- Email: [email protected]
Proposal of a modified, treatment-oriented classification of odontoid fractures
The Spine Journal 5 (2005) 123–129
Clinical Studies
Proposal of a modified, treatment-oriented classification of odontoid fractures Jonathan N. Grauer, MDa,b,*, Bilal Shafi, MDa, Alan S. Hilibrand, MDa, James S. Harrop, MDc, Brian K. Kwon, MDa,d, John M. Beiner, MDa, Todd J. Albert, MDa, Michael G. Fehlings, MD, PhDe, Alexander R. Vaccaro, MDa a Rothman Institute at Thomas Jefferson University, 925 Chestnut St., Philadelphia, PA 19107-4216, USA Department of Orthopaedics and Rehabilitation, Yale University School of Medicine, PO Box 208071, New Haven, CT 06520-8071, USA c Department of Neurosurgery, Thomas Jefferson University, 909 Walnut St., Philadelphia, PA 19107, USA d Combined Neurosurgical and Orthopaedic Spine Program, Department of Orthopaedics, University of British Columbia, 269-828 W. 10th Avenue, Vancouver, BC V5Z 1L8, Canada e Division of Neurosurgery, University of Toronto and Spine Program, Krembil Neuroscience Center, Toronto Western Hospital, 399 Bathurst St., Toronto, Ontario M5T 2S8, Canada
b
Received 1 March 2004; accepted 28 September 2004
Abstract
BACKGROUND CONTEXT: The classification scheme of odontoid fractures described by Anderson and D’Alonzo is the one most commonly used. However, uncertainty exists in the distinction between Type II and “shallow” Type III fractures. Moreover, fractures at the base of the odontoid (Anderson and D’Alonzo Type II) include a spectrum of injury patterns. PURPOSE: To modify the Anderson and D’Alonzo classification of odontoid fractures based on current clinical treatment options. STUDY DESIGN: Proposal of a modified classification system for odontoid fractures. METHODS: A more precise distinction between Type II and III fractures based on the presence/ absence of C1–C2 facet involvement is proposed. A modified classification of Type II fractures based on fracture line obliquity, displacement and comminution is then proposed, because these are factors deemed to influence management. To evaluate the reproducibility of this classification, 52 odontoid fractures were reviewed and classified by four attending spine surgeons and three spine fellows. RESULTS: There was substantial agreement (at least five of seven respondents) in 70% of cases. The overall kappa value for the modified classification system was 0.48, indicating moderate agreement, and there were no differences in kappa values between attending spine surgeons and fellows. CONCLUSIONS: The reproducibility of this system was demonstrated by the moderate agreement observed when applied to odontoid fractures at our institution. The proposed utility of this system is its ability to guide clinical decision making in the treatment of odontoid fractures. Prospective application of this modified classification system and suggested treatment options is now required. 쑖 2005 Elsevier Inc. All rights reserved.
Keywords:
Odontoid process; Axis; Spinal fractures; Cervical vertebrae
Introduction In the adult population, odontoid fractures account for 9% to 15% of cervical spine injuries [1–3]. These injuries usually result from falls in the elderly or high-velocity acciFDA device/drug status: not applicable. Nothing of value received from a commercial entity related to this research. * Corresponding author. Department of Orthopaedics and Rehabilitation, Yale University School of Medicine, PO Box 208071, New Haven, CT 06520-8071, USA. Tel.: (203) 785-4077; fax: (203) 785-7132. E-mail address: [email protected] (J.N. Grauer) 1529-9430/05/$ – see front matter doi:10.1016/j.spinee.2004.09.014
쑖 2005 Elsevier Inc. All rights reserved.
dents in the young and middle-aged. The fracture mechanism is generally agreed to be hyperextension or hyperflexion of the cervical spine. Contemporary treatment of these fractures ranges from collar immobilization [1,4] to operative treatment with posterior fusion or anterior screw fixation [5–10]. However, the most appropriate treatment for specific odontoid fractures remains controversial. The classification of odontoid fractures has evolved over the last three decades. Schatzker et al. [11] initially classified these fractures as high or low depending on where the fractures were in relation to the accessory ligaments. In 1974,
124
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
Anderson and D’Alonzo proposed the classification system still favored today based on fracture pattern (Fig. 1) [4]. Type I fractures are through the upper portion of the odontoid process and are thought to occur by avulsion of the apical and/or alar ligaments. Type II fractures occur at the base of the odontoid, between the level of the transverse ligament and the C2 vertebral body. Type III fractures extend into the vertebral body. Type II fractures are the most common and unstable type of odontoid fracture. Increased patient age, extent and direction of fracture displacement, delay in diagnosis and fracture comminution have all been shown to negatively influence union rates [1,3,4]. With high reported nonunion rates, the literature demonstrates that immobilization is essential for these fractures. The difficulty is determining how such immobilization should be achieved and maintained: collar immobilization [1,4], halo immobilization [3] or anterior [9] or posterior [5–8,10] internal fixation. Sometimes it can be difficult to precisely define when a Type II fracture extends inferior enough to be considered a Type III fracture. Many authors refer to such intermediate fractures as “shallow” or “high” Type III fractures. These injuries fall into a grey zone between Type II and Type III fractures for which some authors suggest fixation [1,7]. Overall, there are two clear limitations of the Anderson and D’Alonzo classification system of odontoid fractures. First, there is not a sufficiently precise distinction between Type II and III fractures. This is borne out in the literature by ambiguous distinctions between these injuries, as well as the authors’ experience in discussing and reporting fracture patterns. Second, Type II fractures represent a broad anatomic spectrum of injuries, which can managed differently depending on such factors as the fracture line obliquity, displacement and comminution. This is highlighted by the different types of fracture fixation now advocated for different subpopulations of such injuries.
