Reproducibility of metabolomic profiles among men and women in 2 large cohort studies.

Background: Rigorous studies are necessary to demonstrate suitability of metabolomics platforms to profile metabolites in archived plasma within epidemiologic studies of human disease, for which attenuation of effect estimates due to measurement error is a key concern.

Methods: Using a liquid chromatography-tandem mass spectrometry platform, we quantified 257 metabolites from archived plasma to evaluate metabolite interassay reproducibility, reproducibility with delayed processing, and within-person reproducibility over time. Interassay reproducibility was assessed with CVs from 60 duplicate plasma samples donated by participants in the Nurses' Health Study and Health Professionals Follow-up Study, and 20 QC pool plasma replicates. Metabolite reproducibility over a 24- to 48-h processing delay (n = 48 samples) and within-person reproducibility over 1-2 years (n = 80 samples) were assessed using Spearman and intraclass correlation coefficients (ICCs).

Results: CVs were <20% for 92% of metabolites and generally were similar by plasma anticoagulant type (heparin or EDTA) and fasting time. Approximately 75% of metabolites were reproducible over delays in processing of blood samples (Spearman correlation or ICC ≥ 0.75, comparing immediate and 24-h delayed processing). Carbohydrates and purine/pyrimidine derivatives were most adversely affected by the processing delay. Ninety percent of metabolites were reproducible over 1-2 years within individuals (Spearman correlation or ICC ≥ 0.4).

Conclusions: For potential use in epidemiologic studies, the majority of plasma metabolites had low CVs and were reproducible over a 24-h processing delay and within individuals over 1-2 years. Certain metabolites, such as carbohydrates and purine/pyrimidine derivatives, may be challenging to evaluate if samples have delayed processing.