Association of Single Nucleotide Polymorphisms in Transforming Growth Factor-β1 Pathway and Risk of Radiation Pneumonitis in Lung Cancer Patients Treated with Thoracic Radiation Therapy

Abstract

Several clinical and dosimetric parameters are known to be associated with radiation-induced lung toxicity, including radiation pneumonitis (RP). Also, single nucleotide polymorphisms (SNPs) of genes in TGF-β1 pathway have a notable association with the RP. However, studies on the predictive value of SNPs for RP are still limited. Herein, we tried to develop a novel integrated predictive model for severe RP in lung cancer patients. A total of 59 patients who were treated with definitive or preoperative radiotherapy for primary lung cancer and had DNA samples were included. Potentially functional and tagging SNPs of TGF-β1 (rs1800469, rs1800471, rs1982073, and rs11466345), BMP2 (rs235768, rs3178250, rs1979855, and rs170986), and BMP4 (rs17563, rs4898820, and rs762642) were genotyped. Logistic regression was performed to build severe (grade ≥2) RP prediction models, and best subset selection algorithm with L0 and L2 regulations was used for variable selection. Only clinical/dosimetric variables were evaluated in model 1, and those variables as well as SNPs were included in model 2. Using beta coefficient obtained by the logistic regression, a scoring system was also developed. With median follow-up of 39.7 months, severe RP occurred in 20.3% of patients. In model 1, age (>66) and PTV volume (≥300 cc) were significant factors (p = 0.016, OR 8.820 [95% CI, 1.730-63.800]; and p = 0.024, OR 7.440 [95% CI, 1.460-52.100], respectively). In model 2, the above two factors (p = 0.010, OR 16.200 [95% CI, 2.440-187.000] and p = 0.025, OR 10.100 [95% CI, 1.610-105.000]) and the AG/GG genotype in BMP2 rs1979855 were significant factors (p = 0.031, OR 7.260 [95% CI, 1.380-59.100]). The AUC was significantly higher in model 2 than in model 1 (0.822 vs. 0.741, p = 0.029). According to developed scoring system, patients with a score >2.8 are more likely to experience severe RP (AUC 0.829). BMP2 rs1979855 could serve as a reliable biomarker for predicting RP while significantly improving predictive power compared to when only clinical factors were used.