Abstract:
|
In the studies for treatment of neurodegenerative diseases, we may use longitudinally collected data from one or more validated instruments, and define the events of interest from these instruments. For example, in order to evaluate the disability progression in patients with multiple sclerosis, we collect EDSS, time to 25-foot walk, and 9-hole peg test at each visit, and derive the composite endpoint of disability progression status based on these instruments. Binary event rates and event rates based on the Kaplan-meier estimates are the frequently used endpoints to assess the event rates at a fixed time point. In this research, we will look at the properties of various techniques based on the choice of proper estimand. In addition, we will compare the estimands in terms of estimation and statistical power, and how they are impacted by the estimation bias due to missingness or censoring from the intercurrent events. Different missing data imputation methods will be evaluated under these two estimands. Simulations will be conducted for the evaluation.
|