OR WAIT null SECS
Effective analytics will eliminate failures, deviations, and non-conformances.
Not long-ago, automation in pharmaceutical manufacture meant robots picking items from a conveyer belt, placing them for different robotic arms to neatly position them for further processing, and, ultimately, folding them into their packaging. Shortly after robotic pipetting became the norm, the X-Y robot platform began to find a home in research to help build combinatorial libraries in drug discovery, or to deposit samples for experiments, and perhaps off-the-shelf commercial microarray platform products. Revolutionary in their day, these now seem antiquated compared with several generations, spurred in part by FDA’s encouragement of process analytical technology (PAT), of sophisticated spectroscopy and software integration systems.
The pharmaceutical automation market is projected to reach $18.2 billion by 2029 at a compound annual growth rate of 12% from 2022 to 2029 (1). This growth is still partly driven by early FDA encouragement alongside White House and other national programs to modernize industry and each nation’s manufacturing competitiveness. The Advanced Research Projects Agency for Health (ARPA-H) and BioMADE come quickly to mind. BioMADE is a manufacturing Innovation Institute (MII) sponsored by the US Department of defense. In the United Kingdom, the High Value Manufacturing (HVM) Catapult or the Medicines Manufacturing Innovation (MMI) Centre stand out. By pooling funds, adoption of novel, complex, or costly equipment can be quickly trialed, assessed, and encouraged to be broadly adopted. All this is done with a mind’s eye, ultimately, on current good manufacturing practice (CGMP) and regulatory compliance. But if the technology is especially novel or effective in a new way, these adoptions can even help drive evolution in the regulations themselves.
In manufacturing tightly regulated healthcare products, human performance has long been second best. Society has oriented itself more toward the magic of silicon chip processors and digitalization, so manufacturing has embraced what has become admiringly termed “Industry 4.0”, with a view to attain end-to-end task automation. As these manufacturing tasks in pharmaceuticals become not only more complex and diverse but also more expensive, analytical tools and sensors have been raised on high as inseparable from “success” itself. While the complicated and costly new analytics provide a healthy positive return on investment in many applications, they are not always the optimal answer. These analytics necessitate a more highly skilled workforce with cross-team integration that includes built-in flexibility and a willingness to continually learn new skill sets, repeatedly.
Mauricio Futran, owner of Pharmaceutical Engineering Solutions, who spent 40 years in Big Pharma, working for Merck, Bristol Myers Squibb, and Janssen, a Johnson & Johnson company, stated in 2023 that “segregation and lack of uniformity have been a traditional issue in the industry. The beauty of continuous manufacturing is that the issue of segregation goes away … we now have process analytical tool (PAT) sensors that can assure you are on the right quality in real time, the systems are super steady without drift, and also, when talking about tableting, you can actually assay every single table with near Infra-red, and have [individual tablet-level] quality assurance and control of the process” (2).Taking this into consideration, and taking a step back to look at the wider-picture implications of this digitalization, Futran went on to conjecture that “this is going to drive an evolution in regulations ... and in concepts of validation ... on top of which you add multivariate analytics, this gives an assurance that what you are doing is right on track of where it should be, compared with all your preceding batches” (2).
In the 1960s, NASA pioneered the use of digital twin technology. This in-silico method of comparison and extrapolation kept humans safe while drastically reducing operation and equipment costs for hard to repeat experiments. To better predict risks and reduce trial and error delays, pharmaceutical equipment designers have begun to embrace and deploy digital twin technologies. A recent high-profile example was GSK using digital twin technology “to fine-tune their manufacturing process when scaling up production of their vaccine adjuvant in response to the C[OVID]-19 pandemic” (3). In her study of its use, Alison Doughty goes on to say, “The Empa research center in Switzerland [The Swiss Federal Laboratories for Materials Science and Technology] is working on digital twins to optimize drug dosage for people afflicted by chronic pain, taking into account factors such as age and lifestyle to help customize the digital twin” (3).
In the hopes of early and less costly optimization, the efforts are typically coupled to machine vision and machine learning (ML) models, along with cloud-based historical batch process performance. In this way much can be learned and tried in a quick and cost-effective manner, especially when first approaching a new molecular therapy or chemical modality. When married to artificial intelligence (AI) and ML, this can be perhaps the most effective way to experimentally churn through the largest number of potential solutions in any given pharmaceutical challenge. But turned to more mundane efforts, such as keeping an eye on a bioreactor process or flow chemistry system for, say, critical quality attributes (CQAs), it has revolutionary potential.
In a presentation at the 2022 PDA–FDA Joint Regulatory conference, Martin Van Trieste spoke about drug quality, Industry 4:0, Six Sigma, and the data deluge, saying, “We’re now using machine learning and AI, and that will really help eliminate failures, eliminate deviations, eliminate non conformances, and get past what I call blind compliance” (4).He went on to say that because of AI, “the future looks really bright for patients with improved safety, reliability, and quality, and all at a lower cost, some of which will be passed on to the consumer. Now if the future is bright for this group of people, it must be bleak for somebody else. And it’s going to be bleak for the dinosaurs of the industry, for the large quality organization that have been built around testing and inspecting the product, because if you are doing AI, you don’t need large testing organizations. The computer will not allow a mistake to happen, and it will safety check everything at the end … It’s really good for the patients. It’s not good for the quality unit, so I suggest that you really learn data analytics, you really learn AI‚ become that resource in the company, because that’s how you’re going to keep your job in the future” (4).
1. Meticulous Research. Pharmaceutical Automation Market by Component (Plant-level Controls, Enterprise-level Controls, Plant Instrumentation), Mode of Automation (Semi-automatic, Fully-automatic), End User, and Geography—Global Forecast to 2029. Market Research Report, May 2022.
2. Mirasol, F.; Spivey, C. Drug Digest: Advancements in the Implementation of Continuous Manufacturing. PharmTech.com, May 5, 2023.
3. Doughty, A. How the Pharma Industry is Using Digital Twin Technology. www.linkedin.com, July 6, 2022.
4. Spivey, C. Six Approaches Making the Most Significant Impact in the Future. Pharm. Technol. 2023, 47 (1), 16–18.
Chris Spivey is the editorial director of Pharmaceutical Technology.
Volume 48, No. 2
When referring to this article, please cite it as Spivey, C.Analyzing Exabytes. Pharmaceutical Technology 2024, 48 (2), 30–31.