Poisson Linear-Quadratic Tumor Control Probability Model Validation in Micro- and Macroscopic Breast Cancer Radiation...

To validate a mechanistic Poissonian linear-quadratic tumor control probability model that incorporates tumor-volume heterogeneity, interpatient radiosensitivity variability, and treatment time effects using published breast cancer radiation therapy (RT) data sets.

Local control (LC) data were obtained from the START A/B, FAST-Forward, and Early Breast Cancer Trialists' Collaborative Group studies. To address diverse scenarios, we combined data on the additive effect of adjuvant RT with outcomes from RT-alone treatments. We fitted the model parameters [α, β, α/β, σα (interpatient heterogeneity), tumor volume scaling, and exponential time-to-failure] against the observed LC rates. We quantified the effect of overall treatment duration on locoregional recurrence and estimated the residual clonogen count after surgery. We also predicted in-breast recurrence rates for various tumor sizes for neoadjuvant tumor-directed RT dose-fractionation schemes.

The model reproduced 5-year LC outcomes across both adjuvant and RT-alone settings. The estimated residual clonogen count after surgery was 125. The calculated effect of overall treatment time for locoregional recurrence was 0.58 Gy/d, closely aligning with the 0.60 Gy/d estimate from the START trials. Simulations showed strong tumor-size dependence for preoperative stereotactic schedules: for a 5-mm diameter (T1a) tumor, the model predicted that a single 23-Gy fraction would result in >99.9% LC, whereas a 10-fraction regimen would yield ∼98% LC. For a 10-mm diameter tumor, single-fraction treatments maintained >98% LC. Control declined with increasing size, and no evaluated regimen achieved >90% LC for tumors >4 cm in diameter.

By incorporating tumor size heterogeneity, radiosensitivity variability, and time effect, the Poissonian linear-quadratic-based population tumor control probability model provides a robust framework for predicting LC in breast cancer RT. With further validation in larger data sets, this model could become a valuable tool to tailor dose regimens to individual patient and tumor characteristics, potentially improving LC rates and optimizing treatment strategies.