Anatomic lobectomy with systematic lymph node sampling or lymph node dissection is the standard treatment for patients with early-stage non-small cell lung cancer (NSCLC). An open approach by thoracotomy remains the preferred operable approach for most practicing thoracic surgeons in the United States. Since the early 1990s, multiple hybrid and totally endoscopic minimally invasive techniques for lobectomy for lung cancer resection have been described.1–3 For the past decade, the da Vinci Robotic System has been increasingly used as a minimally invasive technique for lobectomy.
The learning curve for adoption of the robotic platform for lobectomy for malignant pulmonary disease has yet to be clearly studied or defined. The learning curve for any minimally invasive operation is defined as the number of times a particular procedure must be performed to complete the procedure consistently with high accuracy and precision, within a reasonable operative duration, an equivalent or lower complication rate, and a low conversion to open rate.4,5 Furthermore, the learning curve for a minimally invasive procedure for a malignant pathology should provide an objective account of adequacy of resection based on oncologic principals pertaining to the specific procedure. With respect to lobectomy for NSCLC, an equivalent lymph node sampling or dissection is a prerequisite for a viable minimally invasive alternative to open thoracotomy (OT).
The purpose of this study is to analyze the clinical outcomes when transitioning from an OT to a robotic platform, without significant pre-existing video-assisted thoracoscopic surgery (VATS) experience, for lobectomies in the treatment of stage I or II NSCLC. This study attempts to characterize the learning curve(s) associated with this technical transition that is commonly faced by thoracic surgeons with extensive OT proficiency but limited advanced thoracoscopic or VATS expertise.
Study Design and Patient Cohort
This study presents a retrospective review of all patients who underwent a lobectomy for clinical stage I or stage II NSCLC at a single Veterans Affairs (VA) facility by a single surgeon from January 2007 to December 2014. Figure 1 displays the number of OT and robotic lobectomies performed each year by the surgeon. The surgeon joined the VA facility in 2007, and the robotic platform was adopted by the surgeon in 2011. The surgeon in this study was in clinical practice for nearly 10 years, after the completion of thoracic surgery residency, before embarking on learning the robotic platform in 2010. He had performed nearly 20 VATS lobectomies for a two and a half year period between 2007 and 2010. In mid 2010, our surgeon showed interest in robotics and took an advanced training course at the Intuitive Surgical Inc. headquarters in Sunnyvale, CA USA. This was followed by participation in two separate case observation courses offered at two facilities with significant expertise in the robotic lobectomy technique. He then began performing tier 1 thoracic robotic procedures, such as pulmonary wedge and simple mediastinal tumor resections, under an obligatory six case proctoring requirement at our institution. The surgeon's lobectomy case load nearly doubled in 2011, mainly as result of exclusive referral of patients from a sister VA facility, located 300 miles away, dissatisfied by their regional thoracic surgical coverage and seeking minimally invasive thoracic surgical support. From the adoption of the robotic technique onward, patients undergoing lobectomy for stage I or II NSCLC were consecutively subjected to either robotic or OT lobectomy, with initial robotic cohort benefiting from a selection bias for smaller, more peripheral pulmonary nodules, and more favorable intrapleural anatomy (i.e., lack of fused pulmonary fissures and/or pleural symphysis). After approximately first 40 robotic lobectomies, the selection bias dissipated as greater operational comfort was achieved and use of robotics was preferred and based solely on the availability of the robotic platform. Since the adoption of the robotic platform, 85% of resections for stage I or II NSCLC at our institution were performed using a robotic approach.
A total of 157 patients with clinical stage I and II NSCLC underwent either an OT (n = 57) or robotic (n = 100) lobectomy (22 VATS lobectomies in this period were excluded from this analysis). Robotic lobectomies converted to OTs (n = 13) were included in the robotic cohort. The robotic cohort was divided into five consecutive groups, each consisting of 20 patients. The intraoperative and postoperative outcomes were tabulated. We used six metrics to evaluate learning curve when navigating through the robotic pentiles: operative time, conversion to open, estimated blood loss, hospitalization duration, overall morbidity, and pathologic nodal upstaging. The first five metrics have been used in similar learning curve studies.5–7 Pathologic nodal upstaging was used as a surrogate for completeness of resection in lung cancer surgery and is a unique metric for our learning curve model.
