Although generally accepted that the younger the child, the more potential for significant progression of the deformity, the variety of combinations of congenital deformities seems limitless and hence the prediction about what will happen with growth remains difficult. Nonetheless, expert opinion has recommended early fusion with the assumption that “it is safe to say that it is easier to prevent a deformity than to correct a deformity” [7].

To date there remains a paucity of strong evidence on which recommendations for early in situ fusion for congenital scoliosis can be made. Of the available studies, the majority of the literature remains retrospective in nature with limited follow-up. McMaster’s review of a single type of congenital scoliosis (unilateral bar and contralateral hemivertebrae) described significant variability in the natural history [4]. Nine patients who had an upper thoracic scoliosis had variable rates of progression: two reached curves of 70° and 80° at skeletal maturity while 5 others were ≤50°. One 4 year-old was managed by immediate arthrodesis because of a curve of 48°, yet a 5 year old with a curve of 45° was observed and had a curve of 51° by age 10. Similarly, substantial variability in timing of arthrodesis, surgical approach, number of segments fused, as well as magnitude of pre-operative curve makes it difficult to support the conclusions that early surgery, preferably in the first year of life, is warranted.

Goldberg et al. [8] reported their experience of in situ fusion in the setting of congenital scoliosis and included a retrospective review of 43 patients who were at least 15 years of age at final follow-up. They noted that although the localized fusion was effective in preventing progression of the Cobb angle of the congenitally malformed area, it did not control the overall deformity that developed and progressed with growth. Their case series noted a 25.6 % reoperation rate due to continued progression of the deformity. The type of congenital scoliosis was heterogeneous in this study.

Subsequently Vitale and colleagues [9] reported clinical and radiographic outcomes, also in a retrospective manner on 21 patients. In this study, 5 of the 21 patients had multiple procedures and the authors also reported pulmonary function data as well as Child Health Quality of Life Outcome (QOL) questionnaires. The group compared the results with healthy children and concluded that the cohort of congenital scoliosis patients with early fusions had worse pulmonary function tests (PFTs) and QOL scores compared to healthy peers, cautioning against early fusion in this patient population. The comparative group, however, was that of healthy peers which limits the interpretation of their data, as patients with congenital scoliosis often have other comorbidities. The authors did not report on pre-operative PFTs or QOL data for comparison.

A more recent retrospective case-control study reported on outcomes of instrumented versus non-instrumented spinal fusion in a total of 51 patients [10]. Although the instrumented group had better post-operative curve correction, continued and marked curve progression was noted in both groups with a 25 % reoperation rate (both groups). The authors questioned whether early fusion for congenital scoliosis was meeting the goal of progression prevention.

Karol et al.’s review of in situ fusion in early onset scoliosis although having mixed etiologies, 20 of the reported on 28 patients had congenital scoliosis [11]. They had a 39.3 % revision rate. The authors cautioned against early fusion as it put patients at risk of restrictive pulmonary disease, especially those with proximal thoracic deformity who were fused over more than four segments. Again, pre-operative pulmonary data on these patients were lacking, making it difficult to support their conclusion that early fusions have a significant direct negative effect on pulmonary function in the long-term.

Both earlier and the more recent studies remain retrospective cross-sectional case series with no longitudinal or prospective component and so remain as weak evidence only supporting/refuting early fusion as a treatment for congenital scoliosis.

Convex growth arrest as a treatment for congenital scoliosis has been previously described as effective by a number of authors [12–14]. It has been described for patients with hemivertebrae but others have also reported on its outcomes for failure of segmentation defects and mixed anomalies. Demirkiran et al. [15] most recently described their experience with this technique in a series of 13 patients (mean operative age, 5.4 years; mean follow-up, 4.7 years). The authors’ indication for surgery was a long sweeping curve, including >4–5 segments, and not suitable for single hemivertebrectomy. Although they report promising results with a mean curve correction of 33.5°± 12.4° at final follow-up, from an average preoperative Cobb angle of 49°± 10.9°, none of these patients were evaluated to skeletal maturity. In addition, the authors note that in nine patients the curve improved, in three patients there was no change, and in one patient there was curve progression, illustrating that the procedure may have either an “epiphysiodesis effect” (what it is intended for) or a “fusion effect” where the curve may not improve but also will not progress. They admit that, at present, with the absence of any reliable technique to evaluate the growth potential of the vertebral apophysis, it is impossible to predict the outcome of this procedure. Therefore which patients would be the most appropriate candidates for an instrumented convex hemiepiphysiodesis remains unclear.

In contrast, Thompson et al. [16] reviewed retrospectively their case series of 30 patients (mean age, 6.3 years), 63 % of whom reached skeletal maturity. They noted an improvement in Cobb angle in 23 patients, progression arrested or slowed in 5 patients, and progression in 2 patients. Their technique involved a combined anterior/posterior approach. Marks et al. [17] examined a further 53 patients with a mean follow-up of 8.8 years. Thirty-four patients were skeletally mature. This series reported on the effects of convex hemiepiphysiodesis on a variety of types of congenital anomaly. Those with failure of segmentation defects, as well as complex anomalies, continued to progress despite surgery, with a final Cobb angle increase from a mean of 61°–70°. In contrast, 97 % of hemivertebrae patients had reversal in their curves or slowed progression with an average preoperative Cobb angle of 41° improving to 35° at final follow-up. The authors felt that a younger age at surgery and a hemivertebrae located in the lumbar spine yielded the best results.

