Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Feb 1.
Published in final edited form as: J Mech Behav Biomed Mater. 2021 Nov 23;126:104976. doi: 10.1016/j.jmbbm.2021.104976

Design considerations for piezoelectrically powered electrical stimulation: the balance between power generation and fatigue resistance

ED Krech 1,2, LJ LaPierre 2, Safakan Tuncdemir 3, A Erkan Gurdal 3, EG Haas 4, PM Arnold 2,5, EA Friis 1,2,4
PMCID: PMC8792359  NIHMSID: NIHMS1761435  PMID: 34864397

Abstract

Quality and timing of bone healing from orthopedic surgeries, especially lumbar spinal fusion procedures, is problematic for many patients. To address this issue, clinicians often use electrical stimulation to improve surgery success rates and decrease healing time in patients with increased risk of pseudarthrosis, including smokers and diabetics. Current invasive electrical stimulation devices require an implantable battery and a second surgery for removal. Piezoelectric composites within an interbody implant generate sufficient power under physiologic loads to deliver pulsed electrical stimulation without a battery and have demonstrated promising preclinical bone growth and fusion success. The objective of the current study was to assess the power generation and fatigue resistance of three commercially manufactured piezocomposite configurations in a modified implant design to demonstrate efficacy as a robust biomaterial within osteogenic implants. The three configurations were electromechanically assessed under physiological lumbar loading conditions, and all configurations produced sufficient power to promote bone healing. Additionally, electrical and mechanical fatigue performance was assessed under high load, low cycle conditions. All configurations demonstrated runout with no gross mechanical failure and two configurations demonstrated electrical fatigue resistance. Future piezoelectric implant design decisions should be based on power generation needs to stimulate bone growth, as mechanical fatigue efficacy was proven for all piezocomposite configurations tested.

Keywords: electrical stimulation, piezoelectric composites, fatigue resistance, power generation, implant design

Introduction

Complete bone healing from a variety of orthopedic procedures is very challenging for many populations (Fingar et al., 2014). Lumbar spinal fusions have increased over 200% in the last decade, with at least a quarter of a million procedures performed each year (Deyo, 2015). Despite recent innovation in interbody implant technology, nonunion and pseudarthrosis rates are as high as 47% in lumbar spinal fusion procedures (Einhorn, 1995; Ekegren et al., 2018; Gan and Glazer, 2006; Knox and Chapman, 1993; Tzioupis and Giannoudis, 2007). Because there is no expectation for nonunions to heal spontaneously, surgical or other intervention is necessary to stimulate the healing process (Bhandari et al., 2012). Healing is especially challenging in difficult-to-fuse populations, particularly patients with diabetes and tobacco users, both of which are associated with higher rates of pseudarthrosis, failed fusions and increased time to healing (Berman et al., 2017; Browne et al., 2007; Chung et al., 2012; Kwiatkowski et al., 1996; Marin et al., 2018). As the incidence of diabetes is expected to increase substantially in the next decade (Boyle et al., 2001), as well as a continuing rise in the percentage of patients with comorbidities, a cost-effective, efficient solution is necessary to aide bone healing in these difficult-to-fuse patients (Martin et al., 2007).

To supplement bone healing and improve fracture and surgical fusion success rates, several adjunct therapies are used in addition to the primary implants for stabilization. One of the most common therapies used in spinal fusion is electrical stimulation. Implantable stimulators providing constant direct current (DC) stimulation applied directly to the fusion site has shown great success in long bone fractures and spinal fusions (Paterson et al., 1980; Saxena et al., 2005; Tejano et al., 1996). Improvements in fusion rates and decreased healing times were linked to the constant DC stimulation, but required additional surgery to remove the battery pack, increasing risk of infection and secondary wound healing issues evident with difficult-to-fuse patients (Geerlings and Hoepelman, 1999). However, the stimulating device is placed at the transverse processes rather than the interbody space, producing a less than optimal fusion environment.

Piezoelectric materials emit electric charge from cyclic mechanical loading and have been explored as an alternative to batteries to harvest ambient recurring motion (i.e. walking) and subsequently power auxiliary devices (Li et al., 2014). The accumulated energy from harvesting this cyclic mechanical loading through piezoelectric ceramics is directly proportional to the frequency of input mechanical source. For example, higher frequency sources such as machine vibrations have higher energy than lower frequency mechanical loading such as human body motion. Despite its low frequency nature, harvesting energy from human motion, specifically energy harvesting from walking, attracted many researchers across multiple disciplines (Platt et al., 2005, Antaki et al., 1995, Fan et al., 2017). High force as a function of person’s weight paired with high strain potential as a function of vertical displacement of center of mass promise to generate large energy. In a simplified expression, energy is directly correlated to the force, displacement and frequency of the source. Piezoelectric ceramics have been used successfully to generate energy from walking loads, but the power levels are limited by inefficiencies (Feenstra et al., 2008; Kymissis et al., 1998). At low frequencies, as seen in human motion, the mismatch in device resonance frequency and loading frequency worsens the limited charge density and strain amplitude of piezoceramic materials, making energy transduction less effective (Jaffe et al., 1971). Cofired stacks of piezoelectric elements connected electrically in parallel and stacked mechanically in series significantly lowers source impedance, increasing efficiency at lower resistances (Platt et al., 2005). However, the elastic modulus of the piezoceramic stacked generators is substantially greater than that of bone, causing mechanical impedance mismatch between the bone, which then yields a significant reduction in maximum allowable force and thus energy transfer to the harvester. The ceramics itself are also brittle and could not withstand repetitive multiaxial loads, inhibiting direct use as an implant material.

Goetzinger et al. developed multilayer piezoelectric composite spinal fusion interbody implants by encapsulating lead zirconate titanate (PZT) macro fibers in a medical grade epoxy, which increased the compliance of the harvester that improved the mechanical impedance matching and hence the performance (Goetzinger et al., 2016). The encapsulated PZT fibers improved the overall reliability of the piezoceramics and the multilayer structured enabled an increase functionality and power levels at lower resistance. These implants were designed to transduce cyclic walking loads to DC electrical signals that would be delivered directly to the intervertebral disc space through an attached electrode. The preclinical efficacy of these implants was demonstrated in a pilot ovine study, in which the histology, CT scans, and biomechanical data showed evidence of enhanced fusion in the active piezocomposite implants, compared to the controls that did not fuse and developed soft tissue callus (Friis et al., 2015). Despite promising results, the difficult and unreliable fabrication process limited replication, scalability and implementation as a tough implant material.

