Agreement Between Automated Pedicle Screw Planning and Surgeon-Selected Implant Dimensions in Thoracic and Lumbar Spine Surgery: A Three-Year Single-Center Study
TL;DR
Its systematic diameter underestimation and broad pair-level differences support its use as an initialization and review aid, with final sizing retained as a surgeon decision, but ASP did not provide interchangeable estimates of surgeon-selected dimensions.
Abstract
Background: Automated pedicle screw planning (ASP) can initialize trajectories and implant dimensions, but correlation with implanted screws does not establish agreement. We quantified the association and agreement between ASP recommendations and actual intraoperative implant selection (AIS) and examined pathology, asymmetric planning, and available safety outcomes. Methods: This retrospective single-center study included 103 consecutive patients treated with thoracic or lumbar instrumentation from January 2021 to January 2024. The dataset contained 351 bilateral screw-pair records. Pearson and Spearman correlations, linear regression, and Bland–Altman analyses were supplemented by patient-clustered bootstrap 95% confidence intervals (CIs), pathology-stratified estimates, and a sensitivity analysis excluding a verified 94 mm iliac-fixation observation. Results: Diameter showed moderate correlation (Pearson’s r = 0.635, 95% CI 0.542–0.721; Spearman’s rho = 0.679, 95% CI 0.554–0.767; both p < 0.001). Length also showed moderate correlation (Pearson’s r = 0.675, 95% CI 0.565–0.754; Spearman’s rho = 0.600, 95% CI 0.450–0.714; both p < 0.001). For ASP minus AIS, the diameter bias was −0.75 mm (95% CI −0.86 to −0.64), with limits of agreement from −2.00 to 0.50 mm. The length bias was −0.15 mm (95% CI −1.05 to 0.79), with limits from −12.30 to 12.00 mm. The 94 mm value was verified as a genuine iliac-fixation observation; excluding it changed Pearson’s r only from 0.675 to 0.679. Asymmetric ASP length recommendations occurred in 115/351 pairs (32.8%), usually by 5 mm. Conclusions: ASP tracked overall implant-size trends but did not provide interchangeable estimates of surgeon-selected dimensions. Its systematic diameter underestimation and broad pair-level differences support its use as an initialization and review aid, with final sizing retained as a surgeon decision.