INTRODUCTION
Adolescent idiopathic scoliosis (AIS) is a complex 3-dimensional spinal deformity. The main goal of the surgical treatment of AIS should be to achieve coronal and sagittal balance and preservation of motion segments, while avoiding complications such as curve progression, coronal decompensation, junctional kyphosis, adding-on, and the need for revision surgery [10,14,22,41,52,58,60]. There has been considerable debate about the adequacy of selective spinal fusion for AIS. With the development of classification systems and instrumentation techniques, the guidelines for treating AIS have undergone significant changes. The goal has now shifted to minimizing the number of fusion segments and maximizing the number of unfused motion segments, expecting that the unfused curve will spontaneously accommodate to compensate for the corrected position of the fused curve [4]. Although spontaneous correction of the unfused curve can be achieved after selective fusion, postoperative decompensation remains a troublesome problem. Even with a precise interpretation of the curvature and determination of proper fusion level, there are also many cases of postoperative decompensation. Despite the widespread interest in correcting coronal deformities, there are discrepancies in the literature regarding the definition of coronal decompensation, and relevant risk factors have not been conclusively established. In this study, based on a literature review, we describe the current concepts of selective fusion in AIS, evaluate the available information on the postoperative development of coronal decompensation, and summarize the outcomes of this complication and related factors.
MATERIALS AND METHODS
In this review, published clinical studies and review articles dealing with selective fusion as a means of surgical treatment for AIS were included. PubMed, Scopus, Web of Science, and Google Scholar were searched from inception through May 2020 using a predefined search strategy. The following key words were searched in the databases : “adolescent idiopathic scoliosis,” “selective fusion,” “selection of fusion level,” “coronal decompensation,” and various combinations of these terms. A total 68 studies were identified. Case reports and articles that did not focus on the selection of the fusion level and postoperative changes were excluded. Finally, 28 articles were included in this systematic review. The radiological parameters mentioned in this article are summarized in Fig. 1.
HISTORY AND CRITERIA FOR SELECTIVE FUSION
Corrective surgery for AIS can provide several benefits to affected patients, including improvements in quality of life, disability, back pain, psychological well-being, and breathing function. According to Ward et al. [57], who compared the outcomes of 190 non-operatively treated AIS patients with those of 166 operatively treated patients, statistically significant differences were found in self-image, satisfaction, and total score in favor of the operative cohort. However, in many previous studies, lower back pain and disability occurred over the course of long-term follow-up in patients who had long level AIS surgery. This was related to degenerative disc disease, a late complication. Based on the study of Akazawa et al. [1], who investigated the long-term (average follow-up period : 35 years) incidence of lumbar disc degeneration and Modic changes in the non-fused segments of patients with AIS who previously underwent spinal fusion, it is recommend for the lowest fusion level to be L3 or higher in order to reduce the risk of disc degeneration in midlife. Selective fusion is a concept that has developed in relation to these problems and is defined in this review as selective thoracic fusion (STF) for thoracic curves and selective thoracolumbar/lumbar (TL/L) curves.
In 1958, Moe [36] first introduced the concept of STF for a primary thoracic curve with a compensatory lumbar curve and noted that the curve pattern suitable to STF was characteristically a primary right thoracic curve with a left lumbar curve, being structural but not as inflexible as a thoracic curve with bending to the side. The surgical treatment of AIS has made remarkable progress since the development of the Harrington rod in the late 1950s [15]. King and colleagues [27] stated that successful STF, with the lower instrumented vertebra (LIV) at the neutral vertebra (NV) and stable vertebra (SV), was performed using Harrington instrumentation with spontaneous correction of the lumbar compensatory curve in a King-Moe type II curve. The King-Moe classification of idiopathic thoracic scoliosis is a long-standing system widely used to classify curve patterns and to recommend fusion levels. However, a limitation of this system is that it is based only on the coronal plane, lacks a defined isolated thoracolumbar curve type, and has relatively poor to fair inter-observer and intra-observer reliability [12,31].
