|Committee E55 on Process Analytical Technology: A Suppliers Perspective
by Ian Clegg, Ph.D.
The pharmaceutical industry is subject to a number of pressures, although the overriding drivers remain the research, development and marketing of new therapies and the maintenance of the highest quality standards at point of use. Both these factors are in the direct interest of all stakeholders, i.e., manufacturers, consumers and industry regulators. The R&D groups within the industry concentrate on the development and assessment of new chemical entities (NCEs) and dosage forms, that is, on the delivery of new products, since this ensures that new medicines will become available with associated high profit potential and that the industry will grow.
A direct consequence of the emphasis on NCEs and new product forms has meant that manufacturing, as a whole, was not considered a critical core competence or tool for competitive advantage. Manufacturing assessments are widely practiced in industry in general and there are many standard metrics that allow performance to be compared and contrasted between the various industrial sectors (an assessment produced by ABB Process Solutions is shown in Table 1). Although there is no doubt that pharmaceutical manufacturing processes are able to produce products to specification, these metrics show that manufacturing performance and therefore cost effectiveness lag behind those of other sectors.
There are many potential reasons for these differences in performance but one example is the differences in the regime that ensures that appropriate product quality is achieved. Within the pharmaceutical industry, the existing quality regime in manufacturing is based upon a process of control whereby raw materials, intermediate products and final products are all typically subjected to an extensive range of clearly defined and closely monitored analytical and physical tests. If a material being tested conforms to a pre-determined standard value then it is deemed fit for purpose and processing can continue or product can be released for sale. This regime is traditional practice for the industry and it seems intuitively reasonable, i.e., there are clear checks and balances at all the critical stages during manufacture.
However, in terms of effective use of capital assets, this approach to the control of quality does confer significant disadvantages compromising capacity, cycle time and cost and is clearly out of step with other manufacturing sectors such as semiconductor and aerospace. In these other sectors, product quality is assured through the operation of robust and well-understood processes. Such processes can be relied upon to produce high-quality products quickly and in a cost-effective manner; they have a minimum amount of quality control testing and the overriding emphasis is based on thorough understanding of the manufacturing process embodied in process control mechanisms that are inherently repeatable.
Process Analytical Technology
There are several ways in which this existing position within pharmaceuticals might be improved, although one of the more prominent ones is the application of process analytical technology; this has been the subject of a significant amount of coverage in the technical and trade press in recent months. With PAT-based systems, the parameters that affect the process are monitored, their effect on the process understood, and their influence controlled. PAT systems allow the quality of the products to be assured by enabling highly repeatable and inherently stable processes with immediate indication of non-conformance. In terms of the manufacturing metrics described here, PAT would enable improvements to be made, for example CpK (process capability) would be increased and cycle time radically decreased.
PAT is the direct subject of the recently formed ASTM Committee E55 which has approximately 140 members drawn from pharmaceutical manufacturers, suppliers, regulators, as well as academics and consultants.
In some ways, the acronym PAT is misleading. Although this initiative is based on analytical science and measurement technology, it also includes other important tools (see Table 2). The problem word in PAT is analytical because this implies complete reliance on the analytical aspects of process technologies. It is true that the analytical measurement method is important, as this gives the primary data view into the process and can provide unique insights into absolute process behaviors, stability and dynamics. Indeed, if suppliers fail to provide a wide range of reliable, robust and validated measurement techniques that are able to perform properly under realistic process conditions, then this will inevitably limit the rate of deployment of measurement technologies and the associated improved process understanding within the industry. It is, however, comforting to reflect on the fact that the provision of the analytical measurement system is a reliable and fairly well-proven aspect of the PAT initiative, for example, the substantial number of instruments (predominantly based on near-infrared spectroscopy) that have already been installed within the industry.
Unfortunately, it is also true that the analytical measurement is of limited value to manufacturing (although the value of improved understanding in R&D and scale-up cannot be underestimated) unless it works in concert with other procedures and systems. For it is only after the transition has been achieved from raw data through to information and ultimately to knowledge that true value accrues to the user of PAT-based systems.
