Process Analytical Technology Concepts and Principles

Process Analytical
Technology
Concepts and Principles
Mark L. Balboni
Process analytical technologies are a
multifaceted group of concepts that
comprise one of many initiatives that form
FDA’s “Pharmaceutical GMPs for the 21st
Century.”Yet what constitutes a process
analytical technology or whether it is in fact
a new technology remains unclear. This
article outlines the key concepts that define
process analytical technology and
emphasizes the relevant theory and
applications of chemometrics.
PHOTODISC, INC.
Mark L. Balboni is a senior compliance
consultant at KMI, a division of PAREXEL
International, LLC, 28241 Crown Valley
Parkway, F601, Laguna Niguel, CA 92677,
tel/fax 949.830.2355, mbalboni@belmont.
kminc.com.
The current process analytical technology (PAT) initiative
underway within FDA exemplifies the latest consortium
between FDA and the industry that aims to encourage
the concepts of quality by design, use of computerized
data gathering and evaluation techniques, and process- and
product-monitoring methods through advanced instrumentation
and data evaluation. Although this partnership between
FDA and the industry is relatively new (2001), methods related
to PAT such as chemometrics have been studied and have been
in use for quite some time. Yet, the PAT initiative has raised several
questions: What does PAT really encompass? Is it a new
technology or is it a series of proven scientific principles? How
can PAT be used in a pharmaceutical operation to gain better
process understanding and possibly reduce cycle times and associated
costs?
This article discusses the concepts that embody PAT. Emphasis
is placed on chemometrics, which is the use of mathematical
and statistical models to extract and interpret chemical
data.
What is PAT?
FDA defines PAT as
● a system for the analysis and control of manufacturing
processes based on timely measurements of critical quality
parameters and performance attributes of raw materials and
in-process materials
● a process to ensure acceptable end-product quality at the completion
of the processing (1).
FDA also states that PAT involves
● the optimal application of process analytical chemistry (PAC)
tools
● feedback process-control strategies
● information management tools and/or product–process optimization
strategies for the manufacture of pharmaceuticals
(1).
In summary, PAT can be defined as the optimal application of
PAC tools, feedback process-control strategies, information
management tools, and/or product–process optimization strategies
to the manufacture of pharmaceuticals.
PAT focuses on the principles of building quality into the
product and process as well as continuous process improvement.
A few examples of PAT tools and strategies are as follows:
54 Pharmaceutical Technology OCTOBER 2003 www.pharmtech.com

● at-line, in-line, or on-line measurement of process quality
and performance attributes using a variety of instrumentation
and measurement strategies such as near-infrared (NIR),
vibrational, acoustical, and X-ray spectroscopy
● chemometric approaches such as multivariate statistical and
pattern recognition methods.
● real-time data and information management systems for
process control (2).
FDA’s PAT initiative is supported by many large pharmaceutical
corporations and distinguished members of academia.
Another potential advantage of PAT is the opportunity to
place more reliance on in-process testing as the basis for final
product release. This type of product-release methodology requires
both a significant amount of data to be compiled and
heavy correlations to be determined by analysis and evaluation
(3). The theory is that product release could be based on relationships
(i.e., correlations between observed in-process test results
and predictive qualitative results of the final product) established
during product–process development and confirmed
by both validation and routine review of product–process data
for commercial lots. These relationships coupled with confirmation
testing of the finished product would serve as the basis
for release, or the product could possibly be released on the
basis of the observed in-process results and how a currently
produced lot favorably compares with other previously released
product lots.
FDA’s PAT initiative is being spearheaded by the Center for
Drug Evaluation and Research (CDER), Office of Pharmaceutical
Science. Since 2001, FDA has held and participated in a series
of PAT meetings as part of the Process Analytical Technologies
Subcommittee for the Office of Pharmaceutical Sciences
and with the FDA Science Board. These meetings have included
not only FDA personnel but also pharmaceutical industry representatives,
members of academia, and engineering and consulting
professionals.
The PAT initiative is part of a larger FDA initiative called
“Pharmaceutical CGMPs for the 21st Century: A Risk-Based
Approach”(4). The agency seeks to improve the regulation of
pharmaceutical manufacturing using a science- and risk-based
approach to product-quality regulation while incorporating an
integrated quality-systems approach.
