Professional Documents
Culture Documents
Analytical Letters
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713597227
To cite this Article Krull, Ira S. and Swartz, Michael(1999)'Analytical Method Development and Validation for the Academic
Researcher',Analytical Letters,32:6,1067 — 1080
To link to this Article: DOI: 10.1080/00032719908542878
URL: http://dx.doi.org/10.1080/00032719908542878
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
ANALYTICAL LETTERS, 32(6), 1067-1080 (1999)
MINI-REVIEW
Ira S. Krull*
Department of Chemistry, 102 Hurtig Building, Northeastern University,
360 Huntington Avenue, Boston, MA 02115, USA.
and
Michael Swartz
Pharmaceutical Market Development R&D, Waters Corporation,
34 Maple Street, Milford, MA 01757, USA.
INTRODUCTION
During the course of our scientific careers, and after reading numerous
original research papers and reviews submitted for publication to various journals
or books, it has become apparent to us that many, if not most, academic
researchers have little use for method validation studies or demonstrations. Now,
that may sound a bit harsh or a strong way to start this mini-review, but this
1067
situation has been our collective impression after carefully reading and
commenting on hundreds of submitted journal manuscripts over perhaps the past
25 years.
Many papers that are submitted for publication involve some type of
method development, be that in high performance liquid chromatography
(HPLC), high performance capillary electrophoresis (HPCE), mass spectrometry
(MS) or other analytical techniques. Unfortunately, most papers from academia
Downloaded By: [UNAM Direccion General de Bibliotecas] At: 18:12 8 August 2008
do not usually include attempts to partially or fully validate methods. This is very
unfortunate, because the reader is not apprised as to the status of validation of the
method being described, and most importantly, whether or not the method can
indeed be routinely applied by other analysts in other laboratories with the same
or similar instrumentation and skills. In many manuscripts, there is a glaring
omission of analytical figures of merit, such as limits of detection or quantitation,
calibration plots, accuracy and precision, robustness, and reproducibility. There
is also often an omission of repeat measurements on the very same sample
extract, no indication of the number of such repeats (n = 3, 4, 5, ?), no statistical
treatment of the data, nor any application to simulated or real samples for an
attempt at method validation. Such papers become of questionable and dubious
value to the general reader, especially those analysts that might actually wish to
employ the reported (purported) methods in the very near future. How much
more work will be needed on the part of those readers in order to learn whether
or not the described methods will really work on their own samples in their own
labs with their own instrumentation and skills?
This unfortunate state of affairs in submitted and published manuscripts
has resulted in a huge literature of methods that are of really questionable utility,
applicability, validity, or value, other than to the authors. How can we change
this sorry state-of-affairs, so that journals and authors of future submitted
manuscripts, will indeed provide some degree of method validation and
assurance of applicability and transference of the reported methods? That is
really the gist of this mini-review, to provide the readers with some idea of the
already vast literature in the areas of analytical method validation, for academics
METHOD DEVELOPMENT AND VALIDATION 1069
1 *•[ Precision 1
[ Limit of Detection 1
Method
* •
^- Specificity ]
Validation
Linearity 1
* Ranqe )
Downloaded By: [UNAM Direccion General de Bibliotecas] At: 18:12 8 August 2008
Robustness 1
^- System Suitability 1
they are submitting an IND or NDA for a new drug substance. There are
numerous reviews and overviews in the area of analytical method validation and
USP/ICH requirements, and we will just summarize the more essential features
in these1'15. Figure 1 summarizes the ICH method validation parameters, soon to
be adopted by the USP, and what the FDA expects to see for drug method
submittals1. The amount of method validation studies and number of validation
parameters that must be met changes as a drug proceeds from discovery to Phase
III and then, it is hoped, final FDA approval of an NDA.
