The research team, led by Sai Long and Yao Li of the School of Physical Education at Jiangxi Normal University Science and Technology College in China, conducted the review in accordance with the PRISMA 2020 reporting standard and prospectively registered the protocol in PROSPERO. They searched six major databases—PubMed, Embase, Cochrane CENTRAL, Scopus, Web of Science, and SPORTDiscus—from inception through June 2026, screening 4,694 records and ultimately including 25 trials published between 2003 and 2026. The trials spanned Asia, Europe, North America, Australia, and the Middle East, and the exercise programs they tested were remarkably diverse: aerobic training, resistance training, high-intensity interval training, combined regimens, Pilates, yoga, and walking-based programs, lasting anywhere from six weeks to twelve months.

The headline numbers are striking. Across 15 trials involving 504 participants, exercise reduced alanine aminotransferase, or ALT, by a mean of 7.33 units per liter compared with control conditions, with a 95 percent confidence interval of −10.71 to −3.95. Aspartate aminotransferase, or AST, fell by 5.52 units per liter across 13 trials with 439 participants. Body mass index dropped by a pooled 1.54 kilograms per square meter across 17 trials, and peak or maximal oxygen uptake—VO2peak or VO2max, the gold standard measure of cardiorespiratory fitness—rose by 4.08 milliliters per kilogram per minute across seven trials. These are not trivial changes. ALT and AST are the workhorse blood markers of hepatocellular injury, and reductions of this magnitude are comparable to those seen with some pharmacological interventions studied in fatty liver disease.

But the analysis also delivered a sobering counterpoint: the pooled effects on liver fat itself, insulin resistance measured by HOMA-IR, and longer-term glycemic control measured by HbA1c did not reach statistical significance. Seven trials with 233 participants contributed liver fat or steatosis data, yielding a standardized mean difference of −0.36 with a confidence interval spanning −0.93 to 0.20—suggestive of a benefit but far from conclusive. HOMA-IR, pooled from six trials, showed a non-significant reduction of 0.49, and HbA1c, also from six trials, showed a non-significant drop of 0.18 percentage points. The authors are careful to note that improvements in BMI, glycemic markers, or fitness are clinically relevant in the broader cardiometabolic context of MASLD but should not be read as direct evidence that the liver disease itself has resolved.

The methodological machinery behind these conclusions is worth unpacking, because it reveals both the strength and the fragility of the evidence. The researchers used the Cochrane RoB 2 tool to assess risk of bias and found that no trial was judged at high overall risk; roughly half were rated low risk and the remainder had some concerns, most often involving incomplete reporting of randomization procedures or missing outcome data. They then applied the GRADE framework to rate certainty of evidence, and the results were uneven: moderate certainty for the ALT and BMI findings, low certainty for cardiorespiratory fitness, and very low certainty for AST, liver fat, HOMA-IR, and HbA1c. That gradation matters enormously for how clinicians and patients should weigh each result.

Heterogeneity was a persistent theme. Statistical heterogeneity, quantified by the I² statistic, reached 71.3 percent for ALT, 93.8 percent for AST, and 72.3 percent for VO2peak—meaning that the trials varied substantially in the size of the effects they reported. The authors addressed this in several ways. They ran leave-one-out sensitivity analyses, which generally preserved the direction of the pooled estimates. They conducted duration subgroup analyses, splitting trials into those lasting twelve weeks or less versus longer programs. Intriguingly, shorter interventions showed a larger ALT reduction—10.09 units per liter versus 2.46 units per liter for longer programs, a statistically significant difference. But the authors caution against interpreting this as evidence that shorter programs are physiologically superior; the comparison is observational, and duration may simply proxy for differences in baseline enzyme levels, exercise intensity, supervision, and adherence across trials.

One trial in particular shaped the AST picture. A 2025 study by Kim and colleagues reported an unusually large AST reduction, and when the authors excluded it in a dedicated sensitivity analysis, the pooled effect shrank from 5.52 to 3.25 units per liter while heterogeneity collapsed from 93.8 percent to 33.5 percent. The direction of the benefit held, but the magnitude proved sensitive to a single influential study. For BMI, funnel-plot asymmetry detected by Egger regression—significant at p=0.018—raised the possibility of small-study effects, though the authors stopped short of calling it definitive publication bias, noting that asymmetry can arise from multiple sources and that the BMI dataset included small, imprecise trials. No such asymmetry was detected for ALT or AST.

Perhaps the most conceptually important contribution of the review is its insistence on separating domains of response. ALT and AST are markers of hepatocellular injury, not direct measures of hepatic function, so their reduction signals biochemical improvement rather than histological resolution. Liver fat was measured with non-equivalent methods across trials—some used MRI-derived proton density fat fraction or proton magnetic resonance spectroscopy, which directly quantify intrahepatic lipid, while others used the controlled attenuation parameter from transient elastography, an indirect marker. Pooling these on a standardized scale cannot erase that measurement mismatch. Meanwhile, gains in aerobic capacity are biologically consistent with improved skeletal-muscle oxidative capacity, glucose disposal, fatty-acid utilization, and whole-body metabolic flexibility—adaptations that may reduce metabolic substrate delivery to the liver, but which the non-significant HOMA-IR and HbA1c results show were not uniform across metabolic domains.

The findings arrive amid a rapidly evolving evidence landscape. A 2024 network meta-analysis by Huang and colleagues ranked exercise modalities across metabolic and liver-function outcomes; a 2025 synthesis examined Mediterranean diet combined with exercise; and a 2026 meta-analysis by Wang and colleagues focused specifically on high-intensity interval training versus moderate-intensity continuous training, including dose-response analyses. The new review distinguishes itself by treating structured exercise as a single common exposure and applying one consistent post-intervention analytical framework across seven prespecified outcome domains. It also reports prediction intervals—ranges describing the expected effect in a future, comparable setting—which for ALT, BMI, and VO2peak all crossed the null, a statistical way of saying that the average benefit does not guarantee a uniform response in every clinical context.

For the millions of people living with MASLD, the practical message is cautiously encouraging. Structured exercise emerges as a physiologically relevant component of management, particularly when the goals include lowering elevated liver enzymes, reducing body mass, and building cardiorespiratory fitness. But the authors are explicit that the optimal prescription—the right combination of mode, intensity, and duration to produce reproducible hepatic adaptations independent of weight loss—remains unknown. They call for future trials that report exercise dose and adherence in sufficient detail, use standardized and sensitive measures of intrahepatic fat, and prespecify metabolic and cardiorespiratory outcomes. Until then, the liver enzyme improvements documented here stand as the most consistent signal that exercise does something real for the fatty liver, even as the deeper question of whether it reshapes the diseased organ itself awaits stronger evidence.

Subject of Research: Effects of exercise training on hepatic, metabolic, and cardiorespiratory outcomes in adults with metabolic dysfunction-associated steatotic liver disease

Article Title: Exercise training improves hepatic enzymes, body mass index, and cardiorespiratory fitness in adults with metabolic dysfunction-associated steatotic liver disease: a systematic review and meta-analysis

Article References: Exercise training improves hepatic enzymes, body mass index, and cardiorespiratory fitness in adults with metabolic dysfunction-associated steatotic liver disease: a systematic review and meta-analysis. (n.d.). https://doi.org/10.3389/fphys.2026.1976073

Image Credits: AI Generated

DOI: 10.3389/fphys.2026.1976073

Keywords: MASLD, NAFLD, exercise training, liver enzymes, ALT, AST, body mass index, cardiorespiratory fitness, meta-analysis, liver fat, insulin resistance, HOMA-IR