In 2001, to overcome the limitations of the King-Moe classification, Lenke et al. [31] introduced a new classification system that redefined the way in which arthrodesis levels are selected. The classification combines six coronal curve patterns (1 through 6) with three lumbar modifiers (A, B, or C) and three sagittal thoracic modifiers (minus, normal, or plus), and requires not only standing coronal and lateral full-spine radiographs, but also supine side-bending films. The spinal column regions to be evaluated in this system are proximal thoracic (PT), main thoracic (MT), and TL/L. The major curve that has the largest Cobb angle should always be included in the fusion. If the curve is regarded as being nonstructural (corrects to <25° as measured on side-bending radiographs and/or kyphosis of <20° between T2-5 and T10-L2), it does not have to be included in the fusion. The authors proposed radiographic criteria that should be considered when evaluating a patient for STF or TL/L fusion. The thoracic apical vertebral translation (AVT) is the distance between the C-7 plumb line and the center of the apical vertebral body of the thoracic curve. The TL/L AVT is the distance between the center of the apical vertebral body of the TL/L and the central sacral vertebral line (CSVL) [55]. The other factor that helps in the decision of whether selective fusion is possible is apical vertebral rotation (AVR), which is based on the Nash-Moe grading for vertebral rotation [37]. They recommended STF for structural curves to treat Lenke 1C curves if the ratios of MT : TL/L Cobb angle, AVT-MT : AVT-TL/L, and AVR-MT : AVR-TL/L are >1.2. Conversely, the recommended ratio for these parameters for selective TL/L fusion should be >1.25 (Tables 1 and 2) [33].
However, these guidelines are not routinely accepted and have several limitations. In Lenke 1C curves, Newton and colleagues [38] reported that only two-thirds of experienced surgeons would perform STF, and Crawford et al. [6] noted that only 49% (138/264) of patients underwent STF in their series; thus, the Lenke lumbar C modifier was termed as a “rule breaker.” In addition, the Lenke classification has no lowest or uppermost instrumented vertebra criteria and poor inter-observer reliability in the PT curve.
STF
Upper instrumented vertebra (UIV)
The preservation of motion is a less important factor in the selection of the UIV because the thoracic spine is rigid by nature due to the stabilizing effects of the rib cage and the sternum. Selection of the UIV is important for shoulder balance, proximal curve progression, and proximal junctional kyphosis. Several studies have attempted to establish the best predictors of the need for fusion of the PT curve [19,25]. If the PT curve is structural, should the PT curve be included in fusion? Up to what level should fusion be performed?
Lee and colleagues [29] proposed that both the PT and the MT curves should be fused when the left shoulder was elevated or PT curves were rigid, as the correction and fusion of the lower thoracic curve aggravated shoulder balance. However, this strategy has changed due to advent of strong instrumentation systems. Lenke et al. [32], who used Cotrel-Dubousset instrumentation defined criteria for inclusion of the PT curve in the fusion. These criteria included curve magnitude >30° and >25° on bending radiographs, AVR ≥Nash-Moe grade 1, AVT >1 cm, elevation of the left shoulder or tilt of T1 into the concavity of the PT curve, or location of the transitional vertebra between the two thoracic curves at T6 or lower. In 2000, Suk et al. [49] recommended that the idiopathic thoracic scoliosis with a PT curve of more than 25° and a level or elevated left shoulder should be considered a double thoracic curve pattern and should be treated by fusion of both the proximal and the distal curves. Kuklo and colleagues [28] retrospectively evaluated radiographic parameters associated with the PT curve to determine whether these characteristics could be used to predict postoperative shoulder balance. The authors concluded that spontaneous PT curve correction consistently occurs after instrumented correction of the MT curve. Furthermore, preoperative side-bending radiographs (PT curve flexibility) are positively correlated with postoperative spontaneous PT curve correction. Because of the spontaneous correction of PT curves, some authors remain skeptical about the extension of fusion to T2 or T3 in cases of nonstructural PT curves [5]. Cil et al. [5] reviewed the Lenke criteria in their series of patients with nonstructural PT curves, some of whom had undergone PT fusion. They concluded that there was no difference in terms of outcomes between including a nonstructural PT curve into the area of fusion or solely fusing the MT curve, meaning that extension of fusion to T2 or T3 is unnecessary. However, shoulder balance was not specifically assessed.
