Highlight
– In a large target‑trial emulation using Mass General Brigham EHR data (2010–2023), GLP‑1 receptor agonist (GLP‑1RA) initiation in adults with T2DM and early MASLD was associated with a lower rate of progression to high‑risk FIB‑4 compared with DPP‑4 inhibitor initiators (HR 0.75, 95% CI 0.65–0.87).
– The protective association persisted in per‑protocol and landmark sensitivity analyses and across clinically relevant subgroups (low baseline FIB‑4, BMI <30, statin/aspirin/metformin users).
– No difference was observed for the composite of cirrhosis, hepatic decompensation, hepatocellular carcinoma, or liver transplantation (HR 0.98, 95% CI 0.72–1.34), underscoring the need for longer follow‑up and adjudicated hard outcomes.
Background: disease burden and rationale
Metabolic dysfunction‑associated steatotic liver disease (MASLD, formerly commonly termed NAFLD/NASH) is the hepatic manifestation of the metabolic syndrome and affects an estimated 25–30% of adults worldwide. Among people with type 2 diabetes mellitus (T2DM) the prevalence is higher and rates of progression to clinically meaningful fibrosis and liver‑related morbidity are substantially elevated. Fibrosis stage is the principal histologic predictor of liver‑related outcomes and mortality, and strategies that slow or reverse fibrosis progression are clinical priorities.
Glucagon‑like peptide‑1 receptor agonists (GLP‑1RAs) are now cornerstones of glucose management and weight reduction in T2DM and obesity. Randomized controlled trials have demonstrated GLP‑1RA benefits on liver histology (reduced steatohepatitis and steatosis) in selected NASH cohorts, but evidence on fibrosis progression in routine clinical practice and the effect on longer‑term liver endpoints remains limited. The study by Choi et al. (Liver Int. 2025) uses a target‑trial emulation in an EHR cohort to evaluate whether GLP‑1RA use is associated with lower fibrosis progression in patients with MASLD and T2DM compared with a plausible active comparator, dipeptidyl peptidase‑4 inhibitors (DPP‑4i).
Study design and methods
Choi and colleagues performed a retrospective cohort study across the Mass General Brigham (MGB) system between 2010 and 2023, explicitly emulating a target trial. Inclusion required adult patients with electronic evidence of MASLD and T2DM, initiation of either a GLP‑1RA or a DPP‑4i, and baseline Fibrosis‑4 (FIB‑4) index 2.67) defined and confirmed by two consecutive values ≥90 days apart within a one‑year window. The secondary outcome was a composite of advanced liver events: cirrhosis, hepatic decompensation, hepatocellular carcinoma (HCC), or liver transplantation.
To address confounding by indication and measured baseline differences, the investigators used 1:1 propensity score matching to balance covariates between treatment groups, producing 2,238 matched pairs. They reported incidence rates per 100 person‑years and hazard ratios from time‑to‑event analyses, and performed prespecified sensitivity analyses, including per‑protocol and landmark approaches, plus subgroup analyses by baseline FIB‑4, BMI, and common co‑medications.
Key findings
Main outcome: Over the study period, GLP‑1RA users had a lower incidence rate of progression to high‑risk FIB‑4 compared with DPP‑4i users (3.25 vs. 4.29 per 100 person‑years). The primary time‑to‑event analysis yielded a hazard ratio (HR) of 0.75 (95% CI 0.65–0.87), indicating a 25% relative reduction in the hazard of reaching a high‑risk FIB‑4 after initiation of GLP‑1RA therapy.
Sensitivity and subgroup results: The association remained directionally consistent in per‑protocol analyses (HR 0.80) and in landmark analyses designed to mitigate immortal time bias. Clinically relevant subgroups — patients with low baseline FIB‑4, those with BMI <30 kg/m2, and individuals concurrently using statins, aspirin, or metformin — also showed protective associations, suggesting robustness across metabolic phenotypes and concomitant pharmacotherapies.
Secondary outcome: There was no statistically significant difference between GLP‑1RA and DPP‑4i groups for the composite of cirrhosis, hepatic decompensation, HCC, or liver transplantation (HR 0.98, 95% CI 0.72–1.34). Event rates for hard liver outcomes were low over the available follow‑up, limiting power to detect differences.
Safety and tolerability: Safety endpoints are not the focus of the paper and were not reported in detail; GLP‑1RA tolerability and recognized adverse events (gastrointestinal symptoms, rare risks such as pancreatitis or gallbladder disease) should still be considered in clinical decision‑making.
Interpretation and biological plausibility
The observed association aligns with mechanistic and trial data supporting hepatoprotective effects of GLP‑1RAs. Proposed mechanisms include weight loss and reduced adipose tissue inflammation, improvements in insulin resistance, reductions in de novo lipogenesis and hepatic steatosis, and possible direct hepatic and immune‑modulatory effects mediated through GLP‑1 receptor–dependent and independent pathways. Phase 2–3 randomized trials have demonstrated improvements in steatosis and steatohepatitis with liraglutide and semaglutide; however, effects on fibrosis reversal have been less consistent and often limited by sample size and treatment duration.
Using FIB‑4 as the primary outcome leverages a validated, readily available noninvasive marker of fibrosis risk that predicts liver‑related outcomes in population studies. While not a replacement for histology or transient elastography, FIB‑4 is appropriate for large EHR‑based comparative effectiveness research and for detecting clinically meaningful shifts in fibrosis risk at a population level.
