Anatomical considerations

Thoracic zygapophyseal joints are true synovial joints. The joints and capsules have nerve plexus formations and some free nerve endings capable of nociceptive transmission. The adventitia of the blood vessels supplying the Z-joints and the cancellous bone of the vertebral bodies contain unmyelinated nerve fiber plexuses that give rise to pain when irritated by various mechanisms. Mechanisms include abnormal postural stress caused by kyphosis and scoliosis, edema from direct trauma or inflammation, direct compression of innervated connective tissues by bone fragments, neoplasms, and acute natural inflammatory response or iatrogenic inflammatory responses secondary to injected hypertonic saline solutions.

Nerve fibers converge into medial branch of the primary posterior ramus and relay in dorsal root ganglion prior to entering the spinal cord. Because intersegmental connections in the thoracic spine are not as pronounced as in its cervical and lumbar counterparts, the pain arising in any particular segment of the thoracic spine is more precisely localized than comparable segmental lesions in upper or lower regions of the vertebral column.¹⁴

Encapsulated mechanoreceptors and nociceptors in joint capsules of thoracic and lumbar spine have been demonstrated histologically¹⁵ and immunohistochemically.¹⁶ These mechanoreceptors respond to different states of excursion, provide proprioceptive sense, modulate protective muscular reflexes, and via nociception signal potential tissue damage in the event excessive force is applied. Consistent with neck mobility, need to position the head accurately in space, and the need for coordinated muscle control for protection and posture the cervical spine has more mechanoreceptors than the thoracic spine.

In the experimental study of healthy subjects pain was provoked in 72.5% of 40 tested thoracic T3–4 to T10–11 zygapophyseal joints when injected contrast medium distended the joint capsule.¹⁷ Referral pain patterns, although overlapping between segmental levels, were more localized compared to patterns obtained by stimulation of lumbar and cervical joints. The most intense pain was reported one segment below and lateral to the injected joint (Fig. 73.1). Precise borders could not be delineated. The largest inferior referral was 2.5 segments while lateral referral did not cross the posterior axillary line. Two subjects reported interesting referral patterns. In one case the T3–4 injection produced pain in the back and subject also stated that ‘pain went into my lung behind my sternum.’ In another case the T3–4 injection produced pain in the back and the subject also reported that pain ‘like a quarter-sized cylinder went toward my breast bone.’

Fig. 73.1 Referral pain patterns from stimulation of zygapophyseal joints T3–4 to T10–11.

(Adapted from Dreyfuss et al. Thoracic zygapophyseal joint patterns: a study of normal volunteers. Spine 1994; 19(7):807–811)

Referral pain patterns from stimulation of zygapophyseal joints C7–T1 through T2–3 and T11–12 were described by Fukui et al. in 1997 (Fig. 73.2).¹⁸ A total of 21 joints were injected in 15 patients with previously documented zygapophyseal joint pain. At C7–T1 all patients described pain in paravertebral region extending towards the superior angle of scapula, into interscapular region, and to the inferior angle of scapula. Lateral extension toward the shoulder and suprascapular region was described by two patients. T1–2 stimulation referred into interscapular region and to the inferior angle of the scapula. In two subjects referral was reported into the superior angle of the scapula and suprascapular region. Stimulation at T2–3 joint provoked pain laterally toward the interscapular area and caudally toward the inferior angle of the scapula. T11–12 joint injection referred pain into localized area around the injection and one patient described pain over iliac crest. The authors conclude that referral maps from stimulation of joints at levels of C7–T1 to T2–3 provided such a large overlap that their clinical usefulness in tracking the origin of pain is limited. Anatomical dissection confirming the C7 and C8 medial branches traveling to T2 and T3 level¹⁷ may explain this observation.

Fig. 73.2 Referral pain patterns from stimulation of zygapophyseal joints C7–T1 through T2–3 and T11–12.

(Adapted from Fukui et al. Regional Anesthesia 1997; 22(4):332–336)

Although lumbar facet denervation was introduced in 1974,¹⁹ the procedure was not properly performed until Bogduk²⁰ described the anatomical course of the lumbar medial branches in 1979.

