Statistical methods for studying multiple primary malignant neoplasms.

Case reports and case series dealing with multiple primary malignant neoplasms provided useful criteria for defining and documenting this phenomenon. The formation of tumor registries greatly aided in identifying a sufficient number of multiple primary cancer patients and facilitated case-control comparisons. Reports of two or more neoplasms occurring together in the same individual do not constitute proof of a significant association; the tumors must be shown to occur together more frequently than expected by chance. The person-years approach applied to data derived from a well-defined population makes it possible to compare the observed and expected number of subsequent primary cancers. The results of the most sophisticated procedures are no better than the quality of the data, however, and one must critically examine possible sources of bias before accepting statistical significance as representing biologic significance.