Accepted answer
Check the SURMOUNT-2 inclusion criteria against yourself in that order: entry BMI band, diabetes status, prior weight-loss attempts, and what the run-in excluded. Registration programmes recruit a population selected to show an effect if one exists, which is the right design and a poor basis for generalising. The run-in is the part that is easiest to miss: a programme that drops people during a placebo lead-in has already removed those least likely to tolerate or comply, and the published arms describe the survivors. External validity is not a property of the trial; it is a property of the distance between its population and yours, and that distance is yours to measure.
The short version: the effect is real, the magnitude depends on the population, and the population is usually the part that gets dropped when a result is quoted second-hand.
Trial populations are selected. Exclusion criteria in this class routinely remove people with significant renal impairment, prior pancreatitis and unstable psychiatric illness, which is exactly the population the results are then quoted for.
Put another way, intention-to-treat and per-protocol analyses answer different questions. ITT asks what happens if you offer the treatment; per-protocol asks what happens if it is taken as directed. The gap between the two is a measure of how tolerable the protocol was.
Where a result is quoted from a conference abstract rather than a peer-reviewed publication, the numbers routinely move between the two. It is worth checking which one you are reading.
I am not a clinician and this is not medical advice; it is a reading of a published protocol.
When two sources disagree, the answer is almost always in the methods section of the one you have not read.