Polycystic ovary syndrome: metformin in women not planning pregnancy

Cochrane review

The Cochrane review (Costello et al. 2007, assessed as up-to-date in 2006) included 5 RCTs (1 additional study discussed in the Cochrane review was based on data from 2 of these RCTs: it analysed only lipid levels as outcomes and is not discussed here). All women in the trial had PCOS but inclusion criteria varied between studies.

Three RCTs (Morin-Papunen et al. 2000, Morin-Papunen et al. 2003 and
Harborne et al. 2003) compared metformin monotherapy with co-cyprindiol monotherapy. The dose of metformin was 500 mg 3 times a day in 1 of these RCTs (Harborne et al. 2003) and 500 mg twice a day for 3 months followed by 1000 mg twice a day for 3 months in the other 2 RCTs (Morin-Papunen et al. 2000 and Morin-Papunen et al. 2003).

Two RCTs investigated metformin in combination therapy. One RCT (Elter et al. 2002) compared metformin 500 mg 3 times a day plus co-cyprindiol with co-cyprindiol monotherapy. One RCT (Cibula et al. 2005) compared metformin 500 mg 3 times a day plus oral contraceptive with oral contraceptive monotherapy. The oral contraceptive used in this RCT contained ethinylestradiol 35 micrograms plus norgestimate 250 micrograms (available in the UK as Cilest).

Only 1 RCT (Elter et al. 2002) conducted an ITT analysis and this study was single blind (clinical outcome assessors were blind to the treatment allocation). The other RCTs were open label, and analyses included only those women who completed the trial: withdrawal rates ranged from 6.7% to 44%. Allocation concealment was assessed by the Cochrane reviewers as clear in all but 1 RCT; in that RCT (Elter et al. 2002), the random allocation sequence was generated by computer but the Cochrane reviewers assessed allocation concealment as unclear.

Randomised controlled trials published after the Cochrane review

Cetinkalp et al. (2009)

In this study, women with PCOS (age not stated) were randomised to receive metformin 2000 mg daily or co-cyprindiol. A third group received rosiglitazone; results for this group are not reported here. Data were analysed for women who completed the 4-month study (6 women were lost to follow-up; group allocation was not stated). The allocation method was not discussed and the study appears to have been open label. No power calculation was reported.

Hirsutism (assessed using the Ferriman–Gallwey score) improved to a statistically significant extent in the metformin group (n=47, mean reduction score 2.16 points from 12.71 at baseline, p<0.05) and in the co-cyprindiol group (n=33, mean reduction 3.04 points from 15.07 at baseline, p<0.05), with no difference in the rate of improvement between groups (p=0.475). Acne (measurement scale not reported) improved across all groups from baseline, although this was not found to be statistically significant and there were no statistically significant differences between groups (p values not reported). All women in the trial who had oligomenorrhoea (less than 8 menstrual cycles annually) at baseline reported a statistically significant improvement in menstrual cycle length after 4 months of treatment (p<0.05 for each group), although there was no statistically significant difference between groups (p=0.202).

Luque-Ramirez et al. (2007)

In this study, 34 women (average age 24±6 years) with PCOS were randomised to receive metformin 850 mg twice daily (n=19) or co-cyprindiol (n=15) for 24 weeks. No women who were taking co-cyprindiol left the study, however 7 women who were taking metformin did and were lost to follow-up (2 because of adverse effects, 2 because of protocol violation and 1 because she became pregnant). Data were analysed on an intention to treat (ITT) basis with last observation carried forward: this was the baseline observation in 5 women taking metformin and none of those taking co-cyprindiol. Analysis was also undertaken on data from the participants who completed the 4 month study. Allocation concealment was unclear and the study was open label. A power calculation was conducted based on an outcome of changes in fasting insulin levels: this suggests that the study was underpowered for that outcome.

Hirsutism (assessed using the Ferriman–Gallwey score) statistically significantly improved in both groups (p<0.05, absolute data not reported). The between-group difference was not statistically significant in the analysis restricted to women who completed the 4-month study but was statistically significant in favour of co-cyprindiol (p<0.05) in the ITT analysis. Effects on acne were not assessed. After 24 weeks, menstrual regularity was restored in 50% of women receiving metformin and in all women receiving co-cyprindiol (p value not stated).

Meyer et al. (2007)

In this study, 110 women with PCOS who were also overweight (BMI greater than 27 kg/m²; mean age 31 years) were randomised to receive metformin 1000 mg twice daily, co-cyprindiol, or combination therapy with an oral contraceptive (ethinylestradiol 20 micrograms with levonorgestrel 100 micrograms) plus spironolactone 50 mg twice daily. Data were analysed for those women who completed the 6-month study (n=100). The allocation method was not discussed and the study was open label. A power calculation was reported that suggests the study was adequately powered for the primary outcome of changes in insulin resistance.

Hirsutism (assessed using the Ferriman–Gallwey score) improved in the metformin group (n=36) from a mean of 8.80 at baseline to a mean of 6.1 at 6 months (p<0.05). Improvements in the other groups were also statistically significant but between-group differences were not. Acne was not assessed as an outcome. Menstrual cycle length reduced in the metformin group from a mean of 112 days at baseline to 65 days at 6 months. Improvements in the other groups were also statistically significant but between-group differences were not.

Wu et al. (2008)

In this study, 60 women with PCOS (mean age 25 years, range 19–35 years) were randomised in equal numbers to receive co-cyprindiol only, metformin 500 mg 3 times daily, or a combination of both these treatments. Women were stratified into obese (BMI greater than 25 kg/m²) or non-obese (BMI less than 25 kg/m²) groups, and the proportion of obese and non-obese women was similar in all treatment groups. Analyses were based on the women who completed the study (n=53) and by obesity subgroup within each treatment allocation. The allocation method was not discussed. The study was single blind (the clinical outcome assessor was blind to the treatment allocation). No power calculation was reported.

Hirsutism (assessed using the Ferriman–Gallwey score) improved in the co-cyprindiol and co-cyprindiol plus metformin groups to a statistically significant extent (by a mean of 0.9–1.9 points on the Ferriman–Gallwey score from a baseline of 7.8–8.3) but not in the metformin monotherapy group (mean difference 0.2–0.4 points from a baseline of 7.6–8.1), but the between-group statistical significance was not reported. Menstrual regularity was fully restored in all women receiving co-cyprindiol or co-cyprindiol plus metformin, but in only approximately 28% of those receiving metformin (between-group statistical significance not reported).