The Effect of Multi-Level Laminoplasty and Laminectomy on the Biomechanics of the Cervical Spine: a Finite Element Study

Abstract

Laminectomy has been regarded as a standard treatment for multi-level cervical stenosis. Concern for complications such as kyphosis has limited the indication of multi-level laminectomy; hence it is often augmented with an instrumented fusion. Laminoplasty has emerged as a motion preserving alternative. The purpose of this study was to compare the multidirectional flexibility of the cervical spine in response to a plate-only open door laminoplasty, double door laminoplasty, and laminectomy using a computational model. A validated three-dimensional finite element model of a specimen-specific intact cervical spine (C2-T1) was modified to simulate each surgical procedure at levels C3-C6. An additional goal of this work was to compare the instrumented computational model to our multi-specimen experimental findings to ensure similar trends in response to the surgical procedures. Model predictions indicate that mobility was retained following open and double door laminoplasty with a 5.4% and 20% increase in flexion, respectively, compared to the intact state. Laminectomy resulted in 57% increase in flexion as compared to the intact state, creating a concern for eventual kyphosis - a known risk/complication of multi-level laminectomy in the absence of fusion. Increased disc stresses were observed at the altered and adjacent segments post-laminectomy in flexion.

Introduction

Cervical myelopathy is caused by spinal canal narrowing leading to spinal cord dysfunction. Laminoplasty techniques have become increasingly popular for treating multilevel cervical spinal stenosis, when a clinically relevant diagnosis such as cervical myelopathy is present. By retaining the dorsal elements of the spine, laminoplasty has the potential to preserve spinal stability and alignment and decreases the risk of postlaminectomy kyphosis and instability. Moreover, bone graft and fusion-related complications are avoided^1,2.

Since its introduction, several modifications have been made to the basic procedural theme of laminoplasty. The numerous laminoplasty techniques can be divided into two basic types (1) the midline splitting technique, otherwise known as a double door or “french door” laminoplasty (DDL)³, in which the laminae are opened via splitting of the spinous process and bilateral hinges at the lateral aspect and (2) the open door laminoplasty (ODL)⁴ procedure that consists of an osteotomy on one side of the lamina while a “hinge” is created on the other side, allowing for rotation of the entire lamina away from the lateral mass. Both techniques have demonstrated long-term success in increasing the spinal canal diameter and preventing worsening of the myelopathy⁵. Reported limitations, however, include inadequate decompression on the hinge side, the potential for reclosing of the door, as well as range-of-motion (ROM) restriction^6,7.

Achieving and maintaining an increased diameter of the spinal canal is critical to facilitating neurological recovery. Current laminoplasty techniques may include use of sutures, suture anchors, allograft or autograft bone, synthetic spacers and miniplates^8-10. The non-rigid nature of sutures may lead them to cut out, break, or stretch over time¹¹ and such changes have been associated with premature laminoplasty closures at rates ranging from 1.5% - 34%¹². Bone struts and ceramic blocks have the potential to dislodge, which may also lead to premature laminoplasty closure. Plating is a relatively new laminoplasty fixation technique, and plate-only constructs have increased in popularity. However, potential instrumentation issues, such as broken miniplates and screw back out pose as possible disadvantages.

The majority of studies addressing laminoplasty procedures are retrospective clinical studies. Few experimental investigations have addressed the biomechanical response to the procedure, and hence little is known about the biomechanical benefits and complications of plate fixation. The purpose of this study was to build upon our previous experimental and computational work^13-15 to address multilevel cervical laminoplasty, specifically a plate-only ODL and a DDL, and compare them to the standard laminectomy procedure in a specimen- specific computational model. Computational models enable information to be gathered that is difficult, if not impossible to attain experimentally, let alone clinically (i.e., stresses, facet loads). Consequently, our goal was to investigate not only the overall and intersegmental motions of the cervical spine in response to the laminectomy and laminoplasty procedures, but also to determine the stresses that developed throughout the individual spinal components and surgical implants.

