Exercise and NASH: Biopsy-Proven Gains, Small Trials

Exercise can change NASH on a liver biopsy. That is the important part.

Most lifestyle stories about fatty liver die in a swamp of softer endpoints: better enzymes, less fat on imaging, nicer-looking charts. The more interesting result is harder to fake. In a 2024 randomized trial by Mucinski et al., adults with biopsy-confirmed metabolic dysfunction-associated steatohepatitis, or MASH, completed a 10-month supervised program that combined high-intensity interval training with a modest calorie deficit. The intervention changed liver histology on paired biopsy, not just blood work.

That does not make exercise a magic liver drug. It does make the usual dismissal, that exercise only helps around the edges, much harder to defend.

What the best trial actually showed

The Mucinski trial enrolled 24 adults with biopsy-confirmed MASH, defined by a NAFLD activity score of at least 4 out of 8, plus at least 3 of 5 metabolic syndrome criteria. Sixteen participants entered treatment and eight served as controls. The exercise prescription was specific: HIIT three times per week, using 4 intervals of 4 minutes at 90% to 95% of maximum heart rate, separated by 3-minute active recoveries at about 50% of maximum heart rate, alongside roughly a 500 kcal per day energy restriction. Adherence reached 91% plus or minus 3%, and the treatment group lost 6.2% plus or minus 1.6% of body weight versus 1.2% plus or minus 1.9% in controls.

The headline result was histological improvement on paired liver biopsy after 10 months. That matters because biopsy remains the reference standard for diagnosing and grading steatohepatitis. Imaging can estimate liver fat. Biopsy shows tissue architecture.

The study also did something exercise papers rarely do: it used a two-step hyperinsulinemic-euglycemic clamp with isotope tracers to test mechanism directly. The main metabolic correlate of steatohepatitis resolution was not improvement in hepatic insulin resistance. It was peripheral insulin sensitization, specifically better skeletal muscle glucose disposal relative to circulating insulin. That is a much more interesting answer than "exercise burns calories," which is true in the same way that "fire is hot" is true.

Why muscle insulin sensitivity could help the liver

The logic is plausible and supported by broader mechanistic work. When skeletal muscle handles glucose poorly, more post-meal substrate stays available for the liver. That can feed hepatic de novo lipogenesis and worsen lipid overload. At the same time, adipose dysfunction increases free fatty acid delivery to the liver. A 2022 mechanistic study by Fuchs et al. supports that background model, linking adipose inflammation and impaired insulin signaling to NAFLD biology.

Exercise changes the traffic pattern. If muscle becomes more insulin sensitive, more glucose gets stored as glycogen or oxidized in muscle instead of getting shunted toward the liver. That reduces the substrate pressure driving hepatic fat accumulation and lipotoxicity. The Mucinski trial did not prove every step in that chain, but it did identify peripheral insulin sensitization as the metabolic driver most closely tied to histological improvement.

Exercise-only evidence is real, but it mostly uses softer endpoints

The case for exercise does not rest on one small biopsy trial. A 2022 systematic review and meta-analysis by Houttu et al. pooled 24 studies, including 18 randomized controlled trials, and found that aerobic exercise above 750 METs per week was about 3.5 times more likely than control to produce clinically meaningful liver fat reduction. Both high-intensity interval training and moderate-intensity continuous training showed benefit.

That strengthens the practical message: exercise has hepatic effects even when the protocol is not identical to the Mucinski program. It also sets a rough dose threshold. More than 750 METs per week translates, in real life, to something like several moderate sessions or a smaller number of harder ones.

Still, the endpoint problem does not disappear. Many of the studies in that meta-analysis used MRI-PDFF, ultrasound steatosis scores, or FibroScan liver stiffness measurements rather than paired biopsy. Those tools are useful, but they do not all answer the same question. FibroScan stiffness, for example, can improve when inflammation or congestion improves; that is not the same thing as confirmed fibrosis regression on histology. So the pooled evidence for lower liver fat is solid, while the pooled evidence for fibrosis-stage reversal is much weaker.

