Using PAT in clinical bioprocessing for cell and gene therapies – a framework for success

08 May 2024 19min read

Cell and gene therapies (CGTs) are gaining rapid market entry, boasting at least 24 EU and 32 FDA approvals to date, most of which were granted in the past 5 years. Despite their groundbreaking clinical impact on previously incurable diseases, less than 2% of eligible patients can currently access these therapies.

In a recent article, we delved into the challenges that hinder widespread access to cell and gene therapies, such as high manufacturing costs, regulatory complexities and limited reimbursement opportunities. To date, 28% of EU-approved CGTs have since been withdrawn for commercial rather than clinical reasons, suggesting there is a clear need to improve the commercial viability of this technology.

As more cell and gene therapies begin to move into the clinic, there is a growing attention on the need to improve manufacturing processes in particular, which remains a major barrier for the commercialisation of CGTs. Most current CGT manufacturing systems have the capacity to meet the demand for just a few thousand patients annually and use costly, inefficient and largely manual processes adapted from the manufacture of simpler biologics.

Creating a system that addresses these manufacturing challenges requires a deep understanding of bioprocessing fundamentals and the integration of real-time monitoring, analytics and control. Because of the challenges of input material variability and the complexity of CGT manufacturing, it is imperative to implement a systematic and methodological approach to achieve a successful clinical bioprocess development.

  1. Process mapping
  2. Process monitoring
  3. Process understanding and optimisation
  4. Process control
  5. Commercialising cell and gene therapies – the way forward

A brief overview of clinical bioprocessing

CGT manufacturing involves complex bioprocessing of donated human cells, engineered viruses, or genetic material. This process requires precise manipulation and control of living cells through multiple steps, often involving manual labour and taking weeks to complete.

There are two main types of CGTs: allogeneic and autologous. Allogeneic therapies can be produced in large batches using cells from unrelated donors, while autologous therapies are tailored to individual patients using their own cells. Although autologous therapy offers personalised treatment and often results in better clinical efficacy, it presents challenges in process development and scalability due to patient-specific variability in starting materials and the lack of suitable automated small footprint single batch manufacturing bioprocessing tools.

The bioprocessing of CGTs typically involves several steps, such as purification and gene editing, to transform input materials (e.g., patient cells) into the final product. These steps are conducted under defined settings known as process parameters, which include factors like incubation time and temperature.

An effective methodology for bioprocess development

Here, we present a methodology for successful bioprocess development, which focuses on building a comprehensive process understanding and optimisation, utilising several practical tools that can be applied to achieve a successful outcome.

The methodology aligns with the Process Analytical Technology (PAT) framework endorsed by the FDA. According to PAT, the product quality is described by the Critical Quality Attributes (CQAs), which are characteristics of the product (e.g., uniformity, purity) that must fall within specified limits to ensure desired quality. For example, a product with impurity content beyond the specified limits will have adverse effects on the safety and efficacy of the therapy and is therefore not acceptable.

The definition of Critical Process Parameters (CPPs) are processing factors that affect CQAs and must be monitored or controlled during the manufacturing process. Critical Material Attributes (CMA) describe the characteristics of the input materials (e.g. input concentrations and input material purity) which can also influence the finished product quality.

PAT promotes the idea of monitoring CMAs, CPPs and CQAs and adjusting CPPs to achieve the desired CQAs, all in real-time. For example, this could involve increasing the incubation time (CPP) until the cell concentration measured in real-time has reached a certain value (CQA) rather than choosing a fixed incubation time. This approach can help with compensating for variability both in raw materials and equipment, to produce a consistent product.

To apply the principles of the PAT framework in practice, we have set out the following methodology for effective bioprocess development:

blog 1200 x 500 (shorter)

1) Process mapping: the objective of this step is to summarise the current understanding of the manufacturing process and define the scope for further process development. This is a desk-based exercise which includes outlining critical and candidate critical process parameters (PPs and CPPs) and output quality metrics (or Critical Quality Attributes CQAs), as well as expected sources of noise (unwanted sources of variation).

Key tools:

  • Function tree diagram
  • (parameter) p-diagram

2) Process monitoring: the objective of this next step is to develop and evaluate the appropriate measuring analytical systems needed to measure the PPs and CQAs shortlisted in the process mapping step.

Key tools:

  • Fit-for-purpose approach
  • Decision matrix

3) Process understanding: (experiment and simulate): the objective of this step is to gain a further understanding of the relationship between process parameters and critical quality attributes, through experimentation and simulation/modelling techniques.

