Perspective

pubs.acs.org/ac

Sample Preparation Composite and Replicate Strategy for Assay of

Solid Oral Drug Products
Brent Harrington,*, Beverly Nickerson,*, Michele Xuemei Guo, Marc Barber, David Giamalva,
Carlos Lee, and Garry Scrivens

Pzer Worldwide Research and Development, Statistics, 558 Eastern Point Road, Groton, Connecticut 06340, United States

Pzer Worldwide Research and Development, Analytical Research and Development Department, 558 Eastern Point Road, Groton,
Connecticut 06340, United States

Pzer Worldwide Research and Development, Analytical Research and Development Department, 674 Ramsgate Road, Sandwich,
Kent CT13 9NJ, United Kingdom

Pzer Consumer Healthcare, Analytical Development, 1121 Sherwood Avenue, Richmond, Virginia 23220, United States

ABSTRACT: In pharmaceutical analysis, the results of drug product assay testing

are used to make decisions regarding the quality, ecacy, and stability of the drug
product. In order to make sound risk-based decisions concerning drug product
potency, an understanding of the uncertainty of the reportable assay value is
required. Utilizing the most restrictive criteria in current regulatory
documentation, a maximum variability attributed to method repeatability is
dened for a drug product potency assay. A sampling strategy that reduces the
repeatability component of the assay variability below this predened maximum is
demonstrated. The sampling strategy consists of determining the number of
dosage units (k) to be prepared in a composite sample of which there may be a
number of equivalent replicate (r) sample preparations. The variability, as
measured by the standard error (SE), of a potency assay consists of several
sources such as sample preparation and dosage unit variability. A sampling
scheme that increases the number of sample preparations (r) and/or number of dosage units (k) per sample preparation will
reduce the assay variability and thus decrease the uncertainty around decisions made concerning the potency of the drug product.
A maximum allowable repeatability component of the standard error (SE) for the potency assay is derived using material in
current regulatory documents. A table of solutions for the number of dosage units per sample preparation (r) and number of
replicate sample preparations (k) is presented for any ratio of sample preparation and dosage unit variability.

A nalytical test results are critical to the development and

release of a commercial pharmaceutical product. Test
results are used during all stages of drug development to make
decisions, such as selecting and optimizing clinical and
commercial formulations, optimizing manufacturing process
parameters, and assessing and ensuring active pharmaceutical
ingredient (API) and drug product strength (labeled active
ingredient content), quality, and stability against regulatory
specications.
In order to generate accurate, reliable analytical data, the
analytical method must be appropriate for use. An under-
standing of the uncertainty of the reportable potency assay
value (the measured active ingredient content of the drug
product) is required in order to make risk-based decisions of
the product potency assay measurement. It is this under-
standing that allows the team or company to communicate the
risk in decisions concerning product potency.
ICH Q6A1 species universal tests that are generally
applicable to new drug products to ensure safety and ecacy.
These tests include description, identication, assay, and
impurities. Assay testing typically involves reporting the average
value of replicate preparations of a composite of multiple
dosage units in order to generate an assay value representative
of the entire batch of product. The average value of the
replicates is typically dened as the reportable value. Little or
no guidance is given in regulatory documents to dene the
number of replicates and number of units in a composite that is
appropriate for a given drug product. Australian Therapeutics
Good Order 782 species that, for a tablet or capsule without a
British Pharmacopeia monograph, the average content of each
active ingredient in a pooled sample of not fewer than 20
dosage units should be used to determine the assay value for
the active ingredient. This guidance however does not take into
consideration product or method variability which could
signicantly impact the number of replicates and number of
dosage units in a composite needed to obtain reliable results. A
review of USP monographs3 reveals that the majority of the
monographs for oral dosage forms require not fewer than 20
dosage units to be prepared in a composite sample for assay
Received: September 22, 2014
Accepted: November 12, 2014
Published: November 12, 2014

2014 American Chemical Society 11930 dx.doi.org/10.1021/ac503551r | Anal. Chem. 2014, 86, 1193011936
Analytical Chemistry Perspective

