European Journal of Pharmaceutical Sciences 21 (2004) 155159

Feasibility study for the rapid determination of the amylose

content in starch by near-infrared spectroscopy
Christiane C. Fertig
a
, Fridrun Podczeck
b,
, Roger D. Jee
a
, Mark R. Smith
a
a
The School of Pharmacy, University of London, 2939 Brunswick Square, London WC1N 1AX, UK
b
Sunderland Pharmacy School, HNSS, University of Sunderland, Chester Road Campus, Sunderland SR1 3SD, UK
Received 20 February 2003; received in revised form 2 September 2003; accepted 25 September 2003
Abstract
Near-infrared (NIR) spectroscopy was used to determine the amylose content in six different starches, whose declared amylose contents
ranged from 2 to 95% m/m. The amylose content of starches can vary considerably between batches depending on growth conditions and time
of harvesting. An NIR calibration model was developed for amylose using simple laboratory produced mixtures of amylose and amylopectin
in different ratios. The spectral region at 17001800 nm showed a good correlation to the amylose content of these mixtures. A simple
absorbance ratio calibration model using standard normal variate and rst derivative pre-treated spectra gave a root mean standard error
of prediction of 1.2% m/m. Application to real samples gave amylose contents in reasonable agreement with the average values stated
by the supplier. NIR spectroscopy provides a rapid and non-destructive method for the quantitative determination and standardisation of
amylose in starch and could make a suitable alternative to traditional techniques, such as complex formation of starch with iodine or
n-butanol.
2003 Elsevier B.V. All rights reserved.
Keywords: Amylose content; Near-infrared spectroscopy; Starch
1. Introduction
Near-infrared (NIR) spectra of materials provides both
physical and chemical information about a given sample
(Mark, 2001). It is able to determine the concentrations
of one or more constituents of a compound from a sin-
gle spectrum, and can be likened to a ngerprint of the
material (Sekulic et al., 1998). Not only is it a rapid and
non-destructive technique, but also requires minimal or no
sample preparation. NIR spectra may be measured in either
the transmittance or reectance modes. Liquids are generally
measured by transmittance, while reectance measurements
are more commonly applied to highly scattering liquids or
solid samples (Mark, 2001).
NIR spectra often exhibit a baseline shift caused by varia-
tions in sample compaction, scatter from the particle surface
and the particle size of the material itself. Particle size, for

Corresponding author. Tel.: +44-191-515-2568;

fax: +44-191-515-2568.
E-mail address: fridrun.podczeck@sunderland.ac.uk (F. Podczeck).
example, determines the spectral path length, which can lead
to a substantial effect on the resultant spectrum. To minimize
the inuence of these parameters the raw spectra are usually
subjected to mathematical pre-treatments before developing
calibration models. Frequently used spectral pre-treatments
are the calculation of rst-derivative or second derivative and
standard normal variate (SNV) transformation. Derivatives
remove baseline shifts and sloping background absorption,
which arises from the physical nature of the sample, such
as particle size (Mark, 2001).
Single wavelength absorbance versus concentration plot
calibrations are rarely possible with NIR spectra because of
overlapping absorbances between the various components.
The simplest calibration models are typically based on mul-
tiple linear regression (MLR), using the absorbance at two
or more wavelengths (Mark, 2001). By using more than one
wavelength allowance for overlapping peaks can be taken
into account.
NIR is an important analytical tool in the pharmaceutical
industry where it is used to identify materials and to mea-
sure their properties, such as composition, moisture content,
content uniformity, homogeneity of mixing, particle size or
0928-0987/$ see front matter 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejps.2003.09.011
156 C.C. Fertig et al. / European Journal of Pharmaceutical Sciences 21 (2004) 155159
as in-process control to monitor production processes (Guo
et al., 1999).
Starch is a complex polysaccharide of -d-glucose
units exclusively, which are joined by a sequence of
-d-(1,4)-glucosidic linkages thus giving rise to linear or
helical chains. These are referred to as amylose. The much
less frequent -(1,6)-glucosidic linkages form the branch
points between the chains thereby creating highly branched
domains, which are called amylopectin. Although amy-
lose and amylopectin are closely related polymers they
exhibit different structures which are distinguishable by
NIR spectroscopy. The ratio of amylose to amylopectin
in native starch is not only genetically controlled but also
depends on the growth conditions and time of harvesting.
Starch is widely used in pharmaceutics as a ller, binder,
disintegrant and thickening agent. Also recently it has been
used in biodegradable lms and coatings (Ali, 2002).
The classical and still commonly used method for the
quantication of amylose and amylopectin is the iodine
reaction coupled with potentiometric or amperometric titra-
tion (Banks et al., 1971). The method is based on the
capacity inherent to amylose to accommodate polyiodide
ions, chiey I
5

