Do women lose more weight than men on GLP-1 medications?

On average, yes, in the pooled evidence. A 2026 JAMA Internal Medicine systematic review pooled six trials in 19,906 patients and found weight loss of 10.9% of baseline body weight among women (95% CI, 7.0% to 14.8%) against 6.8% among men (95% CI, 4.6% to 9.0%). The confidence intervals overlap, and this is a group average, not a prediction about any individual.

Why is the difference so much smaller when measured in kilograms?

Because percentage of baseline body weight has each person's starting weight in the denominator, and women start lighter on average. The two analyses that reported the comparison in kilograms found gaps of about one kilogram: 1.04 kg pooled across 14 studies in a 2025 Journal of Diabetes meta-analysis, and 1.1 kg adjusted in a 7,847-person real-world Italian cohort. The same meta-analysis reported the gap as 1.69 percentage points when expressed as a proportion of baseline weight.

Is the sex difference explained by estrogen?

Estrogen synergy is a proposed explanation, not an established finding. The authors of the 2025 Journal of Diabetes meta-analysis suggest that GLP-1 receptor agonists and estrogen could act synergistically on the supramammillary nucleus and could ameliorate leptin resistance, alongside two other hypotheses: pharmacokinetic differences related to lower body weight in women, and greater improvement in inflammatory markers. None of these mechanisms was measured in the pooled trials.

Does this mean women should take a different dose?

No. Nothing in this evidence supports a sex-based dose. The current US prescribing information for Wegovy, Zepbound and Ozempic sets out dosing without a sex adjustment, and each label's pharmacokinetics section states that sex has no clinically significant, relevant or meaningful effect on the drug. Dosing decisions belong to a prescribing clinician.

A woman sits at a kitchen counter in morning light, mid-conversation with someone out of frame, a mug and a bowl of fruit beside her.

Women's Health

Do women lose more weight on GLP-1s than men? What the sex-difference evidence actually shows

Q: Do GLP-1s work differently by age, race or starting BMI?

The 2026 JAMA Internal Medicine review found no significant heterogeneity of treatment effect by age (7 trials, 4,314 patients), race (9 trials, 25,229), ethnicity (7 trials, 8,328), baseline BMI (15 trials, 9,473) or baseline HbA1c (4 trials, 1,886). Sex was the only characteristic of the six where the pooled analysis found a difference.

By the US Health Digest editorial team · Published August 30, 2026 · Every claim linked to its primary source

Yes, on average — and the size of the difference depends almost entirely on whether you measure it in percentages or in kilograms. A Johns Hopkins–led systematic review published in JAMA Internal Medicine in 2026 pooled six trials in 19,906 patients and found weight loss of 10.9% (95% CI, 7.0%–14.8%) among women against 6.8% (95% CI, 4.6%–9.0%) among men — while finding no significant difference by age, race, ethnicity, baseline BMI or HbA1c. But the analyses reporting the same comparison in kilograms put the gap at roughly one kilogram: 1.04 kg pooled across trials, 1.1 kg in a 7,847-person cohort. Women start lighter, so the same kilogram is a bigger percentage. Both numbers are true; only one gets printed.

What the JAMA Internal Medicine analysis actually did

The paper is Alexander et al., "Heterogeneity of Treatment Effects of Glucagon-Like Peptide-1 Receptor Agonists for Weight Loss in Adults: A Systematic Review and Meta-Analysis," JAMA Internal Medicine 2026;186(5):567–577, from the Center for Drug Safety and Effectiveness at the Johns Hopkins Bloomberg School of Public Health. Its question was not how much weight these drugs cause people to lose — that is settled — but whether the answer changes depending on who you are.

The team searched MEDLINE, Embase and the Cochrane Central Register from inception through July 26, 2024, screened 7,705 unique records, and included 41 articles representing 64 randomised trials, using random-effects models and the Cochrane risk-of-bias tool.

One structural detail governs everything that follows. The paper states its outcome measures as "Change in body weight measured in kg (by age, baseline BMI, baseline HbA1c) or percentage change from baseline (by sex, race and ethnicity)." The age and BMI comparisons are in kilograms. The sex comparison is in percentages. So the single most-quoted figure from this paper — 10.9% versus 6.8% — is, by the paper's own design, a ratio, and a ratio has a denominator.

