Continuous pharmaceutical processing has the potential to significantly improve the way in which pharmaceutical products are being produced. However, the complexity of continuous pharmaceutical processing also poses new challenges. The design of an effective plant‐wide control strategy to ensure that the critical quality attributes (CQAs) of the final product remain within specification is one of those challenges. This talk will discuss the design and implementation of a modular plant‐wide control strategy for an end‐to‐end continuous pharmaceutical pilot plant, which is aligned with QbD principles and supported by a first‐principle process model. A hierarchical decomposition strategy is used to classify control objectives at different temporal and structural scales. A plant‐wide dynamic model of the process is used to synthesize and evaluate control loops. The results show that CQAs of the final product can be kept close to specification in the presence of significant and persistent disturbances.

The Process Analytical Technology (PAT) concept initiated by the American Food and Drug Administration, aims at designing a well understood and controlled process which consistently yields the desired product quality, thereby reducing the traditional post-manufacturing quality control burden. Ultimately, the relationships between critical quality attributes (CQAs) of the process (typically amount and quality of product) and critical process parameters (CPPs) have to be well defined and understood and the CPPs should be strictly controlled, ensuring acceptable process performance and product quality. Efforts towards implementations of this quality concept have stimulated innovation in the development of tools enabling PAT, examples ranging from statistical data analysis methods to complex analytical techniques. We present a case study for evaluation of PAT tools on a human cell cultivation process. Up-stream processes are amongst the most complex industrial processes, typically yielding large bodies of data which are often not fully analyzed and explored. Therefore, as a initial step towards building process understanding, multivariate data analysis (MVDA) was used to explore early development data. Thorough analysis of the available dataset highlighted the requirements of industrial data to make use of the full power of MVDA techniques, strongly supporting the need for 1) designing sound statistical experimental sets (i.e. using DoE methodology) and 2) investigating additional (online) analytical measurements which could better capture the process’ CQAs. The outcomes of the resulting experiments will be discussed here.

Andreas Schneider is Vice President Life Science Alliances at Roche Diagnostics. He is also the co-chair of the global PAT Data Management Team at ISPE and a member of the Global PAT CoP (Community of Practice) Steering Committee, the BIO CoP and the PAT Article Review Committe

In biomanufacturing, multiple sensors provide a wealth of data that could be used to enhance process understanding and assist in performance improvement. The challenge is how to move from a data rich environment to one where the data is translated to information and ultimately the creation of knowledge thereby realising process benefits. Addressing the data, information, knowledge chain, is not straightforward as there are a series of compounding issues. For example, the data analysis problem is not restricted to one unit operation as reaction has an associated chain of recovery unit operations. This raises the need to understand, and extract, the inter-unit interactions. The approach adopted depends to a large extent on whether it is the development or manufacturing environment that is being considered. In development, operational policy changes are made and new avenues of operation explored. The volume of data can be limited but the spread of data is large and the objective is to search for robust and effective design and operational areas. In production, the spread of data coverage is limited with the objective being to look for occasional deviations from predominantly ‘consistent’ behaviour. The underlying knowledge extraction philosophy is therefore somewhat different between production and development. Process Analytical Technology (PAT) provides a valuable information source that potentially offers process understanding and fingerprinting in both development and production. By embedding PAT data-based process descriptors within monitoring and optimization algorithms, processes inherently adopt a ‘quality by design concept’ (QbD). This will be demonstrated by drawing on the experiences gained from a number of case studies.

There has been an increasing industrial interest in the development and application of new technology tools for enhancing global Pharmaceutical industry related initiatives such as: Quality by Design (QbD), Process Analytical Technology (PAT), Process Intensification, Green Chemistry, etc. One of the key developments that has enhanced developments in these areas is in the standardization of process sampling, as exemplified by NeSSI™ (the CPAC New Sampling and Sensor Initiative). Continued developments of the NeSSI™ platform have expanded its use from sampling a variety of process streams to becoming a platform for process sensors and communication to the control systems. The data from these measurement tools contributes to developing more effective process control strategies – a goal that many companies are trying to achieve.

