Biomechanics and Mechanisms of Spinal Cord Injury

The pathogenetic concept of SCIWORA is based on the premise that the inherently elastic juvenile spine can accommodate considerable intersegmental displacement without fracture or ligamentous rupture, but at the expense of the underlying spinal cord. Four mechanisms may be involved in the pathogenesis: hyperextension, flexion, distraction, and spinal cord ischemia.

Hyperextension of the cervical spine forces the interlaminar ligaments to bulge forward into the spinal canal and may be responsible for up to 50% narrowing of the canal’s diameter.^55,56 In addition, the cord thickens as it shortens during hyperextension.⁵⁷ Thus, the space available for the cervical cord within the canal is reduced during moderate hyperextension. During violent hyperextension, Taylor and Blackwood⁵⁶ and Bourmer⁵⁸ demonstrated rupture of the anterior longitudinal ligament, shearing of the intervertebral disk from the end plates, and retrodisplacement of the upper vertebrae. Dynamic radiography on fresh cadavers showed that elastic recoil of the displaced segment could result in perfect spontaneous reduction and give a normal radiographic appearance.⁵⁹ With its elastic constituents, the juvenile spine is especially susceptible to this sequence of momentary dislocation and spontaneous reduction. Intersegmental movement in children is further facilitated by their horizontal facet joints,^{21,23–26,32} and the immature vertebral body splits readily from its end plate within the brittle growth zone and leaves no radiographic trace.³³ The elastic elements,^21,32,60 horizontal facets,^21,24–26 wedge-shaped bodies,^21,23 and absence of uncinate processes likewise predispose the juvenile spine to hyperflexion myelopathy.^31,32

The spines of infants and young children are also uniquely vulnerable to distraction.^42,61 Leventhal found that although the elastic spinal column of neonatal cadavers could be stretched up to 2 inches without structural damage, the inelastic spinal cord ruptures if stretched more than 0.25 inch.⁶² The clinical counterpart of Leventhal’s study is the finding of frank rupture of the cord and meninges within a completely intact vertebral column in infants who had undergone forceful breech extraction.^24,63 In nonobstetric SCIWORA, the best evidence of distraction injury is found in patients when the thoracic spine is forcefully distracted in lap belt injuries. Rupture of the dura and spinal cord in these patients is evidenced by the resultant cerebrospinal fluid–pleural fistula.⁶⁴ Finally, the precarious architecture of the neonatal atlanto-occipital articulations predisposes the upper cervical cord to ischemic necrosis from occlusion of the vertebral artery.

The Age Factor

As predicted by biomechanics, age influences SCIWORA in three ways. First, many studies have shown that the incidence of SCIWORA is higher than that of other injury types in young children (0 to 9 years) with traumatic myelopathy, and vice versa in older children (Fig. 224-1).^14,44,46,65 Second, young patients with SCIWORA typically have more severe neurological injuries than do their older counterparts (Fig. 224-2).^5,46 In our own series, 21 of 30 children younger than 8 years suffered complete or severe injuries as compared with only 1 of 25 children older than 8 years (P < .000001).²⁷ Similarly, Osenbach and Menezes noted severe injuries (Frankel grades B and C) in 21 of 22 patients younger than 9 years versus comparable injury severity in only 1 of 9 older children.¹⁸ Third, several studies have shown that young children are far more susceptible to upper cervical SCIWORA whereas lower cervical lesions occur more often in older children. In our series, 9 of 10 upper cervical injuries occurred in children younger than 8 years versus 21 of 33 lower cervical injuries in older children (Fig. 224-3).²⁷ In the series of Osenbach and Menezes, all 7 nonobstetric upper cervical injuries occurred in children younger than 9 years.¹⁸ Unlike cervical lesions, SCIWORA involving the thoracic cord is evenly distributed over the pediatric age range, with no age predilection.^18,27,46 Thus, because young children tend to have upper cord injuries and most of them are complete or nearly complete, the rehabilitation potential of this age group is very limited and the socioeconomic burden is accordingly prohibitive.

FIGURE 224-1 Incidence of spinal cord injury without radiographic abnormality versus overt spinal column injury in children with traumatic myelopathy as a function of age. “Young” children are those younger than 9 or 10 years.

(Data from references^{14–18,34,40,44,75,83}).

FIGURE 224-2 Severity of neurological injury in children with spinal cord injury without radiographic abnormality as a function of age. “Young” children are those younger than 9 years.

(Data from references^{3,14–16,18,44,46,75}).

FIGURE 224-3 Correlation between patient age and level of neurological injury in spinal cord injury without radiographic abnormality. The vertical dashed line denotes the age (8 years) at which the physiologic properties of the juvenile spine change. The upper cervical spine was most prone to injury in infants and young children, whereas the lower cervical spine was at risk in all ages up to 16 years.

Radiographically Occult Instability

It is important to distinguish between the radiographically occult instability in SCIWORA and the overt instability seen in fracture-subluxation and subluxation without fracture (pure ligamentous injury). In SCIWORA, the spinal cord is injured without radiographic evidence of a fracture or subluxation that would indicate extensive ligamentous tearing and gross instability. Yet there had to have been sufficient bone displacement at impact to inflict damage to the cord. This implies that the stabilizing ligaments and fibrocartilaginous structures, although sufficiently elastic to stretch and recoil without rupturing, may have been severely sprained or even partially torn to render the involved vertebral segments vulnerable to repeated stress. The resulting instability or biomechanically precarious state is therefore “occult” rather than overt.^{20,21,24–27,33} The fact that flexion-extension radiographs fail to disclose abnormal movements in SCIWORA does not argue against the existence of occult instability because even gross instability can be masked by muscle spasm.³⁷ Moreover, motion studies executed with deliberate caution cannot compare with the random and unguarded neck movements in real-life activities. It is therefore not surprising that post-SCIWORA dynamic films do not reveal the precarious state of the injured segments.²⁷

Until the advent of MRI, the existence of occult instability in SCIWORA was unproved although strongly suggested by two phenomena: delayed neurological deterioration and recurrent SCIWORA. In a number of children with SCIWORA, neurological deficits are not detectable immediately after trauma but develop and worsen after a latent period of 30 minutes to 4 days, apparently in the absence of further trauma.^{27,40,43,66–68} The mechanism of this delayed deterioration remains speculative. A few cases may be due to posttraumatic occlusion of radicular arteries as a result of thrombosis or spasm, with subsequent delayed infarction of the cord,^43,66 but this must be rare. The majority of cases probably result from repeated “punch-drunk” trauma to the cord during the latent period. This presumption is supported by two observations. First, many of these children recall transient neurological symptoms at the time of the initial trauma, such as subjective weakness, acroparesthesia, or Lhermitte’s phenomenon, thus indicating that the cord had been distracted or hit during the intersegmental displacement. This displacement may well have overstretched or partially torn crucial ligaments and rendered the segments susceptible to repeated shifts during otherwise innocuous neck movements. Second, the incidence of delayed deterioration at our institution has decreased dramatically (13 of 24 cases treated before 1982³⁴ versus 2 of 33 since 1982²⁷) because of the rigorous implementation of prompt neck immobilization and neurosurgical referral for any child with neck pain and transient neurological symptoms. Many of these children have been found to have subtle neurological deficits that could easily have been missed on cursory examination, or they show evidence of myelopathy on somatosensory evoked potential (SSEP) testing. Such patients would presumably have been at risk for reinjury if they had not been immobilized expeditiously.

The phenomenon of delayed neurological deterioration in SCIWORA suggests that the spinal column has been weakened temporarily by the impact, even if the initial neurological symptoms are mild. This concept of a postimpact period of vulnerability is supported by the phenomenon of recurrent SCIWORA, noted up to 10 weeks after initial SCIWORA in 8 of 42 children treated before 1986.⁶⁹ Before this time, it had been our practice to immobilize children in hard collars for 2 months after SCIWORA. Four of the children with recurrent SCIWORA had removed their hard collars and were reinjured while engaging in forbidden sports such as wrestling, gymnastics, or basketball. The others were reinjured while wearing their collars. The neurological deficits resulting from the recurrent SCIWORA were much worse than those of the initial trauma, which in all cases were mild.⁶⁹ It is noteworthy that none of these children had overt instability on dynamic radiographs either at the first SCIWORA or at the time of reinjury. These normal studies belie the precarious state of the “once touched” spinal cord when the unprotected neck is subjected to the erratic motions of a child at play.

Findings on Magnetic Resonance Imaging

The discovery of MRI abnormalities in the soft tissues of the spine represents the first direct evidence of occult ligament damage in SCIWORA.⁷⁰ These abnormalities are extraneural, involving mainly ligaments and intervertebral disks, and neural, involving the cord itself.

Extraneural Abnormalities

The extraneural soft tissue injury demonstrated on MRI is well correlated with the mechanism of injury.⁷⁰ For example, rupture of the anterior longitudinal ligament, seen as loss of continuity of the low-signal prevertebral line, widening of the anterior intervertebral space, and anterior disk herniation (Fig. 224-4), is found in cases of hyperextension. Rupture of the posterior longitudinal ligament, seen as loss of continuity of the retrovertebral line and small posterior disk herniation (Fig. 224-5), is correlated with hyperflexion. Intradiscal hemorrhages are seen when a translational interbody shearing mechanism is suspected. Hemorrhage of the tectorial membrane (tear) has been associated with the violent shaking of child abuse (Fig. 224-6), and hemorrhages in the interspinous and interlaminar ligaments are seen with distraction injury (Fig. 224-7). Furthermore, the level of these soft tissue injuries corresponds exactly to the level of the neurological lesion.¹⁹

FIGURE 224-4 Thirteen-year-old girl who had mild partial C7 myelopathy after a hyperextension injury. Left, Plain film revealing slight separation of the disk space at C6-7. Right, Sagittal magnetic resonance image (TR, 2200; TE, 30) obtained 3 days after injury showing discontinuity of the prevertebral line of signal void, a finding suggestive of a tear in the anterior longitudinal ligament, anterior C6-7 disk herniation (large arrowhead), and possible end-plate separation (small arrowhead).

FIGURE 224-5 A 9-year-old boy sustained a lateral flexion injury and had complete Brown-Séquard syndrome at initial evaluation. A sagittal long TR (TR, 4000; TE, 18) magnetic resonance image shows interruption of the retrovertebral line of the signal void, consistent with a disrupted posterior longitudinal ligament and C2-3 disk herniation (arrow).

FIGURE 224-6 Sagittal magnetic resonance image in an infant with a tear of the tectorial membrane resulting from whiplash-shake syndrome. It is seen as a subacute hemorrhage (hyperintense on short TR) within the soft tissue behind the odontoid process.

FIGURE 224-7 T2-weighted magnetic resonance image in a 6-year-old child with hyperflexion and mild cord syndrome. Widespread edema (arrows) is evident in the interspinous ligaments and muscles of the upper five cervical levels.

These extraneural MRI findings strongly support the hypothesis that SCIWORA is associated with spraining or partial tearing of crucial stabilizing structures of the vertebral column.^19,27 For years, these ligamentous and disk injuries, which are responsible for the post-SCIWORA vulnerable period, had escaped detection by conventional radiology but are now “exposed” by MRI. The continued use of MRI is expected to reveal more information about SCIWORA.

The signal characteristics of hemorrhage within non-neural tissues are detectable within hours of the injury on T1-weighted images as lines of hyperintensity because extravasated blood is converted within hours to methemoglobin (MHb) in non-neural tissues versus days to weeks in neural tissues (see later).^4,30,71,72 Extraneural injuries in SCIWORA can thus be diagnosed by MRI within hours after the injury. The problem is that the adjacent fascial, muscular, and fat planes also produce high signal and can overwhelm the small streaks of MHb. The use of fat suppression sequences reduces the high signal of fat but not that of blood on short relaxation time (TR) (T1-weighted) images and is effective in highlighting the soft tissue hemorrhages in SCIWORA.

Neural Abnormalities

The MRI abnormalities seen within the traumatized spinal cord correspond to the signal characteristics of the various metabolized forms of extravasated hemoglobin (Table 224-4). Fresh blood from spinal cord hemorrhage contains intracellular oxyhemoglobin, which is rapidly converted to deoxyhemoglobin (DHb) within 3 hours. DHb is a stable intracellular compound and appears hypointense on long TR (T2-weighted) images.^{4,10,30,71,73} Gradient echo techniques, which are particularly sensitive to paramagnetic substances such as DHb, detect acute hemorrhage even better than do standard long TR sequences.^{12,30,71,73,74} Oxidation of intracellular DHb to MHb begins after approximately 3 days, although DHb may still be detectable up to 8 days after injury, especially with larger hematomas. Intracellular MHb is hyperintense on short TR images but hypointense on long TR images, whereas extracellular MHb appears hyperintense on both short and long TR images. Subsequent hemolysis around day 7 liberates MHb into extracellular tissues, where it may persist for months. Extracellular MHb is finally converted to hemosiderin after phagocytosis; this substance, which may persist for life, is markedly hypointense on long TR images and slightly hypointense on short TR images.⁷³ The larger the hemorrhage, the longer it takes for all the DHb to be converted to MHb and ultimately for MHb to become hemosiderin. The signal characteristics are thus dependent not only on the time elapsed since injury but also on hematoma size.

