Efficacy and safety of GLP-1 receptor agonists in MASH with fibrosis: A systematic review and meta-analysis

Abstract

Background & Aims

Metabolic dysfunction-associated steatohepatitis (MASH) is a risk factor for progressive hepatic fibrosis and cirrhosis. The role of glucagon-like peptide-1 receptor agonists (GLP-1RAs) in the treatment of patients with MASH with fibrosis remains under investigation. This meta-analysis evaluates the efficacy and safety of GLP-1 RAs in patients with ‘at-risk’ MASH.

Methods

We reviewed and analyzed all randomized controlled trials (RCTs) from PubMed, Embase, and Cochrane databases. Primary outcomes included MASH resolution without worsening fibrosis or fibrosis improvement without worsening of MASH. Secondary outcomes included adverse events (AEs), laboratory, and anthropometric data. Studies evaluating dual agonists of the GLP-1 receptor with glucagon or glucose-dependent insulinotropic polypeptide agonists were also included. We performed meta-regression analyses to assess whether histologic outcomes were mediated by changes in body weight and glycemic control.

Results

Seven RCTs (1,800 patients, mean follow-up: 136.8 weeks) were included. In the population with baseline F2–F3 fibrosis stage, GLP-1 RAs were superior to placebo for histological resolution of MASH without worsening fibrosis (risk ratio 2.96; 95% CI 1.70-5.15, p <0.001; I² = 75.4%) and improvement of fibrosis without worsening MASH (risk ratio 1.59; 95% CI 1.32-1.90, p <0.001; I² = 0%). GLP-1 RAs also improved aminotransferases, MRI-proton density fat fraction, HbA1c, serum lipids, blood pressure, and anthropometric parameters in RCTs of MASH with F2–F3 fibrosis. GLP-1 RA treatment was associated with higher rates of treatment-emergent gastrointestinal AEs, but no significant differences in serious AEs when compared to placebo.

Conclusion

GLP-1 RAs improve MASH and MASH-associated hepatic fibrosis while also improving cardiometabolic risk factors in patients with MASH. Ongoing studies will define whether these surrogate endpoints translate into a decrease in liver-related and all-cause morbidity and mortality in patients at risk for adverse clinical outcomes attributable to MASH.

Impact and implications

Metabolic dysfunction-associated steatohepatitis (MASH) with fibrosis is a major driver of liver-related morbidity and mortality, yet effective pharmacologic options remain limited. This meta-analysis shows that glucagon-like peptide-1 receptor agonists (GLP-1 RAs) significantly increase histologic resolution of MASH without worsening fibrosis and improve fibrosis stage without worsening MASH. Simultaneously, they improve metabolic risk factors, supporting their role as disease-modifying agents in patients with at-risk MASH. These benefits, together with a favorable safety profile, highlight their potential as a dual-target therapeutic strategy for hepatologists and endocrinologists managing this high-risk population. Future trials assessing long-term clinical outcomes will be essential to guide guideline development, inform access policies, and support the extension of GLP-1 RA use.

Graphical abstract

Highlights

• •GLP-1 RAs increase histologic MASH resolution without worsening fibrosis.
• •GLP-1 RAs improve fibrosis stage without worsening MASH, showing disease-modifying effects.
• •GLP-1 RAs reduce hepatic fat fraction and improve cardiometabolic and body measures.
• •GLP-1 RAs have a favorable safety profile, with more GI events than placebo but no increase in serious adverse events.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD) is the most common chronic liver disease worldwide and the most rapidly increasing cause of liver transplantation in the US.¹ Approximately 40% of US adults are expected to have MASLD by 2050, while 23.2 million people (7.9%) will have metabolic dysfunction-associated steatohepatitis (MASH), with around one-third showing moderate or severe (F2–F3) fibrosis.² Advanced hepatic fibrosis (F3) and cirrhosis (F4) correlate with higher all-cause mortality and liver complications,[3], [4], [5] highlighting the need to resolve steatohepatitis and improve hepatic fibrosis, the primary predictor of liver-related morbidity and mortality.⁶

The increasing trend in prevalence and incidence of MASLD/MASH is aligned with the rising prevalence of obesity and type 2 diabetes mellitus (T2DM), two risk factors that are strongly associated with the development of MASH.⁷ Approximately 70% of individuals with T2DM have MASLD, around 35% have MASH, and roughly 7% develop MASLD-related cirrhosis.⁸ In patients with MASH, the presence of T2DM is the strongest predictor of advanced hepatic fibrosis and cirrhosis.⁷ A bidirectional relationship between MASH and T2DM exists. Interestingly, the presence of T2DM and poor glycemic control confers an increased risk for MASH, advanced hepatic fibrosis, and liver-related mortality in patients with MASLD/MASH.^3,9,10 In those with advanced-stage disease, the presence of T2DM is associated with a fourfold higher risk of hepatocellular carcinoma.¹¹ Likewise, the reverse relationship between MASLD is also described, with an approximately two-fold greater risk of T2DM in patients with MASLD compared to those without MASLD.^12,13 Similarly, in patients with obesity, MASLD affects nearly 75%, and about one-third develop MASH.¹⁴ Weight loss, when achieved and sustained, improves liver biochemistry and histologic features of steatosis, steatohepatitis, and potentially fibrosis.¹⁵ In patients with MASH, reducing body weight by at least 5% can reduce hepatic steatosis, a loss of 7% or more may reverse MASH, and losing 10% or more can stabilize or even reverse liver fibrosis.¹⁶ Glucagon-like peptide-1 receptor agonists (GLP-1 RAs), incretin-based therapies, have revolutionized the management of both obesity and T2DM.^17,18 GLP-1 RAs have achieved up to 20% weight loss and 2.1% HbA1c reduction,^19,20 sparking interest in their potential therapeutic role in MASLD, a disease closely linked to obesity and T2DM.

Although previous meta-analyses have examined the use of GLP-1 RAs in MASLD,[21], [22], [23], [24], [25], [26] they were predominantly enriched with studies that included patients with MASLD, without defining the presence of ‘at-risk’ MASH (moderate-to-advanced [F2–F3] hepatic fibrosis), the target cohort of patients eligible to receive FDA-approved pharmacotherapies for treatment of MASH. Only two trials utilizing liver histology with smaller sample sizes, the LEAN²⁷ 2015 and Newsome et al.²⁸ 2020 trials, have been included in prior meta-analyses of GLP-1 RAs for MASLD/MASH. However, several recent publications, including the global part-1 results of the phase III ESSENCE trial (NCT04822181),²⁹ have improved our ability to understand the broader safety and efficacy of GLP-1 RAs on the histologic features of disease activity in patients with ‘at-risk’ MASH.[29], [30], [31], [32], [33]

Patients and methods

Protocol and registration

The systematic review and meta-analysis were conducted and structured in accordance with the Cochrane Handbook for Systematic Reviews of Interventions and the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) recommendations.^34,35 The protocol was registered in PROSPERO (the International Prospective Register of Systematic Reviews) under the identification number CRD42025639458. The meta-analysis will be updated when new evidence likely to influence the results is identified.

