Anterior Surgery for Thoracolumbar Burst Fractures: Rationale and Technique
Thoracolumbar burst fractures occur after the thoracolumbar spine fails in axial compression. The fracture results in comminution of the vertebral body, disruption of one or both of the vertebral end plates, and retropulsion of the posterior vertebral body wall (the middle column), into the spinal canal. This latter feature is the pathognomonic finding in a burst fracture, and it has the potential to cause spinal cord or cauda equina injury (Figure 1a, 1b, 1c, and 1d).
Fig. 1a: Lateral radiograph of thoracolumbar burst fracture
Fig. 1b: Anteroposterior radiograph of thoracolumbar burst fracture
Fig. 1c: Axial CT scan through burst fracture, demonstrating retropulsion of the middle column fragment with splaying of the pedicles and fracture of the lamina.
Fig. 1d: Sagittal reconstruction of burst fracture demonstrating the retropulsed middle column fragment.
There are several classifications for thoracolumbar injuries. The comprehensive classification system (Magerl 1994), further modified by the AO, classifies burst fractures within the severe end of the spectrum of Type A injuries. Type A injuries are axial compression injuries and differ in aetiology from Type B (distraction injuries including flexion-distraction injuries such as Chance fractures), and Type C injuries (unstable three column injuries with rotation in the AP projection). The classic description of burst fractures was provided by Denis (1983), who gave the three-column description of the thoracolumbar spine, and then sub-classified the types of burst fractures (Denis 1984). The Denis classification takes into account the location of fracture comminution in the vertebral body, and any deformity. In Type A there is comminution of the whole body, in Type B the superior body only and Type C the inferior body. Type D injuries have whole body comminution and a lateral translational deformity in the AP projection, suggesting greater mechanical instability. Type E injuries have asymmetrical vertebral body comminution in the coronal plane resulting in a coronal plane angulation.
McCormack et al (1994) described a further classification system and quantify the degree of vertebral body comminution and potential mechanical instability. Although this classification system was specifically designed to guide the need for anterior column reconstruction after posterior short segment pedicle screw stabilization, it is a useful guide to the magnitude of comminution and potential mechanical insufficiency.
Although these classification systems give a guide to the many varieties of thoracolumbar injuries and burst fractures, many combinations exist, mandating the need for careful assessment to define the mechanical failure that has occurred in the injury complex. The most unstable variant of the burst fracture is where significant kyphosis occurs and is associated with posterior ligamentous injury or horizontal posterior element fracture. In this injury type (clinically suspected by marked posterior tenderness, bruising or any palpable gap at the interspinous level) not only has the anterior and middle column failed under axial compression, but the posterior column has failed in tension.
Burst fractures characteristically do not involve the neural arch unless splaying of the pedicles results in a vertical split of the lamina. Both the pedicle separation and vertical split of the lamina may be seen on an AP x-ray and are pathognomonic of a burst fracture, and are confirmed with axial imaging with either CT or MRI scan (Figures 1b, 1c, and 1d).
The biomechanical evaluation of fracture combinations has shown progressive mechanical instability when there is cumulative involvement of the anterior, middle and posterior columns. The addition of a posterior column injury results in maximal mechanical instability in a burst fracture injury (James 1994). The biomechanical failure patterns associated with these injuries should guide treatment planning from the many options available.
Burst fractures are most frequent at the thoracolumbar junction (T10 to L2). They are rare in the proximal thoracic spine because of the stabilization this region receives from the rib cage. They are occasionally seen in the low lumbar spine but the different anatomy of this region means treatment options may differ when compared to the thoracolumbar junction (Robertson).
The thoracolumbar junction represents a transitional zone with a straight spinal segment between the thoracic kyphosis and the lumbar lordosis, and a transition of facet orientation from the coronal facet joints of the thoracic spine to the sagittal facet joints of the proximal lumbar spine. In this region the spine is most exposed to axial compression without protection from the rib cage proximally and the pelvis distally.
The incidence of thoracolumbar burst fractures is bimodal. High-energy injuries from falls and motor vehicle accidents occur in younger active patients. The second group sustain injuries from falls at home, from ladders, doing gardening activities, or performing handyman activities. This second group of patients may have significant medical co-morbidities including osteopenia, and these may influence treatment options.
Retropulsion of middle column bone fragments represents the major risk to the neural structures in thoracolumbar burst fractures. As the cord ends at L1, and the cauda equina roots cascade off the cord over many segments closely approximated to the distal end of the cord, a variety of neurological structures may be damaged - with a variety of neurological abnormalities. Whilst cord injuries are frequently regarded as complete (total loss of cord function), and incomplete, the certainty of outcome of neural function does not exist for cauda equina injuries, where the roots behave more like peripheral nerves with potential for later recovery.
