Epidemiology and Genetics

Primary tumors involving the spinal cord are uncommon¹ and tend to occur in young patients. In adults, they are outnumbered by primary brain tumors at a ratio of roughly 20 : 1; for children, a ratio of 5 : 1 is reported by most centers. Tumors arising from the spinal canal and vertebrae show a more heterogeneous age distribution. The rarity of primary spinal tumors makes it difficult to establish a clear-cut association with specific cytogenetic abnormalities. Some investigators have found an association between spinal and brain ependymomas and the loss of sequences in chromosome arm 22q, suggesting the presence of one or more ependymoma tumor suppressor genes in this location.^2–4 Pediatric ependymomas and astrocytomas have been associated with loss of chromosome arm 17p.^3,4 Some myxopapillary and papillary ependymomas have been associated with bcl-2 oncoprotein expression.⁵ Other ependymomas have been associated with mutations in the MEN1 gene and loss of heterozygosity in the chromosome arm 11q region.⁶

Prior exposure to therapeutic irradiation probably increases the risk for developing spinal canal meningiomas, soft tissue sarcomas, vertebral body sarcomas, and possibly spinal cord gliomas, although the magnitude of this risk is unclear. Neurofibromatosis is strongly associated with an increased incidence of spinal canal neurofibrosarcoma and has been associated with both intracranial and spinal cord ependymomas as well.² This latter association may be indirect, with the actual cause being the loss of a separate ependymoma tumor suppressor gene located near the NF2 gene.³

Anatomy

The spinal cord extends from the level of the foramen magnum to the L1-L2 vertebral level in most adults. Below the termination of the cord, the spinal canal contains the lumbar cistern, an enlargement of the subarachnoid space that surrounds the cauda equina. The spinal subarachnoid space terminates inferiorly at the S2-S3 level and does not extend laterally beyond the spinal canal. The subarachnoid space limits the volume at risk for harboring metastases borne by the cerebrospinal fluid (CSF). Below S2-S3, the spinal canal forms an extradural space, which continues inferiorly into the coccyx.

The cord proper lies within the spinal canal, with the dura intimately applied to the bony and ligamentous structures forming the periphery of the canal. In addition to the cord, the spinal canal includes the subarachnoid space, vessels, nerve roots, and meninges. The spinal canal is encased by the elements of the vertebrae and intervertebral ligaments, with the vertebral bodies located anteriorly; the pedicles, lamina, and transverse processes laterally; and the spinous processes posteriorly.

Tumors involving the spinal cord have often been classified by their anatomic relationship relative to the dura. The majority of extradural lesions are metastatic. Intradural, extramedullary tumors are roughly evenly divided between primary and metastatic tumors, while intradural, intramedullary tumors are most often primary gliomas.

Approximately 10% of all primary tumors involving the cord arise from the vertebrae, 65% from the spinal canal, and 25% from the cord proper. For purposes of treatment, it is useful to group primary tumors based on whether they arise from the cord, spinal canal, or vertebrae (Table 26-1). This system helps to predict whether radical surgery is feasible. It also suggests whether the normal spinal cord must be included within the high-dose irradiation volume or could instead be partially excluded from treatment with the use of sophisticated irradiation techniques.

Table 26-1 Common Primary Tumor Types Involving the Spinal Cord

Pathology

The histology of spinal tumors correlates strongly with location. Primary tumors of bone (osteogenic sarcoma, chondrosarcoma, and chordoma) comprise most primary vertebral tumors aside from myeloma or plasmacytoma. Chordomas in particular are concentrated in sacral locations. Soft tissue sarcomas (particularly nerve sheath tumors such as malignant schwannoma or neurofibrosarcoma) and meningiomas comprise most spinal canal tumors. Astrocytomas and ependymomas account for the majority of primary cord tumors, with oligodendrogliomas occurring rarely. Other histologies such as vertebral body hemangiomas, lipomas, and arteriovenous malformations (AVMs) are occasionally encountered. Cauda equina ependymomas account for 60% of all spinal ependymomas and are generally included with intramedullary ependymomas when radiotherapy results are reported. Although spinal ependymomas (cord and cauda) outnumber astrocytomas 2 : 1, intramedullary tumors are almost evenly divided between ependymomas and astrocytomas.

Grade is an important consideration for several tumor types. Low-grade spinal canal sarcomas appear to be at low risk for both local and distant recurrence if completely resected. High-grade, malignant meningiomas are probably at higher risk for recurrence than the more typical benign meningiomas. The importance of grade is uncertain for ependymomas, as some series have identified grade as a significant prognostic factor,^7–9 and others have not.¹⁰ The overall risk of subsequent intracranial failure for low-grade spinal ependymomas is low, on the order of 7%.¹¹ However, this value is probably higher for histologically malignant ependymomas.^12,13 Myxopapillary ependymomas reportedly have a particularly favorable prognosis.⁹

Grade is clearly the most important factor influencing outcome for spinal astrocytomas.^14,15 Although prolonged survival is generally the rule for patients with low-grade astrocytomas, malignant astrocytomas such as glioblastoma multiforme (GBM) and highly anaplastic (Grade 3) astrocytomas (HAAs) are aggressive tumors characterized by rapid local recurrence following treatment, craniospinal axis dissemination, and short survival.^13–17 The pilocytic astrocytoma subtype reportedly has a better prognosis than the diffuse fibrillary astrocytoma subtype.¹⁸

Clinical Presentation

Patients with primary spinal tumors generally present with local or distal neurologic symptoms produced by spinal cord dysfunction. Local symptoms often include focal pain, segmental or nerve root weakness, and sensory deficits corresponding to a dermatomal distribution. These symptoms and signs usually correspond closely to the level of the lesion within the cord. Distal symptoms are produced by involvement of the cord’s long tracts, creating paresis, sensory deficits, and autonomic dysfunction diffusely below the level of the cord lesion. Depending on the location of the lesion, long tract signs may be relatively lateralized (e.g., the Brown-Sécquard syndrome). Since the cord is substantially shorter than the spine in adults, an intrinsic cord lesion is generally located anatomically a few vertebral bodies above the clinical cord level (e.g., a T8 cord segment astrocytoma may lie at the level of the T6 vertebral body).

