Metformin for Type 2 diabetes mellitus

Metformin has historically been the first-line drug option for Type 2 diabetes mellitus, likely based on the UK Prospective Diabetes Study (UKPDS-34 trial), which showed benefits in mortality and myocardial infarction with metformin compared to diet control in overweight patients.1 The finding, however, was muddied by a UKPDS substudy showing metformin increased mortality when added to a sulfonylurea. Neither finding was explained or replicated.1 To better assess metformin for Type 2 diabetes, we examined the evidence from systematic reviews and randomized controlled trials (RCTs) to determine the benefits and harms of metformin in management of Type 2 diabetes compared to standard care or placebo. The primary source of this evidence-based summary is a 2012 meta-analysis of 13 studies by Boussageon et al.2 including 13,110 patients with Type 2 diabetes mellitus. Study participants had a mean age of 57 years, 47% were female, mean body mass index was 30, and the average hemoglobin A1c was 8.2%. Comparisons included metformin versus diet alone or placebo, sulfonylurea plus metformin versus sulfonylurea alone, and metformin withdrawal versus continuation. The authors found no significant reduction in any efficacy outcome (death, cardiovascular death, myocardial infarction, stroke, peripheral vascular events, and amputation), and no increase in severe hypoglycemic episodes. Adverse events were not reported. Separately, a single trial comparing metformin to rosiglitazone or glyburide showed a 16% absolute increase in gastrointestinal adverse events (number needed to harm [NNH] 6).3 A network meta-analysis published in 2023 by Shi et al.4 also found metformin did not reduce death or cardiovascular outcomes but did not report which trials contributed to this finding. The authors would not provide the information when contacted; therefore, we could not confirm or replicate their results. A Cochrane review from 2020 did not find any qualifying studies comparing metformin to placebo or standard care and reporting major clinical outcomes.5 Three other reviews of metformin were excluded because they either included observational studies6, 7 (without separate reporting of RCTs) or included trials that did not answer the clinical question of interest8 (i.e., included trials of patients without diabetes [e.g., with prediabetes] and pooled together placebo- and active-controlled trials). These findings contrast with those of UKPDS-34, which demonstrated significant reductions in mortality and myocardial infarction compared to diet alone.1 While UKPDS-34 had a low risk of bias, there was uncertainty in the results, with wide CIs for mortality (relative risk [RR] 0.6, 95% CI 0.5–0.9, absolute difference 7% over 11 years, number needed to treat [NNT] 14) and myocardial infarction (RR 0.6, 95% CI 0.4–0.9, absolute difference 6% over 11 years, NNT 16). In a separate UKPDS trial, metformin plus sulfonylurea increased mortality compared to a sulfonylurea alone (RR 2.0, 95% CI 1.0–2.5, absolute risk difference 6% over 7 years, NNH 17), with no difference in myocardial infarctions.1 Regarding safety, UKPDS-34 found no increase in severe hypoglycemia or gastrointestinal events compared to diet alone. The primary source of uncertainty in the metformin evidence seems to be imprecision: the CIs are too wide to rule out clinically important benefits or harms. For instance, in the review by Boussageon et al. the 95% CI for the RR of mortality with metformin was 0.8–1.3.2 Regarding quality of the evidence, the authors rated all trials as 3 or 4 out of 5 on the Jadad score (indicating high risk of bias in at least one domain).9 The three systematic reviews also had notable shortcomings. Boussageon et al. made the questionable decision to pool trials comparing metformin to diet with trials adding metformin to other drugs. The review by Shi et al. lacked transparency. The Cochrane review was unable to find data they considered reliable enough on clinical outcomes. This uncertainty is reflected in some guidelines. The American Diabetes Association recommends either sodium-glucose cotransporter-2 (SGLT2) or glucagon-like peptide (GLP)-1 drugs for improving heart and kidney outcomes in high-risk patients10 while metformin is described as providing "potential benefit." Similar recommendations are made by the American Association of Clinical Endocrinology,11 which recommends SGLT2 or GLP-1 drugs ahead of or without metformin.12, 13 However, metformin continues to be the most prescribed medication for Type 2 diabetes in practice.14, 15 Finally, it is important to note that while metformin may have unclear benefits, few alternatives are better. Within the review by Shi et al., nine of 11 other drugs did not reduce cardiovascular outcomes.4 Furthermore, sulfonylureas and insulin cause weight gain16 and severe hypoglycemia,4 while thiazolidinediones4 and some dipeptidyl peptidase-4 inhibitors17-19 increase heart failure. SGLT2 and GLP-1 drugs are the only ones shown in meta-analyses to reduce death and cardiovascular events, but also cause harms (SGLT2s4 increase genital infections and ketoacidosis and GLP-1s4, 13 cause gastrointestinal events). In summary, while the effects of metformin were not statistically significant, the broad CIs cannot rule out potentially meaningful clinical benefits or harms. We have therefore chosen a yellow color recommendation to reflect the uncertainty of metformin's effects. High-quality RCTs are needed to determine if common practice is benefiting people with Type 2 diabetes prescribed metformin. The authors declare no conflicts of interest.