Use of eGFR Slope Thresholds as End Point Components in a Kidney Disease Progression Hierarchical Composite End Point

Key Points It is feasible to apply eGFR thresholds in kidney progression hierarchical composite end point (HCE) to minimize variability impact on patient comparisons. Use of eGFR thresholds did not change treatment effect estimates of the kidney progression HCE in seven major CKD trials compared with no threshold. Applying eGFR thresholds had minimal effect on statistical power of the kidney progression HCE across various sample sizes. Background We developed and validated a kidney disease progression hierarchical composite end point (HCE) combining time-to-event end points with the rate of eGFR decline (eGFR slope) as a continuous end point. An alternative to this continuous end point is to apply various thresholds on an absolute and relative scale for the pairwise comparisons in the eGFR slope component. We assessed the effect of different thresholds on the treatment effects and statistical power on the kidney disease progression HCE analyzed using win odds. Methods We calculated the win odds in seven international phase 3 CKD trials and compared treatment effects for the original HCE versus HCEs with different eGFR thresholds (0.5, 0.75, or 1.0 ml/min per 1.73 m 2 per year eGFR slope difference), as well as categorical thresholds and thresholds determined by percent differences in eGFR slope. In addition, we estimated the statistical power for these thresholds using a bootstrap sampling procedure to evaluate their impact on trial efficiency. Results For the seven CKD trials, the win odds estimate remained consistent, combined with a minor reduction in statistical power regardless of which eGFR thresholds were applied. For instance, for thresholds 0 (original HCE), 0.5, 0.75, and 1.0, the win odds of the Dapagliflozin and Prevention of Adverse Outcomes in CKD trial were 1.41 (95% confidence interval [CI], 1.32 to 1.52), 1.41 (95% CI, 1.31 to 1.51), 1.40 (95% CI, 1.31 to 1.51), and 1.40 (95% CI, 1.30 to 1.50) with 97%, 92%, 92%, and 93% statistical power for a sample size of 500, respectively. Conclusions Our findings suggest that using eGFR thresholds in the kidney disease progression HCE did not alter treatment effect estimates and had only a minimal effect on statistical power compared with its continuous use.