The UK Medical Research Council ST03 trial compared perioperative epirubicin, cisplatin and capecitabine (ECX) chemotherapy with or without bevacizumab (B) in gastric and oesophagogastric junctional cancer. No difference in survival was noted between the arms of the trial. The present study reviewed the standards and performance of surgery in the context of the protocol-specified surgical criteria. Surgical and pathological clinical report forms were reviewed to determine adherence to the surgical protocols, perioperative morbidity and mortality, and final histopathological stage for all patients treated in the study. Of 1063 patients randomized, 895 (84·2 per cent) underwent resection; surgical details were available for 880 (98·3 per cent). Postoperative assessment data were available for 873 patients; complications occurred in 458 (52·5 per cent) overall, of whom 71 (8·1 per cent) developed complications deemed to be life-threatening by the responsible clinician. The most common complications were respiratory (211 patients, 24·2 per cent). The anastomotic leak rate was 118 of 873 (13·5 per cent) overall; among those who underwent oesophagogastrectomy, the rate was higher in the group receiving ECX-B (23·6 per cent versus 9·9 per cent in the ECX group). Pathological assessment data were available for 845 patients. At least 15 nodes were removed in 82·5 per cent of resections and the median lymph node harvest was 24 (i.q.r. 17-34). Twenty-five or more nodes were removed in 49·0 per cent of patients. Histopathologically, the R1 rate was 24·9 per cent (208 of 834 patients). An R1 resection was more common for proximal tumours. In the ST03 trial, the performance of surgery met the protocol-stipulated criteria. Registration number: NCT00450203 ( http://www.clinicaltrials.gov). © 2019 The Authors. BJS published by John Wiley & Sons Ltd on behalf of BJS Society Ltd.