What this is
- This research evaluates the efficacy and safety of various pharmacological agents for metabolic dysfunction-associated steatohepatitis ().
- It incorporates recent histological data on SGLT2 inhibitors, which were previously underrepresented in meta-analyses.
- The study aims to rank treatments based on their ability to improve liver fibrosis and resolve , while also assessing tolerability.
Essence
- FGF21 analogs and SGLT2 inhibitors show the highest potential for improving liver fibrosis in , while GLP-1/GIP dual agonists excel in resolution.
Key takeaways
- FGF21 analogs ranked highest for liver fibrosis improvement with a relative risk (RR) of 2.22 (95% CI 1.40-3.54).
- SGLT2 inhibitors also demonstrated significant efficacy for fibrosis improvement with an RR of 2.27 (95% CI 1.22-4.17), indicating their potential as an oral treatment option.
- For resolution, GLP-1/GIP dual agonists showed the highest efficacy (RR 5.21), followed by FGF21 analogs (RR 3.52) and SGLT2 inhibitors (RR 2.92).
Caveats
- The evidence for SGLT2 inhibitors is based on limited trials, necessitating further validation in larger Phase 3 studies.
- Variability in treatment duration across studies may introduce bias in comparing outcomes related to fibrosis improvement and resolution.
- The absence of a formal GRADE assessment indicates that the certainty of evidence for SGLT2 inhibitors is currently low.
Definitions
- MASH: Metabolic dysfunction-associated steatohepatitis, a progressive liver disease characterized by fat accumulation and inflammation.
- NMA: Network meta-analysis, a statistical method that compares multiple interventions simultaneously by integrating direct and indirect evidence.
Simplified
Introduction
Metabolic dysfunction-associated steatohepatitis (MASH) represents the progressive phenotype of metabolic dysfunction-associated steatotic liver disease (MASLD) and has emerged as a critical global public health challenge (1, 2). Liver fibrosis is the strongest predictor of adverse long-term clinical outcomes, including cirrhosis, hepatocellular carcinoma, and liver-related mortality, in patients with MASH (3). Critically, fibrosis regression—even by a single stage—is associated with an over 80% reduction in liver-related adverse events, making the reversal of fibrosis the primary surrogate endpoint for regulatory approval (4).
The pharmacological landscape for MASH has evolved rapidly. A historic milestone was achieved with the recent FDA approval of resmetirom, a thyroid hormone receptor-β (THR-β) agonist, based on robust Phase 3 histological data (5). Concurrently, injectable biologics such as FGF21 analogs (e.g., efruxifermin, pegozafermin) and GLP-1-based polyagonists (e.g., tirzepatide) have demonstrated impressive anti-fibrotic and MASH-resolving efficacy in recent clinical trials (6–8). Furthermore, oral pleiotropic agents, notably SGLT2 inhibitors, are being actively explored for their potential to holistically address the metabolic drivers of liver injury (9).
Despite these rapid advancements, critical gaps in the evidence base remain. While recent high-quality network meta-analyses (NMAs) have successfully ranked established agents like resmetirom and GLP-1 receptor agonists (10, 11), these comprehensive reviews largely predated the availability of critical histological data for SGLT2 inhibitors. Consequently, they excluded emerging evidence, such as the recent multicenter trial by Lin et al. (9), potentially underestimating the role of this highly accessible oral antidiabetic class in fibrosis regression. The relative standing of these generic oral agents against potent novel injectables remains ill-defined.
From a mechanistic perspective, SGLT2 inhibitors are hypothesized to influence fibrosis biology through multiple pathways extending beyond glycemic control: (1) attenuation of hepatic de novo lipogenesis via reduced substrate delivery and improved insulin sensitivity; (2) mitigation of oxidative stress through enhanced ketogenesis and fatty acid oxidation; (3) potential modulation of the TGF-β1/Smad signaling pathway, attenuating hepatic stellate cell activation; and (4) reduction of systemic inflammation via weight loss and visceral adiposity reduction. These mechanisms differ from incretin-based therapies, which primarily act through weight-loss-driven substrate reduction, and from hepatocyte-targeted agents like THR-β agonists, which directly modulate hepatic lipid metabolism.
To address these limitations, we conducted a comprehensive NMA to: (1) provide an updated comparative efficacy ranking of pharmacological mechanisms for liver fibrosis improvement and MASH resolution; (2) evaluate tolerability profiles based on treatment discontinuation due to adverse events; and (3) perform a two-dimensional cluster analysis to characterize the benefit–risk balance of each drug class—specifically integrating the latest histological data for SGLT2 inhibitors.
Methods
Study registration and protocol development
This systematic review and network meta-analysis (NMA) was conducted and reported in strict accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Network Meta-Analyses (PRISMA-NMA) statement. The study protocol has been registered in the Prospective Register of Systematic Reviews (PROSPERO) database (CRD420261292638).
Eligibility criteria
The study eligibility criteria were defined based on the PICO(S) framework (Population, Intervention, Comparator, Outcomes, and Study design) as follows:
Population
Adult patients (≥18 years) with biopsy-confirmed metabolic dysfunction-associated steatohepatitis (MASH), formerly known as non-alcoholic steatohepatitis (NASH), accompanied by liver fibrosis (stages F0-F4 with predominantly F2-F3).
Intervention
Any pharmacological intervention aimed at improving MASH or liver fibrosis. The primary drug classes of interest include but are not limited to: GLP-1 receptor-based agents (mono- or dual-agonists), fibroblast growth factor 21 (FGF21) analogues, thyroid hormone receptor beta (THR-β) agonists, peroxisome proliferator-activated receptor (PPAR) agonists, and galectin-3 inhibitors.
Comparator
Placebo. To ensure the connectivity of the evidence network, all included interventions must be connected through this common comparator.
Outcomes
Primary outcome
The proportion of patients achieving an improvement in liver fibrosis, defined as a reduction of at least one stage (≥1) with no worsening of MASH.
Secondary outcomes
Study Design (S): Phase II or III randomized controlled trials (RCTs).
Exclusion criteria
The following were excluded: (1) non-randomized studies (e.g., observational studies, case reports); (2) reviews, editorials, meta-analyses, and conference abstracts (if sufficient data cannot be obtained); (3) animal or in vitro (preclinical) studies; (4) studies without a placebo control group; (5) Phase I clinical trials.
