What is the actual trial incidence of nausea by agent and dose, and how comparable are the numbers?

Every summary I read says nausea is "the most common adverse event" and then quotes a single percentage, usually without saying which trial, which dose or which comparator. That is not useful for setting expectations.

What I want is the incidence figures as they appear in the safety tables, agent by agent and dose by dose, with the placebo arm alongside them, because I assume a fair fraction of the reported nausea in any trial is background rather than drug.

Specific things I cannot resolve from the abstracts:

• Does nausea incidence actually rise with final dose within a trial, or does it flatten out?
• How much of the difference between agents is the molecule and how much is the escalation schedule the protocol happened to use?
• Are these figures "ever reported during 68 weeks" or "reported at any given time"? Because those are wildly different quantities and I suspect the headline numbers are the first.

A table with the numbers and the trial names attached would be more useful to me than another paragraph telling me nausea is common. I am aware that cross-trial comparison is not a head-to-head comparison; I would still rather see the numbers than not.

Adding the timing signature, since the incidence numbers say nothing about when in the week or the month symptoms actually land, and the pattern is specific enough to be diagnostic.

Within the dosing interval

For a weekly agent, nausea tracks the concentration profile with a short lag. Semaglutide reaches maximum plasma concentration somewhere in the one-to-three-day window after a subcutaneous dose and has a half-life of about a week. Tirzepatide has a half-life of roughly five days with a broadly similar time to peak. So the reported pattern is:

• Hours 0-12: usually nothing. If you feel sick within an hour of injecting, that is very unlikely to be pharmacological.
• Day 1 to day 3: the peak. Most people who report nausea report it here.
• Day 4 to day 5: tapering.
• Day 6 to day 7: usually the best days of the week, and for many people the days their appetite returns noticeably.

Daily agents lose this structure entirely, which is one reason liraglutide's symptom burden feels different from its incidence figure: the same number of people report nausea, but there is no good day in the cycle to recover in.

Across a dose step

The signature here is even more distinctive, and it is the thing worth learning to recognise:

1. Step the dose. Nothing much happens for the first day or two.
2. Symptoms appear or worsen over roughly days two to five, peaking in the first week to ten days at the new dose.
3. They plateau through week two.
4. They decline through weeks three and four, often substantially, so that the last week before the next step is the most comfortable.
5. Next step. The cycle repeats, usually with a lower amplitude than the previous one.

The decline in steps 4 and 5 is not you getting used to feeling sick. It is tachyphylaxis of the emetic response, which is a real, measurable, receptor-level adaptation, and it is the entire reason four-week escalation intervals exist. Four weeks is not arbitrary: with a one-week half-life, four weeks is four half-lives, so you reach 1 - (0.5 raised to the 4th) = 1 - 0.0625 = 93.75% of steady state at the current dose before you perturb it again. Stepping at two weeks would mean stepping at 75% of steady state, i.e. stacking an increment on top of an exposure that is still rising.

What does not fit the signature

If your symptoms do any of the following, the dose ladder is probably not the explanation and it is worth an actual clinical conversation rather than a titration adjustment:

• Appear more than four weeks into a stable dose, having previously been absent.
• Are flat across the whole week with no day-6 relief.
• Are worst on the day of injection, before the drug has meaningfully absorbed.
• Worsen progressively over months rather than settling after each step.
• Come with severe or radiating abdominal pain, fever, persistent vomiting, or an inability to keep fluids down. Those are not tolerability questions.

The last group matters because gallbladder disease and pancreatitis both occur on this class at low but real rates, and both can present as "the nausea got worse". A symptom that breaks the timing pattern deserves a different explanation from one that follows it.

A methodological caveat on the whole table above, because these numbers are softer than their decimal places suggest.

Ascertainment is not standardised. Nausea in these trials is a spontaneously reported treatment-emergent adverse event, coded to a preferred term. It is not measured with an instrument, not graded on a validated scale in most protocols, and not actively solicited at fixed intervals in the way an efficacy endpoint is. Whether a participant's "a bit queasy in the mornings" becomes a row in the safety table depends on whether they mentioned it, whether the site investigator recorded it, and how the coding dictionary handled it. Trials with more frequent contact ascertain more events. That alone can move a number by ten points.

Blinding is imperfect. An agent producing 40% nausea and 15% weight loss is functionally unblinded by week twelve for most participants. Symptom reporting after that point is not blind reporting, in either arm. The placebo arm's 17.4% in STEP 1 partly represents genuine background nausea and partly represents people who correctly deduced they were on placebo and reported accordingly, which pushes in the opposite direction to the usual assumption.

Populations were screened. Registration trials exclude people with significant gastrointestinal disease, prior pancreatitis, symptomatic gallstones, and often anyone on medications that slow transit. They also exclude people who cannot tolerate the run-in. Real-world incidence should be expected to exceed trial incidence for that reason alone, and pharmacovigilance data broadly bear that out.

Severity is what got lost. The vast majority of nausea events in every one of these trials were coded mild or moderate. The proportion coded severe is typically 1-3%. A table of ever/never incidence gives equal weight to a fortnight of feeling off after a dose step and to intractable vomiting requiring hospital fluids, which are not the same event in any sense that matters. If you are trying to predict your own experience, the more informative pair of numbers is the severe-event rate and the discontinuation-for-GI-events rate, both of which are low single digits on semaglutide and tirzepatide, rather than the headline incidence.

None of this makes the table wrong. It makes it a description of how often a symptom was mentioned in a screened population under partial blinding, which is a narrower claim than it looks.