The worldwide prevalence of pancreatic cancer has been rising in recent decades, and its prognosis has not improved much. The imbalance of substance and energy metabolism in tumour cells is among the primary causes of tumour formation and occurrence, which is often controlled by the neuroendocrine system. We applied Cox and LASSO regression analysis to develop a neuroendocrine regulation- and metabolism-related prognostic risk score model with three genes (GSK3B, IL18 and VEGFA) for pancreatic cancer. TCGA dataset served as the training and internal validation sets, and GSE28735, GSE62452 and GSE57495 were designated as external validation sets. Patients classified as the low-risk population (category, group) exhibited considerably improved survival duration in contrast with those classified as the high-risk population, as determined by the Kaplan-Meier curve. Then, we combined all the samples, and divided them into three clusters using unsupervised clustering analysis. Unsupervised clustering, t-distributed stochastic neighbor embedding (t-SNE), and principal component analysis (PCA) were further utilized to demonstrate the reliability of the prognostic model. Moreover, the risk score was shown to independently function as a predictor of pancreatic cancer in both univariate and multivariate Cox regression analyses. The results of gene set enrichment analysis (GSEA) illustrated that the low-risk population was predominantly enriched in immune-associated pathways. "ESTIMATE" algorithm, single-sample GSEA (ssGSEA) and the Tumor Immune Estimation Resource (TIMER) database showed immune infiltration ratings were enhanced in the low-risk category in contrast with the high-risk group. Tumour immune dysfunction and exclusion (TIDE) database predicted that immunotherapy for pancreatic cancer may be more successful in the high-risk than in the low-risk population. Mutation analysis illustrated a positive link between the tumour mutation burden and risk score. Drug sensitivity analysis identified 44 sensitive drugs in the high- and low-risk population. GSK3B expression was negatively correlated with Oxaliplatin, and IL18 expression was negatively correlated with Paclitaxel. Lastly, we analyzed and verified gene expression at RNA and protein levels based on GENPIA platform, HPA database and quantitative real-time PCR. In short, we developed a neuroendocrine regulation- and metabolism-associated prognostic model for pancreatic cancer that takes into account the immunological microenvironment and drug sensitivity.