- Email: [email protected]
Contact mechanics of reverse engineered distal humeral hemiarthroplasty implants
Journal of Biomechanics 48 (2015) 4037–4042
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Contact mechanics of reverse engineered distal humeral hemiarthroplasty implants Ryan Willing a,b,n, Graham J.W. King b, James A. Johnson b a Department of Mechanical Engineering, Thomas J. Watson School of Engineering and Applied Science, State University of New York at Binghamton, Binghamton, NY, USA b Bioengineering Research Laboratory, The Hand and Upper Limb Centre, St. Joseph Health Care London, London, Ontario, Canada
art ic l e i nf o
a b s t r a c t
Article history: Accepted 27 September 2015
Erosion of articular cartilage is a concern following distal humeral hemiarthroplasty, because native cartilage surfaces are placed in contact with stiff metallic implant components, which causes decreases in contact area and increases in contact stresses. Recently, reverse engineered implants have been proposed which are intended to promote more natural contact mechanics by reproducing the native bone or cartilage shape. In this study, finite element modeling is used in order to calculate changes in cartilage contact areas and stresses following distal humeral hemiarthroplasty with commercially available and reverse engineered implant designs. At the ulna, decreases in contact area were 3473% (p¼ 0.002), 2771% (po0.001) and 1472% (p¼ 0.008) using commercially available, bone reverse engineered and cartilage reverse engineered designs, respectively. Peak contact stresses increased by 461757% (p¼0.008), 3877127% (p¼0.229) and 165716% (p¼ 0.003). At the radius, decreases in contact area were 2173% (p¼0.013), 1372% (po0.006) and 671% (p¼ 0.020), and peak contact stresses increased by 75752% (p40.999), 241732% (p¼ 0.010) and 61710% (p¼ 0.021). Between the three different implant designs, the cartilage reverse engineered design yielded the largest contact areas and lowest contact stresses, but was still unable to reproduce the contact mechanics of the native joint. These findings align with a growing body of evidence indicating that although reverse engineered hemiarthroplasty implants can provide small improvements in contact mechanics when compared with commercially available designs, further optimization of shape and material properties is required in order reproduce native joint contact mechanics. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Distal humeral hemiarthroplasty Reverse engineered implants Finite element analysis Contact mechanics Cartilage stress Elbow Orthopedic surgery
1. Introduction Distal humerus fractures occur with an incidence rate of 5.7/100,000 per year (Anglen, 2005; Robinson et al., 2003). Distal humeral hemiarthroplasty (DHH) was first described in 1947 (Mellen and Phalen, 1947), and is chosen when open reduction and fixation (ORIF) may have a low likelihood of success, but the patient is not a good candidate for total elbow replacement (TER) due to their higher functional demands (Argintar et al., 2012). Only recently have implants designed specifically for DHH become commercially available worldwide; however none are currently approved for use in the USA. Cartilage erosion following DHH has been reported in several recent studies (Adolfsson and Nestorson, 2012; Burkhart et al., 2011; Hohman et al., 2014; Phadnis et al., 2015; Smith and Hughes, 2013), particularly at the proximal ulna, and likely occurs as a result of replacing the native cartilaginous surfaces with a n Correspondence to: Department of Mechanical Engineering, Thomas J. Watson School of Engineering and Applied Science, Binghamton University – SUNY, P.O. Box 6000, Binghamton, NY 13902-6000, USA. Tel.: þ1 607 777 5038. E-mail address: [email protected] (R. Willing).
http://dx.doi.org/10.1016/j.jbiomech.2015.09.047 0021-9290/& 2015 Elsevier Ltd. All rights reserved.
comparatively rigid metal prosthesis. In a recent study (Smith and Hughes, 2013), moderate to severe cartilage wear at the proximal ulna was observed in 6/16 patients and correlated with worse results in some outcome measures. None of these studies reported DHH failures due to cartilage erosion alone, and insufficient data exists to predict long-term effects of wear progression. Previous in-vitro studies have demonstrated that commercially available (Lapner et al., 2014) and custom (Willing et al., 2014a) DHH implants fail to provide the same contact areas as native articulations. Neither study provided direct measurements of the cartilage stress. While it is generally accepted that contact areas and average contact stresses are inversely proportional, this is not necessarily true for peak contact stresses, necessitating direct measurements or computational analyses. The custom implants reproduced the shape of the osseous anatomy of the distal humerus (therefore called “reverse engineered” implants); this resulted in a systematic undersizing of these implants with respect to the actual cartilage surface, causing a loss of joint congruency (Willing et al., 2014a). The purpose of this study was to compare the contact patterns and cartilage stresses of elbows before and after DHH with commercially
4038
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
available, bone reverse engineered, and cartilage reverse engineered implants. We hypothesized that custom DHH implants reverse engineered from the cartilage geometry of the distal humerus would provide contact patterns and stresses in closer agreement to those of the native joint than the other designs.
Pewaukee, WI, USA) at 120 kV and 200 mA. Bones were oriented parallel with the axial direction of the CT scanner, and the articular cartilage surfaces were exposed to air during the CT scan. This protocol allows us to more clearly delineate the cartilage surface by leveraging a strong contrast against air, and allows cartilage thickness measurements with a mean error of only 0.3 mm compared to direct measurements (Lalone et al., 2015). Scans were performed with in-plane pixel sizes ranging between 0.283–0.391 mm and a 0.625 mm slice thickness. Image data was saved in the Digital Imaging and Communications in Medicine (DICOM) file format.
2. Methods 2.2. Model segmentation 2.1. Specimen preparation and imaging Five unpaired fresh-frozen cadaveric elbow specimens (left arms, three male and two female, mean age of 57, range 21–77) were used for this study. Institutional review board (IRB) approval was not required for this cadaver study. Specimens were disarticulated and denuded of all soft tissue such that the distal humerus, proximal ulna and proximal radius were separated and the articular cartilage was exposed. Each joint was examined and confirmed to be free of any signs of cartilage erosion or osteophyte formation. Exposed cartilage surfaces were loosely wrapped in normal saline-soaked towels to prevent desiccation. Within 24 hours, all specimens underwent computed tomographic (CT) imaging using a GE Discovery CT750 HD scanner (GE Health care,
CT scan DICOM files were imported into Mimics (Materialise, Leuven, Belgium) (Fig. 1a), where a threshold-based segmentation technique was used to label any voxels with a radiodensity greater than or equal to 250 HU as bone, and a threedimensional model was created using a marching cubes algorithm (Lalone et al., 2015; Willing et al., 2014a, 2014b, 2013). The threshold-based technique was used again to create bone plus cartilage models from the CT images by labeling any voxels with a radiodensity greater than or equal to 700 HU (Lalone et al., 2015; Lapner et al., 2014; Willing et al., 2014a, 2014b). All three-dimensional models were wrapped (minimum detail of 0.25 mm, 2.00 mm gap closing) using a built-in Mimics function, which filled small holes present in the bone models, primarily in non-articular regions, and did not affect cartilage thickness distributions. The resulting shell models, composed of triangles sharing edges and vertices representing either the bone or boneþ cartilage exterior shapes, were saved in the standard tessellation language (STL) file format. These models were reconstructed using a radial basis function, to smoothen the geometry and improve triangle shape and size uniformity (Fig. 1b). 2.3. Mesh generation Finite element (FE) mesh models for predicting elbow joint contact mechanics were prepared using a technique which was previously validated through comparison of model-predicted and experiment-measured contact pattern measurements made using casting techniques (Willing et al., 2013). NETGEN (RWTH Aachen University, Germany and Johannes Kepler University Linz, Austria) was used to mesh the boneþ cartilage STL surfaces with four-sided quadrilateral-shaped shell elements, with typical edge lengths ranging between 0.30 and 0.45 mm (Fig. 1c). These elements provided a template to eventually create six-sided hexahedral elements, which are generally preferred over tetrahedrons, to represent cartilage in the FE models. To reduce the solve time during FE analysis, any quad elements
Fig. 1. Bone (left) and cartilageþ bone (right) meshes were created from CT data. (a) CT data was segmented in MIMICS and (b) exported as STL models. (c) STL models were meshed using NETGEN to generate tetrahedral (tet) or quadrilateral (quad) meshes of the bone or cartilage, respectively. (d) The cartilage mesh was cropped to articular regions only. (e) Hexahedral (hex) elements representing cartilage were created based on the quadrilateral mesh, and assembled with the tetrahedral bone mesh.