Percutaneous facet denervation was reported in thoracic Z-joints,³ but the study offered no data on the detailed surgical anatomy of the thoracic medial branches. The targeted points for the thoracic medial branches were at a location that was analogous to the position of the lumbar medial branches proposed previously²⁰ at the junction between the superior articular process and the transverse process. In 1994, Stolker et al. published the data from the anatomical study of two thoracic cadaveric spines, where the cannula was placed under fluoroscopic guidance into a target point at the junction of the base of the superior articular process and the transverse process.²¹ The specimens were then frozen and sectioned with the heavy-duty cryomicrotome. It was observed that although the cannulas were placed reproducibly onto the osseous targets, the medial branch ‘stem’ was never found within the reach of the electrodes. The authors concluded that if the medial branch ‘stem’ is supposed to be the target, a more anterior, more cranial, and more lateral position would perhaps be more effective.

In contrast to the course of the lumbar medial branches, which are at the junction of superior articular process and the transverse process, the medial branch of the thoracic dorsal ramus has a different anatomical location (Figs 73.3, 73.4). Cadaveric dissection of 84 medial branches by Chua and Bogduk revealed the thoracic medial branch arose from the dorsal ramus typically 5 mm lateral to the intervertebral foramen, traversed laterally, dorsally, and inferiorly, and remained posterior to the superior costotransverse ligament.²² After leaving the intertransverse space the medial branch crossed the superolateral corner of the transverse process below (e.g. T3 medial branch and T4 transverse process) and then passed medially and inferiorly across the posterior surface of the transverse process. In its course over the dorsal aspect of the transverse process the nerve was sandwiched between the multifidus muscle anteriorly and the semispinalis thoracis posteriorly. This was the typical course at the levels of T1–4 and T9–10. The T11 medial branch crossed the lateral aspect of base of the superior articular process of T12 vertebra, the transverse process of which is much shorter that other transverse processes. The T12 medial branch had an analogous course to that of the lumbar medial branches at the junction of superior articular and the transverse process of L1. At midthoracic levels (T5–8) the medial branch did not always assume contact with the transverse process and often demonstrated cephalad displacement. The nerve, after making a turn medially in the intertransverse space, descended only slightly inferiorly and remained separated from the transverse process by the fascicles of multifidus muscle.²²

Fig. 73.3 Anatomical location of thoracic medial branches.

(Adapted from Chua and Bogduk. Acta Neurochir 1995; 136:140–144)

Fig. 73.4 Nerve supply of thoracic spine structures.

From Bogduk N. Innervation and pain patterns of the thoracic spine. In: Grant R, ed. Physical therapy of the cervical and thoracic spine, 3rd edn. New York: Churchill Livingstone; 2002:73–81)

Two articular branches were noted to arise from the medial branches. A short ascending branch separated from the medial branch inferior to the Z-joint and innervated the inferior capsule of the joint. A descending articular branch arose from the medial branch at the superolateral border of the transverse process and in its sinuous course through the multifidus muscle reached and innervated the superior capsule of the joint below.²²

The medial branches in upper thoracic segments are musculocutaneous, while lower have only muscular distribution.⁷

Based on these data, the appropriate target for thoracic medial branch block at T1–3 and T9–10 levels is the superolateral corner of the transverse process where the nerve lies against the osseous structure. For radiofrequency denervation the probe, upon contacting the superolateral corner, should be passed over the edge of the transverse process to be in the contact with the medial branch. Targeting of midthoracic medial branches, due to their different positions, would be more challenging and less reliable. T11 and T12 medial branches are blocked in the same fashion as their lumbar counterparts (junction of the transverse process and the superior articular process). Interestingly, multiple interventional textbooks that have recently been published still describe the techniques of thoracic medial branch blocks and denervation not consistent with current knowledge of anatomy.

Thoracic zygapophyseal joint orientation

The articular surface of the joint is inclined anteriorly from the frontal plane at approximately 60 degrees to the horizontal plane, making the inferior portion of the joint more posterior and the cephalad pole more anterior. The joint plane is also rotated along the vertical axis about 20 degrees so that the lateral aspect of the joint is more anterior and the medial aspect of the joint is more posterior. This orientation changes in the low thoracic spine, which is the more vulnerable segment of the thoracic spine. While maintaining the frontal orientation at T10–11, the T11–12 level shows considerable variation with transition to sagittal orientation similar to orientation of zygapophyseal joints in lumbar spine.²³ In a cadaveric study of 37 male spines by Malmivaara et al.,²³ the authors also observed that at T10–11 level pathological changes were mostly disc degeneration, while at the level of T12–L1 level facet and costovertebral joint degeneration were dominant. The level of T11–12 demonstrated both disc and Z-joint and costovertebral joint degeneration.