fied, treatment-oriented classification scheme. The proposed scheme includes Type I, II and III fractures and thus maintains the general architecture of the Anderson and D’Alonzo system with which most physicians are familiar. First, Type II fractures are more precisely defined to distinguish them more clearly from Type III fractures. In the proposed system, Type II fractures are those that are located between the inferior aspect of the anterior C1 ring and do not extend into the superior articular facets of C2 (Fig. 2, left). Based on this definition, fractures that are oblique in the anterior/posterior dimension may extend into the C2 vertebral body and still be considered Type II fractures as long as there is no involvement of the superior C2 facets. If either of the superior C2 facets is involved, a fracture is considered a Type III fracture (Fig. 2, right). Using the above definitions, Type I fractures (without evidence of occipital-cervical instability) and Type III fractures (that include enough of the C2 body to involve the superior C2 facets) are considered to be those that have been treated successfully in the past with collar immobilization. Type II fractures, however, are managed differently depending on the specifics of the fracture anatomy. As such, we propose a stratification of Type II fractures into three subtypes as described in Table 1 and Fig. 3.
To address the limitations of the Anderson and D’Alonzo odontoid fracture classification system, we propose a modi-
• Subtype A: transverse fractures without comminution and with less than 1 mm displacement. One mm or less was chosen to attempt to define fractures that had minimal or no displacement because preliminary reviews of films had revealed that there was not good agreement when such fractures were defined as no displacement at all. Although rare, this fracture type has been shown to have predictably good results with external immobilization [12], which would make these reasonable treatment options. • Subtype B: fractures that pass from anterior superior to posterior inferior or displaced transverse fractures (greater than 1 mm). These fractures have been shown to have good results when treated with anterior screw fixation after fracture reduction, assuming adequate bone density. Although complications and loss of fixation have been reported [1,3], this technique spares much of the axial rotation at the C1–C2 articulation, which is compromised with the atlantoaxial fusion
Fig. 1. Anderson and D’Alonzo’s classification of odontoid fractures. Note that Type I injuries are at the tip of the odontoid. Type II injuries are at the base of the odontoid. And, Type III injuries extend into the cancellous C2 body.
Fig. 2. In our classification, we propose a more precise distinction between Type II and Type III odontoid fractures. Type II fractures have no involvement of the superior articular facets of C2 (A) whereas Type III fractures do have involvement of the superior articular facets of C2 (B). Arrows demonstrate where the fracture line exits C2 superiorly.
Materials and methods Proposal of a modified classification system for odontoid fractures
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
125
Table 1 Proposed classification and treatment algorithm of Type II odontoid fractures Type II fracture subtype
Description
Treatment guideline
Type I Type II Type II A Type II B Type II C Type III
Above inferior aspect of C1 anterior arch Between Type I and Type II fractures Nondisplaced fracture Anterior superior to posterior inferior and displaced transverse fractures Anterior inferior to posterior superior or comminuted fractures Including at least one of the superior articular facets of C2
External immobilization As noted below External immobilization Anterior screw fixation Posterior atlantoaxial fusion External immobilization
techniques. Therefore, many surgeons favor this as the surgical treatment of choice for these fractures. • Subtype C: fractures that pass from anterior inferior to posterior superior or fractures with significant comminution. These are not appropriate for anterior screw fixation. Posterior atlantoaxial stabilization has been shown to yield good, long-term results with these fracture patterns [7,10]. Evaluation of a series To evaluate the reproducibility of applying this proposed classification system, a series of odontoid fractures clinically managed at our institution were reviewed. The Division of Human Subjects Protection (institutional review board) of our institution reviewed and approved the following protocol. The consult logs of our spine service were reviewed, and patients with odontoid fractures who presented to our institution between 1995 and 2001 were identified. Inclusion criteria for the cases were a minimum of an open mouth
and lateral radiograph, a computed tomography scan or tomogram and age greater than 18 years. Fifty-two cases that met these criteria were identified. Example cases are shown in Figs. 4 to 7. Seven surgeons (four fellowship-trained, attending spine surgeons and three spine fellows) were educated to this modified classification system before analyzing the radiographs. Each surgeon then independently reviewed the films and graded each case as Type I, IIA, IIB, IIC or III (Table 1, Figs. 2 and 3). Statistical analyses were performed using the SAS 8.0 computer software package. The first set of analyses sought to describe the overall level of consistency between respondents. Percent agreement was computed to provide an overall picture of the amount of concordance between respondents. Then, kappa statistics were computed to determine concordance between respondents, controlling for chance agreement. The second set of analyses aimed to determine whether differences in agreement existed for individual physicians or for attending physicians versus fellows using average kappa
Fig. 3. Proposed sub-classification of Type II odontoid fractures.
