PURPOSE: To identify dosimetric factors that predict development of radiation pneumonitis (RP) following stereotactic or hypofractionated radiotherapy for lung tumors. METHODS: Seventy-nine consecutive patients with either a planning target volume (PTV)>100 cm(3) (n=69) or prior pneumonectomy or bi-lobectomy (n=13) were identified. Radiation doses (range: 5-50 Gy, with 5 Gy increments) were converted to equivalent doses (EQD(2 Gy)) (α/β=3). Total lung (TL), ipsilateral (IL) and contralateral lung (CL) volumes minus PTV, receiving 5 Gy (V5) up to 50 Gy (V50) and mean lung dose (MLD) were analyzed. Predictors of grade ≥3 RP (CTCAEv4.03) were identified with concordance-statistics (C-statistic) and p-values used to quantify the performance of the model. Factors found to be significant were entered into a recursive partitioning analysis (RPA). RESULTS: Median PTV was 150 cm(3). Grade ≥3 RP was observed in 8 patients (10%). In univariable analysis, CL-MLD, CL-V5-15, TL-MLD, TL-V5-V10 and ITV size were predictive of RP (p<0.05). In multivariable analysis, contralateral MLD (p=.007) and ITV (p=.063) were the strongest predictors of grade ≥3 RP, with excellent discrimination (C-statistic: 0.868). CONCLUSION: Contralateral MLD and ITV size are both strong predictors of grade ≥3 RP post treatment. Planning constraints should aim to keep contralateral MLD below 3.6 Gy.
PURPOSE: To identify dosimetric factors that predict development of radiation pneumonitis (RP) following stereotactic or hypofractionated radiotherapy for lung tumors. METHODS: Seventy-nine consecutive patients with either a planning target volume (PTV)>100 cm(3) (n=69) or prior pneumonectomy or bi-lobectomy (n=13) were identified. Radiation doses (range: 5-50 Gy, with 5 Gy increments) were converted to equivalent doses (EQD(2 Gy)) (α/β=3). Total lung (TL), ipsilateral (IL) and contralateral lung (CL) volumes minus PTV, receiving 5 Gy (V5) up to 50 Gy (V50) and mean lung dose (MLD) were analyzed. Predictors of grade ≥3 RP (CTCAEv4.03) were identified with concordance-statistics (C-statistic) and p-values used to quantify the performance of the model. Factors found to be significant were entered into a recursive partitioning analysis (RPA). RESULTS: Median PTV was 150 cm(3). Grade ≥3 RP was observed in 8 patients (10%). In univariable analysis, CL-MLD, CL-V5-15, TL-MLD, TL-V5-V10 and ITV size were predictive of RP (p<0.05). In multivariable analysis, contralateral MLD (p=.007) and ITV (p=.063) were the strongest predictors of grade ≥3 RP, with excellent discrimination (C-statistic: 0.868). CONCLUSION: Contralateral MLD and ITV size are both strong predictors of grade ≥3 RP post treatment. Planning constraints should aim to keep contralateral MLD below 3.6 Gy.