Automating intelligently: The growth of automation and AI in pharmaceutical manufacturing

Introduction

Many pharmaceutical companies embrace automation and other digital-based technologies such as those connected to data-driven decision-making to create process efficiencies and to remain competitive. To confer the maximum advantages from automation, artificial intelligence can assist with optimizing the automation process. This article serves as an introduction to automation coupled with artificial intelligence for pharmaceuticals, considering the drivers, the technology, the importance of data integrity and the impact upon the types of work roles required for the future.

The automation priority

Process automation is increasing, where the goal has been set to transform inefficient and error-prone tasks in pharmaceutical manufacturing (and hence automation connects neatly with the Pharmaceutical 4.0 paradigm). Speed is another area in which automation steals the limelight. Automation may also become an integral part of businesses seeking to maximize effectiveness with limited personnel. Automation is also crucial in pharmaceutical engineering systems, as the likelihood and cost of human error are high, and human reaction times are slow compared to the system. Robotic forms of automation can reduce error and also reduce human intervention into critical areas, which adds an additional advantage for areas where microbial contamination would trigger automatic batch rejection as with aseptic processing. Automation can be boosted through artificial intelligence, mainly where there is a need to select between options or make predictions about what the next action will be.

Examples of intelligent automation solutions for pharma

Examples of automation adoption in pharmaceutical manufacturing include the use of robots to handle primary pharmaceutical packaging operations. As the first generation concept, applying automation has enabled manufacturers to manage bottle orientation, capping, labelling, and collation systems. As part of the second generation of automation, technologies have enabled firms to monitor the operation on a supervisory level, such as checking for low hopper levels, fallen bottles and low-level supplies. The third generation will include aspects of artificial intelligence.

For example, with real-time Raman spectroscopy instruments (a direct optical method based on inelastic light scattering)

1. Can assess the quality of the finished medicines by measuring the molecular vibration and rotational energy changes of tablets. It ensures the chemical compounds are as desired in each tablet before they are released for distribution. AI can assess the data and be used to decide the quality of individual product items. One case study demonstrates using the technology to make chemometric analyses of ibuprofen using a Partial Least Squares Regression (PLS-R) algorithm. The outcome was assessed using a total error approach and ± 15% as acceptance limits, with success
2. A second study considered the use of automatized polarization-resolved Raman setup, coupled with chemometric analysis, for enantiomeric differentiation of butan-2-ol.
3. Types of machine learning can also assess whether products are packed within the correct bottles and whether the boxes are marked with the correct labelling affixed. The use of robotics can also be applied to dispensing, sorting, kit assembly and light machine-tending. The advantages of being sought include greater speed and accuracy, more flexibility, and more reliability. Developing the process manufacturing theme, in keeping with the process analytical technology (PAT) initiative and the increasing demand for effective and versatile real-time analysers during drug manufacturing. In the last decades, Raman spectroscopy has been adopted in secondary manufacturing of pharmaceutical solid dosage forms, covering the common secondary process steps of a tablet production line
4. In a different example, connected to the reliability, AI can help with automated process adjustments. Consider this case: If a temperature gauge makes a higher than expected reading, the AI machine could detect it and rectify the situation rather than requiring an operator to intervene and make an assessment about the required course of action. Between batch runs, automation and technology can create the opportunity to leverage data and analytics to improve processes. This is sometimes referred to as ‘enterprise Manufacturing Intelligence’
5. Also related to core manufacturing is the idea of “connected manufacturing plants.” One feature of this type of model is to have data visible and available on demand. Connected manufacturing puts in place the required steps in the manufacturing process, like delivery of raw materials or shipment of the finished product, which can be added into the process without disturbing production. For example, it is possible to construct an especially designed mixing unit that can blend active ingredients into a homogenous compound. This process could happen with no interruption, which would save time and reduce costs and lower the possibility for error.
  
