An alphanumeric classification of osteoporotic pelvic ring injuries

The proposed alphanumeric classification (ANC) of OPRI is summarised in Table 1. It is based on injury localisation and morphology. It applies a scheme similar to the widely used alphanumeric AO/OTA fracture classification system and should therefore be familiar to and immediately applicable for orthopaedic trauma surgeons. The classification system differentiates between injury types (A, P and AP), groups (1–3) and subgroups (1–3). The degree of instability increases from group A to group AP and from (sub-)group 1 to (sub-)group 3.

After Institutional Review Board approval, 100 consecutive CT scans of patients with OPRI were retrospectively reviewed. Inclusion criteria were (1) OPRI after low-energy trauma or without history of trauma and (2) patient age of at least 70 years. Exclusion criteria comprised (1) OPRI after high-energy trauma, (2) age < 70 years, (3) osteolytic pelvic lesions and (4) metal implants in the pelvic ring.

Three orthopaedic traumatologists with varying levels of clinical experience (a second-year resident, a first-year consultant and an experienced senior pelvic surgeon) and a musculoskeletal radiology consultant independently reviewed the CT scans. The raters were instructed to analyse the injury pattern in each case and to classify the injury pattern according to both the proposed alphanumeric classification system (ANC, Table 1) and the FFP classification system [14] at the same time to allow for direct comparison of reliability values. The axial pelvic CT dataset as well as standard coronal and sagittal two-dimensional reconstructions were available to the raters in a PACS system. Furthermore, a multiplanar reconstruction tool (IMPAX EE, Agfa Healthcare) allowed for two-dimensional reconstruction of the pelvic ring in any arbitrary plane. In a second session 2 months after the first session, the case order was scrambled and each of the four independent raters repeated the classification process to assess intraobserver reliability. For the description of the patients’ collective, such as the relative distribution of fracture patterns, the classification by the most experienced senior pelvic surgeon was used as a reference.

IBM SPSS v24.0 software (Chicago, IL, USA) was used for all analyses. Inter- and intraobserver reliability were assessed by calculating percentage agreement as well as linear-weighted kappa coefficients (κ). For the assessment of interobserver reliability, data from both sessions were used. Owing to the fact that only ANC type P and AP (but not type A) injuries are further divided into groups and subgroups, group and subgroup ratings were missing in those cases classified as ANC type A injuries. As recommended by De Raadt et al. [15], the missing ratings were dealt with by applying the pairwise deletion method (i.e. complete-case analysis) for calculation of ANC group and subgroup agreement. In addition to the weighted kappa statistics (interobserver agreement between pairs of raters), the Fleiss’ kappa statistic was computed to assess interobserver agreement among all four raters. Interpretation of the kappa values was performed according to the guidelines of Landis and Koch [16], with kappa values of 0.01–0.20 indicating slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and > 0.80 (almost) perfect agreement.

The comprehensiveness of a fracture classification implies that all possible fracture patterns are covered by the classification system. Although the FFP classification covers the vast majority of OPRI in 11 subcategories, there are OPRI that are not definitely classifiable according to the FFP classification. This applies, for example, for bilateral incomplete sacral fractures without anterior pelvic ring involvement, because FFP type IIa injuries include isolated unilateral nondisplaced sacral fractures only. Similarly, nondisplaced iliac fractures without anterior pelvic ring involvement are not covered by the FFP classification. Another issue is that the FFP classification neither differentiates between uni- and bilateral posterior pelvic ring injuries in type IIb and IIc OPRI, nor does it specifically consider transverse sacral fracture components at all.

In comparison, comprehensiveness is one of the strengths of the proposed ANC. It covers all OPRI patterns and combinations in a well-known hierarchic system of fracture types, groups and subgroups, including those not classifiable according to the FFP classification. For example, bilateral incomplete sacral fractures without involvement of the anterior pelvic ring are classified as P2.11 or as P3.11 (depending on the presence of a transverse sacral fracture component). Uni- and bilateral injuries of the posterior pelvic ring in all their possible combinations are clearly distinguished depending on injury location, completeness of the sacral fracture and presence of a transverse sacral fracture component crossing the sagittal midline.

