Learning from our virtual mistakes

The idea of using simulators to teach endoscopy is almost as old as the procedure itself. In 1937, Schindler described maneuvering a gastroscope inside a hollow model to practice examination of the stomach.1 Colonoscopy simulators, ranging in complexity from modified hair dryer tubes to computerized virtual reality environments, have been available for more than 30 years.2 The First International Conference on Endoscopy Simulation, held in 2006, issued consensus recommendations for the use of simulators in both the training and credentialing of endoscopists.3

In this issue of GIE, Haycock et al4 report a randomized, controlled trial that further illustrates the utility of simulators in the teaching of therapeutic endoscopy. The arguments for using simulators early in endoscopy training are straightforward and intuitive: novice endoscopists have higher complication rates5, 6, 7 and lower patient satisfaction scores8 than experienced practitioners. It has been shown that supplementing hands-on training with simulation decreases patient discomfort in subsequent procedures9 and that simulator-trained fellows exhibit better overall performance on their first 30 to 80 colonoscopies than their peers.10

Yet despite more than 3 decades of study and debate, simulators play virtually no role in the training of most gastroenterologists today. One often-stated reason for this is cost. As shown in the article by Haycock et al, a simulator-based training program does involve a significant investment of both time and money. Although 89 candidates expressed interest in the study, the investigators were unable to enroll the desired 32 participants, largely for financial reasons. However, simulators may prove to be cost-saving in the long run because of improvement in the efficiency and quality of procedures performed by fellows. Currently, fellow participation prolongs procedure time by 10% to 37%11 and this effect could be mitigated by allowing fellows to practice in a simulator laboratory. This issue is likely to gain in importance as financial pressures on academic centers increase and more procedures are performed in high-throughput ambulatory surgery centers.

Perhaps a more fundamental reason for the lack of simulator-based training is that we do not know exactly what to expect from simulators and are therefore unsure of how they would fit into a training environment. At first glance, one may envision an “ideal” simulator as a fully immersive computer-generated environment that reproduces the entire experience of performing an endoscopic procedure: sedating the patient, inserting and advancing the endoscope, identifying landmarks, examining around the flexures, recognizing pathology, performing biopsies, removing polyps, and even addressing immediate complications such as postpolypectomy bleeding and small perforations. Various levels of difficulty and pathology, ranging from large polyps to subtle flat and depressed lesions, could be “programmed” into such a computer with a touch of a button. This would allow the trainee to perform endoscopy at a fairly high level with no risk to patients and no impact on endoscopy unit workflow. Not only withdrawal time, but also polyp detection rate and percentage of mucosa visualized could be easily measured and compared with some standard of competency before the trainee moves on to the next step. If one wished to largely replace live teaching with simulation, such a sophisticated and comprehensive simulator would not only be desirable, but essential.

However, on closer examination, it becomes clear that such a system is neither practical nor necessary. First, it would be far too expensive for the average teaching institution, which trains fewer than 3 fellows per year.12 The computing power required to simulate the forces involved in endoscope advancement to the cecum is considerable; to add to this, a physics engine that adequately simulates the behavior of the endoscope, the accessory, and the colonic mucosa during a polypectomy or clip deployment would require an aircraft-grade simulator at a cost of several million dollars. Anyone who has used one of the current computer-based endoscopy simulators can attest that, although they approximate the feeling of endoscope insertion to some degree, “virtual” polypectomies on such devices bear little resemblance to the real thing. Second, it is clear that simulation will always be an adjunct to, not a replacement for, live one-on-one teaching. Even brief supervised training on real patients is superior to the most advanced computer-based simulations available today.13 Undoubtedly, computer technology (most likely propelled by the video game industry) will one day progress to the point where a sufficiency realistic, low-cost, and fully immersive simulator becomes cost-effective, but we need not wait for such a machine to realize the benefits of simulation. Instead, we should redefine our expectations away from global simulation toward simpler models that allow the user to practice and demonstrate proficiency in specific skills that translate into improved outcomes in the real world.

The first question that we must ask, then, is what aspect of endoscopy we wish to simulate. In general, simulation should focus on rare but potentially serious situations, the outcomes of which hinge on specific, teachable skills that improve with repeated practice. Examples of such situations would be the encounter of a spurting vessel at the base of an ulcer or causing a small perforation after endoscopic mucosal resection. Inserting an endoscope and advancing it to the cecum, although teachable and likely to improve with practice, should not be a rare event in a 3-year gastroenterology fellowship. With adequate supervision and instruction, fellows should gain adequate proficiency in diagnostic colonoscopy with minimal risk to patients. Furthermore, most fellows by the end of training far exceed the 100 colonoscopies required for certification. Exposure to polypectomy and hemostasis, conversely, is more variable. Opportunities to practice such techniques are often limited by concerns for patient safety in emergent situations or by patient volume in a particular teaching institution. Even the most open-minded patient and well-intentioned endoscopy teacher cannot allow a fellow to unsuccessfully clip an actively bleeding vessel or attempt to remove a polyp multiple times until he or she gets it “just right,” as could be the case in a simulation environment. This deficiency in exposure is reflected in current guidelines for credentialing endoscopists after a minimum of only 20 polypectomies.14 Although this number of procedures may confer some degree of competence, no one would argue that the 21st polypectomy performed by even the most prodigious fellow is at the same skill level as the 200th. Thus, there is a pressing need for a calm, collected environment to practice such maneuvers, and simulation should focus on these underpracticed aspects of therapeutic endoscopy.

