Background: The choice between coronary artery bypass surgery (CABG) and percutaneous coronary intervention (PCI) for revascularisation in patients with diabetes and multivessel coronary artery disease, who account for 25% of revascularisation procedures, is much debated. We aimed to assess whether all-cause mortality differed between patients with diabetes who had CABG or PCI by doing a systematic review and meta-analysis of randomised controlled trials (RCTs) comparing CABG with PCI in the modern stent era. Methods: We searched Medline, Embase, and the Cochrane Central Register of Controlled Trials from Jan 1, 1980, to March 12, 2013, for studies reported in English. Eligible studies were those in which investigators enrolled adult patients with diabetes and multivessel coronary artery disease, randomised them to CABG (with arterial conduits in at least 80% of participants) or PCI (with stents in at least 80% of participants), and reported outcomes separately in patients with diabetes, with a minimum of 12 months of follow-up. We used random-effects models to calculate risk ratios (RR) and 95% CIs for pooled data. We assessed heterogeneity using I2. The primary outcome was all-cause mortality in patients with diabetes who had CABG compared with those who had PCI at 5-year (or longest) follow-up. Findings: The initial search strategy identified 3414 citations, of which eight trials were eligible. These eight trials included 7468 participants, of whom 3612 had diabetes. Four of the RCTs used bare metal stents (BMS; ERACI II, ARTS, SoS, MASS II) and four used drug-eluting stents (DES; FREEDOM, SYNTAX, VA CARDS, CARDia). At mean or median 5-year (or longest) follow-up, individuals with diabetes allocated to CABG had lower all-cause mortality than did those allocated to PCI (RR 0·67, 95% CI 0·52-0·86; p=0·002; I2=25%; 3131 patients, eight trials). Treatment effects in individuals without diabetes showed no mortality benefit (1·03, 0·77-1·37; p=0·78; I2=46%; 3790 patients, five trials; pinteraction=0.03). We identified no differences in outcome whether PCI was done with BMS or DES. When present, we identified no clear causes of heterogeneity. Interpretation: In the modern era of stenting and optimum medical therapy, revascularisation of patients with diabetes and multivessel disease by CABG decreases long-term mortality by about a third compared with PCI using either BMS or DES. CABG should be strongly considered for these patients. Funding: Canadian Institutes of Health Research, Heart and Stroke Foundation of Canada, and Canada Research Chairs programme.