You are on page 1of 31

USP Technical Guide:

Control Strategies for


Continuous Manufacturing of
Solid Oral Dose Drug Products
USP Technical Guide

How to use this Technical Guide

This document is intended to serve as a helpful illustration


of the process of developing control strategies for solid
oral dose drug products. Although there are guidance
documents available [ICH 2023] for pharmaceutical continuous
manufacturing in general, and for specific process components
such as process analytical technologies, this document
presents a more detailed conceptual and practical discussion of
control strategy limited to solid oral products.

This document is intended for informational purposes only.


The document does not reflect the opinion of any USP Expert
Bodies on future revisions to official text of the USP–NF. The
recommendations herein are not intended to be binding in
any way, nor should they be considered the only acceptable
approach. Parties using this document are solely responsible
for making decisions regarding its suitability for use. USP
will not be responsible for the use or implementation of this
document, or any information contained herein. None of
the recommendations or other information in this document
constitute a legal opinion or advice, or an opinion or advice on
regulatory matters. Parties relying on the information in this
document bear independent responsibility for awareness of,
and compliance with, any applicable federal, state, or local
laws and requirements.

© 2023 The United States Pharmacopeial Convention, 12601 Twinbrook


Parkway, Rockville, MD 20852.

Help us continually improve and update this document. USP


welcomes your valuable feedback, email PCM@usp.org.

2
2
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

Contents
4. PAT in the Development of Control Strategy 22
1. Introduction 4
4.1. Introduction: PAT-based and Parametric
1.1. Pharmaceutical Continuous
Quality Assessment 22
Manufacturing (PCM) 4
4.2. Spectroscopic PAT Tools 22
1.1.1. Continuous Manufacturing for Oral Solid
Dose Drug Products 4 4.2.1. Near-Infrared Spectroscopy 22

1.1.2. Continuous and Semi-continuous 4 4.2.2. Raman Spectroscopy 23

1.2. Regulatory Guidance and Requirements 5 4.3. Chemometrics 24

1.3. Control Strategy: Origins 5 4.3.1. Sampling and Data Quality 24

1.4. Key Terminology 6 4.3.2. Preprocessing 24

1.4.1. Quality by Design 6 4.3.3. Variable Selection 24

1.4.2. Quality Target Product Profile 7 4.3.4. Quantitative Predictive Modeling 25

1.4.3. Critical Quality Attributes 7 4.3.5. Model Optimization and Validation 25

1.4.4. Continuum Approach to CQA Classification 8 4.3.6. Model Maintenance and Lifecycle
Management 25
1.4.5. Critical Process Parameters 9

1.4.6. State of Control 9


5. Application to Oral Solid Dose, Drug
1.5. Control Strategy: Steps 9
Product Manufacturing 26

5.1. Solid Oral Dose CM Methods and their


2. Elements of Control Strategy: Part I 12 Specific Control Strategies 26

2.1. Material Attributes 13 5.1.1. Direct Compression (DC) 26

2.2. Process Monitoring and Control 15 5.1.2. Wet Granulation 26

2.2.1. Process and Unit Operation Modeling 15 5.1.3. Dry Granulation 27

2.2.2. Process Control Systems 16 5.2. Feeding and Blending Unit Operations 27

2.2.3. Operational Procedures 16 5.3. Example Control Strategy 28

2.2.4. Equipment and System Integration 17

6. Appendix 29
3. Elements of Control Strategy: Part II 18

3.1. Risk Assessment 18 7. References 31


3.2. Material Characterization 18

3.3. Experimental Design 19

3.4. PAT Development 20

3.5. Full-Line RTD, Performance, and Modeling 20

3.6. Out-of-specification Determination


and Diversion 20

3.7. Scale-up Considerations 20

3
USP Technical Guide

1 Introduction
1.1. Pharmaceutical Continuous and tablet presses. This process knowledge allows for the
streamlining of experimentation needed to support the
Manufacturing (PCM)
manufacturing process control strategy. Three commonly used
As per the ICH Q13 guideline, a pharmaceutical process approaches for oral solid dosage (OSD) drug products (DP) are:
is considered to be a continuous manufacturing (CM)
• direct compression (one or more blending steps of the
process if it involves a unit operation chain featuring a
drug substance [DS] and excipients followed directly by
“continuous feeding of input materials into, the transformation
tablet compression)
of in-process materials within, and the concomitant removal
of output materials from it.” The guideline further states that • dry granulation (compaction and milling of DS to improve
the process must have a minimum of two connected unit particulate properties prior to blending), or
operations featuring a flow of material from one operation
to the next. Hence, the manufacturing process for simple • wet granulation (water or organic solvent is added to the
solid oral dose drug products (e.g., tablets) could feature a pre-blend of the DS and excipient[s] to agglomerate fine
continuous feeding and mixing step for it to be considered a particles into granules and improve particle properties)
CM process. Further downstream, there are process designs
that differentiate between various solid oral drug product Other less commonly used approaches include hot-melt
process lines (described in detail in Section II[b]) which may extrusion (HME) and direct encapsulation. However, CM for
also run in a continuous mode. oral solids introduces new equipment with new modes of
operation, new means of operating and controlling familiar
While CM has been a modality of choice for many
equipment, new risks, and new fundamental principles that
manufacturing industries such as fine chemicals, foods,
are unique to this method of manufacturing. Due to these
and oil and gas, its implementation in the pharmaceutical
factors, enhanced development approaches (i.e., Quality by
industry is relatively new. Hence, several techniques and
Design [ICH 2009], use of process analytical technology (PAT)
practices are being adapted from other industries, and new
[FDA 2004], or use of process models) have been proposed
ones are being developed. Manufacturing and regulatory
for developing CM control strategies.
experience from the past two decades, however, has resulted
in the development of certain best practices in solid oral
drug product manufacturing. This guide is limited to control 1.1.2. Continuous and Semi-continuous
strategy considerations for CM of solid oral drug products.
A CM process in the pharmaceutical context does not
necessarily have to be “end-to-end”, nor is it expected that it will
1.1.1. Continuous Manufacturing for Oral Solid always run continuously for weeks or months. Most processes
Dose Drug Products have a well-defined operating time range before they are
stopped for cleaning and other maintenance. Often processes
Oral solid dose traditional batch process development
are run in one or two shifts per day. Not all unit operations have
has had the advantage of substantial product and process
to be run in a connected, continuous mode. Many successful
understanding stemming from historical experience with
CM processes have unit operations such as pre-blending,
equipment such as blenders, granulators, roller compactors,
sieving, milling, or tablet coating that are run in batch mode.

4
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

1.2. Regulatory Guidance and 1.3. Control Strategy: Origins


Requirements As per the ICH Q8(R2) Guideline [ICH 2009], control strategy
Regulatory guidance specific to developing CM control is defined as “A planned set of controls, derived from current
strategies has been provided through ICH Q13. In addition, product and process understanding that ensures process
an ASTM standard guide, E2968-14 [E2968], defines key performance and product quality. The controls can include
concepts and principles for the development and operation parameters and attributes related to drug substance and drug
of continuous processing technologies for pharmaceutical product materials and components, facility and equipment
manufacturing. These guidances and standards align with operating conditions, in-process controls, finished product
Quality by Design (QbD) principles and process design specifications, and the associated methods and frequency of
concepts described in the FDA Process Validation guidance monitoring and control.”
[FDA 2011]. The concept of a “state of control”—a condition
The definition is valid for both batch and CM operations. It
that provides continued process performance and product
is encouraged that control strategy be used during process
quality—is defined in ICH Q10 [ICH 2009]. The control
development and manufacturing and included in the regulatory
strategy includes elements to be monitored and parameter
filing in the ICH M4 Common Technical Document (CTD) format.
settings to be adjusted when necessary to maintain the state
It can be included within section 3.2.P.2 of this document,
of control. If necessary, non-conforming material is identified
referred to as the Pharmaceutical Development section.
and diverted from the process.
An OSD DP consists of several components such as the
The regulatory filing dossier for a CM process should be
DS, one or more excipients, and other substances such as
in accordance with the ICH guidance of industry M4:
lubricants, glidants, and stability enhancers. Figure 2 is an
The CTD – Quality (ICH M4Q) [ICH 2002].
illustration of this implementation of CM. This illustration
USP’s Quality Standards for Pharmaceutical Continuous shows a single loss-in-weight feeder equipped with a refill
Manufacturing Expert Panel (2015-2020) noted [Matos 2018]: system, a horizontal continuous blender, a tablet compaction
machine, temporary tablet collection bin, and a continuous
• A CM process is subjected to the same requirements coating machine. A process consisting of these steps
regarding current good manufacturing practice (cGMP) can be classified as a ‘direct compression’ process. Many
compliance as a batch process. formulations require preliminary steps before tableting,
such as wet granulation (followed by drying and milling) or
• Principles of risk management apply equally to batch and
CM processes.

