DISCUSSION

Mechanical performance of the cortical bone analogs

The primary mechanical benchmark for a cortical bone analog is the elastic modulus of natural canine cortical bone, reported across skeletal sites and breeds in the range of approximately 7.5–21.0 GPa [7]. The highest modulus achieved in the present study (E+5%RF: 1,513.12 ± 67.85 MPa) represents approximately 20% of the lower bound of that target range. This shortfall is consistent with the known stiffness ceiling of short-fiber thermoset composites fabricated by manual open-cast molding, in which fiber length distribution, random orientation, and residual porosity limit load transfer efficiency [30, 31, 32]. No equivalent modulus response was observed in the polyester system, in which elastic modulus remained statistically unchanged across all three weight fractions (p > 0.05 for all pairwise comparisons), a finding that distinguishes the two resin matrices and has practical implications for formulation selection in low-resource fabrication contexts.

Figure 4: Stress–strain curves of individual specimens. E = epoxy resin; R = polyester resin; RF = fiber reinforcement weight fraction. The x-axis scales differ between panels to optimize the readability of the individual curve morphology. Absolute strain values for each group are presented in Table 2.

Figure 5: Mechanical property comparison between E+3%RF and PU foam. (A) Elastic modulus. (B) Yield strength. (C) Ultimate stress. (D) Yield strain. (E) Ultimate strain. ns = p ≥ 0.05; * = p < 0.05; *** = p < 0.001; **** = p < 0.0001.

This difference between the two resin systems is most likely attributable to cure shrinkage. Unsaturated R undergoes volumetric shrinkage of 7%–12% during polymerization, compared with <2% for E [36]. In an open-cast, unpressurized format, as used in the present study, this shrinkage generates residual stresses at the fiber-matrix interface, promotes micro-debonding, and reduces the efficiency of load transfer from the matrix to the fiber [37, 38]. The result is that fiber addition in R produced no statistically detectable stiffness gain at any weight fraction tested. Manufacturing consistency followed the same pattern: the coefficient of variation for elastic modulus was 1.4%–4.5% across the epoxy groups and 5.9%–14.5% across the polyester groups at equivalent fiber weight fractions (Table 2). A coefficient of variation of 14.5% means that specimens of nominally identical composition will behave differently during surgical training, which defeats the purpose of a standardized synthetic model. On both counts—stiffness response to reinforcement and specimen-to-specimen consistency—E is the preferable matrix for hand-pour canine bone model fabrication.

Effect of fiber reinforcement on tensile properties

Fiber reinforcement reduced ultimate tensile stress and ultimate strain in both resin systems despite increasing elastic modulus. In E, ultimate stress fell from 33.17 ± 5.24 MPa to 17.71 ± 2.27 MPa at 5 wt%, representing a 46.6% reduction, and ultimate strain decreased from 5.96 ± 0.79% to 1.66 ± 0.35%. R showed a comparable decline, with ultimate stress decreasing from 31.28 ± 8.92 MPa to 12.89 ± 2.00 MPa and ultimate strain decreasing from 4.82 ± 1.08% to 1.73 ± 0.81%, respectively (Table 2). This finding is consistent with fiber agglomeration and void formation at higher weight fractions in hand-mixed systems, generating stress concentrators that initiate premature matrix failure under tensile loading [39, 40]. The reduction in ductility does not, however, necessarily impair training utility. Canine cortical bone has an ultimate tensile strain of approximately 1%–3% [30, 41], and unreinforced E at 5.96% already exceeds this range substantially; epoxy composites containing 3–5 wt% fiber reinforcement (1.93%–1.66%) fall within or immediately below this range, representing a closer approximation to natural tissue. Whether the reduced ultimate stress at higher fiber weight fractions causes premature or inconsistent fracture under drilling and sawing loads cannot be determined from tensile coupon data alone.