Routes of Spread

Primary cord tumors, particularly high-grade ependymomas and malignant astrocytomas, can spread via the subarachnoid CSF throughout the craniospinal axis. Low-grade astrocytomas and ependymomas generally remain localized. Primary tumors arising from the spinal canal and vertebrae occasionally spread hematogeneously, with the lungs being the most frequent metastatic site. Lymphatic metastases are rarely encountered with any primary spinal tumor.

Diagnostic and Staging Studies

The workup for primary cord tumors is chiefly radiographic. Magnetic resonance imaging (MRI) with gadolinium for contrast is the imaging study of choice.¹⁹ Standard or metrizamide-enhanced computed tomographic (CT) myelography is indicated for patients unable to undergo MRI (e.g., pacemaker-dependent patients). Intramedullary tumors usually extend for multiple segments of cord, and sagittal MRI is particularly useful for delineating the rostral and caudal extent of tumor. Imaging of the entire craniospinal axis is required for all ependymomas, because spinal axis metastases develop in 5% to 15% of intracranial ependymomas.²⁰ These spinal “drop metastases” can produce the initial symptoms of a disseminated, primary intracranial ependymoma.

The entire neuraxis should also be imaged for cord HAAs and GBMs. Low-grade spinal astrocytomas and oligodendrogliomas do not require imaging of the entire craniospinal axis. CT scanning of the chest to rule out pulmonary metastases should be performed for vertebral osteogenic sarcomas and chondrosarcomas, as well as for spinal canal neurofibrosarcomas. Other tests such as angiography may be of limited use for AVMs or meningiomas. CSF cytology may be informative, but its use in determining treatment or influencing outcome is controversial. No chemical markers appear to be clinically useful in the diagnosis or follow-up of spinal tumors. The role of the bromodeoxyuridine (BrdU) and Ki-67 labeling indices as prognostic indicators for spinal ependymomas is under investigation.^21,22 An elevated cyclin D1 labeling index, high MIB-1 proliferation index, and p53 immunolabeling might be indicative of high grade in ependymomas, although a correlation with clinical outcome is unclear.^5,23

Staging System

Presently there is no commonly accepted TNM-type staging system for primary cord tumors. Primary vertebral body tumors and spinal canal tumors should be staged using the American Joint Committee on Cancer (AJCC) staging systems for bone and soft tissues, respectively.²⁴

Standard Therapeutic Approaches

Surgical exploration combining maximal resection (preferably en bloc) with minimal risk of iatrogenic injury is the preferred initial treatment procedure. In addition to decompressing the cord, essential histologic information is obtained. There is no role for preoperative irradiation or irradiation without a prior attempt at histologic diagnosis. As many as 4% of radiographically diagnosed “cord gliomas” have proven to be benign upon pathologic examination, with infarcts, demyelinating processes, amyloidosis, and sarcoidosis comprising some of the disease processes which mimic gliomas.²⁵

The anatomic location and histology of the tumor have important implications for treatment. Complete resection of primary vertebral body tumors is infrequent; the radiation oncologist must generally deal with residual macroscopic tumor requiring radiation doses greatly exceeding normal spinal cord tolerance. Charged particle beams are generally employed in this setting. Completely resected, low-grade spinal canal sarcomas and meningiomas generally do not require postoperative irradiation. Subtotally resected meningiomas may be managed with observation, conventional megavoltage irradiation, or charged particle beams. Charged particle beams are preferred to irradiate subtotally resected or high-grade sarcomas of the spinal canal.

The extent of resection is variable for primary cord gliomas. Using modern microsurgical techniques with intraoperative monitoring, complete resection of low-grade gliomas has become more common, and outcome has been successful without the use of adjuvant radiotherapy.^26,27,97 Postoperative megavoltage irradiation provides favorable long-term results following subtotal resection. Craniospinal irradiation is employed in the management of malignant as well as benign, multifocal ependymomas.

The favorable location of ependymomas of the cauda equina usually permits complete resection. Intramedullary ependymomas frequently have tissue planes separating the tumor and cord that facilitate complete resection, as well. Low-grade astrocytomas and oligodendrogliomas are infiltrative and generally lack tissue planes separating the tumor from normal cord. Aggressive resection in this situation can result in substantial neurologic injury, so subtotal resection or biopsy is often performed. However, aggressive resection has been performed successfully in some expert hands.²⁸ If frozen sections suggest HAA or GBM, resection is usually abandoned after biopsy is performed.

The treatment approach for primary spinal canal and vertebral tumors differs from that for primary cord tumors because a small distance (often on the order of millimeters) physically separates the tumor from the cord. This distance often allows aggressive surgery without an unacceptably high risk of cord injury.

