Real-World Evidence in FDA Safety-Related Labeling Changes: Public Documentation, Analytic Use, and Transparency

This dataset contains event-level annotations for FDA Safety-related Labeling Changes (SrLCs) used to characterize the public documentation of real-world evidence (RWE) in postmarketing drug safety labeling changes. The main dataset includes 10,616 FDA SrLC events with drug metadata, application information, label-section changes, public evidence availability, broad public RWE documentation, analytic public RWE documentation, RWE source and study-design categories, relevance of RWE to the labeling change, transparency-score components, annotation provenance, and post-QC issue flags. A curated validation file summarizes human consensus labels for a stratified validation sample, together with machine-versus-human agreement metrics. The public-release dataset is designed for transparency and reproducibility of the manuscript analyses. It excludes raw labeling-change text, free-text annotation notes, supporting-evidence JSON snippets, individual reviewer workbooks, adjudication worksheets, prompt/runtime files, and packet text files. Public FDA event links and derived annotation variables are retained. Included files provide the main public-release CSV, a data dictionary, the annotation codebook, the annotation pipeline summary, aggregate annotation counts, residual QC issue flags, and human-validation summary tables. The event metadata and public document links derive from publicly available FDA sources. The annotation labels and derived variables were generated for this study.