Drug-associated osteoarthritis signals in FAERS: A pharmacovigilance study.

OBJECTIVE: To identify drugs disproportionately reported with osteoarthritis (OA) in the FDA Adverse Event Reporting System (FAERS) and to prioritize candidate molecular features for further validation. METHODS: FAERS reports from the first quarter of 2004 to the first quarter of 2025 were analyzed. Disproportionality analysis was performed using the reporting odds ratio and bayesian confidence propagation neural network methods. Sensitivity analyses were conducted using alternative lower-bound thresholds for the reporting odds ratio confidence interval. Least absolute shrinkage and selection operator regression was applied as an exploratory feature-refinement method. Time-to-onset (TTO) and Weibull shape parameter analyses were performed to describe temporal reporting patterns. Drug-related genes were integrated with OA transcriptomic data, followed by functional enrichment, protein-protein interaction network analysis, and machine-learning-based candidate-gene prioritization. RESULTS: Pharmacovigilance analysis identified 18 drugs or fixed-dose combinations with drug-associated OA reporting signals. Among 2964 of 18,175 OA reports (16.3%) with complete and valid date information, the median TTO was 245 days (Q1-Q3: 29-789 days). Weibull shape parameter analysis yielded a β value of 0.524 (95% CI: 0.509-0.539), indicating an early-failure reporting pattern. Transcriptomic analysis identified 80 OA-related pharmacovigilance-related genes. Subsequent exploratory network and machine-learning analyses prioritized CP and RUNX1 as candidate genes for further investigation. CONCLUSION: This study identified drug-associated OA reporting signals in FAERS and prioritized candidate genes that warrant further validation. These findings should be interpreted as hypothesis-generating rather than as evidence of causal relationships, confirmed molecular mechanisms, or validated diagnostic biomarkers.