Complete descriptions of FDA’s PAT initiative and the “Pharmaceutical
CGMPs for the 21st Century” approach, including
electronic links to associated documents and reference materials,
are provided in References 1 and 4.
PAC and in/at/on-line monitoring
According to Hailey et al., PAC is the technique of “gathering
analytical information in real time at the point of manufacture”
(5). As they noted, PAC “places an emphasis on the process
rather than the final product,” including “an understanding of
the relationship between final product specification [sic] and
the critical variables during the manufacturing process.”
Although the approaches and instrumentation currently
being discussed are in some cases categorized as being novel or
new, real-time measurement (PAC) has existed for some time
(e.g., real-time temperature monitoring of reaction vessels during
the synthesis of active pharmaceutical ingredients). One
particular industry–university partnership, The Center for
Process Analytical Chemistry (CPAC) at the University of Washington,
has been in existence since at least 1986 (6).
Although PAC is not a new approach, many of the related
techniques have been tested and used only on a limited basis
by a very small percentage of the pharmaceutical industry. The
following is a partial list of the various sensors and instrumentation
recently discussed at the 2003 Arden House Conference
either in-use or currently being evaluated for feasible use for
production monitoring:
● NIR spectroscopy for moisture determination
● X-ray spectroscopy
● radio frequency for moisture determination
● microwaves for moisture determination
● RAMAN spectroscopy, with vibrational spectroscopy being
the most common. RAMAN complements IR spectroscopy
and is used for raw-material identification, polymorph differentiation,
and reaction monitoring.
● fluorescence for water quality
● on-line measurement of color
● X-ray fluorescence for the detection of inorganic materials
● photoacoustic spectroscopy.
Because most of these technologies are extremely sophisticated,
one must realize that the key emphasis of PAT is not so
much how to collect the data or what kind of instrumentation
should be used, but rather what data should be collected, what
is done with these data, and what associated conclusions are
reached. Therefore, a complete and thorough understanding of
the manufacturing process is paramount.
Chemometrics
To fully understand PAT, one must first understand the science
behind manufacturing processes, including how these processes
operate, their limitations, and their expected outcomes.
In 1974 Sante Wold, from the Institute of Chemistry at Umea
University (Umea, Sweden) described “the art of extracting
chemically relevant information from data provided in chemical
experiments” as chemometrics (7). He stated that this art is
“heavily dependent on the use of different kinds of mathematical
models” and that it was important to have knowledge in
statistics, numerical analysis, and applied mathematics, including
the challenge to “structure the chemical problem to a form that
can be expressed in a mathematical relation.”
Twenty years later, during a 1994 presentation, Professor Wold
said his definition of chemometrics had not changed much. He
said chemometrics requires us to ask ourselves “How do we get
chemically relevant information out of measured chemical data;
how do we represent and display this information; and, how do
we get such information into data?”
In 1996, Wise and Gallagher stated chemometrics “is the science
of relating measurements made on a chemical system to
the state of the system via application of mathematical or statistical
methods” (8). Similarly, Hardy noted that data are “raw
information, both qualitative and quantitative” (9). He observed
that in and of themselves, “raw data are meaningless” and that
a method of “analysis and a model” are needed to “gain knowl-
56 Pharmaceutical Technology OCTOBER 2003 www.pharmtech.com

edge” from the data. He also identified the clear need for a
process to “convert data to knowledge.”
Dosage forms recently have been described as “complex
multifactorial physiochemical systems” (10), sometimes referred
to as multivariate. Multivariate analysis “consists of
methods of statistical analysis of multivariate data, characterized
as consisting of several observations on each set of objects
or mathematically represented by a collection of points
in a finite-dimensional Euclidean space R P ” (11). So, multivariate
analysis is an important statistical method widely used
by chemometricians.
Described below are a few of the tools commonly used by
chemometricians, all of which are helpful for evaluating processing
data generated during pharmaceutical unit operations
or when synthesizing active pharmaceutical ingredients.
Principle component analysis (PCA). PCA is a technique used to
express variations of many variables by a small number of indicies
(11). Wise and Gallagher describe PCA as a favorite tool
of chemometricians for data compression and information extraction
and note that “PCA finds combinations of variables or
factors that describe major trends in a data set” (8). Sans et al.
