S2797 Comparison of Non-Invasive Indices in Predicting Esophageal Varices in Mexican Population

Introduction: Variceal bleeding has a high mortality rate in patients with cirrhosis. Therefore, timely screening for esophageal varices (EV) is essential. Baveno VII recommends upper endoscopy based on liver stiffness measurement and platelet count; however, these criteria have limitations. Recently, non-invasive indices have been developed to predict the presence of EV. We compare the diagnostic performance of EVendo score, international normalized ratio to platelet ratio (INPR) and Lok score. Methods: We included patients with cirrhosis regardless of the stage (Child-Pugh, MELD), any etiology, >18 years, both sexes, and without prior screening for EV. An observational study was conducted. Statistical analyses were performed using IBM SPSS Statistics. Frequencies and medians were used for population descriptive analyses. Chi-square and Student’s t-test were used for group comparison. ROC curves and Youden index were used to determine the cutoff values with the highest sensitivity and specificity. P-value of < 0.05 was taken as statistically significant. Additionally, positive predictive value (PPV), negative predictive value (NPV), and likelihood ratios were calculated. Results: 139 patients were included: women 51% and men 48% and mean age of 54 ± 13 years. The etiologies included were hepatitis C virus infection 31%, alcohol-related liver disease 26%, autoimmune liver diseases 24%, and metabolic dysfunction-associated steatotic liver disease 13%. Regarding Child-Pugh classification 52 (37%) were Class A, 67 (48%) B, and 20 (14%) C. Endoscopic findings: Twenty-eight percent had large varices, 40% had small varices, and 31% had no EV. When analyzing each of the indices for detecting EV regardless of size, we found that using a cutoff of ≥0.84 for Lok Score had a sensitivity of 74%, specificity of 77%, PPV of 87%, and NPV of 59%. For EVendo Score, a cutoff of ≥5 had a sensitivity of 66%, specificity of 86%, PPV of 91%, and NPV of 54%. Finally, INPR with a cutoff of ≥0.94 had a sensitivity of 83%, specificity of 71%, PPV of 85%, and NPV of 66%. Conclusion: All indices demonstrated acceptable sensitivity and specificity; however, the index with the highest sensitivity for predicting EV was INPR (83%). EVendo Score provided the highest specificity and PPV, regardless of variceal size. The Lok Score did not show any advantage over the other indices. Although none of the scores performed perfectly, they could complement clinical decision-making by helping stratify risk and prioritize the need for endoscopy particularly in resource-limited settings.