The Future is the Present: Artificial Intelligence in Pharmaceutical Manufacturing

Hollywood has long presented artificial intelligence (AI) as a futuristic concept far from reality, but technology has now eliminated the gap between science fiction and science fact. AI is here and here to stay; 50% of global healthcare companies plan to implement AI strategies by 2025 (1). For pharmaceutical manufacturers, AI has the potential to revolutionize process design and control, and thus bring benefits to patients and challenges to regulators.

Owing to its fictionalized past, the term AI can mean wildly different things to different people. FDA’s Center for Drug Evaluation and Research (CDER) describes AI rather inclusively as “a branch of computer science, statistics, and engineering that uses algorithms or models that exhibit behaviors such as learning, making decisions, and making predictions” (2). The National Academies of Sciences, Engineering, and Medicine issued a report on innovations in pharmaceutical manufacturing that highlighted AI’s potential role in the measurement, modeling, and control used for pharmaceutical manufacturing (3).

AI offers potential benefits to pharmaceutical manufacturers in the form of optimized process design and control, and smart monitoring and maintenance, to drive continuous improvement. AI in concert with other innovative technologies might advance pharmaceutical quality, build more resilient supply chains, and improve the availability of medicine for patients (4). In anticipation of these benefits, FDA is proactively preparing for the arrival of AI technologies in pharmaceutical manufacturing.

The AI revolution

AI is an enabler of Industry 4.0, the fourth industrial revolution characterized by integrated, autonomous, and self-organizing production systems (5). The Industry 4.0 paradigm brings the hope of a well-controlled, hyper-connected, digitized ecosystem and supply chain. The COVID-19 public health emergency seems to have accelerated the development of Industry 4.0 technologies, such as AI, that might respond to rapidly changing supply and demand and reduce dependence on human intervention.

So how might manufacturers deploy AI in process design and control?

Process and scale-up optimization. AI models employing machine learning might use process development data to more quickly design and identify optimal process parameters or scale-up strategies, reducing development time and waste.

Process control. Advanced process control might allow for dynamic control of a manufacturing process which, in combination with real-time sensor data, might be used to develop process controls that can precisely predict the trajectory of a process. Pharmaceutical manufacturers are expecting to adopt advanced process control approaches that combine AI techniques with chemistry and physics knowledge to improve manufacturing efficiency and output.

Process monitoring and fault detection. AI methods might be used to better monitor equipment and detect changes from optimal performance. AI-detected deviations might trigger maintenance activities in a manner that minimizes process downtime.

Trend monitoring. AI methods integrated with process performance metrics might offer better trend monitoring, even across products or locations, and consequently allow for proactive corrective and preventive actions to address manufacturing discrepancies before they impact the supply chain, or worse, cause drug shortage.

The potential use of AI to monitor quality doesn’t even end at the process. AI might also be used to monitor a product’s quality after manufacturing, for example, for the integrity of its packaging or presence of particulates. A vision-based quality control system might use AI to analyze images of packaging, labels, or glass vials to detect deviations (6). Beyond the product itself, AI might be used to identify cluster problems from consumer complaints and deviation reports that might contain large volumes of unstructured text. Though the potential applications of AI in pharmaceutical manufacturing will continue expanding, whether a manufacturer employs AI or not, high-quality drugs should be consistently available to patients.

Regulating AI

CDER established a Framework for Regulatory Advanced Manufacturing Evaluation (FRAME) initiative to prepare a regulatory framework to support the adoption of advanced manufacturing technologies that could benefit patients, such as AI. FDA recognizes potential benefits, and risks, of AI throughout the drug product lifecycle. As some manufacturers look to employ AI in their manufacturing processes, the regulatory framework will need to enable the timely adoption of these technologies while also keeping patients safe. CDER recognizes the need to better understand how this new manufacturing paradigm could impact pharmaceutical operations and regulation. The regulatory strategies of the past might not work as effectively in an AI-enabled Industry 4.0. However, regulators do not develop and implement manufacturing technologies. There is a need for the stakeholders developing these technologies to provide FDA a better understanding of where and how AI innovations might be used in drug manufacturing. Industry and regulators will not be able to address the issues related to AI by working separately; true innovation requires science-based collaboration.

