Lateral compression fracture of the pelvis represents a heterogeneous group of complex 3D patterns of displacement

Injury, Int. J. Care Injured (2008) 39, 893—902

www.elsevier.com/locate/injury

Lateral compression fracture of the pelvis represents a heterogeneous group of complex 3D patterns of displacement A. Khoury a,b, H. Kreder a,c, T. Skrinskas a, M. Hardisty a,d, M. Tile a,c, C.M. Whyne a,c,d,* a

Orthopaedic Biomechanics Laboratory, Division of Orthopaedic Surgery, Holland Musculoskeletal Program, Sunnybrook Health Sciences Centre, Toronto, ON, Canada b Orthopedic Surgery, Hadassah Medical Centre, Hadassah, Israel c Department of Surgery, University of Toronto, Canada d Institute for Biomaterials and Biomedical Engineering, University of Toronto, Canada Accepted 24 September 2007

KEYWORDS Pelvis; Lateral compression fracture; 3D image analysis; Displacement pattern

Summary Although clinical and radiological criteria exist to direct the non-operative and operative treatment of other types of pelvic injuries, none exist for lateral compression (LC) fractures. The purpose of this study is to describe the patterns of injury in LC fractures through quantitative 3D radiographic analysis. It is hypothesised that LC fractures represent a spectrum of injuries with a combination of translational and rotational displacements. CT data from 60 patients with unilateral lateral compression fractures were obtained. Quantification of translations and rotations of the fractures was performed using 3D visualisation software. Fractures initially diagnosed as LC actually represent a spectrum of displacement patterns, ranging from a minimally displaced hemipelvis to complex combinations of displacements. Fractures were grouped based on pattern of rotation and translation into 5 distinct groups. 3D analysis of displacement patterns demonstrated a complexity in LC fractures which may explain the variations seen in outcomes associated with this injury. # 2007 Elsevier Ltd. All rights reserved.

Introduction * Corresponding author at: Orthopaedic Biomechanics Laboratory, Sunnybrook Health Sciences Centre, UB 19, 2075 Bayview Avenue, Toronto, ON, Canada M4N 3M5. Tel.: +1 416 480 5056; fax: +1 416 480 5856. E-mail address: [email protected] (C.M. Whyne).

Historically, pelvic fractures have been classified based on the correlation between pelvic stability and the mechanism of injury, as proposed by Pennal and modified by Tile.18,20 Young and Burgess23

0020–1383/$ — see front matter # 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.injury.2007.09.017

894 refined this concept stating that pelvic stability can be judged by fracture pattern, direction of the force of injury, and by knowledge of pelvic ligamentous anatomy. Their classification includes anterior pelvic compression, lateral compression, vertical shear and combinations thereof, subclassifying the LC fractures into types I, II and III (windswept). The OA/OTA classification of pelvic fractures is based on Tile’s classification. Bucholz described an anatomical classification based on the degree of posterior injury to the pelvic ring and an assessment of stability, but the system contained only 3 groupings. In contrast, Letournel classified pelvic ring injuries simply according to site of injury (anterior or posterior). Classification systems have been developed to provide guidance to orthopaedic surgeons in directing appropriate fracture management,4,17,18,20,21 however better understanding of patterns of pelvic injury is needed in order to optimise treatment strategies.10 Lateral compression (LC) pelvic injuries constitute a diverse group of fracture-dislocations that occur after application of laterally directed forces to the pelvis.2,4,5 LC fractures account for more than 50% of all pelvic injuries in most series20 and are most commonly caused by side impacts.9,13,20 This lateral impact mechanism results in fractures which are normally classified as rotationally unstable but vertically stable (OTA Type B2) injuries.17 These injuries usually include a compression fracture of the posterior pelvic ring with impaction of either the sacrum or the sacroiliac (SI) joint, and an ipsilateral or contralateral anterior pelvic ring disruption.15,16,20 Presently, while clear clinical and radiological criteria exist to direct the non-operative and operative treatment of other types of pelvic injuries such as open book injury and specific subtypes of LC fractures (crescent, posterior locked and anterior locked pelvis), no such criteria exist for the majority of LC fractures.18,20 Anecdotally, some surgeons believe the majority of patients with LC fractures do well with non-operative treatment.16 Recent studies however, have shown a correlation between residual displacement, decreased function and increased pain in these injuries, indicating that a subgroup of patients with LC fractures may benefit from an improved outcome with operative treatment.1,14,15 More recently, some orthopaedic trauma surgeons have advocated fixation (internal and/or external) of some LC injuries to improve patient comfort and to allow for earlier mobilisation.19 According to the current standard of care, the decision to offer operative stabilisation to patients with LC fractures is based on a clinical impression of pelvic mobility, leg length discrepancy, pain and a qualitative assessment of the

