Thirty-eight consecutive skeletally mature patients with an unstable burst fracture between T10 and L5 who were admitted to the Department of Orthopaedic Surgery at the University of Texas in Houston were prospectively identified and managed from September 2002 to January 2005. Institutional review board approval and patient informed consent to participate in the study were obtained. Ten patients were lost to follow-up. Unstable burst fractures were defined with use of clinical and radiographic criteria. In order to be considered to have an unstable burst fracture, the patient had to have a type-A3 injury with at least one of five additional clinical or radiographic findings6: (1) neurologic deficit, (2) =50% canal compromise, (3) =50% loss of vertebral body height, (4) =30° of segmental kyphosis, or (5) a combined type-B,A3 or type-C,A3 fracture. Fractures were classified with use of the AO thoracolumbar fracture classification system, which is based on the mechanism of injury7. Type-A injuries are caused by compression, type-B injuries are caused by distraction, and type-C injuries are multidirectional injuries with a rotational component. Type-A injuries are further subclassified into type A1 (wedge compression fracture), type A2 (coronal split fracture), and type A3 (burst fracture). Combined (type-B,A3 or type-C,A3) injuries can occur and usually represent a more severe injury. All patients with a type-A3 fracture were considered for the study, but only patients with unstable type-A3 injuries were included in the study. Exclusion criteria included (1) a stable type-A3 injury, (2) skeletal immaturity, or (3) a senile, osteopenic, or insufficiency fracture. A fracture severity score was also calculated, as described by McCormack et al.5 and Vaccaro et al.8. The percentage of obstruction of the spinal canal was estimated on the preoperative computed tomography scan by dividing the available anteroposterior diameter of the spinal canal at the injured level by the average of the diameter of the canal space at the two uninjured vertebrae cephalad and caudad to the injured level. Segmental kyphosis was determined by measuring the Cobb angle between the superior end plate of the vertebral body cephalad to the injury and the inferior end plate of the vertebral body caudad to the injury. The percentage loss of the vertebral body height was calculated by dividing the height of the fractured vertebra by the average height of the cephalad and caudad vertebrae and then subtracting this number from 100%. No patient refused to participate in the study. Demographic data were collected, and functional outcomes were determined on the basis of a visual analog pain scale, pain medication use, the Short Form-36 (SF-36) score, and the Oswestry Disability Index score9,10. The outcomes data were collected preoperatively; at three, six, twelve, and twenty-four months postoperatively; and then every year thereafter. Preoperative and follow-up neurologic assessments were performed as described by Frankel et al.11. All patients were managed urgently, with an intention to operate within twenty-four hours after admission whenever possible.
Surgical Technique
Initial reduction of the kyphotic deformity was obtained by placing the patient in the prone position on transverse gel rolls across the chest and the pelvis. Schanz pedicle screws with a diameter of 5.2 or 6.2 mm and a thread length of 35 to 40 mm were placed bilaterally in the vertebrae cephalad and caudad to the fractured vertebra through a longitudinal, midline incision measuring approximately 12 cm. The largest-diameter screw that could be accommodated by the pedicle was placed. A wide decompressive laminectomy from pedicle to pedicle was performed in all patients who presented with a neurologic deficit or >50% canal compromise. Direct decompression of the thecal sac initially was performed by tapping the retropulsed fracture fragment into the vertebral body through a posterolateral approach in one patient. In all subsequent patients, the fragments were not exposed or directly manipulated. Some reduction of these fragments may have occurred as a result of ligamentotaxis in patients with an intact posterior longitudinal ligament. However, patients with >50% canal compromise probably had complete disruption of the posterior longitudinal ligament12 and thus did not benefit from ligamentotaxis.
A pedicle finder was used to create a bilateral, transpedicular pathway within the fractured vertebral body under fluoroscopic guidance, creating a pathway parallel to the superior end plate. A working cannula was then placed bilaterally into the posterior third of the fractured vertebral body through the previously created transpedicular pathway. A deflated bone tamp (size 15 or 20-mm, KyphX Xpander; Kyphon, Sunnyvale, California) was placed bilaterally through the working cannulas into the anterior portion of the fractured vertebra.
