Non-Invasive Prediction Models for Esophageal Varices and Red Signs in Patients With Hepatitis B Virus-Related Liver Cirrhosis.

Abstract

Background: Red signs are closely related to esophageal variceal bleeding, and, despite improvements in therapy, the mortality rate remains high. We aimed to identify non-invasive predictors of esophageal varices and red signs in patients with hepatitis B virus-related liver cirrhosis. Methods: This retrospective study included 356 patients with hepatitis B virus-related liver cirrhosis after applying inclusion and exclusion criteria among 661 patients. All patients underwent endoscopy, ultrasonography, laboratory examinations, and computed tomography/magnetic resonance imaging. Univariate and multivariate logistic regression analysis were performed, and prediction models for esophageal varices and red signs were constructed. Results: Multivariate analysis revealed that spleen diameter, splenic vein diameter, and lymphocyte ratio were independent risk factors for esophageal varices and red signs. On this basis, we proposed two models: i) a spleen diameter-splenic vein diameter-lymphocyte ratio-esophageal varices prediction model (SSL-EV model); and ii) a spleen diameter-splenic vein diameter-lymphocyte ratio-red sign prediction model (SSL-RS model). The areas under the receiver operating characteristic curve for the two prediction models were 0.843 and 0.783, respectively. With a cutoff value of 1.55, the first prediction model had 81.3% sensitivity and 76.1% specificity for esophageal varices prediction. With a cutoff value of −0.20, the second prediction model had 72.1% sensitivity and 70.7% specificity for the prediction of red signs. Conclusions: We proposed a new statistical model, the spleen diameter-splenic vein diameter-lymphocyte ratio-red sign prediction model (SSL-RS model), to predict the presence of red signs non-invasively. Combined with the spleen diameter-splenic vein diameter-lymphocyte ratio-esophageal varices prediction model (SSL-EV model), these non-invasive prediction models will be helpful in guiding clinical decision-making and preventing the occurrence of esophageal variceal bleeding.