This paper examines the design of phase II trials in oncology and recommends departing from the traditional uncontrolled trial design. Entrance into phase II clinical evaluation represents a key milestone in the development of any new cancer therapy. As novel molecular-targeted therapies are introduced, whose primary action is to slow the growth of tumors, it will be important to ensure that the clinical trial design will effectively capture any clinical benefit of these agents. The objective of a phase II trial should, in addition to identifying active therapies, be extended to identifying those that are likely to be successful in pivotal trials. It is therefore necessary to quantify the likelihood of either incorrectly halting the development of an active agent or continuing development of an ineffective agent. We believe only randomized studies with comparative intent and including a concurrent active control, can reliably assess these risks corresponding to significance and power. Given that the objective of phase II studies is to identify promising treatments, it is important not be constrained by conventional levels of significance. This paper will review the various approaches to phase II trial design in oncology and provide a framework for fully powered randomized trials of a moderate size. For example, a randomized trial of just 100 patients could lead to the termination of development of 90% of inactive agents whereas at least 80% of agents with a meaningful and realistic increase in progression-free survival would be identified for confirmatory study. We believe randomized studies with progression-free survival endpoints are the most powerful and economical method of determining the clinical activity of new cytostatic agents.