Each year, spine injuries account for 6% of all fractures [4, 9, 14]. Of those, 30% to 60% involve the thoracolumbar spine [3, 4, 6, 21]. The presentation of these injuries can vary based on the fracture pattern, stability, and associated neurologic injury. Because managing these spinal fractures is often predicated on these factors, a classification system that incorporates these variables is important to provide a common language for communicating plans. However, establishing a classification system with universal application and utility has been challenging. The AO thoracolumbar system [20 ] is the most current classification system, which expands on previous systems such as the Thoracolumbar Injury Classification System [19 ], to provide a universal language for communicating about thoracolumbar injuries. However, even with a common algorithm for communicating about fractures, this system still has limitations in achieving high levels of agreement among providers in classifying injury. These limitations must be recognized before it can be widely clinically applied.
To systemize the discussion of spine injuries, in 1970, Holdsworth  described a classification system based on the mechanism of injury and the concept of a two-column spine construct, with an anterior and posterior column resisting compression and tensile forces, respectively. Using CT imaging to build on this classification, McAfee et al.  developed a system based on injury patterns that correlated with spinal stability, starting with more stable wedge compression fractures to more unstable translational injuries. Because of CT imaging’s ability to provide greater detail regarding fracture patterns, a more versatile classification system was needed. Subsequently, Denis  elaborated on the two-column theory and proposed a three-column model in which the anterior column includes the anterior longitudinal ligament, anterior vertebral body, and the annulus; the middle column includes the posterior half of the vertebral body, the annulus, and the posterior longitudinal ligament; and the posterior column contains the neural arch, facets, and posterior ligamentous complex. In this model, an unstable injury involves two or more columns.
In 1994, when the original classification system was developed by Magerl et al. , the focus of the classification systems shifted to include morphology and fracture patterns. It included three groups of fractures: compression (Type A), distraction (Type B), and torsional injuries (Type C), which were divided into nine subcategories and further divided into 25 different subcategories. The intricacy of the original AO thoracolumbar classification made it cumbersome, and it was poorly used by the medical community . Similar to its predecessors, it did not integrate neurologic status, recognize soft tissue injury, or help guide clinical management. Nevertheless, the system became the basis for a morphologic categorization of injury. Around the same time, another classification was developed by McCormack et al. , which also looked vertebral body morphology to help determine the failure rate of posterior arthrodesis. Identifying the need for a simpler classification system that incorporated neurologic status and soft-tissue status to help guide treatment, Vaccaro et al. [18, 19] proposed the Thoracolumbar Injury Classification System (TLICS), which assigned a cumulative numerical score based on fracture patterns, neurologic status, and the integrity of the posterior ligamentous complex. The focus of this review will be on the AO thoracolumbar classification created by the AOSpine forum with the intention of simplifying the Magerl system and supplementing it with clinical attributes alluded to in the TLIC system.
Numerous classification systems for thoracolumbar injuries have been proposed and have focused on the mechanism of injury, morphologic patterns, and neurologic impairment. McAfee et al. , Denis , and Magerl et al.  established classification schemes that have built on one another and focused on injury patterns, and Vaccaro et al.  incorporated neurologic and soft-tissue status into a system called Thoracolumbar Injury Classification System to produce a score to guide management.
However, the TLICS was also met with several criticisms , which included poor reproducibility in properly evaluating the integrity of the posterior ligamentous complex using MRI and the inclusion of PLC injury, which can inflate the aggregate score. Also, the severity scoring system guiding treatment may be influenced by different factors such as culture or region and may not reflect global surgical preferences or an agreed approach to treatment. Therefore, there was still a need for a classification system that was comprehensive but simple and reliably reproducible. The AOSpine forum revised the original AO thoracolumbar classification system to address these needs. The AO thoracolumbar classification system incorporates morphologic and clinical factors including neurologic status to produce a tier-based classification system based on increasing severity.
The AO thoracolumbar system classifies injuries based on morphologic patterns of fractures, neurologic status, and clinical modifiers. Morphologic patterns are grouped based on the failure mode of the spinal column, with Type A involving compression fractures, Type B injuries involving failure of the anterior or posterior tension band without gross translation, and Type C injuries including dislocation or rotational or displacement-type injuries. Within each type, there are subgroups that are based on increasing severity and instability. An A0 fracture is an insignificant fracture of the spinous or transverse processes, and A1 fractures are wedge compression of single endplate without violation of the posterior vertebral wall. A2 injuries include both endplates with a split pattern, without violation of the posterior vertebral wall. A3 injuries include vertebral fractures affecting a single endplate with violation of the posterior vertebral wall but no disruptions of the posterior tension band. A4 injuries are similar to A3 injuries, except both endplates are involved (Table 1).