A 2021 meta-analysis by Babu et al. adds another helpful point: exercise-only interventions, without deliberate weight loss or dietary change, still improved liver enzymes, liver stiffness measures, and metabolic biomarkers. That supports the idea that exercise has liver benefits beyond whatever comes from eating less. It does not erase the confounding in the biopsy trial, where exercise and calorie restriction were bundled together.

FGF21 is interesting. It is not settled.

The second mechanistic story is FGF21, a hormone tied to metabolic stress. In a 2023 post-hoc analysis by Stine et al., 20 adults with biopsy-proven NASH completed a 20-week moderate-intensity aerobic program at 45% to 55% of VO2peak, five 30-minute sessions per week, without weight loss. Twelve exercised and eight served as controls.

Serum FGF21 fell by 22%, or 243.4 ng/mL, in the exercise group, while controls increased by 34%; the between-group difference reached statistical significance with p=0.037. The drop in FGF21 also correlated with higher cardiorespiratory fitness gains, lower liver volume, and lower PAI-1. In multivariable regression, each 1 mL/kg/min increase in VO2peak predicted a 44.5 ng/mL fall in FGF21, with p=0.031.

That is a coherent signal. It fits the idea that chronically elevated FGF21 in NASH may reflect compensation for metabolic stress, and that exercise could normalize the signal as fitness improves. But this result comes with several warning labels. The analysis was post-hoc, not pre-specified. The sample was tiny. The participants did not undergo paired liver biopsy at follow-up, so the study cannot show that falling FGF21 tracked histological improvement in the same people. At this stage, FGF21 is an interesting biomarker and a plausible pathway, not confirmed proof of causation.

The strongest mechanistic claim the evidence supports today is that exercise-driven muscle insulin sensitization is linked to MASH improvement on biopsy. FGF21 normalization is promising, but it is still auditioning for the job.

How much confidence should you have?

Enough to take exercise seriously. Not enough to oversell it.

The main biopsy-confirmed trial is small: 24 participants total, with only 8 controls. The control group also started with significantly higher intrahepatic triglyceride content than the treatment group, 24% plus or minus 4% versus 15% plus or minus 2%, despite similar activity scores. That baseline imbalance complicates interpretation. The intervention also combined HIIT with calorie restriction, so it cannot isolate the independent histological effect of exercise alone.

Generalizability is another issue. The Mucinski cohort was predominantly non-Hispanic white and predominantly female. Patients with more advanced fibrosis may respond differently, in part because cardiorespiratory fitness is already low in this population. Dahmus et al. in 2023 showed that patients with NASH and advanced liver disease had the lowest VO2peak among liver disease groups, which makes fitness gains both meaningful and harder to achieve.

There is also no head-to-head biopsy-endpoint trial comparing exercise with modern drug therapy. That matters because the pharmacology literature is larger and better powered. A 2025 meta-analysis of GLP-1 receptor agonists and dual agonists and a 2024 phase 2b denifanstat trial provide a benchmark for what a mature biopsy-based evidence base looks like: multicenter, placebo-controlled, and much larger. Exercise may prove comparable for some patients. It may also fall short. Current data cannot answer that question.

What this means in practice

The most defensible takeaway is narrower than the hype and stronger than the cynicism. Structured aerobic exercise, especially at moderate-to-high intensity, improves liver fat and metabolic markers across many studies. In one small but unusually rigorous randomized trial, a 10-month program of HIIT plus modest calorie restriction improved MASH histology on paired biopsy, and the metabolic driver most closely tied to that change was peripheral, not hepatic, insulin sensitization.

If you want a protocol that matches the evidence reasonably well, a 2023 review by Barrón-Cabrera et al. supports moderate-to-vigorous aerobic training as the most consistently useful approach for liver fat and metabolic outcomes. The current literature points toward 3 to 5 sessions per week, with at least 20 weeks for biomarker changes and closer to 6 to 12 months if you expect tissue-level change. That is not glamorous. It is also what the data support.

The verdict, then, is simple. Exercise is not a consolation prize for people who cannot get drug treatment. It has real disease-modifying potential in MASH. But the best biopsy-level evidence is still small, the cleanest mechanism belongs to muscle insulin sensitivity rather than any liver-specific miracle, and the FGF21 story needs a larger trial before it graduates from interesting to established.