minerva medica copyright

2 downloads 104 Views 140KB Size Report
fractures of the proximal humerus. To understand the pathoanatomy of prox- imal humeral fractures Codman suggested studying the involvement of four distinct.
M C IN O E P R Y V R A IG M H E T® D IC A

MINERVA ORTOP TRAUMATOL 2009;60:447-60

The Neer classification for fractures of the proximal humerus: a narrative review S. BRORSON 1, L. HENRIK FRICH 2, A. HRÓBJARTSSON 3

In 1970 Charles Sumner Neer, II (b.1917) published his classification system for fractures of the proximal humerus. The authors present the Neer classification and its modifications in a historical context, review the literature on observer agreement and consider comments from leading shoulder surgeons. One of the most eagerly discussed aspects of the classification system has been the low inter- and intraobserver agreement reported in more than a dozen studies. Systematic training seems, however, to improve agreement to an acceptable level, especially among specialists. The authors discuss current attempts to improve the Neer classification and the features of a satisfying classification for proximal humeral fractures. The classification of displaced fractures of the proximal humerus remains a challenge in clinical orthopedics and research. Key words: Observer variation - Shoulder fractures - Classification.

“A [classification] system has scientific validity if the categories within it have their own unique natural history, prognosis, and treatment requirements.” 1 History (1896-1969)

C

lassifications of proximal humeral fractures in the 20th century can roughly be divided into anatomical and etiological classification systems or combinations of those.

Corresponding author: S. Brorson, MD PhD, Department of Orthopedic Surgery, Herlev University Hospital, Herlev Ringvej 75, DK-2730 Denmark. E-mail: [email protected]

Vol. 60, No. 5

1Department

of Orthopedic Surgery, Herlev University Hospital, Herlev, Denmark 2Department of Orthopedic Surgery, Odense University Hospital, Odense, Denmark 3The Nordic Cochrane Centre, Rigshospitalet, Copenhagen, Denmark

Anatomical classification systems

Emil Theodor Kocher (1841-1917) published the last major pre-radiological classification system for fractures of the proximal humerus in 1896,2 the same year as Wilhelm Conrad Roentgen (1845-1923) presented his important work on “a new kind of rays”.3 Kocher’s classification was anatomically based and developed in analogy to his classification of fractures of the proximal femur (Figure 1). Fractures of the proximal humerus were classified into “supratubercular” (intracapsular) fractures through the anatomical neck and “infratuberbular” (extracapsular) fractures below the anatomical neck. The “infratuberbular” fractures were further divided into “pertubercular” fractures involving the tuberosities and “subtubercular” fractures through the surgical neck. Adding the orientation of the fractured humerus to the anatomical description enabled Kocher to distinguish a great number of fracture types (Figure 2). Kocher conducted biomechanical studies inflicting forces on bone specimens to study

MINERVA ORTOPEDICA E TRAUMATOLOGICA

447

BRORSON

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

1 1

a 2 2

M C IN O E P R Y V R A IG M H E T® D IC A

3

Reginald Watson-Jones (1902-1972) 9 combined the mechanism of injury with the radiographic appearance (Figure 3). This principle of classification has been adopted in later classification systems (Figure 4).10 The distinction between abduction and adduction fractures is, however, problematic from a radiological perspective as an angulation can appear radiographically as either abduction or adduction fracture dependent on the rotation of the arm.

b

b

3

4

Figure 1.—Kocher”s anatomical classification for fractures of the proximal humerus and the proximal femur. Reprinted from Kocher.2

The four-segment approach (1934)

the correlation between fracture anatomy and the mechanism of injury. However, he did not take into account the degree of displacement considered important in later classification systems. Other anatomical classifications have been suggested by 20th century authors 4-6 and elements from Kocher”s anatomical classification can be found in most later classification systems, for example in the Neer classification.7 Etiological classifications systems

Ernst Dehne (1905-1983) 8 classified fractures of the proximal humerus according to the mechanism of injury and the deforming muscular forces. Three mechanisms were distinguished (lateral, dorsal and central) with distinct clinical properties.

A

B

C

D

“It is remarkable in studying the literature of fractures in the upper end of the humerus to find how little detail is given to the probable relations which the muscles and bursæ bear to the fragments”.11 Ernest Amory Codman (1869-1940) 12 was a multitalented orthopaedic surgeon and a pioneer in skeletal radiology,13 bone sarcoma,14 medical registry 14 and hospital organization.15 He published an atlas of skeletal radiographs in 1896 and stressed the practice of obtaining at least two radiographic views to visualize bone pathology.13 The eponym “Codman’s triangle” is well known to orthopedic surgeons. During medical school Codman was on an elective in Vienna and was introduced to the subacromial bursa. He developed an early and persistent interest in shoulder pathology,

E

F

G

H

Figure 2.—Examples of fracture patterns classified by Kocher.2 A) Impacted pertubercular fracture in adduction; B) Pertubercular fracture in adduction, extension and anterior displacement (healed fracture); C) Pertubercular fracture in abduction and anterior displacement; D) Pertubercular fracture in extension and anterior displacement; E) Impacted subtubercular fracture in abduction; F) Subtubercular fracture in abduction and anterior displacement; G) Impacted subtubercular fracture in adduction (healed fracture); H) fracture. Combined pertubercular fracture and anatomical neck fracture.

448

MINERVA ORTOPEDICA E TRAUMATOLOGICA

October 2009

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

BRORSON

M C IN O E P R Y V R A IG M H E T® D IC A

epiphyseal union (Figure 5): 1) the greater tuberosity; 2) the lesser tuberosity; 3) the humeral head; and 4) the humeral shaft. Thus, the four-segment approach was radiographically defined, but it also took into account the muscular forces acting on the four segments and the condition of the soft parts. However, from a modern point of view it lacks a conception of displacement. Codman”s approach profoundly influenced later classifications of proximal humeral fractures.

