Are we achieving what the algorithm tells us? Analysis of lumbar pedicle subtraction osteotomies with pre-bent rods

Abstract

Precise restoration of sagittal balance is a critical goal in adult spinal deformity surgery. Computer-assisted planning allows for patient-specific alignment targets and rod pre-bending, theoretically improving the accuracy of surgical correction. However, the correlation between planned and achieved alignment goals using pre-bent rods remains unclear. The aim of this study was to evaluate the accuracy of alignment correction in patients undergoing lumbar pedicle subtraction osteotomy (PSO) using UNiD-derived pre-bent rods, and to compare software-generated preoperative alignment targets with actual postoperative radiographic parameters. A retrospective cohort study was performed of adults who underwent lumbar PSO with long-segment thoracolumbar fusion (≥ 6 levels) at a single academic center between 2018 and 2022. Inclusion criteria required UNiD preoperative planning, PSO performed at the planned level, use of patient-specific pre-bent rods, and complete radiographic data. Planned alignment targets were obtained from the UNiD platform and compared with immediate postoperative standing lateral radiographs. Absolute differences between preoperative-to-planned and preoperative-to-postoperative values were compared using paired t-tests. Effect sizes (Cohen’s d) were calculated and post hoc power analysis was performed, with primary focus on pelvic incidence (PI), sagittal vertical axis, pelvic tilt, and PI minus lumbar lordosis (PI-LL). Twenty patients (60% female, median age 66.8 years) were included. The planned PSO angle closely matched the achieved correction (mean −24.2° planned vs −24.02° ± 7.31° postoperative, p = 0.94). Lumbar lordosis and L4–S1 lordosis exceeded planned correction, with a significant but modest increase at L4–S1 (p = 0.03). Pelvic parameters demonstrated the largest deviations from plan. Pelvic tilt correction exceeded predictions by a mean of 8.97° ± 7.10° (p < 0.01); the sagittal vertical axis was undercorrected by a mean of 36.37 ± 48.30 mm (p < 0.01); and PI changed more than anticipated (p = 0.01). PI-LL improved substantially from a mean of 29.73° ± 15.76° preoperatively to −3.23° ± 10.95° postoperatively (p < 0.001). The planned and achieved L1 pelvic angle did not differ significantly. Computer-assisted planning with patient-specific rods accurately reproduced the intended PSO and segmental lumbar correction but did not reliably predict global sagittal parameters. These findings suggest that current planning tools might require refinement to improve the accuracy of predicted postoperative alignment using pre-bent rods and to minimize the risk of suboptimal outcomes.