Treatment selection method is an important application of adaptive design in clinical studies. This paper will extend the simulation-based approach and Bonferroni-Holm test procedure in Posch et al. (2011) to a normally distributed endpoint assuming varying selection rules. We assume that the study has two active treatments and a control group with one interim analysis, and use the following treatment selection rules based on interim analysis results: (1) select one active treatment with a better response than other active treatment by a pre-specified threshold; (2) early termination is allowed if it meets the pre-specified criteria. How to determine critical boundaries to maintain Type I error under these scenarios, and how to calculate the sample size to achieve the target statistical power will be presented.