A prospective, blinded study of CA19-9 in 2,467 patients having abdominal surgery yielded 356 patients with pancreatic, gallbladder, and biliary disease who submitted coded preoperative serum specimens. In this group, there were 84 patients with pancreatic cancer and 24 patients with gallbladder-biliary cancer; the remainder had benign lesions. The recorded imaging data and marker results were merged with the patients' demographic, clinical, and surgical data and tissue diagnoses for analysis. Receiver operator character calculation suggested that a reference value of 100 U/ml for CA19-9 was appropriate rather than the 37-40 U/ml value most frequently employed and yielded a specificity of 97% in the 467 operated patients with a sensitivity of 8.3% for all nonpancreatic-biliary cancers and 62% overall for these lesions. In the more diagnostically challenging nonicteric patients, CA19-9 sensitivity was 55%, specificity was > 99%, positive predictive value (PPV) was 97%, and negative predictive value (NPV) was 88%. When CA19-9 results were combined with those from endoscopic retrograde cholangiopancreatography, ultrasound (US), or computed tomography (CT), the PPV, and especially the NPV were increased. The addition of carcinoembryonic antigen results did not affect overall results. The addition of CA19-9 results to ambiguous or indeterminant imaging interpretation clearly improved the combined specificity, sensitivity, and PPV, but the change was less impressive, albeit positive, for NPV. The combination of CA19-9 and CT (or US) is a reasonable, cost-effective, noninvasive approach to establishing the diagnosis of pancreatic, cholangitic, or biliary cancer in nonicteric patients. Although no single procedure or combination of procedures was found to detect early, small lesions, CA19-9 is clearly a clinically useful adjunct to imaging in nonjaundiced patients suspected of having these malignancies.