If the drug does not make it through all stages of Phases I-III clinical
trials, then not all of the validation parameters in Figure 1 need to be
demonstrated. This has all been sorted out by the USP and is discussed
elsewhere1'14*16. Whatever the stage of a drug's progress towards
commercialization and introduction, any pharmaceutical analyst must be familiar
with the fundamental terms of Figure 1. Unfortunately, too many academic
analysts and those that develop newer analytical approaches are not always as
familiar with these terms. Since this mini-review is really being aimed at
academics, we need to critically and correctly define these terms. Various
analytical textbooks may also be scoured for more in-depth, academic
METHOD DEVELOPMENT AND VALIDATION 1071
present. Naturally, small deviations from the true value are usually found, no
method is 100% accurate, but the closeness of method accuracy is an important
marker of the overall, final utility and applicability of that method with real
samples.
When the analytical method is repeated on different aliquots of the same
sample several times, so that the number of samples is greater than three, and the
number of repeat injections from each aliquot is also greater than three, then it is
possible to determine precision. It is recognized that no validated analytical
method can possibly be done or repeated on the same sample less than three
times, from start-to-finish, with each workup injected at least three times. In a
regulated industry, this is so commonplace and so accepted that it is never
questioned, especially involving FDA submittals. Only in academia does this
question of the number of repeats come into the discussion and be questioned.
The closer the agreement amongst the different determinations done on the very
same sample, then the better the precision of that method. The goal, of course, is
to demonstrate both high accuracy and precision, within one laboratory and one
analyst, and then between analysts in different laboratories. Accuracy and
precision therefore become the cornerstones of any quantitative analytical
method, and these must be very clearly demonstrated and reproducible. That
means one must have samples of known composition and known levels of the
analytes of interest. The method of demonstrating accuracy and precision can
involve a variety of practices, such as single-blind, double-blind, interlaboratory
collaborative study, and others1'9"10.
1072 KRULL AND SWARTZ
in the same sample that are not the analyte itself. True quantitation cannot be
performed until the method is first shown to be specific, and that the peak of
interest is the analyte of interest, and nothing else.
Linearity refers to the calibration plot and its slope of detector response
vs. concentrations injected. There can be external standard calibration plots,
internal standard, standard additions, and others, with or without the sample
matrix present (e.g., matrix-matched calibration plots)24. Linearity relates to how
Downloaded By: [UNAM Direccion General de Bibliotecas] At: 18:12 8 August 2008
well detector response (peak height or peak area or peak intensity) relates to or
correlates with increasing concentrations of analyte present. Ideally, the
calibration plot should be linear of several orders of magnitude, including the
concentration levels expected to be found in the final samples.
The ICH defines the linear concentration range depending on the nature of
the sample and the type of method1. It is not necessary to demonstrate linearity
beyond the expected concentration ranges, with some extension at both ends, and
surely one is not required by ICH to demonstrate linearity of as many orders of
magnitude as possible, starting from the LOD or LOQ. There are very practical
reasons for demonstrating a rather limited linearity range for real samples, rather
than the perhaps academic notion of demonstrating linearity over 5-7 orders of
magnitude.
Range relates to the linear portion of the calibration plot and its linearity.
That is, a calibration plot will be linear over a certain range of concentrations, as
above. This concentration magnitude or spectrum is then defined as the range of
the method, wherein accurate and precise quantitation should be best realized.
Again, it is not necessary according to ICH to demonstrate a range of 10 orders
of magnitude, but rather to limit that range to the samples of interest and their
expected concentrations.
Robustness is a term that academicians rarely utilize in their papers, but
one that ICH and FDA have taken very seriously. It refers to how well the
method performs over small, intentional changes in the method's operational
parameters. Robustness is actually demonstrated by experimentation, by
1074 KRULL AND SWARTZ
before real samples are then run. If the system suitability sample does not yield
chromatograms (injected several times, in replicate) as expected, then there is
every reason not to proceed with real samples until the problem(s) is defined and
corrected. System suitability samples are run to ensure the proper operation of
the instrumental system itself, not necessarily the entire method including sample
preparation, and thus such a sample does not constitute a real sample as much as
a made-up, make-believe, partial sample (usually without the entire sample
Downloaded By: [UNAM Direccion General de Bibliotecas] At: 18:12 8 August 2008
matrix present).
