Spinal instrumentation is commonly used with posterolateral spinal arthrodesis and serves to stabilize the treated spinal segment until a solid osseous union is achieved. Numerous clinical studies have documented that rigid spinal instrumentation leads to a higher rate of fusion than semi-rigid fixation or arthrodesis without instrumentation3,31,33,34. In addition, a biomechanical study performed with a posterior spinal arthrodesis model showed that rigid fixation provided a stiffer fusion mass15. Thus, it is likely that rigid internal fixation creates a better mechanical and biological environment for fusion. Despite this, anterior vertebral bone may deteriorate as a result of device-related osteoporosis6,7,21,28. Although load-sharing of spinal instrumentation should vary at different stages of the healing process, it remains unclear how the load distribution through the fusion mass and the spinal instrumentation changes as the fusion mass develops.
The objectives of the present study were to investigate the time-related changes in the biomechanical, radiographic, and histological properties of a posterolateral spinal fusion mass and to examine the load-sharing characteristics of the spinal instrumentation and the posterolateral fusion mass throughout the healing process.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Funds were received in total or partial support of the research or clinical study presented in this article. The funding source was Towson Orthopaedic Research Foundation, Towson, Maryland.
†Department of Orthopaedic Surgery, Hokkaido University School of Medicine, Kita-15, Nishi-7, Kita-ku, Sapporo 060, Japan.
‡Orthopaedic Biomechanics Laboratory, The Union Memorial Hospital, 201 East University Parkway, Baltimore, Maryland 21218.
§Scoliosis and Spine Center, O'Dea Medical Arts Building, 7505 Osler Drive, Suite 104, Towson, Maryland 21204. Please address requests for reprints to Dr. McAfee.
Animal Model and Operative Procedure
Twenty-four skeletally mature crossbred Western sheep weighing 600 to 700 newtons were used in this study after the protocol had been approved by the Institutional Animal Care and Use Committee.
Anesthesia was induced with 0.15 milligram of diazepam per kilogram of body weight and ten milligrams of ketamine per kilogram of body weight, both administered intravenously. The animals were then anesthetized with endotracheal inhalation of 2 per cent halothane. Perioperatively, the animals were given prophylactic antibiotic therapy in the form of one gram of cefazolin sodium administered intravenously.
A single midline incision was made through the skin and fascia after sterile preparation. The paraspinal muscles were stripped subperiosteally from the spinous processes and the laminae to the transverse processes to expose the posterior osseous elements, including the transverse processes, from the third to the sixth lumbar vertebra. Destabilization of the posterior spinal column at the motion segments between the third and fourth and the fifth and sixth lumbar vertebrae as well as excision of the bilateral facet joints, spinous processes, and supraspinous and interspinous ligaments was performed in all animals. After the destabilization, transpedicular screw fixation was performed at the segments between the third and fourth and the fifth and sixth lumbar vertebrae with the Texas Scottish Rite Hospital spinal instrumentation system (donated by Sofamor Danek, Memphis, Tennessee), which consists of stainless-steel screws (5.5 millimeters in diameter and thirty millimeters long) and rods (4.76 millimeters in diameter). Corticocancellous bone for a graft was then obtained from the iliac crest through the same skin incision. The spinous processes that had been excised during the destabilization procedure were also used as bone graft. A posterolateral spinal arthrodesis, with twenty grams of autologous corticocancellous bone, was performed randomly at either the motion segment between the third and fourth lumbar vertebrae or the motion segment between the fifth and sixth lumbar vertebrae after decortication of the transverse processes. No bone graft was used at the other segment, which served as the instrumented control (Fig. 1).
The animals were given analgesics (4.5 milligrams of phenylbutazone per kilogram of body weight) orally and prophylactic antibiotic therapy (one gram of cefazolin sodium) intramuscularly each day for the first postoperative week. Six animals were killed with an overdose of pentobarbital at each of four time-periods: four, eight, twelve, and sixteen weeks after the procedure. The spines were removed and were kept frozen at -20 degrees Celsius until mechanical testing.
