New methods have been developed to evaluate targeted therapies, since the classic sequence – phase I, toxicity; phase II, efficacy; phase III, comparison with standard treat ment – is no longer effective for evaluating these new treatments. In traditional cytotoxic chemotherapy trials, we observe a positive correlation between dose toxicity and dose efficacy. In targeted therapy trials, however, high doses can sometimes be well tolerated and increasing the dose beyond a certain level does not increase tumour response. Early clinical trials in targeted therapies therefore need to simultaneously assess toxicity and provide early signals of efficacy, based on biomarkers when available. Phase II primary endpoints have also been questioned, since the RECIST (Response Evaluation Criteria in Solid Tumour) is not well suited to functional modifications in tumours. New phase III trials, with more homogeneous targeted populations, are using more flexible designs, including interim analyses and adaptive designs. These flexible designs allow the sample size, and sometimes the trial design, to be modified during the trial. This article discusses these new methodological challenges for evaluating targeted therapies.