Eligibility criteria

Inclusion in this meta-analysis was limited to studies meeting all of the following eligibility criteria: (1) randomized, placebo-controlled clinical trials; (2) evaluation of the efficacy and safety of GLP-1 receptor agonists (GLP-1RAs); and (3) inclusion of patients with biopsy-proven MASH and F1–F4 liver fibrosis. There were no restrictions on language or publication date. Studies evaluating both single GLP-1 receptor agonists (e.g. liraglutide, semaglutide, exenatide, dulaglutide) and dual agonists targeting the GLP-1 receptor in combination with glucagon or glucose-dependent insulinotropic polypeptide (e.g. tirzepatide, survodutide) were included. Conversely, abstracts, editorials, letters, book chapters, case reports, in vitro studies, animal studies, reviews, studies with overlapping populations, or studies lacking a placebo comparison were excluded.

Search strategy and study selection

We conducted a systematic review of PubMed, Embase, and Cochrane Database from inception to July 10^th, 2025. Our search combined keywords presented in the supplementary information. We examined the reference lists from the included manuscripts to identify any relevant additional studies. Two authors (JMBM and RDSB) independently conducted the search and triaged studies using Rayyan.³⁶ After excluding duplicates and titles/abstracts unrelated to the clinical question, we assessed the eligibility via full-text review. In cases of disagreement, a third reviewer (ESA) was consulted.

Data extraction and outcomes

Two authors (LHDP and AFC) independently extracted the data. The primary outcomes were: (1) histologic resolution of MASH without worsening of fibrosis; (2) the rates of histologic improvement in the liver fibrosis stage without worsening of MASH. The secondary outcomes included the proportion of patients with at least one adverse event (AE), as well as changes in liver stiffness, controlled attenuation parameter, laboratory, imaging, metabolic, and anthropometric parameters. MASH resolution was defined as the presence of isolated or simple steatosis without steatohepatitis, with a NAFLD activity score (NAS)³⁷ of 0-1 for lobular inflammation, 0 for hepatocellular ballooning, and 0-3 for steatosis. Worsening of MASH was defined as any increase in one or more NAS activity subcategories. Improvement in liver fibrosis was defined as a ≥1-stage decrease, and worsening of fibrosis as a ≥1-stage increase, in the NASH-CRN³⁷ fibrosis staging system. Our primary outcomes, as well as the outcome “MASH resolution and improvement in fibrosis”, which was not reported here because these data were not available in the included trials, represent the FDA-recommended histological endpoints considered likely to predict clinical benefit.³⁸

Histologic data were reported by a consensus between two pathologists in the ESSENCE,²⁹ SYNERGY-NASH,³¹ Newsome et al.,²⁸ and LEAN²⁷ studies, and a single pathologist in the studies from Sanyal et al.³² and Loomba et al.³⁰ Because the included studies enrolled patients across a wide spectrum of fibrosis (F0–F4), who may exhibit clinical and biochemical features that are not directly comparable, our primary efficacy analyses were restricted to the ‘at-risk’ MASH population (fibrosis stage F2–F3). The overall study populations were considered only for the safety assessment of GLP-1 RAs in MASH. Notably, the LEAN²⁷ trial included a small subset of patients without fibrosis (stage F0); however, this trial was retained since F0 cases represented only 0.1-0.3% of the total sample in our meta-analysis, a proportion deemed negligible.

Statistical analysis

We used the “meta”, “metafor”, “metagen”, “dplyr”, and “ggplot2” packages on RStudio 4.3.1 for statistical analysis.³⁹ Data synthesis employed DerSimonian and Laird random-effects models. Dichotomous endpoints were analyzed as the number of patients experiencing at least one event relative to the total number of patients in each group, and results were compared using risk ratios (RR) with corresponding 95% CIs. Slight differences in the total number of patients between the efficacy and safety analyses occurred in some studies because the safety population included the entire trial cohort, whereas efficacy outcomes were limited to the subset of patients who underwent biopsy. Continuous endpoints were analyzed as change from baseline, and results were compared using mean differences with corresponding 95% CIs. Statistical significance was set at p <0.05. The two primary efficacy outcomes were analyzed as independent co-primary endpoints, consistent with regulatory agencies' guidance and the design of individual clinical trials in which either outcome is considered an acceptable surrogate of treatment efficacy; therefore, no multiplicity adjustment was performed.[27], [28], [29], [30], [31], [32], [33]^,38 Continuous data reported in the median (interquartile range) were converted to mean and standard deviation using Luo et al. and Wan et al. estimates, respectively.^40,41 Pooled baseline mean values were calculated using a sample-size-weighted mean, and pooled standard deviations were derived using the standard pooled variance formula based on study-level summary statistics, as recommended in the Cochrane Handbook.³⁵ Heterogeneity was evaluated through I² and Cochran’s Q test with significance set at p <0.10 and I² >40%. In cases of significant heterogeneity (I² >40%), sensitivity analyses were conducted to ensure the reliability of the findings.

Trial sequential analysis (TSA) was performed on the primary outcomes using TSA software.⁴² We used the effect measure (RR) and selected a random effects model using the DerSimonian-Laird method. No continuity correction was applied in the case of a zero event. We estimated the required sample size on the calculated effect size for the intervention, considering a type I error of 5% and a power of 80%; benefit, harm, and inner wedge boundaries were drawn using the O’Brien-Fleming spending function. Heterogeneity correction employed the variance-based model.

As a sensitivity analysis for the histologic endpoints, we conducted an arm-based meta-analysis of randomized, placebo-controlled trials enrolling adults with biopsy-proven MASH and reporting prespecified biopsy outcomes at approximately 48-72 weeks. Co-primary outcomes were (1) histologic resolution of MASH without worsening of fibrosis and (2) histologic improvement in fibrosis stage without worsening of MASH, both for patients with baseline F2–F3 stage fibrosis. From each arm, we extracted randomized sample size, number (or percentage) of responders, treatment duration, and mean percent change in body weight at the biopsy time point; when only kilogram change was provided, percent change was computed as (mean change in kilograms ÷ mean baseline body weight in kilograms) × 100. To evaluate weight-dependence across contrasts, we prespecified an arm-level moderator Δweight (%) defined as the absolute percent weight change in the active arm minus the absolute percent weight change in the placebo arm. When only percentages of responders were available, the number of events was reconstructed by rounding (percentage × randomized N ÷ 100). In multi-dose trials with a shared placebo, placebo denominators and events were split equally across active arms to avoid double-counting (per Cochrane guidance).³⁴ For each contrast, 2 × 2 tables were formed and log risk ratios (log-RRs) with sampling variances were computed using a continuity correction of 0.5 in all four cells to handle zero or sparse counts. Pooled effects used random-effects models with restricted maximum likelihood (REML); if REML failed to converge, Paule-Mandel then DerSimonian-Laird were tried sequentially, and when between-study variance was ∼0, a fixed-effect model was reported. Inference used Knapp-Hartung adjustments; we report τ², I², Cochran’s Q, and a 95% prediction interval when a random-effects model was obtained. We then fit mixed-effects meta-regressions of log-RRs on centered Δweight (%). Regarding glycemic control, exploratory moderator analyses used ΔHbA1c, defined as the absolute percentage point HbA1c reduction in the treatment arm minus that in the placebo arm. The results were stratified according to the presence of T2DM. Arms lacking a given moderator contributed to the pooled effects but were omitted from that moderator model. Finally, we applied a leave-one-out analysis to evaluate the influence of individual studies and consistency.

