Prediction of esophageal varices by liver and spleen MR elastography

Objectives We aimed to assess the diagnostic performance of MR elastography (MRE) in predicting esophageal varices (EVs) in patients with chronic liver disease. Methods We prospectively performed liver (LSM) and spleen stiffness measurements (SSM) using MRE and endoscopic screening for EVs to determine if patients with hepatocellular carcinoma were eligible for resection. We investigated whether LSM, SSM, and other non-invasive preoperative parameters were associated with the presence of EVs. In order to predict EVs, 211 patients were divided into training (n = 140) and test (n = 71) groups. A nomogram was built using independent factors based on logistic regression analysis in the training group and its accuracy was validated using an independent cohort. Results Forty-six patients (21.8%) were diagnosed as having EVs (mild, n = 36; severe, n = 10). According to multiple regression analysis, LSM (odds ratio, 2.362; 95% confidence interval [CI], 1.341-4.923; p = 0.001) and SSM (1.489; 1.095-2.235; p = 0.010) were independent predictors of EVs in the training group. The nomogram showed good discrimination, with a C-index of 0.942 (95% CI, 0.892-0.974) through internal validation, and good calibration. Application of the nomogram in the test group still gave good discrimination (C-index, 0.948; 95% CI, 0.868-0.995). Conclusions The combination of LSM and SSM using MRE is an accurate tool to identify patients at risk for EVs.