Two trials of the same molecule, using the same absorption test, reach opposite conclusions about whether it slows the stomach at all. At 1.0 mg after twelve weeks, first-hour emptying was slowed to a ratio of 0.73 against placebo.[1] At 2.4 mg after twenty weeks, there was no first-hour effect and five-hour absorption came out 8% higher than placebo, which the investigators summarized as no evidence of delayed gastric emptying at week 20.[2] The mechanism blamed for every difficult dinner is largest at the beginning and smallest by the time most people reach the dose they stay on.
What the two trials actually measured
Both used paracetamol absorption, which tracks how fast stomach contents reach the small intestine. The first was a randomized, double-blind, placebo-controlled crossover trial in 30 adults with obesity, each taking once-weekly semaglutide escalated to 1.0 mg or placebo for twelve weeks before a standardized meal test. First-hour gastric emptying was delayed, at an estimated treatment ratio of 0.73 (95% CI, 0.61 to 0.87). Overall emptying across five hours was not statistically different between treatments.[1]
The second was a double-blind parallel-group trial in 72 adults with obesity, randomized to semaglutide escalated to 2.4 mg or placebo for twenty weeks. Five-hour paracetamol area under the curve, the primary endpoint, was 8% higher on semaglutide (P = 0.005) — and the paper is careful about that number, reporting it as non-significant once corrected for week-20 body weight (P = 0.12). There was no effect on first-hour absorption, on peak concentration or on time to peak. Meanwhile appetite fell, fullness and satiety rose, and control of eating improved (all P < 0.02).[2]
Read together, those describe a delay that is real, front-loaded within a meal, and fading with continued exposure. The broader argument about how much the stomach slows and which instrument you believe belongs to the gastroparesis article, which is a different question from what happens at a table.
The fade is not complete while the dose is still climbing
Tirzepatide was measured the same way, and the timeline separates two kinds of patient. In diet-induced obese mice, tirzepatide delayed gastric emptying much as semaglutide did, and those acute effects were abolished after two weeks of treatment. In a phase 1 four-week multiple-dose study, once-weekly tirzepatide delayed emptying after a single dose, and the effect diminished after multiple doses in healthy participants. In participants with type 2 diabetes following an escalation schedule of 5/5/10/10 or 5/5/10/15 mg, a residual delay was still observed after multiple doses.[3]
That is the practical distinction. Somebody who has been at the same dose for six months has a stomach behaving close to normal on this measurement. Somebody who stepped up two weeks ago has restarted the part of the curve where the effect is biggest. A review of the clinical consequences of delayed emptying adds a third group: the effect is limited in people whose gastric emptying was already slow before treatment started, so the delay is not simply additive on top of an existing problem.[4]
The calendar predicts the evening better than the menu
Pooled tolerability data put a shape on when trouble arrives. Across STEP 1 to 3, 2,117 participants on semaglutide 2.4 mg were compared with 1,262 on placebo over 68 weeks. Nausea was reported by 43.9% against 16.1%, diarrhea by 29.7% against 15.9%, vomiting by 24.5% against 6.3%, and constipation by 24.2% against 11.1%. 99.5% of those events were non-serious and 98.1% mild to moderate, they were transient, and they occurred most frequently during or shortly after dose escalation. 4.3% of treated participants permanently discontinued for gastrointestinal reasons.[5]
The tirzepatide program reports the same timing from a different angle. Across SURMOUNT-1 to -4, gastrointestinal events were reported by 27.8% to 72.8% of treated participants against 12.2% to 32.5% on placebo, most of them during dose escalation, and first use of antiemetic and antidiarrheal medication was most commonly reported during dose escalation as well.[6]
A holiday is a fixed date. A titration step is a movable one, and the titration article sets out where the steps fall. Two events landing in the same week is the situation both datasets describe, and how long the nausea itself usually runs is in the nausea article.
