Characteristics of Multifragmentary Patellar Fractures: Three-dimensional Computed Tomography-based Fracture Mapping Study

Jae-Woo Cho Korea University Guro Hospital Zepa Yang Korea University Guro Hospital Eic Ju Lim Chungbuk National University Hospital Seungyeob Sakong Korea University Guro Hospital Wonseok Choi Korea University Guro Hospital Whee Sung Son Korea University Guro Hospital Hanju Kim Korea University Guro Hospital An Seong Chang Korea University Guro Hospital Do-Young Lim Korea University Guro Hospital Youngwoo Kim Uijeongbu St. Mary's Hospital Beom-Soo Kim Keimyung University Dongsan Medical Center Jong-Keon Oh (  jkoh@korea.ac.kr ) Korea University Guro Hospital


Introduction
Although patellar fractures represent only 0.5-1.5% of all skeletal injuries 1,2 , they can be signi cant, often resulting in poor functional outcomes. Since these injuries potentially compromise the extensor mechanism, impairment of patellofemoral articulation can directly affect the ability to perform daily activities and the quality of life [3][4][5][6] . Therefore, when treating a patellar fracture, it is important to achieve, as accurately as possible, an anatomic reduction of the articular surface in addition to restore the extensor mechanism.
Patellar fractures requiring surgery are comminuted (multifragmentary) type fractures in up to 55% of the cases 1,7 . A recent computed tomography (CT) study revealed that distal pole patellar comminution was noted in 88% of the cases, indicating a substantial prevalence of comminution among fracture cases 8 .
Furthermore, comminuted patellar fractures were frequently associated with coronal split fractures and inferior pole comminution [8][9][10][11] . This complex fracture type, which is described as a complete articular frontal/coronal multifragmentary patellar fracture, is classi ed as a 34C3 fracture, according to the current AO/OTA classi cation 12,13 . This classi cation system is based on the anterior/posterior view of the patellar fracture without any subgroup categorization, leading to a limited explanation of complex articular fracture patterns.
Detailed knowledge of common fracture patterns and of locations of articular impaction or free articular fragments would aid in simplifying the sequence of fracture reduction and selection of the optimal xation methods in patellar fracture management. However, limited literature is available regarding anatomic patterns of articular injuries in comminuted patellar fractures, thereby, necessitating further evaluation. The present study aimed to de ne the location and frequency of fracture lines on the coronal and sagittal planes of the articular surface and then to map the locations of articular impaction in AO/OTA 34C3 fractures.

