From April 1981 to March 1989, 163 consecutive patients who had neurological deficits because of a burst fracture of the thoracolumbar spine were managed operatively at Hokkaido University Hospital and Hokkaido Orthopaedic Memorial Hospital in Sapporo, Japan. The indication for the operation was a persistent neurological deficit from encroachment on the spinal canal by the retropulsed osseous fragments. In most instances, the operation was performed when the neurological status of the patient had plateaued with only limited recovery.
Thirteen patients were excluded from this investigation. Two patients died. One, a seventeen-year-old boy who had had uncontrolled diabetes mellitus and a severe urinary-tract infection preoperatively, died of pneumonia fourteen days after the operation. The other patient, a fifty-seven-year-old man, died because of a rupture of an esophageal varix two years and seven months postoperatively. The other eleven patients were lost to follow-up. All of these patients had been followed for two to three years postoperatively, had returned to their previous jobs, and were noted to have a fusion of the injured spinal segments at the most recent follow-up examination. Some degree of neurological recovery had been documented in all thirteen patients, and none of the thirteen had a complication during the follow-up period. We reviewed the results for 150 patients who had been followed for a minimum of five years.
There were 109 male and forty-one female patients. They had a mean age of forty-one years (range, thirteen to seventy-two years). The male patients had an average height of 168 centimeters (range, 150 to 178 centimeters), and the female patients had an average height of 155 centimeters (range, 145 to 164 centimeters). The body weight of the patients ranged from fifty-three to eighty-six kilograms; the male patients weighed an average of sixty-seven kilograms and the female patients, an average of fifty-six kilograms.
Most of the patients were transferred to our institutions from other hospitals and, therefore, the injuries were not acute. The interval from the injury to the operation was less than forty-eight hours for seven patients, two to less than fourteen days for thirty-eight patients, two weeks to less than one month for forty-five patients, one to less than six months for forty-one patients, six to twelve months for eight patients, and more than one year for eleven patients. The mechanism of injury included a fall from a height (ninety-four patients), a motor-vehicle accident (thirty-seven patients), a fall from a height during a suicide attempt (nine patients), and a direct impact from a falling heavy object (ten patients). Associated trauma included facial or cranial injury (five patients), thoracic injury (nine patients), urogenital injury (three patients), pelvic fracture (four patients), injury of the upper extremities (twelve patients), and injury of the lower extremities (twenty-nine patients).
All patients had anteroposterior and lateral radiographs of the spine, preoperative myelography with myelographic tomography or myelographic computed tomography, and preoperative and postoperative computed tomography. Myelographic tomography and myelographic computed tomography were used to evaluate the relationship between the epiconus, conus medullaris, or cauda equina and the retropulsed osseous fragments. These radiographic examinations were performed either at the referring hospitals or at our hospitals.
The 150 patients had 158 burst fractures; 143 patients had a fracture at a single level, six patients had two contiguous or non-contiguous vertebral fractures, and one patient had three non-contiguous fractures (Table I). According to the classification system of Denis10, thirty-eight (24 per cent) of the fractures were type A (a fracture of both of the end plates); ninety-two (58 per cent), type B (a fracture of the superior end plate); ten (6 per cent), type C (a fracture of the inferior end plate); eleven (7 per cent), type D (a burst rotation fracture); and seven (4 per cent), type E (a burst lateral flexion fracture). One hundred and twenty-three (78 per cent) of the 158 burst fractures were associated with a fracture of the lamina at the same level. There was no evidence of nerve tissue interposed or trapped within the laminar fractures on the preoperative myelographic or plain computed tomography scans.
The original area of the spinal canal was calculated by averaging the areas of the spinal canal cephalad and caudad to the injured vertebra. The percentage of obstruction of the spinal canal (the ratio of the maximum area of the retropulsed osseous fragment or fragments to the area of the original spinal canal) was estimated on the preoperative computed-tomography scan according to our previously reported method21 (Table II). The nerve-root lesions in the five patients who had a burst fracture of the fourth lumbar vertebra were associated with a smaller percentage of obstruction of the spinal canal. These nerve-root lesions were attributed to retropulsed bone that had narrowed the lateral recesses and resulted in compression of the nerve roots. Fractures of the sixth thoracic and fifth lumbar vertebrae were not associated with neurological deficits. Burst fractures of the fourth or fifth lumbar vertebra associated with neurological deficits have not been treated with anterior decompression and instrumentation at our institutions since June 1985.
