Institutional rev

*Reprint req

Documentation,

J Shoulder Elbow Surg (2012) -, 1-9

1058-2746/$ - s

http://dx.doi.org

www.elsevier.com/locate/ymse

The AO Foundation and Orthopaedic Trauma Association(AO/OTA) scapula fracture classification system: focuson glenoid fossa involvement

Martin Jaeger, MDa, Simon Lambert, FRCSEd(Orth)b, Norbert P. S€udkamp, MDa,James F. Kellam, MDc, Jan Erik Madsen, MDd, Reto Babst, MDe, Jonas Andermahr, MDf,Wilson Li, MDg, Laurent Audig�e, PhDh,*

aDepartment of Orthop€adie und Traumatologie, Universit€atsklinikum Freiburg, Albert-Ludwigs-Universit€at, Freiburg,GermanybShoulder and Elbow Service, The Royal National Orthopaedic Hospital, Stanmore, UKcDepartment of Orthopaedic Surgery, Carolinas Medical Center, Charlotte, NC, USAdOrthopaedic Department, Oslo University Hospital, Ullevaal, NorwayeKlinik f€ur Unfallchirurgie, Luzerner Kantonsspital, Luzern, SwitzerlandfDepartment for Orthopedic and Trauma Surgery, Hospital of the University of Bonn, Mechernich, GermanygDepartment of Orthopaedics & Traumatology, Queen Elizabeth Hospital, Kowloon, Hong KonghAO Clinical Investigation and Documentation, D€ubendorf, Switzerland

Background: Fractures of the glenoid frequently require surgical treatment. A comprehensive and reliablescapula classification system involving the glenoid fracture patterns is needed to describe the underlyingpathology. The AO Scapula Classification Group introduces an appropriate novel system that is presentedalong with its inter-rater reliability and accuracy.Materials and Methods: An iterative consensus process (involving a series of face-to-face meetings andagreement studies) with an international group of 7 experienced shoulder surgeons was used to specifyand evaluate a scapular fracture classification system with a focus on fracture patterns of the glenoidfossa. The last evaluation was conducted on a consecutive collection of 120 scapular fractures documentedby both plain radiographs and computed tomography scans including 3-dimensional surface rendering.Inter-rater reliability was analyzed with k statistics, and accuracy was estimated by latent class modeling.Results: Of 120 scapular fractures, 46 involved the glenoid (38%), with 38 classified as F1 articular rimfractures. The overall median sensitivity and specificity in identifying these fractures were 95% and 93%,respectively. Surgeons’ accuracy in classifying F1 fractures ranged from 86% to 100% (median, 94%).Subsequently, classification of simple F1 fractures resulted in a proportion of 36% of anterior rim fractures,19% of posterior rim fractures, and 45% of short oblique fractures, with accuracies ranging from 85% to98%.Conclusion: This new system for scapular glenoid fractures has proved to be sufficiently reliable and accu-rate when applied by experienced shoulder surgeons. Further validation of the most detailed system, as well

iew board/ethics committee approval: not required.

uests: Laurent Audig�e, PhD, AO Clinical Investigation and

Stettbachstrasse 6, CH-8600 D€ubendorf, Switzerland.

E-mail address: laurent.audige@aofoundation.org (L. Audig�e).

ee front matter � 2012 Journal of Shoulder and Elbow Surgery Board of Trustees.

/10.1016/j.jse.2012.08.003

2 M. Jaeger et al.

as involvement of surgeons with different levels of training in the framework of clinical routine andresearch, however, should be considered.Level of evidence: Level II, Development of Diagnostic Criteria, Diagnostic Study.� 2012 Journal of Shoulder and Elbow Surgery Board of Trustees.

