Assessment of variceal bleeding in cirrhotic patients: accuracy of multi-detector computed tomography

Abstract

Background: Esophageal variceal hemorrhage (EVH) has been shown to be a leading cause of mortality in patients with portal hypertension. Our purpose was to assess the utility of multi-detector computed tomography (MDCT) features in the assessment of esophageal varices (EVs) and esophageal variceal hemorrhage (EVH). This prospective study included 85 cirrhotic patients who underwent MDCT and Upper Gastrointestinal Tract (UGIT) endoscopy within 2 weeks. Four radiologists evaluated the presence of EVs and the presence and size of different collaterals. Multivariable logistic regression analysis was calculated to investigate the significant predictors influencing EV and EVH. Results: Findings of EV with MDCT were the best predictor of EV or EVH. The presence (and/or size) of following collaterals had significant association with both EV and EVH: paraesophageal (p < 0.001, < 0.001), short gastric (p = 0.024, 0.010), gastric varicosities (p < 0.001, < 0.001), coronary (p < 0.001, < 0.001), and main coronary vein (MCV) (p < 0.001, = 0.011). We proposed an imaging-based model (presence of coronary collaterals, main coronary vein size > 3.5 mm, presence of short gastric collaterals, presence of gastric varicosities, size > 1.5 mm) with 97% sensitivity, 91% specificity, and 94% accuracy to predict EVs. We suggested another model (presence of paraesophageal collaterals, presence of short gastric vein (SGC), SGC size > 2.5 mm, main coronary vein size > 3.5 mm, gastric varicosities size > 1.5 mm, size of EVs > 4 mm, and Child C score) to predict EVH with 98% sensitivity, 81% specificity, and 89.5% accuracy. Inter-observer agreement was high in the detection of EVs (W. Kappa = 0.71–0.88). Conclusion: MDCT is an effective modality in the diagnosis of EVs. At MDCT, the presence and/or size of various collaterals including para-esophageal, short gastric, coronary collaterals, and gastric varicosities are accurate predictors for either EVs existence or EVH. We suggested two computed tomography imaging-based models with high reproducibility and acceptable accuracy for the prediction of EV and EVH. With cirrhotic patients, we recommend that radiologists report collaterals in their daily practice.