An agenda for Clinical Trials: clinical trials in the genomic era

The mantra of many clinical trialists has been to do randomized clinical trials with broad eligibility and avoid subset analysis. Developments in cancer research, however, have raised some questions about this approach in the genomic era of molecularly targeted therapeutics. A large body of evidence indicates that cancers of most primary sites are heterogeneous with regard to molecular pathogenesis, genomic signatures and phenotypic properties. Consequently, it is not necessarily reasonable to expect such tumors to have equal sensitivities to a drug that inhibits a particular protein target. The protein target may be driving tumor growth in only a subset of the tumors. As an example, two large randomized clinical trials were recently reported comparing standard therapy to standard therapy plus a new drug, Iressa, for patients with lung cancer [1,2]. Both trials were convincingly negative. Nevertheless, the US Food and Drug Administration approved the drug based on the recommendation of an advisory committee. The approval resulted from evidence of durable partial tumor response in about 10% of patients with advanced lung cancer in an uncontrolled phase II study. Subsequent publications indicated that patients who responded in the phase II trial were those whose tumors had a mutation in the kinase domain of their EGFR gene, rendering those tumors highly sensitive to treatment with an epidermal growth factor receptor inhibitor such as Iressa [3,4]. This result, indicating that Iressa was highly effective for a small subset of cases and that large randomized clinical trials of unselected patients failed to identify the value of the drug, has provided a stimulus to think about clinical trial methodology for the evaluation of molecularly targeted drugs in oncology. Clinical trials in which eligibility is restricted to those patients whose tumors are sensitive to the drug can be substantially more efficient than traditional clinical trials with broad eligibility. Some of the dramatic improvements in the possible efficiency have been indicated by Simon and Maitournam [5]. The improvement in efficiency results because the treatment effect is substantially larger in the focused clinical trial if there is a good assay available for selecting patients likely to respond to the new treatment. Such focused treatment can provide a more favorable benefit to complication ratio and result in a greater proportion of the treated patients benefiting from the treatment. This can also have important economic benefits for society. Currently, for some indications such as stage I estrogen receptor positive breast cancer, fewer than 10% of the patients treated with cytotoxic chemotherapy actually derive benefit from the drugs. The proportion may be even lower for prevention settings of cancer and other diseases. It is important, therefore, to develop predictors of