The learning curve of robotic assisted pedicle screw placement: individual patient data meta-analysis.

BACKGROUND CONTEXT: Robotic-assisted pedicle screw placement (RA-PSP) has grown popular yet is technically challenging. Current estimates of the learning curve of RA-PSP are derived from low-quality investigations and are generally limited to single-surgeon experiences.

PURPOSE: To estimate the learning curve of short-segment (1 or 2 level) RA-PSP, focusing on operating time, screw accuracy, blood loss, and complication rates.

STUDY

DESIGN

Individual patient data meta-analysis.

PATIENT SAMPLE: Patients who have received RA-PSP for treatment of degenerative lumbar pathology.

OUTCOME MEASURES: Primary outcomes included operating time, screw accuracy (Gertzbein-Robbins Grade A), blood loss, and complication rates. Learning curve thresholds were defined as cases or screws required to achieve prespecified performance improvements.

METHODS

Individual patient data were extracted from 14 studies identified from a 2024 systematic review and combined with local retrospective data. Studies were included if they reported outcomes after 1-2 level RA-PSP stratified by case number and referable to an individual surgeon. Abstracts, animal/cadaveric studies, non-English articles and simulator studies were excluded. Two reviewers independently screened articles with disagreement resolved via consensus. Bayesian mixed-effects nonlinear and logistic regressions were applied. Best performing models where selected via cross validation.

RESULTS

The analysis included 824 patients from 17 surgeons. Initial operating time was 212.5 minutes (95% CI, 168.3-259.5), reducing to 96.7 minutes (95% CI, 55.4-128.2) at the plateau. Half of surgeons achieved 80% of the initial time improvement after 46.0 cases (95% CI, 44.0-49.0), with plateauing after 136 cases (95% CI, 131.0-142.0). Screw accuracy improved from 91.0% (95% CI, 87.0-95.1) to 95% after 217 screws (95% CI, 197.0-266.0). Blood loss reduced by 80% after 65 cases (95% CI, 62.0-79.0), plateauing at 196.5 cases (95% CI, 189.0-236.0). Complication rates showed no significant learning curve.

CONCLUSIONS

The learning curve for RA-PSP is longer and more variable than previously reported. Approximately 60 cases may be necessary for half of surgeons to overcome the bulk of the learning curve however to ensure at least 90% of surgeons complete 80% of the learning curve may necessitate a large number of cases. Patient safety implications remain unclear due to a modest complication rate throughout the learning curve and the unknown consequences of minor pedicle screw misplacement.