Technology Overview

The CyberKnife system consists of a lightweight, 6-megavolt (MV) linear accelerator (LINAC) mounted on an industrial robot, a remotely repositionable treatment couch, orthogonally placed digital x-ray cameras, a treatment-delivery computer, and treatment planning stations (Fig. 100-1). During treatment, numerous images are obtained to optimally locate the target before the delivery of 100 to 150 individual treatment beams.¹⁴ The CyberKnife can deliver individual beams to any part of a tumor from nearly any angle and provides highly conformal dosing to complex three-dimensional (3D) targets.¹⁵

FIGURE 100-1 A, The CyberKnife frameless stereotactic system includes a modified 6-MV X-band LINAC mounted on a highly maneuverable robotic manipulator (KUKA Roboter GmbH, Augsburg, Germany). B, Two high-resolution x-ray cameras are mounted orthogonally to the headrest. C, One of the two x-ray sources is mounted in the ceiling projecting onto the camera. D, The treatment couch is mobile, allowing the x-ray sources to image targets at any point along the neuraxis.

Several conventional radiation therapy (RT) systems have recently been modified to provide SRS using techniques originally developed for the CyberKnife™. The BrainLab Novalis, with floor and ceiling mounted x-ray detectors, most directly emulates the CyberKnife™ localization system. The Elekta Synergy, Varian Trilogy, and Novalis TX systems use different combinations of cone-beam CT scanners and orthogonal x-ray cameras. These RT-based systems are constrained by the gimbal design of the gantry which only allows radiation to be delivered along two-dimensional arcs. Nevertheless, multileaf collimators can compensate for much of the constraints that stem from this gantry design.

Treatment Details

Spinal SRS is image guided and completely frameless. Individually molded masks or cradles are fashioned before the planning scans are completed. The patient rests in the mask or cradle during computed tomography (CT) scanning, magnetic resonance imaging (MRI), 3D angiography or positron emission tomography (PET) imaging, and again during treatment. The devices are comfortable, limit movement, and expedite positioning. At Stanford, we use an Aquaplast mask (WFR Corp., Wyckoff, NJ) for upper cervical lesions (Fig. 100-2A). Lower cervical, thoracic, lumbar, and sacral lesions are treated using an AlphaCradle (Smithers Medical Products, Akron, OH) (Fig. 100-2B). Most patients are treated supine but prone and lateral decubitus positioning is possible.

FIGURE 100-2 Simple immobilization devices used during CyberKnife treatment. A, The Aquaplast mask is used in patients with upper cervical lesions. B, AlphaCradle custom body mold is used in patients with lesions below the cervical spine.

A high-resolution, fine-cut CT scan, most often with contrast, is required for all treatment plans. The CT is essential for its excellent spatial accuracy and is needed to calculate the digitally reconstructed radiographs (DRRs) used for real-time targeting. MRI scans may help with visualization but cannot be utilized alone because of the need to generate DRRs. Meanwhile, MRI often suffers from inherent limitations in spatial resolution. In the appropriate clinical situation, PET or angiogram images are also obtained. When a frame-less radiosurgical system is used, the patient does not need to remain in the facility while the treatment plan is developed.

Bony landmarks are used to target spinal lesions and those in adjacent structures. Spinal fusion hardware does not interfere with treatment. The accuracy of CyberKnife approaches ±0.5 mm.^7,16 For lesions not associated with bony landmarks, three or more gold seeds or titanium screws are implanted (Fig. 100-3).^13,15

FIGURE 100-3 Implanted fiducials are marked and numbered. Left, Computed tomography-based digitally reconstructed images from the perspective of the two orthogonal CyberKnife mounted x-ray cameras (A and B). Center, Real-time x-ray images from the two x-ray cameras. Right, Overlay of the reconstructed and actual radiographic images.

CyberKnife treatment plans use Accuray’s MultiPlan software. BrainLab, Varian, and Elekta systems employ similar programs. After uploading the CT and other images to the planning station, the surgeon and radiotherapist “contour” the target and radiation-sensitive structures by tracing their outlines (Fig. 100-4). The dose and number of sessions, as well as dose limits for adjacent structures, are prescribed by the physicians (Fig. 100-5). Physicists then use the planning system, the contour data, and the dose prescriptions to create a 3D representation of the lesion geometry and define sets of treatment beams. An ideal treatment includes evenly distributed beams that target the dose uniformly while limiting exposure to sensitive structures (Fig. 100-6). The neurosurgeon and/or physicist will iteratively perfect the plan by adding or removing constraints, or re-positioning individual beams. A multidisciplinary team reviews and accepts each treatment plan before delivery.

