The first step in developing spine radiosurgery is to determine targeting accuracy. Many phantom studies have demonstrated the clinical feasibility of achieving submillimeter accuracy [6]. The first human study of spine radiosurgery, however, showed the targeting accuracy and reproducibility of patient positioning were within 1.5 mm [7]. The dose fall off from the 90% to 50% isodose line was less than 5 mm towards the spinal cord. The overall procedure of spine radiosurgery includes patient positioning and immobilization, image acquisition, tumor and normal tissue delineation, radiation treatment planning, repositioning and treatment delivery under image-guidance.

The next step of applying radiosurgery to the spine is positioning and immobilization and are involuntary organ motions. There is no perfect method of immobilization. It seems important to make the patient feel stable and comfortable at the treatment position with minimal pressures at any body part. Experience with the spine radiosurgery has shown that breathing-related organ motion exists, but it does not significantly affect treatment outcome. There also potential voluntary or involuntary motions secondary to swallowing, pulsation or coughing. Some of these motions can be reduced with proper premedication. Important to the practice is to minimize any movements that can be controlled by premedication. One notable example is random movement due to spine pain, which can be minimized by proper pain medication before the procedure. Conscious sedation may further reduce most of the voluntary movements.

Although the concept of spine radiosurgery had been conceived even 50 years before the start of cranial radiosurgery, clinical application was limited mainly because of the difficulty with immobilization. In an attempt to address this issue, invasive procedure initially by anchoring hardware to the cervical spine and skull, and recently anchoring of the stereotactic frame to the lumbar spinous process under general anesthesia have been introduced [8]. Other less invasive immobilization devices used a body frame with contour mold fixation [9]. Most recently, a frameless and non-invasive positioning method for spine radiosurgery by Ryu et al. [7] and Yin et al. [10] has been developed. Most institutions now use the newer type of positioning devices for spine radiosurgery. It is important to use a method with which the patient feels comfortable while in the treatment position.

Each vertebral bone is composed of one compact bone marrow with a potential anatomic barrier at the pedicle area during the ossification phase. Although imaging studies sometimes show the gross disease within the vertebral body, it may not represent the full extent of gross tumor volume, i.e., known as GTV. Therefore, the combined volume based on the concept of clinical tumor volume (CTV) incorporating GTV should be the target for spine radiosurgery, which ultimately is the entire vertebral body of the involved spine. The method of target delineation adopted in the RTOG 0631 trial is illustrated in Fig. 1. There can be many different scenarios of spine involvement. In general, spine metastasis is modeled on three different scenarios: the most common case involves the vertebral body alone (Fig. 1A), involvement of the vertebral body extending to the pedicles (Fig. 1B), or involvement of the dorsal elements, such as spinous process and lamina (Fig. 1C). Lesions more extensively involving the vertebral body and pedicles are either treated with a generous margin (dotted line) or treated including both anterior and posterior elements (solid line) as shown in Fig. 1B.

The most critical normal structure for spine radiosurgery is the spinal cord. Planning simulation CT images using 1-3 mm slice thickness are used for delineation of the spinal cord with MR image fusion. T1-weighted contrast-enhanced and T2-weighted images are useful. At the author's institution, the spinal cord volume has been consistently defined as the volume extending from 6 mm from above the target volume to 6 mm below the radiosurgery target. This method of spinal cord volume delineation is diagrammatically shown in Fig. 2. The rationale for this denominator spinal cord volume is based on the distance of dose fall-off (90% to 50% isodose line) being 5 mm, and that the radiosurgical beam arrangement is co-planar [7]. Using this spinal cord volume as denominator, the spinal cord partial volume tolerance dose has been defined as 10 Gy to the 10% of the spinal cord volume defined from 6 mm above to 6 mm below the radiosurgery target volume [3]. When the cord volume receiving greater than 10 Gy was calculated, it was found to be 0.35 mL. A careful precaution is advised in applying the absolute volume tolerance dose because there can be a broad individual variation in spinal cord diameter across patients.

It is also recommended to delineate the normal tissues near the target. These include mucosal structures of laryngopharynx, trachea, and esophagus, as well as bowel and kidney. Recommended normal tissue doses have been also published [11,12].

