GLP-1 and Acid Reflux: the Label Says 5% Against 3% on Placebo, and the Reflux-Adjacent Term the Drug Most Clearly Adds Is Belching

Ask whether a GLP-1 drug causes acid reflux and you will be told about gastric emptying: the drug slows the stomach down, so of course the acid comes back up. The FDA prescribing information for Wegovy does state that semaglutide delays gastric emptying — in section 12.2, under pharmacodynamics. It does not attribute reflux to that mechanism anywhere, and neither do we. What the label does give is a number, and beside it the number that says how much of it to attribute to the drug at all: 5% of adults on Wegovy 2.4 mg reported gastro-oesophageal reflux disease in the weight-reduction trials, against 3% of the people on placebo in the same trials.

This is the fifteenth 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, fatigue, where one Phase 3 trial reports less of it on the drug and dizziness, where the two largest arms in the base have no placebo column at all. As in those, every figure below is quoted from a primary source with the arm and the number of participants it belongs to, and all of them were 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 it says about the two terms next to it

The current US prescribing information for Wegovy (semaglutide), revised June 2026, lists gastro-oesophageal reflux disease among what it calls the *"Most common adverse reactions (incidence ≥5%) in adults or pediatric patients aged 12 years and older"*. Two of its tables carry a row for it:

Table 3 is the source of the 5%. It lists every adverse reaction reported in at least 2% of drug-treated adults *and* more often than placebo, so reflux cleared that bar — but only just, and the gap is the point: three of the five percentage points were also reported on placebo. Measured as the placebo figure's share of the drug figure, reflux sits at 60%, a larger placebo share than nausea (36%), vomiting (25%), diarrhoea (53%), constipation (46%), dyspepsia (33%) or fatigue (45%) in the same table. Only five of the nineteen rows in that table have more of their reported rate already present in the placebo arm: viral gastroenteritis (75%), headache (71%), abdominal distension (71%), flatulence (67%) and gastroenteritis (67%).

A third table gives a checkable absence. Table 4 covers the 7.2 mg trial and lists reactions occurring in at least 2% of patients *and* more often than both the 2.4 mg arm and placebo — nausea, vomiting, dysesthesia, constipation, abdominal pain, fatigue, headache, dizziness, hair loss and flatulence. Reflux is not in it. So the label publishes no reflux figure for the 7.2 mg dose; it did not clear that table's bar. For the 25 mg oral tablet the label says only that in its 204-patient trial *"the types and frequency of common adverse reactions were similar to those listed in Table 3"*.

Now the two terms printed immediately around it, both of which a person searching for "acid reflux" would recognise as the same complaint:

Both of the first two are reported *more often* on the drug than reflux is, and both have a far larger margin over placebo. Belching carries the largest multiple of its own placebo figure in the whole of Table 3: under 1% on placebo against 7% on the drug. So on the label's own evidence, the reflux-family symptom the drug most clearly adds is belching, not reflux — and if the question you actually mean is "will my upper gut feel worse", the 5% figure is the wrong one to quote, because it is the narrowest of the four rows above.

One more line worth knowing, because it is the label's only causal-sounding statement in this area and it is about timing, not mechanism: the reactions it groups with reflux — *"dyspepsia, abdominal pain, abdominal distension, eructation, flatulence, gastroesophageal reflux disease, gastritis, hemorrhoids, and hiccups"* — *"were most frequently reported during dosage escalation."* The label's gastritis row also carries a footnote that keeps a related term out of the reflux count: *"Includes chronic gastritis, gastritis, gastritis erosive, and reflux gastritis."* Reflux gastritis is counted under gastritis, not under reflux. (All quoted from the prescribing information via DailyMed, set ID ee06186f, SPL version 19, effective 18 June 2026, accessed 18 September 2026.)

What the trial registries say, arm by arm

Our database cited 14 distinct studies for a stated acid-reflux rate when this article was written on 18 September 2026. Every one of them is a ClinicalTrials.gov results record, and every row below was read from that trial's own adverse-events module on the same day, including the placebo arms our store does not keep. Eight of the 14 post a placebo arm of at least 20 people. The counts are the registry's; the percentages and the differences are our division and subtraction of them.

Read the difference column rather than the headline and two things hold.

Every placebo-controlled drug arm of more than 300 people sits above its own placebo, by 1.9 to 4.1 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 its placebo. That is a narrower and steadier band than this series has found for most effects, and it is the number to carry away: not "5%", but about two to four points above whatever the background rate is in the population you belong to.

Below 300 people per arm the sign stops being guaranteed, exactly as it does for every other effect. Across the 21 drug arms in the eight placebo-controlled trials, 19 sit above their own placebo, one reads level with it to a tenth of a point (retatrutide 1 mg, 1 of 69 against 1 of 70) and one sits below it (the arm the registry labels *"8 mg LY3437943 (4 mg)"*, 0 of 35 against 1 of 70). Both of those are one report either way. The eight placebo arms themselves run from 0.0% (ATTAIN-J, 0 of 60) to 3.1% (STEP 1, 20 of 655) — so the background rate of reflux reported by people on no drug at all, in these trials, is roughly nought to three percent.

The remaining six sources cannot answer the placebo question, and three of them are interesting anyway:

SURPASS-CVOT (NCT04255433), a 13,299-person cardiovascular-outcomes trial in type 2 diabetes, posts reflux at 337/6,647 = 5.1% on tirzepatide against 296/6,647 = 4.5% on dulaglutide 1.5 mg. Two GLP-1-class drugs against each other, no placebo column anywhere in the record.

