A retrospective study on clinical factors influencing intra-fraction motion using volumetric imaging for spine stereotactic body radiotherapy

Abstract

Objectives

Stereotactic body radiation therapy (SBRT) for the spine is challenging due to high-dose gradients sparing the cord in the treatment plans. We present our findings of initial setup error and intrafraction motion from Cone-beam computed tomography (CBCT) imaging.

Materials and methods

A total of 47 patients treated with spine SBRT with a total of 154 fractions following a fractionation schedule of 16 Gy in 1, 24 Gy in 2, and 30 Gy in 5 fractions were part of this study. Pre-treatment CBCT was used for localization of the target and couch shifts were applied based on target volume matching to the planning CT image set. Post-treatment CBCT was acquired for all fractions. Intrafraction motion (IFM) was calculated by matching post-treatment CBCT to planning CT for the target volume.

Results

The average Intrafraction motion was 1.6 ± 0.9 mm for the study cohort. The average and standard deviation of intrafraction motion were 0.4 ± 1.1 (AP), 0.3 ± 0.9 (SI) and 0.2 ± 1.2 (RL) respectively. The average Initial setup error tabulated from the offline review showed a mean value of 7.8 ± 5.3 mm. The average and standard deviation of the initial setup error were 2.5 ± 5.5 (AP), 2.4 ± 5.3(SI), and 0.8 ± 4.5(RL) respectively. The correlation of intrafraction motion with body mass index (BMI) and the number of consecutive vertebrae levels did not show any statistical significance, however, there was a significant association with gender as women showed more IFM

Conclusions

Our study on intrafraction motion from CBCT images reinforced the importance of immobilization and imaging for positioning spine SBRT patients.

Advances in knowledge

The need for CBCT and imagining for positional errors is emphasized while treating with SBRT spine and the need for proper immobilization techniques

INTRODUCTION

Recent advances in systemic treatment options, such as Immunotherapy, targeted therapy, and newer chemotherapeutic agents, for patients suffering from stage IV cancers, have significantly improved the survival and clinical performance of these patients. Now, it is even more challenging to treat patients with spine metastases with conventional radiotherapy, due to limited long-term pain control as shown by UK/New Zealand Bone Metastases randomized control trial.^1,2,3 Spine Stereotactic Body Radiotherapy (SBRT) has emerged recently as an effective modality to palliate patients with spine metastases.^4,5,6 This technique has become increasingly popular to treat patients with spinal metastases, even in the setting of re-irradiation. A multi-institutional analysis revealed that SBRT can provide effective pain control in painful spinal metastases and prolong progression-free survival.

Although Spine SBRT is a non-invasive, short course, and an advanced technique to treat spine metastases, it requires robust technology in terms of planning, patient setup, and motion corrections using Cone-beam computed tomography (CBCT) or radiographic imaging. It also requires very good immobilization devices to assure avoidance of catastrophic side-effects with respect to the spinal cord and cauda equina, which anatomically lies near the spine.^7,8 Few studies have reported the incidence of radiation myelopathy with SBRT and have recommended safe dose limits for the spinal cord.^9,10 The dose gradient achieved with SBRT is significantly sharper when compared to other conventional techniques of delivering radiotherapy, and also the dose delivered per fraction is significantly high, which demands patient positioning to be very precise to avoid normal tissue toxicity and ensure planned dose delivery to the target. Phase III results from the clinical trial RTOG 0631 published recently showed pain control rates using SRS/SBRT were lower as compared to conventional EBRT for spine metastases with no increase in adverse effects.¹¹

Patient immobilization is one of the most important links in the process for a robust SBRT spine program. Various devices such as thermoplastic masks, custom vacuum bags, full-length vacuum body fixation, and a dedicated stereotactic positioning system are available. The vacuum body fixation is preferred for the middle and lower thoracic/ lumbar spine, as it has been shown to minimize intrafraction motion (IFM)¹² and thermoplastic mask for the cervical/ upper thoracic spine.¹³

3D imaging has been proven to be more precise than conventional 2D on-board imaging for setup verification with a significant impact on target dose and organ as risk (OAR) sparing, as reported by various studies in the past.^11,12,13 A robotic couch with the capability of corrections in 6 degrees (both rotational and translational shifts) in combination with in-room CBCT has been used to improve accuracy in patient positioning.¹⁴ Patient movement during treatment can also significantly impact the dose received by spinal cord/cauda equina, which makes clinical evaluation even more pertinent.

