Deformation and Durability of Soft Three-Dimensional-Printed Polycarbonate Urethane Porous Membranes for Potential Use in Pelvic Organ Prolapse

Pelvic organ prolapse (POP) is a condition in which loss of support to the vagina causes the pelvic organs supported by it to descend from their natural position into the vagina and often through the vaginal introitus. POP results in a feeling of a bulge and organ dysfunction resulting in incontinence and sexual dysfunction [1]. Surgical repair of POP can be performed with native tissues, however, high failure rates of these procedures [2] have prompted surgeons and their patients to turn to biomaterials, most often polypropylene mesh. The mesh is used to bear the physiologic mechanical load and to return the vagina and the pelvic organs to their normal anatomical position. However, use of synthetic mesh has been found to have a high rate of complication, up to 15% in particular procedures [3]. Two of the most common complications are mesh exposure through the vagina and pain.

Currently, it is widely accepted that mesh performance is significantly affected by critical mesh properties such as mechanical behavior [4,5], mesh weight [6], pore size [7], and fiber dimension [8]. Importantly, mechanical behavior, mesh weight, and porosity have been shown to be strongly correlated properties [9]. Higher mesh weight tends to elicit a more pronounced inflammatory response while a lighter mesh with thinner fibers and larger pore size tends to have better tissue incorporation [5]. Pore size and porosity are generally thought to have the highest impact in terms of host response [8]. As pore size increases, bridging fibrosis decreases due to the increased distance between fibers [8]. This reduces the chance of a continuous fibrotic network building between the fibers, a complication that is associated with pain. A minimum of 1 mm pore diameter is often accepted as a critical size for tissue ingrowth [10]. Since an implanted mesh will be placed under load, there is also a need for a stable pore geometry under tensile or multi-axial loading. Studies have shown that the application of tension within the expected physiological range often results in 70–90% of the pores collapsing to <1 mm [9,10] especially when loaded in uniaxial tension. The use of geometries such as reentrant auxetic structures has been suggested as a potential solution as materials with auxetic pores demonstrate an increase in pore size under loading [11]. In terms of filament type, monofilament fibers were found to elicit a better host response compared to multifilament fibers [12,13].

Currently, the most commonly used synthetic meshes are knitted polypropylene (PP) meshes. Although high mesh stiffness was believed to correlate with a negative host response, there are few studies on the effects of the mesh material's mechanical properties. PP has an elastic modulus (E) of about 3 GPa, three orders of magnitude higher than anatomical vaginal tissue (stiffness = 2.5–34 MPa) [14–17]. Although there is a large variation in the literature on measurements of the mechanical properties of vaginal tissue [14], the tissue-mesh discrepancy is strong compared to all measurements. Significant evidence has indicated that the mechanical properties of a material strongly impact implant performance. Highly rigid implants cause high shear stress at their interface with the body, micromotion [18,19], and stress-shielding [20], all correlated with mesh complications. Therefore, there is substantial room for improvement by utilizing a softer material that matches the mechanical properties of the host tissue. Attempts at fabricating a mesh with varied absorbable synthetic materials include the use of knitted polyether-ether-ketone and polyamide [21], electrospun polyurethane [22], and polylactic acid [23]. To our knowledge, there have been very few published effort toward a permanent biocompatible, elastomeric, monofilament device for surgeries to repair pelvic organ prolapse [24].

We propose a three-dimensional (3D)-printed polycarbonate-urethane (PCU) membrane as a potential solution to overcome the limitations of polypropylene mesh. PCU is a subset of thermoplastic polyurethanes (TPU) and has recently found many uses as a soft-tissue replacement device because of its biocompatibility, biostability, and stiffness that more closely resembles many soft tissues [25–28]. Similar to other TPUs, PCU is composed of hard and soft segments that consist of diisocyanate and macrodiols, respectively. In the case of PCU, the macrodiol is polycarbonate. Driven by thermodynamic favorability, the hard segments aggregate into domains that act as a physical crosslink, lending TPUs their thermoplasticity and highly elastomeric behavior [29]. In addition, the mechanical properties of PCU can be tuned by changing the ratio of hard and soft domains, enabling a wider solution space in designing devices [28]. As is common with thermoplastics, the mechanical properties of PCU will depend strongly on the temperature and environment at which it is tested [30].

In recent years, 3D-printing has found numerous applications in biomedical implants. 3D-printing has several important advantages over traditional fabrication methods, such as capability for personalized solution, rapid prototyping, and fabrication of complex shapes that are otherwise impractical or unachievable with traditional approaches. The last aspect is particularly pertinent since there is often a need for intricate and porous structures such as meshes or nonwoven interconnected porous structures in medical implants [31,32]. In the case of polypropylene meshes, these devices are knitted or woven after polypropylene is extruded into thin fibers to decrease device stiffness. As a soft polymer that is compatible with the material properties of the vagina, PCU can be 3D printed. Fused deposition modeling (FDM) is the most commonly used 3D-printing approach owing to its simplicity, low-cost, and wide material possibilities. In FDM, a thermoplastic polymer such as PCU is melted and extruded through a small nozzle. The nozzle is mounted on an x–y gantry which traces the cross section of the desired object, thus building it in a layer-by-layer fashion. In this study, FDM will be used to fabricate PCU membranes with varying architectures using polymers with different base (i.e., material) stiffness values.

