You are on page 1of 18

Anal Bioanal Chem (2010) 398:137–154

DOI 10.1007/s00216-010-3781-x

REVIEW

Process analytical technology (PAT) for biopharmaceutical


products
A. S. Rathore & R. Bhambure & V. Ghare

Received: 26 February 2010 / Revised: 20 April 2010 / Accepted: 23 April 2010 / Published online: 18 May 2010
# Springer-Verlag 2010

Abstract The “Pharmaceutical Current Good Manufactur- Regulatory framework


ing Practices (CGMPs) for the 21st Century—A Risk Based
Approach” initiative announced by the FDA in August Process analytical technology (PAT) is a key element of
2002 to improve and modernize pharmaceutical manufac- the “Pharmaceutical Current Good Manufacturing Practi-
turing facilitated adoption of process analytical technology ces (CGMPs) for the 21st Century—a Risk Based
(PAT) by the pharmaceutical industry. The potential for Approach” initiative announced by the FDA in August
improved operational control and compliance resulting 2002 to improve and modernize pharmaceutical manu-
from continuous real-time quality assurance was highlight- facturing [1]. The PAT initiative was first proposed by the
ed as a likely benefit that would result from PAT United States Food and Drug Administration’s (FDA)
implementation. A considerable amount of work has been Center for Drug Evaluation and Research (CDER) with
done on this topic by academic and industrial contributors the objective of achieving significant health and economic
in the last decade. In this paper, we will start with a brief benefits by application of modern process control and tests
overview of evolution of PAT concepts and a review of in pharmaceutical manufacturing [2, 3]. Shortly thereafter,
their application in the wider pharmaceutical industry. The with an endorsement from the FDA’s Science Board, the
rest of the paper focuses on PAT applications for biotech PAT subcommittee under the Advisory Committee for
processes with emphasis on developments in the last five Pharmaceutical Science (ACPS) was formed in November
years. It is our observation that while significant advances 2001 and consisted of the following four working groups
have been accomplished with regard to our ability to with representatives from the FDA, industrial experts, and
analyze/monitor key process and quality attributes in the academic representatives [3–5]
biotech industry, much more needs to be done with regard
& PAT applications working group;
to utilizing the collected data for subsequent control of the
& PAT products and the process development working
process, to achieve optimum yield and product quality. The
group;
latter is necessary to achieve the benefits that will result
& PAT process and analytical validation working group;
from PAT implementation.
and
& PAT chemometric working group.
Keywords Process analytical technology, PAT .
Bioprocessing . Quality by design The key objectives of the ACPS subcommittee were to
[6]:
& Identify and define the technology and regulatory
A. S. Rathore (*) : R. Bhambure : V. Ghare hurdles, possible solutions, and strategies for successful
Department of Chemical Engineering, implementation of PAT in pharmaceutical development
Indian Institute of Technology, and manufacturing;
Hauz Khas, New Delhi 110016, India
e-mail: asrathore@biotechcmz.com & Discuss the general principles of the regulatory appli-
Url: www.biotechcmz.com cation of process analytical technology including the
138 A.S. Rathore et al.

principles of the method validation, specifications, use, at the end of the manufacturing process. A process is
and validation of the chemometric tools; and generally considered well understood when:
& Discuss the need for general FDA guidance to facilitate
1. All critical sources of variability are identified and
the implementation of the PAT.
explained;
Based on the comments and recommendations suggested 2. Variability is managed by the process; and
by the ACPS PAT subcommittee, the FDA announced the 3. Product-quality attributes can be accurately and reliably
draft version of the PAT guidance “Guidance for Industry predicted over the design space established for materi-
PAT—A Framework for Innovative Pharmaceutical Manu- als used, process conditions, manufacturing, environ-
facturing and Quality Assurance” in September 2003 [7]. mental, and other conditions.
The final version of the PAT Guidance was released in
The objective for PAT implementation could be one or
September 2004 with the help of the Center for the
more of the following [8, 15–19]:
Veterinary Medicine (CVM) and the Office of Regulatory
Affairs (ORA) [8]. The guidance was the result of & Better process understanding
collaborative work amongst several FDA offices includ- & Improved yield because of prevention of the scrap,
ing the CDER, CVM, and ORA. The underlying vision rejects, and reprocessing
was that of utilizing innovation, fundamental scientific & Reduction in the production cycle time by using on-
and engineering knowledge, and quality-management line/at-line or in-line measurements and control
principles for efficient pharmaceutical manufacturing. & Decrease in the energy consumption and improved
The PAT initiative has also been supported by the efficiency from conversion of the batch process into a
European Medicines Agency (EMA), the Japanese continuous process
Ministry of Health, Labor, and Welfare (MHLW), and & Cost reduction because of reduced waste and reduced
by the International Conference on Harmonization (ICH) energy consumption
guidelines [9]. & Real-time release of the batches
The more recently launched initiative “Quality by
From an implementation perspective, perhaps, PAT can
Design” (QbD) is also aligned with the PAT concepts [9–
be visualized as the three-step process illustrated in Fig. 1
14]. Quality by Design has been defined in the ICH Q8
[18, 19]. The design phase starts early in process
guideline as “a systematic approach to development that
development when the given unit operation is being
begins with predefined objectives and emphasizes product
designed and then later optimized and characterized [12,
and process understanding and process control, based on
20]. In this phase, the critical quality attributes (CQA) that
sound science and quality risk management” [9]. The
are being affected by the process step are identified along
approach involves identification of the product attributes
with the critical process parameters (CPP) that have been
that are of significant importance to the product’s safety
determined to affect the CQA. This process understanding
and/or efficacy, design of the process to deliver these
is the essence of PAT and critical for the next two phases. In
attributes, a robust control strategy to ensure consistent
process performance, validation and filing of the process
demonstrating the effectiveness of the control strategy, and, ANALYZE
finally, ongoing monitoring to ensure robust process Ability to
analyze for the
performance over the life cycle of the product. PAT plays pertinent CQA
a key role in creation of a robust control strategy for the as well as
process and for ongoing process monitoring [11]. DESIGN to monitor
Knowing the the CPP
CQA pertinent
for the process
step and the
What is PAT? CPP that impact
them
Process analytical technology (PAT) has been defined as “a CONTROL
Using process
system for designing, analyzing, and controlling manufac- understanding to
turing through timely measurements (i.e., during process- control CPP so as
ing) of critical quality and performance attributes of raw to achieve
consistent
and in-process materials and processes, with the goal of CQA
ensuring final product quality” [8]. A desired goal of the
PAT framework is to design and develop well-understood Fig. 1 The three steps that must be taken for PAT implementation,
processes that will consistently ensure a predefined quality and the objective(s) of each step
Process analytical technology (PAT) for biopharmaceutical products 139

the analyze phase, a suitable analyzer is identified for of the product and separation from HCP, which is a CQA.
monitoring of the CQA and the CPP. PAT application can Thus, use of UV absorbance based pooling in this case is a
be at-line (sample removed, isolated from and analyzed PAT application. Figure 2b illustrates a case where use of the
close to process stream), on-line (sample removed for same analytical tool (UV absorbance at 280 nm) is not a PAT
analysis from process stream and returned to process stream application. In this case also the column separates the
again), in-line (sample not removed but analyzed in place), product from HCP. As mentioned above, the UV absorbance
and off-line (sample removed and analyzed away from at 280 nm cannot differentiate between the protein and HCP
process stream) [8, 18, 19]. For a PAT application, it is and, because the two species are not baseline separated,
necessary for the analytical results to be available in the implementation of this pooling scheme is likely to result in
time-frame necessary to facilitate real-time decision mak- inconsistent product quality with respect to the CQA
ing. Finally, the control phase involves designing a control (variable HCP levels in the pool). Hence, unlike the case
scheme based on the process understanding such that the illustrated in Fig. 2a, use of UV absorbance at 280 nm is not
data from the analyzer can be utilized for making real-time the correct tool for PAT application for this case.
process decisions, and consistent process performance and
product quality can be achieved.
PAT is more than just analysis [18, 19]. This is further PAT applications in the chemical industry
illustrated in Fig. 2, which shows pooling of fractions from
a process chromatography column on the basis of UV The concept of PAT has been applied in the chemical
absorbance at 280 nm, an application that is very industry for several decades and has been the subject of
commonly utilized in biotech processing. Figure 2a illus- several reviews and books [21–29]. Over this time, the
trates an example of what could be a PAT application. In chemical industry has moved towards implementation of
this case, the process chromatographic column is separating process analyzers and process modeling approaches that
the product from host-cell proteins (HCP). The two species allow not only optimization of productivity and product
are well resolved on the column, as is evident from the quality but also provide real-time assurance that the process
baseline separation. Thus, even though UV absorbance at is in control and in case of any deviations, suggest steps that
280 nm cannot distinguish between the product and HCP, would bring the process back in control. Table 1 presents a
the baseline separation enables its use for consistent pooling summary of some of the recent applications that have been
published on this topic [30–36]. Near infrared (NIR)
A spectroscopy, acoustic sensor, nuclear magnetic resonance
UV Absorbance at 280 nm

Product (NMR), Raman spectroscopy, attenuated total reflectance


(ATR), and Fourier transform infrared spectroscopy (FTIR)
have been used as process analyzers in these applications.
Process component analysis (PCA), partial least square
(PLS), and soft independent modeling of the class analogy
Host cell Pooling cuts based (SIMCA) are some of the statistical approaches that have
proteins
p on UV absorbance been used to enable analysis and modeling of the abundant
data that is generally provided by the above mentioned
process analyzers. Near infrared (NIR) and mid infrared
Process time (MIR) spectroscopy have been proposed as tools for
predicting quality of diesel/biodiesel blends (density, sulfur
UV Absorbance at 280 nm

Product
B content, and distillation temperatures) and the performance
has been shown to match that of conventional approaches
that are much more cumbersome with regard to analysis and
portability [30]. NIR has also been used as an analyzer to
Product
related
determine the effect of several operating conditions on
impurity Pooling cuts based recovery, selectivity, and productivity for production of
on UV absorbance
methyl isobutyl ketone (MIBK) [31]. Use of this PAT
approach enabled the authors to perform the necessary
Process time experiments in a time-efficient fashion and resulted in 30%
improved productivity of MIBK. NIR, in combination with a
Fig. 2 Use of UV absorbance at 280 nm for pooling of fractions from
a process chromatography column separating a protein product from
suitable statistical tool, has also been used in other
other host cell proteins. a. Baseline separation of product from HCP. applications such as raw material analysis [33], measurement
b. Incomplete separation of product from HCP of product quality [34], and process monitoring [36].
140 A.S. Rathore et al.

