Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) has long been used as a default reference gene in quantitative mRNA profiling experiments. However, its expression reportedly varies in response to a range of pathophysiological variables (inflammation, oxidative stress, hyperinsulinaemia, hypoxia) which feature prominently in sepsis. We therefore assessed the applicability of using GAPDH as a reference gene for expression studies in sepsis compared to other housekeeping genes (succinate dehydrogenase complex subunit A (SDHA), hypoxanthine phosphoribosyltransferase (HPRT)-1). Severe sepsis resulted in a 42.4-fold increase in median GAPDH expression (P<0.001), whereas median HPRT expression was raised more modestly (2.9-fold; P<0.001), and there was no significant difference in SDHA expression between sepsis and control patients. HPRT was identified by NormFinder to be the most stably expressed single gene. In order to assess the impact of this variability on data interpretation, interleukin (IL)-10 expression was normalised separately to GAPDH and to the geometric mean of HPRT and SDHA. In the former case, there was no significant difference in IL-10 expression between controls and septic patients, whilst in the latter, a significant 8.5-fold increase in median IL-10 expression was noted (P<0.001). GAPDH is thus an unreliable housekeeping gene for normalising gene expression in sepsis which should be replaced by alternative, validated reference genes.