Abstract: I-SPY 2 is a phase 2 drug screening process that matches patient biomarker subsets with therapies on the basis of rate of pathCR. Therapies that perform sufficiently well in I-SPY 2 in that they have a high predictive probability of success in phase 3 (and therapies from other phase 2 trials) “graduate” into I-SPY 3. This is platform trial that conforms to the FDA’s draft guidance. The statistical goals are to confirm the phase 2 results regarding pathCR (for accelerated approval) and to show superiority to standard therapy on the basis of EFS (for full approval). The trial is prospectively adaptive in the sense that sample size to have sufficient power for both pathCR and EFS endpoints is determined during the trial based on pathCR information that has accrued in the trial. The relationship between pathCR and EFS is modeled depending on biomarker subset using the results of the FDA’s recent meta-analysis (presented at the 2012 San Antonio Breast Cancer Symposium and soon to be published). The design is “Goldilocks” in the sense that the sample size is tailored to the results of the trial and is neither too big nor too small. I will consider triple-negative disease as an example. If the pathCR rate advantage of the experimental therapy is minimal or large then the resulting mean sample size is small. If the pathCR advantage is moderate, such as an increment of 20%, then the trial usually attains its maximal possible sample size.