Fracture reduction by postoperative mobilisation for the treatment of hyperextension injuries of the thoracolumbar spine in patients with ankylosing spinal disorders

In this study, we assessed clinical and radiological outcomes after surgical stabilisation of hyperextension injuries of the ankylosed thoracolumbar spine by either the conventional open rigid posterior instrumentation or by following a new treatment concept involving percutaneous less rigid posterior instrumentation and fracture reduction by postoperative mobilisation. The latter treatment strategy was developed in an attempt to (1) improve efficiency of fracture reduction and restoration of pre-injury sagittal alignment to ensure osseous contact at the fracture site which is a prerequisite for osseous healing, as well as to (2) reduce the rate of wound healing complications and infections frequently encountered in these patients, probably as a consequence of atrophic and degenerative trunk muscles resulting from muscle inactivity due to spinal ankylosis.

Spinal fractures typically involve the motion segment (i.e. discs, facet joints and ligaments) and are therefore regarded as articular fractures. Since a reconstruction of the motion segment is not feasible given the available treatment options, the surgical treatment of these injuries generally requires fusion of the motion segment. In the ankylosed spine, however, the motion segments have already spontaneously fused and the spine acts biomechanically more like the diaphyseal part of a long bone. The aim of the surgical treatment of fractures of the ankylosed spine are therefore similar to those of shaft fractures and include reduction and stabilisation of the fracture in order to promote osseous healing [2, 6].

At first glance, plate fixation via an anterior approach seems to be the best option for both reduction and stabilisation of hyperextension injuries of the ankylosed thoracolumbar spine. The exposure provided by the anterior approach may allow for direct reduction of the hyperextension fracture, while anterior plate fixation biomechanically acts as a tension band for the neutralisation of extension forces. Anterior approaches and plate fixation, however, have some major drawbacks in these patients. First, osteoporosis is frequently associated with ankylosing spinal disorders [1, 4, 8, 16] and reduces the strength of screw anchorage. Second, long-lever arms lead to high torques in the ankylosed spine. These moments have to be neutralized by the implant, until healing has occurred. Both facts highly increase the risk of failure of the bone–implant interface with subsequent implant loosening. The use of longer plates with more points of fixation reduces the risk of implant loosening. It may, however, be hard to implant longer plates via an anterior approach in the presence of a rigid thorax and pre-existing kyphosis [2]. In addition, peri- and postoperative pulmonary complications are frequently observed in these patients [7, 9, 17]. In our study, more than half of all patients developed postoperative pulmonary complications. This additionally argues against an anterior approach.

Posterior instrumentation is, therefore, regarded as the standard treatment for hyperextension injuries of the thoracolumbar spine in patients with ankylosing spinal disorders [3, 4, 6, 9]. Implant loosening, however, is a major issue in these patients after posterior instrumentation as well with rates reported as high as 15% [18]. There are several options to reduce the risk of implant loosening. First, cement augmentation of the pedicle screws increases the strength of the bone-implant interface. Second, multisegmental posterior instrumentation distributes the entire load to more points of fixation [19]. The third option is the reduction of the rigidity of the instrumentation. Less rigid constructs absorb part of the energy during mobilisation and, therefore, decrease the strain at the bone-metal-interface. The rigidity of the posterior instrumentation may be reduced by both increasing the working distance of the rods (i.e. the distance between the pedicle screws adjacent to the fracture) [19] as well as by reducing the bending stiffness of the longitudinal rods [20]. In our study, we used two rods of different bending stiffness. We observed no implant loosening in the group with softer rods (PLR group) despite a lower rate of cement augmentation of the pedicle screws (1/6 cases) compared to two cases of pedicle screw loosening in the OR group (cement augmentation in 8/14 cases, Table 2). We, therefore, advocate to reduce the rigidity of the posterior instrumentation by using soft rods. In addition, the working distance of the rods may be increased by leaving at least one vertebral body adjacent to the fracture site without pedicle screws, even if the fracture type would allow the insertion of pedicle screws in all vertebral bodies (Figs. 1b, 2). Moreover, considering the high rate of postoperative wound complications reported for patients with ankylosing spinal disorders (see below), using a percutaneous less rigid instrumentation, and optionally pedicle screw cement augmentation seems to be more favourable to reduce the risk of implant loosening than excessively increasing the instrumentation length at the risk of soft tissue complications.