The neurological deficits were divided into five groups according to the level of the nerve injury: the spinal cord cephalad to the epiconus (spastic) (three patients), the epiconus (flaccid) (thirty-two patients), the conus medullaris only (thirteen patients), both the conus medullaris and the cauda equina (thirty-three patients), and the cauda equina or an isolated nerve-root lesion (sixty-nine patients) (Table I). The pure conus medullaris syndrome was defined as a disturbance in the function of the anal and bladder sphincters with sensory loss in the perineal area only (no loss of motor function or sensation in the lower extremities). Neurological assessment was performed with use of the new grading scale of Frankel et al., as modified by Bradford and McBride6 (Table III). We also used the lower-extremity motor-index scale17 for the evaluation of motor function, which is performed by manual, bilateral testing of five muscles (the iliopsoas, quadriceps femoris, tibialis anterior, extensor hallucis longus, and gastrocnemius). A normal score is 50 points.
The neurological deficits occurred immediately after the injury in 136 patients, and in thirty-four of them the deficits improved by at least one grade between the time of the injury to the time of the operation. However, all 136 patients demonstrated some neurological deficits at the time of admission to our hospitals (Table IV). In fourteen patients, the onset of the neurological deficits was delayed from one to six months after the injury. The delayed onset of neurological loss from the unstable burst fractures was attributed to increased kyphosis or late segmental instability, which caused neurological compression by retropulsed osseous fragments in the spinal canal.
The preoperative function of the bladder was urodynamically evaluated by means of cystometry with estimation of residual urine. The patients were also asked about the sensation of a full bladder, episodes of incontinence, and the need for self-catheterization. Thirty-three patients (22 per cent) had paralysis of the bladder and forty-five (30 per cent) had dysfunction of the bladder on admission to our hospitals.
Seven patients had been managed with a posterior procedure before admission to our hospitals. Posterior instrumentation for reduction and decompression by ligamentotaxis had been performed in three patients; laminectomy and posterolateral arthrodesis with instrumentation, in two; and posterolateral decompression and arthrodesis with instrumentation, in two. All of these patients still demonstrated neurological deficits (grade C in one, D1 in two, D2 in three, and D3 in one, according to the modified system of Frankel et al.6) at the time of the anterior decompression at our hospitals. The posterior instrumentation was removed before the anterior procedure was performed.
Operative Procedure
An extrapleural and retroperitoneal approach with removal of the left tenth or eleventh rib22 was used to expose the thoracolumbar junction (the eleventh and twelfth thoracic and first lumbar vertebrae). The twelfth rib was not removed from most patients because it is too small and weak for use as a strut graft. Three vertebrae (the fractured vertebra as well as one cephalad and one caudad to it) were exposed. The second, third, and fourth lumbar vertebrae were exposed by the retroperitoneal approach with removal of the eleventh rib. The intervertebral discs cephalad and caudad to the injured level were almost completely excised, leaving the anterior longitudinal ligament intact (Fig. 1-A). The vertebral resection for anterior decompression of the spinal canal was done, leaving the anterior and contralateral cortices of the fractured vertebral body intact (Fig. 1-B). The displaced fragments of the vertebral body in the spinal canal were removed from the anterior aspect of the posterior longitudinal ligament with use of long sharp curets. The posterior longitudinal ligament was often found to be disrupted or attenuated as a result of the original injury, but it was not removed in order to avoid bleeding from the epidural venous plexus. The anterior decompression was not judged to be complete until the base of the contralateral pedicle was well visualized.
