Colectomy is a potentially life-saving procedure for patients with severe attacks of UC who fail medical therapy. We aimed to systematically review studies that reported the short-term colectomy rate in severe UC or reported variables that could predict treatment failure. We conducted a systematic literature search for cohort studies and controlled trials published between 1974-2006. Thirty-two studies met the inclusion criteria; 16 reported short-term outcome and predictors of therapy failure, 13 only outcome, and 3 only predictors. In the pooled analysis, 581 of 1991 patients required colectomy (weighted mean 27; 95% confidence interval [CI], 26%-28%), and 22 died (1%; 95% CI, 0.7%-1.5%). In a heterogeneity-controlled meta-regression, colectomy rate did not change during the last 30 years (R(2) = 0.07, P = .8). Cyclosporine was used in only 100 patients, with a 51% (95% CI, 41%-60%) short-term success rate. A second meta-regression failed to demonstrate a dose-colectomy response of methylprednisolone therapy beyond 60 mg daily (R(2) < 0.01, P = .98). More than 20 variables were identified in 19 studies to predict medical therapy failure, but only a few were consistently reproduced: disease extent, stool frequency, temperature, heart rate, C-reactive protein, albumin, and radiologic assessment. The short-term colectomy rate in severe UC has remained stable during the last 30 years, despite the introduction of cyclosporine, which was not used frequently. We could not find any support for administering methylprednisolone at a higher dose than 60 mg/day. Variables that predict outcome of corticosteroid therapy could aid in the development of guidelines for introduction of rescue therapies in severe UC.