Patients and methods
This retrospective study was reviewed and approved by the Ethics Committee of Zhongshan Hospital. All procedures were performed in accordance with the Declaration of Helsinki and strictly adhered to institutional guidelines. A total of 63 patients with anterior pelvic ring injury were enrolled from January 2014 to January 2017, with a minimum follow-up of 13 months. Inclusion criteria were unstable anterior pelvic ring fracture with a stable posterior ring (either intact or recovered after fixation) diagnosed by the senior trauma surgeon, hemodynamic stability, and full consciousness. Patients with an open contaminated wound, who were lost to follow-up before 3 months, had a pathologic fracture, or were <16 years of age were excluded (Fig. 1).
The pelvis of each patient was examined by preoperative radiography (including anterior-posterior [AP], inlet, and outlet views) and computed tomography (CT). Imaging data were analysed by 2 senior orthopedists according to the Arbeitsgemeinschaft für Osteosynthesefragen (AO) and Orthopaedic Trauma Association (OTA) modified tile type classification. Post-surgery clinical outcome was assessed based on Tornetta and Matta grade, Majeed score, and complications.
If required, surgical reduction of the posterior pelvic ring injury was performed prior to fixation as previously described . For MPSRF, a 3- to 4-cm oblique incision was made over each anterior inferior iliac spine (AIIS). A bony tunnel from the AIIS to the posterior superior iliac spine was created with a pedicle finder. A polyaxial pedicle screw with a diameter of 7 mm and length of 60 mm was inserted into the tunnel to a depth of about 2 cm from the bone surface to avoid compression of vascular tissue. A subcutaneous tunnel was created from the incisions at bilateral AIIS to the Pfannenstiel incision over the deep fascia. The curved titanium rod was inserted to connect the 3 bilateral pedicle screws via the subcutaneous tunnel. After confirming that there was sufficient space between the rod and bone by fluoroscopy, the screws were tightened with a torque screwdriver, with those at bilateral AIIS tightened before the one at the pubic tubercle. A representative case is shown in Supplementary Figure 1 (S1).
The INFIX was inserted in the same manner. Briefly, 2 polyaxial pedicle screws were placed at bilateral AIIS. A subcutaneous tunnel was created from 1 side of the AIIS to the other, and the precontoured titanium rod was connected to the 2 screws via the tunnel. Fracture reduction was performed by the same method as described above. A representative case is shown in Supplementary Figure 2 (S2).
Postoperative management and follow-up
Functional exercises of the lower limbs and joints were initiated in bed as early as possible after the operation to prevent deep vein thrombosis; regular wound and dressing care was performed in the outpatient clinic until sutures were removed at 2 weeks’ post-surgery. Crutch-assisted partial weight bearing was permitted at 6 and 10 weeks for AO/OTA type B and C injuries, respectively, as long as the pain was tolerable. Full weight bearing was allowed if osseous union was confirmed by radiography. Physical therapy was prescribed for muscle strengthening and gait training. Hardware removal was performed between 12 and 14 months after the surgery.
Patients were followed up by phone or at the clinic. Routine follow-ups were scheduled for postoperative weeks 4 and 8 and months 6, 12, 18, and 24. Radiographic images included a 3-view (AP, inlet, and outlet) pelvis series and all patients were asked about pain, numbness, and motor ability at each follow-up. Physical and neurologic examinations were performed to evaluate irritation around the implant as well as the condition of the lateral femoral cutaneous nerve (LFCN) and femoral nerves.
Radiographic and functional outcome assessment
The extent of fracture reduction was evaluated by postoperative X-ray examination and graded according to the method of Tornetta and Matta  as excellent (displacement ≤4 mm), good (5–10 mm), fair (10–20 mm), or poor (displacement >20 mm). A grade of excellent or good was regarded as a satisfactory outcome.
Clinical outcome measures included operation time, hospital stay, intraoperative blood loss, and postoperative complications. The Majeed rating system was used to assess functional outcome at 6 months, the time of implant removal (10–14 months), and last follow-up; the variables were pain (30 points), standing (36 points), sitting (10 points), sexual intercourse (4 points), and work (20 points). Aggregate scores were classified as excellent (>85), good (70–84), fair (55–69), or poor (<55) .
In order to define the solid geometry of the pelvis, we constructed a model of the pelvis of a healthy 32-year-old male (height, 175 cm and weight, 73 kg) based on images from a CT scan. The material properties of the model were obtained from previous studies (Table 3) . A 2-cm gap was created at the right superior and inferior rami to simulate injury using Geomagic Studio software (3D Systems Inc, Rock Hill, SC, USA). 3-Dimensional models of the rod and screws were constructed using Creo v3.0 software (Parametric Technology Corp, Needham, MA, USA). The materials for the different models and implants were assumed to be elastic, isotropic, and linear. A value of 0.33 was set as Poisson’s ratio (y) for both cortical and cancellous bone.
Models of the intact pelvis and injured pelvis with the 2 types of fixation were imported into Workbench v.17.0 software (ANSYS Inc, Canonsburg, PA, USA) to analyse equivalent von Mises (VM) stress and displacement. We performed convergent analysis to balance the accuracy and efficiency of the finite element simulation by adjusting element size. Interaction surfaces including sacrum, sacroiliac cartilage ilium, pubic rami, and bone implant were fully constrained and a vertical load of 600 N was imposed on the superior surface of sacrum to simulate upper body weight. The unilateral (left/right) acetabulum, bilateral acetabulum, and bilateral ischium were fully constrained to simulate single-leg (left/right) and dual-leg standing and sitting postures, respectively (Supplementary Figure 3).
Data were analysed using SPSS v20.0 software (SPSS Inc, Chicago, IL, USA). Data satisfying the conditions of normality are presented as means ± standard deviation; non-normal data are presented as medians and quartiles. Differences in categorical variables (e.g. postoperative complications) were assessed with the chi-squared test or Fisher’s exact test, whereas differences in continuous variables were evaluated with the Student’s t test when the assumptions of normality were valid. A P value <0.05 was considered significant.