The liraglutide 3 mg label names headache among the most common adverse reactions, in the group reported by at least 5% of treated patients. Its table gives the figure: 13.6% of 3,384 treated adults. The column beside it gives the number that changes the sentence: 12.6% of 1,941 adults on placebo reported the same thing.[1] Headache is genuinely common on this drug, and roughly one point of it is attributable to the drug.
The same gap, across every program that counted it
The semaglutide 2.4 mg weight-reduction pool recorded headache in 14% of 2,116 treated adults against 10% of 1,261 on placebo — a four-point difference, the widest in the class.[2] In the adolescent trial of the same drug at the same dose, 133 treated against 67 on placebo, it was 17% against 16%.[2] The tirzepatide weight-reduction label, whose adverse-reaction table lists everything reported by at least 2% of treated patients and more often than placebo, does not contain headache at all.[3] Nor does the tirzepatide diabetes table.[4] On the injectable and oral semaglutide diabetes labels, headache appears only in the postmarketing section, where reports carry no frequency.[5]
Read together, the class produces a headache signal that is somewhere between one and four percentage points, or absent, depending on the molecule and the population — against placebo arms reporting 7% to 16%. The headline number a reader is likely to meet is 14%. The portion of that 14% the injection is responsible for is closer to four points, and in three of the five programs it is one point or none.
The dose test, which is the cleanest evidence available
Two 72-week trials randomized 1,311 adults to semaglutide 7.2 mg weekly, 304 to 2.4 mg and 303 to placebo. Headache was reported by 7% on placebo, 8% at 2.4 mg and 9% at 7.2 mg.[2] Tripling the dose moved the rate by one percentage point. Vomiting over those same three arms went 6%, 16%, 22%, and nausea went 13%, 35%, 39%, so the trials were plainly capable of detecting a dose-driven symptom when one existed.
That is close to a direct test of causation, and it fails to find much. A symptom that scales with drug exposure scales; headache here does not. Dizziness, which sits near it on the label and is often reported in the same breath, behaves differently and is covered separately in the dizziness article.
No pooled estimate exists, and that is informative
An umbrella review published in 2026 gathered 60 meta-analyses covering 1,751 randomized trials and 3,580,616 participants and reanalyzed 116 unique adverse health outcomes across gastrointestinal, cancer, fracture, respiratory, neurologic, psychiatric, hepatic and endocrine domains, with the neurologic domain built from three separate meta-analyses. Headache is not among the 116.[6] Nobody has pooled it, in a literature that has pooled nausea, vomiting, diarrhea, fractures, thyroid disease and serious infection.
The likely reason is visible in the label tables above: a one-to-four point difference over a noisy background is not the kind of signal a meta-analysis is usually assembled to characterize. The practical consequence for a reader is that no confidence interval for this symptom exists to be quoted, and anyone quoting one has constructed it.
The two routes that are specific to the drug
Dehydration is the first, and it is the same pathway the volume-depletion warning describes. Reduced intake, plus fluid lost to nausea, vomiting or loose stool, is a plausible and unmeasured contributor — no trial has reported headache stratified by hydration status, so the mechanism rests on physiology rather than on a number from this drug class. The days when those losses are largest are the ones covered in the sick-day article.
Hypoglycemia is the second, and here the labels are specific about who it applies to. In the semaglutide weight-reduction program, hypoglycemia defined as blood glucose under 54 mg/dL was reported in 6% of treated patients against 2% on placebo — in the trial restricted to patients with type 2 diabetes, 403 against 402.[2] The tirzepatide program splits the same way: in its diabetes trial, 4.2% against 1.3%; in its trial of adults with obesity and without diabetes, plasma glucose under 54 mg/dL was recorded in 0.3% of treated patients and none on placebo.[3]
For a cash-pay buyer without diabetes and not taking insulin or a sulfonylurea, that is a three-in-a-thousand pathway, and attributing a headache to low blood sugar is the least likely explanation on the list. For someone who does take insulin or a sulfonylurea, the dose-reduction instruction in both labels exists precisely because that risk is real, and the interaction questions sit in the oral medication article.