TABLE 224-4 Magnetic Resonance Imaging Signal Alterations in Spinal Cord Injury As a Function of the Chemical State of Blood

Five patterns of parenchymal (cord) findings are seen on the post-SCIWORA MRI scan.⁷⁰ (1) Complete disruption of the spinal cord is seen with severe distraction injuries in very young children (Fig. 224-8). This is usually accompanied by hypointensities on gradient echo sequences (DHb) of the rostral cord stump. The common sites of cord disruption are the upper cervical and upper thoracic levels. Severe flexion injury occasionally produces a similar picture. (2) Major cord hemorrhage has occurred when more than 50% of the cross-sectional area of the cord on axial MRI shows evidence of extravasated hemoglobin (Fig. 224-9). Depending on the timing of MRI, the signal characteristics may be those of DHb, intracellular or extracellular MHb, or hemosiderin. This pattern is associated with severe neural deficits and a dismal prognosis. (3) Minor cord hemorrhage has occurred when less than 50% of the cross-sectional area of the cord shows evidence of hemorrhage. It is associated with moderately severe initial deficits but a decent chance for recovery. (4) Edema only, without hemorrhage, is noted if the T1-weighted image shows isointensity or slight hypointensity but the T2-weighted image is intensely bright. No stage of hemoglobin metabolism produces this pattern of signal characteristics. Edema is predictive of a good outcome. (5) Approximately 35% of patients with clinically proven SCIWORA do not have MRI abnormalities in their spinal cords. These patients have an excellent prognosis for complete recovery.⁷⁵

FIGURE 224-8 Two-month-old with a complete C6 injury subsequent to a motor vehicle accident. A sagittal short TR (TR, 800; TE, 30) magnetic resonance image obtained 6 days after injury shows loss of cord continuity consistent with cord transection.

FIGURE 224-9 Twenty-one-month-old boy with a complete C3 injury after suffering physical abuse. A, Computed tomographic myelography performed 15 days after injury shows marked enlargement of the cord; the cause (hemorrhage or edema) could not be discerned. B, Axial short TR (TR, 600; TE, 20) magnetic resonance imaging (MRI) performed 16 days after injury shows intraparenchymal hyperintensity expanding and replacing the cord substance, consistent with methemoglobin. C, Sagittal short TR (TR, 600; TE, 20) MRI at day 16 shows the methemoglobin-suffused segment of the cord extending from C2 to T1. D, Sagittal short TR (TR, 600; TE, 20) MRI performed 3 months after injury shows marked cervical cord atrophy.

It is our policy to repeat the MRI study on days 6 to 9 after the injury, regardless of whether the acute MRI (on days 0 to 3) showed abnormal findings. Edema takes 3 to 4 hours to develop, and small dots of DHb may not be easily detectable until it is converted to MHb, at which point contrast between the spinal cord and hemoglobin is greater. If the initial MRI findings are abnormal, a later study at 3 to 6 months is performed to detect syrinx formation and marked atrophy of the cord.

Thoracolumbar SCIWORA

About 13% of cases of SCIWORA involve the thoracic cord. SCIWORA of the thoracic spine is typically due to violent trauma, such as high-speed automobile accidents or crushing injuries.* Accordingly, these patients have a high incidence of associated thoracic, abdominal, and pelvic injuries. Three discrete subgroups of patients with thoracic SCIWORA are encountered. One consists of children who sustain severe distraction injuries related to lap seat belts and high-speed vehicular collisions.⁶⁴ CT myelography typically shows spinal cord and dural rupture and free flow of contrast material into the mediastinum.^25,42,76,77 The splinting effect of the rib cage normally protects the thoracic spine against horizontal displacement as a result of flexion and extension forces. However, because the sternocostal joints permit a fair amount of “bucket handle” movements, the rib cage provides little protection against longitudinal distraction. In lap belt injuries, there is always a component of forceful distraction caused by forward propulsion of the upper part of the chest while the pelvis is harnessed by the lap belt. The elastic juvenile spinal column then recoils like a spring to restore normal configuration after having stretched enough to rupture the cord and dura. As with upper cervical SCIWORA, thoracic SCIWORA secondary to distraction occurs almost exclusively in children younger than 8 years.

A second subgroup of thoracic SCIWORA includes children who are crushed by slow-moving vehicles.^3,27 In these cases, tire marks are clearly imprinted on the body, so the body position and exact site of crushing can be determined. All thoracic SCIWORA victims injured by crushing were lying prone when run over by the cars that left tire marks on their backs. In contrast, children with tire marks on the abdomen or anterior chest region typically suffer intraperitoneal injuries but rarely SCI. In the group with ventral tire marks, the gentle S curve of the thoracic spinal column, being solidly buttressed by the ground, merely straightens out and is thus protected against excessive intersegmental displacement; the abdominal viscera bear the brunt of the injury. In the group with dorsal tire marks, the spinal column is hyperextended into the softness of the chest and abdominal cavities, thereby resulting in hyperextension cord injury and rupture of retroperitoneal organs while sparing the well-protected intraperitoneal contents.

The last subgroup of thoracic SCIWORA involves a direct hit on a standing child by a high-speed vehicle. These children all had severe or complete cord lesions, and all sustained severe or even fatal extraneural injuries. The presumed mechanism is an egregious self-reducing deformity of the thoracic spine caused by an extreme force that has overcome the protective splinting of the rib cage, which explains the high fatality. Ironically, because attention is often focused on the other life-threatening injuries during initial resuscitation, the paraplegia is not infrequently missed until the child’s condition stabilizes and survival ensured, thereby affording a perversely fertile field for litigation.

Management

The initial management of a child with a presumed SCI begins in the field (Fig. 224-10). Patients with symptoms or mechanism of injury compatible with SCI should be immobilized supine in a hard collar on a fracture board. If the child’s head is large, the body from the shoulders down is propped up with folded blankets or foam sheets to prevent forced flexion of the neck.

FIGURE 224-10 Management algorithm for the evaluation and treatment of children with spinal cord injuries. The main focus is on recommendations for patients with possible spinal cord injury without radiographic abnormality (SCIWORA). ER, emergency room; CT, computed tomography; MRI, magnetic resonance imaging.

When possible, a careful history is obtained from the patient or eyewitnesses to determine the cause and mechanism of injury. In addition, interpretation of the associated non-neural injuries frequently gives clues to the mechanism of the spinal injury. All patients undergo complete cervical spine radiography, and a full spine survey is performed on patients with presumed thoracic cord lesions or severe multisystem trauma. Patients with SCI but without abnormalities on plain films undergo thin-section axial CT with bone algorithms to rule out an occult fracture at the level of the neurological lesion, followed by MRI. The radiographic evaluation is not complete until good-quality flexion and extension films of the affected area of the spine are obtained to rule out overt ligamentous instability. In many patients, severe spasm of the paraspinal muscles precludes adequate flexion for several days; the dynamic study must be repeated a few days later.

In children with severe cord injuries diagnosed within 8 hours of the injury, a 24-hour course of high-dose corticosteroids is initiated. This recommendation is based on the beneficial effects of high-dose methylprednisolone in adult patients reported by the National Acute Spinal Cord Injury Study (NASCIS III).^80–82 There are no equivalent data for children, and our endorsement of corticosteroids for SCIWORA must be understood to be “empirical” rather than evidence-based. Patients with cervical cord injuries are immobilized in a Guilford brace (B.A. Guilford and Sons Orthotic Laboratory Ltd., Cleveland, OH) or other cervical-thoracic brace for 3 months. Mid to low thoracic lesions are treated with a thoracolumbar orthosis (TLSO), and children with upper thoracic lesions are fitted with a TLSO with added chin and occiput supports. Physical and occupational therapies are begun in the intensive care unit (ICU). Those with severe residual deficits are transported to a rehabilitation facility when clinically stable. Patients with mild residual weakness are enrolled in an outpatient physical therapy program. Bracing is maintained for 3 months, during which time both contact and noncontact sports are strictly prohibited. The brace is favored over a hard collar for several reasons: (1) it can be tailored to fit children as young as 1 year; (2) it imparts greater limitation of motion; (3) it produces less skin irritation and chafing than a collar; and most importantly, (4) a child cannot easily remove the brace, which also interferes with play activity, both helping to enforce compliance. It may be helpful to embellish with graphic details the estates of paralysis, wheelchair existence, decubitus ulcers, impotence, and so on to older children and their families as a warning of the dire consequences of protocol violation.

Our decision to immobilize for 3 months is based anecdotally on the fact that one patient treated with just 2 months of wearing a hard collar suffered recurrent SCI with trivial trauma 10 weeks after the initial SCIWORA. Since changing from our old protocol of using just a hard collar for 2 months,³⁴ no recurrent SCIWORA has been reported.⁶⁴ The improved outcome may just as well be due to switching of the collar to bracing as to switching from 2 months to 3 months. Currently, there are institutions that favor the 2-month protocol, and unless adverse results are reported with the shorter immobilization, the choice of 2 versus 3 months must be left to the individual belief of the treating physician. After 3 months, flexion-extension radiographs are obtained with the child out of the brace to check for late instability. In our experience, this is rarely a problem with SCIWORA, nor is late deformity such as kyphosis or scoliosis.^19,20,83

SCIWORA in the Comatose Child

The question is often asked how to “clear the neck” of a comatose child. A normal spine film and CT scan rule out most cases of gross fracture and dislocation, but not SCIWORA. In some ICUs, a comatose child is kept on a hard collar and flat log-rolling regimen as long as the child is unresponsive or being maintained on mechanical ventilation, but this could last weeks and skin breakdown around the chin and occiput can become troublesome long before a valid neurological examination is feasible. In this setting we suggest performing MRI and using SSEPs to detect spinal cord conduction block.⁷⁵ If both MRI and SSEPs are normal, the spinal precautions are removed. If either the SSEPs or MRI is abnormal, strict spine precautions are observed until the patient is able to mobilize, at which time a proper brace is fitted.

Outcome

Several studies,^5,18 including our own from 1989,²⁷ have shown that the main predictor of long-term outcome in children with SCIWORA is the admission neurological status. Children with complete lesions rarely improve. Those with severe but incomplete lesions often improve with time but seldom regain normal function. Only patients with mild to moderate initial deficits can hope for a full recovery (Fig. 224-11). Thus, the overall outcome for SCIWORA can best be improved by preventing serious neurological damage. Primary prevention programs within the community such as ThinkFirst have lowered the overall incidence of SCI in children. Once SCIWORA has occurred, the pediatric neurosurgeon’s responsibility would be to minimize the extent of neuronal damage through three avenues: (1) rule out other (non-SCIWORA) categories of SCI, (2) anticipate and prevent delayed neurological deterioration, and (3) prevent recurrent injury. This means that occult fractures, disk herniation, epidural hematoma, and overt ligament instability must be excluded with dynamic radiographs, CT, and MRI before the diagnosis of SCIWORA can be established because each of these entities mandates more than simple bracing. The results reported by Pang and Pollack²⁷ and Pollack and colleagues⁶⁹ suggest that the incidence of delayed neurological deterioration and recurrent SCIWORA can be lowered by compulsive prehospital care and a low threshold for the use of spinal immobilization in any injured child meeting our diagnostic criteria for SCIWORA.

FIGURE 224-11 Long-term neurological outcome in children with spinal cord injury without radiographic abnormality as a function of initial neurological status.

Prognostic Value of Magnetic Resonance Imaging in SCIWORA

Although preventive measures are still paramount in improving outcome, recent MRI data⁷⁵ suggest that our earlier outcome predictions made without MRI²⁷ have become somewhat obsolete. In our recent series,⁷⁵ we correlated the five neural MRI abnormalities (really four abnormalities plus “normal”) with the four grades of neurological deficit (complete, severe, mild, and normal) at initial evaluation and 6 months later in a much larger population (Fig. 224-12). We found instead that although the general conclusions from the original outcome analysis²⁷ stand, in specific cases MRI surpasses the initial neurological examination in predicting outcome. The implication of cord disruption, once clearly portrayed on MRI, is self-evident. In all patients with this finding, who invariably have complete sensorimotor paralysis, the deficits remain complete. Major hemorrhage is also associated with complete or severe deficits (Frankel grades B and C) at initial evaluation, as well as with very poor long-term outcome.^71,84 Patients with this finding either remain profoundly impaired or achieve only minimal recovery. No one in our series ever ascended from severe to the mild or moderate category (Frankel grade D) of cord syndromes, although two patients initially found to be complete were raised to the severe grade. In short, cord disruption and major hemorrhage are no better predicators of outcome than the admission neurological examination.

FIGURE 224-12 Correlation of findings on magnetic resonance imaging (MRI) with neurologic syndromes at initial evaluation and at 6 months’ follow-up in 50 patients in the authors’ updated series. The numbers in each box correspond to numbers of patients in a syndrome category. The fate of cord disruption or pulping is obvious. All patients with major hemorrhage have complete syndromes, but only a few improved to the severe grade. All minor hemorrhage was associated with either complete or severe grades, but a significant number improved to a mild grade at 6 months. An even larger percentage of patients with edema only improved to the mild grade. Most patients with normal cord findings have mild injuries and improve to normal status at 6 months. The 2 patients who had complete syndrome but became normal within 2 days (cord concussion) had normal findings on MRI.

The situation is different with the other categories of neural MRI findings. All patients with minor hemorrhage have severe initial deficits (Frankel grades B and C) at initial evaluation, but only 60% remained in this grade at 6 months and 40% ultimately recovered to the mild grade (Frankel grade D). Furthermore, about 44% of the patients with edema only on MRI initially had moderate to severe deficits (Frankel grade B and C), yet none of them remained in the severe grades; overall, 75% improved to the mild grade and 25% became normal. Thus, in patients with minor hemorrhage and edema only, the MRI finding is a better predictor of outcome than the initial neurological examination. By and large, edema only and minor hemorrhage, in that order, are associated with a much better prognosis than major hemorrhage, even though the initial deficits may be equally severe in the three groups. Among the four types of neural abnormalities on post-SCIWORA MRI, edema-only remains the best indicator for good recovery.

