Nausea and vomiting are the effects a prescriber warns about, and constipation is the one that outlasts them. Diarrhea sits between the two and gets discussed as though it were interchangeable with either, which it is not: it ranks the molecules in close to the reverse order nausea does, it responds to dose in the opposite direction from constipation, and it is the one gastrointestinal effect with a downstream complication serious enough to appear in a warnings section.
What the labeled trials recorded, next to the placebo arm
The semaglutide 2.4 mg weight-reduction program pooled three randomized placebo-controlled trials. Diarrhea was reported by 30% of the 2,116 treated adults and by 16% of the 1,261 on placebo, which puts it second only to nausea at 44% against 16%.[1] The placebo figure is the half that rarely travels. Roughly one person in six taking nothing at all reported diarrhea over a 68-week trial, so a little more than half the symptom in the treated arm is the background rate of being a person for more than a year.
Tirzepatide was pooled across two trials with a dose split. Against 8% on placebo, diarrhea was reported by 19% at 5 mg, 21% at 10 mg and 23% at 15 mg.[2] The two labels use different pooling and different trials, so the two placebo rates — 16% and 8% — are not a finding about semaglutide against tirzepatide. They are a reminder that the denominator changes, which is why the head-to-head questions belong in the molecule comparison rather than in a subtraction between two package inserts.
The pooled estimate, and how certain it is
An umbrella review published in 2026 gathered 60 meta-analyses covering 1,751 randomized trials and 3,580,616 participants, then reanalyzed each under random effects. Diarrhea came out at an odds ratio of 1.94 (95% CI, 1.52 to 2.49), graded high quality of evidence — the same grade as nausea (OR 2.47) and a better grade than vomiting (OR 2.78, moderate).[3]
That ordering is worth holding onto. Diarrhea carries the smallest of the three effect sizes and one of the two strongest certainty ratings, while the largest effect size in the set rests on weaker evidence. The review also notes that between-study heterogeneity and the prediction intervals leave real residual uncertainty, which is a different statement from the confidence a single percentage implies.
Dose moves diarrhea and constipation in opposite directions
The tirzepatide label reports both symptoms over the same three doses in the same pooled population, which makes it the cleanest available test of the titration story. Diarrhea climbs: 19%, 21%, 23%. Constipation falls over those identical doses: 17%, 14%, 11%.[2]
So the sentence everyone repeats — that gastrointestinal effects worsen as the dose goes up — is true of diarrhea, vomiting and nausea and false of constipation, in one table, in one trial program. Anyone deciding whether to hold a dose because of a bowel complaint is making a different calculation depending on which end of the spectrum the complaint sits on, and the schedule those decisions attach to is in the titration article.
A dose-response network meta-analysis of 39 reports covering 33,354 adults with overweight or obesity and without diabetes adds the shape. For semaglutide, diarrhea risk rose fastest at the lowest doses and then tended to plateau; for liraglutide and orforglipron the increase looked closer to linear.[4] A plateau means most of the diarrhea risk is acquired during the starting doses, which is the opposite of how the nausea curve is usually described to a new patient.
The molecule ranking is not the nausea ranking
That same analysis estimated relative risks agent by agent. For diarrhea, three reached significance: tirzepatide at 3.35 (95% CI, 1.92 to 5.85), liraglutide at 1.82 (1.48 to 2.25) and semaglutide at 1.77 (1.47 to 2.14). Three did not: exenatide at 0.18 (0.02 to 1.34), cagrilintide at 1.20 (0.68 to 2.10) and orforglipron at 2.30 (0.89 to 5.93).[4]
Now set that against nausea in the same paper, where every agent examined raised the risk and orforglipron ranked highest at RR 4.77. The oral agent with the worst nausea signal produced no significant diarrhea signal, and tirzepatide — which sits at the bottom of the significant nausea list at RR 2.90 — tops the diarrhea list.[4] Six drugs acting on the same receptor do not produce the same stool, which is the clearest available argument against reading a single “gastrointestinal events” percentage as though it described a symptom.
