药物警戒的未来：人工智能与药物安全系统的整合

Table of contentsIntroductionThe current regulatory landscapeThe next generation of PVThe IQVIA approachLooking aheadAbout IQVIAAbout the authors 3479101011 The use of artificial intelligence (AI) is rapidly proliferating across industries, andthe pharmaceutical space is no exception. Activating AI across every stage of thedrug lifecycle can yield major benefits in terms of speed, efficiency, compliance,and patient outcomes. However, the relative newness of AI’s widespread usegives many across the healthcare industry pause, particularly when it comesto ensuring compliance with a rapidly evolving regulatory landscape.For pharmacovigilance (PV) teams, AI may be bothappealing and daunting. On one hand, AI can bea major asset in terms of helping experts handleincreasing case volume; on the other, implementingAI in PV requires strict tracking and documentation toensure continued compliance with shifting guidelines.Regulators like the U.S. FDA have demonstrated anongoing effort to educate themselves and provideguidelines for leveraging AI compliant with regulations,which is reflected in several initiatives by the agency toprovide clarity and direction to stakeholders, includingthose in the PV space.Introduction iqvia.com | 3To successfully integrate AI into PV systems, anorganization must align its people, technology, andprocesses to build streamlined workflows supportedby a robust governance structure. If done successfully,these efforts will yield significant efficiencies acrossPV workflows.Activating AI across every stageof the drug lifecycle can yieldmajor benefits. 4 | The Future of Pharmacovigilance: Integrating AI with Drug Safety SystemsThe current regulatorylandscapeDespite a climate of uncertainty, the FDA and otherregulators around the world have been active interms of regulations and guidance around AI fornearly a decade (Figure 1). In 2016, the FDA publisheda paper that highlighted how it planned to reviewregulatory submissions with AI components. From2016 through 2018, the FDA conducted heavy researchon how to activate AI across different stages of thedrug development lifecycle and throughout lifecyclemanagement, particularly in the drug safety domain.And in the last four years, the FDA has organizedinternally and collaborated with other regulatoryagencies to put out discussion papers and guidance.Figure 1: Timeline of FDA responses to AIRegulatoryreviewsFDA reviewingregulatorysubmissions with AIcomponents2016SaMDaction planJan: FDA releaseof SaMD action planfor AI/ML-basedsoftware2021Joint guidingprinciplesOct: FDA publishes jointguiding principles for”Good Machine LearningPractices”2021 May: FDA issuesdiscussion paper ”Useof AI/ML in Development ofDrugs and Biologics” Regulatory guidanceJan: FDA publishesdraft guidance”Considerations forthe Use of ArtificialIntelligence to SupportRegulatory Decision-Making for Drug andBiological Products” Fast forward to today, and the FDA continues to explorehow drug manufacturers can leverage this technologyin development and manufacturing. In January 2025, theagency published the draft guidance “Considerations forthe Use of Artificial Intelligence to Support RegulatoryDecision-Making for Drug and Biological Products,”which aligns closely with the discussion paper theyreleased in 2023. The 2025 guidance focuses on AI usageacross all stages of the lifecycle for both small moleculedrugs and biologics, demonstrating the FDA’s willingnessto not only embrace AI but to guide sponsors through itsusage in a collaborative fashion. It specifically addresseshow to use AI to support regulatory decision-making,including in the context of PV, outlining a risk-basedapproach to define an AI model to support credibilityand accuracy. iqvia.com | 5The FDA’s Center for Drug Evaluation and Research(CDER) recently launched the Emerging Drug SafetyTechnology Program (EDSTP), which intends to:•Be a central point of discussion on the use of AI andother emerging technologies in PV.•Manage and transfer knowledge specific to the contextof AI use in PV.•Better understand the use of AI and other emergingtechnologies in PV to inform regulatory approaches.The FDA continues to explore howdrug manufacturers can leveragethis technology in developmentand manufacturing. Non-bindingDiscussions are non-binding,aimed at mutual learning ratherthan regulatory advice 6 | The Future of Pharmacovigilance: Integrating AI with Drug Safety SystemsOne of the key components of EDSTP is to providesponsors, upon request, with the opportunity to meetwith the FDA to discuss how AI is being used in theirprocesses and to gain regulatory insights into usingAI in the PV domain. The EDSTP meeting aims to createan opportunity for shared learning (Figure 2). Thesediscussions are non-binding and open to applicantswith at least one approved application regulated byCDER, as well as CROs and other relevant parties.Meeting requests are granted quarterly for up to nineparticipants in a 12-month period and submissions a