The identical operative procedure was performed on six fresh-frozen sheep spines, which served as zero-week controls.
Mechanical Testing and Measurement of Strain on the Hardware
The frozen specimens were thawed at room temperature, and the paraspinal soft tissues were removed to obtain the ligamentous spinal specimen (the third through sixth lumbar vertebrae). The specimen was divided into functional spinal units at the third and fourth lumbar vertebrae and at the fifth and sixth lumbar vertebrae. The cephalad and caudad vertebrae of each functional spinal unit were anchored with stainless-steel screws and were secured in metal fixtures with polyester resin. All biomechanical testing was performed with use of a biaxial materials testing machine (MTS 858; Bionix Testing System, Minneapolis, Minnesota) interfaced with an IBM PS/2 computer through a high-speed analog-to-digital converter (DASH 16F; Metrobyte, Taunton, Massachusetts).
To measure the strain on the hardware, uniaxial strain-gauges (model CEA-06-125UN-350; Measurements Group, Raleigh, North Carolina) were affixed with waterproof coating to two opposing surfaces of the 4.76-millimeter-diameter stainless-steel rods. The rods that had been implanted at the time of the operation were replaced bilaterally with the strain-gauge-equipped rods. During the exchange of the hardware, the specimen was kept on the materials testing machine under displacement and rotation controls, which allowed consistent alignment of the spine during the exchange. Care was taken to ensure that no preloads were applied to the functional spinal units. The strain-gauges were positioned equidistant from the screw-rod junctions and were aligned sagittally on one rod and coronally on the other (Fig. 2). These rod arrangements allowed the gauges to measure anterior and posterior surface strains on one rod and medial and lateral surface strains on the other. All strain data were acquired with use of a multichannel signal-conditioning amplifier (2100 System; Measurements Group) interfaced with an IBM-PC computer.
Each functional spinal unit was tested non-destructively under axial compression (at 500 newtons), torsion (±6 newton-meters with 150 newtons of compressive preload), flexion-extension (±6 newton-meters), and lateral bending (±6 newton-meters) with use of a modified loading system18. The rate of loading was 100 newtons per second for axial compression and 1.2 newton-meters per second for the rotational testing modes. The axis of rotation was centered at the junction of the posterior one-third and anterior two-thirds of the intervertebral disc during torsion and flexion-extension testing. Lateral bending testing adhered to the rotational axis determined by the mid-sagittal point of the intervertebral disc. Each test was repeated for five loading and unloading cycles. During axial compression and flexion-extension, the strain on the hardware was measured on the rod with the sagittally aligned gauges. During lateral bending, the strain was measured on the rod with the coronally aligned gauges. The specimens were continuously moistened during the testing with 0.9 per cent sodium chloride irrigation solution.
After each functional spinal unit had been tested with the spinal instrumentation, the instrumentation was carefully removed without damage to the posterolateral fusion mass and the mechanical tests were repeated in the same manner to examine the mechanical stiffness of the posterolateral fusion mass.
Stiffness was calculated as the peak applied load divided by the elastic zone, which is the difference between the range of motion (maximum displacement) and the neutral zone (displacement at the zero-load point)30. The data from the fourth loading cycle were used for the calculation of stiffness. Torsional, flexion, extension, and lateral bending stiffnesses were defined as a ratio of applied torque (in newton-meters) to angular deformation (in degrees). Compressive stiffness was calculated as a ratio of applied force (in newtons) to displacement (in millimeters).
The strain on the rod was measured at the peak load during the fourth loading cycle. In the most common loading situation, the rods may be subjected to two force components: axial and bending. The surface strain on the rod can be described by a combination of the two measured strains due to axial and bending forces11. Thus, the axial stress imposed by the axial load was calculated as the elastic modulus multiplied by one-half the sum of the surface strains. The bending stress imposed by either anterior-posterior bending or lateral bending was the elastic modulus multiplied by one-half the difference between the surface strains.
