Practical Application of the AO Spine Primary Infection Classification System: Clarifying Definitions, Borderline Cases and Common Misinterpretations

Abstract

Study Design

Narrative review.

Objective

To clarify nuances when applying the AO Spine Primary Infection Classification.

Methods

The international validation of the AO Spine Primary Infection Classification System was successful but recurring themes caused confusion. This included the exact definition and reasons for infections to transition between categories, definition of canal involvement or neural compression and whether an epidural infection superimposed on degenerative disease escalates infection severity. There was uncertainty regarding handling of multi-level or junctional infection and clinicians encountered difficulty applying the modifiers.

Results

The difference between a Type 0 and Type 1 infection is reasonably distinct given the former is purely extraspinal whilst the latter involves spondylodiscitis or osteomyelitis without canal involvement. Type 2 infections include those with any canal involvement, regardless of whether CSF effacement occurs, in the absence of neural compression. The spinal canal is defined as the compartment dorsal to the posterior longitudinal ligament and excludes infections that are ventral to this ligament but bow into the canal. Any neural compression of the thecal sac centrally or within the foramen escalates categorization to a Type 3 infection.

Conclusions

Primary spinal infections have the potential to rapidly progress and cause devastating neurological deficits. The introduction of the AO Spine Primary Infection Classification establishes a universal language to communicate disease severity. We clarify common nuances encountered during international validation of this system including definition of canal involvement and relevance of CSF effacement, whether foraminal involvement constitutes neural compression and the nomenclature concerning multi-level or junctional disease.

Introduction

The management of spinal epidural abscesses is nuanced and multi-factorial in nature, with a combination of medical and surgical interventions demanding consideration. On one end of the spectrum is conservative management with antibiotics unless significant neurological deterioration occurs. ¹ This form of maximal medical management may involve broad-spectrum intravenous antibiotics, percutaneous drainage of collections by interventional radiology, and close monitoring of clinical and biochemical status. Proponents of this strategy argue that instituting close observation in the first instance avoids a proportion of patients undergoing unnecessary surgery. ² On the other hand, others contend that early surgical intervention for even non-compressive epidural abscesses is appropriate in order to minimize the risk of potentially devastating deficits such as cord compression or cauda equina compression. ³ Additionally, the purported benefit of reducing the load of the infection with resultant lower morbidity and mortality is an emerging tenet. ³

It is essential to have an established classification system, as in other fields of spine surgery, not only to enable clinicians to communicate in a reliable and consistent manner but also to facilitate accurate categorization of infections in future clinical trials. ⁴ The AO Spine Primary Infection Classification has been introduced to stratify spinal infections in a graded fashion. During internal validation by Delphi consensus and international validation by a survey of participating AO Spine members of this new system, we encountered several recurring areas of confusion requiring clarification. This included queries when attempting to classify multi-level or junctional disease, as well as how to handle superimposed degenerative disease. Furthermore, there was confusion regarding the definition of canal involvement and whether CSF effacement is required to progress to the next tier, as well as whether foraminal disease without canal involvement constitutes neural compression. We address these nuances of the AO Spine Primary Infection Classification system to assist clinicians in daily application.

Methods

A literature review was conducted to identify the current existing available classification systems to grade primary pyogenic spinal infections. We aimed to compare these to the proposed AO Spine Primary Infection Classification and specifically identify prognostic factors which escalate the severity of infections when managed by clinicians. Two independent authors (BTSK and JWT) searched the MEDLINE, OVID, Embase, Google Scholar and Cochrane Database of Systematic reviews from date of inception to June 2026 with the search terms ‘spinal infection’ AND ‘classification’ OR ‘system’ AND ‘outcomes’. Original cohort studies, literature reviews and systematic reviews were included. Exclusion criteria included secondary or atypical infections. The themes and controversies generated form the basis for this literature review.’ Institutional board review approval and informed patient consent was obtained to include publication of individual patient clinical images.

