Women's Health

GLP-1 Drugs and PCOS: What the Evidence Actually Shows

By the US Health Digest editorial team · July 30, 2026

For weight, BMI and waist, the evidence is reasonably consistent; for insulin resistance, the pooled result is negative; for fertility, it is thin. A 2026 network meta-analysis of 16 randomized trials in non-menopausal women with PCOS ranked a GLP-1 plus metformin most effective for body weight — a mean difference of −5.58 kg (95% CI −8.57 to −2.59) — with a GLP-1 alone close behind at −5.22 kg (Omarion et al., Frontiers in Endocrinology, 2026). But none of the interventions studied significantly improved HOMA-IR, the standard blood-based index of insulin resistance, and reproductive outcomes were not assessed at all. The one trial that did measure them enrolled 100 women at a single hospital in China, was open-label, ran a diabetes-range semaglutide dose for 16 weeks, and counted pregnancies only after semaglutide had been withdrawn (Chen et al., 2025). No large phase 3 trial has enrolled PCOS as its population.

Why weight-loss drugs get proposed for PCOS

PCOS is diagnosed by pattern rather than by a single test. The Rotterdam criteria used to enrol the trial described below require two of three features: irregular or absent ovulation, clinical or biochemical androgen excess, and polycystic ovarian morphology on ultrasound (Chen et al., Reproductive Biology and Endocrinology, 2025). Two of those three are reproductive; none is a weight measurement.

The case for metabolic drugs comes from the company PCOS keeps. The three agents compared in the 2026 network meta-analysis — GLP-1 receptor agonists, metformin and inositol — are all proposed on metabolic grounds. The Chen trial measured testosterone, a visceral adiposity index, C-reactive protein and menstrual cycle recovery alongside weight, which shows what researchers hope a weight-lowering drug will move: insulin resistance, androgen excess, and the ovulation those two are thought to disrupt. That is the hypothesis. Keep it separate from the results.

The pooled evidence: consistent on the scale, silent on insulin resistance

The most useful single document is a network meta-analysis published in Frontiers in Endocrinology on July 3, 2026. A conventional meta-analysis pools trials that compared the same two things; a network meta-analysis links trials sharing a common comparator so treatments never directly tested against each other can still be ranked — reach bought at the cost of relying partly on indirect comparison. This one covered 16 randomized controlled trials in non-menopausal women with PCOS, reported to PRISMA-NMA standards and graded certainty with CINeMA (Omarion et al., 2026). On body size the findings held up, and sensitivity analyses confirmed robustness for weight and BMI.

Waist circumference is the one place a GLP-1 alone outperformed the combination: only GLP-1 monotherapy reached significance, with the other three arms moving in the same direction without getting there. Where a cell reads "ranked below," the analysis gives a rank rather than a point estimate we could verify, so we declined to fill in a number — our limitation, not a null result. These are pooled mean differences across 16 trials, not results at one shared timepoint: read them as direction and rough magnitude.

On metabolism the findings did not hold up. None of the interventions showed significant effects on HOMA-IR, and the results were marked by heterogeneity — the trials disagreed with each other more than random variation alone explains. HOMA-IR is the Homeostatic Model Assessment of Insulin Resistance: a calculation from fasting glucose and insulin, used as a proxy for how hard the body is working to keep blood sugar in range. It is the number most often invoked when these drugs are promoted for PCOS, and it did not move significantly for any arm.

The authors' own caveat is the most useful sentence in the paper: "reproductive, endocrine, and patient-centered outcomes were not evaluated in this analysis and therefore no claims of overall superiority in PCOS management can be made." Citing this analysis as evidence that GLP-1s treat PCOS cites it against its own stated scope.

The one trial that measured reproductive outcomes — and its limits

Published in Reproductive Biology and Endocrinology in July 2025, this trial randomized 100 overweight or obese women with PCOS to metformin 1000 mg twice daily, or the same metformin plus semaglutide 1 mg once weekly, for 16 weeks (Chen et al., 2025; registry ChiCTR2400090908).

Mean weight change over 16 weeks was −6.09 ± 3.34 kg in the combination group and −2.25 ± 4.27 kg on metformin alone, all anthropometric comparisons at P<0.01. Those are two separate within-group means; the arithmetic gap of roughly 3.8 kg is not a modelled between-group difference with its own confidence interval, and we are not presenting it as one. The combination group also improved more in testosterone, in a Chinese visceral adiposity index and in C-reactive protein, with higher rates of menstrual cycle recovery.

Then the part that gets quoted without its conditions. From week 16 to week 40, all participants received metformin monotherapy — semaglutide was withdrawn. Over that 24-week window, natural pregnancy rates were 35% in the former combination group versus 15% (P<0.05). The signal therefore describes what happened after semaglutide stopped, in women who had lost more weight while on it. It is not evidence that taking semaglutide while trying to conceive helps, and by design cannot be.

