RA03.07: LEARNING CURVE FOR ROBOTIC AND VIDEO-ASSISTED THORACOSCOPIC ESOPHAGECTOMY: A MULTIDIMENSIONAL ANALYSIS

Abstract

Abstract Background Lymph node dissection (LND) along the left recurrent laryngeal nerve (RLN) is a technically challenging part of esophageal cancer surgery, especially after chemoradiotherapy (CRT). Robotic surgery holds promise to increase its safety and feasibility. Methods Patients who underwent minimally invasive esophagectomy and RLN dissection following CRT were dichotomized according to the use of robotic surgery (robotic esophagectomy [RE] versus video-assisted thoracoscopic esophagectomy [VATE]). Comparisons were made in terms of 1) number of dissected nodes, 2) rates of RLN palsy, 3) rates of perioperative complications, and 4) learning curves. Analysis of learning curves was performed with the cumulative sum (CUSUM) method (target failure rate for left RLN palsy: 15%). Results The RE and VATE groups consisted of 39 and 67 patients, respectively. The intraoperative identification of the left RLN was more common in the RE group (97.4%) than in the VATE group (68.7%, P < 0.001). Postoperative left RLN palsy was significantly more frequent in the VATE group (26.9%) than in the RE group (10.3%, P = 0.042), with a higher rate of pneumonia in the former (16.4% versus 2.6%, P = 0.03). CUSUM analysis revealed a longer learning curve when left RLN LND was performed through VATE. Left RLN palsy rates did not decrease below the target rate with the use of VATE, whereas RE allowed achieving the predefined target rate after 12 operations. Conclusion In the current study, we compared for the first time the learning curves of traditional and robotic surgery for RLN LND (which can be considered as one of the most technically challenging parts of esophageal cancer surgery). Our goal was to determine the minimum number of treated cases required to achieve an acceptable technical competency. Our data indicate that RE significantly facilitated a complex procedure like left RLN LND, resulting in a shorter learning curve compared with VATE. Notably, the left RLN was more easily identifiable with the use of RE, which in turn resulted in lower post-procedural left nerve palsy rates. Although MA analysis revealed that robotic surgery was initially more time-consuming than VATE, RE procedural times improved rapidly and became shorter than those of VATE after treatment of 26 cases. Disclosure All authors have declared no conflicts of interest.