The use of traditional cofired piezoelectric stacks would improve manufacturability, while lowering source impedance and producing more power at lower resistances (Platt et al., 2005), thus overcoming obstacles in practical implementation of the piezoelectric component of the implant. Krech et al. recently studied the power generation capability of compliant layer adaptive composite stacks (CLACS), a PZT stack with interdigitated compliant layers (E. D. Krech et al., 2018; E. D. Krech et al., 2018). The compliant layers increased power generated from physiologic walking loads by 50% when PZT volume was held constant, a significant increase in power output efficiency at low frequencies. Similarly, Cadel et al. found that stacked PZT discs (a cofired stack analog) and CLACS included within an interbody implant design produced sufficient power needed to stimulate bone growth (Cadel et al., 2018). Even within the limited footprint of a transforaminal lumbar interbody fusion (TLIF) implant with a graft window, utilizing PZT discs addressed fabrication concerns with the macro fiber composite, while increasing power production efficiency. Static or dynamic mechanical integrity of the piezocomposite implant material has not yet been assessed. The inclusion of the compliant layer in CLACS may not be cost-effective if it does not improve the power production and fatigue life of the material.

The purpose of this study was to assess the power production capability and fatigue performance of three different piezocomposite generator designs. It was hypothesized that the use of CLACS would increase power production and fatigue resistance under loading conditions seen in spinal fusion implants. The overall goal was to understand the most appropriate PZT stack generator configuration in a modified implant design to produce sufficient power to supply electrical stimulation, while maintaining mechanical integrity under fatigue loading conditions.

Methods

Specimen Fabrication

The present study was devised to design, validate and elucidate the power generation capability and fatigue resistance of cofired PZT stacks and CLACS to test feasibility for their use in orthopedic implants. To mimic scalable manufacturing methods to be used in a final implant design, custom, production level piezoelectric stack generators were designed and fabricated, with and without CLACS technology. The three configurations, as seen in Fig. 1, represent the generic implant design utilized to effectively gauge failure of the interfaces between materials, primarily the polymer encapsulation and brittle ceramic PZT, and electrical connections. Other mechanical stress concentrations common in standard implants were intentionally eliminated, as to isolate the fatigue life comparison of the three PZT configurations.

Figure 1.

Figure 1.

Open in a new tab

Three configurations assessed. A. The external PEEK casing was the same in all three configurations. B. Internal schematic of Configuration 1 (C1). B. Internal schematic of Configuration 2 (C2). Note the compliant layers in between each PZT disc and the epoxy encapsulation were the same material. C. Internal schematic of Configuration 3 (C3). Note that the endplates with the titanium post were encapsulated with the PEEK body and PZT stack. The epoxy encapsulation of each PZT stack is shown in blue.

The piezoelectric generators (cofired stacks and CLACS) were manufactured by QorTek, Inc. (State College, PA) and assembled with the PEEK and endplates by Evoke Medical, LLC (Lawrence, KS). A worst case reasoning was utilized to design the configurations: mechanically the tallest height of an interbody implant was used (15mm), and electrically the height and diameter of the PZT was limited to what would fit within the shortest implant height and footprint. The overall implant height (15mm) and loaded footprint surface area (100mm2) was constant across all configurations. QorTek HT301 composition, soft-type PZT with high energy harvesting figure of merit, d33=460pC/N and g33=27mVm/N, is used for all three configurations. Configuration 1 (C1) included a traditional cylindrical PZT cofired stack generator. Configuration 2 (C2) utilized the CLACS design from Krech et al., utilizing PZT discs and interdigitated compliant layers. Configuration 3 (C3) included the same PZT stack as C1 with a titanium post at each end to enhance load transfer and subsequent power generation.

For C1 and C3, the cylindrical stack was a cofired 6mmØx3.2mm tall stack with 5 layers connected electrically in parallel and sintered together with platinum inner electrodes (Fig. 1B). The PZT was poled through-thickness post sintering in a silicone oil bath at 120°C for 5 minutes under 2.5kV/mm electric field. For C2 (Fig. 1C), CLACS were created with five 0.8mm thick PZT discs, interdigitated with cured EPO-TEK 301 (Epoxy Technology, Billerica, MA) compliant layers of the same thickness of the individual discs. After the PZT discs were electroded with a low-fire silver ink, they were poled through-thickness under the same conditions as the cofired stack, and subsequently connected electrically in parallel using conductive epoxy. For C3, the cylindrical stack with a titanium post, the same stack as C1 (6x3.2mm) was used but was encapsulated with a 6mmØ 4mm titanium post endplate (Fig. 1D). An error in the development of new manufacturing methods for these PZT configurations led to slight differences in PZT volume across the configurations. PZT properties were measured pre- and post- encapsulation to ensure correct electrical connection and material integrity following the potting process.

The overall goal of this study was to mechanically and electrically assess the performance of commercially produced CLACS (C2) as compared to a cofired piezoelectric generator (C1). The third configuration with the titanium post (C3) was an alternate way to enhance power from the stack by directing the load to the piezo stack. The balance between power production amplification and adequate resistance to fatigue failure was critical to understand across the three configurations, thus guiding configuration design decisions.

To capture the desired failure at material interfaces, the three configurations were designed to minimize geometric stress concentrations (implant surgical insertion attachments) that would influence fatigue failure. The PEEK casings were 13mm tall, with a 10x10mm cross section. There was an 8mmØ through-hole to house the PZT stacks. All PZT stacks were potted in the PEEK outer casing with EPO-TEK 301, a medical grade epoxy (Fig. 1). All stacks had a 1mm radial epoxy encapsulation and were centered within the height of the PEEK body, with epoxy filling in the remainder of the volume. Additionally, C1 and C2 had 1mm aluminum end plates adhered to the PEEK with EPO-TEK 301. For C3, the 1mm titanium endplates had a 4mm post extrusion, as seen in Fig 43D and were encapsulated with the PZT and PEEK bodies. Titanium was used in C3 to mimic final implant design and elucidate the effect of stiffness of the post to direct load to the PZT stack. This modified implant assembly represents materials and interface stress concentrations that will be present in final implant designs. All test specimens were x-rayed prior to testing to ensure no visible air bubbles in epoxy or other discontinuities existed in the encapsulation.

Electromechanical Testing

Using an MTS MiniBionix 858 test frame (MTS, Eden Prairie, MN), power production under physiologic compressive loads was compared. A 2.5kN load cell and self-aligning platen was used to ensure pure compressive loads were applied (Fig. 2). Physiologic stress experienced by lumbar spinal fusion implants was normalized to the surface area of the configurations. An 800N compressive preload was applied to ensure specimen were always in compression. Cyclic loads at three amplitudes (67N, 335N and 670N) were applied at three frequencies (1Hz, 2Hz and 5Hz). These loads and frequencies represent conservative estimates for the cyclic stresses experienced by orthopedic implants during walking and other typical human motion (Arshad et al., 2017; Cromwell et al., 1989).

Figure 2.

Figure 2.

