Polycystic ovary syndrome is the commonest cause of anovulatory infertility and is managed with numerous drugs.1 Before starting drug treatment, women should be helped to optimise their body weight to improve the chance of spontaneous ovulation and natural conception, improve the chance of successful response to ovulation induction drugs, and maximise their health in pregnancy.
Extremes of opinion on weight loss have been eloquently outlined: Balen and colleagues suggested that drug treatment should be withheld from women with a body mass index greater than 35 owing to the dangers of pregnancy in obese women,2 and Laredo emphasised the concern of stigmatising overweight women by such withholding of fertility treatment.3 Options for drug treatment may seem more straightforward, but a multicentre placebo controlled trial of metformin reported in this week's BMJ by Moll and colleagues (p 1485) challenges that view.4
First line drug treatment for women with the syndrome who are failing to conceive comprises oral ovulation induction agents. Clomifene citrate is the longest established treatment, with efficacy confirmed through meta-analysis of randomised controlled trials, and aromatase inhibitors have shown promise but their efficacy is yet to be confirmed in randomised controlled trials.5 Insulin sensitising agents such as metformin have been the main focus of many trials, although these have had much clinical heterogeneity—in dose, time of start of treatment, and duration of treatment—and the high levels of effectiveness that were reported by some early trials have not been replicated.4,6
Polycystic ovary syndrome is characterised by increased insulin resistance, which provides the rationale for using metformin to induce ovulation. Metformin is an oral hypoglycaemic agent that has been shown to reduce serum concentrations of insulin and androgens, reduce hirsutism, and improve ovulation rates.7,8
The study by Moll and colleagues shows that metformin should not be used routinely as part of first line treatment for inducing ovulation.4 This conclusion must be taken seriously, as the trial seems to have been conducted and reported in exemplary fashion. One limitation was the stipulation of ovulation as the primary outcome, given the problems of such surrogate outcomes in fertility trials of women with polycystic ovary syndrome.9 None the less, Moll and colleagues' data for similar rates of live births and ongoing pregnancies in the treatment arms are compelling.
Although data derived from meta-analysis7,8 led to widespread recommendations to use metformin in conjunction with clomifene citrate in preference to clomiphene citrate alone,10 these data have been widely misinterpreted. Four of the five randomised controlled trials which found a significant increase in pregnancy rate with combined treatment included only women with known clomifene resistance, and clomifene sensitivity was not stated in the other trial.9 Metformin was not recommended for women recently diagnosed as having the syndrome, owing to the higher incidence of side effects with metformin and no additional increase in pregnancy rates.4
The live birth of a baby is the only clinically meaningful primary end point for a couple with infertility. Although 11 randomised controlled trials of metformin in polycystic ovary syndrome have already been meta-analysed for their benefit in inducing fertility, it is remarkable that only three trials (including only 77 women) reported live birth as an end point.9 While the Peto odds ratio for ovulation, based on seven trials including 310 women, was 3.88 (95% confidence interval 2.25 to 6.69) for metformin versus placebo or no treatment, there were insufficient data to confirm a significant increase in the pregnancy rate (five trials including 127 women, with Peto odds ratio 2.76, 0.85 to 8.98), and almost no data on live births (two trials including 50 women, Peto odds ratio 1.0, 0.13 to 7.79).9
The only published clinical trial to have compared metformin with clomifene citrate as a treatment for primary induction of ovulation (in non-obese women with polycystic ovary syndrome), which claimed that metformin was beneficial, was so methodologically flawed as to render the conclusions invalid.11 The authors did not conduct an intention to treat analysis (despite their claim to the contrary) and used inappropriate denominators for all the outcomes. The raw data from this trial are interesting, but the authors would need to present an appropriate reanalysis of their results and the results would need to be replicated in other settings before metformin could be recommended as a first line treatment.
Much effort has gone into evaluating metformin in women with polycystic ovary syndrome, but sadly much of this work has fallen short of the necessary methodological standard. The robust trial of Moll and colleagues4 confirms that, at present, it is advisable to restrict metformin to second line treatment for women whose anovulatory infertility is resistant to clomifene citrate.