Introduction
Metal-induced artifacts significantly impede postoperative imaging of the spine [
1,
2]. To overcome this shortcoming of conventional metallic pedicle screws, nonmetallic high-strength carbon fiber-reinforced polyetheretherketone (CF/PEEK) pedicle screws have been developed. CF/PEEK is radiolucent and nonmagnetizable, and thus allows to minimize imaging artifacts in CT and MRI imaging [
3‐
5]. This feature greatly facilitates postoperative radiological assessment of fusion and adjacent segment disease, postoperative evaluation of neural structures in case of newly developing neurological symptoms as well as radiological follow-up of spinal tumors after spinal stabilization. Moreover, radiolucency of CF/PEEK improves dose calculation for postoperative radiotherapy of spinal tumors and minimizes radiation scattering and tumor shielding caused by metallic implants [
5,
6]. Beyond radiolucency, CF/PEEK may also be advantageous over standard titanium in terms of pedicle screw loosening due to its unique material properties. However, to the authors’ best knowledge, screw anchorage and loosening of CF/PEEK screws—although critical prerequisites for successful clinical application—have not been biomechanically evaluated yet.
Pedicle screw loosening is one of the most common complications after posterior instrumentation of the thoracolumbar spine. Screw loosening rates of up to 15% have been reported for nonosteoporotic patients and even higher loosening rates of up to 60% have been observed in osteoporotic individuals [
7]. Pedicle screw constructs are widely used to treat fractures, degenerative diseases, tumors and infection of the thoracolumbar spine as well as to correct spinal deformity [
8]. Loosening of pedicle screws, however, may result in loss of reduction, spinal instability, painful nonunion after fusion and possibly even in recurrent pain despite successful fusion, and therefore may necessitate revision surgery [
9‐
12]. There is an ongoing debate on how to improve pedicle screw anchorage and prevent screw loosening, and various strategies have been considered and evaluated. These strategies include various cement augmentation techniques, expandable screws and screw anchors, modifications of pedicle screw and screw thread design, alternative screw trajectories and screw insertion techniques, and perioperative teriparatide administration [
13‐
15].
Until now, none of these strategies has been aimed at using an alternative, nonmetallic screw material, but all of them have rather still been based on and tailored to conventional metallic pedicle screws. Nonmetallic CF/PEEK, however, offers important advantages over titanium or stainless steel beyond radiolucency. CF/PEEK exhibits outstanding biomechanical properties with a modulus of elasticity closer to that of bone as compared to titanium and stainless steel [
3]. CF/PEEK screw/rod constructs may therefore allow to better approximate the normal biomechanics of the spine via increased anterior column load-sharing, and consequently may promote interbody fusion and reduce the rate of adjacent segment disease [
16‐
18]. Moreover, the lower modulus of elasticity theoretically allows to reduce stresses and micromotion at the bone-screw interface and thus to reduce pedicle screw loosening [
18‐
20]. As to material properties, PEEK is a fully biocompatible, sterilizable and mechanically competent nonabsorbable semi-crystalline polymer [
3]. Incorporation of reinforcing carbon fibers permits to tailor the elastic modulus and to substantially improve the mechanical properties of neat PEEK [
3,
21]. CF/PEEK, therefore, is much more like titanium in terms of material properties as compared to neat PEEK and may be a desirable compromise between stiff titanium and distinctly less rigid neat PEEK constructs to act as an adjunct to fusion. Both neat PEEK and CF/PEEK have already been successfully used for spinal applications, such as interbody cages, rods and interspinous implants [
16,
22‐
24], whereas CF/PEEK pedicle screws have only recently become available.
The purpose of this biomechanical study therefore was (1) to compare screw loosening of nonmetallic CF/PEEK pedicle screws to standard titanium screws under cyclic loading and (2) to evaluate whether PMMA cement augmentation enhances screw anchorage of CF/PEEK pedicle screws. The two hypotheses tested were that (1) CF/PEEK pedicle screws resist an equal or higher number of load cycles until loosening than standard titanium screws and that (2) PMMA cement augmentation further increases the number of load cycles until loosening of CF/PEEK pedicle screws.
Discussion
The results from this biomechanical study show that (1) the nonmetallic CF/PEEK pedicle screws resisted a similar number of load cycles until loosening as standard titanium screws and that (2) PMMA cement augmentation of CF/PEEK screws significantly increases the number of load cycles until loosening.