Methods

A detailed three-dimensional (3D) finite element model of the cervical spine (C2-T1) was adopted and modified for this study^14,16. The vertebral bodies were segmented from CT images of a cadaveric specimen and were meshed with hexahedral elements using a multi-block meshing technique (IA-FEMesh)¹⁷. Each vertebral body was divided into cortical and cancellous regions and elastic moduli of 10GPa and 450MPa were respectively assigned¹⁸. The inter vertebral discs were each modeled with distinct annular and nuclear regions (Table 1) ^19,20. The annular grounds were defined using a hyperelastic material definition featuring the Mooney-Rivlin formulation, while the nucleus was represented by 3D fluid elements. The five major cervical spine ligaments consisting of the anterior longitudinal ligament (ALL), posterior longitudinal ligament (PLL), ligamentum flavum (LF), interspinous ligament (ISL), and capsular ligaments (CL) were incorporated into the model. Additionally, the facet gap was modeled using the tabular pressure-overclosure relationship available in ABAQUS (Dassault Systemes), thereby simulating the behavior of the articular cartilage.

Table 1.

Regional hyperelastic material constants for the annulus fibrosis

Segment	Anterior (MPa)		Posterior (MPa)		Lateral (MPa)
	C1	C2	C1	C2	C1	C2
C2-C3	0.56	0.14	0.3	0.075	0.46	0.11
C3-C4	0.3	0.075	0.3	0.075	0.133	0.033
C4-C5	0.2	0.05	0.2	0.05	0.133	0.033
C5-C6	0.2	0.05	0.133	0.033	0.133	0.033
C6-C7	0.2	0.05	0.3	0.075	0.133	0.033
C7-T1	0.8	0.2	0.8	0.2	0.3	0.075

The following sections detail the modifications made to the intact finite element mesh to simulate the laminectomy and laminoplasty procedures at levels C3-C6. These surgical simulations were in accordance with our previous experimental and finite element (FE) investigations^13,15.

Laminectomy (C3-C6) Simulation in the C2-T1 Model

The spinous process and both lamina were resected from the intact model (Figure 1A) for vertebrae C3-C6, while the facet joints remained intact. Additionally, the associated ligaments (ISL, LF) were removed. The final laminectomy model was comprised of 165,865 nodes and 158,957 elements. Figure 2A shows the resulting C2-T1 laminectomy model.

Figure 1. Superior view of a vertebra following: (A) Laminectomy, (B) Open Door Laminoplasty stabilized with plates and screws, (C) Double Door Laminoplasty stabilized with a spacer; highlighting the Laminar Opening Space (LOS) of 10 mm.

Figure 2. C2-T1 finite element model showing (A) Laminectomy, (B) Open Door Laminoplasty (ODL), and (c) Double Door Laminoplasty (DDL) at levels C3-C6.

ODL (C3-C6) Simulation in the C2-T1 Model

A bicortical cut was simulated along the junction of the lamina and the lateral mass of each C3 through C6 intact vertebral mesh by completely removing a layer of elements. On the contralateral side, a hinge of approximately 3-4mm was created along the junction of the lamina and lateral mass by removing elements representing the unicortical layer. The spinous processes of the involved vertebrae (C3-C6) along with the ISL were excised¹⁵. Additionally, to simulate the surgical procedure, the LF at the adjacent levels (C2-C3 and C6-C7) was partially cut on the open side of the lamina to allow for the laminar opening. Two screw holes in the lateral mass and one hole in the lamina were created based on the desired plate position. The lamina of each vertebra (C3-C6) was opened towards the hinge by applying a uniform load until a laminar opening space (LOS) of 10mm²¹ was obtained as illustrated in Figure 1B. Our goal was to account for the stresses that arise as the hinge is opened to accommodate the implant which are ultimately transferred to the plate/screws once im- planted¹⁵. Consequently, the principal stresses developed in the vertebral bodies and surrounding ligaments after laminar opening were extracted from the model and applied to the laminoplasty model as initial conditions¹⁵.

Computer Aided Design (CAD) models (ProE; PTC, Needham, MA) of the titanium plates and screws (Medtronic Sofamor Danek, Memphis, TN) were generated from their respective physical dimensions and were meshed with hexahedral elements using IA-FEMesh. Each component was assigned an elastic modulus of 116GPa and Poisson's ratio of 0.3. Small sliding contact was formulated at the interface between the bone and the laminoplasty plates, while the surfaces of the bone/ screw and the screw/plate were tied during the analysis. Such contact formulations have been used previously to model the contact between the screw and bone²². Figure 2B shows the C2-T1 finite element model with ODL simulated at the C3-C6 levels.

