ESTRO 2025 - Abstract Book

S1351

Clinical - Lung

ESTRO 2025

lymph node irradiation is associated with the end-of-radiotherapy (EoRT) ALC. To examine this, we use mediastinal tissue mean dose ( D mean,med ) as an approximation of lymph node dose.

Material/Methods: A cohort included 180 LA-NSCLC patients treated Jan 2018-Sept 2023 at a single institution. Patients received ≥1 cycle of platinum-based chemotherapy concurrently with RT (60-68 Gy in 2 Gy/fr. or 55 Gy in 2.75 Gy/fr.). Mediastinal tissue segmentation was automated and done using operations with already segmented structures (Figure 1). Continuous EoRT ALC variable was the modeled outcome in univariate and multivariable linear regression models. For that purpose, EoRT ALC was obtained from an exponential fit of patient-specific ALC measurements prior to and during the treatment. A multivariable model with D mean,med and other predictors deemed most clinically relevant for EoRT ALC prediction was predefined while considering inter-predictor correlations. Analogously, a model with estimated dose to immune cells (EDIC) was predefined for comparison. EDIC represents dose to immune cells in circulating blood 1 and was calculated using D mean of heart, lung, and body as per Ladbury et al . 2 Univariate modelling was done for all available predictors.

Results: Only dosimetric predictors, baseline ALC, and planning treatment volume were significant ( p <0.05) in univariate analyses. Results of univariate and predefined multivariable models are in Table 1. Inter-predictor Pearson correlation coefficients of continuous variables were greater for EDIC than for D mean,med . Adjusted R 2 was 0.231 for a multivariable model with D mean,med and 0.257 for the one with EDIC. Residual analyses were comparable. Akaike information criterion modestly favored a model with EDIC (67 vs. 61).