Fig. 2. Loads and boundary conditions applied to FE model. Muscles forces (FBIC, FBRA and FTRI) acted at the corresponding insertion points of the biceps, brachialis and triceps (respectively) on the ulna and radius, and were oriented parallel with the long axis of the humerus. Lateral collateral and medial collateral ligament forces (FLCL and FMCL) were oriented towards the anatomic flexion/extension (F/E) axis of the elbow.
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
4039
Fig. 3. The four contact configurations simulated by the FE model. (a) The native distal humerus was represented by a deformable cartilage layer. (b–d) DHH with a commercially available, bone reverse engineered, and cartilage reverse engineered implant, referred to as the COM, BRE and CRE designs, respectively. DHH implants were modeled as rigid surfaces. which were obviously not on the articular surface were manually deleted, except in the coronoid and olecranon fossa, where potential contact must be permitted (Fig. 1d). The remaining elements were swept to the corresponding bone STL shell models, and three layers of linear hexahedral elements were generated in between. The resulting hexahedral mesh represented the nonuniform thickness articular cartilage layer in the FE models. The bone STL shell models were meshed using linear tetrahedral elements with 2 mm typical edge lengths using NETGEN. 2.4. FE model The bone tetrahedral and cartilage hexahedral meshes were imported into Abaqus (Dassault Systèmes Simulia Corp., Providence, RI, USA). The distal humerus, proximal ulna and proximal radius of each specimen were assembled with the elbow situated between 60–90° of flexion and with neutral forearm rotation. Cartilage elements were assigned Neo-Hookean hyperelastic material properties with a bulk modulus (K) of 0.31 MPa and a shear modulus (G) of 0.37 MPa (Brown et al., 2009; Schenck et al., 1994), which approximates the equilibrium response of cartilage to compressive loading (Willing et al., 2013). Bone was assumed rigid; the effect of this simplification has been shown to be minimal in previous work (Willing et al., 2013). The interface between hexahedral cartilage and tetrahedral subchondral bone elements was simulated using a tie constraint (bonded contact, no sliding). The radius and ulna were modeled as a single rigid body with separate deformable contact surfaces at the proximal articulations, effectively simulating a forearm pinned in neutral rotation, consistent with previous invitro studies (Lapner et al., 2014; Willing et al., 2014a, 2014b). Contact between cartilage surfaces was simulated using surface-to-surface frictionless contact with a nonlinear contact stiffness Penalty algorithm (Willing et al., 2013). The insertions of the triceps, biceps and brachialis tendons, and the medial and lateral collateral ligaments, were identified on the ulna and radius by an orthopedic surgeon. Constant muscle forces of 40 N, 20 N and 20 N were applied at the approximate center of each insertion of the triceps, biceps and brachialis tendons (respectively) and were oriented parallel to the long axis of the humerus (Lapner et al., 2014; Morrey et al., 1991; Willing et al., 2014a, 2014b). Constant ligament forces of 20 N each were applied at their respective insertions, and were oriented towards the corresponding origins on the distal humerus (Fraser et al., 2008; Lapner et al., 2014; Pichora et al., 2007; Willing et al., 2014a, 2014b). These origins were defined as the locations where the elbow anatomical flexion/ extension (F/E) axis (defined by a line passing through the centers of a sphere fit to the capitellum and a circle fit to the trochlea (London, 1981; Shiba et al., 1988)) passed through the corresponding medial or lateral cortex of the distal humerus (Fig. 2). Models were created for each specimen in order to simulate four different configurations (Fig. 3). 2.4.1. Native A native distal humerus was simulated by including the native distal humerus cartilage, modelled using deformable hexahedral elements. 2.4.2. Commercially available DHH (COM) Hemiarthroplasty with a commercially available DHH prosthesis was simulated. The surface geometries of Latitude Anatomic DHH prostheses (Tornier, Texas, USA) were represented by rigid surface meshes of triangle shaped elements. Optimal prosthesis size for each specimen was chosen based on matching the medial–lateral distance between the trochlea and capitellum centers (Lapner et al., 2014). 2.4.3. Bone reverse engineered DHH (BRE) A custom DHH prosthesis based on the distal humerus bone shape for each specimen (Willing et al., 2014a) was simulated by eliminating the cartilage layer and allowing contact between the rigid surface of the implant representing the distal humerus bone and the cartilage surfaces of the proximal radius and ulna. 2.4.4. Cartilage reverse engineered DHH (CRE) A custom DHH prosthesis based on the distal humerus cartilage shape was simulated for each specimen by modeling the distal humerus cartilage mesh as a rigid surface.
Fig. 4. Typical contact stress results for a single specimen. Contours describe the contact stresses across the contact surfaces of the ulna and radius at 15, 60 and 105° of flexion, when articulating with the native distal humerus (Native), the commercially available DHH design (COM), the bone reverse engineered DHH design (BRE), and the cartilage reverse engineered DHH design (CRE). Red fringe plot values denote contact stresses at or above 2 MPa, this upper limit chosen in order to permit better visualization of the entire contact region. Flexion was simulated by rotating the forearm about the F/E axis of the (fully constrained) humerus (15–105° in 5° increments; Fig. 2). All remaining degrees of freedom, including forearm translations, varus–valgus and internal–external rotations, remained unconstrained and were able to change as a result of the applied loads and contact. Contact areas and stresses at the proximal ulna and radius were calculated. Contact areas were reported as a percentage of the total cartilage surface area, normalizing for differences in specimen sizes. The effect of joint condition on contact area and peak contact stress was examined using a repeated measures analysis of variance (ANOVA) at the ulna and radius. Statistically significant differences were reported if po0.05 after a Bonferroni adjustment for multiple comparisons.
3. Results Results for a single specimen at 15, 60 and 105° of flexion for each contact configuration are shown in Fig. 4, and are representative of the typical stress distributions at extension, midflexion, and full flexion. Contact areas include any region where contact stresses are non-zero. 3.1. Contact area Fig. 5 shows the average contact area (expressed as a percentage of the entire contact area) as a function of flexion angle for each joint condition. Contact area was significantly reduced at both the ulna ( 3473%, p¼ 0.002) and radius ( 2173%, p¼0.013) when the COM design was used. When the BRE design was used, significant decreases were still observed at the ulna and radius ( 2771%,
4040
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
Fig. 5. Contact area for each joint condition. (a) Mean ( 71 SD) of the ulna contact areas, expressed as a percentage of the entire ulna contact surface, versus flexion angle. (b) Mean ( 71 SD) of the radius contact areas, expressed as a percentage of the entire radius contact surface, as a function of flexion angle.
