To compare percutaneous tracheostomy with conventional operative tracheostomy. Randomized clinical trial. The medical and surgical critical care units of a large, tertiary-care, private hospital. Twenty-five male and 21 female translaryngeally intubated patients with respiratory failure, in whom tracheostomy was indicated on clinical grounds, were randomly assigned to one of two groups. The 24 patients in group 1 underwent conventional operative tracheostomy, and the 22 patients in group 2 underwent percutaneous tracheostomy. One patient in group 2 required tracheostomy on three separate occasions during a prolonged hospital stay. Patients were examined daily throughout their hospital stays for adverse events related to the tracheostomy. In all patients who survived until decannulation, plain tomography of the trachea was performed within 3 days of decannulation. Repeat physical and tomographic examinations were performed 6 and 12 wks later. Fifty-eight percent (14/24) of the operative tracheostomies were associated with at least one complication, compared with 25% (6/24) of the percutaneous tracheostomies (p < .05, 95% confidence interval 7% to 59%). Predecannulation problems were more frequent in group 1 patients than in group 2 (46% vs. 13%, respectively; p < .01, 95% confidence interval 9% to 57%), as were later sequelae (88% vs. 27%; p < .05, confidence interval 26% to 96%) in survivors. Group 1 patients were more likely to have multiple complications, and their complications tended to be more serious. In this study, percutaneous tracheostomy appeared to be superior to the conventional operation.