The indication of fusion to the pelvis in neuromuscular scoliosis is based on the underlying disease rather than on pelvic obliquity

The goal of scoliosis correction in neuromuscular scoliosis is a straight spine over a level pelvis in order to maintain sitting balance in a mostly wheelchair-bound population. An old controversy is whether to include the pelvis in the fusion [1‐4]. The pelvis is a strong basis for the spinal fusion and its inclusion facilitates the control of the pelvic obliquity and makes it predictable [5‐9]. On the other hand it prolongs the operation, increases blood loss [10] and reduces mobility. Also it increases the risk for infection due to proximity to the rectum in a population where incontinence is a frequent problem [11]. Thus some authors advocate shorter fusions to L5 [3, 12‐15]. Luk et al. already mentioned in 1986 that fusion to L5 could be sufficient to correct the main proportion of the pelvic obliquity in NMS due to the role of the iliolumbar ligament which holds the L5 vertebra in a relatively constant relationship with the sacrum [16]. Takaso et al. postulated that fusion to L5 is safe in patients with SMA in absent of high preoperative pelvic obliquity, but the amount of PO that can be considered safe remains unclear [15]. Takaso et al. suggested that segmental pedicle screw instrumentation to L5 was safe in patients with DMD with stable L5/S1 articulation and a minimal L5 tilt of 15° or less [14].

Most surgeons base their decision on radiologic parameters. It is recommended to include the pelvis if the curve exceeds 60°, the apex is L4 or lower or the pelvic obliquity is above 15° [1, 4, 12‐14, 17]. These recommendations do not consider the cause and origin of the curvature. Often cases of neuromuscular scoliosis are regarded as a single entity and are only looked upon in contrast to idiopathic or congenital types. This neglects the fact that the deformity of the spine is only a symptom of the underlying neurological and muscular disease and that there is a difference between spastic paralysis or muscular diseases.

The Scoliosis Research Society Classification of Neuromuscular Spinal Deformities differentiates between primary neuropathies caused by lesions of the upper or lower motor neuron and myopathies like muscular dystrophies (Table 1) [18, 19]. A lesion of the upper motor neuron as in CP patients comes with a loss of inhibitory input causing spasticity [20, 21]. In patients with lesions of the lower motor neuron (e.g. SMA) or in myopathic conditions (e.g. DMD) the descending inhibitory tracts are still intact resulting in a reduced muscle tone. Therefore, it is sensible to divide neuromyopathic patients in those with spasticity (like CP) and those with a flaccid condition like DMD and SMA.

Furthermore, there was a development and improvement of the surgical techniques and implants since these recommendations to include the pelvis in the fusion were published. Even in textbooks of 2008 [22] the unit rod and Galveston or Dunn-Mc Carthy technique were described to correct NMS. Now that segmental pedicle screws and either ilium screws or S2AI screws are the new standard, we must rethink this. Especially the added value of pelvic fixation surgery remains unclear and patients which are fused to the pelvis have a higher risk for complications [23]. But secondary revision to the pelvis has even more complications than including the pelvis at the index surgery [11]. Thus the question whether to include the pelvis or not still is open to discussion. In a Cochrane review of 2015 Cheuck et al. wrote on this matter that in absence of evidence clinicians may need to consider anecdotal evidence and their personal experience as well as expert opinion as guidance for their decision on the best care for an individual patient [24].

In our hospital we started to base our decision more on the flexibility of the curve and on the muscle tone. In patients with lowered tonicity as in SMA and DMD, we usually stop at L5 and in patients with high tonicity as in CP we usually include the pelvis. After doing so for several years we want to describe our decision pathway and experiences in order to help other surgeons in their decision pathway.

Most patients showed the large C-shaped curvatures typical for neuromuscular scoliosis. The apex of the main curve was thoracic (higher than Th12) in 26.3% of the cases, thoracolumbar (Th12/L1) in 30.0% and lumbar (lower than L1) in 43.8% of the cases.

Primary posterior pedicle screw-based spondylodesis was performed in all patients. In 76 of the 79 cases, instrumentation to the lower lumbar spine (L4/5) or to the pelvis was done. In 3 of the patients, the lowest instrumented vertebra (LIV) was chosen above the L4 level. As these were rather short thoracic curvatures, they were excluded from further analysis.

As shown in Table 3 there was a tendency for patients with cerebral palsy to have a slightly lower apex of curvature than those with DMD or SMA. However, this difference was not statistically significant (p = 0.536). Patients with flaccid palsy had an average major curve of 68° preoperatively in the sitting position, which corrected to an average of 46° in traction images. In patients with CP, the Cobb angle of the major curve was 89° preoperatively in sitting position and 60° under bending/traction. The difference in preoperative curvature between sitting (p = 0.001) and bending/traction (p = 0.021) was statistically significant. Patients with flaccid palsy had a preoperative pelvic obliquity of 16.6° in sitting position and 11.6° under bending/traction. This was similar with CP patients (20.6° in sitting and 11.1° under bending/traction).

