- Email: [email protected]
Proficiency of Robotic Lobectomy Based on Prior Surgical Technique in The Society of Thoracic Surgeons General Thoracic Database
Andrew F. Feczko, MD, Hongwei Wang, MS, Katherine Nishimura, PhD, Alexander S. Farivar, MD, Adam J. Bograd, MD, Eric Valli eres, MD, Ralph W. Aye, MD, and Brian E. Louie, MD Division of Thoracic Surgery, Swedish Cancer Institute, Seattle; and Cancer Research and Biostatistics (CRAB), Seattle, Washington
Background. Robotic lobectomy represents a paradigm shift for many surgeons. It is unknown if a surgeon’s prior operative approach influences development of proficiency. We compared outcomes based on prior lobectomy experience and used cumulative sum analysis to assess proficiency. Methods. Using The Society of Thoracic Surgeons General Thoracic Database we grouped surgeons as de novo, open-to-robotic, or video-assisted thoracoscopic surgery (VATS)-to-robotic. Operative time, blood transfusion, mortality, and major morbidity were primary outcomes. Unacceptable and acceptable thresholds were determined by review of the literature. Proficiency was defined as 20 consecutive cases without crossing an upper control line. Surgeons were assessed individually, and proficiency was assessed by transition group. Results. From 2009 to 2016, 271 surgeons performed 5619 robotic lobectomies for clinical stage I/II non–small cell lung cancer. Of these, 65 surgeons (24%) performed
‡20 lobectomies (4483 cases). Initial proficiency for an operative time target of 250 minutes was 40% for de novo compared with 14% for open-to-robotic and 21% for VATS-to-robotic surgeons, with improvement to 47%, 29%, and 21%, respectively, after 20 cases. Initial and sustained proficiency related to major morbidity was similar for open-to-robotic and VATS-to-robotic but lower for de novo at 40%. After 20 cases most were proficient (de novo, 93%; open-to-robotic, 100%; and VATSto-robotic, 86%). Proficiency for 30-day mortality and blood transfusion was high in all groups. Conclusions. Outcomes among all transition groups improved with experience. Operating room duration proficiency was challenging for all groups. Cumulative sum may be useful to monitor proficiency in not only subsequent studies but in clinical practice.
P
Most surgeons transitioning to robotics underwent a combination of didactic and hands-on training, including online modules, in-person orientation, basic training courses, and proctored cases. Advanced training classes and focused proctoring have become available. Early robotic adopters reported that the “learning curve” on this training pathway was anywhere from 6 cases for VATS surgeons3 to 18 cases for open surgeons6 based on operative time. How surgeons progress after this initial learning curve and the quality of care during the transition to robotic lobectomy are unknown. Our experience suggests that a surgeon’s prior operative experience and approach (open or VATS) might influence development of proficiency during and after the robotic lobectomy learning curve, with open surgeons taking longer to develop proficiency
ulmonary lobectomy is performed using open, videoassisted thoracoscopic surgery (VATS) or robotic approaches.1,2 These recent studies confirmed that minimally invasive techniques are the dominant approach for early-stage lung cancer and interest in robotic lobectomy growing as an alternative. Those who elect to transition to robotics find that it represents a paradigm shift. Certain elements of minimally invasive surgical procedures, small incisions, and access ports are familiar to most. The surgeon–console interface and the remote patient cart, however, can require new skills ranging from direct surgeon control of all instruments and camera to the absence of haptic feedback.3-5
Accepted for publication Apr 11, 2019.
(Ann Thorac Surg 2019;108:1013-20) Ó 2019 by The Society of Thoracic Surgeons
Presented at the Sixty-fifth Annual Meeting of The Southern Thoracic Surgical Association, Amelia Island, FL, Nov 7-10, 2018. Address correspondence to Dr Louie, Division of Thoracic Surgery, Swedish Cancer Institute, Ste 900, 1101 Madison St, Seattle, WA 98105; email: [email protected].
Ó 2019 by The Society of Thoracic Surgeons Published by Elsevier Inc.
Dr Louie discloses a financial relationship with Intuitive Surgical.
0003-4975/$36.00 https://doi.org/10.1016/j.athoracsur.2019.04.046
GENERAL THORACIC
Proficiency of Robotic Lobectomy Based on Prior Surgical Technique in The Society of Thoracic Surgeons General Thoracic Database
GENERAL THORACIC
1014
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
compared with VATS surgeons. The aim of this study was to define metrics to assess proficiency and describe how prior operative experience affects proficiency during transition to robotic lobectomy.
Patients and Methods The institutional review board of the Swedish Medical Center approved this study with a waiver of individual participant and patient consent because of the retrospective use of deidentified data.
Patients A participant user file was acquired from The Society of Thoracic Surgeons General Thoracic Database (STSGTD). Using versions 2.081-2.3 the database was queried for patients undergoing primary lobectomy for non–small cell lung cancer from 2009 to 2016. To maintain confidentiality a participant user file was created with cases ordered by surgery date and date replaced by surgery year. Each surgeon’s identifier number was also replaced with a deidentified surgeon identification for the study. Robotic lobectomies were identified using the “robotic technology–assisted” field. The participant user file was received with 72,924 lobectomies (Figure 1). Patients with missing data or clinical stages III or greater were excluded. Patients undergoing VATS or open lobectomies were set aside and used to define surgeons’ prior experience. This left 5561 robotic lobectomies for clinical stage I/II non–small cell lung cancer. To account for the learning curve of 20 cases as suggested by Veronesi and coworkers,7 cases performed
Ann Thorac Surg 2019;108:1013-20
by surgeons who completed <20 robotic lobectomies were excluded, leaving 4483 lobectomies. Of the 844 surgeons in the STS-GTD, 210 performed at least 1 robotic lobectomy. Of those, 145 surgeons performed <20 robotic lobectomies and were excluded (88 surgeons completed between 11 and 19 robotic lobectomies). The remaining 65 surgeons were divided into transition groups based on their predominant lobectomy approach in the study period before their first robotic lobectomy. Surgeons performing 50% of lobectomies via thoracotomy were considered open-to-robotic surgeons (ORS) and those who performed 50% of lobectomies using a VATS approach were grouped as VATS-to-robotic surgeons (VRS). If a surgeon’s first recorded lobectomy in the database was robotic, they were labeled as de novo surgeons (DNS).
Outcomes Four outcomes modeled on prior evaluations of robotic quality in the STS-GTD were selected as measures of proficiency during the transition to robotic surgery.8,9 The definitions for these outcomes are outlined in the STSGTD manual. Thirty-day mortality was defined as death after surgery within 30 days of operation. Perioperative transfusion was a dichotomous outcome for red blood cell transfusion intraoperatively or postoperatively. Major morbidity was a composite variable including any occurrence of 1 or more complications, including air leak of more than 5 days, chest tube at discharge, postoperative pneumonia, acute respiratory distress syndrome, respiratory failure, pneumothorax requiring chest tube reinsertion, initial ventilatory support of more than
Figure 1. Flow diagram of study population selection. The Society of Thoracic Surgeons General Thoracic Database was queried for all lobectomies between 2009 and 2016. Cases were selected by complete data, clinical stage (cStage) I/II, and robotic approach. Of that group, surgeons who performed 20 robotic lobectomies were included for the final analysis. Yellow boxes indicate the number of lobectomy cases, and white boxes indicate number of surgeons by transition group. Each surgeon’s identifier number was replaced with a deidentified surgeon identification for the study. (DNS, de novo surgeons; NSCLC, non–small cell lung cancer; ORS, open-to-robotic surgeons; VATS, video-assisted thoracoscopic surgery; VRS, VATS-to-robotic surgeons.)
