Diethylstilbestrol and clear cell vaginal carcinoma. Reappraisal of the epidemiologic evidence.

The association between DES and the development of clear cell vaginal carcinoma may have several alternative explanations rather than a cause- and-effect relation. One important possibility is that susceptibility bias, arising from reasons for use of the drug in a problem pregnancy, is a prime source of the observed effect in the case-control studies. Rather than being a cause of clear cell vaginal carcinoma, exposure to DES would serve as a prognostic "marker" to identify women born of problem pregnancies that increased their risk for development of the disease. Since medical advances in prenatal and perinatal care allowed these problem pregnancies to be carried to the term delivery of a viable child, DES may have been associated with an increased occurrence of clear cell vaginal carcinoma--without causing the disease. Another alternative hypothesis is needed to account for the high rates of reported previous exposure to DES in the case subjects of the two case-control studies. These rates may have been elevated by some form of interviewer bias or recall bias. The relatively high incidence of clear cell vaginal carcinoma that would be expected from the results of these case-control studies has not been observed in cohort studies. No instances of clear cell vaginal carcinoma have thus far been found in suitably assembled cohorts of women exposed to DES in utero. The history of science contains abundant examples of fervently held beliefs about cause- and-effect relations that were later found to be erroneous. The existing evidence of a DES/vaginal cancer relation is currently too weak for the causal role of DES to be regarded as established. To get better evidence, the problems of biased comparison and biased data can be addressed in at least three ways. The first is to carry out an appropriately objective new case-control study, with suitably chosen control subjects; the second is to study the issue of susceptibility bias by getting additional information about the occurrence of problem pregnancies in the mothers of patients with clear cell vaginal carcinoma who were not exposed to DES; the third is to review past tissue specimens from previously diagnosed genital adenocarcinomas, searching for clear cell cancers that may have been unrecognized. Until the suspected biases are addressed and either confirmed or refuted, the relation between DES and clear cell vaginal carcinoma remains a statistical association that is unaccompanied by the quality of evidence required for scientific conclusions.