Seventy-five patients with scoliosis who underwent posterior instrumentation and fusion at Seirei Sakura Citizen Hospital, Chiba, Japan, from January 2010 to August 2011, were identified. Fourteen patients were excluded: 9 patients with congenital deformity and 5 with idiopathic scoliosis and no thoracic curve. Of the 75 patients, 61 with a thoracic curve were included in the study. All surgery was performed by two senior surgeons. Two hundred and twenty-two pedicle screws were implanted in 29 patients (6 males and 23 females) using CT-based navigation from Jan 2010 to Oct 2010 (group C) and 416 screws were implanted in 32 patients (5 males and 27 females) using O-arm navigation from Nov 2010 to Aug 2011 (group O). The etiology of scoliosis was mostly idiopathic; however, neuromuscular etiology was identified in 2 group C patients. In group O, mesenchymal and neuromuscular etiology was identified in 1 and 2 patients, respectively. The average Cobb angle of the major curve was 59.2°±11.4° (range, 45°-88°) in group C patients and 54.4°±10.7° (range, 43°-90°) in group O patients (mean±SD). The average numbers of pedicle screws were 7.7±1.7 (range, 6-14) in group C patients and 13.0±4.2 (range, 7-23) in group O patients. The average number of pedicle screws in group O patients was statistically greater than that in group C patients. There were no significant differences in the mean ages at surgery or the mean Cobb angles of the major thoracic curve between groups C and O patients (Table 1).

All patients were operated on by the same team of surgeons (T.K, T.A., and S.M.) using a 6.3 mm titanium rod and screw and hook instrumentation. Posterior correction surgery was performed using spinal cord monitoring for all the scoliosis surgeries. In group C patients, 3 to 4 pedicle screws were used at the apex on the convex side for the correction. Hooks were placed at the upper end of the fusion area, and pedicle screws were placed at the lower end of the fusion area. All screws were inserted using a navigation system. Applying a de-rotation force with convex pedicle screws, a left rod rotation maneuver was performed and 3 ultra-high molecular weight polyethylene tapes (Nespron cable system, Alfresa Pharma, Osaka, Japan) were tied to the concave rod. In group O patients, 3 or 4 pedicle screws were used at the apex on the concave side instead of the ultra-high molecular weight polyethylene tapes because we could confirm the positioning of the screw relative to the smaller concave pedicles during surgery using O-arm navigation.

A navigation system (Stealth Station, Medtronic, Sofamor Danek, Memphis, TN, USA) was used for correct screw placement and fixation of the thoracic and lumbar spine. The procedure for group C patients employed more than 6 registration points for 3 consecutive vertebrae, to shorten the surgical time (multilevel registration) [19]. Diameters and lengths of all pedicles into which the screws would be inserted were measured on preoperative 3D CT images. In group O patients, after a reference clamp was applied to the spinous process of the vertebra, CT data were obtained with automatic registration by 3D fluoroscopy (Fig. 1). After the navigation system was ready, an entry point was determined using a navigation probe, and a burr was used to make an entry hole. We reconfirmed the screw trajectory using a navigation probe, inserted a curved pedicle probe, and then inserted pedicle screws in some vertebrae above and below the vertebra on which the navigation reference frame had been placed. Reconfirmation of the accuracy of navigation was used as a reference for the vertebrae that was apart from the frame, by touching an anatomical landmark with the probe. If we suspected errors between the images on the O-arm system and the real landmarks, we obtained renewed intraoperative data for navigation [20]. The diameters of the pedicle screws were 4.5 mm, 5.5 mm, or 6.5 mm. The diameters and lengths of all pedicle screws were determined intraoperatively using the O-arm images. When pedicle perforation was detected before screw insertion, screw insertion was not performed. The total time from registration until complete insertion of all of the pedicle screws was measured in both groups. In group C patients, the measurement time started from the point merge, which was set by touching the anatomic landmarks. In group O patients, the measurement time started from the beginning of the scanning of the vertebra. Subsequently, the average time needed to insert one screw, including registration, was calculated.

The accuracy of all pedicle screw placement was evaluated one week after surgery using reconstructed CT scan images at each screw axis. As defined by Neo et al. [21], pedicle screw positions were classified into 4 grades: grade 0, no perforation with the screw completely contained in the pedicle; grade 1, perforations <2 mm; grade 2, perforations ≥2 but <4 mm; and grade 3, perforations ≥4 mm, with grades 2 and 3 representing perforation (Fig. 2). Based on this grading system, the screw misplacement rates were determined.

Statistical analysis was performed using PASW ver. 18.0 (SPSS Inc., Chicago, IL, USA). Basic descriptive statistics (chi-squared and Fisher exact tests) were used when appropriate. Comparisons between the groups were made using a Mann-Whitney U test with a significance level set at 0.05.