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
1015
Figure 2. Representative cumulative sum (CUSUM) curves. (A) De novo surgeons (DNS) who achieved proficiency. (B) DNS who did not achieve proficiency. (C) Open-to-robotic surgeons (ORS) who achieved proficiency. (D) ORS who did not achieve proficiency. (E) Video-assisted thoracoscopic surgery-to-robotic surgeons (VRS) who achieved proficiency. (F) VRS who did not achieve proficiency. Graphs show CUSUM score in relation to number of cases. Initial control line indicated by dashed line. Each surgeon’s identifier number was replaced with a deidentified surgeon identification for the study. (OR, operating room.)
GENERAL THORACIC
Ann Thorac Surg 2019;108:1013-20
GENERAL THORACIC
1016
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
48 hours, reintubation, atrial fibrillation, and unexpected intensive care unit admission. Operating room (OR) duration was calculated as the difference between procedure start and stop times.
Statistical Analysis To assess changes in proficiency, cumulative sum (CUSUM) analysis was used. CUSUM is a sequential analysis technique used in manufacturing to track outcomes for a process over time. These methods have been adapted for evaluating performance in surgical settings.10-12 The CUSUM analysis relies on the specifications of a target value and a known or reliable estimate of the standard deviation. Undesirable results cause the CUSUM score to increase (upward deflection in graphic form), and desirable results cause the overall score to decrease (downward deflection) (Figures 2A-2F). Acceptable failure rates are defined a priori and along with Type I and II error rates identify a threshold (“control” or “alarm” lines) that when crossed in an upward direction indicate a loss of proficiency and in the downward direction signal a correction. Because of the cumulative nature of the score, control lines are reset each time a threshold is crossed, with ongoing failure indicated by sequential upward crossing of control lines. Similarly, each time a control line is crossed in a downward direction, the upper control line resets at a lower value.9,13,14 Acceptable and unacceptable failure rates for each of the 4 outcomes were determined by review of the quality and safety literature for robotic and VATS lobectomy. The target rate for 30-day mortality was set at 1%, with an unacceptable rate of 5%. Perioperative transfusion target rate was set at 5%, with an unacceptable rate of 10%. The major morbidity target rate was set at 10%, with an unacceptable rate of 15%. These values were selected from the report by Fernandez and coworkers15 using the STS database and confirmed in other studies using a similar composite for major morbidity.1,8 The target time for robotic lobectomy (all lobes) was set at <250 minutes, with an unacceptable rate of 250 minutes. This was Figure 3. Percent of lobectomy by approach between 2009 and 2016. (VATS, video-assisted thoracoscopic surgery.)
Ann Thorac Surg 2019;108:1013-20
determined from studies comparing operative time for VATS and robotic lobectomy that ranged between 170 and 240 minutes.5,8,16 Type I and II error rates were set at .1 for binary outcomes. A surgeon was defined as proficient when 20 lobectomies were performed without crossing an upper control line.7,16 Initial and sustained proficiency was defined as achievement of proficiency without repeated crossing of upper control lines, with surgeons serving as their own internal controls. All surgeons included in the CUSUM analysis completed 20 lobectomies. A smaller group of surgeons completed 50 robotic lobectomies. Groups were compared using c2 tests for categorical variables and 2-sample analysis of variance for continuous variables. Statistical analyses were performed using Microsoft Office Excel (Microsoft, Redmond, WA) and SAS (SAS institute Inc, Cary, NC). CUSUM analysis was performed using R (The R Foundation, Vienna, Austria).
Results Over the study period robotic lobectomy increased as a proportion of lobectomies from <1% in 2009 to 18.1% in 2016 (Figure 3). Open lobectomy decreased from 53.3% to 28.0%, with the proportion of VATS lobectomies stabilizing between 2014 and 2016. The 65 surgeons completing 4483 robotic lobectomies were divided into transition groups: 15 DNS (1119 lobectomies), 21 ORS (1511 lobectomies), and 29 VRS (1853 lobectomies). A higher proportion of patients in the DNS group were women; however body mass index, age, and performance status across groups appeared to be similar (Table 1). There was not a statistically significant difference in clinical or pathologic stage of resected tumors. There were statistical differences in forced expiratory volume in the first second of expiration percent predicted, diffusing capacity of the lung for carbon monoxide percent predicted, and American Society of Anesthesiologists class among the groups. DNS had the lowest average forced expiratory volume in the first second of
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
Table 1. Study Population Patient Characteristics by Transition Group
Table 2. Overall Outcomes for Robotic Lobectomy by Transition Group
Patient Characteristics
Outcomes
DNS ORS VRS (n ¼ 1234) (n ¼ 1818) (n ¼ 2509) P Value
Female 485 (60.7) 795 (56.3) Median age, y 69 68 26.6 26.8 Median body mass index, kg/m2 Performance status (Zubrod) 0-1 1175 (95.2) 1708 (94.0) >1 59 (4.8) 110 (6.0) pT1-2 tumor 1102 (89.3) 1625 (89.4) 82 84 Mean FEV1, % predicted Mean DLCO, % 76 74 predicted ASA class III 924 (75.0) 1447 (80.0) Never smoker 204 (16.5) 329 (18.1) Chronic obstructive 503 (40.8) 693 (38.1) pulmonary disease Coronary artery 337 (27.3) 451 (24.8) disease Dialysis 4 (.3) 7 (.4) Diabetes 205 (16.6) 365 (20.1)
1110 (55.8) 69 27.0
.02
2430 (96.9) 79 (3.2) 2193 (87.4) 86
.62 <.01 .74 <.01
77
.03
1982 (79.0) 448 (17.9) 877 (35.0)
.01 .56 .02
532 (21.2)
<.01
18 (.7) 485 (19.3)
.18 .08
Values are n (%) unless otherwise indicated. ASA, American Society of Anesthesiologists; DLCO, diffusing capacity of the lung for carbon monoxide; DNS, de novo surgeons; FEV1, forced expiratory volume in the first second of expiration; ORS, open-to-robotic surgeons; VRS, video-assisted thoracoscopic surgery-to-robotic surgeons.
expiration percent predicted (82%), but diffusing capacity of the lung for carbon monoxide percent predicted was lowest for ORS. ORS also had a greater proportion of American Society of Anesthesiologists class III patients at 80%. DNS were more likely to operate on patients with a history of chronic obstructive pulmonary disease or coronary artery disease when compared with ORS or VRS. Overall outcomes for robotic lobectomy among the groups were similar (Table 2), with the only notable difference in median OR duration. DNS were on average 30 minutes faster than VRS and 9 minutes faster than ORS. When all elements of major morbidity were considered, ORS did have higher rates of pneumothorax requiring chest tube insertion, chest tube at discharge, and postoperative respiratory failure. For 30-day mortality the rates of initial and sustained proficiency were high in all transition groups (Table 3). This improved to 100% proficiency in all groups by 50 cases. For perioperative transfusion the rates of initial and sustained proficiency were similarly high across all 3 transition groups (Table 4). Nearly all surgeons were proficient by 20 cases and 100% were proficient at 50 cases. Proficiency rates for major morbidity were more variable (Table 5). Rates of initial and sustained proficiency were lower in all groups, with only 40% of DNS reaching
30-day mortality Length of stay, d OR duration, min Air leak >5 d Chest tube at discharge Pneumonia Adult respiratory distress syndrome Respiratory failure Pneumothorax requiring chest tube reinsertion Initial ventilatory support >48 h Reintubation Atrial arrhythmia requiring treatment Unexpected admission to the intensive care unit
1017
DNS
ORS
VRS
P Value
13/1169 (1)
22/1767 (1)
26/2313 (1)
.92
5.2 247
4.9 256
5.3 277
.26 <.001
153/462 (33) 190/643 (30) 285/898 (32) 135/1219 (11) 238/1719 (14) 220/2353 (9)
.43 <.001
41/462 (9) 5/462 (1)
61/642 (9.5) 14/642 (2)
79/898 (8.8) 9/898 (1)
.88 .12
22/448 (5)
46/565 (8)
41/839 (5)
.02
55/462 (12)
83/642 (13)
77/898 (9)
.02
4/462 (.9)
4/642 (.6)
11/898 (1.2)
.48
11/204 (5) 38/409 (9) 28/465 (6) 114/461 (25) 173/649 (27) 255/898 (28)
.1 .34
44/861 (5)
72/1361 (5)
69/1740 (4)
.18
Values are n/N (%) unless otherwise indicated. DNS, de novo surgeons; OR, operating room; ORS, open-to-robotic surgeons; VRS, video-assisted thoracoscopic surgery-to-robotic surgeons.
