Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis

Gastroesophageal varices (GOV) is a common complication in patients with portal hypertension. We conducted a meta-analysis in attempt to evaluate the diagnostic accuracy of computed tomography (CT) as a noninvasive imaging tool for identifying GOV in reference to esophagogastroduodenoscopy (EGD).

A systemic literature search of multiple databases were conducted to identify articles that investigated the diagnostic performance of CT for GOV, while employing EGD as reference standard. A 2×2 table was conducted according to the available published data for both esophageal varices (EV) and gastric varices (GV) as individual subgroups. The following indices were calculated: pooled sensitivity and specificity, positive and negative likelihood ratio, diagnostic odds ratio, and area under receiver operating characteristics. All statistical analyses were conducted via STATA13.0 and RevMan5.3.

A total of 11 studies were included in this meta-analysis, 10 articles evaluated the diagnostic accuracy of CT for EV (807 subjects) and 7 articles for GV (583 subjects). The pooled sensitivity and specificity for identifying EV were 0.896 (95% CI, 0.841-0.934) and 0.723 (95% CI, 0.644-0.791), respectively, with an AUROC of 0.86. The pooled sensitivity and specificity for identifying GV were 0.955 (95% CI, 0.903-0.980) and 0.658 (95% CI, 0.433-0.829), respectively, with an AUROC of 0.95. A subgroup analysis suggested varying CT technology could serve as a potential source of heterogeneity between included studies. A Deek's funnel plot indicated a low probability for publication bias.

Computed tomography could potentially replace EGD as a primary screening tool for diagnosing GOV, however results should be interpreted with caution given its suboptimal specificity.