Assessing the validity of observed cancer incidence trends.

The available cancer incidence data from the National Cancer Institute surveys, the Surveillance, Epidemiology, and End Results (SEER) Program, and the Connecticut Tumor Registry are examined for the period from the late 1940s to the present. Two indicators of completeness and accuracy of reporting (percentage of cases reported only by death certificate and percentage of cases with histologic confirmation of the diagnosis) show considerable variation, particularly in the past, but no consistent strong patterns indicating noncomparability of the data are apparent. Incidence data for five geographic areas (Atlanta, Detroit, San Francisco-Oakland, Iowa, and Connecticut) are examined for several primary sites by area and age, and are compared with mortality data for the same areas and the entire country. A variety of patterns in the trends are apparent. For several sites, the incidence trends are similar to what would be predicted based on the mortality patterns, but differences are apparent for other sites. Mortality trends may be influenced by shifts in diagnostic specificity on the death certificates and by changes in survival rates, and both incidence and mortality rates may be influenced by increased casefinding, improvements in diagnostic procedures, expansion of the medical care delivery system, and real changes in the prevalence of risk factors. Therefore, both incidence and mortality data should be used when attempting to assess the real trends in cancer occurrence because sole reliance on one or the other may lead to erroneous conclusions.