Impact of Metformin, Oral Contraceptives, and Lifestyle Modification on Polycystic Ovary Syndrome in Obese Adolescent Women: Do We Need a New Drug?

The number of adolescents with polycystic ovary syndrome (PCOS), like adults with PCOS, appears to be on the rise. The evidence for a change in prevalence rates for PCOS, like much in the treatment of PCOS, is inferred from other conditions. There is clear population-based evidence of increased adolescent obesity in the United States (1) and clinic-based or regional studies to show increased rates of type 2 diabetes in adolescents (2); therefore, likely, there is increased prevalence of PCOS in adolescents. Our treatments for adolescents with PCOS are often adopted from treatments used in adults, although the relative efficacy and the risk/benefit ratio are often unknown. Consequently, nowhere in the ever-increasing landscape of PCOS is the current treatment paradigm less evidence based than in adolescents, and nowhere is there greater promise for long-term prevention and amelioration than in this same population.

Thus it is with open arms that I welcome in this current issue of the Journal, the studies of Hoeger et al. (3) of the treatment of obese adolescents with PCOS. They are small, but guided steps up the evidence-based pyramid and models for future studies in their randomization, their placebo-controlled design, their recruitment and relatively high retention, their bold and extensive lifestyle intervention, and their detailed studies of blood, urine, and body image in this traditionally difficult to study population. The authors report on two linked trials with a total of 79 adolescents, a single-agent trial of common therapies, and a subsequent combination trial of therapies used in the single-agent trial, both lasting 24 wk. Adolescents were selected based on their presenting complaints: hyperandrogenism, oligomenorrhea, and obesity. The investigators used therapies that form the basis of care in adults based on our understanding of the pathogenesis of PCOS: oral contraceptive pills (OCP) to suppress ovarian hyperandrogenism and restore menstrual cyclicity; metformin to improve insulin sensitivity (perhaps with some weight loss), suppress ovarian hyperandrogenism, and restore menstrual cyclicity; and finally a lifestyle program of diet and exercise to lose weight, improve insulin sensitivity, suppress ovarian hyperandrogenism, and restore menstrual cyclicity. Certainly as these clinical assumptions depict, the base of the PCOS adolescent treatment pyramid is lifestyle, and this is supported from multiple diabetes prevention trials (4,5,6).

What a surprise, therefore, to see that lifestyle in the single-agent trial had no effect on body mass index or waist circumference and no effect on visceral fat as determined by single-slice computed tomography scan! Admittedly, the numbers were small (n = 8), there were difficulties with retention and compliance, and some benefit was noted: favorable and clinically significant lowering of the free androgen index as well as decreases in the diastolic blood pressure and plasminogen activator inhibitor (PAI)-1 levels. The metformin single-agent arm was also disappointing; the highlight was a trend toward increased ovulatory frequency by urinary progestin levels (75% of cycles ovulatory vs. 50% of placebo cycles) but otherwise appeared nearly identical to placebo in maintaining other aspects of the PCOS phenotype, i.e. weight, androgens, and insulin untouched. Again the small numbers may have contributed to the failure to detect any differences, but the clinical impression of metformin alone is: Why bother? That would leave OCP as the winner by default, and it did deliver predictable menstrual bleeding, normalization of the free androgen index, and significant weight loss, but it also had its drawbacks including increased C-reactive protein (CRP) levels and increased low-density lipoprotein-cholesterol levels. These biochemical results were mitigated by decreased PAI-1 levels, significant decreases in integrated insulin levels during the oral glucose tolerance test, and a decreased triglyceride/high-density lipoprotein (HDL) ratio, findings that have not consistently been associated with OCP use in adult women with PCOS.

There is an evolution to clinical trials just as there is in our clinical practice; we learn from our shortcomings. If a single treatment is not effective, modify it if possible, and certainly when single agents fail, combination therapy is often the next step. Combination therapy is often the only way to control chronic conditions that lead to cardiovascular disease such as hypertension, dyslipidemia, and diabetes. The investigators did both in the second trial. First, they significantly altered the lifestyle intervention: by using a rolling enrollment period (instead of group starts), adding behavior modification, monitoring exercise, and encouraging greater contact between participants outside the groups (especially electronically) with marked improvements in retention and compliance. For combination therapy, the investigators chose lifestyle and OCP for all, arguably the best agents from the single-treatment trial, and tested the benefits of adding metformin. They also increased the dose of metformin from 1700 to 2000 mg/d, hoping for dose-related improvements in response.

