Repeated Stereotactic Body Radiotherapy With CyberKnife ® in Lung Cancer: Toxicity Can Be Reduced by Sparing Irradiation to the Lungs

Background: Data of multiple courses of stereotactic body radiotherapy (SBRT) for lung malignancies is scarce. We aimed to investigate the ecacy and safety of repeated SBRT with CyberKnife and to identify predictors of lung toxicity. Methods: We reviewed patients with primary or metastatic lung cancer who underwent repeated SBRT with CyberKnife® between December 2013 and April 2021. Local control (LC), overall survival (OS) and adverse events were analyzed. We generated a composite plan and investigated the mean lung dose (MLD), V5–V30 of the bilateral lungs, and absolute lung volumes spared from 5–30 Gy (VS5–VS30), based on equivalent dose in 2.0-Gy fractions (α / β = 3). Clinical and dosimetric predictors of Grade 2+ radiation pneumonitis (RP) were examined. Results: Twenty patients underwent repeated SBRT for 27 lesions. The most predominant dose fractionation was 60 Gy in ve fractions. The median follow-up duration from the last SBRT was 18.0 months (range, 3–55 months). The 1-year and 2-year LC of lung lesions in repeated courses were 95.2% and 95.2%, respectively. The 1-year and 2-year OS rates after repeated SBRT were 88.4% and 49.7%, respectively. Five patients (25%) developed Grade 2+ RP, including a Grade 5 RP. There was no association between clinical factors and Grade 2+ RP. The Grade 2+ RP group showed higher composite MLD (p = 0.042) and lower composite VS5 (p = 0.042), VS10 (p = 0.015), VS15 (p = 0.008), and VS20 (p = 0.025), compared with the Grade 0–1 RP group. Conclusions: Repeated SBRT with CyberKnife® showed favorable LC and OS, but a high rate of Grade 2+ RP. Analysis of the composite plans showed that MLD and VS5–VS20 may predict lung toxicity of multicourse SBRT. A p-value indicates a statistically signicant association.


Introduction
Stereotactic body radiotherapy (SBRT) for early-stage primary lung cancers and solitary lung metastases has shown excellent clinical results [1][2][3] and become established as a standard treatment option.
However, local in-eld recurrence and metachronous lung metastasis remain the common patterns of failure [1,2]. It is also known that primary lung cancer patients have up to 10% probability of developing a secondary lung tumor within 5 years after treatment [4]. For these reasons, a signi cant number of patients with primary or metastatic lung cancers require repeated thoracic radiotherapy. However, patients who have previously received SBRT are often ineligible for surgery due to comorbidities. Thus, the demand for repeated thoracic irradiation is high.
Technological advancements in emerging approaches to SBRT enable repeated thoracic radiotherapy. Precise conformal dose distribution enables high dose delivery to the target, while sparing normal organs.
Recent advancements in SBRT, such as volumetric modulated arc radiotherapy and robotic radiotherapy devices have led to more precise treatment, compared with conventional SBRT. It has been suggested that several organs, including the lungs, is tolerant to additional irradiation after recovery from initial radiation damage [5,6]. The feasibility of repeated SBRT is now being increasingly recognized. Several small, retrospective case series have illustrated the potential clinical utility of repeated SBRT [7][8][9][10]. However, optimal methods for repeated SBRT have not yet been established. Concerns about increased adverse events with subsequent irradiations have been raised by clinical data [11][12][13]. Furthermore, clinical and dosimetric factors that may predict lung toxicity are poorly understood in the situation of multicourse thoracic irradiation because of limited reports. Developing effective and safe procedures, and identifying indicators to avoid clinically signi cant toxicity, are important issues that mandate further study.
Our study aimed to investigate the clinical outcomes and dosimetric parameters of patients with primary or metastatic lung cancers who underwent repeated SBRT with CyberKnife® (Accuray, Sunnyvale, CA, USA), and to evaluate its e cacy and safety, along with predictors of lung toxicity.

