Analysis of the Learning Curve in Telerobotic, Beating Heart Coronary Artery Bypass Grafting: A 90 Patient Experience

Analysis of the Learning Curve in Telerobotic, Beating Heart Coronary Artery Bypass Grafting: A 90 Patient Experience Richard J. Novick, MD, Stephanie A. Fox, RRCP, Bob B. Kiaii, MD, Larry W. Stitt, MS, Reiza Rayman, MD, Kojiro Kodera, MD, Alan H. Menkis, MD, and W. Douglas Boyd, MD Division of Cardiac Surgery, London Health Sciences Center, Department of Clinical Epidemiology and Biostatistics, University of Western Ontario, London, Ontario, Canada, Department of Surgery, Cleveland Clinic Florida, Weston, Florida Background. Recent articles have commented on the learning curve in robotic-assisted coronary artery bypass grafting. We systematically studied this phenomenon using standard statistical and cumulative sum (CUSUM) failure methods. Methods. Ninety patients underwent internal thoracic artery (ITA) takedown and an attempt at ITA to coronary bypass on the beating heart using the Zeus telerobotic system from September 999 to December 00. The rates of mortality and predefined major complications were compared in five quintiles of 8 consecutive patients each and a CUSUM curve was generated for the entire cohort. Results. All patients but one underwent successful endoscopic ITA takedown. Thirteen patients had a totally endoscopic anastomosis, whereas in 6 a small mini-thoracotomy or mini-sternotomy was used. Sixteen patients (7.8%) were converted electively to a sternotomy: patients underwent off-pump and 5 patients For editorial comment see page 663 During the past few years important advances have been made in the development and assessment of off-pump methods of coronary artery bypass grafting (CABG) through a sternotomy [ 3], as well as in limitedaccess myocardial revascularization procedures on the beating heart [4 6]. We performed our first closed-chest left internal thoracic artery (ITA) to left anterior descending (LAD) coronary bypass on the beating heart using robotic telemanipulation in September 999 [7]. We and others [4, 8] have noted that closed-chest, roboticassisted CABG procedures are time intensive, require a significant financial outlay for capital equipment, and may have a significant learning curve despite the efforts Accepted for publication April 3, 003. Address reprint requests to Dr Novick, Division of Cardiac Surgery, London Health Sciences Center, University Campus, P.O. Box 5339, 339 Windermere Road, London, Ontario, Canada N6A 5A5; e-mail: rjnovick@uwo.ca. on-pump surgery. There were no deaths; 3 patients (4.4%) incurred one or more of the major complication(s), including 5,,, 3, and in each of the five quintiles (p 0.39). Standard statistical analyses identified a significant decrease in operating room time (p < 0.000), as well as a decrease in the incidence of an occluded graft or wrong vessel grafted from quintiles to 5(p 0.03). On CUSUM analysis, the failure curve was steep for the first 8 to 0 patients, before moderating its slope for the remainder of the experience. Conclusions. Robotic ITA to coronary bypass on the beating heart has a moderately steep learning curve, which is mitigated by further experience. CUSUM analysis complimented standard statistical methods in detecting a cluster of suboptimal results during the early experience with this procedure. (Ann Thorac Surg 003;76:749 53) 003 by The Society of Thoracic Surgeons of a dedicated surgical, nursing, anesthesiology, and cardiology team. Recent publications from our center have examined the learning curve in a 0-year experience with on-pump cardiac surgery [9], as well as in the transition from on-pump to off-pump CABG [0, ]. In all three of these articles we used the cumulative sum (CUSUM) failure method to detect a cluster of surgical failures and successes in advance of standard statistical analyses. Before embarking on our clinical experience with telerobotic CABG in 999, we decided to assess our learning curve prospectively using both standard statistical and CUSUM failure methods. Our a priori hypothesis was that there would be an important learning curve in the development of this procedure, which would be quickly mitigated by subsequent experience. Patients and Methods Patients selected for telerobotic CABG were assessed preoperatively both by a cardiologist and by the senior Dr Boyd discloses that he has a financial relationship with Computer Motion, Inc. 003 by The Society of Thoracic Surgeons 0003-4975/03/$30.00 Published by Elsevier Inc PII S0003-4975(03)00680-5

