Introduction

Vertebral body tethering (VBT) is a new, non-fusion surgical technique aimed at correcting severe scoliotic deformities while simultaneously avoiding many of the postoperative complications associated with PSF. This technique promotes growth modulation of vertebrae [1] and has been proven to be effective in creating spinal deformities in animal models [2, 3]. There are several studies that have reported that this technique is both safe and effective in treating spinal deformities in children [4,5,6] while maintaining postoperative spinal motion and flexibility [7] as well as intervertebral disk and facet joint health [8].

Though numerous published series report successful postoperative outcomes in patients treated with VBT, five to fourteen percent [4,5,6, 9] of patients exhibit significant progression of their deformity requiring conversion to a posterior spinal fusion (PSF). Yet, little has been published on the operative characteristics associated with PSF after VBT. We aimed to assess various perioperative characteristics in this unique patient population with a matched cohort of patients treated only with a PSF. We hypothesized that: (1) we would observe no difference in estimated blood loss, operative time, or postoperative length of stay in patients that receive a PSF following a failed VBT (i.e., PSF–VBT) when compared to those who received only a PSF (i.e., PSF-Only), and (2) we would observe a difference in the type of instrumentation utilized, implant density, and the number of levels fused in patients treated with PSF–VBT than when compared to PSF-Only.

Materials and methods

The procedures outlined below were approved by the Institutional Review Board for each participating Medical Center. We retrospectively assessed postoperative outcomes from three hundred and one patients enrolled in the Harms Non-Fusion Registry, all of whom received VBT. Of those, twenty-two (7.3%; 22 of 301) met the inclusion criteria for this study as they received a PSF following a failed VBT. At the time of data collection and analysis, these twenty-two patients were the only patients to receive a PSF following a failed VBT in the entire Harms Non-Fusion Registry. All preoperative data were derived from the assessment just prior to PSF in PSF–VBT. We also identified twenty-two patients with matching preoperative characteristics from more than fourteen hundred patients listed in the Setting Scoliosis Straight Registry who received a PSF no earlier than 2010. None of the primary perioperative variables for these patients were available to the authors until after they had been selected as an ideal match based on the following characteristics: (1) main thoracic and lumbar deformity measures (i.e., Cobb angle), (2) levels instrumented, and (3) chronological age at surgery (Table 1). Possible matches were identified in a stepwise fashion based on the order of the characteristics listed above. For all patients, we recorded the following perioperative characteristics: (1) operative time (minutes), (2) estimated blood loss (ccs), and (3) postoperative length of stay (days). We assessed each postoperative standing spine radiograph to determine instrumentation type (i.e., all-screw vs. hybrid constructs) and implant density using the standard formula [10] (i.e., total number of vertebrae instrumented divided by the total number of implants utilized; maximum of two). As a secondary clinical aim, we also asked each site-surgeon which levels would have been fused in initial preoperative PSF–VBT assessments if the index procedure had been a PSF. These theoretical levels were compared to the actual number of levels fused following failed VBT. Because this was a retrospective multi-center analysis, the method used to determine theoretical fusion levels varied, but likely included standard preoperative assessment methods (e.g., Lenke Classification, last touched vertebrae, etc.).

Table 1 Relevant preoperative and perioperative characteristics by matched pair for PSF–VBT and PSF-only
Full size table

All statistical analyses were performed in RStudio. The mean and standard deviation of all continuous variables were calculated separately for each cohort. To validate that the cohorts were properly matched, we performed a two-sided paired t test for all three matching characteristics. To determine if there was a difference in any perioperative numerical value between VBT, PSF–VBT, and PSF-Only, we performed either a two-sided paired t test or a two-sided unpaired t test with unequal variance. The threshold for statistical significance was set at p < 0.05.

