Influence of respiratory motion management technique on radiation pneumonitis risk with robotic stereotactic body radiation therapy

Abstract Purpose/Objectives For lung stereotactic body radiation therapy (SBRT), real‐time tumor tracking (RTT) allows for less radiation to normal lung compared to the internal target volume (ITV) method of respiratory motion management. To quantify the advantage of RTT, we examined the difference in radiation pneumonitis risk between these two techniques using a normal tissue complication probability (NTCP) model. Materials/Method 20 lung SBRT treatment plans using RTT were replanned with the ITV method using respiratory motion information from a 4D‐CT image acquired at the original simulation. Risk of symptomatic radiation pneumonitis was calculated for both plans using a previously derived NTCP model. Features available before treatment planning that identified significant increase in NTCP with ITV versus RTT plans were identified. Results Prescription dose to the planning target volume (PTV) ranged from 22 to 60 Gy in 1–5 fractions. The median tumor diameter was 3.5 cm (range 2.1–5.5 cm) with a median volume of 14.5 mL (range 3.6–59.9 mL). The median increase in PTV volume from RTT to ITV plans was 17.1 mL (range 3.5–72.4 mL), and the median increase in PTV/lung volume ratio was 0.46% (range 0.13–1.98%). Mean lung dose and percentage dose–volumes were significantly higher in ITV plans at all levels tested. The median NTCP was 5.1% for RTT plans and 8.9% for ITV plans, with a median difference of 1.9% (range 0.4–25.5%, pairwise P < 0.001). Increases in NTCP between plans were best predicted by increases in PTV volume and PTV/lung volume ratio. Conclusions The use of RTT decreased the risk of radiation pneumonitis in all plans. However, for most patients the risk reduction was minimal. Differences in plan PTV volume and PTV/lung volume ratio may identify patients who would benefit from RTT technique before completing treatment planning.


| INTRODUCTION
"Stereotactic body radiation therapy" (SBRT) or "stereotactic ablative body radiotherapy" (SABR) refers to highly spatially precise radiation therapy with steep dose gradients delivered to an extracranial target, typically completed in 1-5 fractions with higher doses per fraction than conventional radiation therapy. For lung tumors, SBRT has emerged as an effective treatment technique for early stage lung primary malignancies as well as lung oligometastases with outcomes comparable to surgical resection. 1,2 Lung SBRT has the particular challenge of tumor respiratory motion. Some techniques for managing this include accounting for motion within target volumes, temporarily reducing motion by breath hold or abdominal compression, gating dose delivery by respiratory phase, or tracking tumor motion during treatment using fluoroscopy. 3 One of the most common techniques requiring no breath control or imaging during treatment is expanding the target volume to the entire range of tumor motion across the respiratory cycle, known as the internal target volume (ITV). This is delineated using CT scans obtained at maximum inhalation and exhalation, or a set of scans acquired through the course of the respiratory cycle (4D-CT).
ITV technique has the disadvantage of including more normal lung tissue in the target volume, exposing it to high radiation dose. It also cannot account for unpredicted variations in tumor motion and respiratory pattern during treatment. Compensation may be achieved with larger planning target volume (PTV) margins, but this further increases the dose to normal lung tissue. 4 To overcome these issues, real-time tracking (RTT) techniques were developed such as the CyberKnife robotic SBRT system with Synchrony motion management (Accuray Inc., Sunnyvale CA, USA). 5 With this system, tumor motion is tracked during treatment while the patient breathes freely without coaching. A correlation model is built between orthogonal x-ray images acquired every 60-120 s and the positions of light emitting diodes on the patient's chest obtained by infrared camera at 26 Hz. The model is continuously updated and used to move a linear accelerator mounted on a robotic arm, anticipating target location with high accuracy. 6 The original Synchrony system required invasive placement of gold fiducial markers near the tumor for accurate targeting, associated with risks of pneumothorax, bleeding, and additional treatment cost. 7 The subsequently developed XSight Lung Tracking system (Accuray Inc.) allows tracking based on imaging the tumor itself, obviating the need for fiducial marker placement. 8,9 However, this technique is limited to larger and denser tumors that have adequate x-ray contrast with normal tissue. 10 These motion management systems were primarily designed to optimize target coverage. Less is known about their effects on normal tissue doses, and complications have not been directly compared. The most common adverse effects of lung SBRT are due to radiation of the lung parenchyma, which incites a complex reaction leading to depletion of alveolar pneumocytes, interstitial infiltration of immune cells, and fibroblast proliferation. 11 This manifests pathologically as a continuum from subacute radiation pneumonitis to late pulmonary fibrosis, which may be clinically symptomatic. 12 The reported incidence for symptomatic radiation pneumonitis requiring treatment after SBRT ranges from 5% to 30%. [13][14][15][16][17][18][19][20][21][22] This typically presents as a syndrome of dyspnea, cough, and low-grade fever within 12 weeks of the completion of radiation therapy. Symptoms usually resolve with corticosteroids, although permanent pulmonary dysfunction can occur. 12 The lungs function as a classic radiobiological "parallel organ" 23 and symptomatic radiation pneumonitis is generally correlated with critical dose-volumes rather than maximum dose to lung tissue. 14,18,19,22 Normal tissue complication probability (NTCP) models based on mean lung dose have been derived to predict risk of symptomatic radiation pneumonitis from lung SBRT. 16,17,24 At our institution, it is standard for all lung SBRT patients to undergo respiratory 4D-CT imaging at simulation regardless of the intended motion management technique. This allows assessment of target and organ motion and provides a backup method if fiducialor tumor-based tracking fails. In the current study, patients originally treated using RTT were replanned using ITV volume expansion technique. Comparisons of planning parameters, lung dose-volumes, and radiation pneumonitis NTCP from a previously derived model were conducted. Classifications based on predosimetric characteristics were performed to generate practical guidelines predicting which patients could substantially benefit from real-time tracking, without needing to perform time-consuming replanning and NTCP calculation.