Open in a new tab

Electromechanical and fatigue test setup.

For each load and frequency, AC voltage output of the piezo stack was measured across a shunting resistance sweep ranging from 23kΩ-800MΩ. Resistance range was chosen to characterize the behavior of each configuration at lower circuit resistances and capture the voltage output at the optimal resistance (matched impedance) for each PZT stack configuration and frequency. Voltage output was measured for 15 cycles at each load, frequency, and resistance to ensure steady-state behavior. The average amplitude of the middle 5 cycles was used for the remaining analysis, and converted to RMS, VRMS = Vamp/√2 to predict the expected DC output. Power was calculated for each load, frequency and resistance, P = VRMS2/R. Maximum power as a function of configuration was compared using a one-way ANOVA, with a Tukey-Kramer post-hoc analysis to assess differences between groups (α = .05). This initial power production comparison was used to ensure that each PZT configuration produced sufficient power within the given designs.

Fatigue Testing

Utilizing a modified ASTM F2077 testing protocol, the mechanical and electrical fatigue performance of each configuration was evaluated. Fatigue of orthopedic implants, specifically spinal fusion interbody implants, is typically characterized by generating an S-N curve with at least two implants tested to five million cycles at loads between 30-50% of the ultimate strength (ASTM, 2018). At 5Hz, five million cycles would take over eleven days. Therefore, to make the study length tractable, a low cycle, high load fatigue test was conducted. Based on estimates of fatigue performance (predicate S-N curves) and average footprint (surface area) of similar lumbar implant designs, a failure load at 50,000 cycles was estimated to be approximately 6,000N (Peck et al., 2018). This scaled load aims to accurately match predicted physiological stress levels in the final implant design. A ratio of maximum stress to minimum stress of 10 was used, as is consistent with compression fatigue testing of PEEK and PEEK composites (Jen and Lee, 1998; Lee et al., 2012; Rae et al., 2007). The loading frequency was 5Hz and tests were run in load control to match ASTM F2077. In summary, specimens were subjected to a 600 to 6,000N cyclic compressive load at 5Hz for 100,000 cycles, or until failure. The MTS MiniBionix with a self-aligning platen setup was used to avoid unwanted bending of the sample due to coaxial forces (Fig. 2). Nest plates as described in ASTM F2077 were not utilized due to the parallel nature of the endplates on the modified implant specimens.

Mechanical fatigue behavior of each configuration type was compared. Specimens were loaded until gross mechanical failure occurred (i.e. visible exterior crack formation/propagation) or runout with no mechanical issues after 100,000 cycles. As a mechanical failure criterion for success, a displacement limit was set in the loading program to stop cyclic loading if the specimen deformed beyond the failure values (> 10% of elastic deformation). Displacement was recorded for 15 cycles after every decade from 10 to 105 cycles. Average peak-to-peak displacement was calculated for each data set to track mechanical degradation. It was expected that there would be high variability in the behavior and failure in fatigue of the specimen, so if an average difference between configurations was not perceived, a pass or fail analysis was implemented. Mechanical fatigue resistance was considered successful if the number of cycles to failure was more than the cycles to failure of the existing implant data at a comparable stress ratio, as shown by the green check mark in Fig. 3. This analysis was created to understand material interface strength, predict mechanical fatigue resistance and inform future implant design decisions.

Figure 3.

Figure 3.

Open in a new tab

Representative S-N curve to demonstrate mechanical success criteria.

Electrical output fatigue resistance of each configuration was also assessed. Capacitance values pre- and post- testing were recorded for each specimen to assess electrical fatigue performance of the PZT stack. Raw voltage was measured across a 10MΩ resistance for 15 cycles after every decade from 10 to 105 cycles. The resistive load was chosen to match the impedance of the anticipated electrical load in the final implant design. Voltage measurements were used to compare trends of electrical performance deterioration across the three configurations. Average peak-to-peak voltage was calculated for each data set. Electrical fatigue resistance was considered successful if the peak-to-peak AC voltage at 100,000 cycles was at least 50V, the minimum voltage required to pass through a rectifying circuit. Finally, one specimen from each group was soaked in acetone to assess internal crack propagation through the PZT stack.

Finite Element Modeling

A finite element model of the three configurations (Fig. 1) was developed to compare internal stress concentrations due to the geometric and material property differences between the three composite structures to better understand the electromechanical behavior, understand the nuance of internal mechanical degradation and better inform future design decisions. All models were full 3D static simulations of the three configurations solved with FFEPlus in SOLIDWORKS 2019. Voronoi-Delaunay meshing scheme was used to create the general mesh, then a curvature based mesher was used to further refine areas of interest. Both meshers produced high quality 3D parabolic tetrahedral solid elements, each with 10 nodes and 16 Jacobian points. Mesh size and number of elements varied minimally between configurations, but satisfactory mesh convergence was achieved for all conditions. The flex tail circuit connection and slit in the PEEK housing were neglected to simplify the model and isolate the epoxy/PZT interfaces where failure was observed experimentally. The following material properties (E = elastic modulus, ν = Poisson’s ratio) were used to match experimental conditions: PEEK: E=3.9 GPa, ν=0.4; Ti caps: E=105 GPa, ν=0.37, Al end plates: E=72 GPa, ν=0.33; matrix epoxy E=2.3GPa, ν=0.35; PZT 5-H: E=52 GPa, ν=0.31. All materials were modeled as isotropic and all interfaces shared nodes to represent a fully bonded contact between materials.

To mimic the fatigue experimental testing conditions, loading and boundary conditions were consistent across all three models. A 6000 N pure compressive load was applied to an ultra-high modulus platen that was centered above the specimen to ensure uniform axial displacement across the implant cross section, mimicking the self-aligning platen used in the experimental study. The platen was restricted to only allow translation in the loading direction. The loading interface between the platen and the top of the specimen was a no penetration interface with no friction to allow for straining at the top of the implant. All other interfaces within the implant were modeled as bonded. The bottom face was constrained in the axial direction, allowing translation in the transverse direction. To prevent rotation one bottom corner of the specimen was pinned to prevent translation or rotation. For each configuration, simulations were performed to predict maximum and average stress values and approximate locations to inform interpretation of experimental results.

Results

Power Generation

Under physiological loads, the power generation of each configuration was measured to predict the performance of each PZT stack type in a final implant configuration. Table 1 summarizes the maximum power generation of each configuration as a function of the different load amplitudes and frequencies tested. As expected, based on accepted piezoelectric power generation as a function of applied resistance load, the power increased as the applied resistance load increased until the applied resistance load matched the impedance of the specimen, demonstrating maximum power production. At maximum power (670N, 5Hz) C2 produced significantly more power than C1 and C3, respectively (p < .01). The one-way ANOVA revealed no significant differences in C1 power generation in comparison with C3 power generation, although the trend was a 40% increase due to the titanium post in C3 (p = 0.1). Not accounting for variation in the PZT volume or d33, the inclusion of CLACS in the PZT stack design in C2 increased power output 3.6-fold as compared to the cofired stack in C1. Similarly, C2 produced 2.5 times more power than C3.