Results

Preoperative Characteristics: The PSF–VBT and PSF-Only patients included in statistical analyses were well-matched with no significant differences noted in any of the following preoperative characteristics: (1) chronological age at surgery (p = 0.282), (2) main thoracic deformity size (p = 0.651), and (3) lumbar deformity size (p = 0.164; Fig. 1; Table 2). The preoperative flexibility of main thoracic and lumbar deformities, calculated as the absolute deformity measure obtained from bending and traction radiography, was similar for PSF–VBT (main: 34.9° ± 20.1°; lumbar: 20.1° ± 16.7° [n = 7]) and PSF-Only (main: 29.8° ± 14.1°; lumbar: 12.5° ± 12.2° [n = 22]; Table 2). The number and location of vertebrae that were fused were identical in PSF–VBT and PSF-Only except in three patients where the difference was no more than three levels when compared to their matched peer (T5–L3 vs. T5–L4; T7–L2 vs. T7–L1; and T1–T8 vs. T2–T10; Table 1). The principal reason listed for failed VBT in PSF–VBT was progression of the primary deformity (36%; 8 of 22), suspected broken tether and/or implant failure (27%; 6 of 22), progression of the compensatory deformity (18%; 4 of 22), dissatisfaction with postoperative appearance (9%; 2 of 22), distal adding-on (5%; 1 of 22), and overcorrection (5%; 1 of 22). Additional information can be found in Supplementary Table 1.

Fig. 1
figure 1

Matching characteristics: example patients from each cohort as well as the mean ± standard deviation of the continuous variables used for matching; the P values were calculated using a two-sided paired t test. No significant differences were observed in chronological age (yrs.), pre-PSF thoracic Cobb angle (°), or pre-PSF lumbar Cobb angle (°)

Full size image
Table 2 Preoperative characteristics for index VBT, PSF–VBT, and PSF-only
Full size table

Perioperative Characteristics: We observed no difference in operative time (p = 0.752), estimated blood loss (p = 0.065), or postoperative length of stay (p = 0.913) between the two cohorts (Table 3; Fig. 2). Instrumentation type for PSF–VBT included seventeen all-screw constructs (77%) and five hybrid (23%) constructs; conversely, instrumentation type was all-screw (100%) for PSF-Only. Implant density was significantly lower (p = 0.012) in PSF–VBT (1.7 ± 0.3) than when compared to PSF-Only (1.9 ± 0.2). The existing anterior instrumentation remained intact in all but one PSF–VBT. In this patient, the anterior instrumentation was removed due to overcorrection. The compensatory deformity in this patient decompensated following removal of the anterior instrumentation, resulting in a PSF.

Table 3 Perioperative characteristics
Full size table
Fig. 2
figure 2

Perioperative characteristics: box plots representing the mean, first quartile, and third quartile for operative time (OT: A), estimated blood loss (EBL: B), and length of stay (LOS: C) for PSF–VBT (green) and PSF-Only (blue). We observed no difference in OT, EBL, or LOS between the two cohorts. [Note: One matched pair of patients were removed from OT and LOS due to missing data; they were not the same patients.]

Full size image

Next, we aimed to determine if the number of posterior instrumented levels (theoretical vs. actual) differed in patients that failed VBT than when compared to the levels that would have been instrumented had the index procedure been a PSF (Table 4). Of the twenty-two PSF–VBT patients included, eleven (11 of 22; 50%) received a PSF that was 2.2 ± 1.4 (minimum = 1; maximum = 5) levels longer, on average, than if they had received a PSF at the index procedure. In eight of the eleven (73%) patients where this occurred, an average of 2.1 ± 1.1 levels (minimum = 1; maximum = 4) were added to the distal end of the construct (Fig. 3). Distal extension occurred most often in patients that exhibited adding-on (1 of 1), followed by progression of the primary deformity (4 of 7), a suspected broken tether (2 of 6), and progression of the compensatory curve (1 of 4).

Table 4 Instrumentation type, implant density, and levels added
Full size table
Fig. 3
figure 3