126
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
Fig. 4. Example of a Type IIA odontoid fracture.
values at the individual and group level. Although kappa values typically can measure agreement only between two sources of information, a SAS macro was used that computes estimates and tests of agreement among multiple raters based on methodology presented by Fleiss [13]. Statistical significance of these kappa values is based on z tests.
Fig. 6. Example of a Type IIC odontoid fracture.
Between 1995 and 2001, 218 patients presented to our center with odontoid fractures. This accounted for 4% of all emergency spine visits and 11% of all cervical spine fractures. Of the individuals identified, 52 patients met the inclusion criteria for the study. The other patients were excluded because of age (three were younger than 18 years) and/or lack of adequate imaging studies. Of those included in the study, 25 were female (48%) and 27 were male (52%). Ages ranged from 18 to 100 years, with a mean of 62 years. Ten patients (19%) were injured from a fall from a height, 18 patients (35%) were injured from a fall from a standing position, and 24 patients (46%) were injured from a motor accident. The total number of times each classification category within the revised classification system was assigned by the respondents is presented in Table 2. Respondents were most likely to rate a fracture as Type IIB or Type III. None of the fractures was classified as a Type I; consequently, this category was not included in further analysis. Fig. 8 presents cumulative percent agreement between raters. In 31% of cases (16 of 52), there was 100% agreement among respondents in categorizing the fracture (seven of seven). In
50% of cases (26 of 52), there was at least 86% agreement among respondents (six of seven). In 71% of cases (37 of 52), there was 71% agreement among respondents (five of seven). At least four of the seven respondents agreed on the fracture classification for every case. Next, kappa statistics were used to estimate agreement between respondents controlling for chance agreement. In the first set of analyses, kappa values and test statistics were computed based on responses from all seven respondents. The overall kappa value for the system and kappa values for each of the four possible categories (Type IIA, IIB, IIC, III) are presented in Table 2. All kappa values were statistically significant (more likely than chance alone). The overall kappa value of 0.48 corresponds with moderate agreement according to standard criteria [14]. As shown, there was higher agreement between respondents for Type IIB (kappa⫽ 0.54) and Type III (kappa⫽0.57) than Type IIA (kappa⫽ 0.37) and IIC (kappa⫽0.23) fractures. The next set of analyses sought to determine whether any one respondent appeared to be making classifications differently than other respondents, and whether differences existed in agreement between attending surgeon and fellow respondents. To address differences at the individual level, we computed kappa values for agreement between each respondent with the other six respondents. Respondents who had difficulty using the classification system should have lower average kappa values than the other respondents. Results from these analyses are presented in Table 3. As shown, kappa values did not differ substantially at the individual level, with average kappa values for individuals ranging from 0.43 to 0.52. These results suggest that each of
Fig. 5. Example of a Type IIB odontoid fracture.
Fig. 7. Example of a Type III odontoid fracture.
Results
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
127
Table 2 Frequency of category use, category kappa values and overall kappa value for modified classification system across seven respondents Classification
Frequency of use
Type I Type IIA Type IIB Type IIC Type III Overall classification
0 50 167 29 118
(0%) (14%) (46%) (8%) (32%)
Multiple respondent kappa value (SE)
95% Confidence interval
z test
NA 0.37 0.54 0.23 0.57 0.48
NA 0.31–0.43 0.48–0.60 0.17–0.29 0.51–0.63 0.46–0.50
NA 12.10, 17.69, 7.46, 18.68, 24.02,
(0.03) (0.03) (0.03) (0.03) (.02)
p⬍.001 p⬍.001 p⬍.001 p⬍.001 p⬍.001
NA⫽not applicable.
the respondents showed a similar level of agreement with the other respondents and indicate that no one respondent had marked difficulty using the classification scheme. The overall kappa value for the four respondents who were attending surgeons was 0.48 (z⫽13.20, p⬍.001, 95% confidence interval [CI]⫽0.41 to 0.55) and the overall kappa value for the three respondents who were fellows was 0.50 (z⫽9.12, p⬍.001, 95% CI⫽0.40 to 0.61). The overlap in confidence intervals for the kappa values for attending surgeons versus fellows indicates that no significant difference existed.
Discussion Odontoid fractures are common injuries, yet their treatment remains controversial and variable. We hypothesized
that a reproducible, treatment-oriented classification system would facilitate the evaluation and management of these injuries. We thus proposed a modified classification system with suggested treatment recommendations for each fracture type (Table 1). At least five of seven surgeons agreed on the classification of 70% of the odontoid fractures using this revised system. The overall kappa value we observed in this study was 0.48, consistent with moderate agreement. Established standards for interpreting the magnitude of kappa values suggest kappa values between 0.2 and 0.4 indicate fair agreement, kappa values between 0.4 and 0.6 indicate moderate agreement, and kappa values above 0.6 indicate substantial agreement [14]. It is worth noting that the overall kappa value of 0.48 for the proposed odontoid injury classification is comparable to other classifications systems currently used to describe
Fig. 8. Percentage of cases showing varying levels of agreement among respondents.
