A Reappraisal of the Diagnostic Performance of B-Mode Ultrasonography for Mild Liver Steatosis.

Abstract

Previous studies have shown that ultrasonography has high specificity (80%-100%) but low sensitivity (50%-70%) in diagnosing fatty liver; sensitivity is especially low for mild steatosis. In this study, we aimed to reappraise the diagnostic performance of B-mode ultrasonography (B-USG) for fatty liver disease. We performed a retrospective, multinational, multicenter, cross-sectional, observational study (6 referral centers from 3 nations). We included 5,056 participants who underwent both B-USG and magnetic resonance proton density fat fraction (MRI-PDFF) within a 6-month period. The diagnostic performance of B-USG was compared with that of MRI-PDFF as a reference standard for fatty liver diagnosis, using sensitivity, specificity, positive and negative predictive values, diagnostic accuracy, and area under the receiver operating characteristic curve (AUC). B-USG showed a sensitivity of 83.4%, specificity of 81.0%, and AUC of 0.822 in diagnosing mild liver steatosis (6.5% ≤MRI-PDFF ≤14%). The sensitivity, specificity, and AUC in diagnosing the presence of fatty liver disease (MRI-PDFF ≥6.5%) were 83.4%, 81.0%, and 0.822, respectively. The mean PDFF of B-USG-diagnosed nonfatty liver differed significantly from that of diagnosed mild liver steatosis (3.5% ± 2.8% vs 8.5% ± 5.0%, P < 0.001). The interinstitutional variability of B-USG in diagnosing fatty liver was similar in diagnostic accuracy among the 6 centers (range, 82.8%-88.6%, P = 0.416). B-USG was an effective, objective method to detect mild liver steatosis using MRI-PDFF as comparison, regardless of the etiologies and comorbidities.