Inter-reader agreement of magnetic resonance imaging proton density fat fraction and its longitudinal change in a clinical trial of adults with nonalcoholic steatohepatitis

Abstract

Purpose: To determine the inter-reader agreement of magnetic resonance imaging proton density fat fraction (PDFF) and its longitudinal change in a clinical trial of adults with nonalcoholic steatohepatitis (NASH).

Study type: We performed a secondary analysis of a placebo-controlled randomized clinical trial of a bile acid sequestrant in 45 adults with NASH. A six-echo spoiled gradient-recalled-echo magnitude-based fat quantification technique was performed at 3 T. Three independent readers measured MRI-PDFF by placing one primary and two additional regions of interest (ROIs) in each segment at both time points. Cross-sectional agreement between the three readers was evaluated using intra-class correlation coefficients (ICCs) and coefficients of variation (CV). Additionally, we used Bland-Altman analyses to examine pairwise agreement between the three readers at baseline, end of treatment (EOT), and for longitudinal change.

Results: Using all ROIs by all readers, mean PDFF at baseline, at EOT, and mean change in PDFF was 16.1%, 16.0%, and 0.07%, respectively. The 27-ROI PDFF measurements had 0.998 ICC and 1.8% CV at baseline, 0.998 ICC and 1.8% CV at EOT, and 0.997 ICC for longitudinal change. The 9-ROI PDFF measurements had corresponding values of 0.997 and 2.6%, 0.996 and 2.4%, and 0.994. Using 27 ROIs, the magnitude of the bias between readers for whole-liver PDFF measurement ranged from 0.03% to 0.06% points at baseline, 0.01% to 0.07% points at EOT, and 0.01% to 0.02% points for longitudinal change.

Conclusion: Inter-reader agreement for measuring whole-liver PDFF and its longitudinal change is high. 9-ROI measurements have only slightly lower agreement than 27-ROI measurements.

Keywords: Inter-reader agreement; Liver; Proton density fat fraction (PDFF); Quantitative imaging biomarker; Region of interest; Reproducibility.

Fig. 1

shows the liver covered from dome (top left) to tip (bottom right) on the magnitude-based MRI proton density fat fraction (PDFF) maps. The 27 regions of interest (ROIs) placed by each reader are shown, color coded by reader. The primary ROI in each segment has a continuous border. The secondary ROIs have dashed borders. Note that the scale bar reflects the fat-water signal dominance ambiguity using magnitude reconstruction.

Fig. 2

illustrates the results of the Bland-Altman analysis for measuring PDFF at baseline and its change longitudinally for 9- and 27-ROI methods. Plots for measuring PDFF at end of treatment (not shown) are similar to those at baseline.