Pedicle screw (PS) constructs have become the standard instrumentation for posterior correction and spinal fusion of adolescent idiopathic scoliosis (AIS) [1]. PS instrumentation can improve three-dimensional spinal deformity correction and solid fusion to achieve the goal of surgical treatment of AIS [2]. In patients with thoracic AIS, PS constructs have been shown to improve radiographic outcomes compared with hook or hybrid constructs [3–5]. By providing rigid three-column fixation of the vertebra, PS constructs have been shown to obviate the need for anterior fusion, decrease the number of fusion levels, reduce complication and revision rates, and improve pulmonary function values [3,6,7].

PS density is defined as the number of PS per fused level. Many surgeons use high PS density (HD) constructs with maximum density of 2.0 PS per fused level despite limited data to show the optimal PS density constructs [8–10]. There are valid concerns of HD constructs, including a small but significant risk of PS malposition, increased cost, and postoperative hypokyphosis [8,11,12]. If low PS density (LD) constructs are able to improve curve correction comparable to HD constructs, using fewer screws would improve the efficiency and reduce the implant costs of scoliosis surgery. Although PS constructs have been commonly used in the treatment of thoracic AIS, the PS density for optimal outcomes is still controversial. Some studies have demonstrated larger radiographic correction with HD [8,9]. On the contrary, several studies have shown no correlation between PS density and radiographic correction [12–14]. There are many conflicting factors that can affect the surgical results on curve correction, including curve magnitude, curve flexibility, number and points of fixation, fixation apparatus, and posterior release procedure (i.e., Ponte osteotomy) [15,16]. To the best of our knowledge, no previous studies reported the impact of PS density on deformity correction in posterior spinal fusion (PSF) without the use of additional posterior release procedure especially for Ponte osteotomies (POs). Furthermore, the optimal PS density remains unknown, especially when sagittal and axial correction is considered.

This study aimed to determine whether the number of screws per fused level, or PS density, affects coronal curve correction, thoracic kyphosis, and apical vertebral rotation in PSF without POs of Lenke 1 and 2 AIS and investigate whether HD improves three-dimensional deformity correction at the final follow-up.

The study included 122 patients: 99 patients with Lenke 1curve type and 23 patients with Lenke 2 curve type. There were 106 female and 16 male patients with a minimum follow-up of 2 years (average, 40.1 months; range, 24–92 months). The demographic characteristics and clinical features of the patients are shown in Table 1. The mean±standard deviation age at surgery was 15.0±1.9 years. The median Risser grade of the patients was 3 (range, 0–5). The majority (54.9%) of patients had Risser grade ≥3. The mean preoperative major Cobb angle was 55.3°±7.2° and corrected to 18.3°±7.2°) at follow-up (p<0.01) for an average correction of 67.2%±11.7%. The mean preoperative T2–T12 kyphosis angle was 34.1°±8.8° and decreased to 31.6°±8.7° at follow-up (p<0.01). The mean preoperative RI, assessed by thoracic torsion and vertebral rotation, was 1.89±0.29 and corrected to 1.6±0.23 at follow-up (p<0.01) for an average correction of 17.4%±10.1%. There were no significant correlations between PS density and the following parameters: percent MCC (r=0.10, p=0.26) and CI (r=0.16, p=0.07), change in T2–T12 kyphosis angle (r=−0.13, p=0.14), and percent RI correction (r=−0.09, p=0.37) (Fig. 4). There was no correlation between PS density and absolute degree of Cobb angle correction (r=0.06, p=0.51). Moreover, there was a significant correlation between curve flexibility and percent MCC (r=0.24, p=0.0086). For subgroup correlation analyses, there were no significant correlations between PS density and the following parameters: percent MCC and CI, change in T2–T12 kyphosis angle, and percent RI correction in regard to Lenke curve type (Lenke curve type 1 and Lenke curve type 2), patient age (<15 years and >15 years), Risser grade (Risser grade <3 and Risser grade ≥3), and BMI (BMI <25 kg/m² and BMI ≥25 kg/m²) (Tables 2–5).

