Chapter 16 Biology of Spine Fusion
In 2001, more than 185,000 spinal arthrodeses were performed in the United States, the majority of which were posterolateral lumbar intertransverse process fusions. From 1997 to 2003, spine fusions climbed from the 41st most common inpatient procedure to the 19th, with resultant increases in spending on lumbar fusion procedures (from 75 to 482 million dollars).1,2 The number of fusion procedures continues to increase, as does the complexity of devices available for treatment of various spine disorders. Nonunion rates, however, for single-level fusions have been reported to be as great as 35% and even higher in multilevel procedures.3 Pseudarthroses often result in outcomes that are less than optimal, and often necessitate further surgery.
Local Factors
To achieve a successful fusion, multiple factors must work in concert. These include the local environment of the fusion and systemic factors, with or without the use of fusion enhancers (Box 16-1). Mechanical and biologic factors are closely linked, and any cogent discussion of the biology of spine fusion must be limited to a particular mechanical situation (e.g., compressive forces–anterior column, tensile forces–intertransverse process). This chapter focuses primarily on the biology involved with fusion in the submuscular lumbar intertransverse process environment. To cover the differences and details of all potential fusion environments in the spine is beyond the scope of this chapter. Moreover, one must be cautious in extrapolating results of healing and fusion properties for bone graft substitutes in one region of the spine for another.4 Nevertheless, some of the principles are applicable and important for anyone who has dedicated a career to the advancement of spine surgery.
Graft Properties: Osteoinduction, Osteogenicity, Osteoconduction, and Connectivity
The choice of graft material has profound implications for the success or failure of arthrodesis. The ideal graft is osteogenic, osteoinductive, and osteoconductive. A balance of these entities, with or without instrumentation, ensures a favorable environment for fusion. Osteoinduction is the stimulation of multipotential stem cells to differentiate into functioning osteogenic cells. This is mediated by growth factors in the bone matrix itself (i.e., BMPs). Urist et al. introduced this concept in their studies of the osteoinductive properties of demineralized bone matrix (DBM).5,6
Graft Material
Autograft
Autogenous iliac crest bone in the past has been considered the “gold standard” of graft material. Historically, it has been the most successful graft source in spine fusion. Cancellous autograft has the requisite matrix proteins, mineral, and collagen for the ideals of osteoinductivity, osteogenicity, and osteoconductivity. Its large trabecular surface makes it highly connective as well. Donor site complication rates as high as 25% to 30% have been reported, although a rate of 8% seems more realistic and is more commonly cited.7,8 Morbidity may be associated with an increased incidence of blood loss, chronic donor site pain, increased operative time, infection, and nerve injury. Furthermore, the quantity of bone available is limited, and may be insufficient for long-segment fusions, or in patients who have had previous graft harvests.
Demineralized Bone Matrix
Demineralized bone matrix (DBM) is present in a variety of forms, each of which has a variable degree of osteoinductivity.4 Current data support its use as a bone graft extender but not as a pure substitute or enhancer.9 Autogenous and allograft DBMs are osteoinductive due to the presence of low doses of BMPs (~0.1% by weight).10 Collagenous and noncollagenous proteins serve as osteoconductive material that is left after the demineralization process. Preliminary animal studies have shown efficacy of DBM as a carrier molecule for recombinant human bone morphogenetic protein (rhBMP-2) in ectopic bone formation or as a graft alternative in experimental posterolateral arthrodesis.11,12 A study by Louis-Ugbo et al. used a non-human primate posterolateral fusion model to test a specific formulation of DBM, which was more porous than its predecessor. With this new formulation they found robust fusions and suggested that it exhibited properties of both a graft enhancer and extender.13 A recent prospective randomized study by Cammisa et al. looked at 2-year fusion rates in the posterolateral environment in 120 human patients. They concluded that DBM could function as an adequate graft extender and promote adequate fusions when mixed with a small amount of autogenous bone graft.14
Allograft
The desire to avoid donor site morbidity led to increased use of allograft bone in spine surgery. This was made practical by advances in procurement, sterilization, preparation, and storage.15 Although allograft bone is widely used in spine surgery, concerns regarding fusion rates and disease transmission remain. Allograft is not osteogenic, because there are no surviving cells in the graft. Because of the processing and storage requirements of allograft, some of the osteoinductive potential is lost. It also carries a small but real risk of disease transmission and may elicit an immune response from the recipient.16–21
Immunogenicity and maintenance of osteoinductive and osteoconductive properties are affected by these processing and preservation techniques. Bone usually is frozen or freeze-dried as soon as possible after harvest. Both methods decrease immunogenicity and allow for extended storage. Freezing does not diminish the mechanical properties of the bone, and it may be stored at −70°C for 5 years. Freeze-drying further reduces immunogenicity and inactivates viral agents, but it reduces the mechanical strength of the graft.22–24 Freeze-dried bone may be stored under vacuum, at room temperature, for an indefinite period.