A view of the ultimate objectives of PAT is shown in Figure 1. In the left hand side of this figure is a data-centric description of the world (i.e., the process under examination), and the right hand side is a model-centric view of the same world. This diagram describes a situation where processes are developed and a virtuous cycle is set up, whereby data from a PAT system is induced into a model-centric view; the results are then evaluated and modifications are subsequently made to the PAT system (or its data treatment sequence) and further sets of results are induced into the model.
In this way, there is a continuous improvement in the overall data set available to the assessor of the system and the degree of understanding of the system increases as a direct result. Here, the system may be as simple as a small unit operation or as complex as a large manufacturing unit with multiple, integrated stages. Once the model-centric view of the process is complete and in full accordance with that process, and it has been validated as such, then this is the ultimate realization of PAT. In such a situation, the interaction between all significant process variables are known and constraints can be applied to them; product quality can be assured through good design practice and appropriate control of the process.
A Complex Process
Of course, the transitions between data, information and knowledge takes some time to achieve; it is generally a complex process typically requiring input from a variety of technical disciplines, because these manufacturing processes often include many subtleties and incorporate many detailed interactions.
As an example, a typical data flow is shown in Figure 2. Here, raw data is obtained using an instrument (or more likely, a series of instruments) and is mathematically treated and then transformed to generate a process model. The term model can (and does) cover many things, including chemometrics to examine the analytical information, fluid dynamics to simulate particle flows, and dynamic simulations to describe the effects of changing behaviors of process, materials and equipment.
The basic components required for PAT, such as measurement devices and model methods, are widely used in other industries and are available for application in the pharmaceutical industry. Of course, they will need adaptation to be fully compliant with the prevailing regulatory regime, although the key is to be able to deploy them in a highly integrated manner. Integration is important since it confers a number of important advantages. For example, the cost of ownership is lowered, the data/information set is easier to deal with, and this will increase the rate at which the models can be developed and validated. This, in turn, lowers the cost to the manufacturer.
As an example, Figure 3 shows part of a production process and the data flows from left to right, i.e., from the process vessels to the manufacturing execution system. Attached to the process equipment are the analytical techniques that are at the core of the PAT initiative. However, it can be seen from this diagram that a series of data treatment steps, information management layers and control action functions are required. For example, the analytical equipment generally produce spectral data; these can be thought of as arrays of numbers that encode the chemical and physical information in the samples that are being analyzed. These arrays need to be mathematically treated in some way (using a range of techniques that fall under the umbrella term of chemometrics) in order to produce derived data, and they also need to be communicated effectively and stored properly within a process information management system (PIMS).
Although chemometric software programs are available, generally they were developed specifically for use in the laboratory environment (where they are still mainly deployed) and their extension to the manufacturing arena has proven to be difficult. This is not to say that they cannot be transferred, but rather that they must be adapted and this will inevitably take time to achieve before being accepted by a new set of personnel that inhabit the process arena. Even something seemingly as simple to source as PIMS is difficult in a PAT regime because most of these systems currently available have no concept of storing data as arrays of information; they are simply not designed to handle the complexity and the volume of data. Data mining techniques are also required to uncover the hidden interactions between the data and batch control records are required. Static and dynamic models also play a part and must be available. Control systems must also be provided, these may have actions that are taken manually (open loop) or automatically (closed loop), although the latter option is unlikely given the infancy of the PAT initiative. Of course, this whole array would need to be orchestrated, typically by a manufacturing execution system.
So in order for the full potential of PAT to be realized, manufacturers must supply a range of products and services that underpin the initiative and that include all of its detailed technology facets. Users may choose to deploy individual components and integrate themselves, but an important option is that the components should be available as an integrated suite. This minimizes the complexity of systems, their maintenance and extension while maximizing the benefit from integrating information from diverse physical processes,
geographical locations, manufacturing sources and lifecycle stages. //
Copyright 2004, ASTM International