METHODS: Japanese individuals who received annual health checkups including abdominal ultrasonography (n = 15 788, men/women: 10 250/5538, mean age: 49 years) were recruited.
RESULTS: The numbers of individuals with SLD, MASLD, MetALD, ALD, NAFLD, and MAFLD were 5603 (35.5%), 4227 (26.8%), 795 (5.0%), 324 (2.1%), 3982 (25.8%), and 4946 (31.3%), respectively. Clustering analyses using t-distributed stochastic neighbor embedding and K-means to visually represent interconnections in SLDs uncovered five cluster formations. MASLD and NAFLD mainly shared three clusters including (i) low alcohol intake with relatively low-grade obesity; (ii) obesity with dyslipidemia; and (iii) dysfunction of glucose metabolism. Both MetALD and ALD displayed one distinct cluster intertwined with alcohol consumption. MAFLD widely shared all of the five clusters. In machine learning-based analyses using algorithms of random forest and extreme gradient boosting and receiver operating characteristic curve analyses, fatty liver index (FLI), calculated by body mass index, waist circumference, and levels of γ-glutamyl transferase and triglycerides, was selected as a useful feature for SLDs.
CONCLUSIONS: The new nomenclature of SLDs is useful for obtaining a better understanding of liver pathologies and for providing valuable insights into predictive factors and the dynamic interplay of diseases. FLI may be a noninvasive predictive marker for detection of SLDs.
方法:招募了接受包括腹部超声检查在内的年度健康检查的日本人(n=15788,男性/女性:10250/5538,平均年龄:49岁)。
结果:患有SLD的人数,MASLD,MetALD,ALD,NAFLD,MAFLD为5603(35.5%),4227(26.8%),795(5.0%),324(2.1%),3982(25.8%),和4946(31.3%),分别。使用t分布随机邻居嵌入和K均值进行聚类分析,以直观地表示SLD中的互连,揭示了五种簇形成。MASLD和NAFLD主要共有三个集群,包括(i)低酒精摄入量与相对低度肥胖;(ii)肥胖与血脂异常;和(iii)葡萄糖代谢功能障碍。MetALD和ALD都显示出一个与酒精消耗交织在一起的独特簇。MAFLD广泛共享所有五个集群。在基于机器学习的分析中,使用随机森林和极端梯度提升算法以及接收器工作特性曲线分析,脂肪肝指数(FLI),按体重指数计算,腰围,以及γ-谷氨酰转移酶和甘油三酯的水平,被选为SLD的有用功能。
结论:SLD的新命名法有助于更好地了解肝脏病理,并为预测因素和疾病的动态相互作用提供有价值的见解。FLI可能是检测SLDs的非侵入性预测标志物。