12.1 Fracture classification

The jargon of fracture description is learned early in orthopaedic training and helps to facilitate communication by identifying important parameter values in a qualitative way. The selection of adjectives such as ‘a displaced comminuted closed tibia shaft fracture’ identifies characteristics of an injury. Formal fracture classification is an extension of fracture terminology with specification of a number of groups with common and distinguishing factors (Box 12.1.1).

The benefits of classification include: 1) facilitates communication; 2) identifies prognosis and complication risks; 3) directs treatment; 4) enhances research.

The classification should allow for all theoretical and practical patterns. A good example is the Neer classification of proximal humerus features which defines four anatomical parts to be considered. Since these four parts are present in every proximal humerus the classification is all inclusive.

The classification should allow for only one category for each pattern seen. A good example is extra-articular vs partial articular vs total articular. A poor example is the Gustilo and Anderson (1976) classification of open fractures. Type 1 open wound which is low energy, less than 1 cm long, and generally caused by bone piercing from the inside out while a type 2 wound is considered 1–5cm with a moderate amount of soft tissue damage. A given injury which is 0.9cm long with moderate soft tissue injury has qualities of both type 1 and type 2.

Two different physicians should be able to identify the appropriate group or class to which each fracture should be assigned. Therefore it is necessary to select criteria which can themselves be reliably assessed and which physicians can learn, remember, and are willing and able to apply. Reliability can be measured by the kappa statistic which varies from −1 to 1 based upon agreement of two different reviewers: −1 indicates complete disagreement, 0 is random agreement, and 1 is complete agreement (Figure 12.1.2). It is a generally accepted consensus that values of 0–0.2 represent slight agreement, 0.21–0.4 fair agreement, 0.41–0.6 moderate agreement, and 0.61–0.8 substantial agreement (Figure 12.1.3). Other statistical measures of agreement reported in the literature which range between −1 and +1 include weighted kappa, Cronbach’s alpha, and SAV of O’Connell. Weighted kappa gives partial credit for being close in contrast to kappa which is most applicable to a system where observations either agree or not. Alpha is particularly applicable to rank order analysis.

The reliability of this system decreases with increasing complexity and specificity at the group (fracture geometry) or subgroup (comminution pattern) level. The absence of a more reliable classification system is one measure of adequate reliability.

The appropriate classification should be consistent when one physician reviews the same case at different times. Reproducibility can also be measured by kappa or a similar statistic as previously described. The best classification is highly reproducible.

The classification should assist the physician in the care of the patient. Identifying prognosis, associated injuries, complication risks, and directing optimal treatment are the most common forms of clinical usefulness. Associated injuries may not be recognized without clues to their existence. The presence of associated injuries may make a dramatic difference in the significance of a fracture pattern. Recognizing a distal fibula fracture with talar shift as a Lauge–Hansen SER-4 helps identify by deduction the ligamentous injuries to the deltoid, anterior and posterior inferior tibiofibular ligaments which are not directly visualized on the radiographs.

The classification system is better if it is based on a logical algorithm. It should be possible to systematically answer a series of related questions about the fracture to arrive at the specific classification. The number of categories should be the fewest possible to effectively discriminate similarities and differences between groups. Allowance should be given for future developments in the field which may require subsequent revision of the classification system in a logical manner.

A classification system should facilitate verbal and written communication between clinicians. It should use standard terminology that is clearly defined and accepted and the definition should not change over time. Facilitation of communication could be considered a form of clinical relevance but is of sufficient importance to warrant specific inclusion as a desirable quality.

The usefulness of a classification is usually time dependent. In the late 1970s intramedullary nailing of femoral shaft fractures was gaining clinical acceptance. The presence of comminution was associated with a significant incidence of complication of the technique, including poor rotational control and shortening. Winquist proposed a classification of femoral shaft fractures based upon increasing comminution and identified the percentage of circumference of the bone necessary to be in contact after nailing to reduce malunion to acceptable rates. At that time the classification was useful. Prior to intramedullary nailing there was no difference in outcome as a function of amount of circumference comminuted so the classification would not have been useful. Subsequent development of interlocking nails changed the usefulness of the classification from defining an indication for nailing to defining an indication for interlocking. Improvement in ease of interlocking made the technique useful without regard to amount of comminution so the Winquist classification of femoral shaft fractures was no longer of particular significance or usefulness. This illustrates that the Winquist classification of femur shaft fractures was clinically relevant during the time of difficult interlocking nails being prevalent.