The limitations are not footnotes:

• n=100, with 80 completing (80%). If the pregnancy rates were computed on those completers, each arm held roughly 40 women and a 20-point gap is fewer than ten pregnancies' difference. That is conditional: the abstract gives no denominator.
• Open-label. Nobody was blinded, and cycle recovery and pregnancy attempts are exactly the outcomes knowledge of assignment can influence.
• Single centre, one country — the Second Affiliated Hospital of Chongqing Medical University, China. Generalization to a US population is an assumption.
• 16 weeks of drug, short against obesity trials that read out at 68 to 72 weeks.
• The dose was 1 mg weekly, within the type 2 diabetes range. The weight-management product is titrated to 2.4 mg weekly.

Why the dose detail changes the interpretation

It cuts both ways. A 16-week result at 1 mg is not a preview of 2.4 mg, which has never been tested this way in PCOS. Equally, the headline obesity numbers do not transfer. In STEP 1, semaglutide 2.4 mg produced a mean change of −14.9% of body weight at 68 weeks versus −2.4% on placebo across 1,961 adults, 1,453 of them female, mean age 46 (Wilding et al., NEJM, 2021; NCT03548935). In STEP 2, the same dose in adults with type 2 diabetes produced −9.6% at 68 weeks versus −3.4% on placebo, across 1,210 participants (Davies et al., Lancet, 2021). Both trials report that headline as the treatment-policy estimand — the effect regardless of whether a participant stopped the drug — so the two are like for like. Same drug, dose and duration; different population; five percentage points apart. PCOS is a population these trials never characterized — see semaglutide results for women.

The pregnancy problem, stated plainly

GLP-1 medications are not for use during pregnancy, and clinicians advise stopping in advance of trying to conceive — see GLP-1s, pregnancy and fertility. That is a planning problem, not a simple instruction: the mechanism by which a GLP-1 might improve ovulatory function is weight loss, and the drug must be stopped before the pregnancy it may have made possible. The Chen trial is built around exactly that tension.

Regain after stopping is the complication. In SURMOUNT-4, participants randomized to withdraw tirzepatide regained substantially while those who continued kept losing (Aronne et al., JAMA, 2024) — a different drug in a non-PCOS population, so it says nothing about cycles. But "lose weight on the drug, then stop and conceive" has a time dimension nobody has mapped in PCOS.

Body composition, tolerability, and one dated caveat

A 2026 meta-analysis of 36 studies — three in PCOS populations — found GLP-1 receptor agonists reduced lean body mass by a mean difference of −1.51 kg (95% CI −2.00 to −1.01); across both drug classes the review pooled — GLP-1 agonists and SGLT2 inhibitors together — lean tissue accounted for 28% (95% CI 22%–34%) of total weight lost (Jobanputra et al., Diabetes/Metabolism Research and Reviews, 2026). Read it with its date attached: its search ran only to October 2022, predating the semaglutide 2.4 mg and tirzepatide obesity trials. More in GLP-1s and lean mass and GLP-1s for women over 40. On tolerability, in STEP 1 gastrointestinal adverse events led to discontinuation in 4.5% of the semaglutide group (59 participants) versus 0.8% (5) on placebo — see side effects in women.

What is not known

No large phase 3 trial of semaglutide 2.4 mg or tirzepatide has enrolled PCOS as its study population with prespecified reproductive endpoints. The pivotal obesity trials — STEP 1, and SURMOUNT-1, where tirzepatide reached up to −20.9% at 72 weeks on the treatment-regimen estimand (Jastreboff et al., NEJM, 2022) — did not enrol for PCOS. Any PCOS-specific figure derived from them is inference.

What would settle it: an adequately powered, multi-centre, double-blind trial enrolling women diagnosed by consistent criteria, at a weight-management dose, running long enough to observe cycles, with ovulation and live birth prespecified rather than counted opportunistically, and a stated protocol for the withdrawal-before-conception problem. Until then: GLP-1s reliably change body size in women with PCOS, while their effect on the syndrome's defining features remains largely unproven. See also our GLP-1 FAQ for women.

Questions to ask your prescriber

1. What is the contraception plan? If this drug may improve ovulation, an unplanned pregnancy on it is foreseeable. Ask how long before conceiving you would stop.
2. Which dose, and why that one? The PCOS reproductive data used 1 mg weekly semaglutide; the weight-management product goes to 2.4 mg.
3. Is metformin being combined, and on what grounds? The pooled analysis ranked the combination first for weight, but the margin over a GLP-1 alone was small — −5.58 kg versus −5.22 kg — with a wide interval around it.
4. How will response be measured beyond the scale? Cycle regularity, androgen levels and metabolic labs are what would show whether the PCOS is responding.
5. What are we expecting from insulin resistance, specifically, given the negative pooled HOMA-IR finding?
6. What happens at stopping — for pregnancy, cost, or side effects? Ask what the washout looks like and what regain is expected.
7. How will lean mass and nutrition be monitored, particularly if a pregnancy may follow?
8. Is this an FDA-approved product or a compounded one, and why?

None of this is a recommendation for or against treatment, and nothing here substitutes for a clinician who has your history and labs. The aim is narrower: to make the shape of the evidence visible, so that a decision separates what is established from what is inferred.