Patient selection
After obtaining institutional review board approval (Approval No. ERB-c-1646), we investigated all patients with primary lung cancer or lung oligometastasis who underwent a second or third course of SBRT with CyberKnife® from October 2014 to December 2020. Patients who received their rst SBRT at another institution were also included. The eligibility criterion for primary lung cancers was "clinical T1-2N0M0," according to the Union for International Cancer Control, the eighth edition. For lung oligometastases, we included patients who had control over their primary lesion and extrapulmonary lesions before SBRT. We included those with lung lesions limited from one to three at the time of each SBRT course. Patients were excluded if radiotherapy dosimetry data from any course of lung SBRT were not available for compositing dose volume histogram (DVH) analyses, if the intent of the repeat course of SBRT was palliative (equivalent dose in 2.0 Gy fractions [EQD2] < 60 Gy), or if conventional dose fractionations (< 3.0 Gy per fraction) were used in any radiotherapy course. The associated EQD2 was then calculated using the following linear quadratic model, with an α / β ratio of 10 for tumor and 3 for normal tissue: EQD = D (d + α / β) / (2 + d / [α / β]), where D equals the total prescribed dose (Gy) and d represents the dose per fraction (Gy).

Treatment techniques
We used the CyberKnife® G4 system to perform repeat SBRT, using the following procedure. Patients were immobilized using an individually shaped body cast (ESFORM, Engineering System, Matsumoto, Japan). Four-dimensional computed tomography (CT) scanning was performed to generate contours according to the respiratory motion. Gross tumor volume (GTV) was acquired in each respiratory phase. The internal target volume (ITV) was the summed GTVs of all phases; the planning target volume (PTV) was obtained by expanding the ITV by 5 mm in all directions. We selected a collimator size and optimized the beam weight to ensure high dose conformity for PTV. The planned dose was prescribed to GTV, covering 99% of the volume (D99). GTV at rest was used in the Monte Carlo algorithm and in normalization to avoid uncertainty of dose prescription for the lung density area. GTV D99 was required to be 70-90% isodose of the maximum dose.
Lung lesions were classi ed as centrally located if they were located within 2.0 cm of the proximal bronchial tree or if the PTV abutted the mediastinal or pericardial pleura [14]. Peripheral lung lesions were usually treated with 60 Gy in ve fractions. For central lung lesions, dose fractionation was determined on a case-by-case basis to avoid overdosing the adjacent critical organs. Doses to critical organs were within Timmerman's constraint [15]. For more than ve fractions of SBRT, the procedure was modi ed, based on EQD2. For ve fractions, patients were irradiated once every other day. For more than ve fractions, patients were irradiated over consecutive days. To assess SBRT intensity, the biological equivalent dose (BED) was calculated using the following linear quadratic model, with an α / β ratio of 10: BED = nd (1 + d / [α / β]), where n equals the total number of fractions and d represents the dose (Gy) per fraction. The Xsight spine tracking system (CIRS, Norfolk, VA, USA) was used during treatment for position veri cation.

Analysis of dosimetric parameters
The CT scans, structure sets, and dose distributions were sent to our institution's software platform for dosimetric analysis. Composite plans were generated from each SBRT plan for all patients. We investigated cumulative DVH parameters for the bilateral lungs, including mean lung dose (MLD), bilateral lung volumes V5, V10, V15, V20, and V30, and absolute lung volumes spared from 5 Gy (VS5), 10 Gy (VS10), 15 Gy (VS15), 20 Gy (VS20), and 30 Gy (VS30). The procedure for generating a cumulative composite plan was as follows. Two CT scans were aligned rigidly by using an automatic bone match (with translation and rotations). Given the differences in dose fractionation between SBRT plans, lung doses were corrected for EQD2 (α / β = 3). Finally, volumetric doses, converted to EQD2 for each plan, were summed on the CT images taken before the nal course of SBRT.

Patient follow-up
Clinical response was evaluated by physical examination and chest CT scans at 1, 3, 6, 9, 12, 18, and 24 months after SBRT, then every 6 months. Local failure was de ned as a 20% increase in the greatest diameter of GTV on two consecutive CT scans. If local recurrence was seen as equivocal at a given follow-up point, but later con rmed to be recurrence, the time of local failure was backdated to the time of the initial equivocal nding. Local control (LC) for second or third SBRT treatments was evaluated per lesion. The overall survival rate (OS) was calculated from the start date of the second SBRT course to the date of the last follow-up or death from any cause. Toxicities, such as radiation pneumonitis (RP), were recorded using the Common Terminology Criteria for Adverse Events version 5.0.

