Diagnostic Performance of Attenuation to Stage Liver Steatosis with MRI Proton Density Fat Fraction as Reference: A Prospective Comparison of Three US Machines

Background US tools to quantify liver fat content have recently been made clinically available by different vendors, but comparative data on their accuracy are lacking. Purpose To compare the diagnostic performances of the attenuation parameters of US machines from three different manufacturers (vendors 1, 2, and 3) in participants who underwent liver fat quantification with the MRI-derived proton density fat fraction (PDFF). Materials and Methods From July 2020 to June 2021, consecutive participants with chronic liver disease were enrolled in this prospective single-center study and underwent MRI PDFF quantification (reference standard) and US on the same day. US was performed with two different machines from among three vendors assessed. Areas under the receiver operating characteristic curve (AUCs) for the staging of liver steatosis (MRI PDFF: ≥5.5% for grade ≥S1 and ≥15.5% for grade ≥S2) were calculated in test and validation samples and then compared between vendors in the study sample. Results A total of 534 participants (mean age, 60 years ± 13 [SD]; 320 men) were evaluated. Failure of measurements occurred in less than 1% of participants for all vendors. Correlation coefficients with the MRI PDFF were 0.71, 0.73, and 0.54 for the attenuation coefficients of vendors 1, 2, and 3, respectively. In the test sample, AUCs for diagnosis of steatosis grade S1 and higher and grade S2 and higher were 0.89 and 0.93 for vendor 1 attenuation, 0.88 and 0.92 for vendor 2 attenuation, and 0.79 and 0.79 for vendor 3 attenuation, respectively. In the validation sample, a threshold value of 0.65 for vendor 1 and 0.66 for vendor 2 yielded sensitivity of 77% and 84% and specificity of 78% and 85%, respectively, for diagnosis of grade S1 and higher. Vendor 2 attenuation had greater AUCs than vendor 3 attenuation (P = .001 and P = .003) for diagnosis of grade S1 and higher and grade S2 and higher, respectively, and vender 2 had greater AUCs for attenuation than vendor 1 for diagnosis of grade S2 and higher (P = .04). For all vendors, attenuation was not associated with liver stiffness (correlation coefficients <0.05). Conclusion To stage liver steatosis, attenuation coefficient accuracy varied among US devices across vendors when using MRI proton density fat fraction quantification as the reference standard, with some demonstrating excellent diagnostic performance and similar cutoff values. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Dubinsky in this issue.