• Pharmacopeial product specifications should be met Figure 1: Control Strategy is included in the Pharmaceutical
regardless of the mode of manufacture. Development section of the Common Technical Document

• Regulatory expectations for ensuring product quality as


well as reliable and predictive processing are the same
for batch and continuous manufacturing.

However, ICH Q13 provides additional requirements that are


CM-specific. Control strategy is included among these, with
the following specific components:

• Input material attributes


• Process monitoring and control
• System operation
• Material diversion and collection
• Real-time release testing (if implemented)
• Equipment and system integration
This document discusses these topics in detail.

5
USP Technical Guide

dry granulation (followed by milling). As far as feasible, it is Development of the control strategy usually starts with
preferable to use direct compression (DC) due to process specification of the Quality Target Product Profile (QTPP),
simplicity, lower cost, and fewer opportunities for quality which is the desired set of attributes of the product in
deviation. Given the available process understanding and question. This forms a part of the “predefined objectives” of
significant experience with DC, it is no longer considered an QbD, as defined in ICH Q8(R2).
emerging technology by the US FDA [FDA 2022].

1.4.2. Quality Target Product Profile


1.4. Key Terminology In ICH Q8(R2), the QTPP is defined as:
The control strategy, in simple terms, is a strategy that the
manufacturer deploys to ensure that the desired quality
“A prospective summary of the quality
attributes are achieved from the manufacturing process. characteristics of a drug product that ideally will be
The control strategy must be clearly defined and reported achieved to ensure the desired quality, taking into
in the appropriate sections of the Common Technical account safety and efficacy of the drug product.”
Document (CTD).
QTPP encompasses the attributes of the product that must
meet the claims on the label on the container in which
1.4.1. Quality by Design the product is sold. Examples of attributes in the QTPP
are dissolution (immediate release or sustained), dosage
ICH Q8(R2) defines QbD as:
strength, and mode of administration. Once the QTPP is
“A systematic approach to development that determined, the CQAs and associated critical process
begins with predefined objectives and emphasizes
product and process understanding and process
Figure 2: A basic solid oral DP manufacturing line beginning
control, based on sound science and quality risk
with loss-in-weight feeders and culminating in tablet
management.”
coating unit operation.

Basic Solid Oral DP Manufacturing Line


kg/hr
kg/hr
1
1 Feeders Refill System
2 Feeder
2 3 Mixer
4 Intermediate Hopper
5 Tablet Press
6 Tablet Coater
3 7 CM Batches

5
6

6
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

parameters (CPPs) can be determined using quality risk are those that affect product performance, along with the
management (QRM) [ICH 2006] and scientific rationale. severity of harm and the likelihood of occurrence. To identify
This is where development of an effective control strategy CQAs, one must list the quality attributes on the basis of
can begin. prior knowledge or experimentation, and one must consider
the severity of harm in terms of safety and efficacy that the
Since the output material can become the input material particular attribute can cause. How the CQAs link back to
of the downstream unit operation, output quality attributes the QTPP must also be clearly understood, and so should
become material attributes of the next input material. When the impact of any source of variation on the CQAs.
raw materials (DS and excipients) are first introduced into
the process, some material attributes are defined as critical It is best to define the CQAs, the rationale for selecting
material attributes (CMAs) if they impact one or more them, their acceptable limits (or the design space) and
finished or intermediate product’s CQA(s). However, CMAs ranges (high and low), and the protocols for their
are not officially defined in the ICH Guidelines. Attributes measurement. As an example, content uniformity of the
are either sampled or measured directly with inline/online dosage form can be a CQA while the product’s odor may
instruments (i.e., PAT). Test results of attributes are assessed not be. The target profile must be stated, for example,
against control limits (if used as part of feedback or feed whether the content uniformity conforms to USP General
forward control loops to adjust process parameters) or Chapter <905> Uniformity of Dosage Units.
against acceptance limits (if used for rejection or diversion of
non-conforming products). During the manufacturing process, there are many process
parameters that could affect a product’s quality and its
performance. Some parameters have a pronounced effect
1.4.3. Critical Quality Attributes while others may have an insignificant impact (this is
assessed using the criticality analysis). The parameters that
ICH Q8(R2) defines a CQA as: significantly affect product performance can be considered
as the Critical Process Parameters (CPPs). These parameters
“A physical, chemical, biological or microbiological
must be given special consideration when creating the
property or characteristic that should be within an
control strategy, and their mutual interplay and impact on
appropriate limit, range, or distribution to ensure
CQAs must be well understood.
the desired product quality.”
In addition to the DP CQAs, a set of intermediate CQAs can
The ICH guidelines state that thorough risk assessments
be identified. These are the CQAs of intermediate material
must be carried out. The factors that contribute to risk

7
USP Technical Guide

during the manufacturing process. For example, if content strategy on those parameters and material attributes with the
uniformity is the DP CQA, the blend uniformity can be the greatest impact on quality.
intermediate CQA. In a CM process, the associated CPP can
be a mixer parameter such as impeller speed or flow rate.
1.4.5. Critical Process Parameters

1.4.4. Continuum Approach to CQA Process parameters are inputs to the unit operation; if
Classification they impact one or more CQA(s), they are defined as
CPPs. Process parameters are predefined settings that are
A continuum approach may be adopted so that the “impact” determined either in advance or based on the output of a
is interpreted in relative terms of significance and measurable control loop (feed forward or feedback). Even though fixed
effects on CQAs. settings may be predetermined, they may vary around a set
point within a proven acceptable range over time.
The following is a proposed classification of criticality of
material attributes and process parameters: The ICH Q8 guidance only defines CPPs as relative to their
impact on CQAs. CMAs can be similarly defined relative to
• High impact: shift within parameter control limits that are DP CQAs. A binary approach to defining criticality (critical
likely to cause CQA to go out of limits. or not critical) of parameters can be limiting since the ICH
• Medium impact: shift within parameter control limits that guidance uses the imprecise terminology, “whose variability
are likely to cause measurable shift in CQA but remains has an impact.”
within limits.
The Quality by Design approach to control strategy
• Low impact: shift within parameter control limits that are development starts with the development of the QTPP, which
likely to cause minor, but measurable shift in CQA.
• No impact: shift of parameter creates no measurable shift
Figure 3: Relationship between CMAs, CPPs, and QTPP,
in CQA (or has no mechanistic relationship).
exemplified by a coated tablet product. PSD is particle
Applying this approach allows for a simpler way to categorize size distribution.
criticality and also a way to appropriately focus the control

CPPs CPPs CPPs

Unit Operations 1 Unit Operations 2 Unit Operations “n”

Feeding Mixing Coating


Feed Rate Content Unit Coating Drug
Intermediate Operations Product
CMAs Uniformity Thickness
Uniformity CQAs Chain CQA
Shear Coating Quality
Uniformity
Assured
Color
Product

CPPs
Examples of CMAs Drug Product CQAs
f Particle Size f Bulk & tapped f Potency
f PSD density f Efficacy
f Cohesion f Electrostatic f Bioavailability
f Flowability affinity

8
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

includes elements describing how the quality of the product to be negligible or controllable, resulting in no or
is achieved. Examples of QTPP attributes for OSD include minimal impact on product quality. Larger changes
appearance, mode of administration, drug release (immediate in process variables and/or quality attributes may
or extended), dosage strength, degradation pathways, stability, happen when the CM process is in a transient state,
and primary packaging requirements. Additionally, the QTPP such as during start-up, shut down, changes from
includes the first descriptions of CQAs. In general, most one operating condition to another, and significant
CQAs will be included as testing attributes on the finished disturbances (e.g., equipment failure or sudden
product’s specification, but some CQAs may be tested as in- change of raw material attributes).”
process intermediates. CQAs are derived from general drug
requirements (safety, identity, strength, purity, and quality)
and specific dosage form requirements (dissolution, content 1.5. Control Strategy Steps
uniformity, microbial requirements). For the purposes of
this control strategy development exercise, the focus is on While the decision on when to start process development
quality attributes (impacting product quality); there are also can be taken with practical feasibility in mind, several
non-quality or performance attributes such as yield, cost, companies have reported starting it in parallel to the clinical
throughput, and processing time that may need to be taken development of the product. The following are the important
into consideration during the control strategy definition. considerations for designing a control strategy for a process
such as the one shown in Figure 2:
Initially, a manufacturing scheme is chosen to produce the
OSD DP, either via direct compression, dry granulation, or I. Equipment design, configuration, and integration in
wet granulation. Each unit operation of the process has the CM equipment train: Specific attention must be
input materials with defined material attributes, process given towards harmonizing the individual equipment
parameters, and output material with defined quality inputs and outputs and identifying risk spots such as
attributes; see Figure 3. locations of build-up or fouling.

