THE INTERNATIONAL JOURNAL OF MEDICAL ROBOTICS AND COMPUTER ASSISTED SURGERY ORIGINAL Int J Med Robotics Comput Assist Surg 2014; 10: 218–222. Published online 4 December 2013 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/rcs.1554
ARTICLE
Robotic cholecystectomy and resident education: the UC Davis experience
Eric C. Nelson1 Andrea H. Gottlieb3 Hans-Georg Müller3 William Smith1 Mohamed R. Ali2 Tamas J. Vidovszky2* 1
Department of Surgery, University of Tennessee School of Medicine, Chattanooga, TN, USA
2
Department of Surgery, University of California, Davis, CA, USA
3
Department of Statistics, University of California, Davis, CA, USA *Correspondence to: T. J. Vidovszky, Department of Surgery, University of California at Davis, 2221 Stockton Boulevard, Sacramento, CA 95817, USA. E-mail: tamas.vidovszky@ucdmc. ucdavis.edu
Abstract Introduction The popularity of robotic surgery highlights the need for strategies to integrate this technique into surgical education. We present 5 year data for robotic cholecystectomy (RC) as a model for training residents. Methods Data were collected on all RC over 66 months. Duration for docking the robot (S2) and performing RC (S3), and surgical outcomes, were recorded. We used a linear mixed effects model to investigate learning curves. Results Thirty-eight trainees performed 160 RCs, with most performing more than four. One case was aborted due to haemodynamic instability, and two were converted to open surgery due to adhesions. There were no technical complications. The duration of S2 (mean = 6.2 ± 3.6 min) decreased considerably (p = 0.027). Trainees also demonstrated decrease in duration of S3 (mean = 38.4 ± 15.4 min), indicating improvement in technique ( p = 0.008). Conclusions RC is an effective model for teaching residents. Significant and reproducible improvement can be realized with low risk of adverse outcomes. Copyright © 2013 John Wiley & Sons, Ltd. Keywords robotic surgery; cholecystectomy; learning curve; surgical education; da Vinci; laparoscopy
Introduction
Accepted: 10 October 2013
Copyright © 2013 John Wiley & Sons, Ltd.
The minimally invasive surgery (MIS) transformation has enabled surgeons to provide advanced operative care while significantly enhancing patient recovery (1,2). As the next technological iteration in this process, robotic surgery overcomes some of the limitations of traditional MIS (2). Robotic surgery can enhance operative precision in MIS via the magnified, threedimensional (3D) visualization, increased dexterity due to the wristed instrumentation and elimination of natural tremor. From the surgeon’s perspective, ergonomic positioning is greatly enhanced by sitting comfortably at the robotic console (3,4). Yet, despite its advantages, robotic surgery can be more expensive than traditional MIS and is associated with a learning curve, owing to several factors, which include absence of tactile feedback (4). Robotic surgery has been taught using several different models (5). This approach has shown particular promise in pelvic surgery, gaining wide application in prostatectomy and, more recently, gynaecological operations (6,7). We have previously reported our initial experience with robotic cholecystectomy (RC) using the da Vinci robotic surgery system (Intuitive Surgical, Sunnyvale, CA, USA) (8). Since then, we have favoured RC as a model for training surgical residents in robotic surgery. Cholecystectomy is a
219
Robotic cholecystectomy and resident education
common procedure, such that its technical aspects do not detract from learning the robotic platform. The literature regarding resident education in robotic general surgery has focused on simulation training outside the operating room (9,10). In the clinical setting, urology residents have been trained to perform robotic prostatectomy (11,12), but general surgery training has largely occurred in fellowship or later (5,13). In fact, a majority of surgery residents have minimal, if any, exposure to surgical robotics during their training (13). We hold the view that robotic surgery will become an increasingly important component of general surgery education. However, the current literature is relatively limited regarding the clinical learning curve for robotic surgery during general surgery residency. We, therefore, report our experience in training general surgery residents in robotic surgery, using RC as a model.
Materials and methods Protocol In our practice, all patients considered for elective cholecystectomy are candidates for RC unless the da Vinci system is not available on the day of surgery or the patient requires urgent operation. As such, there were no patientspecific exclusion criteria for this study. Demographic and clinical data were collected on all patients during surgical consultation and the preoperative appointment. An attending surgeon with experience in MIS and robotic surgery supervised all cases. RC was performed by chief residents in surgery or fellows in MIS, except for cases when no trainee was available. Each trainee was required to complete robotic training in the laboratory prior to performing RC. The first session was a one hour introduction to the basic concepts of robotic surgery, including positioning, docking and basic use of the da Vinci system. The introductory session was followed by an additional hour in the dry laboratory to learn camera and instrument coordination and operative skills such as dissection and suturing. Each trainee was also required to first-assist at least one robotic case prior to sitting at the console.
