Os Odontoideum 

Updated: Apr 07, 2021
Author: Eeric Truumees, MD; Chief Editor: Jeffrey A Goldstein, MD 


Practice Essentials

Os odontoideum is an anatomic anomaly of the upper cervical spine that may be defined in radiologic terms as an oval or round ossicle with smooth circumferential cortical margins representing a hypoplastic dens (odontoid process) that has no continuity with the body of C2 (the axis).[1]

In 1863, separation of the odontoid process from the body of the axis was first described in a postmortem specimen. In 1886, Giacomini coined the term os odontoideum for this condition.[2]  This entity is clinically important because a mobile or insufficient dens renders the transverse atlantal ligament (TAL) ineffective at restraining atlantoaxial motion. Translation of the atlas (C1) on the axis may compress the cervical cord or vertebral arteries.[3, 4]

Os odontoideum is rare, but the exact frequency is unknown. Many cases are incidentally detected. Others are diagnosed when patients become symptomatic. There are occasional reports of patients with os odontoideum becoming quadriparetic after minor trauma.[5, 6]  To date, no large-scale screening studies have been performed. In one magnetic resonance imaging (MRI) study of odontoid morphology, a 0.7% (one case in 133 patients) incidence was reported.[7]

There are three evolving and controversial aspects of os odontoideum: etiology, surgical indications, and optimal management. Hadley’s 2002 consensus report called for several studies[8] :

  • Population-wide studies of the prevalence of os odontoideum as an incidental finding
  • A follow-up of incidentally noted and untreated os odontoideum, even with C1-C2 subluxation.
  • A cooperative, multi-institutional natural history study of patients with os odontoideum without C1-C2 instability, to provide demographic and clinical factors predictive of the development of subsequent instability
  • A multi-institutional prospective, randomized trial comparing posterior wiring and fusion techniques with rigid C1-C2 screw fixation

The age at diagnosis varies significantly from the first to the sixth decade of life. With increased awareness, however, os odontoideum has been diagnosed in younger patients.[9] While the etiology remains controversial, an increased frequency of os odontoideum has been reported in patients with Morquio syndrome,[10, 11]  multiple epiphyseal dysplasia, and Down syndrome.

There are two main forms of management, as follows:

  • Clinical and radiologic surveillance
  • Operative stabilization

Surgical stabilization may be recommended in the following three settings:

  • Spinal instability
  • Neurologic involvement
  • Intractable pain


Successful treatment of os odontoideum requires an understanding of the unique anatomic characteristics of the cervicocranium (occiput-C2). The bony elements here develop through enchondral ossification.

The tip of the dens and its associated ligaments arise from the fourth occipital through the cervical-0 (C0) somites, which do not ossify until middle childhood.[12, 13] The base of the dens forms from the C0 and C1 sclerotomes as two paired structures that ossify just before birth. The C2 and C3 sclerotomes give rise to the body of the axis, which fuses with the dens at age 4 years. (See the image below.)

The axis has 5 primary and 2 secondary ossificatio The axis has 5 primary and 2 secondary ossification centers. C0, C1, and C2 sclerotomes contribute to various portions of the dens. The principal portion of the dens body arises from the original center of C1.

In a study of human embryos at 8 weeks of gestation, O’Rahilly et al reported that no transverse segmentation formed within the odontoid process at any time. An embryologic anomaly characterized by a complete/partial segmentation of two rostral parts will result in bipartite dens rather than an os odontoideum.[14]

The atlantoaxial joint (C1-C2) consists of biconcave articulations with loose capsules and small contact areas. Stability is therefore conferred by associated ligaments, including the TAL, which is the primary restraint to flexion and extension. A study of 41 patients with irreducible os odontoideum found that the position of the TAL anterior and inferior to the ossicle was the most common factor associated with irreducibility.[15]

Other important restraints include the apical ligaments, the alar ligaments, the tectorial membrane, and the atlanto-occipital membranes

The vertebral arteries are intricately invested in the bony anatomy of the atlantoaxial segment. They pass just inferior to the C1-C2 facet joint, then course laterally through the transverse foramen of C2. Just above the C1 lateral mass, they turn medially and meet to progress cephalad into the foramen magnum.[16] Aberrancy of the vertebral artery course is not rare and may limit fixation options in some patients selected for operative management of os odontoideum. This deviant course may be unilateral or bilateral.


