Integrating diverse metabolomics data for molecular epidemiology analyses provides both opportuni- ties and challenges in the field of human health research. Combining patient cohorts may improve power and sensitivity of analyses but is challenging due to significant technical and analytical vari- ability. Additionally, current systems for the storage and analysis of metabolomics data suffer from scalability, query-ability, and integration issues that limit their adoption for molecular epidemiological research. Here, a novel platform for integrative metabolomics is developed, which addresses issues of storage, harmonisation, querying, scaling, and analysis of large-scale metabolomics data. Its use is demonstrated through an investigation of molecular trends of ageing in an integrated four-cohort dataset where the advantages and disadvantages of combining balanced and unbalanced cohorts are explored, and robust metabolite trends are successfully identified and shown to be concordant with previous studies.