In a 7-year period, 33 children had endoscopically documented duodenal ulcer disease. Of the 33 children, 29 had Helicobacter pylori antral gastritis. All children with H. pylori-associated duodenal ulcer disease were treated with antibiotics but no H2-receptor blocking agents. For the first 3 years of the study, initial treatment was with bismuth subsalicylate or amoxicillin for 6 weeks. For the latter 4 years, therapy with both bismuth subsalicylate and amoxicillin for 6 weeks was used initially; those in whom treatment failed received bismuth subsalicylate and amoxicillin for 6 weeks, and metronidazole for 4 weeks. Follow-up with endoscopic biopsies was performed immediately after each treatment course and at a mean of 6.5 months after clearance of H. pylori from antral biopsy specimens. Data for noncompliant patients and those receiving nonsteroidal antiinflammatory drugs were analyzed separately. For compliant patients, the rates of H. pylori clearance from antral biopsy specimens immediately after each treatment were as follows: bismuth subsalicylate, 6 of 12 children; amoxicillin, 1 of 5 children; double therapy, 9 of 9 children; and triple therapy, 3 of 3 children. At long-term follow-up, the number of children with no H. pylori in antral biopsy specimens were as follows: monotherapy, 1 of 5; double therapy, 4 of 4; and triple therapy, 3 of 3. Of the noncompliant patients, only 1 of 5 had H. pylori eradication with combination therapy and none had H. pylori eradication with monotherapy. Duodenal ulcer disease had healed in all children with H. pylori-negative antral biopsy specimens at follow-up; duodenal ulcers recurred in all children with persistent or recurrent H. pylori infection. We conclude that therapy with two drugs for 6 weeks is the treatment of choice for endoscopically proved duodenal ulcer and histologically proved H. pylori antral gastritis, and that poor patient compliance reduces the rate of success. Our data strongly support a causal relationship between H. pylori and duodenal ulcer disease.