, within its helical structure. Since amy-

lopectin is unable to form such complexes because of its
short chains and branch linkages interfering with the for-
mation of stable structures, these complexes are specic
for the amylose fraction (Hizukuri, 1996). However, the
iodine afnity varies within species, hence compromising
the accuracy of this method.
2. Materials and methods
2.1. Materials
Six starch samples were provided by the National Starch
and Chemical Company, Bridgewater, NJ, USA. They con-
sisted of four commercially available types, Hylon 5

,
Hylon 7

, LAPS

and PURE

, and two experimental sam-

ples, called Melojel

corn starch and Amioca

powder
waxy corn starch. For reasons of briefness the latter two
are subsequently referred to as Corn starch and Waxy
starch. Hylon 5 (batch number: BJ 9960, release date:
19.10.1999) and Hylon 7 (batch number: FG 5514, release
date: 19.10.1999) are unmodied high-amylose starches
containing 50 and 71% m/m amylose, respectively (values
given by the supplier). LAPS (batch number: 374964,
release date: 27.09.2000) is a modied low-amylopectin
starch typically containing less than 10% m/m amylopectin
and most commonly no more than 5% m/m. Derived from
native starches it is made up of around 75% m/m of a
high-molecular-mass amylose and of around 20% m/m of
a low-molecular-mass amylose with the amylopectin bal-
ance being approximately 5% m/m. PURE (batch number:
1180:6B, release date: 19.12.2000) is a pure corn starch with
an amylose content of 95% m/m or higher. Accordingly the
amylopectin content is 5% m/m or less. Corn starch (batch
number: 450801, release date: 17.7.2001) is an untreated
corn starch normally containing 27% m/m amylose and 73%
m/m amylopectin. Waxy starch (batch number: 516002,
release date: 17 July 2001) is essentially composed of amy-
lopectin thus containing 0% m/m amylose. Amylopectin
(batch number: 9561E, EC-number: 9037-22-3, Aurora,
OH, USA) is isolated from corn starch and of analytical
grade.
2.2. Methods
2.2.1. Preparation of the samples
Aset of 29 mixtures were prepared frompure amylose and
analytical-grade amylopectin covering the amylose content
range 0100% m/m in approximately 5% m/m steps. Calcu-
lated proportions of amylose and amylopectin were weighed
into 10 ml scintillation vials and mixed in a Turbula mixer
(Turbula mixer, Type T2C, Nr. 870290, Willy Bachofen AG,
Basel, Switzerland; Blender cage: Type 32422, Glen Cres-
ton, Stanmore, England) at 42 rpm for 10 min. The mixtures
were stored in the scintillation vials.
2.2.2. NIR measurements
NIR reectance (R) spectra were measured using a
FOSS NIRSystems 6500 near-infrared spectrometer tted
with a Rapid Content Analyser (FOSS NIRSystems, Sil-
ver Spring, MD, USA). Prior to measurement, a system
suitability test consisting of checking the wavelength, ab-
sorbance scale and photometric noise, was performed. All
spectra were measured with respect to the instruments
ceramic reectance standard. Samples were measured di-
rectly through the bottom of the scintillation vials and were
handled with latex gloves. Samples were vigorously shaken
before being place on the sample stage. Each recorded
spectrum was the average of 64 scans over the wavelength
range of 11002500 nm (approximately 40 s measurement
time).
2.2.3. Data analysis
Data analysis was performed using the instrument sup-
pliers software (Vision