All data were given as the mean and the standard deviation. Statistical analyses were performed with one-way analysis of variance and the Scheffé F test as a post hoc multiple-comparison procedure. Statistical significance is indicated at p < 0.05.
Radiographic Assessment
Anteroposterior and lateral plain radiographs were made of the spines under consistent conditions of ten milliamperes, fifty kilovolts, and 0.07 second. The status of the fusion was evaluated on the plain radiographs with use of the grading system documented by Lenke et al.19. With this system, A indicates a big, solid trabeculated bilateral fusion mass (definitely solid); B, a big, solid unilateral fusion mass with a small contralateral fusion mass (possibly solid); C, a small, thin bilateral fusion mass with an apparent crack (probably not solid); and D, bilateral resorption of the graft or fusion mass with an obvious bilateral pseudarthrosis (definitely not solid).
The radiographs were assessed independently by three orthopaedic surgeons who were blinded with regard to the mechanical and histological data. The rate of radiographic fusion was calculated at each time-period by averaging the results of the three observers. After mechanical testing and removal of the spinal instrumentation, computerized tomography scans were made to assess the posterolateral fusion mass in cross section and in three dimensions. For each fusion mass, approximately twenty-five sequential computerized tomography scans were made with use of three-millimeter slice intervals under the same magnification and radiographic conditions.
Histological and Histomorphometric Analysis
After the mechanical testing and radiographic examinations had been completed, the specimens were subjected to undecalcified tissue-processing. The posterolateral fusion masses, including the transverse processes, were sliced sagittally into ten-millimeter-thick sections at the thickest part of the mass. The sliced specimens were fixed and dehydrated in 100 per cent ethanol and were stained with Villanueva osteochrome bone stain (Polysciences, Warrington, Pennsylvania). The specimens were infiltrated sequentially with 100 per cent ethanol, 100 per cent acetone, 50 per cent acetone-50 per cent methylmethacrylate monomer, and 100 per cent methylmethacrylate monomer. The specimens were then embedded in polymethylmethacrylate and were kept in a 48-degree-Celsius water bath for gradual polymerization. After polymerization had been completed, the embedded specimens were cut into 300 to 500-micrometer-thick sections with an Isomet low-speed saw (Buehler, Lake Bluff, Illinois) with a diamond cutting wheel. These sections were mounted on acrylic slides and were then ground and polished to 140 micrometers on a microgrinder (EXAKT-MicroGrinding System; Exakt Technologies, Oklahoma City, Oklahoma). The slide-mounted specimens were evaluated with light microscopy. Microradiographs were also made for each slide-mounted specimen with use of the Faxitron x-ray cabinet system (Faxitron X-Ray; Buffalo Grove, Illinois) under consistent conditions (three milliamperes, twenty-five kilovolts, and 120 seconds).
Histomorphometric quantitative analysis also was performed on the microradiographs with use of the Bioquant Image Analysis System (R and M Biometrics, Nashville, Tennessee). The total area of the fusion mass (in square millimeters) and of the trabecular bone (in square millimeters) was measured within a section that was ten by twenty millimeters. These quantitative parameters were used to calculate the relative trabecular bone area (the trabecular bone area divided by the total area of the fusion mass).
Mechanical Testing
The posterolateral fusion masses at eight, twelve, and sixteen weeks were significantly stiffer than the zero-week controls or the fusion masses at four weeks (p < 0.01) (Fig. 3); with the numbers available, we could detect no significant differences among the eight, twelve, and sixteen-week groups. Although the instrumented controls had a tendency to be stiffer at eight weeks and subsequently, no significant difference was found, with the numbers available, between the zero-week controls and the instrumented controls at four, eight, twelve, or sixteen weeks. The posterolateral fusion masses were significantly stiffer than the respective instrumented controls at twelve and sixteen weeks (p < 0.05). With the exception of axial compression, the loading modes (flexion, extension, and lateral bending stiffness) had similar results (Table I). In axial compression, the zero-week control was significantly less stiff (p < 0.05) than all of the fusion masses, but the four, eight, twelve, and sixteen-week fusion masses were not found to be significantly different from each other, with the numbers available.
