1 Development and validation of a new generation for spine injury classification

Spine fracture classification has come a long way since Böhler’s first systematic allocation of spinal fractures [1] and Nicoll’s simple separation of stable vs. unstable fractures [2]. Over more than half a century a contribution to the understanding of spinal fractures has been made by outlining the importance of the mechanism of injury [3] and developing different mechanistic concepts (two [4]- and three column concepts [5]) and, especially, by Magerl’s definition of three basic, progressive, pathomorphological criteria: (A) compression, (B) distraction and (C) axial torque. The concept of fracture subgroups, eg, the 3-3-3 scheme of AO fracture classi-fication [6] has also made an important contribution.

Still, most classification systems have problems in sufficient-ly subsuming all entities and aspects of spinal fractures ranging from simple isolated vertebral fractures to the most complex multilevel thoracolumbar injury and concomitant neurovascu-lar lesions. Therefore, none of them is ideal or generally accepted because of a lack of information about the severity of the injury, insufficient therapeutic indications, prognostic value or impracticality in terms of accurate communication between treating physicians and researchers [5–10].

The Denis and Magerl (AO) Classifications are widely used but both are based on retrospective, individual studies of thora-columbar injuries in more or less representative study populations [5, 6], however, without a well defined methodological concept or scientific validation (11). Their moderate reliability and the tendency for well-trained spine surgeons to classify the same fracture differently on repeat testing is a matter of some concern [9].

The need for a valid and universally accepted spine injury clas-sification system motivated AOSpine International to establish in 2007 an AOSpine Classification Group (AOSCG) under guidance by the AO Classification Supervisory Committee (CSC) and management by AO Clinical Investigation. In this chapter we highlight the objectives of the AOSCG, as well as the driving methodological and clinical issues.

2 Methodological pathway

Injury classification systems are useful for documentation, such as registries, and for valid comparison of results between clinical studies. In clinical settings, they ideally help surgeons in making their decision on the most appropriate treatment modality, provided classification categories reflect injury severity and include prognostic factors for clinically relevant patient outcomes [12]. In traumatology in general, most published reliability studies, showed poor inter-rater reliability of commonly used classifications [13]. In spine in particular, reliability studies also reported poor to moderate results, such as for the AO classification system [7–9], the Denis system [8, 9], the Dens system [14] or a proposed modification [15]. Valid classific at ion s y s tem s shou ld not on ly b e good i n c l i n ica l ter m s (face and content validity), but also be reliable and accurate, and be related to relevant patient outcomes (construct validity) [11, 16, 17]. Hence it is becom ing a standard that injur y c lassi-fication systems be validated before they are promoted for use in practice and research [13, 18–20]. New ideas are welcome [21], but their clinical and scientific value at best remain uncertain without proper evaluation.

In the development and validation of classification systems, Audigé et al [11] suggested implementing a methodological pathway with three successive research phases (Fig 6.7-1). The first development phase involves clinical experts developing proposals for the classification system, as well as defining the classification process (eg clinical information, image modalities, measurement aids, training of surgeons, etc.). This phase defines a common language with which surgeons should be able to see and describe injuries similarly. Successive pilot agreement studies are conducted to ensure that clinical experts can do this. At this stage the predictive clinical value of any proposed system is at best suspected by expert opinion, and its quantification would remain somehow speculative.

The second phase involves a multicenter agreement study to ensure that future users with less clinical experience can also do so. This creates the basis for a classification tool to be used for documentation and evaluation of treatment options. Only after these first two phases have been completed can recommendations for patient care based on the classification be developed in a third phase, following the implementation of a prospective clinical study.

This methodological pathway has been implemented successfully within all classification projects supported by the AO Foundation through its CSC, in particular for the development of the AO pediatric long-bone fracture classification system [22, 23].

3 AOSpine Classification Group

The AOSpine Classification Group (AOSCG) is composed of a core membership of experienced spine surgeons from various geographical and cultural environments. This facilitates iden-tification of cross-cultural differences in training and understanding of basic clinical concepts and definitions and builds upon the strength of AOSpine as a truly global and multispe-cialty spine organization. In addition, an extended AOSCG is being built with all AOSpine surgeons who may contribute to the development process, such as participating in one of the implemented classification sessions (agreement studies) or providing feedback on the clinical relevance of the proposed system. Broad international acceptance of the developed system by spine surgeons is a key targeted outcome of this project.

