Accepted Manuscript: Journal of Pharmaceutical and Biomedical Analysis

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

Accepted Manuscript

Title: Risk-based approach for method development in


pharmaceutical quality control context: A critical review

Authors: Riccardo Deidda, Serena Orlandini, Philippe Hubert,


Cédric Hubert

PII: S0731-7085(18)31214-7
DOI: https://doi.org/10.1016/j.jpba.2018.07.050
Reference: PBA 12121

To appear in: Journal of Pharmaceutical and Biomedical Analysis

Received date: 23-5-2018


Revised date: 24-7-2018
Accepted date: 30-7-2018

Please cite this article as: Deidda R, Orlandini S, Hubert P, Hubert C, Risk-
based approach for method development in pharmaceutical quality control context:
A critical review, Journal of Pharmaceutical and Biomedical Analysis (2018),
https://doi.org/10.1016/j.jpba.2018.07.050

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
1

Risk-based approach for method development in pharmaceutical quality

control context: A critical review

Riccardo Deiddaa,*, Serena Orlandinib, Philippe Huberta, Cédric Huberta

T
a
University of Liège (ULiege), CIRM, Laboratory of Pharmaceutical Analytical Chemistry, Liège,

IP
Belgium
b
University of Florence, Department of Chemistry “U. Schiff”, Via U. Schiff 6, Sesto Fiorentino

R
50019, Florence, Italy

SC
U
N
*Corresponding author: R. Deidda
A
E-mail address: riccardo.deidda@uliege.be
M

Graphical abstract
ED
E PT
CC
A

HIGHLIGHTS
2

 The importance of adopting a risk-based approach has been tackled

 The analytical quality by design is thoroughly described in every step

 Recent papers regarding the application of a risk-based approach have been reviewed

T
Abstract

IP
Pharmaceutical regulatory bodies increasingly require the implementation of systematic approaches in

R
pharmaceutical product development. Quality control methods play a key role in the control strategy of drugs

manufacturing to assure their quality. A risk-based approach in the analytical method development is strongly

SC
recommended to ensure that the method performances fit the purpose of the method during its entire life-cycle. In the

last decade, analytical quality by design (AQbD), as risk management oriented methodology, has been progressively

U
integrated with method development for fulfilling this objective. This approach has successfully allowed the quality to
N
be designed into the analytical processes by obtaining a deep understanding of the procedures. In this paper the AQbD
A
workflow and its application in the development of methods to be used for pharmaceutical quality control have been
M

treated and discussed. Recent publications regarding how AQbD has been applied in separation techniques were

reviewed. The different development strategies have been also showcased, highlighting their advantages and
ED

disadvantages, in order to give a useful overview.

Abbreviations: API, active pharmaceutical ingredient; AQbD, analytical quality by design; ATP, analytical target profile;
PT

BBD, Box-Behnken design; CCD, Central Composite design; CMA, critical method attribute; CMP, critical method

parameter; DD, Doehlert design; DoE, design of experiments; EC, established condition; FCD, face centered design;
E

fFD, fractional factorial design; FFD, full factorial design; FMEA, failure mode effect analysis; ICH, International
CC

Council for Harmonisation; KS, knowledge space; MAA, market authorisation application; MODR, method operable

design region; MPV, mixture-process variable; OFAT, one factor at time; PBD, Plackett-Burman design; QbD, quality
A

by design; QbT, quality by testing; QC, quality control; QRMP, quality risk management process; RA, risk assessment;

Rs, resolution; RSM, response surface methodology; S, separation criterion; SST, system suitability test, TD, Taguchi

design.

Keywords: Analytical Quality by Design; Design of Experiments; Drug analysis; Quality Control; Method Operable

Design Region; Risk Based Approach


3

1. Introduction

What is “risk”? The risk is defined by the International Council for Harmonisation of Technical Requirements

for Pharmaceuticals for Human Use (ICH) as the combination of the probability of occurrence of harm and the severity

of that harm [1]. This generic definition obviously changes according to the field of application. The FDA document

“Pharmaceutical cGMPs for the 21st century-A risk based approach” [2] and the ICH guideline Q9 on quality risk

T
management [1] clarify this concept, besides the importance to adopt risk management strategies to ensure the quality in

IP
pharmaceutical processes. Since quality control (QC) methods play a key role in quality assurance, their capability to

R
fulfil their intended purpose is crucial. Furthermore, the QC methods are submitted by companies to the regulatory

SC
bodies in each country where the pharmaceutical product is distributed to support market authorisation applications

(MAA) [3]. The risk associated to their performance should be accurately assessed and taken into consideration for each

U
stage of the method life-cycle. Moreover, a deep knowledge about their characteristics is essential to control the risk

associated to their use. The analytical method life-cycle, reported in Fig.1, consists not only in design and development,
N
but also in validation, control strategy as well as continual improvement after method approval [4]. Prior knowledge and
A
preliminary experiments are the base from which the method design and development start and progress, and a well-
M

planned risk management methodology should be covering the entire method life-cycle. As the QC methods are also

subject to continuous improvements imposed by evolving legislation and laboratory needs, in this way analytical
ED

chemists should be facilitated in performing this task.

A traditional way to develop analytical methods is by means of the quality by testing (QbT) concept, also called
PT

trial-and-error. QbT consists in testing the different operative conditions generally by varying one factor at time (OFAT

investigation). This unstructured approach often requires a large number of experiments before finding the appropriate
E

conditions. The information obtained is limited to those observed by the operator and the possible interactions between
CC

variables are not studied. Moreover, such strategy can also lead to a false optimum. Indeed, QbT approach leads to the

selection of a working point which is not necessarily the best one. Furthermore, the outline of the selected working
A

point is not known as it was partially or even not explored during development. Such strategy certainly does not

facilitate the possible modifications that can be needed in the future because of an inadequate knowledge of their

potential impact on the method performances. The risk is not entirely understood, and it cannot be properly managed [5-

10]. After the identification of the working conditions, the quality of the procedure is tested during the validation phase.

Quality index parameters such as accuracy and robustness are therefore evaluated only at the end of the method

development process. Notably, the method robustness, defined by ICH guideline Q2(R1) as “a measure of the capacity
4

of the method to remain unaffected by small, but deliberate variations in method parameters” [11], is a key point needed

to prove the quality of an analytical procedure and its assessment cannot be avoided.

Since implementing the quality in the pharmaceutical industry means to reduce the variability within the

processes, systematic approaches and a more science-based decision-making are required. The ICH guideline Q8(R2)

on pharmaceutical development [12] describes in a detailed way a systematic approach, defined as quality by design

(QbD), originally intended for manufacturing processes. Because of its facilitations and many advantages, its use has

been considerably increased over the last decade and extended to analytical chemistry [4,5,7-9]. Indeed, analytical

T
methods should accomplish their intended purpose as well as product requirements for a dosage form. Moreover, QC

IP
methods for the analysis of active pharmaceutical ingredients (APIs) and drug products should be considered as an

R
integral part of the QbD concept, as they are aimed to ensure product quality and consequently patient safety [7]. QbD

applied to analytical methods is commonly named analytical quality by design (AQbD) and it may be defined as a

SC
systematic approach to development that begins with predefined objectives and emphasizes analytical method

understanding and control, based on sound science and quality risk management.

U
As presented in Fig. 2, the adoption of the AQbD concept has been significantly rising over the years. These
N
trends represent the number of publications by year related to QbD and AQbD between 2007 and 2017, using Scopus
A
database to search “quality by design” term. All the results for title, abstract and keywords have been screened to sort
M

papers between QbD and AQbD. AQbD approach includes the incorporation of prior expertise, risk and knowledge

management and the application of design of experiments (DoE) throughout the analytical method life-cycle [12].
ED

Unlike QbT strategy, the quality as well as the robustness is conceived and built at the very preliminary stages and

within the method development process. The use of DoE drastically reduces the number of runs in the plan of
PT

experiments and at the same time makes it possible to gain a higher quality of information about the investigated

procedures [13]. After defining the parameters that may affect the method performance by risk assessment, a computer-
E

assisted multivariate strategy allows identifying interactions between them as well as the computing models relating the
CC

parameters to the method performances. The final outcome is the computation of a multidimensional region where the

method can meet its intended purpose with an established probability of success: the method operability design region
A

(MODR). In this way the risk is properly evaluated from the beginning, and a proper risk management methodology can

be implemented to monitor the method performances throughout all the stages of method life-cycle [5,9,12,14,15].