Considering the repetitive physiological loads supported by an implanted device, there is a need to evaluate its capability in enduring cyclic stress to avoid premature failure. A fatigue study is particularly pertinent here since the printed membrane can be considered an architected structure containing pores and notches that can act as stress concentration points, conceivably accelerating fatigue failure. Although studies on the high-cycle fatigue properties of 3D-printed architected PCU exist, recent review articles [33,34] reveal a dearth of fatigue studies on 3D-printed architected polymer structures as compared to studies on architected metal structures [35]. Here, we are specifically interested in studies on the fatigue response of an architected mesh-like construct fabricated out of 3D-printed PCU, of which there is presently none.

In this study, we present a novel elastomeric, 3D-printed membrane intended for potential use in pelvic organ prolapse repair applications. We have utilized the term membranes as opposed to mesh as it better suits the characteristics of these 3D-printed constructs. Two types of membranes were studied: square-pore membranes and an auxetic bowtie-pore membrane. In addition, a commercial polypropylene mesh was included for comparison. Membranes with different pore sizes, pore shapes, membrane thickness, and membrane fiber dimensions were mechanically characterized via monotonic tensile and tensile–tensile fatigue tests. Changes in membrane pore dimension under tensile load were also characterized. All mechanical characterizations were performed in water conditioned to 37 °C to simulate body temperature. The deformation and durability studies here serve as a template for consideration of other novel mesh or membrane materials. Although other materials have been proposed in the literature, it is critical to evaluate fatigue properties of candidate materials and structures to truly demonstrate potential for commercial success.

Carbothanes AC-4075A (shore hardness: 75A), AC-4085A (85A), and AC-4095A (95A) were obtained from Lubrizol (Cleveland, OH) in pellet form. Before all processing steps, the pellets were dried in a vacuum oven at 95 °C and –25 inHg for a minimum of 2 h. PCU pellets were first processed into 1.75 mm diameter 3D-printing filaments using a filament extruder (Filabot EX2, Barre, VT). Filaments were kept in a box with moisture absorbers. Before 3D printing, the filaments were dried again in a vacuum oven at 95 °C, –25 inHg vacuum for a minimum of 2 h.

All FDM samples were printed using a Prusa MK3S (Prusa Research sro., Czech Republic). 95A samples were printed using stock extruder, while 85A and 75A samples were printed using a specialized flexible filament extruder (Flexion, Diabase Engineering). Simplify3D was used to process CAD files into G-code files used for printing. Both tensile and fatigue samples were printed laying down on their largest surface area to improve printing quality. 95A, 85A, and 75A were printed using 240 °C, 260 °C, and 260 °C nozzle temperature, respectively. Printing speeds were set to 4 mm/s. A 0.25 mm diameter nozzle was used to print membranes with 0.3 mm strut width, a 0.4 mm nozzle for membranes with 0.5 mm strut width, and a 0.6 mm nozzle for membranes with 0.7 mm strut width. The complete set of printing parameters is shown in Supplemental Table 1 available in the Supplemental Materials on the ASME Digital Collection. Due to the nature of one-layer printing used to fabricate the membranes, we found that it was necessary to ensure adequate levelness of the print bed to maintain membrane consistency. This is done by physically leveling the bed and measuring its levelness using a distance sensor equipped with the printer. Bed level was measured at 7 × 7 points equally spaced over the print bed, totaling 49 points. A minimum of 0.10 mm variance in level was imposed before printing membranes. All membranes were printed with a target thickness of 0.5 mm. All printed samples were rested for at least 1 week before testing, in accordance with findings from our previous study [28].

Microcomputed-tomography (micro-CT) was performed to evaluate the internal structure of the materials. Micro-CT analysis was carried out on a XTH 225 ST (Nikon) at 110 kV/80 μA. Exposure time was set to 708 ms and two frames were averaged for each image. Voxel size was maintained at 14.9 μm. CT-data was processed into cross section using CT Agent (Nikon) and reconstructed using CT Pro 3D (Nikon). 3D rendering and analysis of micro-CT scans were carried out using Avizo (FEI Visualization Sciences Group).

Tensile tests were performed to evaluate the mechanical properties of bulk PCU and PCU membranes. Test samples for bulk PCU properties were in the shape of a Type V ASTM D638 dogbone. All dogbones were printed with a 0.4 mm nozzle. Three PCU grades were tested: 95A, 85A, and 75A, in decreasing hard segment content, respectively. Tests were carried out under the guidance of ASTM D638. Tests were performed using TR830 (Test Resources) with a 100 lb. load cell. Samples were tested in a water bath heated to 37 °C. The displacement rate was set to 10 mm/min. Fiduciary markers were placed on the gauge of the dogbone and tracked using a video extensometer to calculate strain. The tracking capability of the extensometer was found to be reliable to only about 80% strain. Therefore, stress–strain curves based on extensometer data are presented up to 80% strain. A complete stress–strain curve showing a point of failure is shown based on nominal strain instead, where $εnominal=crosshead displacementinitial grip−to−grip distance$⁠. For dogbone samples of each PCU grade, the properties of dogbones printed using different raster angles were studied. In addition to testing in a water bath, dogbones were also tested in ambient environment for comparison. A minimum of three samples were used for each test group. Modulus was calculated from the stress–strain curve by taking the tangent of the curve at 10% strain.

