Validation and comparison of non-invasive prediction models based on liver stiffness measurement to identify patients who could avoid gastroscopy

Abstract

Abstract Several non‐invasive tests (NITs) based on liver stiffness measurement (LSM) have been developed to rule out varices needing treatment (VNT), including the Baveno VI criteria (B6C), the expanded Baveno VI criteria (EB6C), the LSM-spleen diameter to platelet ratio score (LSPS), and the VariScreen algorithm. We aimed to validate and compare those NITs in patients with compensated advanced chronic liver disease (cACLD). This retrospective study enrolled 354 patients with cACLD; LSM, platelet count (PLT), international normalized ratio (INR), gastroscopy and spleen diameter (SD) were collected. VNT prevalence was 28.5%. In comparison, patients with VNT included higher LSM, INR, and SD and lower PLT. Gastroscopies were spared for 27.7% of patients using the B6C with 1.0% VNT missed rate, 47.2% of patients using the EB6C with 5.9% VNT missed rate, 57.6% of patients using the LSPS with 9.9% VNT missed rate, and 45.5% of patients using the VariScreen algorithm with 3.0% VNT missed rate. Only the B6C and the VariScreen algorithm could safely avoid gastroscopies, and the VariScreen algorithm spared more gastroscopies than the B6C. The results were consistent with the previous when performed subgroup analysis. In conclusion, the VariScreen algorithm performed the best and can be used in clinical.