A practical approach to PAT implementation

It has been estimated that reducing inventory levels across the pharmaceutical industry to those already achieved by the very best pharmaceutical manufacturing facilities could deliver a one-off cash release of $76 billion. A further reduction to inventory levels achieved in other manufacturing industries would be worth another $15 billion.¹ Process analytical technology (PAT) is a key enabling technology to deliver these savings.

During the last 20 years, the use of on-line spectroscopic techniques for process monitoring and control has become commonplace in the petroleum, petrochemical and chemical industries.² This is not true for the pharmaceutical sector, which has been slow to adopt on-line analysis and advanced process control techniques. Even though robust process analysers are available, the use of spectroscopic techniques such as Fourier transform infrared (FTIR), near infrared (NIR), mass spectrometry (MS) and Raman together with chemometric multivariate analysis tools remains primarily the domain of specialist analytical groups and isolated quality assurance (QA) laboratories in the pharmaceutical industry. Implementation on-line has been led by innovative groups of analysts and engineers who have recognized the potential of these techniques to improve understanding, help resolve existing manufacturing problems and decrease operating costs. Their successes are notable in view of the conservatism of the pharmaceutical industry as a whole, which has tended to act as a barrier to innovation and change, particularly in the manufacturing domain.

The climate for on-line analysis and advanced process control in the pharmaceutical industry is currently undergoing rapid and radical change. "Pharmaceutical cGMPs for the 21st Century: A Risk-Based Approach" and "Innovation and Continuous Improvement in the Pharmaceutical Industry" highlight the support of FDA for continuous improvement in pharmaceutical manufacturing.^3,4 At the same time, the requirement to reduce costs is becoming a high priority in Big Pharma boardrooms.⁵ The resistance to change of a traditionally conservative industry has been much reduced by this combination of a supportive regulatory environment and external cost pressures.

The PAT concept is completely aligned with the wider FDA goal of a science and risk-based approach to current good manufacturing practices (cGMPs). PAT is defined as "a system for designing, analysing and controlling manufacturing through timely measurements of critical quality and performance attributes of raw and in-process materials and processes with the goal of ensuring final product quality".⁶ As a direct consequence of the "cGMPs for the 21st Century" initiative, the pharmaceutical industry is experiencing pressure from the regulator to address concerns around limited process understanding, process inefficiencies and continuous process improvement through the adoption of PAT. This paper will describe four key elements of a PAT implementation:

• building a science-based knowledge base

• process monitoring and control

• validation of a PAT system

• implementing a PAT system — regulatory strategies.

Building a science-based knowledge base

The PAT guidance emphasizes the need to develop a deep understanding of the underlying scientific principles behind pharmaceutical manufacturing processes to determine the parameters critical to process and product quality. The knowledge base provided by the PAT approach is valuable in three main ways:

• It is a foundation for robust process and product design.

• It supports and justifies flexible regulatory paths for innovative new approaches.

• It facilitates continuous learning throughout the product life cycle.

The design of experiments, and the capture and evaluation of analytical measurement data are essential parts of building the knowledge base.

Knowledge capture is recommended at all stages of the product and process life cycle, beginning at the preformulation stage and extending through formulation, lab-scale production, scale-up, pilot plant and full-scale manufacturing (Figure 1).

Figure 1 A knowledge base is the foundation upon which a PAT system is based.

In R&D, the PAT approach is concerned with the capture and evaluation of information from research studies on synthesis characterization, stability and alternative dosage forms using new and conventional methods of analysis. Examples of relevant data include records in laboratory notebooks; measurements recorded by analytical equipment; the results of process modelling; and results from laboratory and pilot-scale experiments. Networked IT systems and data management and analysis software provide the means to store, access and mine the information in a practical and, crucially, traceable manner.

Data collected from scale-up production batches and manufacturing batches is added to the knowledge base as it becomes available during the product life cycle. Typically, this includes data on raw material attributes, process parameters, product quality attributes and environmental conditions, and with time, the full range of variability is represented. Examples of sources of variability that are often only fully revealed with time include

• Variations in the raw material supplier manufacturing processes that impact the chemical and physical attributes of the supplied materials.

• Time-based variations in manufacturing performance (e.g., between equipment maintenance events).

• Long-term equipment ageing and degradation effects.

• Effects linked to planned changes to equipment/analyser hardware and software.

• Individual ways of working (i.e., variations attributable to people).

• Changes in the local environment (e.g., temperature and humidity).

This type of PAT knowledge base is of most benefit when it provides a means to evaluate the applicability of the established multifactorial relationships between process and product attributes in different scenarios; for example, by enabling the prediction of the effect of different combinations of process attributes on product quality. This capability is underpinned by general developments in IT technology, which have led to the decrease in costs for hardware and software, and the increasing availability of sophisticated statistical analysis, data modelling and knowledge management applications. As a consequence, establishing and maintaining a knowledge base is well within the reach of all pharmaceutical companies.

Process monitoring and control

Key differences between current practice in pharmaceutical manufacturing and a PAT approach can be summarized as:

• The use of novel analytical technologies.

