Randomized clinical trials of antibiotic prophylaxis in acute pancreatitis are now warranted in the light of recent evidence of pancreatic penetration of certain antibiotics at therapeutic minimal inhibitory concentrations. The aim of the present prospective clinical study was to investigate whether there are detectable risk factors for pancreatic sepsis in acute pancreatitis that would allow better selection of patients for inclusion in clinical trials. Fifty-nine consecutive patients with acute pancreatitis were recruited and submitted to admission baseline and 48-h determinations of Ranson score, and assay of C-reactive protein at admission and weekly intervals thereafter. Contrast-enhanced computed tomography (CT) was also performed within 24 h of admission. Pancreatic sepsis, defined as infection of pancreatic and/or peripancreatic collections, was demonstrated in all cases by culture of samples obtained by needle aspiration and at laparotomy. Although all prognostic indices correlated significantly with sepsis, multivariate logistic regression analysis showed that the only variables predictive of the risk of subsequent sepsis were the presence and extent of necrosis. Early detection of pancreatic necrosis by CT should be the primary inclusion criterion in future clinical trials of antibiotic prophylaxis in acute pancreatitis.