基于血清半胱氨酸蛋白酶抑制剂C的肾小球滤过率预测模型的建立及评价

目的以慢性肾脏病(chronic kidney disease,CKD)患者为研究对象,建立基于血清半胱氨酸蛋白酶抑制剂C(serum cystatin C,s-cystatin C)的肾小球滤过率(glomerular filtration rate,GFR)预测模型并进行适用性评价。方法收集本院242例CKD患者的有关资料并分为模型建立组和模型验证组,以99mTc-DTPA肾显像快速测定所得的清除率作为GFR检测的参考值(residual GFR,rGFR),同时测定s-cystatin C水平,探讨其与rGFR的关系;在模型建立组利用多重线性回归模型建立和评价的原理方法,以s-cystatin C建立适宜于CKD患者的rGFR预测模型,并在模型验证组与同源的Hoek模型和Orebro模型进行适用性比较和评价。结果将s-cystatin C进行标准倒数变换后,其与rGFR回归后的相关系数为0.773。经多重线性回归分析二次拟合后方程复相关系数、决定系数、校正的决定系数和剩余标准差分别为0.863、0.745、0.742、0.207。经残差的P-P概率图分析残差符合正态分布,残差具方差齐性。预计GFR(estimated GFR,eGFR)=67/s-cystatin C+3。该模型所得eGFR与模型验证组rGFR分布无差异,30%和50%准确性不低于Hoek模型和Orebro模型,精密度高;经Bland-Altman图和ROC曲线分析,适用性较好。结论该模型与同类研究所获GFR预测模型相比,体现了较好的预测能力,可试用于临床CKD患者GFR水平的预测。

Estimating glomerular filtration rate based on serum cystatin C

Objective To develop an estimating formula for glomerular filtration Rate(GFR) based on serum cystatin C in patients with chronic kidney disease(CKD).Methods Clinical characteristics of 242 CKD patients were collected.The patients were randomly divided into modeling group and model validation group.The rGFR obtained from 99mTc-DTPA clearance rate was used as a reference value of GFR.s-cystatin C was detected by latex enhanced immunoturbidimetric method.Preliminary linear regression analysis followed by multiple linear regression were performed to investigate the association between s-cystatin C and rGFR.The validity of the estimation formula was tested in the model validation group in comparison with Hoek formula and Orebro formula.Results With standardised countdown conversion,s-cystatin showed linear correlation with rGFR,with a correlation coefficient of 0.773.The multiple correlation coefficient,determination coefficient,adjusted R square and std.error of the estimation model were 0.863,0.745,0.742,and 0.207,respectively.The residuals P-P probability plot analysis showed that the model residuals fitted into normal distribution with homogeneity of variance.The formula was: eGFR = 67/s-cystatin C +3.No significant difference was found between the distribution of eGFR and rGFR.Our formula had an accuracy of 30% and 50%,which were no less than those obtained from Hoek formula and Orebro formula.The new formula also had acceptable bias and high precision.The Bland-Altman analysis and ROC curve analysis showed good applicability of the new formula.Conclusion The GFR prediction formula we established has a good prediction performance as comparised with other formulae,which could be used in measuring GFR in CKD patients.