The widespread tacit acceptance of most fracture classifications used clinically has been strongly challenged in recent publications. The reliability of applying various fracture classifications by different surgeons or reproducibility of a surgeon on two different occasions has been shown to be much lower than previously assumed. Current studies should and are focusing on the source of this variability. There is clearly a tendency for surgeons to adjust the classification based upon their own experience. In addition, clinicians have a limited and imperfect recall of definitions involved in classifications and this will result in some real error. Consensus classification has been proposed as a method to salvage classification schemes with poor reliability. Although this is appealing in theory, consensus classification has not yet been shown to be reproducible or reliable among different groups.

The radiographic interpretation of fracture lines is highly variable and dependent upon radiographic quality (contrast, completeness, obliquity). Reports of accuracy of radiographic measurements suggest more variability than commonly believed. For example, the imprecision in measuring Bohler’s angle after a calcaneus fracture or the percentage of lost height in a lumbar spine fracture or the millimetres of displacement of a proximal tibia fracture all limit the usefulness of classifications which require those measurements. Another common problem is that the classification requires information not available from radiographs. Malleolar distal tibia fractures with large posterior tibial fragments have been considered for internal fixation based on the percentage of articular surface involved in the fragment. This is a two-dimensional concept, however this fragment is very poorly visualized on the anteroposterior or mortise radiograph and only one dimension can be measured on the lateral (see Figure 12.1.1). As more diagnostic information (computed tomography, magnetic resonance imaging, etc.), about a particular injury becomes available, the initial classification may not stand the test of time.

Not all of the problems are with the classifier. Classifications may give general typical scenarios without precise category definitions such as the Gustilo and Anderson open fracture classification. Open fractures caused by tornadoes are defined as type 3 without regard to their size. That may have helped explain the risk of infection in the patients treated in Minnesota but is not a generally good fracture classification scheme to be utilized worldwide.

A common mechanism of injury has often been applied to patterns of injury and fracture classification, such as the nightstick fracture (isolated ulna shaft) from fending off an attack by someone wielding a nightstick (truncheon); or boot-top fracture (midshaft tibia) from ski boot binding design in the 1970s. This technique is highly variable as the true mechanism of injury for a given patient is seldom known with certainty and hence speculation is usually required.

Eponyms have also been used to designate a certain fracture pattern by the person recognized as first publishing and being referenced on the significance of that particular injury. Thus distal radius fractures are commonly referred to as Colles fractures. Eponyms suffer from lack of precision and rarely mean the same thing to the speaker, the listener, or the original author. Fortunately these shortcomings have become recognized and the frequency of use of eponyms is rapidly fading in orthopaedic traumatology. There are probably better ways to pay tribute to our musculoskeletal forefathers than use of eponyms.

Current classification schemes and outcome measures suffer statistically from the tendency to clump rather than spread data across parameter values. For example, most clinical scores range from 0–100 but the range actually reported tends to be much more narrow, for example 50–89. In the extreme, if all patients score 70 then there can be no correlation of any variable to the outcome. When pain is scored as: no pain = 50 points; occasional mild pain with strenuous activity = 40 points; some pain with some activity = 30 points; severe pain with activity = 20 points; severe pain at rest = 10 points, then there will be a strong tendency toward 30 points. The more the results are clumped the less likely there is to be a correlation between variables and outcomes. Therefore a good fracture classification may poorly predict clinical outcome due to problems with measuring clinical outcome rather than deficiency in the classification system. Some authors have suggested rank order analysis to avoid the clumping problem of other outcome measures. Rank order forces the spreading of parameter values across the entire range. However, a drawback to this form of analysis is the lack of general applicability to patients outside the study.

The 2007 modification and republication of the Orthopaedic Trauma Association (OTA) fracture and dislocation classification compendium provides the current standard for the classification of fractures. This classification has now achieved widespread recognition, understanding, and acceptance in publications and orthopaedic practice. The 2007 version has reconciled differences that existed between the previous OTA and AO classifications by cooperation between the two groups and now the two are identical. This classification has been established as the standard by OTA, AO, SICOT, the Journal of Orthopaedic Trauma, and many other organizations and publications and can now be recommended as the standard fracture classification for general use.

The 2007 OTA fracture classification provides a complete system of fracture classification for all bones in the body that uses consistent methodology throughout the skeletal system. It is based on Muller’s original principle of classification of long bone fractures with designation of a bone, then three bone segments/bone, three types/segment, three groups/type and three subgroups/group. This classification fulfils the criterion for a good classification system (Figure 12.1.2) and provides a wide spectrum of specificity. It can be used as a standard for terminology for very gross fracture types or extremely specific patterns for research purposes.

