INTEROBSERVER REPRODUCIBILITY ASSESSMENT OF THE NEW AOSPINE CLASSIFICATION FOR SUBAXIAL CERVICAL LESIONS

Objective: To evaluate the interobserver agreement of the new AOSpine classification for subaxial cervical fractures. Methods: A des-criptive study, which11 traumatic lesions of the subaxial cervical spine (through radiographic and tomographic images), were evaluated by 16 observers being: 6 senior surgeons, 4 fellows in spinal surgery and 6 physicians residents in Orthopedics and Traumatology by the new AOSpine classification, with subsequent statistical analysis of the results. An agreement analysis was performed using the Kappa coefficient, both individually and in combination, with an interpretation of the index performed using the standardized model for Landis and Koch. To determine the level of significance of the analyzes, values less than 0.05 were considered statistically significant. Results: In general, the level of agreement among the examiners was considered reasonable. The lesions “A0 (F3)”, “A4 (F3)”, “B1”, “B3”, “B3 (F3)”, “C”, “C (F3)” and “F3”showed a low level of agreement between the examiners. The level of reasonable agreement was obtained between fractures “A0”, “A1”, “A4”, “B2” and “C (F4)”. The only fracture that presented a moderate level of agreement was the “C (F4 BL)” lesion. This result indicates that the referred injury was the fracture of the subaxial column that presented the best level of agreement among the 16 examiners in the present study. Conclusions: The results of the study indicate an intermediate agreement of the new AOSpine classification for subaxial cervical lesion and point to the need to carry out studies that seek to evaluate this new classification in order to better evaluate its strengths and weaknesses, contributing for its improvement. Level of evidence III; Diagnostic study - investigation of a diagnostic test. y los El de se las fracturas A1, y La única fractura que presentó un nivel de concordancia la Este indica que la la del Conclusiones: Los resultados del estudio indican un intermedio de la nueva clasificación de AOSpine para cervicales y a la de realizar que evaluar esta nueva clasificación, para evaluar sus puntos fuertes y débil, para


INTRODUCTION
Classifications are important in Orthopedics and Traumatology and are frequently used to guide treatment, prognosis, and communication among care teams, 1 in addition to serving to standardize the international scientific language. 2 To meet these objectives, the classifications must be simple, easy to apply in clinical practice, and reproducible, with high concordance among the surgeons. 3 During elaboration, a classification must go through three research phases before validation for clinical use. The first phase consists of the development of the classification by a group of physicians experienced in the treatment of the particular injuries, based on several pilot studies. At this stage, concordance among the surgeons must be greater than 90%. In the next phase, the classification is tested by examiners with different levels of knowledge and practice. In the final phase it is applied clinically and documented for some period of time to confirm its validity. 4 The most recent classification of subaxial cervical injuries is that of the AOSpine group, 5 which seeks to be comprehensive and include various types of subaxial cervical spine bone and soft tissue injuries, based on the AOSpine classification for thoracolumbar fractures. 6 Earlier classifications for subaxial cervical lesions, such as that of Allen and Fergunson 7 and more recently the SLICS classification, do not present adequate interobserver concordance and leave a lot to be desired in assisting with therapeutic decisions. 8 Bearing in mind that the subaxial cervical spine is the most common location for cervical injuries, accounting for 65% of cervical fractures and more than 75% of cervical fracture-dislocations, 8 which are a serious public health problem and can have devastating results, such as irreversible spinal cord injury and death, 9 a classification with good concordance among surgeons is needed to optimize the treatment of these injuries.
The objective of this study was to evaluate the interobserver concordance of the most recent subaxial injury classification (AOSpine).