QC methods are an essential constituent in the overall control strategy for drug manufacturing and they must be

submitted in specific dossiers to the regulatory bodies. Anyway, it is not rare that in QC laboratories changes and

improvements into analytical methods after their approval are needed. Continual improvement is an important part of

the method life-cycle, hence it should be a straightforward process, guaranteeing the continuation of the quality
5

assurance. Analogously to the concept of product life-cycle, where the companies have the opportunities to evaluate

innovative approaches to improve product quality [12], during the analytical procedures life-cycle it is possible to

benefit from advances in analytical sciences [16]. The ICH Q12 guideline depicts a structured approach defining a set of

specific criteria to be met for post-approval changes to analytical procedures. Such changes are classified according to

the potential effect on the product quality. For the ones with a sufficient risk, a prior-approval is needed; while for the

ones associated to a moderate or low risk, the notification of the change to the regulatory agencies is enough. However,

some characteristics of the analytical procedure are considered as necessary to assure the product quality and are called

T
established conditions (ECs). For the ECs changes a new submission to the regulatory authority is required. The

IP
identification of the ECs for analytical methods is based on the effective knowledge concerning the method, acquired

R
during the development process, and described in the MAA. For instance, in a method presenting a not-established

relationship between critical method parameters (CMPs) and critical method attributes (CMAs), for which the risk

SC
associated to a hypothetical change is unknown, each operation parameter is classified as ECs. In fact, the method is

presented as a working point with its parametric values that must be applied during the routine analysis, and no more

U
information about its contour is given. After approval, changing in such values, even if minimal, is not allowed without
N
a new regulatory approval. After requesting it, the company must wait the confirmation, which can take several months
A
to arrive. On the other hand, an analytical method developed following a systematic approach, such as the AQbD,
M

presents an increased understanding about how the CMPs affect the CMAs, that is well detailed in the MAA. In this

case, the ECs are focused on method-specific performance criteria, such as specificity, accuracy, and precision. It is here
ED

where the advantages of a risk-based approach in method development lie. The deep method understanding, collected

thanks to the AQbD approach, allows a MODR to be defined where the method fits its purpose in each point, and where
PT

the quality criteria of the method are assured with a defined level of probability. This means that in the regulatory

documents the method is not described just as a working point, but it is also presented as a region gathering a set of
E

operating conditions where method performances are met. Therefore, the method will present a high-level description
CC

given to the regulatory bodies during the first MAA, and if a post-approval change (within the MODR) in the analytical

method is needed, its notification to the regulatory bodies will be enough. Obviously, the changes should lead to a
A

method with equivalent or better performances than the original, aspect that should be confirmed by validation results

[16].

A possible change in the method could be due to results of trend analysis on method performance, which may

point out that the method is not operating as expected (e.g. causing out-of-specification results). The investigation of the

causes of variation (e.g. an evolution of the matrix constituents in the pharmaceutical product during the stability study)

could result in a change of the method, and thus improvement of the method performance. In this case, the post-
6

approval changes within the MODR may be considered as adjustments and do not need a new regulatory approval, apart

the notification (with details about the quantitative performances of the new working point, if applicable) [4,6]. On the

other hand, if new or revised specifications (e.g., total impurities, potency) are required, for instance due to a process

change in the synthetic route of the API and to the need to quantify a new impurity, the analytical procedure should be

re-optimized and re-validated. In fact, in this case the analytical target profile (ATP) has changed and no information

about the behavior of new hypothetical CMAs has been previously collected in the knowledge space (KS). Such need of

information could lead to a new optimizing process, for example if the method-specific performances are not met

T
anymore. Anyhow, the operator will start this step with a deeper knowledge that will guide towards the re-optimization

IP
and the MODR re-definition.

R
In this context, a recent Stimuli article to the USP by the USP Validation and Verification Expert Panel [17]

highlights the possible changes that may be required throughout the procedure's life-cycle. Changes may include the

SC
incorporation of an additional control, introducing a new method or technology, changing the intended purpose to

incorporate a new impurity or tighten specifications, or alignment with a procedure in a compendial monograph that has

U
been updated. Different typologies of change require different actions, anyway the USP Expert Panel remarks that
N
changes which are within the proven ranges (within the MODR of the procedure) are considered adjustments and do not
A
require a procedure performance qualification study to be performed before returning to routine monitoring. At the
M

opposite end, if the change involves tightening a specification limit or a change to the intended purpose of the procedure

to measure additional attributes, these changes result in the re-definition of a new ATP.
ED

2. Analytical Quality by Design workflow


PT

As shown in Fig. 2, the AQbD concept has been increasingly considered and applied over the last decade. The

AQbD workflow is represented in Fig. 3 and each phase is detailed hereafter. Given that the AQbD model is cyclic, it is
E

important to highlight that an AQbD workflow has a starting point but not a closure. Indeed, after the method
CC

development, the method life-cycle progresses with validation, control strategies and continual improvements. Thanks

to this strategy, all these steps can be continuously managed considering the knowledge acquired. In order to develop
A

QC methods which are compliant with the AQbD model, all the following aspects have to be taken into account, with

particular focus on ATP definition, risk management strategies, and MODR acquirement. Since the AQbD concept has

been largely applied in separation science, we have principally focused the discussion on LC, SFC and CE techniques.

2.1. Analytical target profile and scouting phase


7

The AQbD workflow starts with the definition of the ATP, which represents the object and the base upon which

the planning of experiments during the method development process starts. It exactly defines the purpose of the method

and all the characteristics that the analytical method should have, therefore it can be described as the fit-for-purpose

concept [5,18-22]. The ATP does not refer to a specific analytical technique or an operative mode; it rather establishes

the criteria required to the technique, as well as the matter of measurement. Many questions are asked to developers.

What is the purpose of the method? What are the characteristics of the samples? What are the method criteria requested

by the regulatory agencies? What are the method desired performances? What are the laboratory resources? The

T
answers to all these questions help the developers to have a precise idea of the method and to select the most suitable

IP
technique [8,9,14,23]. As Burdick et al. defined in a stimuli paper of the USP Statistics Expert Committee, “the

R
procedure should be able to quantify [analytes] in the presence of [X, Y, Z] over a range of A% to B% of the nominal

concentration with an accuracy and uncertainty ensuring the reportable results fall within ±C% of the true value with

SC
quantified guarantees” [24].

During the method scouting phase, prior knowledge is gathered, and preliminary experiments are carried out. In

U
separation procedures, normally the stationary phases and solvents in chromatography, the operative mode/ pseudo
N
stationary phases in CE, as well as sample preparation, are quickly evaluated or selected on the basis of literature
A
search, univariate experiments and/or laboratory’s experience with the aim of approaching the ATP. Considerations
M

regarding the sample preparation, the choice of the target concentration values as well as the preparation of reference

solutions should be done with attention. An understanding of the drug and solvent properties is needed and fundamental
ED

to sample the pharmaceutical product in a representative way [25].


PT

2.2. Critical Method Attributes and Critical Method Parameters for Risk Assessment

Typical requirements for QC methods are linked to the specificity and/or selectivity of the method, to the
E

chromatographic or electrophoretic pattern as well as to the quantitative performances of the method, which are
CC

evaluated by validation process. Other parameters considered by developers could be for example analysis time,

procedure cost, and throughput capacity [4,26]. These requirements are also called CMAs and their values should be
A

within appropriate limits, ranges, or distributions to ensure the desired quality [12]. In fact, they are represented by key

responses directly correlated to a mathematical representation of the quality of the method performances and thus to the

quality of the analytical data. The CMAs can also be prioritized and defined as “must have” or “intend to have” during

the mathematical modelling phase.

In order to start a quality risk management process (QRMP), the first step is the identification of hazards, and the

analysis and evaluation of how they affect the quality of the method. The risk assessment (RA) is fundamental for a
8

risk-based method development. The goal is to assure with high confidence that the analytical method will meet all the

requirements set for the CMAs, under the final selected conditions, through the method life-cycle. During the RA, the

method parameters are screened and investigated to identify the CMPs, namely the ones which could potentially affect

one or more CMAs. The individuation and investigation of the CMPs are fundamental to assess the risk. Method

parameters are generally classified as critical, non-critical or unclassified [27]. An important goal of AQbD in method

development is to convert as many unclassified parameters as possible into critical or non-critical categories, otherwise

it would be necessary to fix their values or to set very narrow ranges for them. At the beginning of the RA, the number

T
of the unclassified parameters could be a conspicuous count. It can be reduced by prior knowledge or univariate

IP
preliminary experiments and by the means of a quality risk management strategy, as proposed in ICH guideline Q9 [1].

R
Among the principal tools used to perform this task there are flowcharts, Ishikawa diagram, failure mode effects

analysis, and many others. The Ishikawa diagram [1,28] is a widely used tool for the identification of the risk. This type

SC
of diagram, also called fish-bone diagram or cause and effect diagram, dissects the risk in various categories related for

example to sample preparation, instrumentation, materials, personnel factors and environmental conditions. The CNX

U
tool can be employed for classifying the Ishikawa parameters in the following: parameters which should be controlled
N
(C), potential noise parameters (N) and parameters which should be experimented to determine acceptable ranges (X)
A
[18,29]. Another important risk assessment tool is the failure mode effects analysis (FMEA). The variables are ranked
M

according to the severity degree of the consequences on the analytical results, the probability of failure occurrence, and

the difficulty of failure detection. Multiplying the values for the three components, a risk priority number is obtained
ED

and it can be used to prioritize the risk [1,14,18]. Both these RA tools can be effectively used for selecting the CMPs to

be further studied by DoE. Further details on applications of FMEA to analytical procedures can be found in Refs.
PT

[30,31].
E

2.3. Design of experiments


CC

DoE is a chemometric tool particularly used in method development and optimization [32] and comprehensive

overviews of its application in chromatography [33,34] and CE [13] can be found in recent literature. It is a structured
A

and cost-effective approach which allows the analyst to gain relevant information about the processes by correlating the

responses to controllable variables (method parameters). In this methodology, a regression model linking parameters to

responses is postulated and a suitable experimental matrix is selected according to the complexity of the model. An

experimental matrix is a table where the columns report the values of the parameters and the rows represent the

experiments, where the values of the parameters are simultaneously changed. The experiments are performed according

to the experimental plan, the responses are measured, and the regression model is calculated by means of multivariate
9

linear or partial least square regressions [13,32]. The designs can be classified according to the objective of the

experiments in two broad categories: screening and optimization (or response surface designs). More details about the

choice of the ideal design are reported in more specialized readings [32-37].

