新型尿液代谢标志物在糖尿病肾脏病诊断中的价值

Novel urinary metabolite biomarkers in diagnosis of diabetic kidney disease

目的探讨新型尿液代谢标志物在糖尿病肾脏病(DKD)诊断中的价值。方法选择2020年1至12月在南京医科大学第二附属医院就诊的健康人群或确诊为2型糖尿病(T2DM)的患者235例作为研究对象。本研究包含两个独立的队列,即发现队列(116例)和验证队列(119例)。发现队列分为健康对照组(37例)、单纯T2DM组(40例)以及T2DM合并DKD组(39例);验证队列分为单纯T2DM组(64例)和T2DM合并DKD组(55例)。收集研究对象的人口学和实验室检查等资料,应用超高效液相色谱-串联质谱方法进行尿液代谢物的全靶向定量检测。采用偏最小二乘判别分析和正交偏最小二乘判别分析进行多维建模。利用基于Boruta算法的机器学习筛选潜在标志物。尿液代谢标志物的两组间比较采用Wilcoxon秩和检验,多组间比较采用Kruskal-Wallis检验。采用二元logistic回归分析进行统计建模,受试者工作特征(ROC)曲线评价诊断效能。结果发现队列中,在晨尿标本中定量检测到160种代谢物,单纯T2DM组和T2DM合并DKD组比较后筛选出62种差异代谢物,主要富集在氨基酸代谢通路,其中17种是诊断DKD的候选标志物。在验证队列中,15种差异代谢物及7种候选标志物(异戊酸、异丁酸、亮氨酸、腺苷高半胱氨酸、丙酸、氧代己二酸和丙酰肉碱)被证实。分别用单一候选标志物或7种候选标志物的组合指标构建诊断模型,结果显示,无论在发现队列还是验证队列中,组合指标的诊断效能[发现队列ROC曲线下面积0.888(95%CI 0.814~0.963);验证队列ROC曲线下面积0.811(95%CI 0.734~0.887)]均明显高于任何单一候选标志物(发现队列、验证队列ROC曲线下面积分别为0.697~0.831、0.628~0.728)。结论单纯T2DM患者与T2DM合并DKD患者的尿液代谢物存在显著差异。从尿液代谢物中筛选出的7种候选标志物构建的组合指标可能成为诊断DKD的新途径。

Objective To explore the application of novel urinary metabolite biomarkers in diagnosis of diabetic kidney disease(DKD).Methods A total of 235 healthy individuals and type 2 diabetes mellitus(T2DM)patients admitted to the Second Affiliated Hospital of Nanjing Medical University from January to December 2020 were enrolled in this cross-sectional study.Two separated discovery cohort(116 cases)and validation cohort(119 cases)were included:the discovery cohort was divided into healthy control group(37 cases),simple T2DM group(40 cases)and T2DM combined with DKD group(39 cases),and the validation cohort was divided into simple T2DM group(64 cases)and T2DM combined with DKD group(55 cases).Demographic and laboratory data were collected,and the urinary metabolites were detected by ultraperformance liquid chromatography coupled to tandem mass spectrometry.Partial least squares discriminant analysis and orthogonal partial least squares discriminant analysis were used for multi-dimensional modeling.Machine learning based on Boruta algorithm was used to screen potential markers.Wilcoxon test was used for comparison of novel urinary metabolite biomarkers data between two groups,and Kruskal-Wallis test was used for comparison between multiple groups.Binary logistic regression analysis was used for statistical modeling,and receiver operating characteristic(ROC)curve was used to evaluate diagnostic efficacy.Results In discovery cohort,a total of 160 metabolites were detected in morning urine samples by targeted quantitative metabolomics,62 differential metabolites were screened out from the discovery cohort,mainly enriched in amino acid metabolic pathway,and 17 of which were candidate markers for diagnosis of DKD.Fifteen differential metabolites and seven candidate markers(isovaleric acid,isobutyric acid,leucine,s-adenosylhomocysteine,propionic acid,oxoadipic acid,propionylcarnitine)were verified in the validation cohort.The diagnostic model of the combined marker was constructed by integrating the 7-candidate urinary metabolic markers.The diagnostic efficacy of the combined marker in both the discovery cohort and the validation cohort(the area under ROC curve of the discovery cohort was 0.888,95%CI 0.814-0.963;the area under ROC curve of the validation cohort was 0.811,95%CI 0.734-0.887)were significantly higher than that of a single metabolic marker(the area under ROC curve of the discovery cohort was 0.697-0.831,the area under ROC curve of the validation cohort was 0.628-0.728).Conclusions There were significant differences in urine metabolites between simple T2DM patients and T2DM combined with DKD patients.The combination of various urine metabolic markers might be a novel strategy to diagnose DKD.