Selection of the optimal cortical bone formulation

The maximum fiber weight fraction at which hand-pour casting into the silicone mold remained feasible was 5 wt%; beyond this concentration, the increasing viscosity of the fiber-laden mixture precluded complete mold filling under the manual open-cast conditions used in the present study. Within the range examined, E+3%RF is the recommended formulation for reproducible hand-pour canine bone model fabrication. Although E+5%RF achieved a higher elastic modulus (1,513.12 ± 67.85 MPa vs. 1,333.06 ± 18.42 MPa), reproducibility of that property across fabricated specimens is a practical requirement for standardized surgical training models [10]. On this criterion, E+3%RF is superior because its coefficient of variation for elastic modulus was 1.4%, the lowest recorded across all nine groups, compared with 4.5% for E+5%RF. The reliability of the E+5%RF estimate is further reduced by the loss of one specimen because of pre-test grip failure, yielding n = 4, the smallest group in the study, which limits confidence in both the mean value and the coefficient of variation at that concentration. Both E+3%RF and E+5%RF reached only 17.8% and 20.2%, respectively, of the lower bound of the canine cortical bone target (7,500 MPa); therefore, a 2.4 percentage-point difference at this scale of approximation does not constitute a meaningful biomimetic advantage for either formulation. E+3%RF therefore offers the best available balance of mechanical performance, specimen reproducibility, and casting workability within the hand-pour open-cast format evaluated in the present study.

Evaluation of PU foam as a cancellous bone analog

The PU foam used as the cancellous bone analog had a manufacturer-specified free-rise density of 27–31 kg/m³. The mold-constrained apparent density, measured gravimetrically from dumbbell specimens in the present study, was 280 ± 4 kg/m³ (0.280 g/cm³), approximately one order of magnitude higher than the free-rise specification, consistent with physical compression of the expanding cellular structure when expansion is resisted by the mold walls. This value lies at the lower boundary of the 0.16–0.64 g/cm³ (160–640 kg/m³) range typical of validated rigid PU cancellous bone analogs used in orthopedic simulation [10], and the modulus overlap therefore corresponds to the lower end of the trabecular density spectrum. The measured elastic modulus of 57.72 ± 7.69 MPa falls within the lower portion of the reported range for canine trabecular bone (approximately 50–14,000 MPa depending on anatomical site and apparent density [33, 34, 35]), a span that reflects the strong dependence of cancellous mechanical properties on apparent density [8]. This mechanical correspondence should be interpreted with caution because it is more consistent with low apparent density or osteoporotic-equivalent trabecular tissue than with normal canine cancellous bone. The foam was selected because of its commercial availability in Thailand, controllable expansion properties, and established use in orthopedic simulation [13, 14], and its elastic modulus is sufficiently distinct from all cortical composite groups to produce a mechanically differentiated bilayer construct, the primary functional requirement at this stage of model development.

Development of a hybrid bilayer canine CBM

The mechanical profiles characterized in the present study provide the material basis for a proposed hybrid bilayer CBM combining a fiber-reinforced epoxy cortical shell with a PU foam cancellous core. The two components differ by more than one order of magnitude (23-fold) in elastic modulus, with E+3%RF exhibiting an elastic modulus of 1,333.06 ± 18.42 MPa and PU foam exhibiting an elastic modulus of 57.72 ± 7.69 MPa. This qualitative difference reproduces the structural hierarchy of natural bone, in which a stiff cortical envelope encloses a compliant trabecular interior [7, 8], although both values remain below the target ranges for canine cortical bone (7,500–21,000 MPa) and mid-range trabecular bone (approximately 100–800 MPa [33, 34, 35]), respectively. Exact mechanical equivalence to natural tissue is not required for effective procedural training; synthetic models with simplified mechanical properties have supported valid skill acquisition in veterinary surgical education [5, 6, 46]. This modulus differential is nevertheless a necessary condition for a bilayer surgical training model because it governs the tactile transition encountered when a drill or implant passes from cortical to cancellous tissue [9], although whether this magnitude is sufficient cannot be confirmed from tensile coupon data alone and requires functional testing. The low coefficient of variation for the elastic modulus of E+3%RF (1.4%) suggests that the cortical component can be fabricated with sufficient batch-to-batch consistency for standardized model production using a hand-pour process, although this inference requires confirmation in a three-dimensional mold geometry. Unlike monolithic synthetic bone constructs or human-focused commercial analogs [9, 10], a hand-pour epoxy-PU foam bilayer platform is, in principle, scalable to canine-specific mold geometries without specialized equipment, making it a candidate fabrication route for low-resource veterinary training settings. However, this scalability has not yet been demonstrated in an integrated construct.