Fig. 6. Peak contact stresses for each joint condition. (a) Mean ( 7 1 SD) of the peak contact stress within the contact patch of the ulna at each flexion angle. (b) Mean ( 7 1 SD) of the peak contact stress within the contact patch of the radius at each flexion angle.
po0.001 and 1372%, p¼ 0.006, respectively). The CRE design also caused decreases in contact area at the ulna and radius ( 1472%, p¼0.008 and 671%, p¼0.020, respectively). Comparing the three DHH designs, the CRE design yielded larger contact areas than the COM and BRE designs at the ulna (þ 2072%, p ¼0.003 and þ14 72%, p ¼0.015, respectively). At the radius, the CRE design provided significantly larger contact areas than the COM design (þ 15 73%, p¼ 0.048), but not the BRE design (þ 77 2%, p ¼0.077). 3.2. Contact stresses Peak contact stresses at the ulna and radius versus flexion angle are shown in Fig. 6. Compared to the native condition, the COM design caused peak contact stresses to increase 2.970.4 MPa (þ 461%, p¼0.008) at the ulna. There was no significant change in peak stress at the radius (average increase of 0.470.3 MPa or þ75%, p40.999). The BRE design had a variable effect on peak contact stresses at the ulna, and the average increase of 2.470.8 MPa (þ387%) was not statistically significant (p¼0.229). At the radius, the BRE design caused peak stresses to increase by 1.470.2 MPa (þ241%, p¼0.010). The CRE design caused peak contact stresses to increase by 1.070.1 MPa (þ165%, p¼0.003) at the ulna and 0.470.1 MPa (þ61%, p¼0.021) at the radius. Comparing the three implant designs, the CRE design provided the lowest peak contact stresses at the ulna and radius, but peak stresses were still higher than what was measured for the native condition.
4. Discussion This study used finite element contact analysis in order to analyze contact patterns and stresses of elbows before and after distal humeral hemiarthroplasty with commercially available, bone reverse engineered, and cartilage reverse engineered implant designs. Our findings primarily support our hypothesis that DHH implants whose shapes were reverse engineered from the distal humerus cartilage geometry would provide contact patterns and stresses in closer agreement to those of the native joint than the other designs. Contact was generally diffuse across the ulna and radius cartilage surfaces when the native joint was simulated. Contact stresses were minimal, rarely exceeding 0.5 MPa (Fig. 4). When DHH with the COM design was simulated, contact area was visibly reduced and appeared to be focused at different locations than the native joints’ contact patterns, and contact stresses were elevated. DHH with the BRE design generally yielded contact patterns which appear more similar to those of the native joint than the COM design, however the contact area was still reduced and contact stresses were still elevated. The CRE design appeared to provide the largest and most natural looking contact distributions among the three implant designs; however, contact stresses were still elevated. Recent studies have shown that cartilage wear is particularly prevalent at the ulna, reporting incidence rates of 1/10 patients (Burkhart et al., 2011), 3/8 patients (Adolfsson and Nestorson, 2012), 10/16 patients (Phadnis et al., 2015), 13/16 patients (Smith
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
and Hughes, 2013) and 7/7 patients (Hohman et al., 2014). Severe cartilage wear, to the point of complete wear-through (grade 2) or bone wear (grade 3) was reported in 6/16 patients by Smith and Hughes (2013), and correlated with worse results in some outcome measures; however DHH failures due to cartilage erosion alone are not reported. Insufficient long-term data exists to support whether or not cartilage erosion would ultimately necessitate revision. Our finite element analysis findings generally agree with our previous in-vitro studies with respect to the effects of DHH on joint contact area using commercially available or custom implants. Lapner et al. (2014) showed that a commercially available implant design caused contact area to decrease by 44% and 4% at the proximal ulna and radius, respectively, compared to decreases of 3473% and 2173% in the current study. Both studies predicted very similar trends at the ulna, but disagreed on the outcome at the proximal radius. A possible explanation for this disagreement is that in the in-vitro study, forearm motion was guided by an investigator, which may have prescribed artificial varus–valgus motion. This motion was unconstrained in our finite element model, and was guided by contact and simulated soft tissue tension only. Our previous in-vitro study (Willing et al., 2014a) investigating cartilage contact following DHH with a bone reverse engineered implant design found that contact area at the ulna and radius decreased by 42% and 41% (respectively) versus differences of 27 71% and 13 72% in the current study. The reason why the decreases in contact area were so much larger during the in-vitro study is not entirely clear; however, it could be caused by prescribed artificial varus–valgus motion in the in-vitro study, varying specimen-specific responses to DHH and small sample sizes (n¼ 5 in both studies). The CRE design provided larger contact areas at the ulna and radius than the other DHH designs. This design also tended to provide the lowest peak contact stresses. By reproducing the shape of the distal humerus cartilage, the CRE design resulted in more congruous contact with the ulna and radius, providing increased contact area and reduced peak contact stresses compared to the other designs. It is also noted that the relative ranking of implant designs in terms of maximizing contact area was consistent across the entire flexion range of motion at the ulna and radius. Although ranking was consistent, the size of the difference was not; for instance, contact areas of the radius were more similar in magnitude at 105° of flexion than at 15°. In terms of peak contact stresses, more variability in terms of the relative rankings of different designs was observed across the flexion range of motion. Although the articular geometries of the BRE and CRE designs are similar, they have different levels of congruency with the native ulna and radius, resulting in marked differences in contact stresses at different angles for these two designs. Additionally, the BRE design is systematically undersized as a result of neglecting the cartilage thickness, causing a less congruent articulation. Using the CRE design, contact area was still significantly decreased, and peak contact stresses were elevated by 165% and 61% compared to the native joint. Since this implant design exactly reproduced the shape of the distal humerus cartilage, this decrease in contact area can be solely attributed to the increased stiffness of rigid implant materials compared with relatively soft cartilage. The optimal distal humerus design may in fact lie somewhere between the shapes of the BRE and CRE designs, and would need to carefully balance maximizing contact area while avoiding over-stuffing (which was not simulated by the current model). Some limitations of this study should be recognized. The sample size of this study was small (n¼5), and the models described in this study were not directly validated against experimental data. Threshold-based segmentation techniques are sensitive to chosen threshold values, which could artificially increase/decrease bone
4041
sizes and increase/decrease cartilage thickness; hence we used values which we have validated (mean errors of only 0.3 mm reported by Lalone et al. (2015)) and are consistent with our previous studies (Irish et al., 2015; Langohr et al., 2015; Willing et al., 2014a, 2014b, 2013). Our FE modeling techniques have also been validated against experimental results (Willing et al., 2013). All specimens in all contact configurations received similar treatments, which means relative comparisons of resulting areas and contact stresses should be reliable. This is supported by the fact that our model has predicted similar trends as our previous experimental work in terms of contact area for the native, COM and BRE conditions (Lapner et al., 2014; Willing et al., 2014a). Future studies should directly measure cartilage contact stresses following DHH with these implant designs, although the limitations of interpositional measurement systems make such studies technically challenging and potentially error-prone (Brown et al., 2004; Stormont et al., 1985). Frictionless contact was assumed in our model for intact and hemiarthroplasty reconstructed joints, which is a simplification employed to narrow our focus on material stiffness and geometric factors. Future work should investigate the influence of friction on DHH contact mechanics. Furthermore, we simplified muscle and ligament insertions to single points, although distributed loads have been used in previous works (Geraldes and Phillips, 2014; Polgar et al., 2003). We assumed constant muscle and ligament loads; this is a simplification of the complexities of actual joint loading, but in agreement with our previous experimental studies on hemiarthroplasty contact mechanics (Lapner et al., 2014; Willing et al., 2014a). We eliminated overstuffing as a potentially confounding variable by assuming that the implant positioning by the surgeon was optimal, since our objective was to compare contact geometries. We also neglected the weight of the arm. We opted to use these minor simplifications under the assumption that they would affect all models equally. Future work should investigate the importance of these simplifications and surgical variables on implant design. Finally, we assumed that cartilage damage is likely to result from increases in cartilage contact stresses (Lizhang et al., 2011). Some recent studies have suggested that shear stresses, particularly at the boundary between the deep cartilage and subchondral bone layers, may also be important to cartilage erosion (Ng et al., 2012). The stresses in this region are highly influenced by the pressurization and flow of fluid within cartilage, and require a more complex material model to accurately represent this behavior (Guo et al., 2015; Pawaskar et al., 2011). Our findings show that the CRE design provides greater contact area and generally lower contact stress than the COM and BRE designs; however it is unable to match the native joint contact mechanics, which may lead to cartilage damage. These marginal improvements may not be sufficient to warrant the added costs associated with developing and manufacturing patient-specific implants, especially in light of the fact that CT based arthrography methods or MRI would be required in order to measure the joint-specific cartilage thickness distribution. Instead, it is more likely that improved DHH prostheses could be developed by considering more compliant biomaterials in combination with a more anatomical, but not necessarily custom, implant shape.
