Finite Element Analysis of Fracture Fixation

Abstract

Purpose of review:

Fracture fixation aims to provide stability and promote healing, but remains challenging in unstable and osteoporotic fractures with increased risk of construct failure and nonunion. The first part of this article reviews the clinical motivation behind finite element analysis of fracture fixation, its strengths and weaknesses, how models are developed and validated, and how outputs are typically interpreted. The second part reviews recent modeling studies of the femur and proximal humerus, areas with particular relevance to fragility fractures.

Recent findings:

There is some consensus in the literature around how certain modeling aspects are pragmatically formulated, including bone and implant geometries, meshing, material properties, interactions, and loads and boundary conditions. Studies most often focus on predicted implant stress, bone strain surrounding screws, or interfragmentary displacements. However most models are not rigorously validated.

Summary:

With refined modeling methods, improved validation efforts, and large-scale systematic analyses, finite element analysis is poised to advance the understanding of fracture fixation failure, enable optimization of implant designs, and improve surgical guidance.

Clinical motivation

The management of bone fractures constitutes a significant societal burden.[1] The impact on an individual patient spans a broad spectrum including pain, loss of work, and temporary disability. In severe fractures, there can be permanent disability. This burden is worsened in situations in which bone union does not occur, or a fracture fixation construct mechanically fails prior to bone union (Fig. 1). These nonunion scenarios often require additional surgery in the form of new implants, with or without biologic augmentation. Nonunion also severely prolongs patient disability[2,3] and imparts an additional significant economic burden.[4,5]

Figure 1.

Radiographs from three patient cases (three rows of images). (A) A 72 y/o man sustained an intertrochanteric hip fracture that was stabilized with an intramedullary nail. After six months, the fracture did not heal and the nail sustained a fatigue fracture where the lag screw enters the neck of the femur (arrow). The patient was unable to walk without a walker. (B) He underwent revision surgery with placement of a different implant in compression (laterally based plate) which resulted in eventual healing of the fracture. (C) A 69 y/o man suffered a fracture above his revision knee replacement following a motorcycle accident. He was initially stabilized with a lateral plate. He was unable to restrict his weight bearing and sustained bending of his plate over three months. (D) Eventually, this plate fractured (arrow), requiring (E) complex reconstructive surgery with two plates and bone shortening, which eventually resulted in union. (F) 78 y/o woman sustained an osteoporotic proximal humerus fracture, initially stabilized with a blade plate and non-locking screws. (G) Early failure of the construct resulted, with pullout of the blade (arrow) and screws. (H) Revision surgery using an anatomic specific plate with locking screws resulted in successful union.

As the population ages, the increased incidence of osteoporosis has resulted in a parallel rise in fragility fractures associated with compromised bone density and bone quality, such as in the proximal femur, proximal humerus, distal radius, and spine.[1,6] The increase in the incidence of total knee and total hip procedures[7] has also resulted in the rapid increase in fractures around these implants, i.e. periprosthetic fractures (Fig. 1C–E), as these elderly patients remain active.[8] An interprosthetic femur fracture in an elderly osteoporotic patient between a total hip and a total knee replacement constitutes a singular challenge in reconstructive orthopaedic surgery.[9]

In developed countries, the vast majority of displaced long bone fractures in adults are surgically stabilized with a metallic implant, such as plate and screws (Fig. 1H), or intramedullary nail (Fig. 1A). A fracture fixation implant is designed and applied to provide two benefits to the patient. The first is to create an appropriate mechanical/biologic condition at the fracture site for healing. The second is to provide enough stability to allow immediate functional use of the injured part and thus minimize the fracture-associated disability discussed above. If a fracture is simple, and the major fragments can be perfectly reduced surgically with compression by the implant, then a situation of “absolute stability” is created in which the bone will heal primarily without the formation of callus. Immediate load sharing by the bone ends will also shield the implant from stress. More commonly, a variable gap will exist at the fracture site following surgery. Fixation of such fractures via bridge plating, intramedullary nailing, or external fixation creates a situation of “relative stability” in the fracture gap which requires the synthesis of secondary healing tissues by the body.

