| アブストラクト | To develop and apply a quarterly pharmacovigilance workflow for the 2025 FDA Adverse Event Reporting System (FAERS) public-use XML releases, with emphasis on temporal persistence, cohort sensitivity, and prioritization-oriented signal characterization. The raw input consisted of 12 quarterly FAERS XML files (approximately 8.58 GB), which were parsed into report-level drug-reaction pairs using a chunked workflow. The main analysis used a PS-clinical cohort including primary suspect drugs from serious reports, whereas a sensitivity analysis expanded this definition to a PS + SS-clinical cohort by including secondary suspect drugs. Quarterly product-reaction pairs were screened using a disproportionality rule requiring n11 >/= 3 , proportional reporting ratio PRR >/= 2 , chi2 >/= 4 , and a lower 95% confidence bound of the reporting odds ratio greater than 1. A more conservative subset was defined by n11 >/= 10 , followed by post-screening refinement for temporal comparison. Under the PS-clinical cohort, the quarterly retained sets ranged from 41,897 to 47,220 pairs, and the annual union contained 82,565 unique retained pairs. Of these, 18,629 (22.56%) were detected in all four quarters. Pairwise Jaccard similarities ranged from 0.4027 to 0.4845, indicating moderate but structured temporal continuity, with the strongest overlap between Q3 and Q4. Fully persistent 4/4 pairs showed higher support counts and larger mean chi2 values than quarter-specific 1/4 pairs, whereas mean PRR did not differ materially between these groups. Within the 4/4 core, most pairs were relatively stable over time, while only a small subset showed marked temporal imbalance. Expansion to the PS + SS-clinical cohort substantially increased both the retained signal space and the 4/4 core, while leaving the overall temporal overlap pattern broadly similar. Quarterly persistence may provide an additional and interpretable dimension for internal FAERS-based signal characterization and prioritization. Rather than replacing disproportionality metrics, it may complement them by highlighting reproducibility, continuity, and temporal structure across quarterly retained signal sets. External validation against independent pharmacovigilance reference standards remains an important direction for future work. |
|---|
| 組織名 | Departamento de Matematicas, Universidad Autonoma Metropolitana-Iztapalapa, CDMX,;Iztapalapa, 09340, Mexico. jamc88@xanum.uam.mx.;Departamento de Fisica, Universidad Autonoma Metropolitana-Iztapalapa, CDMX,;Iztapalapa, 09340, Mexico. |
|---|
