Purpose: Dietary supplements are consumed by over half of US adults, yet post-market safety surveillance remains limited. The FDA’s CFSAN Adverse Event Reporting System (CAERS) contains over 230,000 adverse event reports but has been the subject of only one prior computational signal detection analysis. We applied four disproportionality methods, temporal signal detection, and demographic stratification to the expanded CAERS dataset (2004–2025) to identify and validate safety signals for dietary supplements. Methods: We analysed 48,840 unique adverse event reports from the FDA CAERS database. Four disproportionality methods — PRR, ROR, Gamma-Poisson Shrinker (GPS), and BCPNN — were applied to 4,779 product–PT pairs with N ≥ 3. Cumulative sum (CUSUM) control charts provided temporal signal detection. Demographic stratification by age, sex, and product category used Woolf tests for odds ratio homogeneity. Twenty-four validation and sensitivity analyses assessed signal robustness, including multiple testing correction, positive/negative control validation, bootstrap false discovery estimation, and cross-validation against published international signals. Results: We identified 3,017 robust product-name-level signals detected by three or more methods, of which 2,146 were detected by all four methods. Kratom occupied 7 of the top 10 positions by composite risk score, reflecting the breadth of its multi-organ toxicity rather than independent signals, with Kratom–Death ranked highest (PRR 19.7, N = 178). Hepatotoxicity signals clustered in herbal/botanical and weight loss products. Herbal/botanical supplements carried the highest serious outcome rate (78.0%; adjusted OR 2.08, 95% CI 1.53–2.84). CUSUM detected 451 temporally emerging signals (14.9%), including preliminary hepatic enzyme signals for AG1 and Nutrafol (2023–2025). Of 3,017 robust signals, 97.3% survived false discovery rate correction (α = 0.05); 17 of 21 (81%) established international safety signals were recovered. A total of 148 signals were classified as critical tier by composite risk scoring. Conclusions: This computational pharmacovigilance analysis of dietary supplements identified over 3,000 robust product-name-level safety signals across multiple analytic dimensions; the true number of distinct ingredient-level safety concerns is lower due to product name fragmentation. The results support regulatory prioritisation of herbal/botanical and weight loss supplement categories and identify preliminary emerging signals warranting continued monitoring.