Two molecule-specific meta-analyses restricted to adults without diabetes put numbers on the same gap from the other direction. Nine semaglutide trials covering 11,641 treated and 10,479 placebo participants gave diarrhea an odds ratio of 2.10, against 4.06 for nausea.[5] Six tirzepatide trials gave diarrhea a relative risk of 2.92 (2.53 to 3.37), against 3.11 for nausea.[6] On semaglutide, nausea is roughly twice the magnitude of diarrhea; on tirzepatide they are nearly level. The two papers report different measures — an odds ratio and a relative risk — so the numbers do not subtract, and a side-by-side reading of them overstates the semaglutide figures.
When it settles
STEP 1 named nausea and diarrhea together as the most common adverse events on semaglutide and described them as typically transient, mild to moderate, and subsiding with time. The trial also recorded what happened when they did not: 4.5% of the semaglutide group discontinued for gastrointestinal events, against 0.8% on placebo.[7] A rate that low over 68 weeks is the evidence that most of this resolves without stopping, and the 3.7-point gap is the share for whom it did not.
SURMOUNT-5, which randomized 751 adults with obesity and without type 2 diabetes to the maximum tolerated dose of tirzepatide or of semaglutide for 72 weeks, reported that gastrointestinal events were the most common in both arms, mostly mild to moderate, and occurred primarily during dose escalation.[8] A pharmacovigilance analysis of 123,145 tirzepatide reports adds a rough clock from the other side: among reports with usable onset data, the median time to an adverse event was 13 days, with 67.1% arriving within 30 days.[9]
None of that is a duration, and no trial has published one for this symptom. What the three sources agree on is the window: the first weeks of a dose, not the steady state. Diarrhea that begins for the first time in month six, on an unchanged dose, does not fit the pattern the trials described and is worth taking to the prescriber rather than absorbing.
The complication that makes this more than an inconvenience
Both labels carry a warning for acute kidney injury due to volume depletion, and both state that the majority of the reported events occurred in patients who experienced gastrointestinal reactions leading to dehydration. In the pooled tirzepatide weight trials, acute kidney injury was reported in 0.5% of treated patients against 0.2% on placebo.[2] The labels direct monitoring of renal function during initiation and escalation specifically.
That is the practical reason diarrhea is not simply the least pleasant item on a list. Several loose stools a day on top of an appetite that has collapsed is a fluid deficit arriving from two directions at once, and the people for whom it matters most are those already carrying reduced renal function — the group covered in the kidney article. An acute illness on top of that is its own situation, set out in what to do when you are sick.
What the compounded market adds, and what it does not measure
Every figure above comes from randomized trials of FDA-approved product at labeled doses. Most sellers on this site dispense compounded versions, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed, and no randomized trial has measured a gastrointestinal rate in one.
The closest available reading is a pharmacovigilance analysis of 81,078 GLP-1 reports, of which 707 involved compounded products. Compounded formulations carried higher reporting odds of diarrhea at 1.59 (95% CI, 1.25 to 1.99), abdominal pain at 2.84 and nausea at 1.27, alongside far higher odds of preparation errors (48.92), contamination (19.00) and hospitalization (2.35).[10] A spontaneous reporting database has no denominator and heavy channeling, so those ratios cannot produce an incidence. They also cut both ways in one place worth noting: the same analysis found compounded products carried lower odds of administration errors (0.29) and dosing errors (0.24).
What none of this decides
Diarrhea on this class is common, is worse on tirzepatide than on semaglutide by the pooled estimates, concentrates in the weeks after a dose change, and resolves without discontinuation in the large majority. It is also reported by one placebo participant in six, which is the figure that should temper any personal attribution. The full tolerability picture it belongs to is in what the trials recorded, and the pricing and seller questions that sit underneath the choice are in the seller write-ups.
Bloody stool, fever, severe abdominal pain, or diarrhea that produces lightheadedness or a day without urinating is not the ordinary version of this and belongs with a prescriber the same day. So does diarrhea severe enough that fluids cannot be kept down, because that is the exact combination the volume-depletion warning describes.