The segmental kyphotic deformity was then reduced by connecting the pedicle screws to the rods and using the leverage of the screws and wrenches to restore anatomic segmental alignment13. Partial restoration of vertebral body height was then obtained with distraction of the pedicle screws on the rods. To minimize the risk of overdistracting the neural elements, distraction along the screw-and-rod construct was not performed in patients with a combined compression and distraction (type-B,A3) injury. The bone tamps were then simultaneously inflated at 1-mL increments under fluoroscopic guidance, which further restored vertebral body height and created a cavity for the placement of the injectable calcium phosphate cement (Norian SRS or Norian XR; Synthes, West Chester, Pennsylvania) (Fig. 1). Less than 1 mm of posterior bone displacement into the canal is expected with inflation of the bone tamp14. The Schanz screws were cut to the appropriate length prior to the placement of the cement in order to minimize movement of the construct while the cement solidified.
The volume of contrast medium within the bone tamps was used to approximate the amount of cement required to fill the cavitary defect. Calcium phosphate bone cement (Norian XR or Norian SRS mixed with 2.5 mL of powdered barium sulfate [Biotrace; Bryan Corporation, Woburn, Massachusetts] measured in a 5-mL syringe mixed with 1.2 mL of sterile water per every 10 mL of Norian SRS cement) was placed in a 10-mL syringe, which was used to fill the bone-void-filling cannulas. The balloons were deflated, and the tip of the bone-void-filling cannula was placed in the anterior portion of the cavitary defect. Rapid injection of the cement led to cement crystallization at the tip of the bone-void-filling cannula, which impeded subsequent cement injection. A gentle tap with a mallet disrupted the crystallization at the tip of the cannula and allowed the remainder of the cement to be injected into the cavitary defect. However, when the cement was injected too slowly, the cement solidified within the cannula. A cannula containing solidified cement was not usable and was replaced with a fresh cement-containing, bone-void-filling cannula. In our patients, approximately 2 to 3 mL of cement was injected into the right and left cavitary defects over a thirty to forty-five-second time period. Cement injection was stopped when cement extravasated outside the vertebral body or reached the posterior one-third of the vertebral body. The cement was allowed to cure for at least ten minutes before the patient was moved or the instrumentation was adjusted. Probing of the cement was not routinely performed during this time, to minimize disruption of its crystallization15. Cross-links were not routinely used. Decortication of the facets and laminae was performed to facilitate fusion. Local bone graft and iliac crest bone graft were placed to facilitate posterior fusion. The patient was mobilized the next day with a polypropylene thoracolumbosacral orthosis, which was worn for three months. All patients were asked to refrain from tobacco use. All patients were followed clinically and radiographically for a minimum of two years.
Statistical Analysis
Student t tests and Fisher exact tests were used to determine differences between the group with neurologic deficits and the group without. When the measure of interest was continuous, such as age, the SF-36 score, and the Oswestry Disability Index score, the Student t test was used. When the measure was dichotomous, such as sex or return to work, the Fisher exact test was used. The level of significance was set at p < 0.05.
Source of Funding
The funding source did not play a role in the investigation.
Clinicopathologic Characteristics
The study group included eighteen male patients and ten female patients with a mean age of thirty-eight years (range, fifteen to seventy-three years). The mean duration of follow-up was thirty-five months (range, twenty-four to fifty-five months). The cause of the injury was a fall for eleven patients, a motor-vehicle collision for eleven, a motorcycle collision for three, being struck by an automobile or an object for two, and an all-terrain vehicle collision for one. The level of spinal involvement was T12 for five patients, L1 for sixteen, L2 for four, L4 for two, and L5 for one. All patients had an A3 (burst) injury. Seven patients had a combined type-B (distraction) and A3 injury, and two patients had a combined type-C (rotation) and A3 injury.
Twenty-seven patients were managed at the time of the initial admission. One patient underwent surgery fifty-eight days after the injury. The mean time to surgery was four days (range, zero to fifty-eight days). Twenty-two patients (79%) had a decompressive laminectomy.
The mean and median values for the extent of canal compromise, segmental kyphosis, and loss of vertebral body height in the preoperative, postoperative, and latest follow-up periods are given in Table I. All patients had a load-sharing score5 of 7 or more (mean, 8; median, 9). Twenty patients had a thoracolumbar injury severity score of 5 or more, three had a score of 4, and five had a score of 2 (mean, 7; median, 5). The median correction of kyphosis was 14° (range, 1° to 34°). Resorption of the cement was not detected on plain radiographs at the time of the latest follow-up.
The mean operative time was 116 minutes (range, eighty to 165 minutes). The mean estimated blood loss was 316 mL (range, 50 to 1200 mL). The mean hospital stay was 8.7 days (range, two to thirty-five days).
