Stereotactic Body Radiosurgery for Spinal Metastases: A Critical Review
Stereotactic body radiosurgery (SBRS) has emerged as a new treatment option in the multidisciplinary management of metastases located within or adjacent (paraspinal) to vertebral bodies/spinal cord.
SBRS provides an attractive option to deliver high dose per fraction radiation, and therefore a high biologic equivalent dose (BED), typically in one to five fractions. Randomized controlled trials of conventional radiotherapy including vertebral body metastases with single fraction (BED = 14.4 Gy10) or multifraction radiotherapy (BED = 28–39 Gy10) demonstrate unsatisfactory rates of complete pain response and tumor control and without significant end point differences for single-fraction vs. multifraction treatment The aims of SBRS for spinal metastases, therefore, are to improve on existing rates of clinical response and tumor control and to reduce the retreatment rate by increasing the BED to approximately 43–82 Gy10 (20–24 Gy in a single fraction to 24–27 Gy in three fractions).
For previously irradiated spinal metastases, the focal nature of SBRS provides an otherwise unavailable noninvasive treatment option. Thus the goals of spinal SBRS parallel those of brain radiosurgery, with its improved local control over conventional whole brain radiotherapy and effectiveness in previously irradiated brains
Local control and palliative benefit
Published outcomes of spine SBRS can be grouped into four general categories.
1.Unirradiated patients—spinal metastases in a previously unirradiated volume treated with SBRS.
2.Reirradiated patients—spinal metastases in a previously irradiated volume now containing new, recurrent, or progressive metastatic disease treated with SBRS.
3.Postoperative SBRS patients—spinal metastases treated with SBRS after radical resection (corpectomy), or decompressive surgeries (laminectomy), or after stabilization procedures (vertebroplasty or kyphoplasty).
4.Mixed patients—mixed populations involving patients in the previous three categories in which outcomes are not separately reported.
According to these categories, crude local control data overall indicate excellent results, although with variable follow-up. Local control for unirradiated patients was achieved in 67/77 (87%) tumors treated, in 23/24 (96%) reirradiated patients' tumors, in 49/52 (94%) postoperative patients' tumors, and 568/655 (87%) mixed patients' tumors. As yet, there is no evidence to suggest that local control rates differ amongst the various categories.
One must consider the definition of local control used when trying to compare results among the different reports. Four studies based their tumor local control rates according to progression or recurrence by imaging and did not consider symptomatic progression as a failure (local control ranged from 77% to 100%). Five studies based their definition of local failure considering progression, or recurrence, as determined by imaging and/or clinical symptom progression (local control ranged from 87% to 100%). Actuarial data has only been provided by Chang et al. who report a 1-year freedom from progression rate of 84%, and Yamada et al. who reports an 81% rate of local control based on the median follow-up of 7 months. In both these reports local control was based on imaging criteria alone.
Patterns of failure based on imaging after SBRS have been reported. Gerszten et al. reported no case of tumor progression within the immediate adjacent vertebral levels based on the largest published experience of 500 cases treated (gross tumor volume [GTV] alone treated). Ryu et al. reported failures occurring in 3 of 49 patients treated for solitary metastases (the spinal segment was treated), and no failures in adjacent untreated vertebrae. Unfortunately, neither report actuarial rates of local control. The implication of these findings is that progression in adjacent vertebral bodies is rare, and supports SBRS treatment of the involved spinal levels only. Chang et al. reported their patterns of failure for 74 treated tumors where, by imaging, 17 (23%) demonstrated progression. Typically, they included the entire vertebral body while excluding the posterior elements in the target volume. As in the two previous studies, progression occurring in the adjacent vertebrae was rare and occurred in only one case. Otherwise, two main patterns of failure were identified: (1) failure of nontreated pedicles and posterior elements (3 of 17) and (2) failure occurring at the epidural space (8 of 17) (an area at risk of potential tumor underdosing to spare the spinal cord where the limit was set at 10 Gy). Gerszten et al. report, based on a series of renal cell spine metastases treated where the GTV = planning target volume (PTV), failure occurring at the edge of the contoured treatment volume in 6/60 cases treated. However, the mean volume for these failures was large at 102.6 cm3 and 40% larger than the average for the entire cohort. Preliminary data from University of California San Francisco (UCSF), reported in abstract form, suggest that if the minimum distance from the PTV to the neural contour is <1 mm the risk of local failure may be increased. An example of treatment failure from potential underdosing at the PTV-neural contour interface is provided , as the minimum distance from the PTV to the neural critical structures (NCS) contour was 0.6 mm. Alternatively, an example of an excellent response to a metastasis with significant epidural extension is shown
Based on these data, it is possible that (1) failure in the epidural space may be due to underdosing of tumor because of strict spinal cord constraints, (2) uninvolved adjacent posterior elements should be included in the target volume, and (3) encompassing one vertebral body above and below the diseased vertebrae is unnecessary.
