UC Santa Barbara
Statistical Methods for Treatment Effects with Treatment Switching
- Author(s): Chu, Fang-I
- Advisor(s): Wang, Yuedong
- et al.
Treatment switching occurs frequently in clinical trials and other medical studies. Ignoring switching usually leads to biased and inefficient estimates of the treatment effect. More research is necessary to account for possible correlation between switching time and survival time. This thesis consists of two topics, treatment switching in clinical trials in Chapters 1-4 and treatment switching in dialysis studies in Chapters 5 and 6. Our research work addresses the treatment switching scenario through three approaches: (1) include a frailty term to accelerated failure time models, along with a logistic model to accommodate prognostic status, (2) evaluate treatment effect based on the latent event time and (3) quantify the switch effect using a varying coefficient Cox model. The proposed methods are applied to investigate dynamic effect of change in dialysis vascular access in dialysis patients. The statistical analysis using above methods consistently suggests that hemodialysis (HD) patients with access change from catheter (CATH) to arteriovenous (AV) have lower mortality risk than those without access change and early switching is more beneficial.