Risk of bias and evidence quality assessment

Two independent authors (MABM and GGB) assessed the risk of bias following the second version of the Cochrane Risk of Bias assessment tool (RoB2),⁴³ evaluating bias in five domains for each outcome: (1) randomization process; (2) deviations from intended interventions; (3) missing data; (4) outcome measurement; and (5) selection of the reported results. Disagreements were resolved through consensus after discussing the reasons for the discrepancy.

Results

Study selection and baseline characteristics

The initial search yielded 955 results. After removing 598 duplicates and 346 by text/abstract, 11 studies were fully reviewed. Of these, the design of the SYNERGY-NASH,⁴⁴ and two studies were excluded for not meeting the inclusion criteria. Specifically, the study by Romero-Gómez M et al.⁴⁵ was a quality-of-life analysis performed on the phase II double-blind, placebo-controlled trial of semaglutide in patients with MASH²⁸ included in our meta-analysis. The LEAN-J trial⁴⁶ was a pilot, open-label study conducted in Japan assessing liraglutide in patients with MASH and glucose intolerance. Although it reported outcomes of interest, it was excluded because it was a single-arm study with no control group. Additionally, the study by Armstrong MJ et al.,⁴⁷ which reported biochemical outcomes of interest, was excluded to avoid double-counting patient data because the complete and updated analyses were already included in our meta-analysis as the LEAN trial.²⁷ Ultimately, seven studies[27], [28], [29], [30], [31], [32], [33] met the inclusion criteria (Fig. S1).

The non-overlapping cohort included 1,800 patients with MASH and fibrosis, with a mean follow-up of 136.8 weeks. The population was 57.0% female and 74.2% White, with a mean age of 54.9 ± 11.8 years. Baseline body weight was 98.1 ± 22.9 kg, mean BMI was 35.3 ± 6.2 kg/m², and the prevalence of type 2 diabetes mellitus (T2DM) was 54.5% (mean HbA1c 6.7 ± 1.2%). The distribution of Kleiner liver fibrosis stages was 52.8% F3, 31.9% F2, 10.5% F1, 4.3% F4, and 0.1% F0. Mean alanine aminotransferase was 64.8 ± 44.9 U/L, and mean aspartate aminotransferase was 51.4 ± 35.2 U/L. The lipid profile included HDL cholesterol of 44.7 ± 14.8 mg/dl, LDL cholesterol of 98.0 ± 65.8 mg/dl, and triglycerides of 168.6 ± 104.8 mg/dl. Baseline characteristics for individual study groups are detailed in Table 1.

Table 1.

Study characteristics and baseline demographics.

