Role of Generative Artificial Intelligence in Pharmacovigilance and Post Marketing Drug Safety

Abstract

Pharmacovigilance (PV) plays an important role in identifying, assessing and preventing adverse drug reactions during the entire lifecycle of medicinal products. Traditional PV systems mainly depend on manual data review and rule-based methods. This often become challenging when dealing with large amounts of unstructured clinical and biomedical data. In recent years, Generative Artificial Intelligence (GenAI), particularly Large Language Models (LLMs), has shown strong potential to improve the efficiency and accuracy of drug safety monitoring.   GenAI can assist in several PV activities such as preparation of Individual Case Safety Report (ICSR) narratives, extraction of adverse event information, screening of biomedical literature, and drafting regulatory documents including Periodic Safety Update Reports (PSURs), Development Safety Update Reports (DSURs), and Risk Management Plans (RMPs). Domain specific models like BioGPT and Med-PaLM have demonstrated better understanding of biomedical terminology and context. Additionally, approaches such as retrieval augmented generation help improve reliability and support early identification of potential safety signals. However, challenges such as data bias, hallucination risk, limited explainability, and regulatory concerns still need careful attention. Therefore, combining GenAI with expert human evaluation is essential to ensure accuracy, patient safety, and regulatory compliance while improving overall pharmacovigilance efficiency.