Severe neural injuries above L1 will damage the lower spinal cord resulting in an upper motor neurone picture of spastic paralysis. Similar severe neural injuries below L1 may result in a lower motor neurone flaccid paralysis. Neural injuries between these two extremes may result in complex patterns of injury further complicated when lesions are incomplete. An interesting variation, usually associated with L1 injuries, is the conus paraplegic. Injury to the tip of the cord, the conus, results in paralysis of the sacral segment. Loss of bladder and bowel control occurs, yet the patient’s cauda equina roots originating proximal to the conus may be spared, given near normal lower limb function. If the conus injury is severe the sacral paraplegia will be of a lower motor neurone variety, with disruption of the reflex arc, and thus a different pattern of bladder and bowel control failure. This pattern of injury also results in failure to achieve erection in the male.
These differing patterns of neural injury mandate meticulous neurological evaluation at initial assessment. Some have advocated that the classification of “complete” spinal cord injuries be limited to the upper thoracic spine and cervical spine only, because of the potential for recovery of the differing neural tissues in injuries at the thoracolumbar junction.
The role of surgery in the treatment of both neurologically impaired and intact patients is controversial. Cord injury consists of the primary contusion, secondary injury due to cellular changes at the injury site, and the effects of ongoing neural compression. The first mechanism is amenable only to preventative treatment. The secondary injury response is under intensive investigation for effective agents that may modify this process. The use of Methylprednisolone in the immediate post injury phase has been shown to improve outcomes in the NASCIS studies (Bracken et al 1990), but this improvement has not been substantiated in other studies and its role remains controversial. The role of surgery for any ongoing compression remains controversial. It is intuitively attractive to consider that decompression of damaged and compressed neural structures could reduce cellular and neuronal deformity, decompress vascular structures and decompress cells and neurones in the cord. Clinical studies have suggested that effective cord decompression after injury is associated with improved outcomes. A number of authors have suggested that anterior decompression results in better neural recovery than non-operative treatment or posterior decompression (Clohisy 1992, Bohlman 1992, Bradford 1987, McAfee 1985, Transfeldt 1990). Gertzbein (1992a) summarized this data by suggesting that although non-operative treatment is associated with improved neural function, anterior decompression is associated with more rapid and better neurological recovery. Unfortunately review of these studies reveals that the methodology does not stand up to the vigorous evaluation required for evidence-based practice.
Equally effective neural recovery has been demonstrated with conservative management of these injuries (Katoh 1996) and critical literature reviews have cast doubt about the real benefit of surgical decompression (Boerger 2000). In reality the role of surgical decompression is likely to remain unclear, and the ever present call for randomised prospective studies remain unfulfilled. What is clear is that late decompression, once natural recovery has ended, is associated with further improvement in neural function (Bohlman 1992, 1994, Transfeldt 1990). The role of active decompression is also supported by animal studies (Fehlings 1999), where a meticulously controlled experimental environment can be developed. Such studies have shown benefit from early and late decompression.
Accepting the defects of clinical evidence, many practitioners will use the evidence from late decompression, or from animal studies, to justify early decompression in spinal cord or cauda equina injuries. As noted above the evidence suggests anterior decompression will be more effective for anterior neural compression such as occurs in a burst fracture. The disadvantage of posterior approaches to achieve anterior decompression include the need to resect major portions of the neural arch (often uninjured) to obtain access to the middle column, working around an already damaged neural structure risking increased neural damage, and ineffective anterior decompression as compared with that achievable by anterior approaches. Finally it is difficult to reconstruct the anterior and middle columns after a posterior approach has been used to decompress a burst fracture, and there is a significant incidence of construct failure (McLain 1993).
The management options for thoracolumbar burst fractures include non-operative and operative care. The latter can be divided into anterior, posterior or combined approaches. It is clear that in burst fractures without neural injury, there is little to choose from between non-operative care and posterior surgery with short segment pedicle fixation devices. Non-operative care can be limited to two weeks of bed rest followed by bracing and mobilization. There is often slight settling of the fracture with the development of a mild kyphosis but this does not correlate with inferior clinical results and seldom results in any clinically detectable deformity. Late pain of significance is uncommon and the development of secondary neurological deficits is rare (Weinstein 1988).