Table 1. Compendial Analytical Validation Terms and Denitions

term denition
accuracy The closeness of test results to the true value. The measure of closeness is sometimes termed bias, trueness, or systematic error.
intermediate Variability within a laboratory that may include performing the procedure on dierent days, and/or with dierent analysts, etc.
precision
repeatability Precision of the analytical procedure use within a single laboratory over a short period of time using the same analyst with the same equipment.
standard error The square root of the variance divided by the number of individual results.
(SE)
uncertainty The dispersion of values that could be reasonably attributed to the measurand. The precision of a measurement describes the uncertainty.
variability The degree of agreement among individual test results; a measure of the dispersion among a number of measured values. Usually expressed as a
statistic such as the standard deviation, variance, coecient of variation, relative standard deviation, or the half width of a condence interval.
variance A numerical measure of variability denoted by 2.

Table 2. Factors Aecting the Variability of an Assay Value

method factors
dosage unit
factors sample preparation standard preparation analysis (e.g., HPLC)
(1) content (1) weighing; (2) extraction; (3) dilution; (1) weighing; (2) extraction; (3) dilution; (1) injection precision; (2) separation robustness;
uniformity (4) solution stability; (5) environmental (4) solution stability; (5) environmental (3) detection; (4) integration; (5)
conditions during preparation conditions during preparation environmental conditions during analysis

testing. A few of these monograph assay methods state to use
not less than 10 units. Other monograph methods make no
mention of the number of dosage units and only state to place a
suitable number of tablets or capsules into a suitable volumetric
ask to yield a specic concentration. There exists no
consensus, nor does there appear to be a justied rationale
for the number of dosage units selected in these methods.
An appropriate number of units in the composite sample and
number of sample replicates can be determined as a component
of the variability minimization strategy of the assay result for a
given batch. In cases where the dosage unit has high content
variability, increasing the number of units in the composite
sample preparation will reduce the variability in the assay result.
Testing replicate sample preparations can reduce variability for
cases where there is high assay variability due to the analytical
method itself. It is important to understand the composition of
the variability in potency assays in order to eectively
determine an appropriate number of dosage units in the
composite sample preparation and an appropriate number of
sample replicates to generate the assay value for a given batch.
In this work, a sampling strategy was developed for
preparation of sample replicates and composites of assay test
methods for solid oral drug products by chromatographic
analysis. This strategy accounts for variability due to the
analytical method and dosage form and uses the most
restrictive compendia and regulatory requirements2,46 for
potency assay sampling and variability to determine an
appropriate sample strategy.
Table 2. These components comprise the standard error (SE)
of the reportable potency assay and may be termed as the
repeatability of the assay. Additional variance components enter
into the totality of uncertainty of data generated by an assay
method, such as analyst, instrument, laboratory, etc.7 These
components must also be understood when making a decision
of lot release, for example.
An illustration of these method variance components is
displayed through a uniformity of dosage units (UDU)
experiment. As shown in eq 1, total variability of a typical
uniformity of dosage units experiment consists of at least the
following variance components, sample preparation, injection
precision, standard preparation, dosage unit, etc., and may
include other sources such as day and or analyst.
UDU 2 = samplepreparation 2 + injection precision 2
+ standard preparation 2 + dosage unit 2 + ... (1)
To minimize method uncertainty, it is important to focus on
reducing the largest contributors to this total variability by
understanding the method settings (sample agitation rate,
injection volume, etc.) that minimize these contributors as
illuminated through appropriately designed experiments. When
these variance components can no longer be reduced from a
mechanistic understanding, an additional strategy is to dene a
sampling scheme or replication strategy that minimizes those