The percentage and the kilogram are not the same claim

Percentage of baseline body weight is the standard endpoint in obesity medicine for good reasons: it scales, and clinical benefit tracks it better than raw pounds. But it embeds each person's starting weight in the answer. Two people can lose identical amounts of tissue and post different percentages simply because one began heavier.

The cleanest demonstration is not our arithmetic; it is a second meta-analysis that reported both currencies from the same pooled trials. Yang et al., Journal of Diabetes, March 2025 (protocol CRD42023480167) pooled 14 studies of dulaglutide, exenatide, liraglutide, semaglutide and retatrutide and found that females lost more than males by 1.04 kg (95% CI, 0.70–1.38) and by 1.69 percentage points (95% CI, 0.78–2.61). Same review, same trials, two units. The authors say why in plain terms: "women typically have a lower baseline body weight and, thus, higher absolute weight reduction translates into a proportionally larger percentage of overall body weight reduction."

A worked example

This is arithmetic, not data — a demonstration of how the units behave, not a prediction about anyone. Take two people whose starting weights bracket the 104.78 kg mean baseline body weight recorded in SURMOUNT-1's posted results: a man at 110 kg and a woman at 95 kg.

Give the man an 11.0 kg loss and the woman a 12.0 kg loss — a 1.0 kg gap, close to the 1.04 kg that Yang et al. actually pooled.
He has lost 10.0% of his starting weight. She has lost 12.6%.
A one-kilogram difference has become a 2.6-percentage-point difference, purely from the denominators. Nothing about the drug changed between those two sentences.

Now run it backwards, which is what a reader does when they see the headline. Apply 10.9% to the 95 kg woman and 6.8% to the 110 kg man and you get 10.4 kg against 7.5 kg — an absolute gap of about 2.9 kg, roughly three times the ~1 kg that the two analyses reporting kilograms measured.

And here is the part that cuts against the tidy version of this story: that discrepancy is not purely a units artefact. Yang et al. also found the sex gap grows as total weight loss grows — meta-regression β = −0.19 (95% CI, −0.29 to −0.09) — and in trials where the indication was weight reduction rather than diabetes it widened to 4.21 kg (95% CI, 1.75–6.67). Their pool leans on lower-efficacy diabetes trials; the JAMA pool is a different set. So part of the divergence is denominators and part is that different trials are being averaged. Anyone telling you it is entirely one or the other has not checked.

Both bars are the same estimand from the same paper — pooled percentage change from baseline in the six sex-stratified trials, 19,906 patients — on one linear scale. The kilogram figures discussed above are not plotted, because they come from a different review of a different trial set, and putting two estimands in one picture is how a chart starts lying. The whiskers are the published 95% confidence intervals. They overlap between 7.0% and 9.0%: the point estimates differ by 4.1 percentage points, but the intervals around them are wide.

Consistent across everything except sex — the more interesting finding

Strip out the headline and the Hopkins paper is mostly a story about how little these drugs care who you are. The same review found no significant heterogeneity of treatment effect by age (7 trials, 4,314 patients), race (9 trials, 25,229), ethnicity (7 trials, 8,328), baseline BMI (15 trials, 9,473 across three analyses), or baseline HbA1c (4 trials, 1,886). That is an unusual result in medicine, and arguably better news for more people than the sex finding is.

It also lines up with the manufacturers' own labels — all three of which include sex in the list of things that did not change the result. Read the third carefully: Ozempic is licensed for type 2 diabetes, not weight loss, and the sentence quoted from it describes its glycemic efficacy, not a weight outcome.

What the current US prescribing information says about subgroup response, read on DailyMed on August 30, 2026. DailyMed carries the current label revision; these are direct quotations from each label's Clinical Studies section.
Label (current revision)	Verbatim statement
Wegovy (semaglutide) version 19, June 30, 2026	"A reduction in body weight was observed with WEGOVY irrespective of age, sex, race, ethnicity, BMI at baseline, body weight (kg) at baseline, and level of renal function impairment."
Zepbound (tirzepatide) version 38, May 6, 2026	"A reduction in body weight was observed with ZEPBOUND irrespective of age, sex, race, ethnicity, baseline BMI, and glycemic status."
Ozempic (semaglutide) version 20, June 10, 2026 statement is about HbA1c efficacy in type 2 diabetes	"The efficacy of OZEMPIC was not impacted by age, gender, race, ethnicity, BMI at baseline, body weight (kg) at baseline, diabetes duration and level of renal function impairment."

