The MRI substudy adds an imaging question to SURPASS-3. It does not turn liver fat measurement into a direct measurement of every aspect of liver disease. Primary report
Study snapshot
The substudy included 296 participants and added a fatty-liver-index eligibility criterion. Its primary outcome compared liver fat measured by MRI proton-density fat fraction at 52 weeks. Pooled higher tirzepatide groups had a greater reduction than insulin degludec, with an estimated difference of −4.71 percentage points, 95% CI −6.72 to −2.70. Abdominal fat compartments were also assessed. This was a subset of the parent trial, not a separately recruited replication. Primary report
Our appraisal: preserve the measurement boundary
The phrase “liver improvement” is too broad unless the measured feature is named. Fat fraction, fibrosis, inflammation, symptoms, and clinical events answer different questions. This page therefore retains the imaging endpoint rather than translating it into an unmeasured clinical cure.
Selection into an imaging substudy can also narrow applicability. Readers should inspect who qualified, who had usable scans, and whether missing scans differed between groups. A large parent trial does not automatically supply the effective sample size for each scan-based analysis.
If an evidence review includes both this paper and the parent report, the relationship must be recorded. The papers can contribute different outcomes, but their participants should not be counted as independent populations merely because the journal articles have different identifiers.
Finally, an association between weight change and liver-fat change is not by itself a demonstration of the biological pathway responsible. A causal mechanism requires more than two variables moving together.
Frequently asked questions
Was liver fat measured with MRI?
Yes, using proton-density fat fraction.
Is that a biopsy fibrosis score?
No. The measurements are distinct.
Were all parent-trial participants included?
No. This was a selected substudy.
Is −4.71 a relative percentage reduction?
It is the reported between-group difference in absolute percentage-point change.
Is this independent replication of SURPASS-3?
No. It comes from the same parent trial.
Does correlation establish mediation?
No. A correlated change alone does not establish a causal pathway.
What review is still required?
Full imaging methods, missing-scan handling, and the analysis plan.