We performed a multicenter, randomized, controlled clinical trial of therapeutic peritoneal lavage (2 liters per hour for three days) in 91 patients with severe acute pancreatitis. Patients were entered into the study if severe pancreatitis was indicated by multiple laboratory criteria or diagnostic peritoneal lavage. All patients received full supportive treatment. The median time between the onset of symptoms and randomization was 38 hours. Forty-six patients were assigned to the control group and 45 to the lavage group. There were 13 deaths (28 per cent) and 16 patients with major complications (35 per cent) in the control group, as compared with 12 deaths (27 per cent) and 17 patients with major complications (38 per cent) in the lavage group. Lavage did not appear to modify the length of survival, the incidence of pancreatic collections (pseudocysts or abscesses), or the plasma amylase concentration. Considering the statistical power of the design, we conclude that the outcome of severe pancreatitis was not greatly, if at all, influenced by the regimen of peritoneal lavage used in this study.