The most common sterilization methods are high-dose gamma irradiation and ethylene oxide gas sterilization.18 Both alter the structure of matrix proteins, decreasing the osteoinductive capacity and mechanical strength of the bone.23 Other sterilization methods such as autoclaving are even more destructive, and generally are not used.
Although allograft usually has performed well in both cervical and lumbar interbody fusions, in which the graft is subject to compression, the results in the posterolateral lumbar environment, in which primary tensile forces exist, have not been as favorable.25–33 This result has led many surgeons to use allograft as an autograft expander rather than a pure substitute for posterolateral arthrodeses.
Xenograft
The use of bone graft taken from other species has been reported in the orthopaedic literature, but these alternatives to autograft never were incorporated into widespread use. Despite processing, xenografts remain immunogenic and provoke a host response. The graft may become encapsulated from the host’s response, with resultant blockade to revascularization. Ivory and cow horn resist incorporation into host bone and are no longer used. Bovine bone, both freeze-dried and deproteinized, remains weakly antigenic. Both of these types of xenografts have been used in spine surgery with limited success, and therefore they are not recommended as a graft material.34–37
Ceramics
Calcium phosphate (CaPO4) ceramics, including hydroxyapatite (HA) and tricalcium phosphate (TCP), have been widely used in orthopaedic and spine surgery.38 These osteoconductive, biodegradable materials are compatible with the remodeling of bone necessary to achieve optimal strength of a construct. Other, nonresorbable materials remain in the fusion mass, leaving permanent stress risers and prolonging strength deficiencies.
To be useful as a graft material, synthetic materials must have several properties. They must be compatible with local tissues, remain chemically stable in body fluids, and be able to withstand sterilization. Furthermore, they must be available in useful shapes and sizes, be cost-effective, and have reliable quality control. CaPO4 ceramics qualify, and have been widely used in dentistry and maxillofacial surgery, as well as in animal models for experimental spinal research.21,39–47 A wide body of literature exists discussing the use of these materials in human spine surgery.38,46,48–51
Both HA and TCP ceramics are inherently brittle. They may be prepared as either a compact or a porous material. The greater crystallinity and density of the compact form results in greater strength and resistance to dissolution in vivo, whereas porous versions more closely approximate cancellous bone and enhance bony ingrowth (at the expense of more rapid degradation). Under physiologic conditions, HA is resorbed very slowly, whereas TCP generally is resorbed within 6 weeks of implantation.25
Natural coral has been used to augment or even replace autograft, with some success.52–55 The calcium carbonate (CaCO3) in coral is hydrothermally converted to CaPO4. The structural geometry of coral is similar to that of cancellous bone, making it highly osteoconductive and connective.
Animal Studies
The use of CaPO4 ceramic as a spine fusion bone graft substitute has been studied extensively in animal models. Flatley et al. used porous blocks of a 1:1 ratio of calcium HA and TCP ceramic in a rabbit posterolateral fusion model.21 At 12 weeks, histologic sections demonstrated bone ingrowth reaching the central portion of the block with no fibrous barrier between the new bone and the ceramic. Holmes et al. used coralline HA in a canine posterior facet model.46 Although the distribution of bone ingrowth was similar to that seen in autograft controls, they reported no solid fusions, even at 6 months. Using coral porites (calcium carbonate) and a 65:35 HA:TCP biphasic ceramic, Guigui et al. found a 100% rate of fusion in a sheep model, comparable to the fusion rate of autograft in another study by this same group.56,57
The use of composites of ceramic and an osteoinductive agent such as DBM, autograft, or recombinant BMP also has been investigated (see Growth Factors, later in this chapter).58–60 Ragni and Lindholm, in a rabbit interbody fusion model, found that the addition of DBM enhanced the incorporation of an HA block. Animals treated with an HA/DBM composite showed significantly earlier fusion consolidation than those treated with autograft or either HA or DBM alone. By 6 months, however, results of the autograft were comparable to those with the composite.61 Zerwekh et al. compared a collagen/HA-TCP ceramic/autograft composite with autograft alone in a canine posterior fusion model.62 Histologic comparisons of bone ingrowth were similar in both groups at 12 months, as were the results of biomechanical testing. Working in a canine segmental posterior spine fusion model, Muschler et al. compared fusions with autograft, collagen/HA-TCP ceramic composite, collagen/HA-TCP ceramic/autograft composite, and collagen/HA-TCP ceramic/bone matrix protein composite, and with no graft.63 Autograft had a significantly superior union score. Ceramic composite alone performed no better than the no-graft control. The addition of bone matrix protein, however, improved the union score, making it comparable with the composite/autograft treatment.