We have now defined those USP method validation parameters that any
FDA filing must meet in order to pass muster and avoid rejection of the
application. There is more to regulatory method validation than just the above,
and the reader is urged to approach more intense books and review articles that
discuss this topic in much greater depth. However, since this mini-review is
really aimed at the academic researcher/analyst and their approaches to method
validation, we need to move forward.
demonstrate that peak is pure, homogeneous, and the correct analyte (specificity).
It is necessary to demonstrate the precision of the overall method and how small
changes in its operational parameters may affect the final results, as above. It is
also necessary to demonstrate acceptable (<2-4%) accuracy and precision, in
single or double blind manners, with actual sample matrices. It is necessary to
demonstrate qualitative and quantitative repeatability and reproducibility, with
numerous, repetitive injections of each and every sample preparation. It is
Downloaded By: [UNAM Direccion General de Bibliotecas] At: 18:12 8 August 2008
r i
necessary to demonstrate the ranges of linearity of the calibration plots, whatever
the method of quantitation utilized, the nature of that linearity (correlation
coefficient, y-intercept, slope, etc.).
It is also necessary to demonstrate method optimization, prior to
validation, or as a part of validation, in order to reach complete validation. It is
necessary to demonstrate percent recovery of the analyte of interest from several
real sample matrices, regardless of the method of quantitation being employed. It
is necessary to demonstrate accuracy and precision of quantitation with real
samples, whether one uses single, double or zero blind methods of validation. It
is also necessary to demonstrate the time and effort involved in performing
routine sample analyses with the newer method, if not the actual costs so
incurred. It is also necessary to demonstrate LOD and LOQ for the method, just
so that future users will have a better feel for where the method will and will not
perform in an acceptable manner. It is not necessary to demonstrate complete
ruggedness from lab-to-lab or country-to-country, nor to demonstrate system
suitability methods, since that is not really a part of method validation.
Why then do academic analysts not utilize the above, shortened validation
approaches on a more routine basis? Why do they routinely avoid using statistics
in their data, or to^ndicate the number of repeats used per data point, or how
many times a given sample was assayed via different aliquots? Why do so many
papers appear without any demonstration of true applicability or application to
real world samples, in a true quantitative manner? Why is it that so little
quantitation is demonstrated with real samples using more than a single analyst?
METHOD DEVELOPMENT AND VALIDATION 1077
Why indeed? The reason is really quite simple- these things are not being
required by the reviewers, editors, and journal owners or publishers. The reason
that pharmaceutical analysts do perform such involved method validations is also
quite simple- it is required by FDA in order to get a drug approved and remain in
business. Academics will do what is required from a purely scientific viewpoint
or perspective for publication, but if something, such as robustness, is not asked
or expected by the journal, then it is much easier to just avoid so doing.
Downloaded By: [UNAM Direccion General de Bibliotecas] At: 18:12 8 August 2008
It has been argued that journals do not have the space needed to include
full validation results, and that most readers would not read that data if it were
included. Most readers may be much more interested in the science described,
rather than in any method validation to demonstrate that the science really does
work with real samples. However, to satisfy both camps, it would always be
possible to include the complete validation results via a website version of the
paper, in a supplemental section of the journal available on-line or by writing, or
from the authors on request. The fact of the matter is that the science of
analytical chemistry will progress when any and all new methods of analysis are
demonstrated valid at the time of first publication, with real samples, using some
or all of the guidelines being suggested above. Otherwise, the reader will never
know for certain that the method can indeed be taken from the literature and
directly utilized.
LIST OF REFERENCES
24. I.S. Krull and M.E. Swartz. Validation Viewpoint: Quantitation in method
validation. LC/GC Magazine, 16(12), 1084(1998).
25. I.S. Krull and M.E. Swartz. Validation Viewpoint: System suitability
samples- what they are, why they are, and what they should do? LC/GC
Magazine, in press (January, 1999).
Downloaded By: [UNAM Direccion General de Bibliotecas] At: 18:12 8 August 2008