AST classification of proximal humeral fractures: introduction and interobserver reliability assessment
- 297 Downloads
This article introduces an alphanumeric AST (Articular, Surgical neck, Tuberosities) classification of proximal humeral fractures, based on the number, localization, and displacement of articular and extra-articular fragments. All possible cases of proximal humeral fractures can be assessed from a single figure using this classification. The aim of the study was thus to describe the AST classification and to assess interobserver reliability.
This classification is based on a single figure, allowing an easy description of the anatomic variants of different proximal humeral fractures. The severity of the fracture is determined by the fragment displacement in angular degrees and the major linear displacement in millimeters. AST reproducibility was assessed and compared with Neer, AO, and Duparc classifications, commonly used in clinical practice. The interobserver agreement was measured with Cohen’s kappa coefficients and their 95% confidence intervals.
Thirteen independent observers analyzed a total of 64 X-rays. Overall kappa coefficients were 0.34, 0.29, 0.24, and 0.25 for AST, Neer, AO, and Duparc classifications, respectively.
The AST classification, which is easier to use because it is based on only one figure, is at least as reproducible as other proximal humeral fracture classifications.
KeywordsProximal humeral fractures Classification Interobserver reliability
Proximal humeral fractures show considerable morbidity rates. Classification systems can be used to guide treatment and to select the best possible conservative or surgical management strategy. Unfortunately, classifications of proximal humeral fractures show poor intra- and interobserver reliability [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]. We developed a new proximal humeral fractures classification system, the AST (Articular, Surgical neck, and Tuberosities), which relies on the articular and extra-articular characteristics of the fractures lines and displacement of the fragments. We then conducted an interobserver reproducibility study to assess the concordance between this classification and three other commonly used classifications: Neer, AO, and Duparc [12, 13, 14, 15, 16].
Materials and methods
Neer, AO, and Duparc classifications are three commonly used methods in clinical practice. Neer’s classification was proposed in 1970 and is currently the most used internationally [12, 13]. The Neer scheme uses Codman’s figure, which individualizes four parts (head, two tuberosities, and metaphysis). Neer’s classification is based on one figure, showing the different kinds of fractures. Half of the picture shows fracture dislocation, and the term “Part” is use by Neer for a displacement of at least 45° and/or 1 cm. The AO classification was proposed in 1989 [14, 15]. It is alphanumeric according to usual terminology and distinguishes unifocal extra-articular fractures (A), bi-focal extra-articular fractures (B), and articular fractures (C). A number is associated with each letter to describe the gravity of the displacement. Duparc’s classification was published in 1976 . It sorts fractures into two major groups: articular and extra-articular. Articular fractures correspond to the anatomical neck. “C” stands for cephalic and “T” for tuberosity. They can be isolated or associated with tubercular fractures. Duparc’s classification distinguishes CT1, which are not displaced; CT2, which are displaced and impacted; and CT3, which are disengaged. CT4 are complex articular fractures associated with dislocation. Fractures from the notch are individualized in case of dislocation and are classified in posterior and anterior dislocations. Extra-articular types correspond to isolated lesser or greater tuberosity fractures and under tuberosity fractures, with description of the fracture line in comparison with the humeral head. Under tuberosity fractures with an intermediate fragment of the greater or lesser tuberosity are separated.
A reproducibility study was performed by asking independent observers to characterize radiographic images using the AST, Neer, AO, and Duparc classifications. Patients consecutively operated on at our institution over the study period were included in the analysis. X-rays of proximal humeral fractures were selected and printed on CDs. Most X-rays included frontal and lateral views. Thirteen observers were invited to analyze the X-rays and classify the fractures according to four classifications: AST, Neer, AO, and Duparc. The observers were all orthopedic surgeons with an interest in shoulder surgery. Three were shoulder specialists. Five observers were currently working in the department where the AST classification originated (intra-institutional), and the other eight surgeons were from other institutions (extra-institutional). The first author, who designed the AST classification, did not take part in this study.
In order to guarantee independency of reading between the four classifications, different orders for filling the grid were assigned to each participant (ABCD, BCDA, CDAB, etc.) using a randomized list.
Each classification with a simple Kappa coefficient for binary variables;
The overall classification with a weighted Kappa coefficient for multiple modality variables.
The rule used to interpret reliability was proposed by Landis and Koch in 1977 . Finally, the institution of the participants was investigated as a source of possible confusion. We used a z test to compare Kappa coefficients. The statistical threshold for significance was set at 0.05. Statistical analyses were performed with the SAS software, version 9.1 (SAS Institute Inc., 2002).