Keywords: Scapula fracture; glenoid fossa; fracture classification; diagnostic; reliability; accuracy

Scapular fractures are rare and account for 1% of all frac-tures and 3% to 5% of all shoulder girdle fractures.7 In theSwedish population, the annual incidence was estimated at 10per 10,000 inhabitants.18 According to Goss,13 the scapularbody is involved in up to 45% of all scapula fractures, with theremainder involving the glenoid (approximately 10%), thescapular neck (25%), the processes (approximately 15%), andthe spine (5%). Treatment modalities depend on the fracturelocation, pattern, and displacement on the one hand and thepatient’s age, demands, and comorbidity on the other. Incontrast to fractures of the scapular body, which are mostlytreated conservatively, surgery is more often indicated for 80%of fractures involving the glenoid.24 This is explained by thespecific shapeand functionof theglenoid.Glenoid rim fracturescan result in joint instability, where surface incongruity mayincrease the risk of post-traumatic arthrosis accompanied bypain and limited function. The surgical approach is influencedby the main pathology. Not all glenoid fractures can be suffi-ciently reduced and stabilized by an anterior approach. Poste-rior fracture patterns have to be managed through a posteriorapproach that is more complex and less frequently practiced.

In the clinical environment, diagnosis of scapular frac-tures is based mainly on radiographic examination of theshoulder. Three plains comprising an anteroposterior view,outlet view, and axillary view are very helpful in gatheringessential information. If the glenoid is involved, it isstrongly recommended to obtain an additional computedtomography (CT) scan to precisely evaluate the glenoidsurface. It is also crucial to analyze the fracture systemati-cally with the help of a clinically relevant, reliable, and validfracture classification system.5,9,11,19 Until now, a number ofglenoid fracture classifications have been in use1,10,14-16,20;however, none of these are comprehensive and validated.

A comprehensive scapula classification system has beendeveloped as a collaborative effort between the AO Foun-dation and the Orthopaedic Trauma Association.17 Thepurpose of this study was to create and validate a focusedclassification of the glenoid as an extension of this system.

Materials and methods

Scapula classification group and phase I consensusdevelopment

The international Scapula Classification Group (SCCG)comprising 7 experienced shoulder specialist surgeons was estab-lished to propose a focused system for fractures of the glenoid

fossa, along with definitions and illustrations. This system was tobe descriptive, clinically relevant, sufficiently comprehensive, andintuitive for both inexperienced and experienced surgeons. Undercoordination by a methodologist statistician, a phase 1 consensusdevelopment process was implemented in an iterative manner witha series of 3 agreement studies (classification sessions), followedby face-to-face review meetings to discuss disagreements andimprove the system. Consensus was reached when the SCCGagreed that the system was well defined, clinically relevant, andreliable for its purpose.4 This development process was the first of3 research phases that is required to be completed before theclassification could be considered as validated.

Fracture classification system

The classification system for bony lesions of the scapula consistsof separate codes, that is, one each for the involvement of thearticular segment (glenoid fossa), the body, and any of theprocesses. In addition, each code comprises 2 levels, where level 1represents a basic system for rapid simplified documentation17 andlevel 2 provides a focused system for detailed documentation.

Classification of the scapula is divided into 3 regions: thearticular segment, the processes, and the body. The articularsegment includes the area involving the glenoid fossa and thearticular rim, which is limited by a line joining the superiorarticular rim to the lateral border of the suprascapular notch anda line positioned medially and parallel to the plane of the glenoidfossa that starts cranially at the lateral border of the suprascapularnotch.

In the basic system, fractures of the articular segment (denotedF) are classified in 1 of 3 groups (Fig. 1): F0, fracture of thearticular segment, not through the glenoid fossa, but resulting inthe fossa being detached from any part of the scapula body; F1,fracture involving the glenoid fossa with a simple rim, transverse,or oblique fracture pattern; and F2, multifragmentary joint fractureinvolving the glenoid fossa with 3 or more articular fragments.

The focused system is an extended classification of F1 or F2fractures according to their location in 4 quadrants withinthe glenoid fossa (Fig. 2). These quadrants are defined by theequatorial line and the intertubercular line running from thesupraglenoid tubercle to the infraglenoid tubercle. There are 5subgroups, consisting of 3 for F1 fractures and 2 for F2 fractures:F1(1), simple anterior articular rim or oblique fracture; F1(2),simple posterior articular rim or oblique fracture; F1(3), simpletransverse or short oblique fracture; F2(4), multifragmentaryfracture with more than 1 fracture line exit point; and F2(5),central fracture-dislocation, with no exit line through the rim.

