three exposure categories in the study years 1979-1984. The standardized incidence ratio for females was 1.53 (95 percent CI 1.02-2.21); for males it was 1.00. The authors were unable to explain this gender difference, but it would seem to weaken the argument that the leukemia was caused by exposure to the toxins. However, for males other exposures, such as those in the workplace, could be more important determinants of leukemia than environmental exposures.
A nationwide ecologic study was conducted by Griffith et al. (1989) using the National Priorities Listing (NPL) of hazardous-waste sites developed by EPA and cancer mortality data for 13 major cancer sites by county for 1970-1979. The NPL identified 593 waste sites in 339 U.S. counties in 49 states with analytical evidence of contamination in groundwater that provided a sole source of potable water supply. Significant associations between excess deaths and all counties with hazardous-waste sites were shown for cancers of the lung, bladder, esophagus, stomach, large intestine, and rectum for white males and for cancers of the lung, breast, bladder, stomach, large intestine, and rectum for white females, when compared with the counties that did not have hazardous-waste sites. No adjustment was possible for the effect of industrialization or other confounders at the individual level, such as factors related to lifestyle, including cigarette smoking. Many of the associations were strong and of a similar order in both sexes; there were odds ratios of 5.9 in males and 4.3 in females for cancer of the large intestine. The odds ratio for lung cancer was higher in females (5.2) than it was in males (2.0). These associations could have resulted from factors not considered in the analysis or from other problems with the method of the studies. For example, Greenland (1990) has shown that ecologic estimates can be more sensitive to misspecification and misclassification than are individual-level estimates, primarily because ecologic estimates are based on assumptions and inferences rather than on direct measurements.
Wong et al. (1989) conducted ecologic and case-control analyses to evaluate the possibility that contamination of drinking water in Fresno County, California, with dibromochloropropane (DBCP) increased mortality from gastric cancer and leukemia. The study was conducted because a previous ecologic analysis by the California Department of Health Services for 1970 to 1979 found such an association. Mortality data for 1960-1983 were included in the ecologic analyses, and deaths for the period 1975 to mid-1984 were used in the case-control studies; nongastric cancer or leukemia deaths served as controls. The only positive association was an increased risk of leukemia for farm workers. No relationship was found for gastric cancer or leukemia with DBCP contamination of drinking water.