All patients with no cord abnormality on post-SCIWORA MRI, except two, had mild initial deficits, and all made a complete recovery. The two exceptions initially had complete sensorimotor paralysis but within 2 days had also made a complete recovery.⁷⁰ This phenomenon of “cord concussion” had been described by Quencer, who thought that the “chemical alterations” within the cord that led to the initial paralysis did not generate abnormal MRI signals.⁸⁵ Thus, normal MRI findings always presage an excellent prognosis regardless of the initial neurological grading.

Anatomy and Biomechanics

The occipitoatlantoaxial (O-C1-C2) articulations function as a single unit that procures movements of the head about the spine. The atlas works as a biconcave washer between two spheres of motion. The upper articulation, between the occipital condyles and the atlas, is cup shaped in the sagittal plane and medially tilted in the coronal plane. This orientation allows flexion, extension, and some lateral bending, but minimal rotation. In contrast, the biconvex and laterally tilted facet articulation between the atlas and axis allows a large degree of rotation centered about the dens. Sagittal movement of the anterior arch of the atlas along the slightly pointed dens provides additional flexion and extension.

Movements within the O-C1-C2 unit are facilitated by the redundant articular capsules and thin occipitoatlantal and atlantoaxial membranes, which in turn impart minimal stability.⁹⁸ Rather, stability is provided by the strong internal ligaments between C2 and the occiput, chiefly the tectorial membrane, a well-developed continuation of the posterior longitudinal ligament that straps the body of C2 to the clivus; by the paired alar ligaments that connect the dens to the occipital condyles; by the cruciate ligament; and by the apical dental ligament (Fig. 224-16). Flexion of the basion beyond the tip of the dens is limited by the tectorial membrane, whereas extension is checked by the tectorial membrane and by bony contact between the opisthion and the arch of C1. Lateral bending and rotation are controlled by the alar ligaments.¹⁰⁵ The importance of the tectorial membrane and the alar ligaments was demonstrated by Werne, who found that sectioning of the former results in excessive flexion and extension, as well as vertical separation of the basion from the dens beyond the normal range of 1 cm, whereas sectioning of the latter results in increased lateral bending and rotation.¹⁰⁶ When both the alar ligaments and tectorial membrane are cut, complete occipitoaxial dissociation occurs. The remainder of the internal ligaments—namely, the ascending band of the cruciate ligament and the apical ligament—are ineffectual in limiting motion. Finally, cadaver studies have shown that paraspinal muscle action also contributes to stability of the O-C1-C2 unit.^107,108

FIGURE 224-16 The joints and ligaments of the craniovertebral junction. Top, Posterior view. Bottom, Lateral view. Lig, ligament.

(From Wolf JW, Johnson RM. Cervical orthoses. In: Bailey DW, Sherk HH, eds. The Cervical Spine. Philadelphia: JB Lippincott; 1983:54-61.)

Pathology and Pathogenesis

Traumatic AOD results from high-energy impacts that cause rupture of the tectorial membrane and alar ligaments.^{65,99,101,102,109–111} Pedestrian-vehicle accidents are the most common cause. Three factors make children more vulnerable to AOD than adults. First, children are more frequently involved in pedestrian-vehicle accidents.¹¹² Second, the relatively larger head in children in comparison to the torso generates a greater shearing moment. Third, the occipitoatlantal joint is inherently more unstable in children.¹¹³

The force vectors involved in AOD may be reconstructed by circumstantial evidence. Adams’ study revealed a high incidence of associated submental lacerations and mandibular fractures,⁹⁹ and others have reported posterior atlas fractures,¹¹⁴ both of which point to a hyperextension mechanism. Furthermore, 9 of Adams’ 12 subjects had pontomedullary lacerations,⁹⁹ a lesion that is characteristic of hyperextension and rotation.^115–117 However, approximately 25% of patients with AOD have coexistent anterior atlantoaxial subluxation, and other patients are seen with spinomedullary transection by a posteriorly pointing dens.^99,110 These findings suggest a hyperflexion mechanism.¹¹⁸ Because the alar ligaments—the check ligaments for rotation—are also ruptured in AOD, the forces must be directed obliquely to the head.^{105,119–121} Thus, AOD is most often caused by hyperextension-distraction forces directed obliquely to the chin, but hyperflexion-distraction forces to the occiput are involved in some cases.

Recently, Sun and colleagues¹²² used high-resolution MRI to reveal a progressive pattern of disruption of the structures responsible for the stability of the O-C2 joint. Progressing from minor to severe injuries, the sequence of successive and additive involvement of structures is as follows: suboccipital musculature, atlanto-occipital and atlantoaxial membranes, apical and alar ligaments, tectorial membrane, and bilateral atlanto-occipital joint capsules. These authors postulated that involvement of the tectorial membrane is a critical threshold for instability at the O-C2 joint. Tectorial membrane damage can consist of frank rupture (Fig. 224-17) or stripping of the membrane from the clival or C2 attachments (Fig. 224-18). In both cases, the mechanical advantage of the membrane’s stabilizing function is rendered ineffective. Weakening of this membrane appears to “expose” the O-C1 joint capsules to overt disruption, the denouement of the AOD scenario, with the cord being left vulnerable to traction and crushing injuries. In our recent series of AOD in which 14 children underwent MRI,¹²³ seven tectorial membranes had ruptured, three had been stripped off the clivus, and only four had remained intact.

FIGURE 224-17 Sagittal T2-weighted magnetic resonance image in a 7-year-old child with atlanto-occipital dislocation. The tectorial membrane is completely disrupted. The upper stump (arrow) is still attached to the clivus. The alar ligaments are not visualized above the tip of the dens, thus suggesting that they have also ruptured.

FIGURE 224-18 Sagittal T2-weighted magnetic resonance image in a 9-year-old child with atlanto-occipital dislocation. The tectorial membrane (arrow) is lifted off the clivus. The clival-dens space contains blood or edema fluid.

Traynelis and coworkers classified AOD into three types.¹¹⁸ Type I, the most common, is characterized by anterior displacement of the occiput on the atlas. Type II is a longitudinal separation of the two structures, and type III is posterior displacement of the occiput. The prevalence of anterior AOD seems contradictory to the postulated hyperextension vector, but given the highly unstable nature of this injury, the location of the head relative to C1 may be more a function of postimpact positioning than the dynamics of the impact. As an example, when an injured child lies on a flat board, the occipital prominence of the large head forces the neck into flexion, thereby artificially creating an anterior displacement pattern. Rigid classification of AOD according to the relative position of the head therefore carries no correlative value with regard to mechanism and should be discouraged. Isolated examples of pure lateral dislocation,¹²⁴ as well as pure rotatory subluxation, further attest to the extreme looseness of the disrupted O-C1 joint.^125,126

Clinical Findings

Children with traumatic AOD have signs of brainstem, spinal cord, and cranial nerve injury. Thirty percent of children who survive AOD are apneic or in full cardiorespiratory arrest at the scene of the accident.^47,127–133 Other brainstem findings include pupillary abnormalities, rotatory nystagmus, ocular bobbing, decerebrate posturing, and variable alterations in consciousness.¹²³ The motor deficits include quadriparesis, cruciate paralysis, crossed hemiplegia, or hemiparesis, depending on the level of the pyramidal tract lesions. Brainstem injury in AOD is due to a combination of actual disruption of tissues, spinal medullary compression by bony structures or epidural hematoma, and ischemia secondary to vertebral artery spasm or thrombosis.¹³⁴ Thrombosis of the anterior spinal artery produces the unique syndrome of ipsilateral hypoglossal palsy and crossed hemiplegia.¹³⁵ The caudal six pairs of cranial nerves may also be injured from downward traction of the medulla against the fixed points at their exit foramina.¹³⁶ However, the most common cranial neuropathy in AOD, sixth nerve palsy, is probably caused by the concomitant head injury. Posttraumatic hydrocephalus must be ruled out if the abducens palsy persists.

The true mortality rate associated with AOD is unknown. Of 35 nonfatal cases of pediatric AOD reported in the literature,^{104,111,122,129–133,136–145} only 5 had no initial deficits except for neck pain and torticollis. Of the 30 patients with deficits, 12 recovered fully; 3 remained quadriplegic and ventilator dependent, 2 of whom died 8 months after injury. The remaining 15 patients experienced various intermediate recoveries. Thus, children with AOD who have incomplete initial paralysis often make significant improvements, but in isolated cases, even initially flaccid paralysis does not preclude a good recovery.

Radiographic Diagnosis

The diagnosis of AOD is contingent on a high level of suspicion in trauma victims, especially those with mandibular and facial fractures or with instantaneous cardiorespiratory instability. Casting of the brainstem with subarachnoid blood, seen on 25% of head CT scans in patients with AOD, should immediately suggest a distraction injury at the atlanto-occipital junction. The radiographic “chase” for the diagnosis should be relentless because survival depends on recognition. Although gross separation of the occiput and atlas is obvious on plain x-ray films, horizontal luxations are much harder to diagnose. The literature is replete with imperfect diagnostic criteria. A critical review of these criteria allows the clinician to use multiple methods to augment the positive yield.

One of the oldest criteria is the interval between the tip of the dens and the basion (DB interval) (Fig. 224-19). Wholey and associates examined 480 adult and 120 pediatric cervical spine radiographs and determined that the upper limit of the DB interval is 5 mm in adults and 10 mm in children, the age difference being attributed to incomplete summit ossification of the dens in children.¹⁴⁶ Earlier application of the DB interval revealed considerable overlap between normal subjects and patients with AOD, possibly because of the mixed age groups.¹¹⁰ The most recent use of the DB interval exclusively in children showed that the value in normal children is less than 12.5 mm; the values in 11 children with proven AOD were all greater than 14 mm.¹³⁸ However, several problems are inherent in this method. First, as with any absolute radiographic measurement, the number depends on the target-film distance. Wholey’s original measurements were done on an upright film at a target-film distance of 6 ft. When the DB intervals in 400 subjects were measured on supine films taken at 40 inches, Harris and colleagues found the upper limit to be 12 mm in adults.¹⁴⁷ Second, the variable ossification of the odontoid summit in children makes universal application of a maximal DB interval unreliable. Third, the measurement varies with flexion, extension, and axial traction in normal subjects. Nevertheless, a DB interval greater than 14 mm in a child is almost certainly indicative of AOD, but values between 10 and 14 mm still defy precise designation.

FIGURE 224-19 The dens-basion interval introduced by Wholey and colleagues.¹⁴⁶ An interval of 10 mm is abnormal, and 14 mm is indicative of atlanto-occipital dislocation in children. B, basion; D, dens.

In 1980, Dublin and associates measured the distances from the posterior cortex of the mandible to the anterior arch of C1 (A) and to the odontoid (B) (Fig. 224-20).¹⁰⁹ In anterior AOD, both A and B should be increased. However, because the mandible and the dens are not sagittally coplanar, these measurements vary greatly with the incident angle of the x-ray beam and are not consistently reproducible.³⁴ We do not recommend this method.

FIGURE 224-20 Dublin’s method, which uses the distances between the posterior cortex of the mandible to the anterior arch of C1 (A) and to the odontoid (B). In anterior atlanto-occipital dislocation, both A and B should increase. The normal range of A is −10 to + 9 mm; the normal range of B is +2 to +17 mm.

To circumvent the problems with noncoplanar landmarks, Kaufman and coworkers measured the width of the atlanto-occipital joint itself.⁴⁹ They found that the distance between any two opposing points along the joint surface should not exceed 5 mm in children on lateral or anteroposterior skull films. The problem with their original description is the use of plain films; superimposition of the mastoids and imperfect overlap of the two atlanto-occipital joints often make interpretation impossible.

Two methods using dimensionless numbers have been devised to eliminate the error attributable to radiographic magnification. The Powers ratio is obtained by dividing the distance between the basion (B) and the posterior arch of C1 (C) by the distance between the opisthion (O) and the anterior arch of C1 (A) (Fig. 224-21A).¹¹¹ Survey of a normal population yields a mean BC/OA ratio of 0.77 ± 0.09. In the so-called type I AOD, the BC/OA ratio is increased because BC widens whereas OA shortens. Powers found that ratios greater than 1 are always indicative of anterior dislocation. Although highly predictive of anterior dislocation, the Powers ratio has two major shortcomings: it fails to diagnose distraction and posterior dislocation, and the opisthion is often difficult to pinpoint on the lateral spine films. In addition, anomalies of the posterior arch of C1 and the foramen magnum invalidate the ratio. The use of sagittally reconstructed CT has greatly simplified localization of the opisthion (see Fig. 224-21B).

FIGURE 224-21 Powers’ ratio. A, The ratio is obtained when the distance between the basion (B) and the midpoint of the posterior arch of C1 (C) is divided by the distance between the opisthion (O) and the midpoint of the anterior arch of C1 (A). The mean BC/OA ratio is 0.77. Any value greater than 1 is indicative of atlanto-occipital dislocation (AOD). B, Sagittal computed tomographic reconstruction in a patient with AOD. The BC/OA ratio is greater than 1.

The second dimensionless method uses the X lines proposed by Lee and colleagues¹³⁴ (Fig. 224-22). Normally, BC2L and C2O tangentially intersect the dens and the highest point on C1, respectively. If both lines miss their tangential points, AOD is suspected. This method is not useful in children because its validity depends on atlantoaxial integrity and a fully developed dens.

FIGURE 224-22 Lee’s X line. A line drawn between the basion (B) and the midpoint of the spinolaminar line of C2 (C2SL) should tangentially intersect the posterosuperior aspect of the dens. A line drawn from the opisthion (O) to the posteroinferior corner of the C2 body (C2) should also tangentially intersect the C1 spinolaminar line. If both lines miss their respective intersections, atlanto-occipital dislocation is suspected.