Table 1 Examples of PAT applications in the chemical industry

Application Process analyzer Statistical tool Observation

Prediction of the properties Near infrared (NIR) and mid Partial least-squares (PLS) NIR and MIR spectra have been shown to
of the diesel or biodiesel infrared (MIR) spectroscopy calibration help in predicting distillation temperature
and sulfur content of diesel or biodiesel [30]
Catalysis reaction involving In-line NIR Design of experiments PAT application helped in determination of
conversion of acetone to (DOE), Principle- the factors affecting the productivity,
methyl isobutyl ketone components analysis (PCA), selectivity, and yield of the MIBK and
(MIBK) PLS, and cluster analysis thereby leading to improved productivity
for MIBK [31]
Industrial process for Acoustic sensor with high- PCA and PLS PAT approach used to predict fluidization
granulation of urea during temperature microphone probe airflow, reflux of fines to the reactor,
fertilizer production granule moisture content, and granule
size [32]
Raw material identification NIR K nearest neighbor (KNN) Fast and cost-effective method for raw
and quality control and soft independent material analysis [33].
modeling of the class
analogy (SIMCA)
Analysis of the organic Nuclear magnetic resonance PLS Less time-consuming and the cost-effective
content of waste water (NMR) spectroscopy method for analysis of organic content [33]
Simultaneous determination NIR PLS Non-destructive and non-polluting method
of methanol and ethanol of analysis enables faster detection of
in gasoline methanol adulteration of the
gasoline [34]
Simultaneous monitoring of Raman spectroscopy and PLS PAT approach was utilized to understand
solute concentration and attenuated total reflectance how the feeding strategy for the reactant
polymorphic state of the (ATR) Fourier transform affects the polymorph composition of
crystal infrared (FTIR) spectroscopy L-glutamic acid [35]

On-line determination and NIR PLS with distributed On-line process control led to significant
control of the water content control system improvement in yield [36]
of a continuous conversion
reactor

PAT applications in the pharmaceutical industry of PAT in the pharmaceutical industry [18, 19]. The key
feature of PAT is that quality is built into the product, rather
Innovations in the process analytical chemistry (process than being tested before release of product [8]. The PAT
analyzers) and in our ability to capture and analyze large framework comprises risk management, at/on-line sensors
amounts of data have served as the key drivers for adoption that assist in monitoring/controlling/designing of the pro-

Fig. 3 The different unit


operations that comprise a
typical pharmaceutical process. NIR T NIR PSD NIR
Each step can potentially benefit
from implementation of one or
more PAT applications. Adapted
from Ref. [17]. Copyright Dispensing Blending Unit Op 1 Milling Blending Compressing Coating
permission from Advanstar
Communications. PSD: Particle
size distribution; T: Tempera-
ture; DW: Dry weight; PI: DW PI DW PI PI PI DW PI DW PI
Product impurity

Raw Material NIR Temperature Control Particle Size Distribution Average Tablet Weight
20.5˚C

20.3˚C

20.1˚C

19.9˚C

19.7˚C
Process analytical technology (PAT) for biopharmaceutical products 141

cess and prediction of process performance [37]. A variety reliable testing of raw material quality such that only those
of analytical techniques have been used in the pharmaceu- raw material lots that meet the required specifications
tical industry, including Fourier transform infra-red spec- would be used during processing. In-line temperature
troscopy (FTIR), UV-spectroscopy, gas chromatography, monitoring of the performance of the extrusion step could
high performance liquid chromatography (HPLC), X-ray be used to control the step via a feedback loop to the
diffraction spectroscopy, and NIR spectroscopy. heating/cooling system that controls the temperature.
Figure 3 illustrates a typical tablet manufacturing process Further, particle-size distribution will be continuously
in a pharmaceutical process [17]. It is seen that PAT monitored during milling for process consistency and
approaches could be applied to the various unit operations controlled via feedback or feed forward control. The
in the process: dispensing, blending, milling, compression, weight, thickness, potency, and hardness will be tested at-
and tablet coating. NIR provides a means of quick and line at the tablet press for continuous quality verification

Table 2 Examples of PAT applications in the pharmaceutical industry

Application Process analyzer Statistical tool Observation

Rapid and accurate tablet Acoustic resonance Principle-components A fast and non-destructive method for on-line analysis
identification spectroscopy analysis (PCA) and label comparison before shipping, to avoid
mislabeling of drug [38]
Active determination of content NIR Partial least-squares NIR method was developed and validated for
of uncoated pharmaceutical (PLS) analysis determination of active content ranging from
pellets 80-120% of the usual active content of the
uncoated pharmaceutical pellets [39]
Mechanical property determination Air-coupled excitation Iterative computational Examination of the vibrational resonance frequencies
of the drug tablet and laser interferometric technique can be directly correlated with the mechanical
detection properties of the tablet providing a non- destructive
technique for physical characterization of the
tablet [40]
Analysis of sustained-release tablet Terahertz pulsed – Tablet coating thickness, coating reproducibility,
film coatings using terahertz spectroscopy (TPS) distribution, and uniformity can be easily
pulsed imaging (TPI) determined. The method was validated against
optical microscopy imaging [41]
Roller compaction process of dry Thermal effusivity – Effusivity measurement were used to monitor the
granulation measurement using the roller compaction process [42]
effusivity sensor
Evaluation of content uniformity NIR PCA NIR/PCA was used to predict content uniformity of
for low-dose tablets low-dose tablets manufactured by a direct
compression process [43]
Powder flow characterization NIR PLS Real time information on the flowing cohesive powder
mixture was used to avoid powder segregation or
agglomeration and thus to maintain product
quality [44]
NIR measurement of the potency NIR PLS Potency of heparin active pharmaceutical ingredient
of an API was determined by this non-destructive
method [45]
Active drug identification and NIR PLS NIR method was used for qualitative and quantitative
content determination determination of ranitidine in granules for
compression, cores, and final tablet [46]
Monitoring capsule manufacturing NIR PLS PAT was utilized for testing of identity and quality
at small-scale level of raw materials, for blend uniformity analysis,
and for final content analysis of busulfan pediatric
capsules [47]
Quantification of the active NIR and UV–visible PLS More economical and less time-consuming method
ingredient in pharmaceutical spectroscopy for quantification of the lysine clonixinate [48]
injectable formulations

Prediction of dissolution for a NIR PLS This method was used to identify differences in the
sustained-release dosage form composition of the coating polymer used for a
tablet and thus assist with prediction of dissolution
behavior [49]
Analysis of liquid formulations Laser-induced breakdown – Method does not need any sample preparation and is
containing sodium chloride spectroscopy (LIBS) less time-consuming [50]
142 A.S. Rathore et al.

