Multicenter comparison of totally laparoscopic and totally robotic pancreaticoduodenectomy: Propensity score and learning curve-matching analyses.

BACKGROUND Few studies have compared laparoscopic pancreaticoduodenectomy(LPD) and robotic pancreaticoduodenectomy(RPD) despite emerging use of minimally invasive pancreaticoduodenectomy(MIPD). Present study therefore compared perioperative outcomes of LPD and RPD patients, and evaluated safety and feasibility of MIPD. METHODS This retrospective multicenter analysis evaluated MIPD patients through June 2020 performed by three experienced pancreatic surgeons at three different institutions. Perioperative outcomes were compared before and after propensity score-matching analyses, and learning curves based on operation time were used for additional matching analysis. RESULTS Of 362 patients, 282 underwent LPD and 80 underwent RPD. Open conversion rate was significantly higher in LPD(p=0.001). There were no significant differences in rates of major complications(Clavien-Dindo≥III) and clinically-relevant postoperative pancreatic fistula(CR-POPF). After matching, operation time(p=0.001) and hospital stay(p=0.027) were significantly shorter in RPD, but no differences in major complications and CR-POPF. Propensity score-matched comparison after learning curve attainment showed shorter operation time(p=0.037) and hospital stay(p=0.014) in RPD, and no differences in major complications and CR-POPF. CONCLUSION RPD had several advantages compared with LPD, including shorter operative time and hospital stay, and lower open conversion rate. Postoperative complications including CR-POPF showed comparable results in two groups. Both LPD and RPD seemed feasible and safe approaches in experienced hands.