Safety and learning curve analysis of robotic-assisted pancreaticoduodenectomy: experience of a single surgeon

Although prior studies have discussed learning curves (LC) of robotic-assisted pancreaticoduodenectomy (RPD), a recognized definition is lacking. This study analyzed the clinical outcomes of 85 consecutive RPD cases performed by a single surgeon to evaluate the safety and learning curve of RPD using the da Vinci Xi robotic system. There were 51 male and 34 female patients, with a median age of 64 (20–87) years. The average preoperative body weight and BMI were 64.15 ± 11.43 kg and 23.36 ± 3.33 kg/m2, respectively. The clinical outcomes of each patient were analyzed using the textbook outcome(TO), and the learning curve of the RPD was evaluated by calculating the TO rate of patients using the cumulative sum analysis method (CUSUM).The operation time (OT) was 288.92 ± 44.41 min, and the postoperative hospital stay was 10 (1–134) days. In total, 23.52% (20/85), 5.88% (5/85), 2.35% (2/85), and 5.9% (5/85) experienced grade IIIa, IIIb, IV, and V complications. A total of 46 patients achieved TO outcomes (TO group), while 39 did not (non-TO group). The smoking rate in the TO group was lower (P < 0.05) and the albumin level was higher (P < 0.05) than that in the non-TO group. The TO rate became positive after the 56th case, all patients were divided into a learning improvement group (56 cases) and a proficient group (29 cases). The total bilirubin level in the learning improvement group was lower (P < 0.05) and the bleeding volume was higher (P < 0.05).RPD is safe and effective for carefully selected patients. The learning curve was completed after 56 patients.