Polycarbonate-urethane membranes were tested on a similar setup. Displacement rate was set to 50 mm/min. Only 95A and 75A PCU membranes were fabricated. Two different pore shapes were tested: square and reentrant auxetic “bowtie.” Three different strut sizes were tested: 0.3 mm, 0.5 mm, and 0.7 mm. All membranes were set to 0.5 mm thickness and 15 mm wide. The total porosity of every group was kept constant at approximately 66%. Tensile test results for membranes are represented as load/width versus relative elongation curves. Load/width represents the load that membranes were subjected to during testing divided by its width (resulting in unit of N/cm). Relative elongation (RE) represents the elongation of the membranes divided by their original length.

During tensile testing, changes in membrane pore size were tracked with an optical camera at 20 frames per minute. The frames were then processed with a matlab script (See Supplemental Figure 1 available in the Supplemental Materials on the ASME Digital Collection) over the course of the test until specimen failure. Digital processing was used to determine pore area, major axis length (axis along tensile load direction), and minor axis length (axis perpendicular to tensile load direction).

Fatigue tests were performed with a TR830 test frame (test resources) equipped with a 100 lb. load cell. All tests were performed in a load-controlled manner. Loads were prescribed sinusoidally at a frequency of 0.5 Hz and stress ratio of R = 0.4 (R = Load_max/ Load_min) until sample failure or runout (see Supplemental Figure 2 available in the Supplemental Materials on the ASME Digital Collection). Runout is prescribed at 10⁵ cycles due to time constraint. A minimum of three samples for each load level and a minimum of two runout samples were tested. Load level is reported as load/membrane width (resulting in unit of N/cm). Fatigue data is presented in the form of a pseudo-“S-N” plot in terms of peak load/sample width (N/cm) versus cycles at failure. A similar plot with normalized peak load versus cycles at failure is also presented. Normalized load is a unitless quantity representing load relative to the mean failure load of the sample type (load/width divided by failure load/width).

Statically loading the samples for creep testing was performed with the same setup with relative elongation of the sample tracked throughout the duration of the test. For intermittent testing, samples were subjected to 50 cycles of the assigned load level and then rested for 10 min at 0 N. This pattern is repeated until sample failure.

Statistical analysis was performed using one-way analysis of variance and Tukey's posthoc test to determine statistical significance of different groups in this study. Statistical significance is set at p ≤ 0.05 unless otherwise stated. All statistical analysis was performed using jmp software package (SAS).

Monotonic tensile tests were performed to assess the mechanical properties of 3D-printed PCU dogbones immersed in water at 37 °C. Samples were printed with raster lines parallel to tensile testing direction unless otherwise mentioned. The resulting representative stress–strain curves based on extensometer data are shown in Figs. 1(a)–1(c) depicting a sample with the median failure stress of a group. Due to the limited range of the extensometer, an additional plot of stress versus nominal strain up to sample failure is provided by Figs. 1(d)–1(f). Nominal strain is strain based on crosshead displacement of the tensile test frame: $εnominal=crosshead displacementinitial grip−to−grip distance$⁠.

All tested samples exhibited elastomeric loading behavior, showing an initial toe region followed by a slight plateau and ending with a steeper region before failure. The bulk modulus of all PCU grades evaluated in a 37 °C water bath testing environment was within range (or within one order of magnitude) of vaginal tissue demonstrating the softness and tunability of PCU stiffness to match the stiffness of vaginal tissue, which can vary with aging, pregnancy, and prolapse (Table 1). Immersion of the PCU dogbones in a 37 °C water bath elicited significant changes in the mechanical property of all PCU grades. All three grades of water-bath tested PCU were more compliant compared to samples tested in air at ambient temperature. The average modulus of PCU-95A, PCU-85A, and PCU-75A decreased by an average of 40.7%, 44.2%, and 8.9%, respectively, indicating a greater change for PCU grades with a higher hard-segment content. Along with this, there was an increase in failure strain accompanied by a decrease in failure stress for all grades. Among all grades, failure stress and failure strain in water bath at 37 °C were found to be similar. The significant change in mechanical property after immersion in a heated water bath suggests that for an accurate evaluation of a PCU-based implant, mechanical tests should always be performed at physiological temperatures. However, a limitation of this study is that the tested samples used here were not conditioned in water for a prolonged period before testing. Past studies have shown that conditioning in water for up to 5 months can noticeably affect mechanical properties, albeit not in a catastrophic manner [36].