Conflict of interest statement Graham King receives royalty and consulting payments from Tornier Inc. None of the other authors have any conflict of interest, including any financial or personal relationships with other people or organizations that could inappropriately influence their work.
4042
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
Acknowledgments Funding was received from the Canadian Institute for Health Research (CIHR, 210521) in support of the data collection for this study. The first author was supported in part by the Joint Motion Program — A CIHR Training Program in Musculoskeletal Joint Research and Leadership. The authors would like to acknowledge the contribution of Dr. Michael Lapner, who assisted with specimen preparation for this study.
References Adolfsson, L., Nestorson, J., 2012. The Kudo humeral component as primary hemiarthroplasty in distal humeral fractures. J. Shoulder Elbow Surg. 21, 451–455. Anglen, J., 2005. Distal humerus fractures. J. Am. Acad. Orthop. Surg. 13, 291–297. Argintar, E., Berry, M., Narvy, S.J., Kramer, J., Omid, R., Itamura, J.M., 2012. Hemiarthroplasty for the treatment of distal humerus fractures: short-term clinical results. Orthopedics 35, 1042–1045. Brown, C.P., Nguyen, T.C., Moody, H.R., Crawford, R.W., Oloyede, A., 2009. Assessment of common hyperelastic constitutive equations for describing normal and osteoarthritic articular cartilage. Proc. Inst. Mech. Eng. Part H: J. Eng. Med. 223, 643–652. Brown, T.D., Rudert, M.J., Grosland, N.M., 2004. New methods for assessing cartilage contact stress after articular fracture. Clin. Orthop. Relat. Res. 423, 52–58. Burkhart, K.J., Nijs, S., Mattyasovszky, S.G., Wouters, R., Gruszka, D., Nowak, T.E., Rommens, P.M., Muller, L.P., 2011. Distal humerus hemiarthroplasty of the elbow for comminuted distal humeral fractures in the elderly patient. J. Trauma 71, 635–642. Fraser, G.S., Pichora, J.E., Ferreira, L.M., Brownhill, J.R., Johnson, J.A., King, G.J., 2008. Lateral collateral ligament repair restores the initial varus stability of the elbow: an in vitro biomechanical study. J. Orthop. Trauma 22, 615–623. Geraldes, D.M., Phillips, A.T., 2014. A comparative study of orthotropic and isotropic bone adaptation in the femur. Int. J. Numer. Methods Biomed. Eng. 30, 873–889. Guo, H., Maher, S.A., Torzilli, P.A., 2015. A biphasic finite element study on the role of the articular cartilage superficial zone in confined compression. J. Biomech. 48, 166–170. Hohman, D.W., Nodzo, S.R., Qvick, L.M., Duquin, T.R., Paterson, P.P., 2014. Hemiarthroplasty of the distal humerus for acute and chronic complex intraarticular injuries. J. Shoulder Elbow Surg. 23, 265–272. Irish, S.E., Langohr, G.D., Willing, R., King, G.J., Johnson, J.A., 2015. Implications of radial head hemiarthroplasty dish depth on radiocapitellar contact mechanics. J. Hand Surg. 40, 723–729. Lalone, E.A., Willing, R.T., Shannon, H.L., King, G.J., Johnson, J.A., 2015. Accuracy assessment of 3D bone reconstructions using CT: an intro comparison. Med. Eng. Phys. 37, 729–738. Langohr, G.D., Willing, R., Medley, J.B., King, G.J., Johnson, J.A., 2015. Contact analysis of the native radiocapitellar joint compared with axisymmetric and nonaxisymmetric radial head hemiarthroplasty. J. Shoulder Elbow Surg. 24, 787–795.
Lapner, M., Willing, R., Johnson, J.A., King, G.J., 2014. The effect of distal humeral hemiarthroplasty on articular contact of the elbow. Clin. Biomech. 29, 537–544. Lizhang, J., Fisher, J., Jin, Z., Burton, A., Williams, S., 2011. The effect of contact stress on cartilage friction, deformation and wear. Proc. Inst. Mech. Eng. Part H: J. Eng. Med. 225, 461–475. London, J.T., 1981. Kinematics of the elbow. J. Bone Joint Surg. Am. 63, 529–535. Mellen, R.H., Phalen, G.S., 1947. Arthroplasty of the elbow by replacement of the distal portion of the humerus with an acrylic prosthesis. J. Bone Joint Surg. Am. 29, 348–353. Morrey, B.F., Tanaka, S., An, K.N., 1991. Valgus stability of the elbow. A definition of primary and secondary constraints. Clin. Orthop. Relat. Res. 265, 187–195. Ng, K.C., Rouhi, G., Lamontagne, M., Beaule, P.E., 2012. Finite element analysis examining the effects of cam FAI on hip joint mechanical loading using subjectspecific geometries during standing and maximum squat. Musculoskelet. J. Hosp. Spec. Surg. 8, 206–212. Pawaskar, S.S., Ingham, E., Fisher, J., Jin, Z., 2011. Fluid load support and contact mechanics of hemiarthroplasty in the natural hip joint. Med. Eng. Phys. 33, 96–105. Phadnis, J., Banerjee, S., Watts, A.C., Little, N., Hearnden, A., Patel, V.R., 2015. Elbow hemiarthroplasty using a "triceps-on" approach for the management of acute distal humeral fractures. J. Shoulder Elbow Surg. 24, 1178–1186. Pichora, J.E., Fraser, G.S., Ferreira, L.F., Brownhill, J.R., Johnson, J.A., King, G.J.W., 2007. The effect of medial collateral ligament repair tension on elbow joint kinematics and stability. J. Hand Surg. 32, 1210–1217. Polgar, K., Gill, H.S., Viceconti, M., Murray, D.W., O'Connor, J.J., 2003. Strain distribution within the human femur due to physiological and simplified loading: finite element analysis using the muscle standardized femur model. Proc. Inst. Mech. Eng. Part H: J. Eng. Med. 217, 173–189. Robinson, C.M., Hill, R.M., Jacobs, N., Dall, G., Court-Brown, C.M., 2003. Adult distal humeral metaphyseal fractures: epidemiology and results of treatment. J. Orthop. Trauma 17, 38–47. Schenck Jr., R.C., Athanasiou, K.A., Constantinides, G., Gomez, E., 1994. A biomechanical analysis of articular cartilage of the human elbow and a potential relationship to osteochondritis dissecans. Clin. Orthop. Relat. Res. 299, 305–312. Shiba, R., Sorbie, C., Siu, D.W., Bryant, J.T., Cooke, T.D., Wevers, H.W., 1988. Geometry of the humeroulnar joint. J. Orthop. Res.: Off. Publ. Orthop. Res. Soc. 6, 897–906. Smith, G.C., Hughes, J.S., 2013. Unreconstructable acute distal humeral fractures and their sequelae treated with distal humeral hemiarthroplasty: a two-year to eleven-year follow-up. J. Shoulder Elbow Surg. 22, 1710–1723. Stormont, T.J., An, K.N., Morrey, B.F., Chao, E.Y., 1985. Elbow joint contact study: comparison of techniques. J. Biomech. 18, 329–336. Willing, R., Lapner, M., King, G.J., Johnson, J.A., 2014a. In vitro assessment of the contact mechanics of reverse-engineered distal humeral hemiarthroplasty prostheses. Clin. Biomech. 29, 990–996. Willing, R., Lapner, M., Lalone, E.A., King, G.J., Johnson, J.A., 2014b. Development of a computational technique to measure cartilage contact area. J. Biomech. 47, 1193–1197. Willing, R.T., Lalone, E.A., Shannon, H., Johnson, J.A., King, G.J., 2013. Validation of a finite element model of the human elbow for determining cartilage contact mechanics. J. Biomech. 46, 1767–1771.