(12) observed that PCA can be used to determine the number
of components that influence the data of the process as well as
to identify the chemical meaning of the components. They proposed
that PCA enables one “to approach multivariate chemical
problems in order to obtain underlying information about
the correlation of the raw data and the real meaning interpretation.”
Finally, Wise and Gallagher (8) acknowledged that “generally
there is a great deal of correlated or redundant information
in process measurements,” and “often essential information
lies not in any individual process variable but how the variables
change with respect to one another” or how they co-vary. They
also noted that “some sort of signal averaging” would be useful
in cases in which a large amount of noise is created from the
volume of data and a lack of clear understanding of the data
exists.
However, Wu et al. point out two limitations of PCA:
● When an object is added or removed as displayed in a plane,
principal components (PCs) must be recalculated all over
again by following the process of selection and interpretation
of the PCs.
● No more than two PCs at a time can be viewed (inspected)
in a plane, and this prevents one from using the information
contained in other PCs (13).
The authors also note that their “star-plot” method could be
used as an alternative way to display and analyze multivariate
data.
Application of PCA to chemical processes. Wise and Gallagher (8)
studied data obtained from a slurry-fed ceramic reactor using
thermocouples placed at 20 locations (8). They found a great
deal of correlation as the data generated followed a sawtooth
pattern. In addition, the study revealed the following:
● PCA performed on this data found three factors that captured
approximately 97% of the variance in the data set.
● This previously noted finding allowed 16 variables to be replaced
with three new ones, which were linear combinations
of the original variables.
● The sawtooth pattern was attributed to changes in the level
of molten glass, which was a controlled variable.
● The three factors (PCs) were identified as the level of molten
glass in the reactor, the variation between two groups of measured
locations, and the variation of overall process temperature
(also a controlled variable).
Sans et al. also used PCA to determine stoichiometric models
from on-line spectroscopy for selecting the number of reactions
and the number of chemical absorbing species to better
describe a chemical reaction (12). They chose to use
spectroscopy methods because they can reveal information
about the dynamic evolution of the reaction mass during a
chemical reaction. Although the outcome of this study is not
discussed in this article, Sans et al. did note that “semibatch
processes are examples of complex reaction networks that generally
are difficult to interpret because of the large number of
reactions occurring simultaneously and because of the effects
related to the addition of materials that may cause complex volume
changes during the processing time.”
Multiway principle component analysis (MPCA). An analog of PCA
is what is known as multiway PCA, which is “equivalent to performing
PCA on a very large two-dimensional matrix formed by
unfolding the three-way array X into one of six possible ways,
only three of which are mathematically unique” (8). General PCA
methods do not take into account the ordered nature of the data
sets, meaning that the data were not collected in a sequential
manner. Multiway methods take into account the ordered (sequential)
nature regarding when data were generated “because
the data usually are organized into time-ordered blocks that are
each representative of a single sample or process run” (8).
Nomikos and MacGregor (14) describe another aspect of
MPCA; namely, the use of on-line measurements to identify
“significant deviations from the normal operating behavior by
using SPC charts.” An empirical model is based on data generated
when the process is operating within a state of control.
They noted, “future unusual events are detected by referencing
the measured process behavior against the ‘in-control’ model
and its statistical properties.”
Partial least squares (PLS). Wise and Gallagher described PLS
as a regression “related to both principle component regression
(regression of properties on the principle component scores of
the measured variable) and multiple linear regression (also
known as inverse least squares)” (8). They observed that this
analysis can be used to predict “properties of a system based on
variables that are only indirectly related to the property.” Accordingly,
by using PLS, one attempts to “find factors” that both
“capture variation and achieve correlation.”
Multiway partial least squares (MPLS). Nomikos and MacGregor
observed that MPCA using statistical process-control charts
“only makes use of the process variable trajectory measurements
(X) taken throughout the duration of the batch” (14). In contrast,“multiway
partial least squares (MPLS) can be performed
using both the process data (X) and the product quality data
(Y),” and “focuses more on the variance of X that is more predictive
for the product quality Y” (14).