In March 2023, CDER released a discussion paper to solicit public and stakeholder input on AI in drug manufacturing to better identify areas of policy consideration for AI technologies (2). This discussion paper proposed areas of consideration based on CDER’s evaluation of the existing regulatory framework: standards for developing and validating AI models, clarification on regulatory oversight for AI in pharmaceutical manufacturing, maintenance of cloud applications and continuously learning AI systems that adapt to real-time data, and data management practices commensurate with the volume of data generated.

FDA’s public workshop on the use of AI in drug Manufacturing on September 26–27, 2023, in conjunction with the Product Quality Research Institute, was designed to provide another opportunity for stakeholders to discuss key topics with regulators, such as:

• Artificial Intelligence in Process Development, Process Monitoring, and Commercial Batch Trend Monitoring
• Artificial Intelligence and the Use of Big Data and Data Management within the Pharmaceutical Quality System
• Lifecycle Approaches to Management of Artificial Intelligence.

Discussions on topics like these will help further inform the evaluation, development, and implementation of a regulatory framework that considers the benefits and risks of AI.

The future and AI

Like industry, regulators can employ AI to improve processes. In fact, FDA currently uses AI for translating documents, screening adverse event reports, and forecasting the volume of incoming regulatory submissions. In the future, both manufacturers and regulators might benefit from the extensive data and analysis provided by the expanded use of AI. International regulators have projected using AI to detect false or misleading drug information, scan scientific literature, identify safety signals, and respond to public inquiries (7).

While there are potential benefits of AI, there are also risks. Access to high-quality data is a fundamental requirement for effective AI training or learning. AI can be particularly sensitive to the characteristics of the data used for training, testing, and validation. The process analytical technologies providing data to AI systems must be accurate and representative. For learning purposes, data must represent not only process successes but also process failures. It will be critical to ensure that data used for AI training or learning are fit for use based on quality, reliability, and representativeness.

Humans, and not machines, are ultimately responsible for assuring that high-quality drugs are available to patients. Humans must be able to interpret the information generated by AI enough to ensure, for example, adherence to current good manufacturing practice requirements. Even with the best intentions, there are risks for unintended consequences. Might some even use AI for unethical outcomes?

Still, continued technology advancement is undeniable. There’s already talk of an Industry 5.0 paradigm which focuses on man and machine working together, instead of separately, and considers the manufacturer’s impact on the workforce and environment (8). In looking ahead, it feels only appropriate to ask the uber popular AI chatbot, ChatGPT, how things will turn out, “Will AI improve pharmaceutical manufacturing?” “Yes, AI has the potential to significantly improve pharmaceutical manufacturing.”

The combined efforts of regulators, industry, and scientists can now help to realize this potential while safeguarding against risks to patients.

References

1. USM. Top 5 AI Use Cases in Pharma & Bio Medicine. Usmsystems.com (accessed March 8, 2023).
2. FDA. Artificial Intelligence in Drug Manufacturing. FDA.gov. (accessed April 25, 2023).
3. National Academies of Sciences, Engineering, and Medicine. Innovations in Pharmaceutical Manufacturing on the Horizon: Technical Challenges,
   Regulatory Issues, and Recommendations. The National Academies Press online, DOI:10.17226/26009 (2021).
4. Kopcha, M. Beyond ‘good’ Practices. PharmaManufacturing.com (October 20, 2022).
5. Arden, S. et al. Industry 4.0 for Pharmaceutical Manufacturing: Preparing for the Smart Factories of the future. Inter. J. Pharm. online,DOI:10.1016/j.ijpharm.2021.120554 (March 29, 2021).
6. FDA. Artificial Intelligence/Machine Learning Assisted Image Analysis for Characterizing Biotherapeutics. FDA.gov. (accessed April 13, 2023).
7. International Coalition of Medicines Regulatory Authorities. Horizon Scanning Assessment Report–Artificial Intelligence (2021).
8. EC. Industry 5.0, https://research-and-innovation.ec.europa.eu/research-area/industrial-research-and-innovation/industry-50_en, (accessed April 13, 2023).

About the author

Adam Fisher, PhD is Director of Science Staff and Immediate Office within the Office of Pharmaceutical Quality at the Center for Drug Evaluation and Research at FDA.

Article details

Pharmaceutical Technology
Vol. 47, No. 9
September 2023
Pages: 32–34

Citation

When referring to this article, please cite it as Fisher, A. The Future is the Present: Artificial Intelligence in Pharmaceutical Manufacturing. Pharmaceutical Technology 2023 47 (9).