Surgical technique RC was performed using either a five-port technique [one 12 mm periumbilical port for the camera, two 8 mm ports in the right and left midclavicular lines for the robotic instrumentation, one 12 mm port in the subxiphoid region and one 5 mm port in the right lower quadrant (RLQ)] or a 4-port technique (identical, but without the 12 mm subxiphoid port). After port placement and adhesiolysis (when indicated), the gall bladder was grasped through the RLQ 5 mm port and retracted cephalad to visualize the infundibulum and the triangle of Calot. The patient was then placed in reverse Trendelenburg and left lateral Copyright © 2013 John Wiley & Sons, Ltd.
decubitus position. The da Vinci surgical robot was then positioned and docked. The cystic duct and cystic artery were dissected and clearly identified, using the robotic L-hook cautery and Cadiere grasper, and then clipped and divided. The gall bladder was then dissected from the fossa and placed in an endoscopic retrieval bag. The robot was then withdrawn, and the two 12 mm port sites were closed with absorbable sutures. Additional procedures were performed (umbilical, ventral hernia repair) as necessary.
Statistical analysis Data were prospectively collected at the time of operation for clinical and educational purposes and retrospectively reviewed, under approval from the Institutional Review Board, for the purposes of this study. For analytical purposes, the procedure was divided into four discrete segments: segment one (S1), or set-up time, spanned from the initial skin incision until the decision to position the da Vinci; segment two (S2), or docking time, entailed the set-up of the robot onto the surgical field; segment three (S3), or procedure time, encompassed the entire robotic component of the procedure; and segment four (S4), or closing time, was the remainder of the operation until skin closure. The objective of the statistical analysis was to evaluate the learning curves associated with each of these four segments of RC. For all four segments, we used a linear mixed effects model with log-transformed time as response (14): logðT i Þ ¼ β0 þ b0kðiÞ þ β1 i þ β2 þ b2kðiÞ RkðiÞ þ εi
(1)
where Ti is the response time for the ith patient, 1 ≤ i ≤ n, k = k(i) is the index of the trainee who operates on the ith patient, k = 1, …, n and Rk(i) is the number of surgeries completed by the kth trainee at the time of operation on ith patient and thus accounts for the trainee’s previous surgical experience at the time of each case. Here, β0 is a non-random intercept, and b0k(i) is a random intercept for the kth trainee. β1 is a non-random slope acting on patient number and, thus, reflects overall programmatic learning effects, as surgical experience was gained with increasing number of patients. β2 is a non-random slope acting on Rk(i) and reflects average learning effects across trainees as each trainee gains surgical experience, and b2k (i) is the random portion of this slope, which reflects individual learning effects of particular trainees. Finally, εij is a measurement error, assumed to be independent and identically distributed with finite variance. We assume that all random effects and errors are normally distributed. Significance of both fixed and random effects were tested at the 5% level. Using this model allowed us to investigate learning on three different levels: the individual learning curve of each trainee, variation across resident-specific learning curves, and overall surgical experience of the programme. Int J Med Robotics Comput Assist Surg (2014); 10: 218–222 DOI: 10.1002/rcs