		ESSENCE 2025	SYNERGY-NASH 2024	Sanyal et al. 2024	PROXYMO 2024	Loomba et al. 2023	Newsome et al. 2020	LEAN 2015	Total
Design (NCT)		RCT Phase III (NCT04822181)	RCT Phase II (NCT04166773)	RCT Phase II (NCT04771273)	RCT Phase II (NCT04019561)	RCT Phase II (NCT03987451)	RCT Phase II (NCT02970942)	RCT Phase II (NCT01237119)	N/A
Follow-up (weeks) Mean		240 weeks	52 weeks	48 weeks	19 weeks	48 weeks	72 weeks	48 weeks	136.8 ± 92.9 weeks
Intervention		semaglutide, 2.4 mg once weekly	tirzepatide 5/10/15 mg once weekly	survodutide 2.4/4.8/6.0 mg once weekly	cotadutide 300/600 μg once daily	semaglutide 2.4 mg once weekly	semaglutide 0.1/0.2/0.4 mg once daily	liraglutide 1.8 mg once daily	N/A
Number of patients	Intervention	534 (66.7)	142 (74.7)	219 (74.7)	50 (67.6)	47 (66.2)	240 (75)	26 (50)	1,258 (69.9)
N (%)	Placebo	266 (33.3)	48 (25.3)	74 (25.3)	24 (32.4)	24 (33.8)	80 (25)	26 (50)	542 (30.1)
Patient characteristics
Female sex, n (%)	Intervention	313 (58.6)	82 (57.7)	111 (50.7)	27 (54.0)	31 (66)	150 (62.5)	8 (30.8)	722 (57.4)
	Placebo	144 (54.1)	27 (56.3)	44 (59.5)	14 (58.3)	18 (75)	44 (55.0)	13 (50)	304 (56.1)
Age (years), mean ± SD	Intervention	56.3 ± 11.4	54.7 ± 11.2	50.1 ± 13.2	57.6 ± 11.2	59.9 ± 7.1	55.8 ± 10.4	50.0 ± 11.0	55.0 ± 11.7
	Placebo	55.4 ± 12.0	53.5 ± 11.6	53.0 ± 11.5	52.2 ± 13.5	58.7 ± 9.7	52.4 ± 10.8	52.0 ± 12.0	54.6 ± 11.9
Race, n (%)	Intervention	White: 361 (67.6) Asian: 142 (26.6) Other: 21 (3.9) Black or African American: 3 (0.6) Missing data: 7 (1.3)	White: 121 (85.2) Asian: 17 (12) American Indian or Alaska Native: 3 (2.1) Black: 1 (0.7)	White: 149 (68.0) Asian: 63 (28.8) Other: 7 (3.2)	White: 48 (96) Black/African American: 2 (4) American Indian or Alaska Native: 0	White: 41 (87.2) Asian: 1 (2.1) American Indian or Alaska Native: 1 (2.1) Black/African American: 0 Other: 1 (2.1) Missing data: 3 (6.4)	White: 186 (77.5) Asian: 36 (15.0) American Indian or Alaska Native: 3 (1.2) Black/African American: 1 (0.9) Other: 13 (5.4)	White: 23 (88.5) Asian: 1 (3.8) Black: 1 (3.8) Other: 1 (3.8)	White: 929 (73.8) Asian: 260 (20.7) Other: 58 (4.6) Missing data: 10 (0.9)
	Placebo	White: 179 (67.3) Asian: 74 (27.8) Other: 10 (3.8) Black or African American: 2 (0.7) Missing data: 1 (0.4)	White: 43 (89.6) Asian: 5 (10.4) American Indian or Alaska Native: 0 Black: 0	White: 56 (75.7) Asian: 17 (23) Other: 1 (1.4)	White: 23 (95.8) Black/African American: 0 American Indian or Alaska Native: 1 (4.2)	White: 21 (87.5) Asian: 0 American Indian or Alaska Native: 0 Black/African American: 2 (8.3) Other: 0 Missing data: 1 (4.2)	White: 62 (77.5) Asian: 12 (15.0) American Indian or Alaska Native: 0 Black/African American: 0 Other: 6 (7.5)	White: 23 (88.5) Asian: 1 (3.8) Black: 0 Other: 2 (7.7)	White: 407 (75.1) Asian: 109 (20.1) Other: 24 (4.4) Missing data: 2 (0.4)
Hispanic or Latino ethnic group, N (%)	Intervention	95 (17.8)	51 (35.9)	59 (26.9)	N/A	N/A	31 (12.9)	N/A	N/A
	Placebo	51 (19.2)	18 (38)	22 (29.7)	N/A	N/A	9 (11.2)	N/A	N/A
Body weight (kg), mean ± SD	Intervention	95.4 ± 24.5	101.1 ± 21.4	101.8 ± 22.9	99.2 ± 19.0	95.2 ± 18.7	97.4 ± 21.0	101.0 ± 18.0	97.8 ± 22.8
	Placebo	97.6 ± 24.6	96.0 ± 21.6	98.1 ± 20.8	102.2 ± 18.1	98.6 ± 22.2	101.3 ± 23.3	108.0 ± 18.0	98.8 ± 23.1
Waist circumference (cm), mean ± SD	Intervention	111.8 ± 15.6	N/A	114.1 ± 13.8	N/A	112.9 ± 11.8	N/A	110.0 ± 11.0	N/A
	Placebo	113.1 ± 16.0	N/A	113.0 ± 14.2	N/A	118.3 ± 14.8	N/A	108.0 ± 18.0	N/A
BMI (kg/m2), mean ± SD	Intervention	34.3 ± 7.2	36.2 ± 6.0	35.9 ± 6.4	37.1 ± 6.2	34.6 ± 5.9	35.7 ± 2.3	34.2 ± 4.7	35.2 ± 6.2
	Placebo	35.0 ± 7.1	36.0 ± 6.7	35.5 ± 6.4	37.5 ± 5.1	35.5 ± 6.0	36.2 ± 2.4	37.7 ± 6.2	35.6 ± 6.3
T2DM, n (%)	Intervention	296 (55.4)	82 (57.7)	84 (38.4)	29 (58)	35 (74.5)	149 (62.1)	9 (34.6)	684 (54.4)
	Placebo	151 (56.8)	29 (60.4)	29 (39.2)	12 (50)	18 (75.0)	50 (62.5)	8 (30.8)	297 (54.8)
Arterial blood pressure (mmHg), mean ± SD	Intervention	N/A	N/A	SBP: 129.4 ± 14.7 DBP: 80.5 ± 8.6	SBP: 127.8 ± 9.7 DBP: 76.9 ± 6.9	SBP: 132.6 ± 13.7 DBP: 78.7 ± 9.6	SBP: 133.6 ± 15.1 DBP: 81.3 ± 9.7	SBP: 130.0 ± 13.0 DBP: 79.0 ± 11.0	N/A
	Placebo	N/A	N/A	SBP: 129.4 ± 12.5 DBP: 81.2 ± 8.4	SBP: 124.9 ± 8.6 DBP: 78.6 ± 6.3	SBP: 135.8 ± 14.7 DBP: 87.0 ± 6.6	SBP: 131.0 ± 13.0 DBP: 82.0 ± 9.0	SBP: 133.0 ± 12.0 DBP: 78.0 ± 9.0	N/A
Baseline laboratory tests
ALT (U/L), mean ± SD	Intervention	67.8 ± 42.3	62.6 ± 34.2	57.9 ± 43.5	44.7 ± 30.7	56.1 ± 39.4	70.6 ± 57.4	77.0 ± 34.0	64.9 ± 44.7
	Placebo	67.9 ± 44.7	59.7 ± 30.3	57.3 ± 36.6	48.8 ± 31.2	41.8 ± 23.5	74.8 ± 65.4	66.0 ± 42.0	64.7 ± 45.5
AST (U/L), mean ± SD	Intervention	53.2 ± 28.6	50.0 + 24.5	45.9 ± 34.9	35.1 ± 16.3	51.9 ± 24.2	55.4 ± 52.1	51.0 ± 22.0	51.2 ± 34.7
	Placebo	52.8 ± 33.1	52.3 ± 21.3	51.3 ± 40.9	38.4 ± 24.6	42.9 ± 20.3	54.6 ± 54.2	51.0 ± 27.0	51.7 ± 36.3
GGT (U/L), mean ± SD	Intervention	87.5 ± 90.2	63.9 ± 49.4	N/A	51.9 ± 31.9	126 ± 110.3	91.1 ± 79.9	91 ± 69	N/A
	Placebo	85.6 ± 91.0	97.9 ± 98.1	N/A	53.4 ± 39.6	167.6 ± 229	81.8 ± 71.5	115 ± 174	N/A
HbA1c in patients with T2DM (%), mean ± SD	Intervention	7.3 ± 1.1	N/A	6.88 ± 1.01	N/A	N/A	7.3 ± 1.2	N/A	N/A
	Placebo	7.0 ± 1.0	N/A	7.1 ± 0.9	N/A	N/A	7.3 ± 1.2	N/A	N/A
HbA1c in patients without T2DM (%), mean ± SD	Intervention	5.8 ± 0.5	N/A	N/A	N/A	N/A	5.8 ± 0.5	N/A	N/A
	Placebo	5.8 ± 0.5	N/A	N/A	N/A	N/A	5.8 ± 0.7	N/A	N/A
HbA1c in overall population (%), mean ± SD	Intervention	6.8 ± 1.2	6.5 ± 1.1	6.9 ± 1.0	6.6 ± 1.1	7.1 ± 1.3	6.7 ± 1.2	5.9 ± 0.7	6.7 ± 1.2
