Conventional dose finding studies only focus on safety, and ignore the efficacy when performing the dose search and providing recommendation. Especially when the efficacy is not monotonically increasing with the doses (e.g. targeted therapies or immunotherapies), such design may lead to suboptimal dose being selected in the subsequent phase 2 or 3 studies. Thus, an adaptive design that considers both safety and efficacy is desired. In our work, we discussed and presented trade-off design and utility function design, which directly account for efficacy and tolerability. Trade-off design with different desirability functions is a model-based Bayesian approach that allows for adaptive dose choice for next cohort and non-monotone dose-efficacy relationship. The utility-based design adopts joint utilities of the possible outcome pairs. It offers adaptive randomization to allow for better exploration of possible treatment effects. It can be extended to ordinal responses in a straightforward way as demonstrated in the examples. We also conduct simulations to evaluate model performance under various scenarios and provide recommendations for oncology phase I-II dose finding studies.