GLP-1 drugs and the skin: what was measured

The dermatology literature on GLP-1 receptor agonists is large enough to review and thin enough that nearly every headline rests on a few dozen people. This page walks what was measured, and stops where the studies stop.

One drug class, two opposite directions

Persson and colleagues searched four databases for English-language, peer-reviewed human studies from 2014 to 2025, excluded animal and in vitro work, and followed PRISMA. Fifty-one studies met the criteria. Thirty-four reported adverse dermatologic effects — hypersensitivity, injection-site reactions, pruritus, urticaria, angioedema, and immune-mediated conditions including bullous pemphigoid. Seventeen reported benefit: improvement in psoriasis, fewer hidradenitis suppurativa flares, better wound healing, anti-aging potential, decreased inflammation. On hidradenitis suppurativa, their conclusion is that the role “remains uncertain”. A count of studies is not a count of patients.

Psoriasis: the one real trial

Petkovic-Dabic and colleagues randomised 31 people with obesity, type 2 diabetes and psoriasis who were already on metformin. Fifteen received semaglutide, sixteen were controls. After twelve weeks the median PASI score in the treated group had fallen from 21 (IQR 19.8) to 10 (IQR 6), p = 0.002, and the median DLQI from 14 (IQR 5) to 4 (IQR 4), p = 0.002. Serum IL-6 and C-reactive protein fell significantly, as did BMI and LDL. It was open-label, twelve weeks long, and thirty-one people is thirty-one people.

Nicolau and colleagues ran a six-month prospective cohort in 43 people with psoriasis and obesity, with no diabetes requirement, and found a mean PASI reduction of 48 %. Change in PASI correlated most strongly with the reduction in superficial fat (r = 0.89), then with the DLQI change (r = 0.55) and the Beck Depression Inventory change (r = 0.51). After adjustment for age and weight loss, changes in BMI and glycaemic markers were no longer significantly associated — the skin tracked where the fat came off, not the scale.

The National Psoriasis Foundation’s medical board wrote the tempering clause themselves in JAMA Dermatology in 2026. Relative PASI reductions across published studies run from roughly 40 % to 80 %, but most of those studies are small (7 to 48 patients), short (six months or less) and lack a control group. Their conclusion is explicitly conditional: if confirmed in larger randomized clinical trials.

Hidradenitis suppurativa: a literature made of cases

A 2026 narrative review in the International Journal of Dermatology gathered the human evidence on incretin-based therapy in hidradenitis suppurativa and reported what kind of evidence it is: predominantly case-level reports, plus a limited number of observational cohorts. Disease activity, flare frequency, pain and quality of life improved in parallel with weight loss. The review is direct that evidence for weight-independent anti-inflammatory effects “remains limited and inconclusive”.

Lal and Herringshaw collected the case reports instead: psoriasis, hidradenitis suppurativa, acanthosis nigricans and Hailey-Hailey disease, all improving after GLP-1 therapy, all in people for whom something else had already failed. That handful of cases is close to entirely the evidentiary base for the idea that this class treats skin disease.

The adverse signal, measured at scale

Two analyses of the FDA Adverse Event Reporting System put numbers on the other half. Fat and colleagues took 2018 to 2024 reports for semaglutide, liraglutide, exenatide and dulaglutide, coding the MedDRA terms rash, pruritus, urticaria, alopecia and angioedema. Cutaneous events appeared in up to 8.16 % of GLP-1 RA cases, more often in women, at a mean age of 60. Semaglutide carried the highest rate and dulaglutide the lowest. Against DPP-4 inhibitors as comparator the proportional reporting ratio was 0.27 (95 % CI 0.257–0.284) — proportionally fewer skin events than the comparison drug, not zero. Exenatide carried increased odds (OR 5.01, 95 % CI 4.69–5.35).

Patino and colleagues ran a comparable sweep across six agents. The five most common cutaneous reactions were eczematous, pruritus, drug eruptions, hyperhidrosis and alopecia. Life-threatening cutaneous events accounted for 2.17 % of all cutaneous reactions, with no significant difference between drugs. One asymmetry is worth naming: the class prescribed for type 2 diabetes reported higher rates of alopecia and hyperhidrosis than the same class prescribed for weight management, both printed as P = 0.000. A spontaneous-reporting database gives you what got written down, not a rate.

