Geometry. A set of 20 FE computational model variants was developed, representing diabetic tissue conditions and comparable healthy tissue cases at different foot postures, as specified in Table 1. Each model variant captured the anatomy of the posterior aspect of the heel, including the calcaneus bone, distal Achilles tendon, fat, and skin (see Figure 1a). Flat elastic foam was modeled as the support surface (see Figure 1a). Ten (10) of the model variants did not include the multilayered dressing and were used as reference cases, which then could be compared to cases where the foam dressing was incorporated (see Table 1). Variants 1–4 were set to a neutral foot position (0˚ plantar flexion) and variants 5–8, 9–12, and 13–20 were set to 10˚, 20˚, and 30˚ of plantar flexion, respectively (see Table 1, Figure 1). For each position of the foot tested, either diabetic or healthy soft tissues properties were assigned.
The anatomical structures of the posterior heel from the MRI dataset, as well as geometrical modeling of the dressing, were segmented as described in the authors’ previous work.27 Briefly, 56 T1-weighted axial MRI slices from the suspended left heel of a healthy man were used to develop an anatomically realistic, 3D, undeformed model geometry, using the ScanIP® module of the Simpleware® software package30,31(Simpleware, Exeter, United Kingdom). For clinical relevance, the study dressing was modeled (the specific considerations regarding the geometrical and material modeling of the dressing have been published previously27). The innermost and outermost layers of the dressing were modeled as contact conditions and used according to the manufacturer’s specifications to describe the airlaid (layer 2), nonwoven (layer 3), and polyurethane foam (layer 4) as physical layers in the modeling. The flat foam support (10-mm thickness) was added to all the heel model variants at the preprocessing stage in the PreView module of the FEBio software package.32
Mechanical properties of the tissues and dressing. Constitutive laws and mechanical properties of all the tissues included in the heel model were adopted from the literature (see Table 2). Specifically, the healthy Achilles tendon and calcaneus bone were assumed to be linear-elastic isotropic materials (ie, having a constant stiffness that does not depend on the extent of deformation nor does it depend on the direction of the deformation) with instantaneous elastic moduli of 205 kPa and 7 GPa, and Poisson’s ratios (indicating a material’s deformation at a direction perpendicular to the loading direction) of 0.49 and 0.3, respectively33,34 (see Table 2). Skin and fat tissues were assumed to be nearly incompressible (Poisson’s ratio of 0.495), nonlinear isotropic materials. The large deformation mechanical behavior of skin and fat was described by an uncoupled Neo-Hookean material model35 with instantaneous shear moduli33,36 (see Table 2). Based on the literature, the instantaneous shear moduli of diabetic skin and fat tissues were assumed to be 40% stiffer than the corresponding healthy tissue14; likewise, the diabetic Achilles tendon was assigned a 40% softer instantaneous elastic modulus17 (see Table 2). The flat foam support was assumed to be isotropic linear elastic with a Poisson’s ratio of 0.3 and an elastic modulus of 45 kPa, which are within the range of hospital mattress properties.37 Layers 2, 3, and 4 were assigned instantaneous elastic moduli of 15.3, 75, and 12 kPa, respectively, according to the measurements described previously27 (see Table 2). The Poisson’s ratios of all dressing materials38 were set as 0.258 (see Table 2).
Body loads applied to the model, shear, and friction conditions. The loads applied to the model were chosen to simulate the descent of the calcaneus bone against a flat foam support during supine weight-bearing. For calibration, tissue deformations were measured from a weight-bearing MRI of the heel of the same subject to find the weight force the heel applied on the support. Because the total reaction force should be equal to the weight of the foot/ankle in the supine position, this datum was used to produce the same heel-support reaction forces in all the model variants. The resulting compressive (downward) displacements after imposing the same foot/ankle weight value in all the model variants were in the 4.7-mm to 5.2-mm range. Additionally, in order to represent shearing forces acting on the foot when the head of the bed is elevated, and by doing so to evaluate the importance of subsequent heel repositioning in patients with and without diabetes, model variants 17–20 were used. In these specific variants, a combination of compression and shear loading was employed, where shear displacements were taken as equal to the compressive ones27 (see Table 1). The bottom surface of the elastic foam support was fixed for all motions. Tied (fixed) interfaces were defined between all tissue components as well as between the skin and the polyurethane foam layer of the dressing (layer 4) to account for the adherence properties of the innermost Safetac® layer of the dressing that interfaces the skin (layer 5). Frictional contact between the support and either the skin of the bare heel or the outer surface of the dressing was defined, with the coefficients of friction set as 0.43 and 0.35 for the skin-support and dressing-support contacts, respectively.27,38,39
Creating the computational simulations. The tissues and dressing were meshed using the ScanIP® module of Simpleware®,30 with mesh refinements that were applied in the skin and dressing materials near the contact area with the support (see Figure 1a). Meshing the support was performed in the Preview module of FEBio.32 All of the above FE simulations were set up using the PreView module of FEBio (Version 1.14), analyzed using the Pardiso linear solver of FEBio (http://mrl.sci.utah.edu/software/febio) (Version 1.7.1) and post-processed using the PostView tool of FEBio (Version 1.4).32
Biomechanical outcome measures. The volumetric exposures of the soft tissues to mechanical strains were systematically compared using the effective Green-Lagrange strain as the strain measure.35 The effective Green-Lagrange strain is an adequate measure of large tissue deformations and the consequent risk for HUs because it is simultaneously considering tissue deformations in all loading modes (ie, compression, tension and shear) using a single weighed (scalar) measure. The strain data were pooled together from tendon, fat, and skin tissues for all elements in a consistent volume of interest in the meshes, which was defined by the circumference of the calcaneus and its projection at the retrocalcaneal region, as illustrated in Figure 1b. Additionally, peak effective and maximal shear stresses (shear stresses arise from shape distortion of the tissues due to forces parallel to the cross-sections of the tissues) in fat and skin tissues were compared separately as a second measure of the risk for HUs. For clarification, effective strain and stress measures consider compression, tension, and shear loads altogether, as opposed to just the aforementioned (maximal) shear measures, which do not consider tension and compression loads. The stress data were pooled from all the elements of skin or separately from those of fat. Converged time steps for data collection were chosen so the resulting reaction forces acting back from the support were within <3% difference from the predefined target reaction force (as explained previously). Hence, the biomechanical efficacy of the dressing was evaluated in diabetic and healthy tissue conditions based on the criteria of reduction in volumetric exposures of the soft tissues to critical effective and shear strains and stresses. The verification of the reference model variant based on the calculated interface pressures under the heel and corresponding compressive strains in the support, which have been compared to published experimental data, is described in a previous paper by the authors that fully validates the present modeling framework.27
Data collection and analyses. Excel software (Microsoft Co, Spokane, WA) was used for the post-hoc analyses of the finite element simulations and for plotting and quantitatively comparing the results of all simulations. It should be mentioned that means and standard deviations are not provided because no variability in the modeling outcomes exists when the parameters of the models are set at a certain way (see Sidebar). Accordingly, no variance around the mean and no average values are reported — just a single value of each parameter that is calculated from a predefined set of simulation parameters (eg, peak tissue strain is calculated based on a simulated set of specific tissue mechanical properties and whether a dressing is present or not). The standard deviations are zero; each time the same simulation is run, the same outcome is obtained.