I-SPY 2 is an adaptively randomized, biomarker-driven clinical trial in neoadjuvant breast cancer. It is sponsored by the Foundation for the NIH, which is a consortium of the NIH, the FDA, industry, and academia. It is more a screening process than a trial. Patients are randomized to receive one of a number of experimental agents in addition to standard therapy or standard therapy alone, in the pre-surgery setting. Agents are included in the trial as they become available and provided the accrual rate is sufficient to accommodate them. Patients are categorized by tumor type into 8 biomarker subtypes, and are assigned therapy depending on the performance of the various therapies within their subtype. A biomarker signature is a combination of these 8 subtypes and so there are 255 possible signatures, although only 10 signatures are considered in the trial design. Therapies graduate from the trial when their predictive probability of a successful phase 3 continuation trial in one of these signatures is sufficiently high, and they drop from the trial for futility when all 10 predictive probabilities are sufficiently small. The trial’s primary endpoint is pathologic complete response at surgery, about 6 months after randomization, which has recently been established in an FDA draft guidance as a possible route to accelerated approval. To get information about the patients’ progress earlier than 6 months we use tumor imaging during the treatment phase, relating tumor burden to the primary endpoint. Additional “qualification biomarkers” are assessed in parallel with the trial but do not drive treatment assignment. As of July 2012 we’ve randomized over 240 patients to 5 treatment arms.