Background

Selective thoracic fusion (STF) has been the gold standard for treating primary thoracic and compensatory lumbar curves in adolescent idiopathic scoliosis (AIS), in which both the thoracic and lumbar curves cross midline, and the lumbar curve is smaller and more flexible than the thoracic curve, since Moe had advocated its concept [1, 2]. For this curve pattern, STF induces spontaneous lumbar curve correction and preserves more mobile lumbar segments than does fusing both the thoracic and lumbar curves [3,4,5,6,7,8]. However, postoperative coronal imbalance is a significant problem after STF [7, 9,10,11], which may result in poor surgical outcomes with re-operation. Moreover, reported surgical outcomes have shown that patients with resultant coronal imbalance after surgery tend to have inferior Scoliosis Research Society (SRS) questionnaire score on patients’ satisfaction with treatments, compared to balanced patients [12]. Thus, the postoperative behavior of coronal balance is a major concern in the surgical outcomes for primary thoracic and compensatory lumbar curves in AIS. Causative factors reported for postoperative coronal decompensation include excessive correction of the thoracic curve, improper selection of the lowest instrumented vertebra (LIV), pre-existing coronal decompensation, and inappropriate curve identification [7,8,9,10,11, 13]; however, many of previous studies on postoperative behavior of coronal balance after STF are based on various surgical approaches including anterior approach and/or posterior approach with pedicle screw (PS), hook or hybrid constructs [3, 4, 6, 10, 13], and those with PS construct are scant. Apparently, surgical procedures and corrective maneuvers would influence the postoperative course of unfused lumbar curve and coronal balance.

On the other hand, we also experience some improvement of postoperative coronal imbalance during postoperative period after STF; however, some patients persist coronal imbalance. Compared to causative factors for the onset of postoperative coronal decompensation, the factors associated with the improvement of postoperative coronal imbalance during postoperative period have not been well investigated.

Thus, the purposes of this study were (1) to evaluate the postoperative behavior of coronal balance after posterior thoracic fusion with PS construct for Lenke 1C and 2C AIS (Lenke 1C/2C-AIS), in which the thoracolumbar/lumbar (TLL) curve bends less than 25° on side-bending without 20° or more kyphosis between T10 and L2, and the apical vertebra of TLL curve does not touch the center sacral vertical line (CSVL) and (2) to identify factors related to the onset of immediately postoperative coronal decompensation (IPCD) and changes of coronal balance during the follow-up period. In this study, postoperative change of coronal balance (PCCB) was defined as the change of coronal balance between preoperative and immediately postoperative evaluations, whereas postoperative coronal balance remodeling (PCBR) was defined as the improvement of postoperative coronal balance between immediately postoperative and final follow-up evaluations.

We hypothesized that preoperative radiographic measurements, LIV selection, and amount of main thoracic (MT) curve correction would have an impact on IPCD, PCCB, and PCBR.

Methods

This study was designed as a multicenter retrospective study reviewing radiographs and clinical charts. After institutional review board approval by the ethics committee at the International University of Health and Welfare (No. 5-15-19), radiographic measurements and clinical charts review were conducted. Twenty-one patients (twenty females, one male), who underwent posterior thoracic fusion for Lenke 1C or 2C AIS at three institutes, were enrolled in this study. Inclusion criteria were as follows: (1) primary surgery by posterior thoracic fusion with PS construct of LIV ending at L3 or above for Lenke 1C/2C-AIS and (2) patients with minimum 2 years follow-up.

The mean age and Risser grade at the time of surgery were 15.1 ± 2.7 (11–22) years old and 3.4 ± 1.4 (1–5), respectively. The mean follow-up period was 3.1 (2–7.3) years. Investigated clinical data were Lenke classification, number of fused vertebrae, and level of LIV. Radiographic coronal measurements included the Cobb angles of the proximal thoracic (PT), MT, and TLL curves; apical vertebral translations (AVT) of the MT (AVT-MT) and TLL (AVT-TLL) curves; LIV tilt; coronal balance; and trunk shift on preoperative, postoperative (between 1 and 4 weeks after surgery), and final follow-up standing posteroanterior (PA) radiographs. Preoperative curve flexibilities were evaluated on right and left side-bending films. Radiographic sagittal measurements included the proximal thoracic kyphosis between T2 and T5, thoracic kyphosis between T5 and T12, thoracolumbar kyphosis between T10 and L2, lumbar lordosis between T12 and S1, and sagittal balance on preoperative, postoperative, and final follow-up standing lateral radiographs.