Human Studies
The clinical efficacy of ceramics, either alone or as part of a composite, has yet to be fully elucidated. Studies suggest that these entities do have beneficial effects. Passuti et al., in a study of 12 severely scoliotic patients, used internal fixation and blocks of 3:2 HA-TCP ceramic alone or mixed with autogenous cancellous bone.49 After 15 months average follow-up, radiographs demonstrated fusion in all patients. Histologic examination of biopsy material from two of the subjects revealed new bone formed directly on the ceramic surface and ingrowth into the macropores. Similarly, Pouliquen et al. successfully used natural coral as a graft substitute in 49 patients with idiopathic scoliosis.54 Although the results were favorable, their small patient populations, single diagnosis, and average patient age of 14 years limited these studies. Acharya et al. designed a prospective matched case study examining the effect of a hydroxyapatite–bioactive glass ceramic composite as a stand-alone graft versus autogenous bone in posterolateral spine fusion.64 The study was halted early, because at 1 year fusion was found to be inferior with the bone substitute as a stand-alone graft compared with autograft.
Mechanical Stability
Fusion rate is affected by the mechanical stability of the involved segments.65–70 As a result, internal segmental instrumented fixation has commonly been used to achieve higher rates of fusion, an approach that is supported by various studies in the literature.65–67,70–72 Even in the presence of a rigid construct, nonunion still occurs in up to 10% to 15% of patients, especially when hardware loosening or failure occurs.71–75 Fusion level, number of segments involved, patient weight and activity level, and postoperative bracing all influence the rate of fusion.76
Animal Studies: Spinal Instrumentation
The effects of spinal instrumentation and stability have been investigated in various animal models.77–81 Although this approach can tell us much about short-term effects of instrumentation failure, caution must be exercised in extrapolating this information to the long-term effects in the human body at the bone/instrumentation interface. McAfee et al. created a canine instability model to study both the effect of spinal instrumentation on fusion success and the radiographic incidence of fusion with respect to spinal stability.66–68 At 6 months, radiographs revealed a greater probability of fusion in the instrumented animals than in the noninstrumented animals. The instrumented fusions also were more rigid. Likewise, Zdeblick et al. demonstrated both an increased rate of fusion and a more rigid fusion when anterior instrumentation was used in a canine model of an unstable burst fracture at L5.70 These results also were replicated by Shirado.82 Kotani et al. showed that after solid posterolateral arthrodesis was achieved in a sheep model, transpedicular fixation continued to provide mechanical support.83
The biologic activity of the graft material may partly determine the need for internal fixation. Fuller et al. showed that rigid fixation improved bone ingrowth into a calcium carbonate block in a canine anterior thoracic interbody fusion model.84,85 Because ceramics are not osteoinductive, a mechanically stable environment is crucial for ingrowth. Osteoinductive graft substitutes, on the other hand, may not be as reliant on construct rigidity.
Nagel et al. developed a sheep model of delayed union and nonunion.69 Posterior lumbar laminar and facet fusions with iliac crest graft were performed on seven sheep. Six of the seven sheep developed nonunions at the L6-S1 interspace; all cephalad interspaces fused (21 of 21). Eight normal sheep underwent in vivo flexion-extension radiographs. Five normal sheep spines were studied ex vivo, using displacement transducers to test stiffness, displacement, and strain in flexion-extension. The lumbosacral level demonstrated significantly more motion than the other levels, suggesting that motion was a major factor in determining the success of fusion in this sheep model. Similar observations have been made in dogs.86 The increased stability and decreased motion that instrumentation provides would seem valuable in such instances.
Human Studies: Spinal Instrumentation
Contradictory human studies of the effects of spinal instrumentation have been widely reported. Zdeblick discussed 124 patients undergoing fusion for different conditions.72 Patients were randomized into three groups, all having dorsolateral autograft fusions. Patients in group 1 were not instrumented, those in group 2 were instrumented with a semirigid pedicle screw system, and individuals in group 3 had rigid pedicle screw instrumentation implanted. The rigid group had a significantly higher fusion rate (95%) than the noninstrumented group (65%). The instrumented groups together had 95% excellent or good results, whereas the noninstrumented patients had only 71% good or excellent outcomes (a statistically significant result).
