验证四种基于肌酐的肾小球滤过率估算方程的性能

Comparative Performance of Four Creatinine-based GFR Estimating Equations

【目的】评估四个基于肌酐(Cr)的估算肾小球滤过率(eGFR)方程的预测性能:2012慢性肾脏病流行工作组(CKD-EPIcr)方程、2021CKD-EPIcr方程、湘雅方程和欧洲肾功能协会(EKFC)方程。【方法】本研究纳入198名来自中山大学附属第三医院和澳门镜湖医院的的慢性肾脏病患者。所有参与者通过碘海醇血浆清除率测定肾小球滤过率(mGFR)。eGFR方程计算的结果和mGFR进行对比。研究通过Bland-Altman图、一致性相关系数(CCC)、覆盖概率(CP)和总偏差指数(TDI)描述mGFR和eGFR间的一致性。通过偏差、精确度、均方根误差(RSME)和测量GFR30%偏差范围内的估计百分比(P_(30))评估eGFR方程的估算能力。通过2000次自助抽样的方法获得偏差、IQR、RSME以及P_(30)的95%置信区间(CI)。初步选择最优的eGFR方程后,其他方程和该方程进行统计学检验。统计检验方法如下:①偏差通过Wilcoxon符号秩检验;②P_(30)通过McNemar-Bowker检验;③RMSE和IQR在自助抽样获得2000个样本后,通过独立样本t检验比较组间差异。【结果】总体人群中mGFR和四个eGFR方程(EKFC、2012CKD-EPIcr、2021CKD-EPIcr和湘雅方程)中位数分别是56.2 mL·min^(-1)·(1.73m^(2))^(-1)、67.1 mL·min^(-1)·(1.73m^(2))^(-1)、73.0 mL·min^(-1)·(1.73m^(2))^(-1)、66.9 mL·min^(-1)·(1.73m^(2))^(-1)和63.8 mL·min^(-1)·(1.73m^(2))^(-1)。总体人群中的一致性比较中,Bland-Alt⁃man图表明,EKFC方程的最低平均差和最窄的95%一致性区间。EKFC方程有最优的CCC、TDI和CP,值分别为0.90、24.41和0.50。总体上,EKFC的偏差、精确度、P_(30)和RSME分别为-0.99、14.64、0.80和14.68,95%CI的范围分别为(-2.53,0.94)、(11.82,17.35)、(0.73,0.85)和(12.69,17.35),全面优于其他三个eGFR方程,差异有统计学意义,所有P值均<0.05。mGFR亚组中的结果基本和总体情况一致。【结论】EKFC方程的预测性能全面超越2012CKD-EPIcr方程、2021CKD-EPIcr方程和湘雅方程,且该方程的P_(30)>75%,可以满足临床的诊疗需求。所以推荐EKFC方程用于中国人群的GFR估算,但需要纳入更多的参与者去进一步证明这个结论。

【Objective】To assess the predictive performance of four creatinine-based equations for estimated glomerular filtration rate(eGFR):2012 chronic kidney disease epidemiology collaboration(CKD-EPIcr)equation,2021CKD-EPIcr equation,Xiangya equation and European kidney function consortium(EKFC)equation.【Methods】A total of 198 patients with chronic kidney disease from the Third Affiliated Hospital of Sun Yat-sen University and the Kiang Wu Hospital in Macao were enrolled.We compared the GFR measured(mGFR)by iohexol plasma clearance and the eGFR calculat⁃ed by four equations.The agreement between mGFR and eGFR was analyzed by Bland-Altman plots,concordance correlation coefficient(CCC),coverage probability(CP)and total deviation index(TDI).The performance of eGFR equations,including their bias,precision,root square mean error(RSME),and percentage of estimates within 30%deviation of mea⁃sured GFR(P_(30)),were evaluated.Bootstrap method(2000 samples)was used to calculate bias,interquartile range(IQR),RSME,and 95%confidence intervals(CI)for P_(30).After selecting the optimal eGFR equation as the reference,we statisticlly tested other equations by①Wilcoxon signed-rank test for bias;②McNemar-Bowker test for P_(30);③comparing RMSE and IQR with independent samples t test after 2000 bootstrap samples were obtained.【Results】The median mGFR and four eGFR equations(EKFC,2012CKD-EPIcr,2021CKD-EPIcr and Xiangya equation)in the overall popu⁃lation were 56.2 mL·min^(-1)·(1.73m^(2))^(-1),67.1 mL·min^(-1)·(1.73m^(2))^(-1),73.0 mL·min^(-1)·(1.73m^(2))^(-1),66.9 mL·min^(-1)·(1.73m^(2))^(-1)and 63.8 mL·min^(-1)·(1.73m^(2))^(-1),respectively.The Bland-Altman plots showed that EKFC equation had the lowest mean difference and the narrowest 95%limit of agreement.The EKFC equation had the optimal performance on CCC,TDI and CP with values of 0.90,24.41 and 0.50,respectively.Overall,the bias,accuracy,P_(30) and RSME from the EKFC equation was-0.99,14.64,0.80,and 14.68,respectively,with 95%CI ranging from-2.53 to 0.94,11.82 to 17.35,0.73 to 0.85,and 12.69 to 17.35,respectively,which were superior to those values from other three eGFR equations.The differences were statistically significant(all P<0.05).The results in the mGFR subgroups were basically consistent with the overall trend.【Conclusions】Of the four eGFR equations validated in this study,the EKFC equation comprehensively surpasses 2012CKD-EPIcr equation,2021CKD-EPIcr equation,and Xiangya equation.With P_(30)>75%,the EKFC equation can meet clinical diagnostic needs.Therefore,the EKFC equation is recommended for estimating GFR in a Chinese population,but more participants need be included to further support this conclusion.