Measurements for coronal curvature were performed by the Cobb method. AVT-MT and AVT-TLL were measured with reference to the C7 plumb line or the CSVL, respectively. Coronal balance was measured as the horizontal distance between the C7 plumb line and the CSVL, and trunk shift was measured as the horizontal distance between the center of apex for MT curve and the CSVL. Coronal balance and trunk shift were defined as a negative value when the C7 plumb line or the center of apex for MT curve locates at the left side to the CSVL. Sagittal balance was measured as the horizontal distance between the C7 plumb line and superior-posterior corner of S1 vertebra and was defined as a positive value when the C7 plumb line locates anteriorly to superior-posterior corner of S1 vertebra, and vice versa. Coronal decompensation was defined as the absolute value of coronal balance greater than 2 cm. The ratios of values of MT curve to those of TLL curve in preoperative Cobb angle and AVT were also calculated. Reference vertebrae including the lower end vertebra (EV) and the lower neutral vertebra (NV) of MT curve, the stable vertebra (SV) at thoracolumbar lesion, and the apex of TLL curve (ApexTLL) were recorded, and the gap differences between the LIV and the reference vertebrae were counted (LIV-EV, LIV-NV, LIV-SV, ApexTLL-LIV).

Preoperative radiographic measurements, including the Cobb angles, and flexibilities of the PT, MT, and TLL curves, the AVT of the MT and TLL curves, the ratios of the MT curve to TLL curve in the Cobb angle and AVT, coronal balance, trunk shift and sagittal values, LIV selection, and postoperative correction of MT curve were evaluated whether correlating or not, to PCCB and PCBR.

Additionally, patient-reported clinical outcomes with Scoliosis Research Society-22 (SRS-22) questionnaire scores at the final follow-up were reported.

Surgical procedure

All patients were operated on posteriorly. After exposure of the posterior elements of the spine, removal of inferior facet joints to be fused in all cases was performed. Pedicle screws were inserted bilaterally by free-hand technique, with additional attachment of some hooks in seven cases. For most patients, the LIV was selected at the SV on a PA radiograph; for others, the LIV was selected at a more distal level to partially correct the TLL curve. Corrective maneuver includes rod rotation on the concave side, followed by in-situ rod contouring, placement of a rod on the convex side, and segmental compression and distraction via PS. Intra-operative spinal cord monitoring with motor-evoked potential was routinely performed. Bone grafting was carried out using local bone materials from facetectomies and osteotomies in all cases.

Statistical analyses

Statistical analyses were performed using the software package IBM SPSS Statistics 22 (IBM Japan, Tokyo, JAPAN). Radiographic measurement values at different time points were evaluated by analysis of variance. Unpaired t test was used to compare continuous variables. Correlation analyses were performed using Pearson’s correlation coefficients. A P value less than 0.05 was set to be statistically significant.

Results

The Lenke classification was 1C− in three, 1CN in ten, 1C+ in one, 2C− in two, and 2CN in five. The mean number of fused vertebrae was 9.6 ± 2.4(6–14), and the LIV was T11 in six, T12 in six, L1 in five, L2 in three, and L3 in one (Table 1). Details of each patient are shown at Table 2. The mean Cobb angles for the PT, MT, and TLL curves were 29.4 ± 9.4° (flexibility, 24.7 ± 19.3%), 57.3 ± 12.1° (flexibility, 29.6 ± 16.3%), and 42.3 ± 5.7° (flexibility, 75.0 ± 14.6%) preoperatively and were corrected to 16.1 ± 7.6°, 17.8 ± 7.7°, and 19.3 ± 7.4° immediately postoperatively and 16.9 ± 7.8° (42.8 ± 18.8% correction), 22.8 ± 8.9° (60.4 ± 12.2% correction), and 22.5 ± 7.8° (46.6 ± 17.6% correction) at the final follow-up, respectively (Table 3). The preoperative Cobb angle ratio of the MT curve to the TLL curve was 1.4 ± 0.2. The postoperative changes in the Cobb angles of the PT, MT, and TLL curves were statistically significant (Table 3). The mean AVT-MT and AVT-TLL were 45.2 ± 15.5 mm and 22.0 ± 5.7 mm preoperatively and were corrected to 2.0 ± 11.7 mm and 25.0 ± 7.0 mm immediately postoperatively and 10.0 ± 13.0 mm and 16.8 ± 8.4 mm at the final follow-up (Table 3). The preoperative AVT ratio of the MT curve to the TLL curve was 2.4 ± 1.2. The mean coronal balance and trunk shift were −3.8 ± 10.8 mm and 41.3 ± 20.5 mm preoperatively and were corrected to −21.2 ± 13.6 mm and −19.3 ± 16.7 mm immediately postoperatively and −12.0 ± 11.1 mm and -2.5 ± 17.4 mm at the final follow-up, respectively (Table 3) (Fig. 1). Coronal decompensation was observed in two patients preoperatively, in ten immediately postoperatively, and in three at the final follow-up. No patients required revision surgery. The mean LIV tilt was 24.4 ± 12.3° preoperatively and was corrected to 8.1 ± 6.7° immediately postoperatively and 10.2 ± 5.8° at the final follow-up. The postoperative changes in AVT-MT, AVT-TLL, LIV tilt, coronal balance, and trunk shift at the final follow-up were statistically significant (Table 3).

