A multivariate analysis was performed to determine the outcome, and factors prognostic for outcome, in 57 patients requiring surgical intervention for radiation bowel injury. The actuarial 2- and 5-year cause-specific survival (CSS) was 76 and 74%, respectively, with a median follow-up of 62 months for the survivors. The median time from surgery to death from complications was 4 months. Identified sites of injury were both large and small bowel. The types of injury were defined as stricture, perforation, inflammation, and fistula. At surgery 9 patients had more than one site, and 15 patients had more than one type of injury. Cox proportional hazards regression models relating survival to individual patient characteristics were constructed using surgical procedure, radiation-surgery interval, age, stage, radiotherapy technique and dose, and the individual sites and types of injuries. Only the site of injury was found to be of prognostic significance for CSS (P less than 0.03). However, when the site and type of injury were recoded as single or multiple, Cox regression analysis found both the site (P = 0.008) and the type (P = 0.02) of injury to be statistically significant for CSS (favoring single sites and types). Stepwise multivariate regression analysis found type of injury to be insignificant when site of injury was already in the model. Ileal damage was associated with the lowest CSS of any single site of injury (56%) and also appeared to be responsible for the poor CSS of those with multiple sites of injury (46%).