Systematic literature reviews for drug development programmes — safety background documents, clinical overview sections of regulatory dossiers, evidence packages for indication expansion decisions — typically take six to eighteen months and consume substantial expert time. A significant proportion of that time is spent on literature search: designing search strings, running them across multiple databases, removing duplicates, and iterating when initial results are incomplete. Ontology-driven search does not automate the expert judgement required for evidence synthesis, but it substantially compresses the initial evidence retrieval phase.
The Iterative Search Problem
Keyword-based literature search is inherently iterative. Reviewers run an initial search, examine the results, notice that important known studies are missing, infer additional search terms from those missing studies, add them to the search string, and re-run. This cycle repeats until the reviewers are confident (but rarely certain) that the result set is reasonably complete. Each iteration takes days to weeks. The total time from initial search design to a defensibly comprehensive result set typically represents two to three months of elapsed time even before any evidence assessment begins.
Achieving Comprehensive First-Pass Retrieval
Ontology-driven search achieves comprehensive retrieval in the first pass by automatically expanding the query to include all synonyms, hierarchically related concepts, and formally equivalent terms defined in the relevant biomedical ontologies. A query for evidence about a specific compound automatically includes all approved synonyms, chemical name variants, and CAS registry numbers. A query about a disease class automatically includes all specific subtypes, historical names, and related coding system identifiers. The result is a first-pass result set that would otherwise require multiple keyword iterations to assemble — reducing the search iteration cycle from months to days.
Documentation and Auditability
Regulatory submissions require that literature search methodologies be documented and reproducible. Ontology-driven searches have a significant advantage here: the query expansion is explicit and traceable to specific ontological relationships, making the search methodology fully documentable. Reviewers can provide a complete audit trail of exactly which concept identifiers were queried, which synonym sets were included, and which ontology version was used — a level of methodological transparency that keyword string searches cannot match.
