Modeling radiation pneumonitis risk with clinical, dosimetric, and spatial parameters
Andrew J Hope1, Patricia E Lindsay, Issam El Naqa
1Department of Radiation Oncology, Washington University School of Medicine, Siteman Cancer Center, St. Louis, Missouri 63110-1032, USA.
Purpose:
To determine the clinical, dosimetric, and spatial parameters that correlate with radiation pneumonitis.
Methods And Materials:
Patients treated with high-dose radiation for non-small-cell lung cancer with three-dimensional treatment planning were reviewed for clinical information and radiation pneumonitis (RP) events. Three-dimensional treatment plans for 219 eligible patients were recovered. Treatment plan information, including parameters defining tumor position and dose-volume parameters, was extracted from non-heterogeneity-corrected dose distributions. Correlation to RP events was assessed by Spearman's rank correlation coefficient (R). Mathematical models were generated that correlate with RP.
Results:
Of 219 patients, 52 required treatment for RP (median interval, 142 days). Tumor location was the most highly correlated parameter on univariate analysis (R = 0.24). Multiple dose-volume parameters were correlated with RP. Models most frequently selected by bootstrap resampling included tumor position, maximum dose, and D35 (minimum dose to the 35% volume receiving the highest doses) (R = 0.28). The most frequently selected two- or three-parameter models outperformed commonly used metrics, including V20 (fractional volume of normal lung receiving >20 Gy) and mean lung dose (R = 0.18).
Conclusions:
Inferior tumor position was highly correlated with pneumonitis events within our population. Models that account for inferior tumor position and dosimetric information, including both high- and low-dose regions (D(35), International Commission on Radiation Units and Measurements maximum dose), risk-stratify patients more accurately than any single dosimetric or clinical parameter.