Salazar and colleagues’ 2024 review of the published cases is the place to look: dermal hypersensitivity reactions, eosinophilic panniculitis, bullous pemphigoid, morbilliform drug eruptions. A 2026 report in Dermatology and Therapy records improvement of chronic spontaneous urticaria in two cases.

The face, measured on scans

“Ozempic face” was a phrase for a while before it was a measurement. Facial plastic surgeons wrote about it in 2023, noting that prescribers seldom counsel patients on the facial impact, and that the drug may need to be stopped before general anaesthesia because of delayed gastric emptying. Sharma and colleagues queried hospital records for patients holding a GLP-1 receptor agonist prescription who also had head-and-neck CT or MRI both before and after, 2017 to 2024. Twenty patients had usable imaging; mean exposure was 321 days (SD 291) and average weight loss 11.0 kg (SD 6.9). Median decrease in total midfacial volume was 9.0 % (IQR 3–14 %). Superficial volume fell 11.0 % (IQR 5–15 %); deep volume fell 7.0 % (IQR −20 to 15 %). Weight loss correlated with superficial volume loss (rho = 0.590, p = 0.006) and not with deep volume loss (r = 0.115, p = 0.629). The line their abstract reports is 7 % of facial volume for every 10 kg.

Twenty patients, retrospective, one centre, no control group. Haykal and colleagues are candid about the gap: the timing of any intervention is unstudied and the mechanisms of skin and fat change unresolved.

What is on this page

Semaglutide and Tirzepatide are the two compounds here with human dermatology data, and tirzepatide is the thinner. El-Amawy’s 2026 review found injection-site reactions slightly more frequent across the SURPASS trials than with semaglutide, and the therapeutic literature for tirzepatide limited to case reports and small studies — no randomised trial against a skin disease score. Cedirian and colleagues reach the same conclusion for the class.

Retatrutide is the thinnest of all, and the search terms matter. Every term is tagged to title and abstract; the counts were run on 30 September and re-run on 1 October 2026. retatrutide[tiab] AND (skin[tiab] OR dermatologic[tiab] OR cutaneous[tiab]) returns three records, none of them reporting a retatrutide skin outcome: a longevity and aesthetic-medicine review, a dermatology-practice strategy review, and a pharmacovigilance data-mining analysis of dysesthesia across the ATC class. retatrutide[tiab] AND (psoriasis[tiab] OR hidradenitis[tiab] OR alopecia[tiab]) returns one, that same strategy review; pairing retatrutide with rash, pruritus, urticaria and angioedema returns zero. retatrutide[tiab] alone returns 191. A compound with 191 records and no dermatologic outcome measure among them is a different situation from one with a small adverse literature, and it is the situation most often glossed over.

One record in that set cuts against the simple version. The dysesthesia analysis (42168638) closes by noting that dysesthesias have already been observed in clinical trials of semaglutide, tirzepatide and retatrutide, and that skin burning is a distinctive form of it. No retatrutide study scores psoriasis, hidradenitis, alopecia or any other dermatologic disease measure, but “no skin measurement” is not the same claim as “no skin finding”, and only the first is what these searches support.

What the data do not show

They do not show that any of these compounds is a skin treatment. One randomised trial of 31 people, one uncontrolled cohort of 43, and a set of case reports is the entire human base for the therapeutic claim. They do not separate the drug from the weight loss: the 48 % PASI reduction tracked superficial fat loss at r = 0.89, and after adjustment change in BMI was no longer associated.

They do not show that the adverse signal is rare in the way a trial would show it: the 8.16 % and the 2.17 % are proportions of spontaneous reports, not incidence in a treated population. And they do not show that retatrutide is free of skin effects — dysesthesia has been reported in trials of it alongside semaglutide and tirzepatide. A 2025 review in the Journal of Investigative Dermatology builds the wound-healing case entirely from in vitro and animal models and says clinical trials against standard of care are needed. That is a mechanism, not an outcome.

Research use only. This page describes published clinical and analytical literature for research reference. It is not medical advice and not a suggestion for human use. Nothing we supply is for human or veterinary use.