proficiency compared with 67% and 69% of ORS and VRS, respectively. ORS improved most rapidly and completely in this category, with 100% proficiency at 20 cases. DNS and VRS both showed improvement at 20 cases but had not achieved 100% proficiency at 50 cases. OR duration proved to be the most challenging for all groups (Table 6, Figures 2A-2F). DNS demonstrated the highest initial and sustained proficiency at 40%, whereas ORS were the lowest at 14%, followed by VRS at 21%. By 20 cases DNS improved by 7%, whereas ORS more than doubled the percentage of proficient surgeons. At 50 cases DNS had achieved 100% proficiency. VRS seemed to gain the most proficiency between 20 and 50 cases.
Comment The primary finding in this study is that a surgeon’s previous approach to lobectomy does not exert a uniform effect on development of proficiency during transition to robotic lobectomy. Initial and sustained proficiency varied for each outcome metric, but all transition groups
GENERAL THORACIC
Ann Thorac Surg 2019;108:1013-20
GENERAL THORACIC
1018
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
Ann Thorac Surg 2019;108:1013-20
Table 3. Outcomes of CUSUM Analysis by Transition Group: 30-Day Mortality (Overall Rate 1.1%, Target Rate 1%, Unacceptable Rate 5%)
Transition Group
Initial and Sustained Proficiency
Proficient by 20th Case
Proficient by 50th Casea
93 95 86
93 95 97
100 100 100
DNS ORS VRS
Table 5. Outcomes of CUSUM Analysis Analyzed by Transition Group: Major Morbidity (Overall Rate 18%, CUSUM Target Rate 10%, Unacceptable Rate 15%)
Transition Group
Initial and Sustained Proficiency
Proficient by 20th Case
Proficient by 50th Casea
40 67 69
93 100 86
89 100 92
DNS ORS VRS
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
Values are percentages.
Values are percentages.
CUSUM, cumulative sum.
CUSUM, cumulative sum.
improved with increasing experience, resulting in quality being maintained during the transition. For 30-day mortality and transfusion all groups demonstrated high early and consistent proficiency. Comparatively proficiency for major morbidity demonstrated a more apparent learning curve, with ORS demonstrating the most rapid and complete attainment of proficiency. The most striking difference among groups was the ability to attain proficiency with an operative time target of 250 minutes. Only DNS were proficient in this outcome at the 50-case threshold. Our study provides additional context for surgeons wishing to transition to robotic lobectomy, particularly in terms of operative time. Several single-center experiences have identified 20 cases as an important threshold during transition to robotic lobectomy. Veronesi and coworkers7 identified a learning curve of 18 cases before operative time decreased during transition from open thoracotomy. Others have reported operative times of 256 minutes in the first 21 cases that decreased to 187 minutes in the next 22 cases when moving from thoracotomy to robotic approaches.7,17 Similarly operative times of 240 minutes in the initial 40 robotic cases have been reported during the transition from VATS.18 In a meta-analysis Cao and coworkers19 suggested that 30 cases were needed to reduce conversions and to shorten operative times. These data
suggest that 20 cases is an important hurdle during transition to robotic lobectomy. Comparatively our results suggest that 50 cases are required to develop proficiency for most surgeons regardless of prior experience. Surgeons seeking to transition to robotic lobectomy must be invested for at least 50 cases, particularly those with open and VATS backgrounds. Moreover, previous studies using CUSUM analysis identified 50 cases as an institutional proficiency threshold for VATS lobectomy. Highvolume institutions were also more likely to attain proficiency at around 50 cases.9 Like operative duration, major morbidity (predominantly pulmonary complications) allowed the identification of surgeons who gained proficiency, whereas transfusion and 30-day mortality at current thresholds do not discriminate proficiency very well. Several observations can be made based on surgeon’s prior experience to inform training and research. Open surgeons demonstrated lower operative time proficiency; however their prior experience appears to translate into proficiency for morbidity, mortality, and transfusion. Although major morbidity proficiency was high, ORS also demonstrated more respiratory complications (specifically related to air leak). Focused training to address this might emphasize technical robotic skills such as tissue
Table 4. Outcomes of CUSUM Analysis Analyzed by Transition Group: Perioperative Transfusion (Overall Rate 3.5%, CUSUM Target Rate 5%, Unacceptable Rate 10%)
Table 6. Outcomes of CUSUM Analysis Analyzed by Transition Group: OR Duration (Median Operative Time 253 Minutes, Target Operative Time <250 Minutes)
Transition Group
Initial and Sustained Proficiency
DNS ORS VRS
93 90 90
Proficient by Proficient by 20th Case 50th Casea 100 95 97
100 100 100
Transition Group DNS ORS VRS
Initial and Sustained Proficient by Proficiency 20th Case 40 14 21
47 29 21
Proficient by 50th Casea 100 57 67
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
Values are percentages.
Values are percentages.
CUSUM, cumulative sum.
CUSUM, cumulative sum; OR, operating room.
