34 The HRA was established in 2005 as an alliance that fosters collaboration among nonprofit, nongovernmental funders to support health research and training across a continuum of biomedical science applications that advance health. The HRA also has a shared grants database called Grants in the Health Research Alliance Shared Portfolio (gHRAsp, www.ghrasp.org), which has been previously described[13] . Importantly, gHRAsp includes funded grant information from cancer funders that are not part of the ICRP, including Breast Cancer Research Foundation, and large foundations that are not cancer specific, including the Burroughs Wellcome Fund, Doris Duke Charitable Foundation, Howard Hughes Medical Institute, and others. (For a complete list of HRA members, see www.healthra.org.) Research grants were extracted from the ICRP and HRA databases using combinations of keywords (breast cancer and metastasis, metastatic, metastases, metasta*, advanced or stage IV) followed by manual validation to ascertain their relatedness to MBC, creating a MBC Grants Dataset. Duplicate grants were removed (e.g., grants from the American Cancer Society, Avon, and Komen that were in both databases). For grants in the ICRP database, we limited our analysis to those identified as having at least 50% relevance to breast cancer (vs. relevance to many or all cancers). We then manually reviewed a random sample (n=100) of grants in the MBC Grants Dataset to validate our search and data extraction strategies. The abstracts of the grants within the random sample confirmed to be relevant to MBC were then used to manually classify each grant in the full MBC Grants Dataset according to the categories in Table 1; key information on targets and therapies under study was extracted. A team of 8 volunteer coders manually assigned the grants in the MBC Grants Dataset to the metastatic stage corresponding to key parts of the Steps in Metastasis framework and Hallmarks of Cancer framework. These assignments were reviewed and validated by 2 additional coders who reviewed the entire dataset. Grants were also categorized by model system or study type as preclinical research, technologic development, or therapy/intervention. The research stage (basic, translational, clinical, or cancer control research) was assigned by mapping the framework assignments to CSO codes. These assignments were manually validated. We extracted the grant information into a large spreadsheet with multiple pivot tables and analyzed the number of grants and dollar amount of funding in each category over time. We also developed a comprehensive list of molecular targets, pathways, and therapies identified in abstracts of the funded grants.