With appropriate mechanical conditions, i.e. interfragmentary strain within the fracture gap on a spectrum initially described by Perren et al.[10,11], the healing progresses and the callus gradually stiffens (granulation tissue, cartilage, then bone) eventually resulting in restoration of load transmission through the bone and offloading the implant.[12,13] Moderate compressive axial strains (less than ~ 40% strain) promote healing, whereas low-to-high shear strains are believed to be inhibitory.[14–16] During the healing of a relative stability construct, especially in patients with impaired biology for endochondral and intramembranous bone formation, the implant undergoes a significant cyclic loading history and can suffer a fatigue fracture prior to bone union. Many human and animal studies have supported the importance of interfragmentary strains in fracture healing.[13–15,17] Hypertrophic nonunions show abundant callus volume without bridging bone, indicating a mechanically induced nonunion, e.g. due to a compliant fixation and overstimulated gap. Studies have also reported that atrophic nonunion sites are vascularized and contain viable stem cells[18], supporting the theory that mechanics (e.g. an overly rigid fixation) also can play a role in many of these nonunions.[10] Furthermore there is a lack of approved pharmacological agents for improving fracture healing.[19]

The fixation construct of an implant applied to a fractured bone has a complex mechanical behavior, both at the time of implantation (primary stability) and post-operatively during the healing phase (secondary stability). This complexity is a result of the bone geometry specific to the fracture location, the internal bone architecture and material properties, the specific implant configuration and characteristics, the implant/bone interaction, the post-operative loading profile, and finally the biologic response of the patient. It is precisely this complexity that makes the finite element analysis (FEA) approach so attractive when analyzing these phenomena. Once an FEA model is validated, the opportunities for investigation, design, and education are extraordinary as discussed below. When thoughtfully applied, FEA will provide an increasingly important computational framework for the development of improved implants, more personalized preoperative planning, and better surgeon education,[20–22] resulting in improved fracture care, fewer nonunions and fixation failures, and less fracture-related patient disability.

FEA overview

This paper provides a concise summary of FEA of fracture fixation biomechanics, intended for the general reader of this journal. We have attempted to use plain language and avoid overly technical aspects. Although the second part of the paper focuses on recent applications of FEA toward studying the fracture fixation mechanics of the femur and proximal humerus, it is beyond the scope of this paper to provide a systematic review of the entire research topic, including studies focused on other anatomic locations.

FEA, used widely in many fields, can be utilized for a variety of purposes in orthopaedics, which have been previously categorized as either: (1) fundamental understanding; (2) implant design; or (3) preoperative planning.[21] Additional purposes are emerging including detection of nonunion[23] and surgeon education. In FEA, displacements, strains, and stresses are computed throughout the entire bone and implant structure—a unique capability that is generally unavailable in other computational and experimental methods (Fig. 2).

Figure 2.

Overview of finite element analysis workflow (adapted from [24], with permission from Elsevier).

At the core of FEA is spatial discretization: dividing a complex 3D geometry into a mesh of very small elements connected at nodes. The elements are deformable, unlike in rigid body musculoskeletal models, allowing for predictions of stresses and strains. The force-displacement equations modeling these simpler geometries can be readily computed, then assembled for the entire structure. Most simulations of fracture fixation assume elastic behavior, i.e. a proportional relationship between stress and strain; advanced material models allow damage and plasticity. Interfaces between sliding components can be modeled with frictional contact. FEA numerically solves for the displacements of all the nodes of the model simultaneously, which leads to computed stress and strain distributions. Most FEA studies solve for static equilibrium of the structure for prescribed boundary conditions (external forces and fixed points/surfaces), neglecting dynamic, inertial effects. However, an alternative (forward dynamic) solution process can be used for dynamic problems or ones with very complex contact.

Strengths and weaknesses of FEA

Biomechanical testing with cadavers remains the “gold standard” for providing the ground truth on ex vivo primary stability assessment of fracture fixation constructs. However, this testing is expensive, requires valuable human tissue, has low throughput and is therefore not well suited for optimization of complex problems that require a systematic and efficient evaluation strategy. In turn, FEA has three main advantages over experimental studies. The first advantage is inherent in computational models: they facilitate a complete inspection of the makeup of the model, as well as the predicted results throughout the entire structure. Understanding of cause-effect mechanical relationships is facilitated, and available data is more complete. In experimental models, certain aspects like material behavior are obfuscated, and measurements such as internal strains are difficult or impossible.