The L4-take-off was significantly higher in patients with CP (39.9°) than with flaccid palsy (30.5°) and was corrected to similar values of 25.4° and 21.6° under traction. The truncal shift was higher in CP patients in the sitting position (25.4° to 19.0°) but smaller under traction (10.1° to 12.4°). All in all there were no significant preoperative differences between patients with DMD or SMA and CP concerning truncal shift.

When traction and sitting films were compared, patients with flaccid palsy had a slightly higher flexibility, although the differences did not reach the significance level (Table 3).

Of the patients with degenerative muscle disease or SMA, 84.2% were instrumented to the lower lumbar spine and only 15.8% to the ilium, whereas those with spastic palsy were instrumented to the ilium in 58%. Patients with degenerative muscle diseases had an average postoperative major curve (in sitting) of 27° compared to 40° in CP patients. Although the pelvis was not included in the majority of the patients, the PO and the truncal shift could be corrected to an average of 8°. While patients with CP had an average pelvic obliquity of 12° and a trunk shift of 17°.

In addition, the percentage of operative correction (POC) and the Cincinnati correction index (CCI) were calculated for the main curve (Table 2). Although there was a significant difference in postoperative major curve and trunk shift, there was no statistically significant difference in POC and CCI. This means that the extent of correction was comparable in both subgroups, which can be explained by the fact that CP patients also had a greater initial spinal deformity.

For SMA and DMD patients instrumented to L5 the mean operative time was 202 min, for those instrumented to the ilium it was 258 min and for CP patients instrumented to the ilium we needed an average cut-suture-time of 289 min.

A subgroup analysis was performed on patients who were instrumented to L5 although according to previous criteria fusion to the pelvis should have been done (we called it the “new lumbar group”). This subgroup was compared to patients who were stabilized to the ilium (pelvic group) and those who even according to conventional criteria only required correction to the lower lumbar spine (traditional lumbar group).

This “new lumbar group” of 38 patients was analysed more closely regarding preoperative conditions and postoperative outcome (Table 4). 33 of these patients (87%) had a flaccid palsy and only 5 had cerebral palsy with spasticity (13%).

Patients of the “new lumbar group” had mainly DMD or SMA and were more flexible in traction radiographs. The preoperative curvature of 74° ranged between pelvic group (90°) and traditional lumbar group (50°). Flexibility was not significantly higher in the new lumbar group than in the other two. But the remaining curve in the traction film was still 46° and thus higher than in the traditional lumbar group 5. In patients we fused to the ilium the main curvature remained above the critical mark of 60° even under traction (62°). The difference in the 3 groups was significant (p = 0.002).

The pelvic obliquity under traction corrected to 11° in the “new lumbar group” and thus ranged clearly below the cut-off of 15°. Those who were only corrected to a mean of 14.5° under traction were fused to the pelvis. Similarly clear values were shown regarding the L4 take-off and the trunk shift in bending/traction images (Table 4).

In the “new lumbar group” a correction of 66% to a mean of 26,8° was achieved. This resulted in balanced spine with a pelvic obliquity of 8,1° and a postoperative trunk shift of 8,3°. The postop results were similar to those of the patients which primarily met the criteria to fusion to L5 (Table 4).

After one and two years of follow-up no significant change in the pelvic parameters was seen in one of the three groups (Table 4). The influence of the spasticity is displayed in the boxplot (Fig. 2). None of the patients had an increase in pelvic obliquity of more than 10° after one- or two-year follow-up. One patient with CP had to be revised and fused to the pelvis due to screw loosening in L5. As radiography only reflects a moment, we asked all patients how they get along in daily life and whether they feel stable. None felt an increasing instability or worsening of the sitting position and thus wished for secondary lengthening. An example is given in Fig. 1.