handling without haptic feedback and operative sequence compared with thoracotomy. VATS surgeons comparatively have higher rates of operative time proficiency, but their lower morbidity proficiency may represent greater initial confidence in a minimally invasive surgical platform, leading to taking on more difficult cases earlier in their experience. Their minimally invasive surgical experience could lead to fewer off-camera or traction injuries, reflected as fewer air leaks and/or chest tube complications in the VRS group. Future education and training systems could focus preferentially on these experience-specific challenges, adapting a core curriculum to smooth the learning curve based on a surgeon’s prior operative approach. It is difficult to be definitive about DNS because they may represent new surgeons or experienced surgeons new to the STS-GTD. If we assume (and it is mostly likely) they are new surgeons, their greater success may be the result of several factors. First, they may have been introduced to robotic surgery as trainees in general or cardiothoracic residency. Second, they may have participated in structured robotic training experiences like the American Association of Thoracic Surgery Graham fellowship. Third, senior mentorship in their early experience may have provided individualized feedback to streamline the transition. Better understanding the success with operative duration and the lower initial morbidity proficiency of this group will inform and focus training programs for DNS. A focus on OR duration is not purely academic, and improvements in training are not exclusively for personal optimization. Robotic lobectomy is associated with higher costs, attributable to the fixed costs of equipment, disposable or limited-reuse instruments, and longer OR times.20,21 Reducing operative times is crucial in bringing the cost of robotic lobectomy in line with VATS. Focused instruction or advanced mentorship for ORS and VRS to reduce variability in OR duration proficiency may help to reduce costs. Personalized tracking and earlier identification of proficiency and loss of proficiency using CUSUM analysis would allow surgeons and tutors to rapidly provide corrective feedback and observe successful stabilization. Surgeons who demonstrate and sustain proficiency in 1 element can focus their efforts elsewhere, developing a more individualized and potentially self-driven curricula. Those who do not reach proficiency, on the other hand, could focus on guided training and retraining. These techniques could be paired with existing proctoring/mentoring relationships and peer video review as part of a surgeon’s initial and ongoing training with the robot. It should be noted that proficiency in this context is not a surrogate for robotic credentialing or privileging at the hospital level. Proficiency is a dynamic and multifactorial quality that can fluctuate in response to surgeon skill and patient characteristics. CUSUM analysis monitors discrete elements to track outcomes, and as our data highlight surgeons can be proficient in 1 area while still improving in another. The apparent value of CUSUM
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
1019
data at an institutional level is to provide real-time outcomes of tracking and feedback to enhance surgeon performance. The granularity and complexity of that data would make case-threshold privileging essentially meaningless and argues in favor of comprehensive outcomes monitoring. There are several limitations to this study primarily related to the use of a large database. First, our data are limited to the STS-GTD and therefore does not capture the robotic experience of institutions outside of this database. Second, the data available did not allow us to determine a surgeon’s prior robotic experience (lobectomy or otherwise) before entering the STS-GTD. DNS were therefore considered “new surgeons” to construct reasonable comparison groups. Third, operative duration was calculated from procedure start and stop times and does not reflect all procedures (bronchoscopy, invasive mediastinal staging) performed under the same anesthetic. Tracking console time directly in future studies could address this variability. Fourth, stratification was not performed based on the specific lobectomy performed. Fifth, measures of oncologic quality such as lymph node counts and rates of nodal upstaging were not available in the STS-GTD for all years of study and therefore were not used. Sixth, by aggregating surgeon CUSUM outcomes our analysis fails to capture variations in proficiency that occur at the individual level and have been highlighted in other CUSUMbased studies.9 Finally, we recognize that selection bias is an inherent challenge in assessing proficiency, and our dataset selected for those surgeons sufficiently interested in robotic lobectomy to complete a training course, proctoring, and at least 20 robotic lobectomies. In conclusion, the prior experience of surgeons transitioning to robotic lobectomy translated into different rates of proficiency depending on the outcome measured. Safety and quality as measured by 30-day mortality, transfusion rates, and major morbidity were attained by all surgeons at around 20 cases, whereas proficiency in operative time remains an elusive goal for many surgeons even at 50 cases. Increasing robotic experience over time leads to success at meeting established benchmarks, and tracking these outcomes using CUSUM analysis is a useful method to monitor achievement and maintenance of proficiency. As additional criteria are identified to further define proficiency with the robotic platform, integration of this monitoring methodology into training programs will help to personalize and streamline the transition to robotic lobectomy. The authors wish to thank Kari Chansky for her guidance and advice on our statistical analysis and manuscript preparation and Dr Varun Puri for sharing his methodology and knowledge of CUSUM analysis during our analysis. Dr Louie discloses a restricted research grant from Intuitive Surgical that supported acquisition of the STS participant user file and statistical analysis. Intuitive Surgical had no direct input or involvement in the design or conduct of the research or its conclusions.
GENERAL THORACIC
Ann Thorac Surg 2019;108:1013-20
GENERAL THORACIC
1020
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
References 1. Oh DS, Reddy RM, Gorrepati ML, Mehendale S, Reed MF. Thoracoscopic and open lobectomy : propensity-matched analysis of recent premier data. Ann Thorac Surg. 2018;104: 1733-1740. 2. Rajaram R, Mohanty S, Bentrem DJ, et al. Nationwide assessment of robotic. Ann Thorac Surg. 2018;103:1092-1100. 3. Louie BE, Farivar AS, Aye RW, Valli eres E. Early experience with robotic lung resection results in similar operative outcomes and morbidity when compared with matched videoassisted thoracoscopic surgery cases. Ann Thorac Surg. 2012;93:1598-1605. 4. Louie BE. Robotic lobectomy for non-small cell lung cancer. Indian J Surg Oncol. 2013;4:125-131. 5. Lee BE, Korst RJ, Kletsman E, Rutledge JR. Transitioning from video-assisted thoracic surgical lobectomy to robotics for lung cancer: are there outcomes advantages ? J Thorac Cardiovasc Surg. 2014;147:724-729. 6. Pardolesi A, Park B, Petrella F, Borri A, Gasparri R, Veronesi G. Robotic anatomic segmentectomy of the lung: technical aspects and initial results. Ann Thorac Surg. 2012;94: 929-934. 7. Veronesi G. Robotic lobectomy and segmentectomy for lung cancer: results and operating technique. J Thorac Dis. 2015;7: 122-130. 8. Louie BE, Wilson JL, Kim S, et al. Comparison of videoassisted thoracoscopic surgery and robotic approaches for clinical stage I and stage II non-small cell lung cancer using The Society of Thoracic Surgeons Database. Ann Thorac Surg. 2016;102:917-924. 9. Puri V, Gaissert HA, Wormuth DW, et al. Defining proficiency for The Society of Thoracic Surgeons participants performing thoracoscopic lobectomy. Ann Thorac Surg. 2019;107:202-208. 10. Hu Y, Jolissaint JS, Ramirez A, Gordon R, Yang Z, Sawyer RG. Cumulative sum: a proficiency metric for basic endoscopic training. J Surg Res. 2014;192:62-67.
Ann Thorac Surg 2019;108:1013-20
11. Hu Y, Puri V, Crabtree TD, et al. Attaining proficiency with endobronchial ultrasound-guided transbronchial needle aspiration. J Thorac Cardiovasc Surg. 2013;146:1387-1392. 12. Steiner SH, Cook RJ. Monitoring surgical performance using risk-adjusted cumulative sum charts. Biostatistics. 2000;1:441452. 13. Waller HM, Connor SJ. Cumulative sum (Cusum) analysis provides an objective measure of competency during training in endoscopic retrograde. HPB (Oxford). 2009;11:565569. 14. Bolsin S, Colson M. The use of the Cusum technique in the assessment of trainee competence in new. Int J Qual Health Care. 2000;12:433-438. 15. Fernandez FG, Kosinski AS, Burfeind W, et al. STS lung cancer resection risk model: higher quality data and superior outcomes. Ann Thorac Surg. 2017;102:370-377. 16. Farivar AS, Cerfolio RJ, Vallieres E, et al. Comparing robotic lung resection with thoracotomy and video-assisted thoracoscopic surgery cases entered into the Society of Thoracic Surgeons Database. Innovations. 2014;9:10-15. 17. Oh DS, Cho I, Karamian B, DeMeester SR. Early adoption of robotic pulmonary lobectomy: feasibility and initial outcomes. Am Surg. 2013;79:1075-1081. 18. Jang H, Lee H, Park SY, ZJ. Comparison of the early robotassisted lobectomy experience to video-assisted thoracic surgery lobectomy for lung cancer. Innovations. 2011;6:305-310. 19. Cao C, Manganas C, Ang SC, Yan TD. A systematic review and meta-analysis on pulmonary resections by robotic videoassisted thoracic surgery. Ann Cardiothorac Surg. 2012;1:3-10. 20. Swanson SJ, Miller DL, Mckenna J, et al. Comparing robotassisted thoracic surgical lobectomy with conventional video-assisted thoracic surgical lobectomy and wedge resection: results from a multihospital database (Premier). J Thorac Cardiovasc Surg. 2011;147:929-937. 21. Deen SA, Wilson JL, Wilshire CL, et al. Defining the cost of care for lobectomy and segmentectomy: a comparison of open, video-assisted thoracoscopic, and robotic approaches. Ann Thorac Surg. 2014;97:1000-1007.