	Placebo	6.6 ± 1.0	6.8 ± 1.2	7.1 ± 0.9	6.8 ± 1.5	7.2 ± 1.2	6.7 ± 1.3	6.0 ± 0.9	6.7 ± 1.1
Lipid profile (mg/dl), mean ± SD	Intervention	Total cholesterol: 187.2 ± 44.5 (n = 524. Missing data: 10) HDL: 46.8 ± 11.6 (n = 518. Missing data: 16) LDL: 110.2 ± 44.5 (n = 517. Missing data: 17) VLDL: N/A Triglycerides: 164.7 ± 77.9 (n = 522. Missing data: 12)	Total cholesterol: N/A HDL: 43.8 ± 9.6 LDL: 106.3 ± 36.8 VLDL: N/A Triglycerides: 172.0 ± 90.6	Total cholesterol: 127.5 ± 42.6 HDL: 43.6 ± 11.4 LDL: 26.8 ± 37.8 VLDL: N/A Triglycerides: 122.2 ± 88.8	Total cholesterol: 189.4 ± 44.7 HDL: 44.4 ± 11.6 LDL: 104.4 ± 35.5 VLDL: N/A Triglycerides: 199.3 ± 105.6	Total cholesterol: 177.2 ± 34.9 HDL: 44.7 ± 10.0 LDL: 100.0 ± 34.4 VLDL: 32.5 ± 17.4 Triglycerides: 168.9 ± 98.3	Total cholesterol: 208.5 ± 110.4 HDL: 49.3 ± 25.4 LDL: 128.6 ± 100.2 VLDL: 38.7 ± 28.2 Triglycerides: 207.1 ± 160.2	Total cholesterol: 174.0 ± 42.5 HDL: 42.5 ± 15.5 LDL: 100.5 ± 30.9 VLDL: N/A Triglycerides: 168.3 ± 97.4	Total cholesterol: N/A HDL: 46.1 ± 15.2 (n = 1,002. Missing data: 16) LDL: 97.8 ± 66.5 (n = 1,001. Missing data: 17) VLDL: N/A Triglycerides: 167.9 ± 107.1 (n = 1,006. Missing data: 12)
	Placebo	Total cholesterol: 179.0 ± 42.2 (n = 260. Missing data: 6) HDL: 44.9 ± 11.2 (n = 259. Missing data: 7) LDL: 102.5 ± 43.3 (n = 259. Missing data: 7)VLDL: N/A Triglycerides: 171.8 ± 88.6 (n = 260. Missing data: 6)	Total cholesterol: N/A HDL: 43.6 ± 9.0 LDL: 106.2 ± 39.1 VLDL: N/A Triglycerides: 169.0 ± 85.3	Total cholesterol: 118.2 ± 33.6 HDL: 44.7 ± 12.7 LDL: 19.9 ± 32.2 VLDL: N/A Triglycerides: 113.7 ± 63.5	Total cholesterol: 201.1 ± 34.8 HDL: 46.4 ± 11.6 LDL: 119.9 ± 30.9 VLDL: N/A Triglycerides: 168.3 ± 62.0	Total cholesterol: 163.4 ± 47.5 HDL: 45.8 ± 12.6 LDL: 88.1 ± 41.7 VLDL: 29.6 ± 11.0 Triglycerides: 151.6 ± 56.2	Total cholesterol: 217.6 ± 76.8 HDL: 49.0 ± 23.6 LDL: 145.7 ± 101.3 VLDL: 44.2 ± 43.9 Triglycerides: 228.2 ± 151.1	Total cholesterol: 193.3 ± 46.4 HDL: 50.3 ± 7.7 LDL: 112.1 ± 38.7 VLDL: N/A Triglycerides: 159.4 ± 70.9	Total cholesterol: N/A HDL: 44.7 ± 13.8 (n = 455. Missing data: 7) LDL: 98.5 ± 64.2 (n = 455. Missing data: 7) VLDL: N/A Triglycerides: 170.3 ± 99.3 (n = 456. Missing data: 6)
Baseline non-invasive tests
Liver fat content according to MRI-PDFF (%), mean ± SD	Intervention	N/A	18.5 ± 7.6	19.5 ± 7.5	19.5 ± 7.5	11.3 ± 5.0	N/A	N/A	N/A
	Placebo	N/A	18.2 ± 6.8	19.6 ± 7.6	19.1 ± 8.2	11.6 ± 5.2	N/A	N/A	N/A
Liver stiffness (kPa), mean ± SD	Intervention	12.8 ± 6.6 (by VCTE)	11.7 ± 5.4 (by VCTE)	N/A	N/A	11.9 ± 6.5 (by MRE)	14.6 ± 11.7 (by VCTE)	N/A	N/A
	Placebo	12.9 ± 7.6 (by VCTE)	12.0 + 5.1 (by VCTE)	N/A	N/A	6.1 ± 2.0 (by MRE)	11.7 ± 10.5 (by VCTE)	N/A	N/A
CAP/liver steatosis assessed by VCTE (dB/m), mean ± SD	Intervention	329 ± 45	N/A	N/A	N/A	N/A	338.3 ± 52.7	N/A	N/A
	Placebo	330±49	N/A	N/A	N/A	N/A	348.6 ± 35.2	N/A	N/A
FIB-4 index score, mean ± SD	Intervention	1.6 ± 0.8	1.6 ± 0.8	N/A	N/A	2.6 ± 0.9	N/A	N/A	N/A
	Placebo	1.6 ± 0.9	1.6 ± 0.7	N/A	N/A	2.5 ± 1.2	N/A	N/A	N/A
ELF test score, mean ± SD	Intervention	9.9 ± 0.9	9.8 ± 0.8	N/A	N/A	10.7 ± 0.8	9.8 ± 1.0	9.3 ± 0.9	N/A
	Placebo	9.9 ± 1.0	9.9 ± 0.8	N/A	N/A	10.6 ± 0.7	9.6 ± 0.9	9.4 ± 1.3	N/A
Baseline liver histology
NASH-CRN (Kleiner) liver fibrosis stage, n (%)	Intervention	F2: 169 (31.6) F3: 365 (68.4)	F2: 64 (45.1) F3: 78 (54.9)	F1: 56 (25.6) F2: 90 (41.1) F3: 73 (33.3)	F1: 11 (22) F2: 30 (60) F3: 9 (18)	F4: 47 (100)	F1: 68 (28.3) F2: 50 (20.9) F3: 122 (50.8)	N/A: 3 (11.5); F0–F2: 2; F3–F4: 1 F0: 0 F1: 5 (19.2) F2: 7 (26.9) F3: 9 (34.6) F4: 2 (7.8)	N/A: 3 (0.3); F0–F2: 2; F3–F4: 1 F0: 0 F1: 140 (11.1) F2: 410 (32.6) F3: 656 (52.1) F4: 49 (3.9)
	Placebo	F2: 81 (30.5) F3: 185 (69.5)	F2: 17 (35.4) F3: 31 (64.6)	F1: 14 (18.9) F2: 30 (40.5) F3: 30 (40.5)	F1: 7 (29.2) F2: 14 (58.3) F3: 3 (12.5)	F4: 24 (100)	F1: 22 (27.5) F2: 22 (27.5) F3: 36 (45.0)	N/A: 4 (15.4) - 3 F0–F2, 1 F3–F4 F0: 1 (3.8) F1: 7 (26.9) F2: 0 F3: 10 (38.5) F4: 4 (15.4)	N/A: 4 (0.7) - 3 F0–F2, 1 F3–F4 F0: 1 (0.3) F1: 50 (9.2) F2: 164 (30.2) F3: 295 (54.4) F4: 28 (5.2)
Total NAFLD activity score, mean ± SD	Intervention	N/A	5.2 ± 0.9	5.2 ± 1.0	N/A	4.7 ± 1.0	4.9 ± 0.9	4.9 ± 0.9	N/A
	Placebo	N/A	5.3 ± 1.0	5.2 ± 1.1	N/A	4.9 ± 1.2	4.9 ± 0.9	4.8 ± 0.9	N/A
NAS/NAFLD subscore for lobular inflammation, n (%)	Intervention	0: 0 1: 187 (35.0) 2: 238 (44.6) 3: 109 (20.4)	N/A	0: 0 1: 99 (45.2) 2: 114 (52.1) 3: 6 (2.7)	N/A	0: 0 1: 14 (29.8) 2: 31 (66.0) 3: 2 (4.2)	0: 0 1: 102 (42.5) 2: 128 (53.3) 3: 10 (4.2)	N/A	N/A
	Placebo	0: 0 1: 103 (38.7) 2: 126 (47.4) 3: 37 (13.9)	N/A	0: 0 1: 37 (50) 2: 32 (43.2) 3: 5 (6.8)	N/A	0: 0 1: 6 (25.0) 2: 17 (70.8) 3: 1 (4.2)	0: 0 1: 33 (41.2) 2: 46 (57.5) 3: 1 (1.3)	N/A	N/A
NAS/NAFLD subscore for hepatocyte ballooning, n (%)	Intervention	0: 0 1: 220 (41.2) 2: 314 (58.8)	N/A	0: 0 1: 163 (74.4) 2: 56 (25.6)	N/A	0: 0 1: 18 (38.3) 2: 29 (61.7)	0: 0 1: 160 (66.7) 2: 80 (33.3)	N/A	N/A
	Placebo	0: 0 1: 100 (37.6) 2: 166 (62.4)	N/A	0: 0 1: 49 (66.2) 2: 25 (33.8)	N/A	0: 0 1: 8 (33.3) 2: 16 (66.7)	0: 0 1: 58 (72.5) 2: 22 (27.5)	N/A	N./A
NAS/NAFLD subscore for steatosis, n (%)	Intervention	0: 0 1: 205 (38.4) 2: 315 (59.0) 3: 14 (2.6)	N/A	0: 0 1: 7 (3.2) 2: 125 (57.1) 3: 87 (39.7)	N/A	0: 0 1: 32 (68.1) 2: 12 (25.5) 3: 3 (6.4)	0: 0 1: 73 (30.4) 2: 116 (48.3) 3: 51 (21.3)	N/A	N/A
	Placebo	0: 0 1: 116 (43.6) 2: 142 (53.4) 3: 8 (3.0)	N/A	0: 0 1: 3 (4.1) 2: 43 (58.1) 3: 28 (37.8)	N/A	0: 0 1: 15 (62.5) 2: 7 (29.2) 3: 2 (8.3)	0: 0 1: 17 (21.3) 2: 46 (57.4) 3: 17 (21.3)	N/A	N/A