Initially, os odontoideum was thought to represent a congenital failure of fusion of the dens to the remainder of the axis. As such, the condition is usually grouped with other craniocervical junction abnormalities, such as dental aplasia and hypoplasia.[17, 18, 19, 20, 21, 22, 23, 24]  Today, it seems clear that failure of the secondary ossification center of the dens to fuse with the base of the odontoid represents a separate entity known as persistent ossiculum terminale.[25]

Differentiation between os odontoideum and persistent ossiculum terminale is clinically critical. The ossicle of the ossiculum terminale is much smaller than that of the os odontoideum. More important, that ossicle lies at the level of the atlantal ring above the transverse atlantal ligament. In this cranial location, ossiculum terminale, unlike os odontoideum, is not associated with significant instability.

Some authors speculate that os odontoideum represents a previous fracture of the odontoid synchondrosis before its closure at age 5-6 years.[26, 27, 28]  These authors describe os odontoideum in patients with previously normal cervical radiographs. For example, Schuler et al elegantly described the evolution of an os odontoideum following trauma in a child.[29]  Subsequently, Menezes identified os odontoideum in children younger than 5 years with a previously normal odontoid. He associated the os with unrecognized fractures.[30]

In this model, os odontoideum develops gradually. Following a fracture of the odontoid synchondrosis, with growth, the alar ligaments carry the dens fragment away from the axis base. The cranial portion of the dens fragment continues to receive a blood supply from the vascular arcade. The avascular caudad portion resorbs, leaving the characteristic rounded ossicle.

One paper documented os odontoideum formation after a fracture with serial radiographs.[31]  Another report, in a 4-year-old girl, showed an incompletely ossified, cartilaginous orthotopic os.[32]  The authors argued for a multifactorial etiology related to the embryology and vascular supply of the odontoid process. A case from the year before made a similar argument with a posttrauma computed tomography (CT) scan showing no fracture, followed by os odontoideum 10 years later.[33]

Authors who favor a congenital basis for os odontoideum point out that the craniovertebral junction is one of the most common sites for malformation. Included are clefts or aplasia of anterior and posterior arches of atlas.[17, 20, 21, 28]  On the other hand, unlike most congenital malformations, os odontoideum tends to occur as an isolated entity without other regional anomalies. Garg et al reported a case of os odontoideum in a myelopathic 16-year-old patient with bipartite atlas.[34] They concluded that coexistence of these conditions support the embryologic basis for os odontoideum. In this case, the dens had an “unusual bony projection” on its anterior surface.

Crockard and Stevens reviewed the embryologic and comparative anatomy data of clinical syndromes associated with craniocervical instability.[35] They concluded that os odontoideum is the product of excessive movement at the time of ossification of the cartilaginous dens and is analogous to the unfused type II odontoid fracture. True hypoplasia of the odontoid peg, on the other hand, was found to be part of a wider segmentation defect associated with Klippel-Feil syndrome,[36]  occipitalized atlas, or basilar invagination and rarely was found to be associated with instability.

In 2006, Sankar et al reviewed 519 consecutive patients with radiographic abnormalities in the occipitocervical region.[37] Os odontoideum was confirmed in 16. Only three of those patients had a history of remote trauma. The authors concluded that this supported an embryologic basis for the condition.

The size of the os odontoideum may vary, but it typically is smaller than the normal dens, particularly at its base. Perhaps there are two etiologic groups. Certainly, in patients with other congenital anomalies or odontoid malformations, an embryologic basis may be assumed. Unfortunately, individual correlations do not prove the case one way or the other. A report of three cases of familial os odontoideum was published in 2011.[38]

Regardless of the underlying cause, sound treatment selection requires an understanding of the natural history of this process. It is known that in a subset of patients, the secondary ligamentous restraints become lax. With increasing laxity, anterior motion of the atlas on the axis can become excessive. When the instability has been longstanding, it becomes multidirectional. Less clear are the percentages and rates of progression to instability. As no population studies are available, assumptions are made from case series data of nonoperatively managed patients.

Two types of os odontoideum are described, based on the position of the dens tip: orthotopic and dystopic os odontoideum. In the orthotopic type, the dens fragment lies in an anatomic position. In dystopic os odontoideum, the dens tip is in any other position. Most commonly, the fragment is located near the foramen magnum, where it may fuse with the clivus. Alternatively, the os may be fixed to the anterior ring of the atlas.

Subluxation and instability are described in both types of os odontoideum. Some authors believe that dystopic os odontoideum is more likely to be symptomatic.

In years to come, both posttraumatic and congenital forms of os odontoideum may be identified; however, authors in each camp have often written as if their hypothesis has been all but proved. Understanding the true etiology of this disorder may be helpful in terms of identifying high-risk patients through genetic testing.

