Aims/hypothesisMetformin is the most-prescribed oral medication to lower blood glucose worldwide. Yet previous systematic reviews have raised doubts about its effectiveness in reducing risk of cardiovascular disease, the most costly complication of type 2 diabetes. We aimed to systematically identify and pool randomised trials reporting cardiovascular outcomes in which the effect of metformin was ‘isolated’ through comparison to diet, lifestyle or placebo.MethodsWe performed an electronic literature search of MEDLINE, EMBASE and the Cochrane Library. We also manually screened the reference lists of previous meta-analyses of trials of metformin identified through a MEDLINE search. We included randomised controlled trials of adults with type 2 diabetes comparing any dose and preparation of oral metformin with no intervention, placebo or a lifestyle intervention and reporting mortality or a cardiovascular outcome.ResultsWe included ten articles reporting 13 trials (including a total of 2079 individuals with type 2 diabetes allocated to metformin and a similar number to comparison groups) of which only four compared metformin with placebo and collected data on cardiovascular outcomes. Participants were mainly white, aged ≤65 years, overweight/obese and with poor glycaemic control. Summary estimates were based on a small number of events: 416 myocardial infarctions/ischaemic heart disease events in seven studies and 111 strokes in four studies. The UK Prospective Diabetes Study (UKPDS) contributed the majority of data to the summary estimates, with weights ranging from 52.3% for myocardial infarction to 70.5% for stroke. All outcomes, with the exception of stroke, favoured metformin, with limited heterogeneity between studies, but none achieved statistical significance. Effect sizes (Mantel–Haenszel RR) were: all-cause mortality 0.96 (95% CI 0.84, 1.09); cardiovascular death 0.97 (95% CI 0.80, 1.16); myocardial infarction 0.89 (95% CI 0.75, 1.06); stroke 1.04 (95% CI 0.73, 1.48); and peripheral vascular disease 0.81 (95% CI 0.50, 1.31).Conclusions/interpretationThere remains uncertainty about whether metformin reduces risk of cardiovascular disease among patients with type 2 diabetes, for whom it is the recommended first-line drug. Although this is mainly due to absence of evidence, it is unlikely that a definitive placebo-controlled cardiovascular endpoint trial among people with diabetes will be forthcoming. Alternative approaches to reduce the uncertainty include the use of electronic health records in long-term pragmatic evaluations, inclusion of metformin in factorial trials, publication of cardiovascular outcome data from adverse event reporting in trials of metformin and a cardiovascular endpoint trial of metformin among people without diabetes.Electronic supplementary materialThe online version of this article (doi:10.1007/s00125-017-4337-9) contains a slideset of the figures for download, which is available to authorised users.