A. Khoury et al. severity of radiographic displacement.1,11 Current stability is only defined clinically by physically evaluating the amount of motion of the affected hemipelvis under sedation or anaesthesia. Due to difficulties quantifying pelvic displacement and instability outside the operating theatre, the indications for non-operative and operative treatment of LC fractures remain unclear. In order to identify which patients with LC injuries may benefit from surgery, it is necessary to develop a better understanding of the spectrum of pelvic stability associated with LC fracture. The aim of this study was to describe the patterns of injury to the pelvis in LC type fractures through quantitative 3D radiographic analysis. A more detailed analysis of LC type displacement patterns will elucidate the indicators for operative treatment in this fracture group. We hypothesise that LC fractures represent a spectrum of injury with a complex combination of translational and rotational displacement patterns. This study is phase one of a larger study for the evaluation of lateral compression fractures of the pelvis, which will attempt to prospectively relate 3D displacement patterns with assessments of pelvic biomechanical stability and more comprehensive patient outcome data. This additional information is required to determine the ultimate clinical relevance of our findings. Identifying the complex patterns of displacements and rotation associated with LC fractures is the first step in this process.

Materials and methods Patient data and 3D analysis Sixty patients (from 2002 to 2004) were identified as having a unilateral lateral compression fracture of the pelvis (47 yrs  19, 27 males, 33 females). The diagnosis of LC fracture was established by two pelvic and acetabular surgeons using normal clinical methods including clinical and radiographic criteria. Based on plain film and CT scan, evidence of sacral impaction and/or whether anterior ramus fracture was present.23 The Tile classification was used where lateral compression fractures (type B2) are considered rotationally unstable and vertically stable. All patients had a relevant mechanism of lateral compression and presented with pain or horizontal instability on physical examination. We included only isolated, unilateral pelvic fractures without acetabular involvement and with a pretreatment CT scan. A 3D reconstruction of each pelvis was created from the pretreatment CT data using commercially

Lateral compression fracture of the pelvis available software (Amira, Mercury Computer Systems, Berlin, Germany). The bone was segmented from axial images using a region growing connected thresholding technique. The segmentation was then used to create a 3D triangulated surface of the pelvis for visualisation and analysis. To quantify the differences in position between right and left hemipelvises, the spatial orientation of each hemipelvis was calculated with respect to a mid sacral sagittal plane. The mid-sagittal plane and a plane representative of each hemipelvis were selected by identifying nine distinct anatomical landmarks on the reconstructed 3D pelvis (Fig. 1). The mid-sagittal plane was defined by: the spinous process of S1, the promontory of S1 and the tip of the coccyx. The anterior superior iliac spines (ASIS), posterior superior iliac spines (PSIS) and ischial spines (IS) were landmarked to define two planes representing the post fracture orientation of each hemipelvis. The defining landmarks of each plane were chosen because they represented inflections on the pelvic surface making them easily identifiable on the 3D pelvic model without any ambiguity. Furthermore, their relative positions were spaced far apart to ensure an accurate planar definition. A global plane was used to define a standard orientation and coordinate frame for all analyses (Fig. 1). The global origin was always defined posterior and cephalad to the pelvis and XYZ axes representing medial—lateral, inferior—superior, and anterior—posterior directions, respectively. The spatial orientations of the two planes were calculated with respect to the mid sacral sagittal global reference frame using the fixed axis theorem.3,7 The fixed axis convention relates one reference frame to

895 another through translations and rotations along and about fixed axes, as opposed to moving axes as in the Euler method. Translations and rotations about the defined axes were described using established anatomical terminology (Figs. 2 and 3). The spatial orientations of the hemipelvises and appropriate transformations for the data were calculated in Amira using a custom written Tool Command Language (TCL) script. The script simply automated the use of available Amira modules in order to expedite and simplify the surgeon’s analysis.

