GLP-1 and Dizziness: the Label Says 8%, the Placebo Arm Says 4%, and the Two Largest Arms in Our Base Have No Placebo Column at All

Ask "does Wegovy make you dizzy?" and the usual answer is a mechanism: you are eating less, you are dehydrated from the nausea, your blood pressure has dropped with your weight. All of those may be true and none of them is a number. The prescribing information gives one — about 8% of adults on Wegovy 2.4 mg reported dizziness in the weight-reduction trials — and, in the column printed beside it, the number that tells you how much of that to attribute to the drug: 4% of the people on placebo in those same trials reported dizziness too.

This is the fourteenth 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%, constipation read against the placebo arm, headache read against the placebo arm, abdominal pain, where the label and the registry disagree by a factor of two, nausea, where the label's 44% holds up and the largest trial reads 18%, diarrhoea, where the three trials behind the label's number add up exactly and fatigue, where one Phase 3 trial reports less of it on the drug. 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 18 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, and what its footnote says

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

Table 3 is the source of the 8%. Its dizziness row carries no footnote, so as printed it is the single term. Table 4's row does carry one — *"includes dizziness and dizziness postural"* — two separate adverse-event terms, counted as one, which matters later because trial registries post those terms separately and we count only the first. Table 5, the adolescent table, lists every reaction reported in at least 3% of drug-treated patients *and* more often than placebo, and dizziness is in it at 8% against 3%, the same drug figure as the adults. (Quoted from the prescribing information via DailyMed, set ID ee06186f, SPL version 19, effective 18 June 2026, accessed 18 September 2026.) The label's oral form gets no separate table: it says only that in the 204-patient WEGOVY 25 mg tablet trial *"the types and frequency of common adverse reactions were similar to those listed in Table 3"*.

For comparison, the same Table 3 puts nausea at 44% against 16% on placebo, diarrhoea at 30% against 16%, vomiting at 24% against 6%, constipation at 24% against 11%, abdominal pain at 20% against 10%, headache at 14% against 10% and fatigue at 11% against 5%. Ranked by how large the placebo figure is as a share of the drug figure, that table reads: headache 71% (10 of 14), diarrhoea 53% (16 of 30), abdominal pain and dizziness 50% (10 of 20; 4 of 8), constipation 46% (11 of 24), fatigue 45% (5 of 11), nausea 36% (16 of 44) and vomiting 25% (6 of 24). Dizziness sits in the upper half of that range, tied with abdominal pain: the drug figure is exactly double the placebo figure, and no lower.

So the label's own answer to "is the dizziness the drug?" is: half of the dizziness reported on the drug was also reported on placebo, and the drug adds about four points on top.

What the trial registries say, arm by arm

Our database cited 17 distinct studies for a stated dizziness rate when this article was written at 15:50 BST on 18 September 2026. Ten of those 17 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 18 September 2026. The counts are the registry's; the percentages and the differences are our division and subtraction of them.

The other seven sources in our dizziness base are absent from that table for two different reasons. Three are placebo-controlled but their placebo arms are too small to read a difference from: a 133-person single- and multiple-ascending-dose study of orforglipron in healthy subjects (NCT03929744, placebo arms of 8, 15 and 2 people, all reporting 0), a 62-person Phase 1 study of orforglipron in Japanese participants with type 2 diabetes (NCT05086445, placebo arms of 4 and 11, both 0), and a 23-person study of semaglutide and the immune system in Alzheimer's disease (NCT05891496, 0 of 12 on placebo against 1 of 11 and 0 of 22 across two drug periods). Four post no placebo arm at all — and the first of them is the largest source in our entire dizziness base. SURPASS-CVOT (NCT04255433), a cardiovascular-outcomes trial of 13,299 people with type 2 diabetes, posts dizziness at 462/6,647 = 7.0% on tirzepatide against 407/6,647 = 6.1% on dulaglutide 1.5 mg: one GLP-1-class drug against another, both within a point of the label's 8%, with no placebo column anywhere in the record. The other three are the long-term safety study ACHIEVE-J (NCT06010004, 2.3%, 2.2% and 4.5% across its three orforglipron doses); a drug-interaction study of orforglipron with carbamazepine (NCT06370728, 0/30 on orforglipron alone and 0/28 on the combination — its carbamazepine-only period, 1/29, was withheld from our store on 16 September because carbamazepine is not a GLP-1); and a Phase 1 bioequivalence study of orforglipron tablets against capsules (NCT06440980), which is the subject of a later section.

Two trials a reader might expect in the table are not in our base at all, for the same reason. The label's Table 3 pools three trials, and their arm sizes identify them exactly: STEP 1, STEP 2 and STEP 3. STEP 2 (NCT03552757) posts no dizziness row. ClinicalTrials.gov results tables list non-serious adverse events only above a frequency threshold the sponsor sets — 5% for STEP 2 — so a term that reached 5% in no arm is simply absent from the record, and the label's pooled 8% is the only public trace of STEP 2's dizziness count. SELECT (NCT03574597), the 17,604-person cardiovascular trial, is absent for the same reason under the same 5% threshold. Both facts were checked on the live registry record on 18 September 2026 rather than assumed.