Our experience showed instrumentation and fusion can be performed shorter than the sacrum or pelvis more often than previously assumed, provided that the spine can be centred over the pelvis, good correction is achieved, and high muscle tension or spasticity are absent. As revision surgery shows higher rates of implant failure [11] a thorough consideration of the extent of stabilization is paramount. The criteria commonly used to decide whether to include the pelvis or not only considered the static conditions and the preoperative X-rays [26]. Radiological parameters that are suggested are the pelvic obliquity, L5 tilt, apex of the curve and trunk shift [1, 2, 4, 7, 12‐14, 17, 27‐29]. Mehta et al. also suggest to take the age of the patient into consideration [13]. Takaso et al. also measured preop flexibility but did not draw any conclusions from it concerning the extent of stabilization [14]. Modi et al. and Whitaker et al. suggested to go to the pelvis in patients with pelvic obliquity > 15° or in the presence of severe lower extremity contractures, but did not regard the flexibility and possibility of the patient to stay centred [4, 26]. Ahmady et al. suggest to differentiate between ambulatory and non-ambulatory patients when deciding between distal spinal and pelvic anchors of growth-friendly implants in children with hypotonic NMS [30]. Douleh et al. in their 2021 literature review concluded that fusion shorter than the pelvis can be considered in minimally ambulatory patients with NMS with adequate head control, the absence of hip subluxation or dislocation and with a pelvic obliquity of less than 15° [31]. Moon et al. emphasized the importance of the correction of the pelvic obliquity to address and improve the patients primary needs like sitting and walking and combine posterior correction to the sacrum with anterior lumbosacral junction release and fusion [32].

However, the patient as an entity must be regarded, and this includes both the underlying disease and the flexibility of the deformity. We were able to show in our study that taking these aspects into account, an instrumentation to the lumbar spine is possible in selected patients without having to compromise the sitting balance and the correction.

The traditional recommendation was to fuse from the upper thoracic spine to the pelvis in non-ambulatory patients [3, 5, 6]. In former days correction was achieved mainly by Luque wires and hooks and the pelvis was needed as a base for correction. Those who stayed shorter than the pelvis advocated for very early surgery, as smaller and flexible curves could balance over the pelvis even with shorter fusion [1, 27]. We now showed that this is not limited to early stages and small curvatures of < 30° but is possible even in later surgery if the spine can be centred over the pelvis.

Former publications up to the beginning of this century recommend the unit rod or Galveston technique for NMS [22]. Pedicle screw instrumentation was used in the last decade, but long surgical time and high complication rates were described [2, 4, 27, 28, 33‐35]. With growing experience operative time and complications decreased so these recommendations need a re-evaluation. In our group of patients with lowered muscle tension (mostly DMD or SMA) the segmental pedicle screw constructs down to L5 took an average of 202 min, ranging from 139 to 315 min. The procedure was prolonged by extension to the pelvis using mostly S2AI screws to 258 ± 39 min. This is longer than for most AIS surgeries which we contribute mainly to the more complex preparation in NMS patients. Especially in DMD the thoracic spine is sometimes hard to reach as it is superimposed by the ribs on both sides. But surgical time and blood loss was lower in our all-screw constructs than in earlier reports of screws in NMS. Stiehl et al. (who operated mainly on CP patients) reported an operative time of 344 ± 99 and blood loss of 1503 ml while Modi et al. who reported only on flaccid NMS in DMD and SMA, had 358 min cut-suture-time and 3003 ± 1634 ml blood loss, respectively [28]. The higher blood loss in DMD is typical due to the absence of dystrophin in in arterial smooth muscle and the resulting decrease in contractile function [36]. Bleeding often starts when osteotomies are done or pedicles are opened. Therefore, a quick and subtle technique is necessary. Although our surgical time was quicker, and similar to those reported with Luque wires, [1] we still had an estimated blood loss of 1405 ± 735 ml. Stopping at L5 reduces surgical time and blood loss but should not impair the outcome.

If ilium fixation is not used, a good pre-selection of the patients is necessary considering the flexibility and the expected correction of the main curve and PO. The underlying disease should also be considered, especially whether it is a disease with rather atonic flaccid paralysis (Duchenne muscular dystrophy, SMA, etc.) or with spasticity (cerebral palsy). In our experience, instrumentation can usually be shorter for degenerative muscle diseases or SMA than for spastic paralyses. In our follow-ups 1 and 2 years after the procedure so far, we did not observe any significant deterioration in sitting balance or pelvic obliquity with the exception of one CP patient which had to be fused to the pelvis due to screw loosening in L5.

Unlike idiopathic scoliosis, where the curvature is the result of a mostly pubertal growth disorder, in NMS it is the secondary result of the underlying disease. Patients with SMA or DMD are unable to stabilize themselves. Due to low muscle tone, the spine collapses and the patient tilts to the side. However, the curvature is often flexible, especially if surgery is not performed too late. In these cases, by correction of NMS the spine centres itself over the pelvis and the pelvic obliquity corrects spontaneously. This can be simulated by preop traction radiographs. In 33 of the 42 patients who had a severe pelvic obliquity while sitting, we were able to centre the patient over the pelvis by straightening the spine without including the pelvis in the fusion.