128
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
Table 3 Average kappa values by respondent and respondent position
Individual respondents Respondent 1 Respondent 2 Respondent 3 Respondent 4 Respondent 5 Respondent 6 Respondent 7 Physician position Attending physicians (n⫽4) Fellows (n⫽3)
Average kappa with 6 other respondents
Range
0.48 0.49 0.43 0.52 0.51 0.44 0.51
0.28–0.64 0.40–0.61 0.28–0.64 0.35–0.64 0.46–0.56 0.35–0.53 0.44–.60
orthopedic pathology. For example, in radiographic analyses of the Neer classification of proximal humerus fractures, kappa values of 0.43 and 0.52 have been reported [15,16]. Many of the other classification systems that we routinely use are not even analyzed in this objective and critical fashion. This, of course, becomes a limiting factor for studies reliant on the accuracy of classifications for injuries being described. In the current study, agreement was strongest for fractures categorized as Type IIB or Type III. The weakest agreement was found for Type IIA and Type IIC fractures. We believe that this was because fracture line anatomy and angulation are easily determined, whereas fracture displacement and comminution may be more difficult to quantify. The advantage of our proposed classification system is the potential for fracture type to be closely associated with treatment recommendations. To this end, we offer this paper to propose the described modifications to the currently accepted Anderson and D’Alonzo system and to suggest potential associations with treatment options based on our current clinical practice. The disadvantage of this classification is that there are other factors that might influence the treatment of fractures but are not considered in this classification. For example, poor bone quality may exclude the possibility of anterior odontoid osteosythesis. An increased anterior-posterior diameter of the thoracic cavity may make it technically impossible to achieve the trajectory needed for anterior odontoid screw placement. Concomitant cervical fractures may also exist and need to be considered in the decision of how to treat the odontoid injury (eg, a long posterior internal fixation construct to manage both an odontoid and subaxial cervical injury). Nevertheless, the proposed classification and treatment algorithm is suggested only to help guide treatment. And, as with any classifications system, clinical judgment is required for its application. As techniques for the fixation of odontoid fractures have evolved, so have the factors that must be incorporated into a classification system used to describe them. We believe
Multiple respondent kappa value (SE)
z test
.48 (.04) .50 (.06)
13.21, p⬍.001 9.12, p⬍.001
that the currently proposed classification system represents an evolution of the Anderson and D’Alonzo classification system of odontoid fractures because it takes such variables into account. Clearly, the proposal of such classification modifications is only the first step. Prospective application of this system in both the classification and treatment of odontoid fractures will be required to validate its utility.
Acknowledgment The authors would like to thank Stephanie Milan, PhD, for assistance with statistical analysis.
References [1] Subach BR, Morone MA, Haid RW, McLaughlin MR, Rodts GR, Comey CH. Management of acute odontoid fractures with singlescrew anterior fixation. Neurosurgery 1999;45:812–920. [2] Lee PC, Chun SY, Leong JCY. Experience of posterior surgery in atlanto-axial instability. Spine 1984;9:230–1. [3] Vaccaro AR, Madigan L, Ehrler DM. Contemporary management of adult cervical odontoid fractures. Orthopedics 2000;23:1109–13. [4] Anderson LD, D’Alonzo RT. Fractures of the odontoid process of the axis. J Bone Joint Surg 1974;56A:1663–74. [5] Brooks AL, Jenkins EB. Atlano-axial arthrodesis by the wedge compression method. J Bone Joint Surg 1978;60A:279–84. [6] Gallie WE. Fractures and dislocations of the upper cervical spine. Am J Surg 1939;46:495–9. [7] Jeanneret B, Mageral F. Primary posterior fusion C1/2 in odontoid fractures: indications, technique, and results of transarticular screw fixation. J Spinal Disord 1992;5:464–75. [8] Harms J, Melcher RP. Posterior C1-2 fusion with polyaxial screw and rod fixation. Spine 2001;22:2467–71. [9] Geisler FH, Cheng C, Pokra A, Brumback RJ. Anterior screw fixation of posteriorly displaced type II odontoid fractures. Neurosurgery 1989;25:30–8. [10] Coyne TJ, Fehlings MG, Wallace MC, Bernstein M, Tator CH. C1-C2 posterior cervical fusion: long-term evaluation of results and efficacy. Neurosurgery 1995;37:688–93. [11] Schatzker J, Rorabeck CH, Waddell JP. Fractures of the dens (odontoid process): an analysis of thirty-seven cases. J Bone Joint Surg 1971; 53B:392–405.
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129 [12] Dunn ME, Seljeskog EL. Experience in the management of odontoid process injuries: an analysis of 128 cases. Neurosurgery 1986;18: 306–10. [13] Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York: John Wiley & Sons, 1981. [14] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159–74.
Sixty-five Years Ago in Spine
The first report of therapeutic use of ultrasound in the care of spine-related disorders is attributed to Pohlmann et al. in 1939 [1,2].
129
[15] Bernstein J, Adler LM, Blank JE, Dalsey RM, Williams GR, Iannotti JP. Evaluation of the Neer system of classification of proximal humeral fractures with computed tomographic scans and plain radiographs. J Bone Joint Surg 1996;78A:1371–5. [16] Sjoden GO, Movin T, Guntner P, et al. Poor reproducibility of classification of proximal humerus fractures: additional CT of minor value. Acta Orthop Scand 1997;68:239–42.
References [1] Pohlmann R, Richter R, Parow E. Ueber die Ausbreitung and Absorption des Ultraschalls im menschlichen Gewebe und seine therapeutische Wirkung Ischias und Plexusneuralgie. Dtsch med Wschr 1939;65:251–4. [2] Norman JM, editor. Morton’s medical bibliography, 5th ed. Hants, England: Scolar Press, 1991:316.