The difference, not the headline

Take the difference column seriously and three things stand out.

Every placebo-controlled drug arm of more than 300 people sits above its own placebo, by 0.4 to 7.4 points. There are eight such arms — STEP 1, STEP 3, the three SURMOUNT-1 doses and the three ATTAIN-2 doses — and not one reads at or below placebo. For injectable semaglutide 2.4 mg in adult weight management the gap is four to seven points: Table 3 (8 − 4 = 4), STEP 1 (7.5 − 3.5 = 4.0) and STEP 3 (12.8 − 5.4 = 7.4). Those are not three independent trials — Table 3 is the label's pool of STEP 1, 2 and 3, and the two registry rows are trials inside it — so the point is that the difference survives being computed at both the pooled and the single-trial level. Note again how much the raw drug figure moves while the difference is steadier: 7.5% in STEP 1 and 12.8% in STEP 3, same molecule at the same maintenance dose, with the placebo arms moving alongside them from 3.5% to 5.4%. STEP 3 is the trial in which everyone enrolled, drug and placebo alike, also followed an intensive lifestyle programme, and it reports the most dizziness of any arm in our base on *both* sides of the randomisation.

Below 300 people per arm, the sign is no longer guaranteed. Across the 23 drug arms in the table, 20 sit above their placebo and three sit below it: ATTAIN-J's 6 mg arm (0 of 61 against 1 of 60), and retatrutide's 1 mg arm (1 of 69 against 2 of 70) and its 8 mg arm started at 2 mg (0 of 35 against 2 of 70). Each of those is one or two reports either way. The six retatrutide arms together run from 0.0% to 8.6% around a single 2.9% placebo, which is the honest shape of a Phase 2 dose-ranging study with 33 to 69 people per arm: the drug effect on dizziness is there in the higher arms and invisible in the lower ones, and the registry does not tell you which is noise.

The two largest arms in our base cannot tell you anything about placebo at all. SURPASS-CVOT's 6,647-person tirzepatide arm (7.0%) and its 6,647-person dulaglutide arm (6.1%) enter our pool as a single entry — the mean of the two, carried at n=6,647 — and that one entry is 6,647 of the 10,337 participants of stated sample size across all 17 entries behind our pooled figure, 64% of the weight before the extraction-confidence discount. They are an active-comparator trial: what they show is that two different GLP-1-class drugs, at maintenance dose, in the same population of people with type 2 diabetes, report dizziness within a point of each other and within a point of the label's 8%. What they cannot show is how much of that 7% would have been reported on no drug at all, because nobody in that trial was on no drug. That is worth knowing before quoting our pooled number as if it were a drug-minus-placebo effect: the weight in it comes mostly from a record that has no placebo column.

So the honest one-line answer to "how much dizziness does the drug itself add?" is: in every placebo-controlled trial arm of more than 300 people that posts a dizziness rate, the drug arm sits 0.4 to 7.4 points above its own placebo arm; in smaller arms it runs from 2.9 points below placebo to 6.2 above. The label's four-point gap sits in the middle of that first range.

What our own number counts, and the one thing a reader should know before quoting it

Our dizziness endpoint publishes a pooled clinical estimate of 6.5% (95% CI 3–14%), from 97 eligible stated rates across 17 distinct studies, source diversity "high", as read at 15:50 BST on 18 September 2026. It is lower than the label's 8% for the reason this series has reported for every effect: the pool is the whole class, not one drug at one dose in one population. Our 17 studies span semaglutide, tirzepatide, dulaglutide, retatrutide and orforglipron, at Phase 1 through Phase 4, in populations from otherwise-healthy volunteers to people with type 2 diabetes in a cardiovascular-outcomes trial.

The second thing is specific to dizziness, and it is the kind of thing a hostile reader should not have to find for us. 59 of those 97 stated rates — 61% — come from two Phase 1 studies of orforglipron in healthy volunteers. NCT06440980, a bioequivalence study of orforglipron tablets against capsules in *"participants with obesity or overweight who are otherwise healthy"*, enrolled 533 people and posts dizziness for 38 separate dose-level and formulation arms, from 0.0% to 12.2%; those 38 arms have 3,772 participants between them, 7.1 times the trial's own enrolment, because the same people appear in arm after arm as the study moves them through cohorts and dose levels. NCT03929744, a single- and multiple-ascending-dose study in 133 healthy subjects, posts 24 arms (three of them placebo) totalling 316 participants, 2.4 times its enrolment. Between them: 38 + 21 = 59 rows.

