24h尿蛋白定量预测方程的诊断效能比较

Establishment of equations for predicting 24-hour urine protein excretion and testing of their performance

目的通过分析尿蛋白/尿肌酐(Upro/Ucr)、尿白蛋白/尿肌酐(Ualb/Ucr)、尿蛋白/尿渗透压(Upro/Uosm)及尿白蛋白/尿渗透压(Ualb/Uosm)与24h尿蛋白定量(24-UP)的相关性,建立并验证尿蛋白排泄量预测方程。方法将参与本研究的530例患者随机分成2组:开发组300例,验证组230例。利用简单线性回归的方法建立尿蛋白预测方程,通过Pearson相关性分析、计算相对差值和绝对差值,绘制ROC曲线综合评价方程预测的准确性。结果建立的预测方程分别为E1-UP(mg/24h)=86.1×[Upro/Ucr(mg/mmol)]0.595;E2-UP(mg/24h)=53.5×[Ualb/Ucr(mg/mmol)]0.704;E3-UP(mg/24h)=21.3×[Upro/Uosm(mg/osm)]0.598;E4-UP(mg/24h)=10.9×[Ualb/Uosm(mg/osm)]0.698。Ualb/Ucr预测尿蛋白排泄量的相关性和准确性优于其他3种检测方法。绘制ROC曲线未发现4种方法的诊断效能存在显著差异。在CKD1～4期时,方程预测的蛋白排泄量与24-UP存在极好的相关性,但在CKD 5期时,所有方程预测的准确性均显著下降。结论虽然Ualb/Ucr预测的尿蛋白排泄量与24-UP相关性和准确性略优于其他3种方法,但4种尿蛋白定量替代方法在预测的综合效能上差异并无统计学意义。

Objective To establish equations for predicting the 24-hour urine protein（24-UP）excretion by analyzing the correlation of four parameters including urine protein/creatinine ratio（Upro/Ucr）,urine albumin/creatinine ratio（Ualb/Ucr）,urine protein/osmolality ratio（Upro/Uosm）and albumin/osmolality ratio（Ualb/Uosm）with 24-UP.Methods Totally 530patients were randomly divided into two groups,with 300for establishing equations and the other 230for testing the performance of the equations.The equations were set up by linear regression model.The performance of the equations was assessed by Pearson correlation,BlandAltman plot,calculating the difference and ROC analysis.Results The equations were as follows：E1-UP（mg/24h）=86.1（Upro/ Ucr[mg/mmol]）0.595,E2-UP（mg/24h）=53.5（Ualb/Ucr [mg/mmol]）0.704,E3-UP（mg/24h）=21.3（Upro/Uosm [mg/osm]）0.598,and E4-UP（mg/24h）=10.9（Ualb/Uosm [mg/osm]）0.698.Equation based on Ualb/Ucr achieved a better dependence and accuracy compared with the other 3equations.However,ROC analysis showed no significant difference in diagnostic performance among the 4equations.We also noticed that excellent correlation was found between the predicted urine protein excretion and 24-hour protein excretion when the patients were at CKD 1to 4 stages.When patients were at CKD 5stage,the accuracy and dependence of the equations fell sharply.Conclusion Though the correlation and accuracy of Ualb/Ucr are better than the other 3 methods,the overall performance is not significantly different among the 4 methods