Besides decreasing the risk of implant loosening, the use of soft rods may be a solution for an additional relevant issue in these patients, and that is fracture reduction and restoration of pre-injury sagittal alignment. While prone positioning facilitates the reduction of kyphotic malalignment in the much more common thoracolumbar compression as well as flexion-distraction injuries, this does not hold true for hyperextension injuries which come along with lordotic malalignment. Intraoperative prone positioning of the patient (which is the only way of positioning to allow for pedicle screw insertion and posterior instrumentation) rather impedes reduction of lordotic angular displacement associated with hyperextension injuries. In our experience, efforts to modify intraoperative prone positioning to facilitate reduction of lordotic angular displacement are usually ineffective and without success. We, for example, tried various strategies of padding below the patient at the level of the apex of lordotic angulation, but this did only marginally help to improve fracture reduction and regularly resulted in substantial intraoperative ventilation problems in these commonly aged and often multimorbid patients due to an increase of intraabdominal pressure. Furthermore, brisk reduction manoeuvres additionally may lead to loosening of the pedicle screws and bears the risk of iatrogenic neurological deficits in these patients [8]. Thus, reduction of hyperextension injuries of the ankylosed spine is extremely challenging; in many cases, no adequate reduction can be achieved by the conventional intraoperative reduction strategies, and despite all reduction efforts, a residual lordotic angular displacement often cannot be avoided. Our concept, therefore, includes flat prone intraoperative positioning of the patient and stabilisation of the fracture in this position without actively performing fracture reduction (Figs. 1b, 5b). Postoperative mobilisation then induces bending of the rods and consecutive fracture reduction (Fig. 2) and thus restores pre-injury sagittal alignment. Our data show that fracture reduction by postoperative mobilisation occurs within the first 3 weeks, when using soft rods (PLR group), and within 6 months with the use of more rigid rods (OR group) (Table 4; Figs. 2, 3). Rapid postoperative reduction with the use of soft rods is highly desirable, as reduction not only promotes osseous healing of the fracture by adding compression to the fracture site (Fig. 4), but also increases the stability of the construct by restoring the buttress function of the anterior column of the spine, therefore, decreasing the risk of pedicle screw loosening and facilitating mobilisation of the patient. This additionally argues for the use of soft rods for posterior instrumentation of hyperextension injuries of the thoracolumbar spine in patients with ankylosing spinal disorders. Percutaneous less rigid instrumentation allowed to restore pre-injury sagittal alignment in all of our patients and osteotomy secondary to trauma was necessitated in none of them.

While our concept of postoperative reduction by mobilisation may be new for the treatment of spinal fractures, it is well known in the treatment of shaft fractures of the upper and lower extremity. For example, dynamic locking of intramedullary nails results in compression of the fracture site during postoperative weight-bearing. This concept requires that the direction of axial compression is provided by the nail and the amount of shortening is limited by the cortical contact between the main fragments. Accordingly, dynamic locking is particularly advisable in simple transverse fractures, while comminuted fractures with no cortical contact between the main fragments require static locking. In hyperextension injuries, the instrumentation is aimed to prevent translation and an increase in lordotic angulation, but, in contrast to compression injuries, does not have to protect the anterior column against compressive forces as the vertebral body is not comminuted but horizontally disrupted. In our concept, osseous contact at the fracture site restores the buttress function of the anterior column of the spine and limits further bending of the rods. The bending should additionally be directed by intact posterior walls, which act as a fulcrum (arrows in Figs. 1a, 5a). The implant itself avoids translational displacement of the fracture during mobilisation to prevent spinal cord encroachment. Accordingly, we do not recommend the use of our concept in fractures with translational displacement or distraction, which disallows the posterior walls to act as a fulcrum, and in fractures with comminution of the anterior column, which is, however, very rare in hyperextension injuries.