To the best of our knowledge this is the first study that biomechanically evaluates CF/PEEK pedicle screws. CF/PEEK pedicle screws have only recently become available, and PEEK and reinforced PEEK composites as a material for screw/rod constructs have been suggested to hold the potential to overcome major drawbacks of standard titanium. PEEK is a radiolucent, biocompatible thermoplastic polymer with greater strength per mass than many metals and suitable for load-bearing orthopaedic implants. CF/PEEK is even superior to neat PEEK in terms of wear resistance and mechanical properties [
3]. CF/PEEK with its unique mechanical properties may thus represent a desirable compromise between stiff titanium and distinctly more flexible neat PEEK, and consequently may provide optimal stiffness to act as an adjunct to fusion. Several previous studies [
22] biomechanically evaluated neat PEEK and CF/PEEK rods for posterolateral fusion, posterior lumbar interbody fusion and anterior column reconstruction constructs. PEEK and CF/PEEK rods were found to provide comparable biomechanical stability and restriction of range of motion to titanium rods [
20,
27,
28]. Due to the lower modulus of elasticity, PEEK rods furthermore provided a more physiological anterior column load-sharing profile [
18,
29,
30]. The increased anterior column load-sharing again appears to reduce stresses at the bone-screw interface [
17,
18] and may promote interbody fusion [
20,
31], particularly when the interbody graft is slightly undersized or partially subsided. Moreover, lower stresses, intradiscal pressures and loss of disc height have been found at the cranial adjacent level in PEEK compared to titanium rod constructs [
17,
32], suggesting that PEEK rods might be beneficial to lessen the probability of adjacent level disease. In contrast to these previous studies on PEEK rods, the present study evaluated the biomechanical performance of CF/PEEK pedicle screws and specifically focused on pedicle screw loosening under cyclic cranio-caudal loading.
The etiology of pedicle screw loosening is still insufficiently understood but contributing factors include poor bone quality, stress shielding effects of rigid metallic screw/rod constructs, insufficient anterior support, and possibly wear debris-induced osteolysis and implant-related infection [
7]. PEEK and CF/PEEK screw/rod constructs for posterior instrumentation may hold significant advantages over titanium constructs with regard to screw loosening. First, rigid metallic screw/rod constructs shift the physiological loads posteriorly and unload the anterior column (stress shielding) resulting in increased stresses at the anchoring points of the construct (i.e., the bone-screw interface) [
20]. In contrast, PEEK and CF/PEEK screw/rod constructs allow for increased anterior column load-sharing due to the lower elastic modulus, and thus more closely mimic physiologic loading [
18,
29,
30]. Increased anterior column load-sharing reduces stresses at the bone-screw interface [
17,
18] and consequently may decrease the risk of screw loosening until bony fusion occurs. Second, the elastic modulus of PEEK and CF/PEEK pedicle screws matches that of the surrounding bone more closely compared to metallic screws, which may additionally decrease the risk of screw loosening. The reduced modulus mismatch between screw and bone theoretically results in a more homogeneous stress distribution and less micromotion due to similar strains of the screw material and the surrounding bone, particularly in osteoporotic bone, and may lead to enhanced osseointegration and improved screw fixation strength [
19,
33,
34].
In accordance with our first hypothesis, CF/PEEK and titanium pedicle screws resisted a similar number of load cycles until loosening. Despite the lower elastic modulus, however, CF/PEEK screws were not superior in terms of loosening. A possible explanation for the latter finding may be that the reduced modulus mismatch between screw and bone affects “primary stability” and resistance to early screw loosening only to a lesser extent, but rather improves long-term screw fixation strength via osseointegration and bone remodelling around the screws. In an elaborate study, Shi et al. [
19] were able to show that axial pull-out strength and peri-implant bone formation is significantly enhanced in expandable pedicle screws of low elastic modulus (Ti-24Nb-4Zr-7.9Sn, 42 GPa) as compared to standard titanium screws (Ti-6Al-4 V, 110 GPa) 6 months after implantation in osteoporotic sheep. Their finding that implant modulus affects peri-implant bone remodelling is in accordance with previous reports [
33,
34]. At an early stage after screw placement, however, bone quality and BMD might be the main determining factor of screw fixation strength and risk of loosening [
35]. A potential clinically relevant effect of the modulus differences between titanium and CF/PEEK screws therefore may not come to light in biomechanical testing which evidently cannot take into account osseointegration and bone remodelling around the screws. In vivo animal implantation studies and clinical studies thus will be needed to clarify whether low elastic modulus CF/PEEK pedicle screws are advantageous over conventional high elastic modulus titanium screws as to long-term screw loosening. To the best of our knowledge, there are only two preliminary clinical studies reporting on first experiences with CF/PEEK pedicle screws in a total of 44 patients (thirty-nine with spinal tumors and five with degenerative disease) [
4,
5]. These studies conclude that implantation of CF/PEEK implants was feasible and that postoperative imaging and dose calculation for radiotherapy was facilitated due to reduced artifacts. These preliminary reports, however, cannot answer the question whether pedicle screw anchorage of CF/PEEK screws is comparable to that of standard titanium screws.
As expected, PMMA cement augmentation significantly enhanced screw anchorage of CF/PEEK screws. This finding is well in line with observations from previous studies evaluating the effect of cement augmentation in metallic pedicle screws [
14]. In contrast to Weiser et al. [
35], we only found a weak to moderate correlation (with Spearman’s correlation coefficients ranging from 0.33 to 0.56) between BMD and number of cycles until loosening for both nonaugmented CF/PEEK and titanium screws. The poor correlation between BMD and number of cycles until loosening may be explained by the small range in BMD of specimens tested as well as by the small sample size. Moreover, pedicle width and length were not determined in the present study and may also influence screw anchorage. When combining all nonaugmented CF/PEEK and titanium screws (
n = 30), correlation between BMD and number of cycles until loosening was still moderate (Spearman’s correlation coefficient: 0.45) but reached significance (
p = 0.013). In augmented CF/PEEK screws, this correlation was even weaker (
r = − 0.18,
p = 0.61), which may be a consequence of an interlocking between the cement and the cancellous bone around the screws. The latter probably outweighs the minor differences in BMD among specimens tested [
36].