A

B

Figure 3.—Example of classification according to mechanism of injury: A) adduction fracture B) abduction fracture. Modified from Watson-Jones.9

and published his first cases early in the 20th century.16, 17 In 1934 Codman published the first book on shoulder surgery entitled “The Shoulder: rupture of the supraspinatus tendon and other lesions in or about the subacromial bursa”.11 Focus in this monumental work was on the role played by the soft parts, especially the subacromial bursa and the supraspinatus tendon, in shoulder pathology. Seventy pages were devoted to the role of the supraspinatus and the subacromial bursa in fractures of the proximal humerus. To understand the pathoanatomy of proximal humeral fractures Codman suggested studying the involvement of four distinct anatomical segments along the lines of the

The Neer classification (1970/1975)

Neer first referred to Codman’s classification in 1953 (Figure 6).18 He had observed a general confusion in the literature defining “fracture-dislocations” and “subluxations” of the humeral head (Figure 7).19 Lack of common terminology in the literature precluded consistent reporting of clinical outcome and the establishing of treatment guidelines. Neer attempted to correlate terminology and pathology before going on to the details of treatment and results.18 In his classic 1970 papers 7, 20 Neer retrospectively analyzed a series of 300 “displaced” fractures of the proximal humerus selected “at random”. Roentgen appearance (anteriorposterior and lateral views were defined) was

Tronc Bras

A

Tronc

Bras

45°

Figure 4.—Example of classification according to mechanism of injury: A) adduction fracture B) abduction fracture. Modified from Razemon and Baux.10

Vol. 60, No. 5

MINERVA ORTOPEDICA E TRAUMATOLOGICA

449

BRORSON

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

c

a

a

b

b

d

c

c

a

d

c b

a

b d

c

a

1

b

d

d

M C IN O E P R Y V R A IG M H E T® D IC A

2

c

a

c

b

d

d

a

c

a

c b

a

b

3

c

a b d

b

d

4

d

c

a

a

b d

a

b

d

c

b

d

c

a

c b d

a

c b

d

a

c

b d

Figure 6.—The four segments of the proximal humerus. Reprinted from Neer.18

e

Figure 5.—The four segments of the proximal humerus and the variety of fracture patterns. All but the two patterns in the lower right corner (surgical neck fracture and greater tuberosity evulsion) are intracapsular and held together by the periosteum and rotator cuff (hatched part of the illustration in upper left corner). Evulsion of the facet of the insertion of the supraspinatus (lower right corner) is considered the most severe lesion causing mechanical problems and has a poor potential for healing. Reprinted from Codman.11

Greater tuberosity

Laser tuberosity Head of humerus

correlated to operative findings and photographs of surgically treated patients. Neer reported that as the work proceeded:

“…distinct anatomical categories became evident, [and] the classification became evident”.7

The Neer classification integrated fracture anatomy, biomechanics, and the notion of displacement. It was built upon knowledge of: 1) the effect of muscles attaching to free segments and rotator cuff integrity; 2) the effect of the vascular supply to the humeral head; and 3) the condition of the articular surface. Focus was on displacement rather than anatomical pattern of fracture lines. In Neer’s classification fractures of a similar type should be grouped together and separated from the more serious and less serious lesions.20 Thus, Neer intended to define an

450

Figure 7.—Four-part anterior fracture-dislocation. Reprinted from Neer.19

ordinal system. Neer’s extension of Codman’s four-segment approach was a breakthrough in classification of proximal humeral fractures. Neer defined six groups (Figure 8, horizontal rows) according to the displacement of Codman’s four segments, but regardless of additional fracture lines. In Group I no segment is displaced more than 1.0 cm or angulated more than 45°. According to Neer 80-

MINERVA ORTOPEDICA E TRAUMATOLOGICA

October 2009

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

I MINIMAL DISPLACEMENT

BRORSON

DISPLACED FRACTURES 3 PART

4 PART

M C IN O E P R Y V R A IG M H E T® D IC A

2 PART

II ANATOMICAL NECK

III SURGICAL NECK

B

A

C

IV GREATER TUBEROSITY

V LESSER TUBEROSITY

ARTICULAR SURFACE

VI FRACTURE DISCLOCATION

ANTERIOR

POSTERIOR

Figure 8.—The first Neer classification. Reprinted from Neer.7

85% 7, 21 of proximal humeral fractures are minimally displaced. They represent similar problems in management and prognoses regardless of the anatomical level. Group II includes anatomical neck fracture with displacement of the head segment. Group III includes displaced surgical neck fractures with the rotator cuff intact. They were further divided into 1) angulated; 2) separated; or 3) comminuted fractures. Group IV includes

Vol. 60, No. 5

greater tuberosity displacements with or without neck fracture. They can be two- or threepart fractures. Group V includes lesser tuberosity displacements with or without neck fracture. They can be two- or three-part fractures. Group IV and V merge into the four-part fracture with neck fracture and both tuberosities displaced. Group VI includes fracture-dislocations further divided into anterior and posterior dislocations, two-, three-,

MINERVA ORTOPEDICA E TRAUMATOLOGICA

451

BRORSON

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

segment is a concept rather than a numerical classification.26 Displacement is a continuum and there will be situations where the true condition of the humeral head must be determined at the operation. The editor, Henry R. Cowell,27 supported Burstein’s view and stated that if even experts can only categorize intraoperatively, general orthopedists cannot be expected to categorize proximal humeral fractures reproducibly based on preoperative radiographs. Burstein added that if only shoulder experts have the skills to use the classification system it would diminish its value almost to the point of uselessness.28 Charles A. Rockwood 29 stated that the Neer classification is just one tool of several used in planning of treatment and Louis U. Bigliani 30 further interpreted Neer’s statement from the 1975 paper and suggested that the Neer classification is designed to be somewhat arbitrary, leaving some room open for discussion about different fracture patterns and their appropriate treatment. Choosing a scheme with more categories may allow a greater prognostic stratification, but at the price of less observer agreement.

