Plasma Levels of IL-8 and TGF-β1 Predict Radiation-Induced Lung Toxicity in Non-Small Cell Lung Cancer: A Validation Study

Abstract Purpose/Objectives We previously reported that combination of mean lung dose (MLD), and inflammatory cytokines (IL-8 and TGF-β1) may provide a more accurate model for radiation-induced lung toxicity (RILT) prediction in 58 patients with non-small cell lung cancer (NSCLC). This study is to validate the previous findings with new patients and explore new models with more cytokines. Materials/Methods 142 patients with stage I-III NSCLC treated with definitive radiation therapy (RT) from prospective studies were included. Sixty-five new patients were used to validate previous findings, and all 142 patients to explore new models. Thirty inflammatory cytokines were measured in plasma samples before RT, 2 weeks and 4 weeks during RT (pre, 2w, 4w). Grade ≥2 RILT defined as grade 2 and higher radiation pneumonitis or symptomatic pulmonary fibrosis was the primary endpoint. Logistic regression was performed to evaluate the risk factors of RILT. The area under the curve (AUC) for the Receiver Operating Characteristic (ROC) curves was used for model assessment. Results Sixteen of 65 patients (24.6%) developed RILT2. Lower pre IL-8 and higher TGF-β1 2w/pre ratio were associated with higher risk of RILT2. The AUC increased to 0.73 by combining MLD, pre IL-8 and TGF-β1 2w/pre ratio compared with 0.61 by MLD alone to predict RILT. In all 142 patients, 29 patients (20.4%) developed grade ≥2 RILT. Among the 30 cytokines measured, only IL-8 and TGF-β1 were significantly associated with the risk of RILT2. MLD, pre IL-8 level and TGF-β1 2w/pre ratio were included in the final predictive model. The AUC increased to 0.76 by combining MLD, pre IL-8 and TGF-β1 2w/pre ratio compared with 0.62 by MLD alone. Conclusions We validated that a combination of mean lung dose, pre IL-8 level and TGF-β1 2w/pre ratio provided a more accurate model to predict the risk of RILT2 compared to MLD alone.