Subtypes of B injuries include B1, which is monosegmental osseous failure of the posterior tension band extending from the spinous process through the pedicles and into the vertebral body, classically referred to as a “chance” fracture. B2 injuries consist of an osseoligamentous injury, which often propagates through an intervertebral level and disrupts the posterior tension band, with or without bony involvement. B3 injuries consist of hyperextension injuries that cause violation of the anterior tension band. These injuries are more common in patients with an ankylosed spine (Table 2).
Type C injuries do not have any subtypes because they are unstable injuries that lead to surgical stabilization. Therefore, dividing Type C fractures into subtypes would add unnecessary complexity to the classification system and would not alter clinical management (Table 3).
For injuries that have combined mechanisms, each injury can be classified separately, with the more severe injury written first. For example, an injury that involves a T5-T6 failure in tension of the posterior tension band and a T6 compression fracture involving one endplate may be written as T5–T6 B2; T6 A1 injury.
The second component of the classification focuses on neurologic deficits, with each increase in level corresponding to higher levels of deficits. N0 is neurologically intact, N1 involves transient neurologic deficits that are no longer present, N2 represents radicular symptoms, N3 represents cauda equina or an incomplete spinal cord injury, and N4 is a complete spinal injury equivalent to American Spinal Injury Association Grade A. NX refers to an unknown neurologic status because of sedation or a head injury and a “+” is designated for continued spinal cord compression (Table 4).
The third factor is a case-specific modifier intended to help with the treatment decision. M1 is used to designate fractures with indeterminate injuries to tension bands based on imaging or a physical examination, and M2 represents comorbidities that may affect indications for surgery such as ankylosing spondylitis, skin injuries overlying a fracture, or rheumatologic conditions (Table 5).
To increase efficacy and allow for consistent communication of management plans, a classification system must be simple to learn and accurately reproducible among different clinicians. The AO group that revised the classification system demonstrated among five surgeons reviewing 110 fractures that there was an 83% agreement about the fracture type, with an overall κ coefficient of 0.77 . Type-specific interobserver κ coefficients were 0.81 for Type A, 0.71 for Type B, and 0.81 for type C, with kappa values 0.60 to 0.79 showing moderate level of agreement and 0.80 to 0.90 showing strong agreement. The group then repeated the evaluation with spinal surgeons who were using the classification system for the first time to demonstrate applicability at all levels of training. With 100 spinal surgeons worldwide reviewing 25 patients, there was overall weak interobserver reliability (overall κ coefficient = 0.56) with moderate-to-significant reliability by types (Type A: κ = 0.80; Type B: κ = 0.68; type C: κ = 0.72) . Intraobserver reproducibility was moderate, with κ = 0.68.
One study found that surgeons agreed on the AO thoracolumbar classification for 60% of patients (24 of 40 patients), producing an overall κ value of 0.72 . A further analysis demonstrated that the highest agreement in classification was for Type A injuries (κ = 0.72), followed by Type C (κ = 0.7) and Type B injuries (κ = 0.58). The same study examined intraobserver reliability and showed high reproducibility for the overall fracture classification (κ = 0.77, range 0.6 to 0.97).
Other groups of independent researchers also sought to determine the reproducibility of the new AO classification system. Urrutia et al.  had six reviewers analyze 70 patients, with 54% agreement (κ = 0.62). Similarly, there was moderate reliability for each fracture group (Type A: κ = 0.62; Type B: κ = 0.57; Type C: κ = 0.69). When the fractures were divided by subtype, reproducibility slightly decreased, with a κ value of 0.55. Overall, intraobserver reproducibility had κ values of 0.77 and 0.71 when fractures were divided into subgroups. That study also found no differences in reliability between attending- and resident-physician grading.
Cheng et al.  reviewed 109 patients with thoracolumbar injuries and reported reliability based on six orthopaedic surgeons’ reviews. Overall, the interobserver κ coefficient for all patients was 0.362. When injuries were divided into subgroups, the κ coefficients were 0.385 for Type A injuries, 0.292 for Type B injuries, and 0.552 for Type C injuries. Intraobserver comparisons showed moderate reproducibility for all types of injuries (κ = 0.442 for Type A, κ = 0.485 for Type B, and κ = 0.412 for Type C injuries). These values suggest that there were discrepancies in the surgeon’s classification of an injury more than half the time.