Posterior surgery with pedicle screw constructs over a short segment stabilizes the fracture and allows early mobilization, much as non-operative regimes do. Recent prospective randomised studies comparing these two treatment options suggest there is no clinical advantage of surgery over non-operative care (Wood 2000, Shen 2001, Alanay 2001). Surgery corrects deformity but modest recurrence is common, even with attempts to perform trans-pedicular bone grafting, as the anterior column remains deficient (Alanay 2001). It should be emphasised that this modest recurrence of kyphosis in the operated group is also not of clinical relevance. Patients who have operative treatment of thoracolumbar burst fracture from a posterior approach may require later removal of the instrumentation. This is balanced, in our experience, by a very small number of these patients who, when treated non-operatively, develop disabling pain and need to have late anterior reconstruction.
When neural injury does occur in association with a burst fracture, it is our preference to perform anterior surgery with corpectomy and reconstruction. The rationale for this is that the ongoing neural compression may recover better with effective decompression. Admittedly the evidence and support of the surgical approach, as discussed above, is not absolute. In cases where mild burst fractures occur, such as Denis Type B, without neural damage, most are treated non-operatively with a short period of bed rest then mobilization in a carefully moulded TLSO.
Relative indications for anterior surgery also include severe comminution, fragmentation, kyphosis and AP mal alignment (Denis Type D injuries). In older patients with medical comorbidities, who sustain thoracolumbar burst fractures with significant neurology; we may well prefer a posterior approach, accepting that the patient may be more suited to a shorter less extensive surgical procedure not requiring a thoracotomy.
The advantages of anterior surgery include direct atraumatic decompression of the spinal canal when neural injury has occurred, and the ability to reconstruct the anterior column deficiency. Disadvantages of anterior surgery include the more extensive approach required, lack of familiarity to many spinal surgeons, the potential for thoracotomy pain, and the potential for pulmonary complications.
Contraindications to anterior surgery alone are Type B and Type C injuries. The posterior ligamentous deficiency in the Type B injury is not controlled by anterior surgery and potential recurrence of kyphosis exists. Dislocation of the thoracolumbar spine in Type C injuries is not appropriately reduced from the front. Anterior approaches are contraindicated in acute Type B and Type C injuries. Obviously there may be more complex injury patterns, where the vertebral body fails in axial compression with a burst fracture pattern, and the posterior structures fail in tension like the Type B injury. These “kyphotic” burst fractures may be best managed by a combination of anterior and posterior surgery.
However the anterior approach may be required to osteotomize and reconstruct the anterior column for late management of inappropriately treated Type B and Type C injuries where there is a kyphotic mal-union.
Where anterior surgery has achieved decompression via vertebral body corpectomy, the reconstruction requires an anterior column reconstruction and the application of internal fixation to stabilize the reconstruction, normally from the vertebral body above to that below the corpectomy.
Corpectomy reconstruction options must provide support to the anterior and middle columns as the major biomechanical function. Options include autogenous bone graft, long bone allograft, and titanium mesh cages. Autologous bone graft options include iliac crest, fibula or rib. All of these have problems with donor site morbidity and non-optimum shape for reconstruction. The narrow shape of the rib and fibula may cause subsidence through the adjacent end plates. Iliac crest provides the best option, and has been successfully used in large series (Kaneda 1997). However iliac crest may be limited in supply, of poor structure and shape, and result in major iliac crest deformity. Occasional translational instability after the use of iliac crest grafting (Figure 2) has led to the use of other more appropriately shaped anterior column reconstruction options.
Fig. 2: Late failure of anterior reconstruction with Z plate and tricortical graft. The graft has tilted posteriorly and later posterior stabilization was required.
Long bone allograft segments, packed with autograft, are a useful alternative (Finkelstein 1999). Alternatively the use of titanium mesh cages has allowed optimum biomechanical function for anterior column support, along with bone grafting with cancellous autograft, which is packed into the centre of the cage. This attractive option allows easy contouring of the cage to fit the corpectomy site, the use of the cancellous bone resected from the fracture site, and no graft site morbidity. The cylindrical cages fit around the strong peripheral rim of the vertebral body endplate (Grant 2001), and because of their shape are resistant to translation or toggle. The serrated ends of the cage likely resist axial rotation (Figure 3).
Fig. 3: Post-operative radiograph of burst fracture reconstruction using titanium mesh cage and anterior stabilizing plate for thoracolumbar burst fracture.
