(P13-044-24) Dietary and Physical Activity Biomarker Development Using Metabolomics in the Interactive Diet and Activity Tracking in AARP (IDATA) Study

Objectives: Cross-sectional and feeding studies have used metabolomics to identify potential biomarkers of diet and physical activity. We created a resource for biomarker development in a longitudinal study with multiple dietary and activity assessment tools, reference biomarkers, and biospecimen types. Herein, we describe the technical reproducibility and biological variability of measured metabolites.

Methods: We used ultra-high-performance liquid chromatography with tandem mass spectrometry to measure a broad range of metabolites in serial serum, 24-hour urine, and first morning void (FMV) urine samples, collected six months apart, from 718 Interactive Diet and Activity Tracking in AARP (IDATA) study participants. We used linear mixed models to parse total variability (π) into between-subject variability (π_B), within-subject variability (π_W), and technical variability (π_T) for 921 serum, 1076 24-hour urine, and 1058 FMV urine metabolites. We estimated technical ICCs ([π_B + π_W]/ π) and 6-month ICCs (π_B/[π_B + π_W]) for each metabolite. We estimated Pearson correlation coefficients to ascertain whether metabolite measures in serum and FMV urine, which are available in current and more easily collected in future epidemiologic studies, are reflective of those in 24-hour urine.

Results: Technical reproducibility was high with median ICCs of 0.89, 0.91, and 0.93 for serum, 24-hour urine, and FMV urine, respectively. 6-month median ICCs were lower at 0.77, 0.64, and 0.50 for serum, 24-hour, and FMV urine, respectively. Metabolites measured in FMV and 24-hour urine (n=1,043) were highly correlated (median r = 0.76, IQR=0.66-0.85). Nearly 50% of metabolites measured in serum were measured in urine (n=452) but were less correlated (median r = 0.35, IQR=0.19-0.50). A larger number of urine (n=173) than serum (n=88) metabolites were xenobiotics, which include compounds derived from diet (e.g., trigonelline from coffee).

Conclusions: Using the IDATA infrastructure, we created a valuable resource for biomarker development with high quality serum and urine metabolomics data and comprehensive dietary and physical activity assessments. We encourage interested researchers to use this resource (https://cdas.cancer.gov/idata) to advance science on the role of diet and physical activity in human health.

Funding Sources: NIH Intramural Research Program