ESTRO 2024 - Abstract Book

S1583

Clinical - Lung

ESTRO 2024

Pre-treatment 4DCT data were used to compute ventilation images for 598 non-small cell lung cancer patients: 256 patients received SABR treatment and 342 patients received non-SABR treatment. Ventilation images were calculated using the VESPIR MATLAB-based toolkit. 4 CT ventilation was based on deformable image registration evaluation of volume change (CTVI Jac ) or change in Hounsfield Unit (HU) value (CTVI HU ). Standard dose-volume metrics (e.g. MLD) and ventilation-based dose-function metrics (e.g. functional mean lung dose, fMLD) were computed for each patient. Dose-function metrics were calculated using two approaches: (1) Structure-based: Creating a functional lung contour and calculating various dose metrics inside the functional contour (Figure 1). The ventilation image was thresholded with a cut-off that was varied systematically from the 5 th percentile to 95 th percentile of ventilation values to produce the functional contour; (2) Voxel-based: Weighting each dose voxel by its ventilation percentile before calculating dose metrics of the lung. The ability of the dose-volume and ventilation based dose-function metrics to predict survival was assessed using Cox regression analysis and concordance index metrics.

Figure 1. Structure-based approach of calculating dose-function metrics: lung mask, CT ventilation image, functional contour, and dose distribution (left to right). All data are overlaid on the peak exhale 4DCT.

Results:

For the SABR cohort, at 1, 2, 3, and 5 years, OS was respectively 87.5, 65.2, 49.2, 29.3% (median OS was 36 months). For the non-SABR cohort, at 1 and 2 years, OS was respectively 74.8 and 61.4% (median OS was not reached). Mean PTV was 34.6 cc and 258.5 cc for the SABR and non-SABR cohorts, respectively. Significant prognostic factors of OS in univariate and multivariate Cox regression analysis were ECOG performance status (PS) ≥ 2 (Hazard Ratio (HR) = 1.41, p = 0.03) for the SABR cohort, and age (HR = 1.03, p = 0.02), logarithm of the PTV volume (HR = 1.61, p = 0.001) and PS ≥ 2 (HR = 1.54, p = 0.02) for the non -SABR cohort. Significant correlations (p < 0.05) between either dose volume or dose-function metrics and survival were seen in univariate analysis. However, in multivariate analysis these factors were not significant. Results for the structure-based approach using an 85th percentile cut-off are presented in Table 1 following the work of Faught et al. which found it to be the optimal threshold for predicting grade 3+ radiation pneumonitis. 5 An MLD of ≤ 6.6 Gy was significantly associated with an increase in median OS from 30 months to 40 months for the SABR cohort. The dose metric that resulted in the highest concordance index in univariate analysis was MLD (C-index = 0.56) and V30Gy (C-index = 0.60) for the SABR and non-SABR cohorts, respectively.

SABR dataset

Non-SABR dataset

Univariate

Multivariate

Univariate

Multivariate

Variable

p-value

HR

p-value

HR

p-value

HR

p-value

HR

Age

0.02

1.02

0.10

1.01

0.03

1.03

0.02

1.03