Reconstructed Three-Dimensional vs. Two-Dimensional Visualization in Unilateral Biportal Endoscopic Spine Surgery: A Randomized Crossover Simulation Feasibility Study.

BACKGROUND/

OBJECTIVES

Reconstructed three-dimensional (3D) endoscopic visualization is believed to improve depth perception and surgical performance compared with conventional two-dimensional (2D) systems. However, evidence supporting its use in unilateral biportal endoscopic (UBE) spine surgery remains limited. The objective was to compare procedural performance, task completion times, and subjective workload between 3D and 2D visualization in simulated biportal lumbar endoscopy.

METHODS

This randomized crossover pilot feasibility ex vivo study enrolled 6 neurosurgeons (3 residents, 3 seniors) who each completed a standardized UBE lumbar discectomy simulation (UpSurgeOn Lumbar SpineBox) under both 3D and 2D visualization in randomized sequence. The primary outcome was composite procedural performance (0-100 scale) assessed using a multi-item checklist encompassing technical success, quality, and safety. Secondary outcomes included task completion times for 7 procedural steps and subjective workload. Linear mixed-effects models adjusted for expertise level, spinal segment, operative side, and session order. Within-participant effect sizes (Cohen's d) and a sensitivity power analysis were calculated.

RESULTS

All participants completed both conditions (24 procedures). Composite performance was high with both modalities, with no significant difference for the unweighted (adjusted difference 0.64 points, 95% CI -3.67 to 4.96, p = 0.76, d = 0.13) or weighted score (1.67 points, 95% CI -4.04 to 7.37, p = 0.55, d = 0.26). Subjective workload also did not differ (p = 0.67, d = 0.19). Effect sizes were negligible to small for most outcomes but reached medium magnitude for subjective fatigue (d = 0.54) and situational stress (d = 0.56). Exploratory expertise-stratified comparisons suggested faster lamina drilling with 3D among seniors (p = 0.04) and slower dura identification with 3D among residents (p = 0.02), based on only 3 participants per subgroup. The sample afforded 80% power to detect only very large effects (d ≈ 1.43).

CONCLUSIONS

3D visualization was feasible and well tolerated but did not significantly improve procedural performance, task completion time, or subjective workload in UBE simulation. Because the study was powered to detect only very large effects, these nonsignificant results reflect limited power rather than equivalence, and the medium effect-size signals for fatigue and situational stress are hypothesis-generating. Adequately powered trials are needed to confirm task-specific or expertise-dependent benefits.