The meal itself is larger than the word “meal” suggests
Restaurant portions have been measured rather than estimated. A multi-country study burned 223 meals from 111 randomly selected full-service and fast-food restaurants in five countries in a bomb calorimeter. The weighted mean energy of a United States restaurant meal was 1,088 kcal (95% CI, 1,002 to 1,181), and only China came out lower, at 719 kcal (646 to 799; P < 0.001).[7]
One result inside it reverses the usual assumption. Fast food contained 33% less energy than full-service meals (P < 0.001), and in Finland worksite canteens provided 25% less than either, at 880 kcal against 1,166.[7] The sit-down dinner, not the drive-through, is the large meal. Against a drug whose measured effect is to cut intake at a single test meal by about a third — the figure and its limits are in the eating article — a 1,088-kcal plate is not an indulgence above the baseline. It is the baseline.
Alcohol at the table is a physiology question nobody has asked on this drug
Whether these drugs reduce drinking is a separate matter with randomized trials behind it, covered in the alcohol article. What happens to a drink taken with a meal is a mechanical question, and the only direct measurement was made in people not taking any of this.
Eight healthy young adults, mean body-mass index 22.7, drank 600 mL of full-strength, low-carbohydrate or low-alcohol beer on three separate occasions with concurrent scintigraphic measurement of gastric emptying and plasma ethanol. Half-emptying times did not differ across the three beers (89.0, 79.5 and 74.6 minutes; P = 0.39). What did vary was the relationship between them: plasma ethanol at fifteen minutes ran inversely to the emptying measure, strongly for the low-alcohol beer (r = −0.87, P < 0.01) and as a trend for the other two. The authors’ conclusion was that gastric emptying is a determinant of the plasma ethanol response.[8]
That is eight people with no GLP-1 exposure, and it settles nothing about anyone on this class. What it establishes is only that the rate at which the stomach empties is one of the things that sets how fast alcohol arrives — which makes a drug that changes that rate, unevenly and mostly early, a reasonable thing to raise with a prescriber rather than a reason to write a rule that no study supports.
The sickness is not the mechanism, and both sponsors measured it
A widespread belief is that a bad reaction to a large meal is the drug working. Two mediation analyses, each run by the company that makes the drug, say otherwise. In the semaglutide pooling, mean weight loss was similar in participants without gastrointestinal events (9.6% to 17.1%) and with them (11.4% to 17.7%). Of the additional 7.6% to 14.4% weight loss over placebo, less than one percentage point was mediated by gastrointestinal adverse events.[5]
The tirzepatide analysis reached the same place with a slightly larger number: weight reduction was similar among participants reporting no nausea, nausea alone, or any nausea, vomiting or diarrhea, and mediation attributed up to 3.1% of total weight reduction to those events and dyspepsia.[6] Under both estimates, almost all of the effect happens without the symptom. What the symptom does mark is the threshold where tolerability becomes a safety question, which is where the vomiting article picks it up.
What a compounded prescription changes here
Every rate above came from a trial of an FDA-approved product given on a published escalation schedule. Compounded semaglutide and tirzepatide are not FDA-approved, and the FDA does not review them for safety, efficacy or quality before they are dispensed. The physiology is the same peptide, so nothing above stops applying — but the single strongest predictor of a bad evening in both datasets is position in a dose-escalation schedule, and a compounded vial arrives without the prescribing information that defines one.
That makes the schedule a question to ask a seller directly: what the steps are, how long each is held, and who decides when to move. Sellers that publish a schedule and sellers that leave it to a coaching chat are answering different questions, and which is which is recorded in the provider write-ups.
Where this leaves a dinner reservation
Nothing here is dietary advice, and nothing here says what to order or whether to drink. What the evidence supports is narrower and more useful: the stomach effect is largest early and fades with steady exposure, it restarts with each dose step, difficult episodes cluster around those steps rather than around food, the ordinary restaurant plate is already about 1,088 kcal before anyone calls it a big meal, and feeling terrible afterward contributes almost nothing to the weight outcome. A specific meal, a specific dose week and a specific drink are a conversation for the prescriber holding the schedule.