The CQAs can be expressed as dependent on (a function


II. Material characterization and compatibility checks:
of) all CMAs and CPPs (Figure 3). While the CMAs and CPPs
The formulation (qualitatively and quantitatively) should
are relevant during manufacturing, the resulting CQAs are
be designed keeping in mind the commercial equipment
relevant for the product performance. The final outcomes,
train constraints [particularly the loss-in-weight
safety and efficacy, are ultimately relevant to the patient.
feeders (LIWFs) performance]. The excipients and drug
substance(s) must be characterized in terms of density
1.4.6. State of Control ranges and particle size distribution specifications.
These attributes are usually proven to be critical
USP’s Quality Standards for Pharmaceutical Continuous towards one or several DP CQA(s) as they are often
Manufacturing Expert Panel (2015-2020) noted that demonstrated to impact blend and content uniformity,
homogeneity, process parameters, and quality attributes are as well as dissolution rate. The material characterization
described as a function of time in CM as opposed to batch is conducted using industry-standard in-process control
[Matos 2018]. The panel further elaborated on the state of (IPC) equipment as well as with testing on LIWFs (either
control as: at smaller scale or directly at the commercial scale).
Based on the DP quantitative formulation and the
“A state of control does not imply the process is process throughput range, the process designer will
at steady state. Continuous unit operations can have to demonstrate that the LIWFs can deliver every
be designed to operate at a steady state, but in raw material at the required throughput range. Several
practice, a CM process does not run at a steady- specific conditions should be tested (startup, shutdown,
state condition, but rather at a condition in which LIWF hopper refill, change in line throughput, impact of
a set of critical process parameters and/or quality increased runtime, etc).
attributes are kept within a specified range of
target values (state of control). Deviations from
III. Residence Time Distribution (RTD) model
these target values triggered by disturbances do
implementation: Process dynamics must be understood
generally occur during the normal operation, but
to track the material and to support the development of
they can be detected and are often small enough
sampling and diversion strategies.

9
USP Technical Guide

Even if the RTD can be defined based on in-silico (e.g., content, moisture, PSD, coating thickness)
modeling, confirmatory trials at commercial scale and measurements are taken under varying
are still needed. As these trials are performed at the conditions (e.g., range of temperature, flow-rate
beginning of the CM process establishment, a tracer is disturbances) and compared to a reference. Special
generally used for the RTD definition. It could be either attention is given to the whole signal chain and
a visually detectable colored tracer (e.g., Fe2O3) or a process: speed of measurement/data acquisition,
chemical tracer easily quantified in samples drawn. speed of conveying the data from the probe to the
Supervisory Control and Data Acquisition (SCADA)
Based on the RTD model, the material diversion strategy system, data processing speed (transforming the
can be set up at this stage. As the material tracking is signal into values) and machine response speed
provided, the diversion of potentially non-compliant (for example, opening and closing of a diverting
material is enabled. Calculations can be performed to valve) ensuring that the non-compliant material in
determine the quantity of wet granules, dry granules, eliminated from the equipment train.
uncoated or coated tablets to be discarded at each
diversion point, depending on the line throughput. It b. Once the models are defined statically, the second
is recommended to be conservative while doing these step is to verify them in dynamic conditions
“diverting” calculations to ensure that all the material applying the same mindset. In this step, the
maintained in the equipment train meets the desired granules, blends, uncoated tablets and coated
specifications (in-spec). tablets (compliant or not) will be generated by
running the CM line under different parametric
IV. Risk Analysis: Manufacturing process risk analysis settings. The results generated by the PAT and the
(MPRA) is conducted either in parallel with, or after reference method on specific collected samples are
the RTD tests. The outcome of this exercise is the compared. Ultimately, the diverted material (due to
identification of potential CPPs (pCPPs), and potential the PAT measurement leading to an opening of the
CMAs (pCMAs) based on their magnitude of impact on diverting valve) will have to be analyzed with the
DP performance (DP CQAs). This MPRA step is a paper reference method and its non-compliance will have
exercise which creates a foundation for the final control to be verified. The same treatment is applied to the
strategy. The confirmation of the criticality (or lack “good” material maintained in the CM train right
thereof) of the identified pCPPs and pCMAs must be after the closing of the diverting valve. Identification
demonstrated via the Design of Experiments (DoEs). of the sample to be drawn will be only possible
The DoEs can be conducted before or after the PAT because of the RTD models generated earlier.
models implementation and validation. Depending
on the status of (and level of experience with) the PAT VI. Equipment Qualification: Once the list of CPPs and
models, the responses used for model creation could CMAs is available and the PAT models established, it is
either come solely from sampling and characterization, of primary importance to understand the relevant input
or partially from them (and partially from the PAT factors that affect the process performance. These inputs
models). The MPRA combined with the DoEs (screening are identified thanks to the models (outputs from the
and optimization) will lead to the final list of CPPs and optimization DoE[s]) via a sensitivity analysis. A sensitivity
CMAs. A control strategy is mandatory to ensure the analysis assesses how variability in the model inputs
DP CQAs will always be compliant during the lifecycle contributes to variations in model outputs.
of the manufacturing process.
As a simple example, the dissolution rate of tablets has
V. PAT implementation: Establishment of PAT model(s) is been identified either by simulations or by output from
required to run a PAT-enabled CM process. Experiments the DoE(s) to be a function of water flow rate introduced
for this purpose can start after the characteristics to be during wet granulation, milling speed, compression force
measured are defined, PAT probes selected, and their and compression speed. After carrying out the sensitivity
locations determined. analysis (based on trials conducted on the line), the
model error will be quantified and the contribution
a. Initially the measurements are done statically, i.e., (weight) of each CPP on the DP CQA (dissolution rate)
on fixed samples. The samples are manually taken will be precisely calculated.
from experiments covering a range of parameters

10
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

VII. Verification trials: At this stage, the verification trials shutdown to clean the probes), clogging of the
on the commercial line can be carried out since the filters, etc.
CPPs and CMAs have been identified, and the control
strategy is in place with the support of the IPCs, PAT VIII. The batch size range (or setpoint) will have to be
sensors and control loops. These can include: defined taking into account the previous conclusions
but also other factors like:
a. Design space verification runs (runs at the “edges”
of the design space) i. Supply chain constraints (e.g., launch schedule,
country-specific regulatory considerations,
b. Test of the minimum and maximum throughput on future forecasts)
the line
ii. Economic constraints, such as the resources
c. Disturbance tests (e.g., filling of hoppers, needed to assemble, disassemble, and clean
vibrations from the high-speed tableting machine), the line.
tests of diverting valves operation, checks for
fouling of the probes (defining the need for line The flow chart in Figure 4 demonstrates a stepwise process
for OSD DP manufacturing. The stages that contribute to the
development of the control strategy are indicated.

Figure 4: Typical CM solid oral dose manufacturing process stages with specific steps that contribute to the establishment of the
control strategy.