Developing a classification system that is impressively simple, but also comprehensive, and has an acceptable number of (sub)categories, but does not oversimplify may be like squaring the circle. Accordingly, a classification system invariably represents a compromise between simplicity and selectivity. The FFP classification consists of four main fracture types with a total of 11 subtypes. The definition and distinction of the four main fracture types of the FFP classification (I: anterior only; II: nondisplaced posterior lesions; III: displaced unilateral posterior lesions; IV: displaced bilateral posterior lesions) is clear and simple. This is also true for the distinction of subtypes in type I, III and IV injuries. It is, however, less intuitive for type II injuries, which represent the majority of OPRI [8, 14]. Type IIa injuries include unilateral nondisplaced sacral fractures without anterior involvement, but they explicitly exclude iliac fractures. Type IIb and IIc injuries share an additional injury of the anterior pelvic ring. Type IIb injuries on the one hand cover sacral fractures only, while type IIc injuries on the other hand additionally include nondisplaced iliac fractures and iliosacral disruptions. This appears to be quite confusing at first glance and may decrease simplicity of the FFP classification scheme.

In comparison, the ANC of OPRI consists of 31 categories and thus provides substantially higher selectivity than the FFP classification, which includes 11 categories. Nevertheless, simplicity of the ANC is ensured by the strictly hierarchic system of types, groups and subgroups. Therefore, mastering the simple systematics of the ANC already results in knowledge about all 31 classification categories. Additionally, the ANC applies a scheme similar to the widely used AO/OTA fracture classification systems and should therefore be familiar to and immediately applicable for traumatologists.

A recent study evaluated the reliability of the FFP classification and reported overall moderate to substantial interobserver and substantial to almost perfect intraobserver agreement [14]. For the assessment of complete nondisplaced vs. displaced sacral fractures (FFP IIc, IIIc and IVb injuries), however, inter- and intraobserver reliability was markedly lower (unweighted κ ranging from 0.10 to 0.52 and from 0.28 to 0.66, respectively; linear-weighted κ ranging from 0.17 to 0.64 and from 0.35 to 0.76, respectively) [14]. The latter may be partly due to the relatively vague definition of “displacement” by Rommens and Hofmann (i.e. “deformation of the anatomical landmarks” and “widening and gap formation between fracture fragments”). Another recent study involving 4 raters and 97 OPRI cases found slight to moderate (unweighted κ ranging from 0.18 to 0.47) interobserver and fair to moderate (unweighted κ ranging from 0.29 to 0.55) intraobserver reliability of the FFP classification [17]. The most recent study by Pieroh et al. [18] applied Fleiss’ kappa statistic and involved 13 raters of varying level of expertise, 3 classification cycles by each rater and 60 OPRI cases. The authors reported moderate intraobserver reliability [mean Fleiss’ kappa coefficient: 0.46 (95% confidence interval: 0.40 to 0.50)] and moderate interobserver reliability [0.53 (0.48–0.58)] for the complete FFP classification. For the FFP main group classification, intraobserver reliability was moderate [0.60 (0.53–0.65)] and interobserver reliability was substantial [0.61 (0.54–0.66)].

In comparison, the ANC showed comparable overall intra- and interobserver reliability although it distinguishes almost three times the number of classification categories. Furthermore, the intra- and interobserver reliability for grading OPRI by ANC types, groups and subgroups was substantial to almost perfect (Tables 3 and 4). Nevertheless, one might argue that the ANC does not distinguish between nondisplaced and displaced sacral fractures. However, the vague definition of “displacement” in the FFP classification resulted in poor reliability and therefore reduced its overall usability. In our opinion, even a more precise criterion for displacement, such as a threshold of 2 or 5 mm, is hardly measurable with adequate reliability in OPRI and would be an arbitrary threshold as well. Second, OPRI result from low-energy trauma or without history of trauma and are therefore typically non- or minimally displaced. Third, the vast majority of fracture classification systems do not specifically include displacement as a parameter, which does not compromise their usability.

