Predictors of Discordance Between Controlled Attenuation Parameter and Magnetic Resonance-Proton Density Fat Fraction in Hepatic Steatosis

BACKGROUND/AIMS: Magnetic resonance imaging-proton density fat fraction (MRI-PDFF) is the reference standard for quantifying hepatic steatosis, but the more accessible controlled attenuation parameter (CAP) often yields discordant results, risking misclassification. We aimed to identify clinical factors driving this discordance and develop a risk-based algorithm to guide steatosis assessment. METHODS: This retrospective, cross-sectional study was conducted at a single tertiary hospital. We identified 1,821 patients who underwent both CAP and MRI-PDFF within a 90-day interval. Using MRI-PDFF as the reference, we performed multivariable logistic regression to identify predictors for CAP overestimation and underestimation, defined as a ≥2-grade difference. RESULTS: Significant discordance (≥2-grade difference) was observed in 387 of 1,821 patients (21.3%), with CAP overestimating in 13.0% and underestimating in 8.3%. After adjustment, overestimation was independently associated with older age (adjusted odds ratio [aOR], 1.016; 95% confidence interval [CI], 1.007-1.024), higher body mass index (BMI) (aOR, 1.115; 95% CI, 1.082-1.149), and diabetes mellitus (aOR, 1.448; 95% CI, 1.022-2.050). Conversely, underestimation was associated with younger age (aOR, 0.949; 95% CI, 0.935-0.964) and lower BMI (aOR, 0.946; 95% CI, 0.908-0.985). Predictive models achieved an area under the curve of 0.704 for overestimation and 0.765 for underestimation. CONCLUSIONS: Over- and underestimation of hepatic steatosis by CAP are driven by distinct clinical predictors. Our risk algorithm reliably identifies a "concordant zone" where CAP can be used confidently, while guiding selective MRI-PDFF confirmation for patients at high risk of discordance, thereby optimizing non-invasive assessment strategies.