A trial names its primary outcome before it starts. That is the whole point of naming one. This trial named lab cue-elicited craving at week 6, and the drug did not move it.
Everything reported as a win here came from somewhere else on the list. [1]
What was and was not significant
Fifty adults with moderate-to-severe alcohol use disorder took oral semaglutide or placebo for eight weeks. The dose ran at 3 mg daily for four weeks, then 7 mg daily for four. Eight weeks is a short window even for the outcome this class moves fastest.
Craving in the lab did not differ. Drinks per day did not differ. Those are the two outcomes a reader would most expect a drinking trial to report.
Heavy drinking days fell, b −0.580 with an interval of −1.012 to −0.148. Drinks per drinking day fell, −1.177, interval −2.307 to −0.047. Craving reported in daily life fell, −2.195, interval −4.174 to −0.216. Cannabis use days fell too, and so did alcohol-related consequences.
Why the ordering matters
A trial with many outcomes will usually produce a significant one. That is arithmetic, not evidence. The primary endpoint exists to stop that arithmetic from doing the talking.
So the honest reading is narrow. The prespecified test failed, and a consistent pattern showed up across several secondary measures — which is a reason to run a bigger trial, and not a reason to conclude anything yet. It is the same problem as a top ranking resting on the thinnest evidence in the set.
What moved and what it was made of
Craving in daily life was significant. Craving in the lab was not. Those two measure the same word in very different ways, and the one that moved is the one built from what people said about themselves.
That does not make it wrong. It does mean the measurement carries a reporting step, the way a coded side effect carries one.
How it sits next to the other habits
GLP-1 drugs keep producing results like this across addictive behaviors. Some are stronger. A large cohort found a much higher smoking quit rate, and an honest account of that field also has to include the trials that found nothing.
The authors of this one wrote that continued development is warranted. That is the correct verb for a phase 2 trial that missed its primary endpoint, and nobody selling anything should upgrade it.