DDL (C3-C6) Simulation in the C2-T1 Model

Two bilateral hinges at the junction of each lateral mass and lamina were created as described for the single hinge of the ODL procedure. The spinous process was split along the mid-sagittal plane and opened until a laminar spacing of 10mm was obtained (Figure 1C). The tip of the spinous process was removed with care to preserve enough bone to hold the spacer. Based on the surgical procedure, the ISL at each of the involved levels were resected and the LF was partially removed at the midline from C2 to C7 to allow for the laminar opening. Again, the stresses developed during laminar opening were introduced to the model as initial condi- tions¹⁵. Because these stresses tend to close the lamina back, a 10mm trapezoidal shaped hydroxyapatite (HA) spacer (elastic modulus of 26GPa²³ and Poisson's ratio of 0. 27), meshed with hexahedral elements, was introduced to stabilize the lamina in the open position. Bony union between HA spacers and spinous processes has been observed in many clinical studies²⁴ and was therefore simulated using the TIED command in ABAQUS. Figure 2C shows the final C2-T1 DDL model.

Flexibility Study

The intact and all three surgical models (ODL, DDL and laminectomy) were tested in flexion/extension (±MX), right/left lateral bending (±MZ), and right/left axial rotation (±MY). The inferior nodes of the T1 vertebral body were fixed in all directions and a moment of 2Nm was applied to the superior surface of C2. The analysis was performed using the finite element software ABAQUS 6.9; enabling the biomechanical response of the intact, laminectomy, and both laminoplasty procedures to be compared. The ROM, facet loads, stresses in the annular regions of the intervertebral discs, and the stresses in the cortical regions of the vertebral bodies were analyzed for all four models (intact, laminectomy, ODL and DDL). Stresses in the laminoplasty plates/ screws were also analyzed.

The current FE model was developed from the original experimentally validated specimen-specific C2-C7 model¹⁴. T1 was added for this investigation and the intact flexibility data was compared to in-house multispecimen experimental studies^13,25 as well as to data reported in the literature^19,26,27. These validation efforts ensured that the computational response was within the range of normal cervical spine behavior in each loading direction for an applied 2 Nm moment (Figure 3). Figure 4 compares the ranges of motion of the individual vertebral levels between the finite element model and the experimental data. The intact model predicted motions that were comparable to the experimental data for the majority of loading modes. Furthermore, we focused on the changes in flexibility predicted by the instrumented computational models as compared to the respective instrumented experimental specimens, thereby establishing confidence in the ability of the specimen-specific model to predict the post-surgical response of the spine.

Figure 3. Comparison of C2-T1 ROM of the C2-T1 finite element model with subject-specific experimental data (C2-C7 at 1Nm) and literature data (C2-T1 at 2Nm) in Flexion/Extension²⁵, Lateral Bending²⁴ and Axial Rotation.²⁴.

Figure 4. (A) Comparison of level-by-level ROM for intact finite element model under 2Nm with in-house experimental and literature data in Flexion/Extension^11,23,25, Lateral Bending^11,23,24 and Axial Rotation. ^11,23,24.

Results

The percent changes in C2-T1 ROM post lamino- plasty and laminectomy with respect to the intact state are shown in Figure 5. The greatest change in motion was observed during flexion for all procedures. During flexion, the ODL and DDL resulted in a 5.4% and 20% increase in C2-T1 ROM respectively, while the laminectomy resulted in a substantial 57.5% increase in the C2-T1 motion. For the remaining loading directions, the greatest change in ROM did not exceed 4.3%.

Figure 5. Percent changes in the C2-T1 ROM after Open Door Laminoplasty, Double Door Laminoplasty and Laminectomy when compared to the intact case.