2.51, FOSS NIRSystems, Silver

Spring, MD, USA).
3. Results and discussion
Absorbance (log
10
R) spectra of a number of natural
starch samples are shown in Fig. 1A. Standard normal vari-
ate transformation of the spectra (Fig. 1B) removes the base-
line offsets making it easier to compare spectra. Perhaps the
most notable feature is the water absorption peaks at ap-
proximately 1450 and 1940 nm. Clearly the water content of
the samples vary. The differences due to the varying amy-
lose contents is much less pronounced, however, it can be
more clearly seen in Fig. 2. Fig. 2 shows spectra of pure
C.C. Fertig et al. / European Journal of Pharmaceutical Sciences 21 (2004) 155159 157
1100 1300 1500 1700 1900 2100 2300 2500
Wavelength/nm
S
N
V

a
b
s
o
r
b
a
n
c
e

2.6
0.0
-1. 8
0.3
0.0
-0. 2
A
b
s
o
r
b
a
n
c
e
1100 1300 1500 1700 1900 2100 2300 2500
(A)
(B)
1
2
3
4
5
6 7
Fig. 1. (A) Absorbance (log
10
R) spectra for Hylon 5 (1), Hylon 7 (2),
LAPS (3), PURE (4), Corn (5), Waxy (6) and amylopectin (7) starch
samples. (B) As in part A, but SNV transformed spectra.
amylose (100% m/m) and amylopectin (i.e. 0% m/m amy-
lose). The spectral region 17001800 nm showing the most
important differences. Further spectral interpretation is dif-
cult, however, impurities such as lipids and proteins could
be attributed to the peaks around 1700 and 20002200 nm,
respectively.
1100 1300 1500 1700 1900 2100 2300 2500
Wavelength/nm
S
N
V

a
b
s
o
r
b
a
n
c
e

2.6
0.0
-1. 8
100%
0%
Fig. 2. SNV absorbance (log
10
R) spectra for amylose (100% m/m) and
amylopectin (i.e. 0% m/m amylose) samples.
3.1. Development of the calibration model
Ideally to develop a calibration model to measure the amy-
lose content of starch samples a large number of naturally
occurring starches with a wide range of amylose contents
and containing varying ratios of low and high molecular
mass amylose is required. As such a set of samples was not
available it was decided to carry out a feasibility study by
setting up a model using samples obtained by mixing pure
amylose and amylopectin. Twenty nine mixtures covering
the range 0100% m/m amylose were prepared and then
divided into two sets. A calibration set (21 samples) with
which to develop the model and a validation set (eight sam-
ples) with which to test the model. Examination of the score
plot from Principal Component Analysis of the calibration
set revealed an outlier and this sample was removed leaving
20 samples in the calibration set.
SNV rst-derivative (segment size = 16 nm, gap size = 0
nm) spectra showed particularly good correlation to the amy-
lose content. Fig. 3A shows the SNV rst-derivative spectra
of the calibration set, while Fig. 3B shows the correspond-
ing correlation plot. Good correlation between the spectral
value and amylose content is shown widely across much
of the wavelength range measured. The highest correlations
S
N
V

f
i
r
s
t
-
d
e
r
i
v
a
t
i
v
e

a
b
s
o
r
b
a
n
c
e
0.8
0.0
-0.4
1100 1300 1500 1700 1900 2100 2300 2500
1100 1300 1500 1700 1900 2100 2300 2500
Wavelength/nm
C
o
r
r
e
l
a
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
1.0
0.0
-1.0
(A)
(B)
Fig. 3. (A) SNV rst-derivative absorbance spectra for the calibration set,
(B) plot of correlation between SNV rst-derivative absorbance value and
percentage amylose for the calibration set as a function of wavelength.
158 C.C. Fertig et al. / European Journal of Pharmaceutical Sciences 21 (2004) 155159
Fig. 4. Expanded spectral region showing the dependence of the SNV
rst derivative absorbance on percentage amylose content.
(both positive and negative) are to be found in the region
17001800 nm. Fig. 4 shows this region for blended samples
containing 0100% m/m amylose in 20% m/m steps.
The best-t model for the calibration set was found to be
given by using SNV rst-derivative spectral pre-treatments
and using a MLR model involving the ratio of two spectral
values, Eq. (1).
y = 167.3 111.2

A
1740
A
1704

(1)
where y is the % m/m amylose content, A is the transformed
absorbance at the subscripted wavelength while the coef-
cients 167.3 and 111.2 came from the least squares tting
process. The standard error of calibration (SEC) was 1.4%
m/m. Applying this model to the validation set gave good
NIR predicted values with a root mean standard error of
prediction (RMSEP) of 1.2% m/m, conrming the good pre-
dictive ability of the model.
The SEC and RMSEP were calculated according to
Eqs. (2) and (3).
SEC =

