Gynecological malignancy signal triage: A two-dimensional framework integrating US-derived FAERS data with comparative US–China GBD trajectories
Ying Ma, Dandan Li, Jieru Duan, Zhuowen Sun, Qirong Dou, Tao Wu, Na GaoBackground
Cancer-related adverse-event reports are difficult to interpret because statistical disproportionality, stimulated reporting, and background disease burden are distinct phenomena. To address this, we formalized a two-dimensional prioritization framework that combines FAERS signal credibility with independent GBD burden context. Because the FAERS dataset is overwhelmingly of US origin, we utilized China exclusively as a prespecified macro-epidemiological GBD comparator, strictly avoiding cross-national pharmacovigilance extrapolations.
Methods
We analyzed FAERS reports (2004–2023) and female GBD 2023 estimates for China and the United States. Four disproportionality algorithms (ROR, PRR, IC, and EBGM) were evaluated for 150 high-frequency primary-suspect drug–event pairs (50 per cancer group). We report any-algorithm positivity and four-algorithm concordance alongside case-definition sensitivity analyses. GBD trends, age–period–cohort models, and exploratory projections to 2053 were integrated to provide population-level context, explicitly not as exposure denominators or for causal validation.
Results
FAERS cohorts yielded 19,717 cancer-associated report records. Among 150 high-frequency candidate pairs, 42 met all four algorithmic thresholds. The top-50 candidate screen captured approximately 72.8%–75.8% of primary-suspect records, serving as a variance-reduction step rather than a complete all-drug screen. Drugs like ranitidine and talc demonstrated marked disproportionality but also exhibited reporter and temporal features heavily compatible with stimulated reporting. Time-to-onset distributions reflected recorded administrative reporting intervals rather than biological carcinogenesis. GBD estimates revealed divergent China–US trajectories, notably a rising baseline for US uterine cancer.
Conclusions
The formalized F × G framework operates as a transparent triage tool, not a causal classifier. FAERS disproportionality generates hypotheses; GBD trajectories indicate potential public-health priority. All prioritized signals—particularly those involving consumer products or immunosuppressants—require formal causal evaluation in longitudinal exposure-denominator datasets.