ObjectiveSodium-glucose co-transporter 2 inhibitors (SGLT2-i) are a novel drug class for the treatment of diabetes. We aimed at describing the maximal benefits and risks associated with SGLT2-i for patients with type 2 diabetes.DesignSystematic review and meta-analysis.Data Sources and Study SelectionWe included double-blinded, randomised controlled trials (RCTs) evaluating SGLT2-i administered in the highest approved therapeutic doses (canagliflozin 300 mg/day, dapagliflozin 10 mg/day, and empagliflozin 25 mg/day) for ≥12 weeks. Comparison groups could receive placebo or oral antidiabetic drugs (OAD) including metformin, sulphonylureas (SU), or dipeptidyl peptidase 4 inhibitors (DPP-4-i). Trials were identified through electronic databases and extensive manual searches. Primary outcomes were glycated haemoglobin A1c (HbA1c) levels, serious adverse events, death, severe hypoglycaemia, ketoacidosis and CVD. Secondary outcomes were fasting plasma glucose, body weight, blood pressure, heart rate, lipids, liver function tests, creatinine and adverse events including infections. The quality of the evidence was assessed using GRADE.ResultsMeta-analysis of 34 RCTs with 9,154 patients showed that SGLT2-i reduced HbA1c compared with placebo (mean difference -0.69%, 95% confidence interval -0.75 to -0.62%). We downgraded the evidence to ‘low quality’ due to variability and evidence of publication bias (P = 0.015). Canagliflozin was associated with the largest reduction in HbA1c (-0.85%, -0.99% to -0.71%). There were no differences between SGLT2-i and placebo for serious adverse events. SGLT2-i increased the risk of urinary and genital tract infections and increased serum creatinine, and exerted beneficial effects on bodyweight, blood pressure, lipids and alanine aminotransferase (moderate to low quality evidence). Analysis of 12 RCTs found a beneficial effect of SGLT2-i on HbA1c compared with OAD (-0.20%, -0.28 to -0.13%; moderate quality evidence).ConclusionThis review includes a large number of patients with type 2 diabetes and found that SGLT2-i reduces HbA1c with a notable increased risk in non-serious adverse events. The analyses may overestimate the intervention benefit due bias.