Abstract
•No statistically significant difference between the high- and low dose arm in risk of RP.•Identified risk factors for RP: target volume, infection, dyspnea at baseline, and increase of cough during RT.•number of patients was too small to establish a statistically sound multivariable model.
Prospectively scored radiation pneumonitis (RP) observed in a national, randomized phase II dose-escalation trial for patients with locally advanced non-small cell lung cancer (NSCLC) was investigated.
Patients with stage IIB-IIIB histologically proven NSCLC were treated with concomitant chemo-radiotherapy (oral Vinorelbine 3times/week) at 60 Gy/30fx (A–59pts) and 66 Gy/33fx (B–58pts) from 2009 to 2013 at five Danish RT centers. Grade 2 RP (CTCAEv3.0) was investigated with univariate analysis for association with clinical and dosimetric parameters, including dyspnea and cough at baseline and during RT. Multivariable logistic regression and Cox regression with regularization were used to find a multivariable model for RP ≥ G2.
Despite a tendency of higher mean lung dose in the high-dose arm (median[range] A = 14.9 Gy[5.8,23.1], B = 17.5 Gy[8.6,24.8], p = 0.075), pulmonary toxicities were not significantly different (RP ≥ G2 41%(A) and 52%(B), p = 0.231). A Kaplan Meier analysis of the time to RP ≥ G2 between the two arms did not reach statistical significance (p = 0.180). Statistically significant risk factors for RP ≥ G2 were GTV size (OR = 2.091/100 cm3, p = 0.002), infection at baseline or during RT (OR = 8.087, p = 0.026), dyspnea at baseline (OR = 2.184, p = 0.044) and increase of cough during RT (OR = 2.787, p = 0.008). In the multivariable logistic regression and the Cox regression analysis, the deviances of the most predictive models were within one standard deviation of the null model.
No statistical difference between the high- and low dose arm was found in the risk of developing RP. The univariate analysis identified target volume, infection, dyspnea at baseline, and increase of cough during RT as risk factors for RP. The number of patients was too small to establish a statistically sound multivariable model.