METHODS
Eleven cases of traumatic subaxial cervical injury were selected from the personal files of one of the authors of the study. The associated radiographic and tomographic images contained no patient identification data. The case files were written to CDs together with a copy of the AOSpine classification ( Figure 1) and delivered to the participants with a printed questionnaire, where they wrote down their responses and then returned the material to the researchers within one week.
All the participants signed an informed consent form and their identity was protected in accordance with the parameters prescribed by Resolution 466/2012 of the National Health Council / Ministry of Health, which addresses research involving human subjects. The study was approved by the Institutional Review Board of the institution.
Cases of pathological fractures (tumoral, from infection, or osteoporotic), those without information clear enough to define the classification, and injuries at multiple levels were excluded.
The cases were classified by 16 observers, 6 of whom were senior surgeons, 4 of whom were fellows in spine surgery, and 6 of whom were medical residents in Orthopedics and Traumatology.
The evaluators classified the cases according to the new AOSpine classification into types A, B, C, and F and their morphological subtypes. (Figure 1) No clinical data were provided to the evaluators for the definition of neurological modifiers or the M modifier.
Interobserver observance was evaluated by statistical comparison of the evaluators' responses.
The analysis of concordance was conducted through use of the kappa statistic, both in individual and combined form, and the interpretation of the statistic was conducted using the interpretation model standardized by Landis and Koch (Strength of Agreement). (Figure 2) To determine the level of significance of the analyses, values of p less than 0.05 were considered statistically significant.
All the analyses were conducted using STATA/MP version 14.0 statistical software.
The researchers were responsible for financing the research.

RESULTS
The cases presented were classified in different ways by the study examiners. (Table 1) Only one examiner (in the senior surgeon subgroup) failed to select the type of fracture in one of the study samples (case 11). The other cases were classified through the interpretation of each examiner, taking the image characteristics and the parameters indicated for each classification into account.
In general, the level of concordance among the examiners was considered fair (combined kappa = 0.2385 / p value = 0.0000) ( Table 2).
The only fracture that presented a moderate level of concordance was injury C(F4 BL), for which the kappa statistic value was equal to 0.4275 (p = 0.0000). This result indicates that this particular injury was the subaxial spine fracture with the highest level of concordance among the 16 examiners in the present study.

DISCUSSION
Classifications are useful tools, not only in the diagnosis and description of injuries, but also in defining optimal trauma team conduct, leading to better results for the individuals under care. A classification should be practical enough to be disseminated and routinely applied, but it must also be accurate, reproducible, and adequately associated with the best indication for treatment and prognosis.
In 2007, Vaccaro proposed a classification system for subaxial cervical fractures based on 3 axes: the morphology of the injury, the integrity of the posterior ligament complex, and the neurological evaluation, which generally showed good interobserver concordance, although when analyzed separately the   posterior ligament complex presented low concordance. 8 One negative point of this classification was that it required the use of imaging examinations, which are often not readily accessible in the context of most urgent care facilities. The previously used classification was that of Allen and Fergunson (1982), who observed moderate interobserver concordance as cited by Vaccaro. In a 2008 comparative study of these two classifications, Jorge 10 observed that both presented good intraobserver and moderate interobserver concordance. Currently, the new AOSpine classification proposed by Vaccaro in 2015 has emerged as one of the most widely disseminated classifications and the results of this study offer more evidence of its reproducibility, and, considering that the evaluations made by the physicians in this study indicate highly variable levels of concordance when comparing the different injuries included in the classification, the overall concordance was considered fair.
In 2017, Urrutia 11 demonstrated substantial interobserver concordance for the main fracture types (A, B, and C) and moderate concordance for the subtypes. In 2015, Vaccaro 5 had also demonstrated good interobserver concordance for the main fracture types when describing the classification.
The data obtained show that the greater experience of the examiner does not necessarily imply higher levels of concordance, a fact that indicates that generalized use of this classification for the entire community is effective regardless of the degree of experience with these fractures.

CONCLUSIONS
From this study it was concluded that the new AOSpine classification presented a fair overall level of interobserver concordance.
The evolution of the classification is key to a more accurate approach and one with better results, thus contributing to reduce the costs involved in the process and improve the treatment of these serious injuries.
All authors declare no potential conflict of interest related to this article. Table 1. Analysis of interobserver concordance of 11 cases of traumatic injuries of the subaxial cervical spine, as described by 16 examiners (surgeons, fellows, and residents), according to the AOSpine classification.  Table 2. Analysis of interobserver concordance of 11 cases of traumatic injuries of the subaxial cervical spine, as described by 16 examiners (surgeons, fellows, and residents), according to the AOSpine classification.