ABSTRACT
Clinical trials of new oncology drugs have traditionally been conducted with broad patient populations to avoid discrepancies between the population tested and the population eventually treated with the drug. This approach has, however, resulted in treating many patients who do not benefit. For example, less than 10% of women with estrogen receptor positive breast cancer that has not spread to the axillary lymph nodes require or benefit from cytotoxic chemotherapy although most have traditionally received such drugs. Over-treatment results in a substantial number of adverse events and expense for treatment of patients who receive no benefit. In oncology, accumulating understanding of genomic differences among tumors of the same primary site indicates that most molecularly targeted agents are likely to benefit only the patients whose tumors are driven by deregulation of the targeted pathways. Availability of improved tools for characterizing tumors biologically makes it increasingly possible to predict whether a particular tumor will be responsive to a particular treatment (Paik, Taniyama, and Geyer 2008). Today many cancer drugs are being developed with companion diagnostics that identify the patients who are good
18.1 Introduction .................................................................................... 18-1 18.2 Terminology .................................................................................... 18-2 18.3 Development of Predictive Biomarkers ....................................... 18-3 18.4 Phase II Designs for Oncology Trials ..........................................18-4 18.5 Clinical Trial Designs Incorporating Predictive
Biomarkers .......................................................................................18-5 18.6 Enrichment Designs .......................................................................18-5 18.7 Designs that Include Both Test Positive and Test Negative
18.8 Adaptively Modifying Types of Patients Accrued ................... 18-11 18.9 Biomarker Adaptive Threshold Design ..................................... 18-11 18.10 Multiple Biomarker Design ......................................................... 18-12 18.11 Adaptive Signature Design .......................................................... 18-12 18.12 Conclusions.................................................................................... 18-13
candidates for treatment. Successful prospective codevelopment of a drug and companion diagnostic, however, presents many new challenges. In this chapter we will address some of the issues in the design of clinical trials for new treatments and diagnostic tests that provide both reliable tests of the overall null hypothesis and also provide reliable indication of which patients benefit from the new treatment.
