National minimally invasive pancreaticoduodenectomy outcomes are improving with time: a propensity score matched NSQIP analysis

Abstract

Background: There is a steep learning curve associated with minimally invasive pancreaticoduodenectomy (MIPD), and some past studies have suggested MIPD could be associated with increased rates of pancreatic fistula and mortality. NSQIP is a methodically rigorous and robust method of complication gathering in surgery and is ideal to study MIPD outcomes over time. We sought to compare the outcomes of patients undergoing MIPD in previous years of pancreatectomy targeted NSQIP (2014–2015) to the most recent year (2016). We hypothesized that the national MIPD outcomes would significantly improve with time. Methods: All patients undergoing elective MIPD (robotic and laparoscopic, including open-assisted and unplanned open conversion) in the ACS NSQIP pancreatectomy specific database were included in the study. Patients were stratified into those who underwent MIPD during 2014 and 2015 (MIPD1415) and those who underwent MIPD in 2016 (MIPD16). Patient covariates and outcomes were compared using standard statistical methods. 1:1 propensity score matching was also performed and outcomes compared. Results: 766 patients (7.8%) of 9069 pancreaticoduodenectomy (PD) patients underwent MIPD between 2014–2016. 48.7% of MIPD were robotic-assisted. Compared to MIPD1415, MIPD16 patients were significantly more likely to have a drain left at the time of surgery (97.5% vs 94.5%, p = 0.04) and less likely to have a vascular resection (10.2% vs 17.3%, p = 0.01). There is also a difference regarding the distribution of gland texture between the two groups. MIPD16 patients had a significantly decreased rate of pancreatic fistula (14.9% vs 21.1%, p = 0.03), mortality (0.3% vs 2.4%, p = 0.02), and length of stay (8.2 vs 10.4 days, p = 0.02) compared to MIPD1415 patients. However, there was no difference in these outcomes between patients in the two different time periods undergoing open PD. Among 279 propensity score matched pairs (n = 558), MIPD16 patients had significantly decreased rates of pancreatic fistula (15.8% vs 23.0%, p = 0.03), clinically relevant pancreatic fistula (9.0% vs 16.1%, p = 0.01), postoperative length of stay (8.0 vs 10.5 days, p = 0.02), and mortality (0.4% vs 2.9%, p = 0.02) compared to matched MIPD1415 patients. In a subanalysis excluding patients with pancreas ducts <3mm (n = 564), MIPD16 patients had a significantly decreased 30-day mortality rate (0.5% vs 2.9%, p<0.05) and postoperative length of stay (8.2 vs 10.2 days, p = 0.04) along with a strong trend towards a decreased rate of clinically relevant pancreatic fistula (8.5% vs 13.0%, p = 0.05), shown in Figure 1. However, there was no difference in these outcomes between MIPD1415 and MIPD16 patients with <3mm pancreas ducts. Conclusion: National MIPD outcomes are improving with time. A major contributing factor for this finding is improved MIPD outcomes in patients with larger pancreas ducts.