There has been virtually no treatment algorithm in the evaluation and management of spine metastasis and/or cord compression because it was only approached with the simple thought of palliation. A few treatment algorithms [13,14] focused only on the surgical treatment, and failed to take into account recent advances in treatment thus rendering them less useful. A recent interdisciplinary decision-making framework has been proposed with an acronym of NOMS, which incorporated components of neurologic, oncologic, mechanical instability, and systemic disease [15,16]. The neurologic consideration assesses the presence of myelopathy and functional radiculopathy, but the most critical determinant is the degree of epidural spinal cord compression. The oncologic assessment is predicated on the expected tumor response and durable control to any available treatments as well as consideration of the natural behavior of primary histology. Mechanical instability is a critical decision point as no amount of radiation or chemotherapy will stabilize an unstable spine. A scale of spinal instability secondary to pathologic fractures, called Spinal Instability Neoplastic Score (SINS) has been proposed examining the radiographs and clinical histories with location of neoplasm, characterization of pain, type of bone lesion, radiographic spinal alignment, degree of vertebral body collapse, and involvement of posterolateral spinal elements [17]. Once instability is determined, the initial treatment is typically an interventional procedure such as, open surgical stabilization, percutaneous cement augmentation and/or pedicle screws. However, this also has to be determined based on the extent of systemic disease and medical comorbidities that impact the ability to deliver the proposed treatment and the expected survival.

The primary goal of spine radiosurgery is to maximize local tumor control of the involved spine while preserving neurologic function. Like the brain radiosurgery, solitary spine metastasis ('a') is the obvious indication as shown in Fig. 3. Any soft tissue extension of the tumor causing epidural compression or a paraspinal mass is included in the target volume. Two contiguous spine levels can also be treated with radiosurgery ('b' in Fig. 3). In the same token, radiosurgery can be used to treat separate isolated spine metastases that are not adjacent to one another ('c' in Fig. 3). With more common use of MRI for making diagnosis of spine metastasis, there often are multiple micrometastases within the spinal column. These patients are also eligible for radiosurgery when there are limited (1-3) numbers of gross involvement ('d' in Fig. 3). These 4 clinical conditions are included in the ongoing randomized RTOG 0631 trial. In the author's institution, radiosurgery is also utilized for symptom control (i.e., palliative purpose) in patients with diffuse spine metastasis if the symptomatic site was identifiable radiographically and clinically at one or two spine segments. There may be no limited number of spine metastases that can be treated with radiosurgery. It requires caution since the spinal cord dose tends to be higher when the length of target was >6 cm [2], although it can be minimized with the use of radiation intensity modulation.

In the process of making a treatment decision, it is important to view the natural process as a spectrum ranging from vertebral bone involvement to mild or moderate epidural compression to frank spinal cord compression. One has to also consider rapidity and severity of neurological symptom development, oncological condition, medical and general condition including comorbidity as well as unique spine problems, such as degenerative change and instability particularly in the elderly spines. Recently, MRI has been used to visualize the extent of such involvement. Treatment decision to treat with surgery is commonly made based on the imaging study rather than neurological status. However, imaging studies are not predictive of the patient's neurological condition. With the use of modern radiosurgery, there are currently two grading systems proposed; one is proposed by Bilsky et al. [18] and it is based on MR imaging study, and the other is based by Ryu et al. [19] and it is based on both MR imaging and neurological status. While no grading system is perfect, the proposed systems provide physicians with a common language across disciplines. Since any imaging study alone is not predictive of neurological status, dual grading system is preferred for evaluation of spinal cord compression. The dual radiographic (anatomical) and neurological (functional) grading system can be used to directly assess the patient and select radiosurgical versus neurosurgical candidates. These are summarized in Table 1, and diagrammatically illustrated in Fig. 4 with examples of images.

