A Novel Model to Predict Esophageal Varices in Patients with Compensated Cirrhosis Using Acoustic Radiation Force Impulse Elastography

BACKGROUND & AIMS: Few noninvasive methods can accurately identify esophageal varices (EVs) in patients with compensated cirrhosis. We developed and validated a novel, acoustic radiation force impulse (ARFI) elastography-based prediction model for high-risk EVs (HEVs) in patients with compensated cirrhosis. METHODS: A total of 143 patients with compensated cirrhosis between February, 2010 and February, 2013 (training set) and 148 between June, 2010 and May, 2013 (validation set) who underwent ARFI elastography and endoscopy were prospectively recruited. Independent predictors of HEVs were used to construct a prediction model. RESULTS: Based on multivariate analysis, we developed two new statistical models, a varices risk score and ARFI-spleen diameter-to-platelet ratio score (ASPS), the latter of which was calculated as ARFI velocity × spleen diameter/platelet count. The area under receiver operating characteristic curve (AUROC) of the varices risk score and ASPS to predict HEVs were 0.935 (95% confidence interval [CI] 0.882–0.970) and 0.946 (95% CI 0.895–0.977), respectively. When ASPS, a simpler model with a higher AUROC, was applied in the validation set, acceptable diagnostic accuracy for HEVs was observed (AUROC = 0.814 [95% CI 0.743–0.885]). To detect HEVs, a negative predictive value of 98.3% was achieved at ASPS <2.83, whereas a positive predictive value of 100% was achieved at ASPS >5.28. CONCLUSIONS: ASPS, a novel noninvasive ARFI-based prediction model, can accurately identify HEVs in patients with compensated cirrhosis. ASPS <2.83 may safely rule out the presence of HEVs, whereas patients with ASPS >5.28 should be considered for endoscopic examinations or appropriate prophylactic treatment.