This study comprised two parts: a retrospective clinical investigation and a patient-specific finite element analysis (FEA).
Part 1: Retrospective Clinical Investigation
From December 2013 to December 2017, we conducted a retrospective search for patients being treated for FNFs in one orthopaedic ward. The following patients were eligible: (1) aged between 20 and 60 years old; (2) had Pauwels angle greater than 50° (i.e., vertical-oriented type, VFNF); (3) treated with hip-preserving surgeries; (4) had pre- and postoperative radiography. Pauwels angle was measured using the modified method based on preoperative antero-posterior X-rays[11]. The following patients were excluded: (1) those with severe comorbidities, including osteoarthritis, osteoporosis, cerebrovascular disease, diabetes, among others; (2) those that additionally had femoral shaft, subtrochanteric, intertrochanteric, or contralateral fractures along with a VNF; (3) those with less than 2 years of follow-up. Finally, a total of 204 patients were included in the clinical investigation, and all gave written informed consent to participate. This study was approved by the local institutional ethics review board (No. 2016 − 143), and was carried out in accordance with the World Medical Association Declaration of Helsinki.[12]
Among the 204 VFNFs, 107 patients were treated with three parallel 6.5 mm cannulated screws (Stryker) (denoted as group G-TRI); 65 patients were treated with three parallel screws augmented with an off-axis screw (denoted as group G-ALP); and 32 patients were treated with DHSs (Depuy Synthes) plus one anti-rotational screw (denoted as group G-DHS). In all of the surgeries, closed reduction was attempted first, and then the reduction quality was evaluated based on intraoperative fluoroscopy imaged in the AP and lateral planes. Acceptable reductions (displacement was < 5 mm, angulation was < 10°) were defined using Haidukewych criteria;[13] For fractures with unacceptable reduction, an open reduction was performed using the modified Smith-Peterson approach. After an acceptable reduction was achieved, an incision was made laterally according to the type of internal fixation device used. Postoperatively, patients were discouraged from engaging in any weight-bearing activities during the first 3 months of recovery. Thereafter, patients were allowed to gradually engage in partial weight-bearing activities only if their radiographs revealed acceptable bone union. Patients were scheduled for postoperative follow-up 6 weeks, 3 months, and one year after fixation surgery. Thereafter, they were evaluated once per year.
Demographic characteristics, including age at surgery, sex, and fracture characteristics, were collected as baseline information. Fracture characteristics included initial displacement, Pauwels angle, reduction method, and reduction quality. During the follow-up, fixation failure and avascular necrosis (AVN) were recorded for subsequent comparisons with clinical prognosis. Fixation failures included non-union (NU), femoral neck shortening (FNS), varus deformation (VD), and cut-out.[2] NU was confirmed when bone healing was not achieved within 6 months. FNS was defined as ≥ 10 mm shortening of femoral neck length, while VD was defined as ≥ 10° decrement of femoral neck-shaft angle.[14] AVN was evaluated radiographically by using the method of Ficat.[15]
Part 2: Subject-Specific FEA
We also conducted a patient-specific FEA based on data from 8 healthy volunteers, ranging in age from 20 to 60 years old (Appendix 1). The patient-specific FEA models were developed in Mimics software (Version 19.0, Materialise, Leuven, Belgium) and further digitally osteotomized with a Pauwels angle of 70° in 3-Matic software (Version 11.0, Materialise, Leuven, Belgium). The three internal fixation strategies we tested in these models were (1) triangle fixation (G-TRI); (2) triangle screws with an off-axis screw (G-ALP); and (3) DHS (G-DHS). These were compared in each patient-specific fracture model (Fig. 1). The digital construction of these devices was created in SolidWorks 2017 (Dassault Systèmes SolidWorks Corporation, Waltham, MA, USA) and assembled in 3-Matic software. In order to control for confounding variables of surgery quality, all fixation devices were digitally implanted based on the same standard criteria (Fig. 1). [9, 16]
All assemblies were meshed into 1mm equal sized facets and converted into 4-node linear tetrahedron (C3D4) solid elements in Hypermesh 13.0 (Altair Engineering, Troy, MI, USA). The solid models were then exported into software Abaqus 6.13 (Simulia Corp, USA) as inp format files. The property of all bone and implant models were assumed as linear elastic material. The density of cortical and cancellous bone was determined by calculating their Hu values based on computed tomography (CT) scans using the formula described previously.[17]
ρ(g/cm3) = 0.000968*HU + 0.5
If ρ < 1.2 g/cm3, E = 2014 *ρ^ 2.5 (MPa), ν = 0.2.
If ρ > 1.2 g/cm3, E = 1763 *ρ ^3.2 (MPa), ν = 0.32.
For cannulated screws, we used values for screws made of titanium (Ti-6L-4V), which has a Young’s modulus (E) of 110,000 MPa and a Poisson’s ratio of 0.3. [18, 19] For the DHS device, we used values for stainless steel, which has a Young’s modulus (E) of 193,000 MPa and Poisson’s ratio of 0.31[20]. As shown in Fig. 2, to simulate mechanical nature of these implants, thread-bone/implant interfaces were tied while others were set to slide contact. The sliding contact of the fracture surface was modeled with frictional coefficient of 0.46, and the frictional coefficient of other self-contacts was set as 0.3 [21]. All fixation models were constrained to within 80 mm distal from the lesser trochanter and subjected to 237.7% of body weight loading,[22] along the femoral mechanical axis. The dynamic compression effect of cannulated screws and lag screws was simulated by using a preloading of 224N for cannulated screws and 591N for DHSs, the same value per mm2 as we described previously[9]. The above finite element simulation process was validated using cadaveric bone in our previous biomechanical test[9], showing a relative coefficient of 0.78–0.94.
Stiffness, IFM, and implant stress (von Mises stress) were analyzed as biomechanical parameters. Note that, in order to overcome the drawback of “center point fallacy” in previous measurement,[23] the IFM of all nodes on both fracture surfaces was calculated, and then the mean IFM value of all nodes was the parameter that was compared among three groups[9]. IFM of each paired node were calculated based on the formula in previous study[9].
Statistical Analysis
Both the clinical and biomechanical statistical analyses were performed with SPSS 24 (IBM Corp. Released 2016. IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY: IBM Corp). Differences in baseline information and clinical complications were evaluated using Chi-square tests and one-way ANOVAs. Biomechanical parameters in patient-specific FEA were evaluated using randomised block one-way ANOVAs. Statistical significance was set at p < 0.05, and all tests were two-sided. Means and standard deviations (± SD) were calculated, and counts were tabulated.