Literature retrieval strategy
Computer searches were conducted in databases including PubMed, Embase, Web of Science, Scopus and Cochrane Library. The search period was from the establishment of the databases to August 2025. The search strategy combined subject terms (such as MeSH, Emtree) and free terms. The search revolved around “disease concepts” (such as “MASH”, “NASH”, “Non-alcoholic Steatohepatitis”), “Liver Fibrosis” and “research design concepts” (such as “Randomized Controlled Trial”, “Placebo”, “Clinical Trial”) were constructed and combined. To ensure the sensitivity of the search, the search terms do not contain specific drug names but are identified during the screening stage. The search strategy was independently developed and implemented by two researchers (Y.M. and M.T.). The detailed search formula is shown in Table 1.
| Database | Search strategy |
|---|---|
| PubMed | #1 (“Metabolic Dysfunction-Associated Steatohepatitis”[Mesh] OR “Non-alcoholic Steatohepatitis”[Mesh] OR “Liver Fibrosis”[Mesh] OR “Liver Cirrhosis”[Mesh] OR MASH[tiab] OR NASH[tiab] OR “metabolic dysfunction-associated steatohepatitis”[tiab] OR “non-alcoholic steatohepatitis”[tiab] OR “nonalcoholic steatohepatitis”[tiab] OR “fatty liver”[tiab] OR “liver fibrosis”[tiab])#2 (“Randomized Controlled Trial”[Publication Type] OR “Controlled Clinical Trial”[Publication Type] OR “Randomized Controlled Trial”[Mesh] OR “Controlled Clinical Trial”[Mesh] OR randomi*[tiab] OR placebo*[tiab] OR “clinical trial”[tiab] OR “double-blind”[tiab])#3 #1 AND #2 |
| Embase | #1 ‘metabolic dysfunction associated steatohepatitis’/exp OR ‘nonalcoholic steatohepatitis’/exp OR ‘liver fibrosis’/exp OR ‘liver cirrhosis’/exp OR MASH:ab,ti OR NASH:ab,ti OR ‘metabolic dysfunction-associated steatohepatitis’:ab,ti OR ‘non-alcoholic steatohepatitis’:ab,ti OR ‘nonalcoholic steatohepatitis’:ab,ti OR ‘fatty liver’:ab,ti OR ‘liver fibrosis’:ab,ti#2 ‘randomized controlled trial’/exp OR ‘controlled clinical trial’/exp OR ‘randomization’/exp OR ‘placebo’/exp OR randomi*:ab,ti OR placebo*:ab,ti OR ‘clinical trial’:ab,ti OR ‘double-blind’:ab,ti OR ‘single-blind’:ab,ti#3 #1 AND #2 |
| Web of Science | #1 TS=(MASH) OR TS=(NASH) OR TS=(“metabolic dysfunction-associated steatohepatitis”) OR TS=(“non-alcoholic steatohepatitis”) OR TS=(“nonalcoholic steatohepatitis”) OR TS=(“fatty liver”) OR TS=(“liver fibrosis”) OR TS=(“liver cirrhosis”)#2 TS=(randomi*) OR TS=(placebo*) OR TS=(“clinical trial”) OR TS=(trial) OR TS=(“controlled trial”) OR TS=(“double blind”) OR TS=(“single blind”)#3 #1 AND #2 |
| Scopus | #1 (TITLE-ABS-KEY(MASH) OR TITLE-ABS-KEY(NASH) OR TITLE-ABS-KEY(“metabolic dysfunction-associated steatohepatitis”) OR TITLE-ABS-KEY(“non-alcoholic steatohepatitis”) OR TITLE-ABS-KEY(“nonalcoholic steatohepatitis”) OR TITLE-ABS-KEY(“fatty liver”) OR TITLE-ABS-KEY(“liver fibrosis”) OR TITLE-ABS-KEY(“liver cirrhosis”))#2 (TITLE-ABS-KEY(randomi*) OR TITLE-ABS-KEY(placebo*) OR TITLE-ABS-KEY(“clinical trial”) OR TITLE-ABS-KEY(trial) OR TITLE-ABS-KEY(“controlled trial”) OR TITLE-ABS-KEY(“double blind”) OR TITLE-ABS-KEY(“single blind”))#3 #1 AND #2 |
| Cochrane Library | #1 [mh “Non-alcoholic Fatty Liver Disease”] OR [mh “Liver Cirrhosis”] OR [mh “Fibrosis”]#2 (MASH OR NASH OR “metabolic dysfunction-associated steatohepatitis” OR “non-alcoholic steatohepatitis” OR “nonalcoholic steatohepatitis”):ti,ab,kw#3 (“fatty liver” OR “liver fibrosis” OR “hepatic fibrosis” OR “liver cirrhosis” OR “hepatic cirrhosis” OR steatohepatitis):ti,ab,kw#4 #1 OR #2 OR #3#5 [mh “Randomized Controlled Trials as Topic”] OR [mh “Randomized Controlled Trial”] OR [mh “Clinical Trials as Topic”]#6 (randomi* OR placebo* OR “clinical trial” OR “controlled trial” OR “double blind” OR “single blind” OR blind* OR RCT):ti,ab,kw#7 #5 OR #6#8 #4 AND #7 |
Literature screening and data extraction
All retrieved literature was imported into EndNote, and duplicates were removed. Two independent researchers conducted a two-stage screening process based on titles, abstracts, and full texts. Discrepancies were resolved through consensus or a third-party adjudicator.
To maximize the robustness of the safety and tolerability analysis, trials reporting relevant safety outcomes (i.e., adverse events leading to study discontinuation) but lacking biopsy-proven histological endpoints at the designated follow-up were exclusively incorporated into the safety network. To ensure mechanistic consistency and respond to concerns regarding class purity, we excluded data from dual SGLT1/2 inhibitors (e.g., licogliflozin (12)). The SGLT2 inhibitor node was restricted to selective SGLT2 inhibitors (e.g., dapagliflozin) for both efficacy and tolerability analyses. This methodological approach ensures a broader and more reliable evidence base for the safety profile of the drug class without compromising the strict inclusion criteria for histological efficacy.
Data extraction and quality evaluation
Data extraction
Two researchers (Y.M. and M.T.) independently extracted the information included in the study using a pre-designed standardized Excel data extraction table. The extracted content includes:
Basic research information
First author, publication year, research design (such as Phase II/III, multicenter), sample size, source of research funding.
Baseline characteristics of the patient
Age, gender, BMI, proportion of diabetes mellitus (T2DM), baseline stage of liver fibrosis (proportions of F2, F3, and F4).
Intervention and control information
Drug name, mechanism of action, dosage, treatment duration.
Outcome data
For the above primary and secondary outcomes, the number of respondents (n) and the total number (N) of each group were extracted.
When the data is missing or the report is unclear, we attempted to contact the original author of the study to obtain supplementary information.
Risk of bias assessment
Two researchers independently assessed the Risk of Bias for inclusion in RCTs using the RoB 2 (Risk of Bias 2) tool recommended by Cochrane. The assessment was conducted in five areas: (1) Randomization process; (2) Deviation from the predetermined intervention; (3) Missing outcome data; (4) Outcome measurement; (5) Selectively report results. Each area was rated as “low risk”, “some concern” or “high risk”.
