BACKGROUND: Laparoscopic pancreaticoduodenectomy (LPD) may have potential minimal invasive advantages for selected patients in limited center. However, few studies analyzed the learning curve. This study aimed to analyze the learning curve of this procedure at a large volume set, and share our experience to surmount it. METHODS: All prospectively maintained data of the consecutive LPDs was reviewed retrospectively. The procedures were performed by single surgeon. Patients were divided into four groups according to staged approach with different focuses: Group A (the first 30 patients), Group B (the second 30 patients), Group C (the third 30 patients), and Group D (the fourth and last 30 patients). And the changes of outcomes during different learning periods were analyzed. RESULTS: Between September 2012 and July 2015, 120 patients underwent LPD. One hundred and eleven of them underwent totally LPD, and 9 patients underwent laparoscopic assisted pancreaticoduodenectomy (LAPD). The mean operative time (OT), mean blood loss and average length of hospital stay (LOS) was 359.8±57.6 min, 169.7±152.6 mL and 17.0±9.8 d respectively. A total of 42 (35%) patients developed morbidity with no mortality. The mean overall OT tended to decrease from 370.2±52.8 min in Group A to 342.0±73.1 min in Group D with the accumulating experience of the surgeon. Moreover, mean OT of pancreatojejunostomy and choledochojejunostomy also tended to decrease from 55.0±8.7, 39.8±11.7 min in Group A to 43.6±7.6, 27.7±11.8 min in Group D respectively. Meanwhile, the clinical outcomes tended to get better. Mean blood loss, morbidity and LOS decreased from 219.3±147.9 mL, 43.3%, 18.7±10.0 d in Group A to 140.1±73.6 mL, 23.3%, 14.4±6.2 d in Group C respectively except for Group D. CONCLUSIONS: Routine practice of the LPD procedure was feasible and safe. Gained experience can improve clinical outcomes in 30 to 60 operations by overcoming the learning curve.