Hogan M D, Chi P Y, Hoel D G, Mitchell T J
J Environ Pathol Toxicol. 1979 Jan-Feb;2(3):873-87.
This paper discusses some of the statistical and biological problems that are likely to be encountered when an indirect or ecological approach is used to assess the possible public health impact of general population exposures to environmental agents. For purposes of illustration, the potential association between various site-specific cancer mortality rates and chloroform levels in public drinking water supplies was considered. The analyses that were performed demonstrated that, for the data sets under consideration, there were some definite associations between chloroform levels and cancer mortality for specific sites such as the rectum-intestine and bladder. However, the marked extent to which these results were dependent on (1) the weighting scheme adopted in the analysis, (2) the presumed appropriateness of the data, and (3) the characteristics of the statistical model was also clearly illustrated. Because of these dependencies the quantitative, causal interpretation of results generated from an indirect study would appear to be a very tenuous and questionable practice in most instances.