If os odontoideum is posttraumatic, the increased use of advanced imaging modalities such as spiral CT and MRI may markedly decrease its incidence in the future. Furthermore, better explication of the etiology may better delineate progression risk, leading to decreased x-ray exposure during follow-up and the limiting of operative intervention only to those patients at high risk for progression. These data may also support or eliminate the need for activity restriction during the observation phase.


Outcomes and prognosis data for os odontoideum patients have been limited to scattered case reports and small case series. These series have typically described successful outcomes with both nonoperative and surgical management.[39, 40, 41]

Klimo et al investigated 77 patients with os odontoideum who underwent surgical stabilization.[3]  In all patients, fusion was achieved at a mean duration of 4.8 months. It was reported that 90% of patients experienced improvement in their neck pain or neurologic symptoms. Furthermore, it was reported that 39% of the patients who presented with myelopathy had complete resolution, 50% showed improvement in their spasticity, and 11% had constant and unchanged spasticity.

Spierings et al described their management of 37 os odontoideum patients.[42] At a median follow-up of 7 years, of the 16 with neck pain only, none developed neurologic deficits. Four patients had mild or transient myelopathy and limited radiographic instability and were also followed nonoperatively. With a maximum follow-up of 14 years, three of the four had no recurrence; one had stable monoparesis. Seventeen patients with myelopathy and more instability underwent surgery. The authors found no difference in neck pain between the surgical and nonsurgical groups.

In a series by Dai et al, five patients with os odontoideum without symptoms were treated nonoperatively and were monitored for an average of 6.5 years.[43] None of these patients reported symptom progression during the follow-up period.

In Fielding’s study of 35 os odontoideum patients, 27 were radiographically unstable.[44] Twenty-six of those 27 underwent successful Gallie fusion. They reported “solid” fusions after 2 months of immobilization in children and 3 months in adults. The 27th unstable patient refused surgery and remained well at 2-year follow-up examination. The eight stable patients were managed nonoperatively and remained well at last follow-up. Symptomatically, two thirds of Fielding’s patients had only mechanical pain. These patients reported resolution after fusion.

The neurologic outcome in patients with symptomatic cord compression is less clear. After surgery, typically, symptomatic progression ceases and most patients report significant symptomatic improvement. These patients can be divided into those with acute, incomplete cord syndromes and those with more insidious myelopathic syndromes. Typically, transient neurologic signs following trauma is associated with a good prognosis. Rapid functional return parallels improvement in neurologic signs, and recurrence rates are low.

A retrospective study of 21 patients treated with rigid, posterior instrumentation showed a mean improvement in the Nurick scale for myelopathy of 2.3 before surgery to 0.7 at the time of follow-up.[45]  At a mean 34.7 months of follow-up, Odom's criteria outcomes were excellent in 47%, good in 37%, fair in 11%, and poor in 5%. All patients with preoperative neck pain had symptom relief or improvement, with all of these patients having more than 83.7% improvement in visual analogue scale scores. The authors reported a mean improvement in space available for cord from 9.3 mm to 17.7 mm. There were no pseudarthroses or screw failures.

Zhang et al determined that patients with asymptomatic or myelopathic atlantoaxial instability secondary to os odontoideum were at risk for acute spinal cord injury after minor traumatic injury.[46] They suggested fixation and fusion should be undertaken as prophylactic treatment to avoid these injuries.

Patients with cerebellar or medullary signs exhibit a more progressive course. In that these patients gradually worsen, surgery is most clearly indicated in this setting.

In a retrospective multicenter review, Saarinen et al compared the outcomes of conservative (n = 10) and surgical treatment (n = 34) of idiopathic os odontoideum in children with the outcomes of conservative (n = 10) and surgical treatment (n = 34) of nonidiopathic os odontoideum.[47] ​ The authors found that patients with idiopathic os odontoideum who had a stable atlantoaxial joint at presentation remained asymptomatic and intact during conservative treatment. Children with idiopathic os odontoideum who underwent spinal arthrodesis had significantly lower complication and nonunion rates than children with nonidiopathic os odontoideum.




A significant but unknown percentage of those with os odontoideum remain asymptomatic. In this population, os odontoideum may be detected incidentally after screening or after an unrelated trauma. Given the frequency of asymptomatic os odontoideum, when symptoms do occur, it is difficult to determine with certainty that the os odontoideum is the true cause. Symptoms of os odontoideum may include the following:

  • Local mechanical neck pain
  • Torticollis and headache
  • Neurovascular symptoms

More rarely, patients may present with thoracic pain only.[48]

In patients with cervical instability, hypermobility of C1 on C2 may lead to direct compression on the spinal cord or embarrassment of its blood supply. As a result, neurologic symptoms may develop.[49, 50]  These neurologic symptoms range from a transitory episode of diffuse paresis following trauma to progressive myelopathy to complete spinal cord injury.[51]  Weakness and ataxia usually predominate over sensory changes.