  The ideal form for the connected plant is through platform level automation. Such a system functions as a single-user interface terminal designed to simplify processes for the human operator. The automation facilitates the operator in viewing monitoring, and manipulating multiple systems via a single username/password and with a standard look and feel to the interface
6. Ordering supplies and stock control is also subject to advances with automation. Examples of automated information management include Enterprise Resource Planning (ERP) systems, under which there are Manufacturing Execution Systems (MES) and actual production process management systems, such as Distributed Control Systems (DCS). The information structure can be built to be flexible so that it can be supplemented, if required, according to changing production requirements
7. Intelligent ERP systems can automatically reorder stock and enable management of inventory at velocity.
  
  From the quality control laboratory perspective, advances in chemometrics and data-driven classification approaches have enabled automation as well as a more straightforward interpretation of results. Automation has also enabled parts of tests to be performed by robots, such as nuclear magnetic resonance and high-performance liquid chromatography. A different area of laboratory focus is with developing personalized medicines. Here automation and robotics can help to phase out the manual testing of individuals’ genomes. AI enhanced automated, high-throughput screening technology enables scientists to access a vast amount of data with little manual interaction. This not only helps to keeps costs low, but it also increases the speed at which any system can analyse gene sequences. Automated systems can also analyse information relating to patient medical histories, genotype data, familial inheritance, and biomedical research. The inclusion of AI permits individual clinical and molecular profiles of each patient to be constructed and for these outcomes to be used to develop personalised medicine and care
8. Furthermore, artificial intelligence can aid pharmaceutical drug development by aggregating and synthesizing information, establishing suitable biomarkers, generating data and models, repurposing existing drugs, generating novel drug candidates, and analysing real-world evidence.
  
  Data is critical during the drug development process. The pharmaceutical sector is probably the only area of business where, to get the product from idea to market, the company will need to spend about a decade and several billion dollars to advance a product (in this case a medicine) to the final phase. At any point up to the final registration of the drug, there is about a 90 percent chance of failure. In some cases, artificial intelligence can reduce the molecule screening process by several years and reduce the discovery-to-market process up a decade. AI can also assist with drug development productivity. There are also many false starts, which can occur at any stage of the development process through to clinical trial.

Data integrity

Pharmaceutical regulators continue to be concerned with Records and Data Integrity (RDI), since this is fundamental to the evidence created across all aspects of GMP. It applies equally to automation as it does to other systems used in the industry. The regulators have positioned several key statements to clarify their expectations. For instance, the U.S. Food and Drug Administration (FDA) has increasingly observed Good Manufacturing Practice violations involving data integrity during inspections. The Agency regards this as troubling since ensuring data integrity is an essential component of the industry’s responsibility to ensure the safety, efficacy, and quality of drugs. Moreover, these data integrity-related violations have led to numerous regulatory actions, including warning letters, import alerts, and consent decrees. This reach includes automation and the application of artificial intelligence.

The most common error remains, not considering how records will be captured and tracked, including how any changes are captured in an audit trail as part of the design process. To overcome this, it is necessary to ensure that electronic records’ integrity as part of system validation and implementation will ensure a positive compliance report. Cultural change is essential in parallel with the technological strategy. Records and data integrity governance is the total company policy, procedures, people, and the organization’s quality management system.

AI and the future of work

The extent that AI will lead to new types of jobs is certain; whether AI will lead to fewer jobs is less certain, given the process of technological innovation itself to create a new and varied role. Moreover, there are limitations with AI technology as it is currently configured. AI systems can be good predictive systems and be particularly good at pattern recognition. However, AI systems have a very repetitive approach to sets of data, which can be useful in certain circumstances. Yet, this approach to data can result in AI making errors, and these are sometimes in the form of obvious mistakes to a human. This occurs because AI does not have a sense of context. As people, we have years of experience in the real world. We have vast amounts of contextual data stored in our brains that make it possible to predict and to know the boundaries of the real world so that even if we have never been in a particular situation, we are still able to deal with it. An example is, with the external environment. An HVAC system, for example, maybe influenced by actual or impending adverse weather. Where AI is used to help automate damper controls to maintain the pressure differential cascade, an AI system can neither observe nor utilise information from the external environment to make this type of judgment, whereas the operator may be completely aware (10). The unknown situation is where AI will fall under such circumstances.