For subgroup analysis, the patients were divided into two groups according to the mean PS density for the entire study. The LD group, consisted of 57 patients, was defined as the use of <1.50 screws per fused level, and the HD group, consisted of 65 patients, was defined as the use of 1.50 or more screws per fused level. There were no differences in demographic and preoperative parameters between the two groups (Table 6). The mean PS density was significantly lower in the LD group (1.3 versus 1.7, p<0.01). The number of fused levels was significantly lower in the HD group (10.6 versus 11.6, p<0.01). The number of PSs per construct was lower in the LD group (15.5 versus 17.5, p<0.01). There were no significant differences in the percentage of rod diameter and rod material use, operative time, and estimated blood loss (Table 7). Statistical subanalysis revealed that there were no significant differences between the HD and LD groups with respect to MCC and CI, change in T2–T12 kyphosis angle, and RI correction. The HD group had less magnitude of postoperative thoracic kyphosis (30° versus 33°, p=0.03) (Table 8). Furthermore, the comparison between patients with PS density ≤1.3 (mean PS density of the entire study–1SD; n=21) and patients with PS density ≥1.7 (mean PS density of the entire study+1SD; n=21) confirmed that there were no significant differences between the two groups in terms of MCC and CI, change in T2–T12 kyphosis angle, and RI correction (all p>0.05). At the latest follow-up, the SRS-30 scores were available in 82 (67.2%) of 122 patients. SRS-30 scores were available in 40 (70.2%) of 57 patients in the low-density group and 42 (65%) of 65 patients in the high-density group. There were no significant differences in all five domains studied and the overall score between those of the LD and HD groups (Table 9).

No intra- or postoperative instrumentation-related complications were noted. There were three adding-on phenomena with required revision surgery in three patients, one patient (1.8%) in the LD group and two patients (3.1%) in the HD group (p=1.0). In the LD group, an adding-on phenomenon was observed in a 16-year-old girl with an increased fractional curve below the arthrodesis, performed down to T12. A revision procedure extending the arthrodesis down to L3 was performed at 2 years after the index procedure. In the HD group, the first adding-on phenomenon was observed in a 14-year-old boy after T3–L1 arthrodesis. The patient underwent extending arthrodesis down to L4 at 1 year after index procedure. In another patient, an 11-year-old girl, an adding-on phenomenon was observed after T4–L1 arthrodesis. There was progression of curve at both upper and lower arthrodesis segments. The patient underwent revision extending arthrodesis from T2 to L4 at 3 years after the index procedure. No subsequent complications were noted in all three patients undergoing revision surgery. They were doing well at the latest follow-up. Three patients in the LD group and three patients in the HD group (5.3% versus 4.6%, p=0.72) had curve progression >5° at the time of latest follow-up. However, they had no substantial back pain or decline in their function at that time. Their SRS-30 scores, the overall score and all five domains, were >3 at the final follow-up. No patients needed additional operation. At the latest follow-up, there were no adding-on phenomenon in this cohort. The rates of proximal junctional kyphosis (PJK) were similar between the LD and HD groups (LD group, 8.8%; HD group, 10.8%; p=0.48) [20]. No patients required revision surgery due to PJK. There was no screw loosening or instrument failure in this study. No patient had secondary operation at the final follow-up.

The present study investigated whether higher PS density improved three-dimensional deformity correction in PSF without POs of flexible, moderate-sized Lenke 1 and 2 AIS. There were no significant correlations between PS density and percent MCC and coronal CI, change in T2–T12 kyphosis angle, and percent RI correction. Furthermore, subgroup analysis confirmed that LD groups had equivalent radiographic and clinical outcomes as HD groups.

We obtained mean percentage MCC of 67.2% in 122 patients in our cohort, which is comparable to other studies [8,9,12,13]. We have considered preoperative curve flexibility, which has been shown to predict curve correction [21]. We have found no correlation between PS density and MCC with or without curve flexibility. Our results were in concordance with the findings of Bharucha et al. [12] and Quan and Gibson [13]. On the contrary, Clements et al. [8] found a significant correlation between MCC and implant density, but the authors included the patients from all six Lenke curve types and also included variable types of the anchors (wires, hooks, and screws) in their study. In a large multicenter database study, Larson et al. [9] found that increased anchor density was associated with increases in coronal curve correction for Lenke 1 and 2 curves. Unfortunately, this study did not account for preoperative curve flexibility and the use of POs. In recent literature, the use of POs performed in thoracic AIS curves has been investigated and found to significantly improve coronal curve correction [15,16]. In the sagittal plane, PS constructs have been shown to provide better thoracic hypokyphotic correction in the previous study [22]. On the contrary, several recent studies have reported that PS constructs decreased magnitude of thoracic kyphosis [9,10,12,13]. Additionally, some authors suggested that higher implant density was associated with decreased thoracic kyphosis [8,9]. We found loss of thoracic kyphosis in the entire cohort postoperatively. However, the magnitude of postoperative kyphosis was in physiologic range. Furthermore, the loss of preoperative to postoperative kyphosis was equivalent between HD and LD groups. We believe that this may be due to the hypokyphotic nature of the deformity in adolescent idiopathic thoracic scoliosis. Thoracic kyphosis alignment is sacrificed to achieve more coronal plane correction [13].

