Introduction

The treatment of Lenke 1C curves has long been controversial. Historically only 62% were treated with selective thoracic fusion [1]. Nonselective fusion resulted in increased blood loss, longer operative times, decreased range of motion and increased rates of distal disc degeneration [2,3,4]. The principle of selective thoracic fusion was advanced by Lenke who proposed this was best performed in patients with thoracic to lumbar Cobb angle, Apical Vertebral Rotation (AVR) and Apical Vertebral Translation (AVT) ratios of greater than 1.2 [5]. These criteria were further refined by Newton et al., who initially proposed limited thoracic curve correction so that the thoracic and thoracolumbar/lumbar curves could balance each other and minimize the risk of coronal decompensation [6]. It was subsequently shown that the uninstrumented curve corrected to match the instrumented curve and thus maximal correction of the thoracic curve is now sought to maximize the correction of the compensatory thoracolumbar/lumbar curve [7].

Just as fusion techniques have significantly evolved for Lenke 1C curves, so have non-fusion techniques. Anterior vertebral body tethering (AVBT) is a new and emerging technology which early results have shown may result in motion preservation [8]. As for any new treatment approach, the outcomes must be compared to the existing techniques, in this case selective thoracic fusion. Early reports for AVBT describe a high revision rate, which ranges from 12.3 to 60% [9,10,11,12], suggesting that the learning curve and best application for this technology are still being refined [13, 14]. Appropriate patient selection continues to be controversial. Several studies have attempted to delineate optimal skeletal maturity with Shin et al. showing skeletal immaturity was the number one reason for AVBT revision due to overcorrection, while Newton et al. showed that advanced skeletal maturity was associated with higher rates of revision to a fusion. The concept importance of sufficient remaining growth was recently reinforced by Alanay et al. who reported on the variable rates of growth and remodeling in AVBT associated with different SMS scores [9, 12, 15]. However, skeletal maturity is only one of many factors that will help define the ideal patient selection. To date, the impact of different lumbar curve modifiers on the outcomes of thoracic AVBT has not been examined. In this study, we evaluated the impact of AVBT on the instrumented thoracic and uninstrumented thoracolumbar/lumbar curves as well as on their C7, apical and LIV coronal alignment based in Lenke 1A vs Lenke 1C curve types.

Methods

This was a retrospective study of a prospective multi-center registry of idiopathic scoliosis patients with Lenke 1A or 1C curvatures that underwent selective thoracic AVBT between 2013 and 2018. Institutional review board (IRB) approval was obtained at each site. We included all patients who were Risser stage 0–3 and/or SMS 2–5 and in whom the LIV of L1 or higher, with a minimum 2-year follow-up. We specifically excluded Lenke 1B curves as we wanted to highlight potential differences in the correction or lack of correction of the uninstrumented thoracolumbar/lumbar curve.

Preoperative demographic data included age, sex, ethnicity, etiology, prior treatment and body mass index (BMI). Digital radiographic software was used to assess major and minor Cobb angle, and curve flexibility preoperatively, postoperatively and on subsequent follow-up images. Postoperative evaluation also included suspected presence or absence of tether rupture (defined as an increase in Cobb angle of > 5 degrees between two adjacent vertebral body screws compared with previous radiographs), maximum Cobb angle correction, disc wedging below the LIV and coronal alignment at most recent follow-up [16]. Coronal alignment was assessed by measuring the distance from the CSVL to the midpoint of the LIV, the apical vertebra of each curve and C7. Primary outcomes included the need for secondary surgery and major and minor Cobb angle magnitude. Secondary outcomes included tether breakage, coronal balance and sub-adjacent disc wedging.

Two-way comparisons were made between the Lenke 1A and Lenke 1C groups. Ratios were tested using Fisher’s Exact test. Means were compared using the two-tailed t test or if the variances were unequal a Mann–Whitney U test was used. Statistical significance was set at a p value < 0.05 and clinical significance was set at a difference in Cobb magnitude > 5 degrees. All statistics were done using SPSS.

Results

111 patients were eligible for inclusion, 62 patients met the inclusion criteria, of whom 43 were Lenke 1A and 19 were Lenke 1C, and 49 were excluded due to incomplete data or inadequate follow-up. Complete demographics are listed in Table 1. There was no difference in the average follow-up between the groups with an average of 3.0 years in the 1A group and 3.0 years in the 1C group. Neither was there a difference in the age at index procedure of 13.2 years old (9.8–17.2) in the 1A group vs 13.7 (11.7–16.0) in the 1 C group) or the rate of revision surgery (one patient in each group with an insufficient correction that was revised to a fusion).

