There is exactly one place in the prescribing information for these drugs where a menstrual cycle is measured rather than mentioned, and it is the rat. The semaglutide labels report an increase in estrus-cycle length at all dose levels in female rats, with a small reduction in corpora lutea at and above 0.03 mg/kg/day. The tirzepatide labels report an increase in the number of females with prolonged diestrus and a decrease in mean corpora lutea, also at all dose levels. Both manufacturers then attribute the finding to the same thing, and it is not the drug: an adaptive response secondary to the effect on food consumption and body weight.
Nothing equivalent appears for a human anywhere in those documents. That gap — a cycle effect measured in the species where cycles were counted, and no cycle effect recorded in the species taking the injection — is the whole subject, and it is the reason the question keeps being answered with anecdote.
The label census, and the controls that make it a census
Six current structured product labels were read term by term against PubMed and DailyMed this session: Wegovy, Ozempic, Rybelsus, Zepbound, Mounjaro and Saxenda.[1][2][3] Across all six, the stems menstru-, menorrh-, amenorrh-, oligomenorrh-, uterine and vaginal return zero occurrences — in the adverse-reaction tables, in warnings and precautions, and in the postmarketing sections.
An absence is worth nothing without proof the instrument was working, so the same read was run for terms the labels do carry. It returns nausea 10 to 20 times per label and pancreatitis 16 to 21 times per label. More pointedly, it returns alopecia from the postmarketing experience section of four of the six — the precise section where a post-approval signal is recorded once one is accepted. So the read reaches the right paragraph, and what it finds there is hair loss rather than anything menstrual. The alopecia half of that comparison has its own page in the hair-loss article.
Two caveats keep this honest. The only apparent hits for ovarian are substring artifacts of the word “covariance” in the efficacy tables, not findings. And absence from an adverse-reaction table is not zero: Wegovy’s Table 3 reports reactions occurring in at least 2% of treated adults and more often than placebo, against 1,261 placebo patients.[1] A menstrual event occurring in 1% of women, or in 3% of women and 3% of placebo, would be invisible there. The trials collected these events. None of them cleared the threshold.
Where the reports do exist, the class splits in half
The adverse-event databases are louder. A 2026 disproportionality analysis of the FDA Adverse Event Reporting System covering six GLP-1 receptor agonists from the second quarter of 2022 through the second quarter of 2025 identified 759 reproductive or endocrine-related cases, 71.3% of them in women, and adjusted for age, sex, indication, reporting country and reporter type.[4]
Semaglutide carried positive signals across several menstrual terms: menstrual disorder at an adjusted reporting odds ratio of 1.91 (95% CI, 1.29 to 2.85), menstruation delayed at 2.15 (1.41 to 3.30), oligomenorrhea at 2.55 (1.17 to 5.54), polymenorrhea at 2.34 (1.31 to 4.19) and menstrual clots at 3.24 (1.09 to 9.61).[4] In the subgroup restricted to women aged 15 to 49 the same terms strengthened, with menstrual disorder at 1.97 (1.14 to 3.42) and oligomenorrhea at 4.78 (2.19 to 10.46).
Tirzepatide in the same analysis, the same database and the same quarters ran the other way. In reproductive-age women its adjusted reporting odds ratio was 0.25 (0.10 to 0.61) for menstrual disorder, 0.25 (0.13 to 0.47) for amenorrhea, 0.19 (0.11 to 0.35) for dysmenorrhea, 0.38 (0.24 to 0.61) for irregular menstruation and 0.09 (0.03 to 0.27) for abnormal uterine bleeding — every one of them significantly below the background rate of the rest of the database.[4]
Set that against the animal sections quoted above, where both molecules lengthen the cycle at every dose tested, and the split stops looking biological. The paper’s own authors say as much, describing the tirzepatide result as a distinct reporting pattern rather than a measure of comparative biological effect. Differences between the two molecules that are measured rather than reported are set out in the head-to-head comparison.
A second analysis of the same drug reaches the opposite conclusion
A separate 2026 pharmacovigilance study examined tirzepatide alone in FAERS from the first quarter of 2022 through the first quarter of 2025 — overlapping quarters, the same database — across 67,305 cases, 75.8% of them female, and 137,583 adverse events. It reported 144 statistically significant signals and named menstrual disorder and postmenopausal hemorrhage among the new ones, with a median time to onset of 6.36 days.[5]
One drug, one database, overlapping windows, opposite findings on the same event term. Neither study is wrong in its arithmetic; they chose different comparators and different adjustment strategies, and a reporting odds ratio is a statement about the shape of a database rather than about a rate in a population. Spontaneous reports have no denominator, so none of the numbers above is an incidence, and the distance between the two results is a fair measure of how much weight any of them can carry.
