Multidimensional Analyses of the Learning Curve of Robotic Low Anterior Resection for Rectal Cancer: 3-phase Learning Process Comparison

Eun Jung Park, MD, Seung Hyuk Baik, MD, PhD, Byung Soh Min, MD, PhD, Kang Young Lee, MD, PhD, Nam Kyu Kim, MD, PhD. Section of Colon and Rectal Surgery, Department of Surgery, Yonsei University College of Medicine, Seoul, South Korea.

INTRODUCTION We aimed to investigate the learning curve of robotic rectal cancer surgery by multidimensional statistical methods, and to compare the learning curve phases with respect to perioperative clinical and pathologic outcomes.

METHODS AND PROCEDURES From April 2006 to August 2011, a total of 130 consecutive patients who were diagnosed rectal cancer underwent a robotic low anterior resection (LAR) by a single surgeon at Severance Hospital, Yonsei University College of Medicine, Seoul, South Korea. Robotic LAR was performed on all the patients using the hybrid technique with the da Vinci® surgical system (Intuitive Surgical, Sunnyvale, CA, USA). The moving average method and the cumulative sum (CUSUM) were used to analyze the learning curve. The risk-adjusted CUSUM (RA-CUSUM) analysis by using logistic regression was used for evaluation of the points which showed completion of the surgical procedure in terms of R1 resection, conversion, postoperative complications, harvested lymph nodes less than 12, and local recurrence. Perioperative clinical outcomes and pathologic results were compared among the learning curve phases. Postoperative complications were graded by the Dindo classification. A p-value of less than 0.05 was considered statistically significant.  

RESULTS According to CUSUM, there were two critical points of the surgeon console time, which determined the learning curve phases at the 44th and the 78th cases. Based on these results, the learning curve was divided into three phases: phase 1 [the initial learning period (1st–44th case), n=44], phase 2 [the competent period (45th–78th case), n=34], and phase 3 [the challenging period (79th–130th case), n=52]. From the results of RA-CUSUM, the possibility of surgical failure was minimized at the 75th case. The total operation time including the surgeon console time and docking time tended to decrease after phase 1 (phase 1=229.8±48.5 min, phase 2=189.4±52.3 min, phase 3=181.6±54.0 min, p<0.001). Number of harvested lymph nodes was 16.7±7.9, 15.6±10.0, 15.9±9.6, respectively (p=0.851). Circumferential resection margin involvement rate was 9.1%, 5.9%, 5.8% (p=0.829), and local recurrence rate was 0%, 0%, 5.8% (p=0.116), respectively. Grade III and IV postoperative complication rate was 6.8% (3 cases), 0%, 11.5% (6 cases), respectively(p=0.191).   

CONCLUSION According to CUSUM, the primary technical competence of the robotic procedure for rectal cancer was achieved at phase 1 of the 44th case. However, the learning curve consisted of three phases. Moreover, technical completion of the robotic procedure to assure feasible perioperative clinical and pathological outcomes was achieved at phase 2 of the 75th case by the RA-CUSUM method.