A second advantage of FEA is that it can be efficient and inexpensive relative to experiments. Once a valid modeling procedure is in place, changing numerical parameters including implant or bone material properties can be done with ease. New implant designs can be tested in silico without the need to manufacture them. Parametric studies with many input variables are possible with FEA,[24,25] unlike experiments that, in best case, still rely on paired comparisons, capable of investigating a single selected aspect and involving potential confounding factors related to within-pair symmetry. The possibility to investigate different implant configurations within the same bone is one of the key advantages of FEA simulations over biomechanical testing. It should be noted though that re-meshing models following an implant change can be time consuming, simulation times can range from hours to weeks, and simulation convergence is challenging for certain problems.

A third advantage of FEA is that it can hypothetically be used clinically for patient-specific predictions and planning, although this area is not well developed in fracture fixation analyses. With these advantages, finite element simulations may effectively complement, partially or fully replace biomechanical experiments for some investigations, providing further insights and accelerating research and development. However, validated models and outcome measures should be used to provide meaningful and clinically relevant results.

Limitations of FEA primarily arise from the deviation of models from clinical reality, as discussed further below. FEA is completely virtual and thus introduces simplification in all aspects. Conversely, cadaveric experimental models introduce simplifications in only certain aspects such as fracture geometry, healing status, loads and boundary conditions. The greater abstraction of FEA also makes it more susceptible to misunderstanding and errors in implementation and interpretation.

Aspects of a model

Geometry:

Bone and fracture geometries, like other model properties, can be either generic or subject-specific.[26] 3D bone geometries can be accurately derived from computed tomography (CT) scans using image segmentation software (Figs. 2A,B, 3). Cancellous bone is typically modeled as a continuum, although micro-FE models of cadavers with explicit representation of trabeculae have been derived from micro-CT scans.[27] FEA results are sensitive to some aspects of bone geometry, e.g. locations where loads and boundary conditions are applied, and canal geometry in the case of intramedullary nailing. Fracture geometries are almost always modeled by virtual osteotomies (Fig. 2C), although realistic patterns derived from CT are possible (Fig. 3). Comminuted (highly fragmented) fractures are typically modeled by removing the comminuted zone, with the rationale that the zone has negligible load transfer. FEA models of clinical fracture fixation, like their benchtop counterparts, usually consider only the initial phase following surgery without any primary or secondary bone healing present. Precise implant geometries enable more accurate determination of stresses for the particular implant (Figs. 2D,H); but generic implant geometries can be useful when the study seeks generalizable knowledge. Screw shafts are often modeled as simple cylinders,[28] fixed to bone (Figs. 2D, 3).

Figure 3.

Example of finite element analysis comparing two fixation options for a challenging periprosthetic femur fracture.

Mesh:

Investigators select among element types and sizes with consideration of implant and bone geometric complexity, and desired balance between solution time and accuracy (Fig. 2E). Sophisticated modern software is used to generate finite element meshes in a semi-automated manner, with consideration of element quality parameters (e.g. aspect ratio). Tetrahedral elements are often used because of geometric complexity. Because solution accuracy generally improves with a denser mesh and/or element complexity (e.g. quadratic vs. linear), especially at critical locations, mesh convergence studies should be performed when deciding an appropriate mesh density. With improvements in computational efficiency, the number of elements in typical orthopaedic FEA models has grown to hundreds of thousands, or up to several millions in microFE simulations.

Materials:

Implant materials are engineering materials and can be adequately modeled with homogeneous, isotropic, linear elastic properties. Bone material behavior, however, is much more complex. Although bone is viscoelastic, it is almost always considered elastic, which is probably adequate because of relatively low physiologic loading rates involved.[29] Most models assume linear elasticity and do not attempt to simulate post-failure behavior, with some notable exceptions.[30] Heterogeneity in material properties is sometimes considered; subject-specific models can map local material properties to elements based on quantitative CT scans according to relationships between Hounsfield units, density, and elastic modulus (Fig. 2F).[31–33] Inclusion of bone’s heterogeneity is important for accurate simulations for some problems.[34] In generic homogeneous models, osteoporotic bone is simulated with a decreased elastic modulus. Cortical thinning and periosteal apposition has also been incorporated into geometry.[35] Bone’s direction-dependency in properties is sometimes considered, but the assumption of isotropy is most common, with a Poisson’s ratio of ~0.3.