The effects of raster angle on mechanical properties under uniaxial tensile load were examined by testing 2 groups of dogbones, one printed with a raster line alternating between 0 deg and 90 deg (i.e., along, and perpendicular to tensile testing axis) and the other 45 deg and 135 deg. Past studies have shown a significant change in mechanical properties for samples with differing raster angles, resulting in higher decrease of failure stress and failure strain for raster angles that are closer to perpendicular to tensile load. Conversely, raster angles aligned with tensile load exhibit improved failure stress and strain [37,38]. From Fig. 1 it is apparent that raster angle does not affect modulus in any appreciable manner. A summary of the modulus of each group is shown in Fig. 2(c). Figures 1(d)–1(f) exhibits stress-nominal strain curves of samples with median failure stress value for each group. Figures 2(a) and 2(b) summarizes failure stress and failure nominal strain for dogbones printed with different raster angles, respectively. PCU 95A and 85A in particular were observed to be more affected by variation in raster angle compared to 75A, both exhibiting a noticeable decrease in failure stress (p < 0.0001) while 75A showed minimal change. This suggests that PCU with higher hard-segment content has an increased sensitivity to print raster pattern for solid samples. Past studies have shown a similar trend wherein harder polymers tend to be more sensitive to raster patterns [37,39]. We propose two possible causes:

1. High hard-segment content slows molecular diffusion since the PCU hard-segment tends to possess a stiffer chain and higher steric hindrance. In addition, the hard segment has a much higher tendency to crystallize than soft segments [28]. Crystallization has been shown to hinder molecular diffusion [40]. These factors can limit layer bonding which strongly depends on molecular diffusion and the resulting entanglement [41,42]. This difficulty in bonding manifests in a more raster-sensitive specimen.

2. A higher notch sensitivity for high hard-segment content PCU. The layering process during 3D-printing naturally creates notches and voids inside the test samples, which could give rise to stress concentrations. Since harder materials tend to have higher notch sensitivity [43,44], this could result in a lower failure stress for harder PCU. No change in modulus due to raster angle was observed for any PCU grades.

The observed decrease in mechanical properties due to differences in raster angle indicates that a 3D-printed PCU structure would benefit from an optimized design and print path planning, particularly for PCU grades with high hard-segment content.

Two types of membrane samples were 3D-printed from PCU: a reentrant auxetic structure (bowtie) and a repeating square structure. Figure 3(a) shows the CAD model for both membrane types used for 3D printing input. As elaborated in the Methods section, three strut sizes were printed for each type of membrane: 0.3 mm, 0.5 mm, and 0.7 mm. Strut size is defined as the width of the struts composing the membrane. Thickness was maintained at 0.5 mm for all strut sizes. To achieve the desired 0.5 mm thickness, membranes with 0.3 mm wide struts were printed using a 0.25 mm nozzle in two layers of 0.25 mm thickness each, totaling 0.5 mm. This was done since it is not possible to directly print 0.5 mm thickness using a 0.25 mm print nozzle. 0.5 mm and 0.7 mm membranes were printed directly in one layer using a 0.4 mm and a 0.6 mm-diameter nozzle, respectively.

To evaluate the resulting elastomeric PCU membranes, micro-CT scans were performed. Figure 3(b) shows the reconstructed CT data. The micro-CT dimensional measurements tabulated in Table 2 were measured at three midstrut locations and show that the printed membranes were close to the desired size.

A feature of interest is the lump seen on the top layer of the square membranes at the intersection of fibers, particularly prominent in 0.5 mm and 0.7 mm membranes (marked by a dashed circle). The square membranes were printed by first laying all the horizontal fibers and then layering the vertical fibers (Supplemental Figure 3 available in the Supplemental Materials on the ASME Digital Collection). During the deposition process of the vertical top layer, the extruded molten polymer is hindered by the bottom layer, causing coalescence of the molten polymer at the intersection. This feature of the membrane gives rise to a disparity in the structural properties of the square membrane's two perpendicular axes even though the square membrane was designed to look identical when rotated by 90 deg. This discrepancy is elaborated upon in Sec. 3.3 on structural properties. Bowtie membranes were printed with a different pathing than square membranes (Supplemental Figure 3 available in the Supplemental Materials on the ASME Digital Collection), resulting in joints at certain locations within the membrane. Similar to the raster angle effect discussed in Sec. 3.1, these joints are layers that have inferior strength to the bulk property of PCU and hence act as weak points. These joints are observed to considerably affect the structural properties of printed membranes as discussed in Sec. 3.3 on structural properties. This leads us to conclude that print pathing is a significant factor in the additive manufacture of single-layer planar constructs such as the printed membranes.

Membranes with varying strut sizes were characterized via monotonic uniaxial tensile testing to evaluate mechanical behavior. Both square and bowtie 95A PCU membranes with strut sizes of 0.3 mm, 0.5 mm, and 0.7 mm were tested. The membranes were each tested in three orientations, corresponding to an angle of 0 deg, 90 deg, and 45 deg. While in most prolapse procedures, meshes are subject to tensile loads, we appreciate that multi-axial loading is present and therefore, we sought to test the behavior of the membranes along multiple axes. Figure 4 shows a visualization of these loading orientations. As a benchmark, a commercial polypropylene knitted mesh (Restorelle) was also tested with the same protocol at 0 deg and 45 deg load orientations. For brevity, samples are referred to in the following format: PCU_Grade-Membrane_Type-Load_Orientation (e.g., 95A-bowtie-0 deg refers to 95A bowtie membrane under 0 deg loading orientation). Figure 5 shows the resulting representative load-relative elongation curves for each 95A PCU group. Figure 6 summarizes the important features of each group's load/width-RE curve. Load/width represents load the samples were subjected to divided by the width of the tested membrane.