Because the patients could not be randomly assigned to groups according to whether or not they had neurologic deficits, the mean age and sex proportions of these two groups were analyzed at the time of the latest follow-up. No significant differences were found with respect to age (p = 0.2198) or sex (p = 0.6716).
Motor deficits were present in eighteen patients (64%). Motor function was maintained or improved in all patients after surgery to repair the thoracolumbar burst fractures (Table II). Neurologic function improved by at least one Frankel grade in fifteen (83%) of the eighteen patients with motor deficits. Neurologic function improved by at least one Frankel grade in all thirteen patients with incomplete neurologic deficits. Neurologic function improved by at least one Frankel grade in two of the five patients with complete neurologic deficits; both of those patients had injuries at L1. Neurologic function remained at the preoperative level in the remaining three patients with complete neurologic deficits; those three patients had injuries at T12, L1, and L2.
The mean pain level was 2 of 10 on a visual analog scale at the time of the latest follow-up. Three patients (11%) required oral narcotic analgesics for pain control at the time of the latest follow-up. Of the five patients who had used tobacco preoperatively, two continued to use tobacco postoperatively and three had quit smoking.
Data on work status were available for twenty-four patients at the time of the latest follow-up. Fifteen (62.5%) of these twenty-four patients returned to work at a median of ninety days (range, twenty-one to 365 days) after surgery. Seven of these fifteen patients had neurologic injuries (five had improvement from Frankel grade D to E, one had improvement from Frankel grade C to E, and one had improvement from Frankel grade A to E). Of the nine patients who did not return to work, eight had a neurologic injury (three had improvement from Frankel grade D to E, one had improvement from Frankel grade B to D, one had improvement from Frankel grade B to C, and three remained at Frankel grade A). Overall, seven (47%) of the fifteen patients with a neurologic injury returned to work, compared with eight (89%) of the nine patients without a neurologic injury (p = 0.048).
The Oswestry Disability Index scores for patients with and without neurologic deficits were compared. The median Oswestry Disability Index score at the time of the latest follow-up was 9% (range, 0% to 50%) for patients presenting with neurologic deficits and 2% (range, 0% to 60%) for patients presenting without neurologic deficits. A score of 0% to 20% indicates minimal disability; 21% to 40%, moderate disability; and 41% to 60%, severe disability. At the time of the latest follow-up, the mean SF-36 physical function score was 28.3 for patients with neurologic deficits, compared with 49.8 for patients without neurologic deficits (p = 0.0001). The mean SF-36 physical component summary score for patients with neurologic deficits was 37.6, compared with 49.6 for patients without neurologic deficits (p = 0.0058).
Complications
There were no known complications among any of the patients who were lost to follow-up. Seven of the twenty-eight patients who were followed for a minimum of two years had development of treatment-related complications. One patient had development of a wound dehiscence; in the case of this patient, contributing factors included multiple rib fractures and a soft-tissue injury to the paraspinous musculature resulting from a direct blow to the thorax that was received when the patient landed on the road. Direct communication from the wound to the pleural cavity was noted intraoperatively in this patient. One patient had perioperative cardiac ischemia on an electrocardiogram, and another patient had development of a symptomatic pulmonary embolus. One patient had an unintended durotomy, which was repaired; no known adverse sequelae developed. No patient had development of an early or late wound infection.
One patient had development of a pseudarthrosis of the posterolateral fusion, which was identified after she fell at a supermarket and complained of new-onset back pain. A posterolateral pseudarthrosis was identified on removal of the instrumentation and exploration of the posterior fusion mass. Reinstrumentation and bone-grafting were performed, and the fusion subsequently healed. The patient had 5° of segmental kyphosis at the time of the latest follow-up, compared with 0° before the second procedure. Vertebral body height was maintained.
Two patients had breakage of a screw and progression of kyphosis of >10°. However, smaller-diameter (5.2 as compared with 6.2-mm) screws had been used in both of these patients, which may have contributed to the breakage. In addition, one of these patients had radiographic signs of a preexisting Scheuermann kyphosis, which probably predisposed him to the progressive kyphosis and subsequent screw breakage. He had had a preoperative segmental kyphosis of 39° from T11 to L1, which was reduced to 10° postoperatively. The kyphotic deformity then progressed to 24°. At the time of the latest follow-up, the patient had mild pain and did not desire removal of the instrumentation. The other patient was evaluated by another spine surgeon, who removed the instrumentation. This patient had presented with 19° of segmental kyphosis from T12 to L2, which was reduced to 6° after the first spinal procedure. There was no evidence of pseudarthrosis at the time of exploration of the fusion mass. At the time of the latest follow-up, the patient had no pain related to the back and had 30° of segmental kyphosis.
