Quantitative real-time PCR (qPCR) involves the need of a proper standard for normalizing the gene expression data. Different studies have shown the validity of reference genes to vary greatly depending on tissue, cell subsets and experimental context. This study aimed at the identification of suitable reference genes for qPCR studies using different peripheral blood cell subsets (whole blood (WB) cells, peripheral blood mononuclear cells (PBMCs) and PBMC subsets (CD4(+) T cells, CD8(+) T cells, NK cells, monocytes, B cells and dendritic cells) from healthy controls (HC), patients with relapsing-remitting multiple sclerosis (RRMS) and interferon-β-treated patients with RRMS (RRMS-IFN-β). Eight candidate reference genes (CASC3, EEF1A1, GAPDH, HPRT1, RPLP0, UBC, UBE2D2 and YWHAZ) were analysed using normfinder and genorm algorithms to identify the most stably expressed genes. We found reference gene expression varied most across cell subsets, and less variation between the donor groups. UBE2D2 was the most stably expressed gene across both HC and RRMS patients and across cell subsets, while UBC was the most stably expressed gene between HC, RRMS and RRMS-IFN-β patients. UBE2D2 and HPRT1 was the most stable combination for analyses of cell subsets between HC and RRMS patients, while the combination of UBC and YWHAZ was superior for analysis of cell subsets between HC, RRMS and RRMS-IFN-β groups. GAPDH was generally unsuitable for blood cell subset studies in multiple sclerosis. In conclusion, we found that blood cell subsets result in marked variation in reference gene expression, and we identified suitable reference genes for studies involving PBMC subsets, RRMS patients and interferon-β treatment.