Clinical Studies
Proposal of a modified, treatment-oriented classification of odontoid fractures Jonathan N. Grauer, MDa,b,*, Bilal Shafi, MDa, Alan S. Hilibrand, MDa, James S. Harrop, MDc, Brian K. Kwon, MDa,d, John M. Beiner, MDa, Todd J. Albert, MDa, Michael G. Fehlings, MD, PhDe, Alexander R. Vaccaro, MDa a Rothman Institute at Thomas Jefferson University, 925 Chestnut St., Philadelphia, PA 19107-4216, USA Department of Orthopaedics and Rehabilitation, Yale University School of Medicine, PO Box 208071, New Haven, CT 06520-8071, USA c Department of Neurosurgery, Thomas Jefferson University, 909 Walnut St., Philadelphia, PA 19107, USA d Combined Neurosurgical and Orthopaedic Spine Program, Department of Orthopaedics, University of British Columbia, 269-828 W. 10th Avenue, Vancouver, BC V5Z 1L8, Canada e Division of Neurosurgery, University of Toronto and Spine Program, Krembil Neuroscience Center, Toronto Western Hospital, 399 Bathurst St., Toronto, Ontario M5T 2S8, Canada
b
Received 1 March 2004; accepted 28 September 2004
Abstract
BACKGROUND CONTEXT: The classification scheme of odontoid fractures described by Anderson and D’Alonzo is the one most commonly used. However, uncertainty exists in the distinction between Type II and “shallow” Type III fractures. Moreover, fractures at the base of the odontoid (Anderson and D’Alonzo Type II) include a spectrum of injury patterns. PURPOSE: To modify the Anderson and D’Alonzo classification of odontoid fractures based on current clinical treatment options. STUDY DESIGN: Proposal of a modified classification system for odontoid fractures. METHODS: A more precise distinction between Type II and III fractures based on the presence/ absence of C1–C2 facet involvement is proposed. A modified classification of Type II fractures based on fracture line obliquity, displacement and comminution is then proposed, because these are factors deemed to influence management. To evaluate the reproducibility of this classification, 52 odontoid fractures were reviewed and classified by four attending spine surgeons and three spine fellows. RESULTS: There was substantial agreement (at least five of seven respondents) in 70% of cases. The overall kappa value for the modified classification system was 0.48, indicating moderate agreement, and there were no differences in kappa values between attending spine surgeons and fellows. CONCLUSIONS: The reproducibility of this system was demonstrated by the moderate agreement observed when applied to odontoid fractures at our institution. The proposed utility of this system is its ability to guide clinical decision making in the treatment of odontoid fractures. Prospective application of this modified classification system and suggested treatment options is now required. 쑖 2005 Elsevier Inc. All rights reserved.
Keywords:
Odontoid process; Axis; Spinal fractures; Cervical vertebrae
Introduction In the adult population, odontoid fractures account for 9% to 15% of cervical spine injuries [1–3]. These injuries usually result from falls in the elderly or high-velocity acciFDA device/drug status: not applicable. Nothing of value received from a commercial entity related to this research. * Corresponding author. Department of Orthopaedics and Rehabilitation, Yale University School of Medicine, PO Box 208071, New Haven, CT 06520-8071, USA. Tel.: (203) 785-4077; fax: (203) 785-7132. E-mail address: [email protected] (J.N. Grauer) 1529-9430/05/$ – see front matter doi:10.1016/j.spinee.2004.09.014
쑖 2005 Elsevier Inc. All rights reserved.
dents in the young and middle-aged. The fracture mechanism is generally agreed to be hyperextension or hyperflexion of the cervical spine. Contemporary treatment of these fractures ranges from collar immobilization [1,4] to operative treatment with posterior fusion or anterior screw fixation [5–10]. However, the most appropriate treatment for specific odontoid fractures remains controversial. The classification of odontoid fractures has evolved over the last three decades. Schatzker et al. [11] initially classified these fractures as high or low depending on where the fractures were in relation to the accessory ligaments. In 1974,
124
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
Anderson and D’Alonzo proposed the classification system still favored today based on fracture pattern (Fig. 1) [4]. Type I fractures are through the upper portion of the odontoid process and are thought to occur by avulsion of the apical and/or alar ligaments. Type II fractures occur at the base of the odontoid, between the level of the transverse ligament and the C2 vertebral body. Type III fractures extend into the vertebral body. Type II fractures are the most common and unstable type of odontoid fracture. Increased patient age, extent and direction of fracture displacement, delay in diagnosis and fracture comminution have all been shown to negatively influence union rates [1,3,4]. With high reported nonunion rates, the literature demonstrates that immobilization is essential for these fractures. The difficulty is determining how such immobilization should be achieved and maintained: collar immobilization [1,4], halo immobilization [3] or anterior [9] or posterior [5–8,10] internal fixation. Sometimes it can be difficult to precisely define when a Type II fracture extends inferior enough to be considered a Type III fracture. Many authors refer to such intermediate fractures as “shallow” or “high” Type III fractures. These injuries fall into a grey zone between Type II and Type III fractures for which some authors suggest fixation [1,7]. Overall, there are two clear limitations of the Anderson and D’Alonzo classification system of odontoid fractures. First, there is not a sufficiently precise distinction between Type II and III fractures. This is borne out in the literature by ambiguous distinctions between these injuries, as well as the authors’ experience in discussing and reporting fracture patterns. Second, Type II fractures represent a broad anatomic spectrum of injuries, which can managed differently depending on such factors as the fracture line obliquity, displacement and comminution. This is highlighted by the different types of fracture fixation now advocated for different subpopulations of such injuries.
