ABSTRACT
BACKGROUND: Preoperative risk stratification for postoperative pancreatic fistula in patients undergoing distal pancreatectomy is needed. METHODS: Risk factors for postoperative pancreatic fistula in 220 consecutive patients undergoing distal pancreatectomy at 2 major institutions were recorded retrospectively. Gland density was measured on noncontrast computed tomography scans (n = 101), and histologic scoring of fat infiltration and fibrosis was performed by a pathologist (n = 120). RESULTS: Forty-two patients (21%) developed a clinically significant pancreatic fistula within 90 days of surgery. Fat infiltration was significantly associated with gland density (P = .0013), but density did not predict pancreatic fistula (P = .5). Recursive partitioning resulted in a decision tree that predicted fistula in this cohort with a misclassification rate less than 15% using gland fibrosis (histology), density (HU), margin thickness (cm), and pathologic diagnosis. CONCLUSIONS: This multicenter study shows that no single perioperative factor reliably predicts postoperative pancreatic fistula after distal pancreatectomy. A decision tree was constructed for risk stratification.