FIGURE 100-4 Contour of L3 metastasis in axial, saggital, and coronal projections. The epidural metastasis is in red.

FIGURE 100-5 Contour of L3 metastasis and spinal roots with superimposed isodose lines from treatment plan in axial, saggital, and coronal projections. The epidural metastasis is in red, the spinal roots are blue, and the 80% isodose line is represented by the thin green line.

FIGURE 100-6 Contour of L3 metastasis and cauda equina with superimposed isodose lines from treatment plan in axial, saggital, and coronal projections. The epidural metastasis is in red, the spinal roots are blue, and the 80% isodose line is the smaller green line.

During spinal radiosurgery, CyberKnife™ patients are placed on the operating table in their custom mask or cradle. Images are automatically obtained by the orthogonally placed x-ray cameras and compared with digitally reconstructed radiographs (DRRs) that were precalculated by the planning software. The table is aligned and the computer aims and delivers the first treatment beam. Additional images are obtained, automatic corrections for movement are made, and all planned treatment beams delivered in order. This process is automatic; after the initial targeting verification is performed by the treating neurosurgeon, radiation delivery is simply monitored by a radiation therapist. Not all commercially available systems permit periodic reimaging and beam retargeting. As a result, in the event of patient movement during treatment with such radiosurgical devices, accuracy may be compromised.

Indications

It cannot be emphasized enough that the indications for spinal SRS continue to evolve quickly; spinal SRS is a new subdiscipline within neurosurgery and only now are standardized procedures for its application being developed. At our institution, we most frequently treat metastases, small benign tumors, postoperative residuals, lesions that recur following conventional surgery or radiation, vascular malformations, inoperable tumors, and lesions in those who decline surgery (Tables 100-1 and 100-2).^3,7,13 Spinal lesions appropriate for CyberKnife™ radiosurgery should be reasonably well circumscribed, clearly visible on CT or MRI, and smaller than approximately 5 cm in diameter. We do not insist on obtaining a biopsy in advance of treatment if the diagnosis is clear from preradiosurgical imaging studies.

Table 100-1 Indications and Contraindications for Stereotactic Spinal Radiosurgery

Table 100-2 Treated/Treatable Lesions with CyberKnife Radiosurgery

Spinal SRS is contraindicated in the presence of significant spinal cord compression causing a severe neurologic deficit, especially when the treated lesion is relatively radioresistant. Radiographic evidence of spinal cord compression is not in itself a contraindication to spinal radiosurgery. Spinal SRS can be used as an adjunct if there is evidence of spinal instability. If the adjacent spinal cord has already received the maximum tolerated radiation dose, then surgery and/or chemotherapy may be more appropriate.

Extradural Metastases

The spine is the most common site for bony metastases, accounting for nearly 40% of osseous tumor spread.¹⁷ Forty percent of cancer patients will develop at least 1 spinal metastasis.¹⁸ Historically, spinal metastases have been managed with chemotherapy, radiopharmaceuticals, surgery, and external beam irradiation.^17,19 Conventional irradiation of spinal metastatic tumors is useful for palliation but its effectiveness is limited by spinal cord tolerance.²⁰ Moreover, relapses are common,^21–24 and retreatment with RT is generally impossible.¹⁸ SRS enables much larger biologically effective doses to be delivered by utilizing a more highly conformal plan that protects the cord. Multiple courses of spinal SRS can control multiple asynchronous metastases, and SRS may be used to sterilize a vertebral body before vertebroplasty²⁵ or following a debulking procedure. The presence of spinal fusion hardware is not a contraindication.²⁶ SRS is ideal for those with limited life expectancies or those who need other treatment. Most SRS treatments are completed in a single 1-hour session.