fied, treatment-oriented classification scheme. The proposed scheme includes Type I, II and III fractures and thus maintains the general architecture of the Anderson and D’Alonzo system with which most physicians are familiar. First, Type II fractures are more precisely defined to distinguish them more clearly from Type III fractures. In the proposed system, Type II fractures are those that are located between the inferior aspect of the anterior C1 ring and do not extend into the superior articular facets of C2 (Fig. 2, left). Based on this definition, fractures that are oblique in the anterior/posterior dimension may extend into the C2 vertebral body and still be considered Type II fractures as long as there is no involvement of the superior C2 facets. If either of the superior C2 facets is involved, a fracture is considered a Type III fracture (Fig. 2, right). Using the above definitions, Type I fractures (without evidence of occipital-cervical instability) and Type III fractures (that include enough of the C2 body to involve the superior C2 facets) are considered to be those that have been treated successfully in the past with collar immobilization. Type II fractures, however, are managed differently depending on the specifics of the fracture anatomy. As such, we propose a stratification of Type II fractures into three subtypes as described in Table 1 and Fig. 3.
To address the limitations of the Anderson and D’Alonzo odontoid fracture classification system, we propose a modi-
• Subtype A: transverse fractures without comminution and with less than 1 mm displacement. One mm or less was chosen to attempt to define fractures that had minimal or no displacement because preliminary reviews of films had revealed that there was not good agreement when such fractures were defined as no displacement at all. Although rare, this fracture type has been shown to have predictably good results with external immobilization [12], which would make these reasonable treatment options. • Subtype B: fractures that pass from anterior superior to posterior inferior or displaced transverse fractures (greater than 1 mm). These fractures have been shown to have good results when treated with anterior screw fixation after fracture reduction, assuming adequate bone density. Although complications and loss of fixation have been reported [1,3], this technique spares much of the axial rotation at the C1–C2 articulation, which is compromised with the atlantoaxial fusion
Fig. 1. Anderson and D’Alonzo’s classification of odontoid fractures. Note that Type I injuries are at the tip of the odontoid. Type II injuries are at the base of the odontoid. And, Type III injuries extend into the cancellous C2 body.
Fig. 2. In our classification, we propose a more precise distinction between Type II and Type III odontoid fractures. Type II fractures have no involvement of the superior articular facets of C2 (A) whereas Type III fractures do have involvement of the superior articular facets of C2 (B). Arrows demonstrate where the fracture line exits C2 superiorly.
Materials and methods Proposal of a modified classification system for odontoid fractures
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
125
Table 1 Proposed classification and treatment algorithm of Type II odontoid fractures Type II fracture subtype
Description
Treatment guideline
Type I Type II Type II A Type II B Type II C Type III
Above inferior aspect of C1 anterior arch Between Type I and Type II fractures Nondisplaced fracture Anterior superior to posterior inferior and displaced transverse fractures Anterior inferior to posterior superior or comminuted fractures Including at least one of the superior articular facets of C2
External immobilization As noted below External immobilization Anterior screw fixation Posterior atlantoaxial fusion External immobilization
techniques. Therefore, many surgeons favor this as the surgical treatment of choice for these fractures. • Subtype C: fractures that pass from anterior inferior to posterior superior or fractures with significant comminution. These are not appropriate for anterior screw fixation. Posterior atlantoaxial stabilization has been shown to yield good, long-term results with these fracture patterns [7,10]. Evaluation of a series To evaluate the reproducibility of applying this proposed classification system, a series of odontoid fractures clinically managed at our institution were reviewed. The Division of Human Subjects Protection (institutional review board) of our institution reviewed and approved the following protocol. The consult logs of our spine service were reviewed, and patients with odontoid fractures who presented to our institution between 1995 and 2001 were identified. Inclusion criteria for the cases were a minimum of an open mouth
and lateral radiograph, a computed tomography scan or tomogram and age greater than 18 years. Fifty-two cases that met these criteria were identified. Example cases are shown in Figs. 4 to 7. Seven surgeons (four fellowship-trained, attending spine surgeons and three spine fellows) were educated to this modified classification system before analyzing the radiographs. Each surgeon then independently reviewed the films and graded each case as Type I, IIA, IIB, IIC or III (Table 1, Figs. 2 and 3). Statistical analyses were performed using the SAS 8.0 computer software package. The first set of analyses sought to describe the overall level of consistency between respondents. Percent agreement was computed to provide an overall picture of the amount of concordance between respondents. Then, kappa statistics were computed to determine concordance between respondents, controlling for chance agreement. The second set of analyses aimed to determine whether differences in agreement existed for individual physicians or for attending physicians versus fellows using average kappa
Fig. 3. Proposed sub-classification of Type II odontoid fractures.
126
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
Fig. 4. Example of a Type IIA odontoid fracture.
values at the individual and group level. Although kappa values typically can measure agreement only between two sources of information, a SAS macro was used that computes estimates and tests of agreement among multiple raters based on methodology presented by Fleiss [13]. Statistical significance of these kappa values is based on z tests.
