Epidemiologic approaches to multiple primary cancer: critical review from the standpoints of validity and efficiency

Epidemiologic approaches to multiple primary cancer (MPC) have been reviewed from the standpoints of validity of study design and statistical efficiency. It was shown that the expected number of MPC is often over- or underestimated when the person-year method is used in calculation. This so-called detection bias becomes stronger the poorer the prognosis of patients with their first cancer, and the longer the duration of the preclinical state of their second cancer. The bias is also affected by the probability of detecting the second cancer at the time as the first. An alternative method has been proposed, in which the prevalence rate of preclinical cancer is evaluated. The efficiency with which the increased frequency of second cancer is detected is affected not only by the risk ratio for individuals at high risk to those at low risk but also by the proportion of high risk individuals in the population.