Skip to content
GLP Loss
← Research
Safety

GLP-1 Sulfur Burps: A Top Signal With No Study Behind It

Belching tops the disproportionality table for both semaglutide and tirzepatide reports, and sits nowhere on the frequency table. No trial has ever measured it as an endpoint, and nothing has measured what a belch contains.

Tessa Whitfield8 min read
Two ways to rank the same reportsSpontaneous reports, ranked by how common and by how distinctiveReported most oftenNausea, 27.7%then diarrhea at 12.8%Belching does not placeMost out of proportionEructationtop of 45 signals, at 42.03the lowest signal sits at 1.01Trials measuring belching as an endpointNone. And none has measured what a belch contains.Frequency says what is common. Disproportionality says what isdistinctive. Quoting one alone describes a different drug.

Sulfur burps are one of the most searched side effects of this drug class and one of the least studied. The gap is not an oversight in the coverage; it is the state of the evidence. Belching appears in the record under a clinical term, it has a plausible mechanism assembled from two separate literatures, and nobody has run the study that would join them. What follows is that record, with the missing pieces named rather than filled in.

What the trials actually recorded

Adverse events in these programs are coded to a standard dictionary, and the term that covers belching is eructation. It is collected the way every other volunteered symptom is collected: a participant mentions it, an investigator codes it, and it lands in a table beside nausea and vomiting. It carries no descriptor for smell, taste, timing or duration, because the dictionary has no field for any of those.

That is the whole of it. No trial in this class has made belching a prespecified endpoint, none has asked participants to characterize what a belch smelled like, and none has measured the composition of exhaled or eructated gas in anyone taking one of these drugs. The incidence tables summarized in what the trials recorded are complete on nausea and vomiting and effectively silent here.

Where the signal does show up

Spontaneous reporting databases are a different instrument, and on this symptom they are the only one that has produced a number. A pharmacovigilance analysis of semaglutide identified 5,442 gastrointestinal adverse-event cases and detected 45 distinct signals. Ranked by the lower bound of the reporting odds ratio, those signals ran from 1.01 at the bottom, for oral hypoaesthesia, to 42.03 at the top — and the term at the top was eructation.[1]

Tirzepatide produces the same shape from an independent dataset. In an analysis of reports from mid-2022 onward, nausea (27.7%) and diarrhea (12.8%) were the events reported most often, while eructation and impaired gastric emptying carried the highest disproportionality of any gastrointestinal term.[2]A separate study of 123,145 tirzepatide reports listed eructation among its novel signals not carried on the product labeling — meaning the pattern in the reports was strong enough to flag against a term the labeling did not lead with.[3]

Those two rankings invert, and the inversion is the point. Frequency asks how many people mention a thing. Disproportionality asks how much more often it is mentioned about this drug than about drugs in general. Nausea wins the first and is unremarkable on the second, because nausea is reported about almost everything. Belching is the reverse: uncommon in absolute terms, and about as specific to this class as a spontaneous report gets.

The timing is consistent across all three datasets

Median time to onset was 13 days in one tirzepatide analysis, with 67.1% of events arriving inside 30 days, and 16 days in the other.[3][2] The semaglutide analysis put the median at 23 days for its highest-priority signal and described every disproportionality signal as having early-failure characteristics — risk concentrated at the start and declining with time on drug.[1]

That is the same calendar the nausea data describe, and it is laid out in the timing article. Whatever is driving the belching appears to be a feature of early exposure and dose escalation rather than of long maintenance.

Who the reports come from is not who you would guess

The tirzepatide analysis went further than the headline signal and split its gastrointestinal reports by patient. Events were more common in older adults, in males, and in patients taking concomitant medications, and most arrived within three months.[2]Two of those three cut against the demographic that dominates direct-to-consumer weight-loss prescriptions, where the patient is more often female and younger, and where the differences by sex are set out in the sex-differences article.

The same analysis found the reported gastrointestinal risk greater in patients with type 2 diabetes than in those taking tirzepatide for weight loss, and put tirzepatide’s overall reporting odds below those of the GLP-1 receptor agonists while remaining above non-GLP-1 drugs.[2] Those comparisons are between populations that differ in age, comorbidity and medication count as well as in indication, and the two approvals are not the same clinical situation — a distinction drawn in the two-indications article. A reporting database cannot separate those, which is precisely why it generates hypotheses rather than answers.

The mechanism, in two halves that have never been joined

The first half is the belch itself, and it is well described. Eructation runs in three phases, and the one that starts it is the gastro-esophageal inhibitory reflex: transient relaxation of the lower esophageal sphincter, triggered by distension of stretch receptors in the proximal stomach.[4] Gas leaves because the top of the stomach is stretched. Anything that keeps a meal sitting in the proximal stomach longer therefore raises the number of times that reflex fires — which is a coherent account of why the reports cluster in the first month, when the emptying effect is at its strongest, and thin out later. The same reflex is the one that carries acid upward, which is why belching and reflux travel together in the heartburn article.

