Authors' Reply: Effects of Zibotentan and Dapagliflozin on Fluid Retention in Patients with CKD: Need More Evidence

Xu and Xu1 raise methodological questions about our study2 published in JASN concerning the proportional hazard assumption and interval censoring. We agree that the proportional hazard assumption should always be evaluated and confirm that in our study the proportional hazard assumption was met. Mixed censoring in the Cox proportional hazard regression model might induce bias. In the Zibotentan and Dapagliflozin for the Treatment of CKD (ZENITH-CKD) trial, participants were regularly monitored during the 12-week treatment period with study visits occurring at 1, 3, 6, 9, and 12 weeks after randomization. The Cox model assumes that any potential fluid retention event occurred at the end of a study interval, but we cannot exclude the possibility that the fluid retention actually occurred between visits, which may introduce bias. As suggested, we reanalyzed the time-to-event data using the accelerated failure (AFT) model using the cs2ic() and ic_par() R statistical software functions to accommodate interval-censored data. We obtained hazard ratios based on the shape parameter of the Weibull distribution within the AFT model. The AFT model resulted in similar estimates for the hazard ratio for fluid retention compared with the original Cox proportional hazard regression analysis. For example, the hazard ratio of fluid retention for the zibotentan/dapagliflozin 0.25/10 mg group versus dapagliflozin 10 mg group was 1.12 using the AFT model compared with 1.21 for the original model. We thus conclude that mixed interval censoring did not bias our results.