Metronidazole has been used to treat pouchitis, but there are no controlled data that show it is effective. Chronic unremitting pouchitis is a form of the disorder particularly difficult to manage. Diarrhea is the main symptom of pouchitis, which results from acute inflammation of the mucosa of an ileal reservoir. To test the hypothesis that metronidazole (400 mg thrice daily for seven days) is no better than placebo at reducing stool frequency in chronic unremitting pouchitis, a double-blind placebo-controlled crossover study has been performed. Thirteen patients who had undergone restorative proctocolectomy for ulcerative colitis were studied. The diagnosis of pouchitis was based on clinical, endoscopic, and histological criteria. At entry all patients had symptomatic pouchitis and were passing more than six stools/24 hr or had consistently bloody stools with at least four of six endoscopic criteria of mucosal inflammation. The median frequency of defecation decreased by 3 bowel actions/24 hr (conservative 95% confidence intervals 0-4/24 hr) on metronidazole but increased by a median of 1/24 hr on placebo. The difference between the median number of bowel motions, when treatment with metronidazole was compared to placebo, was 4 motions/24 hr (P < 0.05) in favor of metronidazole. There was no significant change in the endoscopic or histological grade of inflammation, in the serum C-reactive protein level, or symptomatic scores. In a parallel study, metronidazole did not alter stool frequency in asymptomatic patients without pouchitis who had endoscopically normal reservoirs (six polyposis, six colitis).(ABSTRACT TRUNCATED AT 250 WORDS)