Statistical analysis
All statistical calculations were performed using SPSS version 26.0 software (IBM, Armonk, NY, USA). LC and OS were calculated using the Kaplan-Meier method. Comparative analysis was performed to identify clinical and dosimetric factors associated with the development of Grade 2+ RP. Between Grades 0-1 RP group and the Grade 2+ RP group, we compared percentage values using the chi-squared test. Discreet values were compared using the Mann-Whitney U test. All analyses used the p values of < 0.05 level of signi cance.

Patient and tumor characteristics
Twenty patients underwent repeated lung SBRT for 27 lesions. Second course SBRT were performed on 20 patients with 23 lesions, due to local recurrence after SBRT in three patients (three lesions), second primary lung cancer in six patients (six lesions), and the appearance of newly emerged lung metastases in 11 patients (14 lesions). After the second SBRT course, four patients received the third course SBRT for four lesions because of the appearance of newly emerged lung metastases. Patient and tumor characteristics are summarized in Table 1. Our study population included more metastatic lung cancers than primary lung cancers (65.0% vs. 35.0%, respectively).

Adverse events
Six patients (30.0%) experienced at least one Grade 2 or higher toxicity after repeat SBRT. One patient experienced Grade 2 chest wall pain, with Grade 2 rib fractures, which was likely attributable to SBRT.
Grade 2+ RP was observed in ve patients (25.0%), including one Grade 3 and one Grade 5. All Grade 2+ RP occurred after the last course of SBRT. Four patients had Grade 2+ RP after the second SBRT and one patient after the third SBRT. The median time between the last SBRT and the onset of RP was 4.0 months (range, 2-11 months). There was no case with Grade 2+ RP in the interval between initial SBRT and second SBRT.
One Grade 5 patient received second SBRT for a newly emerged primary lung cancer 28 months after initial SBRT. Notably, this patient had a slight interstitial shadow on pretreatment chest CT. The patient developed Grade 3 RP at 11 months after second SBRT. They were administered steroids but died due to an exacerbation of RP (Fig. 3).

Predictors for radiation pneumonitis
An analysis of clinical predictors for Grade 2+ RP are summarized in Table 3. The presence of a history of lung resection was the only factor to show a (weak) trend. No factors were signi cantly associated with the development of Grade 2+ RP. An analysis of the dosimetric predictors for Grade 2+ RP is summarized in Table 4. For the composite plan, MLD (p = 0.042), VS5 (p = 0.042), VS10 (p = 0.015), VS15 (p = 0.008), and VS20 (p = 0.025) in the Grade 2+ RP group were signi cantly lower than in the Grade 0-1 RP group. Fig. 4 shows the correlation between composite MLD, VS5, VS10, VS15, VS20, and the development of Grade 2+ RP.   Abbreviations: RP, radiation pneumonitis; MLD, mean dose of the bilateral lung; Vxx, the volume received by xx Gy of the bilateral lung; VSxx, bilateral lung volume spared from xx Gy.

