Abstract: SA-PO768
GFR Slopes – The Statistical Method Matters
Session Information
- CKD: Epidemiology, Risk Factors, Prevention - III
October 27, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 1901 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- van Rijn, Marieke, Radboud Univ Medical Centre, Nijmegen, Netherlands
- Wetzels, Jack F., Radboud University Medical Center, Nijmegen, Netherlands
- Blankestijn, Peter J., University Medical Center----, Utrecht, Netherlands
- van den Brand, Jan A.J.G., Radboud University Nijmegen Medical Center, Nijmegen, Netherlands
Background
The FDA is considering the use of eGFR decline and eGFR slope as surrogate endpoints in CKD research. Several statistical methods can be used to estimate GFR decline (Coresh 2014, Leffondre 2015, Asar 2015). We questioned if these methods differed with respect to the estimates of GFR slopes.
Methods
We used the MASTERPLAN (Multifactorial Approach and Superior Treatment Efficacy in Renal Patients with the Aid of Nurse practitioners) dataset, which included 762 patients with an eGFR between 20 to 70 ml/min, a median follow-up of 4.8 yr (IQR 4.3-5.3) and 103 ESRD events. We estimated percentage and annual change in MDRD eGFR over a period of 2 years using 3 methods. For annual change: 1) last estimated GFR – first estimated GFR, 2) linear mixed model (LMM) using the first and last measured GFR, 3) LMM using all GFR measures in 2 years’ time. For percentage change: 4) (last estimated GFR – first estimated GFR) / (first estimated GFR) * 100 and 5) LMM with log-transformed GFR using the first and last measured GFR, 6) LMM with log-transformed GFR using all GFR measures in 2 years’ time.
Results
Overall data were available of 6275 visits and 91% of patients had at least 4 visits within 2 years. Both the distribution for annual and percentage change in eGFR varied greatly between the different statistical methods (Figure 1). Using the first-last eGFR method to estimate GFR slope resulted in a 5 fold wider distribution on the annual scale and 10 fold on the relative scale compared to using a LMM.
Conclusion
The statistical method that is used to estimate change in GFR has a great impact on the estimated GFR slopes. The current debate about the use of eGFR declines and eGFR slope as surrogate endpoints in CKD should also consider the statistical method how to estimate GFR decline.
Figure1: Distribution of GFR slopes. The number in the figure corresponds to the statistical method described in the methods. The blue line is the average GFR slope.