A retrospective claims study of 3,094 non-diabetic men with obesity, aged 18 to 50 and matched one-to-one against men who never received the drug, found a new erectile dysfunction diagnosis or a first phosphodiesterase-5 inhibitor prescription in 1.47% of the semaglutide group against 0.32% of the controls. That is a risk ratio of 4.5 (95% CI, 2.3 to 9.0).[4] It is also an absolute difference of 1.15 percentage points. The multiple and the count describe the same men and will not support the same sentence, and almost everything written about this subject quotes one of them.
The labels record nothing at all
Nine terms — libido, sexual, erectile, impotence, orgasm, arousal, sex drive, sexual dysfunction and testosterone — return zero hits across the full current prescribing information for Wegovy, Zepbound, Ozempic, Mounjaro, Rybelsus and Saxenda.[1][2][3] The same read of the same six documents returns nausea between ten and twenty times per label, dysgeusia once or twice, dehydration five to seven times, and the word alopecia from four of the six postmarketing sections — which is the proof the read reaches the exact place a post-approval signal would appear. Saxenda’s adverse-reaction table descends as far as anxiety at 2.0% against 1.6% on placebo, a neuropsychiatric term at the very bottom of the threshold, and still uses no sexual-function word.[3]
This matters because the claims study above opens by asserting that package inserts list sexual dysfunction as a side effect.[4] They do not. The absence is the starting condition, not a footnote to it, and it is recognized in the literature: a 2026 letter in a dermatology journal carries the title “Sexual function is missing from the adverse-effect map of GLP-1 receptor agonists.”[5] Absence from an adverse-reaction table means below that table’s reporting threshold, never zero — but no threshold explains a term appearing in none of six documents while anxiety at 2.0% appears in one. The hormone question that sits behind this one is handled separately in the testosterone article, and the enrollment skew that shapes who was studied at all is in the men’s health article.
How large the evidence base actually is
The indexed crossing of semaglutide, tirzepatide or liraglutide with any sexual-function term — sexual function, sexual dysfunction, libido, the International Index of Erectile Function, the Female Sexual Function Index — runs to 21 records. Restricted to the randomized controlled trial publication type it runs to one, and that trial measured sperm morphology rather than desire.[6] On the identical drug terms, the crossing with nausea returns 604 records, with quality of life 455 (63 of them randomized trials), and with weight loss 3,629. The zero is not an artifact of a narrow query against a small corpus; the same corpus answers adjacent questions in the hundreds and thousands.
Of the 21, two are single-patient case reports — one of anorgasmia in a woman starting a GLP-1 agonist,[7] one of sexual function change in a woman on tirzepatide.[8] Two are readings of a spontaneous-report database. One is a two-page theoretical framework. The rest are reviews, letters, hypogonadism studies and animal work. There is no prospective cohort with a validated instrument, and no trial in this class has ever prespecified one.
The direction reverses on diabetes status
The claims study restricted itself deliberately to men without diabetes, excluding anyone with a prior diagnosis, an HbA1c above 6.5%, or any history of insulin or metformin, and excluding anyone with prior erectile dysfunction or prior testosterone treatment.[4] In that population the risk went up.
Run the same question in men who do have diabetes and the sign flips. A retrospective observational study of 43 obese, diabetic, hypogonadal men already receiving testosterone undecanoate and metformin added liraglutide for a second year to the subset failing their glycemic target. Those men reached the target, lost weight significantly (p < 0.01), and recorded a significant increase in IIEF-15 score. The men who had responded well to testosterone and metformin and simply continued both showed no further IIEF improvement and regained weight.[9] Forty-three patients, no randomization, no placebo arm, and three drugs moving at once — but it is the only record in this literature of an erectile-function instrument improving on a GLP-1.
Two studies, two populations, opposite directions, and the variable that separates them is not the molecule or the dose. It is whether the patient had the disease the drug was originally approved to treat. Both designs are observational, so neither establishes cause, and the honest reading is that the effect of this class on sexual function has not been isolated from the effect of the metabolic state it is treating.
One database, two readings, opposite signals
A 2025 cross-sectional analysis of the FDA Adverse Event Reporting System covering the fourth quarter of 2003 through the first quarter of 2024 pooled six molecules and three event terms — orgasmic dysfunction, erectile dysfunction and decreased libido — and identified 182 cases, most aged 40 to 60, with diabetes the indication in 43.9%. The reporting odds ratio was 0.41 (95% CI, 0.36 to 0.48). Below one. These events were reported less often for this class than for the rest of the database.[10]
A 2026 analysis of the same database narrowed to one molecule and one term and reported 49 semaglutide-associated erectile dysfunction reports at a reporting odds ratio of 1.53 (95% CI, 1.16 to 2.03) — a statistically significant signal above one.[11] Nothing was falsified between those two papers. A pooled class against a single drug, and three event terms against one, are different questions, and disproportionality measures answer whichever one is asked. The lesson is not that one paper is right. It is that a spontaneous-report database will return whatever sign the slicing produces, which is why it generates hypotheses and settles nothing.