Vial Shake Seed Production Harvest via Chromatography Viral lnactivation


Flasks Expansion Bioreactor centrifugation, Column 1 (for Mab product)
depth filtration,
membrane filtration

Drug Product Bulk Ultrafiltration/ Chromatography Viral Filtration Chromatography


Processing/ Filling Filtration Diafiltration Column 3 (for Mab product) Column 2
(UF/DF)

Fig. 4 A typical production process for a biotech product

and feedback control of compression. This approach will resonance spectroscopy [38], air-coupled excitation and
enable both increased process understanding and process laser interferometric detection [40], terahertz pulsed spec-
control. troscopy [41], effusivity sensor [42], and laser-induced
Table 2 reviews some recent PAT applications in the breakdown spectroscopy [50]. It is evident from the
pharmaceutical industry [38–50]. With regard to process diversity of the analytical tools that are available and their
analyzers, NIR is the most popular and is used in a diverse capabilities that a part of the vision outlined in Fig. 3 can be
set of applications including estimation of active content realized, i.e. the process can be designed such that at the
[39, 43, 45, 46, 48], powder-flow characterization [44], raw end of each step, assurance can be provided that the step
material analysis [47], and dissolution rate [49]. Some other performed its function in a satisfactory manner. However,
analyzers that have been reportedly used include acoustic examples of using the information from the analyzers to

Table 3 Illustration of the challenges of executing PAT for biotech processes

Process Major process steps1 Typical process Typical decision Quality Attribute (QA) Analytical Typical analysis, Ratio
segment time, tp (hrs)2 time, td (hrs)3 or Process Attribute (PA)4 Method5 ta (hrs)6 (td/ta)

Upstream Microbial fermentation 24 2 Misincorporation HPLC 1 2


(production)
Mammalian cell culture 240 10 Glycosylation Oligosaccharide 1 10
(production) profile
Harvest Centrifugation 8 1 Recovery HPLC 1 1
Downstream Refolding 20 2 Misfolds HPLC 0.5 4
Chromatography 8 0.5 Aggregation HPLC 1 0.5
Drug product Formulation 6 0.5 Protein concentration A280 0.1 5
processing Lyophilization 24 1 Moisture NIR 0.1 10
Filling 12 2 Foreign particles Visible particles 0.5 4

Assumptions: 1. Biotech processes are recognized as complex processes that utilize a variety of unit operations. In this analysis, we are discussing
only a few major unit operations to illustrate PAT applications. 2. Typical times for a step can greatly vary from the nature of the molecule (fusion
protein vs. monoclonal antibody product, microbial cell line vs. mammalian cell line etc) and from product to product within the same class. Here,
we use typical values to illustrate the PAT concept. 3. Typical time that is available to make a decision before moving to the next process step. 4.
Every step in a biotech process may have multiple quality attributes (QA) or process attributes (PA) associated to it. Here we are just picking a
single attribute to illustrate the PAT concept. 5. Every QA or PA of a biotech product may be analyzed by multiple analytical methods, often
orthogonal to each other. Here we are just picking one assay for illustration purpose. 6. The analysis time for a given analytical method can greatly
vary. Here we are using a typical value for the sake of illustration.
Process analytical technology (PAT) for biopharmaceutical products 143

change the operating conditions and bring the process back 5. Our understanding of the link between each product
in control (control step of Fig. 1) are not too common and quality attribute and the clinical safety and efficacy of
creation of such control schemes is likely to be the focus of the drug is generally limited [15, 51].
the pharmaceutical industry in the future.
As a result of these differences, implementation of PAT
for biotech processes is bit more challenging [11, 18, 19,
53]. The challenges can be understood from the information
PAT applications in the biotechnology industry
presented in Table 3 and Fig. 5. Table 3 shows the process
time for some of the key biotech unit operations defined as
Figure 4 illustrates a typical process for a biotech product. It
the typical time that is available for making decision for a
shows some of the significant unit operations that together
step before proceeding to the next unit operation; quality or
form the process. While the above mentioned discussion on
process attribute that is typically important for the step;
implementation of PAT in the chemical and pharmaceutical
analytical method used for measuring the attribute; time
industries is relevant to applications in the biotech industry
required for analysis; and the ratio of the process time
also, there are some unique considerations that come into
available for decision making to the time needed for
play with biotech processes and products [15].
analysis. Figure 5 illustrates how the ratio changes for
1. Proteins are large, complex and heterogeneous molecules. different unit operations. It is evident that the ratio can be
2. Biotech processes are more complex than a typical used as an indicator of the ease of implementation of PAT
chemical or small-molecule drug-manufacturing pro- for a given unit operation. Process steps (for example
cesses with regard to number of batch records, number mammalian cell culture) where the time window for
of product quality tests, number of critical process decision making is wide and sample analysis for deciding
steps, and the amount and complexity of process data the course of action can be conveniently performed,
that is generated [51]. implementation of PAT is relatively straightforward. How-
3. Protein products are extremely sensitive to the manufac- ever, in the case of unit operations such as process
turing process used for their production. Lot to lot chromatography, where the time available for taking a
variability in product quality is commonly observed even decision is smaller than the time for analysis, implementing
when manufacturing has been performed using exactly a PAT application successfully requires changes to process,
the same process, and variability of source material has equipment, and analytical methods [11, 54–56].
also been known to affect product quality [15]. In this review, we will divide the process into four parts:
4. Raw materials used in manufacturing of biotech products upstream, harvest, downstream, and formulation. In the
can be complex, with lot to lot variability [52]. following sections, we will review PAT applications in each

Fig. 5 Ease of PAT implemen-


tation for some of the commonly
used unit operations in biotech
processes

PAT implementation is relatively easier


Ease of PAT implementation

12.8

6.4

3.2

1.6

0.8
PAT implementation is relatively difficult

0.4

0.2
Bioprocessing unit operations
144 A.S. Rathore et al.

Table 4 Examples of PAT applications in microbial fermentation unit operation

Analytical tool PAT Status Target of PAT application

Luminescence-based pH optrode On-line pH control in fermentation [59]


Nicotinamide adenine dinucleotide (NADH) On-line Biomass measurement and control of physiology of cell in industrial
fluorescence (enzyme assay) fermentation [60]
Fourier transform infrared (FTIR) with MVDA On-line Concentration of components such as glucose, fructose, and ethanol [61]
Optic turbidimetric sensors On-line Biomass/cell concentration and automatic regulation of substrate
and nutrient feed rate [62]
CO2 sensors On-line Dissolved CO2 concentration and control of medium feed rate [63]
Flow-injection analysis (FIA) coupled On-line Sucrose, glucose monitoring and fed batch fermentation control [64]
with biosensors
NIR with dynamic orthogonal projection On-line Alcohol-content measurement for ensuring robustness of calibration
model applied on-line for a fermentation process [65]
NIR and MIR On-line Monitoring of broth concentrations in an industrial fermentation
process [66]
Galvanic decoupled sensors for measuring On-line Metabolic monitoring of growth curve and measurement of electrical
electrical impedance property of cell solution [67]
Flow-injection system (FIS) with on-line sensors On-line CO2 and O2 gas analysis, monitoring of ammonia control of fed
batch penicillin fermentation and control of feed addition rate [68]
Flow-injection system coupled with biosensors On-line Glucose concentration measurement and control of cell growth [69]
FTIR analysis with multiple linear On-line Measurement of concentrations of proteins and polysaccharides for
regression (MLR) solid-state fermentations [70]
Electrochemical analysis (amperometric sensors) On-line Viable cell measurement and control on metabolic activity of cell [71]
Mass spectroscopy On-line Gas analysis (CO2, O2, CH4) and control of metabolic rate [72]
Sequential injection on-line analysis system On-line D-Lactic acid measurement and control of degree of acidification in
fermentation process [73]
Mass spectroscopy membrane sensor On-line Monitoring of acetone, butanol, and ethanol concentrations and control
of fed batch fermentation of acetone–butanol fermentation [74]
2D fluorescence spectroscopy On-line Monitoring of chemometric model and control of metabolic changes in
fermentation [75]
Transflective embedded NIR (TENIR) Semi-on-line Measurement of concentrations of glycerol and volatile fatty acids for
control of anaerobic fermentation [76]
Automatic membrane inlet mass On-line Aromatic compound analysis and monitoring of changes for a
spectroscopy (MIMS) two-phase process [77]
Electronic nose NIR with standard On-line Monitoring of pH, lactate, glucose, and galactose for control of yogurt
bioreactor probe fermentation [78]
Flow-injection analysis with enzyme sensor On-line Monitoring of glucose concentrations [79]
Glucose sensors (enzyme electrodes) On-line Glucose monitoring for development of new feedback-control strategy [80]
Thermal biosensor with FIA system On-line Monitoring and control for penicillin V production [81]
Ion chromatography with double chamber In-line Measurement of cations (Na+, K+, Ca2+, NH4+) for control of cell
acoustic wave detector growth [82]
Ion chromatography with piezoelectric On-line Lactic acid monitoring for control of production rate [83]
quartz crystal
HPLC At-line Analysis of metabolites for control of product formation [84]

of these areas. We will also address PAT applications in the characterization, and the large number of process conditions
field of chemometrics. that can affect the performance of these steps [18, 57]. It
should be emphasized that perhaps this part of the process
is the most critical because this is where product and
PAT applications in upstream unit operations product related impurities are produced. The latter may
include product-related variants that closely resemble the
Application of PAT in upstream unit operations of microbial desired product and have an equivalent potency and safety
fermentation or mammalian cell culture is complicated by profile (e.g. minor post-translational modifications such as
the fact that these unit operations are encumbered by use of C-terminal processing, N-terminal variants, deamidation)
several complex raw materials that have undergone limited and product-related impurities that do not resemble the
Process analytical technology (PAT) for biopharmaceutical products 145