Patients instrumented to the lumbar spine had a mean postop pelvic obliquity of only 8°. This was independent of the preoperative pelvic obliquity in sitting position which was 8° in the L5 subgroup and 19° in the “new lumbar” subgroup. Other authors, which used pelvic fixation on a regular basis reported on similar postop obliquity but longer surgical time [28]. Likewise we did not find a significant difference in the two subgroups in postoperative L4 take-off and trunk shift, which indicates that the preop X-ray in sitting position is a poor prognostic factor for postop balance and pelvic obliquity in patients with flaccid NMS. In this population curvature and pelvic obliquity are a result of the collapse of the spine and not so much of the spine shape itself if surgery is not done too late.

This is different in CP with strong muscular tone and spasticity. In these patients the spine is pulled into a curve by the muscle contractions. Since the muscles are attached to the pelvis and lower limbs, straightening of the spine can increase the pelvic obliquity. Therefore, in these patients, the pelvis must usually be included in the fusion. In our collective their curvatures were more pronounced (89°) and more rigid. Only in a few cases with little and well-treated spasticity instrumentation to L5 was possible. However, of the 54 cases we fused to L5, one had to be lengthened to the pelvis within 2 years due to screw loosening and to improve sitting ability. A big problem with CP patients is that they are often operated on rather late, although it is known that the scoliosis cannot be influenced by conservative measures such as bracing or physiotherapy. This delay causes an increased surgery time. In addition, these patients often have multiple concomitant diseases and a poor nutritional status, which prohibits overly extensive osteotomies. Together with the high rigidity, this explains the poorer correction rate and outcome in these deformities.

Including the pelvis does not only affect surgical time and blood loss but may lead to other complications such as screw impingement, pressure sores, painful hardware, pseudarthrosis of the lumbosacral junction and “windshield wipering” as well as deep wound infections caused by a dissection of muscles reaching down in the diaper area of the patient [3, 23, 28, 33]. Complication rates of up to 59% [37] and even 79% [23] were reported. Therefore, it is paramount to select those patients who really need pelvic fixation surgery carefully.

Some authors advocate that instrumentation to L5 was safe in flaccid NMS [15, 38]. Takaso et al. reported on a spontaneous correction of PO from 13 to 5° in a small series of 20 patients with DMD [14]. But these recommendations are merely based on ap sitting x-rays and they only use this technique for mild curves with apex above L2 and L5 tilt < 15° [38] and low PO [15]. In their small series of 16 patients with SMA Takei et al. could not determine a cut-off value for Cobb angle or PO in which instrumentation to L5 would be effective [15].

Nielsen et al. compared primary and revision surgery to the pelvis and found a similar correction of PO in secondary revision but a higher complication rate [11]. They state that for these patients with borderline PO the benefits of improved lumbosacral mobility may outweigh the risk of future revision surgery, for those which might progress they still recommend fusion to the pelvis. But they cannot define where this border might be [11]. We think that our work can help in defining those patients by taking flexibility, muscle tone and the primary disease into account.

Mubarak et al. had similar PO in fusion to the pelvis or L5 in a small group of patients with DMD with only 29° Cobb angle and thus recommend early surgery [1], as well as Sengupta et al. who fused 19 patients to L5 with an average cobb angle of 19.8° and significantly younger than those instrumented to the pelvis [27]. Mehta et al. found a spontaneous correction of PO from 10.5 to 8.5° when pelvic fixation was not performed and from 25.8° to 11° with pelvic fusion. 15° was used as cut off point for their decision [13]. All available studies report on a relatively small patient population [8, 15, 17] or included patients from different surgeons [27].

The strength of our study is the relatively high number of patients operated on by a single surgeon in a short period compared to other papers. Furthermore, we had a preoperative plan to when to include the pelvis or not and could evaluate this in a prospective manner rather than evaluating ex-post on which group of patients excluding the pelvis worked out or not. Third we distinguished between patients with spasticity and flaccid NMS which we did not see in other papers. Either only DMD patients were considered, or all NMS were seen as a uniform group but the surgical considerations were not based on the primary disease.

The weakness is the short follow-up with a minimum of 2 years and as always larger numbers of patients would be of advantage. Nevertheless, we will progress on our way of selecting fusion levels according to primary disease and flexibility and will hopefully be able to report on long-time results and bigger populations soon.

We showed that if a good correction of the scoliosis in patients with flaccid paralysis was achieved this results in a stable sitting position with a centred spine over a levelled pelvis. This can be simulated with preop traction films. Patients with spasticity or insufficient correction of curve and pelvic obliquity on traction films will still benefit from fusion to the pelvis. But a static x-ray in sitting or prone position is not sufficient to decide on this matter. The underlying disease, flexibility and a clinical examination including subpelvic reasons for obliquity must always be included (Table 5).