Abstract: TH-OR61
Quantifying Individual-Level Uncertainty in GFR Estimation
Session Information
- Preventing Progression and Reassessing Race in GFR Estimation
November 04, 2021 | Location: Simulive, Virtual Only
Abstract Time: 04:30 PM - 06:00 PM
Category: CKD (Non-Dialysis)
- 2101 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Zhu, Xiaoqian, The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Lirette, Seth, The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Rule, Andrew D., Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Mosley, Tom, The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Butler, Kenneth R., The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Butler, Javed, The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Hall, Michael, The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Vaitla, Pradeep, The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Wynn, James J., The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Dossabhoy, Neville R., The University of Mississippi Medical Center, Jackson, Mississippi, United States
- Guallar, Eliseo, Johns Hopkins University Bloomberg School of Public Health, Baltimore, Maryland, United States
- Shafi, Tariq, The University of Mississippi Medical Center, Jackson, Mississippi, United States
Background
Although the differences between estimated GFR (eGFR) and measured GFR (mGFR) are well-recognized, the magnitude and potential clinical implications of these differences at the individual level are not fully appreciated.
Methods
Using data from four US community-based cohorts with mGFR (total N=3,223), we calculated eGFR from serum creatinine alone (eGFRCR) and cystatin and creatinine (eGFRCYS-CR) using the CKD-EPI equations without race coefficients. Using quantile regression, we assessed eGFR's individual-level reliability by calculating a 95% prediction interval (PI), defined as the distribution of 95% of the observed mGFR values at a given eGFR. We also assessed eGFR's population-level reliability using standard metrics, including median difference (eGFR-mGFR). All GFR results are presented as ml/min/1.73m2.
Results
The participants' median age was 61 years, 32% were Black, and 55% were female. The median mGFR was 68 (IQR, 46 to 88). At the population level, the median difference between eGFRCR and mGFR was small (1.4; 95% CI: 0.9 to 1.9). In contrast, the individual-level 95% PI of the eGFRCR was large, ranging from 53 to 120 at eGFRCR 90 and from 19 to 55 at eGFRCR 30 (Figure and Table). Substantial individual misclassification was also noted using eGFRCR; 16% of individuals with eGFRCR <60 and 28% of those with eGFRCR <30 had mGFR above those thresholds. Results were similar for eGFRCYS-CR.
Conclusion
A substantial individual-level discrepancy exists between eGFR and mGFR. The eGFR PI should be included with eGFR reporting. Some clinical decisions may need to be based on mGFR rather than eGFR.
Table 1
eGFR, ml/min/1.73 m2 | 15 | 20 | 30 | 45 | 60 | 90 | 110 |
95% Confidence Interval (of median) | 19, 21 | 24, 26 | 32, 34 | 45, 46 | 58, 59 | 83, 85 | 100, 102 |
95% Precision Interval (of mGFR) | 10, 39 | 13, 44 | 19, 55 | 27, 71 | 36, 88 | 53, 120 | 64, 142 |
Funding
- Other NIH Support