2.3.1 Screening phase

A high number of parameters can influence the chromatographic or electrophoretic behaviour; thus, often it is

necessary to perform screening designs focusing on all the CMPs selected by RA. The screening designs are useful to

T
simultaneously study the effects of both qualitative and quantitative parameters on the selected CMAs with a low

IP
number of experiments [13,32,34]. This stage is particularly convenient for the systematic investigation of

R
discontinuous parameters, such as stationary and mobile phases in chromatography, or type of additive (organic

modifier, surfactant, cyclodextrin) in CE, because their investigation cannot be included in the subsequent phase of

SC
optimization. The screening experiments are highly advantageous because they make it possible to highlight the

parameters which are not influential and to fix their values, as well as to identify parameters for which the screening

U
results clearly indicate the optimal values, which can be set as well. This enables to reduce the number of CMPs to be
N
further studied during the optimization phase, and thus to reduce the number of experiments to be performed. Another
A
important advantage is the possibility of moving the experimental domain of the variables towards the zone that the
M

screening phase has indicated as leading to the best results.

Two-levels designs are often adopted to study as many parameters as possible, keeping a modest number of
ED

experiments. Plackett-Burman (PBD) and fractional factorial designs (fFD) are among the most employed for this

purpose. Another attractive possibility is the use of asymmetric and symmetric screening designs, where the
PT

investigated levels of CMPs are higher (usually from 2 to 4). Usually, the screening phase requires a good

understanding of the selection of input parameters and output responses, as well as a thoughtful selection of a suitable
E

experimental domain for each CMP [7,13]. Sometimes the screening study can be avoided on the basis of preliminary
CC

knowledge and/or preliminary univariate experiments, provided that the available information allows a rational

planning of the subsequent optimization.


A

2.3.2 Optimization phase and response surface methodology

The optimization phase generally consists in applying response surface methodology (RSM) for estimating the

main interaction and/or quadratic effects of the CMPs on the CMAs. In RSM the number of experiments increases with

respect to the screening phase, because at least three levels should be studied for each parameter in order to evaluate the

presence of curvature of the model. However, in this case it is possible to obtain a predictive regression model and thus
10

to draw a map of the predicted CMAs values throughout the experimental domain (the response surface or contour

plot). This phase is generally focused on the study of the effects of continuous CMPs, such as temperature, gradient

conditions, flow rate, buffer pH and concentration, voltage, and so on. The considered CMPs are selected either on the

basis of screening results or directly from RA.

Different symmetrical designs can be used, including full factorial (FFD), Box-Behnken (BBD), central

composite (CCD), face centered (FCD), Doehlert (DD) and Taguchi (TD) designs. When the number of CMPs to be

studied is high, D-optimal design can be used to reduce the number of experiments by using the mathematical design

T
criterion of D-optimality. In the generation of a D-optimal design, the selection of the experimental runs is performed

IP
focusing on minimizing the uncertainties of the model coefficients, thus enabling to obtain a good compromise between

R
number of experiments and quality of information [38]. The resulting D-optimal design can be an asymmetrical design

which forms an asymmetric shape but can also form a symmetrical shape [34]. The final design should limit the number

SC
of experiments to a small amount, but in every case, it also needs to be characterized by good evaluation criteria of the

matrix including its symmetry and performances, such as condition number and G-efficiency [38]. Optimization designs

U
include also mixture designs, where the response is a function of the proportions of the different ingredients in the
N
mixture, and mixture-process variable (MPV) designs, where MPV models are developed to represent the relationship
A
of the responses with both mixture components and process variables [38,39].
M

In any case, some replicates at the centre point or a duplicate/triplicate of each experiment should be performed

to obtain an estimate of the experimental variance, necessary for estimating the validity of the model by ANOVA. After
ED

the model has been established to be significant and valid by ANOVA [32], the effects of the parameters are assessed,

and the response model can be interpreted by means of two-dimensional graphs (contour plots) or three-dimensional
PT

graphs (response surfaces). These plots show the behaviour of the responses predicted at the selected levels, without

guaranteeing that such responses will attain the defined criteria with high probability [34].
E
CC

2.4. Method operable design region

The MODR for analytical methods is the equivalent of the design space for manufacturing processes [9]. The
A

latter is defined by ICH guideline Q8 as “the multidimensional combination and interaction of input variables and

process parameters that have been demonstrated to provide assurance of quality” [12]. The MODR can be considered as

a zone of robustness where the quality of the method performance is guaranteed in each point with a specified

probability level. The larger the MODR, the more robust the method is with respect to fulfilling the CMA requirements.

It represents the core of the AQbD approach and it is limited by the so-called edges of failure, outside which the method

performances are not accepted. When computing the MODR, aspects such as the uncertainty of the model parameters as
11

well as the probability to meet the CMAs specifications must be taken into consideration. Monte-Carlo simulations,

Bayesian modelling and bootstrapping techniques are useful tools to accomplish this task [5,7,40-42]. Moreover, since

“working within the design space is not considered as a change” [12], a more flexible approach for the analytical

method from the regulatory point of view is gained. Further reading about design space computation can be found in

Ref. [10,21,42,43].

2.5. Working point

T
The working point is a specific point within the MODR which is selected as operating point before performing

IP
the validation. It can be chosen according to various criteria, based on operative conveniences or facilities, such as

R
lower amount of solvents or additives, lower costs, lower analysis time, as well as on the basis of statistical criteria such

as higher value of desirability index or higher probability of fulfilling the requirements, and so on [7,44]. The only

SC
condition is that the point is included in the MODR.

2.6. Method validation U


N
The validation step is mandatory for QC methods and recommended for each analytical method. The goal is to
A
evaluate if the quantitative performances of the method fit for the purpose for which it has been developed. Several
M

documents, such as ICH Q2 [11], FDA cGMPs [2] and ISO [45,46], underline the importance of method validation

giving the definition of the criteria to be tested, without furnishing detailed experimental guidelines. In fact, an official
ED

experimental protocol does not exist yet and this leads the different laboratories to work out of a harmonised approach.

Generally, during method validation many parameters are tested to verify if the method fulfils the requirements imposed
PT

by regulatory agencies, according to the aim of the analytical procedure. Among validation parameters there are

specificity-selectivity, precision, trueness, accuracy, linearity and dosing range, LOQ and LOD as well as robustness. It
E

is essential to underline the fact that each time the working point has changed, a new validation step should be carried
CC

out [2]. In recent years DoE has been often applied also to assess robustness and to a lesser extent intermediate

precision.
A

An increasingly used validation strategy has been proposed by the SFSTP commission [47]. This strategy is

based on a predictive tool, the accuracy profile, that using the total error approach allows to assess the risk associated to

the method measurements. Briefly, the total error designates the combination of the systematic and random errors. For

each concentration level, the β-expectation tolerance intervals are computed considering the bias and the estimated

variability of the results. In such manner, considering a predefined risk α, an interval where future results are expected

to fall inside is computed. The so obtained accuracy profile gives a prediction (with a specified probability of success or
12

failure) which is perfectly in accordance with the risk-based AQbD approach. Another advantage of the accuracy profile

strategy lies in a single statistical decisional tool that integrates some of the validation parameters which normally are

evaluated individually [47-49].

Recently, the integration of validation phase within the MODR was evaluated by means of a proof of concept

[50]. Such integration could provide a subsequent step towards a strategy allowing modification of QC method during

routine use without the need of conducting a partial or full validation of the newly selected working point.

T
2.7. Control strategy

IP
The control strategy is defined as “a planned set of controls, derived from current product and process

R
understanding, that assures process performance and product quality” [51]. It directly derives from risk assessment and

is a part of the risk management strategy. Usually, a QC method must necessarily include a control strategy and the

SC
regulators evaluate if it adequately monitors any variability of the outcomes. The system suitability tests (SSTs) are

widely employed to check and ensure on-going performance of analytical methods.