Data analysis
This study conducted a network meta-analysis using R software (mainly netmeta) and Stata software (network command). The primary NMA and ranking (P-scores) were performed using the netmeta package in R; Stata’s network command was used for supplementary consistency checks and funnel plot visualization.
Data merging and modeling
For all binary outcomes (primary and secondary outcomes), the Risk Ratio (RR) and its 95% confidence interval (CI) was used as effect sizes. We adopted a frequentist random-effects model for NMA, which can take into account both intra-study and inter-study heterogeneity simultaneously.
Evidence network diagram
We drew an evidence network diagram to visualize the direct and indirect comparative relationships among all intervention measures. The size of the nodes represents the total sample size, and the thickness of the connection lines represents the number of studies.
Consistency test
To ensure the reliability of NMA results, we assessed the consistency between direct comparison and indirect comparison evidence in the network. The “Node-splitting” model was adopted for local consistency test. If the P value is greater than 0.05, it is considered that the consistency between direct and indirect evidence is good.
NMA results and ranking
Calculate the RR values and 95% CI for all drug pairwise comparisons (including those with placebo). Meanwhile, interventions were ranked using P-scores, which represent the frequentist equivalent to the Surface Under the Cumulative Ranking curve (SUCRA). The P-score ranges from 0 to 1, with values closer to 1 indicating a higher certainty that an intervention is superior to others.
Heterogeneity and publication bias
We assessed the statistical heterogeneity across the entire network. Comparison-adjusted funnel plots were used to evaluate the possible publication bias or small sample effect in the network.
Subgroup analysis
If the data permit, we conducted subgroup analysis based on preset covariates, such as exploring the impact of different baseline liver fibrosis degrees (F2 vs F3) on drug efficacy.
Results
Literature screening results and evidence network
After systematic retrieval and screening, A total of 34 RCTs from 33 eligible publications were ultimately included in the network (Figure 1). The main characteristics of the included studies are shown in Table 2. The network of eligible comparisons for the primary outcome (fibrosis improvement) demonstrates that placebo served as the central node connecting diverse pharmacological mechanisms (Figure 2). Add-on regimens, such as efruxifermin administered on a background of GLP-1 receptor agonist therapy (Harrison 2025, Cohort D), were modeled as separate combination-therapy nodes and are distinguished from monotherapy nodes in Table 2. Of the 34 included RCTs, 32 studies contributed to the fibrosis improvement network, 32 to the MASH resolution network, and 2 exclusively to the safety network (i.e., trials reporting adverse event data but lacking biopsy-proven histological endpoints at designated follow-up).
Flow diagram of literature search and selection process. The study selection procedure complies with the updated PRISMA 2020 guidelines.
Network plot of eligible comparisons for liver fibrosis improvement. The size of the nodes corresponds to the total number of patients assigned to each intervention, and the thickness of the connecting lines correlates with the number of direct head-to-head comparisons available. SGLT2 inhibitors (highlighted in red) are successfully connected to the network via the central placebo node.
| First author & year | Country/region | Study population | Intervention group protocol | Control group protocol | Follow-up duration | Primary outcome measures | Secondary outcome measures | Key study results | Blinding/study design |
|---|---|---|---|---|---|---|---|---|---|
| Harrison, 2024 [MAESTRO-NASH] () [5] | Multinational (15 countries) | Biopsy-confirmed NASH with F1B/F2/F3 fibrosis; n=966 primary population | Resmetirom 80 mg or 100 mg oral once daily; 52 weeks | Placebo oral once daily; 52 weeks | 52 weeks | 1. NASH resolution without fibrosis worsening;2. Fibrosis improvement ≥1 stage without NAS worsening | LDL-C change; liver enzymes, MRI-PDFF, liver stiffness, lipid profiles | Both resmetirom doses superior to placebo for both primary outcomes; reduced LDL-C and liver fat; common AEs: diarrhea, nausea | Double-blind, randomized, placebo-controlled, parallel, phase 3 |
| Harrison, 2021 [BALANCED] () [6] | United States | Adults with NASH (F1–F3 fibrosis, HFF≥10% by MRI-PDFF); n=80 (21 placebo, 19 efruxifermin 28mg, 20 50mg, 20 70mg) | Efruxifermin: 28mg/50mg/70mg, weekly subcutaneous injection, 16 weeks | Placebo, weekly subcutaneous injection, 16 weeks | 16 weeks (treatment); 4 weeks (safety follow-up) | Absolute change in hepatic fat fraction (HFF) from baseline at week 12 (MRI-PDFF) | Relative change in HFF; proportion of HFF responders; NAS response; ALT change; safety and tolerability | Efruxifermin significantly reduced HFF vs placebo; improved liver enzymes, fibrosis markers, metabolic parameters; 55% had fibrosis improvement; mostly mild-moderate GI adverse events | Double-blind, placebo-controlled, parallel-group, phase 2a randomized controlled trial |
| Loomba, 2023 [ENLIVEN] () [7] | USA | Biopsy-proven noncirrhotic NASH with F2/F3 fibrosis; n=222 randomized, 219 treated | Pegozafermin: 15 mg weekly SC, 30 mg weekly SC, 44 mg q2w SC; 24 weeks | Placebo: weekly or q2w SC; 24 weeks | 24 weeks | 1. Fibrosis improvement ≥1 stage without NASH worsening;2. NASH resolution without fibrosis worsening | NAFLD activity score improvement ≥2 points; liver fat (MRI-PDFF), liver enzymes, fibrosis markers, metabolic parameters | 30 mg weekly and 44 mg q2w pegozafermin significantly improved fibrosis and NASH resolution vs placebo; common AEs: nausea, diarrhea | Double-blind, randomized, placebo-controlled, parallel, phase 2b |
| Loomba, 2024 [SYNERGY-NASH] () [8] | 10 countries (US, Mexico, Europe, Israel, Japan) | Biopsy-confirmed MASH with F2-F3 fibrosis; NAFLD activity score ≥4; BMI 27-50; n=190 | Tirzepatide 5 mg, 10 mg, 15 mg: subcutaneous once weekly, 52-week dose escalation | Placebo subcutaneous once weekly, 52-week dose-matched escalation | 52 weeks treatment + 4 weeks safety follow-up | Resolution of MASH without fibrosis worsening at 52 weeks | ≥1 fibrosis stage improvement without MASH worsening; changes in NAFLD activity score, liver fat, liver enzymes, body weight, fibrosis biomarkers | All tirzepatide doses significantly improved MASH resolution vs placebo; fibrosis improvement rate higher than placebo; AEs mainly mild-moderate gastrointestinal events | Double-blind, randomized, placebo-controlled, parallel design |
