Patient cohort

This was a single-centre retrospective comparative cohort study of patients with severe and complex ASD who underwent posterior spinal fusion and 3-CO between January 2020 to January 2023. All surgeries for the cases were performed by the same surgeon with 20 years of experience in treating complex spinal deformities. The inclusion criteria were as follows: (1) patients with scoliosis, kyphosis, or kyphoscoliosis undergoing posterior spinal fusion surgery; (2) the major curve Cobb angle > 80° in coronal plane and/or sagittal plane; (3) the flexibility of spinal deformity ≤ 25% or concomitant congenital/iatrogenic spinal fusion requiring 3-CO; and (4) etiology of congenital, tubercular, traumatic, iatrogenic or ankylosing spondylitis. All patients were followed up for a minimum of 12 months. The exclusive criteria were: (1) patients younger than 18 years old; (2) patients who could not tolerate the surgery; (3) patients who only undergo posterior column osteotomy; and (4) follow-up data was incomplete. Personalized computer-assisted three-dimensional osteotomy simulation was performed for all recruited patients; additional 3D-printed guide template for osteotomy based on the surgical planning were used for patients who were admitted to hospital after November 2021. Therefore, this cohort was divided into template group and non-template group, and the application of 3D-printed guide template was not randomized. This study was approved by the Research Ethics Committee of Beijing Chao-Yang Hospital and informed consent was taken from patients.

Surgical simulation and executionThree-dimensional spine model reconstruction

Patients’ CT scan data of the whole spine were collected with Digital Imaging and Communications in Medicine (DICOM) format (DICOM format data from Siemens CT machine, SOMATOM Sensation 16, Siemens AG, Forchheim, Germany). All the tomographic pictures were imported into Mimics Medical 21.0 (Materialise NV, Leuven, Belgium), and 3D spine model was established with threshold of 226-3071HU.

Osteotomy simulation

3-CO and correction procedures were simulated in 3-Matic Medical 13.0 (Materialise NV, Leuven, Belgium) with cut tool and trim tool in 3D spine model. After trimming the targeted elements, the rotation tool was used to achieve the osteotomy gap closing procedure. The simulation of PSO and VCR were accomplished following the literature and surgeon’s experience. For patients with kyphoscoliosis, correction simulation was performed after sagittal procedure. The planning osteotomy angle was assessed according to the angle change of osteotomy gap in pre- and post-simulation 3D spine model.

Osteotomy guide template design and manufacture

The osteotomy guide plate will consist of three parts, including the adjacent segment pedicle screw implantation guidance part, the lamina surface fitting part, and the osteotomy trajectory guidance part, which has been described in detail in our previous study [1]. Firstly, in Mimics Medical 21.0 (Materialise NV, Leuven, Belgium), the Lasso command was used to fill and repair the bone structure of the spinal model to build a repaired Mask for spine structure. Then, the Cylinder command was used to draw a cylindrical structure that mimics the pedicle screw implantation, and adjust the cylindrical trajectory structure to the optimal position for internal fixation through three-plane views of vertebrae, thereby completing the screw trajectory design of adjacent fixed segments. Import the spinal modification model and pedicle screw trajectory data into the 3-Matic Medical 13.0 (Materialise NV, Leuven, Belgium) software in STL format for osteotomy guide plate fabrication. Use a Wave Brush Mask to draw on the simulated pedicle screw cylindrical structure and nearby lamina surface (including osteotomy segments) to obtain personalized anatomical markers attached to the osteotomy and screw placement segments. Then use the Uniform Offset tool to deviate the contour area to create an attached guide plate model. Afterwards, the length of the osteotomy guide plate was measured using 3-Matic Medical 13.0 (Material NV, Leuven, Belgium), and the guide rail was drawn using ProE (Parameter Technology Corporation, Boston, MA, USA) software, which will serve as the osteotomy trajectory for the ultrasound osteotome. After these programs are completed, position calibration is performed using Mimics Medical 21.0 (Materialise NV, Leuven, Belgium) to ensure that the produced osteotomy guide rail overlaps with the osteotomy plane. Use the “Cut Orthogonal to Screen” command to cut the appropriate length and use the Boolean command to integrate it with the previous guide template. Finally, the computer-designed osteotomy guide plate template is 3D printed using resin as the printing material to create a solid guide plate template. The 3D spine model should also be 3D printed at the original 1:1 scale, and the 3D printed osteotomy guide template will be tested on the 3D printed spine model.