Less frequently, atlantoaxial instability results in vertebral artery compression precipitating neurovascular symptoms. These vascular symptoms arise from the cervical cord and brainstem ischemia and encompass a bewildering array of signs and symptoms. Early sequelae include ataxia, syncope, vertigo, and visual disturbances. Later, cerebellar and brainstem infarcts and seizures are seen.[52, 53]  In one report, onset of Ondine’s curse (central hypoventilation syndrome) in an 18-year-old was ascribed to os odontoideum. Sudden death is rare but can occur.[54]

Physical Examination

When os odontoideum is suspected, a thorough physical examination is mandatory. Begin with a complete neck and cervical spine examination. Evaluate for tenderness, range of motion (ROM), and associated anomalies. A careful neurologic examination should include assessment of cerebellar and brainstem function, gait evaluation, and a Romberg test. In patients with atlantoaxial instability, upper-motor-neuron findings are commonly identified and may include spasticity, hyperreflexia, clonus, and proprioceptive loss.



Plain Radiography

Radiologic evaluation is used to confirm the diagnosis and estimate the degree of spinal instability. Initial evaluation includes open-mouth anterior-posterior and flexion-extension lateral radiographs. Os odontoideum appears as a round or oval ossicle with a smooth uniform cortex separated from the base of the axis by a wide gap (see the image below). The ossicle border does not directly match up with the axis body. The gap separating the os and the axis proper should lie above the level of the superior articular facets.[55, 56, 57]

A coronal reconstruction of an orthotopic os odont A coronal reconstruction of an orthotopic os odontoideum. Note the wide gap between the rounded ossicle and the base of the axis.

Orthotopic os odontoideum (see the image above) may appear free and in a relatively anatomic position, where it may be difficult to differentiate from an unfused neurocentral synchondrosis, odontoid hypoplasia, or odontoid fracture nonunion. In children younger than 5 years, the neurocentral synchondrosis often has not fused. Dynamic lateral radiographs of those with an unfused synchondrosis do not demonstrate motion, whereas radiographs of individuals with an os odontoideum may demonstrate motion.

A dystopic ossicle may be fixed to the clivus or to the anterior ring of the atlas. The remaining axis is hypoplastic as well. With a dystopic os odontoideum (see the image below), the radiographic diagnosis is clear.

A sagittal reconstruction of a CT scan demonstrati A sagittal reconstruction of a CT scan demonstrating a dystopic os odontoideum. Note that the ossicle appears fused to the clivus (anterior portion of the foramen magnum). Also note the smooth corticated border of the ossicle.

A dens (odontoid process) fracture nonunion (see the image below) typically exhibits a narrow gap between the axis base and the dens. The normal shape and size of the dens are preserved on the open-mouth view.

Anterior-posterior tomogram view of a type II dens Anterior-posterior tomogram view of a type II dens fracture. The fracture line is narrow and lower on the waist of the dens, unlike the fracture line of an os odontoideum. No cortication is noted along the fracture line.

With an os odontoideum, hypertrophy of the anterior arch of the atlas may be seen. This hypertrophy is believed to represent osseous reaction to chronic atlantoaxial instability and is unlikely with an acute dens fracture.

Flexion-extension lateral radiographs typically show motion in symptomatic patients. This abnormal motion is typically seen in the anterior-posterior plane. Some patients are unstable in all directions. In one series, the average translatory motion was 1 cm.

Important prognostic indices are as follows:

  • Anterior atlantoaxial translation
  • Posterior atlantodens interval (PADI)
  • Instability index - The change in the space available for the cord in flexion versus extension
  • Sagittal plane angulation - The difference in the atlantoaxial angle between flexion and extension

In some forms of atlantoaxial subluxation, the anterior atlantodens interval (AADI) is used to measure instability. However, with os odontoideum, the os fragment often moves with the atlas. Therefore, the AADI does not reflect the abnormal motion of the segment. Direct measurement of the motion of C1 on the body of C2 is more useful.

Anterior atlantoaxial translation represents the space between a line projected superiorly from the anterior border of the body of the axis and a line projected inferiorly from the posterior border of the anterior arch of the atlas. More than 3 mm of separation is pathologic.