Define initial formulation, establish


1
QTPP, production range

Identify, install, and test equipment,


2
instrumentation
Control
Strategy
3 Characterize material attributes

Define design space and normal


4
operating ranges

Design of experiments: establish


5
CQA, process control strategies CPPs

Production; Data collection CPPs


6 Process monitoring: PAT, modeling
OOS detection & diversion CQAs
CPPs
Real-time release
CPP is critical process parameters,
Continuous process validation QTTP OOS is out of specification.

11
USP Technical Guide

2 Elements of Control
Strategy: Part 1

A control strategy for solid oral dose CM should include [ICH 2023] assessment of input material attributes, process
monitoring and control, operational procedures, and equipment and system integration aspects. These are described in detail
in this section.

Elements of a CM Control Strategy

Control
Strategy

Process Equipment
Material Monitoring Operational Operation &
Attributes & Control Procedures System
(PMC) Integration

Figure 5: Elements of a CM control strategy

12
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

2.1. Material Attributes accumulation. For example, powders sticking to the outlet
of loss-in-weight feeders can result in uneven flow rate
Continuous processes can be more susceptible than batch and accumulation of ingredients which drop into the
processes to failure due to certain material attributes feeder as an agglomerate. Not only does this destabilize
such as adhesion to equipment and electrostatic charge the formulation on an ongoing basis but it can also cause
spikes of potent active material in the flow stream with
catastrophic implications for quality.

Such issues can be avoided by careful selection of


Figure 6: A schematic example of a direct compression equipment such as feeders with agitators, different screw
manufacturing train and associated potential failure designs, use of flow enhancers, and periodic equipment
modes as well as material properties impacting them and cleaning. Issues can also be prevented by developing a
respective methods for their characterization. deep understanding of unit operations, gained through




13
USP Technical Guide

Unit Operation Failure Mode Material Property Characterization Technique

f Tapped density
Flow rate set point
f Density f Bulk density
divergence
Feeding f Permeability

f Cohesion f Shear Cell


Flow rate variability
f Adhesion f Electrostatics measurement

f Surface energy f Inverse gas chromatography

Agglomerate f Particle size


Milling f Droplet shape analysis
cohesion distribution

f Laser diffraction

f Surface energy f Inverse gas chromatography

Poor mixing/ f Particle size


f Droplet shape analysis
segregation distribution

f Laser diffraction
Blending
Choking, jamming,
— ­­ —
discontinuous flow

f Inverse gas chromatography


f Surface energy
Dry coating
or Contact angle
f Droplet shape analysis

f Inverse gas chromatography


f Surface energy
Dry coating
or Contact angle
f Droplet analysis

f Bonding
Failure to comply
f Plastic f Fitting compaction model
breaking force
Tableting
f Elastic

f Bonding
Failure to comply
f Plastic f Fitting compaction model
dissolution
f Elastic

Table 2: Examples of potential failure modes, respective


material properties, and characterization techniques for the
manufacturing line depicted in Figure 6.

14
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

well-planned experimental campaigns. Prior determination 2.2. Process Monitoring and Control
of normal operating range and design space of each unit
operation is encouraged. Other unit operations also have Process monitoring and control can be achieved using a
materials-related failure modes, and a thorough risk analysis combination of tools such as:
must be performed and included in the control strategy.
• Spectroscopic PAT instruments (e.g., near infrared,
Figure 6 depicts common failure modes of a simple direct
Raman)
compression equipment train. Input materials in a typical
OSD DP process include: • Process measurements (e.g., feed rate, mass flow rate,
compression force)
• DS or the active pharmaceutical ingredient (API)
in powdered form • Models (e.g., residence time distribution, process
models)
• Excipients
• Soft sensors
• Lubricants and/or glidants
Section VI: PAT in the Development of Control Strategy
• Binder
describes these tools in more detail. The aim of using these
• Water tools is to achieve and maintain the process in a state of
control. This is facilitated by real-time data collection and
• Coating components processing followed by control decisions being made
(Figure 7).

2.2.1. Process and Unit Operations Modeling

Unlike batch processes, continuous processes are


dynamic. Changes to operating conditions, whether
Figure 7: PAT, soft sensors, and models facilitate real-time data planned (adjustment to a parameter, start up, shutdown)
collection, thereby helping maintain a state of control which or unplanned (unexpected change to material properties
can eventually lead to continued process verification (CPV). or equipment performance), can propagate from one unit

Elements of Process Monitoring and Control

Process Analytical
Technologies (PAT)
State of Control

Real-time data Continuous Process


Soft Sensors
collection Verification (CPV)

Models

15
USP Technical Guide

target/range. Example: Measuring assay at the output of a


blender in order to adjust the feed rate of DS at the inlet.

• Feed forward control: This entails measuring attributes


at the inlet of a process and using a model to determine
the necessary set point for the process. Example:
Measuring the DS prior to a blender and adjusting the
addition rate to target 100% in the blend.

• Multivariate model-based control (also referred to


as parametric control): Attributes that were measured
previously during model development are correlated
with process parameters. The process parameters are
monitored and constrained to ensure that predicted
attribute values remain within acceptable limits.
Attributes may be monitored to compare them to model
predictions. Example: Feed rate and drying temperature
for a continuous dryer are correlated to achieve a target
moisture level, which is then monitored at the exit and
compared to model predictions.

Both feedback control and feed forward control rely on the


operation to the next and disrupt the state of control. response time of measurement instruments and a clear
One approach to understanding process dynamics is understanding of the RTD model for every unit operation.
the concept of residence time distribution (RTD). RTD Otherwise, control responses to disturbances occur too late
characterizes the time available for material transport to be effective. The reliability of the instruments (such as
and transformation, and it is specific to the process, build up and calibration drift) must be considered. Although
composition/formulation, material properties, equipment advanced control approaches such as feedback or feed
design, and configuration [E2968]. RTD is the degree forward control have been recommended by regulatory
of back mixing in that it ranges from plug flow (RTD agencies, their implementation can be complex and should
approaches zero) to fully back mixed. It determines the not be the default approach to control strategies.
quantity of an ingredient of interest (e.g., API) in the output
material for the process (important for material traceability)
and determines when a change to the input material will 2.2.3. Operational Procedures
impact the output material (important for process control
Operational procedures include considerations such as:
and sampling). Process dynamics directly impact the use
sequence of events during start-up, use of surge tanks or
of online monitoring and control (PAT). Understanding
intermediate bulk containers (IBCs) to maintain line rate (if
RTD is critical in the placement of instruments (PAT tools)
any), material transfer procedures (pneumatic, belt conveyed,
and how they can respond to disturbances in real time for
etc.) and the potential impact of these choices on quality, on
adjustments and/or diversion strategies.
how acceptance or rejection decisions are made, etc.

CM start-up and shut-down processes are transient, and


2.2.2. Process Control Systems
unless an achieved state of control has been demonstrated,
In a dynamic process, material attributes, quality attributes, the product cannot be considered acceptable. Therefore, the
process parameters, and equipment conditions will not control strategy should include a description of the procedure
remain static, and therefore a mechanism of control may be for starting up the production line and information on which
needed. Categories of control strategies [E2968] include: signals and calculations are used to conclude the steady state
has been achieved, including how the product is diverted until
• Feedback control: This entails measuring the attributes then. Similarly, a cut-off point must be specified before the
at the exit of a process and comparing them to a target line shutdown begins. Also, the fate of the material in the line
value or range. A model is used to adjust parameter while this process (startup/shutdown) is on must be described.
settings to bring the measured attributes back to

16
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

In many situations, material must be held up at one point divert or not divert should be made. This is an example of
for a certain duration before the next unit operation is ready feed-forward control.
to accept it. This can occur when unit operations require
more time to process material than others, or the product
itself requires some time (e.g., tablets needing a relaxation 2.2.4. Equipment and Systems Integration
period between compression and coating operations). The
Finally, the ICH Q13 guideline states that those aspects
control strategy must describe the procedure, the duration
of the equipment design that are shown to be critical to
of hold-up in an intermediate bulk container, and the mode
the material quality and its control should be described
of transfer of such material.
and justified. For example, the blade configuration in a
Finally, if during normal processing, the control system continuous mixer can be changed so they always face
triggers an alarm that requires material to be temporarily forward, or they can be set to be a mix of forward and
diverted out, the complete sequence of events and backward facing. During the unit-operation characterization
procedures must be listed in the control strategy. For stage, the amount of back-mixing and material hold-up in
example, in a direct compression process, a feeder control the blender is determined for different blade configurations
system triggers an alarm indicating that a big chunk of API at given impeller speeds. The mixer RTD is hence known.
is exiting the feeder and entering the next unit operation If the blade configuration is changed, it can impact the
(mixing). The process must have been characterized a priori amount of time the materials have available to mix and by
to understand the residence time distribution of the API extension, the degree of shear they experience. During
through the mixer and into the tablet press. Data from the mixing of the lubricant, for example, the amount of shear
in-line sensor (which could be installed in the feed frame of is an important factor. Over-lubricated blends can lead to
the tablet press or in a transfer chute after the mixer) must failures in dissolution performance testing. For this reason,
be analyzed to inform the control system, and a decision to any such design changes must be justified.

