Prospective comparison of three non-invasive tests for pancreatic disease.

Ultrasonography, computed tomography (CT), and an oral pancreatic function test were evaluated prospectively in 107 consecutive patients suspected of having pancreatic disease, the final diagnosis being made independently of these tests. Seventeen patients proved to have either chronic pancreatitis or a pancreatic neoplasm. Five had an unfused ventral pancreas and 10 had "minimal change" pancreatograms. Seventy five patients did not have pancreatic disease. Among the 17 patients with pancreatic disease the sensitivities of ultrasound, CT, and the pancreatic function test were 70%, 71%, and 76% respectively; among those without disease values of specificity of the three tests were 75%, 87%, and 72%. All three tests failed to detect most of the patients with unfused ventral pancreases or minimal change pancreatograms. The predictive values of a positive result for the three tests were 43%, 67%, and 34%. The predictive value of a negative result was greater than 90% for all the tests. Results improved when CT was combined with either of the two other tests. These findings show that CT is the most accurate of the three tests in diagnosing pancreatic disease but that a combination of ultrasound and the oral pancreatic function test offers a more widely applicable and almost as accurate alternative.