Publication date: November 21, 2024
In recent years, artificial intelligence (AI) and machine learning (ML) technology have begun to play an increasingly important role in the pharmaceutical industry. AI offers the potential to revolutionize the way we produce medicines, leading to significant improvements in quality, efficiency, and innovation. However, introducing modern technologies into the regulated pharmaceutical environment is associated with challenges related to ensuring quality and compliance with applicable regulations. It is in this context that GAMP5 (Good Automated Manufacturing Practice) is one of the most important tools for managing the life cycle of computer systems in the pharmaceutical or biopharmaceutical industry, including systems based on AI.
GAMP5 is a set of guidelines developed by the International Society for Pharmaceutical Engineering (ISPE) to assist pharmaceutical companies in ensuring compliance with regulations concerning the quality and security of computer systems. This document presents a comprehensive approach to managing the life cycle of computer systems[1], including validation, design, implementation, operation and retirement of systems that are used in environments requiring high quality standards, such as the pharmaceutical industry. GAMP5 places a strong emphasis on quality and compliance with Good Manufacturing Practices (GMP) and other industry regulations. It includes both a risk-based approach and detailed guidance on the validation of computer systems, including systems controlling manufacturing processes.
Artificial intelligence and machine learning offer wide possibilities in the pharmaceutical industry, especially in the area of data analysis, optimization of production processes and improvement of efficiency and quality of manufactured drugs. This allows the use of large data sets (e.g. from production, quality control, clinical trial results) to create predictive models that can predict results, identify irregularities in the production process, or optimize the efficiency of drug production. The use of AI in drug production includes the analysis of clinical data, where artificial intelligence supports the processing of large data sets from clinical trials, helping to detect patterns that may be difficult for humans to notice. In addition, AI/ML technologies enable the optimization of production processes, allowing for real-time analysis of processes and identification of areas for improvement, which contributes to increased efficiency. Artificial intelligence also supports the automation of drug production.
Although AI/ML offer enormous benefits, their introduction into the pharmaceutical environment is associated with the need to ensure full compliance with regulatory regulations and maintain high quality standards. Therefore, the application of AI in drug production requires adherence to rigorous rules, such as those contained in GAMP5.
In the Polish legal system, a documented program that provides a high degree of certainty that a specific process (method or system) will repeatedly lead to results that meet specific acceptance criteria[2] is called validation. This concept also applies to activities related to an IT system or software, and its legal regulations are described, among others, in the Regulation of the Minister of Health of March 13, 2015 on the requirements of Good Manufacturing Practice. According to these regulations, “before starting to use a computerized system, it must be demonstrated, in the form of appropriate validation or verification studies, that this system is capable of achieving the desired results in an accurate, consistent and repeatable manner.”[3] Moreover, the second point of the same regulation tells us about the obligation of “the entrepreneur to prepare (in writing) a detailed description of the system, specifying: methods of procedure, objectives, security measures, the scope of the system and its main features, and the manner of using it and its interactions with other systems, [which is to be] updated on an ongoing basis.” As can be seen, Polish regulations do not directly mention GAMP5, therefore it can be assumed that GAMP is neither mandatory nor legally binding, however it remains an industry standard for validating automated systems used by people who create software used, among others, in the pharmaceutical industry, because it allows for entering the market with greater efficiency and lower risk[4].
Moreover, GAMP5, especially in the new version with the addition D11, discusses in detail the issues related to the integration of AI/ML in regulated environments. This document covers not only the requirements for validation and testing, but also the management of risks related to the quality of data and algorithms used in AI/ML systems. This is crucial to ensure that these technologies provide reliable results that do not introduce irregularities into production processes.
GAMP5 presents a detailed model of the computer systems life cycle, which includes three main phases: concept, design, and operation. Each of these phases requires separate activities related to data management and algorithms, especially in the context of AI/ML systems.
One of the key aspects of implementing AI/ML systems in drug production is ensuring data integrity. GAMP5 places great emphasis on ensuring that the information used to train algorithms is reliable, complete and consistent, and that any change in the data is monitored and validated. Information preparation includes its selection, classification, cleaning and enrichment, which ensures the credibility of the results of AI/ML models. Additionally, the validation of AI/ML systems is based on the GAMP5 principles, which require demonstrating that these systems meet their intended goals in a predictable and repeatable manner. Also, in the context of artificial intelligence, a key challenge is managing the risks associated with the quality of data and algorithms. AI algorithms learn from data, and irregularities in this information can lead to erroneous results, which in the case of drug production can have serious consequences. Therefore, GAMP5 requires that all information is properly controlled and managed throughout the entire life cycle of the system, and the risks associated with its quality are regularly assessed.
In summary, AI and machine learning technology open up new possibilities in the pharmaceutical industry, from optimizing production processes to analyzing clinical data. However, the introduction of these modern tools requires strict adherence to regulations and guidelines to ensure the safety and quality of the manufactured drugs. GAMP5, as a recognized industry standard, provides a framework for managing the life cycle of computer systems that can be effectively applied also in the context of AI/ML systems.
Although GAMP5 is not legally binding, its implementation is a key element in increasing trust in automated systems in the regulated pharmaceutical environment, enabling compliance with GMP requirements and national and international regulations. Through an approach based on risk management, data integrity and system lifecycle, GAMP5 supports the safe and effective use of AI in pharmaceutical manufacturing, contributing to improved quality, innovation and regulatory compliance.
Sources:
[1] Within the meaning of the Regulation of the Minister of Health of 9 November 2015 on the requirements of Good Practice, the life cycle of systems is all stages of the functioning of the system from initial requirements to withdrawal, including design, specification, programming, testing, installation, operation and maintenance.
[2]Legal definition of validation taken from the regulation of the Minister of Health of 9 November 2015 on the requirements of Good Manufacturing Practice.
[3]Text taken from the ANNEX OF GOOD DISTRIBUTION PRACTICE REQUIREMENTS FOR ENTREPRENEURS CONDUCTING WHOLESALE TRADE IN MEDICINAL PRODUCTS, EXCLUDING VETERINARY MEDICINAL PRODUCTS, AND INTERMEDIARIES IN TRADE IN MEDICINAL PRODUCTS
[4] https://www.tricentis.com/learn/compliance-with-gamp-5-guidance-a-checklist