(y y)
2
n
c
2
(2)
RMSEP =

(y y)
2
n
v
(3)
where, y: reference value, y: NIRpredicted value, n
c
: number
of calibration samples and n
v
: number of validation samples.
A plot of NIR predicted amylose content versus refer-
ence amylose content for the calibration set (Fig. 5) yielded
a straight line with a slope and intercept of 0.997 and 0.15,
respectively. The correlation coefcient was 0.9987. Simi-
larly, the plot for the validation set (Fig. 5) gave a straight
line with a slope of 0.979, intercept 1.4 and a correlation of
>0.999.
Fig. 5. Plots of NIR predicted percentage amylose vs. reference amylose
content for both the calibration and validation sets.
3.2. Sample analysis
The natural starch samples were measured and analysed
using the same calibration model. Table 1 summarises the
NIR predicted values for the amylose content along with
Table 1
Amylose content of natural starch samples
Starch
sample
Amylose content
(% m/m)
NIR predicted value (% m/m)
1 2 Mean
Hylon 5 50.0 56.0 56.0 56.0
Hylon 7 71.0 69.0 69.0 69.0
LAPS 95.0 65.0 66.0 65.5
PURE 95.0 98.0 97.0 97.5
Corn 27.0 24.0 24.0 24.0
Waxy 2.0 8.0 7.0 7.5
C.C. Fertig et al. / European Journal of Pharmaceutical Sciences 21 (2004) 155159 159
those provided by the supplier. The amylose content in
Hylon 5 was 56% m/m, which is 6% m/m higher than the
stated value, whereas that for Hylon 7 was 69% m/m, only
slightly below the value stated (71% m/m). By contrast, the
NIR predicted amylose content for LAPS of 65% m/m was
far below the claimed value of 95% m/m. This discrepancy
most probably arises from the fact that LAPS is made up of
75% m/m high molecular mass amylose and 20% m/m low
molecular mass amylose with the rest being amylopectin.
The NIR model developed purely on simple mixtures which
did not include samples of varying low and high molecu-
lar mass amylose could not be expected to predict such a
sample accurately.
Sample PURE gave an amylose content of 97% m/m,
which lies near the declared minimum content of 95% m/m.
Corn starch was found to comprise 24%m/mamylose, which
is close to the claimed value of 27% m/m. Finally, the pre-
dicted amylose content for Waxy starch was low, 7.5% m/m,
though a little higher than the stated value of 2% m/m and
it can therefore be concluded that Waxy starch is essentially
an amylopectin-only material. The analytical-grade amy-
lopectin gave a value of 2.0% m/m, which effectively con-
rms the absence of amylose.
4. Conclusion
This feasibility study has clearly demonstrated that NIR
can provide a simple method for determining the amylose
content in starch samples. Despite the availability of other,
more conventional, methods, such as complex formation
with iodine or n-butanol (Ring et al., 1985), the NIR tech-
nique is more practical in that it is rapid and non-destructive.
Minimum or no sample preparation as well as short sam-
pling time make this technique particularly interesting for
routine analysis. An NIR model set up with a truly repre-
sentative set of natural starch samples could be expected to
provide a precise and accurate assay procedure.
Acknowledgements
The authors are grateful to the National Starch Company
for provision of the starch samples, and to the School of
Pharmacy, London, for a scholarship for CF.
References
Ali, J., 2002. Functional starches. Chem. Ind. 6, 1820.
Banks, W., Greenwood, C.T., Muir, D.D., 1971. The characterization
of starch and its components. Part 3. The technique of semimicro,
differential, potentiometric titration, and the factors affecting it. Strke
23, 118127.
Guo, J.H., Skinner, G.W., Harcum, W., Malone, J.P., Weyer, L.G., 1999.
Application of near-infrared spectroscopy in the pharmaceutical solid
dosage form. Drug Dev. Ind. Pharm. 25, 12671270.
Hizukuri, S., 1996. Starch: analytical aspects. In: Eliasson, A.-C. (Ed.),
Carbohydrates in Food. Marcel Dekker, New York, pp. 347429.
Mark, H., 2001. Fundamentals of near-infrared spectroscopy. In:
Raghavachan, R. (Ed.), Near-infrared Applications in Biotechnology.
Marcel Dekker, New York, pp. 293321.
Ring, S.G., IAnson, K., Morris, V., 1985. Static and dynamic light
scattering studies of amylose solutions. Macromolecules 18, 182
188.
Sekulic, S., Wakeman, J., Doherty, P., Hailey, P., 1998. Automated system
for the online monitoring of powder blending processes using NIR
spectroscopy. Part II. Qualitative approaches to blend evaluation. J.
Pharm. Biomed. Anal. 17, 12851309.