Abstract

Many cancers and other chronic diseases are associated with a long delay between exposure to a putative risk factor and subsequent diagnosis. This presents well recognised problems in the elucidation of suspected risk factors by epidemiological methods. In this paper we discuss the interpretation in epidemiological studies of the effect of a possible risk factor when population exposure is recent and rapidly changing. An important contemporary example concerns the study of early oral contraceptive (OC) use in relation to the subsequent risk of breast cancer. Computer simulations reported here indicate that plausible delays in the manifestation of any effect on breast cancer incidence make it difficult to exclude early OC use as a risk factor for breast cancer, even when large well conducted epidemiological studies show no apparent increased risk. Methods for detecting a 'latent' effect are discussed.