M C IN O E P R Y V R A IG M H E T® D IC A

or four-part fractures. This group also includes articular surface fractures divided into headimpression fractures and head-splitting fractures. Several objections were raised upon the original Neer-classification. Within the six groups fractures have a diversity of prognoses. Displaced fractures of the anatomical neck (Group II) are usually more severe than surgical neck fractures (Group III). Thus, the classification is not ordinal. Neither is it clear whether four-part fracture-dislocations (Group VI) are more severe than other fourpart patterns (Group IV/V).21 Further, it is not clear whether the surgical neck or the anatomical neck are involved in three- and four-part fractures. Finally, the groups are not mutually exclusive as four-part fractures appear as both Group IV and V. Neer revised the classification slightly in 1975.22 Definitions of categories were restated and the numbered groups were eliminated; now referring only to the segments involved. Neer further stated his view on classification of proximal humeral fractures: “The four-segment classification is not a numerical classification but rather a concept that makes it possible to describe any displaced fracture or fracture-dislocation with standard terminology”.22 The debate in Journal of Bone and Joint Surgery (1993-4)

Two observer studies reporting a low level of observer agreement within the Neer classification were published in Journal of Bone and Joint Surgery(Am) in 1993.23, 24 In a critical editorial Albert H. Burstein 25 questioned the validity of the Neer classification. He stated that a fracture classification should help the surgeon to choose an appropriate method of treatment, and provide the surgeon with a reasonably precise estimation of the treatment. Based on the reported levels of observer agreement he concluded that the Neer classification system is a poor tool leading only to confusion and more conflicting clinical results. In his reply, Neer repeated that the four-

452

Neer’s response to the critics (2002)

In a review paper Neer summarized and reflected upon fifteen years of criticism of his classification.1 He suggested that the reported inconsistencies were caused by low quality roentgen images and inexperienced interpreters and concluded that “Knowledgeable interpreters do well and the inexperienced and unpractised interpreters confuse the data with low scores”.1 This statement was not supported by data at the time. Later studies have, however, demonstrated an effect of training of observers.31, 32 Neer also stated that “The more categories there are to select from, the greater the chance of inconsistency”.1 However, the measures of agreement used in the cited observer studies (kappa statistics) correct for the agreement expected by chance. In classification systems using many categories the expected agreement by chance will be lower than in systems using few categories. Since the 1975 version of the Neer classifi-

MINERVA ORTOPEDICA E TRAUMATOLOGICA

October 2009

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

BRORSON

M C IN O E P R Y V R A IG M H E T® D IC A

marize and update data from studies evaluating agreement between and within independent observers classifying fractures of the proximal humerus according to Neer. Kappa as a measure of observer agreement

III

II

IV

I

Figure 9.—Valgus impacted four-part fracture. Reprinted from Jakob.33

cation several studies have reported a prognostically distinct subgroup of four-part fractures, “valgus impacted four-part fractures” (Figure 9), characterized by four segments with variable displacement of the tuberosities and valgus impaction of the humeral head.33, 34 This subgroup was not included in the original Neer classification. In the 2002 revision Neer suggested to consider four-part fractures a continuum of lateral displacement. The valgus impacted four-part fracture can be regarded a borderline “precursor” to the “classical” four-part fracture (the medial periosteum may remain intact). Neer included the valgus impacted four-part fracture pattern into the 2002 version of the classification. He stressed that in the continuum of displacement in fractures there will always be borderline lesions and it may be difficult to agree on borderline lesions. Observer agreement within the Neer classification system: a review of observer studies

Several reviews of observer agreement within the Neer classification system have been published.35-49 Below the authors sum-

Vol. 60, No. 5

The studies included in this review report observer agreement using kappa statistics. Kappa, as defined and interpreted by Landis and Koch,39 is the most commonly used measure of agreement for categorical data. In kappa statistics the observed agreement is adjusted for the agreement expected by chance. Values of kappa range from 1 (perfect agreement) over 0 (expected agreement by chance) to -1 (perfect disagreement). Several limitations of kappa statistics should be mentioned. First, kappa values cannot be compared between studies with different prevalence of the classified items (i.e. Neer group).40 Second, kappa does not take into account the “prognostic difference” between the groups. For example, the “prognostic distance” between a non-displaced fracture and a displaced fracture of the anatomical neck is greater the “prognostic difference” difference between a displaced surgical neck fracture and a displaced fracture of the lesser tuberosity. Assigning weights to kappa can address such differences, but the studies using weighted kappa values do generally not report the weighting factors used. Third, widely different qualitative interpretations of numerical kappa values appear in the literature.35 In a broad review of 44 observer studies on fracture classification Audige et al.35 included eight studies on the Neer classification. They found a variety of interpretations of kappa and stated that direct comparison of kappa values is impossible due to differences in distribution of categories. They requested a systematic methodology for assessing observer agreement within fracture classification systems in general. In a systematic review Brorson and Hróbjartsson 36 included 12 observer studies of the Neer system. No significant difference was found in agreement between experienced and inexperienced observers classify-

MINERVA ORTOPEDICA E TRAUMATOLOGICA

453

BRORSON

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

classification does not improve agreement. They reported disagreement both on the number of fragments and on the assessment on displacement and angulation. The disagreement was not restricted to junior personnel. Brorson et al.37 included five observer studies in their review on observer agreement on displaced four-part fractures. The authors found that observers agreed less on four-part fractures than on the Neer classification in general. In displaced four-part fractures specialists agreed slightly more than non-specialists and advanced imaging modalities (CT or 3D CT scans) seemed to contribute more to agreement than in less displaced fracture patterns. No kappa values, however, exceeded 0.48 and training of observers did not seem to improve agreement on four-part fractures. The auhtors concluded that low agreement poses a problem for interpretation of results in clinical studies, and effects of interventions for displaced four-part fractures should be interpreted cautiously.

