BACKGROUND: Pancreatic surgery is technically complex. We hypothesized that a learning curve existed for pancreaticoduodenectomy even for surgeons who had completed their training.

METHODS: During 1990 to 2004, we studied 650 consecutive patients who underwent pancreaticoduodenectomy by 3 surgeons who began their attending careers at 1 center. Operative time, estimated blood loss (EBL), length of hospital stay (LOS), and the status of resection margins (for pancreatic adenocarcinoma) were analyzed. The chi2, independent t test and Mann-Whitney U test were used to evaluate differences in categorical, normally distributed continuous, and non-normally distributed continuous variables, respectively. Using serial groups of 30 cases, median operative time, EBL, and LOS were calculated and the trend over time modeled using third-order polynomial equations. Trends in retroperitoneal margin positivity (R0/R1) were assessed.

RESULTS: From the first 60 cases per surgeon to the second 60 cases per surgeon, the median EBL dropped (1100 vs 725 mL, P < .001), operative time decreased (589 vs 513 minutes, P < .001), and LOS decreased (15 vs 13 days, P = .004). The proportion of microscopically positive or suspicious margins also decreased from the surgeons' first 60 cases each to the second 60 cases (30% vs 8%, P < .001). Extended analysis of a single surgeon's cases suggested that additional experience provided further incremental improvement (P < .001).

CONCLUSIONS: Pancreaticoduodenectomy has an inherent learning curve. After 60 cases, surgeons achieved significantly decreased EBL, operative time, and LOS, and carried out more margin-negative resections. Improvement in measured outcomes continues during the operative career.