Simonetto, Douglas A

Artificial intelligence (AI) has emerged as a transformative force in liver transplantation (LT), spanning patient selection, donor–recipient matching, intraoperative management, and post-transplant care. Historically, applications have relied on traditional machine learning (ML) and deep learning (DL) techniques trained on structured clinical datasets to improve risk stratification and outcome prediction. Since 2023, however, advances in generative AI (GenAI) and large language models (LLMs) have expanded the scope of AI in LT, enabling multimodal reasoning and the extraction of clinically meaningful information from unstructured data. This review synthesizes recent AI-driven developments across the LT patient journey, from advanced chronic liver disease management to long-term post-transplant outcomes. Representative applications include ML-based models to optimize treatment selection and waitlist prioritization for hepatocellular carcinoma, random forest and gradient-boosting approaches to improve donor utilization and predict graft survival, and DL-based imaging models for graft volumetry and steatosis assessment. Post-transplant innovations encompass neural network models for graft injury classification, recurrent architectures for immunosuppression dosing and fibrosis prediction, multi-task frameworks for complication risk stratification, and integrative omics-based approaches to detect rejection and dysfunction. Emerging GenAI applications include LLM-enabled digital phenotyping of social determinants of health and multi-agent simulations of transplant selection committees. Despite rapid progress, translation into routine practice remains limited by challenges in local validation, workflow integration, and infrastructure requirements. Future opportunities lie in incorporating underutilized data domains, such as psychosocial factors, -omics, and intraoperative physiologic streams, and leveraging GenAI to enhance clinical decision support, research scalability, and patient–clinician communication. AI is poised not only to refine prediction but to reshape the conceptual and operational landscape of liver transplantation.

Introduction: Although the role of cardiometabolic risk factors (CMRF) has been characterized in steatotic liver disease, their role in the severity of alcohol-associated hepatitis (AH) remains unclear. We aimed to evaluate the impact of CMRF on mortality. Methods: Multinational prospective cohort study (2015–2024) including hospitalized patients with severe AH across 32 centers in 14 countries. Diagnosis of AH was made using the NIAAA criteria. Analyses included adjusted competing-risk models by age, sex, ethnicity, history of cirrhosis, CMRF, corticosteroid use, MELD, and ACLF grade, with liver transplantation as a competing risk. Results: 936 participants were included; mean age 48±11.2 years, and 88.9% were male. At least one CMRF was present in 46.6%; median body mass index (BMI) was 24.2 kg/m 2 [IQR: 22.8–28.2], prevalence of diabetes 17.6%, hypertension 16.5%, and dyslipidemia 5.8%. Median MELD was 24.4 [19.3–31.4], 86.7% had severe AH, and 180-day survival was 72.9%. Survival did not differ by CMRF status (log-rank p =0.453). In adjusted competing-risk models, higher age (sHR 1.02; 95%CI:1.01–1.04), greater alcohol intake (per g/day; sHR 1.001; 95%CI:1.000–1.002), MELD (sHR 1.04; 95%CI:1.01–1.06), and ACLF grade 2–3 (sHR 2.34 and 4.34) predicted mortality. However, no individual CMRF independently increased mortality. A prespecified non-linear BMI analysis showed modestly lower mortality between 25–40 kg/m², with a higher risk above 40 kg/m². Conclusion: Among patients with severe AH, metabolic dysfunction was not associated with increased mortality. Although a higher BMI was associated with slightly lower mortality in AH, this may reflect better nutritional status rather than a true protective effect.