U
For quantitative methods, quality control sample, consisting of an independent standard solution with a known
N
concentration of the targeted compound, is usually inserted in the routine analysis sequence allowing to evaluate and
A
keep under control the quantitative performance of the method. Analysing of such quality control samples by means of
M

control chart is largely performed in the pharmaceutical quality control context. A QC method developed by the AQbD

approach is once again advantaged. Indeed, when the risk is assessed and well known it is easier to predict the
ED

behaviour of the method performance and to individuate the source of a possible variation. The CMAs selected for

method development can be used in SSTs as qualitative criteria in order to monitor if the method remains in compliance
PT

with the ATP during the routine use. If the SST shows outcomes out of specification, the analysis must be stopped and

modification actions on the method initiated. Some QC methods, such as the Pharmacopoeias ones, allow changes in
E

parameters within specific ranges without the need of revalidation. If the method is not reported in Pharmacopoeias, the
CC

adjustments in method parameters must be followed by a partial or full validation step before implementing them in

routine.
A

3. Critical review of QC methods developed by the AQbD approach

Analytical procedures are part of the registration application file submitted to regulatory bodies within the EC,

USA and Japan. Companies must necessarily demonstrate that the developed QC methods are suitable for their intended

purposes and properly validated. Typical analytical procedures in a QC framework are quantitative tests for API and

impurities content, limit tests for the control of impurities and identification tests. These QC tests are generally
13

performed on raw materials (APIs and intermediates) and finished products, and for these purposes separation

techniques such as LC, CE and SFC are mainly adopted. Pharmacopoeias report several analytical methods, sometimes

presenting some gaps in such manner that further method development or optimization is needed. In this context, a risk-

based approach in QC method development is strongly recommended. In order to comply with the AQbD concept, the

method development process has necessarily to consider the risk, so that error prediction and propagation must be

necessarily taken into account in the procedure. Several papers regarding the application of AQbD for method

development are presented with the relative characteristics summarized in Tab.1.

T
The ATP definition is the starting point of the AQbD workflow and must be wisely established before the

IP
beginning of method development, by accurately defining the purpose and the desired characteristics of the method.

R
Prior knowledge, preliminary considerations and investigations are decisive aspects to select the proper analytical

technique. Information about analytes behaviours according to the techniques tested is precociously gained and can

SC
facilitate the ATP definition. In a QC context, separation between API(s) and/or impurities as well as excipient(s) is

mandatory to accurately quantify the targeted analytes. Moreover, other performance criteria should be met according to

U
the aim of the analytical procedure, including for instance sensitivity in terms of satisfactory LOQ values for impurities
N
or reproducibility, especially when considering multi-site QC laboratories, and so on. Furthermore, operational
A
simplicity as well as time and cost considerations are criteria to be managed in the design and development of the
M

method. Frequently, during the experimental phases of a pharmaceutical product more evidences are gradually acquired,

and the ATP may be object of several variations. In fact, the ATP precisely defines the objective of the analytical
ED

procedure but cannot foresee the evolution of the analytical needs. For instance, if during the experimentation phases of

a pharmaceutical product a toxicological potential of one impurity is individuated, the required method performances
PT

will change, especially in terms of sensitivity. The ATP will evolve and with it also the CMAs specifications.

Once the ATP is established, the individuation and definition of the CMAs are needed. Since these numerical
E

attributes are directly linked to the quality of the method performances, they should be as representative as possible.
CC

Among the reviewed manuscripts, the overwhelming majority of the selected CMAs are related to the chromatographic

or electrophoretic selectivity, in terms of peak separation as resolution (Rs) and separation (S), as well as retention and
A

migration times, and analysis time. The separation criterion S is defined as the difference between the retention time

measured at the beginning of the second peak and the one measured at the end of the first one. This criterion generally

makes it possible to obtain more accurate predictions than the ones provided by resolution or retention times [43],

which are gradually replaced by S [52-59]. Other selected CMAs in the reviewed manuscripts were represented by peak

shape characteristics related to method sensitivity, such as number of theoretical plates [39,60-63] and tailing and

symmetry parameters [60,63,64].


14

In order to integrate the risk concept into the method development, the CMPs must be identified and selected.

Typically, the risk assessment is performed by employing the Ishikawa diagram. The references show how frequently it

is used in the CMPs selection [29,59-61,65-70]. A typical Ishikawa diagram for LC and SFC methods presents as

principal categories of risk the injection, detection, mobile phase, stationary phase and sample. Instead, for CE

techniques, the main categories are the injection, separation and detection, capillary and background electrolyte.

After choosing the CMPs, their effects are investigated by DoE in a suitable experimental range, namely the KS.

Many authors decided to perform a screening study to acquire more information regarding the effects of such

T
parameters. This phase is not a compulsory step to be compliant with AQbD, and it can be omitted if the developers

IP
possess enough information on the separation system under study.

R
The screening DoE can be carried out by using several matrices, among which the operator can choose which

one best responds to the situation: PBD [55,64,65], fFD [66,71-73], TD [60], symmetric screening matrices

SC
[61,62,68,70], asymmetric screening matrices [59,67,69,74,75]. The asymmetric and symmetric screening matrices

permit to study more levels for each CMA, allowing two types of graphical plots to be obtained. The first describes the

U
effect of changing the levels of the CMPs on each CMA, while the second reports the effects of the different levels of
N
the CMPs on the CMAs. In this way it is possible to obtain a more detailed focus on the separation system, since a
A
higher number of levels of the CMPs can be investigated with respect to two-level designs [70].
M

The study of discontinuous parameters, which cannot be included in the optimization step, has been carried out

by different approaches. Most of the authors fixed this kind of CMPs, for instance type of stationary phase and type of
ED

organic solvent in chromatography, according to the results of preliminary experiments, scouting phase and/or

theoretical considerations [52-54,56-58,63,76-78]. Another possibility is to execute a screening DoE for each of the
PT

discontinuous parameters studied. This approach was followed by Ferey et al., carrying out two screening designs, one

with acetonitrile as organic solvent and one with ethanol [55]; anyway, this approach leads to an increase of the number
E

of experiments. Hence, if possible, it would be advisable to include the discontinuous parameters among the parameters
CC

to be investigated in the screening. The latter strategy has been reported for several CMAs: vial for optimizing sample

preparation [67], chromatographic columns for screening the chromatographic parameters [67,73], organic solvent in
A

the mobile phase [73], neutral cyclodextrin in the background electrolyte [75]. Finally, a special case concerned the

MPV D-optimal design which was used for simultaneously screening the effect of process variables and mixture

components (the constituents of the microemulsion used as background electrolyte) in the case of a MEEKC method

[79].

After the selection of the CMPs and their new experimental domain, the RSM is used to optimize the method.

This methodology firstly involves the choice of polynomial models, including interaction and/or quadratic terms, to
15

approximate the relationships between the CMAs and the CMPs. The selection of the proper design for estimating the

model coefficients is based on the chosen model, on the possibility of easily splitting the experimental domain from the

operational point of view, and on the quality of information desired. BBD is a frequent choice by developers

[29,52,54,61-65], probably due to the fact that it requires only three levels for each parameter and it is usually very

efficient in terms of number of required runs [32]. Other common choices are CCD [55-57,67,76], FCD

[60,66,68,71,72,74] and DD [69,70,75]. Optimal designs have been also used [53]: specially MPV optimal designs have

been employed for optimizing microemulsion separation systems involving both process variables and mixture

T
components as CMPs [39,59,79].

IP
The responses obtained during the experiments are modelled as function of the CMPs using statistical techniques

R
to create the predictive models of the CMAs. The goodness of fit and predictive abilities are central qualities for the

mathematic models, and they can be improved by refining the models, namely deleting some of the interaction/and or

SC
quadratic terms which are not significant [59,79].

The effects of the CMPs on the CMAs are generally visualized by drawing contour plots or response surfaces for

U
each CMA. Since the optimization phase usually regards more than one CMA, it could be useful to overlap them to put
N
in evidence the global optimization area. Sweet spot plots and to a lesser extent overlay plots are then used to highlight
A
the areas where the predicted CMAs fulfil the relative specifications. Notably, for drawing the sweet spot plot a desired
M

value for each CMA is set or put into limits, and this obtained plot is used to point out the various areas where none,

one, more or all the specification(s) are reached for the CMAs. The area where the values of all the predicted CMAs
ED

fulfil the selected specifications is named “sweet spot”. At this level of study no probability of success or failure is

given, but this kind of plot is advantageous to precociously individuate the area where to start performing the Monte-
PT

Carlo simulations, which allow the introduction of the probability concept. Another possibility for identifying the zone

where the target values for the CMAs are reached is the use of desirability function [39,55,63-65]. However, it is crucial
E

to specify that sweet spot plots, overlay plots and desirability functions are based on the mean of the predicted
CC

responses, and do not account for model error and uncertainty in model estimation. Moreover, they do not enable

neither the computation of the probability of success or failure of the method inside the experimental domain, nor risk
A

quantification.

Since the MODR represents the core of AQbD, its appropriate computation is essential to be compliant with such

approach. The MODR can be obtained and plotted in several ways, such as using Monte-Carlo simulations,

bootstrapping techniques or using Bayesian modelling. A widely used strategy to define the MODR is by using the

sweet spot plots and then performing Monte-Carlo simulations [29,52,58,59,61,62,67-72,74,75,77-79]. The probability

of meeting the specifications is calculated by the distribution of the propagated error. This allows performing a risk
16

analysis that considers the propagation of the predictions and the distribution of the responses. Finally, the risk-of-

failure or probability maps are drawn and used to visualize the multidimensional combinations of the CMPs where the

CMAs correspond to the specifications with a defined probability of success. Therefore, the concept of probability is

central to obtain the MODR and a risk-based method. Another strategy applied for computing the MODR is by using

both the Bayesian approach and Monte-Carlo simulations [56,57]. Such approach gives the possibility to study both the

distribution of the error and the uncertainty correlated to the model resulting in a more complete risk analysis. For

further reading it is suggested to refer to Refs. [10,19-21,36]. Summing up, the use of DoE is necessary to define the

T
MODR and can lead to relevant knowledge on the method by sweet spot, overlay and desirability plots, but the

IP
application of DoE alone is not sufficient for being compliant with the risk-based approach. The addition of Monte-

R
Carlo simulations delivers a more throughout approach to the study, and bootstrapping and Bayesian statistics

ultimately complete the risk analysis to the most accurate point.

