duce different diseases, making the determination of their relative role more difficult (Lilienfeld and Lilienfeld, 1980).
Specificity also is diminished when diseases are inappropriately grouped together, or inaccurately classified, obscuring important differences. Thus, a proposed study of vinyl chloride-exposed workers in 1973 would have failed to detect the real effect of increased cancer, because it lumped the relatively rare form of cancer involved, namely angiosarcoma of the liver, into the category of all cancer. This particular study included over 10,000 workers in 37 plants, the majority of the industry at the time. In preparing the analysis of this group, the researchers calculated the SMRs in which the expected numbers of deaths from specific causes were derived from national mortality statistics for all cancers combined. This aggregation of cancer would have obscured the extraordinary finding that one fifth of all recorded incident cases of angiosarcoma for the U.S. in a single year occurred in this group of highly exposed vinyl chloride workers (Utidjian, 1988). Subsequent studies that used appropriate classifications of disease detected significant excesses of angiosarcoma and brain cancer in exposed workers (Utidjian, 1988). Excesses of angiosarcoma have also been detected in residents living near a vinyl chloride manufacturing plant in New York (Brady et al., 1977).
Where a given factor is related to many diseases, its specific causal association with a single disease can prove difficult to demonstrate. Thus, vinyl chloride emissions have also been tied to clusters of birth defects, but the study lacked sufficient power and the findings were not significant (Rosenman et al., 1989). In general, hazardous wastes have been linked in toxicological studies to a wide range of diseases, some of which have long latencies and many of which have multiple causes. Moreover, the common nature of many of the health problems suspected to be caused by exposure to hazardous-waste sites makes the identification of their specific causes problematic. Here again, the problem of multiple comparisons occurs, in that the study of a number of different diseases in different locations will randomly produce some elevations due to chance alone.
Does the relationship between exposure and disease occur regularly in independently conducted studies? To revert to the example of passive smoking and lung cancer, although the RR might be 2.0 or less, this elevated risk was reported consistently in more than 30 different studies conducted in six countries (NRC, 1986a). Even where