Selection of reference genes to normalize mRNA levels between samples is critical for gene expression studies because their expression can vary depending on the tissues or cells used and the experimental conditions. We performed ten cell cultures from samples of prostate cancer. Cells were divided into three groups: control (with no transfection protocol), cells transfected with siRNA specific to knockdown the androgen receptor and cells transfected with inespecific siRNAs. After 24 h, mRNA was extracted and gene expression was analyzed by Real-time qPCR. Nine candidates to reference genes for gene expression studies in this model were analyzed (aminolevulinate, delta-, synthase 1 (ALAS1); beta-actin (ACTB); beta-2-microglobulin (B2M); glyceraldehyde-3-phosphate dehydrogenase (GAPDH); hypoxanthine phosphoribosyltransferase 1 (HPRT1); succinate dehydrogenase complex, subunit A, flavoprotein (Fp) (SDHA); TATA box binding protein (TBP); ubiquitin C (UBC); tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide (YWHAZ)). Expression stability was calculated NormFinder algorithm to find the most stable genes. NormFinder calculated SDHA as the most stable gene and the gene with the lowest intergroup and intragroup variation, and indicated GAPDH and SDHA as the best combination of two genes for the purpose of normalization. Androgen receptor mRNA expression was evaluated after normalization by each candidate gene and showed statistical difference in the transfected group compared to control group only when normalized by combination of GAPDH and SDHA. Based on the algorithm analysis, the combination of SDHA and GAPDH should be used to normalize target genes mRNA levels in primary culture of prostate cancer cells submitted to transfection with siRNAs.