INTRODUCTION: The American College of Gastroenterology assembled a multidisciplinary task force to evaluate the current state and future direction of artificial intelligence (AI) in gastroenterology, hepatology, and endoscopy leading to the development of consensus-based recommendations for responsible AI integration in clinical practice. METHODS: A total of 32 subject-matter experts and 12 industry partners, representing diverse practice settings and expertise, conducted subgroup literature reviews across 5 key areas (endoscopy, practice management clinical applications, training and education, inflammatory bowel disease and liver disease, ethics and equity). Draft statements were developed and rated on a 5-point Likert scale using a modified Delphi process. A consensus was set at ≥70% combined agreement. Nonconsensus items were revised and revoted electronically. RESULTS: A total of 43 statements, 40 (93%) reached consensus in round 1 and the remaining 3 achieved consensus after round 2. Evidence supports computer-aided detection improving adenoma detection rate and miss rate in controlled studies, with mixed real-world impact and insufficient long-term outcomes (e.g., interval colon cancer rate). Recommendations emphasize thorough validation and reduction of bias by heterogeneous Data sets. Outside endoscopy, ambient AI scribes, natural language processing (NLP)-enabled coding, workflow optimization, and previous authorization support show potential. Training recommendations endorse a structured AI curriculum while preserving independent procedural competence to avoid deskilling. In inflammatory bowel disease and hepatology, AI could help improve diagnostic accuracy, help predict risk of disease progression, and help guide therapy. Equity, governance, and reimbursement statements call for chain-of-custody data protections, specialty-society oversight, and payment models that reward quality and cost reduction. DISCUSSION: This consensus outlines how AI can augment rather than replace clinical expertise while promoting safety, transparency, interoperability, and equity. Priorities include pragmatic and prospective trials, multi-institutional data-sharing consortia, bias mitigation, and workforce training to enable trustworthy and clinically impactful AI adoption in gastroenterology, liver, and endoscopy care.

INTRODUCTION: Evidence on the comparative real-world effectiveness of terlipressin vs midodrine plus octreotide (MO) for hepatorenal syndrome-acute kidney injury (HRS-AKI) in the United Kingdom and the United States is limited. METHODS: Using individual-level chart review data for patients across the United Kingdom (2013–2017) and the United States (2016–2019), an indirect treatment comparison was conducted comparing the efficacy of terlipressin (UK cohort) with MO (US cohort). Covariate balancing propensity scoring matched the cohorts on baseline serum creatinine (SCr), presence of encephalopathy and/or ascites, albumin use and duration, age, and sex. The primary endpoint was HRS reversal, defined as achieving SCr ≤1.5 mg/dL by the last day of treatment. RESULTS: At treatment initiation, 90.2% of UK patients received terlipressin (194/215), while 89.2% of US patients received MO (140/157). Concomitant albumin was administered in 67.9% of UK and 98.7% of US patients. In a covariate balancing propensity score-adjusted cohort, HRS reversal was achieved in 53.2% of terlipressin-treated patients (the United Kingdom, weighted effective sample size of 75) compared with 16.9% of MO-treated patients (the United States, n = 89) (adjusted mean difference (95% CI) 36.3% (22.4, 50.2), P < 0.0001). In adjusted analysis, individuals treated with terlipressin experienced an overall reduction in SCr at completion of treatment (SCr decrease 1.00 mg/dL vs increase of 0.08 mg/dL for MO-treated patients, P < 0.0001). DISCUSSION: HRS-AKI treatment and outcomes differ between the United Kingdom and the United States, attributed to the historical standard of care MO in the United States. In adjusted analyses, real-world use of terlipressin was more effective than MO at improving kidney function and achieving HRS-AKI reversal.

Background: Severe alcohol-associated hepatitis (sAH) is a well-characterized disease with high short-term mortality. However, there is limited research on those with a “less severe condition” (moderate AH). This study aims to characterize in-depth patients with moderate AH (mAH), including the performance of mortality scoring systems, key prognostic factors, and survival over time. Methods: A multicenter retrospective cohort study (2009–2019) included patients with mAH (MELD score ≤20 at admission). Cox regression and receiver operating characteristic curves with AUC were used for analysis. Results: We included 1845 patients with AH (20 centers, 8 countries) between 2009 and 2019. mAH was defined as a MELD score ≤20 at admission. Twenty-four percent met the criteria for an mAH episode. Patients with mAH tend to be older and have a higher proportion of females, with a median MELD of 17 (15–19), Maddrey discriminant function (mDF) of 33 (22–40), the trajectory of serum bilirubin of 0.83 (0.60–1.21), and neutrophil-to-lymphocyte ratio (NLR) of 5 (2.96–8.60). The primary causes of death in mAH included multiple organ failure (34.1%) and infections (16.6%). The cumulative survival rates at 30, 90, and 180 days were 94.3%, 90.4%, and 88.2%, respectively. In multivariable analysis, age was the only significant predictor of 30-day mortality (HR 1.49, 95% CI: 1.27–1.76, p <0.001). Mortality prediction models showed poor performance, with AUC for MELD (0.671), mDF (0.726), trajectory of serum bilirubin (0.733), and NLR (0.697). Conclusions: Patients with moderate AH exhibited a mortality of 11.8% at 6 months, primarily driven by multiple organ failure and infections. These patients also exhibit a different clinical profile compared to those with sAH. Tailored models and therapeutic strategies are needed to improve long-term outcomes in mAH.