CT and MRI

Critical evaluation of the bony anatomy of the upper cervical spine is often difficult with plain radiographs alone. A fine-cut (1-mm) sagittally reconstructed computed tomography (CT) scan allows a more detailed depiction of the atlantoaxial articulation.

Previously, the distance between the posterior border of the dens and the anterior border of the posterior ring of the atlas on plain radiographs was termed the space available for the cord (SAC). Currently, this distance is more frequently referred to as the PADI. These terms are occasionally used synonymously.

With magnetic resonance imaging (MRI) or CT myelography, however, the actual space for the cord can be readily measured, and the SAC can be used to refer to the actual anterior-posterior canal dimension. Practically, SAC refers to the PADI minus additional compression from soft tissue. This measurement should be carried from the superior posterior corner of C2 to the posterior ring of C1. Soft-tissue structures such as synovial cysts may further diminish the SAC in some cases of os odontoideum. A PADI of less than 13 mm is associated with neurologic decline.

In one study, plain radiographic measurements of translation and PADI did not accurately reflect clinical status. MRI measurement of cord compression was more predictive of symptomatology. Hadley’s literature review was unable to establish a linear relationship between PADI and neurologic status.[8]

MRI also may delineate pathologic changes within the cord. On T2-weighted MRI (see the image below) sequences, an increased signal in the substance of the cord may reflect edema or myelomalacia.[58, 59, 60]  An altered T1 signal typically portends a more grim prognosis neurologically, in that it may reflect hemorrhage or necrosis within the cord.

T2-weighted parasagittal MRI image of a patient wi T2-weighted parasagittal MRI image of a patient with os odontoideum and mild compression of the upper cervical spine. This patient presented with transient quadriparesis.

A number of dynamic imaging modalities have been recommended as means to more completely understand the degree and nature of abnormal motion in patients with os odontoideum. For example, cineradiographs and dynamic (flexion-extension) MRI have been recommended because of their ability to define the relationship of the os to surrounding bony elements.[59, 60, 61]  As of this writing, these modalities are not widely used. CT angiography (CTA) and magnetic resonance angiography (MRA) are more commonly helpful in understanding aberrant vertebral artery anatomy or persistence of the first segmental artery.[62, 63]



Approach Considerations

In patients with os odontoideum, there are two main forms of management, as follows:

  • Clinical and radiologic surveillance
  • Operative stabilization

There remains some debate about optimal treatment of this condition. A systematic review and meta-analysis by Pommier et al (N = 2463; 30 nonrandomized comparative studies) concluded that surgical treatment apparently was not inferior to conservative treatment; however, the quality of the evidence was low.[64]

Surgical stabilization may be recommended in the following three settings:

  • Spinal instability
  • Neurologic involvement
  • Intractable pain

In this setting, spinal instability is defined as cord compression or excessive motion noted radiographically. In the absence of large-cohort prospective data, radiographic parameters can only be marginally associated with progressive neurologic dysfunction. That said, most authors agree that the following changes on flexion-extension plain lateral cervical radiographs serve as reasonable guidelines for surgery:

  • Posterior atlantodens interval (PADI) < 13 mm
  • Sagittal plane rotational angle >20°
  • Instability index >40%
  • C1-C2 translation >5 mm

Surgical stabilization is least controversial in patients with obvious neurologic or neurovascular involvement. In some small series, on the other hand, continued nonoperative management has been recommended on the basis of the resolution of symptoms following transient paresis. Nevertheless, even in patients with a complete neurologic deficit, axial pain or the possibility of neurovascular compromise to the brainstem may still indicate surgery.

Surgical intervention in patients with axial pain is more controversial. In those with persistent and disabling pain despite appropriate nonoperative management, stabilization may be reasonable. Yet to justify this approach, surgical results must improve on the natural history of this disease state. Some authors discourage surgical intervention for neck pain alone, stating that surgical outcomes are not sufficiently better than medical treatment to justify the risks.

Contraindications for surgery in os odontoideum patients begin with those patients not expected to benefit from stabilization. In a series of patients without spinal cord symptoms, no difference in outcome existed between those treated with surgical fusion and those treated medically.

In patients with neurologic deterioration, there are few reasonable alternatives to fusion. In smaller children without progressive deficits, it may be appropriate to wait until the bony elements have increased in size (eg, until age 6-7 years). Sublaminar wire passage and screw fixation are technically more difficult, and the risk for iatrogenic injury higher, in smaller patients. Other contraindications address one surgical technique or another.[24, 65, 66, 67]

Nonoperative Therapy

Observation is appropriate in most incidentally diagnosed os odontoideum patients, particularly those without radiographic evidence of significant instability. For patients with mechanical symptoms, medical management is indicated. This treatment includes cervical traction, physical therapy, occasional collar use, and anti-inflammatory medications. Activity limitation often is recommended but is difficult to enforce in the pediatric age group.

