Introduction

Spinal instrumentation and fusion surgery have become a common treatment to restore the balance in adult spinal deformity (ASD). Despite advances in surgical techniques and knowledge in spine biomechanics, postoperative complications are still a problem. Proximal junctional kyphosis (PJK) is one of the most frequent complications following long instrumentation. Measured on the sagittal radiograph between the inferior endplate of the upper instrumented vertebra (UIV) and the superior endplate of the vertebra two levels cranial to it [1], a proximal junctional angle (PJA) greater than 10° indicates a pathological kyphotic deformity of the adjacent segment. Yet, PJK can be asymptomatic and does not always require revision surgery. Proximal junctional failure (PJF) is a more severe form, potentially involving acute proximal collapse, junctional compression fracture, retrolisthesis and/or instrumentation failure such as rod breakage or screw loosening [2]. Revision surgery incidence following PJF is up to 47% [2] making it a clinical and economic issue. PJF pathomechanisms are multifactorial with multiple potential risk factors associated, for instance, with the magnitude of correction in the sagittal plane, the ligament disruption procedure or the type of implant used at the UIV [3, 4].

Compared to pedicle screws (PS), the use of a less-stiff fixation at the UIV, such as hooks or supplementary tensioned bands, are increasingly considered to allow for a gradual transition in stiffness between the instrumented spine and the non-instrumented adjacent levels [4]. Several biomechanical studies have evaluated the range of motion (ROM) transition offered with different instrumentation configurations and related PJF incidence [5] with varying findings. Hooks decreased the ROM of the adjacent segment [6], but not always significantly [7, 8]. Sublaminar tapes or bands can allow for a more effective transition in terms of ROM and intradiscal pressure [6, 7, 9]. The spine stiffness decreased at the adjacent segment with a less-stiff rod material [10], but not always [8]. Using lower diameter or transitional diameter rods was also reported to decrease the spinal loads and the ROM above UIV [10,11,12].

Despite the potential design advantages of more flexible rod-to-vertebra fixation to reduce the risk of PJF, it is still unclear how the different types of implants at UIV balance the loads between the anterior spine and instrumentation and affect the risk of PJF.

The purpose of this study was to compare different proximal instrumentation stiffnesses, including vertebra-to-rod types of fixation and rod characteristics, to test the hypothesis that more flexible proximal instrumentation would significantly reduce the load gradient (i.e., load changes) between the instrumented and uninstrumented spine, and the assumption that load change around UIV in the anterior spine and instrumentation might be a mechanical factor indicative of the risks associated with PJF.

Methods

Adult spinal deformity patients

With the approval of the Institutional Review Board, six ASD patients who underwent posterior spinal instrumentation surgery after 2017 were used to build the numerical biomechanical models to test the research hypothesis (Table 1). All the patients had developed a junctional subluxation complication (i.e., proximal junctional acute collapse) [13] without complications related to implant–bone interface and rod failure.

Table 1 Preoperative patients’ demographic data and geometric indices
Full size table

Multibody modeling and simulation

A patient-specific multibody spinal model, previously developed and validated to study PJF pathomechanisms, was used for this study [11, 12, 14, 15]. This model was based on the 3D geometry of the patients, built using vertebral and pelvic landmarks identified on preoperative lateral and coronal radiographs and 3D reconstruction techniques [16]. The multibody model was built using MD-ADAMS 2014 software (MSC Software Corp. Santa Ana, CA). It included vertebrae from C7 to L5 and the pelvis, which were considered as rigid bodies, interconnected with 6-dimensional general springs (e.g., 6-degree-of-freedom stiffness matrix) to globally represent the mechanical properties of each functional spinal unit (FSU) and components such as the ligaments, intervertebral disc, and facet joints. The mechanical properties of the FSU stiffness matrices were determined from reported biomechanical tests [17, 18]. The FSU stiffness matrices were thereafter adapted to represent the removal of posterior elements at the levels where osteotomies were performed as documented in the operation report [19].

Multiaxial pedicle screws (PS) were modeled as rigid bodies for the threaded shaft and screw head, connected by a revolute joint to represent the multiaxial motion. The PS shank was connected to the vertebral pedicle by a nonlinear spring representing the mechanical properties of its anchorage with the bone [15, 20]. The initial shape of the rods was defined from the postoperative radiographs and a calibration algorithm was performed to find their initial unloaded shape, so that following the simulation of the correction of the instrumentation in the erect position, the elastically deformed rods corresponded to those in the reference postoperative radiographs [21]. According to the surgical steps, moments and forces were applied progressively to correct the deformity and the rods were aligned in the implant heads. Cylindrical joints were then set between implant heads and the rods. After the simulation of all correction maneuvers, the rod–implant cylindrical joints were replaced by fixed joints to simulate the tightening of the set screws.