In 1993, Harris and coauthors began using the basion-axial interval (BAI), defined by a line drawn tangential to the posterior cortical surface of the axis (posterior axial line [PAL]).^147,148 The BAI is the distance between this line and a parallel line drawn through the basion (Fig. 224-23A). The normal BAI in adults ranges from +12 mm (basion anterior to the PAL) to −4 mm (basion behind the PAL). With anterior dislocation, the basion moves away from the dens and the BAI exceeds 12 mm (see Fig. 224-23B), whereas with posterior dislocation, the basion moves far behind the dens and the BAI becomes less than −4 mm. In 50 children studied, the anterior limit of the BAI was still +12 mm, but the basion never moves behind the PAL. These authors showed that when used in conjunction with the DB interval, the BAI is more accurate in adults than either the Powers ratio or X lines. In very young children with a pointed dens, however, the PAL becomes excessively oblique forward. This artificially moves the posterior reference point of the measurement forward and, in effect, underestimates the true extent of the anterior basion displacement.

FIGURE 224-23 Harris’ basion-axial interval (BAI). A, The BAI is the distance between the posterior axial cortical line (PAL) and a parallel line drawn through the basion. The normal BAI is between 0 and 12 mm in children. B, Lateral cervical radiograph of a patient with atlanto-occipital dislocation showing the BAI to be greater than +17 mm.

In 2000, Sun and colleagues introduced the C1-2/C2-3 posterior interspinous ratio to assess the integrity of the atlanto-occipital joint.¹²² This ratio is obtained from the lateral plain radiograph by dividing the shortest distance between the posterior C1 ring and the C2 spinous process by the distance between the spinous processes of C2 and C3 (Fig. 224-24). These authors used abnormality of the tectorial membrane as the positive indicator for AOD and found that normal children and injured children with intact tectorial membranes (presumably without AOD) have an interspinous ratio of less than 2.5 whereas seven of eight children whose tectorial membranes had been damaged (presumably with AOD) had a ratio greater than 2.5. The hypothetical mechanism underlying an abnormal ratio in AOD was supposedly related to forward rocking and downward pressing of the occipital condyle on the anterior rim of the C1 facet secondary to a loose atlanto-occipital joint, thereby forcefully “jacking up” the posterior arch of C1 away from C2.¹²²

FIGURE 224-24 The C1-2/C2-3 ratio. Sun and colleagues suggest that the ratio should be less than 2.5 with a stable atlanto-occipital relationship.¹²²

In a recent study comparing the diagnostic sensitivity and specificity of Wholey’s DB interval, Powers’ ratio, Harris’ BAI, and Sun’s interspinous ratio, Sun’s method was found to have the lowest sensitivity (25%).¹²³ There are several explanations. First, AOD has been shown to occur without demonstrable injury to the tectorial membrane,¹²³ which invalidates the use of its intactness as the reference standard for AOD. Second, if the abnormal C1-2 interspinous widening is due to altered contact relationships between the condyle and the facet of C1 after the loss of ligamentous restraint, postinjury handling of the occiput could randomly rearrange this relationship and unpredictably vary the C1-2 distance. If one takes Sun’s logic to extremity, in the most ferocious case of AOD when all the stabilizing ligaments are completely torn and the condyles are freely floating away from C1, C1 should not be pressed or levered by the condyle and the ratio should stay normal. This has in fact been observed in some of our patients.¹²³ Thus, paradoxically, the more complete the disintegration of the O-C1 joint, the more likely would Sun’s criterion be falsely negative.

Despite such a large array of published radiographic criteria, it is ironic that 6 of the senior author’s 16 cases of AOD fell into the diagnostic borderland between normal and abnormal.¹²³ This led to the realization that with the exception of Kaufman and associates’ flawed study,^49,131 all the aforementioned criteria use bony landmarks that do not actually participate in the atlanto-occipital articulation. It stands to reason that direct measurement of the O-C1 joint interval should be a more precise indicator of O-C1 integrity.

Condyle-C1 Interval

Such a study using both coronal and sagittal reformatted CT images through the condyle-C1 joints of 89 normal children was published by us in 2007.¹⁴⁹ We found that the normal occipital condyle-C1 joint in children is held tightly together by strong ligaments, with a mean condyle-C1 interval (CCI) of 1.28 mm (Fig. 224-25). No individual gap measurement in any projection exceeds 2.5 mm. In addition, there is strong left-right joint symmetry in both the CCI and conformational anatomy. Linear regression analysis between age and CCI shows that the mean CCI value is extremely stable through the age span of birth to 18 years and, furthermore, the left-right symmetry is also age conserved.

FIGURE 224-25 Placement of the measurement points (arrowheads) on the coronal (A) and sagittal (B) occiput-C1 joint surfaces of a normal 7-year-old boy.

Based on these findings, we set a new diagnostic criterion for AOD: if the observed mean CCI in a patient is 4 mm or greater, or if the two joints are grossly asymmetric.¹⁴⁹ On testing this new criterion in a series of 16 patients with AOD,¹²³ we found that 14 patients had bilateral O-C1 joint widening, with a CCI of 7 to 20 mm (Fig. 224-26), and 2 patients had unilateral joint widening who qualified for the previously unrecognized entity of unilateral AOD (Fig. 224-27). Gross left-right asymmetry was seen in 7 patients, including complete bilateral joint dissociation and lateral shifting of both condyles on C1 (Fig. 224-28) and wide swinging of one condyle off the ipsilateral C1 facet pivoting on the opposite joint (Fig. 224-29). The most revealing findings were three cases in which all four “standard” published tests (Wholey’s, Powers’, Harris’ and Sun’s criteria) gave manifestly normal readings yet the O-C1 joints on both sides were obviously disrupted on reformatted CT and measured emphatically positive with the CCI criterion (Fig. 224-30).

FIGURE 224-26 Symmetrical severe widening of the condyle-C1 interval (CCI) in atlanto-occipital dislocation. A coronal reformatted CT scan (A) shows a bilateral symmetrically widened CCI (arrows). The left CCI is 12 mm; the right CCI is 10 mm. Right sagittal (B) and left sagittal (C) reformatted CT scans show that both condyles are sprung away and anteriorly displaced from the C1 lateral masses. A midsagittal image (D) gives a Wholey dens-basion interval of 20 mm, a Powers’ ratio of 1.2, and a Harris basion-axial interval (BAI) of 16 mm, all abnormal. Sun’s ratio of 2.5 is normal. A, midpoint of the anterior arch of C1; B, basion; C, midpoint of the posterior arch of C1; D, dens; O, opisthion; PAL, posterior axial line.

FIGURE 224-27 Unilateral atlanto-occipital dislocation in a 17-year-old with right hemiparesis. A, Sagittal CT through the left occiput-C1 (O-C1) joint (arrows) shows a left condyle-C1 interval (CCI) of 3 mm (normal). B, Sagittal CT through the right O-C1 joint (arrows) shows a right CCI of 6.9 mm, and the condyle is slightly anteriorly dislocated. C, Coronal CT showing a narrow left CCI (arrows). D, Coronal CT showing a widened right CCI (arrows). The slightly oblique orientation of the coronal reconstruction explains why both O-C1 joints are not on a single coronal section. E, Midsagittal CT showing normal Wholey’s (11 mm), Harris’ (10 mm), and Sun’s (1.2) criteria, but a high Powers’ ratio of 1.1. A, midpoint of the anterior arch of C1; B, basion; BAI, basion-axial interval; C, midpoint of the posterior arch of C1; D, dens; O, opisthion; PAL, posterior axial line.

FIGURE 224-28 Lateral shifting of the condyles on C1 in a 13-year-old with incomplete C1 myelopathy. A, Coronal CT showing a lateral shift of both condyles toward the left, indicative of complete bilateral occiput-C1 dissociation. Note the chip fracture (arrow) of the upper right C1 facet. B, Right sagittal CT showing a right condyle-C1 interval (CCI) of 6.5 mm and a C1 facet fracture. C, Left sagittal CT showing a left CCI of 9 mm (arrows). The left condyle is also anteriorly displaced on C1. D, Midsagittal CT showing a Wholey dens-basion interval of 19 mm, a Powers’ ratio of 1.33, a Harris BAI of 10.5 mm (normal), and a Sun ratio of 3.2. A, midpoint of the anterior arch of C1; B, basion; BAI, basion-axial interval; C, midpoint of the posterior arch of C1; D, dens; O, opisthion; PAL, posterior axial line.

FIGURE 224-29 Conformational asymmetry in the occiput-C1 (O-C1) joints in a patient with atlanto-occipital dislocation. A, Coronal CT showing a bilateral widened condyle-C1 interval (CCI), but the left O-C1 joint has very little contact surface (arrows) because the left condyle has swung forward from the C1 facet. B, Sagittal images of both joints show the considerable anterior displacement of the left condyle off C1 (arrows) such that hardly any overlapping surfaces are available for CCI measurements. The right joint, although widened, still has fully facing surfaces. C, Midsagittal CT showing an increased Harris BAI (13 mm) but a normal Wholey dens-basion interval (9.5 mm), Powers’ ratio (1), and Sun ratio (1.9). A, midpoint of the anterior arch of C1; B, basion; BAI, basion-axial interval; C, midpoint of the posterior arch of C1; D, dens; O, opisthion; PAL, posterior axial line.

FIGURE 224-30 Bilateral atlanto-occipital dislocation with all four negative “standard tests” in a 6-year-old girl with incomplete C2 quadriparesis. A, A midsagittal CT scan shows all four tests to be negative: a Wholey dens-basion interval of 11 mm, Powers ratio of 1; Harris BAI of 11 mm, and Sun ratio of 1.2. A, midpoint of the anterior arch of C1; B, basion; BAI, basion-axial interval; C, midpoint of the posterior arch of C1; D, dens; O, opisthion; PAL, posterior axial line. B, Coronal CT showing a bilateral widened condyle-C1 interval (CCI) (arrows). C, Sagittal CT showing a right CCI of 9.6 mm and a left CCI of 7.6 mm. D, Sagittal T2-weighted MRI shows the tectorial membrane lifted off the clivus (thick arrow), or it may also have been torn (thin arrow).

These unexpected findings prompted us to perform a detailed comparison of the diagnostic sensitivity of CCI against the four “standard tests” and other AOD-associated signs such as tectorial membrane damage, perimedullary subarachnoid hemorrhage (SAH), C1-2 epidural blood, and neurological signs of injury to the craniovertebral junction (Fig. 224-31). We found that the CCI was by far the most sensitive (100% sensitivity) index for AOD, with the others, including tectorial membrane tear, all below 70% and all of the “standard tests” trailing even further behind. A test of CCI’s ability to differentiate AOD from non-AOD injuries by using data from five categories of non-AOD upper cervical injuries also shows CCI to have by far the highest specificity among all the other tests.¹²² There are several explanations for CCI’s supremacy:

1 Unlike all other tests, CCI directly measures the integrity of the injured O-C1 joints. Because the normal O-C1 joint is held together tightly by ligaments, with a very narrow joint gap,¹⁴⁹ sudden rupture of these ligaments instantly releases this tight hold similar to unfastening a compressed coil, and the joint surfaces spring apart with a widened gap. Because the normal myoligamentous hold on the joint is “active” or tonic, it would be impossible for the joint surfaces to resume tight apposition once the ligaments are disrupted. Thus, a widened CCI can only be made wider but not narrower by postinjury shifting of the head and neck, unlike tests that use remote landmarks whose mutual relationships may change and rechange with patient handling and are therefore susceptible to producing false-negative readings.

2 By converse reasoning, any injury not directly affecting the stability of the O-C1 joint will not cause a widened CCI and give a false-positive reading.

3 It is the only test that uses both sagittal and coronal images and can thus capture all planes of displacement.

4 It involves adjacent and coplanar reference points that are unaffected by rotational and other distortions.¹⁴⁹

5 It is also unaffected by congenital anomalies of the C1 arches, odontoid, clivus, and foramen magnum.

6 Because the normal CCI does not change significantly with age from birth to 18 years,¹⁴⁹ the validity of the test is not limited by age, unlike tests that are affected by osseous maturation of adjacent structures.^{111,122,134,146–148}

FIGURE 224-31 Bar graphs displaying the sensitivity and specificity of the condyle-C1 interval (CCI) and other diagnostic criteria for atlanto-occipital dislocation (AOD), as well as relevant head CT, MRI, and clinical findings. The upper graph concerns only AOD patients. The pink bars are correct positive readings, thus representing sensitivity percentiles. The purple bars are false-negative readings. The lower graph concerns non-AOD upper cervical injuries. The yellow bars are correct negative readings (for AOD), thus representing specificity percentiles for diagnosing AOD. The brown bars are false-positive readings (for AOD). CCI is the only test with absolute sensitivity and specificity. Sun’s ratio has the lowest sensitivity and specificity among the “standard tests.” C1-2 myelopathy, tectorial membrane damage, and cervicomedullary blood all have high sensitivity but low specificity. SAH, subarachnoid hemorrhage.

Diagnostic Paradigm for Atlanto-occipital Dislocation

The diagnosis of AOD begins with heightened suspicion in a patient with high cervical myelopathy or brainstem deficits (or both) after having suffered high-energy trauma. Perimedullary SAH or C1-2 epidural and subdural blood ^122,137 on the head CT scan greatly strengthens the likelihood.^{123,150–153} Lateral cervical radiography is seldom definitively diagnostic but is helpful if O-C1 distraction is obvious or if there is retropharyngeal edema, which is common in, although not specific to AOD.^{123,137,138,154–159} Plain radiographs are also important to rule out concomitant subaxial fracture-dislocations.