Background

Spinal infections are variable in nature and involve a spectrum of disease, ranging from asymptomatic paraspinal muscle collections to compressive multi-level abscesses causing neurological deficit. ⁵ Risk factors include diabetes, intravenous drug use, frailty and immunosuppressed states, which lead to either haematogenous or non-haematogenous spread.^6-9 Haematogenous arterial spread via the segmental arteries supplying the vertebral body is common, whilst seeding by the valveless venous network is more common from pelvic organs. ¹⁰ A feared consequence of conservatively managed spinal infection is progression of neurological deficit, which is due to a combination of direct compression and secondary arterial and venous compromise. ¹¹ Grandidge et al recommended close monitoring of inflammatory markers given a 37.5% rate of deterioration in conservatively managed patients. ¹²

The AO Spine Primary Infection Classification is designed to stratify pyogenic spinal infections in order of escalating severity and increasing likelihood of requiring operative intervention (Figure 1). This constitutes decompression at a minimum, with stabilization more likely to be required where there is bony destruction, which is signified in the 2+ and 3+ categories. Unlike other AO Spine classifications, there is regional division into lumbar (Type A), thoracic (Type B) and cervical (Type C) subtypes within this fully integrated system.^4,13-17

Figure 1.

AO Spine Primary Infection Classification System with the separation into lumbar (Type A), thoracic (Type B) and cervical (Type C) categories. Escalating types of infection are depicted with increasing degrees of potential for neurological compromise and mechanical instability

Nuances in Nomenclature of Multi-Level and Junctional Disease

A common source of confusion is the nomenclature applied when epidural abscesses or spondylodiscitis cross spinal regions and are junctional in nature. In this instance, the most cranial level is utilized to guide classification. For example, a spondylodiscitis at C7/T1 with a compressive collection causing neural compression at C7 but not T1 would be classified as a Type C3 infection, reflecting both the most cranial level involved and the most severe morphology present. The rationale behind this is threefold. Firstly, this prompts clinicians to recognize the most cranial level of neurological deficit and determine if there is clinical correlation. This is rational given it is cranial ascension of infections as well as the superior most level of infection involvement which potentially carries the greatest risk of neurological deficit. Secondly, labelling the most cranial extent of the lesion allows accurate tracking of the infection, with ascent being readily visible due to the classification nomenclature itself changing. In comparison, labelling the caudal most extent of epicentre of the lesion would not offer this same clarity and ability to observe the infection. Finally, this is consistent with established convention in the literature, such as that of the American Spinal Injury Association (ASIA) in which traumatic spinal cord injuries are classified by the uppermost level, and lends a degree of familiarity to clinicians learning this new classification.^18,19

Nuances in Differentiating Type 0 versus Type 1

The distinction between a type 0 and type 1 infection is often challenging. ²⁰ A Type 0 infection is purely extraspinal whereas a type 1 is spondylodiscitis or osteomyelitis without canal involvement. ²¹ Hence, type 0 infections may involve the spinous processes, transverse processes, or paravertebral musculature (Figure 2). Commonly associated psoas abscesses or soft tissue collections are included as a type 0 morphology. This is the most benign form but is not to be dismissed lightly as Zaki et al noted that they are bilateral in 23.3% of cases and still have the potential to spread into the spinal canal if not appropriately managed. ²² Indeed, more than two-thirds of cases eventually require drainage. ²² This represents a crucial lesson in that maximal conservative management does not necessarily completely obviate intervention, especially given radiologically guided drainage of large psoas or paravertebral collections may provide source control and organism identification with subsequent targeted antibiotics. ²³

Figure 2.

