BACKGROUND: The use of housekeeping genes (HKG) as internal controls for real-time qPCR studies of gene expression is based on the assumption of their inherent stability. However, it is unclear whether this stability is maintained in disease states. In order to test this, the present study investigated the expression of specific HKG in the endometrium of healthy and polycystic ovarian syndrome (PCOS) women. METHODS: Endometrial tissue samples were taken from women with PCOS (n 9) and controls (n 10). The stability of nine candidate reference genes in the endometrial tissues were evaluated; four encode mitochondrial proteins [ATP5B, succinate dehydrogenase complex subunit A (SDHA), cytochrome c-1, glyceraldehyde-3-phosphatedehydrogenase], two encode ribosomal protein genes (18s ribosomal RNA, ribosomal protein L13A), one for cell structure (SDHA), one for cell signalling (beta actin, ACTB) and one involved in DNA repair (topoisomerase I, TOP1). The expression stability of these HKGs was calculated using geNORM qbasePLUS software, with stability defined by M-values, where higher M-value indicating less stability. In addition, changes in their cycle threshold values were analysed to determine direction of change between groups, and a MannWhitney U-test was used to determine statistical differences in expression. RESULTS: The most stable HKGs observed across both PCOS endometrium were found to be YWHAZ, CYC1 and ACTB. Further TOP1 demonstrated higher gene expression in the endometrium from PCOS women compared with those from healthy women. CONCLUSIONS: Of the nine HKGs examined, only YWHAZ, CYC1 and ACTB were stable in both control and PCOS endometrium: these should therefore be used as internal controls for quantitative reverse transcription-polymerase chain reaction analysis. Published discrepancies between endometrial gene expression studies may therefore be due in part to in the inappropriate HKG selection, and future gene expression studies should be based on HKG of known stability in both the disease and healthy states to avoid erroneous interpretation of results.