Posterior Approaches to the Thoracic Spine

Thoracic Spine

The thoracic spine comprises 12 vertebrae—T1 through T12 (or D1‒D12 in international literature), each composed of a vertebral body and posterior elements. The posterior elements include the pedicles, facets, transverse processes, laminae, and a spinous process; together with the posterior vertebral body, these components form a protective bony ring enclosing the vertebral foramen. Additionally, the facets play a minor role in axial load-sharing. As at other spinal levels, the facets and vertebral body form a three-joint system. The adjacent vertebral bodies articulate with one another through the fibrous intervertebral disc, and the superior and inferior articular processes interface through synovial joints called the zygapophyseal or facet joints. Within the thoracic spine, the facet joints have a relatively vertical orientation that, along with the rib cage, limits flexion-extension while allowing for moderate amounts of lateral bending and axial rotation.

The rib cage represents a unique feature of the thoracic spine. Each thoracic vertebra articulates with the ribs bilaterally at the vertebral body (i.e., the costovertebral joint) and the transverse process (i.e., the costotransverse joint). Ribs are numbered based upon the vertebrae with which they articulate. For the first, tenth, eleventh, and twelfth ribs, this articulation occurs with a single vertebra, but for the second through ninth ribs, articulation is with both the same-level vertebra and the cephalad vertebra (e.g., the second rib with T1 and T2 vertebrae). Because of this articulation with demifacets on adjacent vertebrae, the second through ninth ribs cross the posterolateral disc space, which has obvious consequences for posterior and posterolateral approaches. Exiting nerve roots are numbered similarly to the ribs and exit below the pedicles at each level via the neural foramina.

The thoracic vertebrae increase in size from T1 to T12, ^, following progressive increases in compressive forces seen at more caudal vertebral segments. More caudal segments also have smaller transverse processes as the vertebrae begin to adopt a more lumbar-like morphology. In contrast, the spinous processes demonstrate no consistent trend with respect to size; however, they do adopt a more vertical orientation in the midthoracic spine, leading to a shingling effect that limits the feasibility of an interlaminar approach. Further limiting interlaminar access is the progressive increase in thoracic laminar height as one progresses in the cephalocaudal direction.

Moving in the cephalocaudal direction, the thoracic vertebral bodies show a gradual transition in morphology. At T1, a more rectangular, cervical-like body shape is adopted. Progressing caudally, there is an initial decrease in vertebral body width, causing the vertebral body to adopt a heart-shaped morphology in the midthoracic spine. However, from T4 to the thoracolumbar junction, the vertebral body progressively widens, giving the vertebral body a more kidney-shaped, or lumbar-like, morphology. The pedicles show a similar transition in anatomy, adopting a more anteroposterior axial orientation in the midthoracic spine and a more oblique orientation at the upper and lower segments. ^, Sagittal pedicle angulation, in contrast, transitions from a steeper angle in the upper thoracic spine (roughly 30 degrees at T1) to a flatter angle in the thoracolumbar region (10–15 degrees). With this change in angulation is a concomitant change in pedicle size. Pedicle sagittal diameter is noted to increase progressively as one moves caudally ; however, transverse pedicle diameter appears to reach a relative minimum in the midthoracic segments (T4), with larger diameters seen at both more cephalad and more caudal levels. Despite this increase in pedicle sagittal diameter, there does not appear to be a concomitant change in cortical bone thickness, which varies between 1 and 1.5 mm. ^, Clinically, this necessitates larger-diameter pedicle screws in the lower thoracic vertebrae to achieve cortical bone engagement. Interestingly, cortical bone thickness appears to be greater along the medial border. Although some have advocated for the adoption of more medial trajectories to take advantage of this thicker cortical bone and to avoid lateral wall deviation, we do not recommend this approach, given the increased risk for canal violation and spinal cord injury. This concern is greatest at the thoracic apex, which possesses both the smallest pedicles and the narrowest vertebral canal.

Both mathematical and in vitro/cadaveric studies have demonstrated that the rib cage plays a significant role in the biomechanics of the thoracic spine. Of note, the semirigid fixation of the rib cage to the vertebrae limits axial rotation by more than a factor of 2. ^, Significant reductions in flexion-extension and lateral bending are also seen, ^, although the degree of stabilization in these axes is less pronounced. Additionally, mathematical modeling has suggested that the rib cage helps reduce compressive forces on the vertebrae.