Error and sensitivity analysis To define the normal anatomical variance in the pelvis, we analysed the differences in orientation between right and left hemipelvises in 10 normal (unfractured) pelvises, (49 yrs  19, 6 males, 4 females). Anatomical differences in spatial orientation existing between hemipelvises in the 10 unfractured cases were calculated. Measurements were repeated three times on each specimen by a trauma surgeon in order to assess the intra observer repeatability of the method. To evaluate interobserver variance, 10% of the fractured pelvis scans (N = 6) were analysed by an additional orthopaedic surgeon following training on the software. The inter and intra observer variances were analysed using interclass correlation coefficients. Error analysis was used to quantify the effect of landmark selection errors on the final comparison of displacement and rotational descriptions of the pelvic planes. From inter and intra observer repeatability measurements, landmark selection was deemed accurate to within 5 mm. This error

Figure 1 (A) 3D reconstruction of the pelvis showing the global axes, several identified landmarks and the plane representing the right hemi-pelvis. (B) A posterior view showing the remaining landmarks. Landmark abbreviations: PSIS, posterior superior iliac spine; IS, ischial tuberosity; ASIS, anterior superior iliac spine; SP, spinous process of S1; P, promontorium; T, tip of coccyx.

896

A. Khoury et al.

Figure 2

This figure illustrates the isolated translations in all directions and the related nomenclature.

is a combination of the CT scanning parameters, user landmark identification error and surface triangulation resolution. User variability was assessed as well in this study and the selected 5 mm value represents a reasonable estimate of error (worst case scenario) from the combination of CT resolution, surface reconstruction density and user variability. To assess the effect of landmark inaccuracy, randomly generated errors between 5 mm and +5 mm (following a Gaussian distribution) were added to the raw data for the entire data set.6 Final translational and rotational errors of the hemipelvis comparisons were defined as the difference between the original data set and the error modified set.

Patient analysis Thresholds above which displacement could be classified as due to the LC fracture were determined from the normal anatomical differences seen in the intact pelvises and the error associated with the described method. In the 60 pelvises classified with

an LC fracture, translational and rotational differences between the intact and fractured sides were quantified and compared to determine patterns of displacement due to the injury.

Clinical data The causes of injury, presence of abdominal injury and treatment (operative/non-operative) were compared between groups. Functional outcome data from the Short Form 36 questionnaire (SF36) collected at 12 months post injury was also compared to Canadian normative data12 based on retrospectively collected information from a pelvic and acetabular database maintained at our institution. The SF36 is a global health questionnaire which assesses the impact of the diagnosis and treatment of the condition on the patient’s overall wellbeing.22 There are 36 questions scored from 0 to 100, with 100 representing the best possible health status. From these scores, two summary scales are calculated for the physical component (PCS–—physical functioning, role physical, bodily pain, general

Lateral compression fracture of the pelvis

Figure 3

897

This figure illustrates isolated rotation around each of the three axes and the related nomenclature.

health perceptions) and the mental component (MCS–—vitality, social functioning, mental health, role emotional).22

Table 1 Average of the inter/intraobserver class correlation coefficient for identification of pelvic landmarks (>0.6 is good, 0.4 < coefficient < 0.6 is moderate) Axis

Interobserver correlation

Intraobserver correlation

M/L I/S A/P

0.91 0.62 0.174

0.46 0.61 0.35

Results Error and sensitivity analysis As previously mentioned, from inter and intra observer repeatability measurements, landmark selection was deemed accurate to within 5 mm. Intra and inter observer agreement was generally high (Table 1). The simulated Gaussian distribution errors on the raw data yielded a mean net rotation error of 2.48 (range of 4.88) and a net translation error of 2.7 mm (range of 5.4 mm). The mean difference in rotation and translation between the 10 normal hemi-pelvises was 1.98 and 4 mm, respectively. Summing the range of analysis errors and normal anatomical variation values, thresholds were defined for differences between fractured and

Poor correlations in the A/P direction are improved with multi-axial viewing of the models during landmark selection.

unfractured sides of less than 10 mm and 78 were considered minimally displaced.

Patient analysis Translations and rotations for the 60 fractured pelvises were calculated between fractured and unfractured sides (Figs. 2 and 3). Of the 60, 30 pelvises exhibited the translational displacement

898

A. Khoury et al.

Fracture patterns Fractures were grouped based on pattern of rotation and translation into 5 groups: (1) minimally displaced fractures, 22 patients (36.6%); (2) pure translation, 8 patients (13.3%); (3) pure rotation, 8 patients (13.3%); (4) combined single axis rotation with translation, 18 patients (30%); (5) combined multi-axis rotation with translation, 4 patients (6.7%).

