Abstract: TH-PO1094
Improved Time to eGFR Decline Outcomes in Glomerular Disease
Session Information
- CKD: Clinical, Outcomes, Trials - I
October 25, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 1902 CKD (Non-Dialysis): Clinical, Outcomes, and Trials
Authors
- Mansfield, Sarah, Arbor Research Collaborative for Health, Ann Arbor, Michigan, United States
- Zee, Jarcy, Arbor Research Collaborative for Health, Ann Arbor, Michigan, United States
- Mariani, Laura H., University of Michigan, Ann Arbor, Michigan, United States
- Gillespie, Brenda W., University of Michigan, Ann Arbor, Michigan, United States
Background
The standard method of using two estimated glomerular filtration rate (eGFR) measures to calculate time to a percentage decline in eGFR may result in inaccurate event times due to eGFR variability and restriction of events to study visit times. We propose fitting a regression line to all observed eGFR measures to improve accuracy and power of time to percentage decline in eGFR outcomes.
Methods
Using data from the Nephrotic Syndrome Study Network (NEPTUNE) of patients with minimal change disease, focal segmental glomerulosclerosis, or membranous nephropathy, we compared time to 40% decline in eGFR using both event time calculation methods. We also used computer simulations to assess power and accuracy of event times under both methods. Parameters for the simulations, such as eGFR variability, percent missing eGFR observations, and correlation between successive eGFR measurements, imitated NEPTUNE data.
Results
380 NEPTUNE patients with a mean of 6.5 eGFR measurements over a mean of 33 months of follow-up were included in the analysis. Among these patients, 91 had a 40% decline in eGFR using the two-point method, while 68 had a 40% decline in eGFR using our proposed linear regression method. For patients who had an event under both methods (n=64), the two-point method estimated earlier event times compared to the linear regression method (Figure 1). Under simulated data conditions, the standard two-point method was less accurate in estimating event times than our proposed regression method, particularly with high eGFR variability or more missingness. The two-point method was also less powerful in detecting time-to-event differences between groups.
Conclusion
Using our proposed regression method to estimate time to a percentage decline in eGFR increases accuracy and power. Reducing noise in outcome estimation with our proposed method increases the ability to discover treatment or biomarker effects.
Kaplan-Meier estimates of the probability of a 40% decline in eGFR using NEPTUNE data.
Funding
- NIDDK Support