| Lin, 2025 [DEAN trial] () [9] | China | Biopsy-diagnosed MASH (with/without T2DM); n=154 | Dapagliflozin 10 mg oral once daily; 48 weeks | Matching placebo oral once daily; 48 weeks | 48 weeks | MASH improvement (NAS ≥2-point reduction or NAS ≤3) without fibrosis worsening | 1. MASH resolution without fibrosis worsening;2. Fibrosis improvement ≥1 stage without MASH worsening;Histological subscores, liver stiffness, metabolic parameters | Dapagliflozin significantly improved MASH resolution and fibrosis regression vs placebo; well-tolerated | Double-blind, randomized, placebo-controlled, parallel |
| Shankar, 2024 [PROXYMO] () [13] | United States, Puerto Rico | Biopsy-proven noncirrhotic MASH with fibrosis (F1-F3); BMI ≥30 kg/m²; n=74 | Cotadutide 300 μg: subcutaneous once daily, 19-week dose titration; Cotadutide 600 μg: subcutaneous once daily, 19-week dose titration | Placebo subcutaneous once daily, 19-week dose-matched titration | 19 weeks treatment + 4 weeks safety follow-up | Incidence of TEAEs and serious TEAEs | Changes in HFF (MRI-PDFF), ALT, AST, GGT, body weight, BMI, liver volume, metabolic parameters, non-invasive liver biomarkers | Cotadutide 600 μg significantly reduced HFF, ALT, AST vs placebo; AEs were mainly mild-moderate gastrointestinal events | Double-blind, randomized, placebo-controlled, parallel design |
| Armstrong, 2016 [LEAN] () [14] | UK | Overweight patients with biopsy-confirmed definite NASH; n=52 | Liraglutide 1.8 mg; subcutaneous once daily; 48 weeks (14-day titration) | Placebo subcutaneous once daily | 48 weeks (treatment); 60 weeks (total) | Resolution of definite NASH with no fibrosis worsening at week 48 | Changes in NAFLD activity score, liver histology components, liver enzymes, metabolic parameters, quality of life | Liraglutide significantly increased NASH resolution vs placebo; reduced fibrosis progression, weight, and HbA1c; well-tolerated | Multicenter, randomized, double-blind, placebo-controlled, phase 2 trial |
| Harrison, 2023 [HARMONY] () [15] | USA | Adults with biopsy-confirmed NASH, F2/F3 fibrosis, NAS ≥4; n=128 | Efruxifermin 28 mg or 50 mg; subcutaneous once weekly; 24 weeks | Placebo subcutaneous once weekly | 24 weeks (primary analysis); 96 weeks (ongoing) | Proportion of patients with ≥1-stage fibrosis improvement and no NASH worsening at week 24 | NASH resolution without fibrosis worsening; fibrosis improvement; changes in HFF, liver stiffness, metabolic markers, bodyweight; safety | Efruxifermin 28/50 mg significantly improved fibrosis resolution and NASH resolution vs placebo; reduced liver fat and improved metabolic profiles; acceptable safety | Multicenter, randomized, double-blind, placebo-controlled, parallel-group, phase 2b trial |
| Cusi, 2016 [Long-Term Pioglitazone for NASH] () [16] | United States | NASH with prediabetes or type 2 diabetes, n=101 | Pioglitazone 45 mg once daily, oral, 18 months (double-blind) + 18 months open-label | Placebo once daily, oral | 36 months | ≥2-point reduction in NAS in 2 histologic categories without fibrosis worsening at 18 months | NASH resolution, histologic improvement, hepatic triglyceride content, insulin sensitivity | Pioglitazone significantly improved primary and secondary outcomes; benefits sustained at 36 months | Randomized, double-blind, placebo-controlled, parallel-group, single-center |
| Loomba, 2024 [FALCON 1] () [17] | United States, Japan | NASH with stage 3 bridging fibrosis, n=197 | Pegbelfermin 10 mg, 20 mg, 40 mg, subcutaneous once weekly, 48 weeks | Placebo subcutaneous once weekly | 48 weeks | ≥1-stage fibrosis improvement without NASH worsening or NASH improvement without fibrosis worsening at week 24 | Histologic fibrosis/NASH improvement, MRI-PDFF, MRE, liver enzymes, PRO-C3, adiponectin | Primary endpoint not met; pegbelfermin improved hepatic fat, liver stiffness, and biomarkers vs placebo; well-tolerated | Randomized, double-blind, placebo-controlled, parallel-group phase 2b trial |
| Harrison, 2020 [STELLAR-3] () [18] | Multinational (26 countries for STELLAR-3; 21 countries for STELLAR-4) | STELLAR-3: NASH with bridging fibrosis (F3), n=808; STELLAR-4: NASH with compensated cirrhosis (F4), n=877 | Selonsertib 18 mg once daily; Selonsertib 6 mg once daily, oral administration, 48 weeks | Placebo once daily, oral | 48 weeks | ≥1-stage fibrosis improvement without NASH worsening at week 48 | Changes in noninvasive fibrosis tests, progression to cirrhosis, liver-related clinical events, NASH resolution | Selonsertib showed target inhibition but no significant antifibrotic effect vs placebo; similar safety profile | Randomized, double-blind, placebo-controlled, parallel-group phase III trials |
| Harrison, 2019 [MGL-3196-05] () [19] | USA | Biopsy-confirmed NASH with F1–F3 fibrosis, MRI-PDFF hepatic fat ≥10%; randomized n=125 (84 resmetirom, 41 placebo) | Resmetirom (MGL-3196), initial 80 mg oral once daily, dose adjusted at week 4, 36 weeks | Placebo, oral once daily | 36 weeks | Relative change in MRI-PDFF hepatic fat at week 12 | Proportion of patients with ≥30% hepatic fat reduction; absolute hepatic fat change; histological improvement; liver enzymes; lipid and fibrosis biomarkers | Significant reduction in hepatic fat at 12 and 36 weeks; improved liver histology, enzymes, and lipid profiles; mild transient GI adverse events | Randomized, double-blind, placebo-controlled, parallel-group phase II trial |
| Harrison, 2020 [EMMINENCE] () [20] | USA | Biopsy-confirmed NASH with F1–F3 fibrosis; total randomized n=392 | MSDC-0602K, 62.5 mg, 125 mg, or 250 mg, oral once daily, 52 weeks | Placebo, oral once daily | 52 weeks | Hepatic histological improvement (≥2-point NAS reduction, ≥1-point reduction in ballooning or lobular inflammation, no fibrosis increase) at 12 months | NAS improvement without fibrosis worsening; NASH resolution; fibrosis reduction; changes in insulin sensitivity, liver injury, fibrosis markers | No significant effect on primary/secondary histological endpoints; high doses significantly reduced glucose, HbA1c, insulin, liver enzymes, and NAS; well tolerated without dose-limiting side effects | Randomized, double-blind, placebo-controlled, parallel-group phase IIb trial |
