This retrospective review of a radiological database of patients with AIS who underwent posterior corrective surgery was approved by the Research Ethics Committee of Osaka University Hospital (no. 15098-5). The research ethics committee of our institution waived receipt of written informed consent, because all clinical and radiographic interventions in the study followed routine assessments and the study was retrospective. Instead, the patients were allowed to opt out of the study based on the research information published on our institution’s website.
Patients and surgical procedure
We enrolled all consecutive female AIS patients who had undergone posterior corrective fusion between February 2017 and August 2019 and whose age at the time of surgery was between 10 and 20 years (n = 30). Three patients who underwent computed tomography (CT) scans without a hydroxyapatite phantom were excluded. Thus, 27 patients were included in this study. The median age at the time of surgery was 16 years (interquartile range [IQR], 14–19 years).
All surgeries were performed through the conventional posterior spinal approach under general anesthesia and neuromonitoring. After exposure of posterior spinal bony elements, pedicle screws were inserted into as many pedicles as possible. If the pedicle screws could not be inserted due to anatomical problems, hooks or sublaminar tapings were substituted. A Grade 1 osteotomy (resection of inferior facet) was performed in each fusion segment and a Grade 2 osteotomy (resection of both inferior and superior facets and ligamentum flavum) was performed in each rigid segment where segmental flexibility was not confirmed in preoperative traction and side-bending plain radiographs. After instrumentation, we corrected coronal and sagittal alignments and directly rotated vertebrae to correct 3-dimensional deformities. The autologous local bone graft and hydroxyapatite were transplanted on the decorticated lamina and articular surfaces. Neither iliac bone grafts nor bone morphogenic proteins were used in any of the cases. The patients were banned from participating in sports activities for 1 year after surgery.
Patients’ demographic data
From medical charts, we obtained each patient’s age at the time of surgery, preoperative body mass index, and levels of lowest instrumented vertebrae.
Each patient’s type of scoliosis was classified according to the Lenke classification on the basis of preoperative, full-length, standing, posteroanterior, and lateral radiographs . Preoperative and 1-year postoperative Cobb angles of the main thoracic (MT) and thoracolumbar/lumbar (TL/L) curves and preoperative Risser grades were measured on a flat-panel monitor at our hospital using built-in imaging software (SYNAPSE 5; FUJIFILM Medical Systems, USA, Inc., Lexington, MA).
The disc wedging angle, which was defined as the angle between the upper and lower endplates adjacent to the disc (left open, +), was measured in every disc below a lowest instrumented vertebra (Fig. 1). In addition, a disc wedging angle index (DWAI), which was defined as the sum of the upper and lower disc wedging angles adjacent to a vertebra, was calculated in every vertebra within the non-instrumented lumbar spine (Fig. 1). The perioperative change in DWAI (ΔDWAI = 1-year postoperative value − preoperative value) was also calculated (Fig. 1).
The patients underwent routine CT scans about 1 week and 1 year postoperatively for the purpose of detecting mispositioning of instrumentation or confirmation of bone union. CT images were acquired using 1 of 2 scanners (Discovery CT750 HD, GE Healthcare Japan, Tokyo, Japan, or Aquilion ONE, Toshiba Medical Systems Corporation, Tochigi, Japan). The scans used a slice thickness of 0.625 mm with the Discovery CT750 HD and of 0.5 mm with the Aquilion ONE; a tube voltage of 120 kVp; a matrix of 512 × 512; and a standard algorithm. The tube current was maintained by an automatic exposure control system. Each CT scan imaged together a standardized spine phantom consisting of 5 rods containing 0, 50, 100, 150, and 200 mg/cm3 hydroxyapatite (HA) (B-MAS 200; Kyoto Kagaku Co., Ltd., Kyoto, Japan). Multiplanar reconstruction was performed by our institution’s built-in 3-dimensional imaging software (Synapse Vincent; FUJIFILM Medical Systems, USA, Inc., Lexington, MA). From the 1-week and 1-year postoperative CT scans, each vertebra within the non-instrumented lumbar spine was superimposed automatically by the built-in application of the Synapse Vincent (Fig. 2); then, spherical regions of interest (ROIs) for measurements of Hounsfield Unit (HU) values were set in each vertebral body, excluding the cortical margin on the superimposed 3-dimensional images so that the ROIs should be in the same area on both scans. The centers of the spherical ROIs were set in each vertebral body as follows (Fig. 3):
- At the center of each vertebral body (ROI representing the whole of each vertebral body);
- At the center of the right half of each vertebral body (ROI representing the right half of each vertebral body); and
- At the center of the left half of the vertebral body (ROI representing the left half of each vertebral body).
The BMD (mg/cm3 HA) of each ROI was calculated by substituting the HU values into the linear regression equation obtained from the measured values of the phantom in each scan. The laterality index (LI) of BMD in each vertebral body was calculated by the following equation:
LI = (BMD of right half of the vertebral body) / (BMD of left half of the vertebral body)
The postoperative change in the LI (ΔLI = 1-year postoperative value–1-week postoperative value) was also calculated.
The statistical analysis was performed using IBM SPSS Statistics Version 25 (IBM, Armonk, NY, USA). The Wilcoxson signed rank test was used to compare Cobb angles of preoperative and 1-year postoperative MT and TL/L curves and 1-week postoperative and 1-year postoperative BMDs. Spearman’s rank correlation coefficient was used for a correlation analysis. Differences were considered statistically significant at P values < 0.05.