Diagnostic injections

Diagnostic blocks used to temporarily denervate the Z-joints are performed using fluoroscopic guidance with local anesthetic injected into the intra-articular space, onto the medial branch, or both. Injecting 0.1–0.2 mL of nonionic water-soluble contrast agent such as iohexol or iopamidol will confirm the proper placement of the needle tip within the joint. False-positive or false-negative findings may result from extracapsular spread into epidural space during intra-articular injection or intravascular uptake during medial branch block. Extravasation of the local anesthetic beyond the intended target will decrease the specificity of the test by blocking the nociceptive transmission via the sinuvertebral nerve, nerve roots, or spinal nerve. Limiting the volume of the anesthetic agent will increase the specificity. Use of volumes larger than 1 mL or even 0.5 mL may be ‘empirically therapeutic,’ but will likely not have a diagnostic value.²⁴

The sensitivity of the diagnostic test is a function of the patient’s pain threshold, pain tolerance, pain level prior to the administered test, experience of the injectionist, ease of the procedure, size of the needle and concurrent soft tissue trauma, and psychological profile.

Wilson reported the outcome following intra-articular injection of bupivacaine and triamcinolone in 17 patients,²⁵ of which two had thoracic kyphoscoliosis, seven had a history of spine injury, and eight had no apparent precipitating factors. After reproduction of their usual pain, 13 patients reported immediate benefit for the expected duration from the local anesthetic. Nine patients reported significant improvement for more than 1 month.

The technique of the intra-articular thoracic zygapophyseal joint injection has been described by Dreyfuss.¹⁷ With the monitored patient in a prone position on a fluoroscopy table the appropriate area is routinely prepped with Betadine and draped in a sterile fashion. A true posteroanterior (PA) projection of thoracic spine is obtained. If the T7–8 zygapophyseal joint, for example, is to be injected, the inferior aspect of this joint will be at the superior aspect of the T8 pedicle. A skin mark is then made at the mid to inferior portion of another pedicle below, i.e. T9. A 25-gauge, 3.5 inch needle is inserted in a 60° angle cephalad toward the target joint. Under intermittent fluoroscopic imaging in PA projection, the needle is advanced towards the superior articular process of T8. The needle should remain on the imaginary line connecting the centers of the T8 and T9 pedicles in order to avoid puncturing the pleura laterally or entering the epidural–subarachnoid space medially. Aiming the needle at the inferomedial aspect of the joint may facilitate the joint entry in difficult cases, as this is the most posterior aspect of the joint. As the tip is seen over the mid to superior aspect of the T8 pedicle, the image intensifier is rotated away from the side injected into almost full lateral view until the outline of the joint is clearly visible. The needle is then advanced into the inferior portion of the joint. If the needle is not seen adjacent to osseous structures, manipulation of the needle is performed in the anteroposterior (AP) projection to assure the needle has not strayed lateral or medial. Once the needle is in position, 0.1–0.15 mL of nonionic contrast is administered to confirm the proper placement of the needle. Contrast agent will be observed in superior and inferior recesses of the joint. Local anesthetic (0.5% bupivacaine or 2% lidocaine) with corticosteroid may be injected, but remember that the joint volume is only about 0.5 mL. When firm capsular endpoint is achieved, injection should be stopped to prevent joint capsule rupture. During capsular distension the patient is asked whether concordant pain is reproduced. After the expected onset of anesthesia the patient is asked about the pain relief and this is then further observed and recorded in a pain diary for up to 8–24 hours.

Dreyfuss reports repeating the joint injection 4 weeks later, up to three times a year if long-term relief is observed. If one can confirm that significant short-duration pain relief consistently occurs after local anesthetic denervation, radiofrequency denervation of the medial branches is the best treatment,¹⁷ if one assumes that there is an efficacy similar to that of cervical and lumbar medial branch neurotomies. However, the lack of anatomically based prospective randomized studies evaluating the efficacy of the thoracic medial branch neurolysis precludes the general acceptance of this treatment modality.

In summary, the data exist about thoracic facet joint anatomy, referral pain pattern maps in normal volunteers, technique of the procedure, but the studies about therapeutic utility and validity of the procedure are still to come, despite a widespread use of those procedures.