Treatment protocols in the published literature vary greatly and there is significant debate regarding the most appropriate treatment margins.²⁷ For purely bony lesions, Amdur et al.²⁷ recommend treating visible tumor plus a 1-cm margin in bone or a 2-mm volume outside the external cortex. For lesions within the canal, the margins are not extended beyond the visible tumor. Many groups irradiate the entire affected vertebral body including the pedicles. Chang et al.¹⁸ recommend treating the pedicles for possible tumor extension, pointing out that 18% of recurrences occurred in the pedicles. At Stanford we typically treat the tumor volume as seen on CT or MRI.¹⁷ There are no studies that show a clear benefit of one approach over the other, but those who treat smaller volumes argue that the cord dose is decreased and that recurrences can be retreated. Dose recommendations are variable with single-session prescriptions ranging from 8 to 24 Gy in the published literature.²⁷ At Stanford we have used from 16 to 25 Gy in up to five fractions, but usually opt to treat in the fewest possible sessions as dictated by the length and dose of irradiated spinal cord.¹⁷

Overall the efficacy of radiosurgery for spinal metastases appears roughly comparable to that for brain metastases.^2,3,17 Tumor growth is arrested by radiosurgery in up to 100% of cases and results are independent of histology (Table 100-3).^4,28 Pain relief ranged from 43% to 96% and unlike standard RT, is generally apparent within days of SRS.²⁷

Because of the delay between SRS and the involution of the tumor, radiosurgery is almost never an alternative to emergency decompression. Neither generalized metastatic involvement of the axial skeleton nor epidural carcinomatosis is an indication for radiosurgery, as SRS cannot cover such broad areas effectively. In the presence of instability, radiosurgery may be used as an adjunct to a stabilization procedure such as vertebroplasty. Delayed post-SRS spinal instability can occur but is not common.

Intradural Metastases

Intramedullary spinal cord metastases, which are rarely seen clinically, are found in 0.9% to 2.1% of autopsies in cancer patients.³³ Such lesions comprise 8.5% of central nervous system metastases³⁴ and their frequency will likely increase with longer patient survival. Wowra et al.²⁹ review their results and the literature. They report that 96% of spinal metastases were well controlled with spinal SRS. The risk of myelopathy was less than 1%.

Primary Intramedullary Lesions

At Stanford, we have treated 92 hemangioblastomas in 31 patients. Sixteen were spinal intramedullary tumors. Patients were treated with a median radiosurgical dose of 23 Gy, and after a median 34 months of follow-up, 15 of the 16 spinal hemangioblastomas in this series either remained stable or decreased in size. Among all hemangioblastomas, those causing edema and those associated with cysts did less well. None of the patients developed radiation myelopathy.³⁵

Some spinal ependymomas prove difficult to resect for reasons of anatomy or associated medical comorbidities. Although not common, tumor recurrence also occurs. There is very little information available regarding SRS. Ependymomas, which may be controlled with conventional radiation therapy, have responded favorably to spinal SRS in the few cases that have been published.^3,36,37 The Stanford experience with CyberKnife radiosurgery has been quite favorable with good local control and no complications (unpublished data). We know even less about the treatment of spinal astrocytomas. For the occasional well-demarcated, biopsy-proven, newly diagnosed or recurrent spinal cord astrocytoma, spinal SRS may be a theoretical alternative if surgery is not possible.

Intradural, Extramedullary Lesions

Meningiomas, schwannomas, neurofibromas, and hemangioblastomas are the benign lesions most frequently considered for treatment with spinal SRS. However, microsurgical resection remains the most appropriate intervention for most patients. Surgical resection is generally curative and in the process, the surgeon both establishes a tissue diagnosis and immediately decompresses the spinal cord.³⁸ Nevertheless, SRS is appropriate for benign lesions that are inaccessible, when lesions are numerous (as in neurofibromatosis or von Hippel-Lindau disease), in patients with significant medical comorbidities, or for patients who decline open surgery.^39–41 Moreover, spinal radiosurgery also has the virtue of posing little risk to the parent motor or sensory nerves in cases of nerve sheath tumors.