Developing a Control Strategy

DP
Ishikawa (Fishbone) CQAs
Diagram

Risk Analysis

Variables ranked, sorted


Control
Strategy

DoE (Design of Experiments) Enhanced Process


and Design Space Understanding

Figure 8: An approach for developing a control strategy that keeps the desired quality attributes as the central focus during
process design.

17
USP Technical Guide

Elements of Control
3 Strategy: Part II

An effective control strategy describes all relevant material relative impact is estimated (using historical process knowledge
attributes that can impact product quality and also and understanding) for relevant CQAs. The relative, expected
demonstrates how the manufacturer will control them. control level is also considered (e.g., if a parameter is fixed,
Furthermore, it identifies the procedures to be used for it has no variability to impact quality, and if a parameter is
in-process controls. PAT is commonly used to assess process tightly controlled, its impact may be relatively low).
performance against predetermined metrics. Mechanistic and
process models, as well as soft sensors, can also be used to Process parameters and material attributes having high and
support real-time decision making. If used, these form a part medium impact are likely to impact CQAs and, therefore,
of the control strategy and should be validated and reported in must be included with process experimentation (DoEs)
the appropriate section. The manufacturing equipment must when a range is desired for these factors. The DoEs are
be qualified to show it is able to perform as expected within designed to determine the actual magnitude of effect on
the desired line rate (flow rate). A unique feature of CM is that one or more CQAs. Consequently, at the conclusion of DoEs,
decisions regarding product acceptance/rejection can be the risk assessments are revisited and re-scored based on
made in real time, and any out-of-specification (OOS) material the additional process knowledge achieved. If high-impact
or product can be rejected using a diversion mechanism built parameters and material attributes remain, then these
into the manufacturing line. unit operations are considered for more rigorous control
strategies including in-process control using PAT.

3.1. Risk Assessment 3.2. Material Characterization


Once a list of process parameters and attributes has been
CM depends on the consistent gravimetric feeding of both
created for each unit operation, the next step is to apply
the DS and the excipients. Material characterization and
first principles and existing process understanding to
particle engineering for properties may be required. Powder
make an initial assessment of the process risks to CQAs. A
build-up, hopper bridging, surface adhesion, agglomeration,
Failure Mode and Effects Analysis (FMEA) is a common risk
and dusting are all potential problems that can interfere with
assessment tool but may be difficult to score when used in
consistent powder flow. Since multiple batches of material
early process development. An alternative, simple risk matrix
may be added to the same feeder, batch-to-batch variability
is shown in Figure 9.
must be minimized. Once a particular excipient has been
This matrix is applied at the unit operation level, first to developed to have optimal flow for a CM process, it may be
score the relative impact of that unit operation on final and used across multiple products without further development.
intermediate CQAs. Note that the intermediate CQAs listed
The accuracy of gravimetric feeding is evaluated in feeding
are also potential in-process controls. Where high or medium
trials. Although feeders are available in a variety of flow rates,
impact is shown, it is highly likely that unit operation has
low-flow-rate feeding remains a challenge for low potency
CPPs and/or CMAs.
DSs and low-proportion excipients (such as magnesium
Once the higher-impact unit operations are identified, stearate and silicon dioxide). Strategies such as pre-mixing of
a parameter (and material attribute) risk assessment is some components or proportional feeding (linking feeders
performed on that unit operation. For each factor, the in a master/slave control scheme) may be required. Refilling,

18
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

Figure 9: Example of simple risk matrix for CM of OSD: How unit operations impact intermediate and final CQAs.

Unit Operation
CQAs
Wet
LIW Feeder Drying LIW Feeder Blending Compression
Granulation

Appearance No Low No No Low Medium

Assay High High No Medium Medium No

Content Uniformity No Low No Medium High Medium

Dissolution No High No No Low Low

Impurities No Medium Medium No No No

Water Content No Medium High No No No

Intermediate CQAs (In-Process Controls)

Blend Uniformity No Medium Low Medium High No

Water Content No High Medium Low Low Low

Weight Check No No No No No High

Hardness No No No No Low High

Risk Of Impact, Small Risk


High = Risk Of OOS Medium = Low =
But In Spec Of Impact

No Potential
No = LIW = Loss-In-Weight
Impact

perturbations, noise, and vibration can all impact feeding operations. In early studies, it is acceptable to substitute a
accuracy and are included as part of these studies. suitable placebo to save the cost of the DS. As operating
ranges are refined, the DS is added to the formulation so
that RTD models can be developed for each unit operation.
3.3. Experimental Design However, a tracer may be used instead of the DS for RTD
studies. Tracer characterization and rationale are critical.
Initial feasibility studies usually begin with a direct
When selecting a tracer, the critical physical attributes should
compression process due to its simplicity and cost-
be as comparable to the DS characteristics as possible.
effectiveness, but if direct compression is not possible,
Excipients should also be included in model development.
dry or wet granulation unit operations are used.
The next step is the application of DoEs to define design spaces.
First, engineering/placebo studies are typically done with
Initial screening designs may use a placebo where possible.
the objective of evaluating the parameter ranges of unit

19
USP Technical Guide

Factorial studies will finalize the design spaces and confirm • Medium: Can affect product quality, but there are
that DP CQAs are met. Once the uncoated tablet properties other engineering controls available to mitigate any
have been achieved, a coating process DoE is performed, if redundancies due to false prediction.
required. Unlike batch processing, different DoE conditions
can be run in succession after allowing for conditions • High: The sole control/assurance of product quality.
to stabilize per the RTD model. This allows for more DoE
Low- and medium-impact models are the most common
conditions and more samples to be tested than is done
types in control strategies. Where the impact is high to
during batch process development.
medium, validation of the model will likely be necessary.

3.4. PAT Development


3.6. Out-of-specification Determination
PAT instrument and control development will proceed in and Diversion
parallel with later DoE studies. Measurements of identity,
Most CM manufacturing trains deploy a diversion system
blend uniformity, content uniformity, loss on drying, and
to direct any material deemed to be OOS to scrap
particle size distribution are commonly done using online PAT
or for further analysis. As described in Section 1.5, a
instruments. However, not all unit operations may require PAT
comprehensive analysis of the process dynamics will lead to
if parametric controls maintain low variability of the attribute.
an understanding of how a disturbance propagates through
Typically, as process understanding increases, the use of PAT
the line. For example, to describe how a disturbance in the
tools decreases, allowing for a robust control strategy that
mass flow rate of an ingredient, often starting from a feeder,
is not overburdensome for data collection and reporting. A
would propagate through the system and affect one or more
risk assessment is appropriate to determine whether PAT is
DP CQAs. Material traceability along the manufacturing
needed for the control of the unit operation.
line is required to divert any non-conforming product. In
simple words, the control system must know precisely, and
3.5. Full-Line RTD, Performance, at all times, “what” material is “where” and “when” in the
and Modeling integrated line.