The presence of any of the aforementioned three warning signs should compel the next level of investigation with thin-cut CT of the upper cervical spine and MRI. An abnormal CCI and conformational disfigurement of the O-C1 joints should be the primary concern on CT, but the lesser “standard tests” should also be performed for corroborative evidence. MRI accurately delineates the extent of ligament damage, as well as the severity of the spinal cord and brainstem injuries, which is of decided prognostic value for functional recovery⁷⁵ (Fig. 224-32).

FIGURE 224-32 Diagnostic paradigm for atlanto-occipital dislocation (AOD). Suspicion is generated by the neurological lesion, perimedullary SAH and C1 extra-axial blood, or suggestive lateral cervical spine radiograph. CT and MRI to evaluate the condyle-C1 interval (CCI) and other criteria should make the diagnosis readily. SAH, subarachnoid hemorrhage.

In clinical practice, four measurements of the O-C1 joint gap are often not necessary or even feasible to make the diagnosis of AOD. In a widely separated joint, one or two measurements should suffice. If any portion of a joint is wider than 7 mm, it is irrelevant whether any other portion of that joint is more closely apposed. If the joint is so widely dislocated horizontally that it has little or no overlapping surfaces, atlanto-occipital instability can be safely assumed without an actual CCI measurement.¹²³ If the CCIs for both sides are between 3 and 4 mm and there is neither asymmetry nor dislocation but clinical suspicion for AOD is strong, a very careful traction stress test with incremental weights may be performed under fluoroscopy to see whether the CCI widens,¹⁵⁴ preferably with the patient awake and under electrophysiologic monitoring.

Management of Atlanto-occipital Dislocation

Once the suspicion of AOD is entertained, however tentatively, traction should immediately be avoided. Traction has been known to worsen neurological deficits and increase O-C1 distraction.^152,154,160 Preliminary stabilization can be maintained with a stiff cervical collar and side sand bags while obtaining CT and MRI scans. Excessive bending at the craniovertebral junction must be avoided by carefully placing padding under the shoulders and back to eliminate the torque effect of the prominent occiput.¹⁶¹ After AOD is confirmed by CT and MRI, a halo vest should be fitted to maintain temporary stability while other life-threatening injuries can be dealt with. Because AOD is such an unstable injury and patients are so susceptible to early deterioration, some form of definitive immobilization should be established as soon as it is medically tenable.

Although a few successful cases have been reported with nonsurgical management,^{111,130,162,163} halo immobilization alone is generally acknowledged to be inadequate for AOD. This conclusion is consistent with the well-known fact that pure ligamentous injuries heal poorly with immobilization.^105,164 Moreover, the craniovertebral junction is the least secure site with the halo, and acute slippage of O-C1 alignment has frequently been reported while in the halo.^{47,103,120,129,165,166} Examples of chronic recurrent O-C1 dislocation causing delayed brainstem compression are known even when initial healing had ostensibly taken place with halo immobilization.^{109,138,160,167,168}

We recommend internal fixation of the occiput to C2 and, in special cases, to C3. The type of O-C1-C2 fixation depends on the patient’s age and the corresponding anatomy of the region relevant to screw purchase, including the thickness of the occipital bone, the size and strength of the lateral mass of C1 and the pars interarticularis (isthmus) of C2, and the adjacency of the vertebral artery to the projected screw paths. Anatomy permitting, screw-plate fixation is biomechanically superior to wire loop techniques whether in providing immediate stability, in resisting loosening of the construct, in preventing vertical settling of the transfixed segments,^169,170 or in reducing the number of segments needed to be incorporated into the fusion. Its more solid fixation often dispenses with the postoperative halo, a notable advantage in view of the potential hazards with the halo in children.^51,171

Thus, in older children we prefer occipital screw-plate fixation using a C1-2 transarticular screw,^172,173 a C2 isthmus screw,¹⁷⁴ or the adapted Harms and Melcher technique of a C1 lateral mass screw and C2 isthmus screws.¹⁷⁵ Although we usually select children older than 8 years for intra-axial screws, the age limit should be flexible to suit the morphology of the prospective screw paths, which should be preplanned with triplane CT reconstruction through the C2 isthmus and C1 lateral mass. In children, the thickest occipital bone buttress for screw anchorage is usually the midline keel and sometimes near the base of the mastoid.

In young children we still prefer sublaminar wiring,^{123,176–178} and if the C1 and C2 laminae are exceedingly delicate, as they are in children younger than 3 years, the fusion is extended to C3 or even to C4 to distribute the stress on the uprights, with minimal additional loss of combined segmental motion.^51,177 When the occipital squama is too thin for regular screw purchase, we have recently used the “inside-outside” screw technique described by Pait and colleagues¹⁷⁹ (see later). A postoperative halo is necessary if sublaminar wiring is used.

Neurological Outcome

AOD has historically been associated with a very poor outcome. Hamilton and Myles found no survivors after AOD in their review of pediatric spinal injuries.^15,16 Dickman and coauthors reported 10 deaths in their series of 14 patients, with only 3 patients showing improvement.¹⁵⁴ High mortality and severe morbidity are similarly reported by Imaizumi and associates¹⁵² and by Saeheng and Phuenpathom,¹⁸⁰ whereas the SCI guidelines published jointly by the American Association of Neurological Surgeons and the Congress of Neurological Surgeons¹⁵⁸ in 2002 documented poor neurological outcome in 42% and death in 5%. There is no doubt that the prognosis of patients with AOD depends on the severity of the initial neurological injury, in this regard not unlike other categories of pediatric SCI.⁷⁵ Patients with complete high cord transection or severe head injuries have a very poor outcome,^{49,114,119,134} and good recovery usually occurs in those with partial or mild initial neurological syndromes.*

Our series documented early and delayed deaths in 19%, complete or severe residual quadriplegia in 12%, but excellent neurological recovery in 69%.¹²³ Overall, our results are considerably better than the historical data. Comparison of the initial and final neurological states in our patients, however, shows no disagreement with the literature for those with initially complete quadriplegia, who tend to remain profoundly disabled. Ventilator-dependent, high-quadriplegic children generally do poorly despite aggressive measures to improve respiratory function, but even in patients with incomplete but severe initial injuries, functional improvements are unfortunately limited.

The real difference that gives our series an overall better outcome in fact to a larger proportion of the AOD patients initially admitted with mild to moderate neurological injuries. Our aggressive and expectant search for this diagnosis, using a sensitive and specific diagnostic test that enabled us to screen every traumatized child fitting the clinical and mechanistic profiles of AOD, probably succeeded in making early diagnoses and thus avoided secondary injury to the cord.

Functional Anatomy of the Atlantoaxial Joint

A unique feature of the atlantoaxial articulation is that the facets are oriented more horizontally than all the other facet joints of the cervical spine, which have much steeper angles. This nearly (but not completely, an important distinction) horizontal orientation of the C1-2 articulation enables the atlas to rotate on the axis with minimal bony impediment, at least within certain limits. Furthermore, even though the osseous contour of the atlantoaxial joint looks concave on radiography, the padding of the joint cartilage converts the actual articulating surfaces to biconvex disks. Rotation of the atlas pivots about the odontoid process, which acts as an upright post to which most of the stabilizing ligaments within the whole occiput-atlas-axis complex are either attached or intimately related. Of these, the alar ligaments function as the check ligaments against excessive rotation of the atlas on the axis. Traditional teaching asserts that the alar ligaments connect the dorsolateral surfaces of the tip of the dens to the occipital condyles, but there is in fact an important extension of the alar ligament that binds the lateral mass of C1 to the odontoid,²⁰⁸ which may explain more fully the yoking between C1 and C2 in rotation. The thickened capsular ligament aids in restricting rotation and flexion of the atlantoaxial facet joints, but the rudimentary (in humans) apical ligament, the atlantodental ligament, the anterior and posterior atlanto-occipital membranes, and the atlantoaxial membrane all serve minor stabilizing functions.

Motion Analysis and Normal Motion Curve

To study the normal dynamics of C1-2 rotation, thin-cut nonoverlapping axial CT scans were used to determine the C1 and C2 angles in relation to the vertical axis (0 degrees) during a full rotational sweep of the head.¹⁸³ The separation angle between C1 and C2, the C1-2 angle, is the algebraic difference between the two angles, or C1 − C2. Seven to eight sets of C1 angles and C1-2 angles are obtained with the head turned to different positions from 0 degrees to one side and then turned fully to the opposite side. Plotting the C1 angle on the x-axis, which denotes the head positions, against the C1-2 angle on the y-axis, which represents the C1 − C2 separation angles, into a motion curve thus displays the instantaneous relationship between C1 and C2 in all head positions during a full head swing; that is, the curve accurately defines the rational dynamics between C1 and C2 for the entire physiologic range of head rotation (to both sides). Combining the individual motion curves of multiple subjects into a composite curve by using the sixth-degree “best-fit” polynomial function on the total pool of C1 and C1-2 angles therefore formulates an accurate mathematical and graphic prediction of the entire spectrum of C1 and C2 relationships during normal axial rotation¹⁸³ (Fig. 224-34).

FIGURE 224-34 The composite motion curve of C1-2 rotation shows that the crossover point between C1 and C2 with turning from one side to the other occurs very nearly at vertical 0 degrees and that the motion curves to each side of midline are roughly mirror images of each other. The three regions of the curve are separated by the arrows. From C1 = 0 degrees to the solid arrow (C1 = 23 degrees), the relationship is linear with a gradient of 1 when only C1 moves (the single-motion phase). Between the two arrows the relationship is nonlinear when both C1 and C2 are moving at different rates (the double-motion phase). From the open arrow to C1 = 90 degrees, C1 and C2 move as a couple at the maximum separation angle (unison-motion phase).

Figure 224-34 depicts the composite curve of 21 children. It very nearly passes through the zero point, which means that the crossover between C1 and C2 when turning from one side to the other occurs with the head (C1) almost exactly at the vertical 0-degree position (i.e., with the nose straight forward). (Below 0 degrees on the y-axis, the C1-2 angle becomes negative, which means that C1 is on the other side of C2.) This indicates that the vertical 0-degree position is the null point of normal rotation. In addition, the portion of the curve within one quadrant describing motion on one side appears to be the exact mirror image of the curve in the diagonal quadrant depicting motion to the opposite side.

Normal and Abnormal Dynamics of C1-2 Rotation

Based on our motion analysis data,¹⁸³ the precise events during C1-2 rotation can now be defined in the context of the functional anatomy of the O-C1-C2 joints. There are three distinct regions on the composite motion curve that reflect three distinct phases of C1-2 rotation. With initial head rotation, say to the right, C1 is “cranked” rightward by the tight bony structures of the roller-in-groove joint between its upper facet and the occipital condyle. In contrast, the lax capsular ligaments and the oblique orientation of the alar ligaments allow C2 to be unperturbed for the first 23 degrees or so of head movement. During this first, or single-motion, phase of rotation when C2 remains immobile and only C1 moves, the C1 and C1-2 angles are identical and the curve is perfectly linear with a gradient of exactly 1 (between 0 degrees and the solid arrow in Fig. 224-34). Beyond 23 degrees, the left alar ligament begins to tighten and pulls the odontoid process in the same direction as C1, thus commencing the second, or double-motion, phase of rotation (when both C1 and C2 are moving), in which the C1-2 angle is always less than the C1 angle. The curve begins to arc downward. With continued head turning, the ligaments become progressively more taut. Further C1-2 separation is increasingly impeded, and C2 is pulled more forcefully and faster toward C1 such that even though the C1-2 angle continues to widen, its rate of change is progressively less with each additional degree of C1 rotation. The gradient of the motion curve accordingly becomes less steep as the arc of the curve flattens to a plateau (between the arrows in Fig. 224-34). As the C1 angle approaches 65 degrees, the ligaments and capsular membranes become maximally stretched and the bony contours in the joint surfaces of the axis and atlas probably also reach their contact limits; further widening between C1 and C2 ceases, thus commencing the plateau, or unison-motion, phase of the curve (see Fig. 224-34). The two bones now turn as a couple with a fixed separation angle of about 43 degrees as the head continues to rotate as far as 90 degrees from the vertical 0-degree position. Head turning from 65 degrees onward is carried out by subaxial motion only.

In AARF, the defining pathologic state is therefore not any absolute angle of separation between C1 and C2 but the abnormal way in which C1 and C2 interrelate during axial rotation.²⁰⁶ It is the fluidity of motion and the degree of freedom in individual movements that characterize the severity of the condition. For example, the normal crossover null point is at or very near 0 degrees.¹⁸³ If on counter-rotation C1 crosses C2 at −15 or −20 degrees from zero on the side opposite the patient’s initial head position, pathologic “stickiness” exists between C1 and C2. Likewise, if on reverse rotation the C1-2 angle narrows much more slowly than normal so that the motion curve is significantly shifted to the left, pathologic stickiness again exists. Certainly, if crossover does not occur at all or, in the extreme case, when the C1-2 angle remains undiminished on forceful counter-rotation, a functional “lock” between the bones has taken place.²⁰⁶

Normal C1-2 rotation does not occur solely and independently in the horizontal plane; “coupling movements” of other types also occur simultaneously. For instance, as the atlas rotates to the left, the orientation of the plane of rotation remains parallel to the horizon only for the first 20 to 25 degrees. Beyond 25 degrees, the right alar ligament begins to tighten, which serves not only to impede further rotation but, by virtue of its expansion to the C1 lateral mass, also to cause the right lateral mass of the atlas to telescope slightly forward and downward. This results in a slight head tilt to the right (completing the full cock robin position). Because the actual articulating surfaces between C1 and C2 are in fact on a slant in the coronal plane and not horizontal, the initial horizontal rotation must at some point encounter resistance. This tilting converts the plane of rotation of the right joint from horizontal to oblique, in better conformity to the angle of the joint surfaces, and this accordingly allows further rotation from 25 degrees to occur up to the expected 43 degrees. It must then follow that from 25 degrees leftward, the primary stress of the rotation is inflicted mainly on the right joint and any bony lock or capsular tear would probably occur in this joint, usually with the lateral mass of C1 popping forward and overhanging C2.¹⁸³

Diagnostic Motion Study and Classification of Atlantoaxial Rotatory Fixation

In patients with painful torticollis who have the classic cock robin appearance, three sets of CT scans are used to study their C1-2 rotatory dynamics: (1) with the head in the initial position, designated the P position; (2) with the head turned to or close to the zero position by the examiner, designated the P₀ position; and (3) with the head turned to the side opposite the initial position as far as the patient can tolerate, designated the P₋ (“corrected”) position. The corresponding C1 and C1-2 angles are obtained for the three positions. In cases in which the child’s neck is extremely painful and rigid and the P₋ position is reached before the nose could line up with the vertical zero position, a third set of scans is performed with the head turned back to approximately halfway between the P and P₋ positions. This third set is then considered the P₀ scan. Mild sedation is used if anxiety and discomfort prevent an optimal study.