In 2008, Ilharreborde et al. [19] agreed with the earlier conclusion by Lee et al. [29] that no relationship exists between T1 tilt and that fusion for the entire PT curve is not necessary for every double thoracic curve, and therefore recommended fusion of structural PT curves and nonstructural PT curves if T1 tilt and shoulder balance are in the same direction and would be worsened with correction of the main curve. If shoulder tilt and T1 tilt are in opposite directions, they suggested that instrumentation of only the part of the upper thoracic curve (T2 or T3) would be possible [19]. Elfiky and colleagues [9] proposed that spontaneous correction of the PT curve after fusion of the MT curve occurs in structural curves greater than 35° and less than 45°. Thus, a non-fusion strategy may be appropriate for PT curves between 35° and 45°. The results of studies on UIV selection in AIS with thoracic curves are summarized in Table 3.
In summary, many studies have been conducted to determine the appropriate UIV of thoracic major curves. Based on the results of the studies so far, several review articles have recommended the following. If the PT curve is structural, the UIV is T2 for preoperative left shoulder elevation, T3 for preoperative level shoulders, and T4 for preoperative right shoulder elevation. If the PT curve is nonstructural, the UIV is the upper end vertebra (UEV) of the thoracic major+2 (2 vertebrae proximal) for preoperative left shoulder, UEV+1 (1 vertebra proximal) for preoperative level shoulder and the UEV for preoperative right shoulder elevation [10,25,26,54]. However, no consensus guidelines about selection of the UIV, surgical methodology, and risk factors for shoulder imbalance and proximal junctional kyphosis have been established. Long-term studies and methodological research are needed to determine the selection of the appropriate UIV for STF.
Lower instrumented vertebra
Although long-term follow-up data on selecting the distal fusion level are scarce, it is generally accepted that the area of spinal fusion should be as short as possible. Selection of the appropriate LIV is important in avoiding distal junctional problems such as distal junctional kyphosis (DJK) and adding-on [11].
Before the development of instrumentation techniques, fusion from the end to end vertebra (EV) with neutral rotation was recommended as the ideal LIV for AIS to prevent the adding-on phenomenon [11,50]. In 1983, King et al. [27] recommended the SV, which is the distal vertebra most bisected by CSVL, as the LIV when using Harrington rods. However, this recommendation became less applicable with the introduction of more powerful segmental instrumentation systems [53]. With the increased use of thoracic pedicle screws, the LIV selection criteria have been discussed by several authors. Suk et al. [50] suggested that the LIV should be selected based on the relative position of the NV. They recommended selecting the NV when it was the same or 1 level distal to the EV of the main curve and NV-1 if 2 or more levels separated the EV from the NV [50]. Parisini and colleagues [40] considered the rotation of the lumbar vertebra just below the lower thoracic EV as an important factor, like the SV position, for determining fusion levels in single thoracic curves where the lumbar compensatory curve does not cross the middle line. If the rotation of the first vertebra just below the EV is in the same direction as the thoracic curve, and if the SV and EV have a difference of >2 levels, then distal fusion to L2 or L3 is recommended. However, if the rotation of the first vertebra just below the EV is in the opposite direction, and if SV and EV show a difference of ≤2 levels, then SV-2 or SV-3 can be selected as the LIV. Wang et al. [56], who previously reported that the selection of the LIV was closely correlated with the presence of the addingon phenomenon, suggested that choosing the first vertebra in the cephalad direction from the sacrum whose deviation from the CSVL is more than 10 mm as the LIV provides the best outcomes, as doing so both prevents adding-on and conserves more lumbar motion and growth potential. Sarlak and colleagues [45] argued that the tilt of L3 and L4 in the coronal plane may play a significant role in determining the distal fusion level in Lenke 1A curves. They recommended that the distal fusion level should be extended to at least lower end vertebra (LEV)-1 in type 1A curves with a neutral L3 vertebra, while it might be necessary to go down to the LEV with L3 vertebral tilt [45]. However, the selection of a proper LIV for thoracic major curves with lumbar modifier C remains controversial. Takahashi et al. [52] focused on the choice of fusion levels in STF of AIS with Lenke type 1B, 1C, and 3C curves. They distinguished the curve patterns according to relative positions of the SV and the lower EV. If the SV was below the EV, the LIV should be chosen at or at least 1 level distal to the SV. If the SV and the EV are same, they recommended choosing the LIV to be 1 level below the SV/EV to achieve the greatest correction of both thoracic and lumbar curves, as well as trunk shift.
Matsumoto et al. [35] investigated the occurrence of and factors related to postoperative adding-on in Lenke 1A curves. They suggested that the LIV should be extended to or beyond the last touched vertebra (LTV; the LIV that is touched by the CSVL) to avoid the development of postoperative adding-on. Fischer et al. [11], who evaluated the optimal LIV on the basis of rotation or CSVL, proposed that the LIV should be the LTV or within 2 levels proximal to the NV. Shen and colleagues argued that both SV and the last substantially touched vertebra (LSTV) can be applicable for the LIV and showed favorable outcomes in the management of Lenke 1A curves [46]. However, in comparison with the LSTV, the SV significantly decreases the occurrence of the adding-on phenomenon. In a recent study by Qin et al. [42], Lenke type 2 curves could be classified as 2A-R (L4 tilt to right) and 2A-L (L4 tilt to left) according to the direction of L4 vertebral tilt. They recommended extending the fusion level to the LSTV in 2A-R curves and to 1 level distal from the LSTV in 2A-L curves to avoid the distal adding-on phenomenon. The results of studies on LIV selection in AIS with thoracic curves are summarized in Table 4.
To summary, numerous studies have been conducted on the selection of the appropriate LIV in STF for avoiding complications, such as adding-on and DJK, but this question remains controversial. Further comparative research on various criteria is required to improve long-term surgical outcomes (Fig. 2).
Selective thoracolumbar or lumbar fusion
UIV
The concept of minimizing the fusion levels and saving motion segments in TL/L curves are not new in the literature. Lenke et al. [31], in their new surgical classification system, suggested that the UIV is often at the UEV of the TL/L curves, and Sanders and colleagues determined that if the ratio of the TL/L to thoracic Cobb angle magnitude is 1.25 or greater, the thoracic curve bends out to 20° or less, and the triradiate cartilages are closed, TL/L selective fusion showed favorable outcomes and did not require further surgery [44].
Shufflebarger et al. [47] reported favorable outcomes of PSF for TL/L curves in AIS using a wide posterior release and posterior pedicle screws from the inferior EV to the UEV. According to long-term data reported by Bennett et al. [2], the UEV has usually been selected as the UIV in TL/L curves. A review article by Trobisch et al. [54] provided guidelines for Lenke 5C curves and recommended that the UEV should be used as the UIV unless it is not at the apex of the thoracic kyphosis. In 2014, Okada and colleagues [39] published the first comparative study addressing whether a short fusion strategy is applicable in Lenke 5C curves. They reported that not using the UEV as the UIV in Lenke 5C curves, but instead having the UIV be 1 level caudal to the UEV, is a reasonable alternative as a short fusion strategy [39]. In 2016, Sudo et al. [48], who agreed with Okada et al. [39], reported that radiological and clinical outcomes did not differ according to whether the UIV was the UEV or 1 level caudal to the UEV. The results of studies on UIV selection in AIS with TL/L curves are summarized in Table 5.
