The Portal Hypertension Decompensation Score – A Validated Predictive Model of Liver Decompensation Related to Portal Hypertension

Background and Aims: There is a need for non-invasive risk stratification in people with compensated advanced chronic liver disease (cACLD) to prognosticate and guide management. We aimed to develop a score that predicts decompensation in people with cACLD without the need for a liver stiffness measurement (LSM). Methods: A multicentre state-wide cohort of patients with cACLD between 2004-2015 were followed until decompensation. A predictive score using serum markers was developed in a training cohort (n=967) using competing risk analysis and internally validated (n=417). Further external validation and comparison to other scores was undertaken in 315 patients between 2017-2024. Results: Decompensation occurred in 172 (17.8%), 64 (15.3%) and 51 (16.2%) in the training and two validation cohorts (p=0.60) after median follow-up of 3.2, 3.4 and 1.9 years respectively. Bilirubin, alanine aminotransferase, alkaline phosphatase, albumin and platelets predicted decompensation and combined into a final model - the Portal Hypertension Decompensation Score (PDS). The PDS was well calibrated with good discrimination for predicting decompensation. In the two validation cohorts, accuracy (time dependent AUC) of the PDS for predicting decompensation was high at 2-years (0.75 and 0.82) and 5-years (0.74 and 0.83). A low score (<-3.348) had a sensitivity of 74-84% in prediction of no decompensation with a negative predictive value of 91-95%, a high score (> -2.828) was 87-93% specific for future decompensation with a positive predictive value of 33-58%. Conclusions: The PDS is an accurate predictor of decompensation in cACLD. It discriminates patients who are low-risk from those who are high-risk and who may benefit from further evaluation or treatment, without requiring the use of LSM.