121 patients with acute pancreatitis thought to be due to gallstones were randomised to treatment with urgent endoscopic retrograde cholangiopancreatography (ERCP) and endoscopic sphincterotomy (ES) or with conventional treatment. They were stratified by predicted severity of the attack, according to the modified Glasgow system. ERCP was done within 72 h, and if common bileduct stones were identified, patients underwent ES immediately to extract the stones. There were fewer complications in the 59 patients who underwent ERCP +/- ES than among the 62 treated conventionally, the difference being confined to those whose attacks were predicted to be severe (6/25 ERCP +/- ES [1 death] compared with 17/28 conventional treatment [5 deaths]). Hospital stay was also shorter for patients with severe attacks who underwent ERCP +/- ES than for those who received conservative treatment (median 9.5 versus 17.0 days).