Clinical data

Figure 4 Frequency distribution of LC pelvic fracture (A) translational displacements and (B) rotational displacements.

above the threshold of 10 mm in one or more of the possible three directions. Of these 30 patients, 73.34% (N = 22) exhibited translations in the medio-lateral (M/L) direction, whereas only 16.7% (N = 5) and 6.7% (N = 2) of patients had translations in the infero-superior (I/S) and antero-posterior (A/ P) directions, respectively (Fig. 4a). Maximum translations were 58 mm, 22 mm and 21 mm in the M/L, I/S and A/P directions, respectively. Considering rotational displacement, 30 patients were displaced above the defined threshold of 78 in at least one of the possible three directions. Of these patients, 20% (N = 6) exhibited rotations in flexion—extension (F/E), 66.7% (N = 20) in internal— external rotation (I/E) and 36.7% (N = 11) in varus— valgus rotation (V/V) (Fig. 4b). The maximum rotation magnitudes were 19.58, 21.88 and 30.78 in the F/E, I/E and V/V directions, respectively.

Additional clinical data was available on the cause of injury, presence of abdominal injury and treatment for 57/60 patients (95%) (Table 2). Abdominal injuries occurred with the highest frequency in patients with combined single axis and multi axis rotation with translation (groups 4 and 5). No abdominal injuries were seen in patients with pure translation (group 2). However, surprisingly there was a rate of abdominal injury of 33% in the minimally displaced fracture group (group 1). Overall, operative treatment was chosen for 21% of patients (12/57) which is representative of the majority of LC fractures being treated conservatively. Operative treatment was most common (75%) for patients with combined multi axis rotation and translation (group 5) perhaps representing the most serious or unstable conformation, although operative treatment choices did span all groups. LC fractures occurred in the majority of cases from motor vehicle collisions and pedestrians hit by motor vehicles in all groups (overall 79%), with the remaining causes falls from heights and industrial accidents. However, patients with pure translation (group 2) recorded more falls from heights (43%) and relatively low speed motor vehicle collisions (43%) as causes of injury. More severe accidents were noted in many patients in groups 3—5, including individuals thrown from a vehicle, dragged, and heavy objects falling on them. Functional outcome scores (SF36) were available for only 22 patients (37%) at an average follow up of 12 months (Table 3). In comparison to mean standardised summary scores for Canadians,12 patients

Table 2 Clinical data by fracture pattern group: number/percentage of patients with abdominal injuries and those who received operative (vs. non-operative) treatment LC injury group

Patients with available clinical data

Abdominal injury present

Operative treatment

1 2 3 4 5

21 8 6 18 4

7 0 1 8 2

3 3 2 1 3

(33%) (0%) (17%) (44%) (50%)

(14%) (38%) (33%) (13%) (75%)

Lateral compression fracture of the pelvis

899

Table 3 SF36 results at 12 months following injury LC injury group

Number of patients

Physical component scale

Mental component scale

1 2 4 5 Canadian Norms

7 5 6 4

55 42 68 36 51

69 50 74 41 52

Only limited data was available (for 22 patients) with no data available for group 3. Summary results from the physical and mental component scales are presented along with comparative data from Canadian Norms.

in groups 1 and 4 demonstrated superior physical and mental component scores than the standard population. In contrast, groups 2 and 5 had poorer physical and mental status, with the worst outcomes in group 5.

Discussion In testing the robustness of our analysis, initial inter and intra observer measurements revealed some repeatability issues in selecting the anatomical landmarks. Low correlations (Table 1) in the A/P direction were the result of poor depth perception when viewing the model on the computer screen. This was remedied in future trials by verifying landmark positioning through extensive multi-axial viewing of the models. In analysing the 3D reconstructed CT’s, we found that fractures initially clinically diagnosed as LC actually represented a spectrum of displacement patterns, ranging from a minimally displaced hemipelvis to complex translational and rotational displacements. Although it is not surprising that LC nomenclature is used for a variety of displacement patterns, our contention is that a more detailed categorisation will result in improved indicators for non-operative versus operative treatment. Current classification systems sort pelvic fractures by the assumed direction of force applied to the pelvis (Young and Burgess based on Pennal) combined with a qualitative assessment of stability (Tile) and sub-classifications based on fracture morphology (OTA Classification). However, in moving to the OTA classification, two distinct injury mechanisms, open book (B1) and lateral compression (B2), were grouped together under Type B. Even at this time, it was felt that further subclassifications were needed to adequately describe the spectrum of LC fractures. The present study has characterised LC fractures through the quantification of displacement patterns without any assumptions about the direction of force or stability. Our findings suggest classical LC fracture classifications may yield an inhomogeneous group of fractures. This may explain

