Nausea Rose From 32% to 50% Across Three Liraglutide Doses in One Trial — of People With Type 1 Diabetes

Our dose-gradient article compared dose arms across eight different trials and found that mixing drugs into the comparison inflates the apparent effect of dose. One trial in our corpus does not need that correction at all: three liraglutide dose arms, one protocol, one population, one results table. It just is not the population most readers of this site have in mind.

ADJUNCT ONE (NCT01836523, Mathieu et al., Diabetes Care, October 2016) randomized 1,398 adults with type 1 diabetes 3:1 to once-daily liraglutide — 0.6 mg, 1.2 mg or 1.8 mg — or placebo, added to their existing insulin therapy, over 52 weeks, double-blind. It was designed to test whether liraglutide on top of insulin improves glycaemic control and reduces insulin dose and body weight in type 1 diabetes; it was not a weight-management trial in the sense every other study in our corpus is, and its patients were not there for GLP-1 therapy on its own.

It entered our corpus on 27 September 2026 as part of a batch of pinned registry trials, and every one of its 22 stored rows was re-verified against ClinicalTrials.gov's own results record the same day: arm, term, numerator and denominator all match exactly. The placebo arm (348 participants in the registry's safety population) is excluded from our corpus by design, per our methodology — we pool active-drug rates only.

Three arms, one trial, a clean dose gradient

Non-placebo arms only, all liraglutide, all added to insulin, all read verbatim from the same results table:

Each percentage-point change and z-value is computed from the raw counts (not the rounded percentages), as a two-proportion test between the 0.6 mg and 1.8 mg arms, Bonferroni-corrected for testing 7 effects at once (critical |z| ≈ 2.69, versus the usual 1.96 for one test on its own). Three effects — nausea, vomiting, reduced appetite — clear that corrected bar comfortably. Diarrhoea clears the uncorrected 5% line but not the corrected one, so we call it suggestive rather than established, the same hedge our SURMOUNT-5 piece used for a similar borderline gap. Constipation, abdominal pain and fatigue move by 3 points or less and are not distinguishable from sampling noise at these arm sizes — fatigue, notably, does not even move monotonically (it peaks at the middle dose), which is what a null usually looks like.

Headache is not in this table, and the reason is ours, not the registry's. Our store carries a single headache row for this trial — 0.6 mg, 16.0% (56 of 350) — because our collector sends at most 80 new registry rows to extraction per run, and ADJUNCT ONE's two upper-dose headache rows were the 81st and 82nd in the queue on 27 September 2026; they are due on the next run. The registry itself posts all three arms: 56 of 350 (16.0%) at 0.6 mg, 46 of 348 (13.2%) at 1.2 mg and 49 of 347 (14.1%) at 1.8 mg — read directly from ClinicalTrials.gov on 27 September 2026, not yet in our store, and therefore kept out of the table above, which is restricted to rows verified in our production data. On those registry figures headache does not rise with dose (z = −0.69, 0.6 versus 1.8 mg); our headache article sets out what placebo arms report for this effect.

Why this belongs in an article, not just a pooled number

These 22 rows already count toward our corpus's whole-class clinical rates — nausea's pooled estimate, for instance, rests on dozens of distinct studies, this one among them. That is not wrong: a liraglutide adverse-event rate is a liraglutide adverse-event rate whatever the background therapy, and our published method pools whole-class rates and discloses the drug and dose-tier mix behind them, not the indication. But it is incomplete. Our predictor and methodology disclose which drugs and dose tiers sit behind a pooled figure; neither currently says which indications the pooled trials were run in — obesity, type 2 diabetes, cardiovascular-outcome, and now, with this trial, type 1 diabetes on background insulin. A reader using our tool to think about their own weight-management side-effect risk has no way to see that a slice of the evidence behind it comes from a different population on a different background therapy. We have filed a proposal to tag pooled rates by indication and publish that mix alongside the existing drug mix; until it ships, this article is where that population is disclosed for this specific trial.

What this trial actually found, and why gut rates are not the headline

We are reporting gastrointestinal side-effect rates because that is what our site tracks. It is not what made ADJUNCT ONE clinically notable. The trial's own published finding was that liraglutide added to insulin in type 1 diabetes increased rates of symptomatic hypoglycaemia (in all three liraglutide groups, though the 0.6 mg estimate's confidence interval crosses 1) and, at 1.8 mg, of hyperglycaemia with ketosis, alongside its benefits to HbA1c, insulin dose and body weight — a safety signal serious enough that it has limited liraglutide's use as an insulin adjunct in this population. Neither hypoglycaemia nor diabetic ketosis is among the 15 effects our corpus tracks (built around the weight-management questions this site answers), so nothing on this site, including this article, measures that signal. If you found this article searching for liraglutide's safety in type 1 diabetes specifically, the gut-effect table above is real but is not the trial's main safety story — read the primary source linked below for that.

Honest limitations

Not a weight-management trial. Every other study behind our pooled clinical rates enrolled people for obesity, type 2 diabetes or cardiovascular risk on a GLP-1. This one enrolled people with type 1 diabetes already on insulin. The gut-effect biology of liraglutide may not differ by indication, but we have not measured whether it does — we are disclosing the population, not asserting the rates transfer.

Headache is one arm, not three, for the reason stated above.

No placebo comparison in this article. A placebo arm exists in the trial and is excluded from our corpus by design; we have not pulled its counts for this piece, so we cannot say how much of any arm's rate is background symptom burden versus drug effect.

52 weeks, titration and maintenance together. Like our other single-trial pieces, these rates span the full treatment period, not an isolated maintenance dose.

One trial. ADJUNCT TWO (NCT02098395), a second, smaller liraglutide-in-type-1-diabetes trial, is queued in the same pinned-trial batch and had not yet reached our store as of this writing.

Every figure in the table above was read from ADJUNCT ONE's ClinicalTrials.gov results record and verified against our production store on 27 September 2026, the day this piece was first published; the three headache figures are registry-only, as stated. These rows already count toward our corpus's pooled estimates at the data API, with the full eligibility methodology at magistra.health/en/methodology. Trial registration: clinicaltrials.gov/study/NCT01836523. Primary publication: Mathieu C, et al. "Efficacy and Safety of Liraglutide Added to Insulin Treatment in Type 1 Diabetes: The ADJUNCT ONE Treat-To-Target Randomized Trial." Diabetes Care, 2016;39(10):1702-1710.