Background: Gene expression profiling is thought to be an important tool in determining treatment strategies for breast cancer patients. Tissues for such analysis may at a preoperative stage be obtained, by fine needle aspiration (FNA) allowing initiation of neoadjuvant treatment. To evaluate the extent of the genetic heterogeneity within primary breast carcinomas, we examined whether a gene expression profile obtained by FNA was representative of the tumor. Methods: Tumors from 12 consecutive cases of early predominantly estrogen receptor positive (ER+) breast cancer patients undergoing primary surgery were split in halves and FNAs were obtained from each half. A tissue biopsy of the tumors was also snap-frozen for comparison. Non-amplified RNA was investigated by the novel qRT-PCR-based technique, Low Density Array (LDA) using 4 reference genes and 44 target genes. Results: Comparison of gene expression at the single gene level in the two FNA samples from each tumor demonstrated various degrees of heterogeneity. However, compared as gene expression profiles, intratumor correlations for 9/12 patients were high and these pairs could in a theoretical blinding of all the FNAs be correctly matched by statistical analysis. High correlations between the gene profiles of tumor FNAs and tissue biopsies from the same patient were observed for all patients. A cluster analysis identified clustering of both the two FNAs and the tissue biopsy of the same 9 patients. Conclusion: The overall genetic heterogeneity of breast carcinomas, as sampled by FNA, does not prohibit generation of useful gene profiles for treatment decision making. However, sampling and analysis strategies should take heterogeneity within a tumor, and varying heterogeneity amongst the single genes, into account.