The aim of this study was to investigate the intratumor heterogeneity of gene expression profiles in epithelial ovarian cancer (EOC). This was done to evaluate whether sampling of a single macrodissected tissue sample from each EOC case would bias the data and result in, eg, prognostic studies based on gene expression microarray experiments. From nine EOCs removed at Odense University Hospital, Denmark, three tumor samples of 200-300 mg each were taken with greatest possible mutual distance. The samples were immediately flash frozen. A parallel section was taken for histopathologic comparison. RNA was extracted from the tissue samples. Five micrograms of each RNA sample was used for labeling. The fragmented biotin-labeled complementary RNA was hybridized to Affymetrix GeneChip Human Genome U133 plus 2.0 arrays, and scanning was performed on the GeneArray scanner 3000 (Affymetrix, Santa Clara, CA). Data were evaluated using hierarchical clustering and intraclass correlation coefficient (ICC) from reliability analysis. All evaluation methods revealed low intratumor heterogeneity. Intratumor ICCs ranged from 0.888 to 0.978. In contrast, "between-tumor" ICC was 0.549 indicating much lower intra- than intertumor heterogeneity. Due to a low degree of intratumor variation, we conclude that it is sufficiently accurate in a clinical setup to use single, macrodissected tumor samples in the evaluation of gene expression in EOCs.