The problem of identifying the lowest effective drug dose and/or the highest safe dose is often an important issue in a drug efficacy study. Stepdown testing procedure is a widely used approach in this area. Under the assumption of a monotone mean configuration, the stepdown testing procedure is particularly powerful and easy to perform. One common shortcoming of the previous research is that the proposed testing procedures have implicitly assumed that the dose study data are randomly collected from the probability distributions. In a clinical dose study with adaptive design, this assumption is not correct because the data collected are not totally random samples. It is necessary to tailor stepwise testing procedures for non-random samples in order to draw valid statistical inference. A simulation-based approach is proposed to address the issue.