Simple anterior (1) and posterior (2) articular rim or obliquefractures are further divided into 3 categories (Fig. 3): 1a/2a,infra-equatorial fracture located in 1 quadrant (same side as themaximum glenoid meridian); 1b/2b, rim fracture anterior/posterior

Figure 1 Fractures of the articular segment. � 2012, Jaeger et al.

Figure 2 Four areas defined within the glenoid fossa. � 2012, Jaeger et al.

Classification of glenoid fractures 3

to the maximum glenoid meridian with exits superior/inferior tothe equatorial line; and 1c/2c, fracture oblique line exiting onthe opposite side (posterior/anterior) to the maximum glenoidmeridian. (Initial definitions used for the last evaluation session arepresented in Appendix I, available on the journal’s website at www.jshoulderelbow.org. A revision and final proposal are presented inFig. 3).

Simple transverse or short oblique fractures (3) are also furtherdivided into 3 categories (Fig. 4): 3a, infra-equatorial; 3b, equa-torial; and 3c, supra-equatorial.

For multifragmentary fractures with a combination of 2‘‘simple’’ fracture lines, the focused codes describing this

combination may be added as an optional specification (Fig. 5);for example, in the case of a transverse and rim fracture, theclassification would be F2(4/1c3c).

Consecutive series and classification sessions

The classification system was evaluated for reliability and accu-racy using a consecutive series of scapula fractures documentedwith CT scans and conventional radiographs obtained from2 centers in Europe and North America. Cases were includedwhen diagnosed with a scapula fracture within 15 days after theinjury in adult patients (mature skeleton). Whereas the first session

Figure 3 Focused classification system of glenoid fossa rim fractures (F1). � 2012, Jaeger et al.

Figure 4 Focused classification system of simple transverse or short oblique fractures of the glenoid fossa (F1). � 2012, Jaeger et al.

4 M. Jaeger et al.

used the 2-dimensional CT series visualized with a standardDICOM (Digital Imaging and Communications in Medicine)viewer, the last 2 sessions considered 3-dimensional (3D) recon-structions for all cases prepared by the primary author (M.J.) on

video sequences (Fig. 6). All diagnostic images were collected inDICOM (Digital Imaging and Communications in Medicine)format, anonymized, and distributed on DVD format to SCCGmembers, who classified fractures independently ‘‘at home’’

Figure 5 Focused classification system of multifragmentary joint fractures (F2). � 2012, Jaeger et al.

Figure 6 Example of a 3D CT reconstruction video used for the last classification session. � 2012, Jaeger et al.

Classification of glenoid fractures 5

within a 3-month period. Fracture codes were collected electron-ically using specifically designed Excel sheets (Microsoft, Red-mond, WA, USA), which were centralized for the analyses. Theresults including reliability and accuracy data were reviewedduring face-to-face meetings to identify the reasons for codingdisagreements and improving the system.

The final session included 120 cases documented with the2-dimensional CT series and 3D CT reconstruction videos, as wellas additional radiographs available from 105 cases. Data weremanaged and analyzed by use of Intercooled Stata, version 11(StataCorp LP, College Station, TX, USA). The basic system wasfirst analyzed to assess the likely distribution of articular segment(F0/F1/F2) fractures in the sample and identify them. The focusedsystem was then applied to this fracture subset and separatelyanalyzed for simple fracture patterns (F1) and multifragmentaryjoint fractures (F2). Interobserver reliability was evaluated bymeans of k coefficients. The k coefficient is commonly used asa chance-corrected measure of agreement, which ranges from þ1(complete agreement) to 0 (agreement by chance alone) to less

than 0 (less agreement than expected by chance). The k coefficientis a useful indicator of reliability, and a value of 0.70 is consideredan adequate sign of reliability. Classification accuracy was esti-mated by latent class modeling22,23 using Latent GOLD software,version 3.0.1 (Statistical Innovations, Belmont, MA, USA). Thistechnique aims at identifying the most likely ‘‘true’’ distribution offracture classes in the population of scapula fractures based on theevaluated sample and the agreement data collected among theparticipating surgeons; for each class, the degree of classificationaccuracy for each surgeon is also estimated.6