After complete decompression, the vertebral plates of the Kaneda device were attached to the lateral aspect of the vertebral bodies cephalad and caudad to the vertebrectomy with the use of a punch and screw fixation (Fig. 1-C). In the fifty-four patients who had the operation before June 1984, the Mark-I model of the Kaneda device (without rod couplers) (Mizuho Ikakogyo, Tokyo, Japan) was used. The Mark-II model of the Kaneda device (with rod couplers) was used in the ninety-six patients who were managed since June 1984. The relationship of the direction of the screws and the vertebral plate was triangular, and the screws had to penetrate the contralateral cortex (Fig. 1-D). Kyphosis was corrected by use of a spreader device applied to the heads of the two anterior screws (Fig. 1-E). Correction of the kyphosis was judged by checking the arrangement of the end plates of the vertebrae cephalad and caudad to the vertebrectomy. In patients who had late presentation of kyphosis, the anterior longitudinal ligament was often scarred and needed to be sectioned at the level of the disc.
After the kyphosis was corrected, the length of the gap was measured from vertebral end plate to vertebral end plate and a tricortical iliac-crest graft was obtained. The defect of the iliac crest was reconstructed with a ceramic substitute (bioactive apatite-wollastonite glass-ceramic; Nippon Electric Glass, Ootsu, Japan)2. The tricortical iliac-crest graft and two or three struts of rib were placed into the defect created by the distracted vertebral resection. In the first 109 patients, the bone grafts were placed with the tricortical portion adjacent to the instrumentation. In the forty-one patients who had the operation in 1987 or later, the tricortical portion of the iliac-crest graft was inserted opposite the Kaneda plate, near the contralateral pedicles (Fig. 1-F). Bone chips from the resected vertebral body were packed into the gap between the iliac-crest graft and the anterior vertebral wall (Fig. 1-F).
After bone-grafting, threaded rods and two sets of rod couplers were applied (Fig. 1-G). When there was a potential for compression of the aorta by an implant located at the eleventh thoracic level (three patients), the implants were covered with a Teflon (polytetrafluoroethylene) sheet. The wound was then closed and drained in the standard fashion. A chest tube was not necessary for most of the patients who had been managed with the extrapleural and retroperitoneal approach. The pleura was checked for air leaks by pouring saline solution into the extrapleural retroperitoneal space just before wound closure. A chest tube was used in nine (10 per cent) of the ninety-three patients who had a burst fracture of the twelfth thoracic or first lumbar vertebra and in whom the pleura had been injured. No chest tube was used in the patients who had a burst fracture of the second, third, or fourth lumbar vertebra. No postoperative pneumothorax occurred in any of the patients who did not have a chest tube.
Postoperative Care
At four to seven days after the operation, the patients were encouraged to walk with a polypropylene thoracolumbosacral orthosis, which was worn for twenty to twenty-four weeks. Eleven patients who had a delayed union or a pseudarthrosis wore the brace for more than six months. The patients were taught to continue active isometric exercises of the trunk muscles while wearing the thoracolumbosacral orthosis. After removal of the brace, the patients who were not manual laborers were allowed to perform the normal activities of daily living without any special restrictions. The manual laborers returned to their jobs six to eight months postoperatively.
Follow-up
The mean duration of follow-up was eight years (range, five years to twelve years and eleven months). The latest clinical and radiographic evaluations were conducted by two of us (H. T. and T. H.) who had not participated in the original operation.
Statistical Analysis
Statistical analyses were performed with use of the Welch t test or the Student t test. A p value of less than 0.05 was considered to be significant.
Neurological Recovery
None of the patients had neurological deterioration after the anterior decompression and stabilization. The patients stayed at our hospitals and at our affiliated rehabilitation hospitals for an average of twenty-four days (range, eleven to fifty-six days) after the operation. The patients who had residual motor dysfunction that was severe (grade A in two patients and grade C in three patients postoperatively) needed a longer period of hospitalization for rehabilitation. The neurological function of 142 (95 per cent) of the 150 patients had improved at least one grade (Table V). Eight patients (5 per cent) had not recovered any neurological function at the time of the latest follow-up evaluation. Preoperatively and postoperatively, the neurological function of two of these patients, who had a burst fracture of the twelfth thoracic vertebra, was grade A; that of one patient, who had fracture of the first lumbar vertebra, was grade C; and that of five patients, who had a fracture of the first lumbar vertebra, was grade D1 (Table V). Two of these eight patients had sustained the spinal cord lesion cephalad to the epiconus; one, in the epiconus; and five, in the conus medullaris. Preoperatively, all eight patients had had complete loss of function of the bladder and the bowel with anesthesia of the perineal area.
