Different from the use of intention-to-treat (ITT) analysis for efficacy evaluation, many pharmaceutical companies currently use treatment emergent (TE) analysis for adverse event (AE) safety analysis. In the TE analysis, study period and AEs occurring after a pre-specified post-treatment window will not be included. One consideration for using the TE AE analysis is that including substantial off-drug period and events in the analysis may dilute the power for detecting safety signals especially if after discontinuation residual treatment effect diminishes quickly. We perform quantitative analyses to compare the unbiasedness and power of the ITT and TE AE analyses under several different settings and metrics (difference in rates and relative risk). Results show that unbiasedness and power are not always in the same direction. The choice of an approach for a particular trial should depend on the focus of the analysis. A data example is used to illustrate these points.