Abstract: SA-OR085
Performance of eGFR-Based Surrogate End Points by Statistical Simulation: A Report from an NKF-FDA-EMA Workshop
Session Information
- New Considerations for Renoprotection Clinical Trials
October 27, 2018 | Location: 1B, San Diego Convention Center
Abstract Time: 06:18 PM - 06:30 PM
Category: CKD (Non-Dialysis)
- 1902 CKD (Non-Dialysis): Clinical, Outcomes, and Trials
Authors
- Greene, Tom, University of Utah, Salt Lake City, Utah, United States
- Ying, Jian, Univrsity of Utah, Salt Lake City, Utah, United States
Group or Team Name
- CKD-Epidemiology Collaboration
Background
Analyses at the individual and trial level support the validity of 30% or 40% eGFR declines and eGFR slope as surrogate endpoints for randomized clinical trials (RCTs) of CKD. We apply statistical simulation to determine conditions in which eGFR-based surrogate endpoints a) increase statistical power, allowing shorter follow-up or reduced sample size compared to the clinical endpoint of doubling of SCR or kidney failure (CEP), and b) preserve low risk of false positive conclusions if the treatment does not affect the CEP (Type 1 error).
Methods
Using a dataset of 59,074 patients from 47 RCTs, we determined input parameters for simulations of eGFR trajectories and the relationships of these trajectories with kidney failure and death. 1,000 independent simulations were performed for > 1,200 scenarios for eGFR trajectories and specific study designs. We compared the total sample size (N) required for 90% power and evaluated Type 1 error for mean eGFR slope from baseline (total slope) and from 3 months follow-up (chronic) and for composite time-to-events endpoints based on 30%, 40% declines or kidney failure vs. the CEP.
Results
The table displays the required N for each endpoint for 9 scenarios defined by the acute effect of the treatment on eGFR and the type of long-term treatment effect (uniform across fast and slow progressors), proportional (greater for fast progressors), or intermediate. Other simulations show that in the absence of a positive acute effect, Type 1 error favoring the treatment is preserved for each method except chronic slope. Type 1 error for chronic slope may be inflated if a negative acute effect attenuates as eGFR declines.
Conclusion
eGFR-based surrogate endpoints can substantially reduce the required N vs. the clinical endpoint, particularly when baseline eGFR is high and there is no acute effect. The optimum eGFR-based endpoint depends on the rate of eGFR decline, type of treatment effect and study design.
Funding
- Private Foundation Support