The learning curves of a validated virtual reality hip arthroscopy simulator

The increasing utilisation of more technically demanding surgical procedures, coupled with decreases in trainees’ caseloads and working hours, has led to difficult learning curves in orthopaedic surgical training [1‐7]. This is particularly true for the rapidly expanding field of hip arthroscopy. The geometry of the joint, combined with its thick soft-tissue envelope, makes hip arthroscopy particularly challenging and long learning curves have previously been described [3, 8‐10]. Concerns regarding patient safety, fuelled by evidence of worse outcomes when arthroscopic procedures are performed by inexperienced surgeons, mandate the development of an additional means of training arthroscopic surgeons [3, 9].

Though virtual reality (VR) simulation training has been increasingly adopted by other surgical specialities, its use in arthroscopic surgical training remains limited [5, 11]. This slow uptake is, in part, the result of a lack of robust evidence as to the benefits of VR training to both the trainee and the patient [5, 11]. One means of demonstrating these benefits to the trainee is through the analysis of the learning curves associated with simulator training—with repeated practice reflecting measurable increases in validated performance metrics. Though virtual reality simulators have been suggested to shorten surgical learning curves, there are currently only reports in the literature on box simulators and low fidelity simulators [11‐13]. Without demonstration of learning curves, it cannot be assumed that the visuo-haptic feedback provided by the VR simulator is sufficiently consistent to allow skill development.

The purpose of this study was to investigate how novice performance on a VR hip arthroscopy simulator’s visualisation module changed with repeated practice. We hypothesised that repeated use of the simulator would lead to significant decreases in time taken, number of collisions and distance travelled by the arthroscope before reaching a plateau, thus demonstrating learning curves.

The mean age of the students was 21.7 ± 1.8 and the male to female ratio was 19:6. 12% of students were in their third year of study, 48% were in their fourth year of study and 40% were in their fifth year of study. The mean number of times using an arthroscope (simulated or real) amongst the students was 0.3 ± 0.7 and no student had ever used a hip arthroscope (simulated or real).

Learning curve were demonstrated by the students, with significant improvements in all parameters following training (Fig. 2). Average total time decreased by nearly 75% across the course—936 s (657–1688) to 261 s (207–379) (p < 0.0001) (Table 2). The first significant increase from the participants baseline was seen in session 3 (p = 0.024) and no significant difference from session 7 (peak performance) was first found at session 5.

Collisions between the arthroscope and soft-tissues decreased across the sessions by nearly 90%—91 (44–126) to 10 (5–27) (p < 0.0001). The first significant improvement from baseline was found at session 3 (p = 0.013) with no significant differences from session 7 found at session 6. Similarly, collisions between the femoral head and the arthroscope also decreased—34 (25–79) to 7 (4–9) (p < 0.0001). These improvements were first seen in session 4 (p = 0.0007) and no significant difference from session 7 was first found at session 6.

Total time of soft-tissue collision was found to decrease from 527 s (369–994) to 111 s (94–144) (p < 0.0001), with the first significant decrease seen in session 4 (p = 0.0003) and no significant difference from session 7 was first found in session 5. Length of femoral head scratches showed similar decreases—190 mm (120–285) to 34 mm (22–57) (p < 0.0001), with the first significant decrease in session 5 (p = 0.005) and no significant difference from session 7 first found in session 6.

The most important finding of the present study was the demonstration a learning curve for a visualisation module on a previously validated hip arthroscopy simulator [15]. Significant increases were seen in six simulator measured parameters after a minimum of three sessions, before a plateau was reached after a maximum of six sessions. These results mimic those of similar studies into the learning curves associated with virtual reality orthopaedic simulators, and support its potential use in the development of basic hip arthroscopic skills such as visualisation [13, 18‐24]. The assessment of learning curves, or ‘training effects’ is an essential aspect of VR simulator validation. It demonstrates that the VR environment generated by the simulator has sufficient visuo-haptic consistency, enabling individuals to develop psychomotor skills within it. Without demonstration of this training effect, it cannot be assumed that the visuo-haptic feedback provided by the VR simulator is effective in enabling dextrous development.

Previous studies have demonstrated learning curves and training effects for other arthroscopic simulators. Pollard et al. demonstrated a learning curve for lateral and supine patient positions in simulated hip arthroscopy using a box simulator, showing improvements in time taken and efficiency of movement [13]. Similar learning curves have been demonstrated using the Sheffield Knee Arthroscopy Training System and a passive haptic knee arthroscopy simulator with medical students [19]. A particularly difficult learning curve for knee arthroscopy was noted in a similar study by Rahm et al. using a passive haptic knee arthroscopy simulator [23]. Additionally, a learning curve has been demonstrated for the insightMIST (GMV, Madrid, Spain) shoulder VR simulator, further supporting shoulder VR simulators’ implementation in surgical training [20]. Furthermore, the retention of the skills acquired during simulation has been investigated by monitoring performance over a more protracted period, demonstrating limited degradation over time [18, 21].

A small number of studies have assessed the real-world benefits of VR simulators, showing improved performance in an operative setting following simulation training [5, 11]. Cannon et al. showed orthopaedic residents randomised to receive VR simulator training outperformed their control group counterparts in an operating theatre [25]. These benefits to knee arthroscopy were also assessed by Waterman et al. who measured the improvements to performance, and compared them to another group trained on cadaveric specimens [26]. Similarly, Banaszek et al.’s study assessed improvements in arthroscopic performance in medical students trained on a VR knee arthroscopy simulator, demonstrating significantly greater performance when compared to controls [27]. Contrary to these promising results, the benefits of knee arthroscopy simulation were not found to be significant in a study by Rebolledo et al. when testing orthopaedic residents’ performances on cadaveric models after simulation training [28]. However, this study did find significant improvements in shoulder arthroscopy performance compared to controls.

The results of this study demonstrate learning curves for a hip arthroscopy simulator, with significant improvements seen after three sessions. All performance metrics were found to improved, demonstrating sufficient visuo-haptic consistency within the virtual environment, enabling individuals to develop basic arthroscopic skills.