GLP-1 and Headache: the Label Says 14%, and the Placebo Arm Says 10%

Ask "does Ozempic cause headaches?" and most answers say yes, quote a percentage, and stop. The number they quote is usually real. What it almost never comes with is the figure printed directly beside it in the same table: the share of people on placebo who reported a headache too. For headache that second number matters more than for any other common GLP-1 side effect we have read this way, because it is nearly the same size as the first.

This is the ninth piece in our evidence series, after which effects have enough evidence to calibrate a model, the corpus-derived pooled rates, what patients themselves report, gallstones end to end, what is inside a personalised risk number, hair loss end to end, why trial vomiting rates run from 6.5% to 58% and constipation read against the placebo arm. Like those, every figure below is quoted from a primary source with the arm and the number of participants it belongs to, and every one of them was read on 16 September 2026 — from the current US prescribing information, from each trial's own results table on ClinicalTrials.gov, or from our own live production API.

What the label says

The current US prescribing information for Wegovy (semaglutide), revised June 2026, lists headache among the *"most common adverse reactions (incidence ≥5%) in adults or pediatric patients aged 12 years and older"*. Its three adverse-reaction tables each give a drug column and a placebo column:

Table 3 is the source of the 14% that gets quoted. Read the whole row and it says something different from the headline: 14 in a hundred on the drug, 10 in a hundred on placebo, in the same trials, over the same 68 weeks, under the same reporting rules. In the 7.2 mg trial the gap is one to two points; in adolescents it is one. (Quoted from the prescribing information via DailyMed, SPL version 19, effective 18 June 2026, accessed 16 September 2026.)

For comparison, the same Table 3 puts nausea at 44% against 16% on placebo, vomiting at 24% against 6%, and constipation at 24% against 11%. Of the four effects this series has now read against the placebo column, headache is the one where the placebo figure is the largest fraction of the drug figure: 10 of 14, against 11 of 24 for constipation, 16 of 44 for nausea and 6 of 24 for vomiting.

What the trial registries say, arm by arm

Our database cited 18 distinct studies for a stated headache rate when this article was written at 15:52 BST on 16 September 2026. Nine of those 18 post a placebo arm of at least 20 people; every row below was read from that trial's own ClinicalTrials.gov adverse-events module on 16 September 2026. The counts are the registry's; the percentages and the differences are our division and subtraction of them.

The other nine sources in our headache base are absent from this table for two different reasons. Four are placebo-controlled but split their placebo across parts of 2 to 15 people, which is too few to read a difference from: two single- and multiple-ascending-dose studies of orforglipron, in healthy volunteers (NCT03929744, placebo parts of 8, 15 and 2) and in Japanese participants with type 2 diabetes (NCT05086445, placebo parts of 4 and 11); a dose-titration study in Chinese participants (NCT06023095, placebo arm of 4); and a 23-person inflammation study of semaglutide in Alzheimer's disease (NCT05891496, placebo arm of 12, reporting 1 headache against 0 of 11 and 1 of 22 on the drug). Five post no placebo arm at all: a tablet-versus-capsule bioequivalence study of orforglipron in 533 adults with obesity (NCT06440980, 38 dose-level arms from 0.0% to 19.4%), a multiple-dose formulation study (NCT05841238), a drug-interaction study of orforglipron with carbamazepine (NCT06370728, 0/30 and 0/28 on orforglipron — see the correction below), the long-term safety study ACHIEVE-J (NCT06010004, 3.0%, 3.7% and 4.5% across its three doses), and SURPASS-SWITCH (NCT05564039), where tirzepatide 15 mg or maximum tolerated dose reads 8/139 = 5.8% against 6/143 = 4.2% on dulaglutide, an active comparator rather than a placebo.

Two of the largest trials of the class are not in our headache base at all, and the reason is itself informative. ClinicalTrials.gov's non-serious adverse-event table lists only terms that reached the trial's reporting threshold, 5% for both of these. SELECT (NCT03574597, 17,604 people, semaglutide 2.4 mg against placebo in a cardiovascular-outcomes population) posts no non-serious headache row, so headache did not reach 5% in either of its arms; its serious-events table lists 7 of 8,803 on semaglutide against 11 of 8,801 on placebo. SURPASS-CVOT (NCT04255433, tirzepatide against dulaglutide in 13,299 people with type 2 diabetes) likewise posts headache only as a serious event, 1 of 6,647 against 4 of 6,647. Under our eligibility rules a serious-events row is admitted, as a floor on incidence, only for an effect the registry posts in no other table anywhere in the corpus, which headache is not, so neither trial contributes a rate — but both records say the same thing the placebo-controlled table above says.