E. C. Nelson et al.
220
Results During the study period, 160 patients underwent RC. The study population was predominantly female (71%) and had mean age of 46 + 14.5 years and mean weight of 93 + 24.8 kg (Table 1). Cholecystectomy was completed robotically in 157 patients. One case was aborted due to physiological instability of the patient prior to use of the robot. Two patients required conversion to open cholecystectomy due to adhesions. No complications referable to the robot occurred. Postoperatively, one patient developed a small port site haematoma that did not require any intervention. One patient developed postoperative pulmonary oedema and required hospitalization. A third patient developed deep venous thrombosis after discharge, which required re-admission to the hospital and anticoagulation. No known biliary injuries or wound infections occurred and there were no mortalities in the series. Overall, 31 residents and seven fellows were trained during the study period. Regression models for residents and fellows were compared using an F-test and no statistically significant differences were noted. Therefore, all trainees were analysed as one group. Trainees completed between 1 and 13 RCs each, with over 50% of the trainees performing four or more surgeries (see Table 2). In several cases, a chief resident was not available to scrub for the case and an attending surgeon performed the case at the console. Therefore the total number of surgeons listed in Table 2 is 42. Overall, the mean duration of S1 (set-up time) was 17.2 + 7.7 min and of S4 (closing time) was 42.6 + 16.3 min. No statistically significant change was identified in S1. S4 did show a statistically significantly decrease in time over the course of the study, but varied widely depending on additional procedures performed. Significant learning was demonstrated by trainees in S2 (docking time) and S3 (procedure time), as demonstrated
in Figures 1 and 2. The duration of S2 (mean = 6.2 + 3.6 min) decreased significantly with increasing experience (p = 0.027; Figure 1). Trainees demonstrated significant decrease in duration of S3 (mean = 38.4 + 15.4 min, p = 0.008; Figure 2). Furthermore, there was no statistically significant variability among trainees in improvement for S3. However, within the group, differences in traineespecific learning were evident for S2 (p = 0.01). Table 3 summarizes significant learning effects.
Discussion Surgical education is currently undergoing major evolution. Debates regarding the spectrum of current general surgery practice, objective evaluation of competence, and methods for safe transfer of advanced operative techniques to trainees are ongoing. The challenges are further amplified by current work-hour limitations on the exposure of general surgery house staff to the full episode of patient care in complex surgical cases (15). This partly explains the surge in popularity of postresidency fellowship training in a variety of fields, including the evolving field of MIS (16). We subscribe to the concept that MIS is inseparable from general surgery and that a well-trained general surgeon must be versed in the techniques of MIS. Numerous surgical educators have addressed these issues with proposals that range from inanimate training models to computerized simulation to standardized evaluation of performance in the operating room (17). Recently, the American Board of Surgery has underscored the importance of MIS in the current scope of clinical practice by including the Fundamentals of Laparoscopic Surgery (FLS) certification among the requirements for general surgery certification. While significant progress has been made in improving ’standard’ MIS education, robotic surgical training remains
Table 1. Demographics of patient population Demographics of Patient Population Mean age (range, years) ± SD Female, n (%) Male n (%) Mean weight (range, kg) ± SD
46 (18–93) ± 14.5 114 (71) 46 (29) 93 (40–173) ± 24.8
Table 2. Number of surgeries/trainee Number of surgeries 1 2 3 4 5 6 7
ARTICLE
Robotic cholecystectomy and resident education: the UC Davis experience
Eric C. Nelson1 Andrea H. Gottlieb3 Hans-Georg Müller3 William Smith1 Mohamed R. Ali2 Tamas J. Vidovszky2* 1
Department of Surgery, University of Tennessee School of Medicine, Chattanooga, TN, USA
2
Department of Surgery, University of California, Davis, CA, USA
3
Department of Statistics, University of California, Davis, CA, USA *Correspondence to: T. J. Vidovszky, Department of Surgery, University of California at Davis, 2221 Stockton Boulevard, Sacramento, CA 95817, USA. E-mail: tamas.vidovszky@ucdmc. ucdavis.edu
Abstract Introduction The popularity of robotic surgery highlights the need for strategies to integrate this technique into surgical education. We present 5 year data for robotic cholecystectomy (RC) as a model for training residents. Methods Data were collected on all RC over 66 months. Duration for docking the robot (S2) and performing RC (S3), and surgical outcomes, were recorded. We used a linear mixed effects model to investigate learning curves. Results Thirty-eight trainees performed 160 RCs, with most performing more than four. One case was aborted due to haemodynamic instability, and two were converted to open surgery due to adhesions. There were no technical complications. The duration of S2 (mean = 6.2 ± 3.6 min) decreased considerably (p = 0.027). Trainees also demonstrated decrease in duration of S3 (mean = 38.4 ± 15.4 min), indicating improvement in technique ( p = 0.008). Conclusions RC is an effective model for teaching residents. Significant and reproducible improvement can be realized with low risk of adverse outcomes. Copyright © 2013 John Wiley & Sons, Ltd. Keywords robotic surgery; cholecystectomy; learning curve; surgical education; da Vinci; laparoscopy
Introduction
Accepted: 10 October 2013
Copyright © 2013 John Wiley & Sons, Ltd.
