Single-arm trials are commonly used in the early development of oncology drugs to evaluate the preliminary anti-tumor activity. Provided compelling activity is seen, single-arm trials can lead to accelerated approval which can significantly shorten the time to market for new drugs. However, there are challenges in interpreting the results obtained from single-arm trials since they may provide a biased estimate of treatment effect because of patient selection and lack of control, and there is always an attenuation effect from smaller single-arm trials to larger randomized trials. In order to address the challenges, statistical methods should be applied to adjust for the bias before moving on to late-phase trials. In this roundtable discussion, we will discuss the sources of bias, statistical methods to adjust for bias, when to do the adjusted analyses and other practical considerations that should be made in conducting single-arm trials, with the aim to provide some practical guidance to the industry.