The difference, not the headline

Take the difference column seriously and the picture is consistent across the sources large enough to read.

In the adult weight-management trials of injectable semaglutide 2.4 mg, the gap is about three points: Table 3 (14 − 10 = 4), STEP 1 (15.2 − 12.2 = 3.0) and STEP 2's 2.4 mg arm (7.7 − 5.0 = 2.7). As with constipation, those three readings are not three independent trials — Table 3 is the label's pool of three weight-reduction trials, and STEP 1 and STEP 2 are trials inside that programme with their own registry records — so the point is that the difference survives being computed at both the pooled and the single-trial level, not that three separate studies agree. What is worth noticing is that the raw drug figures across those same sources run from 7.7% to 15.2% and the placebo figures from 5.0% to 12.2%: the two columns move together, and the drug figure on its own tells you more about the trial than about the drug.

In the tirzepatide and orforglipron trials the gap is close to zero. SURMOUNT-1's three tirzepatide doses sit within 0.7 points of their own placebo arm, and 15 mg reports fractionally *less* headache than placebo. ATTAIN-2's three orforglipron doses sit within 1.6 points of placebo, the Japanese orforglipron trial's three doses all sit *below* placebo, and the third orforglipron Phase 3 trial runs from 0.0 to 2.0 points above it. In the adolescent oral-semaglutide trial the placebo arm is three points higher than the drug arm. Of the nine placebo-controlled sources in the table, six report at least one drug arm at or below its own placebo.

Two rows sit well above the rest, and we are not going to smooth them into the picture. The 120-person appetite-and-eating-behaviour study reports 23.6% against 12.5% — an 11-point gap, in a trial whose own title says it was studying appetite, eating behaviour and psychosocial status, with 72 people on the drug. And one of retatrutide's six dosing arms reports 4 of 35, 11.4%, against 0 of 70 on placebo, while three of its other arms report 0 of 33, 0 of 35 and 2 of 69. Neither is a weight-management trial of an approved drug at an approved dose, and in both a handful of people moves the percentage by double digits; they are in the table because they are in our base, and a table that left them out would be quieter than the evidence.

So the honest one-line answer to "how much headache does the drug itself add?" is: in every placebo-controlled trial arm of more than 300 people that posts a headache rate, the drug arm sits within about three points of its own placebo arm — from 0.3 points below it to 3.2 above — and in adult weight-management trials of semaglutide 2.4 mg specifically, about three points above it. That is a small difference next to nausea (44% against 16% in the same label table) or constipation (24% against 11%), and it is why "headache is a common side effect" and "most of the headaches reported in these trials also happened on placebo" are both true statements about the same table.

Our own pooled estimate, and why it is not the Wegovy number

Our headache endpoint published a pooled clinical estimate of 9.9% (95% CI 4–22%), from 117 eligible stated rates across 18 distinct studies, source diversity "high", when this article was written at 15:52 BST on 16 September 2026. It is lower than the label's 14% for the same reason our constipation figure is lower than the label's: the pool is the whole class, not one drug at one dose in one population.

Our pooled figure weights each source by its own stated sample size (a source stating none counts as 1), multiplied by a discount for how confident our extraction was in reading the figure, after collapsing each study to one entry — the unweighted mean of its posted arms, carried at the size of its largest arm. So the entries with the most weight are STEP 1 (15.2%, n=1,306), SURMOUNT-1 (7.5%, n=636), ATTAIN-2 (5.9%, n=331) and the 533-person bioequivalence study (7.0%, n=215), and beside them sit six Phase 1 pharmacology studies of orforglipron — dose-escalation, formulation and drug-interaction studies dosed for days rather than months, in healthy volunteers or small diabetic cohorts — which are six of the 18 studies and contribute the widest per-arm spread in the base, from 0% to 37.5% across arms of 6 to 215 people. Whether Phase 1 arms of that kind belong in a pooled incidence estimate at all, or should be kept and labelled, is a methodology question we have open at the time of writing; for now they are in, and the live endpoint names every one of them so you can take them out yourself.

