The comparison of an incident case series with an incident series of second primary cancers, using either a case-control or follow-up study design, is proposed as an efficient method for evaluating the relative risk of a rare genetic susceptibility marker and its prevalence in the population, and for evaluating gene-environment interactions. The relative efficiency of this design versus a conventional case-control study is highly dependent on the population prevalence of the marker and its relative risk. However, for relatively rare but highly penetrant genes, the relative efficiency can be very high. In an example presented regarding a planned study of the p16 gene and its role in melanoma, a conventional case-control study may require up to 70 times as many subjects to achieve equivalent precision to the study of second primaries. The use of second primary cancers in this way requires assumptions about the validity of the classification of a new tumor as a second primary, the extent to which risk of a second cancer is influenced by treatment of the first cancer, and the nature and extent of surveillance bias. However, the problems of ascertaining a valid series of population controls are avoided. The study of second cancers represents an important and underused tool in molecular and genetic epidemiology.