In terms of palliative benefit, only three studies provide data for complete pain relief post-SBRS and report a complete response (CR) in 56% (5 of 9 patients) , 33% (6 of 18 patients), and 75% (6 of 8) of patients treated in their respective series. This is difficult to interpret given that the patient numbers are few, and the data are subjectively as opposed to assessment using a formal pain response tool. However, favorable rates of some degree of pain relief have been reported (ranging from 67% to 100%). In most cases, no consistent pain scoring system was used, or none at all, because of the reality of the retrospective nature of most of the series. Therefore it is impossible to gauge the potential benefit in relation to conventional radiotherapy, and this remains as an important question to justify this treatment approach. It is also not possible to quantify in the postoperative setting the benefit of adjuvant SBRS in terms of pain relief, in which resection or decompression has been performed as pain may be more prominent because of postsurgical type pain or alternatively completely resolved.
In the only study to report on quality of life after spine SBRS, Degen et al. report no statistically significant change in the physical well-being component or mental component summary of the SF 36 collected up to 18 months post-SBRS for those patients treated with metastatic disease (unfortunately they do not specifically report the dose to the metastatic tumors). Chang et al. has reported a decline in narcotic usage after SBRS, and reports a decline from 60% to 36% at 6 months after SBRS. Gerszten et al. report long-term improvement in pain in 86% of 294 patients treated (not actuarial) using a visual analogue scale with care to ensure that pain relief was not due to increased narcotic usage
As for neurologic outcomes, Chang et al. reported outcomes using a formal neurologic clinical tool of the McCormick scheme Other listed studies used the National Cancer Institute Common Toxicity Criteria or Radiation Therapy Oncology Group toxicity grading systems. These variations in study design make local control and quantitative assessment of pain relief difficult to characterize in a consistent fashion.
Local control data have been reported on SBRS for a variety of tumor histologies that have ranged from the radioresistant (sarcoma) to the radiosensitive (myeloma). This information is not listed, but will be subsequently discussed.
Target delineation and prescribed dose
There are two general philosophical camps with regards to defining the PTV. The first, as practiced by centers including Pittsburgh, UCSF, and Stanford, contour the GTV (or radiographically visible tumor) without an anatomic applied margin for potential sites of microscopic disease (i.e., no clinical target volume [CTV]) analogous to radiosurgery for brain metastases. The margin for PTV has ranged from no distinct applied margin (i.e., GTV = PTV), to absolute expansions of 2 mm, 3 mm, or 1 cm on the GTV, albeit modified to exclude the neural contour. The second, as practiced by centers including Henry Ford Hospital and the MDACC, contour the GTV plus a CTV margin based on the anatomic routes of spread within the spinal segment . This variation in the defined PTV is described illustrate the two different approaches. The imaging technique also varies from using the CT alone to fused magnetic resonance images (MRI) and the modalities are
A range of prescribed doses have been reported and currently the main fractionation schemes include single fraction SBRS ranging from 8 to 24 Gy, or hypofractionated regimens consisting mainly of 4 Gy × five fractions (frx), 6 Gy × five fx, 8 Gy × three fx, and 9 Gy × three fx. None of the studies describing mixed populations of unirradiated and previously irradiated tumors specified different prescriptions for the two groups, and as yet there is no evidence to support one regimen over another. The initial use of external beam radiation therapy followed by a SBRS boost has largely been replaced with an SBRS alone approach.
Minor and limited toxicities (acute or subacute) have been reported with spine SBRS and include esophagitis, mucositis, dysphagia, diarrhea, lethargy, paresthesia, wound dehiscence, transient pain flare, nausea , vomiting, trismus, non-cardiac chest pain idiopathic vasculitis, mild skin hyperpigmentation, transient laryngitis, and transient radiculitis. Of note, neither at the MDACC nor at UCSF do we routinely premedicate patients with steroids before SBRS.