Chemotherapy is not indicated in the management of low-grade cord gliomas, although concurrent treatment with carmustine (bis-chloroethyl-nitrosourea [BCNU]) or temozolomide might be useful with HAAs or GBMs. Osteogenic sarcomas arising from the vertebrae should be treated with aggressive chemotherapy, as performed in other osseous sites. The role of adjuvant chemotherapy for other soft tissue and bone sarcomas is less clear, but in view of the high risk of systemic recurrence, might be considered for high-grade lesions. Adjuvant chemotherapy with procarbazine, lomustine (chloroethyl-cyclohexyl-nitrosourea [CCNU]), and vincristine (PCV) may be considered for malignant (but not benign) meningiomas.

Role of Radiation Therapy

The radiation oncologist faces four basic clinical situations when evaluating a patient with a primary cord glioma:

The radiotherapeutic approach is different for each situation (Table 26-2). Postoperative irradiation is not indicated for completely resected low-grade gliomas.^26,29,30 Incompletely resected low-grade astrocytomas and ependymomas are generally treated with postoperative focal irradiation. High-grade (malignant) astrocytomas are palliatively irradiated with focal fields. Malignant ependymoma and multifocal, benign ependymomas are treated with curative-intent, craniospinal irradiation.

Table 26-2 Radiotherapeutic Management of Primary Spinal Tumors

HAA, Highly anaplastic (Grade 3) astrocytoma; GBM, glioblastoma multiforme; XRT, x-ray therapy.

* Total dose prescribed using standard (1.8-Gy) once-daily fractions with megavoltage photons unless otherwise specified.

† Includes astrocytoma, ependymoma, and oligodendroglioma.

‡ Megavoltage photon irradiation.

§ Protons or helium ions.

Completely resected spinal canal meningiomas require no additional therapy, since the risk of recurrence is only on the order of 6%.³¹ However, subtotally resected meningiomas have a higher risk of local recurrence, making postoperative irradiation a reasonable consideration.

Nerve sheath sarcomas are approached in a similar manner: postoperative irradiation is generally withheld if the tumor has been completely resected. Particle beam irradiation is given if macroscopic tumor remains. Incompletely removed vertebral chondrosarcomas and chordomas are likewise irradiated with charged particles (see Table 26-2). Osteogenic sarcomas are treated with a combination of chemotherapy, surgery, and postoperative particle beam irradiation if complete resection is not achieved.

Simulation

CT-based simulation (with or without MRI fusion) is presently the preferred technique. When possible, intensity modulated radiation therapy (IMRT) should be used to treat spinal tumors with curative intent, due to the advantages obtained by delivering a more homogeneous dose distribution across the high-dose target volume, and reducing the dose delivered to nearby critical structures, including the heart, lung, kidneys, gastrointestinal tract, and liver. Where IMRT is unavailable conventional treatment techniques are employed, beginning with a 2D or CT-based simulation before the start of irradiation.

Low-grade cord gliomas are treated with megavoltage photon fields encompassing the radiographically apparent lesion (gross target volume [GTV]) with a 3- to 5-cm margin of normal spinal cord (or brainstem for high cervical lesions) both rostrally and caudally (clinical target volume [CTV]). The preoperative sagittal MRI is the most useful study for determining the size and location of the GTV. Field width for non-IMRT cases rarely needs to exceed 7 to 8 cm, since only the cord requires irradiation. Sophisticated immobilization devices are useful when treating with IMRT, and thermoplastic face masks are useful for conventionally-treated tumors located in the cervical spine. A typical clinical, non-IMRT setup for a patient with a thoracic cord glioma is shown in Fig. 26-1.

FIGURE 26-1 • Clinical setup showing cephalad and caudal field borders for a localized low-grade thoracic spinal cord glioma.

Whether to include an associated syrinx (a dilated, fluid-filled intramedullary cavity) in the treatment volume is controversial. At times, the syrinx results from local mass effect causing obstruction and dilation of the central canal of the spinal cord; in this situation, the syrinx is not actually part of the neoplastic process, but instead represents a reaction of normal tissue. At other times, the tumor itself may be from a cystic cavity or syrinx as part of the neoplastic process, and in this situation the syrinx must be included in the GTV treatment volume. Clinically distinguishing these situations from one another is often difficult; consultation with the neurosurgeon regarding his or her impression at the time of surgery is often beneficial. In general, a small syrinx can easily be included in the treatment volume. An extensive syrinx extending for virtually the entire length of the cord can be technically difficult to encompass with conventional treatment fields. IMRT can be particularly beneficial in this situation (Fig. 26-2).

FIGURE 26-2 • A, Beam’s eye view achieved using three-dimensional conformal reconstruction of a thoracic cord glioma target volume and adjacent lungs with a posterior oblique wedged-pair irradiation technique. Custom low melting point alloy blocking or multileaf collimation is designed using the beam’s eye view of the target volume. B, Transverse section showing dosimetry from the treatment plan developed using the target volume derived from A. Note the substantial portion of lung included in the beam’s exit path at this level. C, Dose-volume histogram derived from the treatment plan shown in B using dosimetry information for the entire treatment volume. Virtually the entire target volume receives 100% of the total dose of 50.4 Gy at 1.8 Gy per fraction. Despite the impressive amount of lung included in the exit beam shown in B, the histogram shows that only 10% to 15% of each lung actually receives more than 25 Gy (50% of the total dose) under this plan. Nonetheless, 40% of the left lung and 60% of the right lung receives a dose greater than 15 Gy (30% of the total dose). See also Fig. 26-2. D, Verification simulator film taken with the beam’s eye view geometry based on the treatment plan developed in A, B, and C. E, Port film taken during treatment on the linear accelerator using the treatment field shown in D confirms the accuracy of the daily patient setup.