A Type B0 infection in which there is an extra-spinal focus of infection demonstrated on T2-weighted magnetic resonance imaging in the sagittal (A) and axial (B) planes. There is T2 hyperintensity in the spinous processes (blue arrow), interspinous ligaments and adjacent soft tissue (red arrow) without any disc or canal involvement

As a more severe distinct recognized entity in the AO Spine Primary Infection Classification, Type 1 infections are osteomyelitis with or without discitis, in the absence of canal involvement. This may manifest clinically as significant lower back pain or even referred radicular pain from neural irritation without compression, but despite these symptoms there is preservation of neurological status given the lack of involvement of neural elements. Computed tomography (CT) demonstrates intervertebral disc enhancement with contrast and soft tissue swelling. ²⁴ Magnetic resonance imaging (MRI) may demonstrate low signal on T1-weighted sequences within the disc space and vertebra consistent with fluid and bone marrow oedema respectively. ²⁵ T2-weighted images confirm hyperintensity within the disc space and adjacent endplates, and may also demonstrate hyperintensity within the adjacent psoas musculature. ²⁵ In fact, Ledbetter et al found that hyperintensity within the psoas, labelled the ‘imaging psoas sign’, significantly correlated with spondylodiscitis and was 92% sensitive as well as 92% specific. ²⁶

The natural history of spondylodiscitis is progression from type 1 to type 1+ in which there is bony destruction. ²¹ The discs are usually avascular in a normal healthy individual, with osmotic diffusion utilized to supply nutrients. Infection usually begins in the anterior portion of the vertebral body adjacent to the endplate where blood supply is richest. As a result, the earliest signs may be rounding of the anterior vertebral body on radiographs, with subsequent dorsal spread to the adjacent disc and vertebra resulting in loss of disc height and gradual osteolytic erosion of the subchondral plates. ²⁴

Accordingly, type 1+ includes bony destruction of endplates. ²⁷ The significance of this is that infection is more likely to spread to the foramen or canal and thus cause significant neural compression with progression to deformity. It is important to recognize, therefore, that rarefaction of the endplates is still type 1, whereas progressive osseous destruction with or without pathological fracture represents the more severe type 1+.^28-30 This modifier of ‘+’ represents a source of confusion at times but is independent of the morphology of infection and may be added to any of Types 1 to 3 to denote osseous destruction.

Nuances in Differentiating Type 1 versus Type 2

Definition of Canal Involvement

In the new AO Spine Primary Infection Classification, the difference between type 1 infections and type 2 infections is sparing or involvement of the canal respectively. This remains a common source of ambiguity and is responsible for heterogeneous reporting in the literature.^31,32 This scenario is exemplified in Figure 3. Once spondylodiscitis breaches the canal but does not cause neural compression, this is classified as a type 2 subtype. The canal is defined as being dorsal to the posterior longitudinal ligament (PLL) and vertebral bodies, anterior to the laminae and medial to the pedicles in keeping with standard anatomical definitions. ³³ Classification as Type 2 is based on involvement of the spinal ³⁴ canal, regardless of whether there is cerebrospinal fluid effacement or thecal indentation. Only central canal rather than foraminal narrowing is considered when classifying whether an infection is type 1 versus type 2 (Figure 4).

Figure 3.

An example of a Type A1 infection on T2-weighted magnetic resonance imaging in the sagittal (A) and axial (B) planes which represents an important demonstration of two key concepts. Firstly, this demonstrates the importance of recognizing the focus of the infection is in the disc space and vertebral body with extension into the compartment ventral to the posterior longitudinal ligament (red arrow) and therefore not considered to involve the spinal canal. Secondly, superimposed degenerative disc disease (blue arrow) causes some mild narrowing of the central canal in the lumbar spine but the infection does not contribute to this

Figure 4.

Type A2 infection illustrated on T2 weighted magnetic resonance imaging in the sagittal (A) and axial (B) planes. There is a dorsal T2 hyperintense epidural collection (blue arrow) spanning L3/4 to L5/S1 which is seen to involve the spinal canal with some narrowing of the central canal but without neural compression (red arrow)

Foraminal Involvement

A recurrent theme among clinicians attempting to apply this classification was whether compression of the nerve root, but not the central canal containing the spinal cord or cauda equina, still constituted a Type 3 infection. In this system, neural compression with foraminal narrowing is categorized as a type 3 infection (Figure 5). Separately, there may be underlying degenerative disease from ventral disc prolapse, ossified PLL or flavum hypertrophy causing canal narrowing. However, infections remain classified as type 1 unless the canal is breached secondary to the actual infective process. A type 2+ morphology is defined when there is canal involvement combined with osteolytic bony destruction.

