Objective To analyse the ability of simple clinical and biochemical parameters to predict glucocorticosteroid (GCS) treatment failure in patients with acute attacks of ulcerative colitis. Design/methods Retrospective analysis of clinical and biochemical data. Setting Four Swedish university hospitals. Patients Ninety seven patients with acute attacks of ulcerative colitis severe enough to warrant treatment with intravenous GCS, hospitalized during the years 1988–93. Main outcome measure Colectomy within the first 30 days after hospitalization, defined as 'clinical steroid resistance'. Results Thirty days after admission, 39 patients (40%) were in complete clinical and endoscopic remission while 33 (34%) had undergone colectomy. During follow-up for 24 months, four patients among the 39 initially in remission underwent colectomy. Among the 25 patients (26%) not attaining remission after 30 days, an additional nine patients subsequently required colectomy. Steroid resistance was associated with duration of disease (2.7 vs 8.1 years, P= 0.0037) and steroid treatment before hospitalization (70 vs 42%, P= 0.010). In particular, elevation of body temperature (37.4 vs 36.9°C, P = 0.012), persistence of diarrhoea (6.8 vs 3.6 bowel movements/day, P< 0.0001) and passage of blood (83 vs 42%, P= 0.0003) as well as CRP elevation (36.3 vs 18.0 mg/l, P= 0.007) on day 3 after treatment initiation were identified as predictors of a poor response. CRP ≥ 25 mg/l and >4 bowel movements/day on day 3 of hospitalization independently predicted a high risk for colectomy within 30 days. Conclusions Sustained elevation of body temperature, persistent bloody diarrhoea and continued CRP elevation on day 3 of intravenous GCS treatment strongly predict clinical steroid resistance in acute attacks of ulcerative colitis. In the group of poor or non-responders, colectomy or more aggressive medical treatment should be considered at an early stage.