Clinical staging was achieved through positron emission tomography and computed tomography scans. If the images showed positron emission tomography avid hilar or mediastinal lymph nodes, or enlarged lymph nodes of 1 centimeter or greater, cervical mediastinoscopy was performed to confirm clinical N0 status, just preceding the surgical resection under same anesthetic sitting. Endobronchial ultrasound was sparingly used.
All robotic lobectomies in this study were performed on an Intuitive Surgical da Vinci SI Surgical System (Intuitive Surgical, Sunnyvale, CA USA) using the four-arm completely portal robotic pulmonary resection (CPRL-4) popularized by Cerfolio et al.8 A procedure was categorized as a robotic case if the robotic arms were docked and robotic dissection at the surgeon console was commenced. Open thoracotomies were all performed via standard posterolateral thoracotomy approach, dividing latissimus dorsi muscle and preserving serratus anterior muscle.
Data Collection and Definitions
The specific data collected included patient demographics, age, sex, smoking status, pulmonary function data, tumor characteristics (location, histology, and clinical and pathological stage), operative times, estimated intraoperative blood loss, duration of chest tube drainage, prolonged intubation (>24 hours), postoperative atrial fibrillation, postoperative pneumonia, postoperative length of stay, perioperative morbidity and mortality (at 30 and 90 days), hospital readmissions within 30 days, reoperations within 30 days, number of lymph node stations (N1 and N2) sampled, and pathologic nodal upstaging.
Definitions for the learning curve metrics were the following: (1) operative time was recorded by nursing staff and defined as interval between initial skin incision and patient's repositioning to supine. (2) Conversion was defined as any emergent or elective conversion to OT, anytime after initial docking of the robotic arms, and initiation of robotic-assisted dissection. (3) Estimated blood loss in the OT cohort was recorded by the anesthesia team based on amount of blood collected in the suction canister at the completion of the operative procedure. In robotic lobectomy cases, the total number of saturated large rolled Kittner Sponges, each saturated sponge containing approximately 5 mL of whole blood, was added to estimate the blood loss. In addition, if a suction-irrigator device was used to evacuate intrapleural blood, the amount recovered was added to the total to estimate the final blood loss. (4) Hospitalization duration was defined as the number of days to discharge after surgery. (5) Overall morbidity was a composite variable of any one of the minor or major complications as defined by the Thoracic Morbidity and Mortality Classification System.9 The variable was fulfilled if 1 or more of those complications occurred. (6) Pathologic nodal upstaging was defined as discovering unsuspected N1 or N2 disease in clinically node negative (i.e., cN0) patients. Each lymph node sampled or resected by surgeon during the lobectomy procedure was accordingly marked by the nursing staff based on the International Association for the Study of Lung Cancer lymph node mapping guidelines.
Categorical variables were compared using χ2 tests, and continuous variables were compared using t tests and analysis of variance testing. Learning curve was assessed by plotting operative time, estimated intraoperative blood loss, conversion to OT, length of stay, overall morbidity, and pathologic nodal upstaging for consecutive robotic cases. After plotting these data points, the learning curves were then analyzed using linear and logistic regression. A P value of less than 0.05 was considered significant.
In total, 157 patients met the inclusion criteria for this study, consisting of 100 patients in the robotic group and 57 patients in the OT group. The clinical characteristics of the patients in this study are presented in Table 1. There were no significant differences in age, sex, smoking status, pulmonary function, previous lung resection, primary tumor location, neoadjuvant therapy, or clinical stage between the two cohorts. The robotic cohort had more patient's with clinical stage IA NSCLC than the OT cohort (73% vs. 56%, P = 0.03).
