Accuracy of a dedicated 100 Hz vibration-controlled spleen stiffness measurement for the detection of esophageal varices in naïve patients with compensated advanced chronic liver disease: interim results from a multicentric cohort

Introduction Clinically significant portal hypertension (CSPH) marks a critical step in the natural history of compensated advanced chronic liver disease (cACLD) and may lead to esophageal varices (EV). The Baveno VI criteria suggest liver stiffness measurement (LSM) and platelet count (PLT) for non-invasive identification of cACLD patients not requiring screening gastroscopy. Aim We investigated the accuracy of the novel 100 Hz vibration-controlled transient elastography-based spleen stiffness measurement (SSM) exam for the identification of EV in cACLD patients. Materials and Methods Retrospective study of Mainz, Vienna, Leuven, Rome and Palermo. Patients with cACLD of any etiology (LSM≥10kPa or histological F4 fibrosis), but without previous decompensation (bleeding, encephalopathy, ascites) were included. SSM and LSM were obtained using Fibroscan F630 ≤1 month within screening gastroscopy. Prediction performance between different SSM cut-offs with respect to the Baveno criteria (LSM>20kPa and/or PLT<150 G/L) were compared by logistic regression with 10-fold cross-validation, adjusted for age, gender, BMI, transaminases, INR, albumin, and bilirubin. Backward feature selection based on likelihood ratio test was applied to identify significant confounders. Performance was calculated by balanced accuracy (BA), specificity (SP) and sensitivity (SE). Wilcoxon test was used to evaluate significant performance improvement of SSM cut-offs with respect to Baveno criteria, or to signifcant confounders. Results 343 cACLD patients with a median age of 59 years (60.3% male) and NAFLD as the main etiology (51.3%) were included. 137 had EV with 49 high-risk EV (HR-EV), while median SSM, LSM and PLT were 40.5kPa, 21kPa and 139 G/L, respectively. The figure shows BAs at different SSM cut-offs, compared to Baveno (red line). The best overall performance with all-type EV was at SSM=60 kPa (BA=0.72, SP=0.86, SE=0.58); Baveno: BA=0.66, SP=0.86, SE=0.39. Comparing HR-EV vs. absence of EV, the best cut-off was at 50kPa (BA=0.71, SP=0.95, SE=0.47; Baveno: BA=0.56, SP=0.92, SE=0.29). These SSM thresholds significantly improved BA when significant confounders were considered. Conclusions The novel spleen-dedicated 100 Hz SSM is associated with presence of EV in cACLD patients. In both all-type EV and HR-EV, SSM showed better accuracy than the Baveno LSM-PLT criteria, achieving a better trade-off between SP and SE.