Other statistical tools. Two other statistical tools that may be
useful in PAT efforts are
58 Pharmaceutical Technology OCTOBER 2003 www.pharmtech.com

● capability studies, which measure the ability of the process to
consistently meet specifications by evaluating select process
outputs and calculating the average and ranges over a specified
time on control charts. From these studies, capability indicies
Cp (used to evaluate the variation) and Cpk (used to
evaluate the centering of the process) are calculated.
● design of experiments, which are experiments that involve
changing one or more of the process inputs and measuring
the results to one or more of the process outputs (15).
Rapid microbiology test methods
Another part of FDA’s PAT initiative involves discussion
about the feasibility of developing so-called rapid microbiology
methods (emphasis added, as most microbiological testing
can take anywhere from several days to a couple of weeks
to complete). This topic is of heightened interest within the
industry as manufacturers seek additional ways to reduce cycle
times.
Rapid microbiology testing was discussed during the FDA
Advisory Committee for Pharmaceutical Science meeting on
23–24 October 2002 (16). During that meeting, the committee
identified at least two problems or risks associated with the development
of rapid microbiology methods; namely,
● validation of test methods may not yield results equal to those
for traditional test methods
● acceptance by regulators (e.g., FDA).
The group also categorized microbial determinations as follows:
● qualitative methods (presence or absence of microbes; e.g.,
sterility testing)
● quantitative methods (enumeration of microorganisms present;
e.g., microbial limits tests)
● microbial identification.
FDA did concede that rapid microbiology test methods are
an important part of the PAT initiative but that at this time it
has not yet been discussed extensively (16). FDA also commented
that the general guidance document on PAT would not
specify details about rapid microbiology methods but would
rather cover them in a general sense to encourage their use. Finally,
the group felt that microbial identification probably has
the most rapid method systems presently available, and quantitative
testing probably has the least detection systems.
Real-time data information and management systems
Another PAT tool being discussed is the use of real-time data
information and management systems. Although a detailed
overview of this topic is not provided in this article, some of
the possible concerns include possible 21 CFR Part 11 implications,
adding complexity to already complex computer system
architecture, and the amount of data retention necessary, given
that continuous monitoring will generate very large volumes
of data.
Circle/eINFO 45
60 Pharmaceutical Technology OCTOBER 2003 www.pharmtech.com

With FDA’s recent withdrawal of Part 11 guidance documents,
it would be advisable to first review the anticipated PAT guidance
document for information concerning data information
and management issues and secondly review the 21 CFR Part
11 implementation portion of the “Pharmaceutical CGMPs for
the 21st Century Initiative,” which includes both a Notice of
Availability and a draft guidance on Part 11. The draft guidance
on Part 11 attempts to clarify the scope and applicability of the
regulation, which ultimately may undergo revision to clarify
the scope and requirements.
Benefits and challenges: an industry perspective
Most industry representatives that either were involved in discussions
regarding the feasibility of PAT or who have conducted
successful PAT efforts thus far have had many more positive
than negative comments regarding the advantages of adopting
PAT principles. Positive perceived benefits of PAT include
● decrease in cycle times
● lower costs
● increased efficiency and batch-to-batch consistency
● process fingerprinting (signature) that would be useful for validation,
scale-up, and confirming acceptable handling of changes
● increased process understanding and a decrease in variability,
rejects, and lot failures
● possible continuous processing and the ability to adjust process
on the basis of real-time monitoring data.
Conversely, the most-common perceived or actual challenges
include
● product-approval delays by inclusion of PAT methodologies
into relatively traditional drug development and validation
activities
● lack of a written PAT guidance document from FDA
● an increase in the amount of data being generated, including
not only what one should do with the extra data but also possible
Part 11 implications
● increased pressures to meet aggressive filing timelines, added
costs to make changes, lack of senior management support,
and resource constraints (Source: Reference 1 and 2003 AAPS
Arden House Conference.
Where we go from here
Although no one can clearly predict how widely accepted PAT
will become, it is clear that FDA does support innovation. The
agency is appealing to the industry to take a more active role in
understanding its manufacturing processes and is seeking quick
and effective resolution of problems associated with good manufacturing
practices that result in rejected batches, stability failures,
field alerts, and product recalls.