All seven patients who had been managed with a posterior procedure before the anterior procedure was performed showed some neurological recovery at the time of the follow-up evaluation. One patient had improvement from grade C to grade D2; one, from grade D1 to D2; one, from grade D1 to E; three, from grade D2 to E; and one, from grade D3 to E.
The difference between the preoperative and postoperative scores on the motor-index scale was larger for the group that had a lesion of the spinal cord than for the group that had a lesion of the cauda equina (p < 0.01), but the final score was higher for the group that had a lesion of the cauda equina (Table VI). With the numbers available, we could not detect a significant association between the interval from the injury to the operation and the amount of postoperative neurological recovery.
Thirteen patients who had no recovery of bladder function had had no sensation to pinprick in the perineal area preoperatively (Table VII). The patients who had had decreased sensation to pain in the perineal area preoperatively had incomplete or complete recovery of bladder function.
Functional Results
At the most recent follow-up examination, 101 patients (67 per cent) had returned to their previous occupation, thirteen (9 per cent) had changed to less strenuous work, and thirty-six (24 per cent) were unemployed. The unemployed patients included eleven (7 per cent) who had reached retirement age after returning to their previous occupation, seven (5 per cent) who had retired before the operation, five (3 per cent) who had paraplegia or paraparesis (incontinence), ten (7 per cent) who were managed for schizophrenia before and after the operation, and three (2 per cent) who had alcoholism. Twenty patients (13 per cent) had not been working preoperatively. Therefore, 125 (96 per cent) of the 130 patients who had been employed before the injury returned to work after the operation; 112 (86 per cent) of those 130 patients returned to their previous jobs at full capacity. Patients who had performed light labor returned to work with the aid of a brace a mean of three months (range, one to five months) after the operation. Patients who had performed heavy labor returned to work with or without the aid of a brace a mean of seven months (range, six to nine months) after the operation.
Pain in the back was assessed at the most recent follow-up examination with use of the pain scale of Denis11. One hundred and fifteen patients (77 per cent) were given a rating of P1 (no pain); twenty-one (14 per cent), a rating of P2 (occasional slight pain with no need for medication); ten (7 per cent), a rating of P3 (moderate pain with a need for occasional medication but no interruption of work or major change in activities of daily living); four (3 per cent), a rating of P4 (moderate-to-severe pain with a need for frequent medication and occasional absence from work or a major change in activities of daily living); and none were given a rating of P5 (constant or severe incapacitating pain and a chronic need for medication). The four patients who had a rating of P4 were receiving workers' compensation and demonstrated a complete neurological recovery.
Radiographic Results
The percentage of the spinal canal that was obstructed was evaluated before and after the operation (Table II).
The fusion at the site of the three-level arthrodesis was evaluated on lateral flexion-extension radiographs and tomograms. A pseudarthrosis developed in six (11 per cent) of the fifty-four patients who had fixation with the Kaneda Mark-I device (without rod couplers); two of these patients had had a fracture of the second lumbar vertebra, two had had a fracture of the third lumbar vertebra, and two had had a fracture of the fourth lumbar vertebra. A pseudarthrosis developed in four (4 per cent) of the ninety-six patients who had fixation with the Kaneda Mark-II device (with rod couplers); one of these patients had had a fracture of the first lumbar vertebra, two had had a fracture of the fourth lumbar vertebra, and one had had a two-level fracture at the second and third lumbar vertebrae. Over-all, fusion was achieved in 140 (93 per cent) of the patients. A pseudarthrosis developed in four of five patients who had had a burst fracture of the fourth lumbar vertebra. The mean height, body weight, and duration of hospitalization after the operation of the patients who had a pseudarthrosis were 165 centimeters, sixty-eight kilograms, and twenty-two days, respectively. These values were not significantly different from those for the patients who did not have a pseudarthrosis (p = 0.70, Student t test). All ten patients who had a pseudarthrosis had a successful repair with a posterolateral arthrodesis and posterior instrumentation, and all had a solid fusion at the most recent follow-up examination.