Extraspinal Thoracic Anatomy

The intrathoracic viscera (e.g., heart, lungs) uniquely complicate operative access to the thoracic spine. Given the ventral position of these organs, dorsal approaches to the thoracic spine can allow for less dissection and a more direct approach vector. Moving from the superficial layer inward, the posterior extraspinal tissues are the skin, subcutaneous fat, superficial muscles, intermediate muscles, and deep or “intrinsic” muscles. The superficial muscles include the trapezius (occiput-T12), rhomboids (C7‒T5), latissimus dorsi (T7-sacrum), and serratus posterior inferior (T11‒L2). The intermediate musculature exists only in the superior and inferior thoracic spine, and includes the serratus posterior superior (C7‒T3) and serratus posterior inferior (T9‒12), respectively. The deep muscles run along the length of the spinal column and include the splenius capitis (C7‒T3) and cervicis (T3‒T6) superiorly and the spinalis, longissimus, and iliocostalis along the entire column. The muscles themselves are innervated by the posterior rami and their branches; the exception to this are the rhomboids (innervated by the dorsal scapular nerve), the latissimus dorsi (innervated by the thoracodorsal nerve), and the trapezius (innervated by the spinal accessory nerve).

All of these muscle layers join medially at their connections to the spinous processes, which creates a relatively avascular plane that can be used in the posterior median approach. This allows for a relatively bloodless dissection that also by and large avoids denervation of the paravertebral muscles. Unfortunately, many thoracic spinal pathologies lie ventral to the cord and require a lateral approach. When such an approach is used, many of these muscle layers must be dissected—potentially resulting in devascularization and denervation of these muscles, causing fibrosis and atrophy. ^, These muscles are key to maintaining sagittal balance within the “cone of economy” ^, ; consequently, their atrophy is inevitably associated with progressive kyphosis.

Ventrally, the structures of interest are those of the mediastinum (e.g., heart, trachea, esophagus, great vessels) and the lungs. The aorta and vena cavae immediately abut the anterolateral faces of the vertebral bodies and are neighbored posterolaterally by the paravertebral sympathetic trunks. Other vessels of importance include the azygos and hemiazygos veins on the left side and the posterior intercostal arteries that exit posterolaterally from the aorta and run along the anterolateral aspect of the disc spaces. These posterior intercostal arteries give off intersegmental arteries that feed the spinal cord, chief of which is the artery of Adamkiewicz that typically arises at the level of T10. Although not part of the cardiovascular system, the thoracic duct is an additional vessel of interest, as it drains chyle from the intestines, and its injury can result in a chylothorax and subsequent electrolyte abnormalities and nutritional deficiencies. This structure runs from the anteriorly positioned cisterna chyli at the thoracolumbar junction to the retroesophageal space, where it continues until terminating in the left innominate vein at the cervicothoracic junction. Under normal circumstances, these ventral structures are not at risk of injury; however, when placing instrumentation, they must be kept in mind, as anterior cortex violation has been reported to cause chylothorax and aortic injury.

Also pertinent is knowledge of anatomy considered “nonvital.” Under ideal circumstances, no healthy tissues will be sacrificed; however, it is often necessary to gain adequate exposure for ventral pathologies. In general, sacrifice of the thoracic nerve roots or intersegmental vessels results in only minimal morbidity. The exceptions to this are the C8 and T1 roots, which contribute to intrinsic hand function. Sacrifice of these roots results in significant morbidity, and is unacceptable under most circumstances. When sacrificing nerve roots, we recommend ligating the nerve proximal to the dorsal root ganglion, as work in animal models has suggested that injury distal to the dorsal root ganglion is associated with greater mechanical allodynia.

Intercostal arteries can similarly be sacrificed in most circumstances. When performing this procedure, though, we recommend first temporarily occluding the candidate vessel using an aneurysm clip. Neuromonitoring with motor-evoked potentials and somatosensory-evoked potentials should then be observed for at least 5 minutes to ensure that occlusion of the vessel has not caused spinal cord hypoperfusion. ^, If the signals remains stable for 5 minutes or longer following occlusion, then the vessel should be safe to take. Alternatively, preoperative angiography may be performed to determine which intersegmental vessel gives rise to the dominant segmental medullary vessel feeding the spinal cord.

Selecting an Approach

The primary factors to consider when selecting a surgical approach are the surgical indications, the location of the pathology, the patient’s spine surgery history, and the goals of treatment. Additional considerations include surgeon experience and the anatomic constraints posed by both extraspinal organs and scarring created by prior thoracic surgery. These anatomic constraints may limit the number of approaches that can be safely entertained. In certain cases, multiple approaches may be suitable; however, few high-quality studies have directly compared approaches for thoracic spine indications.

From the perspective of treatment goals, the main categories are decompression, stabilization, or a combination thereof. When decompression is the goal, the point(s) of compression should be identified. In the minority of cases, compression is attributed to a dorsal lesion (e.g., hypertrophied ligamentum flavum, tumor, epidural abscess). In such cases, a midline posterior approach for direct decompression is optimal. However, the majority of thoracic spine pathologies are the result of ventral or ventrolateral pathologies. In these cases, a direct midline posterior approach cannot achieve effective decompression because the thoracic kyphosis and tethering of the cord by the spinal roots prevent the cord from floating away from the site of compression. Similarly, the degree to which the cord can be retracted to reach the pathology is generally inadequate. Consequently, to achieve effective decompression, a posterolateral (e.g., transpedicular) or far-lateral approach should be adopted. Progressively more lateral approaches offer better access to the vertebral body, disc, and ventral epidural space; however, they come at the cost of more extensive dissection.