The most devastating complications are those that are permanent. There are seven cases of reported radiation-induced myelopathy in four separate publications of a total 396 tumors treated. Dodd et al. described a patient with a cervical spine meningioma treated with 24 Gy in three fractions who developed posterior column dysfunction 8 months post-SBRT. He reported the volume of spinal cord receiving >8 Gy (V >8 Gy) as 1.7 mL. Ryu et al. reported a case of myelopathy occurring in a breast cancer patient 13 months' post-SBRT. Based on the detailed dosimetric data provided of the percent volume of spinal cord irradiated, the highest cord point dose reported for this patient was 14.6 Gy in a single fraction (BED of 121 Gy2). Gibbs et al. report on three cases of myelopathy post-SBRS. Two of the myelopathy patients had received prior radiation (a spinal cord maximum dose of 25.2 and 40 Gy), and had received a subsequent BED of 46–81 Gy3. Gwak from Korea, report on 1 patient developing myelopathy following a maximum point dose to the brainstem of (35.2 Gy in three fractions) 243 Gy2, and a second patient who had been previously irradiated (50.4 Gy) whose subsequent maximum point dose was (32.9 Gy in three fractions) 213 Gy2. Other than the Korean group, none of the doses described at which myelopathy was observed fell into the extreme of the range of the doses delivered to the spinal cord in the respective series. The possibility of setup error leading to unintentional over-dosage of the spinal cord is a possible limitation to any current dose–volume toxicity analysis, and to interpret this information is also difficult when the spinal cord contouring practice differs
At the MDACC, one lung cancer patient was reported with L1 vertebral body collapse and intractable pain 12 months after SBRS requiring vertebrectomy and stabilization. Two additional cases of vertebral body compression have occurred after SBRS (data unpublished and provided by Dr. M. J. Sohn, Korea) where in one case percutaneous vertebroplasty was required to prevent further compression and deformity.
In terms of other organs at risk such as the kidneys, lungs, small bowel, and esophagus, there has yet to be any Grade 3–5 late toxicities reported, and in terms of dose thresholds much is left to the clinical judgment of the treating physician. A discussion of dose limits to non-spinal cord organs is beyond the scope of this review.
Spinal cord tolerance and SBRS
It has generally been accepted that a homogeneous dose of 45 Gy (2 Gy/day) results in a <0.5% risk of myelopathy. For single fraction radiotherapy, the tolerance to the spinal cord is unknown, but generally quoted as safe up to 8–10 Gy given a homogeneous exposure . SBRS results in inhomogeneous dose distributions within the tumor volume and organs at risk. Therefore the spinal cord volume is exposed to a gradient of dose. Ryu et al. published partial volume neural structure dosimetric data based on 230 tumors treated, and the average and highest dose to 1%, 5%, 10%, 20%, and 30% of the neural contour volume and the maximum point dose. They consistently defined their spinal cord volume as 6 mm above and below the radiosurgery target, and recommend to limit the volume of spinal cord receiving a dose of 10 Gy or higher to 10% of the spinal cord volume contour . Although this threshold information is helpful, it is dependent on the total volume of cord outlined. Even if one is contouring several slices above and below, this is dependent on slice thickness and kyphotic spine curvature. This contouring approach provides a relative dose–volume description of the neural contour dosimetry and is potentially sufficient for coplanar beam IMRT delivery.
Neural critical structure contouring and dose reporting
As the focus of this section is on the neural critical structure (NCS), defined as the spinal cord and cauda equina, we reviewed reports that also included spinal SBRS for benign tumors. Studies were included if length of follow-up, NCS contouring technique, and NCS dosimetric information (or a threshold by which treatment planning was based on) were reported.
Contouring of the spinal cord or cauda equina has varied in practice in the published literature. Examples of the various methods of delineating NCS structures include delineation of the spinal cord (with or without margin for setup uncertainty), the spinal canal, the thecal sac (for both the spinal cord or cauda equina), and the intrathecal contrast-enhanced thecal sac. Because defining the NCS is not well defined, we compare and contrast the NCS delineation method at the MDACC as opposed to the USCF approach as the methods are distinctly different and an opportunity to discuss the issues.
