Predictors of post-operative mortality following pancreatectomy: A contemporary nationwide analysis.

Abstract

e15706 Background: To date, the only curative option for pancreatic cancer is surgical resection, which only 15-20% of patients are eligible for at diagnosis. The conventional operation is a pancreaticoduodenectomy which carries a mortality risk of 3-7%. We used the National Cancer Database (NCDB) to examine surgical outcomes following pancreatectomy. Methods: We queried the NCDB from 2004-2015 for patients with stage I-III adenocarcinoma of the pancreas treated surgically. Outcomes reported were 30 and 90 day mortality, readmission within 30 days, and prolonged hospital stay. Prolonged hospital stay was defined as 17 days and 8 cases per year was used to define high volume centers. Multivariable logistic regression was used to identify predictors of all outcomes. Propensity matching was conducted to account for indication bias. Results: We identified 24,798 patients meeting eligibility criteria. The median patient age was 66 (40-90). The majority of patients were stage T3 (47%) N0 (65%) with pancreatic head lesions (83%). Fifty-seven percent of patients were treated with a Whipple procedure. Most patients (83%) were treated with surgery up front. Overall unadjusted risk of 30 day and 90 mortality ranged from 1.32-2.54% and 4.1-7.12%, respectively, depending on extent of surgery. On multivariable analysis predictors of 30 day mortality included preoperative therapy, increased age, higher comorbidity score, lower income, type of surgery, and treatment at a low volume center. Ninety day mortality was associated with preoperative therapy, increased age, increased comorbidity score, higher T stage, more extensive surgery, and treatment at a low volume center. Multivariable analysis revealed increased comorbidity score, treatment at an academic facility, and non-partial pancreatectomy as predictive of readmission. Prolonged hospital stay was associated with increasing age, decreased education, lower income, non-Caucasian race, more extensive surgery, and more remote year of treatment. Propensity matched Kaplan-Meier analysis showed improved survival with treatment at high volume centers, median survival of 23 months compared to 21 months (p = 0.0026). Conclusions: In the largest population based series to date, the 30 day mortality rates remain acceptably low with an increase seen at 90 days. The predictors of post-operative mortality seen here (age, comorbidity score, extent of surgery, and center volume) highlight the need for appropriate patient selection and physician experience to help achieve the best possible perioperative outcomes.