NMEs approved by the FDA from 2010–2016 were identified from FDA reports (35 ) and designated “first in class” or “follow-on” based on assessment by the FDA (36 ). NMEs were designated “phenotypic” or “targeted” based on the criteria of Swinney et al. (29 (link), 30 (link)). Known molecular targets for each NME and approved clinical indications were determined from FDA labels (37 ) and other sources as described in SI Methods. For biological products comprising a naturally occurring protein, the target is considered to be the normal counterpart of the biological product.
PubMed searches were performed for each drug (“drug search”) using an ontology of drug name synonyms in ChEMBL (38 ) and the National Center for Biotechnology Information (NCBI) Query Translation. PubMed searches for molecular targets (“target searches”) were performed using Boolean search terms and NCBI Query Translation. The PubMed Identifier (PMID) was recorded for each publication identified in the search.
Data associating publications with specific NIH-funded projects were obtained from the RePORTER/ExPORTER format files catalog (39 ). The “Link Tables for Project to Publication Associations” (hereafter, “Link Table”) associates PMIDs from 1980–present with projects that provided research funding and the PMID year. Each PMID was associated with a funding year corresponding to the project number and year in the Link Table. The Project Data Table provides the fiscal year cost for each project (2000–present). Costs were assigned for each funding year corresponding to the program cost in the year associated with the PMID in the Link Table. For publications with dates 1–4 y after the end of the project, costs for the final year of the project were used. The activity code associated with the core project number indicates the grant type.
Redundant identification of PMIDs and funding years occurred when a publication was identified in different drug or target searches or was cited in more than one supporting project. Consequently, each analysis required two steps, first identifying all PMIDs or project years with the specific properties being characterized and then eliminating duplicates within that subset.
Funding years were categorized as “drug” if one or more of the PMIDs associated with that project were identified in a drug search. Funding years were categorized as “target only” if every PMID associated with that project was identified through target searches. The process is illustrated in a schematic (Fig. S1), and an illustrative example (venetoclax) is shown in Fig. S2.
Data analysis and visualization were performed in PostgreSQL, Excel, and Tableau. All costs are given in constant dollars inflation-adjusted to 2016 using the US Bureau of Labor Statistics’ consumer price index (CPI) (40 ). A more detailed description of the analytical methods is provided in SI Methods. The search terms, summary statistics of each search, and complete dataset of PMIDs and associated funding years are provided in Dataset S1.