Keller et al. [18] reported their experience in 16 patients (mean operative age, 4.8 years; mean follow-up, 4.8 years). Their series had a variety of formation and segmentation anomalies and they found 37 % of curves improved, 42 % were unchanged or progressed less than 7°, 16 % progressed 10°–15°, and 5 % progressed greater than 15°. The best results were in patients with a hemivertebrae, whereas progression despite surgery was seen in patients with an unsegmented bar and a contralateral hemivertebrae. The authors felt that their results were most predictable at producing an arrest of the deformity and the epiphysiodesis effect to be less predictable. The authors conclude that it remains unknown what the final outcome will be, as these patients were not followed to skeletal maturity.

The remaining retrospective reviews pooled data shows improvement in curves in 48 % of cases (range 20–77 %), no change (fusion effect) in 40 % (range 17–70%), and curve progression in 12 % (range 0–21 %) with conclusions that earlier intervention provides the best results [12–14, 19, 20]. Given that all these studies are retrospective single center case series without the benefit of comparative controls, the effect of age as well as type of congenital anomaly on the results of surgical intervention cannot truly be determined and the small number of patients in each study limit broad conclusions to be made. Therefore, there is weak evidence at best to suggest that early hemiepiphysiodesis (before the age of 5 years) is most appropriate for congenital scoliosis.

Hemivertebra excision, initially described as via a combination of anterior and posterior approaches, has recently been refined to an all-posterior approach. Most studies report on the technique and feasibility of hemivertebra excision with expert opinion on the advantages and disadvantages of either a combination anterior/posterior approach or an all-posterior approach [21–24]. Studies reporting on hemivertebra excision in young children have attempted to report on children younger than 10 years of age with a substantial variation in timing of surgery. Crostelli et al.’s 15 cases had a mean age of 5.5 years, with a range from 2 to 9.5 years [25]. Although they reported a percentage curve correction of 72.5 % (mean preoperative curve of 44°) and no major complications, the average follow-up was extremely limited (3.3 years). Chang and colleagues [26] performed a retrospective review of 18 patients less than 10 years of age (mean 6.6 years, range 2.6–9.8 years); they noted an average preoperative curve of 34.4° correcting to 8.4° (75 % correction) immediately postoperatively. This was an average of 12.9° at final follow-up (62.5 % correction). Their series had a mean follow-up of 11.4 years and also reported no complications from surgery.

Chang et al. [27] also reported their results of hemivertebra resection and the effect of age on outcomes. They arbitrarily assigned patients into two groups (nine in each group) based on age at time of surgery: those under 6 years of age (mean age 4.2 years) and those between 6 and 10 years of age (mean age 9 years). The authors hypothesized that those that had surgery before the age of six had better deformity correction with no impact on vertebral or spinal growth. The younger patients had, on average, a preoperative curve magnitude of 32.4° improving to 9.1° at final follow-up, compared to the older age group with a preoperative curve of 36.5° improving to 14.5°. As a retrospective review of patients treated at a single center, without the benefit of true controls, the effect of age on the results of surgical intervention cannot truly be determined. It seems more interesting that although the average difference in age between the groups was nearly 5 years, the mean difference in preoperative Cobb angle was only 4.1°, suggesting that there was not a dramatic risk of progression during this time; therefore a delay in treatment until the child is older could be equally considered.

Bollini and colleagues reported homogeneous cohorts of thoracolumbar and lumbar hemivertebrae resections, in two separate studies [22, 23]. The studies had equivalent mean age at time of surgery (3.5 years for the report on thoracolumbar hemivertebrae resections; 3.3 years for the lumbar resections). The average follow-up was 6 and 8.6 years, respectively. In the thoracolumbar resections, there was a mean improvement of 69.3 % in the segmental curve, from an average 34.8° before surgery to 10.7° at final follow-up. Seven of the thirty-four patients had immediate postoperative complications and late complications occurred in 12 patients (34 %), which included pseudarthrosis in 5 patients and progression of curve in 6 patients. Of the 34 patients, 18 required additional surgery during the follow-up period.

In their report of lumbar hemivertebrae resections, Bollini et al. [23] noted a more significant improvement in the segmental scoliosis curve of 71.4 % with a mean curve of 32.9° preoperatively, improving to 9.4° at final follow-up. Of the 21 patients in this series 3 experienced postoperative complications, which included radiculopathy, wound infection, and acute renal insufficiency. The authors concluded in both studies that hemivertebra excision should be performed as early as possible.

Similarly, retrospective case series from Wang et al. [28] and Ruf et al. [29] conclude that early hemivertebra resection surgery should be considered to prevent the development of severe local deformities and secondary structural curves. The mean preoperative curve in Wang et al.’s study was 36.6°, which improved to 5.1° [28]. Ruf et al. [29] had a preoperative Cobb angle of 36° in patients without a bar formation that improved to 7° following surgery; those with a bar formation had a preoperative curve of 69° improving on average to 23°. The average age was 4.9 and 3.4 years, respectively.

Nakamura et al. [30] reported on five patients over an average 12.8-year follow-up period for hemivertebra resection. Their illustrative case report on one patient described the age at operation to be 13 years 7 months while another case had surgery at the age of 3 years. Despite satisfactory long-term results reported by the authors, definitive conclusions with a sample size of 5 cannot be made. Zang et al. [31] retrospectively reviewed 58 hemivertebra resections in 56 patients with a significant range in age at the time of surgery (mean age 9.9 years with range 1.5–17 years). There was a mean improvement of 72.9 % in this study in the Cobb angle at final follow-up. The mean follow-up was 37.9 months (range 24–58 months). The authors did not analyze the results as a function of age yet conclude that hemivertebra resection is ideal for very young children.