S. Ramesh1, D. Yoon1, M. Rogge1, K. Kidd2, A. Williams2, J. Roignot3, K. Blaskeslee3, A. Bleyer2, S. Kim4; 1University of Florida, 2Wake Forest University School of Medicine, 3Sail Bio, Inc, 4Center for Pharmacometrics and Systems Pharmacology, College of Pharmacy, University of Florida.
Graduate Student University of Florida Orlando, Florida, United States
Background: To optimize the design of clinical trials for Autosomal Dominant Tubulointerstitial Kidney Disease (ADTKD), quantifying the rate and severity of the disease progression is critical. This study developed joint models quantifying progression for two ADTKD genetic variants, Uromodulin (UMOD) and Mucin1 (MUC1), using longitudinal values of estimated Glomerular Filtration Rate (eGFR), age of End Stage Renal Disease (ESRD), and baseline patient characteristics. Methods: The models were developed by leveraging longitudinal individual patient-level data from a natural history study (n=371 for UMOD, n=232 for MUC1, age range>18 years), randomly divided into training and test datasets (4:1) while maintaining the covariate distributions. The longitudinal trajectories of eGFR were initially quantified using the nonlinear mixed effects (NLME) modeling approach, with consideration of how the patient baseline features alter eGFR trajectories. After finding the hazard function that best describes timing of ESRD through the Time-to-Event (TTE) modeling, the NLME models were combined with the TTE models. The final joint models were validated using the test dataset. Explored baseline patient characteristics included age of gout onset, sex, common mutations (including cysteine UMOD mutations), and UMOD and MUC1 promoters. Monolix (version 2021R2) and R were used. Results: For both UMOD and MUC1 variants, sigmoid Imax and Weibull hazard functions best described the longitudinal decrease of eGFR and age of ESRD, respectively. Correlation was observed between random effects of γ (Hill-coefficient) and S0 (extrapolated eGFR at age=~18). Significant covariates for DPT50 (age at which eGFR is half of maximum decrease) and γ included baseline eGFR and baseline age for both variants. The age of gout onset was a significant covariate for UMOD. The eGFR decline in ADTKD-MUC1 patients was approximately 2x faster than decline in ADTKD-UMOD patients. The predicted time-varying eGFR significantly affected Te (Weibull scale parameter), and the connections between the models were established by quantifying the influence of time-varying eGFR in Te. Conclusion: The joint models accurately predict age of ESRD onset and the changes in eGFR trajectories using baseline individual characteristics, allowing subgroup analysis.
