The Journal of Gastroenterology published a subgroup analysis on 28 August 2026 looking at the 116 Japanese participants inside ESSENCE, the phase 3 trial of semaglutide in metabolic dysfunction-associated steatohepatitis.

One of the two primary endpoints separated clearly. The other did not.

That reads like a finding about the liver. It is almost certainly a finding about arithmetic, and the parent trial is what shows the difference.

A framing note. Semaglutide is an approved medicine, this is a study of patients, and semaglutide is not in the Peptra Labs catalogue. Nothing below is guidance about human use.

What ESSENCE is

ESSENCE is an ongoing phase 3, multicentre, randomised, double-blind, placebo-controlled trial, registered as NCT04822181 and funded by Novo Nordisk.

It enrolled 1,197 patients with biopsy-defined MASH and fibrosis at stage 2 or 3, randomised 2:1 to once-weekly subcutaneous semaglutide 2.4 mg or placebo for 240 weeks. A planned interim analysis at week 72 covering the first 800 randomised patients was published in the New England Journal of Medicine on 5 June 2025 by Sanyal, Newsome and colleagues, and is referred to as part 1.

Part 1 had two primary endpoints, both read from liver biopsy. Resolution of steatohepatitis without worsening of fibrosis. Reduction in fibrosis without worsening of steatohepatitis.

In the full part 1 population, both separated. The authors report resolution in 62.9 percent versus 34.3 percent, an estimated difference of 28.7 percentage points with a 95 percent confidence interval of 21.1 to 36.2. For fibrosis they report 36.8 percent versus 22.4 percent, a difference of 14.4 points, interval 7.5 to 21.3. Both at P below 0.001.

Hold on to that second row.

What the Japanese semaglutide subgroup found

Of the first 800 randomised participants, 116 were Japanese: 78 received semaglutide and 38 received placebo. The authors report that compared with the overall population, the Japanese participants were older and had a lower body mass index.

Week 72 endpoint, Japanese subgroupSemaglutidePlaceboEstimated difference
Resolution of steatohepatitis, no worsening of fibrosis63.1 percent36.8 percent26.65 points, CI 7.43 to 45.86
Fibrosis improvement, no worsening of steatohepatitis36.5 percent28.9 percent7.07 points, CI -11.48 to 25.62

The first interval sits entirely above zero. The second crosses it.

On tolerability, the authors report that semaglutide was well tolerated in the Japanese population based on the adverse event profiles of all treated participants, and that age and body mass index distribution did not meaningfully affect efficacy or safety. Their conclusion is that results were directionally consistent with the overall ESSENCE population.

The comparison the paper invites

Put the two sets of numbers side by side and something specific appears.

On the semaglutide arm, the Japanese figures are almost identical to the overall trial: 63.1 against 62.9 percent for resolution, 36.5 against 36.8 percent for fibrosis. Those are the same results to within rounding.

The placebo arm is where the two populations differ. Japanese placebo participants reached resolution 36.8 percent of the time against 34.3 percent overall, and fibrosis improvement 28.9 percent against 22.4 percent.

So the shrinking difference on the fibrosis endpoint did not come from the treated patients doing worse. It came from the untreated patients doing better, in a comparison group of 38 people.

A wide interval is not a negative result

The fibrosis interval in the subgroup runs from minus 11.48 to plus 25.62. That is a span of 37 percentage points.

An interval that wide is compatible with a moderate harm, with no effect, and with a benefit larger than the one the full trial measured. It excludes almost nothing. Reporting it as “no effect on fibrosis in Japanese patients” would be a straightforward misreading, and reporting it as evidence that the compound works differently by population would be worse.

The full trial found a 14.4 point difference on this endpoint. A subgroup of 116 people, with 38 in the comparison arm, does not have the resolution to detect a 14 point difference. The result is what an underpowered slice of a positive trial looks like, and the authors say as much in their own careful wording when they call the findings directionally consistent.

This matters beyond one paper. Subgroup analyses are published constantly, they are read as though each one is an independent test, and a great many of them are too small to answer the question they appear to be asking.

Who wrote it

Three of the eight authors list Novo Nordisk affiliations, at the company’s sites in Denmark and Japan, and the parent trial is company funded. Nakajima, Okanoue, Hiasa and Takahashi are hepatologists at Japanese academic centres, and Sanyal, the senior author, led the parent publication.

None of that invalidates anything. It is standard for a registration programme, and the disclosure is in the paper. It is worth stating plainly, because subgroup analyses of company trials are the part of the literature where framing choices carry the most weight, and the framing here is conservative rather than promotional.

How this differs from yesterday

We reported yesterday’s comparison against sleeve gastrectomy, where surgery removed far more liver fat than the drug did and the relative reduction in liver stiffness came out with a p value of 0.428. Two results, both showing one endpoint separating while another does not. They are not the same kind of result.

That study was adequately sized for the comparison it made, used two readouts from a single instrument, and found a genuine divergence between fat and stiffness. This one is a small slice of a larger trial in which both endpoints separated, and the slice lost the smaller of the two effects. The first is a signal about biology. The second is a signal about sample size.

Telling those apart is most of the work in reading this literature. Both look like “one thing moved and the other did not” in a headline.

The same distinction applies to a 52-week MASH biopsy trial we covered earlier this month, where a combination arm beat monotherapy on every histological number and still missed its primary endpoints, and to a matched cohort where hard liver outcomes came out level despite one arm losing more body mass index.

What this means for the catalogue

Nothing in ESSENCE involves a compound we supply, and the useful part for a laboratory is the reading habit.

Preclinical work on research compounds is almost entirely small. Animal studies with eight or ten per group produce confidence intervals wider than this one, and they are routinely described in secondary coverage as showing that a compound did or did not do something. The correct response to a wide interval is to say that the study could not tell, which is a different statement from either result.

Retatrutide and tirzepatide both have hepatic literature at a much earlier stage than this, which means the gap between what is measured and what is claimed is correspondingly wider for them.

For laboratories, our European research buyer guide covers procurement and documentation, everything is supplied on a research use only basis, and our ranking of the most-studied research peptides tracks the published record for each compound.

References

The products referenced on this site are supplied for laboratory research use only. They are not medicines and are not intended for human or veterinary use. This article summarises published research for informational purposes and is not medical advice. Statements about third-party studies belong to their authors.

author-avatar

About Peptra Labs Research

The Peptra Labs research desk follows peptide science: new peer-reviewed studies, EU and US regulatory decisions, and clinical trial results. Every article cites its primary sources. All compounds discussed are for laboratory research use only.

Newer
BPC-157 Grown in Bacteria: 82.9% Ulcer Inhibition in Rats
Back to list
Older Tirzepatide CV Approval: The Overlooked 0.92 Hazard Ratio

Related Posts