The scale problem finishes the point. The same analysis validated its method with positive controls — levonorgestrel and etonogestrel, drugs that genuinely and predictably alter bleeding — which returned a reporting odds ratio of 45.76 (44.01 to 47.59) for these outcomes. Semaglutide, pooled across the same terms, returned 1.24 (1.06 to 1.43).[4] The instrument can detect a forty-five-fold signal. What it detects here is a twenty-four-percent one.
The confounder is larger than the signal, and points the other way
Substantial weight loss changes menstrual cycles on its own, and the effect has been measured without any drug involved. A retrospective study of 430 women with obesity and without PCOS followed after bariatric surgery found the proportion with irregular cycles fell from 53.5% before surgery to 22.6% at twelve months.[6] A cross-sectional study of 515 premenopausal women after bariatric surgery found irregular menstruation reported by 38.6% before and 25.0% after, a relative risk of 0.65 (95% CI, 0.53 to 0.79), with no significant change in the mean number of periods per year.[7]
Two things follow. The first is that a thirteen- to thirty-one-point swing from weight loss alone dwarfs anything in the reporting data, which makes drug attribution in an individual unresolvable — the exposure and the confounder arrive together, by design, in the same months. The comparison between those two routes to the same weight loss is in the bariatric surgery article.
The second is direction. The best-measured effect of losing this much weight is cycles becoming more regular, not less — which is the opposite of what an adverse-event frame implies, and which is also why a returning period on one of these drugs is not necessarily a problem to be treated. Where that restored ovulation is the intended outcome, the trial evidence belongs to the PCOS article; where it arrives unplanned, the consequence is a contraception question rather than a menstrual one, and that sits in the pregnancy and contraception article.
The one cycle question that has been studied properly runs backwards
There is a well-powered experimental literature on GLP-1 drugs and the cycle. It asks the reverse question. A 2026 study in female rats on a high-fat diet timed liraglutide and semaglutide to specific estrous phases and found that dosing during proestrus and estrus enhanced the intake-suppressing effect relative to dosing during metestrus and diestrus, with semaglutide given only in proestrus and estrus producing greater body-weight loss; brain GLP-1 receptor expression varied by phase across multiple nuclei.[8]
So the cycle appears to modulate how well the drug works, and whether the drug modulates the cycle has not been tested at comparable rigor in anyone. That asymmetry is the honest state of the field, and it is a rodent result that has not been replicated in women. Whether outcomes differ by sex at all is covered in the article on sex differences.
What the census does not cover
Every label figure above describes an FDA-approved product. Most sellers reviewed on this site dispense compounded semaglutide or tirzepatide, which is not FDA-approved and is not reviewed by the FDA for safety, efficacy or quality before it is dispensed, and which therefore carries no adverse-reaction table of its own — there is no compounded equivalent of the six documents read above. Spontaneous reports rarely identify a product as compounded either, so the reporting data has no clean compounded stratum to compare. What a vial actually contains is covered in the compounded vial article, and the sellers themselves are collected on the compounded semaglutide board.
What this leaves
A woman whose cycle changes in the first months on one of these drugs is experiencing something real and something common. What the published record cannot tell her is whether the injection caused it. The labels record no human menstrual reaction above a 2% threshold across six products. The adverse-event databases produce a signal an order of magnitude smaller than a known bleeding-altering drug, split it in opposite directions across two molecules whose animal data behave identically, and then disagree with themselves about one of those molecules. Meanwhile the weight loss itself moves cycle regularity by thirty points, in the favorable direction, in populations that never took the drug.
That is thin evidence, and the thinness is the finding rather than a gap to be filled with mechanism. The practical consequences are narrow and do not depend on resolving any of it: a returning or changing cycle is worth raising with a prescriber, bleeding that is heavy, prolonged or postmenopausal is worth raising urgently regardless of cause, and a period that restarts after years of irregularity is a contraception conversation before it is anything else. The rest of the tolerability picture is in what the trials recorded.