Various research groups developed different methods to investigate pedicle screw anchorage with more elaborate test setups than simple non-physiological pull-out tests. They all applied cyclic loading with an increasing load magnitude to provoke screw loosening. In addition to the compressive load, the present setup also includes a tensional load component and was thereby capable of reproducing the “windscreen wiper motion” of the screw which is usually seen in in vivo screw loosening. Similar to the present test setup, Kueny et al. [
37] and Bostelmann et al. [
25] used a toggle test setup which allowed translation of the vertebra in the transverse plane via an
x‐
y table during loading. While Kueny et al. [
37] introduced the load on the screw head, resulting in an axial loading superimposed by a bending moment; the present test setup applied the load via a rod and an offset rotational axis. The latter results in an axial loading superimposed by a bending moment with an additional shear component. The shear component, however, is neutralized by the translation of the
x‐
y table. In contrast to these single pedicle screw loading tests allowing for pairwise left–right comparisons under comparable conditions, Wilke et al. [
38] used entire assemblies of bisegmentally instrumented spinal motion segments. While this test setup tests the whole clinically applied instrumentation, it does not allow for control of numerous confounding factors related to specimen variation (e.g., BMD, vertebrae morphology, pedicle morphology). Due to multiple confounding factors smaller differences between test groups may be difficult to detect and require an unreasonable high number of test samples. Yet this setup enables a more comprehensive multidirectional loading involving a combination of axial loads and bending moments varying from flexion/extension to lateral bending as well as combinations of those. Recently, Liebsch et al. [
39] reported a test setup which also allows for a pairwise left–right comparison of pedicle screw anchorage. They proposed an elaborate loading with an axial load superimposed by a bending moment as well as by a shear force in direction of the screw axis. The axial load was introduced physiologically through the vertebral body with the construct fixed to the machine base by a rod. This test setup enables a multicomponent loading of the pedicle screw with combined axial and shear forces and a bending moment. However, with adding a shear force component, an additional constraint to restrict the translation of the screw in direction of the screw axis is introduced. By introducing the load through the vertebra, the peak bending moment is more physiologically located at the screw-rod connection and not at the screw entry point in the vertebra as in toggle tests. In this setup, however, screw anchorage might be very sensitive to the exact position of load application in relation to the screw tip position. While this can be well controlled with standardized synthetic bone surrogates, it may pose a challenge to control this parameter in cadaver tests with left–right comparisons and/or with specimens of varying anatomy. To summarize, all of these test setups have been reported to reproduce an in vivo like screw loosening pattern and all test setups described in the literature so far have their advantages and disadvantages. While a toggle test setup simulates axial loading superimposed by a bending moment without constraints, it introduces the load in the screw head. The multicomponent loading test setup represents a more constraint system, however, it allows to also simulate shear force components and applies the load more physiologically in the vertebra.
A limitation of this study is that biomechanical testing of screws implanted into cadaveric vertebrae cannot take into account implant osseointegration, which determines medium- and long-term fixation strength of load-bearing implants. Moreover, despite our elaborate cyclic cranio-caudal loading protocol, biomechanical testing of cadaveric specimens cannot fully reflect but rather only approximate in vivo loading conditions. As screw loosening most commonly occurs in osteoporotic bone, the tested specimens were obtained from elderly individuals with low BMD; extrapolation of our findings to high-density bone, however, should be treated with caution. Finally, our study did not include a comparison between PMMA-augmented CF/PEEK and PMMA-augmented titanium screws as the number of available cadaveric specimens was limited. Such a comparison could add further valuable information. However, as screw anchorage of nonaugmented CF/PEEK and nonaugmented titanium screws did not significantly differ, it may be reasonable to assume that screw anchorage of PMMA-augmented CF/PEEK and titanium screws is comparable as well.
There are several strengths of this study. First, to the best of our knowledge, this is the first study evaluating the biomechanical performance of CF/PEEK pedicle screws. Previous studies only investigated PEEK and CF/PEEK rods connected to titanium pedicle screws. However, as also noted by Bruner et al. [
27], evaluation of screw/rod constructs entirely made of PEEK and CF/PEEK is essential to improve our understanding of potential advantages these materials may offer for spinal instrumentation. Second, the elaborate load protocol of this study included cyclic cranio-caudal loading with stepwise increasing loads until screw loosening or a maximum of 10,000 load cycles. Cyclic cranio-caudal loading is considered more appropriate to assess and simulate in vivo screw loosening than much more commonly used and easier performed pull-out tests [
14,
25,
40]. Third, CF/PEEK and titanium control screws were randomly placed into either the left or right pedicle of the same vertebra, and screws were of different material, but of identical design. This experimental design thus allowed for a standardized pairwise comparison and for controlling for potential confounding factors, such as specimen variation (particularly in bone quality and pedicle morphology).