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Contact mechanics of reverse engineered distal humeral hemiarthroplasty implants Ryan Willing a,b,n, Graham J.W. King b, James A. Johnson b a Department of Mechanical Engineering, Thomas J. Watson School of Engineering and Applied Science, State University of New York at Binghamton, Binghamton, NY, USA b Bioengineering Research Laboratory, The Hand and Upper Limb Centre, St. Joseph Health Care London, London, Ontario, Canada
art ic l e i nf o
a b s t r a c t
Article history: Accepted 27 September 2015
Erosion of articular cartilage is a concern following distal humeral hemiarthroplasty, because native cartilage surfaces are placed in contact with stiff metallic implant components, which causes decreases in contact area and increases in contact stresses. Recently, reverse engineered implants have been proposed which are intended to promote more natural contact mechanics by reproducing the native bone or cartilage shape. In this study, finite element modeling is used in order to calculate changes in cartilage contact areas and stresses following distal humeral hemiarthroplasty with commercially available and reverse engineered implant designs. At the ulna, decreases in contact area were 3473% (p¼ 0.002), 2771% (po0.001) and 1472% (p¼ 0.008) using commercially available, bone reverse engineered and cartilage reverse engineered designs, respectively. Peak contact stresses increased by 461757% (p¼0.008), 3877127% (p¼0.229) and 165716% (p¼ 0.003). At the radius, decreases in contact area were 2173% (p¼0.013), 1372% (po0.006) and 671% (p¼ 0.020), and peak contact stresses increased by 75752% (p40.999), 241732% (p¼ 0.010) and 61710% (p¼ 0.021). Between the three different implant designs, the cartilage reverse engineered design yielded the largest contact areas and lowest contact stresses, but was still unable to reproduce the contact mechanics of the native joint. These findings align with a growing body of evidence indicating that although reverse engineered hemiarthroplasty implants can provide small improvements in contact mechanics when compared with commercially available designs, further optimization of shape and material properties is required in order reproduce native joint contact mechanics. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Distal humeral hemiarthroplasty Reverse engineered implants Finite element analysis Contact mechanics Cartilage stress Elbow Orthopedic surgery
1. Introduction Distal humerus fractures occur with an incidence rate of 5.7/100,000 per year (Anglen, 2005; Robinson et al., 2003). Distal humeral hemiarthroplasty (DHH) was first described in 1947 (Mellen and Phalen, 1947), and is chosen when open reduction and fixation (ORIF) may have a low likelihood of success, but the patient is not a good candidate for total elbow replacement (TER) due to their higher functional demands (Argintar et al., 2012). Only recently have implants designed specifically for DHH become commercially available worldwide; however none are currently approved for use in the USA. Cartilage erosion following DHH has been reported in several recent studies (Adolfsson and Nestorson, 2012; Burkhart et al., 2011; Hohman et al., 2014; Phadnis et al., 2015; Smith and Hughes, 2013), particularly at the proximal ulna, and likely occurs as a result of replacing the native cartilaginous surfaces with a n Correspondence to: Department of Mechanical Engineering, Thomas J. Watson School of Engineering and Applied Science, Binghamton University – SUNY, P.O. Box 6000, Binghamton, NY 13902-6000, USA. Tel.: þ1 607 777 5038. E-mail address: [email protected] (R. Willing).
http://dx.doi.org/10.1016/j.jbiomech.2015.09.047 0021-9290/& 2015 Elsevier Ltd. All rights reserved.
comparatively rigid metal prosthesis. In a recent study (Smith and Hughes, 2013), moderate to severe cartilage wear at the proximal ulna was observed in 6/16 patients and correlated with worse results in some outcome measures. None of these studies reported DHH failures due to cartilage erosion alone, and insufficient data exists to predict long-term effects of wear progression. Previous in-vitro studies have demonstrated that commercially available (Lapner et al., 2014) and custom (Willing et al., 2014a) DHH implants fail to provide the same contact areas as native articulations. Neither study provided direct measurements of the cartilage stress. While it is generally accepted that contact areas and average contact stresses are inversely proportional, this is not necessarily true for peak contact stresses, necessitating direct measurements or computational analyses. The custom implants reproduced the shape of the osseous anatomy of the distal humerus (therefore called “reverse engineered” implants); this resulted in a systematic undersizing of these implants with respect to the actual cartilage surface, causing a loss of joint congruency (Willing et al., 2014a). The purpose of this study was to compare the contact patterns and cartilage stresses of elbows before and after DHH with commercially
4038
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
available, bone reverse engineered, and cartilage reverse engineered implants. We hypothesized that custom DHH implants reverse engineered from the cartilage geometry of the distal humerus would provide contact patterns and stresses in closer agreement to those of the native joint than the other designs.
Pewaukee, WI, USA) at 120 kV and 200 mA. Bones were oriented parallel with the axial direction of the CT scanner, and the articular cartilage surfaces were exposed to air during the CT scan. This protocol allows us to more clearly delineate the cartilage surface by leveraging a strong contrast against air, and allows cartilage thickness measurements with a mean error of only 0.3 mm compared to direct measurements (Lalone et al., 2015). Scans were performed with in-plane pixel sizes ranging between 0.283–0.391 mm and a 0.625 mm slice thickness. Image data was saved in the Digital Imaging and Communications in Medicine (DICOM) file format.
2. Methods 2.2. Model segmentation 2.1. Specimen preparation and imaging Five unpaired fresh-frozen cadaveric elbow specimens (left arms, three male and two female, mean age of 57, range 21–77) were used for this study. Institutional review board (IRB) approval was not required for this cadaver study. Specimens were disarticulated and denuded of all soft tissue such that the distal humerus, proximal ulna and proximal radius were separated and the articular cartilage was exposed. Each joint was examined and confirmed to be free of any signs of cartilage erosion or osteophyte formation. Exposed cartilage surfaces were loosely wrapped in normal saline-soaked towels to prevent desiccation. Within 24 hours, all specimens underwent computed tomographic (CT) imaging using a GE Discovery CT750 HD scanner (GE Health care,
CT scan DICOM files were imported into Mimics (Materialise, Leuven, Belgium) (Fig. 1a), where a threshold-based segmentation technique was used to label any voxels with a radiodensity greater than or equal to 250 HU as bone, and a threedimensional model was created using a marching cubes algorithm (Lalone et al., 2015; Willing et al., 2014a, 2014b, 2013). The threshold-based technique was used again to create bone plus cartilage models from the CT images by labeling any voxels with a radiodensity greater than or equal to 700 HU (Lalone et al., 2015; Lapner et al., 2014; Willing et al., 2014a, 2014b). All three-dimensional models were wrapped (minimum detail of 0.25 mm, 2.00 mm gap closing) using a built-in Mimics function, which filled small holes present in the bone models, primarily in non-articular regions, and did not affect cartilage thickness distributions. The resulting shell models, composed of triangles sharing edges and vertices representing either the bone or boneþ cartilage exterior shapes, were saved in the standard tessellation language (STL) file format. These models were reconstructed using a radial basis function, to smoothen the geometry and improve triangle shape and size uniformity (Fig. 1b). 2.3. Mesh generation Finite element (FE) mesh models for predicting elbow joint contact mechanics were prepared using a technique which was previously validated through comparison of model-predicted and experiment-measured contact pattern measurements made using casting techniques (Willing et al., 2013). NETGEN (RWTH Aachen University, Germany and Johannes Kepler University Linz, Austria) was used to mesh the boneþ cartilage STL surfaces with four-sided quadrilateral-shaped shell elements, with typical edge lengths ranging between 0.30 and 0.45 mm (Fig. 1c). These elements provided a template to eventually create six-sided hexahedral elements, which are generally preferred over tetrahedrons, to represent cartilage in the FE models. To reduce the solve time during FE analysis, any quad elements
Fig. 1. Bone (left) and cartilageþ bone (right) meshes were created from CT data. (a) CT data was segmented in MIMICS and (b) exported as STL models. (c) STL models were meshed using NETGEN to generate tetrahedral (tet) or quadrilateral (quad) meshes of the bone or cartilage, respectively. (d) The cartilage mesh was cropped to articular regions only. (e) Hexahedral (hex) elements representing cartilage were created based on the quadrilateral mesh, and assembled with the tetrahedral bone mesh.