Results

When identifying cases using the basic system to classifya fracture of the articular segment (denoted F), the7 shoulder specialists were in agreement for 73% of the 120cases. The overall k coefficient was 0.79, and the surgeons’pair-wise k (21 pairs) ranged from 0.66 to 0.93. The sample

6 M. Jaeger et al.

identifying the most likely ‘‘true’’ distribution included46 cases (38%) with an F fracture, where the accuracy inidentifying these fractures (diagnosis sensitivity) rangedfrom 91% to 99% (median, 95%). The surgeons’ accuracyin identifying the remaining 74 cases as non–articularsegment fractures ranged from 86% to 98% (median, 93%).

The 46 F fractures were further divided as follows: 1 F0fracture, 38 F1 fractures, and 7 F2 fractures. Surgeonsagreed on this subclassification in 72% of these cases (33),with an overall k coefficient of 0.64. The surgeons’pair-wise k (21 pairs) ranged from 0.43 to 0.92 (median,0.64). Surgeons were 86% to 100% accurate (median, 94%)in terms of classifying F1 fractures. The number of F2fractures was considered too small for estimation of thesurgeons’ classification accuracy with an adequate degreeof precision; nevertheless, 4 surgeons correctly classified6 of the 7 multifragmentary fractures, with a minimum of4 of these fractures correctly classified among the other3 surgeons. The only F0 fracture in the sample wascorrectly classified by 5 surgeons (Fig. 7), whereas theother 2 surgeons classified it as a simple fossa fracture (F1)or simple body fracture (B1).

Surgeons attained full agreement in the classification ofhalf of the 38 simple glenoid fossa (F1) fractures accordingto the higher level of the focused system (Figs. 3 and 4;Appendix I, available on the journal’s website at www.jshoulderelbow.org). The overall k coefficient was 0.66,with the surgeons’ pair-wise k ranging from 0.35 to 0.82(median, 0.67) (Table I). The estimated distribution of caseswithin the sample included 14 F1(1) fractures, 7 F1(2)fractures, and 17 F1(3) fractures with category-specifick coefficients of 0.70, 0.65, and 0.62, respectively. Therewas no central fracture-dislocation among the 7 multi-fragmentary joint fractures (Fig. 6), and all were classifiedas having more than 2 fracture line exit points [code F2(4)]with full agreement among the surgeons.

Classification session data supported the distinction ofsimple fossa fractures into 3 classes with the followingproportions: 36%, 19%, and 45% (Table II). The largestgroup, representing transverse or short oblique fractures(class 3), was classified with 87% to 93% accuracy by5 surgeons and 70% accuracy (median, 91%) for the other2. The second largest fracture group represented anteriorrim or oblique fractures (class 1), showing the highestclassification accuracies from 85% to 98% among thesurgeons (median, 91%). The posterior rim or obliquefractures (class 2) were classified with 80% to 96% accu-racy by 4 surgeons and 64% to 69% accuracy by the other3 (median, 80%). Incorrect classification of rim and obliquefractures as transverse and short oblique fracture typesoccurred more frequently, showing the higher ability todiscern between anterior and posterior rim fractures.

Further classification into the 3 subgroups of simpleglenoid fossa fractures (ie, a, b, and c) showed a lowerdegree of interobserver reliability, with overall k coefficientsof 0.46 and 0.39 based on the combined group of anterior

and posterior fractures (Fig. 3; Appendix I, available on thejournal’s website at www.jshoulderelbow.org) and the groupof transverse fractures, respectively (Fig. 4, Table I).Surgeons’ pair-wise k coefficients ranged from �0.07 to0.89 (median, 0.53) (anterior and posterior fractures) andfrom 0.03 to 0.88 (median, 0.36) (transverse fractures).After reviewing these results, members of the SCCG unan-imously agreed that a revision of categories 1c and 2c wasrequired and a final proposal was made, as presented inFigure 3.

