A randomized double-blind study was carried out in patients with unresolving antibiotic-associated colitis due to Clostridium difficile, to compare the effect of bacitracin (80,000 U/day) with vancomycin (500 mg/day) on the resolution of symptoms, clearance of organism, and prevention of relapse. Forty-two patients with colitis, 9 of whom had a pseudomembrane, were randomized, 21 patients to each treatment group. The two groups were comparable in age, disease severity, and antibiotic exposure. For a 50% reduction in stool frequency the mean times (+/- SE) were 4.1 +/- 0.4 days for bacitracin and 4.2 +/- 0.4 days for vancomycin. Sixteen patients (76%) had symptom resolution after 7 days of treatment with bacitracin, compared with 18 patients (86%) given vancomycin. Patients who failed to respond were crossed over (blind) to the alternative antibiotic, but tended to be refractory to the alternative medication as well. Vancomycin-treated patients had negative toxin (83% vs. 53%, p = 0.04) and negative stool cultures (81% vs. 52%, p = 0.02) more frequently than did those patients given bacitracin. Similar numbers of patients in each group had symptomatic relapse during 1 mo of follow-up, but most of them relapsed yet again after blinded crossover therapy. Although bacitracin was significantly less effective than vancomycin in clearing C. difficile from the stools, both were of similar value in the control of symptoms in a group of patients with predominantly nonpseudomembranous colitis. In view of its low cost, bacitracin is a reasonable first-line alternative to vancomycin in the treatment of antibiotic-associated colitis.