The perioperative results of the entire study cohort are presented in Table 2. The robotic group had significantly less median estimated blood loss (75 vs. 150 mL, P < 0.001), lower incidence of prolonged intubation (0% vs. 7%, P < 0.01), shorter chest tube duration (3 vs. 6 days, P < 0.001), shorter length of stay (6 vs. 10 days, P < 0.001), and lower overall morbidity (35% vs. 56%, P < 0.01) as compared with the OT group. There was a trend toward significance for the robotic group having a lower rate of readmission within 30 days and lower mortality within 90 days as compared with the OT group. Between the two groups, there were no significant differences in intraoperative and/or postoperative blood transfusion, number of lymph node stations sampled, pneumonia or atrial fibrillation rates, reoperation within 30 days, or mortality within 30 days. There was also no significant difference in nodal upstaging rates between the robotic and OT groups (17% vs. 14%, respectively, P = 0.63).
There were 13 (13%) conversions to OT in the robotic group, and 6 (46%) of these 13 conversions occurred during the first 20 cases. The causes associated with the conversions included bleeding from a pulmonary artery branch (2), fused major/minor fissure(s) (3), and large tumor size (1). The most common reason for conversion in the latter 80 cases was the presence of complete pleural symphysis.
Learning Curve for Robotic Lobectomy
Clinical Metric #1: Operative Time
Figure 2 displays the operative times for consecutive robotic lobectomies with a negative regression (P = 0.002). The mean operative times for the first, second, third, fourth, and fifth groups of 20 consecutive robotic cases were 252.2, 200.5, 244.5, 190.5, and 178.0 minutes, respectively. The first group of 20 consecutive robotic cases had a significantly longer mean operative time as compared with the second group of 20 consecutive cases (P = 0.038). The third group of 20 robotic cases had a significantly longer mean operative time as compared with the second group (P = 0.046), as well as the fourth (P = 0.027) and fifth groups (P = 0.004). Detailed review of operative reports of the third group of robotic lobectomies revealed relaxation of the selection criteria ad bias for robotic lobectomy to include patients with higher clinical stage and more challenging anatomy such as fused fissures and more hostile pleural adhesions, possibly resulting in longer operative times. After 60 robotic lobectomies, the operative times approached a steady state phase, which was equivalent to the OT group operative time.
Clinical Metric #2: Conversion Rate to Open
Figure 3 displays the conversion from robotic lobectomy to OT for consecutive robotic lobectomies with a negative regression (P = 0.01). In this study, there were a total of 13 (13%) conversions. The number of conversions associated with the first, second, third, fourth, and fifth groups of 20 consecutive robotic cases were 6, 3, 1, 2, and 1, respectively. The first 40 consecutive robotic cases had a significantly larger rate of open conversions as compared with the next 60 cases (22.5% vs. 6.7%, respectively, P = 0.02). In our series, the learning curve to reduce the rate of conversion to low single digits was approximately 40 patients. Bleeding as a reason for conversion to OT vanished after the first 20 cases.
Clinical Metric #3: Estimated Blood Loss
Estimated blood loss for consecutive robotic lobectomies demonstrated no significant regression (P = 0.86). The median estimated blood loss for the first, second, third, fourth, and fifth groups of 20 consecutive robotic cases were 65, 75, 50, 100, and 75 mL, respectively. Despite two patients needing conversion to OT for pulmonary artery branch bleeding in the first group, there were no significant differences in median estimated blood loss between consecutive robotic groups (P = 0.56). The first group of 20 robotic lobectomies had a significantly less estimated blood loss than the OT group (65 vs. 150 mL, respectively, P = 0.04).
Clinical Metric #4: Duration of Hospitalization
Length of hospital stay for consecutive robotic lobectomies demonstrated no significant regression (P = 0.52). The median length of stay for the first, second, third, fourth, and fifth groups of 20 consecutive robotic cases were 6, 7, 5, 5, and 6.5 days, respectively. There were no significant differences in length of stay for the consecutive groups (P = 0.55). The first group of 20 consecutive robotic cases had a significantly shorter median length of hospital stay than the OT group (6 vs. 10 days, respectively, P = 0.05).