| Harrison, 2020 [Aldafermin Phase II] () [21] | USA | NASH with stage 2–3 fibrosis, NAS ≥4, liver fat content ≥8%; N = 78 | Aldafermin 1 mg, once daily, subcutaneous injection, 24 weeks | Placebo, once daily, subcutaneous injection | 24 weeks | Change in absolute liver fat content (MRI-PDFF) | Markers of target engagement; liver enzymes; fibrosis biomarkers; histological endpoints (fibrosis improvement, NASH resolution); safety | Significant reduction in liver fat content; trend toward higher fibrosis improvement and NASH resolution;combined histological endpoint significantly better with aldafermin; well-toleratedpost-hoc | Double-blind, randomized, placebo-controlled, parallel-group, phase 2 trial |
| Okanoue, 2021 () [22] | Japan | NASH with stage 2–3 fibrosis, NAS ≥4; N = 48 | Apararenone 10 mg, once daily, oral, 72 weeks | Placebo, once daily, oral | 72 weeks (plus 8-week follow-up) | Percent change in serum ALT from baseline to 24 weeks | Changes in AST, fibrosis markers (FIB-4, ELF, type IV collagen 7S, P3NP); histological changes; liver stiffness; safety | Greater reduction in ALT/AST; significant improvement in fibrosis markers; trend toward higher histological fibrosis improvement; safe and well-tolerated; transient mild serum potassium elevation | Double-blind, randomized, placebo-controlled, parallel-group, phase 2 trial |
| Friedman, 2018 [CENTAUR] () [23] | Multinational (Australia, Belgium, France, Germany, Hong Kong, Italy, Poland, Spain, UK, USA) | NASH with liver fibrosis (stage 1–3), NAS ≥4; N = 289 | Cenicriviroc 150 mg, once daily, oral, 1 year | Placebo, once daily, oral | 48 weeks | 2-point improvement in NAS and no worsening of fibrosis | Resolution of steatohepatitis and no fibrosis worsening; ≥1-stage fibrosis improvement and no steatohepatitis worsening; histological changes; inflammatory biomarkers; safety | Primary endpoint not met; significantly more patients had ≥1-stage fibrosis improvement with CVC vs placebo (20% vs 10%); safety comparable to placebo | Double-blind, randomized, placebo-controlled, parallel-group, phase 2b trial |
| Harrison, 2022 [BALANCED-C] () [24] | United States | Adults with compensated NASH cirrhosis (F4); N = 30 | Efruxifermin 50 mg, subcutaneous, once weekly, 16 weeks | Placebo, subcutaneous, once weekly, 16 weeks | 4 weeks (post-treatment) | Safety and tolerability of efruxifermin | Changes in non-invasive fibrosis/liver injury markers, metabolic parameters, histology | Efruxifermin improved fibrosis/metabolic markers; 33% achieved fibrosis improvement, 25% NASH resolution; well-tolerated | Double-blind, placebo-controlled, parallel-group, phase 2a |
| Harrison, 2020 () [25] | Multiple (United States, Spain, Germany) | Adults with biopsy-proven NASH, F1–F3 fibrosis, NAS ≥4; N = 318 | Emricasan 5 mg or 50 mg, oral, twice daily, 72 weeks | Placebo, oral, twice daily, 72 weeks | 4 weeks (post-treatment) | Improvement in fibrosis stage without NASH worsening | NASH resolution, NAS response, changes in liver histology/biomarkers | Emricasan did not improve fibrosis/NASH; may worsen fibrosis/ballooning; reduced serum ALT | Double-blind, placebo-controlled, parallel-group, phase 2 |
| Harrison, 2025 [Cohort D] () [26] | United States | Adults with T2D, MASH with F1–F3 fibrosis, stable GLP-1RA therapy; N = 31 | Efruxifermin 50 mg, subcutaneous, once weekly, 12 weeks | Placebo, subcutaneous, once weekly, 12 weeks | 4 weeks (post-treatment) | Safety and tolerability of efruxifermin + GLP-1RA | Changes in HFF, liver injury/fibrosis markers, metabolic parameters | Efruxifermin + GLP-1RA significantly reduced HFF, improved liver fibrosis/metabolic markers; well-tolerated | Double-blind, placebo-controlled, parallel-group, phase 2b |
| Sanyal, 2024 () [27] | 25 countries (global multi-center) | Biopsy-confirmed MASH with fibrosis F1–F3; n=293; 39% with type 2 diabetes | Survodutide 2.4 mg, 4.8 mg, 6.0 mg; once-weekly subcutaneous injection; 48 weeks (24-week dose escalation + 24-week maintenance) | Placebo once-weekly subcutaneous injection | 48 weeks | Histologic improvement in MASH without fibrosis worsening | Liver fat content reduction ≥30% (MRI-PDFF); fibrosis improvement ≥1 stage; MASH resolution; changes in liver enzymes | All survodutide doses improved primary endpoint vs placebo; 4.8 mg had highest efficacy; dose-related GI adverse events | Double-blind, randomized, placebo-controlled, parallel-group, phase 2 dose-finding trial |
| Francque, 2021 [NATIVE] () [28] | 16 countries (Europe, North America, Australia, etc.) | Noncirrhotic, highly active NASH; n=247; 42% with type 2 diabetes, 76% with significant/advanced fibrosis | Lanifibranor 1200 mg once daily, oral; Lanifibranor 800 mg once daily, oral; 24 weeks | Placebo once daily, oral | 24 weeks | Decrease ≥2 points in SAF-A score without fibrosis worsening | NASH resolution without fibrosis worsening; fibrosis improvement ≥1 stage without NASH worsening; composite of NASH resolution + fibrosis improvement ≥1 stage; changes in liver enzymes, lipid, glycemic, inflammatory and fibrosis biomarkers | 1200 mg lanifibranor significantly improved primary endpoint vs placebo; both doses improved secondary histological and biochemical endpoints | Double-blind, randomized, placebo-controlled, parallel-group, phase 2b trial |
| Loomba, 2021 [ATLAS] () [29] | United States, Canada, Australia, New Zealand, Hong Kong | Patients with NASH-induced bridging fibrosis (F3) or compensated cirrhosis (F4); 392 enrolled, 317 completed treatment | 1. Selonsertib monotherapy: 18 mg, once daily, oral, 48 weeks;2. Cilofexor monotherapy: 30 mg, once daily, oral, 48 weeks;3. Firsocostat monotherapy: 20 mg, once daily, oral, 48 weeks;4. Cilofexor + selonsertib: 30 mg + 18 mg, once daily, oral, 48 weeks;5. Firsocostat + selonsertib: 20 mg + 18 mg, once daily, oral, 48 weeks;6. Cilofexor + firsocostat: 30 mg + 20 mg, once daily, oral, 48 weeks | Placebo, once daily, oral, 48 weeks | 48 weeks | Proportion of patients with ≥1-stage improvement in fibrosis without worsening of NASH at week 48 | Changes in NAS, NASH resolution, liver biochemistry, noninvasive fibrosis markers (ELF, VCTE), MRI-PDFF, ML-assessed histologic parameters, safety parameters | Cilofexor/firsocostat showed highest fibrosis improvement rate (21% vs 11% placebo, p=0.17); significantly improved NAS, liver enzymes, and ML-assessed fibrosis; well-tolerated with pruritus as main AE | Double-blind, randomized, placebo-controlled, parallel-group, phase 2b trial |
