Objectives To assess the utility of MDCT tumor–vascular interface criteria for predicting vascular invasion and resectability in borderline pancreatic cancer (BRPC) patients after neoadjuvant therapy (NAT). Methods This prospective study… Click to show full abstract
Objectives To assess the utility of MDCT tumor–vascular interface criteria for predicting vascular invasion and resectability in borderline pancreatic cancer (BRPC) patients after neoadjuvant therapy (NAT). Methods This prospective study included 90 patients with BRPC who finished NAT, showed no progression in preoperative CTs and underwent surgery. Two radiologists independently assessed preoperative vessel-tumor interface criteria. The area under the ROC curve (AUC) was used to evaluate the diagnostic performance for predicting vascular invasions and resectability using surgical and pathological results as the gold standard. Inter-reader agreement was assessed using the κ coefficient. Results Pathologic vascular invasion was confirmed in 47 (54.7%) veins and 14 (16.3%) arteries. R0 resection was achieved in (82.6%71/86) pancreatic resection. Using criteria of circumferential interface ≥ 180 degrees with contour deformity ≥ grade 3 and/or length of tumor contact > 2 cm to predict vascular invasion, the AUCs for the two readers were 0.85–0.88 for arterial invasion and 0.92–0.87 for venous invasion. Using criteria of circumferential interface ≤ 180° with contour deformity ≤ grade 2 and/or length of tumor contact < 2 cm to predict R0 resection, the AUCs was 0.85–0.86 for the two readers. The overall inter-reader agreement was good ( κ = 0.75–0.80). The κ values for venous invasion, arterial invasion and R0 resection were 0.76, 0.78, and 0.80. Conclusion Tumor–vessel criteria demonstrated good diagnostic performance and reproducibility in the prediction of vascular invasion after NAT in BRPC. These criteria could be helpful in the prediction of R0 resection in cases with only venous involvement.
               
Click one of the above tabs to view related content.