Very few studies comparing the learning curves of RL and CL have been published. In the experimental setting a diversity of parameters, not always well-defined, has been used for analysis of learning curves and only the very beginning of the learning curve is studied. In the clinical setting, an experience bias has been expected due to prior laparoscopic experience of the participating surgeons [
2,
6-
12]. Both experimental and clinical studies show diverging learning curves for robotic surgery. The results of previous studies are not conclusive and to objectively evaluate the learning curve of robotic surgery is difficult.
Our experimental study included participants without any prior experience of open surgery, RL or CL, making the group homogeneous. The performed tasks, well-defined and described, closely mimicked some of the proper surgical procedures used every day in the operating theatre. We used the only robotic surgical system currently on the market and standard CL instruments. The size of our group of participants and the number of repetitions studied was decided after power calculation.
The task of tying a surgical knot was always faster with RL than with CL, even when the participant had gained no experience by carrying out the first part of the study with CL. There are learning curves seen for tying a knot for both RL and CL when comparing trials 1 and 4. The learning curve is steeper for CL but the curves never cross (Figure ). These findings differ from most previous studies where the initial performance with RL is often inferior to the performance with CL [
3]. In a recent publication by Stefanidis et al. the authors reported that robotic assistance significantly improved intracorporeal suturing performance and shortened the learning curve. They also reported that performance of laparoscopic knot tying without robotic assistance did not improve after three repeats [
13]. The first statement is supported by our study but the latter is not since we also saw a significant learning curve for CL. Performing more advanced tasks like tying a knot might be faster for maiden users due to the fact that RL is more "intuitive" with instrument movements mimicking normal hand movements. This is supported by some authors [
4]. The fact that RL is performed with 3-dimensional vision instead of the 2-dimensional vision in CL might also improve the performance, as has been suggested by others [
7,
14].
The transfer effect, with a faster performance if the specific method was used as the second part of the study when the tasks had already been tried by the first method, was seen for continuous suturing and tying a knot with RL, but not for CL. This might be interpreted as the RL method being easier to adapt to once acquaintance had been made with the tasks themselves, at least for maiden users. The study by Blavier et al. showed worse performance when shifting from one method to the other in both directions. The shorter learning curve for RL noted by the same authors is supported by our study [
7].
The learning curve consists of an initial steep phase in which performance improves rapidly. When the change in improvement slows down, the learning curve reaches a plateau phase in which variability in performance is small. The number of repetitions reported here are too low to reach consistency, which characterizes the end of the learning curve. The learning curve for CL was steeper, but the number of repetitions too few to disclose a complete learning curve. This was not the aim of the study. We concentrated on the first phase of the learning curve in order to detect even small changes or differences between the two techniques used. From our data, we can therefore only conclude that it is initially easier for novice subjects to use robotic assistance for the specific tasks using the set performance parameters. Whether or not the curves for RL and CL eventually cross after more repetitions, or when the plateau phase of the learning curve for each technique is reached, remains unclear. This could be the aim of another study in the future.
The objective structured assessment of technical skill (OSATS) described by Reznick et al. is a validated tool widely used in the education literature. The OSATS is feasible, reliable and can be used for testing technical competence with high clinical relevance [
15]. Since we focused on comparing the different repeats and the transfer effect in all three tasks we did not calculate a total score for the time and accuracy parameters.
Dropping the needle was more common in the RL group. Half of the subjects dropped the needle while performing CL and all but two while performing RL. Furthermore, the thread was only torn when using RL. Tactile feedback is not yet possible in RL, which is the most probable explanation for our findings. In spite of these differences, albeit significant, the performance when using RL was not slower in the task "continuous suturing" compared with CL. Without the dropping of the needle in the task "continuous suturing", RL might have been faster. Learning curves are also seen for a continuous suture for both RL and CL when comparing trials 1 and 4 (Figure ). The two figures and express the mean time for each try for the specific task but do not consider in what order the task is performed. As already stated, no difference was noted between RL and CL for "continuous suturing". Clinical reports have indicated that the improved vision in RL seems to make up for the lack of tactile feedback for more experienced surgeons [
5]. The tearing of threads and dropping of needles is probably a greater challenge to the beginner.
The end points of our study, time and accuracy, may not be the best end points to measure. Length of pathway and economy of movement might be better predictors of learning curve and safe performance of laparoscopic surgery. A further possible limitation of our study is that error reduction, an important goal of training, was not measured. The study of Narazaki et al. suggests that both task completion time and distance travelled is shortened with training in novice users [
10].
The suggested advantage of faster laparoscopy in the RL group might not be relevant in clinical surgery since inexperienced users are not supposed to perform advanced laparoscopic surgery. Robotic surgeons today are often senior surgeons and already expert laparoscopists. However, the training to become an expert takes a lot of time and is costly, so learning curves are important also for the future education of young surgeons. If RL is proven easier to master with equal or better results than CL, robotic surgery could be an option for efficient surgical training. The many steps of a surgical intervention each have a learning curve and if learning curves are shorter for RL it may have some clinical relevance even at later stages of training.
RL is still in its infancy but offers great opportunities for the future. Major improvements in the availability of tactile feedback and specifically designed instruments are necessary and expected soon. More research needs to be done to define the exact indications for RL to justify the increased costs and the increased time consumption involved, compared with CL.
Whether or not these features with improved accuracy, dexterity and visualization enhance surgical performance remains unclear.
In conclusion, we found support for our hypothesis that a surgical task, such as tying a knot, was performed faster using RL than with CL, while easier surgical tasks could be performed equally fast with either technique. The lack of tactile feedback in RL is a factor to consider at least for maiden users. Experience from one technique was transferred to the other. Our data do not support the suggestion that considerable CL experience is important for those starting to use RL. On the other hand, previous experience did matter in our study. No difference between the performances of male and female subjects was noted.
1 Johan Larsson,1 Christina C Kockum,1 and Einar Arnbjörnsson1
