Physical and biological predictors of changes in whole-lung function following thoracic irradiation

To develop methods of predicting the pulmonary consequences of thoracic irradiation (RT) by prospectively studying changes in pulmonary function following RT.

100 patients receiving incidental partial-lung irradiation during treatment of tumors in or adjacent to the thorax had whole-lung function assessed via symptoms and pulmonary function tests (PFTs: FEV1-forced expiratory volume 1 s; DLCO-diffusion capacity) before and repeatedly 6-48 months following RT. All had computed tomography-based three-dimensional (3D) dose calculations with lung density heterogeneity corrections for dose-volume histogram (DVH) and normal tissue complication probability (NTCP) calculations. Functional DVHs (DVfH) based on SPECT (single photon emission computed tomography) lung perfusion scans, and serial transforming growth factor-beta (TGF-beta1) levels were available in 50 and 48 patients, respectively. The incidence and severity of changes in whole-lung function were correlated with clinical, physical, and biological factors. Exploratory statistical analyses were performed using chi-square, Pearson correlations, logistic regression, and multiple linear regression.

RT-induced symptoms developed in 21 patients. In the overall group, the single best predictor for the development of symptoms was the NTCP (p < 0.05). Pre-RT PFTs alone were less predictive (p = 0.1 for FEV1, p = 0.08 for DLCO). A multivariate model based on pre-RT DLCO and CT-based NTCP was strongly predictive for the development of symptoms (p < 0.001). NTCPs based on SPECT-derived DVf Hs and TGF-beta1 levels did not appear to provide additional predictive value. The presence or absence of pulmonary symptoms was correlated with the decline in PFT 6 months following RT (p < 0.05). In the overall group, the degree of decline in PFTs was not well correlated with any of the dose-volume variables considered. In patients with "good" pre-RT PFTs, there was a relationship between the percent reduction in PFT and dose-volume parameters such as the percent of lung volume receiving > 30 Gy (p < 0.05).

The extent of alteration in whole-lung function (symptoms or PFT changes) appears to be related to both dose-volume and pre-RT PFT parameters. The data suggest that no one variable is likely to be an adequate predictor and that multivariate predictive models will be needed. Additional studies are underway to develop better predictive models that consider physical factors such as the DVH and regional perfusion, as well as biological/clinical factors such as pre-RT PFTs and TGF-beta1.