Measurement of Strain on the Hardware
During axial compression and flexion-extension, strain data were obtained from the rods with the sagittally aligned gauges. During lateral bending, the data were obtained from the rods with the coronally aligned gauges. Specifically, anterior and posterior surface strains were measured during axial compression and flexion-extension, whereas medial and lateral surface strains were recorded during lateral bending. A positive value represented tensile strain, and a negative value represented compressive strain (Table II). In lateral bending, the strains on the rod at the segment with the fusion mass significantly decreased beginning at eight weeks (p < 0.05). However, with the numbers available, no significant difference was found among the zero-week-control and postoperative groups. In left lateral bending, the rods were always subjected to compressive strain on the medial (left) surface and to tensile strain on the lateral (right) surface; these parameters were reversed for right lateral bending. In flexion and extension, the strain on the rod at the fusion mass was significantly decreased at sixteen weeks compared with the strain at eight weeks (p < 0.05). In flexion, there was compressive strain on the anterior surface of the rod and tensile strain on the posterior surface; these parameters were reversed in extension. In axial compression, the strains on the rod at the fusion mass significantly decreased at eight weeks (p < 0.05), and the rod was always subjected to compressive strain on the anterior surface and tensile strain on the posterior surface.
Analysis of Axial and Bending Stress on the Rod
To better understand the load-sharing changes in the rod, the two stress components—axial and bending stress—were calculated separately from the respective strain components, as previously described (Fig. 4-A and Fig. 4-B). The rod was subjected mainly to bending stress rather than to axial stress in this loading condition. The bending stress on the rod at the fusion mass decreased significantly at eight weeks (p < 0.05). Although the bending stress on the rod at the instrumented control segment had a tendency to decrease by twelve weeks, with the numbers available no significant difference was observed between the zero-week and postoperative control groups. Axial stress on the rod was quite small compared with the bending stress, and axial stress did not change significantly throughout the postoperative period. Analogous findings were obtained for the other loading modalities (Table III and IV).
Radiographic Assessment
According to the qualitative radiographic assessment by the three independent observers, all of the sixteen-week fusion masses were solid osseous unions, whereas bridges of trabecular bone were noted during only ten of eighteen observations at twelve weeks, three of eighteen observations at eight weeks, and none of eighteen observations at four weeks (Fig. 5-A 5-B 5-C and 5-D). In this radiographic assessment, we defined the A and B classifications19 as solid osseous union and the C and D classifications as non-union.
Computerized tomography scans demonstrated that the posterolateral fusion masses consisted of only morselized bone chips surrounded by soft tissue at four weeks, whereas the bone graft was gradually incorporated after eight weeks. A solid osseous fusion mass was confirmed at sixteen weeks (Fig. 6-A 6-B 6-C and 6-D).
Histological and Histomorphometric Analyses
Histological sections of the posterolateral fusion masses were obtained at four, eight, twelve, and sixteen weeks, with a parasagittal section made at the thickest part of the mass. The fusion mass at four weeks primarily comprised morselized pieces of bone graft and surrounding fibrous stroma, whereas partial formation of woven bone was observed around the subperiosteal regions of the transverse processes (Fig. 7-A). At eight weeks, the fusion mass, which partly bridged the transvers processes, predominantly consisted of woven bone (Fig. 7-B). At twelve weeks, a partially trabeculated osseous fusion mass connected the transverse processes from one vertebral level to the next but the bridge was still incomplete (Fig. 7-C). The fusion mass was shown to consist entirely of trabeculated bone at sixteen weeks (Fig. 7-D).
