Overview of studies on endometrial cancer and other types of cancer in humans: perspectives of an epidemiologist.

The epidemiologic studies relevant to the hypothesis that exposure to tamoxifen is associated with risk of cancer of the endometrium and perhaps other organs are described, summarized, and evaluated. One large body of data comes from the randomized trials of tamoxifen as a therapeutic agent against established breast cancer. With respect to endometrial cancer, these studies suggest that an association with tamoxifen use exists. However, the evidence is far from conclusive: the association is not seen in all the reported randomized trials, there appears to be a deficit of endometrial cancer in the comparison groups in two of the most important studies, none of the studies has adequately addressed the problems of confounding by prior hysterectomy and/or hormone replacement therapy, and none addresses the issue of detection bias. One nonrandomized cohort study shows findings similar to those from the randomized studies but also has similar problems, in addition to the problems associated with the lack of randomization of exposure. Case control studies were also considered, with the conclusion that they show results similar to those from the cohort studies (randomized and nonrandomized); that is, that the evidence leans toward the view that there is an association between tamoxifen use and endometrial cancer risk, but that it is incomplete and inconclusive. The issues of detection bias and potential confounding variables are better dealt with in the case control studies than in the cohort studies. One important question raised in one of the major case control studies is whether the apparent association is stronger with total dose of tamoxifen or with duration of use. Some investigators have reported an association between tamoxifen use and cancers other than those of the endometrium, but for none of these cancers is the evidence consistent between studies and the associations lack the theoretical underpinning that might be invoked for endometrial cancer, if in fact the existence of an association were established empirically for the latter site.