SURPASS-SWITCH (NCT05564039), a Phase 4 trial of people switching from dulaglutide, is the one pair in our base that runs the other way: 2/139 = 1.4% on tirzepatide against 8/143 = 5.6% on dulaglutide. Again an active comparator, and with eight events on one side and two on the other it is a small difference between two small counts, not a ranking of the drugs.

ACHIEVE-3 (NCT06045221) puts orforglipron against semaglutide head to head: 24/424 = 5.7% and 26/423 = 6.1% on orforglipron 12 and 36 mg, against 12/426 = 2.8% and 18/425 = 4.2% on the trial’s semaglutide 7 mg and 14 mg arms.

The other three are ACHIEVE-J (NCT06010004), a long-term safety study posting 2.3%, 1.5% and 3.0% across three orforglipron doses with no comparator arm; a 23-person study of semaglutide in Alzheimer's disease (NCT05891496) whose placebo arm is 12 people; and a 52-person Phase 1 multiple-dose study of orforglipron (NCT05841238), which is the subject of the next section.

We also checked our own arithmetic while we were inside those records. Each of the 14 entries in our store is the unweighted mean of that study's posted non-placebo arms, carried at the size of its largest arm, and all 14 reproduce from the live registry to the tenth of a point. No defect found — stated because a negative is only checkable if you say what you checked.

What our own number counts, and the three things to know before quoting it

Our acid-reflux endpoint publishes a pooled clinical estimate of 5.0% (95% CI 3–9%), from 52 eligible stated rates across 14 distinct studies, source diversity "high", as read on 18 September 2026. It happens to land on the same figure as the label's Table 3. That coincidence is not corroboration, and we will not present it as such: the label's 5% is one drug at one dose, whole-trial incidence, in three placebo-controlled weight-reduction trials, while ours pools semaglutide, tirzepatide, dulaglutide, retatrutide and orforglipron across Phase 1 to Phase 4, in populations from healthy volunteers to people with type 2 diabetes in a cardiovascular-outcomes trial. Two different quantities that happen to round to the same number.

First: one 52-person Phase 1 study supplies 35% of our rate rows. NCT05841238 moves 52 healthy overweight and obese volunteers through a sequence of dosing periods and formulations, and the registry posts each of those as its own arm — eighteen of them — 600 arm participants in total, 11.5 times the study's own enrolment, with reflux running from 0.0% to 8.3% across them. That is 18 of our 52 stated rates. A second study (the 23-person Alzheimer's trial) posts overlapping periods too, at just under twice its enrolment. Between them the two supply 20 of the 52 rows. The other twelve studies each post arms totalling their own enrolment or within five of it.

Read as "52 independent observations", that base would mislead. It is not how we pool: our estimate collapses each study to one entry, then weights those entries by stated sample size and by how confident our extraction was in reading each figure. So the Phase 1 study contributes 1 of 14 entries at 1.1%, not 18 observations. The count of 52 is a count of rows, and rows are not people.

Second: nearly two thirds of the weight sits on one trial with no placebo arm. The 14 entries carry 10,512 participants of stated sample size between them, and SURPASS-CVOT alone carries 6,647 of those — 63%, before the extraction-confidence discount. SURPASS-CVOT is the active-comparator trial above. So our pooled 5.0% is emphatically not a drug-minus-placebo quantity: most of its weight comes from a record in which nobody was on no drug. If you want the drug's own contribution, use the difference column in the table above, not our headline.

Third, and the one most likely to matter to a reader: we count one registry term, and it is the narrowest of the family. Our rows are the MedDRA preferred term *Gastrooesophageal reflux disease* and nothing else. We checked the alternative spelling rather than assuming it away — no trial in our corpus posts the American form — but we also counted what else those same records publish, and the answer is substantial: 19 of the 31 trials in our corpus post a separate row for dyspepsia and 14 post one for eructation, both of which the label reports at higher rates than reflux. Neither is in our acid-reflux base, and we publish no pooled figure for either. So our 5.0% is a floor on "upper-gut acid symptoms" in the ordinary-language sense, not a measurement of it.

What patients themselves report

Of the 26 distinct community reports in our corpus, 6 mention acid reflux — a 23% reporting frequency (95% CI 11–42%), from reddit.com and drugs.com. 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 reflux at all. People who post are people with something to say, and the gap between 23% and the trials' 5% is mostly the shape of that selection, not a contradiction of it. See the linked article for why that pool is small and cannot currently grow. Our corpus also holds a single FDA FAERS row under this effect — a share of *orforglipron* spontaneous reports mentioning dyspepsia, from one report — which our API prints in a separate block, with its term visible, precisely because a share of complaints is not a share of patients and because, as the section above says, dyspepsia is not the term our incidence rows count.

If you are on a GLP-1 now

This is educational content, not medical advice. The label puts reflux at 5% against 3% on placebo and groups it with the reactions "most frequently reported during dosage escalation"; the trial arms above put the drug's own contribution at roughly two to four points in trials large enough to measure one. The label states that semaglutide delays gastric emptying and attributes reflux to no mechanism at all — so anyone telling you confidently *why* it happens is going beyond the prescribing information, and so would we be.

What the label does treat as serious is a different thing that can feel similar. It warns that *"Acute pancreatitis, including fatal and non-fatal hemorrhagic or necrotizing pancreatitis, has been observed in patients treated with GLP-1 receptor agonists, including WEGOVY"* and tells patients to *"Stop using WEGOVY and call your healthcare provider right away if you have severe pain in your stomach area (abdomen) that will not go away, with or without nausea or vomiting"*. Persistent burning that does not settle between doses, pain that wakes you, difficulty or pain on swallowing, black or bloody vomit, or heartburn that arrives with severe abdominal pain are all reasons to contact a clinician rather than to reach for another antacid. 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.