The aim of this study was to assess the predictability of the outcome of acute pancreatitis using the Ranson, Glascow, and Acute Physiology and Chronic Health Evaluation (APACHE) II scores, the computed tomography (CT) scan, and several serum markers. Altogether, 137 consecutive patients with acute pancreatitis confirmed by CT scan were prospectively included. Blood samples were obtained daily for 6 days. The predictive value of each parameter was studied by univariate and multivariate analyses comparing mild and severe pancreatitis. A total of 111 attacks were graded as mild (81%) and 26 as severe (19%). Ranson (p = 0.3) and APACHE II (p = 0.049) scores appeared insufficiently predictive in the univariate analysis. Pancreatic imaging by CT scan was insufficiently predictive (p > 0.05), whereas the presence of extrapancreatic fluid collections was more indicative of outcome (p <0.05). With the univariate analysis, the four most reliable serum markers were pancreatic amylase (p <0.001), neutrophil elastase (p <0.05), albumin (p <0.002), and C-reactive protein (p <0.001). Results became homogeneous when the CT results were added; serum albumin plus extrapancreatic fluid collections (negative predictive value 92%-96% and positive predictive value 67%-100%) comprised the best indicator of severity. None of the parameters tested achieved sufficient predictability when used alone. Serum albumin plus extrapancreatic fluid collections comprise the best indicator of severity at the time of admission.