Table 5 Examples of PAT applications for mammalian cell culture unit operation

Process variable Analytical tool Target of PAT application

Glucose concentration Continuous FIA with chemiluminescence Control volume mixing pattern and residence time
assay monitored to indicate metabolic condition to prevent
cell starvation [89]
Biomass concentration Conductance and capacitance-based Specific growth rate and product formation monitored
biomass probe to determine onset of stationary phase and
cell lysis [90–93]
Biomass concentration In-situ microscopes with image processors Specific growth rate used to control product formation
and computer control in agitated cell broth cultivation [92, 94]
Oxygen uptake rate (OUR) and Mass-balance technique for measuring kLa, Control of physiological state of cells and cellular
specific OUR dissolved O2 concentration, gas activity [57]
compositions
Biomass, glucose, and acetate NIR, spectroscopy and electronic nose Control of fed batch cultivation and monitoring of the
concentrations instrument quality of the culture [95]
Cell size, biomass concentration Dielectric spectroscopy Closed-loop control of medium feed rate in fluidized
bed fermenters [96–98]
Glucose and biomass concentrations Multiwavelength fluorescence (MVF), Control of specific growth rate and product
software sensors and MVDA quality [99]
Gas and hydrocarbons Gas sensor array as an electronic nose Detection of early infections in mammalian cell
cultures [100]
Heat fluxes Flow micro calorimeter, dielectric Monitoring of specific growth rate and control of
spectroscopy utilization of media components [101]
Metabolic flux analysis C-Glucose and 2D NMR spectroscopy Control of metabolic activity of Chinese hamster
(isotope tracer method) ovary (CHO) cells in high-density perfusion
cultures [102]

desired product with respect to safety and/or efficacy (e.g. raw material lots. In high cell-density fermentations, NIR
aggregates and highly truncated forms). In several cases the spectroscopy has also been used for at-line control and fault
impurity may not be affected by subsequent steps, thus analysis in fermentation processes [37, 87]. Gnoth et al.
making upstream unit operation the sole process step for have demonstrated the usefulness of mechanistic modeling
controlling the level of the impurity. Many off-line and on- in obtaining consistent total biomass and the product
line techniques are available that can either directly monitor concentration in an E. coli fermentation [88].
actual product formation and/or biomass concentration or Several applications have also been published for mam-
indirectly monitor variables which serve as markers for malian cell culture (Table 5). Use of liquid chromatography
product formation. On-line monitoring has the advantage to quantify the titer of the antibody in the cell culture and,
that it reduces the frequency of sampling and thereby the thus, indicate the time for harvesting, has been reported [58].
potential for microbial contamination [58]. The on-line system is likely to reduce the potential for
Use of PAT for controlling microbial fermentation microbial contamination as repeated sampling for titer
processes has been a topic of interest for several decades. measurement is not necessary. Kussow et al. monitored
Table 4 presents a summary of the several PAT applications growth rate of mammalian cells by utilizing the stoichio-
that have targeted microbial fermentation. Microporous metric correlation between base addition for pH adjustment
ultrafiltration membranes integrated with chromatographic and nutrient consumption (glucose). The perfusion rate was
systems have been used for in-line monitoring of fermen- monitored on-line via this stoichiometric correlation, and
tation [85]. Different variables, for example biomass nutrient concentration was controlled within a set range
production, specific growth rate, product concentration, [103]. The authors suggested that this approach could be
and byproduct formation during batch and fed batch useful in timing of virus infection and result in a higher
fermentations have been monitored for control of the virus concentration. Reinecke et al. have developed an
fermentation process [37]. A combination of NIR spectros- automated immunoassay technique instead of the com-
copy and multivariate data analysis (MVDA) has been monly used enzyme-linked immunosorbent assay (ELISA)
successfully used for screening many basal medium for determination of Immunoglobulin G (IgG) production
powders used in a mammalian cell culture process [86]. in mammalian cell culture monitored via fluorescence. Their
The ability to fingerprint the raw material enabled the system utilized surface plasma resonance to detect antigen–
authors to differentiate between good and poor performing antibody complexes, thus offering a rapid and sensitive assay
146 A.S. Rathore et al.

that could be used for process design and optimization [104]. allow for more time to analyze samples and make process
Different enzyme sensors for detection of glucose, lactate, decisions (Fig. 5); we believe we will see more “true” PAT
and glutamine connected via flow-injection analysis (FIA) applications in the future for this part of the process.
devices have been used, although their large-scale adoption
is hindered by issues regarding their stability, calibration, and
validation [105–107]. On-line cell concentration measure- PAT applications in harvest unit operations
ments via in-situ microscopy (ISM) have been shown to be
useful in capturing images with image processors and As can be seen in Fig. 4, microfiltration, depth filtration,
providing estimates of hybridoma cell concentration in homogenization, centrifugation, and precipitation are some
mammalian cell cultures [108]. of the most commonly used unit operations for harvest of
More recently, use of radio-frequency impedance for on- biotech products. The objective for this part of the process is
line monitoring of viable cell density in cell culture manufac- generally to make the output from the upstream step
turing processes as an accurate and reliable method has been (fermentation broth) ready for further purification by the
suggested [109]. Monitoring of metabolic activity of mam- downstream unit operations. This is achieved via clarification
malian cells and thus estimation of specific growth rate has of the process stream in a manner that maximizes product
been made possible by use of techniques such as NMR, mass recovery across these process steps and avoids any degrada-
spectroscopy, and radioactive tracers [110]. Automated flow tion of the product during the processing. It should be noted
cytometry has been used for monitoring culture homogeneity that the literature on this part of the process is relatively
over time, thus enabling understanding of cell death kinetics sparse in comparison with that on upstream unit operations.
and making the technique useful during scale-up of cell- For depth filtration, the possibility of using a PAT
culture steps [111]. In situ 2D fluorimetry, in combination application for monitoring separation of host cell proteins
with chemometrics, has been suggested as an approach for (HCP) by utilizing absorbance at 410 nm has been suggested
monitoring of key bioprocess variables for mammalian cell- [58]. Host cell proteins (HCP) absorb at 410 nm and thus, by
culture processes [112]. Viable cell density and recombinant monitoring conditions including transmembrane pressure,
protein concentrations could be actively monitored, leading flow rate, optical density (OD) of permeate at 280 nm and
to potential PAT applications utilizing this tool. OD of permeate at 410 nm, blockage of the depth filter can
In summary, the biotech industry has made significant be assessed [113]. This can be used to create a control
advancements in developing analyzers and modeling scheme for the depth filtration step such that HCP level in
approaches that enable us to monitor key process and quality the filtrate is maintained at or below the target level.
attributes in real-time or near-real-time. However, only a few Ultrafiltration units for clarification, integrated with col-
of these references meet the true expectations of PAT quality umn liquid chromatographic systems, have been used for on-
[58, 86, 88], i.e. not only process understanding but also line monitoring of harvest and fermentation processes [58,
process control (Fig. 1) to result in consistency of process 114]. Zhang et al. have used a hollow-fiber membrane for
performance and product. Because upstream unit operations extraction of Mab produced during fermentation. As shown
have a critical effect on product quality and also typically in Fig. 6, the permeate obtained from the hollow-fiber
Fig. 6 Use of an ultrafiltration
membrane in combination with
analytical chromatography for
on-line monitoring of a harvest
process

HOLLOW FIBRE
PUMP
MEMBRANE

PREPERATIVE
AFFINITY COLUMN

FERMENTER

ELUENT

ANALYTICAL
COLUMN
COMPUTER CONTROL
Process analytical technology (PAT) for biopharmaceutical products 147

module was directed either to a chromatographic harvesting variants, deamidation), product-related impurities that do not
column or to an analytical chromatographic column (protein resemble the desired product in respect of safety and/or
G-specific). Data obtained from analysis were used to monitor efficacy (e.g. aggregates and highly truncated forms), and
the harvest process [58]. PAT has also been used for on-line process-related impurities (e.g. host-cell DNA, host-cell
monitoring of biomass-based power plant effluents during proteins, and raw materials from the process). Isolating the
microfiltration. By monitoring fouling behavior by turbid- product of interest from these myriad of species to achieve the
ometry, a suitable regeneration strategy can be applied high purity level expected of a biotech product today is a
during microfiltration [115]. To obtain constant flux, difficult task. As a result, biotech processes consist of several
membrane retention, and turbidity reduction, intermittent unit operations, with each unit operation serving a defined
back flushing with compressed nitrogen gas can help to purpose. For some of the product quality attributes, a certain
avoid cake formation and membrane fouling. This strategy level of redundancy in the process is expected. The
has been shown to result in product retention of more than interdependencies between the performance and expectations
99.5% and turbidity reduction of greater than 95% [116]. of each unit operation are particularly relevant to downstream
Use of near infrared (NIR) spectroscopy for measure- processing. The need for appropriate control of each step and
ment of sedimentation velocity in a centrifuge has been continuous assessment of its performance is also of great
suggested [117]. The authors used a particle-separation importance. In the following text we review PAT applications
analyzer for measurement of separation/sedimentation in some of the most commonly used downstream unit
kinetics of bentonite/xanthan gum mixtures during centri- operations.
fugation. They used an analytical centrifuge with an opto- HPLC has been used as a monitoring tool for a refold
electronic sensor system which measures NIR transmission step [53]. Product purity and the levels of the different
profiles of horizontally inserted samples tubes. Thus, product-related variants and impurities were monitored as a
clarification data enabled optimization of concentration of function of time. The data suggested that it is feasible to
excipients to be added to the final formulation. In another implement a PAT-based control scheme that enables ending
application of NIR, a method has been developed for of refold on the basis of product quality data. This would
monitoring the active pharmaceutical ingredient (API) ensure consist product quality of the refold end pool and
concentration in the different process intermediates during improve operational efficiency by keeping refold time to
harvest steps [118]. The method had an accuracy of 0.01% what is really needed.
(w/w) for an API operational range between 0.20 and Protein crystallization has also been of some interest for
0.45% (w/w). In summary, the harvest steps have not seen PAT applications, because of the increase in structure-based
as much interest in the development of PAT applications as drug design fuelled by the recent developments in genomics
the other parts of the process. This could be because of the and proteomics [119, 120]. In-situ microscopy has been
limited effect harvest steps typically have on final product used to monitor the progress of crystallization processes
quality. Hence, the benefits of implementing PAT are not as and to differentiate among the various crystal forms [119].
significant as they may be for the upstream or the Microfluidics-based high-throughput approaches have been
downstream steps of the process. There may be applica- used for optimization of conditions for crystal formation
tions, though, where, because of unique drivers (such as [121–123]. This is necessary because in order to be used for
product stability) PAT applications such as those reviewed structural analysis, the protein has to be of very high purity.
above may be of interest to the industry. Large crystals of some of the model proteins (lysozyme and
aprotinin) have been produced by controlling the supersatu-
ration level by changing the temperature or the ionic strength
PAT applications in downstream unit operations of the solution [124, 125]. This strategy or a more advanced
control strategy could be used in combination with high-
Commonly used unit operations for downstream processing throughput techniques to improve protein crystal growth. At
of biotech products include refolding, precipitation, filtration this time, protein crystallization remains a technique for
(depth filtration, microfiltration, ultrafiltration/diafiltration, production of protein in small quantities. In the future it is
nanofiltration etc), and liquid chromatography. The objective possible that advances in our understanding of factors that
of purification is to isolate the product of interest from the affect protein crystallization may lead to utilization of this
process stream that comes from the harvest unit operations procedure for large-scale production of proteins [126]. PAT
and contains a variety of impurities besides the product of is likely to play a key role in this evolution.
interest. These impurities include product-related variants that Rathore et al. have used a PAT-based control scheme for the
closely resemble the desired product and have an equivalent diafiltration step in an ultrafiltration/diafiltration (UF/DF)
potency and safety profile (e.g. minor post-translational process. The concentration of each species in the product
modifications such as C-terminal processing, N-terminal stream was measured after each diafiltration volume to assess
148 A.S. Rathore et al.