Overall process control software to coordinate multiple unit


operations will be needed and optimized during full-line 3.7. Scale-up Considerations
performance runs. It is not uncommon for the development
For batch processing, development is done on a smaller
and optimization of the control software to take significant
scale than the expected commercial scale will require. This
time during process development. During these runs, the
allows for performing multiple development experiments
full-line RTD model is verified and the diversion strategy for
while using smaller quantities of expensive materials such
non-conforming materials is finalized.
as the DS. Afterwards, however, a series of scale-up runs
RTD models may be simulated in silico through a digital twin is needed to ensure that the small-scale process control
but are typically developed experimentally using spiking strategy can be translated to the larger commercial scale.
or tracer studies. Consequently, the RTD models cannot CM has three approaches to scale up:
anticipate unexpected disturbances and non-conformities
1. Development uses the same scale equipment
(e.g., poor material flow, vibration, or product build up). More
as commercial, but the run-time is increased for
commonly, a PAT instrument is needed to not only detect
commercial manufacturing.
and respond to these disturbances, but also to provide a
redundant control to the RTD model. 2. Development uses the same scale equipment as
commercial, but the mass flow rate is increased for
When using process models, their impact on product quality
commercial manufacturing.
can be categorized as follows:
3. Development uses smaller-scale equipment than
• Low: Information only, which does not impact process or
commercial, and a scale-up transfer is required for
product quality strategy.
commercial manufacturing.

20
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

Approach #1 provides the shortest development time and A suggested hybrid approach for a new product is to begin
lowest cost, but also requires an investment and installation with smaller-scale equipment, but to use this only for
commitment to equipment sized at the same scale as process feasibility and early development studies. Many
the eventual commercial line. Approach #2 has a distinct CM equipment suppliers have development facilities for
disadvantage, since changing the mass flow rate through this express purpose. For the later process design, the
the equipment impacts the RTD for the process steps, R&D equipment should be scaled to match the eventual
which in turn requires new modeling and control strategy commercial line. This is to ensure that the commercial
development. Approach #3 has a similar disadvantage to #2 line has the same mass flow rate and that scale-up can be
since the control strategy will not transfer directly to larger- achieved by a longer run-time. This turnkey approach should
scale equipment. reduce the cost of transfer to the commercial line.

21
USP Technical Guide

PAT in the Development


4 of Control Strategy

4.1. Introduction: PAT and Quality 4.2. Spectroscopic PAT Tools


Assessment
4.2.1. Near-Infrared Spectroscopy
An OSD CM process may be designed to facilitate the use of
The advent of modern spectroscopy technology can be
process analytical technologies (PAT) such as near-infrared
traced back to the development and utilization of NIR
spectroscopy (NIRS) and Raman spectroscopy. PAT is an
spectroscopy. The NIR region (780-2500 nm) of a spectrum
umbrella term whose meaning can differ from company to
is characterized by the absorption bands from overlapping
company. In OSD DS applications, a common interpretation
overtones and combinations of the fundamental vibrations
of PAT is the application of analytical tools to learn deeply
of functional groups containing hydrogen atoms. The low
about the process in order to devise best practices that lead
absorptivity of NIR light enables direct analysis of samples
to the desired CQAs. A broader interpretation of PAT includes
without prior preparation, as well as measurements
techniques such as statistical process control and modeling
involving scattering effects of diffuse transmittance
[Clegg 2020].
or diffuse reflectance.
A CM process, by design, facilitates in-process monitoring
In CM, NIR is a commonly employed PAT for monitoring
and collection of large amounts of data. The data can be
various stages of the process in real time and providing
used to develop models during the development stage,
valuable information about the quality and consistency
which can be helpful during the commercial manufacturing
of the final product. NIR has several key advantages over
stage.
traditional analytical techniques, including its ability to
In OSD DS processes, common locations for installing analyze the sample noninvasively in real time, its potential
spectroscopic probes are between feeder and blender, for faster and more cost-effective analysis, and its ability
after blender, or in the feed-frame of a tablet press. Some to perform multiple analyses simultaneously. These
modern tablet press equipment allows analyzing every tablet advantages have led to the successful application of NIR in
immediately after compression. Common measurements the continuous monitoring of various key quality attributes.
in such processes include flow rate and its variability, blend For example, NIR can be particularly useful in the continuous
composition and homogeneity, particle size and distribution, blending and granulation processes for predicting blend
impeller torque in the continuous blender, tablet press main homogeneity, particle size distribution, and moisture
compression force, etc. content, which are all important parameters that influence
intermediate flowability and compressibility.
In some special processes, especially semi-continuous
ones, it may be possible to eliminate spectroscopic PAT In the context of PAT methods, soft sensors combining NIR
measurements altogether in the production runs after using with other process parameters can also be used to predict
them extensively to build representative models. Soft-sensor an attribute such as dissolution using tablet weight, main
and gravimetric controls can be used as surrogates for compression force, and NIR-predicted API content. In
spectroscopic PAT, thereby saving resources [Medendorp, addition, the integration of NIR spectroscopy into existing
Shapally, Vrieze and Tolton 2022].

22
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

process control systems has been shown to provide real-time fraction of the scattered light will undergo an energy change
feedback on the manufacturing process, leading to improved (corresponding to the vibrational energy levels of the
process control and increased efficiency. It is important to molecule) that determines the size of the wavelength shift as
note that the implementation of NIR spectroscopy in CM well as the number of shift occurrences.
requires the careful consideration of several factors. First,
one must select NIR spectroscopic instrumentation that As a complementary method to NIR, Raman spectra feature
is suitable and appropriate for the specific manufacturing sharp, well-resolved bands over approximately 50-4000
process. Second, robust and validated analytical methods cm-1, making it suitable for the analysis of crystalline
must be developed to ensure accurate and reliable results. compounds and polymorphs. In addition, the spectrum of
Finally, a formal lifecycle management plan must be in water is extremely weak, which in some scenarios can be a
place so that the equipment and models are monitored desired characteristic, in that it allows direct measurement
and maintained proactively to incorporate new sources of of complex mixtures with limited interference from sample
variation and ensure continued performance and accuracy. moisture content. One of the biggest and most frequent
challenges associated with Raman, however, is the presence
of fluorescence background when collecting quality Raman
4.2.2. Raman Spectroscopy spectra for certain compounds (e.g., microcrystalline
cellulose or MCC). Fluorescence can occur when the
Raman spectroscopy is another widely used PAT tool in CM, excitation wavelength used to generate Raman signal
probing the molecular vibrations in a sample and providing overlaps with the absorption spectrum of the sample. This
detailed information about its chemical composition and can cause some molecules to be excited to higher energy
structure. The basic principle of Raman involves inelastic states, leading to fluorescence emission rather than Raman
scattering of photons by a molecule interacting with the scattering. To mitigate the impact of fluorescence, several
monochromatic incident light. In this process, only a small strategies can be employed: light source wavelength

