GLP-1 and Constipation: the Label Says 24%, and the Placebo Arm Says 11%

Almost every answer to "does Ozempic cause constipation?" quotes a single percentage from the drug's side. Almost none of them quotes the number sitting directly beside it in the same table: the share of people on placebo who reported the same thing. For constipation that omission matters more than for any other common GLP-1 side effect, because the placebo number is large.

This is the eighth 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 and why trial vomiting rates disagree so widely. 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 9 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 constipation among the "most common adverse reactions (incidence >= 5%)". Its three adverse-reaction tables each give a drug column and a placebo column:

Table 3 is the source of the 24% that gets quoted. Read the whole row and it says something different from the headline: 24 in a hundred on the drug, 11 in a hundred on placebo, in the same trials, over the same 68 weeks, under the same reporting rules. (Quoted from the prescribing information via DailyMed, SPL version 19, effective 18 June 2026, accessed 9 September 2026.)

For comparison, the same Table 3 puts nausea at 44% against 16% on placebo, diarrhoea at 30% against 16%, and vomiting at 24% against 6%. Constipation's placebo column is not the highest in the table — nausea and diarrhoea both sit at 16% — but constipation is the effect where the placebo share is the largest fraction of the headline: 11 of 24, against 6 of 24 for vomiting.

What the trial registries say, arm by arm

Our database cited 13 distinct studies for a stated constipation rate when this table was read on 9 September 2026 (20 by 14 September, 22 by 16 September once the registry re-collection completed — see the dated items under "How this number moved" below; the rows here are unchanged and still correct for the trials they name; the placebo-controlled sources that entered later are read in a second table, dated 29 September 2026, below this one). Nine of those 13 post a placebo arm in their public results, and eight of those nine have placebo arms large enough to divide. Every row below was read from that trial's own ClinicalTrials.gov adverse-events module on 9 September 2026; the counts are the registry's, the percentages and the differences are our division and subtraction of them.

The other four sources in our constipation base are absent from this table for two different reasons. Three post no placebo arm at all: tirzepatide versus dulaglutide (NCT04255433), where both arms are active GLP-1 drugs and the constipation rates are 836/6,647 = 12.6% and 771/6,647 = 11.6%, and two all-active orforglipron studies (NCT06010004, NCT06440980). The fourth, a placebo-controlled semaglutide trial in non-alcoholic steatohepatitis (PubMed 33185364), does have a placebo group, but our corpus holds it as a published abstract rather than a registry results table, so we have no per-arm counts on file for it and are not going to quote a difference we have not read. And the ninth placebo-controlled study, NCT05086445, posts placebo arms holding four and eleven participants, which is too few to divide.

Update, 29 September 2026: the placebo-controlled sources that entered later. By 29 September our constipation base held 42 distinct studies. We checked every registry record among them for a placebo arm, and 19 beyond the nine above have one. Three post placebo groups of 15 or fewer (NCT03929744, NCT06023095, NCT05891496), and, as with NCT05086445, we are not dividing them. The other 16 are below. Every row was read from that trial's own ClinicalTrials.gov adverse-events module (the non-serious table) on 29 September 2026, and the method is the same as above.

Two things in this table change what the article said above, and it has been restated where it did. With STEP 3 and STEP UP, our base now holds four weight-management trials of injectable semaglutide 2.4 mg with registry rows (STEP 1, STEP 2, STEP 3 and STEP UP), and their gaps run from 10.4 to 13.9 points, not 11 to 14. Second, STEP 3's placebo arm, which also received intensive behavioural therapy, reports 24.5%, the highest placebo arm in either table. The next highest is 16.7%, in the 48-person appetite study, and no other placebo arm of more than 100 people exceeds 11.3%. Neither record says why, and we are not going to guess. Elsewhere, the cardiovascular-outcomes trials show small gaps like SELECT's (REWIND +3.1, SUSTAIN 6 +1.1 and +5.2). In the weight-management trials of other drugs, liraglutide 3.0 mg lands in or near the semaglutide band (+10.0 to +12.2), orforglipron runs from +9.9 to +20.5, and tirzepatide in SURMOUNT-2 sits lower (+3.9 and +4.9).

