EVIDENCE-BASED ANSWER:
Patients at highest risk for progression to diabetes benefit from metformin.
In patients with metabolic syndrome who are in the highest-risk quartile for progression to diabetes (predicted mean 3-year risk, 60%), metformin, 850 mg twice daily, reduces the absolute risk by about 20% over a 3-year period. Metformin doesn’t reduce the incidence in patients at lower risk of progression (strength of recommendation [SOR]: C, post-hoc analysis of a randomized controlled trial [RCT]).
Intensive lifestyle modification reduces absolute risk in all patients proportionate to risk quartile (from 5% reduction for the lowest quartile to 28% for the highest). Over a 10-year period, intensive lifestyle modification reduces the absolute risk of diabetes by 34% and metformin reduces the risk by 18% for all patients at increased risk (considered as a group)—that is, not separated by risk quartile (SOR: A, large prospective RCTs).
Lower doses or shorter courses of metformin reduce fasting plasma glucose (SOR: C, RCTs with laboratory outcomes) and may reduce the risk of developing diabetes by a smaller amount (SOR: C, flawed RCT).
EVIDENCE SUMMARY
A post-hoc analysis of a prospective RCT (the Diabetes Prevention Program) comprising 3081 patients with impaired glucose metabolism who received metformin, a lifestyle modification program, or no intervention (placebo) found that metformin reduced the risk of developing diabetes only for patients in the highest risk quartile over 2.8 years. Lifestyle modification reduced diabetes risk in all patients.1
Investigators stratified patients who met National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III) criteria for metabolic syndrome into risk quartiles for progression to diabetes using a model they developed based on 7 parameters: fasting plasma glucose, hemoglobin A1c, history of high blood glucose, waist:hip ratio, waist circumference, triglycerides, and height (TABLE1). The model reasonably fit outcomes—the percentage of patients in each quartile who developed diabetes—with an area under the receiver operator characteristic curve of 0.73 (a measure of diagnostic accuracy where 1 is a perfect predictor and 0.5 is random).
Lifestyle modification reduced risk in all quartiles with progressively greater effect as risk increased (lowest risk quartile: ARR=4.9%, 3-year NNT=20.4; highest risk quartile: ARR=28.3%; 3-year NNT=3.5).
There were 2 key weaknesses of the risk model: It wasn’t validated in a separate population and the true incidence of diabetes among patients taking placebo was higher than predicted. The investigators compared their risk prediction model results with the Framingham Risk Score (FRS) for diabetes and found that they correlated well, although the FRS results were consistently about 6% (absolute) higher when corrected for duration. (The FRS calculator is available online at www.framinghamheartstudy.org/risk-functions/diabetes/.)
Lifestyle change reduces diabetes risk more than metformin
The original Diabetes Prevention Program found that intensive lifestyle intervention and metformin reduced the number of diabetes cases over 2.8 years among 3234 patients at risk for developing diabetes.2
Compared with no intervention, fewer patients developed diabetes with either metformin or lifestyle improvement, although lifestyle change had the larger effect (no intervention: 11 cases per 100 person-years; metformin: 7.8 cases; 95% confidence interval [CI], 6.8-8.8; ARR=3.2% per year vs no intervention; lifestyle improvements: 4.8 cases; 95% CI, 4.1-5.7; ARR=6.2% per year vs no intervention).