SC
Every point within the MODR can be essentially selected as working point and several criteria can be considered

by operators in their choice. Practical considerations generally have the priority because they represent the meeting

U
point between the MODR and the laboratory convenience. Before selecting a working point, some authors chose some
N
verification points (test points) to verify the prediction of the models and to validate the MODR. The validation of the
A
MODR, when performed, is often executed with PBD [70], setting as higher and lower levels of the CMPs the
M

respective limits in the MODR.

When a working point is selected, its validation can begin, to test the quantitative performances before its
ED

implementation in routine use. As a matter of facts, the optimization generally concerns the qualitative aspect, but the

quantitative performance of the method must still be tested. Several methodological approaches are available. Almost
PT

all the Authors performed the validation following the ICH guidelines [11,80], while others [56,57,65] used the

accuracy profile methodology [47].


E

Although the MODR is generally considered as a region of theoretical robustness, many authors performed
CC

robustness testing by using DoE in order to verify if the effect on the CMAs of small and deliberate changes of the

CMPs around the working point was significant [29,52,59,62,67-70,72,74,75,77,78,81]. In specific cases, robustness
A

testing could be interesting in order to investigate, among the considered CMPs some method parameters which had

been already fixed in the screening phase or the scouting phase, and thus not evaluated during the optimization phase. In

the case that the MODR is quite limited and comparable to small changes in the CMAs as those generally examined by

robustness studies, it could be possible to use the same DoE for simultaneously validating the MODR and testing

robustness [61].
17

Control strategies are then implemented to ensure that the method is under adequate control during routine use.

As a first line control strategy, SSTs are a standard part of routine application. Normally they are often set during the

validation stage, defining accepted limits for the CMAs on the basis of the values measured during system repeatability

studies. Another important aspect of control strategy can be also represented by possible precautionary statements

derived from robustness study [70]. Alongside, the use of control chart in order to monitor SSTs or quality control

samples is largely in agreement with concepts mentioned by the AQbD approach. Indeed, this graphical tool can be

defined as a continuous process of assessing a quality characteristic in a defined time and thus allowing the monitoring

T
of the analytical performance, which will be in relation with the quality control samples acceptance during routine use.

R IP
4. Conclusions

SC
The implementation of the ICH guideline Q8 in the development of the analytical method is intended to introduce a

more systematic approach to assess the quality for each aspect involved in the pharmaceutical field. Developing QC

U
methods within the AQbD framework results in a wider process understanding, which means facilitation for method

improvements and more flexible regulatory approaches. However, the entire method life-cycle must be considered
N
within the choice of the development strategy. An effective application of AQbD is essential to enjoy its benefits, and
A
for achieving this target the analyst is required to have a deep expertise and sensitivity to better handle risk and error. In
M

conclusion, the Authors strongly suggest and hope for a wider implementation of statistical approaches such as AQbD

for QC method development, which in the future should be completely oriented to error assessment and risk
ED

management, even including validation phase. Not only focused on separation techniques as those discussed in this

paper, these approaches are believed to be further applied in other techniques relevant to the QC area, such as
PT

vibrational spectroscopy, for which some attempts have been made without giving results fully compatible to the risk-

based approach.
E
CC
A
18

References

[1] ICH Harmonised Tripartite Guideline. Quality Risk Management Q9 (2005) International Conference on

Harmonisation of technical requirements for registration of pharmaceuticals for human use.

[2] Pharmaceutical CGMPs for the 21st century – A Risk-Based Approach. Final report (2004) U.S Food and Drug

Administration.

[3] D. Åsberg, M. Nilsson, S. Olsson, J. Samuelsson, O. Svensson, S. Klick, J. Ennis, P. Butterworth, D. Watt, S.

Iliadou, A. Karlsson, J.T. Walker, K. Arnot, N. Ealer, K. Hernqvist, K. Svensson, A. Grinell, P.-O. Quist, A.

T
Karlsson, T. Fornstedt, A quality control method enhancement concept – Continual improvement of regulatory

IP
approved QC methods, J. Pharm. Biomed. Anal. 129 (2016) 273-281.

R
[4] M.K. Parr, A.H. Schmidt, Life cycle management of analytical methods, J. Pharm. Biomed. Anal. 147 (2018)

SC
506-517.

[5] E. Rozet, P. Lebrun, B. Debrus, B. Boulanger, P. Hubert, Design Spaces for analytical methods, Trends Anal.

[6]
Chem. 42 (2013) 157-167.

U
C. Hubert, P. Lebrun, S. Houari, E. Ziemons, E. Rozet, Ph. Hubert, Improvement of a stability-indicating
N
method by Quality-by-Design versus Quality-by-Testing: A case of a learning process, J. Pharm. Biomed. Anal.
A
88 (2014) 401-409.
M

[7] S. Orlandini, S. Pinzauti, S. Furlanetto, Application of quality by design to the development of analytical

separation methods, Anal. Bioanal. Chem. 405 (2013) 443-450.


ED

[8] R. Peraman, K. Bhadraya, Y.P. Reddy, Analytical quality by design: A tool for regulatory flexibility and robust

analytics, Int. J. Anal. Chem. Article ID 868727 (2015) 1-9.


PT

[9] A. Dispas, H.T. Avohou, P. Lebrun, Ph. Hubert, C. Hubert, “Quality by Design” approach for the analysis of

impurities in pharmaceutical drug products and drug substances, Trends Anal. Chem. 101 (2018) 24-33.
E

[10] J.J. Peterson, A Bayesian approach to the ICH Q8 definition of design space, J. Biopharm. Stat. 18 (2008) 959-
CC

975.

[11] ICH Harmonised Tripartite Guideline. Pharmaceutical Development Q2(R1) (2005) International Conference
A

on Harmonisation of technical requirements for registration of pharmaceuticals for human use.

[12] ICH Harmonised Tripartite Guideline. Pharmaceutical Development Q8(R2) (2009) International Conference

on Harmonisation of technical requirements for registration of pharmaceuticals for human use.

[13] S. Orlandini, R. Gotti, S. Furlanetto, Multivariate optimization of capillary electrophoresis methods: A critical

review, J. Pharm. Biomed. Anal. 87 (2014) 290-307.


19

[14] F.G. Vogt, A.S. Kord, Development of quality-by-design analytical methods, J. Pharm. Sci. 100 (2011) 797-

812.

[15] M.H. Rafamantanana, B. Debrus, G.E. Raoelison, E. Rozet, P. Lebrun, S. Uverg-Ratsimamanga, Ph. Hubert, J.

Quetin-Leclercq, Application of design of experiments and design space methodology for the HPLC-UV

separation optimization of aporphine alkaloids from leaves of Spirospermum penduliflorum Thouars, J. Pharm.

Biomed. Anal. 62 (2012) 23-32.

[16] ICH Harmonised Guideline. Technical and Regulatory Considerations for Pharmaceutical Product Lifecycle

T
Management Q12 (2017) International Conference on Harmonisation of technical requirements for registration

IP
of pharmaceuticals for human use.

R
[17] G.P Martin, K.L. Barnett, C. Burgess, P.D. Curry, J. Ermer, G.S. Gratzl, J.P. Hammond, J. Herrmann, E.

Kovacs, D.J. LeBlond, R. LoBrutto, A.K. McCasland-Keller, P.L. McGregor, P. Nethercote, A.C. Templeton,

SC
D.P. Thomas, M.L.J. Weitzel, Stimuli to the Revision Process: Lifecycle Management of Analytical

Procedures: Method Development, Procedure Performance Qualification, and Procedure Performance

Verification Continual Improvement, Pharm. Forum (2013). U


N
[18] P. Borman, M. Chatfield, P. Nethercote, D. Thompson, K. Truman, The application of quality by design to
A
analytical methods, Pharm. Technol. 31 (2007) 142-152.
M

[19] J.J. Peterson, R.D. Snee, P.R. McAllister, T.L. Schofield, A.J. Carella, Statistics in pharmaceutical development

and manufacturing, J. Qual. Technol. 41 (2009) 111-134.


ED

[20] P. Lebrun, Bayesian Design Space applied to Pharmaceutical Development, PhD Thesis (2012) University of

Liège, Belgium.
PT

[21] J.J. Peterson, K. Lief, The ICH Q8 definition of design space: a comparison of the overlapping means and the

Bayesian predictive approaches, Stat. Biopharm. Res. 2 (2010) 249-259.


E

[22] E. Rozet, P. Lebrun, J.F. Michiels, P. Sondag, T. Scherder, B. Boulanger, Analytical procedure validation and
CC

the quality by design paradigm, J. Biopharm. Stat. 25 (2015) 260-268.

[23] M.G. Schweitzer, M. Pohl, M. Hanna-Brown, P. Nethercote, P. Borman, G. Hansen, K. Smith, J. Larew,
A

Implication and opportunities of applying QbD principles to analytical measurements, Pharm. Tech. 34 (2010)

52-59.

[24] R.K. Burdick, D.J. LeBlond, D. Sandell, H. Yang, H. Pappa, Statistical methods for validation of procedure

accuracy and precision, Pharm. Forum 39 (2013).

