The malignant potential among endocrine pancreatic tumors (EPTs) varies greatly and can frequently not be predicted using histopathological parameters. Thus, molecular markers that can predict the biological behavior of EPTs are required. In a previous comparative genomic hybridization study, we observed marked genetic differences between the various EPT subtypes and a correlation between losses of 3p and 6 and gains of 14q and Xq and metastatic disease. To search for genetic alterations that play a role during early tumor development, we have studied 38 small (< or =2 cm) EPTs, including 24 insulinomas and 10 nonfunctioning EPTs. Small EPTs are usually classified as clinically benign tumors in the absence of histological signs of malignancy. Using comparative genomic hybridization, we identified chromosomal aberrations in 27 EPTs (mean, 4.1). Interestingly, the number of gains differed strongly between nonfunctioning and functioning EPTs (3.4 versus 1.5, respectively; P = 0.0526), as did the number of aberrations in the benign (n = 30) and malignant (n = 8) tumors (3 versus 8.4, respectively; P = 0.0022). In the insulinomas, 9q gain (common region of involvement: 9q34) was most common (50%) and in nonfunctioning EPTs, gain of 4p was most common (40%). Most frequent losses in insulinomas involved 1p (20.8%), 1q, 4q, 11q, Xq, and Y (all 16.7%) and in nonfunctioning EPTs, 6q. Losses of 3pq and 6q and gains of 17pq and 20q proved to be strongly associated with malignant behavior in all of the small EPTs (P < or = 0.0219). Our results demonstrate marked genetic differences between small functioning and nonfunctioning EPTs, indicating that these subtypes evolve along different genetic pathways. In addition, our study endorses the importance of chromosomes 3 and 6q losses to discriminate EPTs with a malignant behavior from benign ones.