The result that changed a label has not finished being measured. ESSENCE is a 240-week trial in 1,197 people, and what has been published is a planned interim analysis conducted at week 72 in the first 800 of them. The endpoints it reported are histological — what a pathologist sees on a slide — and the endpoint the trial exists to settle, cirrhosis-free survival, sits in a second part that is estimated to complete in April 2029. That is not a criticism of the trial, which is running exactly as designed. It is a description of what a reader is holding when they read the number. The clinical question of whether a GLP-1 belongs in fatty liver disease at all, and how these histology endpoints relate to one another, is worked through in the fatty liver article. This page is about how ESSENCE itself is built.
Two parts, one denominator each
ESSENCE is a two-part, phase 3, multicenter, randomized, double-blind, placebo-controlled trial. It assigned 1,197 patients with biopsy-defined metabolic dysfunction-associated steatohepatitis and fibrosis stage 2 or 3, in a 2:1 ratio, to once-weekly subcutaneous semaglutide 2.4 mg or placebo for 240 weeks. Part 1 is the planned interim at week 72 in the first 800 randomized: 534 on semaglutide and 266 on placebo.[1]
Part 1 had two primary endpoints, both histological. Part 2’s primary endpoint, recorded in the trial registry, is cirrhosis-free survival from randomization to week 240, and the study is listed as active and no longer recruiting with an estimated completion date of 25 April 2029.[2] The design paper describes the split in the same terms: Part 1 is about liver histology, Part 2 is based on clinical outcomes.[3]
The population that produced the interim is sicker than a general weight-loss cohort. Among those first 800, 250 (31.3%) had fibrosis stage 2 and 550 (68.8%) had stage 3. Mean age was 56 years (SD 11.6), 57.1% were female, mean body-mass index was 34.6 (SD 7.2), and 55.5% had type 2 diabetes.[3] Roughly seven in ten carried bridging fibrosis, and more than half carried a diabetes diagnosis alongside it.
Both endpoints, and what they mean in people
Resolution of steatohepatitis without worsening of fibrosis occurred in 62.9% of the semaglutide group against 34.3% of the placebo group — an estimated difference of 28.7 percentage points (95% CI, 21.1 to 36.2; P < 0.001). A reduction in liver fibrosis without worsening of steatohepatitis occurred in 36.8% against 22.4% — a difference of 14.4 percentage points (95% CI, 7.5 to 21.3; P < 0.001).[1]
Invert those differences and they become the number of people who have to be treated for one to benefit. Roughly one in three and a half for resolution; roughly one in seven for the fibrosis endpoint. Both are strong results by the standards of this disease, and both mean that most people treated did not get the benefit the endpoint measures.
The third figure is the one a patient would ask for. Combined resolution and fibrosis reduction — both at once — occurred in 32.7% against 16.1%, a difference of 16.5 percentage points (95% CI, 10.2 to 22.8).[1] Two-thirds of the treated group did not achieve both.
The placebo column, read on its own
More than a third of the placebo group met the resolution endpoint, and more than a fifth met the fibrosis endpoint, on a placebo injection.[1] Some of that is real change in a disease that fluctuates, some is the effect of trial participation itself, and some is measurement. It is the reason ESSENCE’s headline percentage cannot be read as the drug’s effect: the difference between the columns is the effect, and the difference is roughly half the headline in the case of resolution and well under half of it for fibrosis.
The endpoint is a pathologist’s reading
Both primary endpoints are scored from biopsy slides, and readers of those slides agree with each other less than the endpoints suggest. A study digitized 678 biopsies from 339 patients with paired samples in a separate steatohepatitis trial and had three hepatopathologists score them independently and blinded. Inter-reader unweighted kappa was 0.396 for resolution without worsening fibrosis and 0.366 for fibrosis improvement without worsening steatohepatitis — the same two composites ESSENCE uses.[4]
The same analysis found that 46.3% of patients admitted to that trial on one pathologist’s qualifying read were judged by at least one of the three readers not to meet the histologic entry criteria, and its simulations showed endpoint unreliability driving study power from above 90% down to as low as 40%.[4] That work was not done on ESSENCE and says nothing about ESSENCE’s own reading quality. It says that the instrument is noisy, and noise of this kind attenuates a measured effect rather than inventing one — which makes ESSENCE’s intervals wider in spirit than they look on the page.
