Impact of post-collection freezing delay on the reliability of serum metabolomics in samples reflecting the California mid-term pregnancy biobank

Population-based biorepositories are important resources, but sample handling can affect data quality. Identify metabolites of value for clinical investigations despite extended postcollection freezing delays, using protocols representing a California mid-term pregnancy biobank. Blood collected from non-pregnant healthy female volunteers (n = 20) underwent three handling protocols after 30 min clotting at room temperature: (1) ideal—samples frozen (− 80 °C) within 2 h of collection; (2) delayed freezing—samples held at room temperature for 3 days, then 4 °C for 9 days, the median times for biobank samples, and then frozen; (3) delayed freezing with freeze–thaw—the delayed freezing protocol with a freeze–thaw cycle simulating retrieved sample sub-aliquoting. Mass spectrometry-based untargeted metabolomic analyses of primary metabolism and complex lipids and targeted profiling of oxylipins, endocannabinoids, ceramides/sphingoid-bases, and bile acids were performed. Metabolite concentrations and intraclass correlation coefficients (ICC) were compared, with the ideal protocol as the reference. Sixty-two percent of 428 identified compounds had good to excellent ICCs, a metric of concordance between measurements of samples handled with the different protocols. Sphingomyelins, phosphatidylcholines, cholesteryl esters, triacylglycerols, bile acids and fatty acid diols were the least affected by non-ideal handling, while sugars, organic acids, amino acids, monoacylglycerols, lysophospholipids, N-acylethanolamides, polyunsaturated fatty acids, and numerous oxylipins were altered by delayed freezing. Freeze–thaw effects were assay-specific with lipids being most stable. Despite extended post-collection freezing delays characteristic of some biobanks of opportunistically collected clinical samples, numerous metabolomic compounds had both stable levels and good concordance.