Abstract:
|
In survival data analyses, the occurrence of an outcome of interest is commonly presented with Kaplan-Meier (K-M) estimates at a certain time point t to assess probability of survival or failure (i.e., event) up to time t. For clinical studies in which the goal is to assess the effect of treatment over the referent, absolute difference or ratio of the two K-M estimates at study time t provide the magnitude of the treatment effect non-parametrically. However, inferences of these absolute difference and ratio of K-M estimates have not been sufficiently studied in literature. This presentation discusses inferences of these differences by deriving test statistics and confidence intervals (CIs) at a fixed time t using delta-method approximation. Properties of the derived statistics are studied by performing Monte-Carlo simulations using survival data generated from various survival distribution models. We illustrate the methods with survival data obtained from a clinical trial of cardiovascular disease.
|