A Simplified non-invasive model based on albumin, bilirubin, and platelet count for predicting high-risk varices in compensated cirrhosis
Aim Esophagogastroduodenoscopy (EGD) is the gold standard for detecting high-risk varices (HRV) in patients with liver cirrhosis, but is invasive and resource-intensive. The Albumin-Bilirubin-Platelet (ABP) model is non-invasive, but its performance requires validation in diverse populations. This study aimed to validate and refine the ABP criteria to predict HRV. Methods This retrospective study included 474 patients with compensated cirrhosis, categorized into training ( n = 332) and validation ( n = 142) cohorts. All patients underwent EGD and laboratory evaluation. We evaluated the performance of the original and modified ABP (mABP) criteria, derived from the training cohort, to identify patients in whom screening EGD could be safely avoided. Results The prevalences of HRV were 15.96% and 20.42% in the training and validation cohorts, respectively. In multivariate analysis of the training cohort, only platelet count was an independent predictor of HRV ( p < 0.001). The original ABP criteria spared 21.83% of EGDs in the validation cohort with a 0% missed HRV rate. In the training cohort, the mABP criteria spared 51.51% of EGDs with a 2.34% missed HRV rate. The mABP criteria spared 58.45% of EGDs in the validation cohort, and the missed HRV rate was 4.82%. Conclusion The original ABP criteria provide a simple and safe strategy for identifying low-risk patients with compensated cirrhosis. The modified ABP criteria substantially improve endoscopy avoidance while maintaining acceptable safety, offering a pragmatic alternative in settings where access to advanced diagnostic tools is limited.