Panagiotopoulos Nikolaos, Wolfson Tanya, Harris David T, Batakis Danielle, Agni Rashmi, Ceriani Lael, Covarrubias Yesenia, Hamilton Gavin, Middleton Michael S, Martins Vitor F, Gamst Anthony C, Oechtering Thekla H, Sappenfield Ryan, Horgan Santiago, Grunvald Eduardo, Funk Luke M, Jacobsen Garth R, Lidor Anne O, Goodman James A, Khoury Sami B, Sirlin Claude B, Reeder Scott B
Department of Radiology, University of Wisconsin-Madison, Madison, Wisconsin, USA.
Computational and Applied Statistical Laboratory (CASL), San Diego Supercomputer Center, University of California San Diego, San Diego, California, USA.
Hepatology. 2025 Mar 25. doi: 10.1097/HEP.0000000000001318.
Prior work has shown that MRI-derived proton density fat fraction (PDFF) can diagnose metabolic dysfunction-associated steatotic liver disease (MASLD) noninvasively, but there is a paucity of data on the performance of PDFF to classify more advanced forms of the MASLD spectrum. The purpose of this study was to assess the diagnostic performance of PDFF for the diagnoses of MASLD, metabolic dysfunction-associated steatohepatitis (MASH), and fibrotic MASH in adults with obesity undergoing bariatric surgery, using contemporaneous intraoperative liver biopsy as a reference.
PDFF was evaluated alone and with other potential classifiers (imaging, serum and anthropometric), using Bayesian Information Criterion-based stepwise logistic regression models. Areas under the receiver operating characteristic (ROC) curves (AUC) were computed for all models and single classifiers. Cross-validated sensitivity and specificity were calculated at Youden-based PDFF classification thresholds. Data analysis from 140 patients demonstrated that PDFF was the most accurate single classifier, with high AUC for MASLD (0.95), MASH (0.85), and fibrotic MASH (0.82) (all p <0.001). Multivariable models, including PDFF, outperformed those without PDFF. The Youden-based threshold for PDFF was 4.4% for MASLD (sensitivity: 87%, specificity: 86%), 6.9% for MASH (sensitivity: 77%, specificity: 66%), and 13.5% for fibrotic MASH (sensitivity: 67%, specificity: 85%).
PDFF was the most accurate single classifier for diagnosing MASLD, MASH, and fibrotic MASH. The most accurate multivariable classification models for MASLD, MASH, and fibrotic MASH included PDFF, demonstrating the central importance of PDFF for noninvasive assessment of the MASLD spectrum.
先前的研究表明，磁共振成像衍生的质子密度脂肪分数（PDFF）可用于无创诊断代谢功能障碍相关脂肪性肝病（MASLD），但关于PDFF对MASLD谱系中更晚期形式进行分类的性能数据较少。本研究的目的是，以同期术中肝活检为参考，评估PDFF在接受减肥手术的肥胖成人中诊断MASLD、代谢功能障碍相关脂肪性肝炎（MASH）和纤维化MASH的诊断性能。
使用基于贝叶斯信息准则的逐步逻辑回归模型，单独评估PDFF以及与其他潜在分类器（影像学、血清学和人体测量学）联合评估。计算所有模型和单个分类器的受试者操作特征（ROC）曲线下面积（AUC）。在基于约登指数的PDFF分类阈值处计算交叉验证的敏感性和特异性。对140例患者的数据分析表明，PDFF是最准确的单个分类器，对MASLD（0.95）、MASH（0.85）和纤维化MASH（0.82）具有较高的AUC（均p<0.001）。包括PDFF的多变量模型优于不包括PDFF的模型。基于约登指数的PDFF诊断MASLD的阈值为4.4%（敏感性：87%，特异性：86%），诊断MASH的阈值为6.9%（敏感性：77%，特异性：66%），诊断纤维化MASH的阈值为13.5%（敏感性：67%，特异性：85%）。
PDFF是诊断MASLD、MASH和纤维化MASH最准确的单个分类器。诊断MASLD、MASH和纤维化MASH最准确的多变量分类模型均包括PDFF，这表明PDFF在无创评估MASLD谱系中至关重要。