INTRODUCTION: Building on prior results, we hypothesized that an electrocardiogram (ECG)-enabled machine learning (ML) model could be used to detect advanced chronic liver disease (CLD). METHODS: A cohort with CLD and 12-lead ECGs was matched with controls from electronic health records. A ML model was trained as a binary classifier. RESULTS: There are 12,930 patients with CLD and 64,577 controls in the cohort. The model's discriminative ability to classify CLD showed an area under the receiver-operating characteristic curve 0.858 (95% confidence interval: 0.850–0.866), and at the chosen threshold, CLD ECGs had 12 times higher odds of being classified as CLD (diagnostic odds ratio 12.33, 95% confidence interval: 11.16–13.63). DISCUSSION: An ECG-enabled ML model affords great promise in identifying advanced CLD in low resource areas.

Background and Aims: Clinical hepatology research often faces limited data availability, underrepresentation of minority groups, and complex data-sharing regulations. Synthetic data—artificially generated patient records designed to mirror real-world distributions—offers a potential solution. We hypothesized that diffusion models, a state-of-the-art generative technique, could produce synthetic liver transplant waitlist data from the United Network for Organ Sharing database that maintains statistical fidelity, replicates clinical correlations and survival patterns, and ensures robust privacy protection. Approach and Results: Diffusion models were used to generate synthetic patient cohorts mirroring the United Network for Organ Sharing liver transplant waitlist database between the years 2019 and 2023. Statistical fidelity was assessed using maximum mean discrepancy (MMD) and Wasserstein distance, correlation analysis, and variable-level metrics. Clinical utility was evaluated by comparing transplant-free survival via Kaplan–Meier curves and the MELD score performance. Privacy was quantified using the Distance to Closest Record (DCR) and attribute disclosure risk assessments. The synthetic dataset was nearly indistinguishable from the original dataset (MMD=0.002, standardized Wasserstein distance <1.0), preserving clinically relevant correlations and survival patterns as evidenced by similar median survival times (110 vs. 101 days) and 5-year survival rates (22.2% vs. 22.8%). MELD-based 90-day mortality prediction was maintained (original AUC=0.839 vs. synthetic AUC=0.844). Privacy metrics indicated no identifiable patient matches, and mean DCR values ensured that synthetic individuals were not direct replicas of real patients. Conclusion: Artificial intelligence–generated synthetic data derived from diffusion models can faithfully replicate complex hepatology datasets, maintain key clinical signals, and ensure strong privacy safeguards. This approach can help address data scarcity, enhance model generalizability, foster multi-institutional collaboration, and accelerate progress in hepatology research.

ABSTRACTSeveral dynamic models predict mortality and corticosteroid response in alcohol‐associated hepatitis (AH), yet no consensus exists on the most effective model. This study aimed to assess predictive models for corticosteroid response and short‐term mortality in severe AH within a global cohort. We conducted a multi‐national study of patients with severe AH treated with corticosteroids for at least 7 days, enrolled between 2009 and 2019. Dynamic models—Lille‐4, Lille‐7, trajectory of serum bilirubin (TSB), and neutrophil‐to‐lymphocyte ratio (NLR)—were used to estimate 30‐ and 90‐day mortality. Lille‐7 demonstrated the highest accuracy for both 30‐ and 90‐day mortality.