P51 Hepatocellular carcinoma risk scores for non-viral liver disease: a systematic review

HCC risk prediction models may provide a more personalised and effective approach to surveillance for HCC. Despite an abundance of published risk models, few have been developed for patients with cirrhosis due to ‘non-viral’ aetiologies. We aimed to systematically collate and critically appraise the performance of HCC prediction models in patients with non-viral chronic liver disease.The study was reported in accordance with PRISMA guidelines and prospectively registered with PROSPERO (CRD42022370078). MEDLINE and EMBASE databases were searched using a validated search filter for prediction model studies from inception until October 2022. Two reviewers independently assessed studies for inclusion. The quality of included studies was assessed using the TRIPOD checklist and the risk of bias assessed using the PROBAST tool. Model performance (discrimination and calibration) to identify risk of HCC at a specified time point was described where available.The search strategy identified 7,854 records. After review, 13 studies describing 8 original risk models (development n = 7, validation n = 6) were included. All studies were done retrospectively with 8/13 from single centre cohorts. The study periods ranged from 1997 – 2021, with 62% (n = 8) of studies conducted in western populations. Sample size of cohorts ranged from 269 – 34,932. Follow up periods ranged from 15.1 – 138 months. 46% of included patients had viral hepatitis, 28% alcohol related liver disease (ArLD) and 16% non-alcoholic fatty liver disease (NAFLD). Only 1 model was developed using a competing risk perspective. Between 2 and 7 predictors (median 4) were used to build the models with 15 different predictors used across the 8 models. Age (n = 6 models) and sex (n = 6 models) were the most frequently included predictors. 5 risk models have been externally validated. All studies reported a measure of model discrimination with either AUROC or c-statistic (range 0.61 – 0.82). Measures of calibration were reported in 8 studies (6/7 development studies, 2/6 validation studies). Study compliance with TRIPOD ranged from 50% – 78% with complete reporting of performance measures (33%) and handling of missing data (0%) particularly poor. All studies were rated at high risk of bias.Studies describing risk prediction of HCC in non-viral chronic liver disease are poorly reported and are at high risk of bias. Patients with ArLD and NAFLD remain underrepresented in development and validation cohorts, and the m