Abstract:
|
In randomized trials involving time-to-endpoint, the (stratified) Cox model often has been utilized to estimate the hazard ratio (HR). If covariates or stratified factors are missing, the estimated HR would be biased because subjects with missing data are excluded from the model. In the non-large trials, the exclusion due to missing data would lead to severe influence on the result. Many researchers have proposed some methods to deal with missing data, but those often cannot be applied because of requiring model specification, complicated programming or the sufficient number of data. In this presentation, we show the alternative method which can be applied without such requirements. The main concept of this is based on the two methods that Mehrotra and others proposed to address the issues related to small sample (2001, 2011, 2012). As they suggested that one of the two method would be enhanced by incorporating with another, we think that the incorporating can address the issues due to missing stratified factors in the small-sized trials or strata. To clarify our assumption, we show the simulation results by the incorporating method in some realistic situations.
|