Perioperative Outcomes and Learning Curve of Robot-Assisted McKeown Esophagectomy

Background This study aimed to evaluate the perioperative outcomes of patients undergoing robot-assisted McKeown esophagectomy (RAME) and the learning curves of surgeons performing RAME at a single center. Methods Perioperative outcomes of RAME and video-assisted McKeown esophagectomy (VAME) were compared after eliminating confounding factors by propensity score matching (PSM). The cumulative sum (CUSUM) method was used to evaluate the learning curves of RAME for a single surgical team. Results In general, a total of 198 patients with esophageal cancer (RAME: 45 patients, VAME: 153 patients) were included in this study, and 43 pairs of patients receiving RAME or VAME were matched using 1:1 PSM analysis. Those in the RAME group had more lymph nodes dissected in the total lymph nodes (median 29.0 vs . 26.0, P = 0.011) and the upper mediastinum (median 8.0 vs. 6.0, P < 0.001), especially the left recurrent laryngeal nerve (RLN) lymph node (median 4.0 vs. 2.0, P = 0.001). According to the trend of the CUSUM plot, the learning curve was divided into two stages at the 20th RAME procedure. After mastering the learning curve, RAME harvested a significantly higher number of upper mediastinal lymph nodes (median 9.0 vs. 6.0, P = 0.001), left RLN lymph nodes (median 5.0 vs. 3.5, P = 0.003), and right RLN lymph nodes (median 4.0 vs. 2.0, P = 0.002). Meanwhile, the incidence of postoperative pneumonia in the proficiency phase was significantly lower than that in the learning phase (4.0% vs. 25.0%, P = 0.04). Conclusions RAME improved left RLN lymph node dissection. Surgeons with extensive VAME experience need at least 20 cases to achieve early proficiency in RAME.