Purpose: To investigate the associated Multi-Detector Computed Tomography (MDCT) features for esophageal varices (EVs) and esophageal variceal hemorrhage (EVH), with particular emphasis on different collateral veins.

Materials and methods: All cirrhotic patients who had undergone both Upper Gastrointestinal Tract (UGIT) endoscopy and contrast-enhanced MDCT within 6 months from 2013 to 2019 were included in the study. MDCT of 124 patients, 76 males and 48 females, aged between 21 and 73 years old were evaluated for presence of EV and presence and size of different collaterals. The presence and size of collaterals in patients with high-risk EVs or EVH were compared with others.

Results: Findings of EV in MDCT analysis were the best predictor of EV or EVH, and presence (and/or size) of following collaterals showed a significant relationship with both EV and EVH: coronary (p = 0.006, 0.002), short gastric (SGC) (p = 0.02, < 0.001), and paraesophageal (p = 0.04, 0.01). Those presenting each aforementioned collaterals or with higher collateral size were more likely to develop the EV or EVH. Yet, other collaterals indicated no similar association: para-umbilical, omental, perisplenic, and splenorenal. Main coronary vein (p = 0.02, 0.03) and fundus (p = 0.006, 0.001) varices' sizes were also significantly higher in patients with EV or EVH. Finally, we suggested an imaging-based model (presence of SGC, SGC size > 2.5 mm, presence of EV, and coronary vein size > 3.5 mm) with 75.86% sensitivity, 76.92% specificity, and 76.36% accuracy to predict the presence of EVs according to UGIT endoscopy. Furthermore, we presented another model (presence of SGC, SGC size > 2.5 mm, presence of EV, and MELD score > 11.5 mm) to predict the occurrence of EVH with 75.86% sensitivity, 76.92% specificity, and 76.36% accuracy.

Conclusion: We suggested imaging characteristics for predicting EV and EVH with especial emphasis on the presence and size of various collaterals; then, we recommended reliable imaging criteria with high specificity and accuracy for predicting the EV and EVH.