When IMRT is not utilized, tumors in the upper cervical spine are generally treated with opposed lateral fields to avoid unnecessary irradiation of the aerodigestive mucosa. Tumors located more caudally are often treated with differentially weighted anterior-posterior–posterior-anterior (AP-PA) fields using compensators. Use of beam-split abutting opposed lateral and AP-PA fields to treat tumors at the cervicothoracic junction is discouraged; the risk of overdosing the normal cord and underdosing the tumor owing to setup variation or error far outweighs any advantage in reducing acute morbidity. Again, IMRT is of particular benefit for tumors located in this region, eliminating the problems of possible overlap/underdosing inherent in the use of abutting fields (Fig. 26-3). When the use of abutting fields is unavoidable, a matching beam-split technique using independent jaws is required. The match line should be shifted 1 cm after every 10 Gy to minimize the risk of inadvertent overlap. A medulloblastoma-type “gap” technique should only be used when treating the craniospinal axis, and the gap should never be located over a site of macroscopic disease.

FIGURE 26-3 • The advantages of intensity modulated radiation therapy (IMRT) are highlighted by this difficult low-grade astrocytoma involving the thoracic cord, associated with a neoplastic syrinx extending cephalad through the cervicothoracic junction into the cervical cord. These three images demonstrate the superior dose-homogeneity IMRT provides within the high-dose treatment volume and the substantially greater sparing of adjacent normal structures throughout the treatment volume (compare the isodose curves involving the lung with the discussion in Fig. 26-2). Conventional treatment techniques are not able to provide this level of precise dose-localization, although, theoretically, stereotaxic body radiosurgery would provide similar advantages. A, Transverse image. B, Coronal image. C, Sagittal image. The 5040 cGy isodose line is shown in dark blue.

Wedged pair, posterior, oblique fields offer the theoretical advantage of decreased morbidity by minimizing exit dose in any given location; however, the planning and setup are technically more difficult, and scrupulous verification is necessary to ensure adequate coverage during treatment. Care must also be taken with this technique to ensure that the kidneys, liver, or substantial portions of the lungs are not irradiated beyond tolerance. The use of three-dimensional (3D) conformal treatment planning with dose-volume histograms has greatly improved the reliability and safety of this technique (see Fig. 26-2).

Caudal ependymomas have historically been treated with generous fields extended inferiorly to encompass the entire thecal sac and laterally as far as the sacroiliac joints to encompass the sacral nerve roots. The traditional justification for using such extended fields has been to comprehensively irradiate the subarachnoid space, which many incorrectly believe extends laterally along the sacral nerve roots. However, the subarachnoid space terminates inferiorly at the S2-S3 level and does not extend laterally beyond the confines of the spinal canal. Enlarging the fields laterally beyond the bony canal only serves to unnecessarily irradiate a larger volume of normal tissue. Similarly, CSF-borne metastases from the cauda would be expected to spread from the primary tumor in a cephalad direction as well as caudally. It is logically inconsistent to extend fields caudally well beyond the termination of the subarachnoid space, while arbitrarily stopping fields only a few vertebral bodies cephalad to the primary tumor, where no similar barrier to spread exists.

It is worth noting that a series from Iowa with small numbers of cauda ependymomas found that adjuvant thecal sac irradiation was beneficial when piecemeal rather than en bloc resection was performed.³² However, several other series found no advantage to the use of traditional extended fields over focal fields.^8,33 Following an uncomplicated resection, the use of fields extended inferiorly and laterally to encompass the entire sacrum appear unnecessary from both a clinical and theoretical standpoint.

Craniospinal axis irradiation using a medulloblastoma-type technique is advised for malignant ependymomas, as well as for multifocal, benign ependymomas. Care should be taken to ensure that the matchline does not cut through macroscopic disease, and that the matchline is moved 1 to 2 cm every 10 Gy to minimize inadvertent setup overlap causing overdosage of the normal cord. Focal boost fields with 3- to 5-cm craniocaudal margins should be applied to macroscopic disease sites after the entire neuraxis is treated. Prophylactic craniospinal irradiation is not advised for GBMs or HAAs despite their propensity for CSF dissemination. Since local failure is inevitable, the increased morbidity of neuraxis irradiation to control microscopic, asymptomatic disease is inappropriate. Focal fields encompassing the symptomatic, macroscopic disease with a 3- to 5-cm margin are used instead.

Spinal canal meningiomas are treated with focal fields encompassing the preoperative tumor volume with a 3- to 5-cm margin in cephalocaudal dimension, usually using megavoltage beams. Once again, IMRT is extremely useful in this situation due to improved dose homogeneity across the CTV.

Spinal canal sarcomas as well as sarcomas of bone generally require extremely high doses of radiation, far exceeding the tolerance of the normal cord, for local control to be achieved. Charged particle beams (protons and helium where available) are generally considered to be the standard of care in this setting, as they have the physical ability to minimize the dose delivered to the adjacent normal spinal cord while adequately encompassing the tumor, which is generally located only a few millimeters away. A technical description of such treatment techniques is available in Chapters 68 and 69. The development of image-guided/intensity modulated photon radiation therapy (IGRT/IMRT) offers a new, promising alternative to charged particle beams in this situation.¹⁰¹

Radiation Treatment Plan

Preoperative sagittal MRI and operative findings are the most useful studies defining the volume to be irradiated. When CT-based 3D conformal treatment planning is unavailable, contours should be taken at several levels, particularly in the lumbosacral region, to ensure homogeneity of dose throughout the length of the field. 3D conformal planning using CT scanning with magnetic resonance fusion is mandatory whenever IMRT or charged particle beam therapy is undertaken for spinal canal and vertebral body primary tumors.