In summary, in the management of TL/L curves in AIS, some consensus has been reached regarding the selection of the UIV at 1 level caudal to the UEV as a short fusion strategy that can be an alternative to using the UEV.
Lower instrumented vertebra
For TL/L AIS, a short fusion strategy that saves the distal motion segment and reduces the risk of disc degeneration and low back pain is the goal of surgery. Some authors reported that no clinical and radiological differences were observed according to the distal fusion level (L3 vs. L4, LEV vs. LEV+1) [8,51]. However, it is still believed that preservation of 1 more distal segment may be crucial for achieving a better long-term prognosis.
In 2014, Kim et al. [24] evaluated 66 patients with TL/L idiopathic scoliosis and suggested that L3 can be the LIV when the preoperative L3 crosses the mid-sacral line with rotation of less than grade II on bending radiographs. Otherwise, patients in whom fusion was extended to L4 showed satisfactory outcomes. In a subsequent study, Chang et al. [3] distinguished LIV selection in patients with TL/L idiopathic scoliosis according to the flexibility of the TL curve. If the curve is flexible (L3 crosses the CSVL with rotation < grade II), the LIV should be selected at L3 (LEV), while if the curve is rigid (L3 does not cross the CSVL, or rotation ≥ grade II), the LIV should be lower, at L4 (LEV+1). Lee and colleagues [30] proposed that the LTV was an important factor affecting the optimal correction rate and progression of adjacent disc wedging. They recommended that if the LEV ≥ L3 and the LTV ≥ L4, distal fusion at L3 might be a good choice for saving lumbar motion segments. However, if LEV ≤ L4 and LTV = L5, caution is required in selecting the distal fusion level. Recent studies of TL/L AIS also documented that stopping fusion at L3 or preserving 1 or 2 mobile segments showed similar clinical and radiological outcomes to those obtained with longer fusions [7,20]. The results of studies on LIV selection in AIS with TL/L curves are summarized in Table 6.
Although controversies remain, there is agreement that fusion should be stopped at L3 if certain parameters are met in TL/L idiopathic scoliosis. Further comparative, long-term follow-up studies are required to establish definitive guidelines for the selection of the distal fusion level of TL/L major curves.
Coronal decompensation
Coronal decompensation is a complication of corrective surgery for AIS. The Scoliosis Research Society (SRS) defines “decompensation” as a distance of more than 2 cm between the C7 plumb line and the central sacral vertical line [23,33]. Care should be taken when choosing the UIV and LIV in selective fusion to prevent postoperative coronal decompensation. In the literature, coronal decompensation is thought to result from misinterpretation of the curvature, choice of an improper fusion level, incorrect rod derotation and direct vertebral rotation maneuvers, preoperative lumbar curve characteristics, and overcorrection [43,50,61].
Overcorrection of the primary thoracic or TL/L curve has been considered the most important factor related to coronal decompensation. Winter et al. [59] defined overcorrection as correction exceeding the flexibility of the main curve and stated and that it will cause problems in compensatory curves. Decompensation is believed to be worse when the thoracic or TL/L curve undergoes significant correction during surgery, but the compensatory curve is not spontaneously corrected well enough and is larger than the primary curve after selected thoracic or TL/L fusion. To achieve a well-balanced spine, surgeons should not perform correction in excess of preoperative flexibility and should avoid instrumentation within the “transitional mobile segments” that form the junction between the major and secondary curves [53]. However, male sex, skeletal immaturity, and less correction of the major curve have been reported to be associated with a higher rate of coronal decompensation [13]. Thus, determining the appropriate amount of correction during surgery is important to prevent decompensation.