Failure RE was noted to be similar throughout with a marked increase for 95A-bowtie-90 deg and a marked decrease for 95A-bowtie-45 deg. 95A-bowtie-90 deg exhibited significant unfolding movement, which appears to have been the cause of an increased failure RE. For 95A-bowtie-45 deg, weak points within the membranes that were revealed via micro-CT appear to be loaded at an unfavorable angle, causing a significant decrease in failure RE. All failures during uniaxial tensile testing for both membrane types were observed to occur at junctions, indicating that they are indeed a mechanically weak point.

Failure load is expressed in terms of force reading at membrane failure divided by membrane width (unit of N/cm). Failure load is much higher for square membranes compared to bowtie membranes. In addition, 95A-bowtie-90 deg and 95A-bowtie-45 deg were observed to have a noticeably lower failure load compared to 95A-bowtie-0 deg angle. The reason for these is twofold: (1) The bowtie structure tested is a bending-dominated structure. Bending-dominated structures lose stiffness rapidly along with a decrease in relative density [45,46] as opposed to stretching-dominated structures such as the square membranes. In other words, by the nature of the structure's shape, an equal porosity for both square and bowtie membrane would result in the bowtie membrane having fewer load bearing struts (i.e., struts aligned in the direction of the tensile load) compared to the square membrane. (2) Joints within bowtie membranes act as weak points, especially at certain loading orientations. This is in contrast with square membranes, which do not have these joints when tested with loading direction parallel to the struts. Loading orientations that subject weak points to tensile load will result in lower failure load. These joints are a consequence of print pathing and were examined closely in Sec. 3.2 via micro-CT.

Square membranes with a thicker strut were observed to possess a much higher failure load along with a higher stiffness. This is caused by a significant edge effect because of the small width of the sample (shown clearly in Supplemental Figure 4 available in the Supplemental Materials on the ASME Digital Collection). A wider sample will tend to average out this edge effect, resulting in a more similar load response for all strut sizes. Beyond this, strut size was not found to impact any property significantly.

For square membranes, a membrane that perfectly matches the printing model should have the same properties for both square-90 deg and square-0 deg by virtue of symmetry. However, as a result of the printing artifacts discussed in Sec. 3.2, a lower failure load and stiffness were observed for 95A-square-90 deg (loading parallel to second layer put down during printing) compared to 95A-square-0 deg (loading parallel to first layer put down during printing). This underscores the importance of print pathing and print quality, showing their ability to give rise to tangible effects on structural properties.

Bowtie membranes and 95A-square-45 deg were observed to be much more compliant than 95A-square-90 deg and 95A-square-0 deg. This is indicated by a lower load reading at 20% RE and to a lesser extent, 100% RE (Figs. 6(c) and 6(d)). The effect is more pronounced for bowtie membranes loaded at a 90° angle. Again, this is caused by the bending-dominated structure of the bowtie membrane [11,45]. At 100% RE, most of the bending action reorienting the struts has concluded, allowing the struts to be loaded by stretching. The lower compliance is more pronounced for bowtie membranes with larger strut size. A larger strut size means a larger pore size (i.e., empty space between struts) since membranes were maintained at an equal porosity. A larger pore size allows for more bending action, lowering the membrane's compliance.

Tensile tests were also performed for 0.5 mm 75A PCU membranes. Figure 7 shows a comparison between the load/width-RE curves of 95A PCU membrane and 75A PCU membrane. As expected, the 75A PCU membranes are much more compliant than 95A PCU membranes. This agrees with the dogbone tensile testing result shown in Sec. 3.1, wherein 75A dogbones were observed to be much more compliant than 95A dogbones. Beyond this, the results are similar to 95A membranes where the effect of loading orientation on failure load performance obeys the following order from highest to lowest: 0 deg, 90 deg, 45 deg, for both bowtie and square membrane. However, 75A square membrane exhibits a smaller discrepancy between 0 deg and 90 deg loading direction, indicating 75A membranes as having less defect sensitivity. 75A bowtie membrane also exhibits much higher compliance and failure RE under 90 deg and 45 deg loading compared to 0 deg loading, a property that is not exhibited by its 95A counterpart. Again, this can be attributed to the defect resistance of 75A membranes. This defect resistance is analogous to what is observed in 75A dogbone samples with differing raster angles. As discussed in Sec. 3.1 on monotonic tensile tests of PCU dogbones, we propose two factors as possible reasons for this: (1) improved layer bonding due to the dominance of soft segments and (2) less pronounced stress concentration effect on softer materials.

Samples tested for pore geometry change were 95A PCU membranes with 0.5 mm strut size. For brevity, samples are referred to in the following format: Membrane_Type- Load_Orientation (e.g., bowtie-0 deg refers to 95A BOWTIE membrane with 0.5 mm strut size and 0 deg loading orientation). Figure 8 shows representative images of the membranes at an early and late stage of monotonic tensile testing. Figure 9 shows change of pore shape over the course of testing for 0.5 mm square and bowtie membrane (see Method and Supplemental Figure 1 available in the Supplemental Materials on the ASME Digital Collection for measurement approach). Square-0 deg shows a marked increase in pore area. This increase is mainly driven by major axis elongation while minor axis dimension change is minimal. Conversely, square-45 deg loading shows a significant decrease in pore area driven by minor axis contraction. Major axis elongation for 45 deg loading was not shown on the plot due to difficulty acquiring accurate measurements because of the extremely narrow shape of the pores. Pores were visibly close as the membrane elongated, as shown in Fig. 8. Square-90 deg is assumed to exhibit similar pore-shape change to square-0 deg and, therefore, was not evaluated.