Cost-effectiveness of the proposed canine bone model

The estimated raw material cost for the proposed in-house canine femur model, comprising a 100 g E+3%RF cortical shell and a PU foam cancellous core, is approximately USD 1.90 per unit, representing approximately 7% of the USD 28.25 manufacturer’s suggested retail price of the equivalent Sawbones composite canine femur (Table 3[42]). The projected raw material saving per 10-model training set is USD 263 before international shipping and import costs. E (EP-089) and R (R804) were available in Thailand at USD 19.01/kg and USD 7.56/kg, respectively, and all materials required for fabrication are obtainable without specialized procurement. Cost-effective, locally fabricated training models are increasingly recognized as valuable tools for veterinary and surgical skills education, particularly in resource-limited settings [43–45]. Pink_ et al._ [46] demonstrated a comparable hand-pour approach using PU foam coated with E fabricated from locally sourced materials in a lower-income setting, with drilling and K-wire performance acceptable for procedural training. The present study builds on that work by systematically characterizing the material mechanics and introducing fiber reinforcement to improve cortical stiffness, thereby providing a reproducible material baseline that informal fabrication approaches have lacked.

A. IN-HOUSE MODEL (E+3%RF + PU foam) |  | 
Component | Amount | Cost (USD)
Epoxy resin EP-089 (cortical shell) | 97 g | 1.84
Chopped E-glass fiber, 3 wt % (cortical shell) | 3 g | 0.05
PU foam, 2-component (cancellous core) | ~0.5 mL | 0.01
Raw material cost per model |  | ~1.90
B. COMMERCIAL REFERENCE† |  | 
Model (SKU‡) |  | Price (USD)
Canine Femur, Foam Cortical Shell, Medium (2121) |  | 28.25
Canine Femur, Short Oblique Fracture (2121-24) |  | 35.5
Canine Femur, Long Oblique Fracture (2121-31) |  | 36.75
C. COST COMPARISON vs. SKU 2121 |  | 
Absolute saving per model |  | 26.35
In-house cost as % of commercial price |  | ~7%
Saving per 10-model training set |  | ~263

Limitations and future directions

Mechanical characterization was restricted to uniaxial tensile testing under quasi-static loading in accordance with ASTM D638, providing a reproducible basis for comparing material compositions but not a complete mechanical profile. Canine long bones are predominantly subjected to compression, bending, and torsion during physiological loading [47]; however, none of these loading modalities were evaluated in the present study. Fiber orientation in the hand-poured specimens was not characterized at the microstructural level. Random fiber distribution, the expected outcome of manual mixing and open casting, introduces anisotropy that cannot be resolved using tensile coupon data alone. Fatigue, cyclic, and torsional testing were not performed. Screw pull-out resistance, drilling behavior, plate fixation mechanics, and the effects of humidity exposure and sterilization on composite integrity also remain unquantified. A minimum of five specimens per group was used in accordance with ASTM D638. Although this satisfies the standard's minimum requirement for materials characterization, it represents the lower boundary of statistical power for detecting intergroup differences, and the findings should therefore be interpreted accordingly. The E+5%RF group was further reduced to n = 4 following pre-test grip failure of one specimen, thereby limiting confidence in both the mean value and the coefficient of variation at that concentration.

Three priorities emerge from these limitations. The immediate priority is functional validation of E+3%RF under loading conditions relevant to surgical training, specifically compressive testing, drilling resistance, and screw pull-out testing, because whether the remaining cortical stiffness gap represents a practical limitation depends on material behavior under drilling and implant loading rather than under uniaxial tensile loading alone. If functional testing confirms adequate performance, the hand-pour epoxy platform will be suitable for deployment without additional material reformulation. If functional testing does not confirm adequate performance, reducing the stiffness gap will require continuous or woven carbon fiber reinforcement or aligned fiber lay-up to improve fiber-matrix load transfer [15, 30, 31], because the short-fiber hand-pour format cannot achieve the 7.5–21 GPa elastic modulus of natural canine cortical bone [7] under its current design constraints. Optimization of the cancellous core toward the 0.16–0.32 g/cm³ apparent density range of validated rigid PU cancellous bone analogs, and toward breed- and site-specific mechanical targets, is most practically achieved using a three-dimensional printed lattice geometry, which would decouple apparent density from mold constraint and permit direct alignment with published canine trabecular bone properties [33–35]. Once a construct meeting the functional requirements for surgical training has been fabricated and validated, a blinded comparative study evaluating fracture stabilization performance using synthetic and cadaveric specimens, assessed by expert raters, would determine whether the proposed platform provides training outcomes sufficient to support its recommendation as a cadaver-reduction tool in veterinary surgical curricula.