Artificial Intelligence and Machine Learning in Pharmacovigilance: Adverse Drug Reaction Detection and Signal Management under ICH-GCP Compliance

Pharmacovigilance (PV) is the science and activities relating to the detection, assessment, understanding, and prevention of adverse drug reactions (ADRs) and other drug-related problems. The exponential growth in global drug utilisation combined with the proliferation of electronic health records (EHRs), social media, and spontaneous reporting systems has generated unprecedented volumes of safety data that strain traditional manual review workflows. Regulatory frameworks such as the International Council for Harmonisation – Good Clinical Practice (ICH-GCP) E2A–E6 series, the European Medicines Agency (EMA) pharmacovigilance legislation, and the FDA Sentinel System mandate rigorous, timely, and reproducible signal management processes. This article systematically reviews the application of artificial intelligence (AI) and machine learning (ML) methodologies – including natural language processing (NLP), deep learning, graph neural networks, and Bayesian statistical methods – in ADR detection, signal management, and benefit–risk assessment, with emphasis on regulatory compliance under ICH-GCP. Methods: A structured literature search was conducted across MEDLINE, EMBASE, Cochrane Library, and WHO-VigiBase publication catalogues for the period 2010–2024. Seventy-eight peer-reviewed studies, regulatory guidance documents, and technical whitepapers meeting predefined inclusion criteria were analysed. Methodologies were categorised by AI/ML technique, data source, regulatory context, and performance metrics. AI/ML systems demonstrate superior performance compared with classical disproportionality analyses in ADR signal detection, with AUROC values ranging from 0.82 to 0.97 across validated datasets. NLP-based pipelines applied to EHR free-text achieved F1 scores of 0.78–0.91 for ADR entity recognition. Large language models (LLMs) show promise for automated narrative medical case summarisation and MedDRA coding. Implementation challenges include algorithmic transparency, data heterogeneity, and harmonisation with ICH-E2B(R3) electronic reporting standards. AI and ML are transforming pharmacovigilance by enabling earlier, more sensitive, and scalable ADR signal detection. Successful integration into GCP-compliant workflows requires regulatory-grade model validation, auditability, and explainability frameworks. Prospective collaboration among industry, regulators, and academia is essential to establish harmonised standards for AI-augmented pharmacovigilance.