Surgical Therapy

Several surgical options have been utilized for os odontoideum:

  • Posterior atlantoaxial onlay fusion
  • Posterior atlantoaxial wiring and fusion
  • Posterior occipitocervical wiring and fusion
  • Posterior Magerl screw fixation and fusion
  • Harms-Goel technique of C1-C2 fusion
  • Anterior resection of the os fragment

Onlay fusions are technically straightforward but confer high pseudarthrosis rates. These procedures are best restricted to younger children for whom wire passage is considered high risk. Postoperatively, rigid external immobilization, such as a halo brace or Minerva cast, is recommended.

Historically, posterior atlantoaxial wiring was the standard technique for stabilization of os odontoideum (see the image below). There are a number of variants of this technique, based on the wiring scheme and bone graft placement. 

Lateral radiograph of a dystopic displaced os odon Lateral radiograph of a dystopic displaced os odontoideum 6 months after posterior wiring. After the wiring was performed, the patient had a solid arthrodesis with no motion on flexion and extension. Her neurologic symptoms resolved despite the failure to obtain a reduction. Had she continued to have severe symptoms, anterior odontoidectomy could have been considered.

The oldest variant, Gallie fusion, incorporates wires or cables under the posterior C1 ring and around the C2 spinous process. An onlay autogenous corticocancellous iliac crest graft is held in place by the wire. Hensinger reported that the C2 spinous process in young children often is insufficiently ossified to contain the Gallie wire reliably. He recommended placing a Kirschner wire (K-wire) through the spinous process and wrapping the intervertebral wire around the K-wire and bone.[68, 69, 70]  An, on the other hand, stated that wiring is not needed in young children.[71]

In a Brooks fusion, the wire passes under the C2 lamina as well as under the C1 ring.[72] A structural or tricortical bone graft is wedged between the C1 ring and the C2 lamina, blocking extension. Whereas C2 sublaminar wire passage adds surgical risk, the Brooks fusion is more rigid than the Gallie. Postoperatively, posterior subluxation of the C1 ring and os into the cord are occasionally seen with a Gallie fusion.

Although posterior wiring procedures have a long record of successful atlantoaxial stabilization, their use is declining in this patient population. Shortcomings include the need for postoperative halo or Minerva cast immobilization. The Brooks technique is more stable in extension than the Gallie technique, but neither confers much stability in the patient with multiaxis instability. During sublaminar wire passage, cord injury may occur, especially in patients with an irreducible deformity. Standard Gallie or Brooks techniques are not possible in the absence of the posterior C1 ring.

Before the development of screw-based techniques, these patients required occipitocervical fusion. A modification of the Brooks technique that overcomes this shortcoming has been described.[72] Today, most patients undergoing posterior fusion of an os odontoideum are offered a screw-based rigid stabilization. Preexisting C1 ring deficiency or C1 laminectomy concurrently performed for cord decompression in patients with irreducible deformity does not affect the more laterally placed screws.

The Magerl technique of posterior C2-C1 screw fixation immobilizes the atlantoaxial joint reliably and cost-effectively (see the image below). This technique, though mechanically the most rigid, is also the most technically demanding. The Magerl technique requires a near-anatomic reduction and normal vertebral artery anatomy.

Lateral radiograph demonstrating fixation of a red Lateral radiograph demonstrating fixation of a reduced os odontoideum with Magerl screws in a patient with an incomplete posterior arch of C1.

Introduced by Goel and popularized by Harms in 2001, the many versions of the Goel-Harms technique use C1 lateral mass screws connected to C2 pars, pedicle, or translaminar screws.[73] Today, Goel-Harms has become the common C1-2 stabilization option in adults and older children.[74] This segmental approach and a variety of C2 fixation options ensure its availability even in patients with aberrant vertebral artery anatomy.

Once screws are placed in C1, they can be used to further reduce some cases of os odontoideum. In patients with axial height loss, segmental height may be reestablished by placing allograft or PEEK blocks into the C1-2 facets. Still, careful preoperative measurement of the bony dimensions of the C1 lateral mass, C2 pars, pedicle, and lamina is required before this technique is recommended, especially in children.