The supralaminar hooks (SH) were modeled as rigid bodies, with a nonlinear lamina-to-hook joint representing the compliant junction between the anchor and the vertebra, as described in a previous study [11].

The sublaminar bands (SB) were modeled as two parts. The SB clamp was considered as a rigid body linked to the rod with a cylindrical joint. The band was modeled as a unidirectional spring between the clamp and the vertebra lamina with a stiffness of 410 N/mm [22]. A tensioning force (between 50 and 350 N) was applied to tighten the band, and then the cylindrical joint was replaced with a fixed joint.

The intraoperative prone position was modeled with the pelvis fixed in space and an inline longitudinal joint at the C7 vertebra. The main correction maneuvers were simulated following the operation report for each case (Fig. 1). The erected postoperative posture was simulated by applying downward gravitational forces on each FSU with patient-specific weight following anthropometric data [23, 24]. Springs were used to represent the action of the extensor muscles required to maintain the upright posture and counterbalance the patient’s weight [11]. A bending moment of 5 Nm was then simulated to evaluate a typical functional upper-body flexion [25].

Fig. 1
figure 1

Simulation main steps: (1) instrumentation in ventral decubitus position, (2) postoperative erected posture, and (3) flexion movement

Full size image

Model evaluation

To verify the computational modeling of the instrumentation and model calibration, the simulated correction maneuvers were compared to the actual postoperative radiographs using clinical indices (Cobb angle of the main curvature, T4–T12 kyphosis, and L1–L5 lordosis).

The validation, following ASME V&V40:2018 guidelines, was performed to assess the credibility of this biomechanical model in a previous study [15]. To this, the computed loads corresponding to PJF indicators for a group of asymptomatic patients and patients who have developed acute collapse PJF were compared. The sagittal moment at the adjacent spinal unit was found to discriminate the loads involved in the proximal segment when comparing both simulated groups. A sensitivity analysis and uncertainty quantification highlighted that the mechanical indices used to analyze PJF risks were within physiological ranges for the asymptomatic simulated group [15].

To assess the credibility of the sublaminar bands model, representative experiments of this surgical setting with SB and PS were simulated and compared to the reported data [9]. Five different spinal segments (T7–L2) were instrumented with different implant configurations: 1) PS from L2 to T10, 2) preceding configuration plus bilateral SB at T9 (“1-level SB”) and 3) preceding configuration plus bilateral SB at T8 and T9 (“2-levels SB”). The caudal vertebra was fixed in space and a pure 4 Nm moment for the intact segment and 6 Nm for the instrumented configurations were applied to the cranial vertebra in flexion and in extension. The simulated SB were tensioned using a 350 N force. The intervertebral range of motion (iROM) between UIV and UIV + 2 following the flexion/extension was computed and expressed as a percentage of the motion of the uninstrumented spine configuration (% of intact motion) [9].

Design of experiment to test different proximal instrumentation stiffness fixations

Using a full-factorial design of experiments (DOE), 15 different rod stiffness and vertebra–rod fixation configurations were simulated for each of the 6 cases and 3 simulated phases (i.e., intraoperative instrumentation, postoperative erected posture, and 5 Nm flexion) by combining:

  • UIV implant type: PS, SH, PS with sublaminar bands at the adjacent level with low (SB-50 N) [25] and high tensions (SB-250 N and SB-350 N) [26];

  • Rod stiffness: high (CoCr, 6 mm diameter), medium (CoCr, 5.5 mm), and low (Ti, 5.5 mm).

Several dependent variables were computed to assess the risks of PJF. The proximal junctional angle (PJA), as well as the loads (i.e., forces and moments) held by the proximal anterior functional spinal units (FSU) were chosen as indicators of potential soft tissues disruption. The computed loads at the UIV implant–vertebra interface were used to estimate the risk of bone compaction or screw pull-out [11]. The bending and torsion moment held by the rods at the UIV level were chosen as indicators of rod breakage risk (Fig. 2). These biomechanical variables were post-processed and analyzed using non-parametric Kruskal–Wallis statistical tests for each of the 90 simulations (15 configurations × 6 cases) with Statistica software (TIBCO® Statistica). The results are presented in terms of median, minimum, and maximum for the flexion phase.