The radiographic grade is based on the imaging studies, based on T1-weighted contrast MR image for epidural tumor and T2-weighted image for assessment of the thecal sac and the spinal cord. The radiographic grade of the lesion ranges from 0 to V. Grade 0 is vertebral bone involvement only without canal compromise. Grade I represents minimal involvement of the epidural fat but without thecal sac displacement. Grade II is impingement and/or displacement of the thecal sac. Grade III represents impingement of the spinal cord but without significant distortion or displacement of the spinal cord. Grade IV represents partial spinal cord compression and displacement with cerebrospinal fluid (CSF) still visible within the thecal sac (so-called partial block on the myelogram). Grade V represents severe spinal cord compression with CSF not visible at the level of compression (complete block). At the level of the cauda, an epidural lesion is deemed grade II when less than 50% of the spinal canal is compressed, and grade IV when more than 50% of the canal is compressed.

The neurological grade is based on the neurological symptoms and the status of the patient at the time of presentation. The neurological grade also consists of five scales (from a to e) based on the neurological status: a , no symptoms; b, a focal minor symptom (i.e., axial pain or radiculopathy); c, functional paresis (i.e., the involved neuromuscular elements compromised but remain partially functional) with muscle strength of ≥4 out of 5 either due to compression of nerve roots or spinal cord; d, nonfunctional paresis (i.e., the involved neuromuscular elements are nonfunctional) with muscle strength ≤ 3 out of 5 either due to compression of nerve roots or spinal cord; and e, complete paralysis or urinary and rectal incontinence.

Treatment options for spinal cord compression remain the same. Steroids are used for acute improvement and prevention of neurological symptoms. cEBRT is also used for most of the widespread spine metastases with or without canal compromise. The role of surgery has been demonstrated. Surgery is to achieve decompression of rapidly developing spinal cord compression, and restore the neurological function. The goal of spine radiosurgery is to preserve and improve the neurological status. Therefore, it is reasonable and logical to consider radiosurgery when the patient's neurological condition is well preserved. Ironically, this is quite opposite of what is actually being done in clinical practice at this time, i.e., surgery is used for neurologically intact patient. Comparative neurological outcome results of surgery versus radiosurgery are described in the Section 5.2. Radiosensitivity of the tumor did not make any difference. It is clear that the surgery is indicated when there is neurological deficit, but radiosurgery can be the choice of treatment for any patients who are neurologically intact or ambulatory or have a minor deficit, i.e., for patients with neurological grade c or better.

The usual presenting symptom of spine metastases is back pain. Therefore, the purpose of spine radiosurgery has been primarily for pain control. Pain control of spine metastasis by radiosurgery has initially been reported in the range of approximately 90% using different radiosurgery methods [20,21]. The optimum dose of radiosurgery necessary to achieve pain control is not clearly defined at this time. In the analysis of an early phase II trial of dose escalation at Henry Ford Hospital with dose escalation, there was a strong trend towards increased pain control with a radiation dose of higher than 14 Gy. The one year actuarial pain control rate was 84% with doses higher than 14 Gy delivered in a single fraction [21]. Successful pain control has been reported consistently higher than 90% with the use of various radiosurgery dose in the range of 16.24 Gy regardless of the tumor histology [22,23,24]. This is considerably higher than the average pain control rate of 50%.60% by single dose conventional radiation of 8.10 Gy [25,26]. The current RTOG trial directly compares the conventional 8 Gy (3-dimensional treatment allowed) versus radiosurgery 16 or 18 Gy.

Durability of pain control after radiosurgery is another important parameter. In patients with solitary spine metastasis treated with radiosurgery, the actuarial median duration of pain control was 13 months [21]. The durable pain control after radiosurgery is also superior to pain control after lower dose radiation with 8 Gy where majority of patients experience recurrent pain after 3 months [26], as was demonstrated in RTOG 9714.

Treatment failures after spine radiosurgery can be divided into three different categories. In-field failure refers to tumor regrowth inside the target volume. This may be related to the radiosurgical dose necessary for tumor control or inherent radiosensitivity of the tumor. Marginal failure is a failure within the region of rapid dose fall-off immediately outside of the target volume. Causes of marginal failure could be geographical miss or patient set up error, or underestimation of target volume. Distant failure is due to progressive metastatic spread involving the untreated vertebra in the spinal column.

