C. V. Guthier1,2, C. E. Kehayias1, J. He1, K. M. Atkins3, and R. H. Mak1,2; 1Department of Radiation Oncology, Dana-Farber Cancer Institute/Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, 2Artificial Intelligence in Medicine (AIM) Program, Mass General Brigham, Harvard Medical School, Boston, MA, 3Department of Radiation Oncology, Cedars-Sinai Medical Center, Los Angeles, CA
Purpose/Objective(s): Thoracic radiation therapy (RT) may lead to pulmonary vascular remodeling. In this study we investigated vascular pruning (i.e. regression) of proximal-to-distal pulmonary blood vessels (PBV) after thoracic RT for non-small cell lung cancer (NSCLC). In addition, we investigated the association of PBV dose with outcomes. Materials/
Methods: We retrospectively deployed an open-source pulmonary blood vessel (vein and arteries) auto-segmentation algorithm on RT planning CT scans from a clinical dataset of 878 patients with NSCLC. We divided PBVs into small, medium, and large vessels based on diameter cutoffs of <2.5mm2, 2.5mm2 to 10mm2, and >10mm2, respectively. We split the dataset to measure PBV pruning (Cohort 1, n=179) and PBV dose-outcomes (Cohort 2, n=699). Cohort 1 had paired planning and post-RT CT scans. Pre- and post-RT CT scans were registered to measure changes in PBV total density (PBV-TD) and volume changes for small (PBV-S), medium (PBV-M), and large (PBV-L) vessels, based on regions of dose received and RT modality (3D-conformal RT [3D-CRT], volumetric modulated arc therapy [VMAT], stereotactic body RT [SBRT]). In Cohort 2, patients with locally advanced NSCLC were analyzed for association of PBV dose variables all-cause mortality and major adverse cardiovascular events (MACE), using area under the receiver operating curve (AUC) and Cox and Fine and Gray regressions. Results: The baseline median PBV volumes for cohort 1 vs 2 were: PBVTV 183.3cc (IQR: 67.0cc) vs 165.0cc (IQR:44.7cc), PBV-S 30.0cc vs 32.0cc, PBV-M 121.9vs 125.0cc, and PBV-L 36.0cc vs 37.5. The PBV-S (-2.3%) and PBV-M (-6.8%) significantly decreased after RT, while PBV-L increased (+9.5%), but the total vessel volume (PBV-TD <1.1% change) remained almost constant. There was a dose-dependent effect on PBV-S and PBV-M pruning (p < 0.01 for trend) and PBV-L increase (p < 0.05), and significantly more pruning with SBRT vs 3DCRT and VMAT technique (p < 0.02). The greatest effect was observed in SBRT with 11.7% change. In Cohort 2, dose to the PBV-S and PBV-M were predictive of all-cause mortality on AUC analysis and the strongest predictors were V20 to PBV-S (AUC 0.58) and PBV-M (AUC 0.58). On multivariable analysis adjusting for age, sex, history of cardiac disease, use of surgery, and LAD V15 (but excluding lung V20Gy due to collinearity), PBV-S V20 was associated with mortality (aHR 6.5, 95% CI 1.99-21.20; p=0.002). Though there was no collinearity with planning target volume (PTV), a similar model that included PTV with the above variables resulted in a loss of significance for PBV-S V20 (p>0.05). No PBV DVH parameters were associated with MACE (p>0.05). Conclusion: Radiation dose to the PBVs leads to pruning of small- to medium-sized vessels in a dose dependent manner and is most pronounced with SBRT. Dose to the PBVs may be associated with survival, but not cardiac endpoints. Further analysis of PBV dose and relationship to PTV is warranted to understand the pathophysiological impact of lung vessel pruning.
