricultural runoff, and decomposition of natural organic matter (humus). Other more localized sources include, among oth-ers, landfill leachates, accidental spills of chemicals, and leaking underground storage tanks.
Within the past 10 years, investigators have identified several organic by-products of the chlorination disinfection process. Among these compounds are chloroform (a known animal carcinogen), bromoform, bromodichloromethane, and dibromochloromethane. Chloroform invariably occurs in water which has been chlorinated, while it is absent or present at much lower concentrations in unchlorinated groundwater or in raw surface water prior to chlorination (5). However, nonchlorinated ground water is by no means free of organic chemicals, as shown in Table 2 (6). The significance of this finding is that a number of epidemiologic studies of water quality and cancer risk use populations consuming ground water or nonchlorinated water as a "nonexposed" standard for calculating cancer risk associated with exposure to surface or chlorinated water supplies.
Identification and quantification of the array of organic chemicals in drinking water is a complex and costly procedure and, even now, most of these compounds are rarely measured and none are routinely monitored. Hence, one of the major problems in assessing human health risks associated with chemical contamination of drinking water is the measurement or estimation of hu-man exposure. Can the epidemiologist utilize quantifiable indices of exposure such as levels of chloroform or total organics in water, surrogates of exposure such as chlorinated surface water versus unchlorinated ground water, or direct chemical analysis of specific chemicals?
If the latter, how variable are concentrations of these chemicals over time and place, and what is the relevant time interval between exposure and adverse health effect? Can the epidemiologist utilize any biological markers of human exposure to organic chemicals in water?
Epidemiologic Evidence for Adverse Health Effects Epidemiologic investigations of organic chemicals in water have virtually been limited to considerations of cancer risk. Isolated reports can be found on wells contaminated by leachates from pesticide waste dumps and reputed effects on liver enzyme concentrations in exposed persons (7), but systematic epidemiologic studies of the past decade have focused on cancer outcomes. These studies can be divided into two broad categories which are determined by the unit of observation: aggregate risk studies in which geographical areas (counties, cities, states) are the unit of analysis and individual   (8). b M = mortality, I = incidence. CU = urbanization; I = income; 0 = occupation; D = population density. risk studies. Aggregate risk studies can be performed quite rapidly because public data sources generally provide the required information on cancer mortality, demographic characteristics, and water quality for geographical areas which are the unit of analysis. Fifteen aggregate risk studies were reported for different geographical areas of the U.S. between 1974 and 1982; a recent review of these studies by Callas (8) is summarized in Tables 3 and 4. The most common cancer sites statistically associated with various measures of population exposure to organic chemicals in water were bladder, stomach, colon and rectum although other sites showing statistically significant relationships were esophagus, liver, gallbladder, pancreas, kidney, prostate, lung and breast. There was considerable inconsistency among cancer sites associated with water quality in these 15 studies. The problems with drawing causal causal inferences from aggregate risk studies are many and are inherent to the nature of the aggregate unit of analysis: there is great variability ofindividual exposure within a geographical unit; no individual measurements of exposure are known; potential confounders are seldom known and cannot be linked to individuals; individual exposures cannot be linked to persons who developed cancer. Aggregate risk studies may be useful for justifying more costly epidemiologic investigations of exposed individuals but they leave unresolved the question of causal relationships and provide no basis for quantitative risk assessment. A reasonable conclusion from the reported studies is that there are areas in the U.S. where cancer risk and various (usually surrogate) measures ofwater quality are associated, and the nature of this association should be investigated by conducting individual risk studies.
One historical cohort study relating water source (groundwater, chlorinated water in small towns, chlorinated water in cities) to cancer incidence for liver, kidney, and bladder was reported from an ongoing investigation of residents of Washington County, MD (29). No statistically significant associations with cancer incidence rates were observed, but the power of this study was low, given the relatively small population size (31,000 persons) for calculating incidence rates over a 12-year interval. Thus only 45 cases of liver cancer occurred in the period of followup.
Cantor gave a progress report on a large bladder cancer case-control study conducted by the National Cancer Institute (30). From the 10 cancer registries supported by NCI in each of the SEER areas, 3000 incident bladder cancer cases and 6000 controls were interviewed with respect to lifetime residential history, smoking habits, occupation, coffee and artificial sweetener use, and several other factors. From 1000 public water utilities in each of the 10 study areas, the investigators collected historical information on raw water surface versus groundwater or chlorinated vs. nonchlor- Table   Investigators Harris, 1974 (9) Buncher, 1975 (10   inated water at place of usual residence. Significant association with water quality were found for: mortality from cancer of the bladder in two of five studies, cancer of the colon in three of five, and cancer of the rectum in four of five case-control reports. Five individual risk studies, all of them case-control in design, have been reported in the literature (9) and are reviewed in Tables 5 and 6. Each of the five studies obtained cases and controls from state death registries and utilized data on exposure and potential confounders from information on the death certificate and secondary public sources. Since relatives of decedents were not interviewed, the investigators had no knowledge of smoking or dietary habits, residential mobility of individuals, personal consumption habits, or other factors which may have affected cancer risk. Crude surrogates of exposure to organics in water were utilized, such as sources, treatment practices, and geographical areas served. Approximately 50%o of persons in one area (Iowa) served in 1975 by chlorinated surface or groundwater sources had a different water source in 1955, demonstrating that recent water sources could lead to signif-icant misclassification of exposure status in this area if exposures of 20 years ago were the most relevant to cancer risk. The overall results of this major study have not been reported to date.