Fig. 2. Loads and boundary conditions applied to FE model. Muscles forces (FBIC, FBRA and FTRI) acted at the corresponding insertion points of the biceps, brachialis and triceps (respectively) on the ulna and radius, and were oriented parallel with the long axis of the humerus. Lateral collateral and medial collateral ligament forces (FLCL and FMCL) were oriented towards the anatomic flexion/extension (F/E) axis of the elbow.
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
4039
Fig. 3. The four contact configurations simulated by the FE model. (a) The native distal humerus was represented by a deformable cartilage layer. (b–d) DHH with a commercially available, bone reverse engineered, and cartilage reverse engineered implant, referred to as the COM, BRE and CRE designs, respectively. DHH implants were modeled as rigid surfaces. which were obviously not on the articular surface were manually deleted, except in the coronoid and olecranon fossa, where potential contact must be permitted (Fig. 1d). The remaining elements were swept to the corresponding bone STL shell models, and three layers of linear hexahedral elements were generated in between. The resulting hexahedral mesh represented the nonuniform thickness articular cartilage layer in the FE models. The bone STL shell models were meshed using linear tetrahedral elements with 2 mm typical edge lengths using NETGEN. 2.4. FE model The bone tetrahedral and cartilage hexahedral meshes were imported into Abaqus (Dassault Systèmes Simulia Corp., Providence, RI, USA). The distal humerus, proximal ulna and proximal radius of each specimen were assembled with the elbow situated between 60–90° of flexion and with neutral forearm rotation. Cartilage elements were assigned Neo-Hookean hyperelastic material properties with a bulk modulus (K) of 0.31 MPa and a shear modulus (G) of 0.37 MPa (Brown et al., 2009; Schenck et al., 1994), which approximates the equilibrium response of cartilage to compressive loading (Willing et al., 2013). Bone was assumed rigid; the effect of this simplification has been shown to be minimal in previous work (Willing et al., 2013). The interface between hexahedral cartilage and tetrahedral subchondral bone elements was simulated using a tie constraint (bonded contact, no sliding). The radius and ulna were modeled as a single rigid body with separate deformable contact surfaces at the proximal articulations, effectively simulating a forearm pinned in neutral rotation, consistent with previous invitro studies (Lapner et al., 2014; Willing et al., 2014a, 2014b). Contact between cartilage surfaces was simulated using surface-to-surface frictionless contact with a nonlinear contact stiffness Penalty algorithm (Willing et al., 2013). The insertions of the triceps, biceps and brachialis tendons, and the medial and lateral collateral ligaments, were identified on the ulna and radius by an orthopedic surgeon. Constant muscle forces of 40 N, 20 N and 20 N were applied at the approximate center of each insertion of the triceps, biceps and brachialis tendons (respectively) and were oriented parallel to the long axis of the humerus (Lapner et al., 2014; Morrey et al., 1991; Willing et al., 2014a, 2014b). Constant ligament forces of 20 N each were applied at their respective insertions, and were oriented towards the corresponding origins on the distal humerus (Fraser et al., 2008; Lapner et al., 2014; Pichora et al., 2007; Willing et al., 2014a, 2014b). These origins were defined as the locations where the elbow anatomical flexion/ extension (F/E) axis (defined by a line passing through the centers of a sphere fit to the capitellum and a circle fit to the trochlea (London, 1981; Shiba et al., 1988)) passed through the corresponding medial or lateral cortex of the distal humerus (Fig. 2). Models were created for each specimen in order to simulate four different configurations (Fig. 3). 2.4.1. Native A native distal humerus was simulated by including the native distal humerus cartilage, modelled using deformable hexahedral elements. 2.4.2. Commercially available DHH (COM) Hemiarthroplasty with a commercially available DHH prosthesis was simulated. The surface geometries of Latitude Anatomic DHH prostheses (Tornier, Texas, USA) were represented by rigid surface meshes of triangle shaped elements. Optimal prosthesis size for each specimen was chosen based on matching the medial–lateral distance between the trochlea and capitellum centers (Lapner et al., 2014). 2.4.3. Bone reverse engineered DHH (BRE) A custom DHH prosthesis based on the distal humerus bone shape for each specimen (Willing et al., 2014a) was simulated by eliminating the cartilage layer and allowing contact between the rigid surface of the implant representing the distal humerus bone and the cartilage surfaces of the proximal radius and ulna. 2.4.4. Cartilage reverse engineered DHH (CRE) A custom DHH prosthesis based on the distal humerus cartilage shape was simulated for each specimen by modeling the distal humerus cartilage mesh as a rigid surface.
Fig. 4. Typical contact stress results for a single specimen. Contours describe the contact stresses across the contact surfaces of the ulna and radius at 15, 60 and 105° of flexion, when articulating with the native distal humerus (Native), the commercially available DHH design (COM), the bone reverse engineered DHH design (BRE), and the cartilage reverse engineered DHH design (CRE). Red fringe plot values denote contact stresses at or above 2 MPa, this upper limit chosen in order to permit better visualization of the entire contact region. Flexion was simulated by rotating the forearm about the F/E axis of the (fully constrained) humerus (15–105° in 5° increments; Fig. 2). All remaining degrees of freedom, including forearm translations, varus–valgus and internal–external rotations, remained unconstrained and were able to change as a result of the applied loads and contact. Contact areas and stresses at the proximal ulna and radius were calculated. Contact areas were reported as a percentage of the total cartilage surface area, normalizing for differences in specimen sizes. The effect of joint condition on contact area and peak contact stress was examined using a repeated measures analysis of variance (ANOVA) at the ulna and radius. Statistically significant differences were reported if po0.05 after a Bonferroni adjustment for multiple comparisons.
3. Results Results for a single specimen at 15, 60 and 105° of flexion for each contact configuration are shown in Fig. 4, and are representative of the typical stress distributions at extension, midflexion, and full flexion. Contact areas include any region where contact stresses are non-zero. 3.1. Contact area Fig. 5 shows the average contact area (expressed as a percentage of the entire contact area) as a function of flexion angle for each joint condition. Contact area was significantly reduced at both the ulna ( 3473%, p¼ 0.002) and radius ( 2173%, p¼0.013) when the COM design was used. When the BRE design was used, significant decreases were still observed at the ulna and radius ( 2771%,
4040
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
Fig. 5. Contact area for each joint condition. (a) Mean ( 71 SD) of the ulna contact areas, expressed as a percentage of the entire ulna contact surface, versus flexion angle. (b) Mean ( 71 SD) of the radius contact areas, expressed as a percentage of the entire radius contact surface, as a function of flexion angle.
Fig. 6. Peak contact stresses for each joint condition. (a) Mean ( 7 1 SD) of the peak contact stress within the contact patch of the ulna at each flexion angle. (b) Mean ( 7 1 SD) of the peak contact stress within the contact patch of the radius at each flexion angle.
po0.001 and 1372%, p¼ 0.006, respectively). The CRE design also caused decreases in contact area at the ulna and radius ( 1472%, p¼0.008 and 671%, p¼0.020, respectively). Comparing the three DHH designs, the CRE design yielded larger contact areas than the COM and BRE designs at the ulna (þ 2072%, p ¼0.003 and þ14 72%, p ¼0.015, respectively). At the radius, the CRE design provided significantly larger contact areas than the COM design (þ 15 73%, p¼ 0.048), but not the BRE design (þ 77 2%, p ¼0.077). 3.2. Contact stresses Peak contact stresses at the ulna and radius versus flexion angle are shown in Fig. 6. Compared to the native condition, the COM design caused peak contact stresses to increase 2.970.4 MPa (þ 461%, p¼0.008) at the ulna. There was no significant change in peak stress at the radius (average increase of 0.470.3 MPa or þ75%, p40.999). The BRE design had a variable effect on peak contact stresses at the ulna, and the average increase of 2.470.8 MPa (þ387%) was not statistically significant (p¼0.229). At the radius, the BRE design caused peak stresses to increase by 1.470.2 MPa (þ241%, p¼0.010). The CRE design caused peak contact stresses to increase by 1.070.1 MPa (þ165%, p¼0.003) at the ulna and 0.470.1 MPa (þ61%, p¼0.021) at the radius. Comparing the three implant designs, the CRE design provided the lowest peak contact stresses at the ulna and radius, but peak stresses were still higher than what was measured for the native condition.
