This meta-analysis aimed to evaluate the diagnostic accuracy of computed tomography (CT) for detecting gastroesophageal varices (GEVs) and identify high-risk GEVs in patients with cirrhosis. A comprehensive search of databases identified 28 studies reporting on CT-based diagnosis for GEVs confirmed via endoscopy. Meta-analyses were conducted to calculate pooled sensitivity (SEN) and pooled specificity (SPE), positive likelihood ratio (PLR) and negative likelihood ratio (NLR), diagnostic odds ratio (DOR), and the area under the curve (AUC). Based on the number of patients (or varices), the pooled SEN, SPE, PLR, NLR, DOR, and AUC of CT-based diagnosis were estimated at 0.91 (0.92), 0.81 (0.45), 4.82 (1.67), 0.11 (0.17), 42.47 (10.26), and 0.93 (0.94), respectively, for any GEV and at 0.89 (0.89), 0.90 (0.79), 8.86 (4.28), 0.12 (0.14), 75.71 (30.19), and 0.95 (0.85), respectively, for high-risk GEVs. Subgroup analyses indicated that CT had a higher diagnostic accuracy for esophageal varices compared with gastric varices (AUC: 0.93 vs. 0.89, P < 0.05), and the 64-slice CT yielded superior SEN compared with 16-slice and <16-slice CT (AUC: 0.97 vs. 0.92 and 0.82, respectively, P < 0.05). Prospective studies demonstrated higher diagnostic accuracy than retrospective studies (AUC: 0.95 vs. 0.90, P < 0.05). Regarding variceal size, a cut-off of 3 mm and 5 mm discriminated between low- and high-risk individuals, respectively, with high diagnostic accuracy (AUC: 0.992 vs. 0.997, P > 0.05). CT demonstrates promising diagnostic accuracy for identifying gastroesophageal varices and distinguishing high-risk GEVs in patients with cirrhosis. Further research to validate optimal variceal size cut-offs is warranted to enhance clinical utility. Such a high diagnostic accuracy of CT scans for predicting varices is clinically meaningful for patients with cirrhosis accompanied by portal hypertension. If high-risk varices are identified at CT scans, early intervention would be helpful to reduce the risk of variceal bleeding.
Read full abstract