Anatomical considerations

Thoracic discs and vertebral bodies are innervated by two interconnected nerve plexuses. The ventral nerve plexus is associated with the anterior longitudinal ligament and has a bilateral supply from branches of the sympathetic trunk, rami communicantes, and perivascular nerve plexuses of segmental arteries. It is connected to the nerve plexuses of costovertebral joints. The dorsal nerve plexus is made up of the nerve plexus associated with the posterior longitudinal ligament. This nerve plexus is more irregular and receives contributions from the sinuvertebral nerves. Sinuvertebral nerves (recurrent nerves of Luschka) arise as branches from the superior or anterior aspect of the spinal nerves just distal to the dorsal root ganglion after exiting from the intervertebral foramen.⁸ After 2–3 mm of its course back towards the intervertebral foramen, this somatic root joins with the autonomic root that arises from the gray ramus communicans or sympathetic ganglion at the same segmental level. The sinuvertebral nerve passes through the intervertebral foramen anterior to the spinal nerve and nerve roots. Its branches innervate vertebral lamina, and periosteum of the costal neck. As a part of dorsal plexus it further innervates posterior longitudinal ligament, dura mater, and periosteum of vertebral bodies. Anatomical study of the intervertebral lumbar disc innervation²⁶ also demonstrates distinctive innervation of posterior disc (sinuvertebral nerve) and lateral/anterior disc wall (branches of primary posterior rami and rami communicantes).

The assumption that thoracic pain can be generated by thoracic discs is an extrapolation of our knowledge about discogenic lumbar pain. The scientific data are, however, limited and most studies largely describe pain patterns caused by thoracic disc herniations.

In a retrospective study of 100 symptomatic patients with magnetic resonance imaging (MRI) documented disc derangement who underwent provocative discography, the authors found discs with annular tears, intrinsic degeneration, and/or associated vertebral endplate infractions to be painful 75% of the time.²⁷ Clinical concordance was about 50%. In this series, the authors observed that in addition to thoracic back pain, extraspinal pain, such as chest wall, intrathoracic and upper abdominal pain, was frequently provoked with thoracic disc injections. They note that location of pain upon provocation appears to relate to the anatomic location of annular tears and may refer to anterior extraspinal sites, such as ribs, chest wall, sternum, and visceral structures within the thorax or upper abdomen. Lateral annular defects often produced radicular-type pain either to visceral or musculoskeletal sites. Posterior annular defects produced back pain either locally or diffusely. The location of provoked sensation was not predictable. The authors stated that the disc with partial annular tears (nonprotruding disc derangements) may be painful and clinically significant, as more than 50% of painful discs fell into this category. The authors did not use antibiotics, and report no incidence of discitis in this series using a thin-needle technique.

In this controlled, prospective thoracic discography study, the authors concluded that thoracic discography may demonstrate disc pathology not seen by MRI in asymptomatic individuals and that in truly asymptomatic individual the procedure is not painful. Fifty-five percent of the discs studied in the symptomatic group were concordant, which, as the authors believe, further establishes thoracic discography as an important tool in investigation of patients with thoracic pain if they are considered candidates for surgical treatment.

In addition to history of anaphylaxis to injected agents, pregnancy, inability to communicate the response during the procedure, and concurrent systemic or local infection, the contraindications to the procedure include spinal cord compression with or without myelopathy, bleeding disorder, and anticoagulation therapy. Review of the recent imaging study (MRI, CT, myelography) to evaluate the size of the vertebral canal is necessary prior to the procedure.

As described by Schellhas and Pollei²⁷ and complemented by Tibiletti,²⁸ thoracic discography is performed using fluoroscopy with the patient in a prone position. The skin is prepped twice with iodine solution, followed by two rinses with isopropyl alcohol, and sterile drapes are applied. The disc is accessed from the side opposite to patient’s clinical pain. If the pain is in the midline, the side of injection is selected based on patient’s comfort or individual doctor preferences. A 25-gauge, 3.5 inch spinal needle is advanced through the skin under intermittent fluoroscopic imaging and needle bevel rotation is utilized to control direction. In large patients (over 300 lb) a 5 inch, 22-gauge needle is used. Care is taken to direct the needle lateral to the interpedicular line and medial to costovertebral joints in order to avoid puncture of either the pleura or dura. Needle placement into the upper thoracic discs can be at times technically difficult or even impossible. After annular puncture, the needle is positioned into the center of the disc. Nonionic (e.g. iohexol, iopamidol) contrast agent is injected under fluoroscopic imaging in AP or lateral projection until a firm endpoint is obtained, leakage of contrast medium through annulus is detected, distraction of vertebral bodies occurs, venous opacification is observed, or the patient’s pain response prompts the discographer to terminate the injection. The volume of injected contrast and characteristics of the endpoint (firm, gradual, or no endpoint) are recorded. The amount of contrast agent administered is typically 0.6–1.0 mL unless the disc is incompetent and the injectate leaks outside of the disc. In patients with allergy to contrast, sterile saline solution can be injected instead. Spot images are obtained in AP, lateral, and/or oblique projections at each level. The patient’s responses are recorded. Provoked sensation is judged concordant versus nonconcordant (familiar versus unfamiliar) relative to clinical pain. The patient quantifies the pain on a scale 0 to 10 (verbal rating scale or visual analogue scale). Local anesthetic can be injected into painful discs.

