Validation of a modular low-fidelity training model for biportal endoscopic spine surgery

Spinal endoscopy is associated with a steep learning curve, in part owing to its limited depth perception, which increases the need for training models. This study aimed to investigate construct and predictive validity of the Endobox as a low-fidelity training model for biportal endoscopy.

To assess construct validity, we compared the performances of expertsand novices across four Endobox exercises. Performances were evaluated based on completion time and number of attempts, using a 0º endoscope and rongeur. To assess predictive validity of the Endobox, 16 students (n = 8 per group) were randomly assigned into EB+ (twelve training sessions) or EB- (no training) before performing six simulated discectomies on a non-anatomical simulation model. Completion time, technical skill (modified version of the ASSET score), and cognitive demand (NASA-TLX score) were compared. Statistical analysis was carried out using parametric and non-parametric tests using Stata 18 for Mac.

Construct validity: experts were significantly faster in 3/4 exercises (2. p = 0.02, 3. p = 0.01, 4. p = 0.03), except in exercise 1 (p = 0.43). Experts needed fewer attempts for exercises 3 (p = 0.04) and 4 (p = 0.02), rated most difficult (9/10, 8/10 respectively). Predictive validity: the EB+ group was faster (p < 0.001) and improved across all observations, achieving higher scores for safety (p = 0.007), camera/bimanual dexterity, instrument handling, depth perception, and procedure quality (all p < 0.001). No difference in cognitive load was observed (p = 0.621).

The present study provides preliminary evidence that the Endobox model may develop skills that are relevant to, and potentially transferable to, a simulated procedure representing endoscopic spine surgery.