X. Y. Wu1,2, M. Chen1,2, and J. Y. Chen1,2; 1Shanghai Key Laboratory of Proton-therapy, Shanghai, China, 2Department of Radiation Oncology, Ruijin Hospital, Shanghai Jiaotong University School of Medicine, Shanghai, China
Purpose/Objective(s):To identify thoracic anatomic features that correlate with the benefit of proton therapy (PT) in terms of lung protection compared to photon IMRT in breast cancer (BC) patients and to provide a reference for patient selection. The benefit of PT was assessed by the difference in the normal tissue complication probability (?NTCP) of radiation-induced lung injury (RILI) between proton and photon plans.Materials/
Methods: We retrospectively reviewed 409 postoperative BC patients receiving photon IMRT from Aug 2017 to Feb 2022 and randomly split the dataset into training and validation sets at an 8:2 ratio. A dose-modifying-factors(DMFs)-incorporated Lyman–Kutcher–Burman NTCP model was developed with the training set to predict the risk of grade=1 RILI (CTCAE 5.0) within 1 year after radiotherapy. The DMFs stood for baseline risk factors identified by uni-multivariable logistic regression. NTCP model parameters (n, m, TD50, and DMFs) were derived by maximum likelihood estimation. Model validation used several methods including area under receiver operating curve (AUC). Then, PT plans were generated for 80 patients from the dataset. Linear regression was used to assess the correlation between anatomic features and ?NTCP. Anatomic features were quantified in 6 parameters: the minimum thickness of chest wall soft tissues across all the transverse planes, the volume of ipsilateral lung, the anterior-posterior (AP) to left-right ratio of thorax and ipsilateral lung, the ratio of AP distance of thorax to ipsilateral lung and the Arc Height to Base Ratio (AHBR) (defined by the height to base ratio of the osseous thorax arc within the parasternal to mid-axillary line). The last 4 parameters were measured on 2 transverse planes where the sternal angle and the fourth rib lie. Results: In total, 191 (46.70%) cases of RILI were observed in the whole dataset. Multivariable logistic regression revealed 2 independent risk factors for RILI as BMI=23.52 kg/m² (p=0.049) and chemo-radiotherapy interval= 20 days (p=0.014), which were represented as DMF-BMI and DMF-TIME in NTCP model. The optimal NTCP parameters were: n=0.40, m=0.22, TD50=24.66 Gy, DMF-BMI=0.88, DMF-TIME=0.92. The model performed well in AUC (training set 0.754, validation set 0.733) and other validation tests. Of the 80 patients with photon and proton plans, the mean ipsilateral lung ?NTCP was 57.45±10.51%. Linear regression showed a significant positive correlation between AHBR on the transverse plane of sternal angle and the ipsilateral lung ?NTCP (regression coefficient 56.56, p=0.049), indicating that for every 0.1-unit rise in AHBR, ?NTCP increases by 5.656%. Conclusion: Thoracic anatomic feature helps to predict the benefit of PT in reducing RILI for BC patients. Patients with a larger AHBR at the transverse plane of sternal angle are expected to gain greater benefit from PT in terms of lung sparing.