Fig. 6. Example of a Type IIC odontoid fracture.
Between 1995 and 2001, 218 patients presented to our center with odontoid fractures. This accounted for 4% of all emergency spine visits and 11% of all cervical spine fractures. Of the individuals identified, 52 patients met the inclusion criteria for the study. The other patients were excluded because of age (three were younger than 18 years) and/or lack of adequate imaging studies. Of those included in the study, 25 were female (48%) and 27 were male (52%). Ages ranged from 18 to 100 years, with a mean of 62 years. Ten patients (19%) were injured from a fall from a height, 18 patients (35%) were injured from a fall from a standing position, and 24 patients (46%) were injured from a motor accident. The total number of times each classification category within the revised classification system was assigned by the respondents is presented in Table 2. Respondents were most likely to rate a fracture as Type IIB or Type III. None of the fractures was classified as a Type I; consequently, this category was not included in further analysis. Fig. 8 presents cumulative percent agreement between raters. In 31% of cases (16 of 52), there was 100% agreement among respondents in categorizing the fracture (seven of seven). In
50% of cases (26 of 52), there was at least 86% agreement among respondents (six of seven). In 71% of cases (37 of 52), there was 71% agreement among respondents (five of seven). At least four of the seven respondents agreed on the fracture classification for every case. Next, kappa statistics were used to estimate agreement between respondents controlling for chance agreement. In the first set of analyses, kappa values and test statistics were computed based on responses from all seven respondents. The overall kappa value for the system and kappa values for each of the four possible categories (Type IIA, IIB, IIC, III) are presented in Table 2. All kappa values were statistically significant (more likely than chance alone). The overall kappa value of 0.48 corresponds with moderate agreement according to standard criteria [14]. As shown, there was higher agreement between respondents for Type IIB (kappa⫽ 0.54) and Type III (kappa⫽0.57) than Type IIA (kappa⫽ 0.37) and IIC (kappa⫽0.23) fractures. The next set of analyses sought to determine whether any one respondent appeared to be making classifications differently than other respondents, and whether differences existed in agreement between attending surgeon and fellow respondents. To address differences at the individual level, we computed kappa values for agreement between each respondent with the other six respondents. Respondents who had difficulty using the classification system should have lower average kappa values than the other respondents. Results from these analyses are presented in Table 3. As shown, kappa values did not differ substantially at the individual level, with average kappa values for individuals ranging from 0.43 to 0.52. These results suggest that each of
Fig. 5. Example of a Type IIB odontoid fracture.
Fig. 7. Example of a Type III odontoid fracture.
Results
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
127
Table 2 Frequency of category use, category kappa values and overall kappa value for modified classification system across seven respondents Classification
Frequency of use
Type I Type IIA Type IIB Type IIC Type III Overall classification
0 50 167 29 118
(0%) (14%) (46%) (8%) (32%)
Multiple respondent kappa value (SE)
95% Confidence interval
z test
NA 0.37 0.54 0.23 0.57 0.48
NA 0.31–0.43 0.48–0.60 0.17–0.29 0.51–0.63 0.46–0.50
NA 12.10, 17.69, 7.46, 18.68, 24.02,
(0.03) (0.03) (0.03) (0.03) (.02)
p⬍.001 p⬍.001 p⬍.001 p⬍.001 p⬍.001
NA⫽not applicable.
the respondents showed a similar level of agreement with the other respondents and indicate that no one respondent had marked difficulty using the classification scheme. The overall kappa value for the four respondents who were attending surgeons was 0.48 (z⫽13.20, p⬍.001, 95% confidence interval [CI]⫽0.41 to 0.55) and the overall kappa value for the three respondents who were fellows was 0.50 (z⫽9.12, p⬍.001, 95% CI⫽0.40 to 0.61). The overlap in confidence intervals for the kappa values for attending surgeons versus fellows indicates that no significant difference existed.
Discussion Odontoid fractures are common injuries, yet their treatment remains controversial and variable. We hypothesized
that a reproducible, treatment-oriented classification system would facilitate the evaluation and management of these injuries. We thus proposed a modified classification system with suggested treatment recommendations for each fracture type (Table 1). At least five of seven surgeons agreed on the classification of 70% of the odontoid fractures using this revised system. The overall kappa value we observed in this study was 0.48, consistent with moderate agreement. Established standards for interpreting the magnitude of kappa values suggest kappa values between 0.2 and 0.4 indicate fair agreement, kappa values between 0.4 and 0.6 indicate moderate agreement, and kappa values above 0.6 indicate substantial agreement [14]. It is worth noting that the overall kappa value of 0.48 for the proposed odontoid injury classification is comparable to other classifications systems currently used to describe
Fig. 8. Percentage of cases showing varying levels of agreement among respondents.