The minimally invasive surgery (MIS) transformation has enabled surgeons to provide advanced operative care while significantly enhancing patient recovery (1,2). As the next technological iteration in this process, robotic surgery overcomes some of the limitations of traditional MIS (2). Robotic surgery can enhance operative precision in MIS via the magnified, threedimensional (3D) visualization, increased dexterity due to the wristed instrumentation and elimination of natural tremor. From the surgeon’s perspective, ergonomic positioning is greatly enhanced by sitting comfortably at the robotic console (3,4). Yet, despite its advantages, robotic surgery can be more expensive than traditional MIS and is associated with a learning curve, owing to several factors, which include absence of tactile feedback (4). Robotic surgery has been taught using several different models (5). This approach has shown particular promise in pelvic surgery, gaining wide application in prostatectomy and, more recently, gynaecological operations (6,7). We have previously reported our initial experience with robotic cholecystectomy (RC) using the da Vinci robotic surgery system (Intuitive Surgical, Sunnyvale, CA, USA) (8). Since then, we have favoured RC as a model for training surgical residents in robotic surgery. Cholecystectomy is a
219
Robotic cholecystectomy and resident education
common procedure, such that its technical aspects do not detract from learning the robotic platform. The literature regarding resident education in robotic general surgery has focused on simulation training outside the operating room (9,10). In the clinical setting, urology residents have been trained to perform robotic prostatectomy (11,12), but general surgery training has largely occurred in fellowship or later (5,13). In fact, a majority of surgery residents have minimal, if any, exposure to surgical robotics during their training (13). We hold the view that robotic surgery will become an increasingly important component of general surgery education. However, the current literature is relatively limited regarding the clinical learning curve for robotic surgery during general surgery residency. We, therefore, report our experience in training general surgery residents in robotic surgery, using RC as a model.
Materials and methods Protocol In our practice, all patients considered for elective cholecystectomy are candidates for RC unless the da Vinci system is not available on the day of surgery or the patient requires urgent operation. As such, there were no patientspecific exclusion criteria for this study. Demographic and clinical data were collected on all patients during surgical consultation and the preoperative appointment. An attending surgeon with experience in MIS and robotic surgery supervised all cases. RC was performed by chief residents in surgery or fellows in MIS, except for cases when no trainee was available. Each trainee was required to complete robotic training in the laboratory prior to performing RC. The first session was a one hour introduction to the basic concepts of robotic surgery, including positioning, docking and basic use of the da Vinci system. The introductory session was followed by an additional hour in the dry laboratory to learn camera and instrument coordination and operative skills such as dissection and suturing. Each trainee was also required to first-assist at least one robotic case prior to sitting at the console.
Surgical technique RC was performed using either a five-port technique [one 12 mm periumbilical port for the camera, two 8 mm ports in the right and left midclavicular lines for the robotic instrumentation, one 12 mm port in the subxiphoid region and one 5 mm port in the right lower quadrant (RLQ)] or a 4-port technique (identical, but without the 12 mm subxiphoid port). After port placement and adhesiolysis (when indicated), the gall bladder was grasped through the RLQ 5 mm port and retracted cephalad to visualize the infundibulum and the triangle of Calot. The patient was then placed in reverse Trendelenburg and left lateral Copyright © 2013 John Wiley & Sons, Ltd.
decubitus position. The da Vinci surgical robot was then positioned and docked. The cystic duct and cystic artery were dissected and clearly identified, using the robotic L-hook cautery and Cadiere grasper, and then clipped and divided. The gall bladder was then dissected from the fossa and placed in an endoscopic retrieval bag. The robot was then withdrawn, and the two 12 mm port sites were closed with absorbable sutures. Additional procedures were performed (umbilical, ventral hernia repair) as necessary.