Table 1 Demographics/outcomes for Lenke 1A and Lenke 1C curves
Full size table

Preoperative measures

There was no difference in the preoperative thoracic Cobb angle between the two groups, with an average of 48° (32–79) in the Lenke 1A and 50° (39–60) in the Lenke 1C patients (p = 0.446). However, there was a difference in the preoperative thoracolumbar/lumbar curve magnitude with the Lenke 1A curves averaging 31° (22–44) and the Lenke 1C 40° (30–53) (p < 0.00001). Regarding curve flexibility, the Lenke 1C curves showed increased thoracic curve percent correction on flexibility films 42% (1–77%) vs 55% (29–90) (p = 0.0172). Whereas there was no difference in the thoracolumbar/lumbar percent correction on flexibility films in 65% (11–146) vs 61% (26–97) (p = 0.4902).

Postoperative curve measurements

There was no difference in the first erect thoracic curve magnitude, with the Lenke 1A measuring 27° (10–48) vs the Lenke 1C 30° (16–51) p = 0.193. There was, however, a difference in the first erect thoracolumbar/lumbar Cobb magnitude with the Lenke 1A averaging 19° (4–32) vs Lenke 1C 26° (14–45) p < 0.005.

There was no difference in the maximal Cobb angle correction prior to suspected tether rupture or at the most recent follow-up (in patients with no identifiable tether rupture), with an average thoracic curve magnitude in the Lenke 1A group of 19° (− 34 to 53) and the Lenke 1C of 22° (5–50) p = 0.459. There was a suspected rupture in half of all patients (31/62), but no difference in the rate of rupture between the two groups (20/43 Lenke 1A and 11/19 Lenke 1C patients p = 0.40856). Post-rupture and at the most recent follow-up there was no difference in thoracic Cobb angle in the Lenke 1A vs the Lenke 1C groups [24° (− 34 to 65) vs 29° (5–50) p = 0.322] at the most recent follow-up. There was also no difference in the loss of correction in the thoracic curve after tether rupture in the Lenke 1A vs the Lenke 1C [5° (− 2 to 34) vs 7° (0–19) p = 0.738] or disc angulation below the LIV, [3° (0–9) vs 3° (0–8) p = 0.840] at most recent follow-up. Furthermore, there was no difference in the percent curve corrections from preop to the most recent follow-up for the thoracic curve or the thoracolumbar/lumbar curve, with a percent curve correction in the thoracic curve for the Lenke 1A group of 52% (− 19 to 174) vs the Lenke 1C group of 44% (14–88) (p = 0.45326). Similarly, there was no difference in the thoracolumbar/lumbar curve percent curve correction for the Lenke 1A group at the final follow-up of 41% (− 10 to 96) vs the Lenke 1C group of 32% (− 9 to 76) (p = 0.10524). Complete data is shown in Table 2.

Table 2 Pre-operative, post-operative and most recent follow-up Cobb angle measurements
Full size table

Coronal alignment

As seen in Table 3, there was no difference in the coronal alignment at C7, with the Lenke 1A deviating 8 mm (1–26) from the CSVL vs 2 mm for Lenke 1C (− 23 to 27) (p = 0.057), nor in the thoracolumbar/lumbar apical vertebral translation from the CSVL in Lenke 1A 20 mm (− 16 to 58) vs Lenke 1C 14 mm (− 16 to 56) p = 0.272 at most recent follow-up. However, there was a difference in alignment at the LIV, with less deviation from the midline in the Lenke 1C group [7 mm (− 13 to 31)] vs the Lenke 1A group [12 mm (− 13 to 47)] (p = 0.0355). Figure 1 demonstrates an example of two patients with thoracolumbar/lumbar curves of less than 35 degrees, one where the LIV is well aligned and one where the LIV is poorly aligned.

Table 3 Coronal measurements at latest follow-up
Full size table
Fig. 1
figure 1

Two curves with Thoracic Scoliosis < 35 at 4.4 years and 3.5 years follow-up, respectively. Patient on the left preop and most recent follow-up with alignment < 1.5 cm at the curve apex and LIV. Pt on the right with alignment > 1.5 cm at curve apex and LIV

Full size image

Outcomes at a most recent follow-up

At the most recent follow-up there was no difference in the number of patients with good correction, defined as both the thoracic and thoracolumbar/lumbar curves measuring ≤ 35 degrees, (33/43 Lenke 1A and 14/19 Lenke 1C curves, p = 0.80) (Tables 4 and 5 for Summary of all data for Lenke 1A and 1C curves). The Lenke 1A group corrected an additional 7% (− 66 to 118%) in the thoracic and 3% (− 55 to 53%) in the thoracolumbar/lumbar spine from the first upright. The Lenke 1C group corrected an additional 4% (− 16 to 35%) in the thoracic spine and 5% (− 19 to 33%) in the thoracolumbar/lumbar spine. There was a strong correlation between the rate of correction for the instrumented thoracic and the compensatory thoracolumbar curves/lumbar, with the Spearman’s Rho r = 0.6206 p = 0.00001 for the Lenke 1A and r = 0.81713 p = 0 for the Lenke 1C group. Figure 2 shows patients with similar preoperative curves but differing radiographic outcomes at the most recent follow-up.

