Content area
Abstract
Surrogate endpoints accepted as valid by regulatory agencies provide increased potential for the use of randomized controlled trials (RCTs) in areas of slowly progressing disease. Chronic kidney disease (CKD) is one such setting where the clinical endpoint typically necessitates excessive follow-up, and where promising surrogate endpoints offer the potential to increase the use of RCTs. The prevailing surrogate validating methodology involves a meta-analysis of previously conducted RCTs, where meta-regression models are used quantify the degree of association between the treatment effects on the clinical endpoint and the treatment effects on the surrogate endpoint. There are various shortcomings to the trial-level approaches commonly employed in practice, and advances in meta-regression models are necessary to handle the complexities confronted in surrogate evaluations in the CKD space. Firstly, surrogate evaluations in CKD are typically performed over a broad collection of studies representing a range of disease sub-etiologies, which contrast a variety of interventions. The methods used to evaluate surrogates typically ignore the possibility that the quality of the surrogate can vary across subcategories of trials. We propose a hierarchical model that allows for distinct clinical-surrogate effect relationships within subgroups of trials, and also improves estimation precision over analyses where separate models are fit within subgroups. Second, the traditional trial-level approach requires specification of within-trial correlations between clinical and surrogate effects for each trial, which are frequently at least partially missing for the analysis. We show why mishandling missing within-trial correlations can lead to biased estimates of key meta-regression parameters, which can lead to flawed conclusions regarding the quality of the surrogate. We also provide novel modeling approaches to handle such missingness. Finally, we provide an analysis of surrogate enpdoints in CKD trials where we evaluate whether the surrogate quality changes as a function of the baseline health or rate of disease progression of a trial's patients. This work provides insight into the optimal contrast used to capture the treatment effect on a key surrogate endpoint in the CKD setting.