| MATERIALS AND METHODS
A single institution database was used to retrospectively identify 20 SBRT treatment plans for primary or oligometastatic lung tumors. All treatments were performed on the CyberKnife system with XSight Lung Tracking RTT technique. The original treatment plans were designed to deliver the prescription dose to a planning target volume (PTV RTT ), which was defined by an isotropic expansion from a gross tumor volume (GTV) contoured on the simulation CT. The PTV RTT margin expansion was determined by the treating physician at the time of planning. Originally delivered treatment plans were used for comparisons without modification.  following the movement of the original GTV as much as possible. This was performed by the original treating physician at the time of initial planning if available (2 plans), or by a single physician (C.C.) retrospectively if not available. A new planning target volume (PTV ITV ) was created using 5 9 598 mm (LR 9 AP 9 SI) expansion from the ITV, a previously determined appropriate planning margin for this technique. 4 New treatment plans targeting the PTV ITV were generated using the original prescription doses. Plans were designed using Accuray Multiplan software v.5.2.0. Tissue heterogeneity correction was performed by Monte Carlo algorithm with 1% uncertainty. Planning goals were for prescription dose to cover ≥95% of PTV ITV , and normal tissue doses to be as low as achievable while maintaining target coverage. Doses to organs at risk (spinal cord, heart, esophagus, rib/ chest wall) had to meet TG-101 constraints 25 except for rib/chest wall, which was kept as low as possible while maintaining target coverage. Prescription isodose line ≥60% of maximum dose was used if able to maintain target coverage and normal tissue dose constraints.
The PTV volumes, prescription target coverage, prescription isodose line, conformity index, 26 total MU, estimated delivery time per fraction, and doses to organs at risk were recorded. Figure 1 depicts example RTT and ITV-based plans from a single patient.
The volume of bilateral lungs excluding GTV was defined for each plan. Dose-volume histograms for the bilateral lung volumes were extracted with bin size 0.1 Gy. Doses were converted to the linear-quadratic equivalent dose in 2 Gy fractions (EQD2) using a/ b = 3. 27 After conversion, the mean dose to bilateral lungs was recorded, as well as the following dose volume percentages based on previous publications identifying correlations to radiation pneumonitis risk: V2.5 Gy, V5 Gy, V10 Gy, V13 Gy, V20 Gy, V30 Gy, V40 Gy, and V50 Gy. 16,17,24,28 The normal tissue complication probability (NTCP) for symptomatic radiation pneumonitis requiring treatment (grade ≥2 by NCI-CTCAE v.4) was calculated from a previously published model using the Lyman-Kutcher-Burman formula based on bilateral lung mean dose using TD50 = 20.8 Gy, m = 0.45. 17 For patients with multiple tumors, each tumor was planned separately and doses were considered independently.
Statistical analysis was performed using non-parametric tests in R v.3.3.2 (R Foundation for Statistical Computing). 29 To compare treatment plan dosimetry, pairwise Wilcoxon rank-sum tests were used. To correlate differences in NTCP to tumor and treatment features, Spearman's rank correlation coefficient was used. All statistical tests were two-sided with a significance threshold of P ≤ 0.05. To create a practical guideline to identify patients before dosimetric calculation who would have a meaningfully increased risk of radiation pneumonitis with ITV versus RTT planning, an increase in NTCP of >5% was designated as "clinically significant". This threshold was chosen by agreement that this was the highest increase in NTCP that would be accepted for plans to be considered clinically equivalent. Receiver operating characteristic (ROC) analysis was used to compare the ability of features available before dosimetric calculation ("pre-dosimetric variables") to predict which patients would have these increases in NTCP.

