Dose-volume Factors Predicting Radiation Pneumonitis after SBRT for Ultra-central Lung Tumors

Abstract

The safety of SBRT is uncertain for ultra-central tumors (near the proximal airways or esophagus). Predictors of toxicity in this population are urgently needed. We therefore studied dose-volume correlates of radiation pneumonitis after ultra-central SBRT. 88 patients with tumors abutting proximal bronchial tree (PBT) or PTVs overlapping esophagus (n = 76 and 23; 11 met both criteria) were included. 53 (60%) had primary/locally recurrent lung cancer, and 35 had lung metastases. All had 5, 8 or 15 fractions of image-guided radiotherapy with BED ≥84Gy (α/β=10). Primary endpoints were grade ≥2/3 radiation pneumonitis (RP2+/RP3+) using CTCAE 4.0. Dose-volume histograms (DVHs) using linear-quadratic equivalent doses in 2 Gy fractions (ED) were calculated for lung and heart (α/β= 3 Gy). Volumes of these organs exposed to equivalent doses (VEDX) were tested for correlation with RP2+/RP3+ using Cox proportional hazards models with ED values of 2.5 Gy, then from 5-70 Gy in 5 Gy steps. Age, gender, KPS, COPD, radiographic ILD, smoking, anti-VEGF agents, and GTV size were also tested and included in multivariate analysis with the most significant lung and heart DVH variables. Median follow up was 14.3 months. There were 24 (27%) and 9 (10%) cases of RP2+ and RP3+. Four patients (4.5%) died of RP. Univariately, significant correlations with RP2+ were seen for lung ED from 2.5-70 Gy (p < 0.002), with strongest correlations for ED from 10-20Gy(p < 10-5). One-year rates of RP2+ with Lung VED15< or > 14.7% (the median) were 13% and 46% respectively. Correlations with RP2+ were seen for heart ED from 2.5-50 Gy, with the strongest correlation for VED5(p=0.0001). Significant correlations with RP3+ were seen for lung ED from 2.5-35 Gy, with strongest correlations for VED10(p = 0.005). Correlations with RP3+ were also seen for heart ED from 2.5-15 Gy, with the strongest correlation for VED2.5(p=0.03). One-year rates of RP3+ with Lung VED10< or > 18% (the median) were 5% and 18% respectively. No clinical variables save ILD were significantly correlated with RP2+/RP3+ (p = 0.002 and 0.0001 respectively). Though ILD was statistically significant, only 2 patients in the cohort had ILD (both had RP3+, and one died of RP). The most significant lung and heart DVH variables were correlated with each other (p < 10-4). Multivariate models of RP2+ and RP3+ using Lung VED15and ILD were significant (p <10-5and p=0.02 respectively, for RP2+; p=0.02 and 0.001 respectively, for RP3+). We observed a high rate (27%) of RP after ultra-central SBRT, with a 4.5% rate of fatal RP. Lung dose-volume factors are highly correlated with RP. Presence of ILD warrants extreme caution. Our analysis suggests that RP3+ can be limited to 5% by keeping Lung VED10< 18%, and that RP2+ can be limited to 13% by keeping Lung VED15<14.7%.