In our Unit, experience with these cages suggests that they are a reliable option for corpectomy reconstruction. Kyphosis can be reliably corrected when necessary, and any significant kyphosis recurrence or implant settling into adjacent vertebrae is minimal and clinically unimportant. The only failures we have seen with titanium mesh cages have occurred in late reconstruction when the cage is placed obliquely at the corpectomy site. When this occurs, surgery fails to align the proximal and distal vertebrae in both the sagittal and coronal planes. The cages are then placed in a tilted position and this has led to instrument failure and non-union.
Once cages have been placed, a lateral stabilizing plate or a screw and double rod system will provide additional stability. These devices are placed according to the manufacturers recommendations, with screws achieving bicortical fixation, sparing the discs above and below the proximal and distal vertebrae. Such constructs minimize the reconstruction length and the length of the fused segment.
a) Approach and Positioning
The patient should be in a direct lateral position, shoulders and hips perpendicular to the floor, to assist with accurate screw placement. The kidney rest is at the apex of the thoracolumbar deformity and is elevated so as to prevent the spine sagging between the chest and the pelvis when in the lateral position. In this position the abdomen of more obese patients will sag away from the spine (Figure 4). The approach is from the left side (away from the vena cava) unless surgery is late for a mal-union where there is a coronal plane abnormality. In that situation the fracture should be approached from the concave side so as to distract that side at the time of reconstruction (see below).
Fig. 4: Patient positioning for surgery. The patient is in a right lateral position for a left thoracoabdominal approach.
b) Incision Level
Generally the rib resected in the thoracotomy and thoracoabdominal incision should be two levels above the vertebra above the fracture (i.e. that will be instrumented proximally). It is generally easier to dissect from proximal to distal. Some variation may be required depending on the obliquity of the ribs and this should be assessed from pre-operative lateral x-rays. If the incision is not ideally placed proximal or distal ribs may be osteotomized to improve display. Fractures of T11 and above require thoracotomy only. Fractures of T12 to L2 generally require a thoracoabdominal approach (Figure 5), and fractures of L3 can be managed through a 12th rib incision that does not require opening of the chest.
Fig. 5: Thoracoabdominal approach via a lower rib thoracotomy. The lung and diaphragm are visible. The costal cartilage will be divided and represents a useful landmark for later wound closure.
c) Vertebral Column Dissection
Segmental vessels should be ligated over the three levels (fracture and proximal and distal vertebrae for instrumentation). The segmental vessel should be ligated at the mid-lateral level of the vertebral body. The surgical exposure should demonstrate the anterior and lateral side of the body from where the approach is being made. In the lumbar spine the psoas needs to be mobilized. It can be dissected off the vertebral body where it originates above and below each disc. The use of a Cobb elevator along the disc will allow the psoas to be gradually diathermied away from its origin, allowing the concavity of the vertebral body wall to be displayed and the segmental vessels to be ligated. The psoas muscle may obstruct an adequate view of the vertebral body and the decompression. Once it has been mobilized the muscle can often be held out of the way by a Steinmann pin inserted into the lower vertebra below the fracture.
In the chest, the rib head of the relevant vertebral bodies should be resected, so as to allow identification of the pedicle and the foramen, and to assist with anatomical landmark identification. Anteriorly the vertebral body should be displayed beyond the mid line. Dissection is best achieved at the level of the disc where there is minimal vascularity. The soft tissue can be then raised off the vertebral body as a sleeve. Vertebral body bleeding from adjacent vertebra is controlled with diathermy and bone wax.
Firstly, the disc above and below the vertebra is resected back to the posterior portion of the disc (Figure 6). The vertebral body is then removed at the front and the left side back to the canal. Curettes and Kerrison rongeurs are used, or a high speed burr. Dissection into the canal starts inferiorly and should be carried across to the opposite side so that the neural elements do not bulge out and obscure the view. The maximum compression of the neural elements is normally at the proximal level of the involved vertebral body. Dissection proceeds proximally to resect the fragment that is wedged between the two pedicles. This can be removed with curettes. The remainder of the discs above and below the fracture are resected to demonstrate posterior longitudinal ligament or dura.
Fig. 6: Thoracolumbar spine displayed with mobilization of psoas. Discectomy has been performed at the levels above and below the fracture.
The pedicles on each side must be identified to identify the lateral extent of the decompression. The canal above and below the relevant disc should be palpated to confirm decompression (Figure 7). Dissection must be far enough across the vertebral body to allow central seating of a cage or reconstruction device. If a late decompression has been performed, after a fracture mal-union, then the resection must go all the way across to the opposite side so as to completely release the anterior and middle columns and allow lengthening of these columns. If there is some difficulty resecting bone on the opposite side, perhaps due to blood loss, this remaining wall can be osteotomized back to the canal. Obviously this has to occur below the level of the pedicle. Occasionally, when late reconstruction of a mal union is being performed, the anterior longitudinal ligament may be a tight fibrous band and may require division or resection. Soft tissue planes must be carefully identified to avoid vascular damage. It is important to preserve the end plates of the adjacent vertebra. Haemostasis can be obtained with gel foam, surgical or bipolar diathermy in the canal. The cell saver is often valuable.
