Neoadjuvant Radiation is Causally Linked to Increased Operative Time and Perioperative Blood Transfusion in Pancreatic Ductal Adenocarcinoma

Abstract

Introduction: Blood transfusion in patients undergoing surgical resection for pancreatic ductal adenocarcinoma (PDAC) is associated with worse outcomes. As the treatment paradigm shifts toward neoadjuvant therapy, little is known about how this may influence perioperative blood transfusion.

Materials and methods: The American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database was utilized to identify PDAC patients undergoing pancreaticoduodenectomy. Patients were grouped based on initial treatment type and outcomes were assessed through univariate analysis and multivariate logistic regression. Causal graphical modeling was used to identify key predictors of transfusion (rCausalMGM R package). To assess the impact of neoadjuvant radiation on blood transfusion, causal inference estimators were applied.

Results: From 2014 to 2019, 13,405 patients underwent pancreaticoduodenectomy for PDAC and 11.6% received neoadjuvant radiation. Though patients who underwent neoadjuvant radiation were less likely to have advanced disease, they were more likely to receive a perioperative blood transfusion. On multivariate logistic regression, neoadjuvant radiation treatment was independently associated with perioperative blood transfusion though operative time was suggested to be an outcome mediator. Causal graphical modeling revealed that neoadjuvant radiation therapy leads to perioperative transfusion through operative time. Logistic regression models with Markov Blanket-selected variables suggested that neoadjuvant radiation therapy was independently associated with perioperative blood transfusion, whereas neoadjuvant chemotherapy alone was found to be protective.

Conclusions: Neoadjuvant radiation is independently associated with perioperative blood transfusion in patients with PDAC undergoing pancreaticoduodenectomy. This is in part due to increased operative times, which may reflect difficulty of resection. Future efforts to mitigate the need for perioperative transfusion are warranted to improve outcomes.

Keywords: Causal modeling; Machine-learning; Pancreatic cancer; Pancreaticoduodenectomy; Preoperative treatment.