One hundred and forty two patients with duodenal ulcer who after a short term study had relief of pain and healed ulcers proved endoscopically were allocated at random to double blind maintenance treatment with enprostil (a synthetic dehydroprostaglandin E2) 35 micrograms or ranitidine 150 mg at bedtime for up to 12 months. Patients were monitored every third month and examined by endoscopy at three, six, and 12 months, or more often if warranted. The cumulative relapse rates in the enprostil group at three, six, and 12 months were 37% (25/67), 56% (37/66), and 62% (41/66), respectively. The corresponding rates in the ranitidine group were 8% (6/71), 19% (13/69), and 29% (20/69). These differences were highly significant and further enhanced by life table analysis adjusting for withdrawals and by an "intention to treat" analysis in which absence of proof of non-recurrence was counted as failure, more patients in the enprostil group having been withdrawn because of adverse events or recorded as non-compliant with the protocol. Enprostil 35 micrograms at bedtime cannot be recommended for preventing relapse of duodenal ulcer. Furthermore, the results challenge the clinical relevance of using so called "cytoprotection" for preventing recurrence.