Cancellous bone
FFF technology was chosen due to its wide range of materials and low-cost end devices. Internal structures in the FFF process do not have to be manually designed in contrast to stereolithography, selective laser melting or laser sintering. Instead, different structures as well as the amount of infill can be configured during software-assisted pre-processing.
DA for the manufactured cube-shaped specimen ranges from roughly 1 to 3. Lowest values were observed for the specimen with 40% infill. Augat et al. [26] investigated the anisotropy of human cancellous bone by testing cubes extracted from different bone regions in several spatial directions. Tested specimen from the proximal femur showed a DA of 0.8 and 6.2 and a mean value of 2.2. Comparable results were provided by Goulet et al. [27], where DAranged from 1.1 to 2.5. Thus, the anisotropy of the artificial models corresponds to human bone very well.
All cylindrical specimens have a lower compressive modulus compared to human bone. This is to be expected, due to the large discrepancy between the moduli of the used plastics and that of bone tissue of about 18 GPa or more [24, 28, 29].
When testing human cancellous bone in compression, the stress-strain curve shows a maximum stress at initial failure, which immediately decreases to about 2/3 of the maximum. At higher strains, bone is characterized by a periodic progression of increasing and decreasing stresses around a roughly constant plateau stress, until the specimen becomes compacted, resulting in a rapid stress increase [30]. Typically, the plateau stress is less than or equal to the maximum stress depending on anatomical site and bone density [21, 31]. With increasing relative density compaction occurs at lower strains and thus may be noticeable by a sharp increase in plateau stress. In this sense, the discrepancy between failure stress σy and σp is included for the selection of suitable infills.
The absence of maximum stress at initial failure of the synthetic specimens may be due to the viscoelastic material behavior of the used thermoplastic materials. The material behavior could additionally be caused by the generated gyroid structure being a pure shell structure. In contrast, the bone architecture is a cellular structure, containing both platelike and rodlike struts, depending on the anatomical position [32]. The absence of rods in the gyroid structure could have an effect on the mechanical behavior. This could be one cause of major difference in mechanical behavior of human bone and gyroid structures made of plastic.
Silva et al. [33] investigated the influence of the gyroid structure on the mechanical stability of additively manufactured components in different loading directions (tension, compression, bending, impact strength). Among other things, the influence of the infill density on the mechanical load-bearing capacity of cylindrical specimens with and without an outer wall was analyzed. For PLA specimens with 50% infill σmax/ σy is about 10 MPa and E about 800 MPa [33]. In the present study, cube-shaped and cylindrical specimens made of PLA were tested. Compared with the data of Silva et al., the cube-shaped specimens in the z-direction show a higher average σmax (11.8 ± 3.3 MPa) and E (1084.1 ± 237.4 MPa). In contrast, the cylindrical specimens have both lower σy (8.4 ± 0.3 MPa) and E (427.9 ± 17.6 MPa) than the data of Silva et al.
Applying the gyroid structure as a cancellous bone substitute holds great potential. In addition, there is the possibility of transferring the modelling approach to other body regions since the manufacturing settings can be precisely adjusted and the mechanical properties of the object show good reproducibility. Most of the tested specimen have equivalent σy compared to human bone (s. Fig. 5). For example, mechanical properties of PMMA compression specimens with 30 and 40% infill show very good agreement with cancellous bone specimens from the lumbar spine [34]. In addition, cancellous bone in vertebral bodies exhibits nearly orthotropic material behavior, as do the generated specimens.
The selection of suitable filling grades is based, on the one hand, on the ratio of σp to σy and, on the other hand, on the highest possible modulus. With the exception of PC, applying 40% infill was chosen. For the PC 35% infill was used.
Biomechanical testing
Removing the femoral head with a saw and preparing the medullary canal provided a subjective evaluation of the human bones. The specimens made of PC, PMMA and ABS could be instrumented better than the model made of PLA due to their higher thermal resistance.