750 NOVICK ET AL Ann Thorac Surg ANALYSIS OF THE LEARNING CURVE 003;76:749 53 author. Inclusion criteria were patients with predominantly single vessel coronary artery disease who were experiencing significant angina pectoris despite full medical therapy. Furthermore, these patients coronary anatomy was judged to be not optimally treated by percutaneous coronary intervention at the time of assessment. Exclusion criteria included the need for an emergency operation, a left ventricular ejection fraction less than 30%, a distal coronary target vessel less than.5 mm in size, a previous sternotomy or ipsilateral anterior thoracotomy, and insufficient pulmonary reserve to tolerate one lung ventilation, as assessed on preoperative pulmonary function studies. Patients in this series underwent telerobotic CABG using the ZEUS Robotic Surgical System (Computer Motion, Inc., Goleta, CA). Permission to embark on this trial was granted by the Research Ethics Board of the University of Western Ontario after a detailed review and all patients gave fully informed, written consent for the procedure. The operative methods used have been described in detail in recent publications from our center [7, ]. The outcomes measured in the study included postoperative mortality and the following predefined major complications: perioperative myocardial infarction, reoperation for bleeding, stroke, mediastinitis, sepsis, lifethreatening arrhythmia, new renal failure, need for postoperative intraaortic balloon pump support, the requirement for mechanical ventilation for greater than 48 hours postoperatively, and an occluded graft or wrong vessel grafted on postoperative assessment. The operational definitions of these complications have been defined in previous reports from our center [9, 0]. In addition, we measured operating room times, the requirement for conversion to sternotomy and postoperative intensive care unit and the hospital length of stay. Postoperative coronary and graft arteriography was performed in most patients before discharge home; those who did not undergo postoperative coronary arteriography underwent a detailed Doppler ultrasound assessment of the internal thoracic artery grafts early postoperatively. To more sensitively assess the learning curve in this series, the 90 patients were divided into five quintiles of 8 consecutive patients each. Statistical differences among groups were assessed using Fisher s exact tests for categorical outcomes and analysis of variance for continuous outcomes. Furthermore, a running CUSUM method was used to assess the cumulative failure rate (i.e., the rate of mortality, predefined major complications and poor graft outcomes on postoperative coronary arteriography), as in our recent reports [9 ]. For this study we used a maximum acceptable failure rate of 0% for the sum of all three of these adverse outcomes because this value has been used as a benchmark in our previous studies. Furthermore, recent publications on the early outcomes of CABG procedures have revealed major complication rates of 4% to 5% after on-pump CABG versus 8.8% to 0.6% after off-pump CABG [, ]; our target was directed toward the lower end of this range. In Table. Types of Surgical Procedures in 90 Patients Single direct left ITA to LAD anastomoses 76 Single direct right ITA to RCA or PD 3 anastomoses Right ITA and saphenous vein composite 3 to RCA or PD anastomoses Left ITA and saphenous vein composite to LAD anastomoses Left ITA to LAD and saphenous vein graft to RCA* Saphenous vein grafts to LAD and diagonal* Right ITA and saphenous vein composite to PD and left ITA to LAD* Right ITA to RCA and saphenous vein graft to PD* Left ITA to ramus intermedius and saphenous vein graft to distal LAD* * All double bypasses performed after conversion to sternotomy. ITA internal thoracic artery; LAD left anterior descending coronary artery; RCA right coronary artery; PD posterior descending artery. addition to CUSUM analyses, observed versus expected lengths of postoperative hospital stay were calculated using the multivariable logistic regression model of the Cardiac Care Network of Ontario [3]. A p value less than 0.05 was considered significant. Results From September 999 to December 00, 90 patients were operated on with the intention of undergoing computerassisted, telerobotic CABG on the beating heart at the London Health Sciences Center. The study cohort included 8 males and 8 females, with a median age of 57 years old. The majority of procedures (76/90, 84.4%) were single left ITA to LAD anastomoses. The types of surgical procedures performed in the 4 other patients are listed in Table. In all but one instance the ITA had satisfactory flow after procurement by robotic telemanipulation, whereas in one instance flow was inadequate and a saphenous vein graft to the LAD was used. Thirteen patients had their procedures performed completely endoscopically, with port access and no other incisions. Sixty-one patients required a small minithoracotomy (56 patients) or ministernotomy (5 patients) to facilitate the coronary anastomosis. In 6 patients (7.8%) the procedure was converted electively to a full sternotomy. The reasons for conversion to sternotomy are illustrated in Table. There were no emergent conversions to sternotomy because of hemodynamic decompensation or bleeding. Eleven of 6 converted patients underwent off-pump surgery, whereas 5 patients required on-pump CABG. The operating room times and number of conversions to sternotomy in each 8-patient quintile are illustrated in Table 3. From the first to the last quintile there was a greater than 40% decrease in operating room time, which was highly statistically significant. Furthermore, the stan-