Clinical Metric #5: Overall Morbidity
Overall morbidity for consecutive robotic lobectomies demonstrated no significant regression (P = 0.96). The rates of perioperative morbidity for the first, second, third, fourth, and fifth groups of 20 consecutive robotic cases were 25%, 35%, 35%, 30%, and 30%, respectively. There were no significant differences in morbidity for the consecutive robotic groups (P = 0.96). Despite including patients with open conversions in the robotic arm, the first group of 20 consecutive robotic cases had a significantly less overall morbidity rate than the OT group (25% vs. 51%, respectively, P = 0.045).
Clinical Metric #6: Pathologic Nodal Upstaging
Pathologic nodal upstaging for consecutive robotic lobectomies demonstrated no significant regression (P = 0.31). The rates of upstaging for the first, second, third, fourth, and fifth groups of 20 consecutive robotic cases were 10%, 20%, 20%, 20%, and 15%, respectively. There were no significant differences in pathologic nodal upstaging for the consecutive robotic groups (P = 0.89). The rate of pathologic upstaging for the first 20 patients was statistically equivalent to the OT group (10% vs. 14%, P = 0.64).
Despite the introduction of minimally invasive thoracoscopic techniques in 1990s, OT has remained the most common surgical approach for resection of primary lung cancer, especially in the hands of nonacademic thoracic surgeons. Robotic lobectomy entered the clinical arena in 2004 and has been increasingly embraced by practicing thoracic surgeons for the past decade, currently comprising approximately 15% of all anatomic pulmonary lobectomies performed for malignant disease. Some of the advantages of the robotic platform advocated by the surgeon champions are better ergonomics, superior optics and surgical exposure, greater ease and safety of hilar dissection, and more complete hilar and mediastinal lymph node dissection. However, the learning curve for adoption of a robotic platform for lobectomy by an open thoracic surgeon, previously novice with VATS lobectomy techniques, is inadequately studied and ill-defined. We undertook this single institution-single surgeon study to better characterize this learning curve.
The Major Findings of This Current Study Are Four Folds
1. Completely portal robotic pulmonary lobectomy, even in the initial learning phase, significantly improved clinical outcomes in patients undergoing curative resection of early-stage NSCLC as compared with a contemporaneous group of patients undergoing standard OT by the same surgeon previously novice in advanced VATS techniques. Patients who underwent lobectomy with the robotic approach had less estimated intraoperative blood loss, a lower rate of prolonged intubation, shorter chest tube duration and length of stay, and lower overall morbidity. There was also a trend toward decreased 30-day readmission and 90-day mortality. This indicates that a surgeon with novice level minimally invasive technique skills can adopt a robotic technique and still have significantly improved short-term clinical outcomes for the treatment of early-stage NSCLC. Similarly, in their initial experience with robotic lobectomy, Cerfolio et al.8 demonstrated that the robotic approach had significantly lower estimated intraoperative blood loss, length of stay, chest tube duration, and morbidity as compared with the OT technique.
Our length of hospital stay for both the robotic and OT cohorts is longer than reported in the literature. In 2011, we began accepting patient referrals from one of our sister VA facilities, located 300 miles away in a neighboring state, with no thoracic surgical expertise or support. These patients were treated with conservative discharge indicators to respect their safety in traveling back home and potentially curtail unforeseen complications in their poorly supported index facility. The latter is one of the reasons for our longer length of hospital stay, which was on average 2 to 3 days longer than the length of chest tube duration.
2. Even in learning phase of adoption of a robotic platform, oncologic principals in performing an anatomic lobectomy for lung cancer were not breached. The extent of lymph node dissection and rate of pathologic nodal upstaging in the robotic group were equivalent to the OT group throughout the transition. Nodal upstaging has recently been used as a surrogate measure for the completeness of resection in lung cancer surgery.10–14 Our study showed that there was no difference in the prevalence of nodal upstaging between the robotic and OT groups, signaling that robotic surgery, even in the earliest learning stages, may replicate the completeness of open surgery. Wilson et al.10 demonstrated that the rate of nodal upstaging for robotic lobectomy was similar to thoracotomy but superior to VATS when analyzed by clinical T stage. Similarly, Martin et al.13 found no significant difference in nodal upstaging rates between robotic surgery and thoracotomy but lower upstaging with VATS.
