This article, “The applications and advances of artificial intelligence in drug regulation: A global perspective“, offers an in-depth review of the current status of Artificial Intelligence (AI) integration within drug regulation worldwide. Authored by Lixia Fu, Guoshu Jia, Zhenming Liu, Xiaocong Pang, and Yimin Cui, and published in Acta Pharmaceutica Sinica B in 2025, the paper highlights that while AI has significantly impacted healthcare, its adoption in drug regulation is still in its early stages.
The review aims to provide a comprehensive overview of AI-related policies, initiatives, and practical applications within regulatory agencies globally, while also discussing the challenges and future prospects of AI in this field.
I. The Landscape of AI in Drug Regulation
- AI’s Transformative Role in Healthcare: AI, encompassing techniques like machine learning (ML) and deep learning (DL), has become a transformative force in various healthcare domains, including disease diagnosis, treatment strategy development, and assisting clinical decision-making. It is particularly adept at analyzing “big data” generated in biomedical research.
- The Need for AI in Drug Regulation:
- Drug regulation is a complex and critical component of healthcare systems, responsible for ensuring medication safety, efficacy, and quality.
- Drug research and development (R&D) is a time-consuming, costly, and high-risk process with declining efficiency. Regulatory approval alone can span months to years due to significant manual and repetitive labor in reviewing submissions.
- The emergence of new technologies (e.g., omics, precision medicine) and an increase in drug regulatory submissions present considerable challenges for agencies.
- AI offers a promising solution to streamline regulatory processes, enhance data-driven decision-making, and adapt to these challenges, thereby forming a more scientific and efficient regulatory system.
II. Policies and Initiatives by Major Regulatory Bodies
Many regulatory agencies globally are proactively exploring and implementing AI technologies:
- U.S. Food and Drug Administration (FDA):
- Facing increasing volumes of drug applications and shorter review timelines, the FDA has launched several action plans for modernization.
- Sentinel Initiative (2008): Established an electronic system for postmarket safety surveillance, complemented by the Biologics Effectiveness and Safety System (BEST) which leverages electronic health records (EHRs) and Natural Language Processing (NLP) for automated adverse event (AE) reporting.
- Technology Modernization Action Plan (TMAP) (2019): Aims to update computer hardware, software, data, and analytics, integrating AI to facilitate data usage in regulatory decision-making.
- Innovative Science and Technology Approaches for New Drugs (ISTAND) (2020): Provides a pathway for novel approaches, including AI/ML, to replace or reduce animal testing, assess patients, and assist in study design.
- Data Modernization Action Plan (DMAP) (2021): Clarifies strategies for updated methods, IT, and data usage processes, promoting agency-wide transformation through predictive models and AI.
- Enterprise Modernization Action Plan (EMAP) (2022): Focuses on improving operational efficiency and data utilization by optimizing business processes.
- Cybersecurity Modernization Action Plan (CMAP) (2022): Utilizes AI and ML to enhance network detection and response capabilities against cyber threats and risks to sensitive data.
- FDA Information Technology Strategy (2023): Emphasizes data-driven decision-making using AI and identifying related opportunities and risks.
- European Medicines Agency (EMA) and Other European Agencies:
- The EMA emphasizes AI for advancing regulatory science, improving efficiency, and promoting scientific decision-making.
- Collaborative Groups: The European Medicines Regulatory Network (EMRN), International Coalition of Medicines Regulatory Authorities (ICMRA) (through its Informal Network for Innovation), Big Data Steering Group (BDSG), and Analytics Centre of Excellence (ACE) were established to promote AI in regulation.
- Regulatory Science to 2025-Strategic reflection (2020): Highlighted the transformative impact of big data and AI on regulatory decision-making.
- EMA Network Strategy to 2025 (2020): Aimed to enhance dynamic regulation and automation through AI and digital technologies.
- Multi-annual AI workplan 2023e2028 (2023): Seeks to establish an AI-based regulatory system to enhance data insights and public health decision-making.
- Big Data Workplan 2023e2025 (2024): Proposed enhancing analytical capabilities based on real-world data (RWD) and AI, including establishing a network of analytics centers and validating AI algorithms.
- National Initiatives: Germany’s Federal Institute for Drugs and Medical Devices (BfArM) is building AI computing infrastructure for review, licensing, and research (e.g., using AI for drug epidemiology and text mining for signal monitoring). Switzerland’s Swissmedic launched “Swissmedic 4.0” to apply AI for processing applications and preparing assessment reports, delegating repetitive tasks to machines.