the performance of the DF step as a function of number of In summary, the downstream process steps, in particular
diafiltration volumes. On the basis of the results, monitoring process chromatography, have a lot of gain from imple-
of the pH during diafiltration was proposed as a PAT tool for mentation of PAT. As is evident from this section, this topic
indicating completion of the DF process [53]. This approach is also relatively unexplored when we think of the potential
was suggested to be particularly attractive for cases when an opportunities it presents [53–56, 130]. However, imple-
expensive diafiltration buffer is being used or when the menting PAT for this part of the process will be challenging
diafiltration time must be minimized because of product- because of the short process times typical of these unit
stability concerns, because in both of these cases minimizing operations (Fig. 5). We expect that we will see more
the number of diafiltration volumes and process time would development and implementation of “true” PAT applica-
be beneficial. tions in the future for downstream processing.
Use of high-performance liquid chromatography (HPLC)
as an analytical tool for process monitoring has been proposed
in the literature [127, 128]. Use of on-line reversed phase PAT applications in drug product unit operations
HPLC for separating recombinant human insulin-like growth
factor-I (IGF) and IGF aggregates to facilitate real time In this section we have purposely included some applica-
pooling based on product quality has been successfully tions to traditional small molecules, because the tools and
demonstrated [129]. A similar application using chromato- approaches are likely to be extended to their biotech
graphic separation with protein A-immobilized media has counterparts also.
been shown to reliably control antibody loading on a protein In the formulation step of drug product processing,
A column in a CHO-derived recombinant antibody purifica- different chemical substances including excipients with
tion process [130]. Recently, an approach utilizing a active drug substance are combined to produce the final
combination of on-line size-exclusion HPLC, differential medicinal product. Raman spectroscopy with wide-area
refractometry, and multi-angle laser light scattering analysis illumination (WAI) has been suggested for monitoring of
(MALLS) has been shown to be useful for real-time active pharmaceutical ingredient (povidone) in the eyewash
estimation of product quality of a mutant form of the human formulation, directly through plastic bottles [133]. The
immunodeficiency virus (HIV) [131]. The accuracy of the Raman data were modeled using PLS for quantification of
approach was found to be >89% and intra and inter-day active pharmaceutical ingredient (API) content. A similar
precision were found to be 0.9 and 3.6% RSD, respectively. application used Raman spectroscopy with WAI and PLS for
More recently, Rathore et al. have published a series of determination of acetaminophen in vials containing suspen-
case studies demonstrating the utility of a variety of sion [134]. Laser-induced breakdown spectroscopy (LIBS)
analytical tools for performing analysis that would facilitate has been used for direct and rapid analysis of pharmaceutical
pooling of large-scale chromatography columns. The liquid formulations. In this technique, the atomic line from
analytical tools that have been utilized include HPLC [55, sodium is used to quantify NaCl in formulations such as
56], ultra-performance liquid chromatography (UPLC) [54], injectables, nasal solutions, and bactericidal injections [50].
and fluorescence [132]. These applications are based on the NIR spectroscopy has been used for quantification of lysine
premise that for the cases where the chromatographic clonixinate in an intravenous injection formulation [48].
separation is high-resolution and part of the peak is being Bench-top NMR and MRI (magnetic resonance imaging)
collected to pool the product and exclude the impurity/ have been used for monitoring of emulsions, lipid ingre-
impurities, these approaches offer a way of performing dients, and their adsorption characteristics. This technique
analysis in the required time frame and enable a decision enables non-invasive monitoring of drug-delivery processes
based on product quality. This is in contrast with the often of these formulations [135]. NIR has also been used during
used practice of pooling based on UV absorbance, such as mixing of to illustrate differences in the mixing character-
absorbance at 280 nm, which though an operationally istics of aspirin, citric acid, aspartame, or povidone with
simpler approach to implement, suffers from the fact that avicel, and thus to elucidate the different cohesive properties
absorbance at 280 nm cannot differentiate between the of the particles [136]. The approach was used for under-
product and other product-related impurities. In one case, standing the dynamics of powder mixing and to make
use of on-line HPLC resulted in the pool purity varying quantitative measurements and monitor possible changes in
between 90.2 and 91.6% even though the purity of the feed particle size during blending.
material varied from 62.8–81.6% [56]. Utilization of Lyophilization is one of the most commonly used unit
absorbance-based pooling resulted in variability of pool operations in drug product processing of biotech products.
quality from 81.3–95.4%. This case study clearly highlights In this step, water and other solvents are removed first by
the effect of using PAT-based controls on the consistency of freezing the product and then removal of frozen ice by
product quality. sublimation. In primary drying, frozen ice is sublimated and
Process analytical technology (PAT) for biopharmaceutical products 149

in secondary drying, residual water is removed from the quality attributes in real or near-real time, these data have
porous matrix. Because of the great effect of this process rarely been used for process control (Fig. 1) [138]. Because
step on final product quality and the sensitivity of the step most drug product unit operations allow more time to
to various environmental conditions, lyophilization has analyze samples and make process decisions (Fig. 5), we
been a focus of numerous PAT applications. As seen in believe the industry will start implementing “true” PAT
Figure 7, on-line monitoring of residual water content applications in the future for this part of the process.
during drying, by use of moisture sensors that measure
water vapor pressure, has been used to predict sublimation
end point [137]. NIR spectroscopy and Raman spectrosco- PAT applications in chemometrics
py are most commonly used to determine residual water
during drying. NIR spectroscopy has been used for Chemometrics, the science of extracting relevant informa-
determination of surface and bound water during drying, tion from chemical processes, has had major effect on
and, with the model proposed, used to accurately predict process analytical chemistry. For complex systems, for
the end point for drying [138]. Use of Kalman filter-based example biotech manufacturing, analytical data from on-
observers with mathematical modeling has been suggested line monitoring instruments is typically not amenable to
as an approach that enables in-line monitoring of primary direct interpretation. So chemometric methods are essential
drying phase in vials [139]. Tunable diode laser absorption for effective data analysis [86, 142, 143]. In addition,
spectroscopy (TDLAS) has been applied as a non-invasive chemometrics makes it possible to obtain real-time informa-
technique to monitor water vapor concentration and vapor tion from data. Fast, precise, accurate, and non-destructive
flow velocity during freeze drying, which enables measure- process analyzers in combination with chemometrics are
ment of the total amount of water removed [140]. Another suitable for process analysis and optimization leading to
recent publication has demonstrated how, by integrating the improved product quality, productivity, and efficiency. Che-
various devices (pressure gauge, mass spectrometer, ther- mometric methods help in the development of an empirical
mometer, moisture sensor), an innovative model can be model based on the collected data that can be used for the
developed to enable monitoring of sublimating interface predictive estimation of the properties of a chemical
temperature and heat and mass transfer coefficients during process and hence help in process analysis, optimization,
freeze drying in vials [141]. A mono-dimensional model and control.
capable of giving mass and energy balances in dried layers As can be seen in Fig. 8, developing and implementing a
during primary drying in vials has been validated using chemometric approach is often a multistep process, neces-
experimental data obtained in a pilot scale freeze dryer and sitated by the huge datasets that are to be analyzed and the
used for on-line monitoring [139]. sophisticated statistical tools that are used.
In summary, although, similar to upstream processing,
many advances have enabled us to monitor key process and Preprocessing of the data

Pressure Vaccum chamber Pretreatment of the data by approaches such as standard


sensor normal variate (SNV), orthogonal signal correction (OSC),
and multiplicative scatter correction (MSC) may be a
critical step in creating an accurate chemometric model in
Vials some applications [144, 145].
NIR Probe Heating/Cooling Shelf
Data reduction

Reduction of the multidimensional data can be achieved by


NIR S
Spectrometer
use of regression methods such as the PCA. This enables us
to use the principle components to quantify the interactions
Temperature among the variables by calculating the matrix of the
Control system correlations for the whole data in the form of scores and
Condenser loadings matrix [8, 86, 146].