Key tools:

  • Designing and conducting experiments
  • Machine learning-driven optimisation
  • Modelling and simulations

4) Process control: implementing bioprocessing control measures that will allow development teams to achieve the target values and ranges for critical process parameters and critical quality attributes.

Continuous learning: steps 1-4 can be repeated to further improve the process.

1.Process mapping

Bioprocessing for the manufacturing of cell and gene therapies is complicated to achieve. Process mapping is a helpful way to summarise the current understanding of the manufacturing process and define the scope for further development. By creating a hierarchical breakdown of the overall processing steps and transforming inputs into outputs, development teams can build a clear overview of all functions of the process and reveal noise factors (sources of unwanted variation) as well as interactions. An overwhelming majority of process failures and resulting costs and design iterations are associated with ignoring noise factors during early design stages, making this is a valuable step to undertake.

The outcome of process mapping is a comprehensive list of process parameters and quality attributes, along with a risk-based strategy to prioritise their investigation. Some of the key tools used during process mapping are function tree diagrams and (parameter) p-diagrams.

Function tree diagrams

A function tree diagram is a diagram showing the dependencies between the processing steps of the overall manufacturing process. It breaks these down into fundamental parts, each of which can become the subject of a parameter diagram analysis (see the section below), thereby allowing design and development teams to be more focused and efficient.

The figure below shows a function tree diagram for a theoretical autologous cell therapy manufacturing process, which demonstrates how it may be broken down into its constituent parts.

Function tree diagram example for autologous cell therapy manufacturing:

Blog photos (90)

A (parameter) p-diagram

A (parameter) p-diagram is a structured tool to identify inputs and relate them to the desired outputs for each processing step, while considering the controlled and uncontrolled (noise) factors. It enables development teams to identify material attributes and process parameters that may impact the critical quality attributes.

Diagram demonstrating the purpose of a p-diagram:

Blog photos (91)

The outcome of the p-diagram exercise allows development teams to create a candidate list of critical parameters, which can then be used to shape a list of required measurement systems.

The figure below shows a theoretical example of a p-diagram for the expansion process step of an autologous cell therapy manufacturing process. The p-diagram activity could be performed either for all steps or just for the most critical steps to form the basis of the process monitoring needs.

(Parameter) P-diagram example for autologous cell therapy manufacturing expansion process step:

Blog photos (92)

2.Process monitoring

P-diagrams allow development teams to produce a comprehensive list of the desired measurement systems for monitoring all three types of variables: inputs, outputs and noise. The aim of the process monitoring stage is to develop and evaluate appropriate measuring analytical systems to measure the process parameters and critical quality attributes shortlisted in the process mapping step.

Having identified which parameters need to be measured, the next step is to identify the measuring system key requirements for each of the parameters, to help identify the most suitable method available. We discuss two key tools used in the process monitoring stage, including the fit-for-purpose approach and the decision matrix.

Fit-for-purpose approach

When defining requirements and developing a measuring system, a useful notion to follow is the fit-for-purpose approach. The idea of fit-for-purpose is to consider the intended use of the data acquired from the developed method when defining requirements, developing and validating the measuring system.

For example, a measurement system used to monitor CQAs in real-time based on which CPPs are adjusted to affect the final product, should be fully validated. Whereas a measurement system used to identify critical process parameters during the early stages of process development may not require the same stringent validation.

By following this approach, development teams can guard against over-investing in a method that might have limited utility. The following is a list of questions that can help when defining key early-stage-requirements, using the fit-for-purpose approach:

  • What is the system measuring? Is it measuring the critical parameter/quality attribute directly or an orthogonal metric?
  • How should it be measured? Consider interaction with the sample. Should it be off-line, in-line, at-line, on-line or non-contact and non-destructive? Should it be real-time, or taken at fixed intervals? Accurate in-line, on-line and non-contact measurements are the preferred options for real-time monitoring and dynamic control.
  • Under what conditions? What environmental conditions would the system operate under (temperature/humidity/pressure) and are there any chemical or sub-system compatibility issues? This is an opportunity to consider any user dependent or external noise factors. For example, sensors interfaced directly to a bioreactor must be sterilisable and should not be affected by fouling or interfere with the medium.
  • How well should it be measured? Describe the desired performance metrics in terms of minimum sensitivity (the smallest change in the metric that can be detected), maximum tolerated error (systematic or random deviation from the true metric value) and range (the range of change to detect). Some of the key properties to evaluate for a measurement system include accuracy, precision, linearity, stability and capability.
  • Other considerations? Think of any constraints such as GMP compatibility, size, shape and cost.