The next question is what model is best for practicing therapeutic endoscopy. Three types of simulators exist: artificial tissue, ex vivo animal tissue, and computer based. Procedures on anesthetized animals can also be considered a type of simulation, but, because of cost and other logistical barriers, are unlikely to ever become a major part of training. Each method of simulation has advantages and disadvantages. Artificial tissue simulators that use plastic or similar materials can simulate looping, but do not allow realistic simulation of pathology or therapeutic maneuvers. Ex vivo models, although more expensive and requiring fresh animal organs, provide a realistic feel during endoscopic therapy because they involve using real instruments on real tissue. Although a sufficiently advanced computer-based simulator could potentially perform all these functions in a clean and infinitely reusable environment, this is currently not the case, as described above. As pointed out in the current study, the Erlangen Active Simulator for Interventional Endoscopy (EASIE) and the compactEASIE systems, pioneered by Hochberger in 1996, provide the most realistic training environment for therapeutic endoscopy. The system consists of harvested animal (usually porcine) viscera attached to a plastic frame. Endoscopists may practice closure of artificially created holes and hemostasis of bleeding vessels sutured to the gastric wall and supplied by a pulsatile perfusion system.15 The Erlangen system, with minor modifications, also allows the endoscopist-in-training to ligate artificially created varices, remove submucosal lesions, and even practice therapeutic ERCP and deep enteroscopy.16

After just 1 day of supervised training, this system has been shown to improve interventional skills,17 and it seems that fellows trained on this model have higher success rates and fewer complications in real bleeding cases that they perform subsequently.18 This study adds polypectomy and esophageal dilation to the list of procedures for which simulation enhances the training.

The third question is how to measure proficiency in a simulation environment. Because different people learn at different rates, we need to move beyond a threshold number of procedures and develop a more sophisticated metric for competence. Examples could be the time required to achieve hemostasis in a standardized bleeding scenario or the number of attempts required to close a mucosal defect of a certain size and shape. It is also important that our assessments are free from trainer bias; in other words, we must use an objective standard rather than merely measure how similar a trainee's technique is to that of the evaluator. Several validation studies on computer-based simulators have shown that they can detect the difference between a novice and an expert endoscopist without the bias to which human evaluators may be subject. However, it is somewhat disturbing that novices can reach the expert level after 3 weeks, suggesting either that the computer cannot distinguish subtle differences in technique beyond a basic level of competence or that after 3 weeks, anyone can learn how to handle a simulator well enough to trick it into thinking that he or she is an expert endoscopist.19, 20 The criterion standard for measuring any training program should be improved patient outcomes months or years after completion of training, because this proves that the skills gained translate to real procedures rather than merely reflecting increased proficiency in using a particular simulator.

Finally, we should consider whether simulation should play a role in recredentialing practicing endoscopists. This idea makes sense for many of the same reasons that the American Board of Internal Medicine requires reassessment every 10 years for maintenance of board certification. Just like cognitive gastroenterology, endoscopy is a constantly evolving field, and procedures that were formerly in the realm of surgery are now increasingly being performed by endoscopists. Although not all endoscopists need to perform every interventional procedure, it is important to ensure that all physicians performing endoscopy provide an acceptable standard of care to their patients. It has also been shown that colonoscopy skills can deteriorate with lack of practice,21 so even basic techniques may need to be refreshed at some interval. Because of the number of endoscopists who would be involved in such a program and the need to detect subtle errors rather than teach techniques for the first time, recredentialing would likely require an extensive network of sophisticated computer simulators such as those found in the aviation industry, and it is unclear who would pay for development and maintenance of such a program. Thus, it is unlikely that simulator-based recredentialing will become a reality in the near future.

The training of the next generation of practitioners is one of the most important duties that fall on any profession, and gastroenterology is no different. At the same time, we owe each patient a procedure that is performed competently and safely. The use of simulators may allow us to satisfy both of these goals simultaneously, and the initial investment of time and money could be defrayed by faster learning and improved patient outcomes. This study by Haycock et al provides the latest piece of evidence that, although simulators will never replace live teaching in the endoscopy unit, they can and should have an adjunctive role in teaching specific skills that require practice and to which trainees typically have insufficient exposure.