A1551
Title: Survival analysis with time-varying covariate: A R shiny tool for visualization
Authors: Yimei Li - University of Pennsylvania (United States) [presenting]
Abstract: Survival analysis in clinical research often needs to evaluate the association between time-varying covariates (TVCs) and time-to-event endpoints. Existing methods to analyze TVCs include naive approach (Kaplan-Meier [KM] method and Cox model), extended KM method, landmark analysis, and time-dependent (TD) Cox model. However, pros and cons of various methods are underappreciated and visualization of survival curves with TVCs is not always straightforward. An innovative R Shiny tool, TVCurve, is developed to provide a comprehensive suite for survival analysis with TVCs. The tool incorporates the above methods and generates corresponding survival curves. It also includes a parametric model assuming Weibull distributions. The tool builds in a Simulation panel to allow thorough evaluation of bias in the estimates under different methods. The use of the tool is demonstrated via applications to several clinical trials and extensive simulations are conducted. A summary table of the observed patterns in Simulation results is also provided, as a quick reference for understanding the direction of bias and implication for study conclusions in clinical applications. The findings underscore the importance of employing appropriate statistical methods for analyzing TVC in survival analysis. The easy-to-use R Shiny tool will promote adoption of the correct method and enhance the accuracy and reproducibility of survival analysis in clinical research.