There are very few evidences on vertebral rotation and rib hump deformity correction regarding PS density in scoliotic treatment. PS constructs have been shown to improve vertebral rotation correction and lessen rib hump deformity compared with hook-rod instrumentation [23]. Kuklo et al. [24] assessed the correlation between rib hump deformity and apical vertebral rotation and found the good and high significant correlations between these two parameters. We used rib hump, in terms of RI method, to measure vertebral rotation for patients who did not undergo costoplasty. We found no significant correlation between vertebral rotation, in terms of RI correction, and PS density. Our results are supported by previous studies. Yang et al. [25] evaluated vertebral rotation correction indirectly from photographs of the rib hump. The authors found no significant correlation between implant density and vertebral rotation assessed by photographic parameters. Bharucha et al. [12] evaluated thoracic angle of trunk rotation using a scoliometer. They also found no correlation between implant density and thoracic angle of trunk rotation at 2 years postoperatively.

The present study also reported the change in percent MCC from immediate postoperative radiographs to final follow-up with an average loss of correction of 5% for the entire cohort. Our results are similar to the previous studies, which reported loss of correction rate ranging from 2% to 5% in patients with AIS who underwent PSF with PS constructs [26–28].

In this study, we focused on three-dimensional deformity correction in PSF without the use of POs in thoracic idiopathic scoliosis. All patients underwent a similar posterior surgical release procedure. The exposure was performed with a midline incision and extending to the tips of transverse process. No additional posterior surgical release was employed. We confined the anchor type of the study groups that used only predominantly PS constructs, and all patients only had thoracic scoliotic curve (Lenke 1 and 2 curve types) to render homogeneity of the study. We also accounted for the influence of flexibility on the coronal correction. Our results are supported by previous studies by Cheung et al. [21] and Quan et al. [13]. As regard the biomechanical perspective, Wang et al. [29] and Le Naveaux et al. [30] revealed that low-screw density construct provided similar three-dimensional correction as the higher-screw density construct among patients who underwent PSF for thoracic AIS. Finally, we believe that implant density at some specific regions of the construct, i.e., at the apex and both ends, may be more crucial than overall implant density for three-dimensional deformity correction in thoracic AIS [31,32].

This study has some limitations. First, its retrospective nature makes it vulnerable to various biases. A randomization of anchor density would more adequately address the confounding factors especially for surgeon bias. Second, four surgeons were involved, and the correction techniques were not standardized. Further studies should take into account for identical correction technique so as to remove this confounding factor. Third, our sample size was small. This decreased the power for statistical analysis and undermines the conclusion that there was no correlation between PS density and amount of correction. Fourth, the problem within the fusion area (such as failure of fixation or pseudarthrosis) is one of the major concerns following AIS surgery and may affect surgical outcomes. However, we did not implement the methods and results for evaluating the fusion status in this study. Further study should evaluate the fusion rate or pseudarthrosis after AIS surgery. Although not the primary objective of this study, one of the important issues for thoracic AIS correction is shoulder imbalance correction. However, a previous study found no correlation between increased implant density and shoulder balance correction [25]. The correlation between implant density and shoulder imbalance needs to be further assessed in the future. Finally, the majority (54.9%) of patients in our cohort has Risser grade ≥3 so this may limit the generalizability of our results for patients with less skeletal maturity.

Our study showed that there was no significant correlation between PS density and amount of coronal, sagittal, or axial plane correction. LD constructs may be used as effectively as HD constructs for patients with flexible, moderate-sized thoracic AIS undergoing PSF without the use of POs.