Data from 1312 patients (2010-2021), submitted to upfront resection for resectable PDAC and with at least 15 LNs retrieved/examined were used to build a forecast count-model to predict N0 and the eventual number of metastatic LNs. A negative-binomial regression was adopted through a derivation / validation approach (random selection: 70%/30%) and 5-fold cross validation (CV) in the derivation cohort. Within the derivation cohort (n=918), the count of metastatic LNs was strongly related to head location (p<0.001), radiological largest size (p<0.001), CA19.9 (p<0.001) and the presence of weight-loss (p<0.001). The number of harvested/examined LNs (p<0.001) was also introduced in the analysis. When the model was tested in the out-of-sample CV folds of the derivation cohort, the AUPRC in detecting N0 was 0.885 and the R2 in forecasting the number of LN+ was 0.975. When the count-model was tested in the validation cohort (n=394) the AUPRC in detecting N0 was 0.888 and the R2 in forecasting the number of LN+ was 0.943, confirming the high accuracy of the model.