the wide spectrum of clinical presentation in these patients and as such may necessitate different approaches for optimal outcome. In our series more than one third of the cases (N = 22, 36.6%) were minimally displaced (group 1). While it may be difficult to visualise posterior lesions in undisplaced fractures on plain X-ray film, they are visible on CT. Minimal displacement can be explained only by recoil of the fracture to the primary position or non-displacement. Since the sacrum is compressed, an initial movement is implied. However the final static position seen on CT may be representative of the pelvic post fracture stability. Future biomechanical testing and/or more detailed assessments of clinical outcome are needed to determine if this assumption holds true. Although the intact posterior sacroiliac ligament and the pelvic floor confer an element of stability to the pelvic ring, the amount of displacement during impact may be greatly underestimated by the apparently innocent appearance of these fractures. Such displacements of LC fractures during injury with a subsequent return to their previous position when the injuring force subsides have been previously described by Tile.2 Many apparently undisplaced fractures have caused ruptures of the bladder indicating that significant motion occurred during impact, followed by recoil.2 This concurs with similar findings in our study of an abdominal injury rate of 33% in minimally displaced fractures. However, the low incidence of operative treatment and positive functional outcome scores indicate that this patient population may be doing quite well with conservative treatment. The majority of our pure translation group (group 2, N = 8) demonstrated M/L translations (N = 5). Pure translation occurred along the I/S axis in three cases (5%). By definition, LC fractures are vertically stable indicating that pure vertical translations should not be found in this cohort. These three cases of interest were all displaced superiorly. Furthermore, posterior involvement of the sacrum appeared more extensive than the other LC fractures analysed (in one case an avulsion fracture of the L5 transverse process was identified). These findings suggest that, using existing clinical classi-

900 fication techniques, these fractures may have been misdiagnosed as LC fractures, and really represent type C fractures. While the clinical history of a laterally directed force and the radiologic appearance from the 2D slices may be indicative of LC fracture, a physical examination may reveal a vertical instability due to full posterior element and/or a pelvic floor involvement. Poorer functional outcomes in this group may suggest more instability even without incidence of additional abdominal injury. It is not uncommon that lateral compression fractures diagnosed and treated non-operatively may present displacement in the next clinic visit indicating vertical instability and causing ipsilateral limb shortening. Future wider analyses examining clinical outcome and measuring biomechanical stability are needed to verify if vertical translations identified by 3D analysis are associated with vertical instability.23 If indeed these fractures do identify as unstable, 3D radiological analysis may be vital in preventing fracture misclassification and clinical mismanagement. Furthermore, a higher incidence of falls from heights and relatively low speed motor vehicle collisions in this group may indicate different mechanisms causing pure translational fracture patterns. Group 3 is characterised by pure rotation about one or more of the three axes (N = 8, 13.3%). Internal rotation (3 patients) is consistent with a laterally applied force and the traditional classification of LC fractures. Varus displacement (4 patients) corresponds to a pure laterally applied force to the greater trochanter, whereas valgus displacement (1 patient) may be due to an impact on the cephalad aspect of the ilium. This pattern, while less commonly associated with LC fractures than I/E rotation, could fit with the traditional classification of a bucket handle as described by Tile.2 Two patients had rotations in extension, corresponding to an antero-lateral impact. The remaining 2 patients exhibited rotations about two axes simultaneously, indicating a more complex impact vector; a potential indicator of a higher degree of instability. Group 4 describes the patients (N = 18, 30%) that experienced a combination of rotation about a single axis, paired with translation along one or more axes. Of particular interest in this group were 11 patients who all had internal rotations coupled with lateral translations. These are classic LC fractures with an opening of the sacral fracture accounting for the lateral translation, and an internal rotation resulting from the lateral impact anteriorly. Varus (N = 1) and valgus (N = 2) rotations were also observed with varying degrees of translation, perhaps describing more severe cases of the previously described group 2 displacements. The remaining