23
USP Technical Guide

optimization, sample preparation, instrument configuration, measurements, the NIR or Raman instrument should be
and appropriate data analysis methods. calibrated regularly using reference standards that are
traceable to recognized standards. Other aspects of
In CM, Raman spectroscopy enables real-time process acquisition parameters, sampling interface, sample size,
monitoring and control with rapid, non-destructive analysis and modes of measurement (e.g., at-line, in-line, and on-
of the chemical and physical properties of raw materials, line) should also be considered. Moreover, the reference
intermediate products, and finished products, providing analytical procedure is the “gold standard” when developing
critical information for process optimization and quality quantitative models and thus should be performed on
assurance. Some of the key learnings when Raman the same samples that are used for spectral acquisition
spectroscopy is used as a PAT tool include blend uniformity, whenever possible.
amorphous content, powder flowability, and tablet quality
(e.g., API content, hardness, and disintegration time). It
must be noted that some of these attributes are not directly 4.3.2. Preprocessing
measured by Raman spectroscopy alone, but it contributes
to the measurement process. The aforementioned factors for The goal of preprocessing is to eliminate or reduce variability
NIR implementation, such as instrument selection, analytical that is irrelevant to the property of interest and thereby
method development, and maintenance, also apply to Raman represent the data better for modeling purposes. When
spectroscopy for accurate and reliable measurements implementing Raman or NIR, the complex, high-dimensional
throughout its lifecycle. spectral data are often affected by various factors such
as the quality of the instrument, properties of the sample,
and manufacturing conditions. To obtain reliable, accurate
4.3. Chemometrics spectral data, appropriate preprocessing methods should be
applied to improve signal-to-noise ratio and mitigate artifacts
4.3.1. Sampling and Data Quality and baseline variations. In the manufacturing of oral solid
DPs, variations in powder physical properties such as particle
When developing a calibration model for PATs, it is of utmost
size, packing density, or surface morphology can introduce
importance to ensure that the data are accurate and relevant.
unwanted light-scattering effects manifested in the additive,
The old adage “garbage in, garbage out” holds well when
multiplicative, or wavelength-dependent spectral data.
it comes to chemometric modeling. The total amount of
variation in the dataset consists of explainable variation In addition, as mentioned in the preceding section, a
that potentially can be modeled and random variation or stronger fluorescence signal can overshadow Raman
noise that cannot be captured. A successful application scattering by creating a broad band, which will benefit from
of chemometrics in PAT depends on several factors such data pretreatment to amplify Raman peaks while eliminating
as sample representation, quality of the measurement fluorescence background. The most widely adopted
system, and the model structure. To gather representative preprocessing methods in spectroscopy are mean-centering,
data, samples should be selected across the entire range baseline correction, standard normal variate (SNV),
of expected concentration variation in the process and multiplicative scattering correction (MSC), and derivatives.
should include in the mixture all possible components Other approaches such as extended multiplicative signal
to be tested in the future. In addition, the calibration set correction (EMSC) and generalized least-squares (GLS)
should be large enough to statistically determine the weighting are becoming popular.
mathematical relationship between spectral variables and
concentrations. When preparing samples, the process
should be standardized to ensure that all samples are treated 4.3.3. Variable Selection
consistently, and the actual process being measured is
mimicked accurately. Variable selection is another important aspect of
chemometric modeling when implementing NIR or Raman
Another consideration involves the inherent heterogeneity of spectroscopy as PAT. Careful selection of the spectral
sampling that depends on where in the process, the samples variables most relevant to chemical composition or
are collected. Prior to data collection, instrumentation physical properties can simplify the model and reduce the
elements such as the spectrometer, optics/probes, and noise, leading to more accurate, robust, and interpretable
spectral analysis software should be determined based on prediction results. The choice of the variable selection
the applications. To ensure repeatable and reproducible approach depends on the nature of the data, the complexity

24
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

of the system, and the desired model performance. Based the properties of interest. PLS identifies the latent variables
on the fitted partial least squares (PLS) regression model, for that maximize the covariance between X and Y datasets.
example, loading weights and regression coefficients can be In comparison to MLR, PLS can handle noisy and high-
used as a measure of association between each variable and dimensional data as well as multicollinearity, and it improves
the response. Moreover, variable importance in projection predictive performance.
(VIP) scores are used to assess the importance of each
predictor variable in the PLS model, which takes into account
both the contribution of the variable to the model and its 4.3.5. Model Optimization and Validation
ability to explain the variation in the response variable.
Determining the optimum number of latent variables in PLS
Selectivity ratio is another measure defined as the ratio involves finding the number of components that provide the
between the explained variance of each variable and best trade-off between model complexity and prediction
the residual variance. Genetic algorithm (GA) is a search accuracy. Cross-validation is a widely used technique for
algorithm that mimics the process of natural selection and determining the model dimensionality and evaluating model
genetics to find an optimal solution to a problem. In the predictive performance. It is recommended to use an
context of PLS variable selection, GA works by iteratively independent test set to estimate the prediction error, which
selecting a subset of predictor variables, evaluating the is less biased than the error obtained from the training set.
performance of the PLS model using only those variables, This helps to ensure that the PLS model is not overfitting
and then selecting the best-performing subset for the next to the training set and can be applied to new data with
iteration. This process continues until a stopping criterion is reasonable accuracy.
met or the algorithm converges to a solution.

In addition, interval PLS (iPLS) is a variable selection method 4.3 .6. Model Maintenance and Lifecycle
that aims to identify subsets of predictor variables that are Management
highly correlated with the response variable across a range
Chemometric models require maintenance and lifecycle
of intervals. In iPLS, the predictor variables are divided into
management for continued accuracy and reliability over
intervals based on their values, and PLS models are fitted
time. One of the most important aspects is the quality of the
separately for each interval. The variables that have the
data used to develop the model. Data should be monitored
highest correlation with the response variable across all
regularly and assessed for outliers, missing values, and
intervals are selected for inclusion in the final model.
other quality issues. Data cleaning procedures should be
established to address any issues that are identified. In
4.3.4. Quantitative Predictive Modeling addition, models should be routinely monitored to ensure
that they continue to perform accurately and reliably. Models
The quantitative models commonly used in PAT may need to be retrained or updated with appropriate
implementation include multiple linear regression (MLR), frequency to reflect changes in the process or product being
principal component regression (PCR), and partial least monitored. Lastly, documentation and version control are
squares regression (PLS). MLR models the linear relationship important to ensure that models can be understood and
between a dependent variable and one or more independent maintained easily by other users and that different versions
variables. PCR combines principal component analysis of the model can be tracked and compared throughout the
(PCA) and MLR to develop calibration models. PCA is used to lifecycle.
reduce the dimensionality of the spectral data and then the
principal components are used as input variables to predict

25
USP Technical Guide

Application to Oral
5 Solid Dose, Drug
Product Manufacturing

5.1. Solid Oral Dose CM Methods and there is no intermediate major unit operation between
their Specific Control Strategies blending and tableting.

In all modes of OSD CM discussed here, the following will be


5.1.1. Direct Compression (DC) common:

A typical OSD DC process consists of the following • There is a need to characterize and control input material
components: properties.

• A facility to store, pre-condition, and deliver the input • One must monitor the effects of feeder refill and material
materials to the first unit operation in the process, which property variations on feed uniformity.
is feeding.
• There is a need to characterize blender performance.
• A number of loss-in-weight gravimetric feeders so that
• It is important to consider the feeder-blender
each ingredient can be accurately fed to the next unit
combination and their combined performance.
operation. It is typical to have separate feeders for APIs,
excipients, lubricant, etc. They have a predetermined • Factors related to tableting operation need to be
set-point to maintain the desired ratios of ingredients in considered.
the formulation.

• Blending unit operation that typically features a 5.1.2. Wet Granulation


horizontal or vertical continuous blender. A common
design for the blender is an impeller housed in a A CM line with wet granulation involves feeders, blender(s),
cylindrical shell. The impeller has blades mounted along optional mills, continuous wet granulator, continuous dryer and
its length and as it rotates along its axis, powders are tableting equipment. The wet granulation step can be high or
fluidized, mixed, and conveyed by the blade action. low shear, depending on the process requirements, and it is a
major unit operation between blending and tableting.
• Tableting unit operation involving a rotary tablet press.
The output from the feeder typically passes through a In high-shear continuous wet granulation, added factors will
transfer chute or hopper into the feed-frame of the tablet include binder properties and quantity, rate and amount of
press, wherein rotating paddles disperse the powder into shear imparted, and resulting granule size and uniformity.
dies on a rotating turret to be punched into tablets.
In low-shear continuous wet granulation, a process featuring
• Tablet coating operation using a continuous or traditional twin screw granulators is low shear, wherein the shear
batch-mode pan, or a Wurster coating equipment. may not be a significant factor affecting the intermediate
CQAs. Other parameters mentioned under high-shear wet
The direct compression process (see Figure 2) involves granulation will potentially remain relevant in this case. In
feeders, blender(s), optional mills, and compression and/or some cases, heating is involved, so temperature is another
coating equipment. It is called “direct” compression because key parameter.

26
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

5.1.3. Dry Granulation Process designers carry out experiments using individual unit
operation equipment such as feeders and blenders. These
In this scenario, the compression imparted by rollers will be
experiments help designers understand the performance
a significant factor. Parameters such as roll gap, roll speed
across the design space, which could include the expected
and the feed rate will determine the properties of the rolled
material property and process parameter ranges. It is
material as it progresses to the next unit operation, which
intuitive to carry out experiments with a feeder-blender
typically is milling.
combination, as some feed rate irregularity coming in from
the feeders can be mitigated by the back-mixing action
5.2. Feeding and Blending Unit of the blender. From a risk assessment perspective and to
Operations determine process robustness, it is important to understand
whether this is such a case, and if it is, then what range of
The equipment train is selected based on the desired disturbance can be mitigated. Some key material properties
production rate and it includes a loss-in-weight feeder for in these unit operations are material density, particle size,
each component (e.g., excipients, lubricants), which is and particle distribution, cohesion, and electrostatic affinity.
tuned to maintain the desired ratio corresponding to the The process-related factors and parameters include feeder
formulation. Loss-in-weight feeders are equipped with a refill frequency and quantity of material input, variability
load cell that measures the weight of material inside the between batches of input material, processing rate, blender
feeder at a high frequency and records it via an onboard impeller speed, fill level, blender design, etc.
computer. As the material is fed into the process, the weight
inside the feeder goes down. The feeder control system It is critical that both the feeding and blending unit
adjusts the feeder screw speed accordingly to maintain the operations are well understood and controlled to have
desired set-point in feed rate. This is typically the first point an OSD DP line operating under a state of control.
of process monitoring and control. The sensor acts as a non-
spectroscopic PAT tool.