Background. Robotic lobectomy represents a paradigm shift for many surgeons. It is unknown if a surgeon’s prior operative approach influences development of proficiency. We compared outcomes based on prior lobectomy experience and used cumulative sum analysis to assess proficiency. Methods. Using The Society of Thoracic Surgeons General Thoracic Database we grouped surgeons as de novo, open-to-robotic, or video-assisted thoracoscopic surgery (VATS)-to-robotic. Operative time, blood transfusion, mortality, and major morbidity were primary outcomes. Unacceptable and acceptable thresholds were determined by review of the literature. Proficiency was defined as 20 consecutive cases without crossing an upper control line. Surgeons were assessed individually, and proficiency was assessed by transition group. Results. From 2009 to 2016, 271 surgeons performed 5619 robotic lobectomies for clinical stage I/II non–small cell lung cancer. Of these, 65 surgeons (24%) performed
‡20 lobectomies (4483 cases). Initial proficiency for an operative time target of 250 minutes was 40% for de novo compared with 14% for open-to-robotic and 21% for VATS-to-robotic surgeons, with improvement to 47%, 29%, and 21%, respectively, after 20 cases. Initial and sustained proficiency related to major morbidity was similar for open-to-robotic and VATS-to-robotic but lower for de novo at 40%. After 20 cases most were proficient (de novo, 93%; open-to-robotic, 100%; and VATSto-robotic, 86%). Proficiency for 30-day mortality and blood transfusion was high in all groups. Conclusions. Outcomes among all transition groups improved with experience. Operating room duration proficiency was challenging for all groups. Cumulative sum may be useful to monitor proficiency in not only subsequent studies but in clinical practice.
P
Most surgeons transitioning to robotics underwent a combination of didactic and hands-on training, including online modules, in-person orientation, basic training courses, and proctored cases. Advanced training classes and focused proctoring have become available. Early robotic adopters reported that the “learning curve” on this training pathway was anywhere from 6 cases for VATS surgeons3 to 18 cases for open surgeons6 based on operative time. How surgeons progress after this initial learning curve and the quality of care during the transition to robotic lobectomy are unknown. Our experience suggests that a surgeon’s prior operative experience and approach (open or VATS) might influence development of proficiency during and after the robotic lobectomy learning curve, with open surgeons taking longer to develop proficiency
ulmonary lobectomy is performed using open, videoassisted thoracoscopic surgery (VATS) or robotic approaches.1,2 These recent studies confirmed that minimally invasive techniques are the dominant approach for early-stage lung cancer and interest in robotic lobectomy growing as an alternative. Those who elect to transition to robotics find that it represents a paradigm shift. Certain elements of minimally invasive surgical procedures, small incisions, and access ports are familiar to most. The surgeon–console interface and the remote patient cart, however, can require new skills ranging from direct surgeon control of all instruments and camera to the absence of haptic feedback.3-5
Accepted for publication Apr 11, 2019.
(Ann Thorac Surg 2019;108:1013-20) Ó 2019 by The Society of Thoracic Surgeons
Presented at the Sixty-fifth Annual Meeting of The Southern Thoracic Surgical Association, Amelia Island, FL, Nov 7-10, 2018. Address correspondence to Dr Louie, Division of Thoracic Surgery, Swedish Cancer Institute, Ste 900, 1101 Madison St, Seattle, WA 98105; email: [email protected].
Ó 2019 by The Society of Thoracic Surgeons Published by Elsevier Inc.
Dr Louie discloses a financial relationship with Intuitive Surgical.
0003-4975/$36.00 https://doi.org/10.1016/j.athoracsur.2019.04.046
GENERAL THORACIC
Proficiency of Robotic Lobectomy Based on Prior Surgical Technique in The Society of Thoracic Surgeons General Thoracic Database
GENERAL THORACIC
1014
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
compared with VATS surgeons. The aim of this study was to define metrics to assess proficiency and describe how prior operative experience affects proficiency during transition to robotic lobectomy.
Patients and Methods The institutional review board of the Swedish Medical Center approved this study with a waiver of individual participant and patient consent because of the retrospective use of deidentified data.
Patients A participant user file was acquired from The Society of Thoracic Surgeons General Thoracic Database (STSGTD). Using versions 2.081-2.3 the database was queried for patients undergoing primary lobectomy for non–small cell lung cancer from 2009 to 2016. To maintain confidentiality a participant user file was created with cases ordered by surgery date and date replaced by surgery year. Each surgeon’s identifier number was also replaced with a deidentified surgeon identification for the study. Robotic lobectomies were identified using the “robotic technology–assisted” field. The participant user file was received with 72,924 lobectomies (Figure 1). Patients with missing data or clinical stages III or greater were excluded. Patients undergoing VATS or open lobectomies were set aside and used to define surgeons’ prior experience. This left 5561 robotic lobectomies for clinical stage I/II non–small cell lung cancer. To account for the learning curve of 20 cases as suggested by Veronesi and coworkers,7 cases performed
Ann Thorac Surg 2019;108:1013-20
by surgeons who completed <20 robotic lobectomies were excluded, leaving 4483 lobectomies. Of the 844 surgeons in the STS-GTD, 210 performed at least 1 robotic lobectomy. Of those, 145 surgeons performed <20 robotic lobectomies and were excluded (88 surgeons completed between 11 and 19 robotic lobectomies). The remaining 65 surgeons were divided into transition groups based on their predominant lobectomy approach in the study period before their first robotic lobectomy. Surgeons performing 50% of lobectomies via thoracotomy were considered open-to-robotic surgeons (ORS) and those who performed 50% of lobectomies using a VATS approach were grouped as VATS-to-robotic surgeons (VRS). If a surgeon’s first recorded lobectomy in the database was robotic, they were labeled as de novo surgeons (DNS).
Outcomes Four outcomes modeled on prior evaluations of robotic quality in the STS-GTD were selected as measures of proficiency during the transition to robotic surgery.8,9 The definitions for these outcomes are outlined in the STSGTD manual. Thirty-day mortality was defined as death after surgery within 30 days of operation. Perioperative transfusion was a dichotomous outcome for red blood cell transfusion intraoperatively or postoperatively. Major morbidity was a composite variable including any occurrence of 1 or more complications, including air leak of more than 5 days, chest tube at discharge, postoperative pneumonia, acute respiratory distress syndrome, respiratory failure, pneumothorax requiring chest tube reinsertion, initial ventilatory support of more than
Figure 1. Flow diagram of study population selection. The Society of Thoracic Surgeons General Thoracic Database was queried for all lobectomies between 2009 and 2016. Cases were selected by complete data, clinical stage (cStage) I/II, and robotic approach. Of that group, surgeons who performed 20 robotic lobectomies were included for the final analysis. Yellow boxes indicate the number of lobectomy cases, and white boxes indicate number of surgeons by transition group. Each surgeon’s identifier number was replaced with a deidentified surgeon identification for the study. (DNS, de novo surgeons; NSCLC, non–small cell lung cancer; ORS, open-to-robotic surgeons; VATS, video-assisted thoracoscopic surgery; VRS, VATS-to-robotic surgeons.)
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
1015
Figure 2. Representative cumulative sum (CUSUM) curves. (A) De novo surgeons (DNS) who achieved proficiency. (B) DNS who did not achieve proficiency. (C) Open-to-robotic surgeons (ORS) who achieved proficiency. (D) ORS who did not achieve proficiency. (E) Video-assisted thoracoscopic surgery-to-robotic surgeons (VRS) who achieved proficiency. (F) VRS who did not achieve proficiency. Graphs show CUSUM score in relation to number of cases. Initial control line indicated by dashed line. Each surgeon’s identifier number was replaced with a deidentified surgeon identification for the study. (OR, operating room.)
