The aim of this study was to identify predictor sets of genes whose over- or underexpression in human sporadic adrenocortical tumors would help to identify malignant vs. benign tumors and to predict postsurgical metastatic recurrence. For this, we analyzed the expression of 230 candidate genes using cDNA microarrays in a series of 57 well-characterized human sporadic adrenocortical tumors (33 adenomas and 24 carcinomas). We identified two clusters of genes (the IGF-II cluster containing eight genes, including IGF-II, and the steroidogenesis cluster containing six genes encoding steroidogenic enzymes plus eight other genes) whose combined levels of expression appeared to be good predictors of malignancy. This predictive value was as strong as that of the pathological score of Weiss. The analysis of the population of carcinomas (13 tumors) for genes whose expression would be strongly different between recurring and nonrecurring tumors allowed identification of 14 genes meeting these criteria. Among these genes, there are probably new markers of tumor evolution that will deserve additional validation on a larger scale. Taken together, these results show that the parallel analysis of the expression levels of a selected group of genes on microgram quantities of tumor RNA (a quantity that can be obtained from fine needle aspirations) appears as a complementary method to histopathology for the diagnosis and prognosis of evolution of adrenocortical carcinomas.