Abstract:
|
For the last few years, missing data in controlled clinical trials have been a forefront issue for statisticians due to the concern of biased efficacy results from regulatory agencies. Multiple methods for analyses and data imputation have been suggested to mitigate biased results. These include multiple imputation methodology, model-based methods for example MMRM, and tipping point analyses as sensitivity analyses. But how much missing data is too much ? In this talk, we will explore missing data for Phase III controlled clinical trials. Specifically, various degrees of missing data will be simulated to examine how much will negate a positive efficacy outcome and how much will create a biased false positive. Dichotomous endpoints will be considered first, time permitting, continuous endpoints will also be explored.
|