GENERAL THORACIC
Ann Thorac Surg 2019;108:1013-20
GENERAL THORACIC
1016
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
48 hours, reintubation, atrial fibrillation, and unexpected intensive care unit admission. Operating room (OR) duration was calculated as the difference between procedure start and stop times.
Statistical Analysis To assess changes in proficiency, cumulative sum (CUSUM) analysis was used. CUSUM is a sequential analysis technique used in manufacturing to track outcomes for a process over time. These methods have been adapted for evaluating performance in surgical settings.10-12 The CUSUM analysis relies on the specifications of a target value and a known or reliable estimate of the standard deviation. Undesirable results cause the CUSUM score to increase (upward deflection in graphic form), and desirable results cause the overall score to decrease (downward deflection) (Figures 2A-2F). Acceptable failure rates are defined a priori and along with Type I and II error rates identify a threshold (“control” or “alarm” lines) that when crossed in an upward direction indicate a loss of proficiency and in the downward direction signal a correction. Because of the cumulative nature of the score, control lines are reset each time a threshold is crossed, with ongoing failure indicated by sequential upward crossing of control lines. Similarly, each time a control line is crossed in a downward direction, the upper control line resets at a lower value.9,13,14 Acceptable and unacceptable failure rates for each of the 4 outcomes were determined by review of the quality and safety literature for robotic and VATS lobectomy. The target rate for 30-day mortality was set at 1%, with an unacceptable rate of 5%. Perioperative transfusion target rate was set at 5%, with an unacceptable rate of 10%. The major morbidity target rate was set at 10%, with an unacceptable rate of 15%. These values were selected from the report by Fernandez and coworkers15 using the STS database and confirmed in other studies using a similar composite for major morbidity.1,8 The target time for robotic lobectomy (all lobes) was set at <250 minutes, with an unacceptable rate of 250 minutes. This was Figure 3. Percent of lobectomy by approach between 2009 and 2016. (VATS, video-assisted thoracoscopic surgery.)
Ann Thorac Surg 2019;108:1013-20
determined from studies comparing operative time for VATS and robotic lobectomy that ranged between 170 and 240 minutes.5,8,16 Type I and II error rates were set at .1 for binary outcomes. A surgeon was defined as proficient when 20 lobectomies were performed without crossing an upper control line.7,16 Initial and sustained proficiency was defined as achievement of proficiency without repeated crossing of upper control lines, with surgeons serving as their own internal controls. All surgeons included in the CUSUM analysis completed 20 lobectomies. A smaller group of surgeons completed 50 robotic lobectomies. Groups were compared using c2 tests for categorical variables and 2-sample analysis of variance for continuous variables. Statistical analyses were performed using Microsoft Office Excel (Microsoft, Redmond, WA) and SAS (SAS institute Inc, Cary, NC). CUSUM analysis was performed using R (The R Foundation, Vienna, Austria).
Results Over the study period robotic lobectomy increased as a proportion of lobectomies from <1% in 2009 to 18.1% in 2016 (Figure 3). Open lobectomy decreased from 53.3% to 28.0%, with the proportion of VATS lobectomies stabilizing between 2014 and 2016. The 65 surgeons completing 4483 robotic lobectomies were divided into transition groups: 15 DNS (1119 lobectomies), 21 ORS (1511 lobectomies), and 29 VRS (1853 lobectomies). A higher proportion of patients in the DNS group were women; however body mass index, age, and performance status across groups appeared to be similar (Table 1). There was not a statistically significant difference in clinical or pathologic stage of resected tumors. There were statistical differences in forced expiratory volume in the first second of expiration percent predicted, diffusing capacity of the lung for carbon monoxide percent predicted, and American Society of Anesthesiologists class among the groups. DNS had the lowest average forced expiratory volume in the first second of
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
Table 1. Study Population Patient Characteristics by Transition Group
Table 2. Overall Outcomes for Robotic Lobectomy by Transition Group
Patient Characteristics
Outcomes
DNS ORS VRS (n ¼ 1234) (n ¼ 1818) (n ¼ 2509) P Value
Female 485 (60.7) 795 (56.3) Median age, y 69 68 26.6 26.8 Median body mass index, kg/m2 Performance status (Zubrod) 0-1 1175 (95.2) 1708 (94.0) >1 59 (4.8) 110 (6.0) pT1-2 tumor 1102 (89.3) 1625 (89.4) 82 84 Mean FEV1, % predicted Mean DLCO, % 76 74 predicted ASA class III 924 (75.0) 1447 (80.0) Never smoker 204 (16.5) 329 (18.1) Chronic obstructive 503 (40.8) 693 (38.1) pulmonary disease Coronary artery 337 (27.3) 451 (24.8) disease Dialysis 4 (.3) 7 (.4) Diabetes 205 (16.6) 365 (20.1)
1110 (55.8) 69 27.0
.02
2430 (96.9) 79 (3.2) 2193 (87.4) 86
.62 <.01 .74 <.01
77
.03
1982 (79.0) 448 (17.9) 877 (35.0)
.01 .56 .02
532 (21.2)
<.01
18 (.7) 485 (19.3)
.18 .08
Values are n (%) unless otherwise indicated. ASA, American Society of Anesthesiologists; DLCO, diffusing capacity of the lung for carbon monoxide; DNS, de novo surgeons; FEV1, forced expiratory volume in the first second of expiration; ORS, open-to-robotic surgeons; VRS, video-assisted thoracoscopic surgery-to-robotic surgeons.
expiration percent predicted (82%), but diffusing capacity of the lung for carbon monoxide percent predicted was lowest for ORS. ORS also had a greater proportion of American Society of Anesthesiologists class III patients at 80%. DNS were more likely to operate on patients with a history of chronic obstructive pulmonary disease or coronary artery disease when compared with ORS or VRS. Overall outcomes for robotic lobectomy among the groups were similar (Table 2), with the only notable difference in median OR duration. DNS were on average 30 minutes faster than VRS and 9 minutes faster than ORS. When all elements of major morbidity were considered, ORS did have higher rates of pneumothorax requiring chest tube insertion, chest tube at discharge, and postoperative respiratory failure. For 30-day mortality the rates of initial and sustained proficiency were high in all transition groups (Table 3). This improved to 100% proficiency in all groups by 50 cases. For perioperative transfusion the rates of initial and sustained proficiency were similarly high across all 3 transition groups (Table 4). Nearly all surgeons were proficient by 20 cases and 100% were proficient at 50 cases. Proficiency rates for major morbidity were more variable (Table 5). Rates of initial and sustained proficiency were lower in all groups, with only 40% of DNS reaching
30-day mortality Length of stay, d OR duration, min Air leak >5 d Chest tube at discharge Pneumonia Adult respiratory distress syndrome Respiratory failure Pneumothorax requiring chest tube reinsertion Initial ventilatory support >48 h Reintubation Atrial arrhythmia requiring treatment Unexpected admission to the intensive care unit
1017
DNS
ORS
VRS
P Value
13/1169 (1)
22/1767 (1)
26/2313 (1)
.92
5.2 247
4.9 256
5.3 277
.26 <.001
153/462 (33) 190/643 (30) 285/898 (32) 135/1219 (11) 238/1719 (14) 220/2353 (9)
.43 <.001
41/462 (9) 5/462 (1)
61/642 (9.5) 14/642 (2)
79/898 (8.8) 9/898 (1)
.88 .12
22/448 (5)
46/565 (8)
41/839 (5)
.02
55/462 (12)
83/642 (13)
77/898 (9)
.02
4/462 (.9)
4/642 (.6)
11/898 (1.2)
.48
11/204 (5) 38/409 (9) 28/465 (6) 114/461 (25) 173/649 (27) 255/898 (28)
.1 .34
44/861 (5)
72/1361 (5)
69/1740 (4)
.18
Values are n/N (%) unless otherwise indicated. DNS, de novo surgeons; OR, operating room; ORS, open-to-robotic surgeons; VRS, video-assisted thoracoscopic surgery-to-robotic surgeons.