The OTA classification involves an alpha-numeric designation of the fracture which is particularly important with the current use of computerized search engines to look for information on a specific entity. The first number identifying the BONE (1–9), the second the bone SEGMENT (1–3), then a letter designating TYPE (A,B,C) followed by numbers (1–3) for GROUP and SUBGROUP separated by a decimal point. For example a comminuted fracture of the medial condyle of the distal femur would be coded 33.B2.3 (femur, distal, partial articular, medial, multifragmentary).

There are fundamental differences between various TYPEs of bone SEGMENT fractures. Proximal and distal segments of long bones are classified as extra-articular (A), partial articular (B), and total articular (C). Diaphyseal fractures are classified as simple (A), wedge (B), and complex (C). For the flat bones there are three types based on the same principle of increasing complexity. For example the types of talus (81) fracture are avulsion or process fractures (A), talar neck fracture (B), and talar body fracture (C).

The GROUP designation of fracture pattern typically involves anatomical specificity. For example, partial articular distal humerus (13B) fractures have three groups: lateral or capitellar (13B1), medial or trochlear (13B2), coronal plane (13B3). When anatomical specificity is not relevant, the groups are typically designated by degree and pattern of comminution. For example, total articular distal humerus fractures (13C) are grouped as articular simple, metaphysis simple (13C1), articular simple metaphysis multifragmentary (13C2), articular multifragmentary (13C3). This corresponds with the relative complexity of fracture, the treatment and the prognosis. There are a total of nine categories at the group level for each bone segment and this level of specificity would be more than adequate for the vast majority of clinical situations.

The SUBGROUP level of designation is useful when very detailed and specific categories are desirable as might be the case in research, database, or subspecialty discussions. It is more detail than would be necessary for most clinical situations. For example, distal humerus medial fractures (13B2) is sub grouped by the pattern of fracture as medial or Milch 1 (13B2.1), groove or Milch 2 (13B2.2), multifragmentary (13B2.3). There are three subgroups for each group (for example three subcategories of fractures of the trochlea of the distal humerus). There are a total of 27 categories at the subgroup level for each bone segment (all distal humerus fractures).

The OTA classification has been criticized as being overly complex. However, as illustrated earlier, the level of desired complexity can be chosen to match the needs of the user. If the information for the group or subgroup level is not known or needed, then designation at the level of type can be used. Furthermore, as illustrated by the example of trochlear fractures, the complexity at the level of subgroup is well within the skill set of the average orthopaedic surgeon with good quality plain radiographs, especially with easy accessibility to the standard reference.

The 2007 version of the OTA fracture classification is different from the 1996 version in several areas including hand, foot, flat bones dislocation, and the newly developed AO classification of fractures of bones with open growth plates (paediatric fractures).

A classification system may be imperfect; however it can still be useful. We should use the classifications which are the best available to help direct treatment and predict outcome. Directing treatment is one of the best forms and a classification which correctly partitions a family of related injuries into categories with unique treatments would be ideal. This correlation rarely survives more than a few years. We should continually strive to improve by testing and updating (perhaps once per decade) our classification systems to be logical, reliable, reproducible, and clinically relevant to direct treatment and identify risk.

The 2007 OTA fracture classification is likely to continue to gain acceptance as the standard. New information about this classification in particular will be use to shape future modifications and improvements in the decades ahead. New information about fracture patterns and the predictors of treatment and outcome will also be forthcoming and this information will also result in modification of future versions (2017) of the classification.

Recognitipon of problems with current fracture classifications and the ongoing desire to optimize patient outcome will result in improved classifications in the future. Old and new classifications will be scientifically scrutinized and the reliable, reproducible, logical, clinically relevant ones will be selected for utilization. Ongoing efforts will focus on identifying and minimizing the sources of error or disagreement so as to optimize reliability and reproducibility. Factors will be scientifically and statistically analyzed for predictive value and important factors will be incorporated into fracture classifications.

There will be ongoing efforts to learn from our past and review clinical results as a function of existing fracture classifications. Treatments and outcomes as a function of fracture classification will continue to be reported with the potential for improvement as the tools we use (e.g. prospective fracture classification) are themselves improved.

New methods of classification will be further investigated. Consensus classification by a panel of experienced surgeons may be of benefit. Rank order and other non-parametric statistical analysis may also help solve the problems of poor agreement. The methodology by which we assess classification reliability will be standardized and improved.

There is considerable benefit and interest in generalized fracture classification systems (OTA, AO, SICOT). The ease of computerized and internet accessibility to fracture classification standards and educational tutorial tools will improve the accuracy and generalized utilization of the 2007 OTA fracture classification (e.g. OKO). Increased utilization and widespread acceptance will also depend upon the effectiveness of the classification and its ability to lump similar and distinguish dissimilar injuries in a comprehensible and usable manner.