The risk and predictors for severe radiation pneumonitis in lung cancer patients treated with thoracic reirradiation

Lung cancer is the most prevalent malignancies and the leading cause of cancer death worldwide. Radiotherapy has an important role in the radical treatment of early and advanced lung cancer. However, the rates of recurrent disease after RT still remain high [1, 2]. For recurrent pulmonary tumors after previous thoracic RT, salvage surgery is typically avoided. Second line chemotherapy for such patients has been assessed in several studies, unfortunately the objective response rates and survival are quite poor [3‐5]. With technologic advances, there is growing interest in the use of reirradiation (re-RT) for such diseases. Published studies have demonstrated the effectiveness of re-RT for recurrent, metastatic pulmonary tumors, or a new primary lung tumor after previous thoracic RT [6‐14]. However, radiation pneumonitis (RP) remains to be the most common side effect from retreatment, up to 40% in some cases [6, 12, 14].

Data for dose-volume predictors of severe RP in patients undergoing thoracic re-RT are scarce. V20 ≥ 30% of the composite plans was reported to be associated with severe RP in one small retrospective study [15]. However, deformable image registration (DIR) [16, 17] was not used to generate the composite plans in that study, leading to less accuracy for assessing the composite doses to lung. Besides, re-RT may cause some overlap with the previously irradiated lung volumes. Whether the overlap is related to the risk of RP remains unclear. Moreover, the effect of re-RT characteristics (such as re-RT technique, re-RT location, and Interval between initial and re-RT) on the incidence of RP has not been investigated sufficiently.

In light of these critical issues, we retrospectively investigated the risk and predictors of severe RP in patients undergoing thoracic re-RT.

Published data have suggested that either thoracic re-RT with conventional fractionation [8‐10, 19, 20] or with SBRT [5‐7, 11‐14] is a reasonable treatment option to consider for relapse or progression. However, there is a lack of prospective evidence detailing outcomes after thoracic re-RT. In the present study, 89.6% of patients were retreated with conventionally fractionated 3D-CRT/IMRT. Due to patients with two initial pathologic types (NSCLC and SCLC), with various clinical diagnosis at re-RT (including recurrent, metastatic, or a new primary lung tumor), and only 38.8% of patients with histological confirmation at re-RT, we did not assessed the effectiveness of re-RT.

The CTCAE 4.0 and South-West Oncology Group (SWOG) criteria are frequently used for RP grading in studies. The differences between CTCAE 4.0 and SWOG criteria are mainly in grade 1 and 2 RP. Grade 1 RP graded according to SWOG criteria would change to grade 2 RP if the CTCAE 4.0 criteria would have been applied. There is no obvious difference in grade 3, 4, or 5 RP between the two criteria. In the present study, RP was graded according to the CTCAE 4.0, and rates of grade ≥ 3 RP were 26.9%. Similar finding was also reported in several studies of re-RT with SBRT [6, 14, 15]. Liu et al. [15] analyzed 72 patients receiving re-irradiation with SBRT for recurrent disease and found that rates of severe RP were common (20.8%). Data from a retrospective study conducted at MD Anderson showed that 28% of re-RT patients developed grade 3 RP, and RP was the most common side effect [6]. De Bari et al. analyzed 12 studies on thoracic re-irradiation with SABR after a previous course of high-dose radiotherapy and found that the major toxicity is RP, occurring in around 20% of patients [21]. Ruysscher et al. summarised the available literature-based evidence, and stated that the incidence of severe lung toxic effects did not differ greatly in patients retreated with conventional 3D-CRT compared with patients retreated with SBRT [22]. We also compared the risk of grade ≥ 3 RP between re-RT with conventional fractionation and with SBRT, while no significant difference was observed.