Initially the draft PAT guidance document originally issued by the FDA’s CDER office was only applicable for well defined small molecule drugs. Within 5 years this perspective evolved to include biological products such as monoclonal antibodies. This resulted in the A-mAb case study report last year in which a mock application of a fully PAT/QbD compliant process for the production of a monoclonal antibody was described. The findings and implications of this report reach further to even less defined products such as vaccines. Vaccines are perhaps the most heterogeneous group of pharmaceutical products, including well defined polysaccharides but also whole inactivated bacterial cells. This means that an off-the-shelf PAT solution for vaccines is impossible. However, basic principles and tools such as acquiring a sufficient level of process understanding or applying DoE to explore the process design space are generic for any (bio)pharmaceutical manufacturing process. In this presentation the general approach as followed in the A-Mab case study and its implication for PAT/QbD application on vaccines will be discussed.

Adoption of the Q10 model for a pharmaceutical quality system should facilitate innovation, continual improvement and strengthen the link between pharmaceutical development and manufacturing activities. Industry surveys indicate a potentially $20-$30 billion more in profit gain, when applying QbD initiatives. Significant reduction of Cost of Goods Sold (COGS) and capital expense, increased technical development productivity are key factors to drive the costs down and increase continuous process knowledge to significantly decrease cost of non-compliance. This presentation will highlight how integrated informatics tools are enablers to create start-to-finish knowledge management repository to support successful QbD adoption across all levels of the organization from senior management, through scientists and engineers responsible for products in development and manufacturing, processes, equipment and facilities.

The usual procedure to identify a design space along QbD principles is to set up a multivariate Design of Experiment (DoE) followed by MultiVariate Data Analysis (MVDA). This approach allows relating the factors to the responses in colorful plots answering “how” they are related but now “why”. This contribution aims at showing a roadmap and the benefits analyzing the “why”, achieving mechanistic understanding. The first step is the converting data into information and understanding multivariate correlations. Secondly, we demonstrate using information instead of raw process parameters can even yield to a reduced number of experiments. Finally we show how mechanistic understanding allows for transferring knowledge along development phases and scale up as well as enabling continuous improvement and also allows for extrapolating knowledge back to process development for platform technologies and the product n+1.

The Quality by Design principles (ICH Q8, Q9 and Q10) were used in support of the route development studies for the definition of a Control Strategy for genotoxic impurities. Five genotoxins were identified by the initial Genotox Risk Assessment (GRA), during the route development studies to 1-(3-{2-[4-(2-Methyl-5-quinolinyl)-1-piperazinyl]ethyl}phenyl)-2-imidazolidinone and needed to be under control in the API. Therefore process research studies, by using tools such as Design of Experiment (DoE) and Principal Components Analysis (PCA), were carried out to define the process parameters affecting the formation of these genotoxins and to assess the impact of them on the API quality. The objective was the definition of a robust Control Strategy.

QbD requires a structured science-based approach to process development and routine manufacturing. That, can only happen if – after considering all relevant scientific and technical aspects involved at each critical to quality unit operation – a systems engineering perspective is used to integrate the different components. A whole process analysis connecting raw-materials to processing to end-product properties must be adopted comprehensively, in parallel to a time-wise integration of data, information and knowledge acquired throughout process development, industrialisation and commercialisation. Continuous improvement can only happen if these two perspectives are present and combined under proper data-, information- and knowledge- management systems. The new process validation guidance from FDA clearly shows that companies will have to address the above challenges to be able to (1) comply with the requirement that all manufactured product lots (batches) should be as good as any of the “three golden-batches” previously required for validation; and (2) to be able to control any deviation and explain it in scientific terms backed by deep process understanding. Retrieving, visualising and managing data sources of very different complexity – from simple univariate data (e.g., process parameters) to complex multivariate data (PAT data on quality attributes requiring further processing) both acquired with different sampling frequencies – to: retrospectively demonstrate compliance, do real-time feed-forward control of a process, extract and store process knowledge to feed-back into earlier stage development activities of other processes and support all types of continued process verification, improvement, troubleshooting and product real-time release,will be the main requirements that commercial QbD platforms need to fulfill and provide in the future to companies. The range of tools in such platforms, the need to apply them to each new product and the importance of having the two types of integration above present at all times, will demand a significant change in the current offer of such platforms and also a much wider training offer of both QbD tools and platforms. Here we review existing QbD platforms and describe some of the desirable requirements for QbD tools and platforms that need to be made available to enable companies and regulatory agencies alike to apply QbD throughout a product/process lifecycle.