Table 1.

Average maximum power generated ± standard deviation.

Specimen Type Measured d33 (pC/N) Max Power (μW) 670N/5Hz Resistance of Max Power at 5Hz (MΩ) 670N/2Hz Max Power (μW) 335N/2Hz Max Power (μW) 67N/2Hz Max Power (μW) Resistance of Max Power at 2Hz (MΩ) 670N/1Hz Max Power (μW) 335N/1Hz Max Power (μW) 67N/1Hz Max Power (μW) Resistance of Max Power at 1Hz (MΩ)
C1 (n=9) 1689 ± 390 363 ± 108* 7 152 ± 42 35 ± 11 1.3 ± 0.57 20 76 ± 23 18 ± 6 0.71 ± 0.35 34
C2 (n=5) 1521 ± 608 1190 ± 395 9 560 ± 223 131 ± 58 3.8 ± 2.6 20 280 ± 101 66 ± 29 2.3 ± 1.6 50
C3 (n=7) 1808 ± 397 511 ± 155* 6 216 ± 70 52 ± 20 2.1 ± 1.5 20 110 ± 40 26 ± 11 1.2 ± 1.0 32
Open in a new tab
*

represents significant difference from C2 (p<.05)

As expected, all configurations followed expected trends in PZT power generation: a linear increase as frequency and load increase. The resistance of maximum power was consistent at each frequency load condition for C1 and C3: 6-7MΩ at 5Hz, 20MΩ at 2Hz, and 32-34MΩ at 1Hz. The resistance of maximum power for C2 was the same at 2Hz and was slightly higher for 5Hz (9MΩ) and 1Hz (50MΩ). At a conservative estimate for a walking stress, (335N, 2Hz) all configurations produced satisfactory power at a reasonable resistance for circuit design (10MΩ): C1 produced 31μW, C2 produced 100μW, and C3 produced 45μW.

Because volume was not constant across all configurations, power generation was normalized by the total volume of PZT in each configuration to compare power density as a more accurate figure of merit. The average power densities at 670N, 5Hz and the power densities at a walking load of 335N, 2Hz are shown in Fig. 4. Following the same trends as the overall power output, C2 had the largest power density, followed by C3 and C1, respectively. At lower resistances (< 2MΩ), C2 and C3 had similar power densities, but as resistance increased toward resonance, the average C2 power density increased considerably. The maximum power density was achieved at slightly higher resistances for C2, similar to power generation. At resistance of maximum power for each frequency, C2 increased power density compared to C1 and C3 by 3-fold and 2-fold, respectively. The C3 average power density was 1.4-fold higher than the C1 power density at all frequencies. These trends were consistent for all loads and frequencies tested. Although this measure does not account for the variation in the PZT d33, it is a good indication of the expected power production for each configuration.

Figure 4.

Figure 4.

Open in a new tab

Average power density as a function of configuration type and applied resistance load. A. 670N and 5Hz. B. 335N and 2Hz.

To account for the variation in the PZT manufacturing process across configuration type, power generation was normalized by the measured piezoelectric charge constant, d33. The cofired stacks used in C1 and C3 were manufactured with standard techniques, thus the variation in piezoelectric properties was very consistent. CLACS required an external electrical connection to each layer of the PZT stack to maintain the source impedance, thus the variation in piezoelectric properties was greater. The early-stage manufacturing techniques used in this development work led to greater variation in piezoelectric properties within the C2 group. The constant d33 is a measure of the voltage generated per unit of dynamic force applied, and thus is used to predict expected energy generation for a given PZT material and/or configuration. Assuming the piezoelectric ceramics is a capacitor, energy across a capacitor is proportional to the square of voltage. Therefore, the generated electrical energy per unit of input mechanical energy is directly proportional to the square of d33, so the data presented is normalized by that factor. Fig. 5 presents representative data for an individual specimen from each configuration grouped by similar d33 values at 335N, 2Hz. Fig. 5A compares normalized power for specimen with a d33 in the upper range; C1-26, C2-1, and C3-3 had measured values of 2425, 2143, and 2536 pC/N respectively. Fig. 5B compares normalized power for specimen with an average d33; C1-17a, C2-4, and C3-18 had measured values of 2040, 1920, and 1871 pC/N respectively. Fig. 5C compares normalized power for specimen with a d33 in the lower range; C1-5, C2-7, and C3-1 had measured values of 2040, 1920, and 1871 pC/N respectively.

Figure 5.

Figure 5.

Open in a new tab

Normalized power as a function of configuration type and applied resistance load. Note these are representative data for specimen grouped with similar measured d33 values. A. High d33 group. B. Average d33 group. C. Low d33 group.

Following the same trends as power density, C2 had the highest normalized power, which is mainly due to the improved power generation capability through enhanced mechanical coupling of the CLACS design from Krech et al. Despite the similar trends in normalized power according to volume and piezoelectric charge constant, the difference between the C3 and C1 groups decreased at both the high and low end of the measured d33 values. As evidenced by the variation of normalized power within configurations, especially within C2, the power produced is greatly influenced by the d33 of the PZT stack. Normalization by d33 emphasizes the mechanical effect of CLACS in the C2 structure, as the normalized power presented is a function of the permittivity of the PZT material, the total length of the stack and the improved power generation capability through enhanced mechanical coupling of the CLACS design. When making further design decisions, this variation should be considered, measured and controlled.

Fatigue

The peak-to-peak displacement as a function of number of cycles is plotted for each specimen in Fig. 6. All specimens tested ran out to 100,000 cycles, without reaching the displacement failure limit. The gross mechanical behavior was a pass for all specimens in all configurations. There were no visible cracks in the PEEK body and/or the endplates. Although the displacement did not reach the failure limits, the peak-to-peak displacements can be used to predict likelihood of future failure. For a stable mechanical composite structure, the peak-to-peak displacement should slightly decrease after the initial steady-state is reached. In an unstable structure the trend would be opposite, with an increase predicting future crack formation and potential failure.

Figure 6.

Figure 6.

Open in a new tab

Peak-to-peak displacement as a function of number of cycles. A. Peak-to-peak displacement for all specimen in the C1 group. B. Peak-to-peak displacement for all specimen in the C2 group. C. Peak-to-peak displacement for all specimen in the C3 group.