M C IN O E P R Y V R A IG M H E T® D IC A

ing according to Neer. The authors found kappa values for intra-observer agreement slightly higher than for inter-observer agreement. They found no clear improvement by adding computed tomography (CT) or 3D-CT scans to plain radiographs. No improvement was found by simplification of the classification to two, five, six, 13 or 15 groups. Differences in distribution of categories precluded a statistical pooling of data for observer agreement. Only one study reported kappa values for inter-observer agreement exceeding 0.60 (“substantial agreement” according to Landis and Koch).39 This value was obtained in a randomized study 31 testing the effect of standardized training of observers. Mahadeva et al.38 included seven observer studies in their review. They excluded studies using advanced imaging modalities and stated that variable quality of plain radiographs may contribute to observer variation. This statement was, however, not supported by their data. They found that binary TABLE I.

Inter-pbserver agreement (mean/range of kappa)

Study characteristics

Study

Cases N.

Observers N.

Neerclasses N.

Nondisplaced fractures (%)

Kristiansen 1988 41 Siebenrock 1993 24 Sidor 1993 23 Sidor 1994 42 Brien 1995 43 Bernstein 1996 44 Sallay 1997 45

100 95 50 50 28 20 12

4 5 5 5 4 4 9

5 5 16 16 13 16 6

Sjöden 1997 46 Sjoden 1999 47 Brorson 2002 48

26 24 42

10 7 24

16 16 16

Brorson 2002 31

425

14

16

Schwartz 2003 49 Schrader 2005 32

21 113

11 3

15 16

27 ns ns ns 0 20 20 0.35 (0.24-0.77) 3 12 21 21 0.414 21 0.626 ns ns

Brorson 2009 37

21-429

4-24

210

0-21

Radiographs only

0.30 (0.07-0.48)1 0.40 (0.25-0.51)1 0.48 (0.43-0.58) 0.39 and 0.49 0.48 (0.37-0.75) 0.52 0.58 (0.41-1.00) 2

CT or 3D-CT

Intra-observert agreement (mean/range of kappa) Radiographs or radiographs and CT viewed together on two occasion

0.60 (0.46-0.71)1 0.66 (0.50-0.83)

0.50

0.72 (0.63-0.79)

0.44 0.47

0.34 (0.20-0.85) 0.56 (0.27-0.73)

0.34

0.19-0.737

0.27

0.27

0.17 0.427 and 0.476, 7 0.457, 8 and 0.556, 7, 8 0.16-0.487

1 Weighted kappa, 2 Neck fractures, 3 Tuberosity fractures, 4 Displaced versus non-displaced, 5 Randomized trial, 6 After training, 7Including some CT-scans, 8 Agreement on treatment recommendation, 9 Review of data on four-part fractures from five observer studies, 10 Four-part versus non-four-part.

454

MINERVA ORTOPEDICA E TRAUMATOLOGICA

October 2009

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

BRORSON

observers or by reducing the number of categories. Observer agreement within other classification systems

M C IN O E P R Y V R A IG M H E T® D IC A A

A1

A2

A3

The AO classification

B

B1

B2

B3

C2

C3

C

C1

Figure 10.—The three types and nine groups of the AO classification for proximal humeral fractures. Reprinted from Müller et al.50 Types:A: Unifocal extra-articular (twopart); B: Bifocal extra-articular (three-part); C: Articular (three- or four-part); Groups: A1 to C3; Subgroups (not shown): 27 (three for each of the nine groups): A1.1 to C3.3.

Kappa values for inter- and intraobserver agreement within the Neer classification in 14 observer studies are summarized in Table I. Mean kappa values for inter-observer agreement range between 0.16 and 0.62 for plain radiographs and between 0.34 and 0.50 by adding CT or 3D CT scans. Mean kappa values for intra-observer agreement range from 0.34 and 0.73. No increase in mean kappa values was found among experienced

The AO classification for proximal humeral fractures 50-55 consists of three types, nine groups, and 27 subgroups hierarchically organized according to severity (Figure 10). To use the system one has to ask three questions, each of which has three possible answers.50 Valgus-impacted four-part fractures (C2.1) 33 were initially unique to the AO classification but were included in the Neer classification in the 2002 revision.1 The AO classification has several problems in common with the Neer classification. It is not clearly ordinal according to prognosis. An A1 fracture, for example, may require surgical treatment while an A2 fractures may be treated non-surgically. Not all subgroups are in use and the classification does not always guide treatment. Displaced four-part fractures according to Neer with lateral displacement of the humeral head and displacement of both tuberosities are not captured by the AO classification. Several studies have assessed observer agreement within the AO classification (Table II).24, 46, 47 Mean kappa values for inter-observer agreement range between 0.30 and 0.53. Mean kappa values for inter-observer agreement based on CT and 3D CT scans do not

TABLE II.

Cases N.

Observers N.