For bowtie-0 deg, there is a marked increase in pore area driven by major axis elongation with a slight minor axis increase that stabilizes after around 100% membrane elongation. For bowtie-90 deg, there is also an increase in pore area which is driven by major axis elongation. However, in this loading direction, minor axis length decreased significantly, indicating that the pore shape is narrowing which is confirmed by Fig. 8. For bowtie-45 deg, we observed a significantly reduced pore area, similar to square membranes. Major axis elongation for bowtie-45 deg was not shown in the plot due to difficulty in measuring the extremely narrow shape of the pores. A past study performed on bowtie membrane structures fabricated out of polydimethylsiloxane showed a similar result [47].

These pore shape changes are particularly important to the performance of the membrane when used as a pelvic prolapse implant. From a clinical point of view, larger pores and higher porosity minimize fibrosis, reduce inflammation, and encourage collagen deposition, which results in improved implant integration [8,10]. Explanted meshes from patients with severe complications have shown high mesh density surrounded by a fibrotic capsule which is consistent with mesh deformation [47]. The strong dependence on loading direction and pore shape change for both membrane types also demonstrates that performance of future meshes can be improved by strategically placing pores in a manner that minimizes their deformation when the mesh implant is loaded. This could potentially be designed by simulating loading during real use and placing the pores at every region of the mesh implant in accordance with the simulated load directions.

To evaluate the membrane's response to cyclic loading, uniaxial tensile fatigue tests were performed using 0.5 mm strut bowtie and square membranes. 95A and 75A PCU membranes were tested, each representing both extremes in PCU grade, which correlates to material stiffness. The tests were performed at 0 deg, 90 deg, and 45 deg load orientations (see Fig. 4 for diagram). As a benchmark, commercial polypropylene knitted mesh (Restorelle) was tested with the same protocol along 0 deg and 45 deg. 90 deg orientation is not tested for Restorelle since it is symmetrical with 0 deg (see Fig. 4 for diagram). Fatigue data is presented in the form of a pseudo-“S-N” plot in terms of both load/sample width (N/cm) versus cycles at failure and normalized load versus cycles at failure. Normalized load is a unitless quantity representing load relative to the mean failure load of the sample type (load/width divided by failure load/width). Normalized load Fig. 10 shows fatigue data for Restorelle. Figures 11 and 12 show fatigue data for square and bowtie 95A membranes, respectively. Figures 13 and 14 show fatigue data for square and bowtie 75A membranes, respectively. Arrows indicate the 100,000-cycle runout limit at which point the fatigue test was ended even if samples had not fractured.

Restorelle-0 deg and Restorelle-45 deg exhibit similar S–N curves for both maximum and normalized load, as shown by Figs. 10(a) and 10(b), respectively. Runout was attained at 6.67 N/cm load and about 0.38 normalized load for both test orientations. For 95A PCU membranes, different loading orientations resulted in differing fatigue behavior. For 95A-Square membranes shown in Fig. 11(a), Square-0 deg shows the best fatigue performance as indicated by the higher cycle-to-failure at equivalent load levels. This aligns with their monotonic tensile failure load trend, wherein square-0 deg had the highest failure load followed by square-90 deg and square-45 deg. According to the normalized plot (Fig. 11(b)), square-45 deg loading direction shows the best relative fatigue performance. At up to 100,000 cycles, no plateauing was observed for any of the fatigue curves, indicating no fatigue “limit” in this cyclic regime (1–100,000 cycles).

As shown in Fig. 12, bowtie-0 deg was observed to possess the best fatigue performance followed by bowtie-90 deg and bowtie-45 deg. Similar to the square membranes, this trend aligns well with monotonic tensile failure load data of bowtie membranes. However, when maximum load is normalized by failure load (Fig. 12(b)), bowtie-45 deg shows a superior relative fatigue performance, while bowtie-90 deg and bowtie-0 deg show similar normalized performance.

There appears to be a trend of bending-dominated structures having a superior normalized-load fatigue performance in terms of normalized load. Square-45 deg and bowtie structures are bending-dominated structures and exhibit better normalized-load fatigue performance compared to square-0 deg and square-90 deg, which are stretch-dominated. A past study performed on 3D-printed acrylonitrile butadiene styrene showed a similar result [48], wherein dogbone samples printed via FDM shows the best normalized-load fatigue performance for samples printed with an alternating ±45 deg raster angle, while those printed with a 0 deg raster angle (all fibers aligned with tensile load direction) showed the worst fatigue performance relative to their monotonic tensile strength. Similar to the results shown here, the study also showed ±45 deg samples performing worse in monotonic tension compared to 0 deg samples. The study proposed that the interlocking structures serve to counteract each pair of opposing fibers, thus decreasing the ease with which a crack propagates.

