Senior Quantitative Medicine Developer Critical Path Institute, United States
Disclosure(s):
Zihan Cui: No financial relationships to disclose
Objectives: Autosomal dominant polycystic kidney disease (ADPKD) is part of a spectrum of inherited cystic diseases and is the most common monogenic disorder leading to kidney failure. The clinical course of ADPKD is characterized by a long period of stable glomerular filtration rate (GFR) despite continuous expansion of total kidney volume (TKV). TKV expansion from cyst growth eventually drives GFR decline. Previous efforts by the Critical Path Institute (C-Path) Polycystic Kidney Disease Outcomes Consortium (PKDOC) led to the qualification of TKV as a prognostic enrichment biomarker. This project’s objective is to develop a disease progression model that describes the trajectories of estimated GFR (eGFR) and TKV over time in ADPKD patients, explores the relationship between eGFR and TKV, and examines these endpoints' interactions with other covariates, thereby facilitating modelling of endpoints by sponsors as part of clinical/drug development program.
Methods: Datasets were selected based on their inclusion in the TKV qualification effort, availability of endpoint and covariate data, and relevance to future trials. These datasets included 1,856 subjects in non-active arms from four observational studies and two clinical trials. A disease progression model was developed using Monolix 2023R1, with TKV observations log transformed (logTKV) incorporated as a predictor of eGFR. The following baseline covariates were tested for their effects: age, Mayo class, sex, race, height, proteinuria, hypertension, and anti-hypertensive medication. Models were evaluated by goodness-of-fit criteria, including the Akaike Information Criterion (AIC), diagnostic plots, and visual predictive checks (VPCs). The model was trained and validated using an 80/20 data split.
Results: For the base model, logTKV was best characterized by a linear relationship with baseline logTKV, whereas eGFR was best characterized by a linear relationship with both baseline eGFR and predicted logTKV. An additive residual error model was used for logTKV and a combined residual error model was applied for eGFR. Baseline age and baseline Mayo class were identified as significant covariates. Both had an impact on the eGFR progression slope, while baseline Mayo class also influenced baseline eGFR. VPCs of the training dataset currently predicts both endpoints well at the upper range, yet overestimate eGFR at initial time points of the lower range. Additional covariates and model structures are still being explored at time of writing and model fit is expected to improve as more covariates and structures are considered.
Conclusions: The disease progression model described here effectively characterized observed baseline TKV, eGFR and longitudinal eGFR data. This modeling framework can inform clinical trial design for evaluating potential treatments. Additionally, the model will serve as the basis for an interactive, publicly available clinical trial simulation platform, enabling sponsors to simulate ADPKD progression based on user-defined patient characteristics at study entry.
Citations: [1] Grantham, J. J. et al. Determinants of renal volume in autosomal-dominant polycystic kidney disease. Kidney Int. 73, 108-116 (2008). [2] Monolix 2023R1, Simulations Plus, doi: 10.5281/zenodo.11401936.