| Newsome, 2020 () [30] | 16 countries (global) | Adults with biopsy-confirmed NASH (F1–F3 fibrosis); n=320 (230 with F2–F3) | Semaglutide 0.1 mg, 0.2 mg, 0.4 mg; once daily; subcutaneous injection; 72 weeks | Placebo; once daily; subcutaneous injection; 72 weeks | 72 weeks (treatment); 7 weeks (follow-up) | NASH resolution with no fibrosis worsening (F2–F3 subgroup) | Improvement of ≥1 fibrosis stage with no NASH worsening; changes in liver enzymes, body weight, glycemic/lipid parameters, liver stiffness | Semaglutide 0.4 mg significantly increased NASH resolution vs placebo; no significant fibrosis improvement difference; dose-dependent weight loss and liver enzyme reduction; higher GI adverse events | Double-blind, placebo-controlled, parallel-group, phase 2 randomized controlled trial |
| Younossi, 2019 [REGENERATE] () [31] | Multinational (332 centers in 20 countries) | Adults with NASH, NAS ≥4, fibrosis stage F2–F3 (or F1 with comorbidities); primary analysis n=931 (F2–F3); safety population n=1968 | Obeticholic acid 10 mg or 25 mg, oral, once daily | Placebo, oral, once daily | 18 months (interim analysis) | 1. Fibrosis improvement ≥1 stage without NASH worsening; 2. NASH resolution without fibrosis worsening | Histological improvement of NASH features; NAS improvement; changes in liver enzymes, lipids, body weight | Obeticholic acid 25 mg significantly improved fibrosis vs placebo; NASH resolution endpoint not met; pruritus and LDL increase were common AEs | Double-blind, randomized, placebo-controlled, parallel design |
| Harrison, 2023 [DESTINY-1] () [32] | United States (28 sites) | Adults aged 18–75 years with biopsy-confirmed NASH, MRI-PDFF ≥8%, NAS ≥4, fibrosis F1–F3; n=117 (ITT) | PXL065 7.5 mg, 15 mg, 22.5 mg, oral, once daily | Placebo, oral, once daily | 36 weeks | Relative percentage change in liver fat content (LFC) by MRI-PDFF | Absolute LFC change; histological improvement; fibrosis biomarkers; metabolic parameters (HbA1c, insulin sensitivity); liver enzymes | All PXL065 doses significantly reduced LFC; favorable trends in fibrosis/histology; improved glycemic control; no dose-dependent weight gain/peripheral edema | Double-blind, randomized, placebo-controlled, parallel design |
| Chan, 2017 () [33] | Malaysia | Biopsy-proven NASH (NAS≥4), non-cirrhotic; n=99 | Silymarin 700 mg, oral, three times daily, 48 weeks | Placebo, oral, three times daily | 48 weeks | ≥30% reduction in NAS | Changes in steatosis, inflammation, ballooning, fibrosis, liver stiffness, metabolic parameters | Silymarin did not meet primary outcome; significantly improved fibrosis vs placebo; safe and well-tolerated | Double-blind, randomized, placebo-controlled, parallel, single-center |
| Ratziu, 2016 [GOLDEN-505] () [34] | Europe, United States | Non-cirrhotic NASH patients; n=276 (ITT: 274) | Elafibranor 80 mg or 120 mg, oral, once daily, 52 weeks | Placebo, oral, once daily | 52 weeks + 3 months post-treatment | Resolution of NASH without fibrosis worsening (protocol-defined) | Changes in NAS, liver enzymes, lipids, glycemic parameters, inflammatory markers, fibrosis | 120 mg elafibranor significantly improved NASH resolution (modified definition) vs placebo; improved metabolic profiles; well-tolerated | Double-blind, randomized, placebo-controlled, parallel, multicenter |
| Loomba, 2024 [FASCINATE-2] () [35] | USA, Canada, Poland | Biopsy-confirmed MASH with F2-F3 fibrosis; 168 randomized (112 denifanstat, 56 placebo) | Denifanstat 50 mg oral once daily; 52 weeks | Oral placebo once daily | 52 weeks | 1. ≥2-point NAS improvement without fibrosis worsening;2. MASH resolution with ≥2-point NAS improvement without fibrosis worsening | Fibrosis improvement without steatohepatitis worsening, MASH resolution without fibrosis worsening, liver fat reduction (MRI-PDFF), qFibrosis score change, liver enzymes, lipid parameters | Denifanstat significantly improved both primary endpoints; increased fibrosis improvement and MASH resolution vs placebo; well-tolerated (mainly grade 1–2 AEs) | Multicenter, double-blind, randomized, placebo-controlled, parallel-group phase 2b trial |
| Sanyal, 2025 [ESSENCE] () [36] | 37 countries (global) | Biopsy-confirmed MASH with F2-F3 fibrosis; 800 randomized (534 semaglutide, 266 placebo) in interim analysis | Semaglutide 2.4 mg subcutaneous once weekly (dose-escalated over 16 weeks); 72 weeks | Subcutaneous placebo once weekly | 72 weeks (interim) | 1. MASH resolution without fibrosis worsening;2. Fibrosis reduction without steatohepatitis worsening | Combined MASH resolution and fibrosis reduction, body weight change, SF-36 bodily pain score, non-invasive liver markers, cardiometabolic parameters | Semaglutide significantly improved both primary endpoints; greater weight loss and favorable cardiometabolic changes vs placebo; GI AEs more common | Multicenter, double-blind, randomized, placebo-controlled, parallel-group phase 3 trial |
| Harrison, 2022 [ALPINE 2/3] () [37] | USA | Biopsy-confirmed NASH with F2-F3 fibrosis; 171 randomized (43 placebo, 43 aldafermin 0.3mg, 42 1.0mg, 43 3.0mg) | Aldafermin 0.3mg, 1.0mg, 3.0mg subcutaneous once daily; 24 weeks | Subcutaneous placebo once daily | 24 weeks | ≥1-stage liver fibrosis improvement without NASH worsening | NASH resolution without fibrosis worsening, ≥2-point NAS reduction without fibrosis worsening, liver fat (MRI-PDFF), liver enzymes, fibrosis markers, bile acid parameters | No significant dose-response on primary endpoint; dose-dependent improvements in liver fat, enzymes, fibrosis markers and NASH resolution; well-tolerated | Multicenter, double-blind, randomized, placebo-controlled, parallel-group phase 2b trial |
| Hoofnagle, 2013 [PIVENS] () [38] | United States | Nondiabetic adults with biopsy-proven NASH; n=139 (71 vitamin E, 68 placebo) | Vitamin E (RRR-α-tocopherol): 800 IU once daily, oral administration, 96 weeks | Placebo once daily, oral administration, 96 weeks | 96 weeks (treatment); 24 weeks (post-treatment) | Association between ALT changes and liver histology improvement; effect of weight change on ALT and histology | ALT response rate; NAS score change; fibrosis score change; NASH resolution rate; weight change effect | Vitamin E group had higher ALT response (48% vs 16%); ALT response correlated with NAS improvement; weight loss improved histology; vitamin E effect independent of weight loss | Double-blind, parallel-group, randomized controlled trial (analysis of PIVENS trial)post-hoc |
