Prospective head-to-head comparison of routine non-invasive scores for predicting severe cirrhosis-related morbidity in the general population

Background & Aims: Effective non-invasive risk-stratification tools are essential for the early detection of individuals at high risk for cirrhosis, to enable timely intervention. We conducted a prospective, head-to-head comparison of fibrosis-based and outcome-driven routine blood-based risk scores for predicting cirrhosis-related morbidity in a large community-based cohort. Approach and Results: We first performed a systematic review to identify risk scores derived from routine liver blood tests, and then evaluated them in the UK Biobank. Severe cirrhosis-related morbidity was defined using International Classification of Diseases, Tenth Revision codes. Discrimination and clinical utility were assessed using the Wolbers C-index, time-dependent area under the receiver operating characteristic curve, area under the precision–recall curve (AUPRC), and cumulative incidence accounting for competing risks. The review identified 12 eligible risk scores (10 novel models plus APRI and FIB-4). Among 385,738 participants, the 10-year cumulative incidence of severe cirrhosis-related morbidity was 0.39% (1,498 events). Most novel scores outperformed APRI and FIB-4. LiverRisk showed the highest discrimination at 5 years (C-index 0.847) and 10 years (C-index 0.812), closely followed by CORE (5-year C-index 0.839; 10-year C-index 0.811). In contrast, CORE achieved better enrichment of high-risk individuals, with an AUPRC of 0.088 compared with 0.063 for LiverRisk. At low referral proportions, increasing the CORE threshold yielded greater net benefit than a sequential CORE-LiverRisk referral strategy. Conclusion: CORE and LiverRisk are the most discriminative routine blood-based tools for predicting long-term cirrhosis-related morbidity in the community. When referrals are limited, a higher-threshold CORE-only strategy may outperform a sequential CORE-LiverRisk approach.