Discovering Heterogeneous Treatment Effects on Slope-based Endpoints in Chronic Kidney Disease Trials

Music City Center 

Background: Chronic kidney disease (CKD) is slowly progressive, with clinically-relevant end-points of interest (e.g. kidney failure/dialysis, transplantation, death) occurring many years after diagnosis, making the design of trials to evaluate treatments to slow the progression of kidney disease challenging. Recent work has demonstrated the utility of a 3-year slope in patient's estimated glomerular filtration rate (eGFR) as a high-quality surrogate marker for the clinical end-points of interest, thereby allowing for shorter clinical trials. Existing research has focused on relaxing the linear trend assumption on the eGFR slope, accounting for informative censoring (via fitting a shared parameter model, for example), and evaluating heterogeneous treatment effects (HTE) given predetermined subgroups. Yet, none have explored data-driven subgroup identification and HTE estimation.

Methods: We propose a Bayesian method that incorporated a Bayesian decision tree for HTE into a shared-parameter model that combines a survival model for censoring time with a two-slope spline model that characterizes the rate of decline in eGFR. Our proposed approach simultaneously estimates the eGFR slope in the presence of informative censoring and provides interpretable clinical decisions for subgrouping patients according to their treatment effect on eGFR slope.

Results: We apply our model to analyze the Modification of Diet in Renal Disease (MDRD) Trial, observing strong Bayesian evidence that patients with a baseline eGFR above 34.32 benefit more from the intensive systolic blood pressure control compared to patients with a baseline eGFR below 34.32, with a posterior frequency of 81\% for observing a higher treatment effect estimation in the former group.

Conclusion: Our proposed model can effectively capture even subtle HTE while avoiding over-fitting when no heterogeneity exists, making it valuable for downstream analyses such as treatment recommendations.

Heterogeneous treatment effects

Bayesian statistics