Skin and Injection Site Reactions on GLP-1 Medications: How Common, Which Drug, and What to Do About It

A review by Elgindi and Petronic-Rosic published in Clinics in Dermatology in September 2026 extracted adverse event reports for tirzepatide, semaglutide, liraglutide, exenatide, dulaglutide and lixisenatide from the FDA Adverse Event Reporting System. Among 442,567 reports, 137,412, or 31.0 percent, involved skin and injection site reactions, making it the third most frequently reported adverse event category.

The proportions differed sharply by agent. Exenatide led at 53.1 percent of its reports, followed by dulaglutide at 33.5 percent, tirzepatide at 32.6 percent, liraglutide at 17.1 percent, semaglutide at 12.2 percent, and lixisenatide at 6.9 percent. Most reactions were mild to moderate and were managed symptomatically without stopping treatment.

Adherence to injectable therapy is fragile, and a burning, itchy, or lumpy injection site is the kind of problem that leads people to skip doses rather than call the clinic. Naming these reactions in advance, and explaining that most settle, changes whether someone stays on a medication that is working for them.

The clinical task is triage. Local pain, redness, bruising, and transient nodules are expected and benign. Spreading urticaria, swelling beyond the injection site, wheeze, or throat tightness are a different problem entirely and need urgent attention rather than a change of injection site.

The 2026 review in Clinics in Dermatology combined two kinds of evidence. It reviewed clinical trials, pharmacovigilance studies, and published case reports and series describing non-immunologic skin and injection site reactions with GLP-1 receptor agonists. It also extracted adverse event reports for six agents from the FDA Adverse Event Reporting System, then combined injection site events with skin-related events into a single adjusted category so that agents could be compared on the same basis.

The reactions described are the ones patients actually describe: pain, bleeding, erythema, bruising, a palpable mass, itching, and swelling. The review also catalogued less common findings including dysesthesias, nodules, granulomatous reactions, and hyperhidrosis.

The authors’ summary is measured. Non-immunologic cutaneous adverse events make up a substantial proportion of reported GLP-1 adverse events, rates differ among individual agents, most reactions are mild to moderate and manageable symptomatically, and prospective studies are needed to define actual risk rates, mechanisms, and optimal management. That last sentence is the important one for interpretation.

This is where most coverage of pharmacovigilance data goes wrong, so it is worth being precise. The FDA Adverse Event Reporting System is a voluntary spontaneous reporting database. It collects reports; it does not know how many people took the drug. There is no denominator.

That means 31.0 percent describes the share of submitted reports that mentioned a skin or injection site event, not the share of patients who experienced one. If a drug produces a very large number of gastrointestinal reports, its skin share falls automatically, whether or not its skin reaction rate differs at all.

The same caution applies to the per-agent figures. Exenatide’s 53.1 percent and semaglutide’s 12.2 percent are shares of each drug’s own report volume. Semaglutide has been prescribed to an enormous population under intense public attention, which generates reports across every category and dilutes any single one. Exenatide is an older agent with far fewer total reports, and one of its formulations has a specific and well-documented local reaction pattern.

Spontaneous reporting is also shaped by publicity, litigation, regulatory notices, and how long a product has been marketed. These databases are excellent at detecting signals worth investigating and poor at estimating how often something happens.

A meta-analysis by Taj and colleagues published in Diabetes, Obesity and Metabolism in December 2025 addressed the denominator problem directly by pooling randomized controlled trials. Fourteen trials reporting injection site reactions, covering 4,861 patients and 396 injection site events, showed an increased risk with GLP-1 receptor agonists compared with comparators, with a risk ratio of 3.55, a 95 percent confidence interval of 2.35 to 5.36, and moderate heterogeneity at 41.4 percent.

The second finding in that paper is as useful as the first. Generalized dermatologic events, as opposed to reactions at the injection site itself, were infrequent and not significantly elevated. The signal is local, not systemic.

Their disproportionality analysis of FDA data pointed at the same two agents the 2026 review highlighted. Exenatide showed a strong signal for injection site haemorrhage, with a proportional reporting ratio of 27.6 and an information component of 4.6, and dulaglutide showed a signal for the same event with a proportional reporting ratio of 11.5.

Read together, the two papers tell a coherent story. Injection site reactions are genuinely more common on these drugs than on comparators, by a meaningful multiple. Broader skin disease is not clearly increased. And the reporting share differences between agents are driven substantially by two older products.