Table 1: Feeding-Mixing Unit


Operation and Key Parameters
Material Process
Property-Related Parameter-Related

Powder density Feeder refill frequency

Particle size Refill quantity

Particle size Vibrations or other physical


distribution disturbances

Cohesion Processing rate

Electrostatic affinity Blender impeller speed

Blender fill level

Blender design (vertical,


horizontal, tilt, weir, etc.)

27
USP Technical Guide

5.3. Example Control Strategy

Direct Compression

Process Parameters
Potential Impact Of
Process Step
Equipment Parameters Range Related Dp Cqas
Target Designation
Used Studied Studied

Drug Flow Rate 1 –


Feeder Flow Rate Critical Uniformity of Dosage
Feeding Substance Flow Rate 2
Model(s) (kg/hr) or Not Units (CQA)
PSD (μm) (kg/hr)

(Brand and
Model) Impeller
rpm1-rpm2
Horizontal/ Speed
Vertical/
X kg/h – Content Uniformity
Blending Y/N
Y kg/h (CQA)
Blade Design Forward
Tilted
Continuous
Mixer Blender Tilt X°

Main
Compression kN Range Friability
Force Based on
(Brand and
DoE and
Compression Model) Tablet
Design API Dosage (CQA)
Press Turret Speed/
rpm Range Space
Dwell Time
Disintegration Time

Tablet Load
Rate
(Brand and
Model) Rotation Coating Thickness
Coating
Continuous Speed and Uniformity
Coater
Air
Temperature

28
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

6 Appendix
USP Standards impact Continuous Manufacturing
Relevant General Chapters (Required, below <1000>)

Public
Input Continuous standard
Materials Manufacturing for the product in
Properties Process the market

<429> Light Diffraction Measurement of <429> Light Diffraction Measurement of <905> Uniformity of Dosage Units
Particle Size Particle Size
<711> Dissolution
<430> Particle Size Analysis by Dynamic <430> Particle Size Analysis by Dynamic
Light Scattering Light Scattering Official Product Monographs
<432> Determination of Zeta Potential <696> Characterization of Crystalline Solids
by Electrophoretic Light Scattering by Microcalorimetry and Solution Calorimetry
<467> Residual Solvents <699> Density of Solids
<616> Bulk Density and Tapped Density <776> Optical Microscopy
of Powders <811> Powder Fineness
<695> Crystallinity <846> Specific Surface Area
<699> Density of Solids <856> Near-infrared Spectroscopy
<731> Loss on Drying <858> Raman Spectroscopy
<776> Optical Microscopy <891> Thermal Analysis
<786> Particle Size Distribution Estimation by <921> Water Determination
Analytical Sieving
<941> Characterization of Crystalline and
<811> Powder Fineness Partially Crystalline Solids by X-Ray Powder
<846> Specific Surface Area Diffraction (XRPD)
<856> Near-infrared Spectroscopy
<858> Raman Spectroscopy
<891> Thermal Analysis
<921> Water Determination
<941> Characterization of Crystalline and
Partially Crystalline Solids by X-Ray Powder
Diffraction (XRPD)

29
USP Technical Guide

USP Standards impact Continuous Manufacturing


Relevant General Chapters (Required, above <1000>)

Public
Input Continuous standard
Materials Manufacturing for the product in
Properties Process the market

<1059> Excipient Performance <1039> Chemometrics <1010> Analytical Data-Interpretation


<1063> Shear Cell Methodology for Powder <1058> Analytical Instrument Qualification and Treatment
Flow Testing <1062> Tablet Compression Characterization <1029> Good Documentation Guidelines
<1856> Near-Infrared Spectroscopy—Theory <1174> Powder Flow
and Practice <1039> Chemometrics
<1216> Tablet Friability & <1217> Tablet
<1858> Raman Spectroscopy—Theory and <1079> Good Storage and Distribution
Breaking Force
Practice Practices for Drug Products
<1241> Water Solid Interactions in
<1174> Powder Flow Pharmaceutical Systems Sampling Procedures
<1195> Significant Change Guide for Bulk <1243> Wettability of Pharmaceutical <1099> Limit on Number of Large Deviations
Pharmaceutical Excipients Systems When Assessing Content Uniformity in
<1197> Good Distribution Practices for Bulk <1244> Electrostatic Properties of Large Samples
Pharmaceutical Excipients Pharmaceutical Systems <1220> Analytical Procedure Lifecycle
<1776> Image Analysis of Pharmaceutical <1245> Fitting Compaction Model
Systems <1856> Near-Infrared Spectroscopy—Theory
(Compactor simulator) Characterization
and Practice
<1430> Analytical Methodologies Based on
Scattering Phenomena – General <1858> Raman Spectroscopy—Theory
and Practice
<1711> Oral Dosage Forms – Performance
Tests
<1776> Image Analysis of Pharmaceutical
Systems
<1856> Near-Infrared Spectroscopy—Theory
and Practice
<1858> Raman Spectroscopy—Theory
and Practice

30
Control Strategies for Continuous Manufacturing of Solid Oral Dose Drug Products

7 References
Clegg, I. (2020). Chapter 7 ­– Process analytical technology. ICH (2023). “Q13 Continuous Manufacturing of Drug
Specification of Drug Substances and Products (Second Substances and Drug Products.” from https://database.ich.org/
Edition). C. M. Riley, T. W. Rosanske and G. Reid, Elsevier: sites/default/files/ICH_Q13_Step4_Guideline_2022_1116.pdf.
149-173.
Matos, N. A., S.; Almaya, A.; Biba, E.; Carlin, B.; Cruz, C.;
E2968, A. “E2968 Standard Guide for Application of Hausner, D.; Jayjock, E.; Jensen, K.; Khinast, J.; et al. (2018).
Continuous Processing in the Pharmaceutical Industry.” Book “USP Pharmacopeial Perspective for Pharmaceutical
of standards 14.01. https://www.astm.org/e2968-14.html Continuous Manufacturing: Stimuli to the Revision Process.”
USP Pharmacopeial Forum (USP PF) 44(6): 1-30.
FDA (2004). “FDA Guidance for Industry, PAT – A Framework
for Innovative Pharmaceutical Development, Manufacturing Medendorp, J., et al. (2022). “Process Control of Drug
and Quality Assurance.” from https://www.fda.gov/ Product Continuous Manufacturing Operations—a Study in
media/71012/download. Operational Simplification and Continuous Improvement.”
Journal of Pharmaceutical Innovation 17(1): 85-96.
FDA (2011). “Guidance for industry. Process Validation:
General Principles and Practices.” from https://www.fda.gov/
files/drugs/published/Process-Validation--General-Principles-
and-Practices.pdf.

FDA (2022). “ETP graduates its first technology.” from


https://www.fda.gov/about-fda/center-drug-evaluation-and-
research-cder/news-etp.

ICH (2002). “The common technical document format for


the registration of pharmaceuticals for human use: Quality –
M4Q(R1).” from https://database.ich.org/sites/default/files/
M4Q_R1_Guideline.pdf.

ICH (2006). “Q9 Quality Risk Management (Guidance


Document).” from https://www.fda.gov/regulatory-
information/search-fda-guidance-documents/q9-quality-risk-
management.

ICH (2009). “Q8(R2) Pharmaceutical Development. Tripartite


Harmonised ICH guidance.” FDA Docket FDA-2005-D-0154.

ICH (2009). “Q10 Pharmaceutical Quality System.” from


https://www.fda.gov/media/71553/download.

31

You might also like