RESUMO
PURPOSE: To evaluate the diagnostic performance of previously proposed high-specificity magnetic resonance (MR) imaging-estimated proton density fat fraction (PDFF) thresholds for diagnosis of steatosis grade 1 or higher (PDFF threshold of 6.4%), grade 2 or higher (PDFF threshold of 17.4%), and grade 3 (PDFF threshold of 22.1%) by using histologic findings as a reference in an independent cohort of adults known to have or suspected of having nonalcoholic fatty liver disease (NAFLD). MATERIALS AND METHODS: This prospective, cross-sectional, institutional review board-approved, HIPAA-compliant single-center study was conducted in an independent cohort of 89 adults known to have or suspected of having NAFLD who underwent contemporaneous liver biopsy. MR imaging PDFF was estimated at 3 T by using magnitude-based low-flip-angle multiecho gradient-recalled-echo imaging with T2* correction and multipeak modeling. Steatosis was graded histologically (grades 0, 1, 2, and 3, according to the Nonalcoholic Steatohepatitis Clinical Research Network scoring system). Sensitivity, specificity, and binomial confidence intervals were calculated for the proposed MR imaging PDFF thresholds. RESULTS: The proposed MR imaging PDFF threshold of 6.4% to diagnose grade 1 or higher steatosis had 86% sensitivity (71 of 83 patients; 95% confidence interval [CI]: 76, 92) and 83% specificity (five of six patients; 95% CI: 36, 100). The threshold of 17.4% to diagnose grade 2 or higher steatosis had 64% sensitivity (28 of 44 patients; 95% CI: 48, 78) and 96% specificity (43 of 45 patients; 95% CI: 85, 100). The threshold of 22.1% to diagnose grade 3 steatosis had 71% sensitivity (10 of 14 patients; 95% CI: 42, 92) and 92% specificity (69 of 75 patients; 95% CI: 83, 97). CONCLUSION: In an independent cohort of adults known to have or suspected of having NAFLD, the previously proposed MR imaging PDFF thresholds provided moderate to high sensitivity and high specificity for diagnosis of grade 1 or higher, grade 2 or higher, and grade 3 steatosis. Prospective multicenter studies are now needed to further validate these high-specificity thresholds.