Considering the local force maxima over subsidence (s. Fig. 5), all artificial models are within the range of the human specimens. All specimens show a nonlinearly increasing force peaks with increasing implant subsidence. Specimen HB2 exhibits the steepest increase in the force curve and reaches the highest measured insertion force of 8.5 kN without visible fractures. It can be assumed that this specimen would have reached even higher forces but the test was stopped at this value to protect the load cell.
An explanation for the low values of the ABS specimen may be the low bonding forces between the deposited layers of the additively manufactured core. As a result, the components exhibit significantly lower mechanical properties in the x/y-direction than in the z-direction. Consequently, the subsidence of the prosthesis into the bone has increased and the occurring loads are comparably low over time. Both force and displacement of the ABS model show very good agreement with the specimen HB1 during testing.
The actual forces occuring during the insertion of a femoral stem, as well as the forces leading to calcar fractures are poorly investigated. Some studies report proximal femoral fractures were caused by pressing in an oversized stem using a testing machine, but do not state the required forces [35, 36]. Sakai et al. [37] reported hammering forces of 9.25 kN in their in vitro biomechanical study using artificial bones. However, by using a very rigid experimental setup, without damping properties of human tissue, these forces seem to be too high.
To the authors’ knowledge, the only comparable data were published by Carls et al. [38]. They also pressed femoral stems into human femora utilizing a testing machine and recorded the force progression and subsidence until failure. Their results, similar to those in the present study, show a wide range of failure forces between 1.9 kN and 9.3 kN or subsidences between 2.0 mm and 19.1 mm.
An equivalent increase in surface strain was generally recorded on the inner surface of the femoral stem during insertion. The measured strains are up to 0.14%. An exception is specimen HB1, where inverted strains were observed. The inversion of strain into a local compression could be explained by a lateral fracture of the bone. A crack in the bone causes the femur to bend up radially in cross-section, resulting in local compressive loads on the bone surface. As long as the bone is intact, increasing insertion force leads to an increase of the local strain.
Limitations
Evaluating the anisotropy of the gyroid structure was only performed on specimen made of PLA. It is to be expected that the DA varies among different materials. Based on the collected data, it is assumed that all materials have higher mechanical properties in the z-direction than transversely and that DA varies only in magnitude.
The inhomogeneity of cancellous bone was simulated by two zones of different infill density. In real bone, this homogeneity is much more complex. Nevertheless, the use of AM, especially FFF, allows a much higher variability of mechanical properties compared to PU foams. Another limitation is the small number of specimen used in the biomechanical testing. Although the data do not allow statistically relevant conclusions, the insertion tests still provide information that the model approach is working.
Another issue regarding the manufactured cubic and cylindrical specimen is the ratio of pore size and specimen size. The mechanical response of porous structures loaded in uniaxial compression are dependend on the ration of specimen diameter and pore size. Tekog̃lu et al. [39] showed, that mechanical properties decrease with decreasing ratio of specimen diameter and pore size. Since the mechanical properties of many additively manufactured specimens were lower than the human comparative data, it can be assumed that a wide range of materials and filling grades might be applicable for mimicking human cancellous bone. A reliable parameterization regarding human cancellous bone properties requires more targeted investigations due to the above-mentioned systematic errors.
Manufacturing components using the hand lay-up technique is simple and inexpensive, but it also involves a high probability of defects such as cavities (Fig. 3, top right). The stiffness-reducing influence of such defects cannot be ruled out in the conducted experiments. It can also be assumed that the transversal mechanical stability of the glass fiber laminate does not exceed the stiffness of the polymer matrix (3.2 GPa). Human femoral bone, on the other hand, has a stiffness of 11.5 Gpa [40]. This could be an explanation for the lower insertion forces of the artificial bones. In this study, the glass fibers were intentionally oriented only longitudinally to provoke tearing (periprosthetic fracture) of the stem during implantation, as this is a common complication [41]. An optimization of the glass fiber laminate would be necessary for improving the artificial model in future.