3. Operative time for robotic lobectomy follows a bimodal learning curve. In this study, the operative times of the first group of 20 consecutive robotic cases were statistically longer than OT. The operative times then decreased significantly in the next group of 20 patients, presumably the result of obtaining greater operational comfort and efficiency. However, in the third group of 20 patients, the operative times increased as compared with the second group, without an increase in the rate of open conversions. After the first 40 patients, more challenging patients with higher tumor stage and more hostile anatomy, such as fused pulmonary fissures and extensive intrapleural adhesions were included in the robotic cohort. This resulted in a longer mean operative time for the third group of 20 patients. The operative times for the fourth and fifth groups of 20 patients decreased once again and leveled out, indicating that a steady phase of the learning curve had been reached. In our experience, in a uniform cohort of Veterans, after approximately 60 robotic lobectomies for lung cancer, the procedural efficiency was markedly improved to the point that the obligatory time required for port placement and docking was canceled out by the shorter time required for closure. Our operative time learning curve is longer than that reported in the literature. This difference might reflect the differential patient cohort characteristics, as well as the differential level of technical expertise involved with minimally invasive techniques. Nevertheless, our results are representative of a learning curve for a thoracic surgeon that was previously novice in advanced VATS techniques.
Multiple studies reference a significant decrease in operative time as the most prominent factor of the learning curve for robotic lobectomies. Gharagozloo et al.15 and Veronesi et al.16 found that the first 20 robotic lobectomy cases had significantly longer operative times. However, Gharagozloo et al.15 and Veronesi et al.16 only compared the first 18 and 20 consecutive robotic cases against the next 73 and 80 cases, respectively. By only comparing two groups, they would not have been able to extrapolate a bimodal learning curve even if one was actually present. When Lee et al.17 studied their first 35 robotic lobectomy cases, they showed an operative time learning curve of 15 to 17 cases. This may not have been a large enough sample size to show a bimodal learning curve. Meyer et al.6 observed a learning curve of 15 cases based on operative time; they did not break their consecutive cases into groups but analyzed when the overall regression trendline of their operative times reached a plateau.
4. Our study suggests that approximately 40 robotic lobectomies are required to reach an acceptable single digit rate for open conversion. Furthermore, conversion based on bleeding complications ceased after the first group of 20 cases, demonstrating that development of greater comfort in endoscopic control of bleeding with robotic techniques can be achieved early on in the learning phase of adoption of a robotic platform. Shaw et al.18 showed conversion rates for minimally invasive lobectomy are between 3% and 20% during the learning curve.
Limitations and Strengths
The limitations of our study include biases inherent with any retrospective review. There is a plausible bias when comparing end points such as estimated blood loss, length of stay, and chest tube duration between open and minimally invasive techniques. This bias could be inherent in the knowledge that the minimally invasive technique should have lower and shorter values for these variables as compared with open techniques. Although the clinical characteristics between the open and robotic cohorts were not significantly different, the initial selection bias in robotic cohort containing more stage IA and less hostile intraoperative anatomy/pathology is recognized as an acceptable bias in developing a successful learning curve in adoption of a new minimally invasive technology.
The strengths of our study include its single institution-single surgeon analysis, as well as the uniformity of the patient population, pathologic disease, and procedural characteristics. These factors allowed for consistency in the data throughout the study. Future directions of study should include randomized controlled trials to compare the clinical outcomes of robotic, VATS, and open lobectomy and prospective studies to characterize the learning curves for minimally invasive lobectomy. In addition, a meta-analysis to compile the learning curves of other surgeons would be prudent.
