AB110. SOH25_AB_231. The learning curve of robotic-assisted lumbar fusion—individual patient data meta-analysis

Background: Current learning curve estimates originate from low quality investigations with significant methodological shortcomings and frequently describe only a single surgeon’s progression. The true learning curve of robotic-assisted lumbar fusion (RA-LF) remains uncertain.

Methods: Individual patient data from studies reporting RA-LF outcomes by case number were combined with local data obtained via retrospective review. Bayesian mixed-effects nonlinear and logistic meta-regressions were used to define learning curves.

Results: Data from 14 studies were extracted and combined with local records. This analysis comprised 768 patients from 19 surgeons. The expected operating time for the first case was 212.4 minutes [95% confidence interval (CI): 189.7–235.2], with a plateau time of 109.7 minutes (95% CI: 93.0–124.8). 39.6 cases (95% CI: 34.7–45.8) were needed to overcome 80% of the initial time deficit, with near-plateau performance expected after 113.2 cases (95% CI: 108.1–119.3). The expected accuracy during first pedicle screw placement was 91.2% (95% CI: 87.0–95.1%) with 95% accuracy reached after 238.0 screws (95% CI: 187.5–249.5). 1,408.4 screws were needed for 90% of surgeons to achieve this standard. The risk of surgical complications during the first case was 3.08% (95% CI: 1.00–5.78%), with no obvious learning curve detected.

Conclusions: Improvements in operating time and pedicle screw accuracy are more sustained and likely continue for longer than commonly quoted in existing literature. No difference was found in complication risk at the beginning and end of the learning curve.

Keywords: Fusion; learning curve; pedicle screw; robotic; spine