So there is a real tension, worth naming rather than smoothing over. The labels, describing what each pivotal trial programme found within itself, report no sex effect. The pooled analyses across many trials do. Yang et al. flag exactly this and quote the Ozempic label doing it. The likeliest reconciliation is statistical rather than scandalous: a single trial is powered to detect the drug's overall effect, not a difference between two subgroups inside it, and a gap of roughly one kilogram is easy for one trial to miss and easy for a pool of nineteen thousand patients to see. That is an explanation, not a proof.

Every proposed mechanism is a hypothesis

Coverage of this finding tends to slide from women lost more into here is why, as though the why had also been measured. It was not. The explanations below are the ones Yang et al. put forward in their discussion, reported here as proposals, attributed, with no endorsement from us:

Pharmacokinetics. They suggest higher plasma levels in females may be responsible, citing increased exposure "owing to the lower weight of females compared with males" and decreased clearance compared with weight-matched males. Note that this is partly circular with the percentage problem above — again a claim about body size, not about sex as such. The Ozempic label's pharmacology section supports the first half: "The exposure of semaglutide decreases with an increase in body weight" — while the same paragraph says sex itself does "not have a clinically meaningful effect on the pharmacokinetics of semaglutide."
Estrogen signalling. They propose that GLP-1 receptor agonists and estrogen "could act synergistically to activate the supramammillary nucleus, which controls food intake," modifying food-reward behaviour through pro-opiomelanocortin neurons, and that the combination "may ameliorate leptin resistance." That is a hypothesis drawn from preclinical work, not a finding from any pooled trial.
Inflammatory markers. They point to greater improvement in C-reactive protein and tumour necrosis factor-alpha among females on treatment as a further possible contributor.

One rarely-quoted result bears directly on the pharmacokinetic story. In the GLIMPLES real-world study — 7,847 adults with type 2 diabetes across 18 Italian diabetes centres, median four years of follow-up — the sex difference "remained unchanged after progressive adjustment for variables that differed between males and females, across GLP-1RA molecules, and after accounting for weight-adjusted drug doses." If dose-per-kilogram were doing the work, that adjustment should have shrunk the gap. It did not. One retrospective diabetes cohort does not settle a mechanism, but it counts against the simplest version of the exposure hypothesis.

The subgroup that got studied least is the one with the answer

Here is the finding inside the Hopkins paper that almost no one will report. Of the 48 trials the authors could individually characterise, heterogeneity of treatment effect was evaluated using baseline BMI in 36 (75.0%), HbA1c in 24 (50.0%), age in 21 (43.8%), ethnicity in 12 (25.0%), race in 11 (22.9%) — and sex in 10 (20.8%). Sex was the least commonly examined characteristic of the six, and the pooled sex estimate rests on just six trials.

The trials themselves were overwhelmingly female, which makes the thin subgroup reporting harder to excuse rather than easier. In the posted results, STEP 1 enrolled 1,453 women among 1,961 participants — 74.1%, and SURMOUNT-1, 1,714 among 2,539 — 67.5%. Those are the trials behind the numbers everyone knows: 14.9% at 68 weeks in STEP 1 and up to 20.9% at 72 weeks in SURMOUNT-1. When a trial is three-quarters women, the male estimate is the one built on the smaller, noisier sample.

A related figure circulates that we could not support and are therefore not printing: that 75–80% of GLP-1 users are women. The best nationally representative US estimate we can verify is the RAND American Life Panel survey of 8,793 adults, in which 11.8 percent of adults reported ever using a GLP-1 agonist. Its picture is more textured than one ratio: women aged 30 to 49 were "more than twice as likely to have used a GLP-1 agonist than their male peers" and women aged 50 to 64 had the highest use of any group at 20 percent — but among those 65 and older, "use of GLP-1 agonists is somewhat higher for men than for women." A real skew, yes — but not one flat ratio holding across every age band. More in GLP-1 statistics for 2026.

What the real-world cohorts add

Three observational studies point the same direction as the trials, which is reassuring, and none of them is a substitute for one.