As we know, there is only one retrospective study conducted by Liu et al. [15] that has investigated DVH variables related to RP from 3 sets of plans (initial RT, re-RT, and composite plans). In their study, a wide range of DVH parameters (including V10-V40 and MLD of 3 sets of plans) were assessed, but only a V20 ≥ 30% in the composite plan was significantly correlated with the incidence of grade 3–5 RP in multivariate analysis. In the present study, DIR was performed to create a composite plan, and the correlation between DVH variables from 3 sets of plans (including V5, V20, V50, and MLD) and the risk of grade ≥ 3 RP were evaluated. Different from Liu et al’s results, we found that initial-MLD and composite-V5 were significantly associated with grade ≥ 3 RP in multivariate analysis. One possible explanation for the difference was difference in re-RT technique used for re-RT. All of patients were retreated with SBRT in Liu et al’s study, while most of patients (88.1%) were retreated with conventional 3D-CRT/IMRT in the present study. The unique fractionation schemes and the dose distribution of SBRT might make it different from conventional RT for dose-volume predictors of severe RP. Secondly, DIR was not used in Liu et al’s study when generating the composite plans, which might result in less accurate assessment of the cumulative radiation doses to lung. DIR is an image processing technique, with the potential to account for anatomic changes, including changes in body weight, and early or late normal tissue radiation responses [16, 17]. Our results suggested that not only re-RT but also initial-RT and composite plans should be considered as important options for evaluating retreatment with thoracic RT.

One previous study has shown that patients retreated with RT for in-field relapses experienced lower rates of RP than did those outside the initial radiation field (0% vs. 28%, P = 0.03) [6]. For reason of the difference, the authors hypothesized that previously irradiated areas have already undergone fibrosis and are less susceptible to radiation-induced inflammation. In the present study, relatively lower rates of grade ≥ 3 RP for in-field failures compared with out-of-field failures were observed (18.2% vs. 35.3%), but with no significant statistical difference (P = 0.114). Interestingly, we also found that in-field failures were more frequently observed in patients with higher overlap-V5/re-V5 (0.8–1) and out-of-field failures in patients with lower overlap-V5/re-V5 (0–0.8); and linear-by-linear association testing suggested a significant correlation between location of recurrent tumors (in-field/out-of-field) and overlap-V5/re-V5 (0.8–1/0–0.8) (P < 0.001). Moreover, we found that overlap-V5/re-V5 was an independent predictor of grade ≥ 3 RP in multivariate analysis. However, it was not that the larger overlap-V5/re-V5 was, the lower incidence of grade ≥ 3 RP became. Rates of grade ≥ 3 RP for patients with the largest overlap-V5/re-V5 of 0.8–1 were similar to patients with the smallest overlap-V5/re-V5 of 0–0.4 (18.3% vs. 14.3%, P = 1.000), but were significantly lower when compared with patients with the medium overlap-V5/re-V5 of 0.4–0.8 (18.3% vs. 50%, P = 0.014). These findings suggested that the overlap of volume of lung exposed to low radiation doses between initial RT plans and re-RT plans was associated with the incidence of severe RP, and should be paid more attention when evaluating a plan of re- RT. Moreover, the findings also suggested an indirect association between location of recurrent tumors (in-field/out-of-field) and RP risk via the variable of overlap-V5/re-V5. Not all out-of-field relapses were associated with higher risk of severe RP compared with in-field relapses, and the correlation seemed to be determined by size of overlap-V5/re-V5.

ECOG PS score of 2–3, an FEV1 ≤ 65% were reported to be associated with higher RP in Liu et al.’s study [15]. In our study, none of patients were with ECOG PS score of 3, and there was no significant difference in RP risk between ECOG PS score of 0–1 and 2; the correlation between FEV1 and RP risk was not investigated because of lack of related data. Other potential risk factors of RP such as chemotherapy, the interval between initial and re-RT, location of re-RT (central/peripheral) were also assessed in this study, however, none of them were predictive for severe RP.

There were several limitations associated with this study. First, this is a retrospective, single-institution study, and biases could have been introduced which would affect the conclusions. The multivariate analysis can aid in decreasing these biases but not necessarily negate them entirely. Second, the sample size of this study was relatively small. This makes statistical analysis less reliable and raises the possibility that other important risk factors could be missed. Finally, a few patients who received very low doses of initial RT or re-RT were also included, which might lead to less accurate on the assessment of RP rates after re-RT.

The risk of grade ≥ 3 RP could be predicted by initial-MLD, composite-V5, and overlap-V5/re-V5. Out-of-field failures was associated with higher risk of severe RP compared with in-field failures in some cases, however, the correlation seemed to be determined by size of overlap-V5/re-V5.