Metformin Evidence Base Graded by GRADE: A Clinical Review

How a certainty grade is built, and what this page does instead

A GRADE rating applies to one clinical question, not to a drug in general: does this drug change this outcome in this population. High certainty means further research is unlikely to change the estimate. Low or very low certainty means the true effect may differ substantially from what the data suggest. Randomized trials start high and are downgraded for risk of bias, inconsistency, indirectness, imprecision or publication bias; observational data start low.

A drug that is well proven for one endpoint does not carry that strength to a different endpoint. This page does not assign formal ratings. The sections below compare the design, size and limitations of the evidence behind each claim.

Blood glucose control in type 2 diabetes

Metformin is an oral biguanide. The immediate-release label covers adults and children 10 years and older with type 2 diabetes, as an adjunct to diet and exercise; the extended-release label covers adults only [1][11]. The label says clinical studies have not established conclusive evidence of macrovascular risk reduction with metformin [1].

A 24-week double-blind trial randomized adults with type 2 diabetes (newly diagnosed, treated with diet and exercise only, or previously treated with oral medications) to one of three extended-release regimens or to immediate-release metformin. HbA1c fell 0.73% and 0.74% in the two groups given extended-release metformin 1,500 mg a day (once daily and twice daily), and 0.70% with immediate-release metformin 1,500 mg a day given twice daily [2]. The 2,000 mg once-daily extended-release group fell 1.06% [2]. Fewer patients stopped extended-release for nausea during the initial dosing period than stopped immediate-release [2].

The immediate-release label lists a starting dose of 500 mg twice daily or 850 mg once daily with meals, increased in steps of 500 mg weekly or 850 mg every two weeks to a maximum of 2,550 mg a day in divided doses, with doses above 2,000 mg better tolerated three times daily [1]. The extended-release label lists 500 mg once daily with the evening meal, increased by 500 mg weekly to a maximum of 2,000 mg once daily, with tablets swallowed whole and never crushed, cut or chewed [11].

UKPDS 34 and the 10-year post-trial follow-up

UKPDS 34 took 1,704 overweight adults with newly diagnosed type 2 diabetes (mean age 53) and randomized 753 of them to a conventional diet policy or intensive metformin, with median follow-up of 10.7 years [10]. Median HbA1c was 7.4% on metformin versus 8.0% on conventional therapy. Against conventional therapy, metformin reduced any diabetes-related endpoint by 32% (95% CI 13 to 47), diabetes-related death by 42% (95% CI 9 to 63) and all-cause mortality by 36% (95% CI 9 to 55) [10].

In the 10-year post-trial monitoring, between-group differences in HbA1c were lost after the first year, yet the metformin group still showed risk reductions of 21% for any diabetes-related endpoint, 33% for myocardial infarction and 27% for death from any cause [12].

Preventing diabetes in people who do not yet have it

The Diabetes Prevention Program randomized 3,234 non-diabetic adults with elevated fasting and post-load glucose (mean age 51, mean BMI 34.0, 68% women) to placebo, metformin 850 mg twice daily, or an intensive lifestyle program aiming at 7% weight loss and 150 minutes of activity a week [3]. Over a mean 2.8 years, diabetes incidence was 11.0 cases per 100 person-years on placebo, 7.8 on metformin and 4.8 on lifestyle. Metformin reduced incidence by 31% (95% CI 17 to 43) and lifestyle by 58% (48 to 66); lifestyle was significantly more effective than metformin. The number needed to treat over three years was 13.9 for metformin and 6.9 for lifestyle [3].

In the DPP Outcomes Study, 2,776 participants were followed for a mean of 15 years. Diabetes incidence was reduced by 18% with metformin (HR 0.82, 95% CI 0.72 to 0.93) and 27% with lifestyle, with the benefit shrinking over time. Cumulative incidence at 15 years was 56% with metformin, 55% with lifestyle and 62% with placebo. There was no overall difference in the aggregate microvascular outcome: 13.0% with metformin, 11.3% with lifestyle and 12.4% with placebo [9]. Preventing diabetes in people who do not have it is not a labeled use [1].

Weight

During the double-blind DPP, body weight fell 2.06% with metformin versus 0.02% with placebo, and waist circumference by 2.13 cm versus 0.79 cm [4]. Across 7 to 8 years of open-label follow-up, weight loss remained greater with metformin than with placebo (2.0% versus 0.2%) and was related to adherence; gastrointestinal symptoms were more common with metformin and declined over time [4].

A later analysis looked at the 1,066 participants who lost at least 5% of body weight in year one. Between years 6 and 15, maintained weight loss was 6.2% with metformin, 3.7% with intensive lifestyle and 2.8% with placebo. This is a post hoc analysis of non-randomized subsets, so it does not show that metformin produces more weight loss than lifestyle in general [13].

For scale: in 1,961 adults with overweight or obesity but without diabetes, once-weekly semaglutide 2.4 mg produced a mean 14.9% weight loss versus 2.4% with placebo at 68 weeks [14].

Polycystic ovary syndrome

The 2012 Cochrane review by Tang and colleagues pooled 44 trials in 3,992 women with PCOS, 38 of them of metformin and involving 3,495 women [5]. The review found no evidence that metformin improved live birth rates, either alone (pooled OR 1.80, 95% CI 0.52 to 6.16; 3 trials, 115 women) or when added to clomiphene (OR 1.16, 95% CI 0.85 to 1.56; 7 trials, 907 women). Clinical pregnancy rates were higher with metformin than placebo (OR 2.31, 95% CI 1.52 to 3.51; 8 trials, 707 women) [5]. PCOS is not a labeled indication [1].

