The value of endoscopic surveillance of postgastrectomy patients for the early detection of gastric-stump cancer is controversial. Using data from an Amsterdam postgastrectomy cohort of 2633 patients, we have done a retrospective analysis of the effect on mortality from gastric cancer of endoscopic surveillance in these patients. Between 1976 and 1982, 504 symptom-free patients from the Amsterdam cohort participated in an endoscopic surveillance programme. All patients were followed up until 1988. Relative to the general Dutch population, risk of death from gastric cancer was less among patients who took part in surveillance than among those who did not participate. However, differences were small and similar differences existed for risk of death from lung and colorectal cancers, suggesting the presence of selection. Our study seems to confirm that large-scale surveillance of postgastrectomy patients is not justified.