BACKGROUND: Gene expression measurements became an attractive tool to assess biological responses in epidemiological studies. However, collection of blood samples poses various technical problems. We used gene expression data from two epidemiological studies to evaluate differences between sampling methods, comparability of two methods for measuring RNA levels and stability of RNA samples over time. METHODS: For the PARSIFAL study, PBLC of 1155 children were collected using EDTA tubes in two countries. In the PASTURE study, tubes containing RNA-stabilizing solutions (PAXgene) Blood RNA Tubes; PreAnalytiX) were used to collect cord blood leucocytes of 982 children in five countries. Real-time PCR (conventional single tube assay and high-throughput low density arrays) was used to quantify expression of various innate immunity genes. In 77 PARSIFAL samples, gene expression was measured repeatedly during prolonged storage. RESULTS: In PARSIFAL (EDTA tubes) the median RNA yield after extraction significantly differed between the two centres (70 and 34 ng/microl). Collecting blood into an RNA-stabilizing solution markedly reduced differences in RNA yield in PASTURE (range of medians 91-107 ng/microl). The agreement [Spearman rank correlation (r)] between repeated measurements of gene expression decreased with increasing storage time [e.g., for CD14: r (first/second measurement) = 0.35; r (first/third measurement) = 0.03]. RNA levels measured with either the conventional method or low-density arrays were comparable (r > 0.9). CONCLUSION: Collecting blood samples into tubes containing an RNA-stabilizing solution increases RNA yield and reduces its variability. Long-term storage of samples may lead to RNA degradation, requiring special attention in longitudinal studies.