The recommended strategy for management of dyspepsia is empirical treatment with an H2-blocking drug, followed by endoscopy if the symptoms do not respond or recur. We compared two strategies for the management of dyspepsia--treatment based on the results of prompt endoscopy (group 1) and empirical H2-blocker treatment with diagnostic endoscopy only in cases of therapeutic failure or symptomatic relapse within 1 year (group 2). Eligible patients had symptoms severe enough to justify empirical H2-blocker therapy. Symptoms, drug consumption, and sick-leave days were assessed through monthly diaries. Patients with non-organic dyspepsia diagnosed by endoscopy did not receive ulcer drugs. Of 414 patients randomised, 373 completed 1-year follow-up. Organic disease was found at endoscopy in 68 (33%) of 208 group-1 patients (ulcer in 45). Endoscopy was done in 136 (66%) of 206 group-2 patients. Case selection for endoscopy was not improved by the empirical treatment strategy, since the diagnostic profile was the same as in group 1 and 40% of the expected ulcer cases remained undiagnosed. After 1 year there were no differences in symptoms or quality of life measures. The empirical treatment strategy in dyspepsia was associated with higher costs, due mainly to a higher number of sick-leave days and cost of ulcer drug use. Prompt endoscopy is a cost-effective strategy in dyspeptic patients with symptoms severe enough to justify the current practice of empirical H2-blocker treatment.