There was a mixed response in the C1 group, demonstrated by the widespread data (Fig. 6A). However, on average C1 peak-to-peak displacement decreased by 1% throughout the test. In a few specimens (e.g. C1-5), a slight increase could be an indication of ensuing mechanical failure. None of the C2 specimen had an increase and remained stable after the initial settling period (Fig. 6B). On average, the peak-to-peak displacement decreased a minimal amount, 0.5% in the C2 group. Conversely, all C3 specimens demonstrated an increasing peak-to-peak displacement, indicative of impending failure. The average increase was 4% from the starting value (Fig. 6C).

The electrical fatigue performance of all specimens was measured in comparison to a threshold peak-to-peak AC output voltage of 50V, which is required to pass through a post-processing circuit. Capacitance of each specimen was measured before and after the test as a more specific measure of PZT mechanical failure and/or depolarization of the material. For all configurations, the voltage and capacitance dropped as the number of cycles increased. Table 2 shows the relationship between capacitance and final peak-to-peak voltage output. For C1, the average decrease in capacitance and peak-to-peak voltage was approximately 30%. For C2, the measured capacitance dropped 73%, while the peak-to-peak voltage only decreased 50%. For C3, five of the specimens electrically passed (exceeded the failure criteria) and the capacitance and voltage decreased by 46% and 25%, respectively. However, there was catastrophic failure in two of the C3 specimen, with a sudden drop in voltage output below the threshold before 20,000 cycles, and thus the averages presented in Table 2 represent the specimens that passed. For all specimens, the voltage dropped most significantly between 1,000 and 30,000 cycles but then stayed fairly constant for the remainder of the test. This phenomenon was consistent for all three configurations.

Table 2.

Electrical fatigue performance of each configuration.

Specimen Type Cs Pre Test (nF) Cs Post Test (nF) Percent Decrease (%) Pk-Pk Voltage at 1k Cycles (V) Pk-Pk Voltage at 100k Cycles (V) Percent Decrease (%)
C1 (n=9) 2.4 ± 0.3 1.7 ± 0.2 31 287 ± 48 194 ± 48 32
C2 (n=5) 1.7 ± 0.5 0.46 ± 0.3 73 344 ± 137 172 ± 70 50
C3 (n=5)* 2.3 ± 0.2 1.3 ± 0.2 46 227 ± 74 170 ± 67 25
Open in a new tab
*

Note that two C3 specimen did not pass electrically, and are not included in these average values.

There was no gross mechanical failure of any of the specimen under the fatigue loading conditions tested, however the electrical and displacement results indicated there may have been internal crack formation and propagation within or around the PZT elements in each configuration. To better understand the internal failure modes, a specimen from each group was soaked in acetone to remove the epoxy/PEEK encapsulation (Fig. 7). Fig. 7A shows a brittle fatigue crack propagation through the cofired stack in a C1 specimen. As anticipated, in the C2 specimen the compliant layer between the PZT discs interrupted the crack propagation through the discs, as seen in Fig. 7B. The PZT stack in C3 underwent significant damage (Fig. 7C).

Figure 7.

Figure 7.

Open in a new tab

Internal views of PZT configurations after 100,000 cycles. A. Crack propagation through a C1 stack, side view. B. Crack in one PZT disc show in a C2 CLACS stack, side view. C. Gross mechanical failure of C3 stack, top view.

Finite Element Model

As is inherently true for finite element models, exact stress values and exact locations should be interpreted with caution. However, comparisons were made between models with the same boundary conditions and similar mesh resolution, as well as in relation to experimental measurements to predict pending failure and support geometrical and material design decisions. Table 3 shows the average and maximum von Mises stress in the PZT and the overall composite structure as a function of configuration predicted from the finite element models. Note the specimen maximum stress in the C2 configuration was not in the PZT elements, and was located on the bottom endplate, likely due to the imposed boundary conditions.

Table 3.

von Mises stress predicted by the finite element models for each configuration.

Specimen Type PZT Average Stress (MPa) PZT Max Stress (MPa) Specimen Average Stress (MPa) Specimen Max Stress (MPa)
C1 106.2 258.3 56.9 258.3
C2 112.5 273.7 57.9 307.1
C3 144.5 400.3 57.9 400.3
Open in a new tab

Figure 8 shows the stress distribution throughout the composite structures for each configuration. Consistent with experimental results, the Ti cap in C3 (Fig. 8C) does direct the stress through the PZT stack, increasing stress and power generation as compared to C1 (Fig. 8A). The discs in C2 have a slightly higher stress distribution as compared to C1, also consistent with the increase in power generation. The deformation of the three structures is consistent with the increased stiffness of C3 as compared to the other two configurations.

Figure 8.

Figure 8.

Open in a new tab

von Mises stress for each configuration with displayed 10:1 deformation scale. A. C1, B. C2, C. C3

The stress in the PZT elements of each specimen are shown in Figure 9. The average PZT stress does support the trends in both the experimental power generation and the electrical fatigue results. The stress in the C3 PZT is highest where the stress concentrations from the PZT/epoxy interface and the Ti/epoxy interface would overlap, creating potential for crack formation and propagation and the increased displacement measured experimentally. Additionally, the stress in the PZT discs in C2 is higher than within the stack in C1, consistent with the increase in power generation. This increased stress during the fatigue study could have contributed to the measured decrease in electrical capacitance due to depoling of the PZT discs.

Figure 9.

Figure 9.

Open in a new tab

von Mises stress in the PZT element(s) for each configuration with 1:1 scaled deformation. A. C1, B. C2, C. C3

Discussion

This study was designed to compare three potential scalable PZT configurations for use within a spinal fusion interbody implant to capture enough energy from the available high force low displacement motion, and convert mechanical to electrical energy that will power the conditioning circuit to create DC electrical simulation for bone healing without the use of a battery. The PZT stacks within the implant will generate power under cyclic loads transmitted to the implant through loads similar to walking and other human motion, allowing bone healing DC stimulation to be applied directly to the fusion site. The objective was to correlate power production with fatigue resistance of the three PZT configurations to inform future design decisions.

The use of CLACS to increase power generation has previously been verified; however, comparison to commercially produced cofired stacks has not been presented. Several studies have assessed the power amplification of CLACS through analytical and finite element models (Cadel et al., 2019, E. D. Krech et al., 2018; Pessia et al., 2019). These three different models have compared CLACS to encapsulated piezoelectric elements without interdigitated compliant layers to better understand the strain differences in the PZT discs that enhances power generation. These models have not compared CLACS to a standard cofired stack and have not assessed the stress concentrations when including the titanium cap, as done in this study.

