Two endpoints, two very different results, and the one that did not hold is the one the drug’s liver approval is written around — which makes the order they are read in the whole of what this page has to get right, since the indication is defined by a fibrosis stage.
What this is
A subgroup analysis. Of the first 800 people randomized in the interim efficacy analysis of the ESSENCE trial, 116 were Japanese — 78 given semaglutide and 38 given placebo. [1] They were older and lighter than the trial population as a whole.
That difference is why somebody looked, and it is also why the numbers should not be carried far. A subgroup is not designed to answer a question; it is a slice reported after the fact, and 38 people in a placebo arm is a narrow base for a proportion.
The endpoint that separated
Resolution of steatohepatitis with no worsening of fibrosis reached 63.1% on semaglutide against 36.8% on placebo. The estimated difference in responder proportions is 26.65 percentage points, 95% CI 7.43 to 45.86.
That is a wide interval around a large effect, which is what a subgroup of this size produces, and it clears zero comfortably — unlike the intervals that do not.
What “directionally consistent” means
That is the authors’ phrase for the whole comparison, and it is well chosen. Both point estimates favor the drug, and both are consistent with the overall trial.
It is not a claim that the subgroup confirmed anything. A subgroup that agrees in direction with a larger trial adds reassurance about generalizability and does not independently establish an effect — the difference between not detecting a difference and showing there is none, applied to a slice rather than a whole study.
Why it is worth reading anyway
Because trial populations are rarely the population buying the drug, and a subgroup that differs on age and body size is a small test of whether the result travels.
Here it traveled on one endpoint and was untested on the other. That is a more useful summary than either the headline or the caveat alone, and it is the same reading that the main liver evidence deserves — a drug with a real effect on inflammation whose effect on scarring is the harder, slower, less settled question.