The FFP classification was developed based on the degree of instability. We agree that it is reasonable to assume that OPRI without involvement of the posterior pelvic ring (type I) are more stable than OPRI with posterior pelvic ring involvement (type II–IV), that incomplete sacral ala fractures (type IIb) are more stable than complete fractures (type IIc) and that unilateral posterior pelvic ring injuries (type III) are more stable than bilateral injuries (type IV). However, regarding its relation to injury severity, the main drawback of the FFP classification is that it subsumes OPRI with different degrees of instability into the same FFP type. For example, type IIb and IIc injuries include fractures with uni- or bilateral involvement of the anterior and posterior pelvic ring. Accordingly, these injuries consist of a minimum of two and a maximum of four fractures of the pelvic ring. Moreover, type III injuries are defined as unilateral displaced posterior lesions, but if these injuries are combined with a contralateral nondisplaced posterior lesion, they are still classified as type III. Additionally, type IVc injuries are defined as “combination of different instabilities in the dorsal pelvis” [8]. Finally, transverse sacral fracture components indicating a multiplanar sacral fracture pattern are not considered as a criterion of instability at all.

In comparison, the ANC is based on the degree of instability as well. It clearly differentiates between incomplete and complete sacral fractures, between unilateral and bilateral involvement, and between sacral and extrasacral posterior pelvic ring injuries in all possible combinations. This is enabled by the double-digit subgroup labelling described above. In addition, multiplanar sacral fractures (i.e. H-, U- or Y-type fractures; group 3) with spinopelvic instability are clearly distinguished from sacral fractures without a transverse fracture component (group 1 and 2).

A clinically relevant classification system and its categories not only should be related to injury severity but also should be predictive of relevant patient outcomes and aid treatment decision making. Yet, the predictive value of a classification system can be only suspected at the time of proposal, since its assessment requires a large observational study “with accurate recording of most clinically relevant outcomes and their respective known or suspected prognostic factors, including treatment options” (i.e. phase 3 of validation) as pointed out by Audigé et al. [19]. However, in contrast to Rommens et al. whose classification and treatment recommendations are based on injury morphology only [8, 9], we are convinced that a classification system for OPRI should go beyond morphology and include selected clinical and patient-specific factors relevant to prognosis and treatment decision making. The most important of the latters, in our experience, are those summarised in Table 6 and include restrictions in patient mobility as a consequence of sustaining an OPRI, pain and analgesic requirements due to OPRI, preinjury mobility, time trauma to diagnosis of OPRI, and presence of neurological deficits related to OPRI. Nevertheless, future research is needed to identify the most appropriate clinical modifiers to be incorporated into the ANC with the eventual goal of predicting treatment approaches and outcomes in addition to describing OPRI morphology.

This study has some limitations. First, evaluation of a higher number of cases would have increased the power of our statistical analysis. Second, CT scans but no additional MRI scans were used for reliability assessment. Third, the grading of the 100 OPRI cases by the experienced pelvic surgeon was used as a reference for the description of the patients collective. Fourth, for full validation of the ANC, a pragmatic multicentre agreement study and a prospective clinical observational study have to be conducted in future [19]. Last, the ANC was developed primarily for low-energy OPRI. For the classification of high-energy pelvic ring injuries, the Tile/AO or Young and Burgess classification might be more appropriate even in elderly individuals.

In conclusion, the proposed alphanumeric classification (ANC) of OPRI is a simple, hierarchically organised alphanumeric classification scheme traumatologists may rapidly become familiar with. The ANC provides several advantages over the currently used FFP classification of OPRI, particularly with regard to comprehensiveness, simplicity, and relation to injury severity. Despite the substantially higher number of categories of the ANC, intra- (substantial) and interobserver (moderate to substantial) reliability are comparable to those of the FFP classification. Future research, however, is needed to assess outcomes of different ANC patterns and to clarify indications and most appropriate strategies for operative treatment.