This randomised, double-blind, placebo-controlled study was designed to evaluate the ability of the orally administered matrix metalloproteinase inhibitor, marimastat, to prolong survival in patients with non-resectable gastric and gastro-oesophageal adenocarcinoma. Three hundred and sixty-nine patients with histological proof of adenocarcinoma, who had received no more than a single regimen of 5-fluorouracil-based chemotherapy, were randomised to receive either marimastat (10 mg bd) or placebo. Patients were treated for as long as was tolerable. The primary endpoint was overall survival with secondary endpoints of time to disease progression and quality of life. At the point of protocol-defined study completion (85% mortality in the placebo arm) there was a modest difference in survival in the intention-to-treat population in favour of marimastat (P= 0.07 log-rank test, hazard ratio= 1.23 (95% confidence interval 0.98–1.55)). This survival benefit was maintained over a further 2 years of follow-up (P= 0.024, hazard ratio= 1.27 (1.03–1.57)). The median survival was 138 days for placebo and 160 days for marimastat, with 2-year survival of 3% and 9% respectively. A significant survival benefit was identified at study completion in the pre-defined sub-group of 123 patients who had received prior chemotherapy (P= 0.045, hazard ratio= 1.53 (1.00–2.34)). This benefit increased with 2 years additional follow-up (P= 0.006, hazard ratio= 1.68 (1.16–2.44)), with 2-year survival of 5% and 18% respectively. Progression-free survival was also significantly longer for patients receiving marimastat compared to placebo (P= 0.009, hazard ratio= 1.32 (1.07–1.63)). Marimastat treatment was associated with the development of musculoskeletal pain and inflammation. Events of anaemia, abdominal pain, jaundice and weight loss were more common in the placebo arm. This is one of the first demonstrations of a therapeutic benefit for a matrix metalloproteinase inhibitor in cancer patients. The greatest benefit was observed in patients who had previously received chemotherapy. A further randomised study of marimastat in these patients is warranted.