ALT, alanine aminotransferase; AST, aspartate aminotransferase; CAP, controlled attenuation parameter; DBP, diastolic blood pressure; ELF, enhanced liver fibrosis; FIB-4, Fibrosis-4 index; GGT, gamma-glutamyltransferase; MRE, magnetic resonance elastography; MRI-PDFF, MRI-proton density fat fraction; NAFLD, non-alcoholic fatty liver disease; NAS, NAFLD activity score; NASH-CRN, Nonalcoholic Steatohepatitis Clinical Research Network; SBP, systolic blood pressure; T2DM, type 2 diabetes mellitus; VCTE, vibration-controlled transient elastography. Pooled baseline mean values were calculated using a sample-size-weighted mean, and pooled standard deviations were derived using the standard pooled variance formula.

Primary outcomes

Histologic resolution of MASH without worsening of fibrosis among patients with baseline F2–F3 fibrosis was reported in four studies (1,443 patients) and favored GLP-1 RAs (RR 2.96; 95% CI 1.70–5.15; p <0.001; I² = 75.4%; 4 trials; Fig. 1). Leave-one-out analysis demonstrated homogeneous results and preserved statistical significance following omission of the ESSENCE trial²⁹ (RR 3.63; 95% CI 2.32–5.67; p <0.0001; I² = 14%; 3 trials; Fig. S2). Histologic improvement in liver fibrosis stage without worsening of MASH among patients with baseline F2–F3 fibrosis was also reported in four studies (1,443 patients) and favored GLP-1 RAs (RR 1.59; 95% CI 1.32–1.90; p <0.001; I² = 0%; 4 trials; Fig. 2). Leave-one-out analysis did not identify any individual study as driving the pooled results (Fig. S3).

Fig. 1.

GLP-1 RAs were superior to placebo in histologic resolution of MASH without worsening of fibrosis.

GLP-1 RAs, glucagon-like peptide 1 receptor agonists; MASH, metabolic dysfunction-associated steatohepatitis; RR, risk ratio. Data were analyzed as the number of patients experiencing at least one event relative to the total number of patients in each group. Results were compared using RR with corresponding 95% CI, calculated using DerSimonian and Laird random-effects models. Statistical significance was set at p <0.05.

Fig. 2.

GLP-1 RAs were superior to placebo in histologic improvement in the liver fibrosis stage without worsening of MASH.

GLP-1 RAs, glucagon-like peptide 1 receptor agonists; MASH, metabolic dysfunction-associated steatohepatitis; RR, risk ratio. Data were analyzed as the number of patients experiencing at least one event relative to the total number of patients in each group. Results were compared using RR with corresponding 95% CI, calculated using DerSimonian and Laird random-effects models. Statistical significance was set at p <0.05.

Secondary outcomes

All studies (1,480 patients) reported outcomes related to AEs. GLP-1 RAs were associated with slightly higher rates of any AE (RR 1.08; 95% CI 1.01-1.15; p = 0.016; I² = 49.2%; 7 trials; Fig. 3A) and AEs related to drug or placebo (RR 1.76; 95% CI 1.41-2.19; p <0.001; I² = 0%; 3 trials; Fig. 3B). However, both groups were similar in terms of AEs leading to discontinuation of the study (RR 1.67; 95% CI 0.83-3.38; p = 0.153; I²=39.9%; 7 trials; Fig. 3C) and serious AEs (RR 1.09; 95% CI 0.84-1.40; p = 0.530; I²=0%; 7 trials; Fig. 3D). The GLP-1 RAs group presented significantly higher rates of constipation, decreased appetite, diarrhea, dyspepsia, nausea, vomiting, and fatigue (Figs. S4–S10). Both groups were similar in terms of abdominal pain, abdominal distension and bloating, arthralgia, back pain, COVID-19, dizziness, eye abnormalities, headache, metabolic disorders, musculoskeletal disorders, nervous system disorders, intestinal polyps, psychiatric disorders, urinary tract disorders, and urinary tract infection (Figs. S11–S25). The rate of acute pancreatitis, a rare but clinically important AE associated with GLP-1 RAs, was similar between groups, occurring in 0.4% of patients receiving semaglutide 2.4 mg and 0.5% of patients receiving placebo in the ESSENCE trial.²⁹ The other included studies reported no events of acute pancreatitis in either group.^27,28,[30], [31], [32], [33]

Fig. 3.

GLP-1 RAs were associated with slightly higher rates of any AE and AEs related to the drug or placebo.

However, there was no difference between the groups concerning serious AE and AE leading to discontinuation of the study. AE, adverse events; GLP-1 RAs, glucagon-like peptide 1 receptor agonists; RR, risk ratio. Data were analyzed as the number of patients experiencing at least one event relative to the total number of patients in each group. Results were compared using RR with corresponding 95% CI, calculated using DerSimonian and Laird random-effects models. Statistical significance was set at p <0.05.

Liver stiffness was evaluated in four trials: ESSENCE²⁹ (using vibration-controlled transient elastography [VCTE]), SYNERGY-NASH³¹ (using VCTE), Loomba et al. 2023³⁰ (by magnetic resonance elastography), and Newsome et al. 2020²⁸ (by VCTE). In SYNERGY-NASH,³¹ tirzepatide significantly reduced liver stiffness at week 52, with least-squares mean differences between the drug and placebo of −3.1 kPa (95% CI -5.0 to -1.2), -3.3 kPa (95% CI -5.1 to -1.5), and -3.5 kPa (95% CI -5.3 to -1.7) for the 5 mg, 10 mg, and 15 mg groups, respectively. In ESSENCE,²⁹ semaglutide 2.4 mg also produced a significant reduction in liver stiffness at week 72 vs. placebo (drug vs. placebo regarding percent change from baseline −20.4%, 95% CI -25.9 to -14.4). Additionally, the trial from Newsome et al.²⁸ demonstrated significant reductions in liver stiffness at week 72, with estimated treatment ratios of 0.74 (95% CI 0.62-0.89), 0.70 (95% CI 0.58-0.83), and 0.71 (95% CI 0.59-0.85) for the 0.1 mg, 0.2 mg, and 0.4 mg semaglutide groups, respectively, compared with placebo. By contrast, in the study from Loomba et al.,³⁰ liver stiffness showed no significant difference between semaglutide and placebo after 48 weeks (estimated treatment ratio 0.93, 95% CI 0.80-1.07; p = 0.30). Controlled attenuation parameter was only assessed in two trials. In ESSENCE,²⁹ semaglutide 2.4 mg significantly reduced hepatic steatosis at week 72 compared with placebo (drug vs. placebo regarding change from baseline -30.3 dB/m, 95% CI -39.3 to -21.3). By contrast, the study from Newsome et al. 2020²⁸ reported significant reductions only in the semaglutide 0.2 mg subgroup (estimated treatment difference -27.06; 95% CI -50.34 to -3.78).

