Including Ratio of Platelets to Liver Stiffness Improves Accuracy of Screening for Esophageal Varices That Require Treatment

BACKGROUND & AIMS Based on platelets and liver stiffness measurements, the Baveno VI criteria (B6C), the expanded B6C (EB6C), and the ANTICIPATE score can be used to rule out varices needing treatment (VNT) in patients with compensated chronic liver disease. We aimed to improve these tests by including data on the ratio of platelets to liver stiffness. METHODS In a retrospective analysis of data from 10 study populations, collected from 2004 through 2018, we randomly assigned data from 2368 patients with chronic liver disease of different etiologies to a derivation population (n = 1579; 15.1% with VNT, 50.2% with viral hepatitis, 28.9% with nonalcoholic fatty liver disease, 28.9% with alcohol-associated liver disease, 20.8% with model for end-stage liver disease scores of 9.5 ± 3.0, and 93.0% with liver stiffness measurements ≥10 kPa) or a validation population (n = 789). Test results were compared with results from a sequential algorithm (VariScreen). VariScreen incorporated data on platelets or liver stiffness measurements and then the ratio of platelets to liver stiffness measurement, adjusted for etiology, patient sex, and international normalized ratio. RESULTS In the derivation population, endoscopies were spared for 23.9% of patients using the B6C (VNT missed in 2.9%), 24.3% of patients using the ANTICIPATE score (VNT missed in 4.6%), 34.5% of patients using VariScreen (VNT missed in 2.9%), and 41.9% of patients using the EB6C (VNT missed in 10.9%). Differences in spared endoscopy rates were significant (P ≤ .001), except for B6C vs ANTICIPATE and in missed VNT only for EB6C vs the others (P ≤ .009). VariScreen was the only safe test regardless of sex or etiology (missed VNT, ≤5%). Moreover, VariScreen secured screening without missed VNT in patients with model for end-stage liver disease scores higher than 10. This overall strategy performed better than a selective strategy restricted to patients with compensated liver disease. Test performance and safety did not differ significantly among populations. CONCLUSIONS In a retrospective study of data from 2368 patients with chronic liver disease, we found that the B6C are safe whereas the EB6C are unsafe, based on missed VNT. The VariScreen algorithm performed well in patients with chronic liver disease of any etiology or severity-it is the only test that safely rules out VNT and can be used in clinical practice.