Read as "97 independent observations", that base would be badly misleading. It is not how we pool. Our pooled figure collapses each study to one entry — the unweighted mean of its posted non-placebo arms, carried at the size of its largest arm — and then weights those entries by their stated sample size, multiplied by a discount for how confident our extraction was in reading each figure. So NCT06440980 contributes 1 of 17 entries, at 2.2% and n=215, and NCT03929744 1 of 17 at 2.7% and n=45, not 59 observations; the entries with the most weight are SURPASS-CVOT (6.6%, n=6,647 — the mean of its tirzepatide and dulaglutide arms), STEP 1 (7.5%, n=1,306), SURMOUNT-1 (5.0%, n=636), STEP 3 (12.8%, n=407) and ATTAIN-2 (5.3%, n=331). The collapsing is why our 6.5% is not dragged towards 2% by two Phase 1 studies. But the 97 is a count of rows, and rows are not people: the honest reading of our dizziness base is seventeen studies, six of which post overlapping periods or cohorts of the same participants as separate arms (NCT06440980 at 7.1 times enrolment, the carbamazepine interaction study at 2.9, NCT03929744 at 2.4, the 23-person Alzheimer's study at 2.0, STABLE Wt Loss at 1.6 and the 62-person Japanese Phase 1 study at 1.3). Those six supply 70 of the 97 stated rates between them, so nearly three quarters of the row count in our dizziness base comes from studies that count some participants more than once. The other eleven do not: each one's posted dizziness arms sum to its own enrolment or to within five of it (SURMOUNT-1 2,539 of 2,539, STEP 1 1,961 of 1,961, STEP 3 611 of 611, ACHIEVE-1 559 of 559, ACHIEVE-J 401 of 401, ATTAIN-J 238 of 238, PIONEER TEENS 132 of 132, SURPASS-CVOT 13,294 of 13,299, ATTAIN-2 1,608 of 1,613, OASIS 2 200 of 201, retatrutide 337 of 338). Whether overlapping-period arms belong in a pooled incidence estimate at all is a methodology question we have open at the time of writing; for now they are in, collapsed to one entry per study, and the live endpoint names every source with its URL so you can take any of them out yourself.

There is a third reason our number is not the label's, and it runs the other way. We count only the term "dizziness"; registries post "dizziness postural", "vertigo" and "vertigo positional" as separate rows, and none of those is in our base. Four of our 17 studies post postural dizziness separately (retatrutide: 1/69, 0/33, 0/33, 2/35, 0/35 and 2/62 across its six arms against 0/70 on placebo; ACHIEVE-J: 2/132, 3/135 and 0/134; ACHIEVE-1 and ATTAIN-J one report each on one arm), and four post vertigo. The label groups postural dizziness into its Table 4 row and, as printed, not into its Table 3 row. Counting the related terms the way Table 4 does would raise our figure, not lower it — by less than simple addition suggests, because the label counts a person who reported either term once and a registry table cannot tell you how many people reported both. So our 6.5% and the label's 8% are not two measurements of one quantity that disagree: one is a class-wide pooled estimate of a single term, weighted mostly by a record with no placebo arm; the other is one drug's whole-trial incidence in three placebo-controlled trials. If you want the Wegovy 2.4 mg number, quote the label: 8% against 4% 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 dizziness, and our corpus holds four such rows: 2.9% of 99,460 semaglutide reports, 2.9% of 51,432 liraglutide reports, 2.8% of 94,688 exenatide reports, and 2.9% of 1,154 orforglipron reports, as stored in our corpus (the three marketed drugs' shares were read from FAERS on 8–12 April 2026 and the orforglipron share on 13 August 2026, and are stated as of those dates). 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 dizziness," not "how many people on the drug got it" — so our eligibility rules report these separately rather than average them in. What is notable is how flat they are: three drugs with very different report volumes, and one barely marketed, all within a tenth of a point of each other. The orforglipron figure rests on 1,154 reports of a drug that is not yet widely marketed, and should be read as a count rather than a rate.

What patients themselves report

Of the 26 distinct community reports in our corpus, 2 mention dizziness — a 7.7% 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 dizziness at all, and at two reports it is a count, not a rate. See that article for why the pool is small and cannot currently grow.

If you are on a GLP-1 now

This is educational content, not medical advice. The label lists dizziness among the most common adverse reactions, at double the placebo rate, and the trial evidence above says the drug adds a few points on top of a background rate that runs from one to five percent across the placebo arms above. What the label does *not* do anywhere is state a cause for it — the dehydration and blood-pressure explanations that circulate are plausible mechanisms, and no source in our corpus states a rate that tests either.

The label does, though, name dizziness in two different places, and the second one is the one to know about. In the patient information it appears in its own right — *"the most common side effects of WEGOVY in adults or children aged 12 years and older may include: nausea, stomach (abdomen) pain, dizziness…"* — and it appears again at the head of the list of *"signs and symptoms of low blood sugar"*: *"dizziness or light-headedness, sweating, shakiness, blurred vision, slurred speech, weakness, anxiety, hunger, headache, irritability or mood changes, confusion or drowsiness, fast heartbeat, feeling jittery"*, a risk the label says is increased for *"patients with diabetes mellitus taking WEGOVY in combination with insulin or an insulin secretagogue"* such as a sulfonylurea. It separately warns that *"diarrhea, nausea, and vomiting may cause a loss of fluids (dehydration) which may cause kidney problems"*. Dizziness that comes with sweating, shakiness or confusion, that follows days of being unable to keep fluids down, or that arrives on standing up in someone whose blood-pressure medication has not been reviewed since they started losing weight, is a reason to contact a clinician rather than to wait it out. 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 18 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.