[25] B. Nickerson (Ed.), Sample preparation of pharmaceutical dosage forms: challenges and strategies for sample

preparation and extraction, Springer, New York, 2011.


20

[26] I. Molnár, H.-J. Rieger, K.E. Monks, Aspects of the “Design Space” in high pressure liquid chromatography

method development, J. Chromatogr. A 1217 (2010) 3193-3200.

[27] R.A. Lionberger, S.L. Lee, L.M. Lee, A. Raw, L.X. Yu,Quality by Design: Concepts for ANDAs, AAPS J. 10

(2008) 268-276.

[28] K. Ishikawa, What is total quality control? The Japanese Way, Prentice-Hall, Englewood Cliffs, 1985.

[29] B. Pasquini, S. Orlandini, M. Villar-Navarro, C. Caprini, M. Del Bubba, M. Douša, A. Giuffrida, R. Gotti, S.

Furlanetto, Chiral capillary zone electrophoresis in enantioseparation and analysis of cinacalcet impurities: Use

T
of Quality by Design principles in method development, J. Chromatogr. A, doi: 10.1016/j.chroma.2018.07.021

IP
[30] J.F. van Leeuwen, M.J. Nauta, D. de Kaste, Y.M. Odekerken-Rombouts, M.T. Oldenhof, M.J. Vredenbregt,

R
D.M. Barends, Risk analysis by FMEA as an element of analytical validation, J. Pharm. Biomed. Anal. 50

(2009) 1085-1087.

SC
[31] M.T. Oldenhof, J.F. van Leeuwen, M.J. Nauta, D. Kaste, Y.M. Odekerken-Rombouts, M.J. Vredenbregt, M.

Weda, D.M. Barends, Consistency of FMEA used in the validation of analytical procedures, J. Pharm. Biomed.

Anal. 54 (2011) 592-595. U


N
[32] G. Lewis, D. Mathieu, R. Phan-Tan-Luu, Pharmaceutical Experimental Design, Marcel Dekker, New York,
A
1999.
M

[33] D.B. Hibbert, Experimental design in chromatography: A tutorial review, J. Chromatogr. B 910 (2012) 2-13.

[34] P.K. Sahu, N.R. Ramisetti, T.Cecchi, S.Swain, C.S. Patro, J. Panda, An overview of experimental design in
ED

HPLC method development and validation, J. Pharm. Biomed. Anal. 147 (2018) 590-611.

[35] P. Goos, J. Bradley, Optimal Design of Experiments: a Case Study Approach, John Wiley & Sons, Chichester,
PT

2011.

[36] E. Del Castillo, Process Optimization: a Statistical Approach, Springer Science & Business Media, New York,
E

2007.
CC

[37] R.H. Myers, D.C. Montgomery, C.M. Anderson-Cook, Response Surface Methodology: Process and Product

Optimization Using Designed Experiments, John Wiley & Sons Inc., New Jersey, 2016.
A

[38] L. Eriksson, E. Johansson, N. Kettaneh-Wold, C. Wikström, S. Wold, Design of Experiments – Principles and

Applications, MKS Umetrics AB, Umeå, Sweden, 2008.

[39] G. Piepel, B. Pasquini, S. Cooley, A. Heredia-Langner, S. Orlandini, S. Furlanetto, Mixture-process variable

approach to optimize a microemulsion electrokinetic chromatography method for the quality control of a

nutraceutical based on coenzyme Q10, Talanta 97 (2012) 73-82.


21

[40] I. Krull, M. Swartz, J. Turpin, P.H. Lukulay, R. Verseput, A Quality-by-Design Methodology for Rapid LC

Method Development, Part I, LC GC Am. 26 (2008) 1190-1197.

[41] I. Krull, M. Swartz, J. Turpin, P.H. Lukulay, R. Verseput, A quality-by-design Methodology for Rapid LC

Method Development, Part II, LCGC N Am 27 (2009) 48-61.

[42] P. Lebrun, B. Boulanger, B. Debrus, P. Lambert, Ph. Hubert, A Bayesian design space for analytical methods

based on multivariate models and predicions, J. Biopharm. Stat. 23 (2013) 1330-1351.

[43] P. Lebrun, B. Govaerts, B. Debrus, A. Ceccato, G. Caliaro, Ph. Hubert, B. Boulanger, Development of a new

T
predictive modelling technique to find with confidence equivalence zone and design space of chromatographic

IP
analytical methods, Chemom. Intell. Lab. Syst. 91 (2008) 4-16.

R
[44] K.E. Monks, H.-J. Rieger, I. Molnár, Expanding the term “Design Space” in high performance liquid

chromatography (I), J. Pharm. Biomed. Anal. 56 (2011) 874-879.

SC
[45] ISO 5725, Accuracy (Trueness and Precision) of Measurement Methods and Results - Part 1, General

Principles and Definitions, 1994.

[46] U
ISO/IEC 17025, General Requirements for the Competence of Testing and Calibration Laboratories, 2017.
N
[47] Ph. Hubert, J. Nguyen-Huu, B. Boulanger, E. Chapuzet, P. Chiap, N. Cohen, P. Compagnon, W. Dewe, M.
A
Feinberg, M. Lallier, M. Laurentie, N. Mercier, G. Muzard, C. Nivet, L. Valat, Validation of quantitative
M

analytical procedure, Harmonization of approaches, STP Pharma Pratiques, 13 (2003) 101-138.

[48] M. Elkarbane, M. Amood Al-Kamarany, H. Bouchafra, M. Azougagh, Y. Cherrah, Ph. Hubert, Y. Vander
ED

Heyden, A. Bouklouze, Total error-based validation including the experimental design-based robustness

evaluation of a stability-indicating method for the simultaneous quantification of hydrochlorothiazide and


PT

valsartan in tablet formulations, Acta Chromatogr. 27 (2015) 195-214.

[49] D.H. Shewiyo, E. Kaale, P.G. Risha, B. Dejaegher, J. De Beer, J. Smeyers-Verbeke, Y. Vander Heyden,
E

Accuracy profiles assessing the validity for routine use of high-performance thin-layer chromatographic assays
CC

for drug formulations, J. Chromatogr. A, 1293 (2013) 159-169.

[50] C. Hubert, S. Houari, E. Rozet, P. Lebrun, Ph. Hubert, Towards a full integration of optimization and validation
A

phases: An analytical-quality-by-design approach, J. Chromatogr. A, 1395 (2015) 88-98.

[51] ICH Harmonised Tripartite Guideline. Pharmaceutical Quality Systems Q10 (2008) International Conference

on Harmonisation of technical requirements for registration of pharmaceuticals for human use.

[52] V. Dobričić, D. Vukadinović, B. Jančić-Stojanović, S. Vladimirov, O. Čudina, AQbD-Oriented development of

a new LC method for simultaneous determination of telmisartan and its impurities, Chromatographia 80 (2017)

1199-1209.
22

[53] K. Iliou, A. Malenović, Y.L. Loukas, Y. Dotsikas, Analysis of potential genotoxic impurities in rabeprazole

active pharmaceutical ingredient via Liquid Chromatography-tandem Mass Spectrometry, following quality-

by-design principles for method development, J. Pharm. Biomed. Anal. 149 (2018) 410-418.

[54] J. Pantović, A. Malenović, A. Vemić, N. Kostić, M. Medenica, Development of liquid chromatographic method

for the analysis of dabigatran etexilate mesylate and its ten impurities supported by quality-by-design

methodology, J. Pharm. Biomed. Anal. 111 (2015) 7-13.

[55] L. Ferey, A. Rainbault, I. Rivals, K. Gaudin, UHPLC method for multiproduct pharmaceutical analysis by

T
Quality-by-Design, J. Pharm. Biomed. Anal. 148 (2018) 361-368.

IP
[56] A. Dispas, V. Desfontaine, B. Andri, P. Lebrun, D. Kotoni, A. Clarke, D. Guillame, Ph. Hubert, Quantitative

R
determination of salbutamol sulfate impurities using achiral supercritical fluid chromatography, J. Pharm.

SC
Biomed. Anal. 134 (2017) 170-180.

[57] B. Andri, P. Lebrun, A. Dispas, R. Klinkenberg, B. Streel, E. Ziemons, R.D. Marini, Ph. Hubert, Optimization

and validation of a fast supercritical fluid chromatography method for the quantitative determination of vitamin

U
D3 and its related impurities, J. Chromatogr. A 1491 (2017) 171-181.
N
[58] A. Tumpa, A. Stajić, B. Jančić-Stojanović, M. Medenica, Quality by Design in the development of hydrophilic
A
interaction liquid chromatography method with gradient elution for the analysis of olanzapine, J. Pharm.
M

Biomed. Anal. 134 (2017) 18-26.

[59] S. Orlandini, B. Pasquini, M. Stocchero, S. Pinzauti, S. Furlanetto, An integrated quality by design and
ED

mixture-process variable approach in the development of a capillary electrophoresis method for the analysis of

almotriptan and its impurities, J. Chromatogr. A 1339 (2014) 200-209.


PT

[60] N.K. Yadav, A. Raghuvanshi, G. Sharma, S. Beg, O.P. Katare, S. Nanda, QbD-Based development and

validation of a stability-indicating HPLC method for estimating ketoprofen in bulk drug and proniosomal
E

vesicular system, J. Chromatogr. Sci. 54 (2016) 377-389.


