R. J. Torphy1, C. Friedman1, B. C. Chapman1, M. D. McCarter1, R. D. Schulick1, B. H. Edil1, A. Gleisner1 1University Of Colorado,Department Of Surgery,Aurora, CO, USA
Introduction:
Minimally invasive pancreaticoduodenectomy (MIPD) has been slow to gain acceptance given its complexity and concern for oncologic equivalency when compared with an open approach. The National Cancer Database (NCDB) began documenting surgical approach in 2010. Our objective was to compare oncologic and short-term outcomes of patients with cancer who underwent open vs MIPD (laparoscopic and robotic) from 2010 to 2013 using the NCDB.
Methods:
Adults who underwent pancreaticoduodenectomy from 2010-2013 for cancer were identified after exclusion of patients with metastatic disease, pathologic T0, in-situ disease, or an unknown operating facility. Laparoscopic and robotic approaches were defined as minimally invasive. Multivariable logistic regression that accounted for clustering of patients at facilities was performed to examine the relationship between patient and facility characteristics and the use of MIPD, oncologic outcomes (margin status and lymph node harvest) and short-term outcomes (days to discharge, unplanned 30-day readmission, and 30- and 90-day mortality). The multivariable analyses controlled for demographics, insurance, institutional classification, distance to treating institution, year, Charlson comorbidity score, pathologic tumor stage (pT), nodal stage (pN) and overall stage, grade, histologic diagnosis, and hospital volume of open and MIPD.
Results:
Of the 11,066 patients who underwent pancreaticoduodenectomy for cancer from 2010-2013, 85% (9,406) were performed open and 15% (1,660) were performed minimally invasively. The percentage of minimally invasive cases increased from 11.8% in 2010 to 15.9% in 2013 (P<0.001). Factors independently associated with MIPD included age >80 (OR 1.30, P=0.021), pT2 (OR 0.80, P=0.020), and histologic diagnosis of neuroendocrine tumor (NET) (OR 1.45, P<0.001). Patients undergoing MIPD had decreased odds of a prolonged hospitalization (≥10 days) (OR 0.82, P=0.008). There was no difference in short-term outcomes of unplanned 30-day readmission and 30- or 90-day mortality. Patients undergoing MIPD had decreased odds of positive margins (OR 0.79, P=0.004), and no difference in number of lymph nodes resected. Using our multivariable model, increased hospital volume of open and MIPD was a significant predictor of improved 30- and 90-day mortality, decreased length of stay, and greater number of lymph nodes harvested, with centers in the highest quartile for volume (>21.5 cases per year) performing best.
Conclusion:
MIPD has increased in prevalence in the United States from 2010 to 2013. Patients selected for MIPD were more likely to have a diagnosis of NET and have smaller tumors, demonstrating a selection bias between approaches. After controlling for these differences, short-term and oncologic outcomes are equivalent between open and MIPD. These results also demonstrate an association between improved outcomes with higher hospital volume of pancreaticoduodenectomies for cancer.