Nine RCTs were included in the review (containing 436 participants at the endpoints of the trials). Eight RCTs were double-blind. Overall risks of bias were rated as low for four trials, and moderate for five trials.
From baseline to endpoint, a statistically significant reduction in BMI was observed for the metformin groups compared with the control groups (WMD -1.42kg/m2, 95% CI -2.18 to -0.66; eight trials; Ι²=65%). Removal of the trial whose baseline BMI corresponded with an overweight status, resulted in a reduction of heterogeneity (Ι²=47%), and a slight decrease in the magnitude of the treatment effect (WMD -1.15kg/m2, 95% CI -1.84 to -0.46; seven trials; Ι²=47%). When the four trials that reported BMI z-scores were meta-analysed, a greater decrease in the magnitude of the effect was observed, although no heterogeneity was shown (WMD -0.091.15kg/m2, 95% CI -0.14 to -0.04; Ι²=0%).
Statistically non-significant trends showed that rates of gastrointestinal adverse events and adverse events overall were slightly higher in the metformin groups compared with the control groups.
The metformin groups demonstrated statistically significant reductions in fasting insulin levels (WMD -9.9μU/ml, 95% CI -13.8 to -6.06; six trials; Ι²=42%) and the Homeostatic Model Assessment index (WMD -1.78, 95% CI -3.32 to -0.23; seven trials; Ι²=90.3%), compared with the control groups.
Further results were reported in the review paper.