Pre- and post-treatment CBCT images are routinely employed to quantify positional uncertainties during treatment and as a means for quality assurance of delivering SBRT to the spine with utmost precision.¹² The intrafraction motion also depends on the treatment technique used to deliver SBRT as the time for treatment delivery is extremely short with Volumetric Modulated Arc Therapy (VMAT) when compared to multiple static beams.^15,16 Shorter treatment times with VMAT and quantification of positional uncertainties with volumetric imaging make the SBRT treatments more reliable and accurate.

There are various ways to observe intrafraction motion for spine SBRT besides the CBCT based imaging. On-treatment kV images have been used for tracking the patient motion and one study assessed the use of the conventional 2D x-ray images and learning-based approach was developed to remove soft tissue on the x-rays to improve spine tracking accuracy.¹⁹ There have also been established software application that quantifies intrafraction motion for patient accuracy during treatment delivery with higher accuracy of registering the IFM and helps assess the registration error differences.²⁰ In this retrospective study, we evaluated the setup accuracy by calculating intrafraction motion from CBCT imaging for spine SBRT treatments. Furthermore, we noticed that there were no publications on the correlation of intrafraction motion with clinical and technical parameters. So, our study is the first one to examine the correlation of patient characteristics such as age, gender, Body mass index (BMI), and lesion location with intrafraction motion.

METHODS AND MATERIALS

Patient data

This is a retrospective study, approved by the local Research Ethics Board (REB), for spine patients treated in our institution for nearly 4 years. Eligibility of patients for spine SBRT followed strict criteria that included single or multiple contiguous spine levels. A minimum margin of ≥ 3mm was essential between the edge of the tumor and the spinal cord, for the patients to be eligible for SBRT. Electronic Medical Records (EMR) were reviewed to gather all the data for 47 patients treated during this period. Clinical and dosimetric data were collected using the Varian Aria data management system (Varian Medical Systems, Palo Alto, CA) and the Eclipse^© treatment planning system. Clinical characteristics such as age, gender, BMI, number of consecutive vertebrae treated, number of fractions, and tumor location with respect to the spinal level were recorded. Patient positioning data such as translational and rotational values during treatment were recorded using Offline review from CBCT.

Patient preparation

Clinically eligible patients were immobilized in a stable supine position with comfort as comfortable as possible to allow reproducibility throughout treatment. Uncomfortable (e.g., prone) positions were avoided to minimize movement. BodyFix immobilization system (Elekta) was used for thoracic and lumbar vertebral lesions. The system consists of two layers, a vacuum bag that provides a comfortable and reproducible position for imaging and treatment, and a clear plastic sheet over the patient and when evacuated provides adequate restraint. A Klarity thermoplastic mask for cervical and upper thoracic vertebral lesions (up to T4 vertebra) was used for immobilization.

Patients underwent CT simulation using Philips Big Bore CT simulator with 2 mm slice thickness and scan range encompassing the disease spinal levels and adequate superior and inferior region for accurate dose calculation. Magnetic Resonance Imaging (MRI) comprising T1 post-gadolinium and T2 weighted axial scans was co-registered with the simulation scan, with a focus on the region of Interest (ROI) around the involved lesion to delineate soft tissue tumor component and high-risk OARs.

