Methodology for evaluating the incidence of second primary cancers with application to smoking-related cancers from the Surveillance, Epidemiology, and End Results (SEER) program.

The Surveillance, Epidemiology, and End Results (SEER) database is used to estimate the standardized incidence ratios (SIRs) of second primaries for all pairs of smoking-related cancers and to interpret the results in the context of our knowledge of the known effects of smoking on the incidence of each of the individual cancers. In evaluating the simultaneous incidence of two cancers, one must recognize the inherent duality of the two relevant SIRs linking the two cancers (e.g., A and B), namely, the SIRs of A following B and B following A. Under fairly general assumptions, the two SIRs are seen to be equal, and departures from equality suggest changes in risk status or the introduction of new risk factors after the incidence of the first primary. Based on these observations, a methodological strategy is developed. The data reveal several clear patterns. First, short-term incidence is uniformly much greater than long-term incidence. Second, the SIRs are consistently much higher for women than for men, for every pair of cancers studied. Third, the magnitudes of the SIRs are generally high and often substantially higher than would be expected on the basis of the known risks of smoking. Exceptionally high SIRs are observed between kidney and bladder cancer and between head and neck and esophageal cancer. Various influences may affect these high observed SIRs, including artifactual influences such as diagnostic, surveillance, and misclassification biases, and the effect of different exposure prevalences on subsequent SIRs, which may to some extent explain the strong sex differences. However, these artifacts do not appear to explain the magnitude of the observed SIRs, especially the very strong associations between kidney and bladder cancer and between head and neck and esophageal cancer. It seems likely that other factors play a role, including, possibly, host susceptibility factors or additional common risk factors other than smoking. Although multiple primary cancers are rare, they represent an especially fruitful population for detailed epidemiologic study.