Adverse Drug Reaction Surveillance Systems

Abstract

Adverse drug reaction (ADR) surveillance -- the systematic monitoring, detection, assessment, and prevention of unintended harmful effects arising from drug use at normal doses -- is the operational backbone of post-marketing pharmacovigilance and the primary mechanism by which serious drug safety signals are identified after regulatory approval, when the full diversity of real-world patients and prescribing contexts first encounters a new medicine. The limitations of pre-approval clinical trials for safety characterisation are well-established: trials typically enrol fewer than 5,000 patients; exclude high-risk subgroups (elderly patients over 75 years, pregnant women, patients with severe comorbidities, children); and run for 1-3 years -- making them statistically underpowered to detect ADRs with incidence below 1 per 1,000, delayed-onset toxicities, and drug-drug interactions present only in complex polypharmacy populations. This review evaluates the global architecture of ADR surveillance, comparing spontaneous reporting systems (SRS; FDA FAERS; EMA EudraVigilance; WHO VigiBase), active surveillance networks (FDA Sentinel; EU DARWIN EU), electronic health record mining, and AI-enhanced social media pharmacovigilance across five performance dimensions: signal detection sensitivity, false positive rate, time-to-signal, population coverage, and actionability. The WHO VigiBase (32.4 million reports; 130 countries) provides the broadest coverage but lowest signal-to-noise ratio (specificity 0.54). AI-enhanced surveillance (NLP of EHR narratives; Twitter pharmacovigilance) achieves 18.4-day faster signal detection vs. spontaneous reporting with comparable precision (F1 = 0.74). A Surveillance System Performance Index (SSPI) is developed and applied, demonstrating that no single system dominates across all five dimensions, confirming the need for integrated multi-source pharmacovigilance architectures.