A. Khoury et al. analyses yielded 2 external rotations, 1 internal rotation and 1 flexion rotation with varying translations. As in group 1, there was a relatively high incidence of abdominal injury (44%) yet a low incidence of operative treatment, as well as a positive functional outcome. Group 5 consists of 4 patients with multi axis rotations coupled with a translation. This group consists of four distinct patients, one of whom large rotations about all axes (>198) and a large lateral translation (5.8 cm). The multi-axis displacement in these cases suggests complex impact causing injury. As such, these injuries would not fit easily within previous classification schemes that utilised direction of impact in grouping pelvic fractures. This group is perhaps the very patient set that, although initially diagnosed with stable LC fractures, may benefit most from 3D analysis. These patients also had the highest incidence of abdominal injuries secondary to relatively severe causes of injury. These patients had the highest incidence of operative treatment and demonstrated the poorest functional outcomes, both physical and mental, 12 months post injury. There is a common misconception that a lateral compression force will result only in lateral compression type pelvic fractures. The largest number of rotational displacements occurred in the I/E direction, while the largest translations were along the M/L axis. These findings are consistent with a laterally directed impact. However, it is known that lateral compression forces can result in multiple fracture patterns (ranging from a stable type A fractures to completely unstable type C fractures with full failure of the posterior elements and the pelvic floor). Classic clinical analysis of pelvic fractures relies substantially on 2D axial slice qualitative interpretation, which alone is not always sufficient to decide on non-operative versus operative treatment. Currently, operative treatment is deemed essential at the extremes of instability where there is gross mobility on clinical exam, severe pain or rotation resulting in leg length deformity.1,15,20,21 Non-operative treatment is easily selected for patients with little pain on weight bearing. However, in those patients with significant pain on walking but less severe clinical mobility and varying degrees of sacral impaction, choosing operative treatment is less clear. Visualisation and quantification of displacement patterns in 3D may provide much needed additional information for clinical decision-making. For example, such analyses can pick up subtle effects in minimally displaced injuries and quantify displacements crucial to diagnosis in more severe injuries, such as vertical displacement that may indicate misclassification of a type C fracture (Fig.5).

Lateral compression fracture of the pelvis

901

Figure 5 (A) A posterior view of a 3D reconstruction of a pelvic fracture demonstrating a pure vertical translation. Note the inferior translation of the right (fractured) hemipelvis. (B) A CTscan axial slice of the same pelvis. The arrow indicates the involvement of the posterior elements.

A limitation of the 3D analysis was the high level of thresholding chosen to secure a 100% level of confidence in identifying only pathological displacements versus normal variations in pelvic anatomy. Using these criteria, some displacements associated with LC fracture may have fallen just below this threshold and as such been neglected. Use of a slightly lower threshold would have yielded better group inclusions and explained seemingly uncharacteristic results. This may be accomplished through reduction of analytic errors. The method employed in the current study is operator dependent and requires advanced modelling knowledge. Possible improvements include the use of image registration techniques to replace operator identification of landmarks. The ideal level for thresholding will ultimately be determined by the relationship between the magnitudes and patterns of displacement and biomechanical pelvic stability. Additional work is necessary to understand the relationship between the identified patterns and magnitudes of displacement with pelvic instability and combine these findings with clinical outcomes (pain and function) to determine their potential impact on clinical decision-making.

reported. This indicates a complexity in LC fractures, which may explain the variations seen in treatment protocols and clinical outcomes associated with this injury. Using a single label of ‘‘lateral compression’’ alone cannot fully describe the personalities of these fractures. Characterisation of 3D displacement patterns from imaging data should ultimately aid clinicians in making better decisions as to the need for operative versus non-operative treatment of lateral compression pelvic fractures.

Conflicts of interest There are no financial or personal relationships between the authors and other people or organizations that could inappropriately influence our work.

Acknowledgements This work was supported by grant funding from the University of Toronto and the Sunnybrook Holland Musculoskeletal Program. Additionally, we would like to thank Aimee Gallant for her assistance in compiling the clinical data.