Interactions:

The interactions between surfaces of plates or nails, screws, and bone can be complex and require some simplifying assumptions in FEA models. Connections between bone and simplified threadless screw shafts are most often modeled as fully bonded, with the rationale that there are negligible motions at these interfaces during sub-failure conditions. However, this assumption does affect the local stress environment, and more accurate connections have been proposed.[36] Locking screw threaded connections with plates are also usually modeled as fully bonded, although more accurate compliant connections have been proposed.[37] Traditional (nonlocking) screws are more complex to model; their surgical insertion process develops tension within the screw, higher peri-screw bone strains, compression between the plate and bone, and (with dynamic compression plating) interfragmentary compression. Extramedullary plates and intramedullary nails can separate or slide with bone during mechanical loading, and these interactions can be modeled with Coulomb frictional contact.[38] The presence of frictional contact between bone fragments at the fracture site results in load transfer, which greatly offloads implants. This contact can be intermittent upon loading. There is uncertainty in the above frictional coefficients in physiologic situations, although this may not affect model accuracy unless sliding is significant. Modeling contact necessitates nonlinear, incremental finite element solution procedures which sometimes can encounter convergence challenges. (Geometric nonlinearity, associated with sizeable displacements and rotations, is also inherent in many of these problems.)

Loads and boundary conditions:

Loads applied to fracture fixation constructs in vivo are cyclic and variable in nature. FEA typically focuses on a quasi-static analysis of a peak loading condition during one cycle; the investigator considers the cyclic effects in subsequent analysis outside of FEA. External loads applied to bones are primarily due to joint and muscle forces (Fig. 2G). Fortunately, in vivo studies using instrumented joint replacements have provided reliable data on joint loads.[39] Greater uncertainty exists in muscle forces, which are themselves derived from modeling estimates.[40] Depending on postoperative rehabilitation plans, assist devices, and patient compliance, loads may be reduced in the immediate postoperative period. Boundary conditions entail constraining model degrees of freedom to prevent rigid body motions; a simple example is fixing surface nodes on the bone end that is opposite to where the external forces are applied. But in vivo physiological constraint conditions are not straightforward, and in some problems boundary conditions must be carefully considered as they can have a substantial effect on results.[26,37,41,42]

Basic outcome measures:

The typical FEA solution consists of displacements throughout the structure, and strains and stresses derived from these displacements (Fig. 2H). Interfragmentary displacements at the fracture site influence the course of healing;[14,43] relative displacements between fragment ends can be resolved into axial and shear components and normalized by the fracture gap size to estimate interfragmentary strains. Displacement at the point of load application can be used to determine an overall construct stiffness. Although stiffness does not have clear clinical relevance, it can be measured reliably in experiments for validation purposes. In implants and screws, peak stresses are examined (Fig. 3), as they relate to the potential for static yield or cyclic fatigue failure. Prediction of bone failure more often uses a strain criteria than a stress criteria.[26,44,45] Sub-failure bone stress and strain may also have relevance for potential stress shielding and bone resorption.[46,47]

Advanced predictions

Implant and screw fatigue:

Normal fracture healing with bone bridging results in substantial offloading of fracture fixation implants and screws over time. However, in cases of biologically impaired healing, especially comminuted fractures of the lower limb with minimal cortical contact at the fracture site, implants (including screws) bear substantial cyclic tensile stresses that do not decrease over time. These scenarios carry additional risk of fatigue failure, characterized by a sudden brittle-like fracture of the implant.[48–50] Fatigue failure initiates at stress concentrations such as screw holes.[51] Basic methods to assess fatigue life of engineering materials can use FEA-predicted peak stress, which is the most often used approach in FEA studies of fracture fixation. More accurate predictions involve consideration of the precise implant material properties and surface finish, intraoperative factors like plate prebending,[52] non-reversed and multiaxial nature of physiologic loading, and change in loading over time. In regulatory assessment of medical devices, because of uncertainty in FEA predictions of fatigue of implants, FEA is often used to determine worst case design configurations which are then tested with cyclic loading experiments. Other specific types of hardware failure under cyclic loading, such as failure of variable angle threaded connections between plates and screws,[53] would require alternative prediction approaches.