By convention, the side toward which the chin is initially pointing is assigned the positive sign. Thus, all C1 angles at the P position are positive, and all angles on the opposite side are negative. The diagnostic motion curve is constructed as for the normal motion curve, except that the three data points from the P, P₀, P₋ scan sequences are simply joined with straight lines. The diagnostic curve is analyzed in the context of normal C1-2 rotation by superimposing it on the physiologic composite motion curve, which is made into a normal template by adding 8 degrees to each side of the mean to account for the variance in distribution in younger subjects with physiologic hypermobile joints.²⁰⁶

Three types of motion curves show flagrant departures from physiologic rotation and can be considered representative of three AARF types.²⁰⁶ The antecedent trauma had evidently resulted in pathologic stickiness that prevents closure of the existing C1 − C2 separation angle by manual counter-rotation. In type I AARF patients, the reduction in the C1-2 angle at the maximum tolerated correction (at P₋) is minimal (less than 20% of the C1-2 angle at P); the C1 − C2 separation angles are virtually unchanged even when the head is forcibly rotated way past the midline (Fig. 224-35). All type I motion curves are thus almost straight lines across the right and left upper quadrants of the normal template (Fig. 224-36). Type I AARF represents the abnormal spectrum’s most extreme form, with the two bones locked in the coupled configuration.²⁰⁶

FIGURE 224-35 Type I atlantoaxial rotatory fixation (AARF). A, Simulated C1-2 relationship in type I AARF (bony lock). Top, At initial evaluation (P), C1 is to the right of C2 and separated from it by 45 degrees. Bottom, At a forced corrective position (P₋), C1 has been cranked to the opposite side but is still to the right of C2 with minimal reduction in the separation angle (C1 − C2). B, Three-position CT sequences of a patient with type I AARF. At initial evaluation (P), C1 is to the left of C2 and separated by 42 degrees. At maximum correction, C1 could not be pushed past the midline but rather is still to the left of C2, with minimal reduction in the C1-2 angle (4 degrees, 9.5%), and is maintaining the same coupled configuration. P₀ scans were performed with the head turned back halfway between the P and P₋ positions.

FIGURE 224-36 The motion curves of five patients superimposed on the normal template show type I atlantoaxial rotatory fixation. Each curve is constructed by plotting the C1 angle against the corresponding C1-2 angle in the three test positions. Because C1 is normally tightly coupled to the head, the curve describes the exact relationship (separation angle) between C1 and C2 in the three head positions, thereby revealing how “easily” C1 moves with C2 during counter-rotation. Note that all five curves are practically horizontal lines projecting like cantilevers from the right to the left upper quadrants. Also note that patients 4 and 5 never had their heads (C1) pushed past midline. There is essentially no reduction in the C1-2 angle and hence no change in the coupled configuration of C1-2 with maximum correction.

In type II patients, the maximum reduction in the C1-2 angle is greater than 20% of the initial C1-2 angle (at P₀), but C1 cannot be made to cross over C2. Their motion curves slope downward from right to left but never traverse the x-axis, where C1-2 crossover normally takes place (Fig. 224-37). Type II AARF patient thus have slightly less severe stickiness than do type I patients in that the initial separation angle is reducible but C1 could not be forced to cross C2.²⁰⁶

FIGURE 224-37 The motion curves of seven patients superimposed on the normal template show type II atlantoaxial rotatory fixation. Note that all seven curves slope downward from the right to the left upper quadrants, indicative of a significant reduction in the separation angles; however, none traverse the x-axis, consistent with no crossover between C1 and C2 with maximum correction.

In type III patients, the C1-2 angle narrows with correction and C1 can be forced to cross C2 (all C1-2 angles at P₋ are negative), but always at very abnormal head positions. Their motion curves traverse the x-axis at points where the C1 angle is less than −20 degrees, far left of normal crossover near 0 degrees¹⁸³ (Fig. 224-38). Type III AARF patients therefore show less stickiness than do the other two types because C1 can be forced to cross C2, but only if the head (or C1) is cranked way past the expected null point at 0 degrees. ²⁰⁶

FIGURE 224-38 Motion curves of nine patients showing type III atlantoaxial rotatory fixation. All nine curves traverse the x-axis, indicative of C1-2 crossover, but at points on the negative side of C1 = −20 degrees, far from the normal crossover (null point) at 0 degrees. The resistance against closure of the C1-2 separation angle has cranked C2 far to the opposite side before crossover could occur.

Thus, the degree of pathologic stickiness in AARF can be graded according to the shape of the motion curve: the more horizontal the curve, the more locked are the two bones. For instant diagnosis and classification of AARF, the user superimposes a patient’s motion curve on the diagnostic domain diagram (Fig. 224-39), whose three diagnostic domains correspond roughly to the areas on the template occupied by the three respective types of curves.

FIGURE 224-39 Rough diagnostic domains of the three types of atlantoaxial rotatory fixation (AARF). The boundaries of the various domains are not set in real numbers, except that C1 = −20 degrees for type III, because there is some variability and overlap among the three domains. For example, type II and type III curves will traverse part of the same domain in the upper quadrants; the two are separable only at the x-axis. This figure serves only to remind clinicians of the general shapes of the three types of AARF curves. DGZ, diagnostic gray zone.

To gauge the progress of treatment in a particular patient, the postreduction motion curve can be incorporated on the same template bearing the diagnostic curve; the pretreatment and post-treatment curves on the same template thus provide a quantitative evaluation of the effect of treatment that not infrequently surpasses the bedside impression. The clinician can now select further action based on the altered (or unaltered) dynamics of the C1-2 complex.

Muscular Torticollis and Diagnostic Gray Zone

Patients with muscular torticollis, despite a painful stiff neck, have no actual injuries to the C1-2 joints and therefore manifest physiologic rotational behavior and have motion curves within the normal range. ²⁰⁶

Some patients have stickier C1-2 rotation than normal, and their motion curves cross the x-axis at the ambiguous zone of −8 to −20 degrees yet do not fulfill our AARF criteria²⁰⁶ (Fig. 224-40). They are designated as being in the diagnostic gray zone, which places them in a “wait-and-see” category regarding rendering of treatment. They are given a cervical collar and analgesics and then reimaged in 10 to 14 days. Many “gray” patients display normal curves on restudy, but some will stay symptomatic with marginal dynamics or even progress to type III AARF. The gray zone thus serves as a therapeutic buffer to avoid overtreatment.

FIGURE 224-40 Motion curves of five patients showing the C1-2 relationship in the diagnostic gray zone. All five curves show some left shift from the normal range, and the C1-2 crossover points are between C1 = −8 and −20 degrees. The motion dynamics shows stickier rotation than normal, but not quite qualifying as type III atlantoaxial rotatory fixation by our definition.

The diagnostic paradigm for AARF based on our imaging criteria is shown in Figure 224-41.

FIGURE 224-41 Diagnostic paradigm of atlantoaxial rotatory fixation starting with a child with a cock robin neck and using the three-position CT protocol. C₁C_2P−° is the C1-2 separation angle at the P₋ position. C1-2 angle reduction is in percentile of the C1-2 angle at the initial (P) position.

Consequences of Untreated Chronic Atlantoaxial Rotatory Fixation

Neurological deficits are uncommon in patients with acute AARF. Vertebral artery thrombosis secondary to stretch injury may cause intractable vomiting and vertigo or even brainstem deficits, but this is very rare.²⁰⁹ Chronic untreated AARF, however, frequently leads to generalization of painful spasm to both sternomastoids and the posterior nuchal muscles, thus accentuating the fixed lateral head tilt. This severely distorts the head and neck relationship and, together with the cessation of C1-2 rotation, may theoretically accelerate degenerative changes in the subaxial cervical spine.^210,211 The facial asymmetry in chronic cases is thought to be due to unrelenting and asymmetric muscle pull on the mandible and other growing maxillofacial structures.^210,212 This is at best a cosmetic issue but at worst can lead to disturbed dentition and malocclusion.

Three other changes with chronicity are potentially more sinister²⁰⁶:

Management of Atlantoaxial Rotatory Fixation

Management Outcome

In acute patients, reduction is generally successful with halter traction regardless of the AARF type, but acute type I and type II patients are more likely than acute type III patients to show slippage. Subacute type III patients fared only slightly worse in the number of slippages, the total duration of treatment, and the requirement for halo fixation and fusion. Only chronic type III patients have hopes of preserving normal dynamics. Chronic type I and II patients almost never achieve reduction or stay in reduction despite repeated efforts, and most will require fusion.

Figure 224-42 compares the course and results of treatment of the three types of AARF.²¹⁴ Type I AARF patients surpass type II and III patients in every category of management difficulty and have a much poorer prospect for preservation of physiologic rotation. In turn, type II patients also exceed type III patients in the same categories. The differences between types are particularly marked in the proclivity to repeat slippage, the need for multiple invasive procedures, the duration of treatment, and the likelihood of permanent loss of rotation.

FIGURE 224-42 Comparison of the course and result of treatment of the three types of atlantoaxial rotatory fixation (AARF). Parameters on the x-axis are chosen to reflect the complexity and invasiveness of the treatment procedures. Outcome is measured by the loss of normal rotatory dynamics documented by the last motion curve performed after all treatment steps have concluded (last item on the right of the x-axis). The y-axis on the left denotes percentiles of patients irreducible by traction, needing a halo or fusion, or having lost normal C1-2 rotation. The y-axis on the right denotes the number of treatment procedures, reduction in slippage, weeks of traction, or months of total treatment duration. The “sense” of the numerical value of the y-axes is designed to equate a high number or percentile (i.e., tall point on the graph) with a poor outcome. Type I patients fare worst, type III fare best, and type II patients are in between. Thus, there is a strong correlation between the complexity of management and poor outcome with the severity of the AARF.

Figure 224-43 analyzes the effect of the chronicity of AARF on the course and outcome of treatment. Chronic AARF patients, much more so than acute or even subacute patients, are likely to face a long traction time, suffer repeated slippage, have a greater than 50% rate of nonreduction and fusion, and would very likely end up with absent or permanently restricted rotation at C1 and C2.

FIGURE 224-43 Comparison of the course and result of treatment between acute, subacute, and chronic atlantoaxial rotatory fixation. The x- and y-axes are identical to those in Figure 224-42. Chronic patients did much more poorly than acute patients in every category. Seventy percent of chronic patients lost normal C1-2 rotatory function through either surgical fusion or permanently impaired dynamics versus only 8% of acute patients. Subacute patients have an intermediate course, but more than 40% had loss of normal C1-2 rotation, thus faring almost as poorly as chronic patients.

Combining the effects of chronicity and severity of AARF, the best and worst outcome classes are undoubtedly chronic type I and acute type III patients, respectively.²¹⁴

The recommended management algorithm for AARF is summarized in Figure 224-44.

FIGURE 224-44 Management algorithm for atlantoaxial rotatory fixation (AARF). Identification of the AARF type and determination of pretreatment delay (chronicity) are important requisites for the selection of treatment. Bracing and halo immobilization are maintained for 3 to 4 months. The occiput is not included in the fusion.

Pathogenesis of Atlantoaxial Rotatory Fixation

Any explanation for the pathologic stickiness of AARF must account for the following observations: the severity of the stickiness ranges from a delayed crossover of C1 on C2 to complete bony interlock, the “fixation” is noted only on counter-rotation away from the pointing chin and not toward it, no bony injury in the joint is ever found on radiography, traction reduces some but not all cases, and delay of treatment past 3 months makes reduction much more difficult.

In acute AARF, the stickiness is probably caused by some type of interposing soft tissue within the joint that prevents reverse rotation. This may be a piece of torn joint cartilage^186,209 or infolding synovium.²⁰³ A flap of ripped capsule may be “nipped” by the joint surfaces¹⁸³ and result in a block, or a corner of the C1 facet could have protruded through a capsular rent and is trapped by the “buttonhole” effect. The anteriorly riding C1 facet regularly loses 85% of its contact surface with C2 at end-rotation.²¹⁵ During a particularly violent head swing, the C1 facet could have gone completely beyond the contact points with C2 and its inferior lip now jumps over and becomes tightly locked against the front of the C2 facet.^203,205,210 C1and C2 will now move ineluctably as a couple on forced counter-rotation (type I AARF), but further rotation in the direction of the displacement could still be possible, more so because the restrictive ligaments and capsules have now been torn or sprained. Traction could disengage some C1-2 interlocks and pull infolding tissues out of joints, but not disimpact a “frozen” C1-2 lock, “undo” the buttonholed facet, or dispel a snugly wedged-in fragment of cartilage.