3.A | Plan characteristics and NTCP
We identified 20 RTT plans delivered to 18 patients (two patients had two tumors treated by separate plans). Twelve patients were male and the median age was 73 yr (range 26-95 yr). Fourteen treatments were for lung primary tumors and six for metastatic tumors. Table 1 details the tumor locations, size, and extent of motion, as well as prescription doses (delivered and EQD2) and PTV RTT margins. Two tumors were considered central by RTOG 0813 criteria (within 2 cm of the proximal bronchial tree). 30 Comparisons of RTT and ITV plans are in Table 2. As expected, the PTV volume was significantly larger in ITV than RTT plans (median difference 17.1 mL, range 3.5-72.4 mL, P < 0.001) and the ratio of the PTV volume to total bilateral lung volume (PTV/lung) was also larger (median difference 0.46%, range 0.13-1.98%, P < 0.001).
There were no significant differences in PTV prescription coverage, prescription isodose percentages, or conformity indices. There were also no significant differences in maximum doses to the organs at risk other than the lungs (Table S1).
The mean lung dose was significantly higher for ITV plans, with a median increase in 1.95 Gy (range 0.22-7.37 Gy, P < 0.001). Lung dose-volume percentages were also significantly higher at every level tested, with greater increases at the lower dose-volumes ( Table 2, Example plan comparison with coronal slices of treatment planning CT for RTT plan (a) and ITV plan (b). GTV is shaded blue, ITV is shaded purple, and PTVs are shaded red. From RTT to ITV based-plan, PTV volume increased 36.4 mL, PTV/lung volume ratio increased 0.69%, and NTCP increased 4.1%.

3.B | Pre-dosimetric variable correlations to increase in NTCP
Correlations between predosimetric variables and the increase in NTCP between RTT and ITV plans are in Table 3. Increase in NTCP was most strongly correlated with the increase in PTV/lung volume ratio from RTT to ITV plan (q = 0.79, P < 0.001), and the increase in PTV volume from RTT to ITV plan (q = 0.74, P < 0.001). High correlations were also seen for PTV ITV /lung volume ratio (q = 0.66, P = 0.002), and GTV greatest axial diameter (q = 0.63, P = 0.003).
Statistically significant but weaker correlations were also seen for other measures related to tumor size and motion such as GTV volume, ITV volume, ITV À GTV volume difference, and PTV volumes.
Prescription dose, lung volume alone, and measures of tumor linear motion alone were not statistically significantly correlated with increase in NTCP between RTT and ITV plans.