Figure 5.

Type C3 infection demonstrated on T2 weighted magnetic resonance imaging of the cervical spine in para-sagittal (A) and axial (B) planes. There is evidence of a left sided discitis extending into the foramen ventrally (blue arrow) with resultant narrowing (red arrow) and left C3/4 nerve root compression. Even in the absence of central canal narrowing, the presence of neural compression upgrades this to a Type 3 infection

Nuances in Differentiating Type 2 versus Type 3

Definition of Neural Compression Independent of Canal Diameter

Another common challenge encountered by clinicians is differentiating type 2 infections with canal involvement without neural compression from the type 3 subtype, which is defined by the presence of neural compression. The AO Spine Primary Infection Classification has deliberately defined this transition by the presence or absence of neural compression, rather than by a quantitative percentage of canal compromise as is often seen in radiological grading systems, to ensure there is clinical relevance for three main reasons.

Firstly, it is well known that canal diameter is widely variable. Lee et al noted that spinal canal dimension and canal-to-body ratios have moderate correlation, whilst Vanharanta et al have documented significant variation in inter-articular distance.^35,36 Chazono et al also confirmed there is variation by ethnicity. ³⁷ There is even generational variation, with spinal canal diameters being smaller today than in the 1940s, as Tobin et al noted. ³⁸ Even when utilizing advanced imaging, Lotan et al found that MRI actually underestimated canal size compared to CT and thus recommended multi-modality analysis in this cohort of patients. ³⁹ Rudimentary and arbitrary thresholds have been proposed, such as that of Morishita et al, who found that a cervical spinal canal diameter of less than 13mm was more likely to translate into cord compression. ⁴⁰ It is evident from the work of Baştuğ et al, however, that such a universal rule is not practical, given this figure is likely to vary across the upper, mid and lower subaxial cervical spine and even further caudally as the spinal canal widens in the lumbosacral region itself.^41,42

Secondly, it is implied by the work of Geisser et al in patients with degenerative lumbar spinal canal stenosis that spinal canal diameter itself is not predictive of clinical symptoms. ⁴³ This is rationally explained by Frostell et al who demonstrated that spinal cord diameter itself is also variable. ⁴⁴ A superior predictor of neurological deficit was noted by Montes-Aguilar et al who found that the extent of neural element or spinal cord compression was more predictive in traumatic spinal cord injury.^45,46 Examining specifically spinal tuberculosis infections, Sae-Jung et al arrived at a similar conclusion, in that the predictive factor for neurological deterioration was cord signal change rather than absolute or relative canal narrowing. ⁴⁷ It is for this reason that the AO Spine Primary Infection Classification has elected not to utilize spinal canal diameter narrowing as the means of distinguishing Type 2 and Type 3 infections. More useful is the presence or absence of neural compression, particularly where a multi-compartmental infection with ventral and dorsal components is noted, in which the components in combination cause cord compression but individually may not qualify as a Type 3 rather than Type 2 infection (Figure 6).

Figure 6.

T2-weighted sagittal (A) and axial (B) sequences illustrate a Type B3 infection. However, a crucial concept in the AO Spine Primary Infection Classification System is demonstrated. In isolation, the ventral component centred predominantly on the disc space and vertebral body (blue arrow) would be considered a Type B2 infection with mild narrowing of the canal. In combination with the relatively thick dorsal epidural collection (red arrow), however, this causes circumferential cord compression and therefore the entire infection is upgraded to Type B3

Finally, defining this transition by the presence of compression of the neural elements takes into consideration all structures that may be functionally limiting. These range from the spinal cord and conus medullaris to the cauda equina and nerve roots. Pre-existing degenerative disease may arise for a variety of reasons and may itself lead to canal narrowing and even cord compression, with patients having a resultant lowered reserve in both a mechanical and vascular sense. Nonetheless, any infective insult that leads to new neural compression is classified as a type 3 infection. The threshold at which an infective load causes a neurological deficit may therefore be lower, as Dora et al noted, but this is accounted for by utilizing the AO Spine Primary Infection Classification. ⁴⁸

