Although it has been exciting for lung cancer doctors to observe objective remissions with gefitinib and erlotinib in heavily pretreated NSCLC patients, all of the reported phase III trials testing noncytotoxic, targeted therapies in NSCLC have been negative. Two basic strategies have been employed in developing and conducting these randomized studies. In the case of gefitinib and the matrix metalloproteinase inhibitors, phase III trials were launched based on preclinical data. The second strategy was based on survival results from phase II trials involving regimens consisting of the targeted agent and chemotherapy. Unfortunately, negative results have been observed with the first phase III study (chemotherapy +/- ISIS 3521), which was based on the results of a phase II trial. The initial negative results with targeted agents suggest that a paradigm shift in cancer drug development is needed. Typically, the development of a cytotoxic agent involves determination of the maximum tolerated dose, followed by an assessment of activity as defined by the objective response rate in specific tumor types. "Active" drugs are then moved into phase III testing to determine the effect on survival. Other than targeting the specific tumor type and defining the usual eligibility parameters, no attempt is made to select patients for treatment with new agents. It seems unlikely that there will be significant progress with the targeted therapies unless there is a paradigm shift from this classic model of cancer drug development to a model in which much greater effort is directed toward identifying the target or targets in preclinical models. Intensive effort should be devoted to the development of reliable, clinically applicable assays for the targets that could identify patients who are most likely to benefit from a specific treatment. Rothenberg et al recently made similar recommendations with respect to improving the drug discovery process for cancer. These investigators have emphasized testing new agents in the most appropriate setting, increasing efforts to understand the role of the target, and collection of tissue in an effort to select appropriate patients. Although results from initial randomized trials of targeted therapies in NSCLC have been relatively disappointing, this is not a time to be discouraged. Rather, it is a time to increase the collaborative efforts between basic scientists and clinical investigators.