Inter-observer agreement (mean/range of kappa)

Interobserver agreement (mean/range of kappa)

Siebenrock 1993 24

95

5

0.53 (0.50-0.58) 1 0.42 (0.36-0.49) 2

0.58 (0.54-0.66) 1 0.48 (0.43-0.54) 2

Sjöden 1997 46

263

10

0.31 4 0.30 5

0.16-0.60

Sjöden 1999 47

246

7

0.32 4 0.34 5

0.29-0.74

Study

1 AO-types, 2 AO-groups, 3 All

Vol. 60, No. 5

had CT scans, 4 First classification round, 5 Second classification round, 6All had 3D CT scans.

MINERVA ORTOPEDICA E TRAUMATOLOGICA

455

BRORSON

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

3 Part

Shield fracture

Shieldfracture variants

Isolated greater tuberosity

M C IN O E P R Y V R A IG M H E T® D IC A

2 Part

Valgus

Varus

4 Part

Neutral

Shattered shield

Fracture dislocation Anterior (illustrated)

Posterior (not illustrated)

Head split

Figure 11 Edelson's classification. Reprinted from Edelson.55

exceed 0.34. Mean kappa values for intraobserver agreement are slightly higher than for inter-observer agreement.

21 sets of plain radiographs according to a 17 group Duparc classification and reported mean kappa values between 0.15 and 0.24.

The Duparc classification

Other classification systems

Duparc proposed a classification distinguishing articular and extra-articular fractures and defining five basic fracture patterns: tuberosity fractures, infratuberosity fractures (surgical neck), head fractures, head and tuberosity fractures, and head fractures accompanied by metaphyseal fractures. These patterns are further subdivided. Only a few studies have reported observer agreement within the Duparc classification.49, 54 In Nerot et al.54 seven observers classified 100 set of images (some with CT scans) according to a 10 group Duparc classification. They reported an overall mean kappa-value of 0.5. In Schwartz and Cuny 49 11 observers classified 52, 53

456

Based on museum and clinical specimens Edelson et al. have developed a three-dimensional concept for complex fractures of the proximal humerus.55-57 When four different CT views are assessed together five types of fractures can be defined: two-part, three-part, “shield” fractures and their variants, isolated greater tuberosity fractures, and fracture dislocations (Figure 11). In an observer study 55 10 doctors with different clinical experience classified twenty fractures according to four main categories. Weighted mean kappa value was 0.69 (95 percent CI 0.66 to 0.72). The authors concluded that kappa values for inter-observer

MINERVA ORTOPEDICA E TRAUMATOLOGICA

October 2009

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

on plain radiographs ranged from 0.15 (number of fragments) to 0.89 (varus/valgus without displacement). Mean kappa values for inter-observer agreement with additional CT scans ranged from 0.26 (large/small cephalotuberosity relationship) to 0.79 (extra-articular/articular fracture fragments). Only three parameters showed a significantly improved agreement by adding CT scans (number of fragments, lesser tuberosity displacement, and extra-articular/articular fracture). No overall kappa for the reading system was reported. Nevertheless, the authors concluded that their reading protocol should be used clinically.

M C IN O E P R Y V R A IG M H E T® D IC A

agreement between untrained observers are much higher in a 3D classification than in conventional 2D classifications. Schrader et al.32 studied possible reasons for disagreement within the Neer classification. Three observers identified fracture lines, displacement, and recommended treatment in 113 sets of radiographs (three view trauma series); 32 had additional CT scans. The images were reclassified by consensus after two months and disagreements were discussed. Ten learning points were developed. Reclassification took place after another two months. A small, but significant, improvement on classification according to Neer was identified (mean kappa values 0.42 and 0.47). No improvement by adding supplemental radiographic views or CT scans was found. An improvement in agreement on treatment recommendation was found (mean kappa values from 0.45 to 0.55). It was concluded that the problem is to understand complex fractures, not the classification system itself. Bahrs et al.58 suggested a detailed classification concept condensing Neer and AO with special regard to predefined topographic and morphological criteria. They tested observer agreement within the detailed classification (individual topographic fracture type and morphology) compared to a simple classification (fracture displacement and number of parts). Three observers classified plain radiographs of 22 fractures. Within the simple classification they reported range of kappa for inter-observer agreement on main fragments (0.41-0.60) and on fracture displacement (0.63-0.85). Range of kappa for inter-observer agreement within the detailed classification were 0.25-0.39 (main type and main fragments) and 0.31-0.45 (type of fracture and specification). No overall kappa value was reported. The authors concluded that a simple fracture classification may be more useful than an extensive classification. Mora Guix et al.59 proposed a protocol for structured reading of images of the proximal humerus; 22 parameters for plain radiographs and 11 for CT scans were applied by four observers classifying 30 fractures of which 22 had supplemental CT scans. Mean kappa values for inter-observer agreement based

BRORSON

Vol. 60, No. 5

Classification “anomalies”

Several examples of fracture patterns not captured by the Neer classification have been published. In Ackermann et al.60 three observers classified 108 sets of radiographs by consensus according to Neer and AO. They found that 15.5% of the fractures could not be classified according to Neer as compared to 24.1% according to the AO classification. The valgus-impacted four-part fracture pattern 33 was included in the Neer classification in the 2002 modification1 as a “precursor” of the laterally displaced four-part fracture. In a series of 22 fractures classified intraoperatively Tamai et al.61 found eight fractures not corresponding to any category in the Neer system. They identified fracture patterns where the lesser tuberosity was attached to the upper end of the shaft segment, not to the humeral head, and patterns with both tuberosities displaced and separated from the shaft and the head but not separated from each other forming a butterfly-shaped segment. Anatomically these fractures are threepart fractures, but according to the circulatory status of the head they should be considered four-part fractures.62 Several authors have reported 63, 64 cases with displacement of greater and lesser tuberosities, but with the head and shaft still in continuity. The three latter fracture patterns were all described by Codman (Figure 5) but have not been contained in later classifications.

