Background: We examined the effects of the rs1800469 and rs1800470 single nucleotide polymorphisms (SNPs) of the transforming growth factor-131 (TGFI31) gene and the rs189037 and rs373759 SNPs of the ataxia telangiectasia mutated (ATM) gene on the risk of radiation-induced pneumonitis (RP) in patients who underwent radiotherapy for various thoracic malignancies. Methods: We determined the genotype and allele distributions of rs1800469 (C-509T), rs1800470 (C869T), rs189037 (A-111G), and rs373759 (126713 G>A) in 141 Han Chinese patients who underwent definitive thoracic radiotherapy (50 to 77 Gy, 5 days/wk) for lung cancer (small cell or non-small cell tumors, n = 97), esophageal squamous cell carcinoma (ESCC, n = 27), or mediastinal cancer (n = 17). Clinical variables were evaluated using multivariate logistic regression models to calculate the relative risk of RP associated with the clinical variables, and a Pearson correlation analysis was used to evaluate the relationship between the SNP genotypes and alleles and the incidence of RP for the various risk factors. Results: The T alleles of rs1800470 (CT/TT) and rs1800469 (CT/TT) and the G allele of rs189037 (GA/GG) were associated with the risk of 2 grade-2 RP in the ESCC patients (P = 0.0006, P = 0.0127, and P = 0.0412, respectively), and that the A alleles of rs189037 (AG/AA) and rs373759 (AG/AA) were associated with the risk of grade-2 RP in the patients with mediastinal cancer (P = 0.0063 and P = 0.0003, respectively). None of the SNP genotypes were associated with the risk of RP in lung cancer patients. Conclusion: The T alleles of the rs1800470 (CT/TT) and rs1800469 (CT/TT) SNPs of TGFA/ and the G allele of the rs189037 (GA/GG) SNP of ATM are independent risk factors for RP in Chinese ESCC patients, and the A alleles of the rs189037 (AG/AA) and rs373759 (AG/AA) SNPs of ATM are independent risk factors for RP in Chinese patients with mediastinal cancer. These SNPs might represent useful biomarkers for personalizing radiotherapy regimens for Chinese patients with ESCC or mediastinal cancer to reduce the incidence of RP. Large-cohort studies of these SNPs in thoracic cancer patients are warranted.