Treatment planning

Target volume and Organs at Risk (OAR) were delineated using RTOG 0631 guidelines with the CTV to PTV margin ranging from 0 mm to 3 mm. Planning Organ at Risk Volume (PRV) of 1.5-2 mm was added additionally to Spinal Cord and Cauda Equina for optimization, to further restrict the dose received to these critical structures. The prescribed dose to the target varied depending on the target volume size and treatment type i.e. new or retreatment. There is uncertainty in optimal dose fractionation for spine SBRT, however, there may be a local control benefit of fractionation to limit the risk of potential toxicities such as pain flare, vertebral body fracture, and myelopathy.²¹ We used a dose prescription of 16 Gy in one fraction, 24 Gy in two fractions, or 30 Gy in 5 fractions.

Treatment delivery with image guidance

Varian Clinac 21EX linear accelerator with on-board imaging (OBI) was used for treatments. After prepositioning the patients using lasers and tattoos, CBCT imaging was performed for accurate patient setup. Auto matching was performed with target volume structure for alignment of vertebral bodies followed by manual matching. In manual matching, fine alignment was done in the vicinity of the target to reduce any residual error. Matching of the target to the planning CT was verified in all three planes by the team and then approved for treatment. Figure 1 shows an example of planning CT and CBCT images after correction. Angular deviations (pitch and roll) were minimized using translational motions and yaw was corrected by repositioning the patient if it was more than 2˚. Translational couch shifts along with couch angle from matching were applied. If the positional error exceeded 2 mm in any plane, patients were repositioned by repeating the patient setup and imaging A verification CBCT before treatment was taken for all patients and assessed to proceed with treatment upon approval. Post-treatment CBCT imaging was obtained in the treatment position to quantify the final setup error and deduce intra-fraction motion (IFM).

Figure 1.

An example showing Sagittal and Coronal reconstructed images from planning CT in the top. The bottom images show the axial CBCT image with target volume(left) and the corresponding axial planning CT image (right).

Data analysis

In the Offline review, initial setup errors from the first CBCT couch shifts were noted in a spreadsheet. This data shows the difference between initial patient positioning using lasers and the target anatomy matching using CBCT. Post-treatment CBCT images were matched to planning CT images to extract setup error data in the form of translational and couch angle deviations for a total of 154 fractions. The registration process for data analysis was similar to the treatment procedure with same matching parameters. Clear visibility of bony anatomy as a surrogate for the target helps in reducing any other sources of uncertainty. IFM was calculated using the difference between pre-and post-treatment CBCT matching to planning CT images. Average initial setup error and IFM were calculated for each axis i.e. anterior-posterior (AP), superior-inferior (SI), and right-left (RL). A linear mixed model, which accounts for multiple measurements by the same individual over time, was used to predict factors that may be associated with IFM. The assumption of normality and identification of outliers were examined using residual plots. The assumption of linearity was tested using restricted cubic splines. Analyses were run using the nlme and Hmisc packages in R version 3.4.1.

RESULTS

The patient data and treatment characteristics are depicted in Table 1. Of the total number of 47 patients and 154 fractions of treatment, 10 (21.3%) patients had the disease in the cervical vertebrae, 14 (29.8%) in the Lumbar vertebrae, and 25 (53.2%) patients in the thoracic vertebrae. The mean Body Mass Index (BMI) and age of the study population were 26.4 Kg/m² and 60.6, respectively. Among the study population, 27 (57.4%) were females and 20 (42.6%) were males. The majority of patients (76.6%) were treated naïve to the vertebral level treated with SBRT while 23.4% of patients had prior radiation and were planned and treated to the same level of vertebrae. 24 patients (51.1%) had metastatic lesions involving a single vertebral level followed by 10 (21.3%) patients with disease involvement in 2 vertebrae.

Table 1.

