Magnetic Resonance Imaging Systematically Differs from Histology in Quantifying Macrovesicular Liver Steatosis in...

Background: Liver histology remains the gold standard for assessing liver steatosis (LS); however, non-invasive methods are increasingly being explored in clinical practice. This study aimed to evaluate the agreement between magnetic resonance imaging (MRI) and liver histology in quantifying LS in patients with morbid obesity undergoing bariatric surgery (BS). Methods: This ancillary study is part of a prospective, double-blind, multicenter, randomized placebo-controlled trial investigating the effects of preoperative omega-3 polyunsaturated fatty acid supplementation on liver volume in morbidly obese patients undergoing BS. The parent trial yielded negative results, and randomization arm was retained as a covariate in all analyses. Patients underwent MRI within 2 days before surgery, followed by intraoperative wedge resection and TruCore needle liver biopsy. Agreement between MRI and histology was assessed using the intraclass correlation coefficient (ICC) and Cohen's kappa coefficient (K) for both macro- and microvesicular steatosis. Results: Thirty-seven patients were enrolled; paired MRI and biopsy data were available for thirty-one (83.8%). Moderate and statistically significant agreement was observed between MRI and both TruCore (ICC: 0.52, p = 0.002; K: 0.42, p = 0.007) and wedge-resection (ICC: 0.53, p = 0.001; K: 0.29, p = 0.044) biopsies for macrovesicular steatosis. The MRI-derived values were systematically lower than histological estimates for macrovesicular steatosis (mean MRI: 23.4% vs. histology: 36.7-37.1%). No significant agreement was identified for microvesicular steatosis with either biopsy technique. Conclusions: In morbidly obese patients, MRI demonstrates only moderate agreement with liver histology for macrovesicular steatosis and is unreliable for microvesicular steatosis. The systematic underestimation of macrovesicular steatosis by MRI warrants caution when this modality is used as a standalone decision-making tool in this population. Further studies in larger and more heterogeneous cohorts are needed to better define the performance boundaries of MRI-derived fat-fraction measurement across the spectrum of obesity and metabolic liver disease.