Host factors have been given scant attention in the search for etiology in pancreatic cancer. Several anecdotal reports have identified its familial clustering, whereas a recent population-based case/control study has shown that 6.7% of cases and 0.7% of controls had positive family histories of this disease (p less than 0.001). Forty-seven individuals with pancreatic cancer from 18 families were identified from a review of the medical records of all kindreds on file at our Hereditary Cancer Institute. The observed sex ratio, age of onset, histologic type, and survival were comparable to published data on unselected patients. We did not identify any pattern of extra-pancreatic cancer association. A serious limitation of our study is its lack of a population-based case/control design. Whereas our data are primarily descriptive, they do indicate the need to learn more about the role of familial factors in the etiology of pancreatic cancer. Pancreatic cancer is increasing in incidence, and its prognosis is almost uniformly dismal; identification of persons at high risk may improve cancer control.