For many of these requirements, there will be numerous measurement options to select from, including the option to buy and adapt , or make custom measuring systems. In the early stages of process development, where agility is key, buying off-the-shelf solutions is typically the preferred option. However, it is often the case that there are no off-the-shelf solutions capable of satisfying all the desired requirements, meaning adapting these or making custom solutions is worth exploring early on.

Decision matrix

By gathering the above information on potential measurement options, development teams can then construct a decision matrix to decide on the most appropriate monitoring method by weighing against the identified key-requirements.

The example below shows a comparison of two methods for measuring the “cell density” critical quality attribute (as part of the “expand cell” process-step shown in p-diagram above). Option 1 shows a higher-sensitivity, more expensive and more versatile imaging/AI based method, while option 2 shows a cheaper, optical density-based method.

A suitable strategy for this example would be to utilise the higher specification (option 1) method, to help establish an understanding of the relationship between process parameters and cell density. Once a critical relationship has been confirmed, a lower sensitivity method could then be tested to evaluate whether a lower cost alternative would be capable of the monitoring needs in production.

Once a measurement system has been chosen for implementation, the system can be evaluated to ensure it meets the key requirements and can lead to meaningful results.

Decision matrix example for “cell density” measurement system:

Blog photos (93)

3. Process understanding and optimisation

The objective of this step is to gain further understanding of the relationship between process parameters and critical quality attributes. This involves a process of identifying critical process parameters, characterising their effect on CQAs and optimising their target values and range.

At this stage, development teams can build knowledge on these relationships through designing and conducting experiments, or carrying out mathematical modelling and simulations. Three key tools for this are design of experiment approaches, machine learning and modelling and simulations.

Design of experiment

There are several types of experiments within the Design of Experiment framework that can be used to investigate key relationships in the bioprocess design. Some of the key relationships to investigate include:

  • Criticality and adjustability: Identify the CPPs and CMAs and how their variation affects the CQAs and gain an understanding on how they need to be adjusted to achieve the target outcome.
  • Interactivity: Understand how CPPs and CMAs interact to impact the CQAs, meaning the way by which the effect of one parameter on a CQA is dependent on the level of another parameter. Consider adjusting parameters together instead of in isolation.
  • Sensitivity: Describe the sensitivity of CQAs to the CPPs and CMAs (i.e. for a given change in a CPP or a CMA, what is the resultant CQA change?). Measure sensitivity to define the process tolerances and control strategy.
  • Robustness: Measure how CPPs, CMAs and CQAs are affected by changes in noise factors to implement control measures against them.

Another useful tool early on in developments is One-factor-at-a-time (OFAT) experiments, where a single factor is varied at a time and the effects are analysed. These can help development teams to understand a factor’s impact quickly and cheaply, before investigating it further with more extensive experimentation. However, OFAT is inefficient for understanding complex bioprocessing interactions and carries a significant risk of false conclusions.

The Design of Experiment methodology enables experimentation that can measure the effect of multiple factors (PPs/MAs/Noise) and their interactions on a set of outcomes (CQAs) in a short timeframe. Results from these experiments can be modelled using statistical software to identify optimal factor ranges for a given response.

Designs

Highlighted below are DoE designs that may be useful in the context of cell therapy research and process development.

  • Screening designs are used to evaluate criticality and preliminary adjustability. Here, the aim is to estimate the effect each PP/MA have on the CQAs and distinguish the most critical ones. These designs are most useful where main effects are not well understood. In the context of cell therapy, an example could be transitioning from a manual lab-based process to an automated instrument-based one. Here, demonstrator proof-of-principle rigs with high capacity of integrated measurement systems can be quickly put together, to evaluate the feasibility of transitioning the method and understand which parameters will be of highest risk.
  • Full and fractional factorial designs are used to study interactions (and adjustability) as all factor combinations are tested. In the context of cell therapy, full or fractional factorial designs are suitable after a few critical parameters (3-4) have been established. As part of the factorial design experiment data analysis a mathematical model can be developed for process optimisation and control. However, these experiments can incorporate a high number of runs and may not always be feasible.

An example of using DOE for CGT process development, is the optimisation of the gene delivery and cell transfection step via photoporation. Goemaere et.al have shown that through a DOE design, they were able to achieve optimum delivery yield five to eight-fold more efficiently than when using an OFAT methodology.