In a small series, Hong et al concluded that the Harms C1-2 polyaxial screw-and-rod technique was the most appropriate treatment for patients with os odontoideum.[75]  In Park and coworkers’ series of pediatric atlantoaxial instability, 58% had an anomalous course of the vertebral artery and 42% had anomalous C1-2 bony anatomy.[76]

Several studies have shown that screw placement is possible in most children older than 4 years. In O’Reilly’s series, 11 of 12 pediatric patients were able to undergo transarticular fixation for atlantoaxial instability. Interestingly, the series contained several os odontoideum patients who had failed previous onlay or wired grafting.[14]

In a series of pediatric (mean age, 9.6 years) atlantoaxial fusions, Haque et al were able to achieve stability in all of their patients.[77] In another series, Geck et al took computed tomography (CT) measurements in children aged 2-6 years; C1 lateral mass and C2 pedicle or laminar screws were possible in the majority of cases.[78]  Transarticular screws, on the other hand, were typically not possible, given bony constraints.

Both the Harms and the Magerl technique offer rigid fixation, and postoperative bracing can be safely minimized. Both procedures are effective in patients with deficient posterior C1 arches. These techniques have significantly reduced the need for occipitocervical fusions in this patient population. Percutaneous approaches toward rigid C1-2 fixation have also been described. Holly et al reported that minimally invasive placement of C1-2 instrumentation was technically feasible.[65]

Rarely, symptoms and cord compression persist following posterior stabilization of an irreducible dislocation. In these cases, anterior decompression with removal of the os fragment is recommended through an anterior transoral or retropharyngeal approach.[79, 80] More typically, anterior lesions (eg, synovial cysts) regress following successful posterior stabilization.[58, 81]

Wu et al, in a study of 25 consecutive patients who had os odontoideum with atlantoaxial dislocation, compared the clinical results of posterior fixation and fusion with or without anterior decompression.[82] Sixteen patients with reducible atlantoaxial dislocation were treated with single-level posterior fusion and stabilization; the other nine were treated with this procedure in conjunction with transoral decompression. The authors found that patients with a reducible atlantoaxial dislocation were effectively treated with single-level posterior fusion and stabilization, whereas combined transoral decompression and posterior fusion and stabilization was preferable for those with an irreducible dislocation.

In the presence of progressive neurologic compromise, surgery is clearly indicated. With asymptomatic atlantoaxial hypermobility alone, on the other hand, the decision to proceed with surgery is more debatable. Larger series comparing outcomes between operative and nonoperative management may more clearly support surgery in some patients. Certainly, a better-defined discussion of surgical risks and benefits could be expected.

If surgery has been selected, the merits of each individual technique may be debated. The advent of C1 lateral mass and C2 pars screw fixation has addressed the major concerns with Magerl screw fixation. The Harms technique allows rigid fixation in a wider range of patients, including many with a degree of vertebral artery ectopy or incomplete reduction for which transarticular screw placement would be too risky.

In the future, the current open method of Harms screw placement may be replaced by a more percutaneous approach. Individual “proof of concept” case reports have been published. Until there is evidence proving that the less invasive approaches confer less operative risk, operative indications must not be relaxed.

Operative details

Preoperatively, critical steps include a complete understanding the lesion’s pathoanatomy and attempting reduction of any displacement. For example, during the approach, recognition of posterior C1 ring deficits not only limits surgical options, such as standard wiring, but also renders the cord and vertebral arteries vulnerable during the exposure.

If simple flexion-extension fails to reduce the os odontoideum, a period of skeletal traction may be initiated. Ideally, skeletal traction is used to reduce the atlantoaxial segment while the patient is awake. Hensinger recommended reduction several days preoperatively to decrease cord irritation.[68, 69, 70] One paper suggested 6 weeks of mobile traction for irreducible os odontoideum,[83] followed by occipitocervical fusion.

Halo traction can be used to maintain this positioning. If the reduction is to be performed in the operating room, do so with the patient alert either before or awake after fiberoptic intubation is performed. Some dislocations are irreducible. Displacement of the transverse atlantal ligament (TAL) in front of the ossicle may be an impediment. In these cases, consider fusion in situ rather than a risky operative reduction.

Upper cervical spine surgery in skeletally unstable patients is technically demanding, especially in the pediatric population. Meticulous attention to every detail in reduction, maintenance of reduction, and patient positioning is critical. Monitor somatosensory-evoked potentials (SSEPs) and motor-evoked potentials (MEPs) after intubation, after positioning, and at various stages intraoperatively. Rigidly fix the patient’s head to the operating table; avoid pressure on the eyes. Mayfield tongs or a halo ring with a Mayfield attachment of the Jackson table can be used for this purpose.

