Real-world analysis of depression-related adverse events associated with the adsorbed anthrax vaccine: integrating pharmacovigilance signals, machine learning risk prediction, and transcriptomic mechanisms.

backgroundAnthrax remains a life-threatening zoonotic disease in resource-limited settings. Adsorbed anthrax vaccine (AVA, BioThrax) is the only United States-authorized preexposure prophylaxis for high-risk Bacillus anthracis exposure. Although AVA safety has been described, depression-related adverse events (AEs) remain insufficiently characterized. This study evaluated real-world depression-related safety signals associated with AVA and anthrax-related antibacterial agents.

methodsAVA-related AE reports were extracted from the Vaccine Adverse Event Reporting System (VAERS), and depression-related AE reports for seven anthrax-related antibacterial agents were retrieved from the FDA Adverse Event Reporting System (FAERS). AEs were coded using Medical Dictionary for Regulatory Activities (MedDRA). Disproportionality analyses were performed using four standard signal-detection methods, with sensitivity analysis excluding concurrent vaccination. Seven complementary machine learning algorithms, covering regression, tree-based, boosting, kernel, instance-based, and probabilistic methods, explored predictors, with Shapley Additive Explanation (SHAP) for interpretation. Time-to-onset, clinical priority, and exploratory transcriptomic signatures were also assessed.

resultsIn VAERS, 9446 AVA-related AE reports were identified, including 180 depression-related cases [total depression population n = 180, ROR 5.02 (95% CI: 4.33-5.81)]. Males (78.2%) and individuals aged 31-45 years (49.1%) were most represented, with a median time-to-onset of 10 days and 44.1% classified as serious. Stronger signals were observed in individuals aged 46-65 years [n = 35, ROR 10.31 (95% CI: 7.37-14.42)] and males [n = 142, ROR 6.11 (95% CI: 5.16-7.24)]. Machine learning highlighted serious AEs and age as key predictors. Transcriptomic analysis suggested immune-inflammatory pathway overlap, including neutrophil activation and Yersinia infection. In FAERS, levofloxacin [n = 634, ROR 1.37 (95% CI: 1.27-1.49)] and doxycycline [n = 332, ROR 1.53 (95% CI: 1.38-1.71)] showed positive depression-related reporting signals; "Depression suicidal" had moderate clinical priority.

conclusionsAVA was associated with a potential depression-related pharmacovigilance signal, particularly among males and adults, with early post-vaccination onset. These findings support targeted mental health monitoring after AVA, clearer safety communication in anthrax prevention programs, and vigilance when levofloxacin or doxycycline is used for anthrax therapy. As spontaneous reporting cannot establish causality or incidence, controlled epidemiological studies are needed.