Low-grade astrocytomas and benign ependymomas are treated with daily fractions of 1.8 Gy to a total dose of 50.4 Gy over 5.5 weeks.

Malignant ependymomas and benign multifocal ependymomas are treated with 1.8-Gy fractions. The entire craniospinal axis is electively treated to a dose of 45 Gy, followed by a boost to macroscopic disease sites to 50.4 to 54 Gy total. Treating the craniospinal axis with lower doses (on the order of 20 Gy) provides no benefit compared to using localized radiation fields alone.⁹⁰

Malignant astrocytomas (HAA, GBM) are treated palliatively with focal fields using 1.8-Gy fractions to a total dose of 54 Gy.

Unless observation is chosen, subtotally resected spinal canal meningiomas may be irradiated in one of two ways: either 1.8-Gy daily fractions to a total dose of 50.4 to 54 Gy, or with twice-a-day fractions of 1 Gy (minimum 5- to 6-hour interfraction interval) to 54 Gy. (The hyperfractionated approach has the theoretical advantage of minimizing the risk of radiation injury to the normal spinal cord and has been used safely when incidentally irradiating the cervical cord during treatment for brainstem gliomas.³⁴) Because the risk of long-term local recurrence for subtotally resected intracranial meningiomas is less than 10% when the minimum tumor dose is 52 Gy or greater,³⁵ a reasonable approach is to use a plan in which the maximum dose in the treatment volume is 54 Gy, with a minimum dose to the meningioma of 52 Gy.

Experience with soft tissue and bone sarcomas from extremity sites suggests that doses in the range of 60 to 66 Gy are required to eradicate microscopic disease. Still higher doses are required to control macroscopic disease. These doses clearly exceed cord tolerance, so conventional XRT cannot be used safely in this situation. The physical dose-distribution advantages of charged particle beams can be exploited here. These beams have extremely sharp lateral penumbras as well as Bragg peak behavior that allow accurate placement of the beam within a few millimeters of the spinal cord, without actually including the cord in the high-dose volume. With mixtures of photons, protons (and in the past, helium and neon ions), radiation doses in the range of 65 to 72 GyEq can be delivered to the tumor, while the dose to the cord is limited to less than 45 GyEq. Primary vertebral chordomas, chondrosarcomas, and osteogenic sarcomas are treated similarly. Patients with these types of tumors should be referred to the few national centers (e.g., Harvard, Loma Linda) with particle beam capabilities.

Brachytherapy

There is no established role for brachytherapy in the management of primary cord gliomas. Temporary 192Ir and permanent 125I implants have been used rarely to manage paraspinal sarcomas, occasionally with concurrent placement of a gold foil shield separating the seeds from the cord. Outcome has been poor in situations in which resection of the tumor has required opening of the dura.³⁶ Currently, brachytherapy plays little role in the management of most primary tumors involving the spinal cord.

Radiosurgery

Over the last several years, there have been anecdotal reports of successful outcome using radiosurgical management of spine neoplasms, particularly with the Cyberknife (Accuray, Inc., Sunnyvale, CA). However, to date there is no long-term validated role for radiosurgery in the management of primary cord gliomas.

Critical Normal Tissues (Radiation Myelitis)

The spinal cord and cauda equina are the critical dose-limiting structures in the treatment of primary tumors involving the spinal cord. The cauda equina, consisting of peripheral nerve, is probably more resistant to radiation injury than the cord. Myelopathy was observed to be much more frequent when intrathecal 198Au was used alone or in addition to external beam treatment^20,37; this form of therapy is now obsolete.

It has been postulated that the risk of cord injury depends on the location and length of cord irradiated, but no contemporary data exist to support this widely held notion. Older reports in the literature suggested that the thoracic cord was more sensitive to radiation injury than the cervical cord.³⁸ During the period when this information was gathered, it was common to treat with only one field per day. Since most fields involving the cervical cord were opposed laterals, the dose delivered daily to the cervical cord was the prescribed mid-plane dose. The thoracic cord, on the other hand, was usually irradiated with AP-PA opposed fields, and the radiation dose it received on “PA days” was substantially higher than the mid-plane dose. Since fraction size dominates as a risk factor for radiation cord injury, the radiobiologic dose received by the cord under the “one-field-a-day” approach was much greater than the prescribed mid-plane dose. Thus, the reputed anatomic variability in cord sensitivity appears to be a radiobiologic artifact created by an obsolete treatment technique.³⁹ Unlike the brain, where a relationship apparently exists between the volume irradiated and the subsequent risk of radiation injury, the length of cord irradiated appears to have little impact on the risk of radiation injury.^40,41

The radiation tolerance of the normal spinal cord using once-daily fractions of 1.8 to 2 Gy has been traditionally listed as 45⁴² to 50 Gy.^38,43 The 45-Gy value reportedly entails a 5% risk of myelitis at 5 years (TD5/5), which seems a gross overestimate. A University of Florida review of head and neck cancer patients whose cervical cords were incidentally irradiated found a 0.4% incidence of radiation myelitis with total doses between 45.01 and 50 Gy (2 of 471), compared with a 0% incidence with 40.01 to 45 Gy (0 of 514 patients), and 0% incidence (0 of 75 patients) with doses of more than 50 Gy.⁴⁴ A 6% incidence of cervical myelitis was reported in 72 head and neck cancer patients whose cords were treated with at least 55 Gy.⁴⁵ In that series, the minimum follow-up was 2 years, radiation fraction sizes ranged from 1.5 to 2 Gy, and 26 patients received total doses in excess of 60 Gy. Based partly on these data, contemporary experts have suggested that the TD5/5 for human spinal cord is actually on the order of 57 to 61 Gy, and the TD5/50 is in the range of 68 to 73 Gy.^40,41,46,47 The spinal cord is particularly vulnerable to injury when larger fraction sizes are used; the Medical Research Council Lung Cancer Working Party reported on radiation myelopathy developing in 1048 lung cancer patients undergoing thoracic irradiation with palliative intent.⁴⁸ No radiation myelitis developed using 3 Gy in 10 fractions (30 Gy total); but myelitis developed in 2.5% of patients receiving 3 Gy in 13 fractions (39 Gy total).

