Background: The consistency in steatosis grading between magnetic resonance imaging-based proton density fat fraction (MRI-PDFF) and controlled attenuation parameter (CAP) before and after treatment remains unclear. This study aimed to compare the diagnostic accuracy of steatosis grading between MRI-PDFF and CAP using liver biopsy as standard and to evaluate the value of monitoring changes in steatosis grading with CAP during follow-up utilizing MRI-PDFF as a reference.

Methods: Consecutive patients from a biopsy cohort and a randomized controlled trial were included in this study and classified into 3 groups (the biopsy, orlistat treatment, and routine treatment subgroups). Hepatic steatosis was measured via MRI-PDFF and CAP at baseline and at the 6th month; the accuracy and cutoffs were assessed in the liver biopsy cohort at baseline.

Results: A total of 209 consecutive patients were enrolled. MRI-PDFF and CAP showed comparable diagnostic accuracy for detecting pathological steatosis [⩾S1, area under the receiver operating characteristic curve (AUC) = 0.984 and 0.972, respectively]; in contrast, CAP presented significantly lower AUCs in grades S2-3 and S3 (0.820 and 0.815, respectively). The CAP values correlated well with the MRI-PDFF values at baseline and at the 6th month (r = 0.809 and 0.762, respectively, both p < 0.001), whereas a moderate correlation in their changes (r = 0.612 and 0.524 for moderate-severe and mild steatosis, respectively; both p < 0.001) was observed. The AUC of CAP change was obtained to predict MRI-PDFF changes of ⩾5% and ⩾10% (0.685 and 0.704, p < 0.001 and p = 0.001, respectively). The diagnostic agreement of steatosis grade changes between MRI-PDFF and CAP was weak (κ = 0.181, p = 0.001).

Conclusions: CAP has decreased value for the initial screening of moderate-severe steatosis and is limited in monitoring changes in steatosis during treatment. The confirmation of steatosis grading with MRI-PDFF remains necessary.