Overall, PCU 95A membranes reached runout at similar load levels to Restorelle. Restorelle-0 deg and Restorelle-90 deg reached runout at 6.67 N/cm. Square-0 deg and square-90 deg reached runout at 6.67 N/cm while Square-45 deg reached runout at 5 N/cm. Bowtie-0 deg and bowtie-90 deg had similar runout load levels to square membranes while bowtie-45 deg was weaker. Bowtie-0 deg reached runout at 6.67 N/cm, bowtie-90 deg reached runout at 5 N/cm, and bowtie-45 deg reached runout at 4.3 N/cm. This indicates that square PCU 95A membrane is likely to not catastrophically fail from cyclic loading during use and could be a viable replacement for commercial meshes depending on RE levels following the reconstruction which may be related to patient body mass index and stage of prolapse. Bowtie-45 deg warrants a closer investigation to ensure physiological load will not cause it to fail during human use.

In contrast with 95A square membranes, 75A square membranes were observed to fail similarly for all loading orientation, as shown in Fig. 13(a). This observation aligns with tensile failure load data for 75A membranes (Fig. 7) which also shows a smaller discrepancy between loading orientations compared to 95A membranes. Similar to 95A membranes, square-45 deg exhibits a superior normalized load performance compared to square-90 deg and square-0 deg (Fig. 13(b)). Overall, 75A square membrane appear to possess superior normalized-load performance to 95 square membranes.

In Fig. 14(a), it is shown that 75A bowtie membrane had the following order of performance in terms of load orientation: 0 deg, 90 deg, 45 deg, aligning well with static tensile test results. However, bowtie-45 deg exhibited a superior normalized-load performance compared to bowtie-90 deg, which in turn was superior to bowtie-0 deg. This observation again showed that the bending-dominated structures possess superior fatigue performance relative to their own monotonic strength. However, 75A bowtie membranes appeared to have a much flatter S–N curve.

At low cycle numbers, the normalized-load performance of 75A membrane is observed to exceed 1 for some samples, meaning the samples were subjected to peak cyclic load that is higher than its tensile failure load. During the initial stages of fatigue testing, the cyclic load is ramped up to the desired level cyclically instead of linearly, hence periodically exposing the sample to low loads in the process. This could have allowed the soft segments of 75A to relax, allowing it to withstand loads higher than its tensile failure load for low number of cycles.

Figure 15 shows peak RE at the end of each load cycle plotted against number of cycles for Restorelle-0 deg and 95A-square-0 deg. The curves exhibit RE accumulation under cyclic loading (i.e., ratchetting), showing increasing total RE over the course of the cyclic test. 95A PCU membranes under different loading orientations show similarly significant ratchetting (data not shown). 95A PCU membrane ratcheted significantly more than Restorelle, reaching up to 400% RE at failure while Restorelle failed at less than 100% RE. In addition, samples under high load were observed to fail at a significantly higher RE than samples under lower cyclic load. These observations are in line with past studies on ratcheting of polymer samples, showing ratcheting dependence on stress patterns and material properties [49,50]. Excessive increase in RE under cyclic loading (even without fracture) could cause a mesh implant “failure” since the mesh would be incapable of holding organs in their intended position. This observation emphasizes the need for future studies on fatigue of PCU or other membranes to be conducted with a similar load profile to physiological loading.

Figure 16 shows RE level of 95A-square-0 deg subjected to a static hold compared to one subjected to sinusoidal cyclic load to the same maximum load as the static hold. Static loading represents a pure creep case in which strain accumulation is termed creep. Cyclic loading represents a combination of fatigue and creep, in which strain accumulation is termed ratchetting. Cyclic loading was performed with R = 0.4 and an identical setup to the fatigue tests performed in Sec. 3.5. In Fig. 16(a), the samples were statically loaded at a constant 20 N/cm load and cyclically loaded at a peak load of 20 N/cm, while in Fig. 16(b), the samples were statically loaded at lower load level of 13.3 N/cm and cyclically loaded at a peak load of 13.3 N/cm. As shown in Fig. 16(a), counterintuitively, samples subjected to constant load exhibits a much longer time to failure (∼600 s versus ∼100 s) even though they are exposed to high loads for a longer duration than cyclically loaded samples. This effect is more pronounced with a lower load, as indicated by Fig. 16(b), showing that the statically loaded samples did not fail at 100,000 s, whereas the cyclically loaded samples typically failed at less than 5000 s. This indicates that the cyclically loaded samples indeed failed due to fatigue in addition to the observed creep. For polymer samples under tensile–tensile load-control cyclic loading such as this, past studies have modeled failure to be a consequence of both fatigue and creep damage working synergistically [51–53] as opposed to being two linearly summed independent damage sources. Statically loaded samples were also observed to fail at a higher failure RE (5.12 ± 0.07 for 20 N/cm load) compared to cyclically loaded samples (4.65 ± 0.22 for 20 N/cm load) with the difference likely caused by the previously mentioned creep and fatigue interaction.

