Risk of target coverage loss for stereotactic body radiotherapy treatment of synchronous lung lesions via single‐isocenter volumetric modulated arc therapy

Abstract Treating multiple lung lesions synchronously via single‐isocenter volumetric modulated arc therapy (VMAT) stereotactic body radiation therapy (SBRT) improves treatment efficiency and patient compliance. However, aligning multiple lung tumors accurately on single pretreatment cone beam CTs (CBCTs) can be problematic. Tumors misaligned could lead to target coverage loss. To quantify this potential target coverage loss due to small, clinically realistic setup errors, a novel simulation method was developed. This method was used on 26 previously treated patients with two metastatic lung lesions. Patients were treated with 4D CT‐based, highly conformal noncoplanar VMAT plans (clinical VMAT) with 6MV‐flattening filter free (FFF) beam using AcurosXB dose calculation algorithm with heterogeneity corrections. A single isocenter was placed approximately between the lesions to improve patient convenience and clinic workflow. Average isocenter to tumor distance was 5.9 cm. Prescription dose was 54 Gy/50 Gy in 3/5 fractions. For comparison, a plan summation (simulated VMAT) was executed utilizing randomly simulated, clinically relevant setup errors, obtained from pretreatment setup, per treatment fraction, in Eclipse treatment planning system for each of the six degrees of freedom within ± 5.0 mm and ± 2°. Simulations yielded average deviations of 27.4% (up to 72% loss) (P < 0.001) from planned target coverage when treating multiple lung lesions using a single‐isocenter plan. The largest deviations from planned coverage and desired biological effective dose (BED10, with α/β = 10 Gy) were seen for the smallest targets (<10 cc), some of which received < 100 Gy BED10. Patient misalignment resulted in substantial decrease in conformity and increase in the gradient index, violating major characteristics of SBRT. Statistically insignificant differences were seen for normal tissue dose. Although, clinical follow‐up of these patients is ongoing, the authors recommend an alternative treatment planning strategy to minimize the probability of a geometric miss when treating small lung lesions synchronously with single‐isocenter VMAT SBRT plans.


| INTRODUCTION
Stereotactic body radiation therapy (SBRT) has become a standard of care for selected early-stage nonsmall cell lung cancer (NSCLC) patients. [1][2][3][4] Furthermore, SBRT of solitary primary or metastatic lung lesions is a fast, safe, and effective treatment option with a high control rate comparable to surgery. 4 For elderly medically inoperable patients, SBRT treatment has been shown to be effective. 5 However, elderly patients or those with poor pulmonary function and multiple oligometastastic (<5 lesions) lung lesions may not retain their treatment position for long SBRT treatment times. Traditional SBRT treatment to lung lesions requires an individual plan for each lesion with a separate isocenter placed in each, prolonging patient setup and treatment time. Treating multiple lung lesions synchronously with a single-isocenter plan, either using intensity-modulated radiation therapy (IMRT) or volumetric arc therapy (VMAT), has been studied. [6][7][8][9] Single-isocenter/multilesion VMAT lung SBRT treatments have been shown to be fast and efficient, improving patient comfort. [10][11][12][13][14] Additionally, treatment efficiency and dose buildup at the tumor interface is improved with use of a flattening filter free (FFF) beam. [15][16][17][18] This faster treatment option could potentially reduce intrafraction motion errors and improve patient compliance. 10 Despite the growing interest in single-isocenter/multilesion VMAT lung SBRT treatments, there is a decrement in accuracy when treating multiple lesions synchronously compared to treating the lesions individually. When each lesion is treated separately, the treatment plan has an isocenter in the center of the lesion, and daily This comparison was undertaken to quantify the dosimetric impact of residual setup errors on target coverage and collateral dose to adjacent organs at risk (OAR) for single-isocenter VMAT SBRT treatments of multiple lung lesions. Lung SBRT literature suggests that a biological effective dose (BED10) of ≥100 Gy (with α/ β = 10 Gy) to each lesion is required for optimal tumor local control (LC) and overall survival. 19

2.A | Patient setup and contouring
Patients were immobilized using the Body Pro-Lok TM

2.B | Clinical single-isocenter VMAT plans
For all 26 patients, single-isocenter VMAT lung SBRT plans were generated in Eclipse TPS for treatment on a Truebeam Linac (Varian Medical Systems, Palo Alto, CA, USA) consisting of standard millennium 120 MLC and 6 MV-FFF (1400 MU/min) beam. A single isocenter was placed approximately between the two tumors. Doses were 54 Gy or 50 Gy in 3 or 5 fractions, respectively. Both PTVs (PTV 1 and PTV 2) were planned with dose prescribed to the 80% isodose line and optimized such that 95% of each PTV received 100% of the prescription dose. The maximum dose to the PTV fell inside the GTV. Full arcs (coplanar) were utilized for bilateral lung tumors and partial noncoplanar arcs utilized for uni-lateral lung tumors, with AE5°-10°couch rotations, if possible. Optimal collimator angles and jaw tracking were chosen to reduce MLC leakage between each arc. Dose was calculated using the Boltzmann transport based AcurosXB algorithm for heterogeneity corrections with dose to medium reporting mode. [21][22][23] Planning objectives followed RTOG guidelines. 24,25 Each of the clinical VMAT plans were delivered every other day to the patient in the clinic.

