Evaluating Time-Varying and Heterogeneous Efficacy of COVID-19 Antiviral Drugs

The efficacy of many COVID-19 antiviral drugs has been shown to vary across different demographic and clinical subpopulations. It is therefore of great interest to understand this heterogeneous drug efficacy across broader subpopulations defined by various covariates. Motivated by a large-scale study of COVID-19 antiviral drugs in Hong Kong, we develop a novel method for evaluating time-varying, covariate-specific drug efficacy against severe COVID-19 outcomes. We formulate the effects of drug administration and other covariates through a varying coefficient Cox model, and combine B-splines with an efficient global kernel method for estimation. The performance of the proposed approach is evaluated through extensive simulation studies. In our real data application, we find that drug efficacy is strongly associated with the gap time from COVID-19 infection to drug administration, which highlights the importance of timely intervention.