Currently there are many actions a company can take to
prepare for what FDA is calling the “GMPs for the 21st Century.”
By reading the contents provided on FDA’s Web site regarding
PAT (http://www.fda.gov/cder/OPS/PAT.htm), which
Circle/eINFO 47
62 Pharmaceutical Technology OCTOBER 2003 www.pharmtech.com

includes committee meeting minutes and dozens of presentations,
one will see that many large pharmaceutical firms already
have made significant accomplishments through the
successful implementation of PAT principles. In spite of the
industry’s dwindling profit margins, fewer new products in
the pipeline, and a more competitive marketplace, these companies
already have recognized the positive aspects of implementing
PAT. The following are some active steps that a company
can take right now.
Develop a plan for the future.
● Wait for FDA’s much anticipated guidance document about
PAT before making significant decisions regarding how your
company will handle certain PAT principles (e.g., validation
of new analytical methods for a unique in-process test).
● Analyze existing product lines and determine which may benefit
most from PAT. For example, the compounding and filling
of a well-established product that is a true solution composed
of 99.5% water is probably not the best candidate for
PAT. The manufacture of complex dosage forms such as
tableted products, which also includes in-process monitoring
using process control charts, would probably be a better
choice. Products with recurring quality problems are also
good candidates because process deviations or exceptions are
sometimes results of a less-than-complete understanding of
the process rather than more-obvious causative factors.
● Obtain/retain employees with education, training, and experience
in disciplines necessary to be successful in PAT efforts.
For example, FDA is seeking to hire persons experienced in
in-process/chemical engineering, PAC, chemometrics, and
industrial pharmacy. For your company’s efforts to be successful,
similar expertise will be needed.
● Gain executive management’s support for PAT. Although it
may be easier to make the scientific case, management must
understand how PAT could reduce cycle times and costs and
add value. Make a business case for adopting PAT.
Benchmark with industry and academic partners. Speak with
company representatives from firms that already are using PAT
principles. Many major universities also are heavily involved in
the PAT initiative and could provide guidance, including, but
not limited to, the following schools that have had representatives
at several FDA committee meetings about PAT:
● MIT Pharmaceutical Manufacturing Initiative
● Purdue University, School of Pharmacy
● Center for Process Analytical Chemistry at the University of
Washington.
● Texas A&M Department of Chemistry and Chemical Engineering.
Gather information about the topic. A significant amount of information
concerning PAT can be found on FDA’s Web site (1).
Additional new information continues to be added, and plenty
of information already is available on many university and industry
association Web sites regarding topics such as PAT, PAC,
Circle/eINFO 49
64 Pharmaceutical Technology OCTOBER 2003 www.pharmtech.com

PCA, and chemometrics. Internet keyword searches can yield
significant additional information.
Attend future PAT workshops, seminars, or symposia. Find out
what FDA and others are saying first-hand. If the workshops
are sponsored by an industry association, you may need to attend
in person to get the handouts and presentations because
all of that information may not be published on FDA’s or the
association’s Web site.
Provide comments directly to FDA using the Internet. If you have
substantial comments regarding FDA’s PAT initiative and you
feel it is information that others could benefit from, consider
providing electronic comments to FDA. The agency already has
established a Web-based feedback tool for this purpose (www.
fda.gov/ohrms/dockets/). You will need to search for docket
number 03N-0059. Alternatively, you could send questions or
comments regarding the PAT initiative by email to
pat@cder.fda.gov.
Ensure quality is designed into your products and processes.
● Confirm compliance with good manufacturing practices. Remember,
CGMP is the minimum standard according to 21
CFR 211.1(a). Most of the successful companies strive for and
achieve substantial compliance with these regulations as well
as carry out “best practices.” Poorly developed manufacturing
processes, untrained employees, or equipment that has
not been properly qualified will hinder any PAT efforts.
● Complete thorough product and process development work
to ensure processes are adequately defined. Include robust
specifications and critical process-control points.
● Confirm that active drug substances are well characterized.
● Use validation efforts to demonstrate consistency and reproducibility,
not as a means to conduct range finding, process
adjustments, or enhancements. These are development tasks,
not validation.
● Optimize processes and improve yields after successful validation,
not during validation.