CLACS, as implemented in C2, significantly increased power production by over 60% as compared to a traditional cofired stack (C1), even when accounting for variations in volume and d33. Electrically, this technique utilizes principles of piezoelectric stacked generators to lower source impedance, while also improving the mechanical impedance matching between source to piezoelectric elements as well as increasing lateral strain on the piezoelectric elements due to compliant layer expansion, increasing power output (Ember D. Krech et al., 2018). Mechanically, the toughened piezocomposite closely matches the compliance of bone to prevent stress shielding and resist brittle fatigue failure of the implant, indicated by the stable displacement in the fatigue study. The power amplification of this material will allow small stacks to be included within existing orthopedic implant shapes, while still generating enough power to provide electric stimulation to the gap healing site. This will not change surgical techniques or require development of new instrumentation, increasing likelihood of clinician adoption.

Although C2 did produce the most power, there was more variation within the group, especially the d33 values, as compared to both C1 and C3. It is important to note the CLACS were fabricated with crude manufacturing techniques to validate early stage product concepts. However, this could be indicative of future reliability and scalability concerns and should be balanced with the increased power production. Although the mechanical fatigue performance of C2 was superior, the electrical fatigue resulted in the highest drop in both capacitance and peak-to-peak voltage. Although no specimen had catastrophic electrical failure, and produced more voltage than the threshold, the drop in capacitance was expected. The CLACS structure enhances the mechanical loading on discs, which may instigate de-poling of the PZT over time. While the CLACS produces more power for a given volume of PZT, the height of the overall energy harvesting structure is larger due to the space the compliant layers occupy. This could be a design limitation in implant design, the available volume within an implant must be balanced with desired power production.

As expected, there was also an increase in power generation due to the titanium post in C3. Assuming uniform strain across the cross section and because the moduli of elasticity of PEEK and epoxy are about 3.3% that of the titanium, the load directed to the stack from the titanium post is expected to be approximately 12 times the load in PEEK/epoxy encapsulation region. The increase in power generation in C3 as compared to C1 followed this reasoning. However, mechanical and electrical fatigue performance was affected by the additional stress concentration at the material interface and modulus mismatch, demonstrated in the decreased mechanical and electrical properties due to fatigue of the C3 group.

Although C3 maintained adequate gross macroscopic fatigue performance of the implant assembly, the internal PZT stack was significantly damaged with multiple cracks and degradation. The titanium post configuration would not be suitable for use under repetitive high loads because of the stress concentration overlap, resulting in severe mechanical failure of the stack which would lead to premature electrical failure and further crack propagation for mechanical failure (Fig. 7C). However, the inclusion of the compliant layer stopped crack propagation through the CLACS structure in the C2 specimen (Fig. 7B). Although not for the same purpose, scientists found that in electrical fatigue testing, the adhesive interlayer within a stack decreased stress crack formation and propagation, mirroring these results (van den Ende et al., 2009). In future design, consideration of internal crack propagation and inclusion of a compliant layer could increase both power generation and resistance to fatigue failure.

In traditional composites fatigue testing, the low cycle fatigue limit is considered a reliability estimate for safety, insurance risks and life cycle estimates (Harik and Bogetti, 2003). The stress-life (S-N) approach for fatigue analysis is used primarily for materials screening and is useful for initial process of materials selection of implant materials subjected to high cyclic loading conditions, as seen in all orthopedic implants (Teoh, 2000). Fatigue behavior of piezoelectric stacks has also been analyzed using this approach (Platt et al., 2005). For these reasons, the S-N approach was used to assess the fatigue resistance of the three configurations of implant designs. The results of this study illustrate the mechanisms of premature failure within the different stack types and a clear necessity to consider both electrical and mechanical resistance to cyclic fatigue loading, to prevent brittle failure of the piezoelectric stacks. It will also be critical to ensure a complete encapsulation of the PZT ceramic to prevent crack propagation through the external implant design.

All three PZT configurations produced sufficient power (>30μW) under simulated walking loads to supply bone healing DC power through a rectifying circuit. Although different manufacturing processes were used, power densities in this study compare to similar studies assessing power generation of PZT stacks in spinal fusion implants (Cadel et al., 2018). In order to deliver appropriate DC current density levels through an attached electrode, a rectifying and conditioning circuit will be necessary to convert the AC voltage produced by the PZT in response to cyclic loads to a DC signal. The circuit design and subsequent resistance will be an important consideration when designing the most efficient PZT generator. As the total capacitance of the PZT stack and CLACS increase by increasing the number of layers per unit volume and surface area, the resistance at which maximum power occurs decreases (Cadel et al., 2018; Platt et al., 2005). Thus, circuit design should also influence the selection of the most effective PZT configuration in final implant design.

The use of a piezocomposite within an implant provides an integrative approach to stimulate bone healing. Piezoelectric materials only produce power under cyclic loading, and that rectified power would stimulate an additional bone healing response. However, if the implant is not cyclically loaded or the stack can no longer produce sufficient power, the implant will still function as a standard interbody implant, providing stabilization at the fusion site. C1 and C2 showed promising electrical fatigue performance and based on the power generation measured at runout under the high load, low cycle conditions in the present study, it is predicted that these configurations would provide stimulation long enough to initiate and promote bone healing. However, because the electrical stimulation would be an addition to the mechanical stabilization provided by the implant, the mechanical behavior of the material is more critical as it could affect adverse patient outcomes. Although, all three configurations studied in this work passed the defined mechanical criteria for success, the internal failure mechanisms could lead to external premature mechanical failure and should be considered in future integrated implant designs and a full-length fatigue study. In addition, as the electrical stimulation function helps bone grow, the load on the implant will be reduced and therefore the mechanical fatigue strength requirements will be reduced even further. These results validate the use of piezocomposites as load bearing power generators within implants.

Fatigue of PEEK biomaterials has primarily been approached in the context of specific implant designs (Kurtz and Devine, 2007), and the goal of this study was to give rationale for final implant design by elucidating early failure mechanisms of the proposed design configurations. The loads used in this study were considerably high loads for this test and were only used to assess the gross failure of the three configuration designs. The most common failure in spinal fusion interbody implants occurs at the insertion mechanism (Peck et al., 2018). This study purposely removed mechanical stress concentrations (i.e. insertion points) to best understand the failure modes related to the material interfaces. However, future studies should account for both material interfaces and mechanical stress concentrations. Repeat testing of the final implant design and PZT configuration at reasonable loads and longer cycles will need to be completed to characterize mechanical and electrical fatigue of final implant designs.

Conclusions

A piezoelectrically powered implant as proposed here would eliminate the need for an implantable battery bone growth stimulator, decreasing potential for infection, substantially decreasing cost for patients, hospitals and insurance companies, and eliminating the need for a second surgery. Higher risk of infection, secondary wound healing issues and repetitive intervention decrease value of current implantable electrical stimulation devices. Although current methods of providing DC stimulation have shown clinical efficacy, the drawbacks of current clinical devices have resulted in limited adoption. This study compared the power production and fatigue resistance of three PZT configurations in a modified implant design. All configurations produced enough power to supply bone healing DC stimulation under walking loads and maintained gross mechanical stability under the criteria defined in this work. Thus, fatigue is not the limiting factor in PZT configuration design. Future PZT generator design decisions can be based on power requirements and DC output acceptance criteria, although as demonstrated, internal microscopic failure could be an issue to consider and additional testing should be performed to characterize the entire structure’s full fatigue life. Because of their adaptability, PZT stacks are able to withstand physiological loads, and if electrically adjusted to enhance effectiveness at low frequencies, would allow successful incorporation into orthopedic implants to provide internal DC stimulation at the healing site without a battery in order to address a large clinical need.