Fatty liver disease

A review of nine trials in biopsy-proven non-alcoholic steatohepatitis included three metformin trials and one trial involving both metformin and glitazones. Metformin did not improve any pooled histological or biochemical outcome. Glitazones, by contrast, improved steatosis, hepatocyte ballooning and ALT [15]. Fatty liver disease is not a labeled use [1].

Cancer

A 2013 meta-analysis of 20 publications covering 13,008 people with cancer and type 2 diabetes found better overall survival (HR 0.66, 95% CI 0.55 to 0.79) and cancer-specific survival (HR 0.62, 95% CI 0.46 to 0.84) in metformin users than in users of other glucose-lowering drugs [6]. A comparison of that kind cannot show that metformin caused the difference.

The NCIC CTG MA.32 trial investigated metformin against placebo in early breast cancer; this report covers a planned six-month analysis of metabolic measures in nondiabetic patients, not cancer outcomes [16]. Among the first 492 patients with paired blood samples, decreases were larger with metformin in weight (3.0%), glucose (3.8%), insulin (11.1%), leptin (20.2%) and hs-CRP (6.7%) [16].

Aging and longevity

Metformin will be tested in the TAME (Targeting Aging with Metformin) trial, described as the initial step in developing more effective next-generation drugs against aging [7]. An observational comparison included 78,241 people started on metformin, 12,222 started on a sulfonylurea and 90,463 matched people without diabetes. Adjusted median survival was 15% lower in the matched non-diabetic controls (survival time ratio 0.85, 95% CI 0.81 to 0.90) and 38% lower in the sulfonylurea group (0.62, 95% CI 0.58 to 0.66) than in the metformin group [8]. An observational comparison cannot separate a drug effect from the differences between the people who receive each treatment.

Safety: lactic acidosis

Metformin carries a boxed warning for lactic acidosis. Postmarketing cases have resulted in death, hypothermia, hypotension and resistant bradyarrhythmias, with a subtle onset of malaise, myalgias, respiratory distress, somnolence and abdominal pain. The laboratory picture includes blood lactate above 5 mmol/L, anion gap acidosis, a raised lactate-to-pyruvate ratio and metformin plasma levels generally above 5 mcg/mL [11]. The extended-release label names these risk factors: renal impairment, certain drugs such as carbonic anhydrase inhibitors like topiramate, age 65 or older, radiologic contrast studies, surgery and other procedures, hypoxic states such as acute congestive heart failure, excessive alcohol intake and hepatic impairment [11]. The immediate-release label also lists mitochondrial disease as a lactic-acidosis risk factor [1]. The label says cases primarily occurred in patients with significant renal impairment, and that when lactic acidosis is suspected metformin is stopped and supportive care given in hospital, with prompt hemodialysis recommended [11].

Trial and cohort data give a different view of how often this happens. A 2010 Cochrane review pooled 347 comparative trials and cohort studies and found no cases of fatal or nonfatal lactic acidosis in 70,490 patient-years of metformin use. Using Poisson statistics, the upper limit for the true incidence was 4.3 cases per 100,000 patient-years in the metformin group, against 5.4 in the non-metformin group [17].

Safety: kidney function, contrast and B12

The label says to check eGFR before starting and at least yearly. Metformin is contraindicated below 30 mL/min/1.73 m2, initiation is not recommended at 30 to 45, benefit and risk should be reassessed if eGFR later falls below 45, and the drug is stopped if it falls below 30 [1]. The label specifies when metformin is withheld around iodinated contrast: at the time of or before the procedure in patients with eGFR 30 to 60, a history of hepatic impairment, alcoholism or heart failure, or intra-arterial contrast, with eGFR rechecked 48 hours afterward and metformin restarted only if renal function is stable [1].

On vitamin B12, the label reports a fall to subnormal levels in about 7% of patients in 29-week trials and says to measure hematologic parameters annually and vitamin B12 every 2 to 3 years [1]. In the DPP Outcomes Study, low B12 (203 pg/mL or less) was found in 4.3% of metformin users versus 2.3% of placebo users at 5 years, and 7.4% versus 5.4% at 13 years (not significant); low plus borderline-low B12 (298 pg/mL or less) was 19.1% versus 9.5% at 5 years and 20.3% versus 15.6% at 13 years. The odds of deficiency rose 13% for each year of metformin use (OR 1.13, 95% CI 1.06 to 1.20) [18]. In a separate trial of 390 patients with type 2 diabetes on insulin, metformin 850 mg three times daily for 4.3 years lowered B12 by 19% versus placebo, with absolute risk of deficiency 7.2 percentage points higher and a number needed to harm of 13.8 over 4.3 years [19].

Safety: stomach and bowel effects

In a US trial of immediate-release metformin up to 2,550 mg a day, diarrhea occurred in 53% versus 12% on placebo, nausea or vomiting in 26% versus 8%, flatulence in 12% versus 6%, and diarrhea led to discontinuation in 6% [1]. In placebo-controlled trials of the extended-release tablet (781 patients), diarrhea occurred in 10% versus 3% and nausea or vomiting in 7% versus 2%, with diarrhea leading to discontinuation in 0.6% [11]. The label notes that adverse reaction rates from different clinical trials cannot be compared directly [1].

Hypoglycemia risk rises when metformin is combined with insulin or an insulin secretagogue such as a sulfonylurea, and the label says lower doses of those drugs may be required [1]. The label also warns against excessive alcohol intake, which potentiates metformin's effect on lactate metabolism [1].