Patient and treatment characteristics

Variable		No.	Percentage of the total variable (%)
Age	mean ± SD	60.6 ± 14.0
Gender	Female	27	57.4
	Male	20	42.6
BMI	mean ± SD	26.4 ± 5.4
Spinal level treated	Cervical	10	20.4
	Thoracic	25	51.0
	Lumbar	14	28.6
No. consecutive vertebrae treated	1	24	51.0
	2	10	21.3
	3	6	12.7
	4+	7	14.9
Previous RT to treated spinal level	Yes	11	23.4
	No	36	76.6
Fractions	1-2	25	53.2
	5	22	46.8

The average Intrafraction motion, calculated using post-treatment CBCT and verification CBCT prior to treatment, was 1.6 ± 0.9 mm for the study cohort. The average and standard deviation of intrafraction motion were 0.4 ± 1.1 (AP), 0.3 ± 0.9 (SI) and 0.2 ± 1.2 (RL) respectively. The average Initial setup error tabulated from the offline review showed a mean value of 7.8 ± 5.3 mm. The average and standard deviation of the initial setup error were 2.5 ± 5.5 (AP), 2.4 ± 5.3(SI), and 0.8 ± 4.5(RL) respectively.

Results from univariable linear mixed models are presented in Table 2. Gender was the sole patient characteristic showing a statistically significant association with Intrafraction motion (IFM). Women as compared to Men had higher Intrafraction motion (95% CI 0.02-0.10, p-value 0.004).

Table 2.

The linear mixed model predicting intrafraction motion

Variable		Univariable
		coefficient	95% CI	p
Fraction	(per fraction)	0.00	−0.01 - 0.01	0.936
Age	(per 10 years)	−0.01	−0.02 - 0.01	0.491
Gender	Female	0.06	0.02 - 0.10	0.004
	Male	(reference)
BMI	(per 10 units)	−0.03	−0.07 - 0.02	0.244
Vertebrae	(per vertebrae)	0.01	−0.01 - 0.03	0.288
Location	C	(reference)
	C-T	0.02	−0.10 - 0.15	0.859
	L	0.02	−0.05 - 0.09
	T	0.03	−0.03 - 0.10
	T-L	0.02	−0.10 - 0.14
Reirradiation	Yes	−0.01	−0.06 - 0.05	0.832
	No	(reference)

BMI was recorded before the first fraction for all patients and the average was utilized for linear regression analysis. Our study showed a non-significant association between IFM to BMI, 95% CI of 0.07-0.02 and a p-value of 0.244. Similarly, age (95% CI 0.02-0.01, p-value 0.491) and fractionation (95% CI 0.01-0.01, p-value 0.936) also did not reveal any clinically significant correlation with IFM. Analysis of single vs multiple vertebrae did not yield any significance with respect to intrafraction motion.

Tumor location with respect to the Cervico-thoracic, thoracic, lumbar, and thoracolumbar vertebral levels appeared to have a non-significant influence over IFM, as shown in Table 2. Re-irradiation did not prove to affect IFM with 95% CI 0.06-0.05 (p-value 0.832).

DISCUSSION

Image guidance and motion management are essential in delivering the planned dose to the target volume for stereotactic treatments. The relatively stable spine anatomy helps in minimizing tumor motion compared to thoracic and abdominal soft tissue tumors that can be impacted by respiratory motion. We investigated the intrafraction motion by comparing the positional errors between planning, verification, and Post-treatment CBCT images during treatment delivery. Large translational positioning errors beyond 2 mm could lead to a>5% decrease in target coverage with the prescribed dose and a>25% increase in the dose to OAR according to Wang et al²². In addition to the calculation of intrafraction motion, the statistical analysis looking into the correlation of clinical and technical parameters with intrafraction motion gives another dimension to this work.

The first part of our work investigating intrafraction motion has been studied by others as well and their publications have shown similar results. Hyde et al²³ reported target localization within 1.2 mm and 0.9 degrees using strict repositioning thresholds, near rigid body immobilization, and intra-fraction CBCT every 15-20 minutes. In another study by Finnigan et al, the target was localized within a 3 mm PTV margin with the median translational and rotational positional errors of 0.5 mm and 0.25 degrees respectively without being significantly influenced by vertebral level.²⁴ Rotational and translational setup accuracies play a significant role in assuring precise delivery of high doses of radiation. Any compromise to patient positioning could lead to serious adverse events due to the proximity of the spinal cord, which is radiobiologically more sensitive to high doses per fraction. In this study, we found that the two immobilization systems and CBCT-based image matching have played a crucial role in the acceptable initial setup error which has been corrected using couch shifts before the start of treatment, and intrafraction motion as shown in Fig 2. Though the 3D vector of intrafraction motion is 1.7 mm, the average error in each translational axis is much less suggesting no systematic shift of patients and better positional accuracy for treatment delivery.