Another difference between the two techniques used in this study is the open vs. percutaneous approach. An increased perioperative blood loss during spinal surgery in patients with ankylosing disorders has been described in several studies [8, 21]. It is reasonable to assume that a percutaneous approach may be beneficial in terms of reducing intraoperative blood loss particularly for multisegmental instrumentation of the spine. Due to the retrospective nature of this study, however, we were not able to assess the amount of blood loss in our patients collective. In addition, the risk of impaired wound healing and infection is relatively high in these patients. Backhaus, for example, reports an infection rate of 14% [18]. Atrophic and degenerative trunk muscles as a result of inactivity due to the ankylosed spine may account for this finding. We had a relative risk of impaired wound healing of 21.4% (3/14) after performing an open procedure (OR group), while there was no case of impaired wound healing in the percutaneous group (PLR group). A potential drawback of a percutaneous procedure is a higher rate of pedicle screw misplacement, as there is no clinical control of pedicle screw insertion using landmarks, but radiological control only. The clinical landmarks, however, may be hard to identify even after using an open approach especially in patients suffering from ankylosing spondylitis with ossification of the posterior ligaments and joints. In our study, there was no significant higher rate of pedicle screw misplacement in the PLR group (12.5 vs 9.2%, p = 0.78, Table 2). We, therefore, think that percutaneous pedicle screw placement is safe in these patients when performed by surgeons, who are familiar with the procedure.

Our data show that the patients in the PLR group had less non-surgical postoperative complications (Table 3) and a shorter length of hospital stay despite more co-morbidities according to the Charlson Index. We explain these findings by the lower rate of surgical complications and the smaller surgical trauma due to the percutaneous approach in this group. An interesting finding of our study is that clinical outcomes at final follow-up were very good in both groups. The loss in VAS Spine Score, Parker Mobility Score, and Barthel Index (as compared to pre-operative level) was only minimal (Table 5). This is in contrast to studies assessing, for example, the outcome after thoracolumbar burst fractures. Knop reported an average VAS Spine Score loss after combined posterior–anterior stabilisation of 19.7 points [22], while Reinhold even found a loss of 32.2 points after non-operative treatment [23]. In our study, the VAS score loss was 0.9 and 3.2 points, respectively. These findings may be explained by both the lower demand in a predominant geriatric patient population and by the fact that thoracolumbar burst fractures lead to loss of motion segments, while fracture healing restores the pre-traumatic state in patients with ankylosing spinal disorders.

This study has several strengths worth mentioning. First, and in contrast to the previous reports, our study comprises a relatively high number of patients and includes a control group, whereas the currently available literature on hyperextension injuries of the ankylosed thoracolumbar spine is limited to case reports and small case series solely describing patients’ characteristics, mixing up patients with and without neurological deficits and often without specifying surgical stabilisation. Moreover, these studies are lacking a control group as well as a comparison of different treatment strategies. Second, our study not only reports radiological but also clinical outcomes assessed by a set of validated outcome scores to quantify pain and function, mobility, social independency, and quality of life after a respectable mean follow-up period of 29.2 months, the longest reported so far. Third, we proposed a new treatment approach for these rare but clinically challenging injuries and, for the first time, comprehensively outlined and discussed important aspects to be considered when choosing the optimal treatment strategy for these patients.

Some limitations of our study, however, have to be noted. First, this is a retrospective study with all limitations associated with this study design. For example, we were not able to retrospectively assess the amount of intraoperative blood loss. Second, a higher number of patients would have increased the power of the statistical analysis and may have revealed more significant differences between the two groups; however, given the rarity of this injury pattern, small sample size is a limitation inherent to studies addressing thoracolumbar fractures in patients with ankylosing spinal disorders [24]. Small sample size in this study is, furthermore, a result of exclusion of patients with neurological deficits and additional fractures which was inevitable to be able to compare clinical outcome between groups. In addition, there were more patients in the OR group than in the PLR group. The authors, therefore, are aware that the data of this study cannot form a sound basis to guide treatment approaches, but, nevertheless, represent a “proof of principle” of a new concept to address this type of injury, considering the specific biomechanical characteristics of the ankylosed thoracolumbar spine. Third, the choice of instrumentation did not follow institutional guidelines, but was made by the involved surgeon. Fourth, functional scores (i.e. VAS Spine Score, Parker Mobility Score and Barthel Index) for the time prior to the injury were assessed at final follow-up. Finally, patients with two different ankylosing spinal disorders (AS and DISH) were included in the study. Besides several differences in aetiology and radiographic appearance, however, these two ankylosing spinal disorders share surgically relevant clinical and biomechanical features, which legitimates the merging of both diseases into one study group for a surgery-related study [6].