Episode 20, “The Legend of I-SPY 2 - Part A”
I-SPY 2 set a new standard in breast cancer trials by shifting to a neoadjuvant model—administering therapies before surgery and evaluating pathologic complete response. Dr. Donald Berry and Dr. Scott Berry describe how, unlike adjuvant studies with years-long endpoints, this approach allows direct assessment of tumor response in real time.
The trial implemented adaptive multi-arm bandit randomization: patients were classified into eight molecular subtypes, with allocation probabilities adjusted based on early efficacy signals in each group. I-SPY 2 systematically evaluated 10 biomarker signatures, selected from 255 possible subtype partitions, using Bayesian modeling to refine treatment indications as data accumulated.
Integration of MRI data—through predictive modeling and multiple imputation—enabled estimation of pathologic complete response before surgery, significantly expediting therapy evaluation. This methodology led to rapid, data-driven decisions, such as the graduation of pembrolizumab (Keytruda) based on early predictive outcomes. These design elements not only increased statistical and operational efficiency but also aligned treatment assignments with patient-specific biology, improving outcomes for patients in the trial.
Learn more on the full episode: https://www.berryconsultants.com/resource/20-the-legend-of-i-spy-2-part-a