Subjects
This retrospective study included 142 patellar fractures that were surgically treated at a single university hospital between January 2015 and December 2018. We obtained the CT scans and plain radiographs of each patient's knee injury. Preoperative CT was performed in 1-mm-thick intervals. The medical records were retrieved to identify the patient demographic characteristics, injury mechanisms, and associated injuries around the knee. The present study excluded patients aged < 18 years and > 71 years and those with a previous fracture or dislocation history of the patella or bipartite patella. Of the 142 patellar fractures recorded, we identi ed 66 multifragmentary patellar fractures in 66 patients. Using the AO/OTA classi cation, we classi ed the multifragmentary patellar fractures as type 34C3 fractures.
Coronal articular surface fracture mapping All Digital Imaging and Communications in Medicine (DICOM) les of the preoperative CT scans were imported into Mimics® software (Materialise NV, Belgium) to create a 3D model of each fracture. In the 3D patella model, fractures with over 1 mm displacement were segmented independently. After completing fracture segmentation, the 3D-modeled stereolithography le was imported into 3-matic® software (Materialise NV, Belgium) to reassemble fracture fragments. Each fragment was moved threedimensionally to t between nearby fragments. We used fracture margins, the patella's outer surface continuity, and the shape of the contralateral patella on radiographs as references for fracture reduction (see Supplementary video 1, which demonstrates the Fracture fragment reduction using 3D software).
Images of the reassembled articular surface were captured and retrieved in Adobe Photoshop software (Adobe, CA, USA). To superimpose the fracture pattern onto the standard template, we rotated each captured image along the axis of the patellar vertical ridge. Next, we resized the entire articular surface according to facet locations to maintain the image proportions. We then drew the fracture lines in each image on a superimposed template using the pencil tool in Photoshop (Fig. 1). The coronal articular fracture line drawings were created for all 66 cases and merged together to create a coronal articular fracture map Denomination and classi cation of the fracture lines in coronal articular fracture map Each fracture line was classi ed as "primary" or "secondary" and "horizontal" or "vertical" based on the initial displacement and orientation, respectively. The "primary" fracture line was de ned as "the largest displaced fracture line on the preoperative 3D CT scans considering both fracture gap or step" and the "secondary" fracture lines as "other displaced fracture lines on the preoperative 3D CT scans." Fracture lines oriented at an angle < 45° were classi ed as "horizontal", while those oriented at a greater angle were classi ed as "vertical".
We used the "Fracture line denominating and mapping program" to classify each fracture line and to determine the average value of the classi ed fracture lines. In this program, each fracture line was sectioned from its start point to its exit point or point of intersection with another fracture line. The sectioned lines were designated their respective categories according to the independent observers' judgement. The result of each classi cation was saved as a categorical digital value to make reliability analysis possible.
Two independent observers, trained through an orthopedic trauma fellowship, were involved in the classi cation process. They denominated the fracture lines in 66 cases individually, according to our working de nition of the classi cation, and repeated the same classi cation twice with an interval of 14 days. The nal denomination of each fracture line was con rmed by reviewing the results of the denominated fracture map by arriving at a consensus with two senior surgeons.
The fracture lines, classi ed as "primary horizontal-", "primary vertical-", "secondary horizontal-" and "secondary vertical-" fracture lines in the nal denominated coronal articular fracture map, were digitized into coordinated values on the X and Y axes of a graph, and the lines were subtracted according to their classi cation. The mean value of each classi ed fracture line was calculated by averaging the coordinated value on the graph separately.
Midsagittal fracture mapping We selected the sagittal DICOM images at the patellar vertical ridge to create a midsagittal fracture map. Based on our presumption of high incidence of coronal fragments at the boundary of the medial and lateral facets (without any selection bias), we selected the patellar vertical ridge as the reference because the vertical ridge can be easily identi ed by judging the axial scout images of CT scans. The identi ed fracture fragments on the midsagittal image were reduced by proper rotation and aligning along the anterior cortical and posterior articular margins. The fracture lines in each case were then individually reproduced on a standard patellar template of the midsagittal plane using Photoshop, similar to coronal articular fracture mapping (Fig. 2). After combining the data of the 66 fractures, the sagittal fracture map was superimposed onto a template.
Mapping of the coronal split fragment zone Cases in which the articular surface step-off was > 2 mm, when the anterior cortex was perfectly reduced in 3D model, were suspected of having a coronal split fragment. By reviewing all CT images, we con rmed the presence of coronal split fragments and the con guration of the split (free articular or impacted type) (Fig. 3). For mapping, these coronal fragments were segmented and reduced to achieve an anatomical reduction of the articular surface. Fragment locations were marked and painted on the template by the superimposition of articular surface images. After putting all the cases with coronal split fragments together in a single template, a combined mapping of the coronal split fragment zone was created (Fig. 4).

Data analysis
The morphologic fracture map analysis was descriptive. For each categorical variable, we used Cohen's kappa statistics to determine the interobserver and intraobserver reliabilities of the primary and secondary fracture lines. We also analyzed the factors possibly associated with the presence of speci c fracture patterns (coronal split fracture or inferior pole fracture) using a logistic regression analysis. The statistical signi cance was set at a P-value < 0.05. All statistical analyses were performed using SPSS version 20.0 software (IBM, Armonk, NY, USA).

Demographic characteristics
Among the 43 males and 23 females examined, the average age was 52 years (range, 19-70 years) and the average body mass index (BMI) was 23.08 kg/m 2 (range, 15.8-34.6 kg/m 2 ). The main injury mechanism observed among the patients observed was falling with the knee in a bent position (n = 49, 74.2%). "Tripping over while running" was the most common cause of injury. Associated injuries, besides patellar fractures, were noted in 18 patients. Detailed data, including age distribution, are described in Table 1. The interobserver agreement of the classi cation of fracture lines had a Cohen's kappa value of 0.787 for the primary and 0.725 for the secondary fracture lines. Intraobserver agreement, performed by the rst observer, resulted in a Cohen's kappa value of 0.8216. There was substantial agreement with interobserver reliability and strong agreement with intraobserver reliability. Midsagittal fracture mapping The study identi ed an X-shaped fracture line representing primary and secondary horizontal fracture lines. Between the primary and secondary horizontal fracture lines, a fracture connection called the "main coronal fracture line" exists. This fracture line is mostly located at the anterior third of the midsagittal view. The development of coronal split fragments can be identi ed in the combination of primary horizontal, secondary horizontal, and main coronal fracture lines (Fig. 6a).
Mapping of coronal split fragment zone The coronal split fragment was observed in 58 cases (87.9%). The most common area of the coronal split fragment was the inferior facet of the lateral and medial facets. The fragments were of the free articular type in 32 cases (55.2%) and the impacted type in 26 cases (44.8%) (Fig. 6b).
Factors affecting speci c fracture patterns Age, gender, BMI, injury mechanism, and the presence of primary horizontal and secondary vertical fracture lines were not statistically signi cant concerning the occurrence of inferior pole or coronal split fractures. The presence of the free articular type of coronal split fragment was signi cantly related to inferior pole fractures (odds ratio, 8.15; 95% con dence interval [CI], 2.39-27.6; P = 0.01).

