|
Abstract:
|
In cancer trials, it is common to study the impact of time-dependent variables determined during the study follow-up (such as transplant status or the responder status on a short-term outcome) on the long-term clinical endpoints. With the naïve Cox model or the Kaplan-Meier estimate, we face the issue of the immortal-time bias when evaluating the association between the responders or transplanted patients and long-term outcomes such as overall survival. In our work, we conducted simulation studies under various scenarios to compare the naïve Cox model, landmark analysis, and the time-dependent Cox model in terms of the reduction of the immortal-time bias. The aim of the project is to use proper statistical methods to solve practical issues in survival analysis for oncology clinical trials.
|
