Contributions of autopsy to population-based cancer epidemiology: targetted intervention to improve ascertainment.

The importance of autopsy data in cancer epidemiological investigation is well appreciated by researchers at the Mayo Clinic. The Rochester Epidemiology Program Project is a complete population-based collection of morbidity and mortality data in Olmsted County, Minnesota (population, 100,000), funded by the National Institutes of Health since 1964. Comprehensive ascertainment of cancer in this community cohort has been augmented by initial diagnosis at autopsy for 12% of colorectal cancers, 35% of intracranial neoplasms, 45% of phaeochromocytomas and 37% of renal-cell carcinomas. These rates are in stark contrast to autopsy diagnoses in our cancer registry--usually less than 1%. The difference is attributable to the denominator in these comparisons, a concept of vital importance to the epidemiological interpretation of autopsy findings. Despite the enormous value of the autopsy in our research, social and economic factors diminished autoptic rates in Olmsted County from 63% in 1970 to 39% in 1984. These rates are well above that for the USA as a whole but parallel our national decline in frequency of autopsy. The major factors responsible for this reduction appear to be the increasing proportion of deaths in nursing homes and the advent of declining reimbursement in US health care.