• The establishment of multifactorial relationships between materials, process and environmental conditions, and an understanding the consequences of these relationships for product quality and process robustness.

• The use of knowledge management tools.

Following a PAT approach, the understanding of the interaction between process and product is the basis for the design of the process monitoring, process control and QA strategies used in manufacturing.

It is important to realize that process analysers and PAT tools that are used only for monitoring incoming raw materials, process variables and process end-points do not constitute a PAT approach. Rather, PAT is an integrated approach in which the results obtained from the real-time analysis of critical process control points are used to control the process in some way. During manufacturing, process parameters are adjusted (within clearly defined limits) to produce the desired product quality attributes at the process end-point (this may mean a process unit operation or the process as a whole). The automation systems required for this level of process control are available today and are used extensively in the chemical and petrochemical industries. Figure 2 shows an example of the degree to which the integration of process control sensors has already been realized. Indeed, the same automation systems are already in use in the pharmaceutical industry, but they are rarely integrated with process analysers. Data management is also extremely fragmented, with individual storage, analysis and review tools for each process analyser, preventing the evaluation of trends, and cause and effect relationships.

Figure 2 Advanced extended process control systems are available and can be integrated with process analysers to realize integrated PAT systems for manufacturing.

Another outcome of using a PAT approach in manufacturing is a move away from end product quality control to a real-time QA approach, which gives rise to a change in the definition of quality decision criteria, and this includes acceptance criteria. The most important differences are:

• A move away from time-based end-points towards the attainment of defined quality attributes.

• The use of rigorous statistical principles to define acceptance criteria for product attributes based on the nature of the test and the sample size.

Figure 3 shows the installation of a NIR process analyser probe in a fluid bed dryer. The drying time is controlled by the measurement of the moisture content of the product, moving away from a time-based end-point. Taking the PAT approach a stage further, the dryer operating parameters can be controlled within predefined process limits to ensure that the dryer is operating in the most efficient drying regime. With this approach, drying cycle times can be reduced by 25–50% while delivering product with a more repeatable particle size distribution, as well as moisture content, to the next processing step.

Figure 3 NIR probe mounted in a fluid bed dryer used to monitor and control the drying process (photo courtesy of Glatt Air Technologies).

An example of the use of rigorous statistical principles is the real-time release (RTR) of final product based on on-line or in-line analysis. The PAT guidance indicates that on-line analytical methods for real-time release should be managed as an alternative analytical procedure for product release according to existing guidelines and regulations, and a number of pharmaceutical companies are investigating or have already submitted NIR-based methods for release testing to FDA. It follows that batch records will also need to be reported using methods of evaluation that are appropriate for this type of data; that is, as acceptance ranges, confidence intervals and distribution plots showing variation between and within batches.

Validation of a PAT system

FDA encourages the use of the PAT approach for new drug applications (NDAs) and for approved products. In the case of approved products, data can be collected during manufacturing from process sensors and on-line analysers, and evaluated using PAT tools to establish the scientific basis for moving to a PAT approach. The PAT guidance indicates a variety of risk-based approaches to managing the introduction of on-line analysers into existing processes with the aim of minimizing the regulatory burden for the manufacturer and encouraging innovation. The guidance clearly states that data collected in this way from existing processes, for the purpose of improving process understanding, would be considered research data and would not normally be relevant for regulatory inspections.

To demonstrate the validity of the in-process measurement results and process monitoring and control systems, the elements of the PAT system will need to be validated according to the appropriate GxP regulations. Existing FDA guidelines on computer systems validation and software validation are applicable to PAT systems.⁷ The International Society for Pharmaceutical Engineering's (ISPE) good automated manufacturing process (GAMP) guide for the validation of automated systems also provides relevant information on good practice.⁸ ISPE is setting up a Community of Practice specifically to consider PAT systems. The validation plan for a PAT system will typically include the validation of process analyser hardware and software; software packages for data analysis; process control software; and IT systems for the management, storage and backup of results.

In comparison with laboratory-based methods, analytical method validation for in-process analysers requires the consideration of a greater breadth of sources of error arising from the measurement system and on-line sampling issues: various approaches are required depending on the analytical technique and sampling method used. The PAT guideline highlights the value of using ASTM standards, citing various standards for the multivariate statistical analysis of analytical data, for standard practice for the comparison of test methods and for the on-line spectroscopic analysis of petroleum and petrochemical products in a process stream. The US Pharmacopeia includes general information on a number of spectroscopic techniques that are highly appropriate for the measurement of powders, intact tablets and solid samples on-line; for example, NIR spectroscopy. However, the analyst must currently rely on the scientific literature for guidance with respect to on-line implementation and appropriate statistical measures for system performance. Currently, there is no standardized approach and in each case the validation approach must be justified based on scientific principles and solid statistics. The ASTM E55 committee has been established to provide further guidance on good practice in this area.⁹

The PAT guidance specifically discusses novel analytical techniques that measure and evaluate a process signature, which would be the characteristic response of the system to a given stimulus. Acoustic spectroscopy is perhaps the most familiar of these techniques. In such cases, comparison with the analytical method to a conventional compendial method is often difficult and even inappropriate, and the PAT guidance clearly indicates that test-to-test comparisons are not mandatory in such circumstances. The regulatory agency is prepared to review the use of such techniques based on a comprehensive statistical and risk analysis supported by additional information such as a mechanistic explanation of causal links between the measurement results and observed process variability. The predictive ability of the established correlation functions and multifactorial relationships is considered a key indicator of process understanding in such cases.

