Analysis Code and Computed Tables for a Reproducible Census of Declared Inactive Ingredients in U.S. Oral Drug Listings

Deterministic, standard-library Python code and the computed tables behind the manuscript "Declared inactive ingredients in U.S. oral drug listings: a reproducible census of prevalence, labeling coverage, and within-name variation." This is the analysis companion to the source dataset Declared Inactive Ingredients in U.S. Oral Drug Product Listings: A Census of 51,065 Eligible Distinct Product NDC Listings (Zenodo 10.5281/zenodo.21482069).

Contents: the avoidance-relevant excipient panel as explicit FDA Unique Ingredient Identifier (UNII) sets with inclusion/exclusion rationale (Supplementary Table 1); the panel prevalence and within-generic-name declaration-consistency pipeline (which SHA-256-verifies the census data before computing); the characterization pipeline (excipient burden, most-prevalent excipients, co-occurrence, qualitative-profile diversity, structured-declaration availability, and Established Pharmacologic Class (EPC) stratification); the EPC-crosswalk builder plus the frozen product-NDC-to-EPC map from the 2026-07-23 openFDA export for byte-exact reproduction; and every computed output table together with the two machine-readable number files the manuscript is filled from.

All code is deterministic and standard-library only; re-running it against the source census reproduces every output bit-for-bit. Derived from public-domain U.S. government data (the openFDA National Drug Code Directory and DailyMed Structured Product Labeling); this analysis is independent and not endorsed by FDA or NLM. Code is released under MIT; the excipient panel and computed tables are additionally released under CC BY 4.0.