The intersegmental motions in response to the six loading modes were also compared following the surgical procedures. During flexion, after ODL the adjacent levels C2-C3 and C6-C7 showed a 39% and 20% increase in the motion respectively; while no substantial changes were observed at the altered levels (Figure 6). The percent increase in motion after DDL varied from 4.3% to 34.6%. Compared to the intact model, laminectomy at C3- C6 led to a profound increase (37.5% to 79.6%) in motion across the levels C2-C3 to C6-C7. During extension, the superior adjacent level C2-C3 showed an increase in motion of 8.5% and 28.8% after ODL and DDL respectively, while minimal changes were seen at the other levels. For left lateral bending, a decrease of 11.7% and 20.3% in motion was observed at the inferior adjacent level C6-C7 after ODL and DDL respectively, while minimal changes were seen at the other levels. Similarly, left axial rotation resulted in 13.2% and 15.1% decrease in motion at C6-C7 after ODL and DDL respectively. After laminectomy, both lateral bending and axial rotation led to minimal changes in the motion (<5%).

Figure 6. Percent change in intersegmental motion after the three surgical techniques for the six loading modes.

Figure 7 shows the percent changes in the annular stresses (von Mises) of the intervertebral disc after the simulated surgical procedures. After ODL, the adjacent discs (C2-C3 and C6-C7) showed an increase in the stress values while DDL and laminectomy resulted in an increase in the stresses across the surgically altered levels (C2-C3 to C6-C7) during flexion (Figure 7). Minimal changes in the disc stresses were observed at most of the levels during the other loading modes. During flexion, no facet loads were recorded as they were not engaged. In the other loading modes, the inferior (C6- C7) and superior (C2-C3) unaltered levels recorded ˜ 30% change following the open and double door laminoplasty. The changes observed at the other levels post lamino- plasty and laminectomy was less ˜ 10%.

Figure 7. Percent change in stresses in the annular regions of the intervertebral discs after the three surgical techniques during the six loading modes.

During all six loading modes the von Mises stresses in the laminoplasty constructs, namely the screws and plates (˜250MPa) (Figure 8) of the ODL and the HA spacer (˜ 200MPa) of the DDL, were within the yield strength of the respective materials. Both laminoplasty models (ODL and DDL) resulted in increased vertebral cortical body stresses at the C3-C6 levels during laminar opening, with the posterior vertebral body demonstrating higher stresses than the anterior regions (Figure 9).

Figure 8. Representative von Mises stress (MPa) distribution in a titanium miniplate.

Figure 9. Comparison of maximum von Mises stress recorded in the anterior and posterior cortical regions of vertebral bodies after opening the lamina for ODL and DDL.

Discussion

Understanding the effect of surgical procedures on the biomechanics of the spine may help a clinician better treat, and perhaps prevent spinal instability. Various experimental studies have demonstrated concerns for instability of the spine after laminectomy compared to multi-level laminoplasty. Subramaniam et al.²⁸ reported that laminectomy resulted in a 14.2% increase in the ROM (for ±1.5Nm moment) during flexion/extension when compared to intact state. Under a moment of 1.5Nm, Kubo et al.²⁹ tested fresh cadaveric cervical specimens and observed an increase of 2.6% in motion in flexion/extension, 6.2% in lateral bending, and 8.4% in axial rotation after four-level DDL. Nowinski et al.³⁰ tested nine cervical spines after C3-C6 open door laminoplasty to show a 4% increase in flexion/extension, 2.3% increase in lateral bending, 14% increase in axial rotation. This increase in axial rotation after laminoplasty could be attributed to altering the facet capsules for stabilizing the lamina in an open position using sutures after lami- noplasty. In an in vitro study¹³, we reported that ODL stabilized with titanium miniplates led to insignificant (p>0.05) changes in motion (<1% decrease in Flexion/Extension; 3.6% increase in Lateral bending; 6.3% increase in Axial rotation) while laminectomy led to a significant (p<0.05) increase (20% in Flexion/Extension; 8.2% in Lateral Bending; 15% in Axial rotation) in motion during the primary three loading modes.

