Adolescent idiopathic scoliosis (AIS) is a three-dimensional deformity characterized by the lateral deviation of the spine, axial rotation, and thoracic hypokyphosis [1]. Surgery aims to restore coronal and sagittal balance; however, some patients remain hypokyphotic postoperatively [2]. In particular, posterior spinal fusion (PSF) may lead to further iatrogenic loss of thoracic kyphosis (TK) [3]. Persistent thoracic hypokyphosis is associated with spinopelvic malalignment: loss of lumbar lordosis [3], cervical kyphosis [4], and spinopelvic incongruence, leading to extensor muscle fatigue [5] and lower back pain [6]. Poor sagittal alignment in adulthood may lead to early cervical [7,8] and lumbar spine degeneration [3]. Moreover, thoracic hypokyphosis was positively correlated with diminished pulmonary function [9,10].

Intraoperative TK correction is affected by the inherent sagittal flexibility of the spine. Luk et al. [11] utilized fulcrum bending radiographs to identify the coronal sagittal coupling behavior of the spine in vivo: coronal deformity correction would lead to corresponding changes in the sagittal plane. TK tends to normalize under fulcrum bending correction of the coronal curve, and kyphosis correction on fulcrum bending was correlated with the direction of postoperative correction. No difference was found in the correction between the screw–rod constructs and the hook–rod systems; thus, in sagittal correction, the natural coupling behavior plays a more important role than the implant system.

Surgical factors can influence TK correction. Lower implant density [12], stiffer rod material [13], larger rod diameter [14], and certain curve correction maneuvers such as posteromedial translation and rod rotation techniques [15] could improve sagittal correction. Some studies have suggested the use of Ponte osteotomies [16–18] or releases [19,20] to improve sagittal correction; however, its indications remain unclear because of conflicting evidence on its efficacy and safety [21,22]. Although Ponte osteotomies improved coronal and sagittal correction for patients who were hypokyphotic, it increased operating times and blood loss [23] but did not improve patient-reported outcomes [24]. Consequently, the routine use of Ponte osteotomies in patients with hypokyphosis remains controversial because of its limited benefits and potential surgical risks. Thus, we hoped to define the indications for the use of Ponte osteotomies and releases by utilizing fulcrum bending radiographs to denote when TK would fail to normalize postoperatively.

This study aimed to investigate the relationship between fulcrum bending and postoperative kyphosis correction, identify the surgical risk factors and risk factors on fulcrum bending radiographs for postoperative hypokyphosis, and establish a cutoff of the identified risk factors for postoperative hypokyphosis. The hypothesis was that a low kyphosis on lateral fulcrum bending radiographs is a risk factor for postoperative hypokyphosis.

In total, 156 patients were included in this study; the inclusion and exclusions are detailed in Fig. 2. In this study, 75.6% of the patients had Lenke 1A, 16.7% had Lenke 1B, and 7.7% had Lenke 1C curves. At the 2-year follow-up, the proportion of patients with hypokyphosis decreased from 66.0% (n=103) preoperatively to 43.6% (n=68). Preoperatively, there was no significant difference in demographic and surgical factors between patients with different kyphotic profiles at 2-year follow-up (p>0.10). Patients with hypokyphosis achieved significantly more kyphosis correction on fulcrum bending (p<0.001). Detailed comparisons of the surgical, demographic, and radiographic characteristics are presented in Table 1.

Regarding surgical characteristics, the mean implant density was 1.16±0.14. No difference was found in kyphosis correction with different rod materials (p=0.429): 96.8% (n=151) used titanium, 2.6% (n=4) used cobalt-chromium, 0.6% (n=1) used cobalt-chromium rod on the concave side and titanium rod on the convex side.

T5–T12 kyphosis correction on fulcrum bending was proportional to postoperative correction (immediate postoperative, r=0.540, p<0.001; 2-year follow-up, r=0.369; p<0.001). A negative correlation was found between preoperative T5–12 kyphosis on standing and fulcrum bending (r=−0.497, p<0.001); patients with more severe hypokyphosis could achieve more kyphosis correction. T5–12 kyphosis tended to normalize postoperatively (immediate postoperative, r=−0.737, p<0.001; 2-year follow-up, r=−0.619, p<0.001), as predicted on fulcrum bending. However, significant differences were noted between the T5–12 kyphosis on fulcrum bending and postoperatively (p<0.001).