Formulation explains a good deal of this. The extended-release formulation of exenatide delivers the drug in microspheres made of poly(lactic-co-glycolic acid), a biodegradable polymer, injected weekly. The polymer persists in the tissue while it degrades, and tissue responds to persistent foreign material the way tissue does.

The result is a recognized pattern of pruritus, transient subcutaneous nodules, and foreign body reaction at injection sites. A case published in the Journal of Cutaneous Pathology in 2022 documented the mechanism unusually well: a 63-year-old woman with type 2 diabetes developed recurrent painful abdominal nodules at exenatide injection sites over a year. Punch biopsy showed septal panniculitis containing amorphous material with a mixed inflammatory infiltrate, stains for organisms were negative, and infrared spectroscopy of the biopsy tissue matched the spectral signature of the polymer itself.

That is a nodule with an identified cause, not an allergy. It is also the sort of finding that explains why one agent’s reporting profile looks so different from another’s, and why a patient with recurring lumps on a depot formulation needs a different conversation from a patient with a transient red patch on a solution formulation.

For patients, the practical translation is that a lump at an old injection site is usually a local tissue response to the product, not a sign of infection or of an allergy developing.

Two other categories come up often enough to address directly. The first is hair loss. A 2024 pharmacovigilance study in Cardiovascular Drugs and Therapy by Nakhla and colleagues queried four international adverse event databases and found that GLP-1 receptor agonists were the most reported drug class for hair loss among the diabetes agents examined, with semaglutide, liraglutide, and dulaglutide leading. Crucially, none of these drug-event pairs met the study’s threshold for a positive disproportionality signal.

That distinction is worth holding onto. Being reported most often is not the same as showing a statistical signal above background. Rapid weight loss itself is a recognized trigger for telogen effluvium, a diffuse and reversible shedding that follows a physiological stressor by two to three months, which gives a plausible non-drug explanation for many of these reports.

The second category is true hypersensitivity. The 2026 review deliberately confined itself to non-immunologic reactions, and that boundary is clinically important. Immediate systemic hypersensitivity, including urticaria spreading beyond the injection site, angioedema, wheeze, or anaphylaxis, is rare with these agents but is a medical emergency and a reason to stop the drug and seek care immediately, not a reason to rotate injection sites.

Everything in the mild-to-moderate category described in these papers is local: pain, redness, bruising, itching, swelling, a nodule, sometimes altered sensation at the site.

The honest starting point is that both the 2026 review and the 2025 meta-analysis call for prospective studies to define optimal management. There is no randomized evidence establishing the best way to prevent or treat these reactions. What exists is standard subcutaneous injection practice and the observation, repeated across these papers, that most reactions resolve symptomatically without discontinuing therapy.

Standard injectable technique addresses several of the contributors. Rotating sites between abdomen, thigh, and upper arm rather than reusing one spot allows tissue to recover between doses. Letting a refrigerated pen sit at room temperature before injecting reduces stinging, which is a common and easily fixed complaint. Allowing an alcohol swab to dry fully before the needle goes in prevents a burning sensation that patients often attribute to the medication. Avoiding massage of the site after injection and not injecting through irritated or bruised skin are both sensible.

For symptom relief, a cool compress for local discomfort, an oral antihistamine for itching, and a short course of a topical corticosteroid for a persistent inflamed patch are the usual measures, and should be discussed with the prescriber or pharmacist rather than assumed. Persistent nodules on a depot formulation generally need site rotation and patience rather than treatment.

The threshold for calling the clinic is worth stating plainly: a reaction that spreads beyond the injection site, involves the face, lips, or throat, causes breathing difficulty, produces blistering or skin breakdown, or is accompanied by fever or expanding warmth and tenderness suggesting infection. None of those belong in the self-management category.

The last point is the most practical one. These medications are meaningfully effective for the conditions they treat, and stopping one because of a manageable local reaction is a poor trade. Bring the problem to the prescriber before abandoning the therapy.

The two sources complement each other in a useful way. The FAERS analysis is broad, real-world, and captures rare and unusual presentations that trials would never see, but it cannot estimate how often anything happens. The randomized meta-analysis has denominators, controls, and a pooled effect estimate, but it is limited to what trials chose to record, and injection site reactions are often collected casually in trials designed for glycemic or weight endpoints.