Discrimination of data
Computer
Fig. 7 On-line monitoring, by NIR spectroscopy, of moisture content Discrimination methods help in the grouping of the data
(both unbound and bound water) during freeze drying in vials according to the specific properties of the data [130].
150 A.S. Rathore et al.

Sample removal using Regression analysis of the data (multivariate data calibration)
Sampling technique
Sampling Techniques
and Process Analysis This step helps in correlating the data with quantifiable
Process Analyzer properties of the sample. PLS is the regression method most
commonly used for quantitative prediction of the dependent
Data collection variables [146, 147].

Validation of the chemometric model


Preprocessing of Data
This is a key step, because chemometric analyses often lack
Chemometrics and any mechanistic basis. Thus, this step enables accurate
Data Reduction Multivariate Data determination of the principle components affecting the
Analysis process and reliable estimation of the noise and error in the
analysis. While test-set validation may be used when a
Data Discrimination
large number of data is available, cross validation is often
used when the number of samples is limited. Model
Regression Analysis of Data performance is measured by use of correlation coefficients
which indicate relationships between the measured refer-
ence and predicted reference [142, 143, 148, 149].
Multivariate Data Calibration Application of chemometrics to experimental data in
process analytical technology has two key objectives.
First, qualitative data analysis using chemometrics aids
Validation of Chemometric identification of patterns in data, via trends analysis, and
Model helps to eliminate outliers. Second, quantitative data
Process Control
analysis using regression methods is useful for predicting
Process Control relationships between independent and dependent varia-
bles. Figure 9 illustrates some of the commonly used
analytical tools that together with the appropriate chemo-
Successful PAT Application
metric technique form the basis for a PAT application.
Process analyzers could be hardware sensors or soft
Fig. 8 The various steps that must be taken for development and sensors and may be implemented on-line, in-line, at-line,
implementation of a chemometry-based PAT application
or off-line. Of the hardware sensors, NIR is one of the most
commonly used analyzers in the pharmaceutical industry.
Other commonly used process analyzers include Raman
Unsupervised classification methods, for example the spectroscopy, image probe, mass spectroscopy, biosensors,
Gaussian mixture model and principle-components analysis dielectric spectroscopy, 2D fluorescence spectroscopy,
(PCA), and supervised classification methods, for example biosensors, and enzyme-linked immunosorbent assay
partial least-squares (PLS) analysis and soft independent (ELISA). Soft sensors are software that assist in the
modeling of class analogy (SIMCA), are commonly used in processing of several measurements. Digital signal pro-
biotech applications [52, 86, 147]. cessing is a key part of software sensors. Kalman filters

Fig. 9 The various combina-


tions of analyzers and statistical Statistical and mathematical methods • Design of experiment
• Pattern recognition
tools that together form a PAT
• Multivariate calibration
application
Analytical instrumentation • Near infrared spectroscopy
• Raman spectroscopy
• High performance liquid chromatography
Software assistance • Matlab
• Excel
• SIMCA
Applications • Process analysis and control
• Process optimization
Process analytical technology (PAT) for biopharmaceutical products 151

and fuzzy controllers are the most common examples of Summary


soft sensors [146, 150].
Several applications involving use of chemometric The biotech community agrees that it is mutually beneficial
tools in biotech production have been published recently. to the industry and to its regulators to use PAT applications
PCA has been applied to near-infrared spectral informa- to improve control processes, to deliver more consistent
tion derived from scanning unprocessed culture fluid product quality, and, perhaps, to raise operational efficien-
samples from a complex antibiotic production process cy. Implementing PAT, however, requires a higher level of
for process diagnosis and rapid assessment of process process understanding which results in an immediate need
performance [152]. An integrated on-line multivariate for greater investment of time and resources. Hence, the
statistical process monitoring, quality prediction, and fault pros and cons of implementing a PAT application must be
diagnosis framework for batch processes has been applied analyzed before taking the decision to proceed [53].
for simulated fed-batch penicillin fermentation [153]. The In this paper, we hope we have captured the evolution of
approach enabled prediction of the values of the end-of- the concepts of PAT and reviewed the major developments
batch quality during progress of the batch; this was useful that have occurred with regard to implementation of PAT in
for anticipating effects of excursions on final quality, and biotech processes. Wold et al. have classified the PAT
suggested control strategies. Use of robust statistical applications into five levels as follows [136]:
methods to reduce or remove the effect of outlying data
PAT Level 1 Determination of the concentration and other
points, in order to enable the good data to primarily
properties of component from the spectrum
determine the result, has been suggested [154]. A success-
PAT Level 2 Raw material quality check and the classifica-
ful case study involving modeling of data from a
tion from spectral data
pharmaceutical process and using the model for making
PAT Level 3 Batch process monitoring by means of the
predictions has been presented [151]. The authors empha-
multivariate measurement
sized the importance of using design of experiments (DOE)
PAT Level 4 Using the batch monitoring data and process
during process development and then later using a well
information to predict the quality of the
managed system of data collection, data bases, data
product
analysis, and networks to connect these pieces.
PAT Level 5 Feedback control of the process to maintain
More recently, Kirdar et al. have examined the feasibility of
the quality of the product
using chemometrics for supporting key activities required for
successful manufacturing of biopharmaceutical products, It is our observation that although significant advances have
including scale-up, process comparability, process character- been accomplished with regard to our ability to analyze/
ization and fault diagnosis [155]. Gunther et al. have monitor key process and quality attributes in the biotech
demonstrated application of a flexible process-monitoring industry (Levels 1–3), much more needs to be done with regard
method to industrial pilot plant cell culture data for fault to utilizing the collected data for subsequent control of the
detection and diagnosis [156]. In a more recent publication, process, to achieve optimum yield and product quality [4, 5].
application of chemometrics as a diagnostic tool for Of all the references cited in this paper, only a handful achieve
identifying the root cause of a problem and designing the latter objective. Thus, more needs to be done to achieve the
experimental conditions to demonstrate and correct this root benefits that will result from true PAT implementation.
cause has been published [86]. Another recent paper from
Kirdar et al. reports a case study involving use of a combined
approach of near-infrared (NIR) spectroscopy and chemo- References
metrics for screening of lots of basal medium powders based
on their effect on process performance and product attributes 1. U.S. Department of Health and Human Services, Food and Drug
Administration (2002) http://www.fda.gov/Drugs/Development
[52]. A combined NIR/chemometrics approach was shown ApprovalProcess/Manufacturing/QuestionsandAnswersonCurrent
to make it possible to fingerprint the raw materials to GoodManufacturingPracticescGMPforDrugs/UCM071836
distinguish between good and poorly performing media lots. Accessed 10 Feb 2010
In summary, chemometrics has become almost a 2. U.S. Department of Health and Human Services, Food and Drug
Administration (2001) A presentation at the Advisory Committee
necessary counterpart to the process analyzer in any for Pharmaceutical Science by Dr. Ajaz S. Hussain http://www.fda.
successful PAT application for biotech processes. As the gov/ohrms/dockets/ac/cder01.htm#Pharmaceutical%20Science
industry moves, in general, from “analysis for monitoring” Accessed 10 Feb 2010
to “analysis for monitoring followed by process control to 3. Hinz DC (2006) Anal Bioanal Chem 384:1036–1041
4. Miller RW (2002) Am Pharm Rev 5(2):25–29
get to the desired end point”, we will see an even bigger 5. Miller RW (2003) Am Pharm Rev 6(1):52–61
role of chemometrics in process modeling necessary for 6. U.S. Department of Health and Human Services, Food and Drug
connecting the analysis with process control. Administration (2002) Federal Register 67 (25) – Notice http://
152 A.S. Rathore et al.

www.fda.gov/OHRMS/DOCKETS/98fr/020602e.pdf Accessed 33. Seasholtz MB (1999) Chemometr Intell Lab 45:53–63


