|
Abstract:
|
Binary data can be estimated the treatment effect in a variety of ways. There is a growing concern about binary rate estimation when analyzed without taking into consideration the dropout in the long term safety trials. In this presentation we compared the following seven statistical approaches for determining a appropriate method to handle dropout during a long term follow-up: Binary rate based on ITT, Binary rate based on complete case, Event rate adjusting denominator based on adequate follow-up, Binary rate with single imputation, Binary rate with opposite imputation, Survival analysis for time to first event, and Events per person years. The relative risk estimate, bias, and statistical power were evaluated via Simulation. Simulation results suggest that time to event survival analysis and events per person years method generally are more appropriate for a long term MACE rate.
|