Discussion
We investigated the clinical outcomes and dosimetric parameters of patients with primary or metastatic lung cancer who received multiple courses of SBRT, to evaluate its e cacy and safety, along with predictors of lung toxicity. The second and third courses of SBRT (performed using the CyberKnife®) showed favorable LC and OS, which suggests the effectiveness of such treatment. After repeated SBRT, Grade 2+ RP was frequently observed; a fatal case was experienced. We showed that the MLD and absolute lung volume spared from low-dose irradiation can predict lung toxicity by analyzing composite plans based on EQD2.
In a review by De Bari et al., the 1-year and 2-year LC rates after second thoracic SBRT for primary or metastatic cancer ranged between 59-95% and 50-92%, respectively [16]. A large case series reported that a second course of radiotherapy with 50 Gy in four fractions (BED10, 112.5 Gy) showed a 1-year LC of 95.0% [11]. We performed repeated SBRT (median BED10, 132.0 Gy) and demonstrated excellent LC, which indicated that CyberKnife® -based SBRT, as a repeated radical treatment, provides durable tumor control. The 1-year and 2-year OS rates following the last course of SBRT for patients with primary or metastatic cancer, have been reported to be in the range of 59-80% and 29-74%, respectively [16], which is comparable with our results.
Lung SBRT has been considered a safe treatment with minimal toxicity. For single course SBRT, the rates of Grade 2+ and Grade 3+ lung toxicity have been reported to be 9.1% (7.15-11.4%) and 1.8% (1.3-2.5%), respectively [17]. However, the safety of multicourse SBRT is not well understood. Several retrospective studies of thoracic re-irradiation have reported high rates of Grade 2+ and Grade 3+ RP, of 13.6-50.0% and 3.0-28.0%, respectively [7,10,12,16,18]. In our study, Grade 2+ and Grade 3+ pulmonary toxicity after repeated SBRT was observed in 25.0% and 8.0% of patients, respectively, which is consistent with previous reports. Our data also suggest a concern for increased pulmonary toxicity with multiple courses of lung SBRT, compared with single course SBRT.
We experienced a fatal RP that developed after repeated SBRT. Previous reports have identi ed few Grade 5 RPs in patients who received SBRT after previous conventional fractionated radiotherapy [11,13]. There are no previous reports of fatal RP following multicourse SBRT. Analysis of a large case series reported that Grade 5 RP after a single course of SBRT occurred very rarely (1.3%) [19]. In that study, interstitial changes were observed retrospectively in 73.7% of Grade 5 RP patients, indicating a potential risk factor for fatal pulmonary toxicity. In our study, Grade 5 RP occurred in one of three patients with interstitial shadow; it was not observed in patients without interstitial shadow (33.3% vs. 0.0%; p = 0.015). Even in the context of multicourse SBRT, interstitial changes may be a risk factor for fatal toxicity.
Several risk factors for pulmonary toxicity in repeated thoracic radiotherapy have been identi ed. Out-ofeld relapse [7], short interval between initial radiotherapy and second SBRT [18], poor performance status, impaired lung function, and initial PTV location in the bilateral mediastinum [11]  In previous SBRT studies, lung dose has been assessed using irradiated relative volume (%). Our report is the rst to show a signi cant association between absolute lung volume spared from low-dose irradiation and SBRT-induced lung toxicity. A correlation between decreased VS5 (<1,500 cc) and the development of Grade 3+ RP in conventional fractionated irradiation for primary lung cancer has been previously shown by Tsujino et al. [20]. Our entire cohort of Grade 2+ RP had VS5 < 1,500 cc; reduced VS5 correlated with the development of Grade 2+ RP. However, among the DVH parameters we examined, VS15 was the most strongly associated with RP. Due to SBRT's unique fractionation scheme and dose distribution, the DVH parameter that predicts RP in SBRT may differ from that of conventional fractionated radiotherapy.
This study has several limitations. Due to it being a retrospective analysis at a single institute, there are inherent biases, a small number of events, and an inhomogeneous study population. This may limit the power of the analyses and cause under-reporting of adverse events. Also, we included patients with primary lung cancer and with metastatic lung cancer, which makes it di cult to draw reliable conclusions about survival. Prospective studies with larger populations are needed for more precise assessment. Moreover, deformable image registration (DIR) was not available for generating the composite plans of our study. DIR is an image processing technique that can provide a more accurate assessment of the cumulative radiation doses to the lungs by accounting for anatomical changes [21,22].

Conclusions
Lung SBRT using CyberKnife® was effective as a repeated treatment option for primary or metastatic lung cancer. High rates of lung toxicity were observed, including fatal cases. By analyzing composite plans, we showed that MLD and absolute lung volume spared from low-dose irradiation may be useful in predicting lung toxicity. Computed tomography images from a 78-year-old man who developed fatal RP after a second course of SBRT. The patient had a 50 pack/year smoking history but quit before the initial SBRT; his respiratory status was satisfactory. A slight interstitial shadow was observed at the base of lung (arrow) on the chest CT taken before the second SBRT. The primary tumor in the left upper lobe, which had been treated 28 months before the second SBRT, represented scar-like brosis (arrowhead). Eleven months after the second SBRT for newly emerged primary lung cancer in the right upper lobe, the patient developed RP with hypoxemia. In-hospital pulse steroid therapy resolved the symptoms and pulmonary shadow, but a relapse of RP occurred during steroid tapering, which resulted in the appearance of diffuse interstitial in ltration and severe hypoxemia. No speci c infectious cause was identi ed. The patients died from respiratory failure 13 months after the second SBRT.