Correlation of dosimetric factors and radiation pneumonitis incidence for non-small-cell lung cancer (NSCLC) patients in a recently completed dose escalation study

Abstract

To determine dosimetric factors for lung, lung subvolumes and heart that correlate with incidence of RTOG Grade 3 or higher radiation pneumonitis (G3RP) in all 78 evaluable patients from a completed Phase I dose escalation study of 3-dimensional conformal radiation treatment (3D-CRT) of NSCLC. Between 9/91 and 3/03, 104 inoperable NSCLC patients were enrolled in a phase I 3D-CRT dose escalation study at our institution. Dose levels began at 70.2 Gy and culminated at 90 Gy (Rosenzweig et al, IJROBP 2003;57,S417). 78 patients satisfied the criteria for our study: (1) treatments based on a single CT data set; (2) treatment plans restorable from electronic archive; (3) either developed G3RP within 6 months following treatment or survived that time without it. There were 10 instances of G3RP. Lungs (apex to diaphragm, excluding soft-tissue parts of GTV) and heart (inferior of 0.9–1 cm below the carina fork) were contoured. Also, the lungs were divided lengthwise into upper and lower halves. Dose-volume histograms (DVH) were generated for the total lung volume, ipsilateral (containing majority of PTV) and contralateral lungs, upper lung, lower lung and heart. Parameters evaluated for univariate correlation with G3RP using logistic regression included: mean dose, the Lyman-Kutcher-Burman (LKB) effective uniform dose (deff) and normal tissue complication probability (NTCP), the parallel model fraction damaged (fdam) parameter and Vd, the % structure volume receiving at least dose d for d = 5, 10, 13, and then 20–70 Gy in 10 Gy increments (V05, V10, etc). Univariate correlation of G3RP with heart mean and maximum doses and dose encompassing the hottest 5% of heart (D05) was evaluated via Fishers Exact Test. We compared this analysis of the lung subvolumes to our earlier study of the 49 patients treated to ≤81 Gy (Yorke et al, IJROBP 2002; 54, 329–339) and of V13 to the study of Seppenwoolde et al (IJROBP 2003; 55, 724–35). The most significant variables were Vd for d < 20 Gy in the ipsilateral lung: p-values for V5, V10 and V13 were between 0.005 and 0.006 (Vd was significant for d < 50 Gy). Ipsilateral lung p-values for fdam (which is increasingly sensitive to doses between 10–40 Gy) and V20 were 0.009. In contrast, p-values for the parameters which are more sensitive to volumes at prescription dose - mean dose, deff and NTCP- were 0.023, 0.029 and 0.149 respectively. Similarly, in total lung, Vd for d < 20 Gy were most significant: p-values of V5, V10 and V13 were between 0.018 and 0.023 (Vd was significant for d < 50 Gy); p-values for fdam and V20 were 0.019 and 0.035 respectively; p-values for mean dose, deff and NTCP were 0.029, 0.036 and 0.053 respectively. For the lower-lung, the pattern is slightly different. The most significant variables were V30, V40, and V50, with p-values of 0.021, 0.014, 0.024 respectively (Vd was significant for 5 Gy ≤ d < 60 Gy); p-values for fdam and V20 were 0.029 and 0.050 respectively; p-values for mean dose, deff and NTCP were 0.036, 0.045 and 0.101 respectively. In comparison with our earlier study, the p-values of all significant dosimetric variables except lower lung NTCP decreased and all variables for contralateral and upper lung remained insignificant. Of the heart variables, only D05 showed a trend towards significance (p = 0.09). This analysis confirms the correlation between V13 and severe pneumonitis reported by Seppenwoolde et al. In whole and ipsilateral lung, V05, V10 and V13 were more strongly correlated than the figures of merit for the lung in current clinical use (V20, mean dose, deff, LKB NTCP, and fdam) with the strongest correlation being observed for ipsilateral lung. This analysis also strengthens the correlation between G3RP and V20, mean dose, deff, and fdam in the whole, ipsilateral and lower lung that were observed in our earlier study.