Table 4 Summary of Lenke 1A data
Full size table
Table 5 Summary of Lenke 1C data
Full size table
Fig. 2
figure 2

A, B 14 yo Risser 0 female on the left with a 58° thoracic and 35° thoracolumbar Lenke 1AR curve PA and lateral preoperative radiographs, C first erect PA 22° thoracic and a 24° thoracic and D, E PA and lateral radiographs showing a 21° thoracolumbar at > 3 years follow-up. F, G another 14yo Risser 0 female with a 52° thoracic and 29° thoracolumbar Lenke 1AR preoperative PA and lateral radiographs, H first erect PA radiograph 41° thoracic and I, J PA and lateral radiographs showing 57° thoracic and 33° thoracolumbar at > 3 years follow-up

Full size image

Finally, there was no difference in the rate of revision surgery, with one patient in each group treated with fusion surgery (p = 0.546).

Discussion

This is the first study to look at outcomes of AVBT based on curve type. In this study, we found an equivalent correction of the thoracic curves in both lenke 1A and 1C curves, but less correction of the uninstrumented thoracolumbar/lumbar curves in the Lenke 1C patients. With an average correction of only 25% in the uninstrumented Lenke 1C thoracolumbar/lumbar curves it would be easy to suggest that these patients are not good candidates for AVBT, but there was considerable overlap in these results with curve correction ranging from − 9 to 76%. Additionally, much like the reported results for selective fusion in 1C curves, we found that the correction of the uninstrumented thoracolumbar/lumbar curve correlated with the correction of the instrumented thoracic curve [7]. This would suggest that for ABVT, maximal correction of the instrumented thoracic curve will result in maximal correction of the thoracolumbar/lumbar curve as well.

The challenge is that unlike a fusion where the correction of the instrumented curve is fixed, AVBT can result in increasing correction or overcorrection with continued growth. In this study, we showed that the instrumented and uninstrumented curves corrected an additional 3–7% after their first erect radiograph depending on the curve location and type, with wide variation from − 66 to 118%. Given the additional correction these patients may undergo, the concerns for overcorrection and under correction shown in Shin et al. and Newton et al.’s studies [12, 16] appear validated. Interestingly, despite the variable correction rates, both curve types in our study had similar rates of revision surgery and no patients who underwent revision surgery due to overcorrection. We did have one patient in each group undergo revision surgery and in both cases showed initial correction of 39% for the Lenke 1A patient and 12% for the Lenke 1C patient and neither patient growth modulated. We suspect that these numbers will increase with increased follow-up as more recent studies on AVBT have suggested increased revision rates with longer-term follow-up. Particularly because we noted that 50% of all patients had an apparent tether rupture with an average loss of correction of 5° though this had a large range from − 2 to 34°, and thus this number will potentially increase with time.

This is also the first study to look at the coronal alignment outcomes in patients after AVBT. The concept of the Cone of Economy (COE) has been used to define balance, with unbalanced curves or those deviating > 1.5–4 cm from the CSVL experiencing an increased risk of trunk shift, adding on, pain and increased energy expenditure [17,18,19]. In addition to the correction of both the instrumented and uninstumented curves to less than 35° we found that 77% of patients in each group showed coronal deviation less than 2 cm of the CSVL, for both C7 and the apical vertebra.

Rates of adding on after selective thoracic fusion varies from 6 to 21% in the literature, with increased rates seen in patients with 1AR subtypes that were hypokyphotic [20, 21]. We anticipate that future studies will include an analysis of these subtypes in the results of selective AVBT.

Due to the retrospective registry nature of this study there are several limitations. Including the fact that there was a large number of patients with incomplete data and due to the Covid-19 pandemic we were unable to obtain that additional data. Leading to the exclusion of 44% of patients that could be potentially included in this study. We also acknowledge that some surgeons would elect to tether the thoracolumbar curve in Lenke 1c curves, but as our goal was to study the response of the uninstrumented curve we elected to not include those patients. Furthermore, as much of the data was collected prior to the routine use of SMS, Risser stage was also included in the analysis. Additionally, we did not have complete patient height data, nor were all the images calibrated, so we were unable to evaluate changes in height clinically or radiographically. We were also not powered to detect a difference in the rates of correction or adding on based on LIV. Finally, as we relied on a change of more than 5° between adjacent levels to assess for probable tether rupture, and it is possible we are underestimating the true incidence of rupture in these patients.

At present, AVBT has a high rate of revision surgery compared to selective thoracic fusion and remains in the experimental stage both from a technical and patient selection standpoint. As we elucidate the appropriate indications and intraoperative techniques for AVBT we will be able to better predict the outcomes for skeletally immature patients with thoracic and thoracolumbar/lumbar scoliosis. Meanwhile, judicious patient selection, extensive patient counseling and vigorous outcomes analysis are essential for its application.