Lumbar interbody fusion (LIF) is a well-accepted treatment for low back pain symptoms that emerge from segmental mechanical instability (Fritzell et al., 2001; Bhalla et al., 2017). During LIF surgery, the intervertebral disc (IVD) of the affected segment is replaced by an interbody fusion cage. Interbody cages provide immediate mechanical support and serve as scaffold to facilitate bone growth in the intervertebral space and fuse the two adjacent vertebrae (Bagby, 1988). Although cages are usually enriched with bone graft (substitute) to foster bone formation (Duarte et al., 2017), both material and design of the inserted cage dominate the mechanical interplay and define the initial interface between host tissue and cage. Current interbody fusion cages still render suboptimal fusion rates following LIF treatment (Meng et al., 2021). For this reason, novel interbody cages are still being developed and introduced into the clinic.

One specific technique utilized to manufacture a new generation of interbody cages is metal additive manufacturing (Arts et al., 2020), commonly known as 3D printing. It builds an object layer-by-layer by selectively adding material where needed, thus enabling production of tailored porous implant designs that are biomechanically optimized (Tan et al., 2017; Pobloth et al., 2018). Examples of such novel 3D printed metal interbody cages are trussed titanium interbody fusion cages (Hunt et al., 2021). Trussed cages encompass a network of linear beam elements (struts) that join at several intersections within the design. These highly porous cages provide an open architecture to accommodate bone ingrowth and may deliver bone stimulating mechanobiological strains to the cells attached to the strut surfaces.

Previous ex vivo research quantified the strain in all the struts of a trussed cage under moderate (1,000 N) and strenuous (2,000 N) axial compressive loads, by using high resolution micro computed tomography (CT) imaging (Caffrey et al., 2016). Assuming that strain amplitudes over 200 µε (microstrain, 10⁻⁶ strain) are relevant to both preserve bone homeostasis and stimulate bone formation (Duncan and Turner, 1995), it was concluded that physiological loading of the cages induced strut strains consistent with those reported to maintain bone balance. Accordingly, it was demonstrated that cage design (e.g. diameter of struts) could be adjusted in order to tailor the strains induced by physiological mechanical loads (Caffrey et al., 2018).

Although the aforementioned ex vivo investigations provide valuable insights into the size and distribution of strut strains under physiological loading conditions and allow to explore design modifications, the experimental set-up entailed several limitations. Firstly, loading protocols were limited to static axial compression to allow for microCT image analysis. Secondly, strain magnitudes were quantified per strut, based on the change in total strut length, disregarding local strains within the struts that potentially arise from bending behavior. Thirdly, the actual in vivo strain regimes may depend on many additional factors, including cage placement and patient-specific factors such as weight, bone density, degenerative state of the spine, and spinal curvature (Polikeit et al., 2003a; Abbushi et al., 2009; Naserkhaki et al., 2016; Galbusera et al., 2021). The influence of patient-specific variations on local strain regimes thus remains unknown.

Therefore, this study aimed to determine patient-specific variations in local strain patterns on the surface of a trussed titanium interbody fusion cage. Finite element (FE) modeling enables simulation of several physiological loading conditions and quantification of local strain values within spinal (sub) structures as well as within the cages (Goel et al., 2006; Gustafson et al., 2017). Additionally, the effect of patient-specific factors can be examined by studying the variation between different patient-specific models. Patient-specific FE models of four patients eligible for spinal fusion surgery were modified to simulate a posterior lumbar interbody fusion (PLIF) treatment with trussed titanium cages. Functional dynamics of the non-operated lumbar spinal segments, and the local cage strains and endplate stresses at the operated segment, were evaluated under physiological extension/flexion movement of the lumbar spine.

Figure 1 provides an overview of the demographic data of the four patients included in this study (one male and three female). As patients were selected to fit the same cage size, the population comprised a relatively narrow weight and length range (60–74 kg and 164–172 cm). Lumbar lordotic angles ranged from 29 to 40° while degenerative state of the non-operated discs varied from Pfirrmann grade II to IV. Different gray value intensities and distributions in the trabecular bone regions indicate the differences in bone density between vertebrae and patients.

Figure 2 displays the total lumbar spine motion, the angular motion per FSU, and the IDP per IVD during extension/flexion movement. The four S1-L1 patient-specific models showed comparable asymmetrical extension/flexion flexibility profiles but differed in terms of reaction moment magnitudes at 20°, in both extension (range −9.9 to −7.7 Nm) and flexion (range 17.3–25.8 Nm). These patient-specific differences were also reflected in the angular motion per FSU, especially in flexion at L5-S1. L5-S1 was also the disc with most patient-to-patient variability in terms of degenerative state (Figure 1). The IDP over all discs in neutral position, under follower load compression, ranged from 0.4 to 0.8 MPa. In general, it increased more in flexion (range 1.3–2.7 MPa) than in extension (range 0.6–1.1 MPa). Again, patient-specific variations were most pronounced in flexion at the L5-S1 level. The Pfirrmann grade III L3-4 IVD model (patient 3) led to clearly lower IDP values than the grade II L3-4 IVD models of the other patients, during the flexion movement.