Discussion

The SCCG was established to develop and validate a clini-cally useful comprehensive scapula classification system. Ina first phase, special interest was targeted at testing itsreliability to allow its integration in a clinical setting. Thisstudy focuses on glenoid fractures.

In our series of scapula fractures, the involvement ofarticular segment fractures was estimated for 38% of the120 cases. This incidence is much higher compared withdata reported by Goss15 suggesting that fractures of theglenoid cavity account for up to 10% of all scapula frac-tures. Ideberg et al18 also reported involvement of theglenoid cavity in about 30% of scapula fractures,whichdsimilar to our own studydmight be due to the caseselection process. Our scapular fracture cases wereconsecutively collected from 2 major level I trauma centersin the United States and central Europe regardless of theirtreatments. To achieve a precise diagnosis, both plainradiographs and CT scans were obtained. Obtaining thelatter, especially in combination with 3D volume rendering,is believed to increase the accuracy in detecting glenoidfractures and, therefore, might also explain the higherproportion of glenoid fractures detected in our studycompared with those from 3 more recent publications.2,8,21

In an earlier evaluation session using only radiographs for33 cases, 8 cases (24%) were diagnosed with glenoidinvolvement. The misdiagnosis of glenoid fractures maylead to either joint instability or joint incongruence, withthe potential consequence of secondary arthrosis, so it ismandatory to identify these fractures precisely. We stronglyrecommend performing a CT scan routinely in all cases ofscapular fractures.

Using the entire scapula classification system providedby the SCCG, the involved shoulder surgeons wereextremely accurate (>90%) in identifying all fracturesinvolving the articular segment (code F). According to thefocused classification system, the incidence of anterior rimfractures was 2 times higher compared with posteriorfractures in our series. Single transverse or short obliquefractures were slightly more frequent compared with ante-rior rim fractures. Multifragmentary fractures with centralfracture-dislocation were rare; this type of fracture patternwas not found in any of the 120 cases. The use of the

Figure 7 Fracture of the articular segment not involving the glenoid fossa (F0). � 2012, Jaeger et al.

Table I Coding agreement and reliability for glenoid fossa fractures according to focused system

Fracture subsets Categories) No. of casesy Full agreement k Coefficient

Overall Pair-wise (21 pairs)

Minimum Median Maximum

Simple fossa (F1) 50% 0.66 0.35 0.67 0.821 14 0.702 7 0.653 17 0.62

Multifragmentary fossa (F2) 4/5 7/0 100%F1(1) and F1(2) combined a/b/c 4/9/8 24% 0.46 �0.07 0.53 0.89F1(3) a/b/c 3/8/6 29% 0.39 0.03 0.36 0.88

) The categories of 1 to 5 represent the following: 1, anterior simple rim or oblique fracture; 2, posterior simple rim or oblique fracture; 3, simple

transverse or short oblique fracture; 4, multifragmentary fossa with more than 2 fracture line exit points; and 5, multifragmentary fossa with central

fracture-dislocation (no exit line through the rim).y Estimated by latent class analysis and the distribution of the surgeons’ codes.

Classification of glenoid fractures 7

detailed subclassification seems to be reproducible withhigh accuracy (>80%) in all categories of F1 (1, 2, and 3).A more detailed description of rim fractures with respect totheir location to the equator becomes less reliable and ismost pronounced when differentiating short oblique frac-tures [F1(3)]. This cannot be done without a CT scan andwas still not highly reliable with 3D volume rendering. Onereason for this outcome might be the method used in thisstudy. To provide the same data for all participatingsurgeons, 1 shoulder surgeon performed the volumerendering of all CT scans. The result was presented to thesurgeons as a movie sequence in a 360� horizontal and 360�

vertical rotation; this setting was chosen based on commonpractice (usually of the radiologist) in presenting thesetypes of data for further evaluation. Subsequently, the othershoulder surgeons did not perform their own 3D analysis,that is, they were unable to rotate or focus the 3D scans

according to their own needs, and as such, their accuracymay have been limited. On the other hand, these resultshighlight the difficulty in distinguishing fractures borderingbetween 2 categories. Because the distribution of all gle-noid fractures can be considered a continuum of possiblepatterns, any category is inherently associated with somelevel of imprecision for adjacent fracture pathologies.

