Personalizing neoadjuvant chemotherapy regimens for triple-negative breast cancer using a biology-based digital twin.

Despite advances triple negative breast cancer treatment, ~50% of patients will not achieve a pathological complete response prior to surgery with standard of care neoadjuvant therapy (NAT). We hypothesize that personalized regimens for NAT could significantly improve patient outcomes, which we address with a patient-specific digital twin framework. This framework is established by calibrating a biology-based model to longitudinal magnetic resonance images with approximate Bayesian computation. We then apply optimal control theory to either (1) reduce the final tumor cell number with equivalent dose, or (2) reduce the total dose of NAT with equivalent response. For (1), the personalized regimens (n = 50) achieved a median (range) reduction in the final tumor cell number of 17.62% (0.00-37.36%). For (2), the personalized regimens achieved a median reduction in dose delivered of 12.62% (0.00-56.55%) when compared to the standard-of-care regimen, while providing statistically equivalent tumor control.