Considerations

Process understanding experimentation for CGT process development can be costly, time consuming and particularly challenging due to the process complexity and input material variability. To increase the likelihood of positive outcomes it is worth considering the following:

  • Experimental setup: To make sure the experimental outcome is of use in product development, it is essential to ensure the experimental methods and results can be translated to a GMP manufacturing environment. This can be achieved using agile, modular rigs with integrated measurement systems as well as the potential for GMP conformity and scalability. This can help to avoid investing too much in optimising in a design space that might not capture all the variables of the final process.
  • Variability and noise: Existing data can be analysed to provide insight into variation in a process outcome and advise on sample size and the statistical power of the Design of Experiments. A significant consideration for cell therapy DoE is capturing variation of starting material and noise factors. Monitoring and/or actively varying these factors (compounding), allows teams to evaluate their effect and test process robustness. It can also prevent them from screening effects from factors of interest.

Machine Learning (ML) driven optimisation

Whilst the DoE approach is a successful and well-known tool for the development and optimisation of bioprocessing, in the field of CGT, interpreting these results can be challenging due to the input material variability mentioned earlier. Machine learning (ML) pipelines are a recent alternative for process optimisation.

Academic studies have already demonstrated the efficient optimisation of T cell culture medium for various donors using a ML pipeline, applied into a DoE based on data from a single screening DoE experiment. This new approach could help overcome the challenges of heterogeneous biological material and lead to process optimisation across donors from one dataset, without the need for sequential experiments (e.g. screening followed by optimisation experiments).

By using machine learning, process optimisation could be achieved without the need of thorough model interpretation. This can also lead to a major reduction in the experimental effort and shorten development times substantially, thereby helping to optimise robust autologous cell products for viable therapies for every particular patient.

Modelling and simulations

Besides hands-on experimental work, mathematical modelling and simulations based on the fundamental understanding of the process elements can be used to derive conclusions and advise experimental planning in a quick and cost-efficient manner.

Computational fluid dynamics (CFD), for example, is a cost-effective tool for the development of bioreactors. Simple models can be used to optimise hydrodynamic environments that increase cellular culture performance. For instance, they can advise on geometries and flow rates that limit forces such as shear and pressure gradients that could otherwise damage cells, while operating the reactor in cost-effective conditions.

woman-coding-at-a-desk

4.Process control

Understanding the relationship between process outcomes (CQAs) and process parameters (CMAs and CPPs) can help achieve the target outcome with minimum variation by controlling the process.

This control may involve adjusting parameters or maintaining certain values constant. Mathematical relationships derived from experiments and simulations can determine target values and acceptable deviations from them, ensuring acceptable variation in drug quality attributes.

In the case of autologous cell therapies, with the inherent variability of the input material, control might be a challenge since the input material itself is part of the equation. One possible solution could be the derivation of formulations and process parameters that are optimum for multiple donors.

An alternative possible solution could also implement the PAT approach. This would require manufacturing an instrument capable of characterising the CMAs onboard and monitoring CPPs and CQAs throughout the process. Measurements could then feedback information in real-time to dynamically control and adjust the process parameters accordingly to achieve the desired outcome. A custom manufacturing “recipe” could be applied to each individual donor to produce a personalised drug with the best efficiency.

Commercialising cell and gene therapies - the way forward

The drive towards the commercialisation of cell and gene therapies requires robust manufacturing strategies. Autologous cell therapies present a paradigm shift in personalised medical treatment and it is imperative to embark on the design and development of bespoke solutions tailored precisely to the unique requirements of this field.

Achieving this requires a deep understanding of bioprocessing fundamentals and the integration of real-time monitoring, analytics and control directly into the manufacturing process and small footprint bioprocessing tools. Because of the challenges of input material variability and the complexity of CGT manufacturing, a systematic and methodological approach is needed to achieve successful outcomes.

Our recommendation to CGT developers is to draw from the principles of PAT and quality by design – start small and iterate for a manageable upfront investment. It is not necessary to employ analytical tools for every aspect from day one. An iterative and systematic approach, addressing the most critical aspects first, will instead allow for a gradual adoption. Investing in or developing user-friendly, adaptable PAT elements that evolve with changing manufacturing needs is also a useful approach to take.

The methodology discussed in this article aims to enable efficient navigation through the intricate landscape of cell and gene therapy manufacturing, facilitating informed decision-making to ensure optimal, flexible and efficient outcomes.

Join the conversation

Looking for industry insights? Click below to get our opinions and thoughts into the world of
medical devices and healthcare.