Former scapula classifications, especially of the glenoid,were conducted based on the assessment of plain radio-graphs.5,15,24 All of them were published without furtherreliability assessment. To our knowledge, this is the firststudy providing an evaluation of classification reliabilityand accuracy. Armitage et al3 recently provided a frequencymap of surgically treated scapula fractures. To our knowl-edge, this was the first article until now describing the useof 3D CT volume rendering to enhance the mapping ofscapula fractures. In contrast to our study, Armitage et al

Table II Classification accuracy for patterns of simple frac-tures of glenoid fossa

Surgeons and fracture categories Fracture classes

1 2 3

Class size 36% 19% 45%Surgeon 11: Anterior rim or oblique 85%) 34% 29%2: Posterior rim or oblique 0% 64%) 0%3: Transverse or short oblique 15% 2% 71%)

Surgeon 21: Anterior rim or oblique 88%) 1% 1%2: Posterior rim or oblique 1% 80%) 13%3: Transverse or short oblique 11% 18% 87%)

Surgeon 31: Anterior rim or oblique 91%) 2% 1%2: Posterior rim or oblique 0% 96%) 6%3: Transverse or short oblique 9% 3% 93%)

Surgeon 41: Anterior rim or oblique 98%) 2% 9%2: Posterior rim or oblique 0% 69%) 0%3: Transverse or short oblique 2% 29% 91%)

Surgeon 51: Anterior rim or oblique 98%) 18% 1%2: Posterior rim or oblique 0% 64%) 6%3: Transverse or short oblique 1% 18% 93%)

Surgeon 61: Anterior rim or oblique 98%) 2% 19%2: Posterior rim or oblique 1% 80%) 13%3: Transverse or short oblique 1% 18% 68%)

Surgeon 71: Anterior rim or oblique 91%) 2% 6%2: Posterior rim or oblique 8% 96%) 0%3: Transverse or short oblique 1% 2% 93%)

Classification accuraciesMinimum 85% 64% 68%Median 91% 80% 91%Maximum 98% 96% 93%

) Percentages indicate surgeon classification accuracies (percent-

ages of correct classification).

8 M. Jaeger et al.

mainly focused on scapula body and neck fractures.Although they did not propose a classification system forglenoid fractures, they documented involvement of theglenoid in 17% of all scapula fractures. This incidence issurprisingly low considering the fact that these authorsincluded fractures undergoing operative treatment.According to the current treatment decision process,24 thefracture population of Armitage et al should have a biastoward including proportionally more glenoid fractures. Onthe other hand, Ada and Miller1 reported glenoid fracturesin 10% of their series based on assessments using plainradiographs. Such heterogeneity in the proportion of gle-noid fractures may be a reflection that the distribution offracture patterns is likely to differ between settings and, assuch, a comparison between scapula fracture studiesrequires adequate classification and documentation.

The combination of a glenoid fracture and scapula bodyfracture is frequent, so it is also important to havea combined classification considering both the glenoid andthe rest of the scapula. This classification system should bedescriptive and comprehensive, as well as aid in thedecision-making process for individual therapy, especiallyif minimally invasive approaches are intended.12 To ourknowledge, the current scapula classification system ismost comprehensive as well as highly reliable and accurate.Other existing scapula fracture classification systems lacka more focused description of glenoid fractures. Thecommonly used classification for glenoid fracturesaccording to Ideberg et al18 and Goss15 does not considerthe rest of the scapula; simple glenoid fracture linesbetween anterior and posterior rim fractures (types Ia andIb) and transverse fractures (types II-V) are distinguished,and type VI fractures represent multifragmentary fractures.The classification of scapular fractures by Ada and Miller1

distinguishes among 4 types of fractures including type IIIglenoid fractures; however, further subclassifications ofglenoid fractures do not exist.