95A-square-0 deg samples were loaded intermittently to study its effect on cycle-to-failure. Briefly, the samples were cyclically loaded for 50 cycles, then rested for 10 min at 0 N, and the process repeated until sample failure. 95A-square-0 deg samples were cyclically loaded at two load levels: 30–12 N and 25–10 N. Table 3 shows the resulting cycle-to-failure count for intermittently loaded samples and continuously loaded samples. There is an obvious discrepancy between the different loading patterns. Counterintuitively, continuously loaded samples failed at a higher cycle count compared to intermittently loaded samples. The effect appears to be more pronounced for lower load levels, as indicated by the higher discrepancy for 25–10 N load level compared to 30–12 N load level. A study by Ford et al. [54] on solid PCU samples observed a similar effect. In this study, continuously loaded samples failed at about 4× more cycles compared to intermittently loaded samples. The authors propose that there is a shift in phase separation in the area immediately surrounding the crack tip. The shift itself is attributed as the cause of hysteretic heating or cyclic strain at the zone ahead of the crack tip. Changes in tip microstructure due to loading have been corroborated in past studies. For example, a study by Scetta et al. [55] showed that compared to pristine samples there are indeed microstructural changes at the crack tip when a sample is cyclically strained. Other past studies have also established the importance of loading history toward the fatigue performance of polymeric samples, such as by comparing different multistep loading patterns before cyclic testing [53]. The discrepancy in continuous and intermittent sample shown here further emphasizes that although the simple sinusoidal-loading fatigue tests performed in Sec. 3.5 provide some preliminary perspective on the performance of PCU membranes, there is a need for cyclic testing to be performed using a loading pattern akin to physiological loading for a more accurate view of their performance.

In this study, we characterized the mechanical behavior of PCU of different grades in air and in a 37 °C water bath. In the 37 °C water bath, PCU exhibited a significantly more compliant mechanical response than expected. Print raster angle and its effect on the mechanical properties of PCU dogbones were also evaluated. Softer PCU grades showed smaller changes due to differences in raster angle while harder PCU grades were observed to change more significantly. This observation alludes to the importance of print pathing in membrane fabrication, particularly for harder PCU grades.

Three-dimensional-printed PCU membranes were evaluated using micro-CT, showing unintended artifacts as a result of the layer-by-layer 3D printing process. These features are later shown to affect the structural properties of the PCU membranes, further emphasizing the importance of print pathing. Mechanical evaluation via monotonic tensile testing of PCU membranes shows a more compliant response from the bowtie membranes and a stiffer response from the square membranes. Strut size was observed to have minimal effect on mechanical response. Overall, we observed significant differences in compliance, failure load, and failure RE of different loading orientations. Pore shape under tensile loading also varied significantly depending on the loading orientation. Stable pore geometries with loading are a key feature to afford tissue ingrowth into the device.

Fatigue performance of PCU membranes at different loading orientations largely followed the trend of monotonic tensile failure load. Both 95A and 75A square membranes exhibited a stronger fatigue resistance in terms of absolute load compared to bowtie membranes. However, when cyclic load was normalized by monotonic tensile failure load, it was observed that bending-dominated structures show superior fatigue performance.

Square membranes were found to possess comparable fatigue resistance to a commercial polypropylene mesh. However, under this study's cyclic loading pattern, the membranes exhibit significant creep over the course of the fatigue test. By comparing membranes under cyclic loading to membranes under pure static load, we conclude that failure of the membranes is likely due to a combination of both creep and classical fatigue failure working synergistically. Membranes that were loaded intermittently were observed to fail at a lower cycle count than those that were loaded continuously. These observations suggest a strong need for future fatigue tests to be performed with a load similar to one encountered physiologically during implantation, i.e., in terms of load level, profile, environment, etc. The current work should serve as a template for evaluating other new materials for this application relative to mesh materials that were used in clinical application.

One limitation of this study is that the tested samples were not conditioned in water for a prolonged period before testing. Past studies have shown that conditioning in water for up to 5 months can noticeably affect mechanical properties, albeit not in a catastrophic manner [36]. A future study with long-term exposure to an aqueous environment integrated with fatigue testing could provide further insight into the nature of long-term in vivo fatigue. Another limitation is that only uniaxial tensile loading was implemented. An extension of this study could also involve the use of multi-axial loading states such as ball-burst testing. A ball burst test has the advantage of being somewhat closer to a physiologically relevant loading state. However, ball-burst testing does not capture anisotropy that is otherwise observed in uniaxial testing. Although neither testing protocol can fully capture how a mesh will behave in the highly variable in vivo environment perfectly, insight into how a mesh behaves from both protocols may correlate with its clinical performance if interpreted with the appropriate understanding of the different boundary conditions [6]. A further limitation is the clamping approach used during testing. In clinical settings, meshes are implanted using suture lines, whereas this study assumes a close-to-perfect clamping (i.e., all boundaries have 0 displacement) by using a grip-type clamp. The suture-graft interface is a high stress location and may appreciably impact mechanical performance [56].

This work was performed in part at the Duke University Shared Materials Instrumentation Facility (SMIF), a member of the North Carolina Research Triangle Nanotechnology Network (RTNN), which is supported by the National Science Foundation (award number ECCS-2025064) as part of the National Nanotechnology Coordinated Infrastructure (NNCI). Research efforts for this publication were also supported by the National Institutes of Health under award no. R01 HD097187.

• National Institutes of Health (Grant No. R01 HD097187; Funder ID: 10.13039/100000002).

• National Science Foundation (Grant No. ECCS-2025064; Funder ID: 10.13039/100000001).

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.