Of all preoperative factors (Table 2), T5–12 kyphosis on fulcrum bending was the only significant risk factor for postoperative hypokyphosis (immediate postoperative odds ratio [OR], 0.870; 95% confidence interval [CI], 0.826–0.917; p<0.001; 2-year follow-up OR, 0.916; 95% CI, 0.876–0.959; p<0.001) (Table 3), whereas a 10% increase in convex side implant density was associated with a 74.9% increase in the risk of postoperative hypokyphosis at 2-year follow-up (OR, 1.749 per 10% increase; 95% CI, 1.056–2.897; p=0.03) (Table 3). The cutoff of T5–12 kyphosis on fulcrum bending for 2-year postoperative hypokyphosis was 12.45° (area under the curve, 0.773; 95% CI, 0.661–0.820; sensitivity, 0.773, specificity, 0.662).

Among the 14 patients with a lordotic thoracic spine preoperatively, one patient became normokyphotic at 2-year follow-up. However, there was no significant difference in the baseline demographic and radiological characteristics between the normokyphotic and hypokyphotic groups (p>0.10) (Table 4).

Although coronal correction remains the primary objective in patients with AIS undergoing surgery, there is increasing emphasis on restoring sagittal balance to prevent early degenerative changes caused by spinopelvic malalignment [35]. Ideal TK targets have been widely discussed in the literature as arbitrary cutoffs [35,36] or patient-specific targets according to spinopelvic parameters [37]. However, the introduction of excessive kyphosis to meet these ideal targets could result in adjacent segment disease [38]. Nonetheless, regardless of the “ideal” kyphosis, assessing whether patients will become normokyphotic (≥20°) following coronal curve correction and instrumentation is challenging. This study evaluated the surgical and preoperative risk factors on fulcrum bending radiographs for postoperative hypokyphosis in patients with Lenke 1. A low T5–12 kyphosis on fulcrum bending and high convex side implant density were risk factors for postoperative hypokyphosis (Figs. 3, 4).

Various techniques have been suggested to assess coronal curve flexibility to predict postoperative outcomes including fulcrum bending [26,39–41] and side bending [42]. Specifically, fulcrum bending radiographs can also be used to predict sagittal correction. Luk et al. [11] demonstrated the natural coronal sagittal coupling of the spine on fulcrum bending radiographs, where correcting the coronal curve resulted in some degree of sagittal correction. A positive relationship was found between the coupled sagittal changes and postoperative sagittal correction, indicating that the inherent sagittal flexibility of the spine hinders sagittal curve correction. Furthermore, patients with hypokyphosis tended to achieve more sagittal correction, demonstrating the tendency of the spine to “self-normalize [11].” This is in line with the findings of this study: sagittal correction on fulcrum bending was proportional to postoperative outcomes at immediate (r=0.540) and 2-year postoperative (r=0.369) periods; TK tended to normalize postoperatively (r=−0.737).

To the best of our knowledge, this study is the first to demonstrate the predictive value of fulcrum bending radiographs for postoperative hypokyphosis: high T5–12 TK on fulcrum bending significantly decreases the risk of hypokyphosis by 13.0% and 8.4% at immediate and 2-year postoperative periods, respectively. Patients with rigid sagittal profiles on coronal bending are unlikely to achieve complete kyphosis restoration by coronal curve correction with alternate pedicle screw constructs alone. In the long term, normalization of TK would cause reciprocal changes in unfused segments: improving lumbar lordosis and cervical kyphosis and preventing degenerative changes [43]. A previous study also reported low TK and small coronal curve magnitude as risk factors for hypokyphosis at 2-year follow-up [14]. The present findings propose a correlation between coronal curve flexibility (r=0.187) and low kyphosis (r=0.462) with postoperative kyphosis; however, it was not a significant risk factor after adjusting for other preoperative factors. Moreover, previous studies were limited by the inclusion of multiple curve types (Lenke 1, 2, 3, 4, and 6) with more than one structural curve. This may cause heterogeneity in the results as patients undergoing longer fusions have a higher risk of postoperative hypokyphosis [44].