Screw-bone interface and its failure:

In many constructs prior to healing, forces are transmitted between bones and implants primarily through the bone-screw interface, stability of which is therefore of high importance. FEA is ideal for providing insights into these details that are not assessable experimentally. The mechanical environment in the cortical bone region around screws was investigated with FEA to understand the influence of modeling choices[54,55], quantify the induced strains[56], evaluate the effects of fixation concept[35] and pilot hole size[57], and to explain peri-screw bone resorption[46] and screw loosening.[58] Most FEA studies on screw fixation in trabecular bone focused on optimizing primary stability in the challenging osteoporotic condition.[59,60] Using continuum FEA models, initial stiffness[61], peri-screw bone strain[62] and ultimate force[63] were found to predict failure. MicroFE models, including the trabecular microarchitecture and screw threads, are generally better suited for such analyses,[64,65] although they are computationally expensive and thus challenging for whole bone-implant constructs. Most microFE studies focused on stress distributions[66] and fixation stiffness[67,68] that can be accurately predicted when considering the damage caused during insertion.[69] Pull-out was the most frequently investigated failure mode; nevertheless, the failure process was usually not simulated as surrogate measures from linear elastic analyses showed good predictions.[70,71] Simulating more complex failure processes such as perforation requires the inclusion of post-elastic material properties and contact, which can improve prediction accuracy over linear analyses.[30]

Fracture healing effects:

Despite the importance of fracture healing in mitigating risk of fatigue failure, few FEA models of clinical fixation constructs have included consideration of it, instead modeling a worst-case stress scenario prior to healing. Inclusion of a 3D fracture callus having finite elements of increasing stiffness was used recently to show the importance of healing in preventing fatigue failure of high tibial osteotomy plates.[72] Simpler spring elements at the fracture gap have also been used.[73] FEA has also been investigated as a potential tool to assess the quality of ongoing fracture healing.[23,74–76]

Verification & validation

Verification, validation, and other activities related to model credibility, as described in depth elsewhere,[77–80] have fundamental importance in orthopaedic computational models with potential to impact patient care and safety. These activities should be guided in part by the model’s purpose, including context of use and associated risks.[80] Verification includes e.g. analysis of simplified benchmark problems of known solution, and the aforementioned mesh convergence testing. Sensitivity studies and uncertainty quantification are appropriate especially for model inputs with substantial uncertainty and large impact on the outcomes, such as bone modulus and load direction.

Validation requires a careful selection of a comparator. For FEA models of clinical fracture fixation, cadaver experiments are considered gold standard; although their loads/boundary conditions are a simplified abstraction of clinical reality and involve some undesirable mechanical compliance, and certain measurements cannot be made or are noisy. Outcome measurements that can be compared to FEA results include interfragmentary displacements, overall construct stiffness,[25] bone or implant[81] surface strains, and construct failure under cyclic loading.[62] Experiments with synthetic bones can be appropriate for investigations focused on implant mechanics and interfragmentary displacements, when screw-bone interface failure is not of primary consideration.[25,81,82] Scientific rigor and statistical approach (if any) vary substantially among studies that include a benchtop validation component.[83]

Comparison of FEA predictions to clinical data, which some consider the ultimate validation, is a developing area with few examples in the literature. Potential clinical data include presence of construct failure, healing status, and data from instrumented implants.[84] Construct and interfragmentary displacements have been recently measured using weightbearing CT[85] and radiostereometric analysis.[86,87] Models comparing treatment options can be compared to statistical differences observed clinically between study groups from a cohort study or from large-scale registry data. Or, patient-specific FEA models can be compared to clinical data at the individual subject level, such as callus formation viewed on radiographs.[14]

Applications in femur fracture fixation

Rates of nonunion/failure of femur fracture fixation have been reported as exceeding 10%.[88] Higher rates have been observed for complex femur fractures, including periprosthetic femur fractures (Fig. 3).[89,90] Nonunion and revision surgery for femur fractures imposes substantial burden on the patient (Fig. 1).[3] The femur is subject to loads of several times bodyweight during normal ambulation.[39] Due to the eccentricity of load application through the femoral head, load-induced stresses are further exacerbated through bending, especially for lateral extramedullary implants. Most FEA studies of femur fracture fixation focus on unstable extra-articular fractures.