Since the two inserted PLIF cages (left, right) demonstrated similar deformations within one patient for each of the loading conditions, Figure 3 illustrates only the calculation outcomes in the right cage of patient 1. In neutral position, only small strain values (<200 µε) were calculated in the cage. In extension, strains shifted to the posterior side of the cage and values increased up to approximately 300 µε. In flexion, strains shifted to the anterior side of the cage and locally, values exceeded 500 µε. For all loading conditions, both compressive (negative) and tensile (positive) strains were present at the struts. In the enlarged inset, struts show compressive strain on the one side and tensile strain on the other side, indicating inwards bending of the struts during flexion. The bar chart demonstrates that the relative amount of surface exceeding 200 µε varied more between loading conditions than between patients, whereas the coefficient of variation in flexion was 8.3%. The peak von Mises stresses within the PLIF cages ranged from 248 to 304 MPa over the different patients, which is far below the yield stress of 3D printed titanium.

Since the two caudal contact layers (left, right) within one patient showed similar behavior for each of the loading conditions, Figure 4 shows only the graphical output at the right caudal contact layer of patient 1. Comparable to the strain distribution in the cages, the caudal contact pressure shifts posteriorly and anteriorly in extension and flexion, respectively. Highest caudal contact pressures were observed anteriorly in the flexion configuration. Like the relative amount of surface exceeding 200 µε, the mean caudal contact pressure varied more between loading conditions than between patients (see bar chart). The coefficient of variation of the mean caudal contact pressure in flexion was 9.9% between patients.

The main purpose of this study was to determine patient-specific variations in local strain patterns on the surface of trussed titanium interbody fusion cages. These variations were analyzed in a specific subgroup allowing for implementation of one specific PLIF cage size for all included patients. The results demonstrate that within this specific subgroup patient-specific factors such as weight, bone density, degenerative state of the spine, and spinal curvature did affect local strain regimes; however, loading conditions in this group had a much more prominent impact on both size and distribution of the strains. The same trend was observed for the mean contact pressure of the cages at the caudal vertebral endplates. It should however be emphasized that the patient dataset in current research did not include any patients with comorbidities like previous lumbar surgery, heavy smoking, drug use or other conditions affecting bone or disc metabolism, osteoporosis, obesity, or scoliosis. Therefore, this patient cohort does not reflect the broad patient population undergoing spinal fusion surgery. Inclusion of patients with these comorbidities could provide additional insight as to what patient-specific constraints need to be taken into consideration and how to optimize implant design to address these conditions.

Although the lumbar spine models used in this research were validated in earlier studies, further validation would be warranted, particularly because the earlier studies did not include the instrumentation modeled in this research. A logical next step would be to validate the predicted strains within the cages using ex vivo spine testing of operated spines such that the FE results can be verified against the ex vivo observations. Subsequently, a patient study in which strain results, obtained from patient-specific models, are correlated with the postoperative progression of bone growth would provide further clinical evidence.

The operated patient-specific lumbar spine models in this study are somewhat stiff in flexion and somewhat compliant in extension (Panjabi et al., 1994; Dreischarf et al., 2014). This behavior might be caused by the surgical modifications to the intact models, as spinal fusion surgery generally increases the stiffness of the spine more in flexion than in extension movement (Molz et al., 2003). Since the rotational contribution of FSU L4-5 is known to be proportionally larger in flexion compared to extension (Pearcy et al., 1984), this effect might have been enhanced as all patients were scheduled for L4-5 interbody fusion. Additionally, the implementation of the follower load might have had a minor effect on angular motion per FSU (Rohlmann et al., 2001). The differences observed between patients on total lumbar motion level mainly emerged from FSU L5-S1, whose degenerative state varied most substantially between patients (range II to IV on the Pfirrmann grading system). Because the reduced disc height of a degenerated disc presumably increases the stiffness of the FSU (Muriuki et al., 2016), different load-deflection curves between segments could be expected. Mean IDP values in neutral position were consistent with in vivo data and increased, in accordance with previous literature, more substantially in flexion than in extension (Wilke et al., 2001). As IDP is known to increase more significantly in flexion in the fused spine compared to the native spine (Weinhoffer et al., 1995), IDP values were found to be in the high-end regime. The differences in IDP observed for disc L3-4 and L5-S1 in flexion correspond with loss of water in the more degenerated discs, inducing lower IDP values (Sato et al., 1999). Overall, all patient-specific models revealed physiologically realistic functional dynamics of the operated spine.