Ann Thorac Surg NOVICK ET AL 003;76:749 53 ANALYSIS OF THE LEARNING CURVE Table. Reasons for Conversion to Sternotomy in 6 Patients Intramyocardial left anterior descending artery 9 More extensive coronary disease than revealed on angiography Inability to tolerate single lung ventilation Thickened pericardium and/or dense adhesions Inadequate intrathoracic working space (obese patient) 75 dard deviations of the mean operating room times decreased by half from quintiles to 5, indicating that these times became more predictable with increasing experience. There was no clinical or statistical difference among quintiles in conversion rates. There were no inhospital or postdischarge deaths in this series. A total of 5 perioperative complications occurred in 3 patients (4.4%). These complications included: 4 perioperative myocardial infarctions, reoperations for bleeding, patients with postoperative mediastinitis, minor stroke, and patient with a bradyarrhythmia requiring insertion of a permanent pacemaker. Five patients experienced poor graft outcomes, as documented on postoperative coronary arteriography. These included 3 patients with occluded grafts (only one of whom experienced a postoperative myocardial infarction), as well as patients in which a parallel diagonal branch of the LAD was grafted inadvertently. All 3 patients with an occluded graft underwent regrafting through a sternotomy during the same hospital stay, with a satisfactory postoperative outcome. The patients with inadvertent grafting of a large diagonal underwent a successful angioplasty and stenting of the LAD lesion, which had involved the origin of the diagonal in question on preoperative coronary arteriography. The incidence of major complications and poor graft outcomes in each 8-patient quintile is illustrated in Table 4. Four of 5 patients with adverse graft outcomes were operated in the first quintile and only patient in quintiles two to five. As indicated in Table 4, there was no statistical difference among groups in the incidence of major complications or adverse graft outcomes. However, if the two types of poor graft outcomes were combined, the difference among quintiles was significant (p 0.03). The median postoperative intensive care unit stay was.0 days in the 90 patients (mean. 0.6 days, range to 5 days). Despite the routine use of postoperative Fig. CUSUM analysis of clinical experience with telerobotic CABG. X axis denotes consecutive patients undergoing this procedure from September 999 to December 00. Y axis denotes the number of adjusted cumulative failures (death or any of the prespecified major complications), assuming a maximum acceptable failure rate of 0%. (CABG coronary artery bypass graft; CUSUM cumulative sum.) cardiac catheterization and the need for reoperation in 5 patients (3 for regrafting and for postoperative bleeding), the median duration of hospital stay was only 4.0 days. Furthermore, the observed versus expected ratio for total postoperative length of stay, based on the logistic regression model of the Cardiac Care Network of Ontario, was 0.76 (95% confidence interval, 0.68 to 0.84). The CUSUM failure curve for the entire cohort is illustrated in Figure. CUSUM analysis revealed a relatively steep curve early in our experience, with five failures (i.e. patients with major complications) in the first 8 patients. The slope of the curve then moderated significantly, such that by the end of the experience there were a total of four adjusted CUSUM failures in the entire cohort, using a maximum acceptable failure rate of 0% as a benchmark. Comment The results of this study demonstrated that there is a significant learning curve in the performance of telerobotic CABG, especially during the first 8 to 0 patients. Complications such as stroke, reoperation for bleeding, and mediastinitis were infrequent in all patient quintiles, whereas 4 of 5 patients with poor graft outcomes were Table 3. Operating Room Times and Conversions to Sternotomy (OPCAB and On-Pump CABG) in Each 8-Patient Quintile Variable Quintile Quintile Quintile 3 Quintile 4 Quintile 5 p Value Operating room time (min) 537 9 a 47 7 47 4 347 70 307 56 0.000 Conversion to OPCAB 3 4 0.47 Conversion to on-pump CABG 0 0.7 a The denotes standard deviations. CABG coronary artery bypass grafting; OPCAB off-pump coronary artery bypass grafting.