The difference, not the headline

Take the difference column seriously and the picture is tighter than the raw rates suggest — inside one setting, and only inside it.

In the weight-reduction trials the label itself pools, and in their own registry records, the gap is 11 to 14 points in every source that reports it: Table 3 (24 − 11 = 13), Table 4's 2.4 mg column (19 − 8 = 11), STEP 1 (23.4 − 9.5 = 13.9) and STEP 2's 2.4 mg arm (17.4 − 5.5 = 11.9). The raw drug figures across those same four sources run from 17.4% to 24% and the placebo figures from 5.5% to 11%, so the two columns move together. Those four readings are not four independent trials, and we are not going to present them as if they were: 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 overlap is deliberate — the point is that the difference survives being computed at both the pooled and the single-trial level, not that four separate studies agree. Oral semaglutide at 50 mg lands in the same band (+11.8), and tirzepatide's two lower doses in SURMOUNT-1 sit just inside it (+12.0 and +12.5).

Outside that setting the difference is not stable, and we are not going to present it as if it were. SELECT, dosed identically at 2.4 mg, records 8.1% against 2.6% — a 5.5-point gap with both arms about a third of the STEP 1 values. The most likely reason is not that the drug behaves differently but that a cardiovascular-outcomes trial in 17,604 people ascertains and reports a mild gastrointestinal event differently from a 68-week weight-management trial that is looking for it; that is a hypothesis about trial conduct, not something either record states. At the other end, the three orforglipron arms in the Japanese dose-ranging study sit 20 to 28 points above their own placebo, and the mechanistic appetite study 36 points above its own — small trials in specific populations on an oral small molecule, not the injectables the label describes. And tirzepatide 15 mg (+6.8) reports less constipation than tirzepatide 10 mg (+12.5) in the same trial, which is a reminder that these are counts of reported events, not a dose-response curve.

So the honest one-line answer to "how much constipation does the drug itself add?" is: in adult weight-reduction trials of injectable semaglutide 2.4 mg, about 10 to 14 people in every hundred, on top of a background of 5 to 11 in every hundred who report it on placebo — except in STEP 3, whose placebo arm, which also received intensive behavioural therapy, reported 24.5%. Any other setting needs its own number. (Restated 29 September 2026 from "about 11 to 14 … on top of a background of 5 to 11", after STEP 3 (+12.4 points on a 24.5% placebo arm) and STEP UP (+10.4 points at 2.4 mg) entered our base; see the update under the table above.)

Why our own pooled figure is about half the label's, and why you should not quote it as the Wegovy rate

Our constipation endpoint published a pooled clinical estimate of 11.8% (95% CI 1–56%), from 25 eligible stated rates across 13 studies, source diversity "high", when this article was written on 9 September 2026; as of 21 September 2026, and until later registry batches moved it, it published 11.8% (95% CI 2–43%), from 128 eligible stated rates across 24 distinct studies, still "high"; as of 20:48 UTC on 26 September 2026, and until later registry batches moved it, it published 12.3% (95% CI 3–42%), from 142 eligible stated rates across 31 distinct studies, by then "very high"; as of 20:45 UTC on 29 September 2026, and until later registry batches moved it, it published 11.7% (95% CI 3–40%), from 170 eligible stated rates across 42 distinct studies, still "very high"; as of 12:48 UTC on 30 September 2026, and until later registry batches moved it, it published 11.8% (95% CI 3–40%), from 178 eligible stated rates across 46 distinct studies, still "very high"; as of 19:15 UTC on 3 October 2026, and until later registry batches moved it, it published 11.7% (95% CI 3–39%), from 197 eligible stated rates across 55 distinct studies, still "very high"; as of 17:45 UTC on 16 October 2026, and until later registry batches moved it, it published 11.7% (95% CI 3–36%), from 216 eligible stated rates across 63 distinct studies, still "very high"; as of 14:49 UTC on 21 October 2026 it publishes 11.7% (95% CI 3–36%), from 219 eligible stated rates across 64 distinct studies, still "very high" — the base grew during the registry re-collection described in the update below and the registry batches that followed it. Either way it is roughly half the label's 24%, and the reason is entirely in what the pool contains.