This evaluation is limited by its sample size, although it isone of the largest ever used in such a project. There is oftena tradeoff between increasing the case collection and makingthe evaluation feasible when busy clinicians are involved.Results should be considered with caution when the numberof cases in any category is low, such as in the more detailed a,b, and c category system. In addition, nothing can be saidabout the reliability of coding multifragmentary glenoidfossa fractures with central fracture-dislocation. Althoughclassification of F1 simple fossa fractures according toanterior rim, posterior rim, or transverse patterns appearsreasonably reliable and accurate, location of these fractureswith regard to the 4 quadrants remains open for furtherclarification and evaluation. The wide range of surgeons’pair-wise k coefficients shows that different applications ofthe definitions occurred among surgeons; thus, furtherclarification and specification of the system are required.Transverse or short oblique fractures typically separatea process (ie, coracoid 3c) or involve the body (3a and 3b).The 3a and 3b fractures are typically located near theequator. In contrast, 1c and 2c rim fractures are much smallerbecause they are located near the glenoid edge.

The classification process and evaluation sessions wereperformed by experienced surgeons only, which may notreflect the validity of the system in general practice. Asadvocated by Audig�e et al,4 a further study is required toassess the reliability and accuracy of this classificationsystem when used by surgeons of varying education andexperience. Such evaluation would also consider the finalchange made to the rim fracture categories. This study couldalso be influenced by the preparation of the 3D volumerendering datasets by only the main author. Furthermore, thisreport is not intended to make recommendations regardingany treatment modalities based on this classificationsystem at this early stage; rather, our intent is to provide

Classification of glenoid fractures 9

a comprehensive, reliable, and accurate classification systemas a base for further clinical examinations.

Conclusion

In summary, it was possible to create a comprehensivescapula classification system focusing also on fractures ofthe articular segment. This classification system is believedto be clinically relevant and appears reasonably reliableand accurate with respect to glenoid fractures at both thebasic and more detailed levels. Further validation of themost detailed system in a larger case samplewith surgeonsof different training backgrounds should be considered.

Acknowledgments

The authors thank the following persons for their partici-pation and support in the development of the basic scapulaclassification system: Edward Harvey (Montreal GeneralHospital, Montreal, Quebec, Canada), Dolfi Herscovici(Florida Orthopaedic Institute, Temple Terrace, FL, USA),and JulieAgel andSeanNork (HarborviewMedicalCenter,Seattle, WA, USA). The authors also thank M. Wilhelmi,PhD (AO Clinical Investigation and Documentation), forthe preparation and copy editing of this manuscript.

Disclaimer

The authors, their immediate families, and any researchfoundations with which they are affiliated have notreceived any financial payments or other benefits fromany commercial entity related to the subject of this article.

The surgeons who participated in the successive face-to-facemeetingswere supportedwith regard to their travelexpenses and received a small per-diem payment. Supportfor this study was provided by the AO Foundation.

Supplementary data

Supplementary data related to this article can be foundonline at http://dx.doi.org/10.1016/j.jse.2012.08.003.

References

1. Ada JR, Miller ME. Scapular fractures. Analysis of 113 cases. Clin

Orthop Relat Res 1991:174-80.

2. Anavian J, Conflitti JM, Khanna G, Guthrie ST, Cole PA. A reliable

radiographic measurement technique for extra-articular scapular

fractures. Clin Orthop Relat Res 2011;469:3371-8. http://dx.doi.org/

10.1007/s11999-011-1820-3

3. Armitage BM, Wijdicks CA, Tarkin IS, Schroder LK, Marek DJ,

Zlowodzki M, et al. Mapping of scapular fractures with

three-dimensional computed tomography. J Bone Joint Surg Am 2009;

91:2222-8. http://dx.doi.org/10.2106/JBJS.H.00881

4. Audig�e L, Bhandari M, Hanson B, Kellam J. A concept for the vali-

dation of fracture classifications. J Orthop Trauma 2005;19:404-9.