75 NOVICK ET AL Ann Thorac Surg ANALYSIS OF THE LEARNING CURVE 003;76:749 53 Table 4. Major Complications and Poor Graft Outcomes in Each 8-Patient Quintile Variable Quintile Quintile Quintile 3 Quintile 4 Quintile 5 p Value Patient with major complications* 5 3 0.39 Patients with occluded grafts 0 0 0 0.30 Patients with wrong vessel grafted 0 0 0 0 0.3 * Predefined major complications included perioperative myocardial infarction, postoperative delirium or stroke, reoperation for bleeding, mediastinitis, sepsis, new renal failure, need for postoperative intraaortic balloon pump support, mechanical ventilation for greater than 48 hours, life-threatening arrhythmia or cardiac arrest, and an occluded graft or wrong vessel grafted on postoperative angiography. If the two categories of poor graft outcomes are combined, the difference among quintiles was statistically significant (p 0.03). operated in the first quintile. All 5 of these patients underwent reintervention during the same hospital stay, 3 by reoperation using a sternotomy and by percutaneous coronary intervention. All 5 patients ultimately had satisfactory outcomes, although patient incurred a small perioperative myocardial infarction. Although the cardiac surgeon is usually the catalyst for the development and refinement of telerobotic CABG, the importance of a dedicated anesthesiology, nursing, cardiology, and postoperative care team cannot be overemphasized. Furthermore, in our institution, surgical endoscopic skills learned from prior experience of harvesting the ITA using a voice-controlled robotic endoscopic positioner and the harmonic scalpel [6] facilitated the transition to full, telerobotic CABG. This previous operative experience and the efforts of a dedicated team led to a highly significant decrease in operating room times throughout our 90 patient experience. Moreover, our anesthesiology and postoperative care teams facilitated the fast-track convalescence of these patients, resulting in relatively brief stays in the intensive care unit and in the hospital, even though most of these patients underwent postoperative cardiac catheterization. The median postoperative length of stay of our patients, at 4.0 days, was approximately 5% less than predicted using the multivariable logistic regression model of the Cardiac Care Network of Ontario. Furthermore, this length of stay was day shorter than that of patients undergoing multivessel off-pump CABG through a sternotomy at our institution []. Moreover, almost all of the 74 patients who underwent limited access, telerobotic procedures were able to return to full activities within weeks of surgery. Other centers have reported their early experience with endoscopic robot-assisted CABG using either onpump, arrested heart methods [8, 4 6], or by closedchest beating heart techniques [4, 7]. Damiano and associates [6] reported an intraoperative anastomotic revision rate of 9% and an incidence of 9% of reoperation for bleeding, versus a 3.3% anastomotic reintervention rate and a.% reoperation for bleeding rate in our series. Recently, Dogan and associates [8] reported an intraoperative conversion to minithoracotomy or sternotomy rate of.3%, which decreased to 5% in the last 0 patients. These investigators noted further that the majority of complications occurred in the first 0 patients of this series and are clearly attributable to the learning curve [8]. The recently reported experience from Leipzig recorded an 8.5% conversion rate to sternotomy or thoracotomy when totally endoscopic robotic-assisted CABG was performed on the arrested heart [4]. On the other hand, 6 of the first 8 patients undergoing an attempt at totally endoscopic CABG on the beating heart required conversion to sternotomy or minithoracotomy. Thus, it is clear that even in the most experienced hands