Routine postoperative computed tomography scans were not made. However, one patient with persistent motor weakness (with improvement from Frankel grade A preoperatively to Frankel grade D postoperatively) desired a computed tomography scan two years after the index procedure to determine if there was any residual canal compromise. However, canal remodeling16 had occurred and no substantial residual canal stenosis was detected (Figs. 2-A through 2-D).
The current study demonstrates that unstable thoracolumbar burst fractures in patients with or without neurologic deficits can be decompressed and circumferentially stabilized through a single posterior incision. Reconstruction of the anterior column was performed with use of a transpedicular, balloon-assisted reduction of the fractured vertebral body followed by the injection of calcium phosphate cement. The cement provided internal structural support of the fractured vertebral body, which allowed healing of the surrounding fracture fragments. Posterior reconstruction was achieved with use of pedicle screws placed in the vertebrae cephalad and caudad to the fractured vertebra, whereas decompression of the neural elements was accomplished with the combination of a near-anatomic reduction of the segmental kyphotic deformity and a wide decompressive laminectomy. Maintenance of or improvement in neurologic function was seen in all patients, and the rates of screw breakage and progressive deformity were low compared with those cited in previous reports4,5.
The use of short-segment posterior spinal instrumentation without restoration of the anterior column for the treatment of unstable thoracolumbar burst fractures has been associated with a high rate of early instrumentation failure and progressive deformity4,5. McLain et al.4 reported progressive kyphosis or screw breakage in ten (77%) of thirteen patients who had been managed with short-segment posterior instrumentation alone. McCormack et al.5 also observed a high prevalence of screw breakage (53%; ten of nineteen patients) after short-segment posterior instrumentation without anterior column support. On the basis of their findings, both McLain et al. and McCormack et al. recommended anterior column reconstruction for most patients with unstable thoracolumbar burst fractures that are treated with short-segment posterior instrumentation. In the current series, the prevalence of screw breakage was 7% (two of twenty-eight patients), which suggests that reconstruction of the anterior column with this transpedicular approach may decrease the prevalence of screw breakage in comparison with short-segment posterior spinal instrumentation without anterior column reconstruction.
Mermelstein et al.17 provided biomechanical evidence that short-segment posterior instrumentation combined with a transpedicular reconstruction of the anterior column with injectable calcium phosphate cement decreases pedicle screw bending moments in comparison with those seen in association with posterior instrumentation alone. Specifically, screw bending moments were decreased by 59% in flexion and 38% in extension. Moreover, the overall stiffness of the construct was increased by 40%. This decrease in screw bending moments and increased stiffness of the construct after anterior column reconstruction may explain the lower prevalence of screw breakage and progressive kyphosis demonstrated in the current series.
Cement extravasation is a well-known complication of vertebral body augmentation with cement. However, none of the patients in the current series had development of clinically important spinal cord compression related to cement extravasation, despite disruption of the posterior wall of the vertebral body in all of our patients and the presumed discontinuity of the posterior longitudinal ligament in many of them. The lack of clinically important cement extravasation is supported by the findings of the in vitro studies performed by Verlaan et al., who showed that discontinuity of the posterior longitudinal ligament may not be a risk factor for cement extravasation14.
There have been several reports of death after the injection of polymethylmethacrylate and calcium phosphate cement into fractured vertebral bodies18,19. Nussbaum et al. reported on twelve patients who died after treatment with vertebral body augmentation with use of polymethylmethacrylate cement18. However, many of those patients were elderly and were managed with cement injection into three or more vertebral body fractures. The cause of death in those cases remains unclear but may have been associated with the release of cement monomer or emboli of fat or cement20. In general, the patients in the current series were relatively young and healthy. Additionally, cement use was limited to one vertebral level. There were no deaths in the current series or among the >200 patients managed with calcium phosphate for the reconstruction of fractures of the distal part of the radius, the calcaneus, the tibia, the proximal part of the femur, and the vertebrae in several large studies21-24.
