New GAMP®Guide Provides Framework to Achieve High-Quality AI-Enabled Computerized Systems for Life Sciences

Extensions to existing guidance, new concepts and details are discussed in the Guide including:

• Risk management: Quality Risk Management (QRM) is key to maintaining patient safety, product quality, and data integrity. The Guide reflects specific risk management aspects of AI-enabled computerized systems and a framework for identifying risks and establishing suitable control strategies.
• Scalable life cycle activities: Promoting a risk-based approach and use of critical thinking, the Guide encourages scalable life cycle activities regarding all phases from concept to retirement and key supporting processes.
• Regulated company and supplier collaboration: Fostering effective collaboration between suppliers and regulated companies, guidance is provided on supplier assessment and monitoring specifically to the context of AI-enabled computerized systems.
• Roles and responsibilities: While established roles are augmented, new roles are also required. These roles and their associated responsibilities are considered, including activities of the quality assurance unit (QAU) safeguarding comprehensive decision-making and support in achieving compliance of AI-enabled computerized systems.
• Data and model governance and management: Providing concepts like fit-for-purpose data, considerations on data quality and an extended view on data governance, the Guide helps to build robust data foundations for AI-enabled computerized systems. Furthermore, model governance strategies promote traceability of decisions during development and use of models in AI-enabled computerized systems.
• Ideation, design, and proof-of-concept: To support the decision-making process regarding the setup of AI projects, the Guide elaborates on concept phase aspects including critical elements of a proof-of-concept serving as a foundation for transition to the project phase.
• Model development: Guidance on typical development steps, relevant risks, decision-making, and relevant information and records throughout the model development life cycle are provided.
• Testing of models and AI-enabled computerized systems: Various testing approaches, leveraging the concept of fit-for-purpose-data and key performance indicators, as well as human assessment of usability and suitability, and dedicated considerations like testing for robustness and fairness are included.
• Ongoing monitoring: The collection and use of live data for ongoing monitoring are considered as a basis for ensuring control and facilitating continual improvement.
• Change management: The Guide provides a basis for comprehensive decisions on change management processes, considering both business and quality related drivers.
• Incident and Corrective Action and Preventive Action (CAPA) management: Typical incident scenarios are described, including approaches for remediation as part of incident management processes; further considerations on CAPA management relevant to AI-enabled computerized systems are provided.
• Knowledge management: Guidance on how to integrate new specialized expertise critical for successful implementation of AI in a complex environment, concerning established roles like data and system owners or new roles like data scientists and machine learning engineers. The Guide also reflects on the role of AI literacy and provides guidance on achieving such competencies across the organization.
• Trustworthy AI: Guidance on trustworthy AI is provided, including aspects such as transparency, human oversight, and control and mitigation of bias, thus facilitating responsible and ethical use of AI.
• Explainable AI: Guidance is provided on how to effectively support the human-AI-team in a GxP regulated process.
• Dynamic systems: Expanding on ISPE GAMP® 5: A Risk-Based Approach to Compliant GxP Computerized Systems (Second Edition), “Appendix D11” notion of dynamic systems, guidance is included regarding dynamic systems throughout concept, project, and operational phases.
• Cybersecurity: Particularities of cyber security are covered, including adversarial attacks on data and models.
• AI as and in medical device: The Guide embeds typical medical device processes like design reviews and device-specific risk management practices to general considerations in achieving safe and effective AI-enabled computerized systems.

Figure 1: Overview of New and Extended Concepts Developed in the AI Guide

Conclusion and Summary

Producing the ISPE GAMP^® Guide: Artificial Intelligence for industry has been an enlightening journey. It involved the integration of existing concepts and establishing a comprehensive understanding of subject matter by close collaboration between various disciplines. It not only highlights the complexity of AI-enabled computerized systems but also underscores the importance of interdisciplinary teamwork.

The AI Guide team looks forward to witnessing the value that the ISPE GAMP® Guide: Artificial Intelligence will bring to industry in implementing AI-enabled computerized systems. Looking forward, the team believes that the GAMP® Guide: Artificial Intelligence is only the starting point of a longer journey. While technological advancements may accelerate, the industry faces many challenges across GxP areas: The amount of data will grow, the complexity of processes is likely to rise, and the vision of personalized medicine is getting closer. The team believes that AI can offer means to address those challenges and unlock new horizons in life sciences, if AI is understood as an integral part of a set of digital approaches; not a solves-it-all tool.

All interested stakeholders are invited to join this journey—collaborating, co-creating, and co-inventing the future of life sciences. The ISPE AI Community of Practice (CoP) or the GAMP® CoP Software Automation and AI Special Interest Group (SIG) are open to all ISPE members.

Learn more and get the guide

Acknowledgements

The ISPE GAMP^® Guide: Artificial Intelligence has been produced by an international team of volunteers, under the leadership and direction of Brandi M. Stockton (lead), Eric Staib (co-lead), and Martin Heitmann (co-lead). Stockton was the originator and driving force of the AI Guide initiative, whose leadership and vision laid the foundation for its development.

The Guide was sponsored and supported by the ISPE GAMP CoP Software Automation and AI SIG.

The following individuals contributed to the Guide, representing a diverse set of organizations within life sciences; individuals marked with an asterisk (*) served as a lead for one or more chapters: Rolf Blumenthal, Taylor Chartier*, Aude Chetwynd, Joanne Donald*, Stephen Ferrell*, Gail Francis, Carsten Jasper*, Stuart Jones, Ian Lucas, Michael Martone*, Stefan Münch*, Meher Muttanapalli*, Peyton Myers, Raj Nandhan*, Tatum O'Kennedy, Mohammed Arif Rahman*, Rick Rambo, Laila Rasmy*, Doug Shaw, Tomos Gwyn Williams, PhD*, Kathy Zielinski.

The AI Guide Leadership Team acknowledges the efforts and contributions of the many individuals and companies from around the world who reviewed and provided comments during the preparation of the Guide. The team also thanks the ISPE Global Documents Committee and GAMP Editorial Review Board for their valuable comments on this Guide.

Special thanks go out to Mark Newton for his mentorship throughout the development of the Guide and providing his insights from the Global Documents Committee perspective, and Siôn Wyn for his guidance during the proposal stage of this Guide.