Figure 2.

Average initial setup error and intrafraction motion in each cardinal axis is plotted. Besides, the figure depicts a 3D vector of the initial setup error and intrafraction motion.

The second part of the study focused on collecting all the clinical and technical parameters and performing statistical analysis to find any significant association of the variables with intrafraction motion. Based on the association and influencing parameters, the objective was to work on the relevant parameters toward improving the precision of treatments. This study found no significant association of clinical characteristics such as BMI, age, and tumor location with positional uncertainties obtained from imaging. Another study reported by Jeon et al, though looking at only BMI, reported similar non-association in their findings. We found the only variable which showed statistical significance was gender, female patients were observed to have higher IFM than males (Figure 3). O. Hay et al²⁵ studied the dimorphism of spine curvature between males and females and concluded that females’ lumbar spine showed greater curvature than the males. Increased curvature of the lumbar spine in females could be one of the plausible explanations for IFM to be higher in women than men, for our study. Cervical vertebrae are highly mobile and yet, the motion during treatment was within our institutional tolerance limit. Li et al. concluded in their report that with appropriate immobilization, a 2mm PTV margin is adequate to encompass the target with the prescribed dose.¹⁴ This phenomenon can be explained by underscoring the importance of the use of a thermoplastic mask for cervical and upper thoracic vertebral lesions. Although our study was not statistically powered enough to show the difference with respect to the site of treatment and the immobilization device used, it certainly highlights the robustness of immobilization.

Figure 3.

Gender plot showing intrafraction motion data for both genders. Females showed larger intrafraction motion compared to males.

The combination of ExacTrac system with stereotactic body frame has shown better setup accuracy and intrafraction stability which is comparable to that of the mask-based cranial radiosurgery. Similarly, another retrospective analysis evaluated the intrafraction motion in spine SBRT using the flattening filter free (FFF) beam delivery and cone beam CT image guidance (CBCT) and found that the positional errors were significantly smaller with CBCT-based verification and thus, concluded that including CBCT prior to spine SBRT treatment helps increase the overall positional accuracy.²⁷ The intrafraction target motionduring spine SBRT were investigated in a retrospective analysis with use of orthovoltage kV images that were taken 3-5 times during each treatment and they reported that, though the occurrence of such shifts were less but they did find these shifts during their treatment delivery, thus emphasizing the need for intrafractional monitoring during treatment delivery.²⁸ In another study, evaluating the amplitude of translational and rotational movements and the importance of intrafractional imaging and the impact of treatment time, noted that there was an improvement in the intrafractional movement with integrated SBRT solution consisting of a SBRT table top, an Orfit™ AIO system, and a vacuum cushion. They also concluded that there was indeed a correlation between the treatment time and corrections to be applied thereby leading to usage of imaging modalities to allow reduced fraction time.²⁹

Our study does have its limitations aside from it being a retrospective analysis. One of the potential limitations is that we deduced the intrafractional motion based on pre and post CBCT and not based on the acquisition of imaging information acquired during the treatment. Hence, there was a possibility of the patient moving after the treatment but before the post-CBCT which cannot be documented. However, many other studies have evaluated intrafraction motion similar to our methodology using CBCT. Also, we did not use 6D couch correction for the patient setup which is a very common practice in spine SBRT. Thus we could assess only the translational shifts in our study and not the rotational motions.

CONCLUSION

Quantifying the setup errors in positioning the target volume is paramount to precisely delivering spine SBRT treatments. Our study reinforces the importance of a robust immobilization system and proper image guidance in minimizing the intrafraction motion to less than 2 mm. Except for gender, there were no strong association of any other clinical or technical parameters with intrafraction motion. However, further studies with a large patient population are needed to explore these aspects.