components. For example, standard and sample preparation

METHODS
Denition of Analytical Variability. Analytical variability
or potency assay measurement variability is composed of
several compositional sources. These sources of variability
include the components of dosage unit and the analysis method
which consists of sample preparation, standard preparation, and
sample analysis. Analytical variability and other typical
analytical validation terms used throughout this paper are
dened in Table 1.
In the development of an assay method for drug product, the
variability of a single reportable assay value is composed of
those variance components attributed to the individual analysis
method and standard and sample preparations as detailed in
11931
variability, injection variability, and dosage unit variability all
can be reduced through such a sampling scheme, since a
potency assay may consist of multiple sample and/or standard
preparations, injections, and/or dosage units per sample
preparation. In this work, we focus on method variability and
dosage unit variability due to the importance these variables
have on measurement uncertainty. Equation 1 therefore can be
rewritten as eq 1a. It is assumed that the contributions of
injection repeatability and other system specic variability as
well as the eect of the standard preparation scheme have been
optimized in other experiments.7 In this paper, the term
method refers to the variability composed of the system and
sample preparation.
dx.doi.org/10.1021/ac503551r | Anal. Chem. 2014, 86, 1193011936
Analytical Chemistry
UDU 2 = method 2 + dosage unit 2 (1a)
Why Are We Concerned with Uniformity of Dosage
Units in a Potency Assay? To determine a sampling scheme
that minimizes, to an appropriate level, the contribution of the
method and dosage unit components to the overall potency
assay variability, consider a potency assay method experiment
that produces a nal reportable value (Y). A potency assay
consists of a number of dosage units (duj) prepared into a
sample (Si) of which there may be a number of replicate (r)
sample preparations composited to form the reportable assay
value. Scheme 1 illustrates such an experimental run.
Scheme 1. Illustration of the Samples Prepared to Determine
Assay
Each sample preparation (Si) consists of a preparation of k
dosage units that may be added as whole units or ground into
powder and equivalently weighed. The analyte amount derived
from the sample matrix calculation can be represented by eq 2.8
Yij = + i + ij i = 1 to r ; j = 1 to k (2)
Equation 2 illustrates the signal representing the amount of
an analyte (Y) with a true mean () value and individual signals
which vary about this mean according to the contribution of
method (i) and dosage unit (ij) eects. The eects i and ij
are assumed to be independent of one another and are eects
from a random sample with population mean of 0 and variances
2 and 2, respectively. The variability of such an assay is then
illustrated in eq 2a, where (Y) is the reportable value for the
potency assay that consists of the average of r sample
preparations each composed of k dosage units.
1 1
var(Y ) = 2 + 2
r rk (2a)
The variability of the reportable potency assay value consists
of a contribution from both method (2) and dosage unit
variability (2). Because the reportable value (Y) is the only
value observed (i.e., the individual dosage units are not
observed since they are dissolved in the composite sample
solution), the variability attributed to the potency assay cannot
be decomposed into the error attributed to dosage units and
method using observed potency assay values alone. That is,
these variability components are inseparable.
The variability estimates of the method (2) and dosage
units (2) are required to determine the number of replicate
sample preparations and dosage units per sample preparation in
a sampling scheme: r and k in eq 2a. How do we obtain these
estimates if they cannot be determined from the observed
potency values alone? The answer is through uniformity of
dosage unit and other method development experiments such
as an accuracy assessment experiment.9,10 Uniformity of dosage
unit experiments provides an estimate of total reportable
potency assay value variability: method plus dosage unit as
illustrated in eq 1a, above. The section below illustrates a
typical uniformity of dosage experiment to see the variance
contributions in these experiments.
11932
Perspective
Total Variability: Experimental Model, Uniformity of
Dosage Units. In a uniformity of dosage (UDU) experiment,
an individual dosage unit (dui) is prepared into an individual
sample preparation (Si). Scheme 2 illustrates such an
experimental run.
Scheme 2. Illustration of the Samples Prepared to Determine
Uniformity of Dosage Units
Yi = + i + i i = 1 to r (3)
Equation 3 illustrates the signal (Y) representing the amount
of an analyte with a true mean () and individual signals which
vary about this mean according to the contribution of method
(i) and dosage unit (i) eects. The eects i and i are
assumed to be independent of one another and are eects from
a random sample with population mean of 0 and variances of
2 and 2, respectively. The variability of an UDU value is
then as follows in eq 3a.8
var(Y ) = ( 2 + 2) (3a)
Note: 2 can be composed of both potency and weight of
dosage unit variability.
As in the potency assay, the error attributed to method and
dosage units is inseparable: a single measurement is made on
each sample preparation that consists of a single dosage unit.
To determine a sampling scheme (solving for r and k in eq 1a),
an estimate for each individual component must be obtained. A
general solution (If the UDU and potency assay method are the
equivalent method, then a simple algebraic solution exists for
the solution of method and dosage unit variance estimates. The
viability of this technique is dependent upon the correctness of
the potency assay variability estimate.) that works for any
combination of UDU and potency assay method techniques is
dened by obtaining independent variance component
estimates.
In the development of the potency assay method, there are
several experiments that provide estimates of the method
variability (2) alone. With an estimate of method variability
(2), an estimate of the dosage unit variability (2) can be
solved via subtraction in eq 1a. One possible resource for
providing the estimate of method variability (2) is the
accuracy assessment experiment during the methods develop-
ment and validation. This seems to be a logical choice since an
accuracy assessment is usually executed early in the develop-
ment of a method, hence providing the means to determine a
sample composite strategy as early as the rst uniformity of
dosage unit experiment is completed. See Figure 1 for a
owchart of a possible sampling scheme executed workow. At
least one other strategy exists to estimate the method and
dosage unit variance components; however, it requires two
things. The rst is that the UDU and potency assay method are
equivalent. Second, some minimum number of potency assays
is completed on the same homogeneous material of the UDU
experiment such that the method variability is viable for
calculations. Due to random chance, this might not always be
dx.doi.org/10.1021/ac503551r | Anal. Chem. 2014, 86, 1193011936
Analytical Chemistry Perspective