| Anstee, 2024 [AURORA] () [39] | Multinational (454 international sites) | Adults aged 18–75 years with NASH and stage 2/3 liver fibrosis; total randomized n=1778 (part 1: n=1293; part 2: n=485) | Cenicriviroc (CVC) 150 mg, oral, once daily | Placebo, oral, once daily | Part 1: 12 months; Part 2: up to 60 months | Proportion of patients with fibrosis improvement ≥1 stage without worsening of NASH at month 12 | Fibrosis improvement ≥2 stages without NASH worsening; fibrosis improvement ≥1 stage regardless of NASH; complete NASH resolution without fibrosis worsening; changes in ALT, AST, FIB-4, NFS, monocyte count | CVC did not achieve superior efficacy vs placebo for primary/secondary endpoints; similar safety profile between groups | Double-blind, randomized, placebo-controlled, 2-part parallel design |
| Noureddin, 2025 [SYMMETRY] () [40] | United States, Puerto Rico, Mexico | Adults with MASH-induced compensated cirrhosis (stage 4 fibrosis, Child–Pugh A); n=181 randomized | Efruxifermin 28 mg or 50 mg; once weekly; subcutaneous injection; 96 weeks | Placebo; once weekly; subcutaneous injection; 96 weeks | 96 weeks | Reduction of ≥1 fibrosis stage without MASH worsening at week 36 | Reduction of ≥1 fibrosis stage without MASH worsening at week 96; MASH resolution; changes in noninvasive fibrosis markers, liver enzymes, metabolic parameters; safety | Efruxifermin did not meet primary outcome at week 36; 50 mg efruxifermin improved fibrosis without MASH worsening at week 96; improved liver injury, fibrosis markers, and metabolic profiles; mainly mild-moderate GI adverse events | Double-blind, placebo-controlled, parallel-group, phase 2b randomized controlled trial |
Transitivity assessment and baseline characteristics
To assess the transitivity assumption for valid indirect comparisons, we constructed a Baseline Characteristics Audit Table (). Key potential effect modifiers were examined: mean age (range 46–61 years across trials), BMI (30–38 kg/m² in most trials), T2DM prevalence (variable by drug class; SGLT2 inhibitor trials enrolled predominantly diabetic populations), and fibrosis stage distribution (most trials required F2-F3 at baseline). 1
While inherent clinical heterogeneity exists, core metabolic parameters demonstrated sufficient similarity to uphold the transitivity assumption. We acknowledge that T2DM prevalence differences for SGLT2 inhibitors represent a potential effect modifier, addressed in the Discussion.
Risk of bias assessment
Overall, the methodological quality of the included RCTs was generally acceptable, as summarized in the risk of bias summary plot (). Across most domains, the majority of studies were considered to carry a low risk of bias, though inherent challenges in biopsy-driven MASH trials—including missing paired biopsies and protocol deviations—should be noted when interpreting these assessments. A limited number of trials were flagged with “some concerns,” primarily originating from minor deviations from intended interventions (Domain 2) or missing outcome data (Domain 3) in certain trial arms. Notably, none of the final included RCTs were evaluated as having a high overall risk of bias. The comprehensive, study-level risk of bias evaluations (traffic light plot) detailing each domain for the individual trials are provided in. 1 1
Primary outcome: liver fibrosis improvement
The comparative efficacy of all interventions against placebo for fibrosis improvement is illustrated in the forest plot (Figure 3). Several monotherapies showed statistically significant efficacy. FGF21 analogs ranked as the most effective monotherapy class (RR 2.22, 95% CI 1.40–3.54), followed closely by SGLT2 inhibitors (RR 2.27, 95% CI 1.22–4.17). Other classes also demonstrated significant benefits, including GLP-1/GIP dual agonists (RR 1.79), THR-β agonists (RR 1.61), and GLP-1 RAs (RR 1.51). In contrast, Caspase inhibitors and ASK1 inhibitors failed to show significant improvement. The league table (Supplementary Table 1) further corroborates that SGLT2 inhibitors consistently generated favorable point estimates against other metabolic modulators.
Forest plot of relative risks (RRs) for liver fibrosis improvement (Primary Outcome). Data is presented as RRs with 95% confidence intervals (CIs) derived from the random-effects network meta-analysis. An RR > 1 favors the active treatment over placebo. Interventions are sorted in descending order by their respective P-scores.
Secondary outcome: MASH resolution
For MASH resolution without worsening of fibrosis, the NMA revealed distinct efficacy profiles (Figure 4). GLP-1/GIP dual agonists exhibited the most profound efficacy (RR 5.21, 95% CI 1.18–62.52). FGF21 analogs (RR 3.52) and SGLT2 inhibitors (RR 2.92, 95% CI 1.18–7.26) also showed substantial and statistically significant efficacy. Standard GLP-1 RAs were effective but numerically inferior to the dual agonists and SGLT2 inhibitors for this specific endpoint.
Forest plot of relative risks (RRs) for MASH resolution (Secondary Outcome). Data is presented as RRs with 95% CIs compared with placebo. An RR > 1 favors active treatment. Interventions are sorted in descending order by P-score.
Cluster analysis: efficacy vs. tolerability profile
To comprehensively evaluate the benefit-risk profile, a two-dimensional cluster analysis was performed using P-scores for fibrosis improvement and tolerability (avoidance of discontinuation due to adverse events) (Figure 5). Notably, SGLT2 inhibitors numerically occupied the upper-right quadrant (“Optimal Zone”), suggesting a potentially favorable balance of anti-fibrotic efficacy and tolerability; however, this observation should be interpreted cautiously given the sparse underlying data. In contrast, while FGF21 analogs exhibited comparable efficacy (Cluster 2: Fibrosis-Dominant Efficacy), they were associated with a lower tolerability P-score, reflecting higher AE-related discontinuation rates.
Cluster plot analysis of efficacy versus tolerability profile. The two-dimensional plot illustrates the grouping of interventions based on their P-scores for tolerability (x-axis, defined as the avoidance of treatment discontinuation due to adverse events) and fibrosis improvement (y-axis). The upper-right quadrant indicates the “Optimal Zone,” representing a highly favorable balance of both high efficacy and high tolerability. Note: The tolerability assessment for the SGLT2 inhibitor node is based on a single biopsy-proven RCT (n=1 trial), following the exclusion of dual SGLT1/2 inhibitors to maintain a pure-class mechanistic evaluation.