Short-term control rates for intradural, extramedullary spinal tumors appear comparable to those of similar intracranial lesions. At Stanford we have treated 110 patients with 117 lesions (unpublished data). Greater numbers and longer follow-up periods confirm earlier published observations.⁴² Following SRS, 56% of schwannomas and meningiomas stabilized and 44% regressed radiographically. Neurofibromas did less well with 11% enlarging radiographically, over 50% causing increased pain, and 80% showing a worsening in at least one examination finding. Even including neurofibromas, most myelopathies and radiculopathies improved with SRS, whereas bowel and bladder dysfunction did not. Two of our patients eventually required surgery for tumor enlargement and three required surgery for persistent or progressing symptoms. Only one patient developed radiation myelopathy. Other centers have reported similar results.^39–41

Vascular Malformations

Steinter et al.⁴³ published the first description of SRS for cranial arteriovenous malformations (AVMs) in 1972. The successes in treating the intracranial AVMs inspired the use of SRS for spinal vascular malformations. Spinal AVMs have been described using various systems. Most commonly they have been divided into four types.⁴⁴ Types I and IV are dural and perimedullary fistulas and are better treated with embolization and resection. Most AVMs treated with SRS are type II, or glomus AVMs, with a well-defined nidus. Some type III, or juvenile AVMs, may be amenable to SRS when well-focal and well-defined (Fig. 100-7). SRS causes endothelial damage that leads to the obliteration of the vascular lumena.⁴⁵ Low-flow lesions, such as cavernous malformations, are rarely candidates for spinal SRS.⁴⁶ Of 29 patients treated at Stanford between 1997 and 2009, 22 were followed more than 24 months (unpublished data). Most had glomus AVMs and one had a type III lesion. Sixteen patients presented with hemorrhage and 8 had more than one bleed. Twelve lesions were cervical, 8 were thoracic, and 3 were in the conus. The usual treatment dose was 16 Gy in one session or 20 Gy in two sessions (with 10 Gy delivered to adjacent spinal cord). Ten patients (43%) were previously embolized. All postoperative MRIs in treated patients showed a reduction in volume. Of the 8 patients who had post-SRS angiography, 3 had complete obliteration. None of the treated patients suffered a rebleed, including those where the AVM was not obliterated by MRI or angiography. There was no mortality. Symptoms improved in over 50% and only 3 patients reported worsening of symptoms. There was only 1 case of radiation myelopathy (3%).

FIGURE 100-7 A, Spinal cord AVM prior to treatment. Note the compact nidus. B, Spinal cord AVM 3 years after SRS. Complete angiographic obliteration noted.

Treatment Failures and Complications

Treatment failures fall into several groups. “In-field failures” refer to tumor regrowth within the treated volume and may be related to inadequate dosing. “Marginal failures” involve regrowth at the edges of the treated volume and may be related to poor imaging, underestimation of the tumor volume, or inaccuracies in position or setup. “Distant failures” are not complications but rather involve new lesions in untreated areas. For vertebral metastases, the chance of an asynchronous metastasis in an adjacent level is only 5%.⁴⁷ Furthermore, neurologic damage can be divided into three groups: (1) acute complications that occur within 1 month are usually due to edema, are transient, and are treated with steroids; (2) subacute complications that occur 3 to 6 months after treatment, may be secondary to demyelination, and usually recover; and (3) radiation myelopathy, which is the most feared complication and usually occurs after 6 months. The latter risk more than any other limits the radiosurgical dose used for most paraspinal lesions.⁴⁸ In conventional radiotherapy, it is generally believed that treatment with 45 Gy in 22 to 25 fractions is associated with an incidence of myelopathy of only 0.2%. Meanwhile, among the first 1000 patients treated with the CyberKnife for spinal lesions at both Stanford and the University of Pittsburgh over the past decade, only 6 developed myelopathy (0.6%). In large part due to this 10-year track record with radiosurgery, a re-evaluation of the conventional wisdom and long-standing radiotherapy guidelines pertaining to spinal cord tolerance is taking place. Nevertheless, at Stanford we generally seek to avoid exposing more than 1 cm³ of spinal cord to greater than 10 Gy in single-session plans.⁴⁹

Other complications are rare and, fortunately, less severe. Skin reactions are seen most commonly when the posterior elements are radiated; nausea pharyngitis, esophagitis, and diarrhea are related to gastrointestinal tract exposure. Renal complications, occasionally related to thoracolumbar SRS, are rare.

Summary

While the complexity of spinal lesions and their close association with the cord make operative treatment difficult, it also makes them ideal candidates for spinal SRS. SRS, although a recent development, is supported by a rapidly expanding literature. For many lesions of the vertebral bodies, and some intradural, extramedullary lesions, CyberKnife radiosurgery is clearly both safe and effective. For vascular lesions, the treatment is superior to embolization and surgery for AVMs. Early results show that treatment of selected intramedullary lesions is also possible. It is likely that the indications will expand and the quality of the results will improve as our experience increases.