Finally, GLP-1 RAs were superior to placebo in terms of improvement in liver aminotransferase levels (Fig. S26), MRI-proton density fat fraction (Fig. S27A), ELF (enhanced liver fibrosis) score (Fig. S27B), and NIS4 (non-invasive score 4) (Fig. S27C) glycemic control (Fig. S28), lipid profile (Fig. S29), blood pressure control (Fig. S30), and anthropometric measures (Fig. S31). However, GLP-1 RAs were similar to placebo regarding changes in the Fibrosis-4 index (Fig. S27D).

Sensitivity analysis

Across 10 randomized contrasts (48-72 weeks), GLP-1 RA-based therapy increased the probability of histologic resolution of MASH without fibrosis worsening (RR 2.56, 95% CI 1.98-3.33; p <0.001; REML; Q = 8.95 on 9 df; I² = 25.1%; τ² = 0.0578; 4 trials; Fig. S32). The 95% prediction interval was 1.40-4.68, indicating that a benefit would be expected in most future, similar settings. For fibrosis stage improvement without MASH worsening, the pooled effect was smaller but clearly favorable (RR 1.53, 95% CI 1.35-1.74; p <0.001; REML; Q = 3.46 on 9 df; I² = 0%; τ² ≈ 0; 4 trials; Fig. S33); the prediction interval (1.35-1.74) essentially overlapped the confidence interval, indicating high consistency.

In mixed-effects meta-regression by Δweight (%), the effect on MASH resolution without fibrosis worsening increased by 0.0938 log-RR per 1% greater treatment-placebo weight-loss difference (SE 0.0394; p = 0.045; Fig. S34), equivalent to ≈1.10 × higher RR per 1% and ≈1.60 × higher RR for a 5% larger weight-loss gap; residual heterogeneity was low (I² ≈ 10%) and Δweight explained ∼63% of between-arm variability (R² ≈ 63%). In contrast, for fibrosis stage improvement without MASH worsening, the Δweight slope was null (−0.0027 log-RR per 1%; SE 0.0249; p = 0.916; residual I² = 0%; Fig. S35), indicating no detectable weight-dependent gradient over 48-72 weeks.

Regarding glycemic control, exploratory ΔHbA1c moderators were not statistically significant: in the model of patients with T2DM, slopes were positive but imprecise (+0.93 log-RR per 1 percentage point; p = 0.278 for MASH resolution without fibrosis worsening; +0.18 log-RR per 1 percentage point; p = 0.716 for fibrosis improvement without MASH worsening; Figs. S36 and S37). Similarly, in models of patients without T2DM, the associations were non-significant with wide CIs (Figs. S38 and S39). Leave-one-out analyses and prediction intervals supported the robustness of both endpoints.

Overall, these moderator results reinforce that degree of weight loss, not glycemic change per se, seems the dominant correlate of histologic resolution across trials in this time frame, while fibrosis improvement appears to show no clear dependence on either Δweight or ΔHbA1c. TSA showed that the cumulative z-line crossed the boundary for effect and reached the required sample size from both primary outcomes (Figs. S40 and S41), suggesting that the pooled effect is significant and that the results are robust, making it unlikely that additional studies would change this conclusion.

Risk of bias and evidence quality assessment

The seven included RCTs followed a similar design and were considered at low risk of bias in all domains of RoB2 (Fig. S42). Regarding the GRADE assessment of evidence quality (Table S1), histologic resolution of MASH without worsening of fibrosis was rated as low certainty. It was downgraded for serious inconsistency due to clinical heterogeneity across study populations, including differences in fibrosis stages, GLP-1 RA molecules, and treatment durations, which may lead to varying responses to treatment and contribute to the observed statistical heterogeneity (I² = 66.3%). This outcome was further downgraded because of the small number of included trials, which limited our ability to assess publication bias and small-study effects, and the exclusion of abstracts, which may have increased the risk of dissemination bias. Histologic improvement in liver fibrosis stage without worsening of MASH was rated as moderate-certainty evidence, downgraded due to the limited number of trials and the potential for publication and dissemination bias.

Discussion

This meta-analysis analyzed the efficacy and safety of GLP-1 RAs in patients with MASH and liver fibrosis. We found seven high-quality, similarly designed RCTs that compared this therapy with a placebo. Our main findings were: (1) GLP-1 RAs were superior to placebo with regard to histological resolution of MASH without worsening of fibrosis and fibrosis improvement without worsening of MASH; (2) GLP-1 RAs presented lower rates of fibrosis progression; (3) in terms of safety, GLP-1 RAs presented with higher rates of constipation, decreased appetite, diarrhea, dyspepsia, nausea, vomiting, and fatigue, but no difference from placebo regarding serious AEs; and (4) GLP1-RAs were superior on improvement of liver aminotransferases, MRI- MRI-proton density fat fraction, glycemic control, lipid profile, blood pressure control, and anthropometric measures.

Since the liver does not have GLP-1 receptors, GLP-1 RAs cannot be considered to have a direct effect on liver injury. Instead, their hepatic benefits may involve several indirect mechanisms. GLP-1 RAs promote weight loss and enhance insulin sensitivity, which is crucial for reducing hepatic steatosis and inflammation, key factors in improving liver histology and MASH.^48,49 Moreover, these agents decrease hepatic lipotoxicity by lowering de novo lipogenesis and increasing fatty acid oxidation, thereby reducing hepatic fat accumulation and associated cellular injury.⁵⁰ GLP-1 RAs exert anti-inflammatory effects by inhibiting the NLRP3 inflammasome pathway and reducing the production of pro-inflammatory cytokines, thereby mitigating hepatic inflammation and improving liver histology.^51,52 Further, research indicates that GLP-RAs promote autophagy and mitophagy, which assist in degrading damaged cellular components and minimizing oxidative stress.^53,54

Fibrosis severity is a key prognostic factor in MASH, with advanced fibrosis (F3–F4) increasing the risk of liver-related complications and mortality, and a reduction in fibrosis stage indicating disease regression and improved prognosis.^31,55,56 We found a significant benefit of GLP-1 RAs in the histological improvement of liver fibrosis stage without worsening of MASH. In our sensitivity analysis evaluating whether weight change and glycemic control predict treatment response, our model suggests that fibrosis stage improvement without MASH worsening seems largely independent of the magnitude of weight or glycemic change. Additionally, the TSA reinforced our findings, indicating the required information size. Nevertheless, further studies, particularly with subgroups of baseline fibrosis stage, are warranted to assess whether the effect of GLP-1 RAs also applies to advanced-stage fibrosis.