A positive response is the provocation of 7/10 or higher concordant pain with a negative control disc above and/or below the positive level.

Postprocedural care after thoracic discography does not generally differ from care after lumbar or cervical discography with the exception of watching for possible signs and symptoms of pneumothorax. If one stays behind the rib during needle insertion, a pneumothorax should not, however, occur.

Anatomical considerations

The radiate ligament fans out superiorly, anteriorly, and inferiorly from the head of the rib and attaches the rib to the vertebral body (bodies). The intra-articular ligament bonds the apex of the rib head to the intervertebral disc and divides the synovial cavity into upper and lower halves. Neurohistological documentation of mechanoreceptors and nociceptors has been previously described in the articular tissue of costovertebral joints.²⁹ Innervation of the costovertebral joint is provided by nerve plexuses receiving fibers from the sympathetic trunks and perivascular nerve plexuses in bisegmental fashion, i.e. from the same level and from the adjacent level above.³⁰ Pain referral maps are not available for costovertebral joints. Radiographic changes in CV joints,³¹ symptomatic CV joint derangement in ankylosing spondylitis, seronegative spondyloarthropathy, rheumatoid arthritis, and functional costovertebral dysfunction was previously suggested in several reports.^32–36

Costovertebral joint arthrography was described first in 1988.³⁷ Five cases of CV arthropathy with pain referred into the abdomen and the loin are reported. At T11 and T12 levels, the authors performed costovertebral arthrography, which reproduced concordant pain. Injection of corticosteroid that followed resulted in long-term relief of pain.

The joint space is visualized using an oblique discography projection. The needle is advanced dorsomedial to the rib, lateral to the lamina, above the transverse process, and above the superior aspect of the disc, slightly lateral to the discography entry point in the oblique view. Less than 1 mL of solution is injected, including contrast medium, local anesthetic, and corticosteroid.

Benhamou et al. studied 28 patients with costovertebral arthropathies who presented with pseudovisceral pain.³⁶ The patients were worked up for chest, abdomen, and flank pain thought to represent the disease of genitourinary, gastrointestinal tract, heart, lung, and pleura. The initial working diagnoses were renal colic, angina pectoris, pulmonary embolism, and spinal tumor. After visceral pathology was ruled out, the musculoskeletal diagnosis was established by clinical examination, reproduction of pain with stressing of the CV joint mechanically, radiological studies, costovertebral arthrography, and anesthetic and corticosteroid injection of the joint. Pathology responsible for costovertebral arthropathy included ankylosing spondylitis, osteoarthritis, diffuse idiopathic skeletal hyperostosis, and degenerative arthropathies other than osteoarthritis.

Anatomical considerations

Articulation between the facet of the transverse process and the tubercle of the rib forms the costotransverse joint at the levels of T1–10. The lower levels of T11 and T12 lack this joint. T1–5 joint surfaces are reciprocally curved,⁶ while the lower levels are planar. The joint plane at the rib tubercle is convex and the facet on the transverse process is concave. The costotransverse joint is a synovial articulation and has a thin capsule. The joint is reinforced by the costotransverse, lateral costotransverse, and superior costotransverse ligaments. Innervation is from the articular branches of dorsal ramus and is not strictly segmental.²⁹ Furthermore, each joint is provided with an accessory innervation through twigs derived from the segmentally related intercostal nerves and intramuscular branches of the nerves supplying the multifidus muscle.²⁹ Neither studies in volunteers for the purpose of obtaining referral pain maps nor studies demonstrating pain relief after injecting the painful joint with local anesthetic are available to date.

As described by Lau and Littlejohn,³⁸ a needle is directed into the joint using an oblique fluoroscopic imaging at 45–60 degrees ipsilaterally. Due to the high contrast between the spine and lung tissue, coning is recommended to improve visualization. After administration of local anesthesia into skin, a 22-gauge spinal needle is inserted into the joint. The patient is asked to rotate slowly from side to side and the needle tip is observed to be in the same relationship to the joint. After confirmation of proper position of the needle tip in the joint, 0.2 mL of contrast agent with 1 mL of local anesthetic with corticosteroid is administered.