While several experimental studies have addressed the biomechanical effects of single and multi-level laminectomy procedures, previous computational models have been limited to single-level studies. Kumaresan et al.³¹ and Wan et al.³² used 3D validated finite element models of cervical spines to show increased ROM and stresses at the adjacent level post-laminectomy. Since only a three-level finite element model (C4-C6) was used by Kumaresan et al.³¹, the loading and boundary conditions might affect the applicability of the results. Also, both the above mentioned studies have addressed only single-level laminectomy. It is well-known that the posterior surgical technique involving laminoplasty or laminectomy may be preferred when multiple levels of cervical spine are involved in a compressive disorder. This is the first computational study looking at the differences in terms of flexibility and stress distribution in the implants, vertebral bodies and intervertebral discs using a specimen-specific model.

Table 2 compares the percent changes of the C2-T1 ROM (relative to the intact motion) with our in-house experimental data as well as literature data after ODL, DDL and laminectomy. The findings from the above mentioned studies of laminoplasty and laminectomy are consistent with our finite element results where we observed a 3.4% increase in the ROM after C3-C6 ODL and 30% increase in the motion after C3-C6 laminectomy in flexion/extension. The DDL finite element model predicted an approximate 12% increase in C2-T1 ROM during flexion/extension, thereby explaining the role of lamina-ligamentum flavum complex in the stability of spine.³³ Changes less than 4% were observed in other loading modes (lateral bending and axial rotation) after C3-C6 ODL and DDL. The current findings after laminectomy (changes < 3% in lateral bending and axial rotation) also agreed with clinical observations where post laminectomy deformity is more predominant in flexion- extension than in lateral bending and axial rotation^34,35.

Table 2.

Percent change in range of motion relative to the intact spine.

	Flex/Ext	Lat. Bend	Axial Rot.
Laminectomy 4-Level
Subramaniam et al. (1.5Nm)	14.2± 10	NA	NA
Kode et al. (2Nm)	20.8±20.5	8.2±6.5	15±9.4
Current FE model (2Nm)	30	0.2	1.8
ODL 4-level
Subramaniam et al. (1.5Nm)	2.5 (↓)± 5	NA	NA
Nowinski et al. (C2-C7 and 1.5Nm)	4.1± 4	2.3± 4	14.3± 3
Kode et al. (2Nm)	1(↓) ± 4.9	3.6±5.6	6.3± 5.5
Current FE model (2Nm)	3.4	1.2 (↓)	1.6 (↓)
DL 4-level
Kubo et al. (C2-C7 and 1.5Nm)	2.6±8	6.2±7	8.4±12
Current FE model (2Nm)	12	2.4 (↓)	3.6 (↓)

In addition, the current finite element predictions demonstrated similar trends compared to our in house studies, as well as to the data available in the literature. Minor differences between the current finite element predictions and the literature data could be attributed to several factors. During laminoplasty, the initial stresses developed in the ligaments while opening the lamina may have restricted the ROM during lateral bending and axial rotation. The current study did not resect any capsular ligaments or facet joints to simulate either of the surgical procedures. In contrast, Nowinski et al.³⁰ altered the facet capsules for stabilizing the lamina in an open position using sutures after laminoplasty. They along with Kallemeyn et al.¹⁴ have shown that alteration of the capsular ligaments or facets could significantly increase the ROM.

Since we have established confidence in the intact and surgically altered finite element models through comparisons with experimental data, the models can be used to evaluate changes in the internal biomechanical responses that are otherwise difficult to obtain. For example,the increased disc stresses seen at the altered and adjacent levels after multi-level laminectomy correspond well with previous biomechanical studies^31,32 and may be clinically correlated to the process of disc degeneration.

A limitation of this study is the absence of muscles contributing to spinal stability, both experimentally and computationally. In addition, the tied contact interactions assigned to the laminoplasty constructs assumes a structural rigidity (i.e. fusion) that was not matched in the in vitro conditions. Moreover, a force controlled assessment rather than a hybrid control was chosen so as to allow a direct comparison to the existing experimental studies conducted after laminoplasty and laminectomy.

The current study showed laminoplasty as superior to laminectomy in terms of ROM at the altered and unaltered levels. Finite element predictions suggest the preservation of ROM after open door laminoplasty, which was in agreement with our experimental results. It also addressed the role of ligaments in maintaining the stability of the cervical spine as extensive ligament resection could substantially affect the motion following a DDL. It also showed that unilateral resection of ligaments during ODL results in asymmetric distribution of stresses/motions during lateral bending and axial rotation.