Although it seems logical to assume that cord injured by growing tumor and surgical trauma would be more sensitive to radiation injury than incidentally irradiated normal spinal cord, many series suggest that cord involved by tumor may be safely irradiated to standard tolerance levels. In the Massachusetts General Hospital (MGH) series, 26 patients with various histologic conditions were postoperatively irradiated to the equivalent of 40 to 50 Gy with standard fractionation and none developed radiation myelopathy.⁴⁹ Likewise, none of 29 Medical College of Virginia spinal cord ependymoma and astrocytoma patients treated with doses between 45 and 55 Gy developed radiation-related neurologic deficits.⁵⁰ In the series from the University of California at San Francisco (UCSF), 39 patients with primary cord and cauda tumors received total doses ranging between 45 and 54.7 Gy and only one case of radiation myelitis developed.¹³ Higher total radiation doses have been well tolerated; four ependymoma patients were treated at the Mayo Clinic to doses of 55 Gy and greater without complications.¹²

Outcome

Treating primary spinal tumors with postoperative irradiation is generally successful in terms of survival, less so in terms of local control, and difficult to evaluate in terms of long-term neurologic outcome. Survival of patients with low-grade spinal gliomas is substantially better than that achieved with histologically similar intracranial gliomas, probably because salvage of locally recurrent tumors with additional surgery is often effective. The exact degree of neurologic impairment caused by irradiation is difficult to determine because injury due to direct tumor growth as well as surgical trauma must be taken into account. As a general rule, it seems that a majority of patients either improve neurologically or stabilize with irradiation.⁵¹

Spinal ependymoma patients do well, with 5- and 10-year survival rates in the 70% to 100% range (Table 26-3). Unfortunately, local recurrence ultimately develops in roughly one third of irradiated patients. Delayed local failures are far from rare; most series include several late recurrences developing more than 5 or 10 years after initial treatment, indicating the need for long-term follow-up. The majority of recurrences develop within the irradiated volume.

A Mallinckrodt series of 37 patients with primary cord tumors found a statistically significant dose–response relationship: Survival was 23% for those receiving less than 40 Gy total dose, compared with 83% for those with total doses of 40 Gy or greater.³⁷ This analysis combined ependymomas, astrocytomas, unbiopsied tumors, and lymphoma together.

When ependymomas have been analyzed exclusively, no series has found a statistically significant dose-response relationship in the 40 to 54 Gy range.* In the Mayo Clinic series, four patients were irradiated to total doses of 55 Gy or greater, with one in-field failure still noted at this dose level.¹² The UCSF series included nine ependymoma patients treated with doses between 45 and 50.4 Gy and another nine with doses between 50.4 and 52.9 Gy. The local recurrence rate was 33% in both dose ranges.¹³ Presently, there are no convincing data that demonstrate improved outcome with total doses above 50.4 Gy.

Results for low-grade spinal astrocytomas are similar to those achieved with ependymomas, with long-term survival in the 60% to 90% range but a relatively high incidence of local recurrence (Table 26-4). Patients with pilocytic astrocytomas probably derive little survival benefit from postoperative irradiaton.¹⁰⁰ Most failures develop within 2 to 3 years of treatment. Low-grade astrocytomas tend to fail locally, although craniospinal axis metastases are occasionally encountered.^11,13 Like ependymomas, there is no evidence suggesting improved local control or survival with total doses exceeding 50.4 Gy.^11,13,92

Postoperative irradiation for malignant astrocytomas provides palliation, but long-term results are dismal. Extremely rapid local recurrence, CSF dissemination, and short survival times are the rule, regardless of the total dose of radiation used (Table 26-5).

Published results regarding irradiation of subtotally resected spinal meningiomas are virtually nonexistent. However, extrapolation from the experience with intracranial meningiomas potentially supports the use of postoperative irradiation in this setting.

Spinal canal sarcomas respond favorably to charged particle irradiation. Castro and coworkers^53,54 at Lawrence Berkeley Laboratory (LBL) treated 14 patients with neurofibrosarcoma and other soft tissue sarcoma histologies whose tumors abutted or encircled the spinal cord. All had macroscopic residual tumor when irradiated. With a mix of photons and helium and neon ions, a median dose of 65 GyEq was delivered to the tumor, but the cord dose was generally limited to less than 45 GyEq. The 4-year actuarial local control rate was 56%, and the 4-year actuarial survival rate was 77%. There was one case of radiation myelitis in a patient whose cord dose reached 55 GyEq.