Conclusion In this study we developed a novel technique for the quantification and 3D radiological analysis of displacement in previously identified lateral compression fractures. We identified five groups describing a wide spectrum of pelvic rotations and translations, some patterns of which have not been previously

References 1. Bellabarba C, Ricci WM, Bolhofner BR. Distraction external fixation in lateral compression pelvic fractures. J Orthop Trauma 2006;20(Suppl. 1):S7—14. 2. Burgess AR, Eastridge BJ, Young JW, Ellison TS, Ellison Jr PS, Poka A, Bathon GH, Brumback RJ. Pelvic ring disruptions:

902 effective classification system and treatment protocols. J Trauma 1990;30:848—56. 3. Caraig JJ. Introduction to robotics: mechanics & control. Reading MA: Addison-Welsley Publishing; 1989. p. 15—42. 4. Dalal SA, Burgess AR, Siegel JH, et al. Pelvic fracture in multiple trauma: classification by mechanism is key to pattern of organ injury, resuscitative requirements, and outcome. J Trauma 1989;29:981—1000. 5. Dawson JM, Khmelniker BV, McAndrew MP. Analysis of the structural behavior of the pelvis during lateral impact using the finite element method. Accid Anal Prev 1999;31(1/ 2):109—19. 6. Duck TR, Ferreira LM, King GJ, Johnson JA. Assessment of screw displacement axis accuracy and repeatability for joint kinematic description using an electromagnetic tracking device. J Biomech 2004;37(1):163—7. 7. Dunning CE, Lindsay CS, Bicknell RT, Patterson SD, Johnson JA, King GJ. Supplemental pinning improves the stability of external fixation in distal radius fractures during simulated finger and forearm motion. J Hand Surg Am 1999;24(5):992— 1000. 9. Gokcen EC, Burgess AR, Siegel JH, et al. Pelvic fracture mechanism of injury in vehicular trauma patients. J Trauma 1994;36:789—96. 10. Gordon RO, Mears DC. Lateral compression injury of the pelvis. J Bone Joint Surg 1991;73A:1399—401. 11. Hirvensalo E, Lindahl J, Bostman O. A new approach to the internal fixation of unstable pelvic fractures. Clin Orthop Relat Res 1993;297:28—32. 12. Hopman WM, Towheed T, Anastassiades T, Tenenhouse A, Poliquin S, Berger C, Joseph L, Brown JP, Murray TM, Adachi JD, Hanley DA, Papadimitropoulos E. Canadian normative data for the SF-36 health survey. CMAJ 2000;163(3): 265—71.

A. Khoury et al. 13. Lewis PR, Molz FJ, Schmidtke SZ, Bidez MW. A NASS based investigation of pelvic injury within the motor vehicle crash environment. In: Proceedings of the 40th stapp car crash conference; 1996. p. 1—7. 14. Lindahl J, Hirvensalo E, Bostman O, Santavirta S. Failure of reduction with an external fixator in the management of injuries of the pelvic ring. J Bone Joint Surg Br 1999;81(6): 955—62. 15. McLaren AC, Rorabeck CH, Halpenny J. Long-term pain and disability in relation to residual deformity after displaced pelvic ring fractures. Can J Surg 1990;33(6):431—2. 16. Miranda MA, Riemer BL, Butterfield SL, Burke CJ. Pelvic ring injuries: a long term functional outcome study. Clin Orthop 1996;329:152—9. 17. OTA Committee for Coding and Classification. Fracture and dislocation compendium. J Orthop Trauma 1996;10(Suppl. 1): 66—69. 18. Pennal GF, Tile M, Waddell JP, Garside H. Pelvic disruption: assessment and classification. Clin Orthop Rel Res 1980;151: 12—21. 19. Routt ML, Simonian PT, Ballmer F. A rational approach to pelvic trauma. Clin Orthop 1995;62—74 (318). 20. Tile M. Fractures of the pelvis and acetabulum, 3rd ed., Baltimore: Williams & Wilkins; 2003. p. 3—21, 32—45, 116— 128, 153—161. 21. Tile M. Pelvic fracture: operative versus non-operative treatment. Orthop Clin North Am 1980;11:423—64. 22. Ware JE, Snow KK, Kosinski M, Gandek B. SF-36 health survey: manual and interpretation guide. Boston: The Health Institute, New England Medical Centre; 1993. 23. Young JW, Burgess AR, Brumback RJ, Poka A. Lateral compression fractures of the pelvis: the importance of plain radiographs in the diagnosis and surgical management. Skeletal Radiol 1986;15(2):103—9.