Significant differences remain between sagittal correction on coupling and postoperatively (p<0.001). Additional correction can be attributed to surgical factors. Angelliaume et al. [45] reported better T4–12 kyphosis restoration by cobalt-chromium than titanium rods, with less loss of correction over time [46]. In the present study, although titanium rods were used in most patients, the differences in rod materials yielded no difference in kyphosis correction (p>0.40). This could be due to the large differences in the number of patients in each group, which limited the statistical power.

Furthermore, intraoperative correction maneuvers, additional osteotomies, and releases could improve kyphosis correction. Pesenti et al. [15] compared four correction maneuvers, including in situ bending, rod derotation, cantilever, and posteromedial translation techniques. They reported that the posteromedial translation improved and the cantilever technique significantly worsened postoperative kyphosis correction. Conversely, evidence on the effect of Ponte osteotomies and releases on postoperative sagittal correction is conflicting [16–18]; thus, these procedures are not routinely used in all patients because of the increased operative risk [21,22]. Nevertheless, because the surgical procedure was standardised in this study, the effects of correction maneuver, Ponte osteotomies and releases were not evaluated. The proposed T5–12 TK cutoff of <12.5° on fulcrum bending could be a viable indication to use Ponte osteotomies or releases to improve kyphosis restoration in patients at risk for hypokyphosis.

The effect of implant density on kyphosis restoration remains controversial. Previous studies have discovered that high implant density on the concave side could prevent the flattening of the recontoured kyphotic rod, which improved postoperative kyphotic outcomes [12,44,47]. Larson et al. [48] reported decreased TK in high implant density fusions at 1-year follow-up. This is in line with the present findings that a high convex side implant density increased the risk of hypokyphosis at the 2-year follow-up. This could be due to the compression of the convex anchors following rod insertion: a higher anchor density would lead to stronger compressive forces, causing more posterior column shortening and kyphosis loss postoperatively.

TK restoration is crucial in lordotic thoracic curves to prevent future degenerative changes. Newton et al. [49] studied 60 patients with AIS presenting with >7° lordosis treated with PSF. They demonstrated that higher postoperative kyphosis in those undergoing segmental screw fixation, Ponte osteotomies and those using rigid stainless-steel rods. The present study only identified 14 patients with preoperative lordosis, with one returning to normokyphosis postoperatively, resulting in insignificant differences in preoperative and surgical parameters between the hypokyphotic and nomokyphotic groups. Nonetheless, this finding demonstrates that TK normalization is possible in patients with lordosis using alternate pedicle screw constructs without additional osteotomies or releases, although unlikely.

This study has certain limitations. First, given the retrospective design, some patients underwent surgery in the 2000s. Although no significant change was noted in the surgical protocol and techniques throughout the years, improvements in surgical implants with more rigid fixation may affect sagittal correction. Furthermore, this radiographical study did not analyze correlations with clinical outcomes. Lung function tests are not routinely performed for patients with AIS in Duchess of Kent Children’s Hospital. However, patient-reported outcomes such as Scoliosis Research Society-22 scores were not introduced until the late 2000s and were not routinely assessed in patients undergoing surgery in the earlier years. Moreover, many patients were excluded because of unclear T5–12 TK on lateral view fulcrum bending films, as visualization of the coronal curve on fulcrum bending was the main concern. Finally, an arbitrary cutoff for thoracic hypokyphosis was used. This contradicts the increasing emphasis on using patient-specific TK targets calculated by spinopelvic parameters. However, this was limited by the older long cassette radiographs that did not fully expose the pelvis, hindering the measurement of these parameters. Future prospective studies can ensure clear visualization of T5–12 vertebral levels on lateral view fulcrum bending films, utilize spinopelvic parameters to set patient-specific kyphosis targets, and correlate radiographical outcomes with improvements in clinical and patient-reported outcomes.

To our knowledge, this is the first study that demonstrated the clinical utility of lateral view fulcrum bending radiographs to predict postoperative hypokyphosis. Low T5–12 kyphosis on fulcrum bending increases the risk of postoperative hypokyphosis at both the immediate and 2-year postoperative periods. A high convex side implant density was a risk factor for 2-year postoperative hypokyphosis, whereas other surgical factors did not affect kyphosis correction. The present findings suggest that T5–12 kyphosis of <12.5° on fulcrum bending is associated with a high risk of postoperative hypokyphosis; Ponte osteotomies, releases or a decrease in convex side implant density could improve kyphosis restoration in patients with hypokyphosis.