Fig. 2
figure 2

Mechanical loads associated with risks of PJF in A functional spinal unit, B implant–vertebra connection, and C rod

Full size image

Results

Model evaluation

The mean difference between the simulated and actual postoperative instrumentation correction was of 2–3° for the regional curve angles, and 4° for the more local PJA angle (Table 2) for the six simulated cases, which is below the reported accuracy threshold corresponding to clinically relevant differences derived from 3D reconstructions from biplanar radiographs [27].

Table 2 Simulated vs. actual postoperative correction
Full size table

Compared to reported experiments for similar implant configurations, the simulated iROM using the SB and PS models were in general within 15% [9] (Table 3). For instance, the simulated flexion with 1-level SB was in good agreement with the in vitro study with an iROM reduction by 54% at UIV + 1 compared to PS (vs. 53% in the reference study [9]).

Table 3 Effect of sublaminar bands vs. pedicle screws only on intervertebral range of motion (iROM) after 6 Nm flexion and extension: published experimental tests of Viswanthan et al. compared to same simulations with our model (% difference of iROM)
Full size table

Design of experiment study of different proximal instrumentation stiffness fixations

The tension of the bands pulled the UIV + 1 toward the rod and the PJA was reduced on average by 2.3°, 5.5°, and 5.3° for the tensions of 50, 250, and 350 N, respectively, compared to PS, while the simulations with the simulated PJA with SH were reduced by 1.4° (p > 0.05). (Figs. 3 and 4).

Fig. 3
figure 3

Proximal junctional angle and UIV/UIV + 1 angle before and after 350 N tension of the sublaminar bands for a representative case (n°4)

Full size image
Fig. 4
figure 4

PJA for different implant configurations measured after the simulated instrumentation (median, 25–75% values and min–max for the 18 simulations per implant type: 6 cases × 3 rod stiffnesses)

Full size image

The sagittal moment at UIV + 1 was significantly decreased with higher tension SB (SB-250 N and SB-350 N) as compared to PS (1.3 Nm and 2.5 Nm vs. 15.6 Nm, respectively), but it was significantly increased at UIV + 2 (17.7 Nm and 17.7 Nm vs. 15.5 Nm (p < 0.05). Low tension SB (SB-50 N) allowed a smoother load transition by decreasing a lesser amount of the moment at the UIV + 1 (8.1 Nm vs. 15.6 Nm, p > 0.05) and slightly increasing it at UIV + 2 (16.8 Nm vs. 15.5 Nm, p > 0.05). Using SH instead of PS at the UIV did not change the loads held by the proximal junctional spinal segment (PJSS) (Fig. 5).

Fig. 5
figure 5

Sagittal moment (in Nm) held by the functional spinal units around the proximal junction for different implant types (*p < 0.05)

Full size image

SH significantly increased the caudo-cranial and torsion moments at the bone-implant interface compared to PS (2.6 Nm vs. 0.7 Nm; 2.7 Nm vs. 1.0 Nm, p < 0.05) (Table 4). SB with low tension (SB-50 N) had no impact on the loads held by the implant nor the rods at UIV compared to PS. With increased tension SB (SB-250 N and 350 N), the medio-lateral force at the bone-implant interface increased compared to PS (238 N and 297 N vs. 65 N, p < 0.05). It also decreased the pullout force compared to PS but thus increased the compression force on the bone-implant interface (− 237 N and − 236 N vs. − 14 N, p < 0.05). The torsional and sagittal bending moment held by the rods were increased with high tension rods for SB-250 N and 350 N, respectively (5.2 Nm and 6.1 Nm vs. 2.8 Nm, p < 0.05; 5.4 Nm and 5.7 Nm vs. 0.6 Nm, p < 0.05) (Table 4).

Table 4 Simulated loads at the proximal junctional segment (illustrated in Fig. 2) after 5 Nm flexion for all implant types at UIV
Full size table

Using low-stiffness rods (Ti/5.5 mm) significantly reduced the PJA compared to CoCr/5.5 mm and 6 mm rods (20.0° vs. 24.5° vs. 23.3°, p < 0.05), as well as the caudo-cranial moment at the bone-implant interface (0.7 Nm vs. 0.9 Nm vs. 1.0 Nm, p < 0.05) without affecting the loads held by the rods and the anterior spine (Table 5).