Perhaps the most significant finding from the placebo-controlled trial of metformin is the greater efficacy of the lifestyle therapy in its second rendition, at least in combination therapy. Both groups experienced significant decreases in weight, both groups had complete normalization of the free androgen index, and unlike in the first trial, which also lasted 24 wk, both groups had significant decreases in their hirsutism scores (arguably more clinically significant than biochemical changes in androgen levels). What did the addition of metformin contribute? From a clinical standpoint, precious little other than a decreased waist circumference; from a biochemical standpoint, a mixed message: compared with placebo, there were favorable changes in HDL-cholesterol levels and presumably the triglyceride/HDL-cholesterol ratio, but compared with baseline, a significant increase in CRP levels (clearly an unexpected and puzzling finding) as well as the expected increase in low-density lipoprotein-cholesterol levels most likely due to OCP. As a single agent in the first trial and as a combination agent in the second trial, metformin had no effects on integrated glucose or insulin levels during an oral glucose tolerance test. What then, if anything, is metformin doing? The two trials leave the impression that metformin is, at best, a second-tier therapy for adolescents with PCOS, and many would say that is a generous interpretation.

Is the failure of metformin in adolescents just another in a recent series of disappointments of metformin in PCOS? (7,8,9). Certainly, there are other trials in select adolescent populations where metformin appears effective (10,11), but a recent larger trial in an obese adolescent U.S. population with lifestyle and metformin (12) was consistent with the trials of Hoeger et al. (3) with no clear benefit to adding metformin. The positive take-home message from these well designed studies is that a properly constructed lifestyle therapy in combination with OCP therapy can effectively improve both the reproductive and cardiovascular risk phenotype in adolescents with PCOS. Let there be no sacred cows, including lifestyle, and from that viewpoint, these trials leave me questioning the overall effectiveness, the practicality, and the cost/benefit ratio of an intensive lifestyle intervention in the treatment of adolescents with PCOS in an outpatient setting. To play the devil’s advocate, the addition of lifestyle did little over the 24 wk of the combination trial compared with OCP alone in the first trial. A next step up the pyramid would be a head-to-head trial of OCP therapy with standard lifestyle advice (as in the single-agent trial) vs. the altered lifestyle intervention (of the combined trial). These investigators have shown us the proof of concept in an understudied and challenging population. Given the doubt that always accompanies the results of pilot trials, we can hope as these investigators hope that the experience from these trials will guide the choice of treatments, their implementation, and the sample size selection in adequately powered multicenter trials. These studies are evidence to the upper limits of numbers of adolescent subjects that an accomplished center with dedicated staff can contribute to a single trial, which is not enough to answer the big questions that lie ahead.

Further questions that are raised but not answered by these trials is how do we follow cardiovascular risk in adolescents with PCOS? The investigators chose lipid levels and CRP levels as their biochemical markers. Do favorable or unfavorable changes in lipid levels while on hormonal therapy mean anything for cardiovascular event risk? Certainly the results of the Women’s Health Initiative and tibolone trials in postmenopausal women argue that cardiovascular events are not prevented by favorable changes in lipids (13,14), nor are events prevented with even greater beneficial changes in lipids with nonhormonal drugs such as inhibitors of cholesteryl ester transfer protein (CETP) (15). CRP lowering on drug therapy may also not correlate with events, and what is one to make of the movement of inflammatory markers in trials such as these that diverge with combination therapy or within a single treatment? (Does a PAI-1 decrease trump a CRP increase?) These changes in serum markers may reflect first-pass hepatic effects of sex steroids on hepatic proteins and may not represent a heightened inflammatory state. These considerations diminish my enthusiasm for the selected markers. Should we be monitoring relative change in validated subclinical markers of atherosclerosis such as carotid intimal medial thickness or coronary artery calcification? I suspect a priori that they will likely not cross any disease threshold. What if the link with cardiovascular events is via changes in blood clotting for which we lack adequate surrogate markers? Perhaps the best marker of cardiovascular disease risk to follow based on these studies is body mass index, a marker that is also of great concern and anguish to obese adolescents with PCOS (16), whose improvement is likely to breed compliance and retention as the second trial supports, and which correlates with events in adults, especially in combination with other biometric risk markers (17).

My frustration with this condition in young women and with the results of our current therapy has awakened in me the Sturm and Drang of adolescence. Huey Lewis rings out from the past “I want a new drug, one that does what it should, one that won’t make [them] feel too bad, one that won’t make [them] feel too good.” These studies show we haven’t found it yet for obese adolescents with PCOS, and it may not, if lifestyle therapy (or dare I mention bariatric surgery) triumphs, be a drug.