In contrast to the increased requirement for analyser and IT system validation and more sophisticated statistical approaches for analytical method validation, it can well be imagined that the current practice of '3 batch' process validation will be rendered superfluous by the PAT approach. Using a PAT approach, the process will be continuously validated in real-time through control of the process critical control points and process critical end points.

Implementing a PAT system — regulatory strategies

To facilitate innovation and encourage the adoption of the PAT approach, FDA has consciously sought to remove the perception that technological innovation inevitably carries a heavy regulatory burden. PAT is a joint initiative of the Center for Drug Evaluation and Research (CDER), Office of Regulatory Affairs (ORA) and the Center for Veterinary Medicine (CVM) within the "cGMPs for the 21st Century" framework. A PAT policy development team of four subject matter experts has been established to work with industry to facilitate discussion on proposed PAT approaches at an early stage and support FDA's science and risk-based approach to PAT. Additionally, a PAT review and inspection team has been established, which includes reviewers, compliance officers and inspectors who have been trained and certified in the PAT approach.

The PAT guidance emphasizes that PAT implementation plans should be risk-based and the risk is regarded as being significantly lower when the process is well understood. This is considered to be the case when:

• All critical sources of variability are identified and explained.

• Variability is managed by the process.

• Product quality attributes can be accurately and reliably predicted.

The regulatory options available for implementing PAT are shown in Figure 4 and range from introduction under the facility quality management system (minor change) to implementation under existing scale-up and post-approval changes (SUPAC) regulations (prior approval supplement for major change).¹⁰ For NDAs, the comparability protocol approach offers flexibility to move to a PAT approach once the manufacturing process has been implemented and additional data from on-line analysis has been collected and analysed.¹¹ FDA's PAT team expects the owner of the PAT system to assess the most appropriate approach and be able to justify the decision made.

Figure 4 FDA regulatory paths for the implementation of a PAT system for new and existing NDAs and ANDAs as described in the draft PAT guidance.

FDA is also aware that initially there will be uncertainty about which regulatory path is appropriate and recommends discussion of the options with the PAT policy development team. This degree of intimacy with the regulator will be considered an impossibility by many regulatory affairs departments, but is a genuine attempt by FDA to embrace technological innovation and realize the manufacturing efficiency advantages offered by novel analysers, PAT tools and advanced process control techniques, as already achieved in other industries.

Conclusion

The aim of a PAT approach is to implement robust processes that are flexible enough to accommodate a defined level of variability in process materials (physical and chemical attributes) through adjustment of the process conditions.

A knowledge base created through the collection, analysis and evaluation of research, development and manufacturing data facilitates the design and implementation of a PAT system. The knowledge base also provides the justification for a science and risk-based approach to analytical method validation and process monitoring and control.

This opportunity for dialogue with the PAT policy development team, and the dedicated PAT review and inspection team, represents a dramatic shift in attitude to technological innovation, including process analysers, within FDA. The interest in the efficiency and cost structure of the pharmaceutical industry within the political establishment in the US and other countries is also a wake up call to a manufacturing community that has fallen behind other industries in its use of advanced process control systems and on-line measurement techniques for continuous QA. It can be speculated that without voluntary movement towards improved process understanding, and associated improvements in process efficiency and reductions in operating costs, further pressure will be brought to bear.

Jennifer Methfessel is principal consultant at ABB Engineering Services, UK.

Jean-René Roy is PAT lead competence centre manager at ABB Bomen, Canada.

References

1. R.S. Benson and J.D.J. McCabe, Pharm. Eng. July/Aug 2004.

2. J.M. Chalmers and P.R. Grffiths (Eds.), Handbook of Vibrational Spectroscopy (John Wiley and Sons, Chichester, UK, 2002).

3. www.fda.gov/oc/guidance/gmp.html

4. www.fda.gov/cder/gmp/gmp2004/manufSciWP.pdf

5. An Overdose of Bad News, The Economist, 17 March 2005.

6. www.fda.gov/cder/guidance/index.htm

7. General Principles of Software Validation; Final Guidance for Industry and FDA Staff, US Food and Drug Administration (CDER & CBER), January 2002.

8. Good Automated Manufacturing Practice (GAMP) Guide 4.0, International Society for Pharmaceutical Engineering (ISPE), December 2003.

9. I. Clegg, ASTM Standardization News 32(5), 28–31 (2004).

10. www.fda.gov/cder/guidance/index.htm

11. www.fda.gov/cber/gdlns/cmprprot.htm