spiked into pure formulation placebo or a mix of product

excipients to more closely mimic a typical product sample
preparation.

Yi = + i* + i* i = 1 to r (4)
Equation 4 illustrates the signal representing the amount of
an analyte (Y) with a true mean () and individual signals
which vary about this mean according to the contribution of
method (i*) and measured spiked content (i*) eects. The
eects i* and i* are assumed to be independent of one
another and are eects from a random sample with a
population mean of 0 and variances of *2 and *2,
respectively. The variability of potency assay results from an
experiment described above is dened in eq 4a.
var(Y ) = ( *2 + *2) (4a)
The estimate of *2 in eq 3a represents the variability about
measured concentrations that are exceedingly precise. It is
logical, therefore, to assume that *2 is approximately zero.
Then, the variance of the observed accuracy values (eq 4a)
consist of only variance due to the concentration preparations
(*2). This value is thought to be a good approximation to 2
in eqs 2a and 3a as long as there exists no unrealized eect of
the dosage unit manufacture that is not captured in this spiking
experiment; i.e., *2 in eq 4a is not an exaggerated
underestimate of 2 in eqs 2a and 3a. Other experiments
may be utilized to determine an estimate of the method
variability (2); however, for most methods, a well-dened
accuracy experiment is the most complete and ecient means
for obtaining this estimate.
Illustrating the Sampling Strategy. With estimates of
*2 (from an accuracy experiment as dened above) and
uniformity of dosage unit variability consisting of the sum of
(2 + 2), the eect of increasing the number of dosage units
(k) and/or sample preparation replicates (r) can be determined
through eq 2a. See the case study below for an example of all
the data and calculations necessary to achieve this.
Figure 2 displays the eect of increasing the number of
dosage units (k) and sample preparation replicates (r) in eq 2a

Figure 1. Flowchart to determine sampling scheme.

the case; hence, the authors recommend the workow

illustrated in Figure 1.
Experimental Model: Accuracy Assessment of Drug
Product Assay. To evaluate the accuracy of an analytical
method, spikes of known analyte amount (Spi) are prepared
into an individual preparation (Ci), usually in replicates of three
or more dierent concentrations. Scheme 3 illustrates such an
experimental run. The preparation (Ci) may be an analyte Figure 2. Eect of sample preparation estimates on the standard error
of a drug potency assay.
Scheme 3. Illustration of the Sample Prepared for
Determination of Accuracy upon the variability of a potency assay as measured by the
standard error of potency (sqrt of eq 2a). As the values of k and
r increase, the variability of the potency assay is reduced. The
amount of reduction is predicated upon the ratio of the
components 2 and 2 to the total assay variability. The
11933 dx.doi.org/10.1021/ac503551r | Anal. Chem. 2014, 86, 1193011936
Analytical Chemistry Perspective