To compare the reproducibility of the AO thoracolumbar classification and the Thoracolumbar Injury Classification System, 11 spine surgeons from six different programs were asked to review 50 fractures and classify them according to both classifications . The AO thoracolumbar classification had moderate interrater (κ = 0.59) and substantial intrarater (κ = 0.68) reliability for grading the fracture type compared with fair interobserver (κ = 0.29) and moderate intraobserver (κ = 0.44) reliability for the Thoracolumbar Injury Classification System. Both systems had substantial interrater and intrarater agreement for neurologic involvement (Thoracolumbar Injury Classification System: intrarater κ = 0.90; interrater κ = 0.85; AO thoracolumbar: intrarater κ = 0.91; interrater κ = 0.85) .
Although the AO classification addresses the need for a simplified yet comprehensive system, there are limitations in its interrater reliability, and its critics claim that it falls short in guiding clinical management. As some reliability studies have shown thus far, there are still challenges in achieving high levels of agreement in identifying these fractures using the AO system. Since the classification of morphology is often done by looking at static images, morphology can appear similar on imaging even though mechanism of injury is actually different. For example, a flexion distraction injury (Type B2) that also involves a compression injury (Type A) should be classified as Type B fracture with Type A component. However, a posterior tension band injury could sometimes go unnoticed and then the fracture would be classified incorrectly as a Type A. Examples like this in turn can contribute to differences in interrater reliability for the classification. For this reason, using this classification system to communicate and plan for management must be done prudently because there is still a period of adjustment and familiarization with how to classify these fractures, which may lead to variable interventions.
Because the classification system is still being implemented, there is still a need to address its feasibility in guiding clinical management. Multiple research groups are implementing the classification system to determine if there are clinically important differences in outcomes. More work is needed to assess the efficacy of this system in helping clinicians decide on indications for nonsurgical versus surgical management.
The revised AO thoracolumbar classification distills a previously complex system into a simpler structure that enables providers to communicate using the same language. Compared with the TLIC, the AO thoracolumbar system enhanced reproducibility yet maintained neurologic status and patient comorbidities for clinical applicability. With its types and subtypes, the AO classification system also depicts the morphology of the fracture more clearly than the TLIC system, which uses a composite score. However, the use of the AO thoracolumbar system is still being standardized, and there is still groundwork necessary to improve on the interrater reliability. Although more education and consistency in defining the components of the classification system is needed, there is potential for application in daily clinical management. Currently, prospective studies are required to evaluate the efficacy of this classification system in appropriately treating injuries.
We thank Arien L. Cherones of the Department of Orthopaedics and Sports Medicine, University of Washington for his assistance in preparing illustrations for this manuscript.
1. Cheng J, Liu P, Sun D, Qin T, Ma Z, Liu J. Reliability and reproducibility analysis of the AOSpine thoracolumbar spine injury classification system by Chinese spinal surgeons. Eur Spine J. 2017;26:1477-1482.
2. Denis F. The three column spine and its significance in the classification of acute thoracolumbar spinal injuries. Spine (Phila Pa 1976). 1983;8:817-831.
3. Gertzbein SD. Scoliosis Research Society. Multicenter spine fracture study. Spine (Phila Pa 1976). 1992;17:528-540.
4. Hasler RM, Exadaktylos AK, Bouamra O, Benneker LM, Clancy M, Sieber R, Zimmerman H, Lecky F. Epidemiology and predictors of spinal injury in adult major trauma patients: European cohort study. Eur Spine J.
5. Holdsworth F. Fractures, dislocations, and fracture-dislocations of the spine. J Bone Joint Surg Am. 1970;52:1534-1551.
6. Hu R, Mustard CA, Burns C. Epidemiology of incident spinal fracture in a complete population. Spine (Phila Pa 1976).
7. Kaul R, Chhabdra HS, Vaccaro AR, Abel R, Tuli S, Shetty AP, Das KD, Mohaptra B, Nanda A, Sangondimath GM, Bansal ML, Patel N. Reliability assessment of AOSpine thoracolumbar spine injury classification system and Thoracolumbar Injury Classification and Severity Score (TLICS) for thoracolumbar spine injuries: results of a multicentre study. Eur Spine J.