Discussion
The fracture mapping technique described by Cole et al. 14 21 . However, this is the rst study to analyze the articular fracture patterns systemically in the coronal and sagittal views to improve the understanding of complex articular injuries of the patella. We also believe that actual fracture patterns and comminution in the coronal articular and the sagittal planes have not been re ected in the current classi cation. Thus, this system may not be suitable in the clinical setting for understanding the complex articular injuries of the patella, besides the establishment of the sequence of fracture reduction and selection of optimal xation methods in complex patellar fracture management. Therefore, our ndings may be used to provide substantial information for the development of new AO/OTA classi cation systems for patellar fractures.
In this study, we created coronal articular and midsagittal fracture maps from images of CT scans of the 66 multifragmentary (AO/OTA 34C3) patellar fractures in our cohort to analyze articular fracture patterns systemically in the coronal and sagittal planes. Our descriptive analysis of the coronal articular fracture map revealed that multifragmentary patellar fractures tend to have three fracture lines: a primary horizontal fracture line that compromises the largest displacement at the middle and inferior facet levels of the patellar body; a secondary horizontal fracture line at the inferior margin of the articular facet; and a secondary vertical fracture lines which compromise satellite fragments on medial odd facet, on lateral facet or on both. Furthermore, the midsagittal fracture map identi ed that these fracture lines result in a combined inferior pole fracture and coronal split fragment at the inferior facet of the lateral and medial facets, with the collaboration of the main coronal fracture line.
There are several clinical implications of improved understanding of multifragmentary patellar fracture morphology, including the morphology described in this study, with respect to facilitating successful treatment of multifragmentary patellar fractures (Fig. 7). First, the presence of a primary horizontal fracture line that runs through the middle or inferior facet can be used as a guideline for main fracture xation. Although many surgical options for stabilizing multifragmentary patellar fractures have been introduced, there remains no consensus regarding fracture xation 24,25 . The presence of primary horizontal fracture lines can require the use of tension-band construct xation even in multifragmentary fracture patterns 1,26 . However, comminuted articular fractures need to be reconstructed anatomically before xing the main fractures to ensure fracture site compression in tension-band construct xation 27 . Tension-band xation augmented with other supplementary xatives, including mini plates or screws, multiplanar tension-band wiring, and plating on the tension side, are good examples for which this rationale is used 9,28 .
The second implication of our study relates to the presence of coronal split fracture fragments. Most cases (up to 87.8%) had coronal split fragments of either the free articular or the impacted type. These ndings highlight the di culties and risks involved in articular reduction. The most widely used surgical approach is the anterior approach, in which articular reduction is judged by reducing the anterior cortex indirectly, either by intraoperative C-arm imaging or by palpating articular reduction through the retinacular tear 26 . If coronal split fractures are not managed before main fracture reduction, there is a high risk of articular malreduction. There are many surgical strategies for managing coronal split fragments, including changing the surgical approach. An alternative surgical approach is the lateral parapatellar approach. Gardner et al. 29 suggested using the lateral parapatellar approach and direct reduction to manage comminuted articular fractures. By everting the patellar articular surface, direct visualization for the treatment of complex articular fractures is achieved. An additional method involves reducing the coronal fragment through the fracture window. The coronal split fragment is created by primary and secondary fracture lines. Through the fracture window made from the main primary horizontal fracture line, reduction and xation of the coronal fragments can be achieved. Moreover, these coronal split fragments can be classi ed into two types: free articular and impacted. According to the coronal split fragment type, reduction and xation can differ. If fragments are of the free articular type, they can be reduced back to either proximal or distal main fragments through the fracture window and xed with imbedded mini-screws or through anterior cortical plating. If fragments are of the impacted type, the fracture can be dis-impacted through the fracture window and the defect can be lled with allograft bone chip grafts 9 .
The third implication is the involvement of inferior pole fractures. Similarly with the rate of comminuted inferior pole fracture in the previous report 8 , the results obtained in our study demonstrate that most cases (83.3%) were associated with inferior pole fractures. Moreover, more than half of them (54.5%) showed comminution with a sagittal split. This nding suggests that caution must be taken to manage associated inferior pole fractures in order to prevent unexpected breakdown of the extensor mechanism during the active rehabilitation phase. Several supplementary surgical options for managing associated inferior pole fractures and their coverings have to be planned preoperatively; these include adding cerclage wiring, suturing distal pole fragments, rim plate augmentation, basket plating, and multiplanar plating 10,29−34 . Furthermore, these ndings have to be re ected in the design for precontoured plates in patellar fractures 34 .
In our study, the free articular type of coronal split fragment was statistically associated with the presence of an inferior pole fracture. This nding can be explained in the midsagittal fracture map as the coronal fragment, which is mostly located on the inferior facet and has a distal fracture split exiting on the articular side as a free articular fragment. This split line is associated and connected with the secondary horizontal fracture line, resulting in an inferior pole fracture. This nding suggests that an associated injury can be predicted based on fracture morphology. Therefore, if free articular coronal fragments are observed on plain radiographs, the associated inferior pole fractures should be carefully examined using CT scans, when surgical treatment is performed 8 .
There are several limitations of this study. First, the possibility of selection bias cannot be disregarded. Most injuries in our cohort resulted from falling down on a bent knee. Patients sustained multifragmentary patellar fractures by tripping over while running, tumbling down the stairs, and through other moderate energy causes of injury involving a bent knee position. However, enrolling a comparable number of high-energy injuries associated with direct hit, such as dashboard injuries in motor vehicle collision, would be complicated. Thus, our results are applicable only for cases involving low-to moderate-energy trauma. Second, the reduction of fractures in the model using 3D software was manually performed. In current 3D software systems, no auto-reduction function exists in the fracture model. To limit the risk of technical errors, all procedures were performed by quali ed technicians, and each process was supervised and con rmed by two senior surgeons. Third, our analysis of fracture patterns lacks the clinical relevance of proven clinical data. There is still some doubt regarding how this fracture mapping technique can have notable clinical impact. Prospective cohort studies comparing the clinical results between groups treated based on 3D CT fracture pattern analysis and control groups treated based on conventional classi cation should be conducted in the future. We believe that articular fracture mapping along the articular surface, rather than frontal imaging of gures as in the conventional classi cation system, can enhance understanding of the articular fracture pattern con guration and systemize the stabilization strategy, with an emphasis on articular reduction.
To our knowledge, this is the rst study to elucidate common articular fracture lines and associated fracture patterns, including coronal split fragment and inferior pole fractures, using a 3D fracture mapping technique for AO/OTA 34 C3 multifragmentary patellar fractures.