The second half is the smell, and here the literature is about a different organ entirely. Hydrogen sulfide is produced in the colon by sulfate-reducing bacteria fermenting sulfur-containing substrate, much of it from protein. A crossover study fed 11 healthy volunteers a week of an animal-based low-fiber diet and a week of a plant-based high-fiber diet. Median hydrogen sulfide production was higher after the animal-based week — a median difference of 29 ppm/g (95% CI, 16 to 97; P = 0.02).[5]

The result reverses in a meaningful minority. Two of the 11 participants, 18%, produced more hydrogen sulfide on the plant-based diet, and the authors used the responder split to describe distinct production patterns.[5] So even the dietary lever that gets recommended most often — eat less protein, more plants — went the wrong way in nearly a fifth of a very small sample. High-protein eating is also what this class of drug is usually prescribed alongside, for the reasons set out in the diet article.

What nobody has established

Hydrogen sulfide is made in the colon. A belch comes out of the stomach. Nothing in the published record demonstrates that colonic hydrogen sulfide reaches the mouth as an eructation in a person taking a GLP-1, or that the belches these drugs generate differ in composition from any other belch. The mechanism circulating in forums and repeated on seller pages is an assembly of two true literatures with no measurement joining them.

The absence is institutional as well as experimental. The 2026 expert consensus on nutritional and lifestyle support during treatment produced 52 statements and addressed nausea, vomiting, diarrhea and constipation. Belching is not among the symptoms it covers, and the panel notes its statements were primarily derived from indirect evidence with direct evidence urgently required.[6] A symptom that tops a disproportionality table and appears in no guidance document is the exact shape of an unstudied problem.

What can honestly be said

Three things. Belching is a real and distinctively reported effect of this class rather than a folk complaint. It arrives early and, on the reporting data, recedes. And the two molecules are not identical on gastrointestinal reporting generally, a difference traced in the molecule comparison — though no dataset resolves whether that extends to this particular symptom.

Everything past that is untested. Dietary adjustments aimed at sulfur load are reasonable to try and have never been evaluated in anyone on this drug class. Persistent belching accompanied by severe abdominal pain, vomiting that prevents keeping fluids down, or a sudden change after months of stability is not the ordinary version and belongs with a prescriber. And every figure above was generated from branded product, while most sellers listed here dispense compounded versions, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before dispensing — so even the pharmacovigilance record does not cleanly describe what is in the vial. The bowel end of the same physiology is in the constipation article.

Frequently asked

Are sulfur burps a recognized side effect of GLP-1 drugs?
Belching is, under the clinical term eructation. It is the single most disproportionately reported gastrointestinal event for semaglutide, topping a list of 45 signals, and it carries the highest disproportionality for tirzepatide as well. The sulfur quality specifically has never been recorded in any study, because adverse-event coding has no field for how something smells.
Why do the burps smell like eggs?
The usual explanation is hydrogen sulfide, which sulfate-reducing bacteria produce in the colon from sulfur-containing substrate, much of it protein. That is a real literature about a real gas, but it describes the colon rather than the stomach, and no study has shown that colonic hydrogen sulfide leaves the body as a belch in anyone taking one of these drugs.
When do they usually start and do they go away?
The reporting data put median onset at 13 to 23 days depending on the dataset, with 67.1% of events in one tirzepatide analysis arriving within the first 30 days. Every disproportionality signal in the semaglutide analysis showed early-failure characteristics, meaning the risk concentrates at the start of treatment and declines with continued exposure.
Does eating less protein help?
It has never been tested in anyone on this drug class. In a crossover study of 11 healthy volunteers, a week of animal-based low-fiber eating produced more hydrogen sulfide than a week of plant-based high-fiber eating, by a median of 29 ppm/g. Two of the 11 went the other way, so the lever is not reliable even in the population where it was measured.
Is belching a reason to stop the drug?
Nothing in the published record suggests it signals harm on its own, and the reporting data describe it as an early effect that recedes. Belching that comes with severe abdominal pain, with vomiting that prevents keeping fluids down, or that starts abruptly after months of stability is a different situation and warrants contacting the prescriber.

Sources

  1. [1] Shu Y, He X, Wu P, et al. (2022). Gastrointestinal adverse events associated with semaglutide: A pharmacovigilance study based on FDA adverse event reporting system. Front Public Health. PMID 36339230
  2. [2] Shen P, Peng MS, Kim SJ, et al. (2026). Gastrointestinal adverse events associated with tirzepatide: A bibliometric and pharmacovigilance analysis. PLoS One. PMID 41894422
  3. [3] Guo X, Zhang J, Li Q, et al. (2026). Safety Profile of Tirzepatide in Real-World Clinical Practice: A Pharmacovigilance Study Using the FAERS Database. Diabetes Obes Metab. PMID 42443144
  4. [4] Lang IM (2016). The Physiology of Eructation. Dysphagia. PMID 26694063
  5. [5] Teigen L, Mathai PP, Lopez S, et al. (2022). Differential hydrogen sulfide production by a human cohort in response to animal- and plant-based diet interventions. Clin Nutr. PMID 35500315
  6. [6] Sievenpiper JL, Ard J, Blüher M, et al. (2026). Nutritional and lifestyle supportive care recommendations for management of obesity with GLP-1 - based therapies: An expert consensus statement using a modified Delphi approach. Obes Pillars. PMID 41502845

More in Safety