Over the last decade, the focus of anticancer drug development has shifted from empirical cytotoxic chemotherapy to mechanism-defined molecularly targeted agents, for which appropriate patient selection at the earliest possible point in drug development is rational and critical to success. With the recently legislated "breakthrough product" definition in the U. S., it may be possible to plan a single trial for registration purposes to confirm a major clinical effect observed in phase I. However, most phase I trial designs remain excessively conservative and are driven by criteria developed for cytotoxic agents with the goal of identifying a "maximum tolerable dose" with acceptable risks to patients. This focus on empiric "most dose with acceptable risk" may be misguided for mechanism-targeting new agents, and this could lead to unnecessary delays, increased costs and even higher risk of missing important signals of activity and benefit. There is a compelling need to modify phase I trial designs to facilitate enrichment in molecularly selected patients who are the most likely to harbour disease driven by the targeted pathway and to avoid unjustified exclusions based on obsolete criteria so that the right subset of patients can participate. After discussion of the main inconsistencies of current phase I designs, we propose a new strategy to facilitate the inclusion of molecularly selected patients, in order to accelerate and mitigate risks in drug development as well as to increase the chance of benefit among trial participants. (C) 2013 Elsevier Ltd. All rights reserved.