The experience with primary tumors arising from the vertebral bodies is similar. Twenty-four LBL patients were irradiated with particles for spinal chordoma and chondrosarcoma.⁵⁴ Mean doses were 72 GyEq (chondrosarcoma) and 65 GyEq (chordoma); gross tumor was present at the time of irradiation in 92% of cases. The 3-year actuarial local control rate for chondrosarcoma was 83%; the 3-year survival rate was 69%. Corresponding values for chordoma were 33% local control and 48% survival. There were no neurologic complications. Fourteen patients with sacral chordomas did particularly well, with a 5-year actuarial survival rate of 85% and a 55% local control rate.⁵⁵ Similar results have been reported with the use of protons at MGH.⁵⁶ The 5-year survival rate for proton treatment of cervical chondrosarcomas and chordomas using doses of 56.8 to 80 GyEq was 58%; as with the LBL series, better outcome was seen with chondrosarcomas than with chordomas.⁵⁷ Investigators in Hannover, Germany also reported a benefit from high-dose (60 to 70 Gy) postoperative radiotherapy after subtotal resection of spinal chordoma, although results were best when en bloc resection could be performed.⁵⁸

Experience with vertebral osteogenic sarcomas is limited. Five such patients were treated at LBL; local control was achieved in two without treatment-related complications.⁵⁴ Fifteen osteogenic sarcoma patients treated at MGH with mixed protons and photons achieved a 5-year local control rate of 59%, and a 5-year overall survival rate of 44%.⁵⁹

Treatment of Recurrent Spinal Cord Gliomas

Only minimal information is available regarding management of recurrent, previously irradiated cord and cauda tumors, and what does exist is largely anecdotal. Many patients suffer devastating neurologic impairment from recurrent disease and do not benefit from heroic attempts at salvage. More localized recurrences in patients with reasonable neurologic and performance status should be managed aggressively, since prolonged survival is often achieved.

Surgery is the mainstay of salvage treatment. Low-grade tumors can be treated with reexcision including cordectomy, if necessary. These tumors appear to be particularly appropriate candidates for radical resection. In the experience at New York University, 29 patients (26 of whom had recurrent intramedullary tumors, with 18 previously irradiated) underwent attempted radical resection.¹⁷ Complete resection was obtained in 14, with “99% removal” in another seven. Although all three patients with malignant astrocytomas died from neuraxis dissemination, only one local recurrence was reported for low-grade tumors. Follow-up was short (range: 6 to 36 months), so long-term local control and survival rates are unknown. Lower extremity neurologic function appeared satisfactory, with stabilization or improvement in 72% of patients. However, patients unable to stand or walk preoperatively did not improve. Long-term disease-free survival following recurrence has also been reported following total removal or cordectomy in both the Mayo Clinic and UCSF series.^12,13 Salvage cordectomy has been reported to result in protracted (6-year) survival following local failure in one patient with a spinal cord GBM, although ultimately the patient died from CSF-borne intracranial metastasis.⁹⁸

Reirradiation has been attempted sporadically. Both experimental and clinical data suggest that the spinal cord and lung, unlike the heart and kidney, partially recover from subclinical radiation injury, suggesting that reirradiation of the cord might therefore be safely attempted.⁶⁰ In the UCSF series, 16 patients with spinal cord tumors suffered local failure. Four underwent repeat subtotal resection and reirradiation. No patients with multifocal or high-grade disease benefitted from retreatment. Two others with low-grade recurrences survived several years with high Karnofsky Performance Status values (≥80) following reirradiation. Both had initially undergone subtotal resections (one localized low-grade astrocytoma, one localized low-grade ependymoma) followed by 50.4 Gy to focal fields. After developing in-field recurrence and undergoing repeat subtotal resection, both patients were reirradiated with focal fields with 1-Gy fractions given twice daily to doses of 30 and 27 Gy. The cumulative doses were 77 Gy to the cauda equina for the ependymoma and 80 Gy to the cervical cord for the astrocytoma. No long-term radiation injuries developed.

Natural History

Epidural spinal metastases develop in 5% of patients with systemic cancer at some time during their illness. The incidence of spinal cord compression has decreased over the past 15 years, likely due to earlier detection and treatment of vertebral metastases as MRI has become more widely available. In 1992, it was estimated that there were 20,000 cases of cord and cauda compression annually in the United States,⁶¹ that number decreased to 12,700 by 2007.⁸⁸ Primary carcinomas originating from the breast, lung, and prostate account for about half of all cases of cord compression. Other histologic types (including lymphoma, renal cell carcinoma, myeloma, melanoma, gastrointestinal carcinomas, gynecologic carcinomas, and sarcomas) account for the remainder.

Cord compression is most commonly caused by local progression of a vertebral body or neural arch osseous metastasis, which ultimately grows into the spinal canal. Less commonly, paraspinal metastases extend through vertebral foramina to involve the cord. The majority of lesions (85%) are located anterior to the cord, and often multiple epidural metastases exist simultaneously.

Ninety percent of patients have been diagnosed with cancer before they develop cord compression. Depending on histology, cord compression can develop soon after diagnosis (e.g., lung cancer: average 6 months) or with a prolonged interval between diagnosis and compression (e.g., breast cancer: average 4 years). Cord compression has been reported to develop as the first site of recurrence 27 years after initial diagnosis and treatment of breast cancer.

Symptoms

Pain is the most common initial symptom (present in 90% to 95% of patients) and can be either local, referred, or radicular in nature. It usually precedes all other symptoms by several weeks to months. The site of cord compression often is located away from the perceived site of pain; cervical cord compression can present with mid-back pain, thoracic cord compression can manifest as lumbosacral pain, and vice versa. Weakness is rarely the first symptom (2%) but is common at diagnosis (75%). Sensory loss is present in 50% of patients at diagnosis, but is rarely the initial symptom. Autonomic dysfunction is usually a late manifestation associated with an unfavorable prognosis, but is present in upwards of 50% of patients at the time of diagnosis. Cauda equina compression is notorious for presenting with a quartet of symptoms including pain, lower extremity weakness, sensory disturbances involving the buttocks, posterior thighs, and perineum, and autonomic dysfunction including loss of anal sphincter tone and urinary retention with overflow incontinence.