Acknowledgements

This work was supported by an NIH SBIR Phase II grant (R44AR070088) as well as by the Kansas INBRE, P20 GM103418. The authors would also like to thank the Madison and Lila Self Graduate Fellowship for generous funding of graduate research at The University of Kansas.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Declarations of interest: none

Declaration of interests

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

All co-authors have a conflict of interest with Evoke Medical, LLC, the company to whom the research has been licensed from the university. This conflict has been managed appropriately and has not influenced the integrity of the work presented.

References

  1. Antaki JF, Bertocci GE, Green EC, Nadeem A, Rintoul T, Kormos RL, Griffith BP, 1995. A Gait-Powered Autologous Battery Charging System for Artificial Organs. ASAIO Journal 41, M588. [DOI] [PubMed] [Google Scholar]
  2. Arshad R, Angelini L, Zander T, Di Puccio F, El-Rich M, Schmidt H, 2017. Spinal loads and trunk muscles forces during level walking - A combined in vivo and in silico study on six subjects. Journal of Biomechanics. 10.1016/jjbiomech.2017.08.020 [DOI] [PubMed] [Google Scholar]
  3. ASTM F2077-18, Test Methods for Intervertebral Body Fusion Devices, ASTM International, West Conshohocken, PA, 2018, www.astm.org [Google Scholar]
  4. Berman D, Oren JH, Bendo J, Spivak J, 2017. The Effect of Smoking on Spinal Fusion. Int J Spine Surg 11. 10.14444/4029 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bhandari M, Fong K, Sprague S, Williams D, Petrisor B, 2012. Variability in the definition and perceived causes of delayed unions and nonunions: a cross-sectional, multinational survey of orthopaedic surgeons. J Bone Joint Surg Am 94, e1091–1096. 10.2106/JBJS.K.01344 [DOI] [PubMed] [Google Scholar]
  6. Boyle JP, Honeycutt AA, Narayan KMV, Hoerger TJ, Geiss LS, Chen H, Thompson TJ, 2001. Projection of Diabetes Burden Through 2050: Impact of changing demography and disease prevalence in the U.S. Diabetes Care 24, 1936–1940. 10.2337/diacare.24.ll.1936 [DOI] [PubMed] [Google Scholar]
  7. Browne JA, Cook C, Pietrobon R, Bethel MA, Richardson WJ, 2007. Diabetes and Early Postoperative Outcomes Following Lumbar Fusion. Spine 32, 2214. 10.1097/BRS.0b013e31814blbc0 [DOI] [PubMed] [Google Scholar]
  8. Cadel ES, Frazer LL, Krech ED, Fischer KJ, Friis EA, 2019. Analysis of how compliant layers and encapsulation affect power generated from piezoelectric stacked composites for bone healing medical devices. J Biomed Mater Res A 107, 2610–2618. 10.1002/jbm.a.36767 [DOI] [PubMed] [Google Scholar]
  9. Cadel ES, Krech ED, Arnold PM, Friis EA, 2018. Stacked PZT Discs Generate Necessary Power for Bone Healing through Electrical Stimulation in a Composite Spinal Fusion Implant. Bioengineering 5, 90. 10.3390/bioengineering5040090 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Chung HY, Machado P, van der Heijde D, D’Agostino M-A, Dougados M, 2012. Smokers in early axial spondyloarthritis have earlier disease onset, more disease activity, inflammation and damage, and poorer function and health-related quality of life: results from the DESIR cohort. Ann. Rheum. Dis 71, 809–816. 10.1136/annrheumdis-2011-200180 [DOI] [PubMed] [Google Scholar]
  11. Cromwell R, Schultz AB, Beck R, Warwick D, 1989. Loads on the lumbar trunk during level walking. J. Orthop. Res 7, 371–377. 10.1002/jor.1100070309 [DOI] [PubMed] [Google Scholar]
  12. Deyo RA, 2015. Fusion surgery for lumbar degenerative disc disease: still more questions than answers. The Spine Journal 15, 272–274. 10.1016/j.spinee.2014.ll.004 [DOI] [PubMed] [Google Scholar]
  13. Einhorn TA, 1995. Enhancement of fracture-healing. Journal of Bone 77, 940–956. [DOI] [PubMed] [Google Scholar]
  14. Ekegren CL, Edwards ER, de Steiger R, Gabbe BJ, 2018. Incidence, Costs and Predictors of Non-Union, Delayed Union and Mal-Union Following Long Bone Fracture. Int J Environ Res Public Health 15. 10.3390/ijerphl5122845 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Fan K, Yu B, Zhu Y, Liu Z, Wang L, 2017. Scavenging energy from the motion of human lower limbs via a piezoelectric energy harvester. International Journal of Modern Physics B 31, 1741011. 10.1142/S0217979217410119 [DOI] [Google Scholar]
  16. Feenstra J, Granstrom J, Sodano H, 2008. Energy harvesting through a backpack employing a mechanically amplified piezoelectric stack. Mechanical Systems and Signal Processing 22, 721–734. 10.1016/j.ymssp.2007.09.015 [DOI] [Google Scholar]
  17. Fingar K (Truven HA, Stocs C (AHRQ), Weiss A (Truven HA, Steiner C (AHRQ), 2014. Most Frequent Operation Room Procedures Performed in U.S. Hospitals, 2003-2012 (HCUP Statistical Brief #186). Agency for Helathcare Research and Quality, Rockville, MD. [PubMed] [Google Scholar]
  18. Friis EA., Galvis S, Arnold PM, 2015. DC Stimulation for Spinal Fusion with a Piezoelectric Composite Material Interbody Implant: An Ovine Pilot Study. [Google Scholar]
  19. Gan JC, Glazer PA, 2006. Electrical stimulation therapies for spinal fusions: current concepts. Eur Spine J 15, 1301–1311. 10.1007/s00586-006-0087-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Geerlings SE, Hoepelman AI, 1999. Immune dysfunction in patients with diabetes mellitus (DM). FEMS Immunol. Med. Microbiol 26, 259–265. 10.1111/j.1574-695X.1999.tb01397.x [DOI] [PubMed] [Google Scholar]