Strengths of the study
– Large, real‑world cohort with modern diabetes therapies and a long accrual period (2010–2023).
– Explicit target‑trial emulation and selection of an active comparator (DPP‑4i) reduce confounding by indication compared with comparisons to non‑users.
– Propensity score 1:1 matching balanced measured baseline covariates, and sensitivity analyses (per‑protocol, landmark) addressed common observational biases.
– Robust primary outcome definition requiring confirmatory FIB‑4 measures improves outcome specificity.
Limitations and sources of potential bias
– Residual confounding: As with any observational analysis, unmeasured confounders (health‑seeking behavior, clinician preference, socioeconomic factors, baseline physical activity, dietary patterns, exact weight loss over time) may influence treatment choice and outcomes. Differential weight loss driven by GLP‑1RA could mediate the observed effect rather than direct hepatic actions.
– Heterogeneity within drug classes: The GLP‑1RA class includes agents with different pharmacokinetics, receptor potency, and dosing regimens (e.g., exenatide, liraglutide, dulaglutide, semaglutide). The study reports class‑level associations but cannot confidently parse agent‑specific effects.
– Outcome measurement: FIB‑4 is an indirect marker influenced by age, platelet count and transaminases; age‑related increases can spuriously raise FIB‑4. Although widely used, FIB‑4 cannot replace elastography or histology for staging fibrosis. The lack of detailed elastography, imaging, or biopsy data reduces granularity.
– Follow‑up and event rates: The lack of difference in hard liver outcomes likely reflects low absolute event rates and limited follow‑up duration for events that accrue over many years. Longer observation is required.
– Generalizability: The cohort reflects patients treated within a single integrated healthcare system; external validation in diverse populations and other health systems is needed.
Clinical and research implications
For clinicians managing patients with MASLD and T2DM, these findings add real‑world evidence that initiating GLP‑1RA therapy may attenuate progression of fibrosis risk compared with DPP‑4i therapy. This information is useful when balancing glycemic efficacy, weight benefits, cardiovascular effects, cost, access, and patient preferences. However, causality cannot be assumed from observational data alone, and decisions should be individualized.
For policymakers and guideline panels, these data suggest potential disease‑modifying benefits of GLP‑1RAs at the population level, reinforcing the need to consider liver outcomes in diabetes pharmacotherapy recommendations. For researchers, the findings support the rationale for randomized trials powered for fibrosis progression and longer‑term hepatic endpoints and for mechanistic studies that separate weight‑dependent from weight‑independent effects.
Where evidence stands: selective references
– Armstrong MJ, Hull D, Guo K, et al. Liraglutide safety and efficacy in patients with non‑alcoholic steatohepatitis (LEAN): a multicentre, double‑blind, randomized, placebo‑controlled phase 2 study. Lancet. 2016;387(10019):679–690.
– Newsome PN, Buchholtz K, Cusi K, et al. A placebo‑controlled trial of subcutaneous semaglutide in nonalcoholic steatohepatitis. N Engl J Med. 2021;384:1113–1124.
– Chalasani N, Younossi Z, Lavine JE, et al. The diagnosis and management of nonalcoholic fatty liver disease: practice guidance from the American Association for the Study of Liver Diseases. Hepatology. 2018;67(1):328–357.
– Sterling RK, Lissen E, Clumeck N, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology. 2006;43(6):1317–1325. (Original FIB‑4 index derivation and validation work.)
– Hernán MA, Robins JM. Using Big Data to emulate a target trial when a randomized trial is not available. Am J Epidemiol. 2016;183(8):758–764. (Conceptual framework for target‑trial emulation.)
Conclusion
In this large target‑trial emulation from a single integrated health system, initiation of GLP‑1RA therapy in patients with MASLD and T2DM was associated with a 25% relative reduction in the hazard of progressing to a high‑risk FIB‑4 compared with initiation of DPP‑4i therapy. The association was robust across multiple sensitivity analyses and subgroups, but no difference was detected for hard liver outcomes during the available follow‑up. These results strengthen the biological plausibility and real‑world signal that GLP‑1RAs may slow early fibrosis progression, but randomized controlled trials with histologic endpoints and longer follow‑up will be required to confirm causality and to guide practice. Clinicians should integrate these data with individual patient considerations, adverse‑effect profiles, and cost/access constraints when selecting glucose‑lowering therapies for patients with MASLD.
Funding and clinicaltrials.gov
The primary article reports funding and disclosures; readers should consult Choi et al. (Liver Int. 2025) for detailed funding sources and conflict of interest statements. This was an observational EHR‑based study rather than a registered interventional trial; no clinicaltrials.gov registration applies.
Acknowledgements
We acknowledge the original authors (Choi J, Kamath T, Nguyen VH, Przybyszewski E, Song J, Carroll A, Michta M, Almazan E, Simon TG, Chung RT) for generating the data and analyses summarized here.
References
Selected key references are provided above; readers seeking the full reference list and primary data should consult the original publication: Choi J, Kamath T, Nguyen VH, Przybyszewski E, Song J, Carroll A, Michta M, Almazan E, Simon TG, Chung RT. GLP‑1RA and Liver Fibrosis Progression in MASLD and Type 2 Diabetes: Target Trial Emulation Using Propensity Score Matching. Liver Int. 2025 Dec;45(12):e70447. doi: 10.1111/liv.70447. PMID: 41250965; PMCID: PMC12625798.