Twenty-four recent, relevant publications that used FEA analyses to investigate the biomechanical competence of femur fracture fixations were identified in our non-systematic review (Supplementary Table 1). [14,38,47,73,81,82,91–108] In the following, the number of studies associated with each characteristic is indicated in parentheses. Studies mostly focused on fractures of the femoral neck (6), intertrochanteric region (7), and distal femur (8). Computer simulations were applied to analyze three main scenarios: I) comparison of different implants (8); II) the optimal use of existing plates, nails, and screws (11); and III) options to improve existing implants by optimizing their design (4). The most often examined implant type was intramedullary nail (13), followed by locking plate (7), cannulated screws (5), sliding hip screws (4), the combination of intramedullary nail and locking plate (1) and the combination of screws and medial buttress plate (2). Subject-specific bone geometry was used in 13 studies; however, nine of these used only a single patient. Continuum modelling of bone was used in all studies. Bone material properties in models were mostly separated homogeneous compartments for cortical and trabecular bone (14), or using CT-bone mineral density (BMD) based heterogeneous properties (8). The assumption of bone isotropy was used in all papers but one, which used a transverse isotropic model for cortical bone. Only a single study considered bone’s post-elastic material behavior to simulate failure.[98] The bone-screw interface was modelled as bonded in 12 studies and with contact in four studies, with five studies not describing this interaction and others not including screws. Loading and boundary conditions of the fixation constructs included hip joint force with additional muscle forces (6), or without muscles (16). Multiple loading cases were included in two studies. All models were solved with an implicit FEA approach. Systematic parametric analysis, with hundreds of models, was performed in two studies.[38,105] Main outcome measures in the above studies included fracture gap motion (10), implant stress (18), and overall construct displacement/stiffness (11).

Intertrochanteric femur fractures, a prevalent fracture in elderly patients, are typically fixed with intramedullary nailing. Eberle et al.[81] reported that for cephalomedullary nail fixation, subtrochanteric fractures resulted in the highest nail stresses, followed by pertrochanteric, then followed by lateral neck fractures. Sensitivity analyses showed dependence of stresses on angle of insertion of the nail in the frontal plane. The recent results of Tucker et al.[38] were in general agreement with the findings of Eberle et al. regarding effect of fracture locations. Tucker et al. reported that larger, closer fitting nail diameters reduced axial and shear interfragmentary motions. They also reported on reduced stability associated with unreduced/comminuted fractures. Ali et al.[92] reported that for nail fixation of more realistic (although simulation-derived) intertrochanteric fracture geometries, load transfer and nail stress were substantially influenced by inter-subject variability in fracture geometry. Goffin et al.[96,109] reported studies focused on risk of screw cut-out through the femoral head, as determined by local bone strains. In one subject-specific model fixed with a sliding hip screw, they reported lower strains with inferior positions of the lag screw.[109] They also reported reduced strains associated with bone compaction around a helical blade in simulated osteoporotic bone.[96] Recent studies also investigated alternative designs and materials for proximal femur nails. Wang et al.[106] simulated the effects of lower modulus nails. Although not specifically demonstrated in the study, reduced modulus implants may have some advantages with respect to cut-out in poor quality bone. Li et al.[100] examined the mechanics of a modified nail design with an additional feature intended to better stabilize the medial cortex in unstable fractures.

