The intrinsic nature of tumour behaviour (stable vs progressive) and the presence of liver metastases are key factors in determining the outcome of patients with a pancreatic endocrine tumour (PET). Previous expression profile analyses of PETs were limited to non-homogeneous groups or to primary lesions only. The aim of this study was to investigate the gene expression profiles of a more uniform series of sporadic, non-functioning (NF) PETs with progressive disease and, for the first time, their liver metastases, on the Affymetrix human genome U133A and B GeneChip set. Thirteen NF PET samples (eight primaries and five liver metastases) from ten patients with progressive, metastatic disease, three cell lines (BON, QGP and CM) and four purified islet samples were analysed. The same samples were employed for confirmation of candidate gene expression by means of quantitative RT-PCR, while a further 37 PET and 15 carcinoid samples were analysed by immunohistochemistry. Analysis of genes differentially expressed between islets and primaries and metastases revealed 667 up- and 223 down-regulated genes, most of which have not previously been observed in PETs, and whose gene ontology molecular function has been detailed. Overexpression of bridging integrator 1 (BIN1) and protein Z dependent protease inhibitor (SERPINA10) which may represent useful biomarkers, and of lymphocyte specific protein tyrosine kinase (LCK) and bone marrow stromal cell antigen (BST2) which could be used as therapeutic targets, has been validated. When primary tumours were compared with metastatic lesions, no significantly differentially expressed genes were found, in accord with cluster analysis which revealed a striking similarity between primary and metastatic lesions, with the cell lines clustering separately. We have provided a comprehensive list of differentially expressed genes in a uniform set of aggressive NF PETs. A number of dysregulated genes deserve further in-depth study as potentially promising candidates for new diagnostic and treatment strategies. The analysis of liver metastases revealed a previously unknown high level of similarity with the primary lesions.