If you want the Wegovy 2.4 mg number, quote the label: 14% against 10% on placebo, N=2,116 and N=1,261, Table 3. If you want the whole class with its heterogeneity visible, quote ours, with its interval and its source count. Neither is the other.

What the FDA's adverse-event reports show

A different kind of number sits outside the incidence pool on purpose. FDA FAERS publishes the share of *spontaneous adverse-event reports* for each drug that mention headache: 3,420 of 82,377 semaglutide reports (4.2%), 2,018 of 43,057 liraglutide reports (4.7%), 2,086 of 94,688 exenatide reports (2.2%) and 57 of 1,154 orforglipron reports (4.9%), as read from our own endpoint on 16 September 2026. A share of complaints is not a share of patients — it answers "of the people who filed a report about this drug, how many mentioned a headache," not "how many people on the drug got one" — so our eligibility rules report these separately rather than average them in.

What patients themselves report

Of the 26 distinct community reports in our corpus, 1 mentions headache — a 3.8% reporting frequency. That is not an incidence rate and we never treat it as one: it is the share of people who chose to post about their GLP-1 experience who mentioned headache at all, and at one report it is a count, not a rate. See that article for why the pool is small and cannot currently grow.

Correction, 16 September 2026 — one row withheld before this article was published

Reading every arm of the 18 sources before quoting them turned up one that is not a GLP-1 rate. The drug-interaction study NCT06370728 doses its 30 healthy participants with orforglipron alone (days 1–5), then with carbamazepine alone (days 6–35, except day 15), then with both together (day 15); the registry reports headache for each period, and our collector had stored the carbamazepine-only period — 3 of 29, 10.3% — as an eligible rate, and had done the same for seven other effects from the same trial (nausea, vomiting, abdominal pain, constipation, reduced appetite, dizziness and fatigue). Carbamazepine is an anticonvulsant with no GLP-1 activity, so those eight rows were withheld from both our stores at 15:50 BST today, before this article was published, and the collector now skips that arm by name. The effect on published figures is one stated rate fewer for each of the eight effects (headache 118 → 117; site-wide 1,133 → 1,125 eligible rates from the same 30 studies), no pooled percentage moved, and two intervals widened by a point (headache 4–21% → 4–22%, fatigue 1–23% → 1–24%). We are describing it rather than quietly editing it because a comparator that is not the study drug is exactly the kind of row a hostile reader would ask about first, and this is the second such arm we have found this month — the first, an "Other Oral Glucose-lowering Medication" arm in a semaglutide trial, was withheld on 8 September.

The same day's registry re-collection also finished: the 19 trials whose rows had been collected under a rule that kept only each trial's five highest-percentage arms — a selection bias disclosed on 10 September — are now stored in full (every tracked arm row the registry posts, re-verified against the live registry this afternoon), so the per-study means in the headache base, including SURMOUNT-1's and ATTAIN-2's, are the means of every posted arm and not of the highest ones.

If you are on a GLP-1 now

This is educational content, not medical advice. The label lists headache among the most common adverse reactions, at rates a few points above placebo. Two other places it appears in the patient information are worth knowing rather than paraphrasing: the label lists headache among the *"signs and symptoms of low blood sugar"* to watch for, a risk it says applies *"especially"* to people with type 2 diabetes who also take insulin or a sulfonylurea, and it separately warns that *"diarrhea, nausea, and vomiting may cause a loss of fluids (dehydration)"*. A headache that comes with confusion, slurred speech or a fast heartbeat, or one that follows days of being unable to keep fluids down, is a reason to contact a clinician rather than to take a painkiller and wait. Nothing here should be used to start, stop or change a medication or its dose without your prescriber.

Every figure in this article was read on 16 September 2026 from the cited primary source or from our live production API, and is stated as of that date. Query everything behind it — all 15 effects, both tracks, every source with its URL — at magistra.health/en/data-api, with the full eligibility methodology at magistra.health/en/methodology.