Fusion extended distally in three example PSF–VBT patients; the images included represent the: (1) preoperative assessment, (2) first erect assessment following VBT, (3) pre-revision assessment after VBT and just prior to PSF, and (4) postoperative assessment following conversion to PSF from right to left. A–D represent a patient that presented with an 50° main thoracic and 30° thoracolumbar deformity that was instrumented from T6 to L1. The main thoracic and thoracolumbar deformities were reduced to 42°–24°, respectively, at the first erect assessment. Over time, the main thoracic deformity progressed to 62°, which resulted in a PSF from T2-L4. The hypothetical fusion levels for this patient, had they received a PSF instead of a VBT at the time of the index procedure, were T3–L2. The resultant fusion levels were T2-L4. E–H represent a patient that presented with a 65° main thoracic and 40° thoracolumbar deformity that was instrumented from T6–T11. The main thoracic and thoracolumbar deformities were reduced to 49°–40°, respectively, at the first erect assessment. The compensatory deformity continued to progress, resulting in a 61° main thoracic deformity and a 59° thoracolumbar deformity. This patient received a PSF from T2-L4, with hypothetical fusion levels of T3–T12. I–L represent a patient that presented with a 48° main thoracic deformity and a 29° thoracolumbar deformity that was instrumented from T5–T12. At the first erect assessment, the main thoracic and thoracolumbar deformities were reduced to 44°–33°, respectively. The main thoracic and thoracolumbar deformities continued to progress and were 61°–51°, respectively, prior to PSF. The hypothetical fusion levels for this patient were T3-L1, with actual instrumentation being placed from T3-L3

Full size image

Discussion

Vertebral body tethering (VBT) is an effective surgical option for the treatment of progressive adolescent idiopathic scoliosis in many patients [5, 11]; however, up to 15% of patients [5] may require a PSF due to continued deformity progression. The perioperative characteristics of patients that received a PSF following a failed VBT (i.e., PSF–VBT) are essential in defining the postoperative challenges associated with VBT, particularly as these cases become more common. To our knowledge, this is the first study to report such characteristics in a large multi-center cohort. Our data indicate that operative time, estimated blood loss, and postoperative length of stay are similar in both cohorts. However, an average of two additional levels were fused in 50% of patients that received VBT prior to PSF, most of which were added to the distal end of the construct.

The ability to predict intraoperative blood loss in patients that have had a prior VBT is an essential component in adequately preparing to convert to a posterior fusion. The average amount of intraoperative blood loss recorded in the literature for adolescent idiopathic scoliosis patients undergoing fusion is approximately 900 ccs [12, 13], which closely mimics the values obtained in this study for PSF-Only (825.0 ± 513.0 ccs). Although intraoperative blood loss was not significantly lower (P = 0.065) in PSF–VBT (556.0 ± 443.0 ccs), the mean value between the two cohorts differed by 269 ccs, which included one PSF–VBT patient who lost 2,185 ccs. Such a difference may be clinically meaningful in immature adolescent patients.

It may be expected that operative time would increase in patients that had previously received VBT, as posterior instrumentation would need to be carefully placed around existing anterior instrumentation. Our data indicate, however, that performing a PSF following a failed VBT does not increase operative time when matched by levels instrumented. The literature shows that the average operative time for a PSF is dependent on the number of attending physicians present as well as their experience level, with increased physician number and experience leading to shorter operative times. Chan and Kwan [14] found that operative time was reduced by 71.4 min, on average, when surgical instrumentation was performed by two (176.6 ± 27.0) attending physicians as opposed to one (248.0 ± 49.9). In a different multi-center study performed by Cahill et al. [15], operative time was 458 min, on average, for novice attendings and 265 min, on average, for veteran attendings. The mean operative time observed in the PSF–VBT cohort was 302.0 ± 100.0 min. This value is well within the reported range of single (or dual) experienced physicians indicating that performing a PSF after a failed VBT is not significantly more time-consuming than when compared to an index PSF.

The knowledge of immediate postoperative expectations following a PSF are invaluable to both patients and their families when preparing for such a significant surgery, especially as it relates to length of stay. Based on a nationwide study of 23,279 patients, the average length of stay for an adolescent idiopathic scoliosis patient following a PSF is approximately 5.5 days [16], with the number of levels instrumented and wound-related complications significantly influencing values. Postoperative length of stay in two additional studies on more than 26,000 patients exhibited an average of 4.7 ± 2.5 [17] and 5.4 ± 5.0 [18] days. Our data are nearly identical, indicating that the length of time spent in the hospital following PSF is not significantly influenced by a prior VBT. Thus, immediate postoperative expectations for an index PSF are an accurate representation of what a patient and their family should expect when receiving a PSF following a failed VBT.