Table 3. Sampling Scheme for Drug Product Potency Assaya

a
Note: The red star pertains to the case study discussed below.

stopping criteria, labeled Target SE in Figure 2, can be Table 4. Uniformity of Dosage Unit Results for Product A
derived using the logic outlined in the Discussion section, Capsules
below.

uniformity of dosage unit (UDU) experimental results

RESULTS average % label standard deviation (SD) of n = 10 dosage units
batch claim per batch
A composites and replicates strategy was developed to
determine the number of dosage units (k) in a sampling 1 99.3 3.27
2 101.2 2.21
composite and the number of composite sample replicates (r)
3 101.4 2.18
required on the basis of the composition of method and dosage
4 100.6 2.54
unit variability. The variability of potency equation (eq 2a)
5 101.4 2.23
demonstrates that assay variability (var.(Y)) consists of method
6 98.1 2.99
variability (2) and dosage unit variability (2). A sampling
7 98.2 1.90
scheme that increases the number of sample preparations (r)
8 100.8 2.63
and/or number of dosage units (k) per sample preparation will
9 100.0 2.66
reduce the assay variability as illustrated Figure 2.
average 100.1 2.51
A table of solutions for (r) and (k) from eq 2a was calculated
for several ratios of method and dosage unit variability. A
simplied version for ease of use in standard practice is shown Table 5. Results of Accuracy Experiment for Product A
in Table 3 below. The number of units in the composite is Capsules
rounded to increments of ve for ease of method % recovery
implementation in routine testing. Table 3 illustrates only a
replicate
fraction of the entire solutions to eq 2a; however, it represents
the vast working range for most products based on our % of nominal concentration 1 2 3 mean (%RSD)
experience and could be expanded to accommodate additional 70 100.5 100.9 100.2 100.5 (0.3)
variance component ratios. 100 100.4 100.0 100.6 100.3 (0.2)

CASE STUDY
The composites and replicates strategy was applied to a capsule
130 99.6 99.3 99.5

overall
99.5 (0.1)