The number that undercuts both
The same 2026 analysis ran three drug classes side by side. Semaglutide returned a reporting odds ratio of 1.53. Sitagliptin, a DPP-4 inhibitor, returned 1.52. Dapagliflozin, an SGLT-2 inhibitor, returned 1.51.[11] Three pharmacologically unrelated classes, agreeing to two decimal places. The one thing they share is the population that receives them. Within those same classes, dulaglutide, linagliptin and metformin produced no significant signal at all, which is not what a class effect looks like either.
A time-to-onset analysis in that paper found a random failure-type distribution for all three drugs, meaning reports did not cluster after initiation the way a drug-induced effect usually does.[11] The authors call their own findings preliminary and hypothesis-generating. That is the correct weight to give them.
What the surgical literature implies, and why it cuts the other way
The closest thing to an answer comes from a population that did not take these drugs. A 2026 systematic review and meta-analysis of 21 bariatric-surgery studies covering 695 men with obesity found IIEF-5 scores rising by a mean difference of 6.45 and IIEF-15 by 9.03 after surgery (p < 0.001), with BMI falling by 13.86. Erectile function improved most (SMD 0.76), and the sexual desire domain improved significantly too.[12] A 2022 meta-analysis of 16 observational bariatric-surgery studies in 953 women found total Female Sexual Function Index scores improving (p = 0.0005) and the odds of female sexual dysfunction falling by 76% (OR 0.24; 95% CI, 0.17 to 0.33).[13]
These are surgical cohorts, not drug cohorts, and none of them is randomized. But they establish the direction that large weight loss alone pushes: up, in both sexes, including desire specifically. If a GLP-1 worked on sexual function purely by removing weight, the claims study should have found erectile dysfunction falling. It found the opposite. That gap is the strongest available argument that something other than weight is involved — and it is an argument, not a measurement. One detail in the female meta-analysis sharpens it: BMI was not a significant covariate for the improvement in FSFI score (β = 0.395, p = 0.1), so even in surgery the benefit did not track how much weight came off.[13]
Mechanisms proposed, none of them tested
A 2025 letter in a sexual medicine journal set out a theoretical framework in which this class decreases sexual desire through serotonergic signaling — the same pathway by which SSRIs blunt desire, a comparison developed in the antidepressant article.[14] The anorgasmia case report proposes instead that smooth muscle vasoconstriction reduces genital blood flow, with a secondary hypothesis about hypothalamic dopamine and norepinephrine signaling.[7] Neither has been tested against the other, and neither has been tested against the obvious confounders: fatigue, nausea, reduced caloric intake and mood change, covered in the fatigue article.
The one systematic look at what patients say found the reverse yet again. A mixed-methods analysis of 5,859 threads and comments across six Reddit communities, posted between December 2019 and June 2023, reported that sexual drive and libido were described as increased by several users.[15] Self-selected social media posts are the weakest evidence on this page and are named as such, but they are also the source most of the public commentary draws on, and they point the opposite way from the claims data.
What the compounded market adds
Every figure above was generated by FDA-approved product. Most sellers listed on this site dispense compounded semaglutide or tirzepatide, which carry no FDA approval and undergo no FDA review of safety, efficacy or quality before a pharmacy ships them, and which are covered in the compounded comparison. A pharmacovigilance analysis of 81,078 GLP-1 adverse event reports, 707 involving compounded product, reported elevated reporting odds for abdominal pain (2.84), diarrhea (1.59) and nausea (1.27), and did not report any sexual-function term among its findings.[16] That is silence in a spontaneous database rather than a measured absence. The figures from the same paper that do transfer are reporting odds of 48.92 for preparation errors and 19.00 for contamination.
What can honestly be said
No randomized trial of any drug in this class has measured sexual desire or function. The labels do not record it. One claims study of non-diabetic men found erectile dysfunction risk rising 4.5-fold on an absolute difference of 1.15 points; one small diabetic cohort found erectile function scores improving; two readings of one adverse-event database returned a signal on opposite sides of one; and the surgical literature on large weight loss points firmly upward in both sexes. Anyone who reports a change in desire after starting is reporting something the evidence base cannot confirm or refute, which is a different statement from saying it is not happening. The broader tolerability picture is in what the trials recorded, and the seller-level questions are in the provider write-ups.
A change in desire arriving alongside low mood, loss of interest in everything else, or sleep disturbance is more likely to be the mood picture than the genital one and belongs with a prescriber rather than a urologist. New erectile dysfunction in a man under 50 is a cardiovascular screening prompt in its own right, independent of any drug, and the same visit should check what the metabolic numbers are doing.