http://dx.doi.org/10.1097/01.bot.0000155310.04886.37

5. Audig�e L, Bhandari M, Kellam J. How reliable are reliability studies

of fracture classifications? A systematic review of their methodolo-

gies. Acta Orthop Scand 2004;75:184-94. http://dx.doi.org/10.1080/

00016470412331294445

6. Audig�e L, Hunter J, Weinberg A, Magidson J, Slongo T. Development

and evaluation process of a paediatric long-bone fracture classification

proposal. Eur J Trauma 2004;30:248-54. http://dx.doi.org/10.1007/

s00068-004-1428-3

7. Bahk MS, Kuhn JE, Galatz LM, Connor PM, Williams GR Jr. Acro-

mioclavicular and sternoclavicular injuries and clavicular, glenoid, and

scapular fractures. J Bone Joint Surg Am 2009;91:2492-510.

8. Bartonicek J, Tucek M, Fric V. Radiographic evaluation of scapula

fractures [in Czech]. Rozhl Chir 2009;88:84-8.

9. Burstein AH. Fracture classification systems: do they work and are

they useful? J Bone Joint Surg Am 1993;75:1743-4.

10. Euler E, Habermeyer P, Kohler W, Schweiberer L. Scapula

fracturesdclassification and differential therapy [in German]. Ortho-

pade 1992;21:158-62.

11. Garbuz DS, Masri BA, Esdaile J, Duncan CP. Classification systems in

orthopaedics. J Am Acad Orthop Surg 2002;10:290-7.

12. Gauger EM, Cole PA. Surgical technique: a minimally invasive

approach to scapula neck and body fractures. Clin Orthop Relat Res

2011;469:3390-9. http://dx.doi.org/10.1007/s11999-011-1970-3

13. Goss TP. Frequency of scapular fractures. In: Rockwood CA,

Matsen FA, Wirth MA, Lippitt SB, editors. Fractures of the scapula.

3rd ed. Philadelphia: Saunders; 2004. p. 413-54.

14. Goss TP. Scapular fractures and dislocations: diagnosis and treatment.

J Am Acad Orthop Surg 1995;3:22-33.

15. Goss TP. Fractures of the glenoid cavity. J Bone Joint Surg Am 1992;

74:299-305.

16. Hardegger FH, Simpson LA, Weber BG. The operative treatment of

scapular fractures. J Bone Joint Surg Br 1984;66:725-31.

17. Harvey E, Audige L, Herscovici D, Agel J, Madsen JE, Babst R, et al.

Development and validation of the new international classification for

scapula fractures. J Orthop Trauma 2012;26:364-9. http://dx.doi.org/

10.1097/BOT.0b013e3182382625

18. Ideberg R, Grevsten S, Larsson S. Epidemiology of scapular fractures.

Incidence and classification of 338 fractures. Acta Orthop Scand 1995;

66:395-7.

19. Martin JS, Marsh JL. Current classification of fractures. Rationale and

utility. Radiol Clin North Am 1997;35:491-506.

20. Mayo KA, Benirschke SK, Mast JW. Displaced fractures of the gle-

noid fossa. Results of open reduction and internal fixation. Clin Orthop

Relat Res 1998:122-30.

21. Tadros AM, Lunsjo K, Czechowski J, Corr P, Abu-Zidan FM.

Usefulness of different imaging modalities in the assessment of

scapular fractures caused by blunt trauma. Acta Radiol 2007;48:71-5.

http://dx.doi.org/10.1080/02841850601026435

22. Uebersax JS, Grove WM. Latent class analysis of diagnostic agree-

ment. Stat Med 1990;9:559-72.

23. Vermunt J, Magidson J. Latent class cluster analysis. In:

Hagenaars JA, McCutcheon AL, editors. Applied latent class analysis.

Cambridge: Cambridge University Press; 2002. p. 89-106.

24. Zlowodzki M, Bhandari M, Zelle BA, Kregor PJ, Cole PA. Treatment

of scapula fractures: systematic review of 520 fractures in 22 case

series. J Orthop Trauma 2006;20:230-3. http://dx.doi.org/10.1097/

00005131-200603000-00013