Objective: We aimed to assess trends in patient selection and outcomes to evaluate safety in current practice of pancreatoduodenectomy. Methods: NSQIP database from 2005 to 2014 was used to identify patients undergoing a pancreatoduodenectomy. Preoperative optimization measure was created using lack of >10% weight loss, albumin>3 and bilirubin<2. Annual trends in patient characteristics and outcomes were described in univariate and logistic or linear regression analyses for binary and continuous variables, respectively. Results: 23,295 pancreatoduodenectomies were identified. There was no significant trend of age (overall median 65 years, trend p=0.31), severity of systemic disease increased (ASA≥3 in 73% overall, trend OR=1.07, p<0.001), but so did the rate of preoperative optimization (52% overall, trend OR=1.03, p<0.001). Malignant indication percentage was increasing (75% overall, trend OR=1.04, p<0.001) as was rate of associated vascular resection (8% overall, trend OR=1.09, p<0.001), but not multivisceral resections (4% overall, trend p=0.10). The adjusted odds of minor and major complications increased over time (overall 31.2% and 27.7%, trend OR=1.32 and 1.11 respectively, p<0.001 for both). In contrast, unadjusted and adjusted length of stay was shortening (median 9 days, decrease by 1.60% and 1.44% respectively, p<0.001 for both) and mortality continued to decrease (2.5% overall, trend OR=0.95, p=0.002). Conclusion: There has been increasing severity of systemic disease as well as more advanced primary disease in patients undergoing pancreatoduodenectomy. Despite increasing complication rates, preoperative optimization and management of complications were likely improved, as suggested by shortening length of stay and decreasing mortality.