3.C | Sensitivity analysis
The size of the PTV RTT margin used in the original plans was correlated with tumor size and was potentially a confounder. A sensitivity analysis was performed limiting the comparisons to the 14 plans with an original 5 mm PTV RTT margin, which tended to be used for smaller tumors. The median GTV was 12.9 mL for tumors with 5 mm PTV RTT margin versus for 14.5 mL for all tumors (Tables   S2-S4). The increase in NTCP from RTT to ITV plans remained significant, with a median increase in 1.2% (P < 0.001). The increase in PTV volume and PTV/lung volume ratio from RTT to ITV plans also remained significantly correlated with increase in NTCP (increase in PTV volume: q = 0.57, P = 0.03; increase in PTV/lung volume ratio:

3.D | ROC analysis
For the purposes of identifying tumors that would derive a clinically meaningful reduction in radiation pneumonitis risk with an RTT plan versus an ITV-based plan, a >5% difference in NTCP was designated as "clinically significant". Of the 20 plans, five had clinically significant increases in NTCP using ITV versus RTT. ROC analysis was used to identify predictive thresholds in the four predosimetric variables that were most highly correlated with change in NTCP from RTT to ITV plans (difference in PTV/lung volume ratio, difference in PTV volume, PTV ITV /lung volume ratio, and GTV greatest axial diameter). ROC curves are depicted in Fig. 3.
Increase in PTV/lung volume ratio from RTT to ITV plan was the most accurate predictor of clinically significant increase in NTCP, with ROC area under curve (AUC) of 100%. The most sensitive threshold was an increase in PTV/lung volume ratio of 0.973% [sensitivity 100%, specificity 100%; Fig. 4(a)]. The next best predictor was the increase in PTV volume from RTT to ITV plan, with AUC 95%. The most sensitive threshold was an increase in PTV

| DISCUSSION
In this series of 20 lung SBRT plans using a real-time tracking (RTT) technique for respiratory motion management, we found that <0.5%. NTCP was not calculated, but the expected risk reduction would be small. However, the mean lung doses in that study were also low regardless of planning technique, approximately 3.0-3.5 Gy versus a median of 5.5 Gy for RTT plans and 8.2 Gy for ITV plans in the current study.
These comparisons highlight some of the limitations of this work.
First, the influence of tumor size on radiation pneumonitis risk makes the results dependent on the studied population. We limited our study to patients with tumors visible on orthogonal X rays using XSight Lung Tracking technique, thus having on average larger tumors that would benefit more from RTT. 10  would have contoured. The RTT and ITV plans appeared to be similar quality as measured by target coverage, conformity, and heterogeneity (  (Table S1). All planning constraints were met, and some ITV-based plans had even lower maximum organ doses than the matching RTT plans, likely due to differences in planning techniques and optimization goals. The lack of significant differences across the group is also due to tumors being in various locations throughout the lungs. However, as can be seen from the conditions also must be considered. Lung SBRT for oligometastatic disease may be given for prophylaxis or palliation of symptoms, and treatment-related toxicities may be considered less acceptable from a risk-benefit perspective. Oligometastatic patients are also more likely to receive radiation to multiple lung tumors increasing total lung dose, and to receive systemic agents that may also increase pneumonitis risk. Two patients in this study had multiple tumors treated using SBRT, however, this was not taken into account when calculating radiation pneumonitis risk and may be expected to raise it substantially.
Thus, while also considering these other factors, the capability to identify patients who would benefit from RTT is valuable. If they would benefit, clinicians may recommend attempts at XSight lung tracking, fiducial implantation, or other strategies for motion management. If they would not benefit, clinicians could opt for ITVbased treatment either on the Cyberknife or a conventional linear accelerator. In the present study we both quantify the range of expected differences in radiation pneumonitis risk between motion management techniques and propose some variables that may predict this benefit without requiring dosimetric calculation.

| CONCLUSION
The use of a real-time tumor tracking technique decreased radiation pneumonitis risk for all patients, although for most this absolute risk reduction was small (<5%). Difference between plans in the ratio of the target volume to total lung volume was the best predosimetric predictor of whether patients would derive a clinically significant CHAPMAN ET AL.

ACKNOWLEDGMENTS
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

CONF LICTS OF INTEREST
None.

SUPPORTING IN FORMATION
Additional Supporting Information may be found online in the supporting information section at the end of the article.