New methods for clinical trials with complex outcomes using surrogate and prioritized composite endpoints
Embargo Date
2023-10-05
OA Version
Citation
Abstract
In clinical trials, surrogate and composite endpoints can be used to increase the efficiency of the trial. The first section of this dissertation focuses on the validation of surrogate endpoints where current methodology encounters challenges in the presence of collinearity between the treatment and surrogate endpoint. The proposed path likelihood reduction factor, LRF_P, adapts current methodology into the path analysis framework to quantify the validity of a surrogate endpoint while providing an improved interpretation of treatment effect by estimating both direct and indirect treatment effects. Simulations show LRF_P is less biased and more robust than current methodology in presence of collinearity. LRF_P can be extended to multiple correlated surrogate endpoints while maintaining similar benefits.
Commonly used in cardiovascular trials, composite endpoints combine information across multiple outcomes to present a comprehensive and efficient assessment of treatment effect. In section 2 of this dissertation, we analyze the case of terminal and non-terminal events. Conventional composite endpoint methodology assumes equal importance of each event, however, typically the non-terminal event will occur first. Consequently, the analysis will be driven by the non-terminal rather than terminal outcome. Prioritized endpoints avoid these issues by creating a hierarchy of outcomes. Utilizing prioritized endpoints we propose the adjusted win ratio (AWR) to assess treatment effect while adjusting for covariate imbalance. Simulations show AWR provides a less biased and more precise treatment effect estimate compared to current methodology even in the presence of a large covariate imbalance.
In section 3, we examine a case typically seen in studies of chronic disease with a composite endpoint comprised of terminal and non-terminal recurrent events. Here, composite endpoints provide an insightful assessment of treatment effect on both death and disease burden. The proposed generalized adjusted win ratio (GAWR) utilizes prioritized endpoints to assess treatment effect while mitigating bias through adjustment for prognostic covariates. Simulations show GAWR provides the least biased and most precise treatment effect estimate in the presence of a large covariate imbalance compared to current methodology without sacrificing power. Further, the proposed methods are generalizable to any type of secondary outcome.