Fig. 7: Corpectomy with display of the dural tube partially covered with gel foam.
The screws of the stabilization device must be placed above and below the vertebra resected. They should be placed parallel to the relevant end plates. Additional information regarding the alignment of the discs above and below that resected can be gained by placing of a needle in the next distal disc. It is important to check the correct position of the trunk, so as that screws are placed in the correct plane. When screws are placed it is important that the screws are longer rather than shorter, so that the opposite cortex of the body is engaged to obtain maximum biomechanical strength. The corpectomy site should be distracted and this is best done with vertebral body spreaders rather than overloading the screws. The screws can be used to stabilize the distraction when the reconstruction device is placed. Cages offer an excellent option and can be sized to fill the corpectomy space. They must be filled with bone, normally obtained from the vertebral body resection, or augmented with rib graft (Figure 8). A cage should be placed anteriorly and centrally and obviously must be clearly away from the canal (Figure 9). Watson-Jones or other probes will allow palpation to see placement is anterior to the canal. At this point the lateral stabilization device can be applied with further screws depending on which device is used (Figure 10). Normally the removal of the distraction instruments will allow adequate compression but further compression can be achieved stabilizing the vertebrectomy reconstruction device.
Fig. 8: Titanium Mesh Cage packed with autogenous cancellous bone resected from the fracture site.
Fig. 9: Cage placed at the corpectomy site. Bolts for the lateral stabilizing plate are in situ at the vertebra above and below. The canal and dural tube are obscured by gel foam.
Fig. 10: Intraoperative picture of the cage and lateral plate after anterior reconstruction.
Chest drains (two) are required if thoracotomy has been performed. Strong non-absorbable circum-costal sutures can be useful to re-approximate the ribs and this is assisted by lowering of the kidney rest. If such sutures are used they must be passed subperiosteally on the inferior side to avoid entrapment of the subcostal vessels and nerves. Marcaine to the intercostal interspaces is useful to assist with post-operative pain management. An intra-pleural catheter can be placed to allow ongoing post-operative analgesia.
g) Post-Operative Management
The patient should be nursed supine and log rolled for comfort. The chest drains are removed when x-rays show that the lung is expanded and the drainage reduced. A TLSO is then applied. Because anterior fixation requires stabilization in the relatively weak vertebral bodies, post-operative immobilization is routine. A moulded TLSO or an “off the shelf” TLSO can be used depending on fit. These are maintained for three months after surgery.
In old burst fractures where there has been mal-union, or flexion distraction injuries, or Type C injuries, mal-united anterior and middle columns may require treatment with resection and realignment. In these cases there must be clear analysis as to the nature of any deformity in the sagittal plane. Mal-union following a stable burst fracture likely involves shortening of only the anterior and middle columns. If late surgery is required, complete resection of the vertebral body will assist decompression of the canal and allow reconstruction using anterior reconstruction and stabilization only.
If there has been mal-union of a flexion distraction injury or Type C injury, then often there is both shortening of the anterior and middle columns, and lengthening of the posterior column. In this situation surgery to address sagittal malalignment would involve resection of the anterior and middle columns and lengthening of these columns, followed by osteotomy of the posterior columns and shortening of the posterior structures. In this situation the best option is to resect the anterior and middle columns and insert a cage that recreates anterior height. Anterior stabilization devices are not used.
A secondary procedure is performed with a posterior approach, osteotomy and shortening of the posterior column and stabilization with pedicle screws to the level above and below. If these injuries are combined with any coronal plane deformity, where an acute angular mal-union complicates the situation, then the anterior approach should be performed from the concavity of the coronal plane mal union. This allows resection of the anterior and middle columns in the concavity and then distraction not only of the kyphotic deformity, but also of the concave aspect of the coronal plane deformity. If one were to mistakenly attempt such surgery from the convexity of the coronal plane deformity, there would be the conflicting desire to lengthen the anterior and middle columns, so as to correct the kyphosis, but at the same time to compress the anterior and middle columns from the convex side, so as to correct the coronal plane deformity. For this reason it is important to approach late complex two plane deformities from the concavity of both deformities.