Histomorphometric analysis was performed on the microradiographs. The results demonstrated that the trabecular bone area of the fusion mass increased linearly even after eight weeks (Fig. 8). Importantly, although the posterolateral fusion mass was mature biomechanically at approximately eight weeks, mineralization continued until sixteen weeks.
We used sheep for the posterolateral spinal arthrodesis model because, although the anatomy of a sheep spine differs from that of a human spine (a sheep has six lumbar vertebrae, larger transverse processes, and a smaller vertebral diameter), the biology of bone-graft healing is similar5,9. In addition, because the pedicles in sheep have a large diameter, human-sized transpedicular screws could be used for the posterolateral spinal arthrodesis.
We tried to determine when solid osseous union actually occurs after a posterolateral spinal arthrodesis. Clinically, numerous radiographic studies have documented the time to union after posterolateral spinal arthrodeses in humans, but the interval until evidence of a successful spinal fusion has varied from six to ten months postoperatively10,13,17,33. In addition, a few investigators have operatively explored the site of an arthrodesis to determine the status of the fusion early in the postoperative period14,22-24,26. Outland et al.26 performed such an operative exploration and documented, four to five months postoperatively, solid fusion at the site of sixty three (89 per cent) of seventy-one arthrodeses that had been performed to treat scoliosis. In a study by James14, operative exploration revealed successful fusion, five months postoperatively, at the sites of seventy (80 per cent) of eighty-seven arthrodeses that had been performed to treat scoliosis. Hence, there is a discrepancy between the results of radiographic assessment and those of operative exploration, and the time to union after a posterolateral spinal arthrodesis remains controversial1,4,25. Biomechanical and histological examinations are currently the only methods with which to evaluate the status of the fusion site accurately, but they are difficult to perform in the clinical setting.
Our study demonstrated maturation of the posterolateral fusion mass approximately eight weeks postoperatively, preceding the solid osseous union as confirmed radiographically. Thus, mineralization in the fusion mass continued after the mass had achieved sufficient mechanical strength. These results seem to be consistent with those of the biochemical study by Slater et al.27, who showed the posterolateral fusion mass in a sheep model to be composed of homogeneous materials, such as cancellous bone graft, at twelve weeks postoperatively. Although several investigators have performed histological analyses during sequential stages of bone-graft healing, the healing mechanism after posterolateral spinal arthrodesis remains controversial and unclear2,12,16,29. Furthermore, it remains poorly understood how the histological properties of the posterolateral fusion mass correlate with its mechanical strength during the healing process. We characterized the fusion mass histologically as consisting of immature woven bone at eight weeks postoperatively, despite the fact that the mass had obtained substantial mechanical stiffness. This suggests that immature woven bone can provide sufficient strength to the fusion mass.
Failure of the implant is one of the major complications of posterolateral spinal arthrodesis with instrumentation; the rates have been reported to be 7 per cent (410 of 5756 patients)32, 8 per cent (168 of 2153 patients)33, and 10 per cent (twelve of 120 patients)20. Although pseudarthrosis often leads to failure of the implant, the use of bone graft to augment the anterior column was shown to decrease markedly the rate of failure of transpedicular screws used to fix thoracolumbar burst fractures8. Thus, the load-sharing characteristics of spinal instrumentation and the posterolateral fusion mass are quite relevant to the clinical setting. We demonstrated that the load-sharing properties of spinal instrumentation decreased concurrently with the development of the spinal fusion mass. In the early period after posterolateral spinal arthrodesis with instrumentation, the load across the fused segment was mainly borne by the hardware, indicating that the fusion mass made no contribution to the initial strength or stability of the treated segment. As the fusion mass developed and became stiffer with time, the load across the fused segment was distributed primarily to the solid fusion mass, resulting in unloading of the instrumentation. Importantly, these changes in load-sharing occurred when the fusion mass became mechanically mature, before solid fusion was confirmed radiographically or histologically.
NOTE: The operations on, and the postoperative care of, the animals were performed at the Thomas D. Morris Surgical Research Facility, Reisterstown, Maryland.