proficiency compared with 67% and 69% of ORS and VRS, respectively. ORS improved most rapidly and completely in this category, with 100% proficiency at 20 cases. DNS and VRS both showed improvement at 20 cases but had not achieved 100% proficiency at 50 cases. OR duration proved to be the most challenging for all groups (Table 6, Figures 2A-2F). DNS demonstrated the highest initial and sustained proficiency at 40%, whereas ORS were the lowest at 14%, followed by VRS at 21%. By 20 cases DNS improved by 7%, whereas ORS more than doubled the percentage of proficient surgeons. At 50 cases DNS had achieved 100% proficiency. VRS seemed to gain the most proficiency between 20 and 50 cases.
Comment The primary finding in this study is that a surgeon’s previous approach to lobectomy does not exert a uniform effect on development of proficiency during transition to robotic lobectomy. Initial and sustained proficiency varied for each outcome metric, but all transition groups
GENERAL THORACIC
Ann Thorac Surg 2019;108:1013-20
GENERAL THORACIC
1018
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
Ann Thorac Surg 2019;108:1013-20
Table 3. Outcomes of CUSUM Analysis by Transition Group: 30-Day Mortality (Overall Rate 1.1%, Target Rate 1%, Unacceptable Rate 5%)
Transition Group
Initial and Sustained Proficiency
Proficient by 20th Case
Proficient by 50th Casea
93 95 86
93 95 97
100 100 100
DNS ORS VRS
Table 5. Outcomes of CUSUM Analysis Analyzed by Transition Group: Major Morbidity (Overall Rate 18%, CUSUM Target Rate 10%, Unacceptable Rate 15%)
Transition Group
Initial and Sustained Proficiency
Proficient by 20th Case
Proficient by 50th Casea
40 67 69
93 100 86
89 100 92
DNS ORS VRS
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
Values are percentages.
Values are percentages.
CUSUM, cumulative sum.
CUSUM, cumulative sum.
improved with increasing experience, resulting in quality being maintained during the transition. For 30-day mortality and transfusion all groups demonstrated high early and consistent proficiency. Comparatively proficiency for major morbidity demonstrated a more apparent learning curve, with ORS demonstrating the most rapid and complete attainment of proficiency. The most striking difference among groups was the ability to attain proficiency with an operative time target of 250 minutes. Only DNS were proficient in this outcome at the 50-case threshold. Our study provides additional context for surgeons wishing to transition to robotic lobectomy, particularly in terms of operative time. Several single-center experiences have identified 20 cases as an important threshold during transition to robotic lobectomy. Veronesi and coworkers7 identified a learning curve of 18 cases before operative time decreased during transition from open thoracotomy. Others have reported operative times of 256 minutes in the first 21 cases that decreased to 187 minutes in the next 22 cases when moving from thoracotomy to robotic approaches.7,17 Similarly operative times of 240 minutes in the initial 40 robotic cases have been reported during the transition from VATS.18 In a meta-analysis Cao and coworkers19 suggested that 30 cases were needed to reduce conversions and to shorten operative times. These data
suggest that 20 cases is an important hurdle during transition to robotic lobectomy. Comparatively our results suggest that 50 cases are required to develop proficiency for most surgeons regardless of prior experience. Surgeons seeking to transition to robotic lobectomy must be invested for at least 50 cases, particularly those with open and VATS backgrounds. Moreover, previous studies using CUSUM analysis identified 50 cases as an institutional proficiency threshold for VATS lobectomy. Highvolume institutions were also more likely to attain proficiency at around 50 cases.9 Like operative duration, major morbidity (predominantly pulmonary complications) allowed the identification of surgeons who gained proficiency, whereas transfusion and 30-day mortality at current thresholds do not discriminate proficiency very well. Several observations can be made based on surgeon’s prior experience to inform training and research. Open surgeons demonstrated lower operative time proficiency; however their prior experience appears to translate into proficiency for morbidity, mortality, and transfusion. Although major morbidity proficiency was high, ORS also demonstrated more respiratory complications (specifically related to air leak). Focused training to address this might emphasize technical robotic skills such as tissue
Table 4. Outcomes of CUSUM Analysis Analyzed by Transition Group: Perioperative Transfusion (Overall Rate 3.5%, CUSUM Target Rate 5%, Unacceptable Rate 10%)
Table 6. Outcomes of CUSUM Analysis Analyzed by Transition Group: OR Duration (Median Operative Time 253 Minutes, Target Operative Time <250 Minutes)
Transition Group
Initial and Sustained Proficiency
DNS ORS VRS
93 90 90
Proficient by Proficient by 20th Case 50th Casea 100 95 97
100 100 100
Transition Group DNS ORS VRS
Initial and Sustained Proficient by Proficiency 20th Case 40 14 21
47 29 21
Proficient by 50th Casea 100 57 67
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
a Number of surgeons who completed 50 cases: de novo surgeons (DNS), 9; open-to-robotic surgeons (ORS), 7; video-assisted thoracoscopic surgeryto-robotic surgeons (VRS), 12.
Values are percentages.
Values are percentages.
CUSUM, cumulative sum.
CUSUM, cumulative sum; OR, operating room.