Inter-observer reliability for observers belonging (intra-) or not (extra-) to the institution where the AST classification was designed
Kappa [IC 95%]
0.24 [0.22; 0.26]
0.28 [0.24; 0.32]
0.20 [0.18; 0.22]
0.34 [0.32; 0.36]
0.38 [0.34; 0.42]
0.32 [0.30; 0.34]
0.25 [0.23; 0.27]
0.38 [0.34; 0.42]
0.31 [0.29; 0.33]
0.29 [0.27; 0.31]
0.29 [0.25; 0.33]
0.27 [0.23; 0.31]
The potential clinical usefulness of the AST classification is mainly based on its simplicity of use and on the evaluation of fracture gravity by quantification of the displacement of fragments. The simplicity of use is due to the utilization of a single figure (Fig. 1), representing all possible proximal humeral fractures. This figure includes pictures and an alphanumeric denomination of the fractures. It is especially interesting that four out of 10 pictures (A2, A3, S, ST) represent most fractures encountered in clinical practice. This classification is based on the articular or extra-articular localization of the fracture line, rather than the number of fragments. Type A fractures are articular fractures, and Type S fractures are extra-articular fractures. Regarding articular fractures, the severity of lesions is assessed by the displacement, rather than the number of fragments [16, 18, 19]. Usually, articular fractures are associated with both tuberosities, but exceptions may occur. Three-fragment fractures can also be encountered in clinical practice, as described by J. Duparc and Tamai [11, 16]. The number of fragments is not essential in the AST classification, which is better in terms of simplicity. Another potential clinical relevance is the measurement of displacement. Displacement is evaluated by two numbers, which are the cephalic tilting angle for an articular fracture, or the angulation for an extra-articular fracture, and the major linear displacement of the fragments in millimeters. These two numbers, which are easy to define, allow surgeons to quantify the severity of the lesion in addition to the common and simple denomination of each case. In our institution, we usually operate on the fracture when there is at least 30° of tilt and/or 1 cm of linear displacement, compared with the 45° and 10 mm proposed by Neer’s classification. Another study showed that a tuberosity fragment must be operated on when there is at least 5 mm of displacement .
According to our results, the AST classification is at least as reproducible as the three other classifications evaluated in this study. The kappa indexes obtained should be compared with those reported in the literature, which are all rather poor [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 23]. Some authors insist on the relevance of a training course in order to improve comprehension and reproducibility in Neer’s system [3, 4, 24]. In particular, two authors stated that the poor reproducibility observed with Neer’s classification was due to its complexity [4, 25]. They proposed a simplification but did not show significant improvement considering reproducibility. Others authors considered that tomodensitometry does not improve comprehension of these fractures [2, 7, 23]. On the contrary, Edelson proposed a classification based on 3D CT pictures for five types of fractures . Another study analyzed the reproducibility of a complex classification called MTM, although the authors concluded that it was unusable because of the poor reproducibility and that a simplification was necessary . Many authors insist that displacement is a good indicator of the gravity of these fractures [11, 18, 19]. It seems that reproducibility is acceptable for minor displacements, but it becomes unacceptable for major displacements. Tamai et al. after studying 22 operated complex fractures divided into three and four parts, noticed that eight out of 22 fractures did not correspond to any usual classification type—particularly some of the three fragments fractures . This argument supports the use of the AST classification because it is not based on the number of fragments.
The AST is a classification system developed by the authors for the last 10 years and is already being used in clinical practice for proximal humeral fractures. This new classification is easy to use and is based on a single figure (Fig. 1) including all types of proximal humeral fractures. The severity of the fracture is assessed by the displacement of fragments. This study shows an inter-observer reproducibility at least as good as the three other classifications most often used in current clinical practice.
We would like to thank the physicians who participated in this study for their extraordinary help, and for the reading, classification, and analysis of four different classifications of a total of 64 X-rays of proximal humeral fractures: Dr Beau—Metz, Dr Berrichi—Metz, Dr Boughrebi—Amiens, Dr Camus—Dunkerque, Dr Dujardin—Orléans, Dr Favreul—Lyon, Pr Galois—Nancy, Dr Ionescu—Metz, Dr Irrazi—Metz, Dr Kany—Toulouse, Dr Khalife—Metz, Pr Roussignol—Rouen, Dr Schenk—Strasbourg, Dr Schwartz—Colmar, Dr Scarlat—Toulon, Dr Seboiu—Metz, Pr Simon—Lyon, Dr Travers—Lyon, as well as the talented Mrs L. Collette—Bruxelles, for the figures.
Conflict of interest
The authors have no conflict of interest to declare.
- 13.Neer CS. Four Segment Classification (2008) In shoulder reconstruction. W.B. Saunders, Philadelphie, pp. 366–370Google Scholar
- 14.Jakob RP, Kristiansen T, Mayo K, Ganz R, Müller ME (1984) Classification and aspects of treatment of fractures of the proximal humerus. Bateman and WelshGoogle Scholar
- 17.Boehler J (1935) Boehler (The treatment of fractures), 4th edn. Williams Wood, BaltimoreGoogle Scholar