Evidence strengths and limitations

Advice

None of the RCTs considered in this evidence summary compared metformin, alone or in combination, with placebo; most compared it with co-cyprindiol. None of the RCTs was double blind. Three of the RCTs reported significant differences between baseline characteristics of the treatment and control groups despite randomisation and concealment (Cibula et al. 2005, Morin-Papunen et al. 2000 and Morin-Papunen et al. 2003). Allocation was concealed in 4 RCTs but allocation concealment was unclear or not stated in others. The studies showed a significant degree of heterogeneity in terms of treatment duration, assessment methods and inclusion criteria. The trials were predominantly of short duration, with only 1 trial (Harborne et al. 2003) longer than 6 months, and all were small.

Although improvement in hirsutism from baseline was statistically significant in most of the trials, mean absolute reductions in the 36-point Ferriman–Gallwey scores were 2–3 points or less, so these changes may have limited clinical significance for women with PCOS. Effects on acne were reported in only 2 RCTs (Harborne et al. 2003 and Cetinkalp et al. 2009) and the method of reporting makes drawing conclusions difficult. Only 1 of the RCTs assessed the effect of metformin on development of type 2 diabetes, but this was in only a small number of women over 6 months; none of the RCTs assessed cardiovascular events or endometrial cancer events. Long-term comparison effects on these clinical outcomes as well as hirsutism and acne are lacking.