Another causative factor that contributes to decompensation is selection of an inappropriate fusion level. When performing STF, coronal decompensation is a major concern when the LIV is selected distal to the SV. Therefore many authors have suggested selecting the LIV at the SV in the thoracic major curve to avoid postoperative coronal decompensation [21,46,52]. Preoperative lumbar characteristics are also considered causative factors of coronal decompensation, but the details remain controversial. Liu et al. [34] claimed that preoperative LIV translation and LIV tilt are two important parameters that predict the immediate postoperative coronal balance. During postoperative follow-up, UIV tilt may play a highly important role in postoperative coronal decompensation in TL/L curves [34]. Hwang et al. [16] proposed that less flexibility of the TL/L curve, greater TL kyphosis, and greater distal junctional angle were predictive factors for immediate coronal imbalance in Lenke 5C curves. However, coronal imbalance was corrected spontaneously with a decrease in the LIV tilt. Hence, they suggested that early postoperative coronal decompensation was not a significant factor affecting clinical outcomes.
Hyun and colleagues reported risk factors for distal adding-on or DJK in AIS treated by posterior spinal fusion (PSF) to L3 with a minimum 2-year follow-up [17,18]. In their series, the prevalence of distal adding-on or DJK at the final follow-up was 13.1%. Multiple logistic regression analysis indicated that a preoperative Risser grade of 0 or 1 (odds ratio [OR], 1.8; p=0.014), SV-3 at L3 in standing and side-bending (OR, 2.1 and 2.8; p=0.003 and p<0.001, respectively), total stability scores of -5 or -6 at L3 (OR, 4.4; p<0.001), a rigid disc at L3-4 (OR, 3.1; p<0.001), LIV rotation >15° (OR, 2.9; p=0.001), and LIV deviation >2 cm from CSVL (OR, 2.2; p=0.006) were independent predictive factors. Thus, the authors suggested that the CSVL should touch L3 on upright and bending films, the TS score should be -4 or less, the L3/4 disc should be flexible, L3 should be neutral (<15°) and ≤2 cm from the CSVL, and patients should be ≥ Risser 2 to prevent adding-on or DJK following PSF to L3. Interestingly, they also reported that the prevalence of adding-on or DJK at the final follow-up for AIS with LIV at L3 was significantly higher in patients who underwent anterior spinal fusion. In their series, the ultimate SRS-22 scores were significantly better in patients who underwent PSF. Thus, the authors suggested that if the LIV is L3, PSF is recommended to achieve the greatest correction of both thoracic and lumbar curves, as well as trunk shift, and to prevent adding-on or DJK following fusion to L3. However, due to the lack of research on definite radiological parameters and long-term follow-up data, the effects of preoperative parameters on coronal decompensation are still being debated.
CONCLUSION
The goal of corrective surgery for AIS is provide balanced curve correction while leaving the maximal possible number of mobile vertebral segments. Despite the continuing evolution of classification systems, improvements in instrumentation, and advances in surgical techniques, the choice of the correct fusion levels in AIS is still complex and a consensus has not been established. Based on the results of the studies that have been conducted so far, some agreement has been reached regarding the selection of the UIV in thoracic major curves and the UIV and LIV in TL/L major curves. However, the selection of an appropriate LIV in STF remains controversial, with considerable differences found among authors depending on their preferences in terms of radiographic parameters and complications to avoid. Definitive guidelines for choosing fusion levels in AIS have not been standardized.
Coronal decompensation is a critical complication after corrective surgery in AIS. Causative factors that contribute to decompensation are male sex, skeletal immaturity, undercorrection or overcorrection, selection of an inappropriate fusion level, and preoperative clinical and radiographic characteristics. Therefore, spine surgeons should carefully consider the factors mentioned above when planning surgery.
Further research is needed to establish definitive guidelines for the selection of the fusion level in corrective surgery while minimizing coronal decompensation.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