The State of Knowledge and Future Directions
At this point in time, the epidemiologic evidence on the issues of organic chemicals in water and cancer risk is no more than embryonic. Some associations found for aggregate units of analysis were also observed in a few case-control studies. Exposure to a crude measure of organic water quality appears to be associated with a small but signficantly increased risk of cancer of the colon, rectum and/or bladder. The studies to date are neither consistent in finding a relationship with one or the same set of cancer sites, nor did they convincingly control for potential confounding factors. Hence one cannot conclude that a causal relationship is by any means established. However, the generally positive nature of  [1960][1961][1962][1963][1964][1965][1966][1967][1968][1969][1970][1971][1972][1973][1974][1975] county group source for residence a Data of Callas (8). b UT = total urinary tract cancer, GI = total gastrointestinal cancer. cU = urbanization, 0 = occupation. The question to be addressed is, what further evidence should we seek, or where should new epidemiologic investigations go to shed light on the issue. Further aggregate risk studies are likely to be nonproductive. Clearly, we need some large case-control interview studies and cohort studies that might be fortuitously superimposed on other research objectives. Given the rare frequency of bladder and rectal cancer, a cohort study will have to involve many person-years of followup, since the incidence of these tumors is on the order of 10 to 20 per 100,000. However, the complexity of assessing exposure to organics in water may entirely preclude the conduct of any cohort study of cancer risk.
Assessment of relevant individual exposure to organics in water remains the thorny barrier to progress. The usual surrogates of exposure based on surface versus groundwater or chlorinated water versus unchlorinated water sources are unsatisfactory, given accumulating evidence for an array of organics in ground and unchlorinated drinking water. The resulting misclassification of exposure status, if nondifferential among cases and controls, will bias risk estimates towards the null hypothesis, and this bias may be severe. If surrogate measures are to be used (and alternative measures may not be justifiable until much more is known about the chemistry of organic contamination of water supplies), the epidemiologist will have to hammer out a better exposure index by consulting with knowledgeable water chemists and engineers. Questions to be addressed include: cancer; NS = no significant association. odds ratio.
* How can surface water be graded with respect to potential contamination with organic chemicals? * How can knowledge of upstream dischargers (for which data sources are available) be incorporated into the above question?
* Is the method of sedimentation, filtration, and other water treatment relevant to the formation of Dotentially harmful by-products of chlorine disinfection?
* Is there a predicable relationship between amount of chlorine added and chlorine residual (for which historical data exist) and concentration of chlorinated by-products?
* Is well depth relevant to potential exposure to organic chemicals? * Does proximity of wells or of underground aquifers to agricultural runoff affect the water quality of these sources? * Is there considerable temporal variation in organic constituents and concentrations for the same surface and groundwater source over the span of one year and of several years?
If some or all these questions were even partially answered, it would be possible to design interview studies which, combined with historical data available at each water utility, could lead to much more refined sur-rogate measures of exposure to organic chemicals in water. These surrogates might then reasonably be validated, at least for currently obtainable water samples, by comparison with direct chemical analyses.
Exposure estimates based on direct identification and quantification of organics in water seems an unfeasible approach to epidemiologic studies, given the large array of chemicals, the uncertainty of cancer risk related to nearly all of the chemicals, the multistage process of carcinogenesis, and the possibility of interaction between chemicals. Complexity and cost of chemical analyses would also seem to argue against direct measurements for individual risk studies.
A potentially fertile approach for population studies, especially for the limited sample sizes required in casecontrol studies, is that ofbiological markers of exposure. More than 100 volatile organics can be readily isolated from human blood and other body fluids (3). Wallace et al. (31) claim that breath samples, analyzed for volatile organics, reflect personal exposures to target chemicals in water and in air. Profiles of chemical adducts to PrA can also be obtained from individual blood samples. uonsiderable work needs to be done to related intake of known organics to these biological markers before the markers themselves could be judged feasible for epidemiologic investigations.
It appears that the immediate need is for a collaborative research effort, between epidemiologists, analytical chemists, and water quality engineers, to characterize human exposure to organics in water. This collaboration needs to be directed by the epidemiologists's research objectives, specifically relating cancer risk at least to an ordinal ranking of cancer relevant exposures. If a biological marker could best characterize this exposure, all the better since this would provide a personal integrated exposure estimate. Further epidemiologic studies of cancer risk and water quality, in the absence of progress of exposure estimation, seems unlikely to advance our knowledge of the nature of this relationship.