Histological resolution of MASH is associated with a decreased risk of progression to advanced fibrosis and cirrhosis.^57,58 Reduction in lobular inflammation and hepatocellular ballooning is associated with decreased liver injury and improved liver function, which is critical for MASH resolution.^59,60 As fibrosis is the primary determinant of liver-related morbidity and mortality, its regulation significantly improves survival and long-term outcomes.^57,58 GLP-1 RAs outperformed placebo in histological resolution of MASH without worsening of fibrosis. Although the ESSENCE phase III trial,²⁹ which reported the highest MASH resolution rate (62.9%), contributed substantially to heterogeneity in the primary analysis (I² = 75.4%), this likely reflects differences in sample size and study design rather than inconsistency in the therapeutic effect of GLP-1 RAs. ESSENCE²⁹ enrolled a much larger cohort, thereby exerting the greatest statistical weight, and had a considerably longer treatment duration (240 weeks) compared with prior trials. Additionally, the use of different GLP-1 RA molecules across studies, which may display different response dynamics, could also have contributed to slight differences in clinical outcomes. In our sensitivity analysis evaluating whether weight change and glycemic control predict treatment response, our model suggests that histologic resolution of MASH without fibrosis worsening appears to scale with the magnitude of weight loss, but not with improvements in glycemic control. Additionally, the TSA reinforced our findings, indicating an adequate sample size with robust, reliable, and consistent data.

The NAS is a histological scoring system developed by the Pathology Committee of the NASH-CRN to evaluate the severity of MASLD, formerly known as nonalcoholic fatty liver disease (NAFLD), and its changes over time, particularly in clinical trials. It combines steatosis (0-3), lobular inflammation (0-2), and hepatocellular ballooning (0-2), totaling 0-8.⁶¹ A NAS ≥4 reliably indicates steatohepatitis (85% sensitivity, 81% specificity).⁶² Brunt et al.⁶³ demonstrated that a ≥2-point improvement in NAS was associated with better fibrosis outcomes. Although the individual studies included in this meta-analysis evaluated NAS improvement from baseline, all of them, except SYNERGY-NASH³¹ did not specify the cut-off used to define improvement. Therefore, we did not pool this outcome in the meta-analysis. To ensure clinically relevant outcomes, future trials should report NAS improvement using a ≥2-point cut-off.

Our findings align with previous studies, which show that GLP-1 RAs increase gastrointestinal AEs, especially abdominal pain, abdominal distention, constipation, decreased appetite, diarrhea, nausea, and vomiting.[20], [21], [22]^,[64], [65], [66] While this may limit use in some patients, GLP-1 RAs did not increase serious AEs or lead to unexpected safety concerns. This suggests that these drugs are safe for patients with MASH and liver fibrosis, but regular follow-ups are recommended to monitor tolerability.

Strengths of our study include: (1) inclusion of only high-quality, double-blinded, placebo-controlled RCTs; (2) sensitivity analyses to evaluate the indirect effect of weight loss and glycemic control on MASH outcomes; (3) TSA to ensure reliability and control for random errors; (4) data supporting GLP-1 RAs as promising for fibrosis improvement without worsening MASH, and MASH resolution without fibrosis worsening. However, our study has limitations. First, the small number of studies may limit statistical power, reliability of our conclusions, in-depth subgroup analyses, and publication bias assessment, though TSA was employed to assess the consistency of our conclusions. Second, the inclusion of patients with different baseline fibrosis stages may have contributed to the heterogeneity of some findings. However, 84.7% of our data were from patients with F2–F3 stages, and we restricted our primary efficacy outcomes to this population to improve the reliability of our findings. Third, since this meta-analysis evaluates a drug class, the treatment arm includes different types of molecules, which may have varying effects on the outcomes. For example, survodutide can exert a direct hepatic effect through glucagon activity, whereas semaglutide has primarily an indirect effect on the liver.⁶⁷ This variability could introduce spectrum bias, potentially leading to overestimation or underestimation of the findings. Nevertheless, we conducted a leave-one-out sensitivity analysis to assess whether the results were driven or disproportionately influenced by any single study. Fourth, the predominance of White participants may limit generalizability to other populations. Finally, the histologic assessment in the trial from Sanyal et al.³² (15.45% of efficacy data) relied on a single pathologist; therefore, our results should be interpreted cautiously.

This systematic review and meta-analysis evaluated the efficacy and safety of GLP-1 RAs specifically in patients with MASH and liver fibrosis. Our findings suggest that GLP-1 RAs may improve histologic features of steatohepatitis and hepatic fibrosis in individuals with ‘at-risk’ MASH. Although gastrointestinal AEs were more frequent among patients receiving GLP-1 RAs, the rates of serious AEs were comparable between treatment groups, supporting an overall favorable safety profile. However, these conclusions should be interpreted with caution. The certainty of the evidence is limited by moderate-to-high heterogeneity, differences across study populations and designs, and the small number of available trials, which restricts the precision and generalizability of the pooled estimates. While the consistent direction of benefit across studies supports a potential disease-modifying effect, additional long-term RCTs with comprehensive subgroup analyses are needed to confirm these findings, identify predictors of treatment response, and determine whether GLP-1 RA therapy can alter disease progression and improve clinically meaningful outcomes, including liver-related and all-cause morbidity and mortality.

Abbreviations

AEs, adverse events; GLP-1 RAs, glucagon-like peptide-1 receptor agonists; MASLD, metabolic dysfunction-associated steatotic liver disease; MASH, metabolic dysfunction-associated steatohepatitis; NAS, NAFLD activity score; RCTs, randomized controlled trials; REML, restricted maximum likelihood; RR, risk ratio; T2DM, type 2 diabetes mellitus; TSA, trial sequential analysis; VCTE, vibration-controlled transient elastography.

Authors’ contributions

Conceptualized and designed the study: RSB, MVF. Development of the search strategy: RSB, MVF. Conduction of the search: RSB, JMBM. Study triage: RSB, ESA, JMBM. Data extraction: LHP, AFC. Risk of bias assessment: MABM, GGB. Statistical analyses: RSB, MVF. Graphical representation of the results: RSB, LHP, AFC, JMBM, MABM. Drafting/revision of manuscript: RSB, ESA, LHP, AFG, JMBM, GGB, ASD, MVF, MFA. Final manuscript review and approval: all authors. Guarantor of the article: Rafael Dos Santos Borges.

Data availability

The primary data that support the findings of this study are derived from published literature and the relevant citations are provided in the references section. The raw datasets analyzed and collection forms are available from the corresponding author upon reasonable request.

Financial support

The authors did not receive financial support for the submitted work.

Conflicts of interest

MFA receives research funding (paid to institution) from 89Bio, Akero, Hanmi Pharmaceuticals, Madrigal, and Inventiva, and serves as a consultant/advisor to 89Bio, Akero, Boehringer-Ingelheim, Hanmi, Inventiva, Madrigal, Merck, and Novo-Nordisk. All other authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.

Please refer to the accompanying ICMJE disclosure forms for further details.

Supplementary data

The following are the Supplementary data to this article.