Validation of scores of PRO‐C3 to predict liver‐related events in alcohol‐related liver disease

Background and Aims Risk prediction in alcohol‐related liver disease (ArLD) is an unmet need. We aimed to assess PRO‐C3 models to predict liver‐related events (LRE) in patients with a history of excessive alcohol use without an established diagnosis of chronic liver disease. Methods A prospective cohort study of 462 patients with ArLD, split into a derivation cohort of 221 secondary care patients and a validation cohort of 241 primary care patients. Baseline variables, including fibrogenesis marker PRO‐C3, were used to develop a prediction model. Prognostic accuracy was compared to enhanced liver fibrosis (ELF), fibrosis‐4‐index (FIB‐4), transient elastography (TE) and ADAPT. Results In the derivation and validation cohorts, 67 (30%) and 19 (8%) experienced an LRE during a median follow‐up of 5.2 years (IQR: 3.2‐6.8) and 4.0 years (IQR: 2.7‐5.6). On top of PRO‐C3 and ADAPT score, we generated a model (ALPACA) of independent predictors of LREs (PRO‐C3, AST/ALT, platelets). ALPACA had high prognostic accuracy with a C‐statistic of 0.85 in the derivation cohort, comparable to ELF (0.83) and TE (0.84) and significantly higher than FIB‐4 (0.78), PRO‐C3 (0.80) and ADAPT (0.81). In the validation cohort, all tests had comparable C‐statistics. Compared to low‐risk patients (ALPACA ≤11), high‐risk patients (>11) had a subhazard ratio for LREs of 12.6 (95% CI 5.9‐26.8, p