MINERVA ORTOPEDICA E TRAUMATOLOGICA

457

BRORSON

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

Discussion

M C IN O E P R Y V R A IG M H E T® D IC A

Low agreement between observers classifying proximal humeral fractures according to the Neer classification was consistently reported in 14 observer studies included in this review. No improvement in mean kappa values for inter-observer agreement was found by: a) reducing the number of categories, even to a binary classification, b) including only experienced observers, c) selection of high quality radiographs, or d) by adding CT or 3D CT scans to radiographs. One study 31 reported a significant improvement in interobserver agreement by training of doctors. An ideal classification for fractures of the proximal humerus should be an all-inclusive tool, the categories should be mutually exclusive, it should be logical and clinically useful, and demonstrate observer agreement on an acceptable level. Categorization should be possible from preoperative radiographs and the stepwise progression should be one from a lesser to a greater severity. The classification should also serve as a mean for reporting in the scientific literature, assist documentation in clinical databases, and enable pooling of results across studies. In the absence of a gold standard for fracture classification observer studies may serve as an indirect evaluation of validity and accuracy of a classification system. On the one hand, high levels of observer agreement do not necessarily mean high accuracy as observers can agree on a wrong diagnosis. One the other hand, low levels of observer agreement preclude a high accuracy. If observers cannot agree with each other and themselves, they cannot all be right. However, the value of the Neer classification cannot be assessed by measures of agreement exclusively. Some basic aspects of image interpretation and clinical decision making should be considered.

appear to be easier for observers to agree on than others. Therefore, observer agreement will depend not only on the number of groups in the classification, but also on the type of assessments to differentiate between the groups. Two phases in the classification process can be distinguished: 1) an initial interpretation of radiographs to produce a mental image of a fracture in three dimensions; and 2) a matching of the three dimensional image to the system of classification. It is not clear a which stage the agreement is lost.65 Dirschl and Adams 66 found interobserver agreement on the adequacy of radiographs to be poorer than agreement on the classification itself in a reliability study of classification of fractures of the tibial plafond. By eliminating from the classification process the task of identifying the articular fragments, by having the fragments premarked, inter-observer agreement improved significantly.

Perception and image interpretation Agreement of fracture classifications may be limited by raters” ability to agree on basic radiographic assessments. In evaluating a fracture radiograph, certain basic assessments

458

Agreement on treatment plan and future studies

Fracture classifications are used clinically to guide treatment, estimate prognosis, and predict the risk of complications. It should also provide a reliable language of communication between clinicians making decisions. This property has not been studied empirically. Chan et al.67 found that additional CT-scans did not increase agreement on the classification of tibial plateau fractures. However, it did increase the agreement on treatment plan. In classifying femoral neck fractures according to the Garden system Oakes et al.68 reported that changes in classification grouping had only little impact on proposed treatment plan. Patient age was the critical factor for changes of treatment plan and played as much of a role than the fracture pattern in the choice of treatment. This aspect has not been studied in proximal humeral fractures. It could be addressed in a prospective observer study. It would also be valuable to clarify to what extent preoperative planning in proximal humeral fractures is based on the classification. The Neer classification is hampered by the

MINERVA ORTOPEDICA E TRAUMATOLOGICA

October 2009

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

4. Drapanas T, McDonald J, Hale HW, Jr. A rational approach to classification and treatment of fractures of the surgical neck of the humerus. Am J Surg 1960;99:617-24. 5. Knight RA, Mayne JA. Comminuted fractures and fracture-dislocations involving the articular surface of the humeral head. J Bone Joint Surg Am 1957;39-A:134355. 6. Tanton J. Fractures, 1. Fractures en général, fractures des membres, membre supérieur. Paris; 1915. 7. Neer CS. Displaced proximal humeral fractures. I. Classification and evaluation. J Bone Joint Surg Am 1970;52-A:1077-89. 8. Dehne E. Fractures at the upper end of the humerus. Surgical Clinics of North America 1945;25:28-47. 9. Watson-Jones R. Fractures and joint injuries. 3rded. Edinburgh: E.&S. Livingstone Ltd.; 1944. 10. Razemon JP, Baux S. [Fractures and fracture-dislocations of the upper extremity of the humerus]. Rev Chir Orthop Reparatrice Appar Mot 1969;55:387-496. 11. Codman EA. The Shoulder: rupture of the supraspinatus tendon and other lesions in or about the subacromial bursa. Boston: Thomas Todd; 1934. 12. Mallon WJ. E. Amory Codman, surgeon of the 1990s. J Shoulder Elbow Surg 1998;7:529-36. 13. Codman EA. The Roentgen Ray and its Relation to Fractures. In: Scudder CL, editor. The Treatment of Fractures. Philadelphia: W.B. Saunders; 1900. 14. Codman EA. Registry of bone sarcoma. Surg Gyn Obst 1928;42:381-93. 15. Codman EA. A Study in hospital efficiency, as demonstrated by the case report of the first five years of a private hospital. Boston; 1916. 16. Codman EA. Some points on the diagnosis and treatment of certain neglected minor surgical lesions. Boston Med Surg J 1904;150:371-4. 17. Codman EA. Bursitis subacromialis, or periarthritis of the shoulder joint. Boston Med Surg J 1908;159:533-7. 18. Neer CS, Brown TH, Jr., McLaughlin HL. Fracture of the neck of the humerus with dislocation of the head fragment. Am J Surg 1953;85:252-8. 19. Neer CS. Articular replacement for the humeral head. J Bone Joint Surg Am. 1955;37-A:215-28. 20. Neer CS. Displaced proximal humeral fractures. II. Treatment of three-part and four-part displacement. J Bone Joint Surg Am 1970-A;52:1090-103. 21. Trupka A, Wiedemann E, Ruchholtz S, Brunner U, Habermeyer P, Schweiberer L. [Dislocated multiple fragment fractures of the head of the humerus. Does dislocation of the humeral head fragment signify a worse prognosis?]. Unfallchirurg 1997;100:105-10. 22. Neer CS. Four-segment classification of displaced proximal humeral fractures. Instructional course lectures 1975;24:160-8. 23. Sidor ML, Zuckerman JD, Lyon T, Koval K, Cuomo F, Schoenberg N. The Neer classification system for proximal humeral fractures. An assessment of interobserver reliability and intraobserver reproducibility. J Bone Joint Surg Am 1993;75:1745-50. 24. Siebenrock KA, Gerber C. The reproducibility of classification of fractures of the proximal end of the humerus. J Bone Joint Surg Am 1993;75-A:1751-5. 25. Burstein AH. Fracture classification systems: do they work and are they useful? J Bone Joint Surg Am 1993;75-A:1743-4. 26. Neer CS. Correspondence. J Bone Joint Surg Am 1994;76-A:789. 27. Cowell HR. Patient care and scientific freedom. J Bone Joint Surg Am 1994;76-A:640-1. 28. Burstein AH. Correspondence. J Bone Joint Surg Am 1994;76-A:789-90.