128
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129
Table 3 Average kappa values by respondent and respondent position
Individual respondents Respondent 1 Respondent 2 Respondent 3 Respondent 4 Respondent 5 Respondent 6 Respondent 7 Physician position Attending physicians (n⫽4) Fellows (n⫽3)
Average kappa with 6 other respondents
Range
0.48 0.49 0.43 0.52 0.51 0.44 0.51
0.28–0.64 0.40–0.61 0.28–0.64 0.35–0.64 0.46–0.56 0.35–0.53 0.44–.60
orthopedic pathology. For example, in radiographic analyses of the Neer classification of proximal humerus fractures, kappa values of 0.43 and 0.52 have been reported [15,16]. Many of the other classification systems that we routinely use are not even analyzed in this objective and critical fashion. This, of course, becomes a limiting factor for studies reliant on the accuracy of classifications for injuries being described. In the current study, agreement was strongest for fractures categorized as Type IIB or Type III. The weakest agreement was found for Type IIA and Type IIC fractures. We believe that this was because fracture line anatomy and angulation are easily determined, whereas fracture displacement and comminution may be more difficult to quantify. The advantage of our proposed classification system is the potential for fracture type to be closely associated with treatment recommendations. To this end, we offer this paper to propose the described modifications to the currently accepted Anderson and D’Alonzo system and to suggest potential associations with treatment options based on our current clinical practice. The disadvantage of this classification is that there are other factors that might influence the treatment of fractures but are not considered in this classification. For example, poor bone quality may exclude the possibility of anterior odontoid osteosythesis. An increased anterior-posterior diameter of the thoracic cavity may make it technically impossible to achieve the trajectory needed for anterior odontoid screw placement. Concomitant cervical fractures may also exist and need to be considered in the decision of how to treat the odontoid injury (eg, a long posterior internal fixation construct to manage both an odontoid and subaxial cervical injury). Nevertheless, the proposed classification and treatment algorithm is suggested only to help guide treatment. And, as with any classifications system, clinical judgment is required for its application. As techniques for the fixation of odontoid fractures have evolved, so have the factors that must be incorporated into a classification system used to describe them. We believe
Multiple respondent kappa value (SE)
z test
.48 (.04) .50 (.06)
13.21, p⬍.001 9.12, p⬍.001
that the currently proposed classification system represents an evolution of the Anderson and D’Alonzo classification system of odontoid fractures because it takes such variables into account. Clearly, the proposal of such classification modifications is only the first step. Prospective application of this system in both the classification and treatment of odontoid fractures will be required to validate its utility.
Acknowledgment The authors would like to thank Stephanie Milan, PhD, for assistance with statistical analysis.
References [1] Subach BR, Morone MA, Haid RW, McLaughlin MR, Rodts GR, Comey CH. Management of acute odontoid fractures with singlescrew anterior fixation. Neurosurgery 1999;45:812–920. [2] Lee PC, Chun SY, Leong JCY. Experience of posterior surgery in atlanto-axial instability. Spine 1984;9:230–1. [3] Vaccaro AR, Madigan L, Ehrler DM. Contemporary management of adult cervical odontoid fractures. Orthopedics 2000;23:1109–13. [4] Anderson LD, D’Alonzo RT. Fractures of the odontoid process of the axis. J Bone Joint Surg 1974;56A:1663–74. [5] Brooks AL, Jenkins EB. Atlano-axial arthrodesis by the wedge compression method. J Bone Joint Surg 1978;60A:279–84. [6] Gallie WE. Fractures and dislocations of the upper cervical spine. Am J Surg 1939;46:495–9. [7] Jeanneret B, Mageral F. Primary posterior fusion C1/2 in odontoid fractures: indications, technique, and results of transarticular screw fixation. J Spinal Disord 1992;5:464–75. [8] Harms J, Melcher RP. Posterior C1-2 fusion with polyaxial screw and rod fixation. Spine 2001;22:2467–71. [9] Geisler FH, Cheng C, Pokra A, Brumback RJ. Anterior screw fixation of posteriorly displaced type II odontoid fractures. Neurosurgery 1989;25:30–8. [10] Coyne TJ, Fehlings MG, Wallace MC, Bernstein M, Tator CH. C1-C2 posterior cervical fusion: long-term evaluation of results and efficacy. Neurosurgery 1995;37:688–93. [11] Schatzker J, Rorabeck CH, Waddell JP. Fractures of the dens (odontoid process): an analysis of thirty-seven cases. J Bone Joint Surg 1971; 53B:392–405.
J.N. Grauer et al. / The Spine Journal 5 (2005) 123–129 [12] Dunn ME, Seljeskog EL. Experience in the management of odontoid process injuries: an analysis of 128 cases. Neurosurgery 1986;18: 306–10. [13] Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York: John Wiley & Sons, 1981. [14] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159–74.
Sixty-five Years Ago in Spine
The first report of therapeutic use of ultrasound in the care of spine-related disorders is attributed to Pohlmann et al. in 1939 [1,2].
129
[15] Bernstein J, Adler LM, Blank JE, Dalsey RM, Williams GR, Iannotti JP. Evaluation of the Neer system of classification of proximal humeral fractures with computed tomographic scans and plain radiographs. J Bone Joint Surg 1996;78A:1371–5. [16] Sjoden GO, Movin T, Guntner P, et al. Poor reproducibility of classification of proximal humerus fractures: additional CT of minor value. Acta Orthop Scand 1997;68:239–42.
References [1] Pohlmann R, Richter R, Parow E. Ueber die Ausbreitung and Absorption des Ultraschalls im menschlichen Gewebe und seine therapeutische Wirkung Ischias und Plexusneuralgie. Dtsch med Wschr 1939;65:251–4. [2] Norman JM, editor. Morton’s medical bibliography, 5th ed. Hants, England: Scolar Press, 1991:316.