Improving workflow

Petter Mörée discusses the importance of process analytical technologies.

What challenges are pharmaceutical companies currently facing in terms of streamlining their manufacturing processes?
Petter Mörée. Based on Umetrics experience it is the interaction, or rather lack of interaction, between software, hardware, measurement systems and infrastructure that is the major challenge in streamlining the manufacturing process. For existing processes there is a strong drive to integrate software and measurement systems into hardware, such as granulators. Umetrics works with a number of suppliers to ensure that all available data from a process can be collected and used for multivariate process analytical technology (PAT) models. An additional challenge is that many processes consist of a series of unit operations that are often not one-to-one concerning size, therefore splitting and pooling of material is necessary.

What are the four levels of PAT, as defined by Professor Svante Wold? Why do they require different multivariate tools?
PM. Level 0 - Design of Experiments (DoE) also referred to in ICH Q8. DoE is often used in development of new processes and products and lead to an increase of information and decrease of experiments. With the design space concept there has been an increase in DoE activities and also a need for improvements in DoE software so that the risk of failure can be accurately calculated and visualised.

Level one - Off-Line PAT is the use of analytical techniques that are faster than traditional wet chemistry. A common technique is NIR that requires multivariate calibration but many of the newer methods are also spectroscopy based and benefit from the use of MVA.

Level two - At-Line PAT, could for instance be the use of raw material characterisation. If knowledge is gained for raw materials or intermediate products, operations can be adopted to compensate for these variations. At-line PAT means working very close to the process, usually by process operators, measurements and software need to be adopted to a process environment.

Level three - On-Line PAT, use of real-time tools for one or few unit operations with the ability to react to process variation in real time rather to react to variation in quality attributes.

Level four - Entire process PAT (super model). This is the combination of all above levels where one focus is to enable prediction of final quality attributes using relevant data from entire process.

For level three and four there is a need for very flexible software that can handle both different types of data and large data amounts. On-line software adds an additional level of complexity where multivariate technology is one part in a total infrastructure that has to function seamlessly.

In what specific ways could a pharmaceutical company use multivariate technology software to improve its workflow?
PM. The goal of the development projects has been to establish a design space that gives a reliable production and thereby avoid time-consuming analytical measurements. The feedback, although somewhat limited, is that the strategy works but one still needs to establish the exact balance in how much should be handled by multivariate models from the process and how much by traditional analytical techniques.

For old processes Umetrics have established multivariate models that can be used in realtime (in- at- or online) for monitoring and prediction and, in some cases, control purposes. These models open up the possibility for release of intermediates. Dependent on the process, saving can be quite substantial, one example 40 percent yield improvement on a 30-year old process.

How do you see the use of PAT developing in the future in the pharmaceutical industry?
PM. Based on Umetrics work in non-regulated areas, which we believe will also increase in importance within pharmaceutical productions. PAT control strategies where the process adjusts for variations in environment, equipment and materials to keep output constant and optimal, not keeping the process locked. In the same way as pharmaceutical industries have interacted with and learnt from other industries for batch processes, Umetrics sees that this happens for continuous processes. Most of the tools and the software is already available for PAT within continuous processes and, in Umetrics view, the transition to continuous process will accelerate.

Measuring the entire particle size distribution and defining quality with the d50 is throwing away information. Quality described by PCA or related methods is slowly gaining acceptance. Umetrics have a customer that uses PCA plots to decide which batches to be sent to which customers. In our view this will become more common.

Petter Mörée is Director of Online Products for Umetrics. Mörée has a technical background with a M.Sc. in technical chemistry with a specialisation towards chemometrics. After his M.Sc. Petter joined Umetrics in 1998 as an application specialist and has for the last nine years worked towards PAT and QbD with increased responsibilities.