A multitude of reasons exist as to why there is yet to be an evidence-based algorithm guiding clinicians in this area. ¹¹ The majority of studies conducted have been retrospective rather than prospective in nature.^3,49,50 This introduces significant selection bias, given that clinicians who opted for surgical intervention would almost certainly have selected only patients deemed fit for surgery with a clear indication for decompression with or without stabilization. Interestingly, this observation has the potential to bias findings in either direction. The neurological and functional outcomes for these patients undergoing spinal surgery for epidural abscesses may be poorer given their poor neurological baseline. ³

Conversely, the outcomes for these patients in some instances may be superior to medical management, given that patients who are deteriorating yet are not deemed robust enough to undergo surgery are also likely to be significantly impaired. ⁵¹ This interesting dichotomy translates into ambiguity when attempting to interpret the literature. For example, Nabizadeh et al examined 3,197 patients with pyogenic spondylodiscitis. ⁵² Those with more severe Frankel A or B deficits (22%) had the lowest likelihood of complete recovery (16%). ⁵² In comparison, those with less severe Frankel D deficits (21%) were more likely to make a normal recovery (53%). ⁵² This review comprised exclusively retrospective data, however, and is therefore susceptible to the aforementioned biases. ⁵²

Future Advancements in Spinal Infection

There are a number of other increasingly recognized risk factors for progression of epidural abscesses which future iterations of the AO Spine Primary Infection Classification may consider. In patients with spinal epidural abscesses without neurological deficit, diabetes mellitus and female sex were identified conferring a higher risk of delayed deterioration. ⁵³ Additionally, Staphylococcus aureus is the most frequently implicated organism in spondylodiscitis, with the methicillin-resistant rather than methicillin-sensitive strain having been noted to be a negative prognostic factor for severe neurological deficit. ⁵⁴ Sircar et al therefore suggested that early surgical treatment may be reasonable for patients who are neurologically intact but possess these confirmed negative prognostic factors. ⁵³ Future versions of the classification system and the eventual proposed treatment algorithm may assign these high risk features additional weighiting towards surgery given the tendency for failure with medical management.^53,54

Furthermore, Khanh attempted to risk stratify these infections into primary infections, existing only within the epidural space, and secondary infections, which are associated with osseous or muscular involvement. ⁵⁵ This distinction is crucial given primary cases demonstrate much more favourable outcomes than secondary ones. ⁵⁵ Finally, this system currently examines only extradural infections involving extrinsic compression of the central thecal sac or nerve roots, rather than the rarer intradural infections. This entirely discrete category of non primary pyogenic spinal infections may be considered and incorporated into subsequent iterations of the AO Spine Primary Infection Classification.

Conclusion

Spinal infections are devastating and accelerate neuronal death in a multi-factorial but progressive manner. ⁵⁶ The introduction of the AO Spine Primary Infection Classification as a universal means of communicating and documenting severity of infections is an important landmark in enabling the design of future trials. Just as trauma requires prevention of secondary spinal cord injury through neuroprotective measures, equal vigilance is required in caring for patients with spinal infection. This involves regular clinical, biochemical and radiological assessment, with the need for operative intervention being carefully and continuously evaluated. We clarify common nuances and sources of confusion encountered during international validation of this system, including definition of canal involvement, need for CSF effacement, whether foraminal involvement constitutes neural compression and also nomenclature concerning multi-level or junctional disease.

ORCID iDs

Barry Ting Sheen Kweh https://orcid.org/0000-0001-9130-1543

Mohamed Aly https://orcid.org/0000-0001-5221-8288

Mohammad El-Sharkawi https://orcid.org/0000-0001-6177-7145

Charlotte Dandurand https://orcid.org/0000-0002-4520-6454

Brian Karamian https://orcid.org/0000-0003-0512-6019

Richard Bransford https://orcid.org/0000-0003-2293-5774

Andrei Fernandes Joaquim https://orcid.org/0000-0003-2645-0483