● Consider process or equipment changes carefully, especially
postapproval, because this usually will result in much unanticipated
work and undoubtedly more development activities
and data. Complete risk or impact assessments on the
basis of data and not on opinions or theories.
Many years ago, Alexander Hamilton said,
Experience teaches that men are often so much governed
by what they are accustomed to see and practice, that the
simplest and most obvious improvements, in the most
ordinary occupations, are adopted with hesitation, reluctance,
and by slow gradations. Men would resist
changes, so long as even bare support could be ensured
by an adherence to ancient courses, and perhaps even
longer.
Sometimes changes do take time, so we should look to the
future potential of PAT to enhance our pharmaceutical processes
Continued on page 254
Circle/eINFO 51
66 Pharmaceutical Technology OCTOBER 2003 www.pharmtech.com

Continued from page 66
and applaud FDA for encouraging the application of new technologies
and for their willingness to work with industry on this
important initiative.
Acknowledgment
The author would like to acknowledge the following individuals
for their assistance in the writing of this article: KMI senior
compliance consultant Eric S. Weilage and PAREXEL senior
biostatistician Chunzhang Wu, PhD.
References
1. CDER, Office of Pharmaceutical Sciences, “Process and Analytical
Technologies Initiative,” http://www.fda.gov/cder/OPS/ PAT.htm.
2. A. Hussain and J. Famulare, “FDA’s PAT Initiative and its Role in the
Proposed Drug Quality System for the 21st Century,” presented at the
AAPS Arden House Conference, 27 January 2003.
3. T. Layloff, “Process Analytical Technology (PAT) Subcommittee Report,”
presented at the ACPS meeting 21 October 2002.
4. FDA,“Pharmaceutical CGMPs for the 21st Century: A Risk-Based Approach,”
http://www.fda.gov/cder/gmp/index.htm.
5. P. Hailey et al.,“Automated System for the On-line Monitoring of Powder
Blending Processes Using Near-Infrared Spectroscopy Part I. System
Development and Control,” J. Pharma. Biomed. Analysis 14,
551–559 (1996).
6. D. Illman,“CPAC: An Industry–University Cooperative Research Center
for Process Analytical Chemistry,” TrAC: Trends in Analytical Chemistry
5 (7), 164–172 (1986).
7. S. Wold,“Chemometrics: What Do We Mean with It and What Do We
Want from It?” presented at InCINC ’94, Institute of Chemistry, Umea
University (Umea, Sweden, 1994).
8. B. Wise and N. Gallager, “The Process Chemometrics Approach to
Process Monitoring and Fault Inspection,” J. Proc. Ctrl. 6 (6), 329–348
(1996).
9. J. Hardy, “Special Topics: Chemometrics,” lecture presentation associated
with 3150: 710 (University of Akron, 2000), available at
http://ull.chemistry.uakron.edu/chemometrics/introduction.
10. FDA,“Emerging Issues in Pharmaceutical Manufacturing: Process Analytical
Technology,” science board meeting presentation (16 November
2001).
11. N. Sugakkai,“Iwanami Sugaku Ziten,” original publication by Iwanami
Shoten Publishers (Tokyo, Japan, 1954), copyright by Nihon Sugakkai
(Mathematics Society of Japan); English translation provided by the
Massachusetts Institute of Technology (1977).
12. D. Sans, R. Nomen, and J. Sempere, “Interactive Self-Modelling of
Chemical Reaction System Using Multivariate Data Analysis,” supplement
to Comput. Chem. Eng. 21, S631–S636 (1997).
13. W. Wu et al.,“The Star-Plot: an Alternative Display Method for Multivariate
Data in the Analysis of Food and Drugs,” J. Pharma. Biomed.
Analysis 17 (6-7), 1001–1013 (September 1998).
14. P. Nomikos and J. MacGregor,“Multiway Partial Least Squares in Monitoring
Batch Processes,” Chemometrics and Intelligent Laboratory Systems
30, 97–108 (1995).
15. FDA Global Harmonization Task Force Study Group #3, draft Process
Validation Guidance (1 June 1998).
16. FDA Advisory Committee for Pharmaceutical Science transcripts, 23
October 2002. PT
Jump to
254