Hence, we will, for the foreseeable future at least, continue to need operators as we still need to understand the environment around us, which may contain an infinite number of possibilities. A further example of this is with patient safety and assay results. No organization recommends that we ever forgo having a human being validate findings and provide context when the decision could lead to the difference between life or death.

The issue at hand is more to do with the new skills required. The ability of employees to work with digital technology – drawing in the ‘human factors.’ It is central to the concept of ‘digital IQ,’ which considers how an organization adapts to change and utilises emerging technology to advance company goals. It remains that too many employed in the pharmaceuticals sector lack essential technology skills.

Summary

Through this introductory article, automation has been presented as a technology that can create efficiencies within many aspects of pharmaceutical manufacturing. Beyond efficiencies are important quality and research improvements when automated systems are connected with appropriately trained AI. The introduction of AI is not straightforward in terms of machine training, and both automation and AI can present data integrity challenges. They will, by continuing on the current trajectory, reshape the composition of the pharmaceutical sector workforce. Each of these aspects has been explored.

Whatever the current digital maturity level, automation is being introduced, often rapidly and often to scale. There are some manufacturers, however, who see this new technology as something too complex to implement. It relates to cost, ease of use, and acceptance by staff. This can be counterbalanced with arguments connected to return on investment, energy savings, flexibility, high-speed production, and increased quality. These divisions lead to an uneven path towards adoption; however, the intelligent automation paradigm continues to make inroads.

Tim Sandle is the author of the book Digital Transformation and Regulatory Considerations for Biopharmaceutical and Healthcare Manufacturers, Volume 1: Digital Technologies for Automation and Process Improvement. available via the PDA Bookstore: https://www.pda.org/bookstore/product-detail/5897-digital-transformation-volume-1

References

Long, D.A. (2002) The Raman effect: a unified treatment of the theory of Raman scattering by molecules. John Wiley & Sons, Inc., USA
Mansouri .M. A. et al (2020) Quantitation of active pharmaceutical ingredient through the packaging using Raman handheld spectrophotometers: A comparison study, Talanta, 207: 120306
Rullich, C.C. and Kiefer J. (2019) Principal component analysis to enhance enantioselective Raman spectroscopy. Analyst;144:2080–6
Esmonde-White KA, et al (2017) Raman spectroscopy as a process analytical technology for pharmaceutical manufacturing and bioprocessing. Anal Bioanal Chem. 409(3):637–49
Marshall, T. (2018) Automation in Pharmaceutical Manufacturing, Contract Pharma, at: https://www.contractpharma.com/issues/2018-03-01/view_features/automation-in-pharmaceutical-manufacturing/
Guest, P. (2018) How Can Your Biomanufacturing Facility Benefit from Process Automation? Pharmaceutical Online, at: https://www.pharmaceuticalonline.com/doc/how-can-your-biomanufacturing-facility-benefit-from-process-automation-0001
Åberg,J. (2020) Flexibility and Efficiency in Industrial Automation, Pharmaceutical Online, at: https://www.pharmaceuticalonline.com/doc/flexibility-and-efficiency-in-industrial-automation-0001
Wilkins, J. (2017) The benefits of automation in pharmaceutical manufacturing. The Engineer, August 2017,at: https://www.theengineer.co.uk/automating-pharmaceuticals/
Sandle, T. (2017) Pharmaceuticals facing up to data integrity concerns, Digital Journal, 14th July 2017, at: http://www.digitaljournal.com/life/health/pharmaceuticals-facing-up-to-data-integrity-concerns/article/497654
Oldfield, M. (2020) Will AI Take Your Job? The Tech Magazine, 1(1): 18