Development of a nomogram model based on spleen volume change to predict high-risk esophageal varices in patients with liver cirrhosis

TL;DR

This study creates a non-invasive tool using spleen volume changes to predict high-risk esophageal varices in liver cirrhosis patients, reducing the need for uncomfortable endoscopies.

Contribution

A novel nomogram model based on spleen volume changes and clinical factors to predict high-risk esophageal varices in liver cirrhosis patients.

Findings

The nomogram model achieved an AUC of 0.793, outperforming existing models like platelet-to-spleen volume ratio.

The model demonstrated good calibration and clinical utility with a C-index of 0.779 after internal validation.

The model's sensitivity and specificity were 0.797 and 0.671, respectively, at a probability cutoff of 0.421.

Abstract

Esophageal variceal (EV) rupture is a life-threatening complication of liver cirrhosis. Although upper gastrointestinal endoscopy is recommended for routine screening and risk assessment of EV bleeding, it is an invasive and often unpleasant procedure. This study aims to develop a non-invasive nomogram model based on spleen volume changes to predict the presence of high-risk esophageal varices (HREVs).

A total of 150 patients with liver cirrhosis (mean age 62.3 ± 10.0 years; 95 men and 55 women) who underwent upper gastrointestinal endoscopy were retrospectively included. Spleen volume was measured using abdominal computed tomography. Predictors were identified through multivariate logistic regression and subsequently used to construct a nomogram model. The discriminative ability, calibration ability, and clinical utility were assessed. Internal validation was performed using 1,000 bootstrap resampling iterations.

Based on endoscopic findings, 74 patients were categorized into the HREV group and 76 patients were categorized into the non-HREV group. Multivariate regression identified three independent predictors of HREV: the presence of ascites [odds ratio (OR) = 2.656, 95% confidence interval (CI): 1.224–5.763], prothrombin time (OR = 1.217, 95% CI: 1.043–1.419), and spleen volume enlargement rate (OR = 1.589, 95% CI: 1.276–1.979). These variables were incorporated into the nomogram model. The area under the receiver operating characteristic curve of the nomogram model was 0.793 (95% CI: 0.723–0.863), outperforming previously reported models, such as the platelet-to-spleen volume ratio (0.724), platelet-to-spleen diameter ratio (0.673), aspartate aminotransferase-to-platelet ratio index (0.590), and aspartate aminotransferase-to-alanine aminotransferase ratio (0.558). At a probability cutoff of 0.421, the nomogram demonstrated a sensitivity of 0.797, a specificity of 0.671, a positive predictive value of 0.702, a negative predictive value of 0.773, and an accuracy of 0.733. Internal validation yielded a C-index of 0.779 (95% CI: 0.714–0.853). Overall, the nomogram model exhibited good calibration and favorable clinical utility.

The nomogram incorporating ascites, prothrombin time, and spleen volume enlargement rate effectively predicts HREVs in patients with liver cirrhosis. This non-invasive and user-friendly tool offers an efficient approach for timely HREV evaluation and preventive treatment of variceal bleeding.

Figures