- China National Medical Products Administration (NMPA):
- China recognizes the urgent need to leverage IT to enhance foresight, targeting, and timeliness of drug regulation.
- Chinese Drug Regulatory Science Action Plan (2019): Aims to introduce new systems, tools, standards, and methods to accelerate the modernization of drug governance.
- Action Plan for Accelerating Smart Drug Regulation (2019): Promotes the application of big data, AI, and blockchain to expedite the regulatory system’s transformation and upgrade.
- “14th Five-Year Development Plan for the Pharmaceutical Industry” (2021) and “14th Five-Year National Drug Safety and High-Quality Development Plan” (2021): Encourage the exploration of AI, cloud computing, and big data in pharmaceutical R&D for efficiency, data sharing, and early risk warning.
- “14th Five-Year Plan for Drug Regulatory Cybersecurity and Informatization Construction” (2022): Focuses on strengthening risk management capacity and applying AI to improve the national adverse drug reactions (ADRs) monitoring system.
- International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH):
- The ICH has emphasized AI in various policy documents, including its Reflection Paper on Patient-Focused Drug Development (PFDD) (2021), which mentioned AI/ML and wearable devices for data collection.
- The “Future Opportunities & Modernization of ICH Quality Guidelines” (2021) highlighted that advancements in new therapeutic approaches and AI will increase data volume and complexity, proposing improved knowledge management through data clouds and structured data formats.
- AI/ML are identified as future focus areas for ICH, suggesting their use to improve understanding of products and for areas like ADR monitoring (ICH E2B(R3)) and improving CTD document quality (ICH M4Q).
- Pharmaceuticals and Medical Devices Agency (PMDA) in Japan:
- Unlike other agencies, PMDA currently lacks a unified AI reform plan but encourages independent research projects through grants.
- The Ministry of Health, Labour and Welfare (MHLW) funds research topics related to AI and drug regulation. While AI/ML is used in clinical applications, there’s a current lack of mature cases for AI in drug regulatory systems.
III. Practical Applications of AI in Drug Regulation
Regulatory agencies are exploring AI adoption in several aspects:
- Safety Surveillance (Pharmacovigilance – PV):
- Drug regulatory agencies face the immense challenge of monitoring drug safety, with, for example, the US FDA receiving over 2 million postmarket reports annually.
- FDA: The Center for Drug Evaluation and Research (CDER) uses AI to process and evaluate Individual Case Safety Reports (ICSRs) in the FAERS, including ML models for automated AE report classification. The Sentinel System utilizes NLP and ML to extract features from unstructured EHRs for postmarket safety analysis. The Information Visualization Platform (InfoViP) combines AI/ML (NLP) and visualization to categorize ICSRs, detect duplicates, and assist in analyzing reported AEs.
- International Efforts: The ICMRA identifies AI for PV data processing. Swissmedic developed LiSA, an AI-based literature search engine to identify ADRs. Sweden’s Medical Products Agency (SMPA) uses NLP triage models in the PhaVAI project to identify serious AEs. The WHO Collaborating Centre developed WHODrug Koda, an AI-based drug coding engine, to enhance automated coding of drugs in AE reports.
- Workflow Optimization:
- FDA:
- Knowledge-aided Assessment and Structured Application (KASA) tool (2018): Developed by the Office of Pharmaceutical Quality (OPQ) to address increasing Abbreviated New Drug Applications (ANDAs), KASA significantly improves review efficiency, objectivity, and consistency by using structured assessment templates and risk-level-based algorithms. It has been extended to new drug and biologic applications.
- Computerized Labeling Assessment Tool (CLAT) (2020): Employs AI to automatically review labels.
- An AI-based component is being developed for the FDALabel database to allow customized querying of labeling documents for drug repurposing studies.
- RxBERT: An AI-based language model optimized for analyzing U.S. drug labeling documents to support safety reviews.
- EMA:
- The HMA-EMA BDSG suggests AI for internal regulatory processes, such as using NLP to process text, categorize eCTD submissions, and perform quantitative review of image data.
- AI activities include text analysis, document processing, and analysis of text and images leveraging NLP and Optical Character Recognition (OCR) to extract, classify, and cluster content, automating document categorization and reviewer assignment.
- The EMRN plans to roll out an AI-enabled tool called “Scientific Explorer” in 2024 to facilitate searches within regulatory documents.
- The ACE has implemented pilot projects, including a tool to automate application registration and a speech-to-text chatbot “ASK-EMA”.
