Dose and Side Effects: What the Labels Show When the Placebo Is Taken Out

Every GLP-1 starts low and climbs. Every label gives the same reason for it, and it is not effectiveness — the escalation exists to limit adverse reactions. All twelve approved schedules, and the shared wording behind them, are collected on our titration schedules page.

That raises an obvious question the schedules themselves never answer: how much difference does the dose actually make? This page answers it from the only documents that can, which are the approved labels' own placebo-controlled tables, and the first finding is about which labels those are.

Which labels publish the gradient

A correction first. From 21 to 23 September 2026 this section said that only three current US GLP-1 labels report adverse reactions by dose, all of them for diabetes, and that no weight-management label does. That was wrong. Two of the copies we read were years-old repackager versions returned first by FDA's openFDA search — the Wegovy one dated April 2024 — and the census also missed labels we had read correctly elsewhere on this site. Re-read on 23 September 2026 from the manufacturers' current labels, the picture is close to the opposite:

So the dose gradient for the weight-loss indication can be read from approved labels, and it turns out to have a different shape from the diabetes one. Both are below.

What the diabetes tables show

Rates are the percentage of patients reporting the reaction at least once, from each label's placebo-controlled pool.

Subtract the placebo column, and the shape appears

A raw rate mixes the drug's effect with whatever the trial population reports anyway. Subtracting the placebo rate leaves the excess attributable to the drug, and dividing one dose's excess by the other's says whether the reaction scales with dose. Both operations are ours; the labels publish only the raw percentages.

Three things follow.

Where the dose doubles, nausea and vomiting more than double. Dulaglutide's excess vomiting rises 2.81-fold on a 2-fold dose step; nausea on the semaglutide tablets rises 2.80-fold. This is supra-proportional, and it is the quantitative fact behind every label's insistence on a slow climb. When those episodes happen and how many days each one lasts are on the nausea timing page.

The effect is concentrated in the gut. Outside the gastrointestinal reactions, the only clear step in these three diabetes tables is dulaglutide's fatigue. How each label defines that fatigue row — one term on some, four on others — is on our page on whether GLP-1 makes you tired. Liraglutide's nasopharyngitis moves 8, 9, 10 and its upper respiratory tract infection 6, 7, 6 — the movement of background illness in a trial population, not of a drug effect.

And it is not universal. Liraglutide's 1.5-fold step lifts excess nausea by only 15%, far less than the dose. Whether that is a property of the molecule, of its daily schedule, or of the fact that both arms are already well past the threshold where nausea appears, these tables cannot say.

The weight-management labels: a much flatter shape

The same arithmetic on the three weight-management labels that publish dose columns gives a different answer. Excess is the drug rate minus the same label's placebo rate; the factor compares the highest dose with the lowest.

Zepbound from its Table 1 (placebo nausea 8%); Foundayo from its Table 1 (placebo nausea 10%); Wegovy from its Table 4, the two 72-week trials with 7.2 mg, 2.4 mg and placebo arms (placebo nausea 13%, dysesthesia 0.3% in the text).

Where the dose triples, excess nausea rises by a fifth to a half. That is the opposite of the diabetes pattern above, where a doubling more than doubled it. Zepbound's share reporting any gastrointestinal reaction is 56% at all three doses; what rises is the share who stopped because of one, from 1.9% to 4.3%.

One design difference probably explains most of the gap, and the Trulicity label states it. Its placebo-controlled pool "evaluated TRULICITY 0.75 mg and 1.5 mg without dose escalation" — people were started directly on the dose they stayed on. Every arm in the weight-management trials climbed to its dose over weeks or months. A table that starts people at full strength measures the cost of the dose; a table built on escalation measures the cost of the dose after the climb has absorbed much of it. The labels do not make this comparison, and nothing here proves the escalation is the cause. It is, though, the one difference between the two sets that the documents themselves record.

The steepest gradient in a current weight label is not in the gut. Wegovy's dysesthesia — tingling, burning, skin pain and sensitivity — went from 6% to 22% across the same step that moved nausea from 35% to 39%, and the label says its incidence "increased with increasing dosage and drug levels in the blood". Our Wegovy side effects page sets out what happened to the people who had it. The recovery, dose-reduction and recurrence figures behind that 22% are on skin sensitivity on GLP-1 drugs.

Three reactions that go the wrong way

The most informative rows in the set are the ones that fall.

• Constipation, semaglutide tablets: 2% placebo, 6% at 7 mg, 5% at 14 mg.
• Decreased appetite, VICTOZA: 1% placebo, 10% at 1.2 mg, 9% at 1.8 mg.
• Headache, VICTOZA: 7% placebo, 11% at 1.2 mg, 10% at 1.8 mg.

Each of these is well above placebo at both doses, so the drug is doing something. But none rises with the dose, and one of them is decreased appetite — the effect the drug is taken for, showing no gradient across a 50% dose increase in this pool.

Single percentage points in tables of a few hundred patients are inside the noise, and none of these differences should be read as a real decline. Read correctly they say something more useful: a reaction that does not track the dose is not being driven by the dose. That is a different kind of fact from a rate, and it is invisible unless the label publishes both columns. Zepbound's constipation, which halves across its doses, is the clearest example in a weight-management label.

What this page does not establish

These are separate placebo-controlled pools in different populations — diabetes and weight management, fixed-dose and escalated — and adverse reaction rates from one drug's trials cannot be compared with another's. Nothing here transfers to a dose outside those tabulated, or to any individual.

The excess-over-placebo figures and the factors are arithmetic on published percentages, not a reanalysis of patient data; they carry no confidence intervals, because the labels publish none, and they should be read as the shape of a relationship rather than as estimates.

Nothing on this page is guidance about what dose anyone should be on. That decision belongs to the prescriber who wrote the prescription, and our editorial standards explain why we leave it there.

Sources and dates

• TRULICITY prescribing information, Table 1, adverse reactions in the placebo-controlled pool, SPL effective 2023-08-05 — read 2026-09-21 via openFDA.
• OZEMPIC and RYBELSUS prescribing information, semaglutide tablets adverse reaction table, SPL effective 2026-01-30 — read 2026-09-21 via openFDA.
• VICTOZA prescribing information, adverse reactions in placebo-controlled trials — first read 2026-09-21 via openFDA from a 2023-02-06 repackager copy; re-read 2026-09-23 against Novo Nordisk's current label (effective 2025-10-14, DailyMed), Table 1 unchanged.
• SAXENDA prescribing information, SPL effective 2026-02-25 — read 2026-09-21 via openFDA.
• Added 2026-09-23: WEGOVY prescribing information (Novo Nordisk, SPL effective 2026-06-18, via DailyMed), Tables 3 and 4 and the dysesthesia text; ZEPBOUND (2026-08-28) and FOUNDAYO (2026-07-29) Table 1 via openFDA; current TRULICITY (Eli Lilly, 2026-06-16) and OZEMPIC injection (Novo Nordisk, 2026-06-01) via DailyMed; MOUNJARO (2026-08-27, Eli Lilly's copy, re-checked 2026-09-24) via openFDA.
• Corrected 2026-09-23: the Trulicity copy first read (2023-08-05) and the Wegovy copy (2024-04-23) were repackager versions returned first by an openFDA brand-name search, not the manufacturers' current labels. The Trulicity placebo-pool figures above are unchanged in the current label.