How to clarify the Baveno VI criteria for ruling out varices needing treatment by noninvasive tests

Background & Aims: Baveno VI criteria enabled the screening of varices needing treatment (VNT) without endoscopy but created confusion by not stating the method used to calculate the 5% missed VNT limit, resulting in different calculations across validation studies. We analysed those calculations to clarify their diagnostic meaning. Methods: (a) Literature review and recalculation of the missed VNT rates according to the three definitions encountered. (b) Contingency table comparison of these latter to determine their diagnostic meanings. (c) Real case analysis. 4/Simulation of variations in the three main statistical descriptors (VNT, missed VNT or spared endoscopies). Results: Missed VNT rates in the three definitions varied five- to 10-fold across 7 papers. The contingency table showed that the definitions based on VNT prevalence and spared endoscopy as reference corresponded, respectively, to sensitivity and negative predictive value (NPV). The whole population-based definition corresponded to diagnostic accuracy (not pertinent in that setting). Real case analysis showed that concerning liver stiffness, the 95% sensitivity and NPV cut-offs for VNT were, respectively, 14.1 and 26.5 kPa. The VNT-based definition offered a more statistically powerful paired comparison between diagnostic tests, whereas the definition based on spared endoscopies was hampered by an unpaired comparison. Case simulation showed that the VNT-based definition was the most sensitive to descriptor variations. Conclusion: The definitions of missed VNT rate placing VNT or spared endoscopy as the denominator are appropriate, providing, respectively, sensitivity and NPV for VNT. We privilege the first since it corresponds to the true proportion of missed VNT.