Adoption of a robotic platform for lobectomy for NSCLC is safe and feasible without significant previous VATS experience. In comparison with OT, robotics, even during the early learning phase, resulted in a significant reduction in overall perioperative morbidity, estimated blood loss, chest tube duration, and length of stay for clinical stage I and II NSCLC patients. Robotics, even during the early learning phase, permitted equivalent nodal sampling when performing lobectomies for clinical stage I and II patients. The learning curve for the conversion rate and operative time to reach an acceptable, steady level phase required approximately 40 and 60 robotic lobectomies, respectively.
1. Yang HX, Woo KM, Sima CS, et al. Long-term survival based on the surgical approach to lobectomy for clinical stage I non-small cell lung cancer
: comparison of robotic, video-assisted thoracic surgery, and thoracotomy lobectomy. Ann Surg
2. Kent M, Wang T, Whyte R, et al. Open, video-assisted thoracic surgery, and robotic lobectomy
: review of a national database. Ann Thorac Surg
3. Oh DS, Cho I, Karamian B, et al. Early adoption of robotic pulmonary lobectomy: feasibility and initial outcomes. Am Surg
4. Zhang L, Gao S. Robot-assisted thoracic surgery versus open thoracic surgery for lung cancer: a system review and meta-analysis. Int J Clin Exp Med
5. Hopper AN, Jamison MH, Lewis WG. Learning curves in surgical practice. Postgrad Med J
6. Meyer M, Gharagozloo F, Tempesta B, et al. The learning curve
of robotic lobectomy
. Int J Med Robotics Comput Assist Surg
7. Toker A, Ozyurtkan MO, Kaba E, et al. Robotic anatomic lung resections: the initial experience and description of learning in 102 cases. Surg Endosc
8. Cerfolio RJ, Bryant AS, Skylizard L, et al. Initial consecutive experience of completely portal robotic pulmonary resection with 4 arms. J Thorac Cardiovasc Surg
9. Seely A, Ivanovic J, Threader J, et al. Systematic classification of morbidity and mortality after thoracic surgery. Ann Thorac Surg
10. Wilson JL, Louie BE, Cerfolio RJ, et al. The prevalence of nodal upstaging during robotic lung resection in early stage non-small cell lung cancer
. Ann Thorac Surg
11. Lee BE, Shapiro M, Rutledge JR, et al. Nodal upstaging in robotic and video assisted thoracic surgery lobectomy for clinical N0 lung cancer. Ann Thorac Surg
12. Licht PB, Jørgensen OD, Ladegaard L, et al. A national study of nodal upstaging after thoracoscopic versus open lobectomy for clinical stage I lung cancer. Ann Thorac Surg
13. Martin JT, Durbin EB, Chen L, et al. Nodal upstaging during lung cancer resection is associated with surgical approach. Ann Thorac Surg
14. Medbery RL, Gillespie TW, Liu Y, et al. Nodal upstaging is more common with thoracotomy than with VATS during lobectomy for early-stage lung cancer: an analysis from the national cancer data base. J Thorac Oncol
15. Gharagozloo F, Margolis M, Tempesta B, et al. Robot-assisted lobectomy for early-stage lung cancer: report of 100 consecutive cases. Ann Thorac Surg
16. Veronesi G, Agoglia BG, Melfi F, et al. Experience with robotic lobectomy
for lung cancer. Innovations
17. Lee BE, Korst RJ, Kletsman E, et al. Transitioning from video-assisted thoracic surgical lobectomy to robotics for lung cancer: are there outcomes advantages? J Thorac Cardiovasc Surg
18. Shaw JP, Dembitzer FR, Wisnivesky JP, et al. Video-assisted thoracoscopic lobectomy: state of the art and future directions. Ann Thorac Surg
Gallagher and the group from the VA Long Beach Healthcare system have provided an insightful approach on adopting new technology, the robotic-assisted lobectomy. At times, it can be a struggle for surgeons already in practice to become proficient in new technologies that were not available during their residency or fellowship training. This article has laid out a step-by-step approach to adopting robotic lobectomy, while maintaining oncologic principles and clinical standards. It is a valuable contribution to the literature.