Background In Japan the surgical approach to treatment of potentially curable gastric cancer, including extended lymphadenectomy, seems in retrospective surveys to give better results than the less radical procedures favoured in Western countries. There has, however, been no evidence from randomised trials that extended lymphadenectomy (D2 gastric resection) confers a survival advantage. This question was addressed in a trial involving thirty-two surgeons in Europe. Methods In a prospective randomised controlled trial, D1 resection (level 1 lymphadenectomy) was compared with D2 resection (levels 1 and 2 lymphadenectomy). Central randomisation (200 patients in each arm) followed a staging laparotomy. Findings The D2 group had greater postoperative hospital mortality (13% vs 6·5%; p=0·04 [95% Cl 9-18% for D2, 4-11% for D1] and higher overall postoperative morbidity (46% vs 28%; p<0·001); their postoperative stay was also longer. The excess postoperative morbidity and mortality in the D2 group was accounted for by distal pancreaticosplenectomy and splenectomy. In the whole group (400 patients), survival beyond three years was 30% in patients whose gastrectomy included en-bloc pancreatico-splenic resection versus 50% in the remainder. Interpretation D2 gastric resections are followed by higher morbidity and mortality than D1 resections. These disadvantages are consequent upon additional pancreatectomies and distal splenectomies, and in long-term follow-up the higher mortality when the pancreas and spleen are resected may prove to nullify any survival benefit from D2 procedures.