Objective The study is to investigate the predictive values of dosimetric parameters and patient related factors in severe acute radiation pneumonitis (SARP) after concurrent chemoradiotherapy in non-small cell lung cancer (NSCLC). Methods In all, 147 NSCLC patients treated with concurrent chemotherapy and 3DCRT between 2006 and 2010 was collected. Independent sample t test was used to compare parameter values between patients with SARP and those without SARP. Logistic regression was used to identify significant determined factor. Predictive value of each parameter was tested by ROC analysis. Pearson correlation was used to analyze correlations between parameters. Represent factors were identified by factor analysis. Results The incidence of SARP was 9.5%(14/147). The means lung dose (MLD), V₂₀, V₃₀, V₄₀, and V₅₀(χ²=4.87－6.84,P=0.009－0.025, respectively) were determining factors for SARP. Our datasets shows that for SARP<5%, MLD, V₂₀, V₃₀, V₄₀ and V₅₀ should be ≤16.77 Gy, V₂₀≤34.15%,.V₃₀≤23.62%,.V₄₀≤18.57%, V₅₀≤13.02%. ROC analysis show that areas under MLD, V₂₀, V₃₀, V₄₀ and V₅₀ curves was corresponding to 0.678, 0.661, 0.667, 0.677, and 0.651, respectively. In addition, the sensitivity and specificity of each parameter at cutoff values are:78.0% and 48.1% for MLD;42.9% and 82.0% for V₂₀;78.6% and 52.9% for V₃₀;71.4% and 61.7% for V₄₀, and 57.1% and 67.7% for V₅₀. Factor analysis suggest that we can choose 1 or 2 parameters from MLD, V₂₀, or V₃₀, and another from V₄₀ or V₅₀ for predicting. The incidence of SARP was greater in patients withtumors in right lower lung than other locations (22.2% vs 6.7%,χ²=6.19,P=0.023). ConclusionsThe MLD, V₂₀, V₃₀, V₄₀ and V₅₀ are determining factors for SARP. As predictive value of each parameter alone is relatively week, using two or more parameters to predict SARP is recommended.