Whether or not the incidence of ulcer relapse varies according to the drug used to produce initial healing is a controversial matter. We tackled this problem using data from 15 eligible trials from 25 published controlled trials in patients followed up for six to 12 months. Pooled estimates of differences in ulcer relapse incidence between patients initially healed with H2-antagonists and patients initially healed with non-H2-antagonist drugs were calculated. The overall incidence of relapse in patients healed with comparator drugs is 11 percentage units lower at six and 12 months, than that observed in H2-antagonist-healed patients. The confidence intervals are +/- 8% at six months and +/- 7% at 12 months. These data suggest the existence of a different effect on relapse incidence for the entire class of comparator drugs taken as a whole, compared with H2-antagonists. On considering the non-H2-antagonists singly, this conclusion holds good only in the case of tripotassium dicitrato bismuthate.