Femoral neck fractures have a variety of fixation options including cannulated screws, sliding/dynamic hip screw, and locking plates. Related FEA studies generally have more application to younger patients; in elderly patients, displaced femoral neck fractures are treated often by hemiarthroplasty. Most recent FEA studies of these constructs focused on implant stresses and fragment displacements under loading. A combination of sliding hip screw with additional cannulated screws was more effective than sliding hip screw alone to prevent shear rotation and separation of fragments and reduce the stress in the implants.[94,104] Li et al. reported that applying a medial buttress plate improved stability and reduced shear rotation of the proximal fragment.[99]

Distal femur fractures can be repaired with either plates or nails. For plate fixation, studies have demonstrated the importance of the plate material and working length (distance across the fracture between the inner-most screws) in controlling the axial and shear interfragmentary motions important for healing.[14,47,73] Recent studies have also investigated innovative design modifications of the plate and locking screws, intended to increase axial interfragmentary motions.[82,110]

Applications in proximal humerus fracture fixation

Another clinically challenging topic is the treatment of osteoporotic proximal humerus fractures. Implant fixations enable early functional recovery; however, their benefit versus conservative treatment has been debated.[111–113] Despite the development of the state-of-the-art locking plate technology, the rate of mechanical failures has remained high, up to 36%[114], with screw cut-out and perforation being the most frequent implant-related complications.[115] These outcomes may be improved by enhancing implant designs and clinical guidelines for applying these fixations. Implant fixation of proximal humerus fractures is a complex and multifactorial problem, with age, bone mineral density, fracture complexity and reduction quality being prominent risk factors.[116] A further challenge for fixation strategies is the highly heterogeneous bone stock distribution within the humeral head.[117–119] FEA may help to accelerate research and development towards improved implants and treatment strategies.[120,121]

Thirty relevant publications that used FEA to investigate the biomechanical competence of proximal humerus fracture fixations were identified in our non-systematic review (Supplementary Table 2).[24,27,30,36,62,65,67,68,122–143] In the following, the number of studies associated with each characteristic is indicated in parentheses. Computer simulations were applied to analyze three main scenarios: I) comparison of different implants (9); II) the optimal use of existing locking plates (13); and III) options to improve existing implants by optimizing their design (3). Others focused primarily on the effect of bone stock quality (2), modelling approaches (2), or experimental validation (1). The most often examined implant type was locking plate (25), followed by single screw (5), intramedullary nail (2) and K-wires (1). Subject-specific bone geometry was used in 24 studies; however, six of these used only a single patient. Continuum FE (cFE) modelling of bone was used in most studies (25); the microarchitecture of trabecular bone was considered in microFE simulations investigating screws (4) only and in a single study on construct stability. Bone material properties in cFE models were mostly simplified as being homogeneous (4), separated for cortical and trabecular bone (10), or CT BMD based heterogeneous (12). Considering heterogeneity was found to be influential at fracture fixation simulations of other anatomical locations.[34] All microFE models used homogeneous properties on the tissue level; inclusion of heterogeneity was shown not to be influential.[144] The assumption of isotropy was used in all papers; however, meso-scale anisotropy is inherently included in microFE models. Only a single study considered bone’s post-elastic material behavior to simulate failure.[30] The bone-implant interface was modelled primarily as bonded in most cFE studies (26), contact or a pseudo-threaded approach was used only in three papers. MicroFE simulations used tied interfaces, except for one study applying a soft per-implant bone region simulating damage during screw insertion[69] and another non-linear approach utilizing contact conditions.[30] Loading and boundary conditions of the fixation constructs mimicked physiological activities (11) or simplified conditions from experimental setups (19). Multiple loading cases were included in 13 studies. All cFE models were solved with implicit approach. MicroFE simulations utilized special parallel solvers to efficiently handle large model sizes and one study used an explicit technique to solve the nonlinear problem.[30] Systematic parametric analysis and optimization, with up to thousands of models, was performed in nine studies, requiring fully automated simulation workflows.[24]