Conclusion
Mapping multifragmentary patellar fractures revealed that comminuted patella fractures have a primary articular horizontal fracture line with secondary vertical and horizontal fracture lines, resulting in coronal split fragments (87.8%) and inferior pole fractures (83.3%).

Declarations
Con ict of interest: The authors declare that they have no con ict of interest. The software employing the method of mapping three-dimensional reassembled fracture models was used for coronal articular surface fracture mapping (a). Segmented fracture fragments were reassembled (b). Fracture lines that were present on the articular surface (red lines) were drawn on a standard template by superimposing captured images (c).  Type of coronal split fragment, the free articular type was diagnosed when the coronal fragment was completely separated with a loss of bone continuity from the proximal or distal axial fragments requiring coronal, sagittal, and axial reduction of the fragment. (yellow arrow head represents the exit of fracture line of coronal split fragment) (a). The impacted type was diagnosed when the fragment only had one plane deformity with an intact distal subchondral cortical hinge (white arrow head) (blue bi-arrowed line represent the width of proximal fragment, red bi-arrowed line represent the width of distal fragment, the dotted white lines represent a step off between two lines) (b) Figure 4 The method of mapping the coronal split fragment zone. In a case with articular step-off presented after the anterior cortex was perfectly reduced, the coronal split fragment was identi ed in a midsagittal cut and articular surface imaging (a). The zone of the coronal split fragment was marked on the coronal articular surface template (b).   A 69 year old male sustained a multifragmentary patellar fracture of the Rt. knee (a), Coronal articular fracture mapping shows the Primary horizontal fracture line (black dotted line) is separating the proximal and distal main fragments, Secondary horizontal fracture line (blue dotted line) creates an inferior pole fracture fragment (yellow oval symbol) and Secondary vertical fracture line (yellow dotted line) creates satellites fragments (pink heart symbol). The combination of these fracture lines results in a coronal split fracture fragment (blue star symbol) (b), Coronal split fragment was xed with two embedded 1.5mm screws through the primary horizontal fracture window. Upper satellite fragment and proximal main fragment was xed with anterior cortical plating. Primary and Secondary horizontal fracture lines were reduced with a pointed reduction clamp. Tension band plating was applied over the inferior pole fracture fragment (c), At 3 months follow up, fracture was united and range of motion was recovered fully (d).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. supple1.mp4