|
Abstract:
|
Proportional hazard assumption is usually made in the design and analysis of time to event endpoints. The log-rank test possesses some desirable properties only under this assumption. In many clinical studies a delayed effect is expected due to the nature of the study design or due to the action mechanism of the treatment. If delayed effect is not properly considered in the design of such trials then it can severely impact the power, duration and study outcome. In the presence of known delay time, the power of the log rank test, impact on sample size and use of weighted log rank tests has been reported. But, in order to use weighted log rank tests the weights have to be pre-specified based upon the delay time assumed at the design stage. If the observed delay at the analysis stage is different then the use of weighted log rank test with pre-specified weights can adversely affect the power of the study. We present the design considerations applicable to the studies with delayed effect. In addition, we investigate through simulations, the impact of the delay time misspecification at the design stage when the log rank or the weighted log rank test with pre-specified weights are used.
|
