Abstract Background Western patients with gastric cancer often present with incurable disease. The role of palliative surgical resection is still debatable. Non-curatively treated patients from the Dutch Gastric Cancer Trial were studied to define more accurately which patients might benefit from palliative resection. Methods In the Dutch Gastric Cancer Trial 285 (26 per cent) of the randomized patients were found to have incurable tumours at laparotomy. Four signs of incurability were noted: irresectable tumour (T+), hepatic metastasis (H+), peritoneal metastasis (P+) and distant lymph node metastasis (N4+). Patients had either an explorative laparotomy, a gastroenterostomy, or a resection (partial or total). In the analysis, particular attention was paid to the prognostic factors of age, number of metastatic features, and a combination of these. Results Overall survival time was greater if a resection was performed (8·1 versus 5·4 months; P < 0·001). For patients aged over 70 years there was still a survival advantage of about 3 months if resection was carried out. Morbidity and perioperative mortality rates in this older age group were, however, high (50 and 20 per cent respectively). For patients with one metastatic site a resection was of significant benefit (survival 10·5 versus 6·7 months; P = 0·034). For patients with two or more metastatic sites resection had no significant survival advantage (5·7 versus 4·6 months; P = 0·084). Combination of these factors indicates that patients aged less than 70 years with one metastatic site will benefit significantly from a palliative resection, in contrast to other combinations of factors. Conclusion Age as well as the number of metastatic sites should be taken into account when a palliative resection is considered. Palliative resection may be beneficial for patients under 70 years of age if the tumour load is restricted to one metastatic site.