Department of Internal Medicine, Seoul National University Hospital, 101 Daehak-Ro, Jongno-Gu, Seoul, 03080, South Korea.
Department of Internal Medicine, Gyeongsang National University College of Medicine, Jinju, South Korea.
J Nephrol. 2024 Apr;37(3):681-693. doi: 10.1007/s40620-023-01883-7. Epub 2024 Feb 12.
The adoption of the 2021 CKD-EPIcr equation for glomerular filtration rate (GFR) estimation provided a race-free eGFR calculation. However, the discriminative performance for AKI risk has been rarely validated. We aimed to evaluate the differences in acute kidney injury (AKI) prediction or reclassification power according to the three eGFR equations.
We performed a retrospective observational study within a tertiary hospital from 2011 to 2021. Acute kidney injury was defined according to KDIGO serum creatinine criteria. Glomerular filtration rate estimates were calculated by three GFR estimating equations: 2009 and 2021 CKD-EPIcr, and EKFC. In three equations, AKI prediction performance was evaluated with area under receiver operator curves (AUROC) and reclassification power was evaluated with net reclassification improvement analysis.
A total of 187,139 individuals, including 27,447 (14.7%) AKI and 159,692 (85.3%) controls, were enrolled. In the multivariable regression prediction model, the 2009 CKD-EPIcr model (continuous eGFR model 2, 0.7583 [0.755-0.7617]) showed superior performance in AKI prediction to the 2021 CKD-EPIcr (0.7564 [0.7531-0.7597], < 0.001) or EKFC model in AUROC (0.7577 [0.7543-0.761], < 0.001). Moreover, in reclassification of AKI, the 2021 CKD-EPIcr and EKFC models showed a worse classification performance than the 2009 CKD-EPIcr model. (- 7.24 [- 8.21-- 6.21], - 2.38 [- 2.72-- 1.97]).
Regarding AKI risk stratification, the 2009 CKD-EPIcr equation showed better discriminative performance compared to the 2021 CKD-EPIcr equation in the study population.
2021 版 CKD-EPIcr 方程的采用为肾小球滤过率(GFR)的估算提供了一个无种族差异的 eGFR 计算方法。然而,其对急性肾损伤(AKI)风险的判别性能尚未得到充分验证。本研究旨在评估三种 GFR 估算方程在 AKI 风险预测或重新分类能力方面的差异。
本研究为回顾性观察性研究,于 2011 年至 2021 年在一家三级医院进行。根据 KDIGO 血清肌酐标准定义急性肾损伤。肾小球滤过率估计值通过三种 GFR 估算方程计算:2009 年和 2021 年 CKD-EPIcr 方程和 EKFC 方程。在三种方程中,通过接受者操作特征曲线下面积(AUROC)评估 AKI 预测性能,并通过净重新分类改善分析评估重新分类能力。
共纳入 187139 人,其中 27447 人(14.7%)发生 AKI,159692 人(85.3%)为对照。在多变量回归预测模型中,2009 年 CKD-EPIcr 模型(连续 eGFR 模型 2,0.7583[0.755-0.7617])在 AKI 预测方面优于 2021 年 CKD-EPIcr 模型(0.7564[0.7531-0.7597],<0.001)或 EKFC 模型(0.7577[0.7543-0.761],<0.001)的 AUROC。此外,在 AKI 的重新分类中,2021 年 CKD-EPIcr 和 EKFC 模型的分类性能均劣于 2009 年 CKD-EPIcr 模型。(-7.24[-8.21--6.21],-2.38[-2.72--1.97])。
在该研究人群中,与 2021 年 CKD-EPIcr 方程相比,2009 年 CKD-EPIcr 方程在 AKI 风险分层方面具有更好的判别性能。