Bridwell et al. described 44 patients with degenerative spondylolisthesis.71 Patients were individualized into three groups: no fusion; noninstrumented posterolateral fusion; and pedicle screw instrumented posterolateral fusion. Patients with more than 10 degrees or 3 mm of motion were automatically assigned to the instrumentation group. There was an 87% fusion rate in the instrumented group versus a 30% rate in noninstrumented patients, yet there was no significant clinical difference in successful outcomes between the noninstrumented and unfused groups (30% vs. 33%). Successful outcomes in the instrumented group (83%) were significantly greater than in the nonfusion group. Fischgrund et al. also demonstrated a markedly increased rate of fusion in their patients with instrumentation (83% vs. 45%), yet found no difference in clinical outcome.
In their meta-analysis, Mardjetko et al. reviewed 25 papers describing 889 patients with degenerative spondylolisthesis.87 Five of the included studies described patients undergoing decompression and posterolateral arthrodesis with pedicle screw instrumentation. Although there was a trend toward an increased rate of fusion in the instrumented versus noninstrumented patients (93% vs. 86%), it did not reach significance (P = .08). The clinical outcome was better in the uninstrumented group: 90% versus 86%. However, the authors acknowledged several limitations of their review: data from different treatments over 20 years; variable study designs and quality; and possible dilution of data from the stronger, better-designed studies that suggested an advantage to instrumentation.
Fusion success is also affected by the physical stresses placed on the graft.88 In human beings, 80% of the load at a motion segment is transmitted through the intervertebral disc. Graft placed ventrally, in the interbody region, is thus primarily subjected to compression. This compressive force promotes fusion, presumably by stimulating vascular ingrowth and the proliferation of mesenchymal cells. Dorsally placed graft experiences tensile forces, as does graft placed in the intertransverse process region. Under these less favorable mechanical conditions, fusion is more dependent on biologic factors.
Facet preparation for fusion has been shown to increase motion of the involved segment. Although many surgeons routinely include facet fusion in posterolateral intertransverse process arthrodeses, biomechanical studies have demonstrated a resultant decrease in stability.89,90 The developing fusion preparation decreases the surface area incorporated into the fusion mass, and may result in a less rigid fusion. Rigid instrumentation allows the facets to be prepared and incorporated without sacrificing stability. However, in the osteoporotic patient, the screw-bone interface often is weak. Even with instrumentation, facet preparation may not be appropriate in these individuals.
Overall, it is generally agreed that spinal instrumentation decreases the rate of pseudarthrosis. However, in some situations, especially with single-level fusions, no significant clinical benefit may be obtained. Additionally, although a positive relation exists between radiographic fusion and clinical outcome, no absolute convincing correlation has been demonstrated.91 Currently, prospective randomized blinded clinical trials examining the effects of instrumentation have not yet been completed.
Graft Site Preparation
Preparation of the bony anatomy into which the graft is placed is of paramount importance in achieving a successful fusion. The exposed area of viable, vascular bone should be maximized. This is done by decortication, which may be accomplished with curettes, rongeurs, osteotomes, or a power bur precision drill. Use of a high-speed bur may result in thermal necrosis, which may be minimized with continuous irrigation, use of a drill with deep flutes, and minimizing contact time between bur and bone. As the surface area of decorticated bone increases, so does the connectivity and the availability of osteogenic cells and exposed matrix proteins. Furthermore, a large surface is helpful in forming a bony bridge strong enough to carry the mechanical load.
Soft Tissue Bed
Hurley et al. evaluated the role of these factors in a canine dorsal spine fusion model.86 Thirty-seven animals underwent a modified Hibbs fusion (control), a Hibbs fusion with a fluid-permeable/cell-impermeable membrane interposed between fusion site and muscle mass, or with a membrane impermeable to both cells and fluids. Fusion was achieved in all 12 animals with the semipermeable membrane, whereas no animals with the impermeable membrane had a successful fusion.
Radiation has a detrimental effect on a healing spine fusion that is especially pronounced in the first few postoperative weeks. This effect may be caused by cytotoxicity, but is probably a product of the resultant vasculitis and inhibition of angiogenesis that follows radiation treatment. In the long term, radiation also can induce osteonecrosis and dense hypovascular scars, creating a poor fusion environment. Studies suggest that a 3- to 6-week delay in radiation would be beneficial to the fusion process.92,93 Use of vascularized grafts anastomosed to nonirradiated vessels also may increase the chance of successful fusion.
