We read with great interest the study by Yeo et al. comparing major adverse liver outcomes (MALO) following initiation of baclofen or acamprosate in patients with newly diagnosed compensated alcohol-related cirrhosis 1. The authors should be commended for this large cohort study utilising a vast real-world database with comprehensive propensity score matching across 60 covariates. The study authors' principal finding suggests baclofen is associated with greater risk of incident MALO relative to acamprosate in patients with newly diagnosed compensated alcohol-related cirrhosis, predominantly driven by hepatic encephalopathy. We suggest that the findings be interpreted with caution. Baclofen, even without underlying liver disease, is associated with encephalopathy, particularly in renal dysfunction 2. The use of ICD-10 codes to identify cases of hepatic encephalopathy introduces misclassification risk, with alternative diagnoses such as baclofen-induced encephalopathy, delirium and intoxication possibly being incorrectly attributed to hepatic encephalopathy. Baclofen remains the only relapse prevention medication with randomised controlled trial-level evidence in cirrhosis, although use is typically avoided in those at risk of hepatic encephalopathy 3, 4. Despite propensity score matching, confounding by indication may persist, with clinicians preferentially prescribing acamprosate over baclofen in patients perceived to be at higher risk of hepatic encephalopathy, thereby potentially contributing to the lower rates observed in the acamprosate group. Differential efficacy between each medication is another consideration. A systematic review and meta-analysis examining baclofen efficacy in alcohol use disorder in those with liver disease demonstrated mixed results, revealing no clear benefit compared with placebo 5. Hence, if we consider baclofen possesses lower efficacy with consequent ongoing alcohol use and liver injury, this could independently drive higher hepatic decompensation rates. This seems less plausible given no differences were observed in other individual decompensation events beyond hepatic encephalopathy. Further limitations include limited data on patients' alcohol intake, co-interventions such as enrolment in alcohol relapse prevention programs, medication adherence and of course the retrospective, non-randomised design of the study, all of which can introduce residual confounding. Although the study cohort exceeded 1100 patients, there was only a modest absolute between-group difference in significant study outcomes (~6%). Naturally, prospective randomised controlled trials are required in this area; however, the practicalities of recruiting a sufficiently large sample size to demonstrate statistical significance is likely to present a real-world challenge, necessitating large international, multicentre collaboration. Alcohol-related liver disease remains a substantial burden, with alcohol accounting for 42% of cirrhosis-related inpatient admissions in recent German registry data 6. Additionally, a 2025 meta-analysis demonstrated there have been no significant improvements in short-term mortality from severe alcohol-associated hepatitis in those managed with supportive care over the past four decades 7. A recent analysis of metabolic-dysfunction and alcohol-related liver disease patients found persistent binge drinking is independently associated with advanced fibrosis and all-cause mortality 8. Achieving abstinence or sustained reduction therefore remains the pivotal modifiable determinant of prognosis, and baclofen retains an important evidence-based role in alcohol relapse prevention in liver disease. Prematurely discouraging its use based on a hepatic encephalopathy signal susceptible to the confounders outlined above risks narrowing treatment options in an area with currently limited options. Karl Vaz: conceptualization, writing – review and editing. Dorothy Liu: conceptualization, writing – review and editing. Sarah Lucas: conceptualization, writing – review and editing. Andrew W. Nguyen: conceptualization, writing – review and editing, writing – original draft. The authors have nothing to report. The authors declare no conflicts of interest. This article is linked to Yeo et al. papers. To view this article, visit https://doi.org/10.1111/apt.70619. Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Nguyen et al. (Wed,) studied this question.