In long-neglected rotatory fixation, permanent changes occur that would explain the difficulties in reducing chronic AARF. Contractures of joint capsules and periarticular ligaments and tendons,¹⁸⁶ osteoarthritis,²¹⁰ fibrous adhesion bands,^216,217 and even osseous cross-unions^210,218 between the articulating surfaces have been found during open surgery.

Halo Fixation Device

Among all the available cervical orthoses, the halo vest provides the most rigid immobilization of the cervical spine in children.^{51,94,228,246–248} It also permits the most precise positioning of the neck to achieve and maintain alignment. It is the only external orthosis that does not have a chin support, so it interferes the least with mandibular motion and eating.²⁴⁹ Early mobilization enables these children to participate in rehabilitation.^250,251 In our experience, children tend to sabotage the immobilization if they are not allowed some physical activity. However, the rate of complications associated with the halo is also higher in children than in adults,²⁵⁰ and most are related to the thinness of the child’s calvaria. The anterior pins should be placed immediately above the eyebrows to take advantage of the thick lateral supraorbital buttress²⁵⁰ and laterally to avoid injuring the supraorbital and supratrochlear nerves. The torque used for pin tightening should be between 2 and 4 inch-lb, depending on the size of the child. For children 10 months to 3 years of age, “tiny tot” haloes are available with a small ring and up to 10 small pins that should be only finger-tightened.²⁵² However, one great disadvantage of the multipin system is that it leaves few unused holes on the ring for pin changes in the event of pin site infection or loosening (see later).

The frequency of pin loosening is much higher in children than in adults.^250,253 Some pins migrate cephalad because they were inserted above the skull equator, where the contour of the skull is upward sloping. Pins may also loosen because the correct degree of tightness was not rechecked. As a result of the viscoelastic properties of the skull, the pins frequently loosen up to 4 inch-lb after the vest and uprights are attached to the ring, and they must be retightened to the correct torque 24 hours later. If a loose pin is discovered 2 to 3 weeks after insertion, management depends on the age of the patient. Garfin and colleagues found that with 6 inch-lb of torque, the halo pin only partially penetrates the outer table of the adult skull and there is usually a solid margin of cortical bone to allow safe retightening.²⁵³ Thus, it is usually safe to retighten loose pins in older children, provided that definite resistance can be felt on tightening. The skull tables in young children are much thinner. Accelerated osteoclasis around the pin site softens the bone so much that retightening can cause full-thickness penetration and an intracranial abscess.²⁵⁴ Loose pins should be replaced and not tightened in these children.⁵¹

Pin site infections are also much more common in children than in adults. Baum and coworkers reported a 31% incidence of pin infection in children versus 6% in adults.²⁵⁰ The pin sites should be cleansed thoroughly with 10% hydrogen peroxide solution 4 times daily, and all scabs and crust around the pins must be removed to allow drainage. If the infection is mild and superficial, with only minimal surrounding scalp erythema and drainage, and the pin remains tight, oral antibiotics are prescribed. If the scalp is red and swollen, the drainage becomes copious and purulent, or the child complains of constant local pain or headaches, the pin should be removed and intravenous antibiotics administered to prevent the dreaded cranial osteomyelitis. A CT scan should be performed to rule out an intracranial abscess.^254–257

Children in haloes also have a greater likelihood than adults to show slippage of alignment, partly because of unrecognized pin loosening and partly because children do not protect their haloes. Direct contact of the ring and posterior uprights with the bedding should be avoided to minimize transmitted flexion stress. We use a large foam wedge to prop up the back of the vest so that the child can recline comfortably with the head and halo ring hanging free and still be able to watch television, read, play board games, and self-feed.⁵¹ In addition, children sometimes undergo rapid and drastic changes in body weight and contour after SCI, so the original vest may become ill fitting and need to be recustomized.

Finally, children tend to be very unforgiving if the head position is not adjusted exactly to their liking. Because most unstable injuries are unstable in flexion, it is always tempting to set the neck in slight extension. However, mechanical dysphagia develops in children when the neck is even slightly extended because of restricted upward movement of the larynx during the pharyngeal phase of swallowing. Moreover, if the child’s visual axis cannot command a full frontal view of the surroundings when the head is tilted too far back, the child will simply stop walking. Thus, the neck should be put in the “military salute” position, in which posterior translation is substituted for extension. For young children, the top-heavy halo bars and vest in effect shift the center of gravity upward, and they must be patiently coached to relearn the basic skills of walking and navigating.

Minerva Cast

On rare occasion, the Minerva cast has been used on young children who do not tolerate the halo pins. The thermoplastic jacket is constructed of Polyform, a splinting material made from the polyester polycaprolactone, and Polycushion, a closed-cell foam for padding.²⁵⁸ This lightweight material solves the problems of hyperthermia and encumbered ambulation. Clinical trials have also found that the thermoplastic Minerva jacket actually restricts intersegmental “snaking” movements in the cervical spine better than the halo vest does. It deserves further evaluation for adaptation for use in children.

Guilford Brace

The Guilford brace has been used extensively in our institution on children with SCIWORA.^19,27,69,252 This brace has padded mandibular and occipital supports that interlock to form a rigid metal ring below the ears. The ring is connected by strong metal strips to anterior and posterior thoracic plates that are attached to each other with shoulder and axillary straps. The Guilford brace is comparable to other cervicothoracic braces in its ability to limit flexion, extension, and rotation, especially of the mid to lower cervical segments.^228,247 Like the other orthoses, the Guilford brace is not as effective in controlling lateral bending. Its main advantage over other braces is that it can be custom-made within 24 hours to fit any body size down to that of a 1-year-old. With deepening of the chin and occipital plates to restrict rotation, it is also suited for maintaining reduction in patients with AARF.

Molded Body Splint

There is no ideal orthosis for long-term immobilization of infants with grossly unstable necks. One example of this condition is diastrophic dwarfism, in which affected infants have ligamentous laxity at multiple segments and are usually seen in the first few months of life with apnea and quadriplegic spells because of multilevel subluxation and cervicomedullary injury. These neurological manifestations are completely reversible as long as the neck can be kept in a neutral or slightly extended position. Because no orthosis can reliably eliminate intersegmental motions in these infants, wiring and bone graft fusion often end in nonunion, and we usually do not attempt surgical fusion until these patients reach 12 to 18 months of age. In the interim, the infant is immobilized in a thermoplastic body and head splint molded to the contour of the thorax, head, and neck in the desired degree of extension.⁵¹ The body section incorporates the pelvis and hip joints and has a crotch strap that interlocks with cross-shoulder straps over the chest and abdomen, based on the same principle as an infant car seat. The head section has a padded headband that is strapped across the forehead to secure the head. The back slab can be fenestrated if hyperthermia develops, which happens not infrequently (Fig. 224-45). Diastrophic dwarfs are not usually ambulatory for many months after birth anyway, and with proper skin care, they can be kept safely in this molded body splint for up to 3 years (Fig. 224-46), at which time extensive surgical fusion has a better chance of success.

FIGURE 224-45 Thermoplastic body splint for long-term immobilization of the head and neck in young infants. A, Note the Velcro head and chest straps, fenestrations in the back slab for ventilation, and the flat shelf under the back section for more stable positioning on a bed or hard surface and to provide a secure hand grip for manual handling of the body splint. B, Body splint molded to the contour of the infant’s neck, head, and thorax. The head position is adjustable, usually to slight extension.

FIGURE 224-46 A, Lateral radiograph of an infant with diastrophic dwarfism showing multilevel severe anterior subluxation of the upper cervical spine. The infant had apnea in this position. B, Moderate reversal of the kyphotic curvature maintained by the molded body splint. The infant was asymptomatic in this position.

General Principles

As in adults with displaced fractures, it is always preferable to reduce any malalignment in children as soon as possible. Cord compression should be relieved promptly, and subsequent fusion can be rendered with minimal manipulation of the unstable segments. In older children, preoperative reduction can be accomplished with skull traction. In children younger than 5 years, reduction is best done in the operating room manually under fluoroscopic guidance and general anesthesia. The halo should be applied to maintain the reduction before surgical fusion is begun.

It is common practice at many SCI centers to intubate adult patients with unstable injuries while they are awake so that any sudden changes in the patient’s neurological status can be recognized. Awake intubation is seldom tolerated by children. Young children, in particular, tend to resist vigorously and may actually jeopardize the unstable injury further. We prefer to perform fiberoptic nasotracheal intubation on children after induction of general anesthesia with intravenous thiopental and vecuronium bromide. In this manner, the airway can be quickly and atraumatically secured with minimal head movement. Spinal cord function is monitored with real-time SSEPs throughout intubation, halo application, and the surgical fusion that follows.⁵¹

It is our impression and that of others that bone bank bone has less chance than autologous bone grafts of inducing successful fusion in children.²⁵⁹ The amount of available bone is obviously limited by the size of the child, and both iliac crests should be prepared for use. Other bone sources include ribs, fibula, and split-thickness cranial bone from the occiput.²⁶⁰

Permanent stability at the fusion site ultimately depends on the amount of new bone that manages to form across the fusion surfaces. The larger the fusion surface available for new bone formation, the more exuberant the callus and the stronger the subsequent bony bridge. Although methyl methacrylate incorporated into a wiring construct enhances the immediate stability of the construct,²⁶¹ it also diminishes the available fusion surface for new bone formation and ultimately weakens the permanent strength of the fusion.²⁶² This is a particularly relevant issue in children, in whom the fusion surface is already limited. Moreover, acrylic does not bind to bone, weakens with time, and remains a permanent foreign body.^263,264 Methyl methacrylate is therefore not used in children except when a malignancy is involved.

Continuous micromotion between fusion segments is strongly correlated with nonunion. As mentioned, halo immobilization in children is sometimes complicated by pin loosening and slippage in position. Non–halo-type orthoses provide even less satisfactory external stabilization.²²⁸ Thus, for children with unstable injuries, the burden of immediate stabilization rests largely on the immediate strength of the fusion construct; this is in contrast to adult patients, in whom the halo can be depended on to provide more reliable external immobilization. Thus, one should always select the construct that gives maximal immediate internal stability, that is, the one that is most effective in eliminating intersegmental motion. Fusion techniques for children should be judged on this important principle.

Injuries Involving C2-7

For unstable subaxial injuries in children 1 to 5 years of age, we favor the sublaminar wiring technique.⁵¹ Braided, MRI-compatible titanium Codman or Songer cables are used because their malleability, sliding quality, and crimp-lock mechanism enable safe passage and strenuous tightening. For very young children with delicate dura, Codman cables are preferred because of their soft pliability. Iliac crest bone grafts are then secondarily wired to the fusion surface or tightly pressed down against it by strapping sutures strung across the deep muscles. This extremely strong and stable construct ensures a high rate of union in children. Regardless of the age of the child, the laminae are still the strongest bone buttress to support any wiring. Even in children 1 to 3 years old, 18-gauge cables can be used without producing laminar fracture. Tightening the cables usually makes the two adjacent laminae crunch against each other to produce a more stable friction block than either the interfacetal or interspinous wiring techniques.²⁶¹

Posterior Mid to Lower Cervical Fusion with Instrumentation

Posterior cervical fusion with instrumentation is not commonly used in very young children, mainly because the delicate bone structures will not support internal bracing with large pieces of rigid metal. The latter also reduce the bone surface available for new bone formation. However, cervical instrumentation is occasionally indicated in children with postlaminectomy kyphosis when laminar wiring is not practicable or when a necessary multilevel fusion exceeds the holding capacity of a strut bone graft.

Kyphotic deformity involving more than two vertebral segments requires a corrective strut that can withstand rigorous stress. Particularly in children with connective tissue disease, such as Ehlers-Danlos syndrome and diastrophic dwarfism, the chronic curvature is not generally correctable with the halo device, and the internal fixation must bear most of the deforming strain. In such cases, a metal strut tailored to the length and contour of the deformed segments is fastened to the laminae with sublaminar cables. Choices of strut include a titanium Steinmann pin bent to an appropriately sized rectangle or prefabricated Luque rectangles (Zimmer USA, Warsaw, IN) and Codman subaxial rectangles. Autologous cancellous bone grafts are laid over the exposed fusion surfaces. Because of the unstable nature of these conditions, a halo is used postoperatively for at least 4 to 5 months to reinforce the fusion construct.

The ultimate challenge lies in attempting to stabilize a young child with kyphotic instability after extensive laminectomy. Not only are the fusion surfaces drastically reduced, but the best support for the wiring, the laminae, are also gone. The use of interfacetal wiring or facetal strut grafts is limited by the fragility of the facets in young children. For children younger than 5 years, we use a contoured Luque rectangle with the top and bottom bars fastened to one or two intact laminae at each end of the laminectomy defect, but the fusion necessarily includes at least two extra levels beyond the laminectomy defect. For older children, the best alternative is lateral mass screws and plates.²⁶⁵ These posterior plates were not used extensively in children because of the variable size of their articular pillars and the closeness of the vertebral artery to the screw path, but recently available thinner and shorter screws have broadened their use in children, albeit selectively, depending on the size of the vertebrae and the age of the child. They are the preferred technique because the resulting rigid construct dispenses with the need for the halo.

Injuries Involving the C1-2 Joint

Different fusion techniques are used for the atlantoaxial articulation because of its unique anatomic and biomechanical characteristics. For example, the atlas does not have a spinous process for interspinous wiring. Because of the wide gap between the posterior arches of C1 and C2, simple sublaminar wiring does not impart stability in extension unless the two arches are actually made to approximate. Moreover, the inclination of the atlantoaxial facet joints and the interfacetal ligaments is teleologically “suited” to a 45-degree range of axial rotation^96,257 and a high degree of translational freedom. Any fusion technique for this extremely mobile joint must therefore withstand all four planes of motion—flexion, extension, rotation, and translation.