- FDA:
- Regulatory Science Research:
- FDA: Sponsors projects using AI and RWD to study health disparities, predictive toxicological models (e.g., drug placental permeability using 3D printing and ML), AI-driven cancer risk prediction, and AI models simulating pregnant women to forecast drug and vaccine impacts.
- National Center for Toxicological Research (NCTR) at FDA: Launched the AI4TOX program to develop new AI tools for regulatory science and product safety review, including AnimalGAN (predicting animal toxicology data), SafetAI (deep learning for toxicological endpoints), BERTox (AI-powered NLP for document analysis), and PathologAI (histopathological data analysis). They also developed DeepDILI for predicting drug-induced liver injury.
- National Institute of Health Sciences (NIHS) in Japan: Established a chemical mutagenicity prediction model using Quantitative Structure-Activity Relationship (QSAR) and AI.
- International Exchange:
- FDA officials frequently present insights on AI-supported regulation at scientific forums, emphasizing AI/ML’s potential for postmarket data collection, understanding drug safety/efficacy, modernizing regulation, and fostering global collaboration.
- The NCTR’s “AI4PharmcoVig” research employs AI models for document screening and classification in pharmacovigilance.
- A joint HMA/EMA workshop on AI in medicine regulation facilitated sharing research progress and practical applications of AI in drug regulation, including innovative pharmacovigilance methods utilizing statistical and predictive models, ML, and NLP.
IV. Challenges of AI Applications in Drug Regulation
Despite its vast potential, AI application in drug regulation faces unique and complex challenges:
- Data Quality and Reliability:
- Drug development generates massive amounts of diverse data from various sources, leading to issues like uneven quality, fragmentation, and multiple formats.
- Successful AI algorithms require a large dataset of high-quality, semantically structured data for training.
- Much of the data submitted to regulatory agencies is in the form of unstructured text documents, posing a challenge for AI processing.
- Data can be company-sensitive, hindering sharing even within regulatory bodies. Regulatory agencies need to prioritize preparing high-quality and accurate, sufficiently balanced data.
- Technical Limitations:
- The wide variety of AI algorithms (ML, DL, NLP) presents a challenge in selecting the appropriate algorithm for regulatory applications.
- Adaptability of models when retrained on unseen data is crucial.
- “Black Box” Issue: Many ML algorithms produce models with low interpretability, making it difficult for humans to understand the reasons behind AI decisions. Resolving this issue is critical for reducing bias, improving interpretability, and validating model credibility.
- AI can introduce bias and prejudice through iterative updating. Thorough verification of reliability, repeatability, interpretability, and traceability of models is essential.
- Talent Shortage:
- There is a significant shortage of high-end interdisciplinary talent capable of bridging data science, computer science, and pharmaceuticals.
- Professionals in pharmacovigilance, for instance, need to be trained in entirely new skill sets.
- Given AI’s nascent stage in drug regulation, there’s an urgent need to establish stable interdisciplinary teams.
- Lack of Standards:
- Currently, there are no established standards for AI-related data exchange or for the application of AI in regulatory activities. This hinders communication and the full utilization of AI’s potential.
- The Danish Medicines Agency (DMA) proposed key considerations for static AI/ML algorithms in GxP (Good Practice) scenarios, focusing on datasets, bias, metrics, and result interpretation.
- International associations are encouraged to issue official validation guidelines or establish corresponding gold standards, potentially referencing existing guidelines like GAMP (Good Automated Manufacturing Practice).
AI has played a critical role in optimizing regulatory workflows, improving efficiency and accuracy, reducing workload, and accelerating patient access to therapeutic products, ultimately benefiting public health.
While challenges persist, the prospects for AI in drug regulation remain broad, particularly as technology advances and policies are refined. Future applications include:
- Leveraging AI for knowledge management and providing relevant consultations to patients, healthcare professionals, and drug developers.
- Utilizing AI to conduct regulatory scientific research, such as assessing drug safety in specific populations, establishing regulatory standards for new technology treatment products, and analyzing RWD.
To fully realize AI’s potential, concerted efforts and close collaboration from academia, industry, and regulatory authorities are essential. Strengthening international cooperation, exchange, and refining pertinent regulations and policies will promote the robust development of AI technology in this domain, thereby enhancing the effectiveness of drug regulation.
Reference: Fu, L., Jia, G., Liu, Z., Pang, X., & Cui, Y. (2025). The applications and advances of artificial intelligence in drug regulation: A global perspective. Acta Pharmaceutica Sinica B, 15(1), 1–14. https://doi.org/10.1016/j.apsb.2024.11.006