Work-Up

History, physical, and a high index of suspicion are essential for making the diagnosis of cord compression. All patients with back pain and a history of cancer should be evaluated for possible spinal cord compression. One prospective series of cancer patients suggested that the likelihood of cord compression was approximately 30% when any one of the following risk factors was present: back pain, abnormal neurologic examination, or evidence of vertebral metastases on plain x-rays.⁶² When two factors were present, the probability of cord compression was 60% to 70%; when all three features were present, the likelihood of cord compression was more than 90%.

Imaging

MRI with gadolinium is the radiographic study of choice to evaluate possible malignant spinal cord compression. Patients unable to undergo MRI should be evaluated by metrizamide-enhanced CT myelography. In addition to the clinical area of interest, the study should evaluate the entire spine to determine the complete extent of the lesion and to detect possible synchronous lesions located elsewhere along the spinal axis. One retrospective series reported that multiple levels of cord compression existed in 39% of patients with apparently clinically localized spinal cord compression, and that more than two thirds of patients with multiple levels of spinal cord compression had more than one spinal region involved.⁶³

Diagnostic Approach

The following approach is suggested to evaluate the cancer patient with new back pain: if the neurologic examination is abnormal, a screening sagittal MRI of the entire spine is performed. If the neurologic examination is normal, a screening sagittal MRI of the entire spine is still performed.

When the sagittal screening MRI detects epidural disease, a full MRI of the affected area is performed. This approach results in more rapid performance of the definitive study compared with the old approach of first performing plain films with or without bone scans, followed by MRI when radiographic or scintigraphic abnormalities were noted.⁶⁴ It also results in a less expensive MRI examination if no epidural disease is detected and has proven to be cost-effective compared with the traditional approach.⁶⁴ This approach might be criticized as excessively “high-tech,” and it probably detects far more cases of painful bony metastases than actual cord compression. Nonetheless, the patient’s long term functional status is clearly superior when a painful bony metastasis is irradiated early on, rather than later when neurologic deficits from cord compression have developed.

A widespread belief held among radiation oncologists is that referral for radiation treatment of spinal cord compression occurs most frequently on Fridays, generally late in the afternoon. The validity of this belief is supported by a retrospective study of 443 spinal cord compression patients from the Netherlands, which showed that 30% of such referrals came on Friday, 12% on Monday, 17% on Tuesday, 15% on Wednesday, 20% on Thursday, 5% on Saturday, and 1% on Sunday (P < .002).⁶⁵

Treatment

Management options vary among irradiation, neurosurgical decompression with or without postoperative irradiation, and chemotherapy. In the past, irradiation and corticosteroids were the mainstay of treatment,⁶¹ but improvements in neurosurgical technique during the past decade have now made initial surgical debulking and decompression followed by postoperative radiotherapy the current gold standard of care where available.⁸⁹

Surgery

The traditional approach, decompressive laminectomy, offers poor exposure for the majority of tumors that arise from anterior bony structures. A successful outcome (defined as ambulation after treatment) is achieved in about 30% of patients. Historically, complications have been high, with 10% surgical mortality, 10% significant morbidity, and 10% of patients experiencing further neurologic deterioration.⁷⁰ More recently, selective surgery with vertebral body resection and stabilization has been performed for anterior tumors.⁷¹ This approach appears to have substantially improved results with approximately 80% to 90% of patients ambulatory postoperation, although mortality and serious morbidity rates remain in the 5% to 15% range.

A randomized prospective study comparing the combination of modern selective surgery plus postoperative irradiation with palliative radiotherapy verified the benefit of combined treatment.⁸⁹ In that study, 101 patients with newly diagnosed spinal cord compression were given 100 mg dexamethasone immediately, followed by 24 mg q6hr until treated with either radiation or surgery. Patients were randomized to either radiotherapy (30 Gy in 10 fractions) within 24 hours of diagnosis or surgical debulking/decompression within the same time period. The surgical patients all received postoperative radiotherapy (30 Gy in 10 fractions) beginning within 14 days of surgery. The combined modality patients demonstrated superior post-treatment ambulation rates (84% vs 57%); this was true both for patients who were ambulatory at the time of treatment (94% vs 74%) and those unable to walk at study entry (62% of the combined modality group regained the ability to walk, compared to only 19% in the radiation arm). Additionally, the median duration of ambulatory status was significantly longer in the surgery plus x-ray therapy (XRT) group (153 days) than in the radiotherapy alone group (54 days). Median survival was also significantly better in the combined treatment group (126 days vs 100 days for XRT alone). Putting this palliative benefit of combined treatment into perspective, it appears that, on average, the combined treatment patients preserved their ambulatory status for the rest of their lifespans, while the average radiation-alone patient preserved the ability to walk for only 50% to 60% of his or her remaining lifespan before becoming bedridden.

Unfortunately, the 24/7 combined modality approach requires neurosurgical resources that often exceed the capabilities of most medical communities, and are generally limited to large academic medical centers. Given the lack of available resources in most community settings, many patients with spinal cord compression will still need to be treated with immediate glucocorticoids and radiotherapy rather than a combined surgical/radiation approach.

Nonetheless, there are several situations where irradiation is contraindicated: