Validation of PH and Varices Risk Scores for Prediction of High-Risk Esophageal Varix and Bleeding in Patients with B-Viral Cirrhosis

Esophageal varices (EVs) can be accurately predicted using PH and varices risk scores. We aimed to validate their prognostic performances. Methods: We enrolled patients with B-viral cirrhosis as the training cohort (n = 503). Areas under receiver operating characteristic curves (AUROCs) for HEV were calculated for PH (=−5.953 + 0.188 × liver stiffness (LS) + 1.583 × sex (1:male/0:female) + 26.705 × spleen diameter/platelet count ratio) and varices (=−4.364 + 0.538 × spleen diameter −0.049 × platelet count −0.044 × LS + 0.001 × LS × platelet count) risk scores, and compared to LSPS (=LS × spleen diameter/platelet count). An independent cohort was recruited for further validation (n = 222). In the training cohort, the varices risk score showed the highest AUROC (0.926), followed by the PH risk score (0.924) and LSPS (0.924), but without any statistically significant differences. For varices risk scores ≤−1.70 and ≥1.48, a 95.0% negative predictive value (NPV) and 91.2% positive predictive value (PPV) were observed, respectively. At PH risk scores ≤2.25 and ≥7.71, 95.0% NPV and 90.0% PPV were observed, respectively. At LSPS ≤1.73 and ≥13.9, 95.3% NPV and 95.0% PPV were observed, respectively. The EV bleeding (EVB) risk during follow-up increased stepwise and significantly when stratified by PH, varices risk scores, and LSPS (all p < 0.001). In the validation cohort, NPVs were generally similar when stratified by PH (88.2%), varices risk scores (93.2%), and LSPS (88.9%); however, corresponding PPVs were suboptimal. PH and variceal risk scores are reliable for predicting HEV and future EVB. Patients with PH and varices risk scores ≤2.25 and ≤−1.70, respectively, may avoid endoscopy safely. For convenience, LSPS might be a good alternative, with comparable prognostic performance to these two models.