ESTRO 2024 - Abstract Book

S1754

Clinical - Lung

ESTRO 2024

Material/Methods:

We analysed a set of non-resected LA-NSCLC patients treated with arc therapy given a prescribed dose of 50 Gy to 66 Gy. For each patient, we contoured automatically through the MVision Software a set of 9 cardiac substructures: aorta, left anterior descending artery (LAD), pulmonary artery, left and right atrium, left and right ventricle, superior and inferior vena cava. In the analysis, we also included the dose to the lungs and the dose to the PTV. For this preliminary analysis, we derived each structure's mean dose and 2cc (D2cc) dose. The chemotherapy scheme was included as a feature. Moreover, we retrieved the information on the overall survival at 2 years (OS2), which was used as the study endpoint. Then, we applied a feature selection to define the variable mostly related to the endpoint. A random forest model was applied based on features contributing to and limiting the OS2. The data set was divided into training and test cohorts. Model performances were reported for the test cohort. Finally, we ranked the importance of features to identify the most radiosensitive structures.

Results:

We studied the treatment data of 88 patients; 40% were treated with RT alone, 30% with sequential CH-RT and 30% with concomitant CH-RT. The application of immunotherapy was limited and not analysed in this cohort. The OS2 was 42%. The survival was lower in sequential CH-RT (37%) than in radical and concomitant RT (44.5% in both options, p-value 0.06). No differences in OS2 were reported based only on PTV dose, suggesting a possible role for survival limiting toxicity leading to adverse non-oncological events. The HSs associated with the survival-limiting toxicity were D2cc to the aorta and left ventricle, suggesting the importance of multiple HSs with a serial behaviour to radiation response. Also, the mean dose to the left lung was associated with OS2, confirming its parallel characteristic. Box plots for those variables are depicted in Figure 1.