M C IN O E P R Y V R A IG M H E T® D IC A

limitations of two-dimensional classification systems. Armamentaries as CT and 3D CT scans have contributed little to increase diagnostic accuracy. These modalities may, however, be helpful to elucidate fracture morphology in preoperative planning. The fracture patterns seen on scans may also help to identify fractures at risk of non-union or avascular necrosis.

BRORSON

Conclusions

Low observer agreement within the Neer classification represents a persistent problem for reporting outcome from clinical studies. The classification of complex fractures of the proximal humerus remains a challenge. Riassunto

La classificazione di Neer delle fratture dell’omero prossimale: una review narrativa

Charles Sumner Neer, II (nato nel 1917) pubblicò nel 1970 il suo sistema di classificazione per le fratture dell’omero prossimale. Gli autori presentano la classificazione di Neer e le sue modifiche in un contesto storico, analizzano la letteratura sulla concordanza tra osservatori, e riportano i commenti dei maggiori esperti di chirurgia della spalla. Uno degli aspetti di questo sistema di classificazione maggiormente discussi è la concordanza inter- e intra-osservatore riportata da più di una dozzina di studi. Un’applicazione sistematico sembra tuttavia elevare la concordanza a un livello accettabile, soprattutto tra gli specialisti. Gli autori discutono sia gli attuali tentativi di migliorare la classificazione di Neer che le caratteristiche che dovrebbe avere una classificazione delle fratture dell’omero prossimale per poter essere soddisfacente. La classificazione delle fratture scomposte dell’omero prossimale rimane una sfida per la pratica e per la ricerca ortopedica. Parole chiave: Variazione dell’osservatore - Fratture della spalla - Classificazione.

References

1. Neer CS. Four-segment classification of proximal humeral fractures: purpose and reliable use. J Shoulder Elbow Surg 2002;11:389-400. 2. Kocher T. Beiträge zur Kenntniss einiger Praktisch wichtiger Fracturformen. Basel: Carl Sallmann; 1896. 3. Röntgen W. On a new kind of rays. Nature 1896:2746.

Vol. 60, No. 5

MINERVA ORTOPEDICA E TRAUMATOLOGICA

459

BRORSON

THE NEER CLASSIFICATION FOR FRACTURES OF THE PROXIMAL HUMERUS

49. 50.

cation; only moderate agreement on displacement, even between specialists. Int Orthop 2002;26:271-3. Schwartz C, Cuny C. Fractures of the proximal humerus: A prospective review of 188 cases. Eur J Orthop Surg Traumat 2003;13:1-12. Müller ME, Nazarian S, Koch P, Schatzker J. The comprehensive classification of fractures of long bones. Berlin: Springer; 1990. Müller ME, Allgöwer M, Schneider R, Willenegger H. Manual of internal fixation. 3rd edition. Berlin: Springer; 1991. Duparc J, Largier A. [Fracture-dislocations of the upper end of the humerus]. Rev Chir Orthop Reparatrice Appar Mot 1976;62:91-110. Duparc J. [Classification of articular fractures of the upper extremity of the humerus]. Acta Orthop Belg 1995;61(Suppl 1):65-70. Nerot C, Thoreaux D, Hannouche D. Classification et étude inter-observateurs. Rev Chir Orthop Reparatrice Appar Mot 1998;84(Suppl 1):129-35. Edelson G, Kelly I, Vigder F, Reis ND. A three-dimensional classification for fractures of the proximal humerus. J Bone Joint Surg Br 2004;86-B:413-25. Edelson G, Safuri H, Salami J, Vigder F, Militianu D. Natural history of complex fractures of the proximal humerus using a three-dimensional classification system. J Shoulder Elbow Surg 2008;17:399-409. Edelson G, Saffuri H, Obid E, Vigder F. The threedimensional anatomy of proximal humeral fractures. J Shoulder Elbow Surg 2009;18:535-44. Bahrs C, Schmal H, Lingenfelter E, Rolauffs B, Weise K, Dietz K et al. Inter- and intraobserver reliability of the MTM-classification for proximal humeral fractures: a prospective study. BMC Musculoskelet Disord 2008;9:21. Mora Guix JM, Gonzalez AS, Brugalla JV, Carril EC, Banos FG. Proposed protocol for reading images of humeral head fractures. Clin Orthop Relat Res 2006;448:225-33. Ackermann C, Lam Q, Linder P, Kull C, Regazzoni. [Problems in classification of fractures of the proximal humerus]. Z Unfallchir Versicherungsmed Berufskr 1986;79:209-15. Tamai K, Hamada J, Ohno W, Saotome K. Surgical anatomy of multipart fractures of the proximal humerus. J Shoulder Elbow Surg 2002;11:421-7. Tamai K, Ishige N, Kuroda S, Ohno W, Itoh H, Hashiguchi H et al. Four-segment classification of proximal humeral fractures revisited: A multicenter study on 509 cases. J Shoulder Elbow Surg 2009 Mar 16. [Epub ahead of print]. Takase K, Shinmura K, Yamamoto K. A combined fracture of the greater and lesser tuberosity with head shaft continuity in the proximal humerus. Arch Orthop Trauma Surg 2007;127:895-8. Wainwright AM, Williams JR, Carr AJ. Interobserver and intraobserver variation in classification systems for fractures of the distal humerus. J Bone Joint Surg Br 2000;82:636-42. Dirschl DR, Adams GL. A critical assessment of factors influencing reliability in the classification of fractures, using fractures of the tibial plafond as a model. J Orthop Trauma 1997;11:471-6. Chan PS, Klimkiewicz JJ, Luchetti WT, Esterhai JL, Kneeland JB, Dalinka MK et al. Impact of CT scan on treatment plan and fracture classification of tibial plateau fractures. J Orthop Trauma 1997;11:484-9. Oakes DA, Jackson KR, Davies MR, Ehrhart KM, Zohman GL, Koval KJ et al. The impact of the garden classification on proposed operative treatment. Clin Orthop Relat Res 2003;409:232-40.