Table 1 Patients’ demographics
Full size table
Table 2 Details of each patient
Full size table
Table 3 Radiographic coronal measurements
Full size table
Fig. 1
figure 1

Postoperative behavior of coronal balance and trunk shift. Preoperative coronal balance and trunk shift were significantly shifted to the left immediately postoperatively; however, both of them regained balance at final follow-up

Full size image

Preoperative sagittal measurements did not change significantly at the final follow-up (Table 4).

Table 4 Radiographic sagittal measurements
Full size table

The mean gap differences in LIV-EV, LIV-NV, LIV-SV, and ApexTLL-LIV were 1.0 ± 1.4, 1.1 ± 1.5, 0.6 ± 1.4, and 1.9 ± 1.2, respectively.

Factors associated with IPCD, PCCB, and PCBR

In 20 of 21 patients, coronal balance shifted to the left immediately after surgery. Comparative study showed that among preoperative radiographic measurements, LIV selection and the amount of MT curve correction, the preoperative coronal balance and trunk shift and LIV-SV were significantly different between patients with IPCD and those without (−11.0 ± 7.8 mm vs 2.7 ± 9.0 mm, p < 0.01; 30.1 ± 15.1 mm vs 51.5 ± 19.9 mm, p < 0.05; 1.3 ± 1.5 vs −0.1 ± 0.9, p < 0.05, respectively) (Table 5). PCCB had significant correlation to the LIV selection (LIV-EV r = −0.58, p < 0.01; LIV-NV r = −0.47, p < 0.05; LIV-SV r = −0.48, p < 0.05; ApexTLL-LIV r = 0.61, p < 0.01), whereas PCCB had no significant correlation to the preoperative radiographic measurements or the amount of MT curve correction (Table 6).

Table 5 Comparative analyses between patients with IPCD and those without
Full size table
Table 6 Correlation analyses on PCCB and PCBR
Full size table

On the contrary, PCBR had no significant correlation to the preoperative radiographic measurements, LIV selection, or the amount of MT curve correction, whereas PCBR had significant correlation to the immediately postoperative coronal balance (r = −0.61, p < 0.01) and trunk shift (r = −0.57, p < 0.01) (Table 6) (Fig. 2).

Fig. 2
figure 2

A representative case (Case No. 15). A 15-year-old female with Lenke 1CN adolescent idiopathic scoliosis underwent selective posterior thoracic fusion from T6 to T11. The preoperative coronal balance of −4 mm (a) worsened to −24 mm immediately after surgery (b), but recovered to −9 mm with remodeling, as seen 2 years after surgery (c)

Full size image

SRS-22 questionnaire scores

Domain scores of SRS-22 questionnaire were 4.4 ± 0.2 in Pain, 4.4 ± 0.5 in Function, 3.4 ± 0.9 in Self-Image, 3.9 ± 1.0 in Mental Health, 3.9 ± 0.7 in Satisfaction, and 4.0 ± 0.4 in Total at the final follow-up.

Discussion

The surgical outcomes for a primary thoracic curve with a compensatory lumbar curve remain one of the most controversial issues in surgical AIS treatment. Advantages of STF for this curve pattern are saving more mobile lumbar segments and inducing spontaneous lumbar curve correction; however, the risks of postoperative coronal decompensation and resultant marked lumbar curve magnitude compared to correction and fusion of both the thoracic and lumbar curves are considered to be shortcomings in STF. The primary thoracic and compensatory lumbar curves are generally classified as King type II or Lenke 1C/2C curves, in which the MT curve is larger and more rigid than the TLL curve and both the MT and TLL curves cross midline. In these curve patterns, coronal balance are usually prone to shift to the left preoperatively, and previous studies have shown that coronal balance tends to shift further to the left after STF, which may result in coronal decompensation in some patients [7]. Whereas, it is also known that patients with immediately postoperative coronal imbalance usually regain coronal balance to some degrees during the follow-up period; however, some patients persist coronal imbalance after surgery.