Surgical procedure

The pedicle screw trajectory template was placed firstly on the surface of the lamina and spinous process, ensuring firm bone contact. Specific anatomical landmarks could warrant the perfect match of the 3D-printed guide template with patients. The Kirschner wires were inserted into the pedicle through the foraminule in the guide template to fix the guide template. Then the ultrasonic osteotome was used to resect the lamina along the guide rail of the 3D-printed osteotomy guide template. Next, the 3D-printed guide template is removed and conventional PSO or VCR procedures were performed with a temporary rod. The osteotomy gap was gradually closed by repeated manual compression; however, overshortening of the spinal cord should be avoided. In the non-template group, the 3D spine model would be used to assist the osteotomy procedure. Two representative cases were demonstrated in Figs. 1 and 2.

Fig. 1

A 22-year-old male with congenital kyphoscoliosis undergoing three-column osteotomy at T12 and posterior spinal fusion surgery. (A) personalized computer-assisted three-dimensional simulation; (B) Execution of simulation through three dimensional-printed guide template; (C) The pre- and postoperative full-spine radiography

Fig. 2

A 34-year-old female with congenital scoliosis undergoing three-column osteotomy at T5 and posterior spinal fusion surgery. (A) personalized computer-assisted three-dimensional simulation; (B) Execution of simulation through three dimensional-printed guide template; (C) The pre- and postoperative full-spine radiography

Data collection

The baseline clinical characteristics of the patients were collected, including age, gender, BMI, and diagnosis. The operating time (ORT), estimated blood loss (EBL), fusion levels, the location of 3CO, and the deviations as well as the match ratio of simulated and executed osteotomy angle in the coronal/sagittal plane were collected. Executed osteotomy angle was evaluated in the postoperative CT scan. Events of intraoperative neuromonitoring signal decline, postoperative neurological deficit, and postoperative non-neurological complications were recorded. Neurological deficit was defined as the decline of lower extremity motor score (LEMS) after surgery compared with the baseline LEMS. For radiographic parameters, the preoperative and postoperative main curve Cobb angle, FK, lumbar lordosis (LL), pelvic tilt (PT), sacral slope (SS), sagittal vertical axis (SVA), distance between C7 plumb line and centre sacral vertical line (C7PL-CSVL), and pelvic incidence (PI) were measured on the standing full-length spine radiographs. The health-related quality of life for patients in the template group was also recorded using Scoliosis Research Society-22r (SRS-22r).

Statistical analysis

All statistical analyses were performed utilizing SPSS 24.0 (Chicago, IL, USA). To minimize the effects of confounding factors, the propensity score matching (PSM) was performed with a match tolerance of 0.01 according to the age and gender, based on previous literature [19, 20]. A 1:1 neighbour matching algorithm was applied with a calliper width of 0.2 in patients in the template group and non-template group. After matching, the absolute standardized mean difference (ASMD) was used to judge the balance of each covariate between groups. An ASMD ≤ 0.1 could be considered as balance between groups. The Shapiro-Wilk test was performed to determine whether continuous variables had a normal distribution. Continuous variables with a normal distribution are presented as the mean ± standard deviation; otherwise, the median and interquartile range are used. The counts and percentages are presented for categorical variables. For the comparison of continuous variables between groups, the independent-sample t-test or nonparametric Mann-Whitney U test was performed; for categorical variables, the Pearson chi-square test or Fisher’s exact test was applied. A two-sided P value of less than 0.05 was considered statistically significant.