In children, limit the posterior exposure of the cervical spine to prevent inadvertent extension of the fusion to subjacent levels. Since most os odontoideum is unstable in extension as well as flexion, when applying fixation, avoid overtightening the segment. Overtightening of posterior wiring may lead to posterior translation of the C1 ring and the ossicle into the canal and against the cord.

Fluoroscopy may be useful to assess intraoperative motion of the affected segments. When the use of Magerl screws is attempted, image-guided surgery tools may be useful. Weng et al reported the use of three-dimensional (3D) fluoroscopy with Harms and Magerl techniques in the treatment of 19 patients with symptomatic os odontoideum.[45] Of 60 screws placed, only three slightly penetrated the transverse foramen, without clinical consequences.

Occipital fusion in children must include careful awareness of the thin occipital squama. In that the midline bone of the skull is thicker and stronger, plate and screw convergence toward the midline ensures better purchase.

The midline incision must be long enough to allow medialization of screws, or stab incisions can be made for screw passage. Typically, an incision extending at least 2 cm above and below the C2 spinous process is required. The C1 ring must not be exposed too laterally. The vertebral artery groove becomes shallow more than 1.5 cm from the midline, so only the inferior aspect of the ring should be exposed more laterally.[84]

A plexus of veins surrounds the C2 nerve root. Some authors recommend routine sacrifice of this nerve to help control bleeding and to provide better access to the C2-3 facet and the C1 lateral mass.[85]

Postoperatively, serial radiographs are obtained to ensure progression of fusion and maintenance of stability. Wound healing may be problematic in some settings and requires a layered closure with attention to careful matching of skin edges.


Complications have been reported after both operative and nonoperative management of os odontoideum. Rare, but concerning, are occasional reports of sudden death from cord compression in patients with unsuspected atlantoaxial instability.

In patients with known but stable os odontoideum, nonoperative management may be complicated by progression of instability. These changes may be merely radiographic or may signal that significant and irreversible neurologic problems may develop. In the literature, however, such irreversible neurologic decline is rare.

Once os odontoideum has been identified, these patients should be followed at regular intervals. Clear instruction as to the warning signs or “red flags” for myelopathy must be discussed with the patient. Klimo et al reported three patients with neurologic deterioration who were initially diagnosed and treated conservatively for os odontoideum. The authors concluded that the risk of late neurologic deterioration should be considered during patient counseling.[3]

The morbidity and complications associated with operative intervention vary widely in the numerous small series available. With traditional wiring techniques, failure of fusion is often reported. Pseudarthrosis rates vary in the literature but overall are quite low (0-4%).

Juhl et al reported on atlantoaxial fusion in six patients; the fusion rate was 100%, and no complications were noted.[86] If reduction cannot be obtained or the posterior C1 arch is not intact, an occipitocervical fusion may be performed. Dai et al reported on 33 patients with satisfactory results and a 100% fusion rate.[43] The ossicle may fuse to the base of the axis following successful posterior occipitocervical fusion.

In cases of known pseudarthrosis, revision with Magerl or Harms screws is highly effective. In the rare case in which C1 fixation cannot be achieved, an occiput to C2 fusion is indicated. In a case-control study, Taggard et al found that transarticular screw fixation was 21 times more likely to lead to solid fusion than wiring techniques in patients with atlantoaxial instability.[87]

Neurologic decline may occur after surgery from a number of insults, including drops in blood pressure, excessive motion during positioning, and direct neural tissue trauma. A higher risk for perioperative problems is reported in patients who are highly unstable and in those with fixed dislocations, ongoing cord compression, or inherited ligamentous laxity.

Neurologic injury may occur during implant placement. In Spierings and Braakman’s series of 17 operative cases, two patients died and one worsened neurologically. They reported a surgical morbidity and mortality rate of 18% (3/17).[42] In a series reported by Smith et al, one of 11 pediatric patients declined neurologically after sublaminar wire passage.[88]

Injury to the vertebral artery may occur during an overly wide exposure of the C1 ring in wiring or screw stabilization cases, but it is more likely during screw placement. If the vertebral artery is injured on one side, screw placement must not be attempted on the contralateral side. Overpenetration of bicortical screws anteriorly may result in injury to the internal carotid artery and cranial nerve XII.[88, 89] Less severe, but not insignificant, complications include vascular injury, posterior cervical wound infections, anesthesia complications, and ongoing muscular neck pain. Of these, injury to the vertebral arteries during screw placement is the most dangerous.