19 Feb 2010 34. Fernandes HL, Raimundo IM, Pasquini C, Rohwedder JJR
7. U.S. Department of Health and Human Services, Food and Drug (2008) Talanta 75:804–810
Administration (2003) Guidance for industry: PAT—a frame- 35. Qu H, Alatalo H, Hatakka H, Kohonen J, Louhi-Kulanen M
work for innovative pharmaceutical, manufacturing and quality (2009) J Cryst Growth 311:3466–3475
assurance http://www.fda.gov/ohrms/dockets/ac/03/briefing/ 36. Ward HW, Sistare FE (2007) Anal Chim Acta 595:319–322
3996B1_03_GFI-Pat.pdf Accessed 28 Dec 2009 37. Junker BH, Wang HY (2006) Biotechnol Bioeng 95(2):226–
8. U.S. Department of Health and Human Services, Food and Drug 261
Administration (2004) Guidance for industry: PAT—a framework 38. Medendorp J, Lodder RA (2006) AAPS Pharm Sci Tech 7(1):
for innovative pharmaceutical development, manufacturing and E1–E9
quality assurance http://www.fda.gov/downloads/Drugs/Guidance 39. Mantanus J, Ziemons E, Lebrun P, Rozet E, Klinkenberg R,
ComplianceRegulatoryInformation/Guidances/ucm070305.pdf Streel B, Eyrard B, Hubert P (2009) Talanta doi:10.1016/j.
Accessed 28 Dec 2009 talanta.2009.10.019
9. Guidance for Industry: Q8(R2) Pharmaceutical Development, 40. Akseli I, Cetinkaya C (2008) Int J Pharm 359:25–34
US Department of Health and Human Service, Food and Drug 41. Ho L, Muller R, Romer M, Gordon KC, Heinamaki J,
Administration (FDA). August, 2009. http://www.ich.org/LOB/ Kleinebudde P, Pepper M, Rades T, Shen YC, Strachan CJ,
media/MEDIA4986.pdf Taday PF, Zeitler JA (2007) J Control Release 119:253–261
10. Rathore AS, Winkle H (2009) Nat Biotechnol 27:26–34 42. Ghorab MK, Chatlapalli R, Hasan S, Nagi A (2007) AAPS
11. Rathore AS (2009) Trends Biotechnol 27(9):546–553 Pharm Sci Tech 8(1) Article 23 doi:10.1208/pt0801023
12. Van Hoek P, Hamrs J, Wang X, Rathore AS (2009) In: Rathore 43. Li W, Bagnol L, Berman M, Chiarella R, Gerber M (2009) Int J
AS, Mhatre R (ed) Quality by design for biopharmaceuticals: Pharm 380:49–54
Perspectives and case studies. Wiley Interscience 44. Benedetti C, Abatzoglou N, Simard JS, McDermott L, Leonard
13. Arora T, Greene R, Mercer J, Tsang P, Casais M, Feldman S, G, Cartilier L (2007) Int J Pharm 336:292–301
Look J, Lubiniecki T, Mezzatesta J, Pluschkell S, Rosolowsky 45. Sun C, Zang H, Liu X, Dong Q, Li L, Wang F, Su L (2009) J
M, Rathore AS, Schenerman M, Schofield T, Sheridan S, Smock Pharm Biomed Anal doi:10.1016/j.jpba.2009.11.022
P, Anliker S, McGarvey B, Meiklejohn B, Precup J (2009) 46. Rosa SS, Barata PA, Martins JM, Menezes JC (2008) Talanta
Biopharm Int 22(11):26–36 75:725–733
14. Arora T, Greene R, Mercer J, Tsang P, Casais M, Feldman S, 47. Paris I, Janoly-Dumenil A, Paci A, Mercier L, Bourget P, Brion
Look J, Lubiniecki T, Mezzatesta J, Pluschkell S, Rosolowsky F, Chaminade P, Rieutord A (2006) J Pharm Biomed Anal
M, Rathore AS, Schenerman M, Schofield T, Sheridan S, Smock 41:1171–1178
P, Anliker S, Atkins L, McGarvey B, Meiklejohn B, Precup J 48. Lopez-Arellano R, Santander-Garcia EA, Andrade-Garda JM,
(2010) Biopharm Int 23(1):26–36 Alvarez-Avila G, Garduno-Rosas JA, Morales-Hipolito EA
15. Rathore AS (2009) Trends Biotechnol 27(12):698–705 (2009) Vib Spectrosc 51:255–262
16. Scott B, Wilcock A (2006) J Pharm Sci Technol 60(1):17–53 49. Tabasi SH, Moolchandani V, Fahmy R, Hoag SW (2009) Int J
17. Rathore AS, Gerhardt AS, Montgomery SH, Tyler SM (2009) Pharm 382:1–6
Biopharm Int 22(1):36–44 50. St-Onge L, Kwong E, Sabsabi M, Vadas EB (2004) J Pharm
18. Read EK, Park JT, Shah RB, Riley BS, Brorson KA, Rathore AS Biomed Anal 36:277–284
(2009) Biotechnol Bioeng 105(2):276–284 51. Schellekens H (2009) Nephrol Dial Transplant 2(Suppl 1):i27–i36
19. Read EK, Park JT, Shah RB, Riley BS, Brorson KA, Rathore AS 52. Kirdar AO, Chen G, Rathore (2010) Biotechnol Prog 26:527–
(2009) Biotechnol Bioeng 105(2):285–295 531
20. Seely J (2005) In: Rathore AS and Sofer G (ed) Process validation 53. Rathore AS, Sharma A, Chillin D (2006) Biopharm Int 19:48–57
in Manufacturing of Biopharmaceuticals Taylor and Francis 54. Rathore AS, Wood R, Sharma A, Dermawan S (2008)
21. Freund H, Sundmacher K (2008) Chem Eng Process 47:2051–2060 Biotechnol Bioeng 101(6):1366–1374
22. Koch MV (2006) Anal Bioanal Chem 384:1049–1053 55. Rathore AS, Yu M, Yeboah S, Sharma A (2008) Biotechnol
23. Simon H (1956) Process Applications of Analytical Instruments. Bioeng 100(2):306–316
Instrum Autom 29:1141–1145 56. Rathore AS, Parr L, Dermawan S, Lawson K, Lu Y (2010)
24. Clevett KJ (1986) Process analyzer technology. Wiley- Biotechnol Prog 26:448–457
Interscience, New York 57. Konstantinov K, Chuppa S, Sajan E, Tsai Y, Yoon S, Golini F
25. Koch KH (1997) Process analytical chemistry. Springer, New (1994) Tibtech 12:324–333
York 58. Zhang J, Zhou H, Ji Z, Regnier F (1998) J Chromatogr B
26. Chalmers JM ed (2000) Spectroscopy in process analysis. CRC 707:257–265
Press, Boca Raton, FL 59. Chan C, Lo W, Wong KY (2000) Biosens Bioelectron 15:7–11
27. Bakeev KA ed (2005) Process analytical technology: spectro- 60. Guenneugues P, Chibois S, Trystram G (1990) Food Control 1
scopic tools and implementation strategies for the chemical and (4):236–241
pharmaceutical industries. Blackwell Publishing, Oxford 61. Armenta S, Garrigues S, Gauradia M, Rondeau P (2005) Anal
28. Beebe KR, Blaser WW, Bredeweg RA, Chauvel JP, Harner RS, Chem Acta 545:99–106
Lapack M, Leugers A, Martin DP, Wright LG, Yalvac ED (1993) 62. Salgado AM, Folly ROM, Valdman B (2001) Sensor Actuator B
Anal Chem 65(12):R199–R216 75:24–28
29. Workman J, Koch M, Veltkamp DJ (2003) Anal Chem 75:2859– 63. Chen Y, Krol J, Huang W, Cino JP, Vyas R, Mirro R,
2876 Vaillancourt B (2008) Process Biochem 43:351–355
30. De Fatima L, de Lira B, Vasconcelos FVC, Pereira CF, Paim 64. Ferreira LS, Desouza MB, Folly ROM (2001) Sensor Actuator B
APS, Stragevitch L, Pimentel MF (2010) Fuel 89:405–409 75:166–171
31. Prinsloo NM, Engelbrecht JP, Mashapa TN, Strauss MJ (2008) 65. Zeaiter M, Roger JM, Bellon-Maurel V (2006) Chemometr Intell
Appl Catal A-Gen 344:20–29 Lab 80:227–235
32. Halstensen M, Bakker P, Esbensen KH (2006) Chemometr Intell 66. Triadaphillou S, Martin E, Montague G, Norden A, Jeffkins P,
Lab 84:88–97 Stimpson S (2007) Biotechnol Bioeng 97(3):554–567
Process analytical technology (PAT) for biopharmaceutical products 153