Table 5 Simulated loads at the proximal junctional segment (illustrated in Fig. 2) after flexion for all types of simulated rods
Full size table

Discussion

This biomechanical numerical study presents the impact of implant and rod stiffness configurations on load sharing at the proximal junctional level, which distinguishes itself from previous studies that mainly used iROM to evaluate different instrumentation techniques [5]. Compared to the experimental study by Viswanathan et al. [9], SB modeling was found to have similar quantitative effects on iROM compared to PS alone for the configuration SB-1 level in flexion with a maximum difference of less than 0.5°, considered negligible.

The biomechanical simulations did not show any difference in FSU loading between the simulated SH and PS. While some clinical studies reported a lesser occurrence of PJK using hooks [28], this outcome is not consistent [29]. Our numerical study showed that SB smooth the loading transition, thus decrease the load gradient, between the instrumented and non-instrumented spine, which may reduce the risk of PJF. This transition, however, was influenced by the level of SB tension previously reported above 200 N [25]. The low-tension bands (SB-50 N) provided a more gradual transition through the PJSS in terms of iROM, as reported in [6, 7], but also for the sagittal moments held by the FSUs. This smoother transition may prevent proximal junctional acute collapse as seen in clinical studies [28, 30]. High-tension bands (SB-250 N and SB-350 N), on the other hand, decreased the moment at UIV + 1 but did not smooth the load distribution. The increased tension may decrease the risk of acute collapse directly between the UIV and UIV + 1, but it could increase the risk of disruption cranially. Increasing the tension led to a significant increase of medio-lateral force and compression force at the bone-implant interface. The axial forces held by the bone-implant interface tend to be mainly compressive, but their variation also include pullout forces. The order of magnitude of those forces could still be considered safe compared to the published evidence about pullout strength with different anchors [31]. Moreover, decreasing implant axial load to achieve more compression than pullout may be seen less risky in the context of PJF, where the implant–bone failure mode was predominantly reported in pullout. Increasing SB tension changed the load-sharing between the anterior spine and the instrumentation by releasing the sagittal moment on the FSU but increasing it at the rods level.

Using low-stiffness rods decreased the PJA compared to high stiffness rods, yet the impact of rod stiffness on the anterior spinal loading was not found. However, decreasing rod stiffness decreased the caudo-cranial moment held by the bone-implant interface without affecting the loads held by the rods themselves. Using Ti rods decreased the risk of bone compaction or PS failure and proximal collapse compared to CoCr rods. Considering their ultimate strength value, it also increased the risk of rod breakage since the loads on the rods remain similar. Clinical studies similarly reported that Ti rods reduce the risk of PJK, but with an increased number of cases of rod fracture [32] compared to CoCr rods, which are more prone to PS fracture and PJK [33].

This computational study has potential limitations that should be recognized. Even if the numerical model was comprehensively validated for relative assessment of surgical strategies against PJF risks, the simulated mechanical loads should be interpreted on their relative effect rather than their absolute values [15]. Although the six cases used in this study reflect some of the variability encountered in terms of spinal deformities, the small number limits the generalizability of the current study. Nevertheless, it must be kept in perspective that the objective of the study was to systematically analyze the first-order influence of the different instrumentation variables in play related to PJF at UIV using a parametric study, which resulted in a large number of simulations (90) to compare and evaluate their effects. The load gradient reduction between the instrumented and proximal uninstrumented spine as a biomechanical factor indicative of PJF risk [4] remains to be correlated with clinical cases.

Based on this biomechanical analysis, the ideal combination of rod–vertebra anchor depended on several factors. In these computational models, using less-stiff rods decreased the risk of implant failure, but not in combination with high tension SB due to the risk of rod breakage. For the simulated cases requiring correction of a relatively high thoraco-lumbar junction kyphosis, high tension SB combined with high stiffness rods addressed the needed correction. The rod contouring to the native regional kyphosis between the UIV and UIV-2 would also be expected to decrease the PJK risk post-op. The rod curvature around the UIV was also a consideration since the combination of CoCr rods with high tension SB may increase the risk of failure one level cranial to the instrumentation. The possibility of establishing a more gradual transition of loads, as allowed by SBs with low tension, is a biomechanically interesting avenue to mitigate the risk of PJF, which remains to be studied in greater detail in clinical practice.

Conclusions

This study biomechanically evaluated the effects of different proximal fixations and instrumentation stiffnesses on load-sharing in specific ASD cases, allowing the inference of recommendations to consider reducing the risk of PJF. Simulated sublaminar bands at the proximal adjacent level of the instrumentation only with low pretension allow smoothing of the anterior load distribution without unproperly increasing the loads on the instrumentation. Decreasing the proximal rod stiffness also decreased the risk of bone failure but increased the risk of rod breakage.