4. Discussion This study used finite element contact analysis in order to analyze contact patterns and stresses of elbows before and after distal humeral hemiarthroplasty with commercially available, bone reverse engineered, and cartilage reverse engineered implant designs. Our findings primarily support our hypothesis that DHH implants whose shapes were reverse engineered from the distal humerus cartilage geometry would provide contact patterns and stresses in closer agreement to those of the native joint than the other designs. Contact was generally diffuse across the ulna and radius cartilage surfaces when the native joint was simulated. Contact stresses were minimal, rarely exceeding 0.5 MPa (Fig. 4). When DHH with the COM design was simulated, contact area was visibly reduced and appeared to be focused at different locations than the native joints’ contact patterns, and contact stresses were elevated. DHH with the BRE design generally yielded contact patterns which appear more similar to those of the native joint than the COM design, however the contact area was still reduced and contact stresses were still elevated. The CRE design appeared to provide the largest and most natural looking contact distributions among the three implant designs; however, contact stresses were still elevated. Recent studies have shown that cartilage wear is particularly prevalent at the ulna, reporting incidence rates of 1/10 patients (Burkhart et al., 2011), 3/8 patients (Adolfsson and Nestorson, 2012), 10/16 patients (Phadnis et al., 2015), 13/16 patients (Smith
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
and Hughes, 2013) and 7/7 patients (Hohman et al., 2014). Severe cartilage wear, to the point of complete wear-through (grade 2) or bone wear (grade 3) was reported in 6/16 patients by Smith and Hughes (2013), and correlated with worse results in some outcome measures; however DHH failures due to cartilage erosion alone are not reported. Insufficient long-term data exists to support whether or not cartilage erosion would ultimately necessitate revision. Our finite element analysis findings generally agree with our previous in-vitro studies with respect to the effects of DHH on joint contact area using commercially available or custom implants. Lapner et al. (2014) showed that a commercially available implant design caused contact area to decrease by 44% and 4% at the proximal ulna and radius, respectively, compared to decreases of 3473% and 2173% in the current study. Both studies predicted very similar trends at the ulna, but disagreed on the outcome at the proximal radius. A possible explanation for this disagreement is that in the in-vitro study, forearm motion was guided by an investigator, which may have prescribed artificial varus–valgus motion. This motion was unconstrained in our finite element model, and was guided by contact and simulated soft tissue tension only. Our previous in-vitro study (Willing et al., 2014a) investigating cartilage contact following DHH with a bone reverse engineered implant design found that contact area at the ulna and radius decreased by 42% and 41% (respectively) versus differences of 27 71% and 13 72% in the current study. The reason why the decreases in contact area were so much larger during the in-vitro study is not entirely clear; however, it could be caused by prescribed artificial varus–valgus motion in the in-vitro study, varying specimen-specific responses to DHH and small sample sizes (n¼ 5 in both studies). The CRE design provided larger contact areas at the ulna and radius than the other DHH designs. This design also tended to provide the lowest peak contact stresses. By reproducing the shape of the distal humerus cartilage, the CRE design resulted in more congruous contact with the ulna and radius, providing increased contact area and reduced peak contact stresses compared to the other designs. It is also noted that the relative ranking of implant designs in terms of maximizing contact area was consistent across the entire flexion range of motion at the ulna and radius. Although ranking was consistent, the size of the difference was not; for instance, contact areas of the radius were more similar in magnitude at 105° of flexion than at 15°. In terms of peak contact stresses, more variability in terms of the relative rankings of different designs was observed across the flexion range of motion. Although the articular geometries of the BRE and CRE designs are similar, they have different levels of congruency with the native ulna and radius, resulting in marked differences in contact stresses at different angles for these two designs. Additionally, the BRE design is systematically undersized as a result of neglecting the cartilage thickness, causing a less congruent articulation. Using the CRE design, contact area was still significantly decreased, and peak contact stresses were elevated by 165% and 61% compared to the native joint. Since this implant design exactly reproduced the shape of the distal humerus cartilage, this decrease in contact area can be solely attributed to the increased stiffness of rigid implant materials compared with relatively soft cartilage. The optimal distal humerus design may in fact lie somewhere between the shapes of the BRE and CRE designs, and would need to carefully balance maximizing contact area while avoiding over-stuffing (which was not simulated by the current model). Some limitations of this study should be recognized. The sample size of this study was small (n¼5), and the models described in this study were not directly validated against experimental data. Threshold-based segmentation techniques are sensitive to chosen threshold values, which could artificially increase/decrease bone
4041
sizes and increase/decrease cartilage thickness; hence we used values which we have validated (mean errors of only 0.3 mm reported by Lalone et al. (2015)) and are consistent with our previous studies (Irish et al., 2015; Langohr et al., 2015; Willing et al., 2014a, 2014b, 2013). Our FE modeling techniques have also been validated against experimental results (Willing et al., 2013). All specimens in all contact configurations received similar treatments, which means relative comparisons of resulting areas and contact stresses should be reliable. This is supported by the fact that our model has predicted similar trends as our previous experimental work in terms of contact area for the native, COM and BRE conditions (Lapner et al., 2014; Willing et al., 2014a). Future studies should directly measure cartilage contact stresses following DHH with these implant designs, although the limitations of interpositional measurement systems make such studies technically challenging and potentially error-prone (Brown et al., 2004; Stormont et al., 1985). Frictionless contact was assumed in our model for intact and hemiarthroplasty reconstructed joints, which is a simplification employed to narrow our focus on material stiffness and geometric factors. Future work should investigate the influence of friction on DHH contact mechanics. Furthermore, we simplified muscle and ligament insertions to single points, although distributed loads have been used in previous works (Geraldes and Phillips, 2014; Polgar et al., 2003). We assumed constant muscle and ligament loads; this is a simplification of the complexities of actual joint loading, but in agreement with our previous experimental studies on hemiarthroplasty contact mechanics (Lapner et al., 2014; Willing et al., 2014a). We eliminated overstuffing as a potentially confounding variable by assuming that the implant positioning by the surgeon was optimal, since our objective was to compare contact geometries. We also neglected the weight of the arm. We opted to use these minor simplifications under the assumption that they would affect all models equally. Future work should investigate the importance of these simplifications and surgical variables on implant design. Finally, we assumed that cartilage damage is likely to result from increases in cartilage contact stresses (Lizhang et al., 2011). Some recent studies have suggested that shear stresses, particularly at the boundary between the deep cartilage and subchondral bone layers, may also be important to cartilage erosion (Ng et al., 2012). The stresses in this region are highly influenced by the pressurization and flow of fluid within cartilage, and require a more complex material model to accurately represent this behavior (Guo et al., 2015; Pawaskar et al., 2011). Our findings show that the CRE design provides greater contact area and generally lower contact stress than the COM and BRE designs; however it is unable to match the native joint contact mechanics, which may lead to cartilage damage. These marginal improvements may not be sufficient to warrant the added costs associated with developing and manufacturing patient-specific implants, especially in light of the fact that CT based arthrography methods or MRI would be required in order to measure the joint-specific cartilage thickness distribution. Instead, it is more likely that improved DHH prostheses could be developed by considering more compliant biomaterials in combination with a more anatomical, but not necessarily custom, implant shape.