The literature where the sign flips
The same receptor that shows up on a label as a headache cause is under active investigation as a headache treatment. A randomized, placebo-controlled, double-blind trial used telemetric intracranial pressure catheters in women with active idiopathic intracranial hypertension — intracranial pressure above 25 cmCSF with papilledema. Sixteen were recruited and 15 completed. Exenatide lowered intracranial pressure by 5.7 cmCSF at 2.5 hours (p = 0.048), 6.4 at 24 hours (p = 0.030) and 5.6 at 12 weeks (p = 0.058), against a significance threshold the investigators set a priori at 0.1.[7]
A 2025 systematic review and meta-analysis assembled two clinical trials and two registries covering 1,550 patients with the same condition, 768 of them treated with a GLP-1 or dual GIP/GLP-1 receptor agonist. Treatment was associated with a lower risk of papilledema (RR 0.25; 95% CI, 0.15 to 0.43) and of visual disturbance or blindness (RR 0.41; 0.18 to 0.92), and with a reduction in monthly headache days of 3.64 (95% CI, −6.26 to −1.03) at three months and 4.82 (−8.80 to −0.85) at the end of follow-up. No association with body mass index was detected.[8] The visual endpoints there connect to the ophthalmic questions in the eye health article.
Migraine has one prospective test and it is small and open-label. Thirty-one patients with obesity and high-frequency or chronic migraine, unresponsive to at least two preventive treatments, took liraglutide 1.2 mg daily for 12 weeks. Monthly headache days fell from 19.8 to 10.7 — a mean difference of 9.1 days (95% CI, 5.41 to 12.84; p < 0.001; Cohen’s d 0.90). BMI moved from 34.0 to 33.9, which was not significant (p = 0.060), and the reduction in BMI did not predict the reduction in headache frequency.[9] A systematic review of GLP-1 and pain screened 833 records and included 42 studies across four domains, placing headache alongside inflammatory, neuropathic and visceral pain as an area where the class shows activity.[10]
None of that is a reason to expect a weekly injection to treat a headache disorder. An open-label pilot of 31 people has no placebo arm in a condition with a large placebo response, the randomized evidence is 15 women in a rare disorder, and the meta-analysis pooled registries alongside trials. What the body of work does establish is that a headache arriving on this drug has an equally published mechanism for going the other way, which is an unusual position for a labeled adverse reaction to be in.
What compounded product does not change and does not tell you
Every figure above comes from trials of FDA-approved product at labeled doses. Most sellers reviewed here dispense compounded versions, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed, and no study has counted a headache rate in one. A pharmacovigilance analysis of 81,078 GLP-1 reports, 707 of them compounded, reported elevated odds for abdominal pain, diarrhea, nausea, suicidality and cholecystitis; headache is not among the outcomes it reports.[11] What the same analysis does report — reporting odds of 48.92 for preparation errors and 4.46 for prescribing errors — is the reason an unexpected symptom on a compounded vial carries a question a branded pen does not: whether the dose was the one intended.
What this leaves a reader with
Headache on a GLP-1 is common and mostly not caused by the GLP-1. The largest attributable gap in any program is four percentage points, three programs show one point or none, tripling the dose moves it by one, and no meta-analysis has pooled it. Dehydration is the mechanism most likely to apply and has never been measured here; hypoglycemia applies at 0.3% to anyone without diabetes. The rest of the tolerability picture is in what the trials recorded.
Two patterns do not belong to this list. A sudden, severe headache reaching maximum intensity within seconds needs emergency assessment regardless of what medication preceded it. And a headache that worsens lying down or on waking, or that arrives with transient visual blackouts or a pulsing sound in the ear, describes raised intracranial pressure — the condition this drug class is being trialed as a treatment for, and a reason to be examined rather than to stop and wait.