The percentage of the cage that experienced strains consistent with those reported to maintain bone balance under physiological loading in the current study were slightly less than those found in the aforementioned study that investigated a different trussed titanium cage under moderate (1,000 N) axial compression (Caffrey et al., 2016), i.e., up to 50% of the free struts was loaded beyond 200 µε in the ex vivo study versus up to 30% of the total surface in current in silico approach. These differences, however, might be explained by the fact that in the ex vivo study a larger cage was used, which has a relatively smaller screw insertion block. In the prototype PLIF cage of the current study the screw insertion block carries a relatively large part of the load, thereby reducing the load on the struts. Therefore, it can be expected that for a similar cage design the actual strain values would compare well to those reported in the earlier study.

In current research, a value of 200 µε was adopted in order to quantify the percentage of surface that experienced a strain value relevant for preserving bone homeostasis and stimulating bone formation. It has however been described before that the exact strain threshold for maintaining bone mass is a nonlinear function of the daily loading cycle number (Rubin et al., 2001). Stimuli with magnitudes of 200 µε are estimated to require approximately 35,000 loading cycles per day (once every 2–3 s) to maintain bone mass, whereas for mechanical stimuli with a frequency in the order of 10⁶ to 10⁷ per day (10–100 cycles per second) even strain values lower than 10 µε are suggested to be capable of stimulating bone formation (Qin et al., 1998). Patient-specific spinal motion profiles may therefore be required to interpret the strain values more appropriately. Moreover, it is worthwhile to emphasize that these reference strain values originate from bone remodeling research and it is unknown to what extent these values can be directly translated towards a former intervertebral disc, i.e., a cartilaginous environment. Once interbody fusion has progressed between the vertebrae, these values would be directly applicable. This would, however, require extension of the FE models to include the formation of bone within the cages and was outside the scope of this research.

Although the FE models were intended to predict strain values at the surface of trussed cages on submillimeter scale, they do not provide a full characterization of the mechanical stimuli the attached cells might perceive in an in vivo situation. This is because the exact micro-to nanoscale surface features at the struts, the way cells could be attached to the struts (bridged versus non-bridged), and other mechanical stimuli like fluid flow and hydrostatic pressure in the cages were not involved in current FE models (Kapur et al., 2003; Zhao et al., 2015; Hasegawa et al., 2020). Also, the trussed titanium cage model did not take surface micro-to nanostructure and strut composition in consideration. This simplified cage model therefore provides only a limited representation of the cage properties regarding its in vivo mechanobiological response. In order to accurately represent this response, multi-scale modelling will be required including microstructural features that include cell-strut interaction and fluid flow within the cage.

Since the posterior side of the PLIF cage is shielded more by the pedicle screw and rod instrumentation than the anterior side, higher strains could be found in the anterior part of the cage under extension/flexion movement. Assuming the higher strains will indeed accelerate bone formation at the anterior side of the cage, this would be favorable from a biomechanical point of view as PLIF segments that are partially fused anteriorly are found to be more stable than those partially fused posteriorly (Bono et al., 2007).

Current research used a subset of patient-specific FE models to predict the impact of patient-specific factors on cage level. The same framework could also serve as a platform to evaluate several design modifications of the interbody cages iteratively. Design modifications might be considered in order to target higher surface strains or to distribute the strains more homogenously across the whole cage. However, the ultimate strength of the proposed design modifications should also be continuously monitored as interbody cages should also withstand high peak forces in the lumbar spine (Ledet et al., 2018). Additionally, modified designs could change the amount of direct cage to endplate contact thus affecting the stresses on the endplate and the risk for cage subsidence (Steffen et al., 2000). Cage design optimization algorithms would therefore require a cost function that assesses a combination of several output metrics. In the future, development of such algorithms may facilitate interbody cage design optimizations.

It should be noted that the current study analyzed the behavior of one specific trussed titanium cage geometry used for LIF treatment with a posterior approach (PLIF cage) and that results might be different for other cage geometries. In fact, the choice for another surgical approach, like anterior lumbar interbody fusion (ALIF), would affect the output on the cage level by multiple means. ALIF surgery requires only one large cage as the anterior approach provides full access to the ventral side of the operated spinal segment (Mobbs et al., 2015). Since each single cage contains a screw anchoring point to enable cage insertion during surgery, one trussed ALIF cage contains relatively more struts than two trussed PLIF cages. Additionally, one ALIF cage generally comprehends a larger footprint on the vertebral endplate than two PLIF cages do. Moreover, ALIF surgery can be performed as stand-alone procedure, which generally means there is some additional fixation that can be instrumented anteriorly directly after cage placement (e.g., an anterior fixating plate), but there is no pedicle screw and rod instrumentation or other supplemental posterior fixation involved (Manzur et al., 2019). The different types of additional fixation in PLIF and ALIF surgery obviously result in different loading patterns on the interbody fusion construct (Choi et al., 2013).

In conclusion, this study demonstrates the applicability of patient-specific FE models to quantify the impact of patient-specific factors as weight, bone density, degenerative state of the spine, and spinal curvature on interbody cage loading. As the resulting surface strains were very similar for the different patient-specific models in the selected patient group, it can be concluded that the trussed design is rather robust from a mechanobiological perspective. In the future, the same framework might be further developed in order to establish a pipeline for interbody cage design optimizations.