The bony anatomy available for posterior instrumentation may be limited by small vertebral body size and existing anterior instrumentation. These restrictions have the potential to influence instrumentation type and implant density in PSF–VBT. The majority of PSF–VBT (77%) exhibited all-screw constructs. In the five patients (23%) that exhibited hybrid constructs, only one was all-hook, indicating that many vertebral levels could accommodate at least one pedicle screw. Implant density was lower in patients with a failed VBT than when compared to those that received a PSF for their index procedure. This finding validates the hypothesis that bony anatomy may be limited at several vertebral levels in PSF–VBT. In such cases, surgeons should be prepared to use additional sublaminar bands and/or hooks. The use of fewer implants in PSF–VBT may also account for the slight difference observed in estimated blood loss between the two cohorts.

The number and location of instrumented levels in PSF has a huge impact on postoperative outcomes, including spinal mobility and the development of adjacent segment disease [19, 20]. The potential benefit to non-fusion surgical techniques for the treatment of adolescent idiopathic scoliosis, such as VBT, is that negative effects often associated with a PSF may be mitigated or eliminated. However, when VBT fails and a PSF is indicated, the number of levels required to be instrumented may differ than if a PSF was selected as the index procedure. Our data indicate that 50% (11 of 22) of PSF–VBT in the analyzed cohort ended up with 2.2 ± 1.4 additional levels added to their construct. Most of the additional instrumented levels were added to the distal end of the construct, which has significant implications for a variety of postoperative outcomes. Nohara et al. [21] found that the incidence of disk degeneration increased substantially in patients with increasing distal instrumentation, particularly in the lumbar spine. In our patient population, 40% (6 of 15) received instrumentation at L3 or L4 that would have otherwise not been instrumented if PSF was selected as the index procedure. The literature indicates that up to 25% of patients treated with a selective thoracic fusion may experience distal adding-on, 7% of whom may require surgical revision [22]. The rate of distal construct extension is still significantly lower (7%) than in patients that who received a PSF following a failed VBT (36%: 8 of 22). The rate of distal extension observed here may be biased as flexibility films prior to PSF were not available for all patients; this missing data may have influenced the ability of a site-surgeon to determine if a distal level could have been ‘saved.’ These data may also reflect patient and/or surgeon anxiety associated with the risk of requiring an additional revision procedure should the initial fusion construct not be sufficient to adequately correct the remaining deformity. Regardless, the distal extension of a fusion construct remains as a potential negative effect of failed VBT that should be discussed with patients and their families when deciding whether VBT or PSF is the best surgical option for the treatment of their scoliotic deformity.

There are several limitations to this work. Because this is a retrospective study design, some perioperative characteristics (e.g., operative time [n = 1 pair] and length of stay [n = 1 pair]) were not available for all patients. Moreover, postoperative assessments available for analysis following PSF were limited and highly variable (1.3 ± 1.2 yrs.) for PSF–VBT. As such, we were unable to determine if additional revision procedures were required, particularly those that may have led to an additional distal extension (e.g., decompensation, adding-on, etc.). These data, however, represent all the patients enrolled in a large multi-center registry that have received a PSF following a failed VBT—making these data a crucial resource for pediatric orthopedic practitioners. The largest hurdle we faced was obtaining pre-PSF bending films for PSF–VBT, with data from only seven patients available for statistical analysis. The values provided herein should be interpreted with caution as they may not accurately reflect postoperative deformity flexibility in patients that failed VBT. This study also included data from multiple sites across the USA, Canada, and Turkey, which may have increased perioperative data variability (e.g., physician experience, surgical technique, patient population, etc.). Nevertheless, these data are likely representative of the perioperative outcomes observed at other institutions around the world.

This is the first multi-center, retrospective assessment of all available adolescent idiopathic scoliosis patients originally treated with VBT who exhibited progression of their deformity resulting in a PSF. Through a matched cohort design, we found that operative time, estimated blood loss, and postoperative length of stay were similar in both cohorts, despite lower implant density and the use of more hybrid constructs in PSF–VBT. An average of two additional levels were instrumented in 50% of patients that received VBT as opposed to PSF for their index procedure. Unfortunately, many of those additional levels were added to the distal end of the construct. These additional levels may decrease postoperative mobility and increase the risk of adjacent segment disease, all of which patients are trying to avoid with their initial non-fusion surgical procedure. These findings may be integral in clinical decision making for both index and revision procedures in patients with progressive adolescent idiopathic scoliosis.