handling without haptic feedback and operative sequence compared with thoracotomy. VATS surgeons comparatively have higher rates of operative time proficiency, but their lower morbidity proficiency may represent greater initial confidence in a minimally invasive surgical platform, leading to taking on more difficult cases earlier in their experience. Their minimally invasive surgical experience could lead to fewer off-camera or traction injuries, reflected as fewer air leaks and/or chest tube complications in the VRS group. Future education and training systems could focus preferentially on these experience-specific challenges, adapting a core curriculum to smooth the learning curve based on a surgeon’s prior operative approach. It is difficult to be definitive about DNS because they may represent new surgeons or experienced surgeons new to the STS-GTD. If we assume (and it is mostly likely) they are new surgeons, their greater success may be the result of several factors. First, they may have been introduced to robotic surgery as trainees in general or cardiothoracic residency. Second, they may have participated in structured robotic training experiences like the American Association of Thoracic Surgery Graham fellowship. Third, senior mentorship in their early experience may have provided individualized feedback to streamline the transition. Better understanding the success with operative duration and the lower initial morbidity proficiency of this group will inform and focus training programs for DNS. A focus on OR duration is not purely academic, and improvements in training are not exclusively for personal optimization. Robotic lobectomy is associated with higher costs, attributable to the fixed costs of equipment, disposable or limited-reuse instruments, and longer OR times.20,21 Reducing operative times is crucial in bringing the cost of robotic lobectomy in line with VATS. Focused instruction or advanced mentorship for ORS and VRS to reduce variability in OR duration proficiency may help to reduce costs. Personalized tracking and earlier identification of proficiency and loss of proficiency using CUSUM analysis would allow surgeons and tutors to rapidly provide corrective feedback and observe successful stabilization. Surgeons who demonstrate and sustain proficiency in 1 element can focus their efforts elsewhere, developing a more individualized and potentially self-driven curricula. Those who do not reach proficiency, on the other hand, could focus on guided training and retraining. These techniques could be paired with existing proctoring/mentoring relationships and peer video review as part of a surgeon’s initial and ongoing training with the robot. It should be noted that proficiency in this context is not a surrogate for robotic credentialing or privileging at the hospital level. Proficiency is a dynamic and multifactorial quality that can fluctuate in response to surgeon skill and patient characteristics. CUSUM analysis monitors discrete elements to track outcomes, and as our data highlight surgeons can be proficient in 1 area while still improving in another. The apparent value of CUSUM
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
1019
data at an institutional level is to provide real-time outcomes of tracking and feedback to enhance surgeon performance. The granularity and complexity of that data would make case-threshold privileging essentially meaningless and argues in favor of comprehensive outcomes monitoring. There are several limitations to this study primarily related to the use of a large database. First, our data are limited to the STS-GTD and therefore does not capture the robotic experience of institutions outside of this database. Second, the data available did not allow us to determine a surgeon’s prior robotic experience (lobectomy or otherwise) before entering the STS-GTD. DNS were therefore considered “new surgeons” to construct reasonable comparison groups. Third, operative duration was calculated from procedure start and stop times and does not reflect all procedures (bronchoscopy, invasive mediastinal staging) performed under the same anesthetic. Tracking console time directly in future studies could address this variability. Fourth, stratification was not performed based on the specific lobectomy performed. Fifth, measures of oncologic quality such as lymph node counts and rates of nodal upstaging were not available in the STS-GTD for all years of study and therefore were not used. Sixth, by aggregating surgeon CUSUM outcomes our analysis fails to capture variations in proficiency that occur at the individual level and have been highlighted in other CUSUMbased studies.9 Finally, we recognize that selection bias is an inherent challenge in assessing proficiency, and our dataset selected for those surgeons sufficiently interested in robotic lobectomy to complete a training course, proctoring, and at least 20 robotic lobectomies. In conclusion, the prior experience of surgeons transitioning to robotic lobectomy translated into different rates of proficiency depending on the outcome measured. Safety and quality as measured by 30-day mortality, transfusion rates, and major morbidity were attained by all surgeons at around 20 cases, whereas proficiency in operative time remains an elusive goal for many surgeons even at 50 cases. Increasing robotic experience over time leads to success at meeting established benchmarks, and tracking these outcomes using CUSUM analysis is a useful method to monitor achievement and maintenance of proficiency. As additional criteria are identified to further define proficiency with the robotic platform, integration of this monitoring methodology into training programs will help to personalize and streamline the transition to robotic lobectomy. The authors wish to thank Kari Chansky for her guidance and advice on our statistical analysis and manuscript preparation and Dr Varun Puri for sharing his methodology and knowledge of CUSUM analysis during our analysis. Dr Louie discloses a restricted research grant from Intuitive Surgical that supported acquisition of the STS participant user file and statistical analysis. Intuitive Surgical had no direct input or involvement in the design or conduct of the research or its conclusions.
GENERAL THORACIC
Ann Thorac Surg 2019;108:1013-20
GENERAL THORACIC
1020
FECZKO ET AL ROBOTIC LOBECTOMY PROFICIENCY
References 1. Oh DS, Reddy RM, Gorrepati ML, Mehendale S, Reed MF. Thoracoscopic and open lobectomy : propensity-matched analysis of recent premier data. Ann Thorac Surg. 2018;104: 1733-1740. 2. Rajaram R, Mohanty S, Bentrem DJ, et al. Nationwide assessment of robotic. Ann Thorac Surg. 2018;103:1092-1100. 3. Louie BE, Farivar AS, Aye RW, Valli eres E. Early experience with robotic lung resection results in similar operative outcomes and morbidity when compared with matched videoassisted thoracoscopic surgery cases. Ann Thorac Surg. 2012;93:1598-1605. 4. Louie BE. Robotic lobectomy for non-small cell lung cancer. Indian J Surg Oncol. 2013;4:125-131. 5. Lee BE, Korst RJ, Kletsman E, Rutledge JR. Transitioning from video-assisted thoracic surgical lobectomy to robotics for lung cancer: are there outcomes advantages ? J Thorac Cardiovasc Surg. 2014;147:724-729. 6. Pardolesi A, Park B, Petrella F, Borri A, Gasparri R, Veronesi G. Robotic anatomic segmentectomy of the lung: technical aspects and initial results. Ann Thorac Surg. 2012;94: 929-934. 7. Veronesi G. Robotic lobectomy and segmentectomy for lung cancer: results and operating technique. J Thorac Dis. 2015;7: 122-130. 8. Louie BE, Wilson JL, Kim S, et al. Comparison of videoassisted thoracoscopic surgery and robotic approaches for clinical stage I and stage II non-small cell lung cancer using The Society of Thoracic Surgeons Database. Ann Thorac Surg. 2016;102:917-924. 9. Puri V, Gaissert HA, Wormuth DW, et al. Defining proficiency for The Society of Thoracic Surgeons participants performing thoracoscopic lobectomy. Ann Thorac Surg. 2019;107:202-208. 10. Hu Y, Jolissaint JS, Ramirez A, Gordon R, Yang Z, Sawyer RG. Cumulative sum: a proficiency metric for basic endoscopic training. J Surg Res. 2014;192:62-67.
Ann Thorac Surg 2019;108:1013-20
11. Hu Y, Puri V, Crabtree TD, et al. Attaining proficiency with endobronchial ultrasound-guided transbronchial needle aspiration. J Thorac Cardiovasc Surg. 2013;146:1387-1392. 12. Steiner SH, Cook RJ. Monitoring surgical performance using risk-adjusted cumulative sum charts. Biostatistics. 2000;1:441452. 13. Waller HM, Connor SJ. Cumulative sum (Cusum) analysis provides an objective measure of competency during training in endoscopic retrograde. HPB (Oxford). 2009;11:565569. 14. Bolsin S, Colson M. The use of the Cusum technique in the assessment of trainee competence in new. Int J Qual Health Care. 2000;12:433-438. 15. Fernandez FG, Kosinski AS, Burfeind W, et al. STS lung cancer resection risk model: higher quality data and superior outcomes. Ann Thorac Surg. 2017;102:370-377. 16. Farivar AS, Cerfolio RJ, Vallieres E, et al. Comparing robotic lung resection with thoracotomy and video-assisted thoracoscopic surgery cases entered into the Society of Thoracic Surgeons Database. Innovations. 2014;9:10-15. 17. Oh DS, Cho I, Karamian B, DeMeester SR. Early adoption of robotic pulmonary lobectomy: feasibility and initial outcomes. Am Surg. 2013;79:1075-1081. 18. Jang H, Lee H, Park SY, ZJ. Comparison of the early robotassisted lobectomy experience to video-assisted thoracic surgery lobectomy for lung cancer. Innovations. 2011;6:305-310. 19. Cao C, Manganas C, Ang SC, Yan TD. A systematic review and meta-analysis on pulmonary resections by robotic videoassisted thoracic surgery. Ann Cardiothorac Surg. 2012;1:3-10. 20. Swanson SJ, Miller DL, Mckenna J, et al. Comparing robotassisted thoracic surgical lobectomy with conventional video-assisted thoracic surgical lobectomy and wedge resection: results from a multihospital database (Premier). J Thorac Cardiovasc Surg. 2011;147:929-937. 21. Deen SA, Wilson JL, Wilshire CL, et al. Defining the cost of care for lobectomy and segmentectomy: a comparison of open, video-assisted thoracoscopic, and robotic approaches. Ann Thorac Surg. 2014;97:1000-1007.