Investigators from several centers have reported their short-term results associated with the use of short-segment posterior instrumentation combined with transpedicular cement reconstruction of the anterior column through a posterior approach for the treatment of stable and unstable thoracolumbar burst fractures25,26. In the study by Cho et al., the outcomes for twenty patients who had been managed with short-segment posterior instrumentation combined with polymethylmethacrylate vertebroplasty were compared with those for fifty patients who had been managed with short-segment posterior instrumentation alone25. The prevalence of instrumentation failure was 0% (zero of twenty) in the former group, compared with 22% (eleven of fifty) in the latter group. Anterior vertebral body height and correction of the kyphotic deformity also were maintained in patients who had been managed with the circumferential reconstruction, which was not the case in those who had undergone short-segment posterior instrumentation alone.
In the current study, calcium phosphate cement rather than polymethylmethacrylate was used to reconstruct the anterior column. Both polymethylmethacrylate and calcium phosphate cement can provide immediate structural support to the vertebral body. However, because calcium phosphate cement is a bioactive cement (i.e., it is composed of two of the primary elements contained in bone), it can remodel into bone. Indeed, the biocompatibility and osteoconductive properties of calcium phosphate cement have made it useful for the reconstruction of osseous defects of the distal part of the radius, the calcaneus, the tibia, the proximal part of the femur, and the vertebra21-24,26,27. Polymethylmethacrylate, on the other hand, cannot remodel into bone and may actually prevent healing between fracture fragments26,28. Another benefit associated with the use of calcium phosphate cement is that calcium phosphate from the reconstructed vertebral body, as well as cancellous bone, can be used as local graft material if a corpectomy is deemed necessary because of the failure of the reconstruction or the persistence of neurologic deficits combined with residual canal stenosis24,29. Biomechanically, calcium phosphate cement-reinforced vertebral bodies have been shown to demonstrate nearly identical deformation properties under cyclical loading and failure loads under compression as do vertebrae reinforced with polymethylmethacrylate30. Both calcium phosphate cement and polymethylmethacrylate have a tendency to fracture when subjected to shear forces31.
Similar to what we have described in the present study, Verlaan et al. reported their preliminary results, after a mean duration of follow-up of seventeen months, of short-segment pedicle screw fixation combined with anterior column reconstruction with calcium phosphate cement after balloon-assisted reduction in twenty patients with stable thoracolumbar burst fractures26,27. Unlike the current study, however, that study excluded patients who had a neurologic injury or a disrupted posterior longitudinal ligament. Korovessis et al. also reported on a group of patients with thoracolumbar type-A3 injuries who were managed with long and intermediate-segment instrumentation combined with anterior column reconstruction with calcium phosphate cement32. The severity of the injuries was unclear as the authors did not report clinical or radiographic inclusion criteria or a fracture severity score. The patients in the current series probably had more severe injuries in that nine (32%) of the twenty-eight patients had type-B or type-C injuries combined with type-A3 injuries and eighteen patients (64%) had neurologic deficits. None of the patients in the study by Korovessis et al. had a combined injury, and only six (26%) of twenty-three patients had neurologic deficits.
The current study had several limitations. The structural benefits of anterior column reconstruction with calcium phosphate cement could not be completely determined because the study did not directly compare the outcomes for patients managed with the current short-segment posterior instrumentation and anterior column reconstruction with calcium phosphate cement with those for patients managed with short-segment instrumentation alone. In addition, the apparent decrease in morbidity, blood loss, operative time, length of hospital stay, and costs associated with a transpedicular, anterior column reconstruction could not be directly compared with those associated with procedures requiring a corpectomy through an anterior approach. Moreover, the neurologic recovery seen in patients managed with the current technique involving a decompressive laminectomy and circumferential stabilization cannot be directly compared with that in patients managed with an anterior decompression with a corpectomy. Prospective, randomized studies may be required to adequately investigate these issues.
In summary, the current study has shown that unstable thoracolumbar burst fractures with or without associated neurologic deficits can be decompressed and circumferentially stabilized with use of short-segment posterior instrumentation and transpedicular, balloon-assisted vertebral body reduction and anterior column reconstruction with injectable calcium phosphate cement performed through a single posterior incision. This approach reconstructs and stabilizes the anterior column, restores vertebral body height, indirectly and directly decompresses the thecal sac, reduces the kyphotic deformity, and stabilizes the posterior column through a posterior approach. Moreover, this procedure eliminates the need for an anterior approach in most patients with unstable thoracolumbar burst fractures, which decreases the inherent morbidity associated with anterior thoracic and abdominal approaches and also decreases the risk of injury to visceral and vascular structures. Excellent reduction of the fracture combined with anterior and posterior column reconstruction also decreased the prevalence of instrumentation failure and the loss of correction associated with short-segment posterior instrumentation without anterior column reconstruction. 