The 2‐in‐1 adaptive design allows seamless expansion of an ongoing Phase II trial into a Phase III trial to expedite a drug development program. Since its publication, it has generated a lot of interest. So far, most of the related research focused on type I error control. Similar to most adaptive designs, 2‐in‐1 design could also pose a great challenge on estimation of treatment effect due to the data‐driven adaptation. In addition, the use of intermediate endpoint for interim adaptive decision‐making is a less well‐studied field. In this paper, we investigate the bias and variances in estimation for 2‐in‐1 design and some of its extensions, and propose some bias‐adjusted estimators for 2‐in‐1 design. The properties of the proposed estimators are further studied theoretically and/or numerically, so as to provide guidance on how to interpret the estimated treatment effect of 2‐in‐1 design.