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 — so the two largest studies dominate it: SELECT (8.1%, n=8,803) and the tirzepatide-versus-dulaglutide trial (12.1% across its two active arms, n=6,647 each). Both are large, both are real, and neither is a semaglutide 2.4 mg weight-reduction trial. Mixed in beside them, as read on 9 September 2026, are three orforglipron studies of an oral small molecule, a retatrutide dose-ranging study, an oral semaglutide tablet trial and a mechanistic study in 72 people (the 14 September base adds further orforglipron, semaglutide and tirzepatide registry records — see the update below). Pooling all of that answers the question "across everything we can cite about GLP-1-class drugs, what does the constipation literature look like?" It does not answer "what happens on Wegovy", and the 95% interval (1% to 56% on 9 September, 3% to 40% on 29 September) is the honest statement of how much those sources disagree.

29 September 2026: SELECT and the tirzepatide-versus-dulaglutide trial — SURPASS-CVOT in its registry record — are still the two largest entries, and the third is now REWIND, dulaglutide's cardiovascular-outcomes trial (7.3%, n=4,943). All three are cardiovascular-outcomes trials, and so is SUSTAIN 6 (7.6%, n=826); together the four carry 21,219 of the 34,343 stated-n participants across the base's 42 entries (62%, before the confidence discount) at a size-weighted 9.1%, against 16.0% for the other 38. That, more than any one drug, is why the pooled figure sits at about half the label's. We have not decomposed the move from 12.3% on 26 September to 11.7% on 29 September.

If you want the Wegovy 2.4 mg number, quote the label: 24% against 11% 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. Quoting our pooled percentage as though it were the label's number would be wrong in the direction of understating, which is the kind of error that survives longest because nobody complains about it.

Nobody's source says what causes it

It is close to universal in GLP-1 content — and it was on our own site until this week — to explain constipation by saying the drug slows gastric emptying. The label does say semaglutide delays gastric emptying: it is a named subsection of the Pharmacodynamics section, "Gastric Emptying: Semaglutide delays gastric emptying" (section 12.2), and the Drug Interactions section warns that this may affect the absorption of oral medicines taken alongside it (section 7.2). What the label never does is connect that statement to constipation. We searched every occurrence of the word in the current SPL: it appears in the highlights list of common reactions, in each of the three adverse-reaction tables, in the paediatric narrative and in the postmarketing list — and not once with a mechanism attached.

We had published a sentence attributing this effect to slowed stomach emptying in the description our API and predictor serve for it. It was a plausible sentence that no source we hold states, and it was withdrawn on 9 September 2026; the field now quotes the label's figures and stops there. We are describing the correction rather than quietly editing it because a confident mechanism claim with no citation is exactly the sort of sentence this series exists to remove — and it is worth knowing that the version you may have read on a dozen other sites has, as far as the prescribing information is concerned, no stated source either.

How this number moved twice in three days, and again during the re-collection

Anyone citing a live figure deserves to know when it moved and why. Constipation's published estimate changed twice in the same week, both times because of defects in our own pipeline, both documented in our methodology's corrections log — and then kept moving as the third item below re-collected the registry:

Neither correction was found by an outside reader, and both are the reason the figures in this article are stated with the date they were read. The corpus updates daily; the live endpoint is always the current answer.

What patients themselves report

Of the 42 distinct community reports in our corpus, 9 mention constipation — a 21.4% reporting frequency. (Restated 25 September 2026 from 7 of 26 (26.9%) when ten GLP-1 communities were admitted to the report pool, taking it from 26 to 42.) 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 constipation at all. 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. Two things in the label are worth knowing rather than paraphrasing. Constipation is listed among the most common adverse reactions in the weight-reduction trials, at rates well above placebo. Separately, the postmarketing section — reports collected voluntarily after approval, where the label itself says it is "not always possible to reliably estimate their frequency or establish a causal relationship to drug exposure" — lists "ileus, intestinal obstruction, severe constipation including fecal impaction" among reported gastrointestinal reactions. Constipation that does not resolve, or that comes with abdominal pain, vomiting or an inability to pass gas, is a reason to contact a clinician rather than to add more fibre. 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 9 September 2026 from the cited primary source or from our live production API, and is stated as of that date, except where a later date is given beside it (the 14, 16 and 29 September 2026 updates above). 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.