From a clinical viewpoint, the current development process by the AOSCG is based on the following criteria:

• 
  

  Addresses traumatic spinal injuries only

  
  
• 
  

  Includes the whole spine and identifies four anatomical areas:

  
  
  
  
  • 
    

    – C0–C2 upper cervical spine (UCS),

    
    
  • 
    

    – C2/3–C7/T1 subaxial cervical spine (SACS),

    
    
  • 
    

    – C7/T1–L1/S1 thoracic and lumbar spine (TLS),

    
    
  • 
    

    – L5/S1–S5 sacrum (S)

  
  
• 
  

  Includes soft tissue injuries (to be considered as a separate code)

  
  
• 
  

  Considers a hierarchical system from very simple for all injuries to more detailed

  
  
• 
  

  Describes injury location and morphology based on Xrays and/or CT scans (MRI may be added)

  
  
• 
  

  Is simple, reliable, helps treatment choice and provides prognostic information

  
  
• 
  

  Reaches international acceptance

The first phase of development and validation is anticipated to be long and resource-intensive. Because it uses an iterative process, it is difficult to predict how many validation-and-re-view iterations will be required to achieve broad consensus, understanding of definitions and reliability among experienced surgeons. With simple systems however two to three years are anticipated.

A first proposal was drafted regarded the SACS and TLS areas. The first classification session was completed, whereby 7 surgeons classified 115 consecutively-collected cases from one hospital. Preliminary results highlighted important definition issues such as the distinction between single and multi-level injuries, the identification of the i njur y level(s), the d istinc tion between vertebral body and inter-vertebral body injuries, and a few other definitions about fracture morphology. The aim of the next round will be to clarify the reasons for discrepancies, and reach higher classification reliability.

4 Classification system versus severity score

A classification system implies the use of categories. Like any diagnostic test, these categories might be related to categorical items (eg, presence/absence of a fracture) or derived from continuous item par a meter s (eg , degree of fracture d is placement). In the later context, the choice of the most relevant cut-off value(s) in the definition of the categories must be addressed. Several categorized diagnostic items can be combined to form a classification system. Most fracture classification systems are built on items describing fracture location and morphology. The resulting categories might be indicative of some specific injury characteristics such as severity, ie, they can reliably distinguish groups of injuries that differ regarding the complexity of their treatment, or the quality of their outcome(s). In the development of a classification system is it of utmost importance to identify the most relevant items and therefore define the purpose of the system.

An injury severity score implies the use of a continuous diagnostic process that has a predictive value for treatment option and/or outcome(s). Similarly to a classification system, such a score is built on the basis of several diagnostic items. For instance Vaccaro et al [19, 24, 25] developed a Thoracolum-bar Injury Classification and Severity Score (TLICS) based on items such as fracture morphology, neurologic findings, and ligamentous complex integrity. The Spine Trauma Study Group completed what Audigé et al [11] consider to be phases I and II of development and multicenter reliability / agreement study, respectively. The TLICS was developed to predict and support the treatment decision for surgical treatment. This severity score is obtained by summing points allocated to each diagnostic item deemed to guide treatment decisions in practice. The point system was elicited by expert opinion, so there remains uncertainty as whether the final score is adequately formulated.

In term of validation process, similarly to a classification system, the development of a severity score implies that a reliable diagnostic process exists for each selected items and that the outcome(s) are defined. In addition the weighting of diagnostic items to generate the score should be created with sound statistical foundation, ie, the score calculation algorithm requires clinical data to create prognostic models. This process has been applied in the context of outcome analysis systems to provide norms for trauma care [26–28].

One limitation of a severity score is that the score itself may not help a clinician fully understanding the related injury beyond the provided prognostic information. Recently a severity score was proposed by the AO Craniomaxillofacial fracture Classification Group [29], however this group refocused its objective primarily in the development of a clinically meaningful classification system that can be translated subsequently i n a severity score following a Phase-III documentation project.

Within the research pathway followed by the AOSCG, such a severity score will not be developed in Phase I (reliability studies about the diagnostic items). The AOSCG pursues a path to first gather scientific evidence about the prognostic value of each reliable diagnostic item for targeted outcomes before a severity score is proposed in phase III after clinical study data have been collected.

5 References