limited-access, telerobotic CABG involves a significant learning curve. Nonetheless, as demonstrated by the Leipzig data [4] and the results of our own study, procedural times and adverse clinical events can be significantly reduced with increasing experience. The CUSUM method was introduced in the 950s as a procedure for detecting changes in the frequency of failures, by the repeated application of a sequential probability ratio test [8, 9]. The CUSUM technique recognizes the importance of time as a hidden variable in clinical studies [0] and avoids statistical problems associated with repeated significance testing []. We have used CUSUM methods since 999 to evaluate patient outcomes and learning curves in adult cardiac surgery [9 ]. Our earlier studies demonstrated that CUSUM methods were a more sensitive indicator of a cluster of surgical failures than standard statistical techniques. In this study the CUSUM curve had a steep slope during the first 8 to 0 patients before moderating. Standard statistical analysis did not reveal a significant difference in cumulative major complication rates between quintiles, although the incidence of combined poor graft outcomes on postoperative coronary arteriography was significantly higher in quintile patients. Furthermore, analysis of variance did demonstrate statistically significant decreases in procedural times from quintiles to 5. Thus, in this study, CUSUM and standard statistical methods provided complimentary information to assist members of the surgical team in gauging the rapidity of the learning curve in a complex, highly innovative surgical procedure. As noted in a recent review on computer-enhanced and telerobotic CABG [], the current role of this procedure has not been fully defined. Although the further development of robotic technology has immense longterm potential to minimize morbidity and improve the outcomes of CABG procedures, only highly specialized centers can presently afford the extensive financial investment that is required to use robotic systems in this setting. It is our hope that further reductions in perioperative morbidity and postoperative length of stay will

Ann Thorac Surg NOVICK ET AL 003;76:749 53 ANALYSIS OF THE LEARNING CURVE make telerobotic CABG surgery more cost effective. Furthermore, the application of telecommunication technologies to this field will facilitate information transfer between surgical centers and accelerate the diffusion of new robotic surgical techniques []. These new surgical methods will, in turn, need to undergo thorough and dispassionate analysis using standard statistical and CUSUM methods to ensure that short-term outcomes meet current expectations. Moreover, patients undergoing these procedures must be followed indefinitely to further define long-term graft quality and freedom from adverse clinical events. The authors acknowledge the secretarial expertise of Elizabeth Millar, as well as the technical assistance of Mike Carson in the training and operating room setup of the ZEUS system. They also acknowledge financial assistance for this project by grants from the Ivey Foundation, the Lawson Health Research Institute, Canada Foundation for Innovation, Ontario Innovation Trust, and Ontario Research and Development Challenge Fund. The support of Canadian Surgical Technologies and Advanced Robotics is also acknowledged. References 753. Plomondon ME, Cleveland JC Jr, Ludwig ST, et al. Off-pump coronary artery bypass is associated with improved riskadjusted outcomes. Ann Thorac Surg 00;7:4 9.. Cleveland JC Jr, Shroyer ALW, Chen AY, et al. Off-pump coronary artery bypass grafting descreases risk-adjusted mortality and morbidity. Ann Thorac Surg 00;7:8 9. 3. Mack M, Bachand D, Acuff T, et al. Improved outcomes in coronary artery bypass grafting with beating-heart techniques. J Thorac Cardiovasc Surg 00;4:598 607. 