Statistical analysis Data were prospectively collected at the time of operation for clinical and educational purposes and retrospectively reviewed, under approval from the Institutional Review Board, for the purposes of this study. For analytical purposes, the procedure was divided into four discrete segments: segment one (S1), or set-up time, spanned from the initial skin incision until the decision to position the da Vinci; segment two (S2), or docking time, entailed the set-up of the robot onto the surgical field; segment three (S3), or procedure time, encompassed the entire robotic component of the procedure; and segment four (S4), or closing time, was the remainder of the operation until skin closure. The objective of the statistical analysis was to evaluate the learning curves associated with each of these four segments of RC. For all four segments, we used a linear mixed effects model with log-transformed time as response (14): logðT i Þ ¼ β0 þ b0kðiÞ þ β1 i þ β2 þ b2kðiÞ RkðiÞ þ εi
(1)
where Ti is the response time for the ith patient, 1 ≤ i ≤ n, k = k(i) is the index of the trainee who operates on the ith patient, k = 1, …, n and Rk(i) is the number of surgeries completed by the kth trainee at the time of operation on ith patient and thus accounts for the trainee’s previous surgical experience at the time of each case. Here, β0 is a non-random intercept, and b0k(i) is a random intercept for the kth trainee. β1 is a non-random slope acting on patient number and, thus, reflects overall programmatic learning effects, as surgical experience was gained with increasing number of patients. β2 is a non-random slope acting on Rk(i) and reflects average learning effects across trainees as each trainee gains surgical experience, and b2k (i) is the random portion of this slope, which reflects individual learning effects of particular trainees. Finally, εij is a measurement error, assumed to be independent and identically distributed with finite variance. We assume that all random effects and errors are normally distributed. Significance of both fixed and random effects were tested at the 5% level. Using this model allowed us to investigate learning on three different levels: the individual learning curve of each trainee, variation across resident-specific learning curves, and overall surgical experience of the programme. Int J Med Robotics Comput Assist Surg (2014); 10: 218–222 DOI: 10.1002/rcs
E. C. Nelson et al.
220
Results During the study period, 160 patients underwent RC. The study population was predominantly female (71%) and had mean age of 46 + 14.5 years and mean weight of 93 + 24.8 kg (Table 1). Cholecystectomy was completed robotically in 157 patients. One case was aborted due to physiological instability of the patient prior to use of the robot. Two patients required conversion to open cholecystectomy due to adhesions. No complications referable to the robot occurred. Postoperatively, one patient developed a small port site haematoma that did not require any intervention. One patient developed postoperative pulmonary oedema and required hospitalization. A third patient developed deep venous thrombosis after discharge, which required re-admission to the hospital and anticoagulation. No known biliary injuries or wound infections occurred and there were no mortalities in the series. Overall, 31 residents and seven fellows were trained during the study period. Regression models for residents and fellows were compared using an F-test and no statistically significant differences were noted. Therefore, all trainees were analysed as one group. Trainees completed between 1 and 13 RCs each, with over 50% of the trainees performing four or more surgeries (see Table 2). In several cases, a chief resident was not available to scrub for the case and an attending surgeon performed the case at the console. Therefore the total number of surgeons listed in Table 2 is 42. Overall, the mean duration of S1 (set-up time) was 17.2 + 7.7 min and of S4 (closing time) was 42.6 + 16.3 min. No statistically significant change was identified in S1. S4 did show a statistically significantly decrease in time over the course of the study, but varied widely depending on additional procedures performed. Significant learning was demonstrated by trainees in S2 (docking time) and S3 (procedure time), as demonstrated
in Figures 1 and 2. The duration of S2 (mean = 6.2 + 3.6 min) decreased significantly with increasing experience (p = 0.027; Figure 1). Trainees demonstrated significant decrease in duration of S3 (mean = 38.4 + 15.4 min, p = 0.008; Figure 2). Furthermore, there was no statistically significant variability among trainees in improvement for S3. However, within the group, differences in traineespecific learning were evident for S2 (p = 0.01). Table 3 summarizes significant learning effects.
Discussion Surgical education is currently undergoing major evolution. Debates regarding the spectrum of current general surgery practice, objective evaluation of competence, and methods for safe transfer of advanced operative techniques to trainees are ongoing. The challenges are further amplified by current work-hour limitations on the exposure of general surgery house staff to the full episode of patient care in complex surgical cases (15). This partly explains the surge in popularity of postresidency fellowship training in a variety of fields, including the evolving field of MIS (16). We subscribe to the concept that MIS is inseparable from general surgery and that a well-trained general surgeon must be versed in the techniques of MIS. Numerous surgical educators have addressed these issues with proposals that range from inanimate training models to computerized simulation to standardized evaluation of performance in the operating room (17). Recently, the American Board of Surgery has underscored the importance of MIS in the current scope of clinical practice by including the Fundamentals of Laparoscopic Surgery (FLS) certification among the requirements for general surgery certification. While significant progress has been made in improving ’standard’ MIS education, robotic surgical training remains
Table 1. Demographics of patient population Demographics of Patient Population Mean age (range, years) ± SD Female, n (%) Male n (%) Mean weight (range, kg) ± SD
46 (18–93) ± 14.5 114 (71) 46 (29) 93 (40–173) ± 24.8
Table 2. Number of surgeries/trainee Number of surgeries 1 2 3 4 5 6 7