67. Hofmann MC, Ellersiek D, Kensy F, Buchs J, Mokwa W, 103. Kussow CM, Zhou W, David M, Gryte S, Hu W (1995) Enzyme
Schnakenberg U (2005) Sensor Actuator B 111–112:370–375 Microb Tech 17:779–783
68. Johansen C, Christensen LH, Nielsen J, Villadsen J (1992) Comp 104. Reinecke M, Scheper T (1997) J Biotechnol 59:145–153
Chem Eng 16:S297–S304 105. Renneberg R, Trott-Kriegeskorte, Lietz M, Jager V, Pawlowa M,
69. Tothill IE, Newman JD, White SF, Turner APF (1997) Enzyme Kaiser G, Wollenberger U, Schubert F, Wagner R, Schmid RD,
Microb Tech 20:590–596 Scheller FW (1991) J Biotechnol 21:173–186
70. Gordon SH, Green RV, Wheeler BC, James C (1993) Biotechnol 106. Vanderpol J, Spohn U, Eberhardt R, Gaetgens J, Biselli M,
adv 11:665–675 Wandrey C, Tramper J (1994) J Biotechnol 37:253–264
71. Pescheck M, Schrader J, Sell D (2005) Bioelectrochemistry 107. Dremel BAA, Li SY, Schmid RD. Biosens Bioelectron 7:133-
67:47–55 139
72. Heinzle E, Oeggerli A, Dettwiler B (1990) Anal Chim Acta 108. Guez JS, Cassar JPh, Wartelle F, Dhulster, Suhr H (2009)
238:101–115 Process Biochem 45(2):288–291
73. Shu H, Hakanson H, Mattiason B (1995) Anal Chim Acta 109. Carvell JP, Dowd JE (2006) Cytotechnology 50:35–48
300:227–285 110. Henry O, Kamen A, Perrier M (2007) J Process Contr 17:241–
74. Chauvatcharins S, Seki T, Fujiyama K, Yoshida T (1995) J 251
Ferment Bioeng 79(3):264–269 111. Sitton G, Srienc F (2008) J Biotechnol 135:174–180
75. Rhee J, Kang T (2007) Process Biochem 42:1124–1134 112. Teixeira AP, Portugal CA, Carrinhas N, Dias JM, Crespo JP,
76. Holm-Nielsen JB, Lomborg CJ, Oleskowicz-Popiel P, Esbensen Alves PM, Carrondo MJ, Oliveira R (2008) Biotechnol Bioeng
KH (2008) Biotechnol Bioeng 99(2):302–313 102:1098–1106
77. Tarkiainen V, Kotiaho T, Mattila I, Virkajarvi I, Aristidou A, 113. Yigsaw Y, Piper R, Tran M, Shukla AA (2006) Biotechnol Prog
Ketola R (2005) Talanta 65:1254–1263 22:288–296
78. Cimander C, Carlsson M, Mandenius C (2002) J Biotechnol 114. Lorenz JK, Reo JP, Hendl O, Worthington JH, Petrossian VD
99:237–248 (2009) Int J Pharm 367:65–72
79. Appleqvist R, Hansen EH (1990) Anal Chim Acta 235:265–271 115. Maheshkumar S, Madhu GM, Roy S (2007) Sep Purif Technol
80. Brooks SL, Ashby RE, Turner APF, Calder MR, Clarke DJ 57:25–36
(1987/1988) Biosensor 3:45–56 116. Kuberkar V, Davis R (2001) J Membrane Sci 183:1–14
81. Rank M, Danielsson B (1992) Biosens Bioelectron 7:631–635 117. Kuentz M, Rothlisberger D (2003) Eur J Pharm Biopharm 56
82. Xie Y, Zeng H, Yao S, Wei W (1999) Talanta 50:1019–1025 (3):355–361
83. Zhang J, Xie Y, Dia X, Wei W (2001) J Microbiol Meth 44:105– 118. Rodrigues LO, Cardoso JP, Menezes JC (2008) Biotechnol Prog
111 24:432–435
84. Morovjan G, Szakacs G, Fekete J (1997) J Chromatogr A 119. Bluma A, Hopfner T, Rudolph G, Lindner P, Beutel S, Hitzmann
763:165–172 B, Scheper T (2009) J Cryst Growth 311(17):4193–4198
85. Merbel NC (1997) Trends Anal Chem 16(3):162–172 120. Fujiwara M, Nagy ZK, Chew JW, Braatz RD (2005) J Process
86. Kirdar AO, Green KD, Rathore AS (2008) Biotechnol Prog Contr 15(5):493–504
24:720–726 121. Sharff A, Jhoti J (2003) Curr Opin Chem Bio 7:340–345
87. Macaloney G, Draper I, Preston J, Anderson KB, Rollins MJ, 122. Woerd M, Ferree D, Pusey M (2003) J Struct Biol 142:180–187
Thompson BG, Hall JW, McNeil B (1996) Trans I ChemE 74 123. Hansen CL, Skordalakes E, Berger JM, Quake SR (2002) Proc
(c):212–220 Natl Acad Sci 99:16531–16536
88. Gnoth S, Jenzsch M, Simutis R, Lubbert A (2007) J Biotechnol 124. Shall CA, Riley JS, Li E, Arnold E (1996) J Cryst Growth
132:180–186 165:299–307
89. Huang YL, Li SY, Dremel BAA, Bilitewski U, Schmid RD 125. Jones WF, Wiencek JM, Darcey PA (2001) J Cryst Growth
(1991) J Biotechnol 18:161–172 232:221–228
90. Sarra M, Ison AP, Lilly MD (1996) J Biotechnol 51:157–165 126. Shenoy B, Wang Y, Shan W, Margolin AL (2001) Biotech
91. Guez JS, Cassar JPh, Wartelle F, Dhulster, Suhr H (2004) J Bioeng 73:358–369
Biotechnol 111:335–341 127. Low DKR (1986) J Chem Tech Biotechnol 36:345–350
92. Fehrenbach R, Comberbach M, Ptre JO (1992) J Biotechnol 128. Cooley RE, Stevenson CE (1992) Process Control Qual 2:43–53
23:303–314 129. Fahrner RA, Lester PM, Blank GS, Reifsnyder DH (1998) J
93. Taya M, Hegglin M, Prenosil JE, Bourne JR (1989) Enzyme Chromatogr A 827:37–43
Microb Tech 11:170–176 130. Fahrner RA, Blank GS (1999) J Chromatogr A 849:191–196
94. Ruffieux P, Stockar UV, Marison IW (1998) J Biotechnol 63:85– 131. Barackman J, Prado I, Karunatilake C, Furuya K (2004) J
95 Chromatogr A 1043:57–64
95. Navratil M, Norberg A, Lamben M, Mandenius C (2005) J 132. Kim M, Chung H, Woo Y, Kemper MS (2007) Anal Chim Acta
Biotechnol 115:67–79 587:200–207
96. Noll T, Biselli M (1998) J Biotechnol 63:187–198 133. Park SC, Kim M, Noh J, Chung H, Woo Y, Lee J, Kemper MS
97. Mishima K, Mimura A, Takahara Y, Asami K, Hanai T (1991) J (2007) Anal Chim Acta 593:43–53
Ferment Bioeng 72(4):291–295 134. Rathore AS, Li X, Bartkowski W, Sharma S, Lu Y (2009)
98. Mishima K, Mimura A, Takahara Y (1991) J Ferment Bioeng 72 Biotech Prog 25:1433–1439
(4):296–299 135. Mets H, Mader K (2008) Int J Pharm 364:170–175
99. Odman P, Johansen CL, Olsson L, Gernaey K, Lantz AE (2009) 136. Nordon A, Littlejohn D, Dann AS, Jeffkins PA, Richardson MD,
J Biotechnol 144:102–112 Stimpson SL (2008) Analyst 133:660–666
100. Bachinger T, Riese U, Eriksson R, Mandenius C (2002) Biosens 137. Genin N, Rene F, Corrieu G (1996) Chem Eng Process 35:255–
Bioelectron 17:395–403 263
101. Kemp KB, Guan Y (1998) Thermochim Acta 309:63–78 138. Zhou GX, Ge Z, Dorwart J, Izzo B, Kupura J, Bicker G, Wyvratt
102. Goudar C, Biener R, Boisart C, Heidemann R, Piret J, Graaf A, J (2003) J Pharm Sci 92(5):1058–1065
Kon stantino v K (2009 ) Metab Eng doi:10 .1016 /j. 139. Velardi SA, Hamouri H, Barresi AA (2009) Chem Eng Res Des
ymben.2009.10.007 87:1409–1419
154 A.S. Rathore et al.

140. Gieseler H, Kessler WJ, Finson M, Davis SJ, Mulhall PA, Bons 149. Johnson R, Kirdar AO, Annamalai A, Ahuja S, Ram K, Rathore
V, Debo DJ, Pikal MJ (2007) J Pharm Sci 96(7):1776–1793 AS (2007) Biopharm Int 20:130–144
141. Barresi AA, Pisano R, Fissore D, Rasetto V, Vallan A, Parvis M, 150. Kansakoski, Markku, Kurkinen, Marika, Weymarn, Niklas,
Galan M (2009) Chem Eng Process 48:408–423 Niemela, Pentti, Neubauer, Peter, Juuso, Esko, Eerikainen, Tero,
142. Martin EB, Morris AJ (2002) J Biotechnol 99:223–235 Turunen, Seppo, Aho, Sirkka and Suhonen, Pirkko (2006) VTT
143. Kourti T (2004) Process Anal Technol 1:13–19 Working Papers 1459-7683 http://www.vtt.fi/publications/index.
144. Barnes RJ, Dhanoa MS, Lister SJ (1993) J Infrared Spectrosc jsp Accessed on 14 January 2010
1:185–186 151. Wold S, Cheney J, Kettaneh N, McCready C (2006) Chemometr
145. Roggo Y, Chalus P, Maurer L, Lema-Martinez C, Edmond A, Intell Lab 84:159–163
Jent N (2007) J Pharm Biomed Anal 44:683–700 152. Vaidyanathan S, Arnold SA, Matheson L, Mohan P, McNeil B,
146. Teixeira AP, Oliveira R, Alves PM, Carrondo MJT (2009) Harvey LM (2001) Biotechnol Bioeng 74:376–388
Biotechnol Adv 27:726–732 153. Undey C, Ertunc S, Cinar A (2003) Ind Eng Chem Res 42:4645–4558
147. Geladi P (2003) Spectrochim Acta B 58:767–782 154. Moller SF, Von Frese J, Bro R (2005) J Chemometrics 19:549–563
148. Pedersen DK (2002) PhD Thesis Spectroscopic and chemo 155. Kirdar AO, Conner JS, Baclaski J, Rathore AS (2007)
metric exploration of food quality submitted to Department of Biotechnol Prog 23:61–67
Dairy and Food Science The Royal Veterinary and Agricultural 156. Gunther JC, Conner JS, Seborg DE (2007) Biotechnol Prog
University Denmark, 16-29 23:851–857

You might also like