The development of high-throughput metabolite measurement technologies has enabled the use of metabolomics for epidemiologic studies by profiling metabolite concentrations in large cohorts of human blood samples. Standard protocols are necessary to obtain unbiased profiles through multiple runs over long periods of time and to allow reliable statistical analyses. This study assessed the effects of sampling procedures and storage conditions on the stability of metabolomic profiles in plasma and serum. Charged metabolomic profiles were determined by capillary electrophoresis-mass spectrometry (CE-MS) and compared by multivariate analyses. The effects of pre-analytical procedures, including times for clotting and incubation of serum and plasma, respectively; incubation temperatures; and number of freeze-thaw cycles, were assessed. Overall, inter-individual differences in profiles were larger than intra-individual differences, and profiles in plasma showed better stability than those in serum. These quantified datasets of metabolites, along with their stability and variation, may help in interpreting data from long-term cohort studies.