Quantification of Hepatic Steatosis by Ultrasound: Prospective Comparison With MRI Proton Density Fat Fraction as Reference Standard

BACKGROUND. Multiple ultrasound platforms now provide quantitative measures of hepatic steatosis. One such measure is the ultrasound-derived fat fraction (UDFF), which combines attenuation and backscatter quantification. OBJECTIVE. The purpose of this study was to characterize agreement between UDFF and MRI proton-density fat fraction (PDFF) measurements. METHODS. This prospective cross-sectional study enrolled 56 overweight and obese adolescents and adults (age >= 16 years) who underwent investigational ultrasound (deep abdominal transducer) and MRI examinations of the liver during a single visit from August 2020 to October 2020. Ultrasound examinations included three UDFF acquisitions of five measurements each (15 measurements total), and an overall median of medians was computed (UDFFoverall). MRI examinations included three PDFF acquisitions with calculation of an overall median PDFF. Spearman rank-order correlation was computed between UDFF and MRI PDFF measurements. Intraclass correlation coefficients and Bland-Altman difference plots were used to assess agreement. ROC curves were used to assess diagnostic performance of UDFF for detecting MRI PDFF of 5.5% or more. RESULTS. Median participant age was 32.5 years (IQR, 24.0-39.0 years); 40 participants were female, and 16 were male. A total of 34 (60.7%) participants had an MRI PDFF of 5.5% or more. UDFFoverall was 10.5% (IQR, 5.0-20.0%); median MRI PDFF was 6.1% (IQR, 3.4-13.7%). UDFF (overall) was positively associated with MRI PDFF (rho, 0.82; p <.001; intraclass correlation coefficient, 0.84 [95% CI, 0.59- 0.93]). Mean bias between UDFF and PDFF was 4.0% (95% limits of agreement, -7.9% to 15.9%), with similar bias if summarizing UDFF by the first five measurements (4.4%), first three measurements (4.4%), or first measurement (4.6%). UDFFoverall AUC was 0.90 (95% CI, 0.79-0.96) for MRI PDFF of 5.5% or more; AUC was not significantly different when it was based on the number of UDFF measurements (p =.11-.97 for all pairwise AUC comparisons). UDFF (overall) cutoff of more than 5% had sensitivity of 94.1% and specificity of 63.6% for diagnosing MRI PDFF of 5.5% or more. CONCLUSION. Measurements of hepatic steatosis using UDFF show strong agreement with measurements by MRI PDFF. A UDFF (overall) cutoff of more than 5% provides high AUC and sensitivity, albeit low specificity, for detection of MRI PDFF of 5.5% or more. CLINICAL IMPACT. UDFF may have a clinical role in detection of hepatic steatosis. A reduced number of individual measurements is likely sufficient for determining an overall UDFF value.