Fifty-two patients with severe Crohn's disease were enrolled in this study. Six (11.5%) were intolerant of the medication and had to be excluded. The remaining 46 patients were treated with rifabutin in combination with a macrolide antibiotic (clarithromycin or azithromycin). Patients were treated for a mean of 18.7 (range 6-35) months and followed up for 25.1 (range 7-41) months. Of the 19 patients who were steroid dependent at the start of this study, only two continued to require steroids when treatment was established. A reduction in the Harvey-Bradshaw Crohn's disease activity index occurred after 6 months' treatment (P = 0.004, paired Wilcoxon test) and was maintained at 24 months (P < 0.001). An improvement in inflammatory parameters was observed as measured by a reduction in erythrocyte sedimentation rate (P = 0.009) and C-reactive protein (P = 0.03) at 18 months compared with pretreatment levels, and an increase in serum albumin at 12 months (P = 0.04). When subsets of the study population were analysed, patients with pan-intestinal disease achieved better remission at 2 years than did those with less extensive involvement (P = 0.04, Mann-Whitney U-test). No difference in treatment response by age, disease duration, the presence of granulomas on histology, or the occurrence of drug-induced side-effects, was observed. These data suggest that treatment with rifabutin and clarithromycin or azithromycin may result in a substantial clinical improvement in Crohn's disease and justify the conduct of a randomized controlled trial.