Results of failure after radiosurgery showed an infield failure rate of 6%. More importantly, there was only a 5% incidence of failure at the immediately adjacent spine (marginal failure) [32,33]. The low incidence of infield or marginal failure justifies the use of radiosurgery for localized spine metastasis. Though persistent or progressive pain may appear to be attributed to tumor progression at the treated spine, other causes such as spine instability and degenerative disorder should also be ruled out. It is advisable to minimize the likelihood of developing symptoms secondary to structural problems in the spinal column.

Acute worsening of pain can occur in about 20% range. This pain flare may occur within 24 hours of single dose of radiosurgery, or after a few days with fractionated regimen [34,35]. It is usually transient. It is seen as a local sharp pain or radicular in nature. Short-term use of glucocorticoid can control this pain flare. Prophylactic use of steroid is not indicated. The cause of pain flare is not understood. It may be due to acute radiation-induced inflammation of the nerve roots. It is advised to define the nerve roots and minimize the radiation dose in these areas.

The potential consequences of radiation-induced spinal cord injury can be severe. The best way to prevent radiation-induced myelopathy is to avoid unnecessary or excessive radiation to the spinal cord. The tolerance dose of the spinal cord using traditional external beam radiation is 45.50 Gy in 25 fractions given across the entire diameter of the spinal cord. Biologically effective dose (BED) calculation has commonly been performed to estimate the radiosurgical tolerance dose. However, this is not applicable to spine radiosurgery because there is a sharp dose fall-off within the spinal cord.

In a large clinical trial of 230 patients with careful documentation of spinal cord dose, the spinal cord partial volume tolerance dose has been defined as 10 Gy to the 10% partial volume of the spinal cord defined from 6 mm above to 6 mm below the radiosurgery target volume [2]. When the absolute volume receiving higher than 10 Gy was calculated, it was found to be 0.35 mL. A careful precaution is advised in applying the absolute volume tolerance dose because there is potentially a large individual variation in spinal cord diameter across patients. See Section 3.3 above. The ongoing RTOG 0631 randomized trial adopted this dose constraint. There was one incidence of spinal cord damage clinically and radiographically 13 months after 16 Gy radiosurgery for C1 and C2 epidural compression from the primary breast cancer. The patient improved with steroid treatment. Since spinal cord complications can occur after a long latency period, long-term follow-up is certainly needed.

Vertebral compression fracture has been reported with approximately 20% incidence. It is more often seen in lytic metastases on imaging studies involving more than half of the vertebral body, with radiosurgical dose >20 Gy [36,37]. For compression fractures, one has to also consider whether it is caused by oncological condition or other degenerative changes particularly in elderly spines, and whether it is symptomatic or not. It is important to remember that compression fracture is not a contraindication of spine radiosurgery. Instead, it is important to consider the entire clinical picture, which includes spine stability and general oncological condition.

Normal tissues should be delineated depending on the level of the involved vertebral body, such as the larynx, pharynx, esophagus, bowel, lung, or kidney, etc. The tolerance dose of these organs to radiosurgery is not defined. Experience of certain organ tolerances have been reported, such as for mucositis of the pharynx or esophagus in cervical or upper thoracic spine treatments [11,12]. These symptoms manifest as odynophagia or dysphagia. Acute side effects associated with mucositis generally resolve within two weeks. Long-term tracheoesophageal fistula has also been reported in an immunocompromized patient with multiple myeloma after multiple chemotherapy regimens and bone marrow transplantation. Skin reactions are seldom seen after radiosurgery unless the tumor involves the posterior element extending close to skin. Every effort should be given to minimize any unnecessary radiation dose to the adjacent tissues.

There are many factors that may influence radiosurgical complications. These may include proximity and the extension of the tumor to the adjacent normal tissues, concurrent systemic therapy, compounding comorbidities such as acute infection, prior surgery, and host factors such as diabetes, collagen vascular disease, or any genetic predisposition to the radiosensitivity. A careful assessment of the patient's clinical condition and dosimetry of radiosurgery are necessary in order to reduce the potential risk of radiation-induced complications.