Conflict of interest statement Graham King receives royalty and consulting payments from Tornier Inc. None of the other authors have any conflict of interest, including any financial or personal relationships with other people or organizations that could inappropriately influence their work.
4042
R. Willing et al. / Journal of Biomechanics 48 (2015) 4037–4042
Acknowledgments Funding was received from the Canadian Institute for Health Research (CIHR, 210521) in support of the data collection for this study. The first author was supported in part by the Joint Motion Program — A CIHR Training Program in Musculoskeletal Joint Research and Leadership. The authors would like to acknowledge the contribution of Dr. Michael Lapner, who assisted with specimen preparation for this study.
References Adolfsson, L., Nestorson, J., 2012. The Kudo humeral component as primary hemiarthroplasty in distal humeral fractures. J. Shoulder Elbow Surg. 21, 451–455. Anglen, J., 2005. Distal humerus fractures. J. Am. Acad. Orthop. Surg. 13, 291–297. Argintar, E., Berry, M., Narvy, S.J., Kramer, J., Omid, R., Itamura, J.M., 2012. Hemiarthroplasty for the treatment of distal humerus fractures: short-term clinical results. Orthopedics 35, 1042–1045. Brown, C.P., Nguyen, T.C., Moody, H.R., Crawford, R.W., Oloyede, A., 2009. Assessment of common hyperelastic constitutive equations for describing normal and osteoarthritic articular cartilage. Proc. Inst. Mech. Eng. Part H: J. Eng. Med. 223, 643–652. Brown, T.D., Rudert, M.J., Grosland, N.M., 2004. New methods for assessing cartilage contact stress after articular fracture. Clin. Orthop. Relat. Res. 423, 52–58. Burkhart, K.J., Nijs, S., Mattyasovszky, S.G., Wouters, R., Gruszka, D., Nowak, T.E., Rommens, P.M., Muller, L.P., 2011. Distal humerus hemiarthroplasty of the elbow for comminuted distal humeral fractures in the elderly patient. J. Trauma 71, 635–642. Fraser, G.S., Pichora, J.E., Ferreira, L.M., Brownhill, J.R., Johnson, J.A., King, G.J., 2008. Lateral collateral ligament repair restores the initial varus stability of the elbow: an in vitro biomechanical study. J. Orthop. Trauma 22, 615–623. Geraldes, D.M., Phillips, A.T., 2014. A comparative study of orthotropic and isotropic bone adaptation in the femur. Int. J. Numer. Methods Biomed. Eng. 30, 873–889. Guo, H., Maher, S.A., Torzilli, P.A., 2015. A biphasic finite element study on the role of the articular cartilage superficial zone in confined compression. J. Biomech. 48, 166–170. Hohman, D.W., Nodzo, S.R., Qvick, L.M., Duquin, T.R., Paterson, P.P., 2014. Hemiarthroplasty of the distal humerus for acute and chronic complex intraarticular injuries. J. Shoulder Elbow Surg. 23, 265–272. Irish, S.E., Langohr, G.D., Willing, R., King, G.J., Johnson, J.A., 2015. Implications of radial head hemiarthroplasty dish depth on radiocapitellar contact mechanics. J. Hand Surg. 40, 723–729. Lalone, E.A., Willing, R.T., Shannon, H.L., King, G.J., Johnson, J.A., 2015. Accuracy assessment of 3D bone reconstructions using CT: an intro comparison. Med. Eng. Phys. 37, 729–738. Langohr, G.D., Willing, R., Medley, J.B., King, G.J., Johnson, J.A., 2015. Contact analysis of the native radiocapitellar joint compared with axisymmetric and nonaxisymmetric radial head hemiarthroplasty. J. Shoulder Elbow Surg. 24, 787–795.
Lapner, M., Willing, R., Johnson, J.A., King, G.J., 2014. The effect of distal humeral hemiarthroplasty on articular contact of the elbow. Clin. Biomech. 29, 537–544. Lizhang, J., Fisher, J., Jin, Z., Burton, A., Williams, S., 2011. The effect of contact stress on cartilage friction, deformation and wear. Proc. Inst. Mech. Eng. Part H: J. Eng. Med. 225, 461–475. London, J.T., 1981. Kinematics of the elbow. J. Bone Joint Surg. Am. 63, 529–535. Mellen, R.H., Phalen, G.S., 1947. Arthroplasty of the elbow by replacement of the distal portion of the humerus with an acrylic prosthesis. J. Bone Joint Surg. Am. 29, 348–353. Morrey, B.F., Tanaka, S., An, K.N., 1991. Valgus stability of the elbow. A definition of primary and secondary constraints. Clin. Orthop. Relat. Res. 265, 187–195. Ng, K.C., Rouhi, G., Lamontagne, M., Beaule, P.E., 2012. Finite element analysis examining the effects of cam FAI on hip joint mechanical loading using subjectspecific geometries during standing and maximum squat. Musculoskelet. J. Hosp. Spec. Surg. 8, 206–212. Pawaskar, S.S., Ingham, E., Fisher, J., Jin, Z., 2011. Fluid load support and contact mechanics of hemiarthroplasty in the natural hip joint. Med. Eng. Phys. 33, 96–105. Phadnis, J., Banerjee, S., Watts, A.C., Little, N., Hearnden, A., Patel, V.R., 2015. Elbow hemiarthroplasty using a "triceps-on" approach for the management of acute distal humeral fractures. J. Shoulder Elbow Surg. 24, 1178–1186. Pichora, J.E., Fraser, G.S., Ferreira, L.F., Brownhill, J.R., Johnson, J.A., King, G.J.W., 2007. The effect of medial collateral ligament repair tension on elbow joint kinematics and stability. J. Hand Surg. 32, 1210–1217. Polgar, K., Gill, H.S., Viceconti, M., Murray, D.W., O'Connor, J.J., 2003. Strain distribution within the human femur due to physiological and simplified loading: finite element analysis using the muscle standardized femur model. Proc. Inst. Mech. Eng. Part H: J. Eng. Med. 217, 173–189. Robinson, C.M., Hill, R.M., Jacobs, N., Dall, G., Court-Brown, C.M., 2003. Adult distal humeral metaphyseal fractures: epidemiology and results of treatment. J. Orthop. Trauma 17, 38–47. Schenck Jr., R.C., Athanasiou, K.A., Constantinides, G., Gomez, E., 1994. A biomechanical analysis of articular cartilage of the human elbow and a potential relationship to osteochondritis dissecans. Clin. Orthop. Relat. Res. 299, 305–312. Shiba, R., Sorbie, C., Siu, D.W., Bryant, J.T., Cooke, T.D., Wevers, H.W., 1988. Geometry of the humeroulnar joint. J. Orthop. Res.: Off. Publ. Orthop. Res. Soc. 6, 897–906. Smith, G.C., Hughes, J.S., 2013. Unreconstructable acute distal humeral fractures and their sequelae treated with distal humeral hemiarthroplasty: a two-year to eleven-year follow-up. J. Shoulder Elbow Surg. 22, 1710–1723. Stormont, T.J., An, K.N., Morrey, B.F., Chao, E.Y., 1985. Elbow joint contact study: comparison of techniques. J. Biomech. 18, 329–336. Willing, R., Lapner, M., King, G.J., Johnson, J.A., 2014a. In vitro assessment of the contact mechanics of reverse-engineered distal humeral hemiarthroplasty prostheses. Clin. Biomech. 29, 990–996. Willing, R., Lapner, M., Lalone, E.A., King, G.J., Johnson, J.A., 2014b. Development of a computational technique to measure cartilage contact area. J. Biomech. 47, 1193–1197. Willing, R.T., Lalone, E.A., Shannon, H., Johnson, J.A., King, G.J., 2013. Validation of a finite element model of the human elbow for determining cartilage contact mechanics. J. Biomech. 46, 1767–1771.