Oral contraceptives and cervical neoplasia: pooled information from retrospective and prospective epidemiologic studies.

A positive association between the use of oral contraceptives and the risk of cervical neoplasia has been suggested by different sources. This paper examines epidemiologic evidence on this issue through pooled computation of relative risks emerging from 21 studies, subdivided into three main categories: studies based on routine cytologic screening programs, case-control investigations and prospective studies. The pooled estimates of the relative risks for ever vs never use of oral contraceptives were broadly similar and slightly above unity (1.4 from cytologic screening programs, 1.1 from case-control and 1.4 from prospective studies). The risk increased with duration of use and, generally, lower relative risk estimates derived from the older studies, necessarily based on short-term use of oral contraceptives. A particularly limited increase in risk was apparent from case-control investigations. Furthermore, when allowance was made for the major covariates (mostly indicators of sexual habits), a noticeable decrease in the excess risk was evident. In conclusion, although statistical significance of moderate differences in risk can be obtained by pooling data from several studies, a risk of the magnitude of that for the association between oral contraceptives and cervical neoplasia may well be due to bias and confounding. In addition, since sexual behavior only indicates the probability of having been exposed to some sexually transmitted agent (most likely human papilloma virus), the greatest additional contribution may come from a case-control study in which adjustment for such exposure is possible.