At the MDACC, the intramedullary spinal cord (and thecal sac for cauda equina) is contoured with no applied margin. These practitioners believe that delineating the true cord yields the most accurate normal tissue complication probability information. However, a 2-mm expansion is applied to the spinal cord volume for dose–volume histogram (DVH) analysis. The 2-mm margin is composed of 1 mm to account for setup uncertainty (not accounting for rotational error), and 1 mm for pixel size to account for contouring uncertainty. Thus the spinal cord + 2-mm DVH yields information on the amount of spinal cord irradiated in a worst case scenario, and is used as an avoidance structure for inverse treatment planning.
The criticism of the canal/thecal sac delineation method, as practiced at UCSF, lies in leading one to believe that the spinal cord can withstand artificially high radiation doses. Dosimetric data based on the spinal cord with no applied margin for positional errors will also not accurately reflect the true dose to the NCS. This point is highlighted in a recent analysis by Guckenberger et al., who performed an extensive analysis of potential dosimetric effects to the spinal cord based on various translational and rotational errors. Additional patient specific factors will also affect the true dose delivered to the cord such as intrafractional organ motion (the mean thoracic spinal cord motion range by dynamic MRI was recently reported to be within 0.5 mm), and intrafractional patient movement. Prior reporting of dose to the spinal cord was based on large fields and homogeneous exposure and therefore did not have to address most of the issues involved with determining true cord dose.
At UCSF, the thecal sac based on the planning CT is contoured separately for both the spinal cord and cauda equina, and dose constraints are based on this contour. This serves to provide some margin on the actual organ for dosimetric uncertainty considering that SBRS is performed with the CyberKnife, which has an excellent reported accuracy). Furthermore, intrafractional imaging at the interval of 30–60 s allows for clinical reassurance that patient motion is being accounted for.
The NCS must also be contoured sufficiently above and below the target volume (at UCSF, no defined length is specified as we visually determine it based on the isodose line profile) to accurately capture the low-dose region of the DVH curve and potential islands of high dose beyond the limited dose matrix which typically is set to encompass the tumor and adjacent organs at risk. This issue is highlighted, as a limited dose matrix box was intentionally created for the coronal view to limit the calculation matrix in the superior, inferior, and lateral directions. One can appreciate that the low-dose isodose lines are cut off, and will not contribute to the DVH. Therefore one must be aware of the dose matrix calculation grid in all dimensions such that the resulting DVH is accurate. Lack of this appreciation could lead to serious toxicity. The imaging technique to delineate the NCS is also not well defined and some centers base it on the planning CT, fusion with MRI images, or by CT myelogram. Of note, if MRI fusion is to be used then one has to accept the potential additional error associated with the fusion.
The reporting of dose to the NCS has also varied as indicated. At UCSF, we support reporting the absolute doses to absolute volumes and to report these data for the spinal cord and cauda equina separately. Further stratification is required if previous radiotherapy had been performed. Separating the cord from the cauda is important, because the tolerance for the cauda equina is greater than that of the spinal cord
With new concerns over the potential interaction of radiotherapy with new targeted chemotherapy agents emerging and the lack of radiobiologic understanding as to the implications of inhomogeneous dose distributions within the organs at risk, clearly there is much to learn as to how to best report and biologically model partial volume doses. This is particularly relevant when the NCS has been previously exposed to radiation, and tolerance in this circumstance is even less defined. If a consensus can be reached as to the contouring and reporting of absolute dose then we will gain invaluable information to develop further spine SBRS safety guidelines and dose escalation.
The field of spinal SBRS is in its formative years. Studies from a limited number of centers suggest that SBRS used in the treatment of spinal metastases appears to be safe and effective both in terms of radiographic tumor control and pain relief. This is particularly important as a viable noninvasive option for the previously irradiated patient with painful spinal metastases, as these patients have not had many noninvasive therapeutic options available in the past. To define the role of spine SBRS for patients with unirradiated spinal metastases, and justify this more costly palliative therapeutic approach (SBRS costs approximately three to four times the cost of conventional radiotherapy as broken down in, randomized controlled trials would be helpful to determine if the high BED delivered is necessary in comparison with conventional radiation delivery and its doses.