Subjects
After hospital institutional review board approval was obtained, the medical records of patients treated in our hospital between 2019 and 2021 were retrospectively reviewed. Eligibility for this study required patients to have computed tomography (CT) data available for the injured knee. The patients were categorized into one of two groups: (1) those with a noncontact ACL injury and (2) those who had a fracture of the tibial plateau resulting from a violent injury (control group). To be included as a case in the ACL injury group, patients were confirmed via clinical examination, magnetic resonance imaging (MRI), and arthroscopic visualization at the time of ACL reconstruction by two experienced orthopaedic surgeons. A noncontact ACL injury was defined as an event not occurring due to direct contact between the ACL-injured knee and the ground, another athlete, or other object.
Our inclusion criteria were as follows: noncontact ACL injury or fracture of the tibial plateau, CT scan for the injured knee, 18–45 years old and having a body mass index (BMI) between 18 and 45 kg/m2. Our exclusion criteria were as follows: dysplasia of the knee joint, evidence of osteoarthritis, prior knee injury, or inadequate CT images (such as CT scans without intact femoral condyles). The patients were classified according to type of injury, either noncontact ACL injury or fracture of the tibial plateau. Subjects were excluded from the ACL injury group if they had additional ligamentous injury (medial collateral ligament, lateral collateral ligament, posterior cruciate ligament, or medial patellofemoral ligament). In addition, tibial plateau fractures were also excluded from the ACL group, such as ACL avulsion fractures, so that the ACL injuries were exclusively ACL body injuries. After the medical records were reviewed for eligibility, 74 noncontact ACL-injured cases (34 females, 40 males) were identified from the Department of Orthopaedics at our hospital. The control data were obtained from patients treated in the trauma centre of the same hospital and matched to ACL-injured patients by age and sex. Subjects were excluded from the control group if they had a prior ligament injury (medial collateral ligament, lateral collateral ligament, posterior cruciate ligament, and medial patellofemoral ligament). Although tibial plateau fractures are often associated with avulsion of the ACL, PCL, MCL and PLC, the mechanisms of noncontact ACL injuries are different from those of tibial plateau fractures, which are high-energy violent injuries. The control group was composed of 74 individuals (34 females, 40 males). Figure 1 shows the flow diagram of patient enrolment in the study.
Three dimensional model reconstruction
CT scanning was performed using a 64-slice CT scanner (Somatom Sensation 64, Siemens, Erlangen, Germany) with the knee in extension following surgery to evaluate surgical outcomes. To obtain an accurate sagittal view, a three-dimensional model of the distal femur was created with Digital Imaging and Communications in Medicine (DICOM) CT images, which were obtained using the image processing software Mimics (21.0 Materialise, Leuven, Belgium). The threshold of all cases was set at 226 HU, and the femoral mask was automatically separated using the “Region Grow” function. The three-dimensional model of the femur was reconstructed using the “Calculate Part” function, and the optimal quality was chosen. Then, three-dimensional rotation of the femoral model was performed using the “Pan” and “Rotate” functions for accurate realignment. To obtain the nonorthogonal, sagittal imaging plane, rotation of the femoral three-dimensional model was performed as described by Howell et al. [16]. This was defined as the sagittal imaging plane of the distal femur. The sagittal imaging plane of the medial distal femur was considered to be plane a, and the sagittal imaging plane of the lateral distal femur was considered to be plane b.
Measurement methods
Measurements for both study groups were obtained from a sagittal view image by two independent blind observers and consisted of the lateral femoral posterior radius (LFPR), medial femoral posterior radius (MFPR), lateral height of the distal femur (LH), medial height of the distal femur (MH), lateral femoral anteroposterior diameter (LFAP), and medial femoral anteroposterior diameter (MFAP). Two circles were centred on the femoral shaft to determine the long axis of the distal femur. A line passing through the centre of both circles was considered the long axis of the distal femoral shaft. The LFPR and MFPR were determined using a circle-fitting technique in which the femoral condyle was assumed to have a single radius of curvature in flexion from 10° to 160° as described [15, 16, 20]. The line crossing the centre of the femoral posterior circle and perpendicular to the axis of the distal femoral shaft was used to determine the LFAP and MFAP. The distance from the intersection of those lines to the distal femoral condyle was used to determine the LH and MH. The LFPR was divided by the LFAP and multiplied by 100%, and this ratio was defined as the lateral femoral posterior radius ratio (LFPRR). The MFPR was divided by the MFAP and multiplied by 100%, and this ratio was defined as the medial femoral posterior radius ratio (MFPRR) (Fig. 2). The interobserver and intraobserver reliabilities were calculated by using the intraclass correlation coefficient (ICC). To assess intraobserver reliability, each patient was remeasured > 1 week after the initial measurements by the first blinded observer. To determine interobserver reliability, an additional blinded and independent observer repeated the set of measurements.
Statistical analyses
Statistical analyses were conducted using SPSS software (24, IBM, Chicago, USA). The mean, standard deviation, range and frequency were calculated for continuous variables and percentages. The ICC was calculated to ensure interobserver and intraobserver reliability. According to the normality of the measurements, the Mann-Whitney U test and 2-sample t-test were performed to detect significant differences in all continuous variables, including age, height, weight, BMI, LFPR, MFPR, LH, MH, LFAP, MFAP, LFPRR, and MFPRR, between the ACL-injured group and the control group. The odds ratio (OR) was calculated to determine whether an increased LFPRR and increased MFPRR were risk factors for noncontact ACL injury. A receiver operating characteristic (ROC) curve was used to determine the association between LFPRR and ACL injury and the association between MFPRR and ACL injury. The cut-off was determined at the maximal Youden index with autofit sensitivity and specificity.
Power analysis was performed using G*Power (3.1.9.2, Kiel, Germany) to determine the sample size. According to the preliminary results [11, 12], to achieve a power of 0.95 (a, 0.05; effect size, 0.65), a total of 126 patients (63 per group) were required for this study.