Systemic Factors
Nicotine
Smokers have a higher rate of pseudarthrosis than do nonsmokers.27,28,73,94,95 Cigarette smoke retards osteogenesis and inhibits graft revascularization. Tobacco smoke extracts calcitonin resistance, increases fracture end resorption, and interferes with osteoblastic function.8,96,97
A direct relation between systemic nicotine and spinal pseudarthrosis has been demonstrated in a rabbit model. Silcox et al. performed L5-6 posterolateral intertransverse process arthrodeses with autologous iliac crest graft in 28 rabbits.98 The animals were implanted with osmotic mini-pumps, delivering either saline (control) or nicotine equivalent to a human who smokes 1 to 1.5 packs per day. At 5 weeks, 56% of control animals had a solid fusion by manual palpation; no solid fusions were seen in the nicotine-exposed animals (P = .02).
Drugs
Drugs taken during the perioperative period can have a detrimental effect on the process of fusion. Chemotherapeutic agents administered in the early postoperative period inhibit bone formation and arthrodesis.99–101 Nonsteroidal anti-inflammatory drugs (NSAIDs) suppress the inflammatory response, and may inhibit spine fusion.102
Dimar et al. performed three-level dorsal fusions in 39 rats. Half the animals received indomethacin, 3 mg/kg/day, on 6 of 7 days, and the other animals received saline.103 Treatment was started 1 week preoperatively, and continued for 12 weeks after surgery. In the control rats, 27 of 60 levels achieved solid or moderate fusions, whereas only 4 of 42 levels were similarly fused in the indomethacin group (P < .001). Weaknesses of this study included the following: the experimental model used had not been well characterized, fusion assessment was not rigidly defined, and the indomethacin dose was significantly greater on a milligram-per-kilogram basis than that used in human beings.
Glassman et al. performed a retrospective review of 288 patients who had undergone L4-S1 instrumented, autologous iliac crest graft spine fusions.104 Ketorolac had been administered to 167 of them; the remaining 121 did not receive NSAIDs. Using surgical exploration, hardware failure, and tomograms to determine fusion, they found 4% pseudarthroses in the control group, versus 17% in the ketorolac group (P < .001). The odds ratio indicated that nonunion was approximately five times more likely in those individuals who received ketorolac. There are several problems with this retrospective study: the number of surgeons involved in the cases varied, and the patients received varying numbers of ketorolac doses, beginning at different postoperative times. Their results, however, are supported in a more controlled animal study by Martin et al., who, working in a rabbit model, compared fusion in animals receiving ketorolac or saline.105 They found 35% fusions in the ketorolac-treated animals versus 75% in the controls (P = .037).
Cyclo-oxygenase 2 (COX-2) inhibitors are specific for the isoform of the enzyme targeted by NSAIDs. Long et al. investigated the effect of orally administered celecoxib on spine fusion in the rabbit model.106 They compared rabbits receiving celecoxib, 10 mg/kg daily, with groups receiving either indomethacin, 10 mg/kg, or saline. They found a significant difference between the rate of fusion in controls versus that of the indomethacin group, while animals that received celecoxib fused at an intermediate rate. The study was limited by its small size and the use of a relatively high dose of indomethacin compared with that used in humans.
Osteoporosis
The most common metabolic bone disease in the United States, osteoporosis commonly is assumed to be a negative factor in bone healing. The decreased bone density that is the hallmark of osteoporosis makes stabilization with instrumentation difficult in this population. Additionally, there may be changes in marrow quality and bone turnover rate. Older animals have a decreased capacity for osteoinduction.107 In terms of fusion potential, a decrease in the number of osteogenic stem cells in elderly patients may be more important than absolute bone mass.
Hormones
Hormones affect bone formation both directly and indirectly and probably influence spine fusion as well. These chemical messengers have complex interactions, both positive and negative, with bone-forming and bone-absorbing cells.
Growth hormone, via somatomedins, exerts a stimulatory effect on cartilage and bone formation.108,109 In vivo experimental research has revealed that growth hormone stimulates bone healing by increasing gastrointestinal absorption of calcium, as well as by increasing bone formation and mineralization.110,111 Thyroid hormone, which acts synergistically with growth hormone, is required for somatomedin synthesis by the liver. Furthermore, thyroid hormone has a direct stimulatory effect on cartilage growth and maturation, thereby positively influencing bone healing.
Corticosteroids have been shown both experimentally and clinically to be detrimental to bone healing, increasing bone resorption and decreasing bone formation. They inhibit and promote osteoblastic differentiation and also decrease the synthesis of viable bone matrix.112–114