CC

[61] S. Furlanetto, S. Orlandini, B. Pasquini, C. Caprini, P. Mura, S. Pinzauti, Fast analysis of glibenclamide and its

impurities: quality by design framework in capillary electrophoresis method development, Anal. Bioanal.
A

Chem. 407 (2015) 7637-7646.

[62] S. Orlandini, B. Pasquini, C. Caprini, M. Del Bubba, S. Pinzauti, S. Furlanetto, Analytical Quality by Design in

pharmaceutical quality assurance: Development of a capillary electrophoresis method for the analysis of

zolmitriptan and its impurities, Electrophoresis 36 (2015) 2642-2649.


23

[63] R.N. Dash, M. Habibuddin, T. Humaira, A.A. Patel, Application of quality by design for the optimization of an

HPLC method to determine ezetimibe ina supersaturable self-nanoemulsifying formulation, J. Liq.

Chromatogr. R.T. 38 (2015) 874-885.

[64] D. Awotwe-Otoo, C. Agarabi, P.J. Faustino, M.J. Habib, S. Lee, M.A. Khan, R.B. Shah, Application of quality

by design elements for the development and optimization of an analytical method for protamine sulfate, J.

Pharm. Biomed. Anal. 62 (2012) 61-67.

[65] C.K. Zacharis, E. Vastardi, Application of analytical quality by design principles for the determination of alkyl

T
p-toluenesulfonates impurities in Aprepitant by HPLC. Validation using total-error concept, J. Pharm. Biomed.

IP
Anal. 150 (2018) 152-161.

R
[66] N.R. Wingert, J.B. Ellwanger, L.M. Bueno, C. Gobetti, C.V. Garcia, M. Steppe, E.E.S. Schapoval, Application

of Quality by Design to optimize a stability-inidcating LC method for the determination of tigacrelor and its

SC
impurities, Eur. J. Pharm. Sci. 118 (2018) 208-215.

[67] L. Nompari, S. Orlandini, B. Pasquini, C. Campa, M. Rovini, M. Del Bubba, S. Furlanetto, Quality by design

U
approach in teh development of an ultra-high-performance liquid chromatography method for Bexsero
N
meningococcal group B vaccine, Talanta 178 (2018) 552-562.
A
[68] B. Pasquini, S. Orlandini, C. Caprini, M. Del Bubba, M. Innocenti, G. Brusotti, S. Furlanetto, Cyclodextrin-
M

and solvent- modified micellar electrokinetic chromatography for the determination of captopril,

hydrochlorothiazide and their impurities: A Quality by Design approach, Talanta 160 (2016) 332-339.
ED

[69] S. Orlandini, B. Pasquini, R. Gotti, A. Giuffrida, F. Paternostro, S. Furlanetto, Analytical quality by design in

the development of a cyclodextrin-modified capillary electrophoresis method for the assay of metformin and
PT

its related substances, Electrophoresis 35 (2014) 2538-2545.

[70] S. Furlanetto, S. Orlandini, B. Pasquini, M. Del Bubba, S. Pinzauti, Quality by Design approach in the
E

development of a solvent-modified micellar electrokinetic chromatography method: Finding the design space
CC

for the determination of amitriptyline and its impurities, Anal. Chim. Acta 802 (2013) 113-124.

[71] S. Niedermeier, G.K.E. Scriba, A quality by design-based approach to a capillary electrokinetic assay for the
A

determination of dextromepromazine and levomepromazine sulfoxide as impurities of levomepromazine, J.

Pharm. Biomed. Anal. 146 (2017) 402-409.

[72] S. Krait, M. Douša, G.K.E. Scriba, Quality by Design-guided development of a capillary electrophoresis

method for the chiral purity determination of ambrisentan, Chromatographia 79 (2016) 1343-1350.
24

[73] P.N. Patel, V.S. Karakam, G. Samanthula, S. Ragampeta, Quality-by-design-based ultra high performance

liquid chromatography related substances method development by establishing the proficient design space for

sumatriptan and naproxen combination, J. Sep. Sci. 38 (2015) 3354-3362.

[74] Enantioseparation and impurity determination of ambrisentan using cyclodextrin-modified micellar

electrokinetic chromatography: Visualizing the design space within quality by design framework, J.

Chromatogr. A 1467 (2016)363-371.

[75] S. Orlandini, B. Pasquini, M. Del Bubba, S. Pinzauti, S. Furlanetto, Quality by Design in the chiral separation

T
strategy for the determination of enantiomeric impurities: Development of a capillary electrophoresis method

IP
based on dual cyclodextrinssystems for the analysis of levosulpiride, J. Chromatogr. A 1380 (2015) 177-185.

N. Milošević, A. Vemić, J. Čolović, N. Kostić, A. Malenović, Design of experiments - design space approach

R
[76]

SC
for development of chaotropic chromatography method for determination of trimetazidine dihydrocloride and

two impurities, Chromatographia 80 (2017) 585-592.

[77] M. Jovanović, T. Rakić, A. Tumpa, B. Jančić Stojanović, Quality by design approach in the development of

U
hydrophilic interaction liquid chromatographic metod for the analysis of iohexol and its impurities, J. Pharm.
N
Biomed. Anal. 110 (2015) 42-48.
A
[78] J. Terzić, I. Popović, A. Stajić, A. Tumpa, B. Jančić-Stojanović, Application of Analytical Quality by Design
M

concept for bilastine and its degradation impurities determination by hydrophilic interaction liquid

chromatographic method, J. Pharm. Biomed. Anal. 125 (2016) 385-393.


ED

[79] S. Orlandini, B. Pasquini, C. Caprini, M. Del Bubba, L. Squarcialupi, V. Colotta, S. Furlanetto, A

comprehensive strategy in the development of a cyclodextrins-modified microemulsion electrokinetic


PT

chromatographic method for the assay of diclofenac and its impurities: Mixture-process variable experiments

and quality by design, J. Chromatogr. A 1466 (2016) 189-198.


E

[80] ICH Harmonised Tripartite Guideline. Impurities in new drug substances Q3A(R2) (2006) International
CC

Conference on Harmonisation of technical requirements for registration of pharmaceuticals for human use.

[81] R. Mallik, S. Raman, X. Liang, A.W. Grobin, D. Choudhury, Development and validation of a rapid-ultra high
A

performance liquid chromatography method for the assay of benzalkonium chloride using a quality-by-design

approach, J. Chromatogr. A 1413 (2015) 22-32.


25

Figure legends

Fig. 1. Schematic description of the analytical QC method life-cycle.

T
IP
Fig. 2. Number of publications by year related to QbD (left) and AQbD (right), searching “quality by design” term for

R
title, abstract and keywords and sorting papers between QbD and AQbD, Scopus database (2007-2017).

SC
U
N
A

Fig. 3. Representation of AQbD workflow.


M
ED
E PT
CC
A

Table 1
26

Applications of AQbD to the development of QC analytical methods.

T C Screening Sc Optimizati Op M R
echniqu Analytes MAs CMPs reening on timization ODR ef.
e de CMPs design
sign
H Telmisart R NAa N ACN B S [
PLC an and 3 etention A content in the first BD weet spot 52]
impurities times; gradient step; ACN plots;
P content in the M
eak second gradient onte-
capacity; step; Gradient time Carlo
S simulatio
value ns

T
H Rabepraz S NA N ACN D- M [
PLC ole and 2 value; A content; pH; Conc. oD onte- 53]

IP
genotoxic C of the buffer salt; Carlo
impurities apacity Column simulatio
factor of stationary phase ns
the last

R
eluting
peak

SC
H Ezetimib R NA N ACN B D [
PLC e etention A content; pH; Flow BD esirabilit 63]
time; rate; Injection y
Tailing volume; function
factor; Temperature

heoretica
T
U
N
l plates
H Alkyl p- R Injection P Flow rate; B D [
PLC toluenesulfonate s values; volume; Flow rate; BD Gradient slope; BD esirabilit 65]
A
impurities in Analysis Column Initial ACN y
Aprepitant time; temperature; content function
P Gradient slope;
M

eak Initial
efficienci ACN content;
es H3PO4 conc.
H Ticagrel R Temperatur fF Flow rate; FC M [
ED

PLC or and 3 s values; e; Flow rate; ResVD ACN content D onte- 66]
impurities A pH; Salt Carlo
nalysis conc.; ACN simulatio
time; content ns
R
PT

etention
factor;
M
aximum
E

pressure
H Iohexol, S NA N ACN B S [
CC

PLC its endo-isomer electivity A content; pH; BD weet spot 77]


and 3 impurities α; ammonium acetate plots;
Retentio conc. M
n factors onte-
of last Carlo
A

eluting simulatio
peak ns
H Olanzapi S NA N Temperatur R S [
PLC ne and 7 related values A e; Initial aqueous D weet spot 58]
substances phase content; plots;
Gradient time M
onte-
Carlo
simulatio
ns
H Bilastine R NA N ACN B S [
PLC and 2 impurities etention A content; pH; BD weet spot 78]
factor; Ammonium plots;
27