100.1 (0.5)
formulation of product A, a product under development. As
shown in Table 4, uniformity of dosage unit data was available
for nine manufactured batches. Table 5 shows the results of Table 5. (b) Variance of method = method2 = (0.5 0.5) = 0.25.
accuracy experiments. The data in Tables 4 and 5 are used to (3) Estimate dosage unit variance (dosage unit2): (a) Dosage unit
calculate method and dosage unit variability as shown below. variance is obtained by subtraction using eq 1a.
This information is then used to determine the appropriate UDU 2 = method 2 + dosage unit 2
sampling scheme from Table 3.
Product A Capsules: (1) Calculate variance of UDU (UDU2):
6.3 = 0.25 + dosage unit 2
(a) Overall %RSD for UDU from Table 4 = 2.51; (b) Variance
of UDU = UDU2 = (2.51 2.51) = 6.3. (2) Calculate variance
6.05 = dosage unit 2
of method (method2): (a) Overall Method %RSD is 0.5% from
11934 dx.doi.org/10.1021/ac503551r | Anal. Chem. 2014, 86, 1193011936
Analytical Chemistry
(4) Estimate method variability and dosage unit variability as a
percentage of total variability:
component 2 (% total) = (component 2/UDU 2) 100
method 2 (% UDU) = (0.25/6.3) 100 = 4%
dosage unit 2 (% UDU) = (6.05/6.3) 100 = 96%
(5) Determine sampling scheme (number of sample prepara-
tion replicates and number of dosage units per sample
preparation): (a) Using the overall %RSD from Table 1
(2.5%) and the ratio of % sample preparation variability (4%)
and % dosage unit variability (96%) from step 4 above, nd the
number of replicates and units from Table 3 (see star location).
(b) Sampling scheme is 1 replicate of composite of 5 dosage
units (k = 5, r = 1).
RETROSPECTIVE ANALYSIS
A retrospective evaluation of 26 in house immediate release and
controlled release tablet and capsule drug products at dierent
stages of development was performed using the approach
described in this paper. On the basis of the sampling scheme in
Table 3, 20 of the drug products require one replicate of a
composite preparation containing either 5 or 10 units for assay
testing. Three drug products were outside the bounds of Table
3 due to high method variability (e.g, greater than 30%), but
these products had low UDU (%RSD) (e.g., 1.1% to 1.6%).
Using the calculations described in this paper, it was
determined that for these three products one replicate of a
composite preparation containing either 5 or 10 units was
suitable for assay testing. The remaining three products were
outside the bounds of Table 3 due to high UDU (%RSD) (e.g,
between 4.2% and 4.7%). Using the calculations described in
this paper, it was determined that these three products required
either one replicate of 10 units or two replicates of 10 units or
more. On the basis of this retrospective analysis of 26 drug
products, the majority of the products (>75%) were within the
bounds of Table 3 and demonstrate that the sampling scheme
proposed in Table 3 is suitable as tool for sample preparation
method development. For products outside the bounds of
Table 3, analysis can be performed specically for these
products to dene the sampling scheme using the principles
outlined in this manuscript. This approach also shows that the
majority of the product (>85%) requires only one replicate of a
composite preparation containing either 5 or 10 units.
DISCUSSION
A sampling scheme was developed as shown in Table 3 to
determine an appropriate number of sample replicates and
dosage units in the composite sample preparation for a potency
assay. This sampling scheme is based on the sample preparation
variability, dosage unit variability and product UDU (%RSD),
and a criterion of SE of NMT 1.3. The basis for the criterion of
the SE is discussed below.
Criterion. In order to determine a table akin to the one
illustrated in Table 3 above, a criterion for assessing the
adequacy of potency assay variability must be established. This
criterion should contain the most restrictive elements of current
regulatory guidance in order to maximize the condence in
decisions made by such an assay.
Establishing a Minimal Criterion for the Standard
Error of a Drug Product Assay. The sampling strategy for
11935
Perspective
the number of sample replicates and dosage units in the
composite sample preparations for a potency assay method was
selected on the basis of regulatory criteria for uniformity of
dosage units.2,46
To develop the most restrictive criterion, the standard error
of the drug product assay should be minimized. As dened by
eq 2a, the greater the number of dosage units prepared in a
sample, the greater is the reduction in the variability of a
potency assay. This is a square root relationship; therefore, a
balance between number of dosage units (and hence impact on
sample preparation eorts due to volumes and subdilutions
required) and net gain is needed. The greatest number of
dosage units dened in a regulatory document is k = 20 dosage
units as recommended in TGO 78.2 A greater number of
dosage units per sample preparation would further reduce the
variability of the potency assay; however, the choice of k = 20 is
the most conservative with respect to current regulatory
guidance. Since there exist no known criterion for the number
of sample preparation replicates (r) in regulatory guidance, we
will maintain that r = 1.
From USP <905>,7 we can determine a maximum variability
of dosage units (variance of UDU) allowed utilizing the
criterion contained in that document. As outlined in USP
<905>, for solid oral dosage forms, the acceptance value for
uniformity of dosage units (AV) is calculated by the formula
AV = |M X | + ks and must be less than 15, where the terms
are dened below in Table 6.
Table 6. Denitions for Terms Used in Calculating the
Acceptance Value (AV)
variable denition and condition value
k if n = 30, then k = 2.0
n
i = 1 xi
X sample mean
n
n
i = 1 (xi X )2
S sample standard deviation
n1
M if 98.5% X 101.5%, then M = X
if X < 98.5%, then M = 98.5
if X > 101.5%, then M = 101.5
Table 7 illustrates the calculated maximum standard
deviation of dosage units that would still pass USP < 905>
Table 7. Maximum Observed Standard Deviation Allowed in
USP <905> Acceptance Value (AV) Criterion when n = 30
and Batch Mean is 95% or 105% Label Claim
xbar = 95 98.5 < xbar < 101.5 xbar = 105
5.75 7.50 5.75
criterion of AV < 15. The values in Table 7 are derived by
solving for the standard deviation term in the AV equation for n
= 30 dosage units and a specied xbar (s = (AV |M X |)/k).
A batch mean of 95% (105%) label claim was chosen as this
is a typical solid oral dosage form drug product release
specication. Then, the most conservative (smallest value)
criterion for SE of an assay can be computed as 5.75/sqrt(20) =
1.3, as illustrated in Table 8.
The above illustrates the basic components germane to
determine a criterion for the standard error of a potency assay.
Criterion for the standard error of a potency assay other than
dx.doi.org/10.1021/ac503551r | Anal. Chem. 2014, 86, 1193011936
Analytical Chemistry Perspective