Kyphosis was measured by the 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. This measurement was not performed for eighteen patients who had a burst fracture of the third or fourth lumbar vertebra because of the lack of a preoperative kyphotic deformity. The average kyphosis for the remaining 132 patients was 19 degrees preoperatively, 7 degrees at the time of discharge from the hospital, and 8 degrees at the most recent follow-up examination.
Complications
Seven of the ten patients who had a pseudarthrosis reported back pain and three did not; none of them showed neurological deterioration. Complications other than pseudarthrosis included failure of the device in nine patients who had a pseudarthrosis (breakage of a screw or screws in eight patients and breakage of the screws and the rod in one) intraoperative laceration of the inferior vena cava in one patient who had an old burst fracture of the second lumbar vertebra with severe kyphosis, a deep wound infection in one, a superficial wound infection in three, postoperative urinary-tract infection in three, postoperative atelectasis in ten, transient dysesthesia in the distribution of the genitofemoral nerve in five, and sympathectomy effect on the ipsilateral lower extremity in fifteen patients who had had exposure of the fourth or fifth lumbar vertebra, or both. All four postoperative infections were treated successfully with intravenous administration of antibiotics (cephalosporin) for two to four weeks. The sympathectomy effects, which subsided spontaneously without any treatment within six to twelve months postoperatively, were hotness (twelve patients) and dryness (three patients) of the lower extremity, including the foot, on the side of the operation. None of the patients had an iatrogenic neurological injury, retrograde ejaculations, late vascular injury, or loosening of the hardware other than the nine patients who had pseudarthrosis and failure of the device. None of the implants needed to be removed.
The indication for decompression of the spinal canal in patients who have a thoracolumbar burst fracture is a neurological deficit with radiographic evidence of obstruction of the spinal canal (Fig. 2-A through 2-F). The compressive tissues after a thoracolumbar burst fracture are invariably located in the anterior portion of the spinal canal. Decompression of the spinal canal can be direct6,23,24,26 or indirect9,14,15. Indirect decompression is accomplished with use of posterior instrumentation, with reliance on distraction, correction of any kyphosis, and ligamentotaxis to clear the canal of the displaced anterior bone and soft tissues. Ligamentotaxis is not always successful, and it has been reported that, at best, this method produces incomplete decompression of the spinal canal8. With use of ligamentotaxis, Bradford and McBride found that the mean percentage of stenosis of the spinal canal, as determined with computed tomography scanning, was 26 per cent (range, 0 to 50 per cent) postoperatively6. If ligamentotaxis is to be successful, the technique should be performed within forty-eight to ninety-six hours after the injury8. Crutcher et al.9 reported that type-B fractures, according to the classification system of Denis10, were often resistant to indirect reduction by ligamentotaxis, resulting in an incomplete decompression of the spinal canal.
Many investigators have found that better results can be obtained with direct removal of the retropulsed fragments and soft-tissue debris from the spinal canal6,12,23,24. One of the two methods of direct decompression is a transpedicular decompression through pedicular resection. It is difficult to decompress the far side of the canal with unilateral posterior transpedicular decompression, and there is a risk of iatrogenic neurological injury or incomplete decompression3,20,28. Bilateral transpedicular decompression provides increased exposure of the canal, but it also increases vertebral instability3. A few of the reported series in which transpedicular decompression and cancellous bone-grafting to the fractured body by means of the transpedicular approach were used have shown good results with slight loss of correction of the kyphosis, low rates of implant failure, and no increase in neurological complications13. However, it is difficult to conceive how a cancellous non-structural graft through the pedicle provides any additional support to the anterior column, which might collapse with early loading before incorporation of the graft. With a longer duration of follow-up and a larger series, it is possible that the expected failure of instrumentation and loss of correction of the kyphosis due to an unstable anterior column would become evident. Newer instrumentation systems that include pedicle screws have allowed a decrease in the number of fused segments and greater correction of kyphosis. Despite these advantages, failure to support the anterior spinal column after posterior correction and instrumentation has led to the failure of the posterior spinal instrumentation in many patients13,28,30,32.
The second method of direct decompression is an anterior approach with partial or complete resection of the vertebral body23,24,26. This anterior decompression has to be combined either with anterior stabilization and instrumentation (a one-stage anterior procedure)12,23,24 or with posterior instrumentation and arthrodesis (a combined anterior and posterior procedure)26.