4. Mohr FW, Falk V, Diegeler A, et al. Computer-enhanced. robotic cardiac surgery: experience in 48 patients. J Thorac Cardiovasc Surg 00;:84 53. 5. Kiaii B, Boyd WD, Rayman R, et al. Robot-assisted computer enhanced closed chest coronary surgery: preliminary experience using a harmonic scalpel and ZEUS. Heart Surg Forum 000;3:94 7. 6. Boyd WD, Kiaii B, Novick RJ, et al. RAVECAB. Improving outcome in off-pump/minimal access surgery with robotic assistance and video enhancement. Can J Surg 00;44:45 50. 7. Boyd WD, Rayman R, Desai ND, et al. Closed-chest coronary artery bypass grafting on the beating heart using a computer-enhanced surgical robotic system. J Thorac Cardiovasc Surg 000;0:807 9. 8. Dogan S, Aybek T, Andressen E, et al. Totally endoscopic coronary artery bypass grafting on cardiopulmonary bypass with robotically enhanced telemanipulation: report of fortyfive cases. J Thorac Cardiovasc Surg 00;3:5 3. 9. Novick RJ, Stitt LW. The learning curve of an academic cardiac surgeon: use of the CUSUM method. J Card Surg 999;4:3 0. 0. Novick RJ, Fox SA, Stitt LW, et al. Cumulative sum failure analysis of a policy change from on pump to off pump coronary artery bypass grafting. Ann Thorac Surg 00;7: S06.. Novick RJ, Fox SA, Stitt LW, et al. Effect of off pump coronary artery bypass grafting on risk-adjusted and cumulative sum failure outcomes after coronary artery surgery. J Card Surg 00;7: 9.. Boyd WD, Kodera K, Stahl KD, Rayman R. Current status and future directions in computer-enhanced video- and robotic-assisted coronary bypass surgery. Semin Thorac Cardiovasc Surg 00;4:0 9. 3. Naylor CD, Rothwell DM, Tu JV, et al. Outcomes of coronary artery bypass surgery in Ontario. In: Naylor CD, Slaughter PM, eds. Cardiovascular health and services in Ontario: an ICES atlas. Toronto, Canada: Institute for Clinical Evaluative Sciences, 999:89 97. 4. Reichenspurner H, Damiano RJ, Mack M, et al. Use of the voice-controlled and computer-assisted surgical system ZEUS for endoscopic coronary artery bypass grafting. J Thorac Cardiovasc Surg 999;8: 6. 5. Loulmet D, Carpentier A, d Attellis N, et al. Endoscopic coronary artery bypass grafting with the aid of robotic assisted instruments. J Thorac Cardiovasc Surg 999;8:4 0. 6. Damiano RJ, Tabaie H, Mack MJ, et al. Initial prospective multicenter clinical trial of robotically-assisted coronary artery bypass grafting. Ann Thorac Surg 00;7:63 9. 7. Kappert U, Cichon R, Schneider J, et al. Closed-chest coronary artery surgery on the beating heart with the use of a robotic system. J Thorac Cardiovasc Surg 000;0:809. 8. Page ES. Continuous inspection schemes. Biometrika 954; 4:00 5. 9. Kenett R, Pollak M. On nonsequential detection of a shift in the probability of a rare event. J Am Stat Assoc 983;78:389 95. 0. Altman DG, Royston JP. The hidden effect of time. Stat Med 988;7:69 37.. McPherson K. Statistics: the problem of examining accumulating data more than once. N Engl J Med 974;90:50. Notice From the American Board of Thoracic Surgery Regarding Trainees and Candidates for Certification Who Are Called to Military Service Related to the War on Terrorism The Board appreciates the concern of those who have received emergency calls to military service. They may be assured that the Board will exercise the same sympathetic consideration as was given to candidates in recognition of their special contributions to their country during the Vietnam conflict and the Persian Gulf conflict with regard to applications, examinations, and interruption of training. If you have any questions about how this might affect you, please call the Board office at (847) 475-50. Peter C. Pairolero, MD Chairman The American Board of Thoracic Surgery 003 by The Society of Thoracic Surgeons Ann Thorac Surg 003;76:753 0003-4975/03/$30.00 Published by Elsevier Inc