CA 19-9 is a promising radioimmunoassay for the detection of pancreatic cancer, but its clinical role and cost-effectiveness are not yet known. To investigate these factors, we used clinical decision analysis to study diagnostic strategies for patients with suspected pancreatic cancer presenting as pain or weight loss. Comprehensive diagnostic strategies were developed to reflect current and future patterns of practice utilizing CA19-9 radioimmunoassay (RIA) to yield biopsy-proved cancer or confidently exclude its presence. The performance of the strategies beginning with CA19-9 RIA and ultrasonography were equivalent in positive and negative predictive values over a range of prevalence of pancreatic cancer from 0.02 to 0.15. At higher prevalence, the negative predictive value of the ultrasonography strategy became significantly better. The CA19-9 RIA strategy used fewer noninvasive tests, endoscopic retrograde cholangiopancreatographic procedures, and invasive radiologic studies than did the ultrasonography strategy at each prevalence. The health care costs ranged between $848 and $1413 per patient for the CA19-9 RIA strategy and $1186 and $1848 per patient for the ultrasonography strategy. We conclude that the CA19-9 RIA is a useful, cost-effective initial test for the examination of patients with suspected pancreatic cancer.