The decision to place instrumentation is dictated by the preoperative spinal instability and the anticipated iatrogenic instability (necessary to achieve the primary goals of surgery, such as lesion resection and/or neural element decompression). Evaluation of baseline instability has been well defined in the trauma and oncology literature. Within trauma, the Thoracolumbar Injury Classification System (TLICS) is currently the scoring method of choice and evaluates lesions based upon injury morphology, integrity of the posterior ligamentous complex, and the patient’s neurological status. In this algorithm, injury morphology determines immediate mechanical stability; meanwhile, the integrity of the posterior ligamentous complex determines the long-term stability. The AOSpine Thoracolumbar Spine Injury Classification System is an updated version of TLICS and includes a prescribed surgical algorithm based upon the scored instability. A similar system exists for the evaluation of stability in the setting of spinal metastases, which most commonly localize to the vertebral bodies of the thoracic spine. The Spinal Instability Neoplastic Score (SINS) assigns points based upon spinal level, extent of vertebral body involvement, presence of posterior element involvement, presence of pain, presence of deformity, and presence of lytic or blastic characteristics on radiographic studies. ^, Low scores (i.e., 0–6) are thought to be stable and do not require instrumentation, whereas high scores (i.e., 13–18) identify lesions that would likely benefit from surgical fixation. Additionally, more recent studies have suggested that higher-scoring lesions (i.e., 10–12) that fall into the SINS “potentially unstable” category may also benefit from stabilization.

In contrast to preoperative instability, no good quantitative scoring system exists for evaluating iatrogenic instability. However, for decompression alone it has been shown that decompression of more than two levels is associated with postlaminectomy kyphosis, especially where the decompression bridges a junctional segment (i.e., C7‒T1 or T10‒L1). ^, Consequently, patients undergoing decompression of more than two levels, especially at junctional segments, should receive instrumentation to avoid delayed sagittal deformity. Other procedures that may warrant instrumentation to avoid iatrogenic instability include corpectomy resection of more than 50% of the vertebral body and unilateral or bilateral facetectomy removing more than 75% of the joint. ^,

If instrumentation is to be placed, it is imperative that the surgeon evaluate the patient’s underlying bone quality. Conventionally, this has been done with dual-energy x-ray absorptiometry (DEXA), a radiograph-based study comparing measured bone density to a population of healthy controls. However, some have suggested that DEXA may be an ineffective predictor of bone quality; therefore, some groups have pushed for the use of studies based on computed tomography or magnetic resonance imaging. Where preoperative bone quality is poor, the decision to instrument must be carefully considered. If surgery is necessary and cannot be performed without instrumentation, cement augmentation can be used to improve screw pullout strength and construct rigidity.

Other considerations in terms of surgical approach include the patient’s surgical history and the feasibility of minimally invasive surgery (MIS). MIS approaches—essentially, those that minimize dissection of soft tissue—have been associated with shorter operative times, lower complication rates, and shorter lengths of stay. Because of these attractive benefits, many patients prefer MIS to traditional open surgery. Theoretically, the less-extensive muscle dissection associated with MIS may also reduce postoperative pain and help lower the risk of adjacent segment disease by preserving the posterior tension band. However, the benefits of MIS are limited in patients undergoing revision surgery, and some authors suggest MIS may actually have higher complication rates in this population. Patients who have previously undergone posterior procedures may have substantial scarring that obscures anatomic landmarks; these patients are also at increased risk for wound infection and dehiscence. Additionally, epidural scarring increases the risk of incidental durotomy, which may challenge postoperative wound healing and may be associated with an increased risk of neural element injury. A history of prior nonspinal thoracic surgeries may also impact surgical planning. Specifically, prior intrapleural procedures are a relative contraindication to the use of a transthoracic approach, as the associated intrapleural scarring makes the surgical approach extremely risky. Surgeons may therefore be forced to adopt a posterolateral approach for ventral pathologies that they might otherwise treat with a thoracoscopic, transpleural, or transthoracic approach.

The last major approach consideration is specific to patients undergoing pedicle screw instrumentation. Screw placement in the thoracic spine, especially at the thoracic apex, is notoriously more difficult than lumbar instrumentation because both the vertebral canal and pedicle diameters are smaller in the thoracic spine. One option to address this issue is to use image guidance or robotic assistance, both of which have been demonstrated to increase instrumentation accuracy. These techniques are discussed in detail in Chapter 81 . Below, we delineate the conventional posterior approaches to the thoracic spine—namely, the midline approach, transpedicular approach, costotransversectomy, and lateral extracavitary approach ( Fig. 116.1 ). Additional aspects of anterolateral approaches and minimally invasive approaches are discussed in detail in Chapter 117, Chapter 118, Chapter 119 .