Most studies used construct stiffness and/or axial fracture gap motion as the main outcome measure (14), generally aiming to seek maximum rigidity. Less than half of those studies (6) validated stiffness experimentally or used a different measure for validation such as bone surface strain.[125] While the relevance of rigidity has been shown for single screws, the clinical relevance of stiffer constructs has not been clearly demonstrated, especially considering that overly stiff fixations can be inhibitory for secondary healing.[145] Some studies aimed to minimize shear-type gap motion (3), which is believed to be detrimental for healing.[14–17] Interfragmentary contact pressure in well-reduced fractures was used in a single paper as measure of stability.[132] Another frequently used metric was implant stress, being subject to minimization, aiming to avoid implant damage (12) but without validation. Nevertheless, failure of the metal components is not the most often occurring complication in the proximal humerus fixations.[115] Several studies based their conclusions on one of these metrics or their combinations. Stability of single screw fixations in pull-out and the effectiveness of cement augmentation were shown to be strongly determined by the volume fraction of the peri-implant bone region.[67,68] Construct stability was improved, and implant stress reduced by implant selection[124], combining lateral and medial plating[129,130], placing calcar screws[142], establishing medial cortical contact[136], fracture reduction with impaction[123], as well as via augmentation of locking plates with cement at the screw tips or with additional intramedullary strut increase stability.[122] Design features of plates were shown to be influential. Plate holes that allow either a locking or nonlocking screw placement (combi holes) were associated with higher stress concentration and higher shear motion in the gap compared to locking holes.[143] Construct stiffness was increased by more proximal orientation of the calcar screws[131] and longer threading of the screw.[134]

Another frequently used outcome measure was the stress or strain in the bone region in the direct or indirect vicinity of the screws. Stress of screw-bone interface was found to be better distributed in double plating versus a single lateral locking plate[133] and reduced by the addition of calcar screw, achieving medial cortical contact[142], augmenting the screw tips[132] or the humeral head with cement[125], changing the angles of medical screws[131], and selecting more elastic implant materials.[125] Stress within the bone cannot be measured directly and therefore none of these findings were validated. Several studies investigated bone strain. Deformations were shown to be highly affected by bone quality and loading mode.[135] Detailed analyses indicated that the accurate strain distribution within the peri-screw bone region cannot be reproduced when using homogeneous properties[65] or bonded bone-screw interfaces[36] in cFE models. Nevertheless, the assumption concerning the interface was shown not to affect the statistical findings between different implant configurations when the average compressive principal strain is evaluated in the bone regions around the locking screws within the humeral head.[36] Moreover, the latter measure was demonstrated to be an experimentally validated surrogate for cyclic screw cut-out failure[62] that is one of the most frequent failure modes of these fixations.[115] Using this strain measure, it was shown that the predicted cut-out risk could be reduced by using longer screws[126], appropriately selecting the screw number and configuration in comminuted and non-well reduced fractures,[36,139] especially by maximizing screw spread.[128] Furthermore, cement augmentation of the screw tips was demonstrated to be beneficial[24] especially when prioritizing the calcar screws.[141] Using locking plates that sit more proximally or even proximalizing the implant indicated lower cut-out failure risk.[127] Parametric optimization found improvements with more proximally oriented screw trajectories within the humeral head.[138] Novel plate designs including a medial strut decreased the predicted cut-out risk.[140] Only a single computational study investigated screw perforation failure, showing that non-linear microFE models provide accurate predictions.[30]

Conclusion and future directions

FEA is by far the most widely used computational tool for investigating internal mechanics associated with orthopaedic surgery, including fracture fixation. Sophisticated FEA software has been refined and verified over many years across a wide variety of industries and fields. With improved approaches for applying FEA to orthopaedics, and increased computational power, there is growing acceptance of the use of FEA for development and governmental regulation[146] of implants, scientific investigation, and surgical decision making. In silico trials hold potential for systematic, efficient evaluation of fracture fixation methods in a scientifically rigorous manner that considers ‘biologic’ variability.[79] However, FEA is particularly prone to nuanced method variations, and even errors, that can provide believable but misleading results. With model validation being absent in most studies, or performed only in ex vivo settings in others, the results must always be interpreted with care and require clinical corroboration. Moving forward, established best practices for biomechanical model verification, validation, and credibility can be adapted to the fracture fixation field. Validation efforts should be based on the clinical question and corresponding model outcome(s). Carefully designed benchtop experiments with reliable outcome measurements can be used to validate FEA-predicted mechanics, and innovative approaches will bring us closer to true clinical validation. FEA studies hold great potential to advance the understanding of fracture fixation failure, help to improve guidelines, contribute to surgical education, and support the development of new, optimized implant designs via systematic analyses.