The impact of dose algorithms on tumor control probability in intensity-modulated proton therapy for breast cancer

Purpose To evaluate the radiobiological impact of PB algorithm versus MC algorithm in intensity-modulated proton therapy (IMPT) plans for breast cancer treatment. Methods 20 breast cancer patients who received IMPT to the breast/chest wall and regional lymphatics were included in this study. For each patient, 2 IMPT plans were generated: a PB-optimized plan and a MC-optimized plan. The radiobiological and dosimetric impact to the CTV of the dose algorithms was assessed. The Poisson Linear-Quadratic model was used to estimate the tumor control probability (TCP). The influence of the model parameter uncertainties on the TCP was tested against different sets of published model parameters. Results The PB-optimized plans significantly under-dosed the target as compared to the MC-optimized plans. The median (range) differences in CTV D95% and CTV Dmean were 1.9 (1.2–3.1) Gy and 1.2 (0.5–1.9) Gy, corresponding to 3.8% (2.4−6.2%) and 2.4% (1.0−3.8%) of the prescription dose. The TCP was lower in the PB-optimized plans than the MC-optimized plans. The median (range) of the TCP differences (ΔTCP) were 4% (2−6%), 3% (2−5%), and 2% (1−3%), respectively, when calculated using 3 different model parameter sets. The ΔTCP correlated with the CTV dose difference, and moderately correlated with the CTV volume. Conclusion Due to the inaccurate dose modeling, PB-optimized plans under-dose the target and therefore yield a lower TCP compared to MC-optimized plans in breast IMPT. The magnitude of the resulting difference in TCP reached 6% in our study.