Studies reporting a sex comparison in GLP-1 weight response. The measures are different objects — a pooled percentage, a pooled kilogram difference, an adjusted kilogram difference, a completer-cohort percentage — and must not be read against one another as a single quantity.
Study	Design, n	Measure and estimand	Result
Alexander et al., JAMA Intern Med, 2026 (PMID 41770554)	Meta-analysis of RCT subgroups; 6 trials, 19,906 patients	Pooled percentage change from baseline, random-effects	Women 10.9% (7.0–14.8) vs men 6.8% (4.6–9.0)
Yang et al., J Diabetes, 2025 (PMID 40040445)	Meta-analysis; 14 studies	Mean difference, female minus male, reported both ways	1.04 kg (0.70–1.38); 1.69 percentage points (0.78–2.61)
Marassi et al., Pharmacol Res, 2025 (PMID 40706779)	Retrospective longitudinal, 18 Italian centres; 7,847 adults with type 2 diabetes, median 4 years	Adjusted mean difference in weight change, mixed model for repeated measures	−1.1 kg favouring women (p < 0.001); ≥5% loss 66.5% vs 58.0%; ≥10% 40.0% vs 30.7%
Castaneda et al., Obesity, 2026 (PMID 42290037)	Retrospective cohort, Mayo Clinic; 1,039 adults on tirzepatide who completed ≥12 months of continuous therapy	Total body weight loss % at 15 months — a completer population by design, not intention-to-treat	Women 15.1% vs men 10.7% (p < 0.001); age not an independent predictor
Inceu et al., Medicina, 2025 (PMC12388165)	Single-centre retrospective case series, Cluj-Napoca; 114 adults with type 2 diabetes (47 women, 67 men)	Proportion reaching weight-loss thresholds at 12 months	≥5%: 51.1% vs 28.4% (p = 0.019); ≥10%: 29.8% vs 9.0% (p = 0.006)

Three scope notes. Each meta-analysis names only GLP-1 receptor agonists — the JAMA objective lists semaglutide, liraglutide, exenatide, lixisenatide and dulaglutide; Yang's lists dulaglutide, exenatide, liraglutide, semaglutide and retatrutide — so the tirzepatide evidence here is retrospective only, and the Mayo cohort's 15.1% versus 10.7% is again a percentage of baseline. The effect is also not uniform across the class: Yang et al. found significant sex differences for dulaglutide (0.88 kg; 95% CI, 0.63–1.12) and semaglutide (1.04 kg; 95% CI, 0.45–1.63) but none for exenatide. And two of the five studies — GLIMPLES and the Cluj-Napoca series — are type 2 diabetes populations, with Yang's pool leaning on diabetes trials as well; that is a setting where weight loss is smaller and the comparison behaves differently. For the trial-by-trial picture in women, see semaglutide results in women and Wegovy versus Zepbound for women.

What this changes for you, and what it does not

This is evidence reporting, not medical advice. It describes group averages from published research and cannot predict any individual's result, and nothing here is a reason to start, stop or change any medication.

It is not a dosing instruction. Nothing in these papers supports a different starting dose, titration schedule or target dose by sex. All three labels above set out dosing without a sex adjustment, and each states in its pharmacokinetics section that sex has no clinically significant, relevant or meaningful effect on the drug — those are the three labels' own three different words for it.
It is a group average, not a forecast. The intervals in the chart overlap. A 10.9% pooled mean spanning 7.0% to 14.8% describes where an estimate probably sits, not where you will land; variation between individuals dwarfs the gap between the sexes.
A bigger percentage is not automatically a better outcome if a meaningful share of it is lean tissue — the question we take up in GLP-1s and muscle loss and, for midlife specifically, in GLP-1s in perimenopause and menopause.
Nothing here speaks to tolerability. This literature is about weight outcomes; adverse events by sex are a separate evidence base, covered in GLP-1 side effects in women.

Questions for your prescriber

When you quote me an expected result, is that a percentage of my starting weight or a number of pounds? Can I have it both ways, since my starting weight sits inside any percentage?
The pooled estimate for women is 10.9%, with an interval from 7.0% to 14.8%. What would count as an adequate response for me, and at what week would you assess it?
Does any of this sex-difference evidence change my dose or titration schedule? The labels do not adjust dosing by sex — is there a reason yours would?
How will we track body composition or physical function, rather than only the number on the scale?
Much of the sex-comparison data comes from type 2 diabetes, where weight loss is smaller. Does my diabetes status change what you expect here?
Which symptoms should I report before the next appointment rather than save up for it?

Related reading: GLP-1s for women over 40 and the GLP-1 FAQ for women.