Heterogeneity, inconsistency, and sensitivity analysis
The assessment of inconsistency using the node-splitting method revealed no significant differences between direct and indirect evidence (all P > 0.05) (Supplementary Table 4), supporting the validity of the network. Furthermore, sensitivity analysis excluding studies with small sample sizes did not materially alter the primary findings, with SGLT2 inhibitors maintaining a consistent and favorable efficacy profile (Supplementary Table 2). However, visual inspection of the comparison-adjusted funnel plot (Figure 6) revealed some degree of asymmetry. This was quantitatively supported by a significant Egger’s test (p = 0.0319), indicating the presence of small-study effects across the network. As is common in emerging therapeutic fields like MASH, this suggests that smaller, early-phase trials may have reported more pronounced intervention effects, underscoring the need to interpret these promising estimates with appropriate caution.
Comparison-adjusted funnel plot for the network. The plot evaluates potential publication bias and small-study effects for the primary outcome of fibrosis improvement. The vertical solid line indicates the comparison-specific estimated summary effect, and the dashed lines represent the pseudo 95% confidence limits. The observed asymmetry, quantitatively supported by a statistically significant Egger’s test (p = 0.0319) indicates the presence of small-study effects within the network.
Discussion
Summary of main findings
To our knowledge, this is the first network meta-analysis to comprehensively compare distinct pharmacological mechanisms specifically incorporating the latest histological data regarding SGLT2 inhibitors, which were largely absent in concurrent reviews. Our analysis yielded pivotal findings: FGF21 analogs emerged as the most robust monotherapy for reversing liver fibrosis (RR 2.22, 95% CI 1.40-3.54), while SGLT2 inhibitors showed a promising preliminary signal (RR 2.27, 95% CI 1.22-4.17) that warrants further validation. Furthermore, our cluster analysis revealed a distinct therapeutic divergence: metabolic-focused agents (like GLP-1/GIP) excel at resolving inflammation, whereas liver-targeted agents (like FGF21) and pleiotropic agents (like SGLT2) appear highly effective for structural fibrosis remodeling.
The highly promising frontier: SGLT2 inhibitors for fibrosis
A notable finding of our study is the promising preliminary signal from SGLT2 inhibitors for fibrosis improvement (RR 2.27). While traditionally viewed as antidiabetic agents, recent histological trials suggest their benefits extend far beyond glycemic control. The histological benefits observed are biologically plausible: SGLT2 inhibition reduces hepatic de novo lipogenesis, mitigates oxidative stress, and promotes a shift toward fatty acid oxidation. Similar to the metabolic benefits historically seen with thiazolidinediones, the systemic improvement in insulin resistance and reduction of lipotoxicity likely halt the fibrogenic cascade at its upstream metabolic origin.
This finding is pivotal because, unlike the injectable peptides dominating current rankings (e.g., FGF21, GLP-1), SGLT2 inhibitors are oral, generic, and widely accessible. Given their established cardiovascular and renal benefits, our findings suggest that SGLT2 inhibitors deserve prioritization for future Phase 3 histology-driven trials, rather than immediate therapeutic re-ranking in clinical practice.
The “resolution-fibrosis” disconnect
Interestingly, GLP-1/GIP dual agonists dominated the ranking for MASH resolution but were ranked after SGLT2 and FGF21 for direct fibrosis improvement. This discrepancy suggests a potential time-lag effect: potent weight-loss agents rapidly resolve steatohepatitis (inflammation and ballooning) by reducing substrate delivery, but the resorption of established collagen fibers may require longer treatment durations to manifest statistically. This observation connects to trial design heterogeneity across drug classes. Incretin-based therapies predominantly resolve steatohepatitis through adiposity reduction and substrate depletion, which can occur relatively rapidly. In contrast, fibrosis improvement represents a downstream structural outcome more sensitive to baseline fibrosis severity, central histology reading protocols, and timing of repeat biopsy.
Comparison with prior network meta-analyses
Our findings extend prior NMAs in this field. The comprehensive NMA by Koh et al. (11) ranked THR-β agonists and GLP-1 RAs favorably for fibrosis improvement but excluded SGLT2 inhibitors due to absent histological endpoints at the time of their search. Inclusion of the Lin et al. (9) dapagliflozin trial represents a meaningful update. Notably, adding SGLT2 inhibitor data did not substantially alter relative rankings of established agents--FGF21 analogs and THR-β agonists maintained favorable positions, consistent with prior syntheses.
Limitations
Despite these promising signals, several limitations should be acknowledged. The efficacy ranking of SGLT2 inhibitors for fibrosis regression in our network is currently driven by a highly limited number of trials with biopsy-proven histological endpoints. The wide confidence intervals reflect this restricted pooled sample size. Furthermore, the tolerability network for SGLT2 inhibitors is based on a single biopsy-proven trial (9), as we strictly limited this node to pure SGLT2 inhibitors to ensure mechanistic consistency, excluding the SGLT1/2 inhibitor licogliflozin. While this approach preserves class purity, it limits the precision of the tolerability estimate for this node. Larger, adequately powered Phase 3 trials are needed to confirm these findings.
Additionally, treatment duration varied markedly across trials (16–96 weeks), introducing time-on-treatment heterogeneity. Since fibrosis improvement and MASH resolution are time-sensitive outcomes, indirect comparisons between agents studied over different follow-up periods may be subject to bias.
We acknowledge the absence of formal CINeMA/GRADE assessment. Applying GRADE principles informally, the certainty of evidence for SGLT2 inhibitor comparisons should be downgraded due to serious imprecision (wide CI), indirectness (single-study node), and potential publication bias. Accordingly, conclusions regarding SGLT2 inhibitors should be interpreted as hypothesis-generating rather than practice-changing.
Furthermore, to ensure network connectivity and sufficient statistical power, we pooled different dosages of the same active agent into single drug-class nodes (e.g., grouping various doses of efruxifermin into the ‘FGF21 Analog’ node). Consequently, this class-level analysis precludes the evaluation of dose-response relationships.
Conclusion
FGF21 analogs demonstrated the most robust evidence for improving liver fibrosis in MASH, while SGLT2 inhibitors showed a promising preliminary signal based on limited histological data, and GLP-1/GIP dual agonists are superior for MASH resolution. Specifically, preliminary evidence from a single pivotal trial suggests SGLT2 inhibitors may represent a promising oral option warranting further investigation in adequately powered Phase 3 studies. Given the sparse evidence base, the certainty of evidence for this drug class should currently be considered low. These findings support a more phenotype-informed approach to pharmacotherapy selection in MASH.