Advice

Overview for healthcare professionals

Overview for healthcare professionals

Metformin is commonly used for controlling blood glucose in people with diabetes. It reduces glucose production in the liver and improves the insulin sensitivity of other cells. Many women with polycystic ovary syndrome (PCOS) have insulin resistance so metformin has also been used to treat PCOS.

Regulatory status of metformin

Metformin is not licensed in the UK for treating PCOS and so its use for PCOS is off-label. In line with the guidance from the General Medical Council (GMC), it is the responsibility of the prescriber to determine the clinical need of the patient and the suitability of using metformin outside its authorised indications.

Metformin is licensed for the treatment of type 2 diabetes mellitus, particularly in people who are overweight, when dietary management and exercise alone do not result in adequate glycaemic control.

Evidence statements

Five small randomised controlled trials (RCTs) included in a Cochrane systematic review, and 4 RCTs published after the Cochrane review, provide the evidence for this summary. The usual comparator treatment was co-cyprindiol (ethinylestradiol 35 micrograms plus cyproterone 2 mg) alone. A small number of studies assessed metformin versus metformin plus cyprindiol. No RCTs compared metformin with placebo. Only outcomes that are directly relevant to patients (and not surrogate outcomes such as biochemical markers) are discussed in this evidence summary.

  • There was conflicting evidence regarding the effects of metformin on hirsutism compared with co-cyprindiol. Four RCTs found no statistically significant difference between metformin monotherapy and co-cyprindiol monotherapy on hirsutism. One RCT found that metformin monotherapy was statistically significantly more effective than co-cyprindiol monotherapy. One RCT found no statistically significant difference between co-cyprindiol plus metformin and co-cyprindiol monotherapy. Results from 2 other RCTs suggest that co-cyprindiol or co-cyprindiol plus metformin were superior to metformin monotherapy, but they are not clearly reported.

  • Two small studies found no statistically significant difference between metformin and co-cyprindiol on acne but the assessment methods used were not reported.

  • Metformin monotherapy was less effective at improving menstrual regularity compared with co-cyprindiol monotherapy in 2 RCTs and not statistically significantly different in 2 other RCTs. In 2 further RCTs, metformin monotherapy was observed to be less effective than co-cyprindiol monotherapy (both RCTs) or metformin plus co-cyprindiol (1 of the RCTs) but statistical analysis was not performed.

  • There was no or insufficient data in the studies included in this evidence summary from which to draw conclusions on the effectiveness of metformin for long-term outcomes such as preventing type 2 diabetes, cardiovascular events or endometrial cancer in women with PCOS.

  • Gastrointestinal adverse effects, which occasionally caused women to stop treatment, were more commonly reported with metformin than co-cyprindiol. Other adverse effects that caused women to stop treatment (weight gain, high blood pressure, depression, chest pain and headache) were more commonly reported with co-cyprindiol than metformin.

Summary of the evidence

This section gives a brief summary of the main evidence. A more thorough analysis is given in the Evidence review section.

Of the total of 9 RCTS, no RCTs compared metformin, alone or in combination, with placebo. 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. Allocation was concealed in 4 RCTs but allocation concealment was unclear or not stated in others. Only 2 RCTs conducted an intention-to-treat analysis (ITT; Elter et al. 2002 and Luque-Ramirez et al. 2007); the other RCTs analysed results only from those women who completed the trial (when the number of participants is given, this is the number of women analysed unless stated otherwise). All except 2 were open label (Elter et al. 2002 and Wu et al. 2008 were single blind). Withdrawal rates were up to 44%. The RCTs showed a significant degree of heterogeneity in terms of treatment duration, assessment methods and inclusion criteria. Only 1 trial was longer than 6 months and all were small (n=28–100).

Efficacy

Hirsutism

All RCTs which reported this outcome used the Ferriman–Gallwey score. Three RCTs (Morin-Papunen et al. 2000 and Morin-Papunen et al. 2003, combined n=35; Cetinkalp et al. 2009, n=80) found no statistically significant difference between metformin and co-cyprindiol. One study (Harborne et al. 2003, n=34) found a statistically significant reduction in hirsutism with metformin compared with co-cyprindiol, measured using the Ferriman-Gallwey score and also by patient self-assessment. A fourth study (Luque-Ramirez et al. 2007) found no statistically significant differences between metformin and co-cyprindiol in an analysis based on the 27 women who completed the study, but a statistically significant difference in favour of co-cyprindiol based on an ITT analysis with last observation carried forward (n=34); this was the baseline observation in 5 women taking metformin and none of those taking co-cyprindiol.

Elter et al. (2002) (n=40) found no statistically significant difference in hirsutism score with metformin plus co-cyprindiol compared with co-cyprindiol alone. Meyer et al. (2007) (n=100) found no statistically significant difference between metformin, co-cyprindiol, and combination therapy with an oral contraceptive plus spironolactone. Wu et al. (2008) (n=60) found that hirsutism improved in women taking co-cyprindiol monotherapy and co-cyprindiol plus metformin but not in those taking metformin monotherapy (between-group statistical significance not reported).

Although reductions from baseline were statistically significant in most of the trials, mean absolute reductions on the 36-point Ferriman–Gallwey score were 2–3 points or less.

Acne

Harborne et al. (2003) (n=34) used a patient-self-assessed acne score and found no statistically significant difference in score between metformin and co-cyprindiol. Cetinkalp et al. (2009) (n=80) also found no statistically significant difference in acne from baseline or between metformin and co-cyprindiol, but the assessment methods used were not reported.

Menstrual regularity

Two RCTs (Morin-Papunen et al. 2000 and Morin-Papunen et al. 2003) (combined n=35) found metformin to be statistically significantly less effective than co-cyprindiol in improving menstrual regularity. Cetinkalp et al. (2009) (n=80) found no statistically significant difference between metformin and co-cyprindiol in effect on menstrual regularity and Meyer et al. (2007) (n=100) found no statistically significant difference between metformin, co-cyprindiol, and combination therapy with an oral contraceptive plus spironolactone. In the study by Luque-Ramirez et al. (2007), menstrual regularity was restored in 6 of the 12 women receiving metformin and in all 15 women receiving co-cyprindiol (p value not stated). Wu et al. (2008) (n=60) found that menstrual regularity was fully restored in all women receiving co-cyprindiol monotherapy and co-cyprindiol plus metformin, but in only approximately 28% of those receiving metformin monotherapy (p value not stated).

Other outcomes

Only 1 RCT reported on the development of type 2 diabetes in women with PCOS (Morin-Papunen et al. 2000). Data was available for 18 women over a 6-month period) and found no statistically significant difference between metformin and co-cyprindiol in the rates of type 2 diabetes. No RCTs included in this evidence summary reported the effects of metformin on the risk of cardiovascular events or endometrial cancer in women with PCOS.

Safety

Metformin is associated with gastrointestinal adverse effects (nausea, vomiting and diarrhoea), which can be severe. The Cochrane review (Costello et al. 2007) found that metformin caused a significantly higher incidence of gastrointestinal adverse effects that were severe (leading to treatment discontinuation) compared with co-cyprindiol, and a significantly lower incidence of other severe adverse effects (weight gain, high blood pressure, depression, chest pain and headache). Among all 9 trials there was significant heterogeneity in the rates of treatment discontinuation, which was not always because of adverse effects.

Cost effectiveness and cost

No studies on cost effectiveness were identified. The annual cost of metformin at a dose of 1.5–2 g per day ranges from £30.03 to £83.20, depending on whether standard or modified-release tablets are prescribed (costs taken from the Drug Tariff January 2012).