The Gallie method, which uses a single midline wire loop around the arch of C1 and the spinous process of C2, does not meet the desired objectives because it eliminates neither extension nor rotation.²⁶⁶ In contrast, the Brooks fusion uses bilateral wires passed around the arch of C1 and the lamina of C2 and then tightened separately over a corticocancellous bone graft compressed against the C1 and C2 rings.¹¹¹ A modification of the Brooks technique by Griswold and colleagues involves the use of an almost T-shaped bone graft wedged between the C1 and C2 bony rings; the flat part of the T prevents the graft from being displaced into the spinal canal.²⁶⁷ With the modified Brooks construct, there is stability in both flexion and extension: flexion is eliminated by the wires, and extension is stopped by the buttressing bone graft,⁹⁶ which also serves as a friction block between the two bony arches and offers stability against axial rotation and translation. The modified Brooks fusion is very effective in children with atlantoaxial instability, especially when the C1 arch is fixed in an upward tilt that leaves a wide gap between it and the C2 neural arch (Fig. 224-47). A newer modification of the Brooks method in which two separate interposing bone grafts are used, one on each side of the midline, allows the Brooks principle to be used in cases in which the middle third of the C1 arch has to be removed for decompression, such as when C1-2 realignment is less than perfect.^51,268

FIGURE 224-47 Modified Brooks’ fusion in an 8-year-old girl with untreated C1-2 rotatory fixation. The right lateral mass of C2 is rotated to the left and “stepped off” in front of the C2 facet, thereby preventing reduction despite prolonged traction. Because of the unreduced C2 lateral mass, the posterior arch of C1 is “jacked up” in relation to the C2 laminae. A, Lateral radiograph demonstrating the persistently jacked-up position of the C1 arch. B, Intraoperative exposure showing the wide gap between the C1 arch and the C2 laminae. The C1 arch did not budge with earnest attempts to bring it down with instruments. C, Posterior view of the tricorticate T-shaped interposition graft from the iliac crest. Its natural curvature fits the transverse contour of the neural arches, and a notch was cut for the spinous process of C2. The inset drawing is an end-on sagittal section of the T-shaped graft interposed between the C1 and C2 arches. D, Lateral view of the graft showing the T-shaped fitting for the adjacent neural arches. E, Bone graft wired into place with two wires looping around both neural arches and the graft. F, Postoperative radiograph showing the fusion construct.

There is, however, a caveat when using the Brooks method in young children. It seems that the incidence of rapid graft absorption is higher in young children than in adults, possibly because halo fixation in young children does not provide the same degree of stability and the continued micromotion at the fusion site promotes graft resorption and retards bone formation, thereby ultimately leading to graft shrinkage, which then loosens the wire loops.⁵¹ Thus, in children younger than 3 or 4 years, we prefer to pass bilateral sublaminar wires around C1 and C2 and then tighten them until the two arches touch.^20,51 This is quite easily and safely achieved in most young children without placing excessive strain on the arch of C1 (Fig. 224-48). The conjoined bony rings offer stability in both flexion and extension, and the two-point fixation provided by the double wires also increases friction between the twin rings and reduces axial rotation and translation. On-laid bone grafts are then compressed tightly against the fusion surface with deep sutures. The inherent strength of this construct is undiminished by any interim changes in the grafts, and fusion is almost guaranteed as long as the bone rings and wires hold up against breakage.

FIGURE 224-48 Posterior wire fusion of C1 and C2 in a 6-year-old child. Note the approximation of the bony rings of C1 and C2 by the two sublaminar wire loops that are tightly twisted below the C2 ring. A, Drawing. B, Intraoperative appearance.

Atlantoaxial Fusion with Instrumentation

In older children, other options are available for instrumentation of the C1-2 joint. Magerl pioneered the use of transarticular screw fixation of C1-2 in 1979, and he and Seemann reported their experience in a large series of adults in 1986.¹⁷² Modern adaptations of the original technique using newer drill assemblies, precise planning of the trajectory, and guidance with frameless stereotactic navigation have been described.²⁶⁹ The chief advantages of this technique include excellent fixation against translation and rotation; when supplemented with an interposing bone graft between the arches of C1 and C2, which prevents flexion and extension, use of the halo postoperatively can be avoided.²⁷⁰ In addition, it is useful even when the arches of C1 or C2 have been removed for decompression. Reported successful fusion rates are 95% to 98%,^271,272 as compared with 80% to 86% when wire-graft techniques are used with the halo.^268,271 Complication rates for vertebral artery injury, ischemic stroke, and neural injury are exceedingly low when proper precautions are taken.^{269,271–273} With the addition of bilateral vertical plates, the C1-2 transarticular screw construct can be extended to incorporate lateral mass screw fixation of the subaxial spine.

The bulk of the literature on C1-2 transarticular screw fixation concerns adults. In the last 5 years, this technique has been applied to an increasing number of children, and emerging data suggest great promise (Fig. 224-49). Wang and colleagues reported a series of 13 children (the youngest being 4 years old) with a 100% fusion rate, no complications, and no postoperative halo use,²⁷⁴ as opposed to an 84% fusion rate with a wire-graft method and halo reported by others.¹⁷⁷ Recently, Gluf and Brockmeyer reported equally good results in 67 children, and their experience and insight have underscored a number of special precautions in children that must be heeded.²⁷⁵

FIGURE 224-49 Lateral radiograph showing C1-2 transarticular screw fixation in a 12-year-old child with congenital C1-2 instability. Flexion and extension motions are eliminated by the simultaneous Sonntag-type posterior wiring.

In some adults, the bony groove for the vertebral artery at C2 may be so large that the width of the pars interarticularis (isthmus) is reduced to below safety limits for passage of a 3.5-mm screw. Madawi and associates²⁷⁶ and Paramore and coworkers²⁷⁷ reported a 20% exclusion rate for at least one isthmus because of the vertebral artery anatomy, and in 4% to 5%, neither side was deemed safe. The isthmus in children is accordingly narrower than in adults (Fig. 224-50). Especially in younger children, the absolute dimension of the isthmus must be premeasured and the screw path preplanned on a virtual image before surgery. Triplane computer reconstruction with the cursor traversing the long axis of the isthmus (simulating the screw path) should be done on every child. If only unilateral screw placement is feasible, a postoperative halo should be used. Madawi and associates reported a higher incidence of screw malpositioning if the C1-2 subluxation is poorly reduced with considerable offset in the facet contact surfaces.^276,278 Good alignment is therefore very important in pediatric patients.

FIGURE 224-50 Drawings of C1-2 transarticular screw fixation. The desired screw trajectory is through the C2 isthmus and the C1-2 articulation. The medial side of the C2 pedicle serves as a direction guide. In children and in some adults with a wide vertebral artery groove, the isthmus may be too narrow for the path of the screw.

(From Apfelbaum R. Posterior Transarticular C1-C2 Screw Fixation for Atlanto-Axial Instability. Aesculap Scientific Information No. 25. South San Francisco, CA: Aesculap; 1993.)

One final word on C1-2 instrumentation: anterior odontoid screw fixation has been used in a few children with type II odontoid fractures that were perfectly reduced. The advantage of this technique over a transarticular screw is preservation of C1-2 rotation, which is especially desirable in children. Further experience is needed before the use of this technique in children can be widely endorsed.

Occipital-Cervical Fusion with Instrumentation

Occasionally, an anteriorly dislocated C1 arch cannot be reduced by traction or manipulation. This happens in some cases of os odontoideum, chronic posttraumatic anterolisthesis, and several forms of mucolipidosis in which abnormal tissues are interposed between the dens and the anterior arch of C1. Sublaminar wire fusion of the unreduced ring of C1-2 may add to the spinal cord compression. In such instances, it is safer to remove the arch of C1 to decompress the cord before attempting to fuse the occiput, C2, and C3 together. For this situation and for the progressive kyphosis sometimes seen after occipitocervical decompression for Chiari malformation, we use a Steinmann pin shaped in an inverted U with the bend contoured to conform to the occipital boss.⁵¹ The two vertical bars are wired to the top two available laminae, and the curved U bend is wired to the occiput through bur holes on both sides of the midline (Fig. 224-51). Use of the Luque rectangle adds a measure of inferior stability by fixing the lower rim of the rectangle to the lower intact lamina (Fig. 224-52).²⁰ Cancellous bone grafts are packed into the lateral gutters at C1, on the decorticated laminae, and on the occipital squama.

FIGURE 224-51 Steinmann pin fixation for occiput-C2-C3 fusion. The steel or titanium pin is first bent in an inverted U, and the bend is contoured to the occiput. The bend is wired to the occiput through burr holes, and the vertical arms are fixed to C2 and C3 with sublaminar wires.

FIGURE 224-52 Occipitoatlantoaxial fusion with the Luque rectangle fixed to the occiput with titanium cables through burr holes and to the C1 and C2 arches with sublaminar cables.

The chief disadvantage of the rigid U-bar technique is that sliding motion between the metal bar and the metallic cables cannot be totally eliminated. Completely rigid fixation is not obtained, and because of the great angular momentum of the child’s heavy head working against the cable-bar contact, a halo vest is always recommended postoperatively. Plates have been affixed more rigidly with screws to the occiput and the cervical spine,^279–288 but with children, bone thickness of the occiput is of great concern in terms of both providing adequate bony purchase for the traditional outside-to-inside screws and avoiding penetration of the dura. In fact, the unevenness of bone thickness in the occipital squama of a child is such that screw purchase is secure only into the midline keel and far laterally near the mastoid process.²⁸⁹

Pait and colleagues described the “inside-outside” technique, which obviates most of the problems related to screw purchase in thin skulls and is eminently suitable for occipital-cervical fusion in children.¹⁷⁹ In this technique, a lateral mass plate is contoured to fit the shape of the occiput and the cervical spine, and the plate is affixed to the occiput by means of a flat-headed screw placed through a bur hole in the calvaria so that the head of the screw is in the epidural space and the threads face outward—hence the sobriquet “inside-outside.” The burr hole marks the “entry site” of the inside-outside screw. After the dura is separated from the inner table of the skull, a “keyhole” is cut with a craniotome foot-plated saw blade from the entry site some distance away to the premarked “exit site” for the screw. The size of the exit site is cut to the size of the screw stem, always much smaller than the burr hole. The inside-outside screw is attached to the guide rod (Dyna-Lok T-bolt and nut; Sofamor Danek, Memphis, TN) and then carefully placed epidurally into the entry site and guided to the exit site (Fig. 224-53A). The reconstruction plate is placed “through” the outpointing screw stem, and a nut is semitightened on the exposed screw threads with a socket wrench (see Fig. 224-53B). After the C1-2 transarticular screw (or C2 pedicle screw) and all the subaxial cervical lateral mass screws have been inserted and tightened on the plate, the occipital inside-outside screw is tightened firmly against the nut with the socket wrench while a screwdriver is inserted through the socket wrench into the hex in the hollow inside-outside screw to provide counter-torque (see Fig. 224-53C).

FIGURE 224-53 Pait’s “inside-outside” screw technique. A, The inside-outside screw is attached to a guide rod and inserted into the entry burr hole. B, The reconstruction plate is placed over the inside-outside screw, and the nut is loosely attached to the screw. C, After adjustments for the suboccipital screws on the reconstructed plate, the nut is firmly tightened on the inside-outside screw with a socket wrench while a screwdriver is inserted through the wrench into the hex of the hollow screw head to provide counter-torque.

(From Pait TG, Al-Mefty O, Boop FA, et al. Inside-outside technique for posterior occipitocervical spine instrumentation and stabilization: preliminary results. J Neurosurg Spine. 1999;90:1-7.)

Different plate systems can be linked to the inside-outside screw, depending on the size and age of the child. Children as young as 2 years have been treated with AO/Synthes Posterior Cervical Plates (Synthes Spine, Paoli, PA), but with this plate system, the locations of the apertures for the cervical screws are restricted. With the newer Star Lock CerviFix System, the screw heads on the cervical spine are attached to the vertical bars via sliding links. The inside-outside screws, as well as all the cervical screws, can be placed individually before application of the vertical elements and final adjustment of alignment, thereby eliminating a great deal of the “fidget factor” of the procedure (Fig. 224-54). With this technique, children as young as 12 months can safely be instrumented for occipital C1-2 fusion. The thinness of the occipital bone in young children can be compensated for by choosing screw heads with larger diameters to distribute the crushing forces. A modified axis plate with a custom-tapered perforated vertical stem (Sofamor Danek, Memphis, TN) is designed especially for sublaminar wiring to small delicate laminae (Figs. 224-55 and 224-56). Postoperative halo devices are not usually necessary except for the very young, when redistribution of stress from the fragile laminae to the external orthosis is still necessary. Preliminary fusion rates are very promising.^123,179

FIGURE 224-54 Inside-outside screw fixation construct in a 6-year-old child. A, Vertical bars and sliding screw head links for the cervical screws. B, Lateral radiograph.

FIGURE 224-55 Modified axis plate for inside-outside screw fixation in young children. A special slender perforated stem (between the arrows) has been adapted for sublaminar wiring onto delicate laminae.

FIGURE 224-56 Inside-outside screw-plate fixation for occipital-cervical fusion. Final nut tightening onto the vertical axis plate is accomplished with a socket wrench (black arrow), and a screw driver (white arrow) is inserted into the hollow hex of the inside-outside screw stem to provide counter-torque. Note the slender perforated lower stem for sublaminar cables (already fastened with crimpers).