EDITORIALS Pesticide Residues and Breast Cancer?

As if we needed it, the Journal brings another reminder of the caution with which the results of a single epidemiologic study, or even a handful of them, should be regarded. Hard on the heels of the largest epidemiologic study, until now, showing a positive relationship between breast cancer risk and serum levels of DDE (the major metabolite of DDT), and published last year in the Journal (7), comes the even larger study reported in this issue in which no relationship was found (2). An editorial (3) also accompanied the 1993 report, clearly stressing the need for further research to confirm the provocative finding before considering its "far-reaching implications for public health intervention worldwide" (7). The two studies are similar in a number of respects. Both are case-control studies nested in cohort studies—the 1993 study in the prospective New York University Women's Health Study (7) and the 1994 study in the Kaiser Foundation multiphasic health examination cohort in the San Francisco Bay Area (2). In both studies, the sera were drawn prior to the diagnosis of breast cancer. In the New York study, as the authors did recognize, breast cancer was probably present at the time the blood was drawn, since it was diagnosed within 6 months of the draw. However, the Kaiser study has a number of features that favor its conclusions over those of the New York report. Although the largest up to that time, the New York study had only 58 breast cancer cases; the new study has 150. The New York study had slightly more control women (171) than did the new study (150), but there was careful matching of controls to cases in both studies, and the greater number of cases in the new study is a substantial strength. Moreover, because of the much larger numbers of potential cases available, the Kaiser investigators were able to select three groups of equal size from each of three major racial groups; in New York, 80% of the cohort were Caucasian and it is not indicated that race was matched in the selection of controls. The Kaiser data indicate that race may be a potential confounder of this relationship, although it is not clear whether the same would be true on the East coast, and the direction of the confounding cannot be predicted with confidence. Race may have some intrinsic relationship to DDE levels, but, in the present context, it is of more likely concern as an indicator of socioeconomic status. Perhaps most significantly, the Kaiser study was focused on women who enrolled in the years 1964-1969, a period prior to the federal restrictions on use of DDT and 1-26 years before their diagnoses of breast cancer. This intake period is probably reflected in the substantially higher mean levels of DDE in the Kaiser study—four to five times those of the New York women (analyzed in the same laboratory). However, other factors, such as geography and socioeconomic status, may also be involved. Incidentally, although there are no data on the relative breast cancer incidence rates in the two cohorts, it is difficult to imagine that they could differ by the magnitude that would be expected if DDE levels had any substantial etiologic role and differed by the factors suggested by the DDE levels measured in these studies. What is the possibility that DDT or its metabolites are involved in breast cancer development, but at a late stage, so that the association would be seen in samples taken within 6 months of diagnosis of the disease but not in women whose sera were sampled many years before? It seems unlikely. First, DDE is extremely persistent in fat tissue, and women with high exposures, even 20 years before the diagnosis of their disease, would still be likely to have high levels during the later stages of the development of their breast cancers. Second, the Kaiser report gives data for 14 women with sera collected less than 5 years prior to diagnosis and 20 women with collections 5-9 years prior to diagnosis; neither group suggests a relationship of breast cancer risk to DDE levels. The investigators explored a number of other hypotheses linking exposure to particular temporal phases in the development of breast cancer, but to no avail. Clearly, we shall hear more on this topic, but, for the moment, we must conclude that the available epidemiological evidence overall is not supportive of an association between exposure to DDT and increased risk of breast cancer. I suspect that I was asked to write this editorial not simply to recapitulate the observations and conclusions that can be derived from the accompanying report, but to ponder on why the conclusion from the perfectly respectable study of Wolff et al. (7), and of a few previous, less powerful studies appear, at this point, to have been incorrect. The spectrum of man's diseases is complex, and his environment labyrinthine. Their interaction forms a thicket that is made only more dense by the addition of biochemical and other laboratory markers of exposure and/or disease. The imaginative investigator looks for patterns in this thicket, as others look for pictures in clouds. Investigators' own ambitions, as well as the demands of the systems in which we work, have led to widespread belief that on perceiving a pattern in the thicket it is better to report it, even if it turns out to be only a bunch of leaves, than to fail to report a pattern that someone else later discovers to be a pheasant. Many investigators, including the writer of this editorial, have proceeded on the basis of this belief (4). If science operated in a vacuum, the tendency to over-report suspicious patterns would not necessarily be a bad one—it gives