Practice has moved in the same direction. The 2025 update to the AASLD practice guidance recommends selecting patients for semaglutide using non-invasive tests rather than biopsy, giving thresholds on transient elastography, magnetic resonance elastography and the enhanced liver fibrosis score, and calling biopsy impractical and unnecessary for most patients.[5] The trial ran on an instrument that the guidance built on it no longer asks clinicians to use.
The secondary that did not move
Three secondary outcomes were included in the plan adjusting for multiple testing. Two produced results: the combined histology endpoint above, and a mean body-weight change of −10.5% against −2.0%, a difference of 8.5 percentage points (95% CI, −9.6 to −7.4). The third did not: mean changes in bodily pain scores did not differ significantly between the groups.[1]
That is worth sitting with. Over seventy-two weeks, the liver improved on a slide, the scale moved by ten percent, and the patient-reported measure that the trial had prespecified and protected from multiple-comparison error did not separate. Histological benefit and felt benefit are not the same currency, and this trial measured both and found only one.
The weight column also runs below what the same molecule at the same dose produced in a weight trial. STEP 1 reported −14.9% against −2.4% over 68 weeks, a treatment difference of 12.4 percentage points.[6] ESSENCE’s 8.5-point difference comes from an older, heavier-diseased cohort in which more than half had type 2 diabetes, and the comparison is across two trials rather than inside one. A reader budgeting for weight loss should use the weight trials, collected in the semaglutide weight article, not this one.
The safety columns ran backwards
The abstract reports only that gastrointestinal adverse events were more common with semaglutide.[1] The rates themselves sit in the paper’s safety tables, which are behind the journal’s paywall and are not deposited in PubMed Central. Two independent summaries quoting those tables give an adverse event of any kind in 86.3% of the semaglutide group against 79.7% of the placebo group, and adverse events leading to discontinuation in 2.6% against 3.3%. More side effects on the drug, and fewer people stopping because of them. Those four figures are the only numbers on this page not read from a source anyone can open, and they are reported here as secondhand.
The practice guidance adds a specific negative that matters in a liver population: there were no discontinuations attributable to liver enzyme elevations, and routine hepatic panels are recommended only as clinically indicated.[5] It also lists what still needs watching — acute kidney injury from dehydration, symptomatic gallbladder disease, pancreatitis, retinopathy progression and lean mass loss — and notes that combining semaglutide with resmetirom has not been studied at all.[5]
What ESSENCE has not shown
It has not shown that anyone lives longer, avoids cirrhosis, avoids a transplant or avoids decompensation. That is Part 2’s endpoint and Part 2 has not reported.[2][3] It has not reported at all on the 397 patients randomized beyond the interim cohort of 800, and it has not reported anything past week 72 on anybody.
It excluded cirrhosis by design, enrolling only fibrosis stages 2 and 3,[3] so it says nothing about the patients furthest along. It did not compare semaglutide against any other drug. And because the entry criterion was a qualifying biopsy, it describes a population identified by a procedure most people with fatty liver never have — which is the gap the non-invasive thresholds in the guidance are trying to bridge.[5]
One more limit is regulatory rather than scientific. The approval this result supports is an accelerated one, conditioned on those long-term outcomes, for one branded formulation and one fibrosis range.[5] What an approval of that shape does and does not guarantee is set out in the FDA approval article.
What was in the syringe
Every figure here belongs to branded, FDA-approved semaglutide at 2.4 mg weekly, in patients enrolled on a qualifying liver biopsy and followed with repeat biopsies. The branded products that carry semaglutide and how their labels differ is covered in the Wegovy-versus-Ozempic article. Sellers on the semaglutide board largely dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed.
The distance is wider here than for a weight claim. A liver indication was granted to a specific product, for a specific fibrosis range, on the strength of an interim analysis, with the confirmatory outcome years away. A compounded preparation carries none of that, has never been studied in steatohepatitis, and is dispensed to people whose fibrosis stage nobody has established. The difference between a compounded vial and the labeled product is set out in the compounded-versus-brand article.