Table 8. Maximum Standard Error of a Potency Assay Using (8) Kutner, M. H.; Nachtsheim, C. J.; Neter, J.; Li, W. Applied Linear
Standard Deviation of 5.75% and k = 20 Dosage Units Per Statistical Models, 5th ed.; McGraw-Hill/Irwin: Boston, 1996.
Preparation (9) ICH Q2(R1) Validation of Analytical Procedures: Text and
Methodology; ICH: Geneva, 1994.
xbar = 95 98.5 < xbar < 101.5 xbar = 105 (10) ICH Q2B Validation of Analytical Procedures: Methodology; ICH:
1.29 1.68 1.29 Geneva, 1996.

1.3 can be obtained utilizing probability assessments, of passing

the USP <905>, for example, in conjunction with company lot
release standard practices and in concert with the organizations
risk and control strategies.

CONCLUSION
A sampling strategy for the number of sample replicates and
dosage units in the composite sample preparation for a potency
assay method was developed. This sampling strategy takes into
account the sample preparation variability, dosage unit
variability, and product UDU. It uses the criterion that the
standard error is not more 1.3. A table of solutions for the
number of sample replicates and the number of dosage units
per sample preparation was derived for any scenario of an
analytical method variability composition from such a criterion.
The sampling scheme is demonstrated on a case study. The
procedure illustrated allows a company to dene a sampling
strategy for drug product potency assay methods that conforms
to an internally dened minimal risk-based standard and allows
uncertainty to be communicated in regards to potency assay
decisions. This work provides the assay method basis of
variability understanding upon which additional components of
method variability (e.g., instrument, analyst, environment, etc.)
can now be determined through additional experimentation
such as during the analytical method transfer exercise.

AUTHOR INFORMATION
Corresponding Authors
*E-mail: brent.harrington@pzer.com.
*E-mail: beverly.nickerson@pzer.com.
Notes
The authors declare no competing nancial interest.

ACKNOWLEDGMENTS
The authors would like to thank Debbie Krause, Kim
Vukovinsky, and Loren Wrisley for their encouragement and
support of this work.

REFERENCES
(1) ICH Q6A Specications: Test Procedures and Acceptance Criteria for
New Drug Substances and New Drug Products: Chemical Substances;
ICH: Geneva, 1999.
(2) Therapeutic Goods Order No. 78 Standard for Tablets and
Capsules; Australian Government Department of Health and Ageing,
Therapeutic Goods Administration: Woden, Australia, 2008.
(3) USP 36-NF31 through Second Supplement; The United States
Pharmocopeial Convention: Rockville, MD, 2013.
(4) <905> Uniformity of Dosage Units in USP 36-NF31 through Second
Supplement; The United States Pharmocopeial Convention: Rockville,
MD, 2013.
(5) EP 2.9.40 Uniformity of Dosage Units in European Pharmacopoeia,
Supplement 8.2 to the 8th ed; European Directorate for the Quality of
Medicines & Healthcare: Strasbourg, France, 2014.
(6) 6.02 Uniformity of Dosage Units, in Japanese Pharmacopoeia
Supplement II, 16th ed.; The Ministry of Health, Labour and Welfare:
Tokyo, 2014.
(7) Ermer, J.; Agut, C. J. Chromatogr., A 2014, 1353, 7177.

11936 dx.doi.org/10.1021/ac503551r | Anal. Chem. 2014, 86, 1193011936