We reviewed the clinical results that have been reported for short-segment pedicle-screw fixation (with the screws placed one level cephalad and one level caudad to the fracture) and for other anterior decompression procedures in the treatment of thoracolumbar burst fractures (Table VIII). The loss of correction of kyphosis and the rate of failure of instrumentation were greater in the series that had fixation with pedicular screws5,7,13,14,28-30,32,33 than in our series. The mean loss of correction of kyphosis ranged from 3 to 12 degrees in the reported series that we reviewed, whereas the mean loss of correction in our series was only 1 degree. The rate of failure of posterior instrumentation ranged from 9 to 54 per cent in the other series, whereas the rate was 6 per cent in the present report. Failure of pedicle screw fixation included breakage, bending, or loosening of the screw because of inadequate support of the anterior column13,28,29,32. Despite the high rates of failure of the instrumentation in the series that had pedicle screw fixation, the rate of non-union was surprisingly low (0 per cent) except for that reported by Ebelke et al. (15 per cent)13.
The rate of failure of the Kaneda device depends on proper positioning of the anterior strut graft consisting of tricortical iliac-crest bone and two or three sections of rib. Biomechanical studies of the Kaneda device have shown adequate stability for most loading situations1,18,31,34. With respect to height, body weight, and duration of hospitalization after the operation, we detected no significant difference (with the numbers available) between the group that had a solid fusion and the group that had a pseudarthrosis. All pseudarthroses occurred in patients who had had poor placement of the anterior strut graft. The graft had not provided stable and symmetrical anterior load-sharing in the spinal column because of placement of the cortical portion of the iliac-crest strut graft on the same side as the Kaneda device. Consequently, we think that one of the most important factors that cause pseudarthrosis is the use of poor technique for the placement of the anterior strut bone graft (Figs. 3-A through 3-D). The Kaneda device relies directly on load transmission through a stout, strong tricortical iliac-crest graft for secure fixation. Since 1987, we have used our present technique of placement of the anterior strut graft with the tricortical portion beyond the contralateral pedicle (Fig. 1-F) in fifty-one patients; pseudarthrosis developed in only one patient. Failure of the Kaneda device may be more of a problem for heavier, non-Japanese populations if anterior strut-grafting does not provide adequate supplemental support. We also noted that a larger percentage of pseudarthroses occurred with burst fractures of the fourth lumbar vertebra (four of five patients). The increased rate of pseudarthrosis at this level is most likely related to the difficulty in providing a sufficient amount of bone graft as well as to the large strut graft that is needed to bridge the third, fourth, and fifth lumbar vertebrae. The location of the fourth lumbar vertebral body and the shape of the fifth lumbar vertebral body make instrumentation technically more demanding. Because of the high rate of non-union of burst fractures of the fourth lumbar vertebra, we have not treated this fracture with anterior fixation since 1985.
In our series, the most important predictive factor with regard to neurological recovery was an injury involving the epiconus or conus medullaris. For patients who had such an injury, a complete lack of bladder function with loss of sensation to pinprick in the perineal area preoperatively was indicative of a poor prognosis for recovery of bladder function. Complete paralysis of the bladder associated with incomplete sensation of pain in the perineal area was associated with a better prognosis for recovery of bladder function after decompression.
It is important to note that, although the operation was performed more than forty-eight hours after the injury in 143 (95 per cent) of our patients, the neurological recovery was acceptable. Thus, we believe that when anterior decompression of the spinal canal is performed in patients who have a thoracolumbar burst fracture with neurological deficits, recovery of neurological function is not always dependent on the timing of the operative intervention.
The goals of an operation for a thoracolumbar burst fracture with associated neurological deficits should be decompression of the spinal canal, restoration of spinal alignment, and successful arthrodesis of the injured spinal segments. These goals were accomplished in more than 90 per cent of our patients, who had low perioperative and postoperative rates of complications. We attribute our excellent results to the use of the anterior approach, the design of the Kaneda device, and our close attention to operative details.
NOTE: The authors wish to acknowledge Paul C. McAfee, M.D., of the Spine and Scoliosis Center in Baltimore, Maryland, for his help in reviewing and editing this manuscript.