8. Kepler CK, Vaccaro AR, Koerner JD, Dvorak MF, Kandziora F, Rajesekaran S, Aarabi B, Vialle LR, Fehlings MG, Schroeder GD, Reinhold M, Schnake KJ, Bellabarba C, Cumhur Oner F. Reliability analysis of the AOSpine thoracolumbar spine injury classification system by a worldwide group of naive spinal surgeons. Eur Spine J.
9. Laurer H, Maier B, El Saman A, Lehnert M, Wyen H, Marzi I. Distribution of spinal and associated injuries in multiple trauma patients. Eur J Trauma Emerg Surg.
10. Magerl F, Aebi M, Gertzbein SD, Harms J, Nazarian S. A comprehensive classification of thoracic and lumbar injuries. Eur Spine J.
11. Magnusson E, Spina N, Fernando ND. Classifications in Brief: The Thoracolumbar Injury Classification. Clin Orthop Relat Res.
12. McAfee PC, Yuan HA, Frederickson BE, Lubicky JP. The value of computed tomography in thoracolumbar fractures. An analysis of one hundred consecutive cases and a new classification. J Bone Joint Surg Am. 1983;65:461-463.
13. McCormack T, Karaikovic E, Gaines RW. The load sharing classification of spine fractures. Spine (Phila Pa 1976) 1994;19(15):1741–1744.
14. Oliver M, Inaba K, Tang A, Branco BC, Barmparas G, Schnuriger B, Lustenberger T, Demetriades D. The changing epidemiology of spinal trauma: a 13-year review from a Level I trauma centre. Injury.
15. Oner FC, Ramos LM, Simmermacher RK, Kingma PT, Diekerhof CH, Dhert WJ, Verbout AJ. Classification of thoracic and lumbar spine fractures: problems of reproducibility. A study of 53 patients using CT and MRI. Eur Spine J.
16. Reinhold M, Audige L, Schnake KJ, Bellabarba C, Dai LY, Oner FC. AO spine injury classification system: a revision proposal for the thoracic and lumbar spine. Eur Spine J.
17. Urrutia J, Zamora T, Yurac R, Campos M, Palma J, Mobarec S, Prada C. An independent interobserver reliability and intraobserver reproducibility evaluation of the new AOSpine Thoracolumbar Spine Injury Classification System. Spine (Phila Pa 1976). 2015;40:E54-E58.
18. Vaccaro AR, Baron EM, Sanfilippo J, Jacoby S, Steuve J, Grossman E, DiPaola M, Ranier P, Austin L, Ropiak R, Ciminello M, Okafor C, Eichenbaum M, Rapuri V, Smith E, Orozco F, Ugolini P, Fletcher M, Minnich J, Goldberg G, Wilsey J, Lee JY, Lim MR, Burns A, Marino R, DiPaola C, Zeiller L, Zeiler SC, Harrop J, Anderson DG, Albert TJ, Hilibrand AS. Reliability of a novel classification system for thoracolumbar injuries: the Thoracolumbar Injury Severity Score. Spine (Phila Pa 1976). 2006;31:S62–S69.
19. Vaccaro AR, Lehman RA Jr, Hurlbert RJ, Anderson PA, Harris M, Hedlund R, Harrop J, Dvorak M, Wood K, Fehlings MG, Fisher C, Zeiller SC, Anderson DG, Bono CM, Stock GH, Brown AK, Kuklo T, Oner FC. A new classification of thoracolumbar injuries: the importance of injury morphology, the integrity of the posterior ligamentous complex, and neurologic status. Spine (Phila Pa 1976). 2005;30:2325-2333.
20. Vaccaro AR, Oner C, Kepler CK, Dvorak M, Schnake K, Bellabarba C, Reinhold M, Aarabi B, Kandziora F, Chapman J, Shanmuganathan R, Fehlings M, Vialle L; AOSpine Spinal Cold Injury & Trauma Knowledge Forum. AOSpine Thoracolumbar Spine Injury Classification System: fracture description, neurological status, and key modifiers. Spine (Phila Pa 1976).
21. Wang H, Zhang Y, Xiang Q, Wang X, Li C, Xiong H, Zhou Y. Epidemiology of traumatic spinal fractures: experience from medical university-affiliated hospitals in Chongqing, China, 2001-2010. J Neurosurg Spine.