A nomogram to predict pathologic lymph node positivity in clinical stage I-II pancreatic adenocarcinoma.

Abstract

382 Background: Clinical nodal staging in PDAC is inaccurate. Most pts are cN0, but > 70% are pN+. We hypothesize that preoperative variables are associated with pN+ and could be used to create a predictive nomogram. Methods: The NCDB was reviewed from 2010-13 for pts with clinical stage I-II PDAC. Exclusions were neoadjuvant therapy, < 12 nodes examined, and missing data for clinical/pathologic stage, size, and number of nodes examined/positive. Logistic regression assessed factors associated with pN+ and an interaction was included for extrapancreatic extension and cN stage. A logistic regression based nomogram was constructed and 10-fold cross validation evaluated model discrimination. Results: Of 7,475 pts, 28% were cN+ but 74% were pN+ (P < 0.001). Associations of preoperative factors with pN+ are shown. Size was pathology based. We recommend multiplying imaging based size estimates by 1.33 for use in the nomogram based on studies showing that imaging underestimates size by 25%. Interestingly, extrapancreatic extension was protective for cN0 pts but associated with increased odds of pN+ for cN1 relative to cN0 ps. A nomogram was created to predict pN+ using these variables. The 10-fold cross validated AUC was 0.77. Conclusions: Our nomogram has good discrimination to preoperatively predict lymph node positivity for patients with resectable PDAC. It could potentially be useful in identifying biologically aggressive resectable PDAC pts at higher risk of pN+ in order to select pts for neoadjuvant therapy. [Table: see text]