OBJECTIVE To compare the efficacy of cimetidine and tripotassium dicitrato bismuthate (TDB) in arthritic patients who had developed gastric (GU) or duodenal (DU) ulceration while taking non-steroidal anti-inflammatory drugs (NSAIDs). METHODS Eighty-six rheumatoid arthritis (RA) patients affected by endoscopically proven DU (n = 44) or GU (n = 42), and on chronic NSAID therapy which was not suspended during anti-ulcer therapy, were randomized to cimetidine (400 mg t.d.s.) or TDB (120 mg q.d.s.). A repeat endoscopy was planned after 4 weeks (and 8 weeks, in case of failed healing). The patients who were unhealed after 8 weeks of therapy were allocated to the alternative anti-ulcer drug for a further 8 weeks without interrupting the anti-inflammatory therapy. RESULTS At week 4 of therapy. 14/24 (58%) DU and 9/20 (45%) GU patients treated with cimetidine were healed, compared with 12/20 (60%) and 10/22 (45%) TDB-treated patients (N.S.). At week 8 of therapy, the DU healing rates were 15/24 (63%) with cimetidine and 14/20 (70%) for TDB. The corresponding GU healing rates were 12/20 (60%) with cimetidine and 13/22 (60%) for TDB (N.S.). At week 16, complete healing with cimetidine was observed in 67% of DU and 57% of GU patients unhealed with TDB; the corresponding figures in the patients crossed to TDB were 83% for DU and 63% for GU patients (N.S. vs. cimetidine). CONCLUSIONS No statistically significant difference was found between the healing activities of cimetidine and TDB in rheumatoid arthritis patients with peptic ulcer who did not interrupt their NSAID treatment for arthritis. This trial showed that the continued consumption of NSAIDs appears to slow the ulcer healing process, especially in GU patients.