  21. Goetzinger NC, Tobaben EJ, Domann JP, Arnold PM, Friis EA, 2016. Composite piezoelectric spinal fusion implant: Effects of stacked generators. J. Biomed. Mater. Res 104, 158–164. 10.1002/jbm.b.33365 [DOI] [PubMed] [Google Scholar]
  22. Harik VM, Bogetti TA, 2003. Low Cycle Fatigue of Composite Laminates: A Damage-mode-sensitive Model. Journal of Composite Materials 37, 597–610. 10.1177/002199803029731 [DOI] [Google Scholar]
  23. Jaffe B, Cook WR jr., Jaffe H, 1971. CHAPTER 12 - APPLICATIONS OF PIEZOELECTRIC CERAMICS, in: Piezoelectric Ceramics. Academic Press, pp. 271–280. 10.1016/B978-0-12-379550-2.50016-8 [DOI] [Google Scholar]
  24. Jen M-HR, Lee C-H, 1998. Strength and life in thermoplastic composite laminates under static and fatigue loads. Part I: Experimental. International Journal of Fatigue 20, 605–615. 10.1016/S0142-1123(98)00029-2 [DOI] [Google Scholar]
  25. Knox B, Chapman T, 1993. Anterior Lumbar Interbody Fusion for Discogram Concordant Pain. Journal of Spinal Disorders 6, 242–244. [DOI] [PubMed] [Google Scholar]
  26. Krech ED, Barrett RM, Cadel ES, Friis EA, 2018. Power Generation Amplification and Stack Toughening via Compliant Layer Interdigitation V002T07A010. 10.1115/SMASIS2018-8043 [DOI] [Google Scholar]
  27. Krech ED, Cadel ES, Barrett RM, Friis EA, 2018. Effect of compliant layers within piezoelectric composites on power generation providing electrical stimulation in low frequency applications. Journal of the Mechanical Behavior of Biomedical Materials 88, 340–345. 10.1016/jjmbbm.2018.08.027 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Kurtz SM, Devine JN, 2007. PEEK Biomaterials in Trauma, Orthopedic, and Spinal Implants. Biomaterials 28, 4845–4869. 10.1016/j.biomaterials.2007.07.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Kwiatkowski TC, Hanley EN, Ramp WK, 1996. Cigarette smoking and its orthopedic consequences. Am J. Orthop 25, 590–597. [PubMed] [Google Scholar]
  30. Kymissis J, Kendall C, Paradiso J, Gershenfeld N, 1998. Parasitic power harvesting in shoes, in: Digest of Papers. Second International Symposium on Wearable Computers (Cat. No.98EX215). Presented at the Digest of Papers. Second International Symposium on Wearable Computers (Cat. No.98EX215), pp. 132–139. 10.1109/ISWC.1998.729539 [DOI] [Google Scholar]
  31. Lee W-T, Koak J-Y, Lim Y-J, Kim S-K, Kwon H-B, Kim M-J, 2012. Stress shielding and fatigue limits of poly-ether-ether-ketone dental implants. J. Biomed. Mater. Res. Part B Appl. Biomater 100, 1044–1052. 10.1002/jbm.b.32669 [DOI] [PubMed] [Google Scholar]
  32. Li H, Tian C, Deng ZD, 2014. Energy harvesting from low frequency applications using piezoelectric materials. Applied Physics Reviews 1, 041301. 10.1063/L4900845 [DOI] [Google Scholar]
  33. Marin C, Luyten FP, Van der Schueren B, Kerckhofs G, Vandamme K, 2018. The Impact of Type 2 Diabetes on Bone Fracture Healing. Front Endocrinol (Lausanne) 9, 6. 10.3389/fendo.2018.00006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Martin BI, Mirza SK, Comstock BA, Gray DT, Kreuter W, Deyo RA, 2007. Are Lumbar Spine Reoperation Rates Falling With Greater Use of Fusion Surgery and New Surgical Technology? Spine 32, 2119–2126. 10.1097/BRS.0b013e318145a56a [DOI] [PubMed] [Google Scholar]
  35. Paterson DC, Lewis GN, Cass CA, 1980. Treatment of delayed union and nonunion with an implanted direct current stimulator. Clin. Orthop. Relat. Res 117–128. [PubMed] [Google Scholar]
  36. Peck JH, Kavlock KD, Showalter BL, Ferrell BM, Peck DG, Dmitriev AE, 2018. Mechanical performance of lumbar intervertebral body fusion devices: An analysis of data submitted to the Food and Drug Administration. Journal of Biomechanics 78, 87–93. 10.1016/jjbiomech.2018.07.022 [DOI] [PubMed] [Google Scholar]
  37. Pessia ZR, Cunningham CA, Krech ED, Friis EA, 2020. Power amplification via compliant layer interdigitation and dielectrophoretic structuring of PZT particle composites. Smart Mater. Struct 29, 075004. 10.1088/1361-665X/ab85a5 [DOI] [Google Scholar]
  38. Platt SR, Farritor S, Haider H, 2005. On low-frequency electric power generation with PZT ceramics. IEEE/ASME Transactions on Mechatronics 10, 240–252. 10.1109/TMECH.2005.844704 [DOI] [Google Scholar]
  39. Rae PJ, Brown EN, Orler EB, 2007. The mechanical properties of poly(ether-ether-ketone) (PEEK) with emphasis on the large compressive strain response. Polymer 48, 598–615. 10.1016/j.polymer.2006.11.032 [DOI] [Google Scholar]
  40. Saxena A, DiDomenico LA, Widtfeldt A, Adams T, Kim W, 2005. Implantable Electrical Bone Stimulation for Arthrodeses of the Foot and Ankle in High-Risk Patients: A Multicenter Study. The Journal of Foot and Ankle Surgery 44, 450–454. 10.1053/jjfas.2005.07.018 [DOI] [PubMed] [Google Scholar]
  41. Tejano NA, Puno R, Ignacio JM, 1996. The use of implantable direct current stimulation in multilevel spinal fusion without instrumentation. A prospective clinical and radiographic evaluation with long-term follow-up. Spine 21, 1904–1908. [DOI] [PubMed] [Google Scholar]
  42. Teoh SH, 2000. Fatigue of biomaterials: a review. International Journal of Fatigue 22, 825–837. 10.1016/S0142-1123(00)00052-9 [DOI] [Google Scholar]
  43. Tzioupis C, Giannoudis PV, 2007. Prevalence of long-bone non-unions. Injury 38 Suppl 2, S3–9. [DOI] [PubMed] [Google Scholar]
  44. van den Ende DA, Bos B, Groen WA, Dortmans LMJG, 2009. Lifetime of piezoceramic multilayer actuators: Interplay of material properties and actuator design. J Electroceram 22, 163–170. 10.1007/sl0832-007-9411-0 [DOI] [Google Scholar]

RESOURCES