Comparison of screening strategies with two new tests to score and diagnose varices needing treatment

Background and aims: We aimed to improve non-invasive screening of varices needing treatment (VNT) and compare different screening strategies. Methods: 2,290 patients with chronic liver disease were included in a retrospective study. Etiologies were: virus: 50.0%, NAFLD: 29.5%, alcohol: 20.5%, VNT: 14.9%. Test descriptors were performance (spared endoscopy) and safety (missed VNT). VNT tests were evaluated according to their safety levels either for individual screening (95% negative predictive value (NPV)), population screening (95% sensitivity) or undifferentiated screening (100% sensitivity/NPV) without missed VNT. The tests provided three categories: missed VNT <5%, VNT 100% specificity (new category), both sparing endoscopies, and intermediate (endoscopy required). Results: Independent VNT predictors (etiology, sex, age, platelets, prothrombin index, albumin, ALT, liver stiffness) were included in two tests: VNT virus alcohol NAFLD test (VANT) and varice risk score (VARS). We report results of the whole population. Considering population screening, performances were, Baveno VI criteria: 24.1%, Anticipate: 24.7%, VariScreen: 35.3%, VANT: 40.2% (p<0.001 vs other tests). VANT spared 58.0% more endoscopies in the whole population than Baveno criteria in compensated advanced chronic liver diseases. Considering individual screening, VARS performance was, in all patients: 62.0% vs 42.9% for the expanded Baveno VI criteria (p<0.001), and, in NAFLD: 72.8% vs 65.1% for the NAFLD cirrhosis criteria (p<0.001). Considering undifferentiated screening, VARS performance was 12%. The VARS score estimated VNT probability from 0 to 100% (AUROC: 0.826). Conclusion: VANT and VARS spared from 12% (undifferentiated screening) to 40% (population screening) or 62% (individual screening) of endoscopies in main-etiology patients without ascites.