Selectivit acetate conc. M


yα onte-
Carlo
simulatio
ns
H Ketoprof T Mobile T Mobile FC O [
PLC en heoretica phase ratio; Buffer D phase ratio; Buffer D verlay 60]
l plates; pH; Flow rate; pH plots
P Injection volume;
eak Temperature;
tailing Column
dimension;
Wavelength
H Dabigatr S NA N Initial B M [
PLC an etexilate values A ACN content; BD onte- 54]

T
mesilate Final ACN Carlo
content; Gradient simulatio

IP
time ns
H Trimetaz R NA N pH; ACN C M [
PLC idine and 2 etention A content; HClO4 CD onte- 76]

R
impurities factors; conc. Carlo
Selectivit simulatio
yα ns

SC
H Protamin R pH; Flow P pH; Flow B D [
PLC e sulfate peptides s values; rate; MeOH conc.; BD rate; Temperature BD esirabilit 64]
Tailing Injection volume; y
factors Temperature function

HPLC
U

Bexsero
Protein
antigens of apacity
factor;
C Screening
1: Vial type;
Sample conc.
Sc
reening 1:
Asymmet U Starting
ACN conc.; Ramp
time; Temperature
CD
C
weet spot
plots;
S
67]
[
N
meningococcal Rs Screening ric matrix M
group B vaccine values; 2: Injection 3141//12 onte-
Peak volume; Column Sc Carlo
A
areas type; Starting ACN reening 2: simulatio
conc.; Ramp time; A ns
M

Temperature symmetri
c matrix
2134//16
U 16 APIs R Screening P Optimizati C D [
ED

HPLC for multiproduct etention 1: ACN as organic BD on Design 1 with CD esirabilit 55]
QC activities times; S solvent. ACN: pH; gradient y
values Screening slope; flow rate.
2: ethanol as Optimizati
organic solvent. on Design 2 with
PT

pH; Flow ethanol: pH;


rate; Gradient gradient slope;
slope; Isocratic temperature.
step volume;
Temperature
E

U Naproxe R pH; fF Flow rate; fF O [


CC

HPLC n and its 6 s values; Column chemistry; D Temperature; D verlay 73]


impurities, N Type of Gradient plots,
sumatriptan and umber of organic modifier time. process
its 5 impurities peaks capability
index
A

with
Monte-
Carlo
simulatio
ns
U Benzalko R 10 columns F Mobile Fu O [
HPLC nium chloride s value usion phase A pH; sion AE® verlay 81]
AE® ACN/MeOH ratio software plots
software in mobile phase B;
Starting mobile
phase composition;
Gradient time
S Salbutam S NA N Initial C R [
28

FC ol sulfate and 5 value A %MeOH; Gradient CD etention 56]


impurities time; Back times
pressure; modeling
Temperature ,
Bayesian
model
and
Monte-
Carlo
simulatio
ns
S Vitamin S NA N Temperatur C R [
FC D3 and 3 value A e; Gradient time; CD etention 57]
impurities Final %ethanol times
modeling

T
,
Bayesian

IP
model
and
Monte-

R
Carlo
simulatio
ns

SC
C Zolmitri R Voltage; S Temperatur B S [
ZE ptan and 5 s values; Temperature; ymmetric e; Buffer conc.; BD weet spot 62]
impurities A Buffer conc.; matrix Buffer pH plots;
nalysis Buffer pH 34//9 M
time;
Theoretic
al plates U onte-
Carlo
simulatio
N
ns
C Glibencl T Voltage; S Voltage; B S [
ZE amide and 2 heoretica Temperature; ymmetric Buffer conc.; BD weet spot 61]
A
impurities l plates; Buffer conc.; matrix Buffer pH; plots;
Rs value; Buffer pH; 35//16 Injection M
M

Analysis Injection time time onte-


time Carlo
simulatio
ns
ED

C Metform R Capillary A Injection D S [


yD-CZE in and 5 s values; length; Injection symmetri time; CMβCyD D weet spot 69]
impurities A time; Temperature; c matrix conc.: Buffer plots;
nalysis Voltage; 2334//16 conc.; Buffer pH M
time CMβCyD onte-
PT

conc.; Buffer Carlo


conc.; simulatio
Buffer pH ns

C Levomep S HPγCyD fF HPγCyD FC S [


E

yD-CZE romazine and its electivity conc.; Buffer pH; ResVD conc.; Buffer pH; D weet spot 71]
impurities α values; Buffer Buffer conc. plots;
CC

dextromepromaz A Conc.; M
ine, nalysis Temperature; onte-
levomepromazin time Voltage Carlo
e sulfoxide simulatio
ns
A

C Ambrise C Voltage; fF Voltage; FC S [


yD-CZE ntan and its hiral Rs; Temperature; ResVD Temperatur D weet spot 72]
enantiomeric A Buffer pH; e; plots;
impurity (R)- nalysis Buffer conc.; Buffer M
ambrisentan time γCyD conc. onte-
conc. Carlo
simulatio
ns
C Cinacalc C NA N Voltage; B S [
yD-CZE et and 3 hiral Rs; A Buffer pH; MeOH BD weet spot 29]
impurities Analysis conc; HPγCyD plots;
including the time conc. Monte
enantiomeric Carlo
29

impurity (S)- simulatio


cinacalcet ns
D Levosulp C Type of A Buffer pH; D S [
ouble iride and its hiral Rs; neutral CyD; symmetri SβCyD conc.; D weet spot 75]
C enantiomeric Analysis Buffer conc.; c matrix MβCyD conc.; plots;
yD-CZE impurity time Buffer pH; SβCyD 2135//16 Voltage M
conc.; Neutral CyD onte-
conc.; Voltage Carlo
simulatio
ns
M Amitript R Voltage; S Voltage; D S [
EKC yline and 4 s values; Buffer conc.; ymmetric ACN conc.; n- D weet spot 70]
impurities A SDS conc.; matrix butanol conc.; plots;
nalysis ACN conc.; n- 36 Urea conc. M
time butanol conc.; //16 onte-

T
Urea conc. Carlo
simulatio

IP
ns
C Captopril R Temperatur S Voltage; FC S [
yD- and 1 impurity, s values; e; Voltage; ymmetric Buffer pH; Sodium D weet spot 68]

R
MEKC hydrochlorothiaz A Buffer matrix cholate conc.; plots;
ide and 3 nalysis Conc.; Buffer pH; 37//16 n-butanol M
impurities time Sodium cholate conc.; γCyD conc. onte-

SC
conc.; Carlo
n-butanol simulatio
conc.; γCyD conc. ns
C Ambrise C Capillary A Buffer pH; FC S [
yD-
MEKC
ntan and 4
impurities
including the
hiral
resolutio
n;
length;
Temperature;
Buffer conc.;
symmetri
c matrix
2235//16 U
γCyD conc.;
Voltage
D weet spot
plots;
M
74]
N
enantiomeric Analysis Buffer pH; γCyD onte-
impurity R- time conc.; SDS conc.; Carlo
ambrisentan Voltage simulatio
A
ns
M Almotrip S Process A Process M S [
M

EEKC tan and 3 values; Variables: Voltage; symmetri Variables: PV D-oD weet spot 59]
impurities A Temperatur c matrix Voltage; plots;
nalysis e; Buffer conc.; 22 Temperature; M
2
time Buffer pH 3 //9 Buffer conc.; onte-
ED

Buffer pH; Carlo


Mixture simulatio
components: ns
% Buffer;
% Oil; %
PT

Surfactant/cosurfac
tant
M Coenzy T NA N Process M D [
EEKC me Q10, heoretica A Variables: PV I-oD esirabilit 39]
ascorbic acid, l plates Voltage; y
E

folic acid Buffer conc.; function


Buffer pH;
CC

Mixture
components:
% Buffer;
% Oil; %
Surfactant/cosurfac
A

tant
C Diclofen R Process M Process M S [
yD- ac and 5 s value; Variables: PV D-oD Variables: PV D-oD weet spot 79]
MEEKC impurities Analysis Voltage; Voltage; plots;
time Temperature; Temperature; M
Buffer Buffer pH; onte-
conc.; Buffer pH; MβCyD conc. Carlo
MβCyD Mixture simulatio
conc. components: ns
Mixture % Buffer;
components: % Oil; %
% Buffer; Surfactant/cosurfac
30

% Oil; tant
%
Surfactant/cosurfac
tant

a
NA: not available.
Acetonitrile (ACN); active principal ingredients (APIs); Box-Behnken design (BBD); carboxymethyl-β-
cyclodextrin (CMβCyD); cyclodextrin-modified capillary zone electrophoresis (CyD-CZE); concentration (conc.); D-
optimal design (D-oD); face centered design (FCD); fractional factorial design (fFD); fractional factorial resolution V
design (fFResVD); γ-cyclodextrin (γCyD); hydroxypropyl-γ-cyclodextrin (HPγCyD); I-optimal design (I-oD); methanol
(MeOH); methyl-β-cyclodextrin (MβCyD); mixture-process variable (MPV); resolution (Rs); Plackett-Burman design

T
(PBD); Rechtschaffen design (RD); separation criterion (S); supercritical fluid chromatography (SFC); sodium dodecyl
sulphate (SDS); sulfated-β-cyclodextrin (SβCyD); Taguchi design (TD).

R IP
SC
U
N
A
M
ED
E PT
CC
A

You might also like