Where they agree, confidence is reasonable: injection site reactions are genuinely more frequent on GLP-1 receptor agonists than on comparators, they cluster at the injection site rather than appearing as generalized skin disease, and exenatide and dulaglutide stand out.

What neither can deliver is an incidence figure a patient can use. Nobody can currently say what percentage of people starting semaglutide will develop an injection site reaction, because no adequately designed prospective study has measured it. That is the gap the review’s authors name explicitly.

The headline percentage invites misreading, and the misreading runs in a specific direction. Describing 31 percent of reports as though it meant 31 percent of patients would overstate the problem substantially, and the same arithmetic would make exenatide look roughly four times worse than semaglutide when the comparison is between report volumes shaped by very different prescribing populations and marketing histories.

Combining injection site events with skin-related events into one adjusted category was a reasonable analytic choice for comparing agents, but it merges two different clinical problems. A bruise at an injection site and a widespread rash are not the same event, and the pooled category obscures that.

Spontaneous reporting is also vulnerable to attention effects. Drugs under heavy public and media scrutiny, which describes semaglutide and tirzepatide over the past several years, generate reports at rates unrelated to their underlying safety profile.

The randomized pooled estimate has its own limits. Fourteen trials with 396 events is a modest evidence base, the confidence interval from 2.35 to 5.36 is wide, and trials use trained staff and standardized technique, which likely understates what happens when people inject themselves at home.

These data do not establish how likely any individual patient is to develop a reaction on any specific agent. Reporting shares are not rates, and no prospective incidence study exists.

They do not show that GLP-1 receptor agonists cause generalized skin disease. Randomized data found dermatologic events outside the injection site infrequent and not significantly elevated.

They do not establish that these medications cause hair loss. GLP-1 agents were the most reported class for hair loss among the agents examined in a 2024 pharmacovigilance study, but no drug-event pair met that study’s threshold for a positive signal, and rapid weight loss itself is a recognized cause of reversible telogen effluvium.

They also do not tell clinicians which management strategy works best. The review states directly that prospective studies are needed to define mechanisms and optimal management, so current practice rests on general subcutaneous injection principles rather than on comparative trial evidence.

Injectable medicines have had this problem for as long as they have existed, and the pattern here is familiar from insulin, from depot antipsychotics, and from biologic therapies. Local reactions are common, mostly benign, and disproportionately responsible for people quietly abandoning effective treatment.

What is different now is scale. GLP-1 receptor agonists have moved from a diabetes specialty drug class to one of the most widely prescribed categories in medicine, including large numbers of patients treated for obesity by clinicians who may not have counseled on subcutaneous injection technique before. Problems that were previously distributed across a small population are now visible everywhere.

There is also a formulation trajectory worth watching. Oral GLP-1 receptor agonists in development remove the injection site question entirely, and if they reach broad use they will change which side effects dominate this conversation.

In practice, this is mostly a counseling problem rather than a pharmacology problem. Patients who were told to expect some local redness and itching handle it. Patients who were not told assume something is wrong, stop the medication, and often do not mention it until the next visit, by which point weeks of treatment have been lost.

So I set the expectation up front. Some stinging, redness, or a small lump at the injection site is common and usually settles. Rotate where you inject. Let the pen warm up out of the refrigerator before you use it. Let the alcohol dry. Most of the complaints I hear are fixed by those three things, and it is faintly embarrassing how often the burning a patient attributed to the drug was wet alcohol under the needle.

The line I want patients to know is the one that matters. Local means local. If a reaction spreads past the injection site, involves the face or throat, or causes trouble breathing, that is not an injection site reaction and it does not get managed at home.

I also read these FAERS percentages the way I read all pharmacovigilance data. They are a smoke detector, not a thermometer. They tell me what to look for. They do not tell me how likely it is.

Skin and injection site reactions account for 31 percent of reported adverse events across six GLP-1 receptor agonists in FDA data, and randomized trials confirm injection site reactions occur roughly three and a half times more often than with comparators. Almost all are local, mild to moderate, and manageable without stopping treatment. Systemic reactions involving the face, throat, or breathing are a different problem and need urgent care.

Separate reporting share from incidence every time you see a pharmacovigilance percentage. The 31 percent figure describes what fraction of submitted reports mentioned a skin or injection site event, in a database with no denominator, shaped by publicity and market age. The randomized pooled risk ratio of 3.55 is the number that actually compares these drugs against something, and even that comes from 396 events across 14 trials.