2.C | Simulated single-isocenter VMAT plans
To evaluate patient setup uncertainties, clinically observable setup errors in all 6DoF were simulated in Eclipse TPS. Evaluation of pretreatment CBCT scans for our previously treated single-isocenter VMAT treatments for thoracic lesions allowed for determination of clinically representative random interfraction setup errors to be within AE5 mm in the x-, y-, and z-direction and within AE2°for pitch, yaw, and roll. The translational errors were defined for isocenter displacements. The rotational errors were defined for patient rotations relative to the isocenter around the right-left (pitch), anterior-posterior (yaw), and superior-inferior (roll) directions. For single-isocenter VMAT treatment, our current clinical practice is that, if we observed these setup errors larger than AE5 mm in any translational and AE2°in any rotational direction, we re-setup the patient and reimage for better alignment. Since demonstrating the loss of target coverage due to setup errors in current Eclipse TPS in all 6DoF was not readily accessible, an in-house MATLAB (Math Works, MA, USA) script was written. This simulation method was developed and integrated into Eclipse TPS in order to achieve the desired transformations and recompute the simulated VMAT plan for each fraction. To reproduce the interfraction setup errors, this script allowed the boundary conditions to confine the randomly generated setup uncertainties within AE5 mm in each translational direction and AE2°in each rotational direction with respect to the single-isocenter location as described above. The in-house script utilizes a RE DICOM file that is created with an image registration in Eclipse. This RE DICOM file consists of the patient CT registered to itself, thus the transformation matrix between the two images is null. The MATLAB script utilizes a random number generator to rewrite the transformation matrix of one of the identical patient CT datasets to apply translations and rotations within the determined range of possible shifts.
The random number generator utilized creates uniformly distributed random numbers, thus the transformation could simulate the worstcase scenario for patient setup errors. The image registration workspace in Eclipse TPS allows for visualization of these rigid transformations in all 6DoF. This is repeated for the number of fractions, with the original plan copied to the transformed image. The result of the simulation process is a plan summation of all three or five randomly transformed treatment fractions that mimics day-to-day clinical scenarios, allowing for evaluation of a clinically representative single-isocenter/multitumor VMAT lung SBRT treatment. Figure 1 below demonstrates the steps taken to achieve a complete simulated VMAT plan. Figure 2 demonstrates randomly transformed CT images, used for one treatment (out of five fractions) of a representative plan.

2.D | Plan comparison
All plans were compared per RTOG guidelines for target coverage along with maximum and volumetric dose to the adjacent OAR. Normal tissues that were evaluated included maximum dose to 0.03 cc of ribs, spinal cord, heart, bronchial tree, esophagus, and skin. Lung doses were evaluated using the mean lung dose (MLD), percentage of lung receiving 10 Gy (V10Gy) and 20 Gy (V20Gy) or more. Distance to isocenter was determined by utilizing the coordinates of the geometric center of each PTV as described above. In addition to the OAR doses, both plans were rigorously evaluated using the following metrics:

Parameters
Mean AE STD (range or n = no. of patients) • Heterogeneity Index (HI): Evaluates the dose heterogeneity inside the PTV, • per fraction to the GTV. An α/β ratio of 10 Gy was used for the pulmonary tumor and for n = number of treatments, the BED10 was calculated using the following formula:   Table 2).

2.E | Statistical analysis
For all 52 lesions, the average GTV coverage loss following the random transformations was 0.6%. However, for PTV volumes less than 10 cc, the GTV coverage loss was the greatest at an average of    Table 2, the uninvolved lung V20Gy, V10Gy, and MLD did not change significantly suggesting that the doses intended for the PTVs were not subsequently deposited in the uninvolved lungs.
Dose to the bronchus did not change significantly between clinical VMAT and simulated VMAT plans. However, the largest increase in maximal dose to bronchus was 3.7 Gy for the example patient shown in Fig. 7, although still acceptable per RTOG-0813 protocol. 25 Figure 8 shows the DVH associated with this patient.    VMAT SRS to multiple brain metastases using a "third party" software. 31 It was reported that minimizing rotational setup errors was essential for adequate target coverage, even more so for small lesions in the brain and lesions far from the isocenter location. This Placement of a single-isocenter at patient's mediastinum will avoid potential patient collisions and provide greater degree of noncoplanar arc geometry. It will eliminate the need of additional couch movements during CBCT imaging (couch centering is required for Varian Linac for lateral offsets of >5 cm, potentially introducing an additional source of error) and minimize the need for therapists to enter the treatment room for multiple couch positions.

| CONCLUSION
A novel and simple method for demonstrating isocenter misalignment in six dimensions and the resulting dosimetric impact for single-isocenter VMAT lung SBRT plans for two lesions has been presented. Clinically representative patient setup errors may result in large deviations (up to 72% loss) from planned target coverage.

CONF LICT OF I NTEREST
The author have no other relevant conflict of interest to disclose.