Several causative factors have been reported for postoperative coronal decompensation; however, many of these causative factors were based on not only various surgical approaches but also the evaluation with the change of coronal balance between preoperative and final follow-up evaluations. Detailed analyses to seek the factors associated with the changes of coronal balance between preoperative and immediately postoperative follow-ups and between immediately postoperative and final follow-ups are still lacking. PCBR, defined as the improvement of coronal balance between immediately postoperative and final follow-ups, is another aspect of the postoperative behavior of coronal balance, compared to the onset of postoperative coronal decompensation.

Thus, we attempted to define the correlative factors to IPCD, PCCB, and PCBR after posterior thoracic fusion with PS construct in Lenke 1C/2C-AIS.

As demonstrated in the results, coronal balance tends to shift to the left preoperatively, and it shifts further to the left immediately postoperatively in 20 of 21 patients. To detect the causative factors for the onset of IPCD, we performed comparative and correlation analyses. From the comparative study, preoperative coronal balance and trunk shift, and LIV-SV were found to be significantly different between compensated and decompensated patients immediately postoperatively, whereas correlation analysis found that PCCB was significantly correlated to the LIV selection relative to the EV, NV, SV, and ApexTLL. Regarding LIV selection, only the LIV-SV was found to be significantly associated with both the onset of IPCD and PCCB. These results indicate that the immediately postoperative change of coronal balance depends only on LIV selection, and not on preoperative radiographic measurements or the amount of MT curve correction, and that a fixation more distal to the SV shifts the coronal balance even further to the left immediately postoperatively. Moreover, a patient with a pre-existing coronal imbalance to the left tends to be decompensated to the left immediately postoperatively if the LIV is selected at a level more distal to the SV. It is important to note that coronal balance shifted further to the left immediately postoperatively in majority of patients even when the LIV was selected at the SV, indicating that STF may itself produce an immediately postoperative leftward shift of the coronal balance [11]. Accordingly, patients with preoperative coronal imbalance further to the left and inappropriate selection of LIV, such as fixation distal to the SV would be at a high risk of IPCD. These findings suggest that selecting the LIV at the SV would minimize the risk of IPCD for patients with Lenke 1C/2C-AIS.

Regarding PCBR, this study demonstrated that PCBR occurred in majority of patients more than 5 mm during postoperative periods and the deterioration of immediately postoperative coronal balance occurred in only limited cases (four cases) at the final follow-up and also found that PCBR had significantly negative correlation to the immediately postoperative coronal balance and trunk shift. These results indicate that patients with coronal balance shifted further to the left immediately postoperatively tend to compensate more balance during the follow-up period, which may be attributed to a postural reflex, potentially existing in the relatively flexible lumbar curves (Lenke 1C/2C-AIS), although three patients persisted coronal decompensation at the final follow-up. That is, PCBR is dependent on the immediately postoperative condition in coronal balance and trunk shift. Coronal decompensation at the final follow-up was observed in three patients, so comparative analysis to detect the causative factors for coronal decompensation at the final follow-up was not performed because of the limited number of patients, although the LIV-SV tended to be larger in patients with coronal decompensation at the final follow-up, compared to those without (2.3 vs 0.3). Patients with coronal decompensation at the final follow-up were all decompensated immediately postoperatively, and none of the compensated patients immediately postoperatively were decompensated at the final follow-up. Accordingly, optimal curve identification and surgical strategy to minimize the risk of IPCD may be more beneficial to prevent coronal decompensation at the final follow-up. The current study suggests that selecting the LIV at the SV minimizes the risk of postoperative coronal decompensation after posterior thoracic fusion using PS construct for Lenke 1C/2C-AIS and results in acceptable patients-reported outcomes.

Further studies to determine the optimal candidate for STF based on the preoperative characteristics of the TLL curve and coronal balance are needed to prevent postoperative coronal decompensation in the primary thoracic and compensatory lumbar curves.

This retrospective study has several limitations, including the small sample size and relatively short follow-up period. As for the follow-up period, observation with longer postoperative period is critical in the evaluation of clinical and radiographic outcomes, especially in treating young patients. However, the postoperative spontaneous unfused lumbar curve correction and remodeling of trunk shift after STF are reported to occur usually within 2 years after surgery [14], so a minimum 2-year follow-up observation may minimize these concerns.

Conclusions

The onset of IPCD was associated with the preoperative coronal balance and trunk shift, and the LIV selection relative to the SV, whereas PCBR was associated with the immediately postoperative coronal balance and trunk shift in patients treated by posterior thoracic fusion using PS construct for Lenke 1C/2C-AIS. Selecting the LIV at the SV would be optimal in treating Lenke 1C/2C-AIS in terms of avoiding postoperative coronal decompensation.