Cloud Hosting

Regulators across the globe have been preparing for the arrival of new artificial intelligence (AI) technologies and advances in real-world data (RWD) they say will become a part of regulatory science in the coming years.

That was the main topic of discussion at the 11th Global Summit on Regulatory Science annual conference, where regulators from Brazil, Canada, India, Italy, Japan, Germany, Switzerland, Singapore, the UK and the US presented the ways they are integrating AI and RWD into the operations and regulatory mechanisms of their agencies. The meeting was held virtually in October 2021 and sponsored by the Global Coalition for Regulatory Science Research (GCRSR).

The proceedings were recently summarized by Shraddha Thakkar, PhD, MSc, MS, of the Center for Drug Evaluations and Research (CDER) at the US Food and Drug Administration (FDA), and colleagues from regulatory agencies in the above countries in the journal Regulatory Toxicology and Pharmacology.The regulators discussed in a series of debates, workshops, and presentations how AI and RWD could be applied to food and drug safety assessments, whether regulatory science was prepared for the arrival of AI, how data science tools could better align to regulatory applications, and the future of regulatory science research.Continued progress in AI and RWD provide enormous opportunities for regulatory application with two significant aspects, improving the agencies operation and preparing regulatory mechanisms to review and approve products utilizing these innovations, according to the authors. This is especially important to drug development which usually spans many years and comes with a huge cost, where AI and RWD have demonstrated the ability to improve drug safety and review.The regulators noted that they see the potential for AI and RWD in food safety, pattern recognition, and foodborne outbreaks, which primarily relies on a manual analysis of images, spectrometric data, genomic data, chemical compositions, and identification of contaminants, the authors said. AI and machine learning (ML) have the potential to reduce review times and human variations in manual processes. In many ways, AI and RWD are already here, with agencies like the FDA and Canadian Food Inspection Agency incorporating AI and RWD methodologies into existing programs. AI and RWD can also serve as augmentation tools for existing information aids, such as in the case of Swissmedic considering using serious adverse drug reactions in hospital admissions as RWD to develop automated pharmacovigilance signal detection. Another example is crowdsourcing, which the National Institute of Health Science of Japan used to develop a quantitative structure-activity relationship model for Ames mutagenicity prediction.In two debates, presenters argued that the regulatory community may be prepared and/or unprepared for the advancement of AI and RWD in the domains of scientific knowledge and assessment practices. One presenter argued that AI plays an increasing role in drug discovery and development and that some regulators, like the FDA, are developing programs like the Innovative Science and Technology Approaches for New Drugs (ISTAND) initiative to prepare. Other considerations debated were the role of AI in clinical applications and the extent to which patients may be comfortable using AI-enabled applications in various contexts.Regulatory science could play a critical role in developing a regulatory structure and framework for evaluation of AI application, including promoting trustworthiness and reliability in these technologies, the authors wrote.A workshop where regulators detailed their data analytics tools was another opportunity for AI, Thakkar and colleagues noted, because it has the potential to automate manual reading processes for text associated with safety and efficacy of food and drug products. The vast majority of data used in regulatory decision-making are presented in text document, where AI could be of significance to facilitate the review process, they wrote. Globally, regulatory agencies have not only reviewed vast quantities of submitted application, papers, and/or literature data, but have also generated a plethora of documents during the product-review process. It is typical that these types of records are unstructured text and often do not follow the use of standard vocabulary.Due to lack of standardization and fragmentation of data, leveraging AI to interpret datasets is a substantial regulatory challenge, Thakkar and colleagues explained. The biggest challenge the research community faces is the current fragmentation of data in many repositories with multiple formats and definitions, they said. Another challenge is that, in some cases, the data codes are not uniform. Each data source has a coding system, and different ways of assigning codes to medicines are employed without national or international standardization.The future of regulatory science research in relation to AI and RWD is one where AI augments the work of human clinicians but does not replace them. One of the most significant benefits of AI/ML resides in its ability to learn from real-world use to improve its performance, the authors noted. However, as an emerging technology, AI should be constantly evaluated to actively facilitate the use of these new tools in regulatory settings, they said.Regul Toxicol Pharmacol

Go here to see the original:

Global regulators discuss how they are adapting to AI, real-world ... - Regulatory Focus