Relevance between patients' clinical factors and 4D-CT lung ventilation heterogeneity for evaluating the adoption of functional avoidance thoracic radiotherapy with a large patient database.

Objective: Four-dimensional computed tomography (4D-CT)-guided functional-avoidance radiotherapy reduces radiation pneumonitis, but the lack of patient selection criteria limits its adoption. This study aimed to identify clinical predictors of lung ventilation heterogeneity and determine which patients derive the greatest dosimetric benefit from functional-avoidance planning. Methods: We retrospectively analyzed 274 lung cancer patients. 4D-CT ventilation imaging was generated via deformable registration and a density-change algorithm, and the coefficient of variation (CoV) across six lung zones quantified heterogeneity. Clinical factors were assessed using Mann-Whitney U, Spearman, and multivariable logistic regression for three ventilation outcomes. Thirty-two patients were selected from CoV quartiles and re-planned with functional-avoidance plans; dosimetric parameters were compared using the Wilcoxon signed-rank test, with False Discovery Rate (FDR) correction applied to address multiple comparisons. Results: Univariate analysis showed chronic obstructive pulmonary disease (COPD) and airway obstruction reduced bilateral mean ventilation, while central tumors impaired ipsilateral ventilation. Higher CoV was observed in stage III-IV versus I-II, central versus peripheral tumors, and patients with airway obstruction. Multivariable analysis identified airway obstruction as an independent predictor of elevated CoV (adjusted OR = 2.914, 95% CI: 1.315-6.454, P<0.05). Central location and advanced stage showed borderline significance, while COPD trended toward lower CoV. Dosimetrically, functional-avoidance plans reduced functional lung metrics in a heterogeneity-dependent manner. After FDR correction, key metrics (fMLD, fV10, fV20, fV30) remained significant in Group 1 (highest CoV; Q = 0.040-0.049), while significance diminished in lower CoV groups. Leave-one-out (LOO) sensitivity analysis confirmed fMLD and fV20 as robust endpoints. Furthermore, beam arrangement modifications were most frequent in Group 1 (5/8 patients) and absent in Group 4 (0/8), reflecting clinical optimization needs rather than confounding. Conclusions: Ventilation heterogeneity is driven by tumor burden and airway patency. Patients with advanced stage, central tumors, or airway obstruction exhibit the strongest heterogeneity and may represent appropriate candidates for functional-avoidance radiotherapy in future prospective trials. However, as this replanning study lacked clinical follow-up for radiation pneumonitis, these findings reflect dosimetric rather than confirmed clinical patient-level benefit. Notably, CoV quartiles were derived from internal median splits within this cohort and require external validation. This study provides a dosimetric basis for hypothesis-generating patient selection.