M C IN O E P R Y V R A IG M H E T® D IC A

29. Rockwood CA. Correspondence. J Bone Joint Surg Am 1994;76-A:790. 30. Bigliani LU, Flatow EL, Pollock RG. Correspondence. J Bone Joint Surg Am 1994;76-A:790-1. 31. Brorson S, Bagger J, Sylvest A, Hrobjartsson A. Improved interobserver variation after training of doctors in the Neer system. A randomised trial. J Bone Joint Surg Br 2002;84-B:950-4. 32. Shrader MW, Sanchez-Sotelo J, Sperling JW, Rowland CM, Cofield RH. Understanding proximal humerus fractures: image analysis, classification, and treatment. J Shoulder Elbow Surg 2005;14:497-505. 33. Jakob RP, Miniaci A, Anson PS, Jaberg H, Osterwalder A, Ganz R. Four-part valgus impacted fractures of the proximal humerus. J Bone Joint Surg Br 1991;73-B:2958. 34. Resch H, Povacz P, Frohlich R, Wambacher M. Percutaneous fixation of three- and four-part fractures of the proximal humerus. J Bone Joint Surg Br 1997;79B:295-300. 35. Audige L, Bhandari M, Kellam J. How reliable are reliability studies of fracture classifications? A systematic review of their methodologies. Acta Orthop Scand 2004;75:184-94. 36. Brorson S, Hrobjartsson A. Training improves agreement among doctors using the Neer system for proximal humeral fractures in a systematic review. J Clin Epidemiol 2008;61:7-16. 37. Brorson S, Bagger J, Sylvest A, Hrobjartsson A. Diagnosing displaced four-part fractures of the proximal humerus: a review of observer studies. Int Orthop 2009;33:323-7. 38. Mahadeva D, Mackay.D.C., Turner SM, Drew S, Costa ML. Reliability of the Neer classification system in proximal humeral fractures: a systematic review of the literature. Eur J Orthop Surg Traumatol 2008;18:415-24. 39. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:15974. 40. Gjorup T. The kappa coefficient and the prevalence of a diagnosis. Methods Inf Med 1988;27:184-6. 41. Kristiansen B, Andersen UL, Olsen CA, Varmarken JE. The Neer classification of fractures of the proximal humerus. An assessment of interobserver variation. Skeletal Radiol 1988;17:420-2. 42. Sidor ML, Zuckerman JD, Lyon T, Koval K, Schoenberg N. Classification of proximal humerus fractures: the contribution of the scapular and axillary radiographs. J Shoulder Elbow Surg 1994;3:24-7. 43. Brien H, Noftall F, MacMaster S, Cummings T, Landells C, Rockwood P. Neer's classification system: a critical appraisal. J Trauma 1995;38:257-60. 44. Bernstein J, Adler LM, Blank JE, Dalsey RM, Williams GR, Iannotti JP. Evaluation of the Neer system of classification of proximal humeral fractures with computerized tomographic scans and plain radiographs. J Bone Joint Surg Am 1996;78:1371-5. 45. Sallay PI, Pedowitz RA, Mallon WJ, Vandemark RM, Dalton JD, Speer KP. Reliability and reproducibility of radiographic interpretation of proximal humeral fracture pathoanatomy. J Shoulder Elbow Surg 1997;6:60-9. 46. Sjoden GO, Movin T, Guntner P, Aspelin P, Ahrengart L, Ersmark H et al. Poor reproducibility of classification of proximal humeral fractures. Additional CT of minor value. Acta Orthop Scand 1997;68:239-42. 47. Sjoden GO, Movin T, Aspelin P, Guntner P, Shalabi A. 3D-radiographic analysis does not improve the Neer and AO classifications of proximal humeral fractures. Acta Orthop Scand 1999;70:325-8. 48. Brorson S, Bagger J, Sylvest A, Hrobjartsson A. Low agreement among 24 doctors using the Neer-classifi-

460

51.

52. 53.

54. 55. 56.

57. 58.

59.

60.

61. 62.

64.

65.

66.

67.

68.

MINERVA ORTOPEDICA E TRAUMATOLOGICA

October 2009