How we checked this page

We reached the Hopkins paper through its PubMed record (PMID 41770554), and every headline figure attributed to it second-hand checked out. We could not read the full text — the PubMed Central deposit, PMC12954602, returns HTTP 403 to this system and Europe PMC lists the article as "Subscription required" — so we claim nothing beyond its published record, and every mechanism above is attributed to Yang et al., whose full text is open access. One inconsistency in that abstract is worth logging: it states 48 trials could be individually characterised, then reports "51 trials were parallel (98.1%); 51 multicenter (98.1%)". Fifty-one of 48 is impossible, and 98.1% implies a denominator of 52. We use the 48 figure and the subgroup counts consistent with it.

One claim here is an absence, and it is scoped: we could not stand up the widely repeated figure that 75–80% of GLP-1 users are women, and found the reverse pattern in one age band. We do not claim no dataset supports it, only that we could not verify it — so we did not print it.

Sources

Alexander GC, Xiao X, Dilek S, et al. Heterogeneity of Treatment Effects of Glucagon-Like Peptide-1 Receptor Agonists for Weight Loss in Adults: A Systematic Review and Meta-Analysis. JAMA Internal Medicine. 2026;186(5):567-577. doi:10.1001/jamainternmed.2025.8222. PubMed 41770554. Johns Hopkins Bloomberg School of Public Health; databases searched through July 26, 2024. Full text not accessible to us; cited from the published record only.
Yang Y, He L, Han S, et al. Sex Differences in the Efficacy of Glucagon-Like Peptide-1 Receptor Agonists for Weight Reduction: A Systematic Review and Meta-Analysis. Journal of Diabetes. 2025;17(3):e70063. PubMed 40040445; open-access full text (PubMed Central). Protocol CRD42023480167.
Marassi M, Cignarella A, Russo GT, et al. Sex differences in the weight response to GLP-1RA in people with type 2 diabetes. A long-term longitudinal real-world study. Pharmacological Research. 2025;219:107866. PubMed 40706779. GLIMPLES study investigators; 7,847 participants, 18 Italian diabetes centres.
Castaneda R, Bechenati D, Rivera Gutierrez RJ, et al. Sex, Not Age, Predicts Weight Loss Outcomes With Tirzepatide: A Retrospective Analysis. Obesity (Silver Spring). 2026, online ahead of print June 14, 2026. PubMed 42290037.
Inceu GV, Crăciun AE, Ciobanu DM, et al. Real-World Sex Differences in Response to Treatment with Glucagon-like Peptide-1 Receptor Agonists: Analysis of Single-Center Outpatient Case Series. Medicina (Kaunas). 2025;61(8):1343. PubMed 40870388; open-access full text (PubMed Central).
WEGOVY (semaglutide) injection and tablet, US prescribing information, Novo Nordisk. DailyMed SPL version 19, published June 30, 2026. DailyMed label. Read August 30, 2026.
ZEPBOUND (tirzepatide) injection, US prescribing information, Eli Lilly. DailyMed SPL version 38, published May 6, 2026. DailyMed label. Read August 30, 2026.
OZEMPIC (semaglutide) injection, US prescribing information, Novo Nordisk. DailyMed SPL version 20, published June 10, 2026. DailyMed label. Read August 30, 2026.
Wilding JPH, Batterham RL, Calanna S, et al. Once-Weekly Semaglutide in Adults with Overweight or Obesity (STEP 1). New England Journal of Medicine. 2021;384(11):989-1002. PubMed 33567185.
Jastreboff AM, Aronne LJ, Ahmad NN, et al. Tirzepatide Once Weekly for the Treatment of Obesity (SURMOUNT-1). New England Journal of Medicine. 2022;387(3):205-216. PubMed 35658024.
ClinicalTrials.gov posted results, read August 30, 2026: NCT03548935 (STEP 1; 1,453 of 1,961 participants female) and NCT04184622 (SURMOUNT-1; 1,714 of 2,539 female; mean baseline body weight 104.78 kg, SD 22.12). National Library of Medicine.
Bozick R, Donofry SD, Rancaño KM. New Weight Loss Drugs: GLP-1 Agonist Use and Side Effects in the United States. Rand Health Quarterly. 2025;13(1):3. Open-access full text (PubMed Central). RAND American Life Panel, n = 8,793.

This article is for information only and is not medical advice. Prescription weight-loss medication requires evaluation by a licensed clinician. See our medical disclaimer.