Survival Mediation Model: Applications to Medicine and Genomics

Abstract

Mediation analyses have become a popular approach for studying the effect of a risk factor on an outcome through a mediator. In the first part, I will illustrate the utility of mediation analyses with a study regarding hepatitis B and C viruses in relation to hepatocellular carcinoma (HCC). Hepatitis B and C viruses are well-established risk factors for HCC. Both statistical interaction and mediation analyses were applied to study their coordinated etiologic mechanism (Huang et al., J Clin Oncol 2011; Huang et al., Epidemiology 2016). Limitations of conventional statistical interaction will be discussed and used to demonstrate the advantage of mediation analyses. Despite its advantage, currently literature on mediation analyses with survival outcomes largely focused on settings with a single mediator and only quantified the mediation effects in certain scales (e.g., hazard ratio). In the second part of the talk, I will introduce a multi-mediator model for survival data that employs a semiparametric probit model (Huang and Cai Biometrics 2016). We characterize path-specific effects (PSEs) and derive closed form expressions for PSEs on the survival probabilities. Statistical inference on the PSEs is developed using a nonparametric maximum likelihood estimator and the functional Delta method. I will illustrate the utility of our method in an integrative genomic study of glioblastoma multiforme survival.