Factors predicting radiation pneumonitis in lung cancer patients: a retrospective study.

PURPOSE: To evaluate clinical and lung dose-volume histogram based factors as predictors of radiation pneumonitis (RP) in lung cancer patients (PTs) treated with thoracic irradiation.

METHODS AND MATERIALS: Records of all lung cancer PTs irradiated at our Institution between 1994 and 2000 were retrospectively reviewed. Eighty-four PTs with small or non-small-cell lung cancer, irradiated at >40 Gy, with full 3D dosimetry data and a follow-up time of >6 months from start of treatment, were analysed for RP. Pneumonitis was scored on the basis of SWOG toxicity criteria and was considered a complication when grade> or =II. The following clinical parameters were considered: gender, age, surgery, chemotherapy agents, presence of chronic obstructive pulmonary disease (COPD), performance status. Dosimetric factors including prescribed dose (Diso), presence of final conformal boost, mean lung dose (Dmean), % of lung receiving > or =20, 25, 30, 35, 40, and 45 Gy (respectively V20-->V45), and normal tissue complication probability (NTCP) values were analysed. DVHs data and NTCP values were collected for both lungs considered as a paired organ. Median and quartile values were taken as cut-off for statistical analysis. Factors that influenced RP were assessed by univariate (log-rank) and multivariate analyses (Cox hazard model).

RESULTS: There were 14 PTs (16.6%) who had > or =grade II pulmonary toxicity. In the entire population, the univariate analysis revealed that many dosimetric parameters (Diso, V20, V30, V40, V45) were significantly associated with RP. No significant correlation was found between the incidence of RP and Dmean or NTCP values. Multivariate analysis revealed that the use of mitomycin (MMC) (P=0.005) and the presence of COPD (P=0.026) were the most important risk factor for RP. In the group without COPD (55 PTs, seven RP) a few dosimetric factors (Dmean, V20, V45) and NTCP values (all models) were associated with RP in the univariate analysis (P< or =0.06). According to the multivariate analysis, the use of MMC was independently associated with RP (P=0.007), while Dmean approached statistical significance (P=0.082).

CONCLUSIONS: In this study the use of mitomycin or the presence of COPD is associated with a higher risk of RP. In the entire population NTCP values were not significantly correlated with the incidence of RP. Mean lung dose shows a clear trend toward statistical significance in the patient group without COPD.