The use of artificial intelligence (AI) lures with the prospect of new, potentially revolutionary business models and innovative approaches to make processes faster and more efficient. Especially in a strictly regulated GxP environment, however, the risks of AI applications must also be taken into account.
Thescon therefore organized a seminar on artificial intelligence in the pharmaceutical industry together with PTS Training Service. The aim of this seminar was to demonstrate the possibilities and limitations of AI applications in a regulated environment and to discuss them with experts and representatives from the medical device and pharmaceutical industries.
Prof. Dr. Andreas Lischka from the Ingenium Institut für Unternehmensentwicklung und innovative Medien GmbH defined what constitutes an AI and presented possible future scenarios for its use. He predicted that within the next two years about half of all standard administrative tasks will be performed by AI applications. He considers the right attitude, the 'Digital Mindset', to be an important prerequisite for the successful application of AI in a company, especially at management level. It is not important how managers find Facebook or WhatsApp; it is more important that they know the digital world and what they can achieve with it. Prof. Lischka argued that hierarchies should be abolished, as they are inherently demotivating, and recommended leaving one's comfort zone to evolve digitally.
Thescon partner Ingo Baumann provided insights into the challenges of validating AI applications in his presentation. While the environment surrounding an AI application can be validated or qualified according to the usual procedures, he believes that AI applications push us to the limits of what we can validate. The hurdles for validating AI applications are (still) numerous: On the one hand, no dedicated regulatory requirements for AI applications exist yet. Furthermore, the requirements established for conventional programs are not necessarily suitable for AI programs, and data and its management are far more important for AIs. Repeated testing, as it is applied in conventional CSV, is more damaging than it helps, because it might make the AI application an 'expert idiot' (“overfitting”). In order to still be able to use AI applications in a regulated environment, he recommended that they be addressed not only within a company, but also in dialogue with suppliers, customers and authorities. AI-specific aspects must be taken into account in the specifications for the development and control of AI models. In his opinion, a clearly defined intended use, a comprehensive risk analysis and special requirements for data management are indispensable foundations for the use of AIs in regulated areas.
Based on this, Dr. Jörg Stüben of Boehringer Ingelheim International GmbH presented numerous application examples for AIs in non-critical, non-GxP-relevant areas. Using these examples, he showed that artificial intelligence can already be successfully used throughout the drug development chain, and which prerequisites must be met to ensure that these applications are not under obligation to be validated. As long as a human control authority checks and approves the results of an AI application and the AI application cannot make any decisions on its own, it is not necessarily subject to validation in the GxP area. Thus, AI models can be used, for example, in drug development to characterise and compare molecules, in the optimisation of clinical studies to predict the reliability of test subjects, in document management for the automatic analysis and indexing of documents, or in drug safety for the analysis of social media to detect side effects. Dr. Stüben predicted an increase in AI applications also in the GxP area, and emphasized the importance of including data scientists as experts in such projects.
The event was accompanied by workshops and lively discussions where participants could share their experiences and difficulties with the use of AI applications in the regulated area and where the prerequisites for a successful integration of Artificial Intelligence in companies were discussed.
Do you have questions about artificial intelligence in a regulated environment? Please do not hesitate to contact us!