Food Chemistry 158 (2014) 351–357

Contents lists available at ScienceDirect

Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem

Analytical Methods

The characterization of caffeine and nine individual catechins

in the leaves of green tea (Camellia sinensis L.) by near-infrared
reflectance spectroscopy
Min-Seuk Lee a, Young-Sun Hwang b, Jinwook Lee c, Myoung-Gun Choung b,d,⇑
a
Sulloc Cha R&D Center, Jangwon Co., LTD, Seogwipo, Jeju 699-920, Republic of Korea
b
Department of Herbal Medicine Resource, Dogye Campus, Kangwon National University, Hwangjori #3, Dogye-up, Samcheok 245-907, Republic of Korea
c
USDA-ARS, Tree Fruit Research Laboratory, 1104 N. Western Ave., Wenatchee, WA 98801, USA
d
Department of Biology, University of Texas-Arlington, Arlington, TX 76019, USA

a r t i c l e i n f o a b s t r a c t

Article history: Near-infrared reflectance spectroscopy (NIRS) was used to determine the contents of caffeine and nine
Received 16 January 2013 individual catechins in tea leaves. A total of 665 samples were scanned by NIRS, and also by high perfor-
Received in revised form 13 October 2013 mance liquid chromatography coupled to a diode array detector to determine the contents of caffeine and
Accepted 23 February 2014
nine individual catechins. The calibration models for caffeine, EGC, C, EGCG, EC, ECG, and total catechins
Available online 6 March 2014
had high r2 (more than 0.90) and RSP (the ratio of standard deviation of reference data to SEP(C) in the
external validation set) values (more than 4.1), indicating a good correlation between reference values
Keywords:
and NIRS predicted values. In contrast, the calibration models of GC and EGCG-3Me had low r2 and
Green tea
Near-infrared reflectance spectroscopy
RSP values (below 0.8 and 2.0). Therefore, these results suggest that NIRS could be applied for the rapid
Caffeine determination of the contents of caffeine, EGC, C, EGCG, EC, ECG, and total catechins in tea leaves for
Catechins breeding programs that develop high-quality tea plants.
High performance liquid chromatography Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Tea is consumed worldwide and ranks only second to water as a
beverage. This beverage is prepared from leaves of tea plants,
grown in the tropical and subtropical zones of various continents
(Lee, Prabhu, Meng, Li, & Yang, 2000). Tea is usually distinguished,
viz. Chinese tea (Camellia sinensis) and Assamic tea (Camellia ass-
amica). These two species differ not only morphologically, but also
chemically (Mahanta, 1996). Most tea product consumed in the
world can be categorized into two types, green tea and black tea,
which account for approximately 20% and 80% of the world’s tea
production, respectively (Lee et al., 2000). Green tea is made by
drying the tea leaves, a process which preserves most of the cate-
chins, while black tea is the ‘‘fermented’’ product, in which the cat-
echins have been extensively oxidized into theaflavins (TFs),
thearubigins and other oligomers (Balentine, Wiseman, & Bouw-
ens, 1997; Graham, 1992). Recent interest of polyphenols in green
and black tea has increased due to their antioxidant activities and
their possible roles in the prevention of cancer and cardiovascular
⇑ Corresponding author at.: Department of Herbal Medicine Resource, Dogye
Campus, Kangwon National University, Hwangjori #3, Dogye-up, Samcheok
245-907, Republic of Korea. Tel.: +82 33 540 3321; fax: +82 33 540 3329.
E-mail addresses: cmg7004@kangwon.ac.kr, mchoung@uta.edu (M.-G. Choung).
diseases (Dreosti, Wargovich, & Yang, 1997; Hollman, Tijburg, &
Yang, 1997; Schulz, Engelhardt, Wengent, Drews, & Lapczynski,
1999; Wiseman, Balentine, & Frei, 1997; Yang & Wang, 1993). Fur-
thermore, overwhelming supportive studies indicate that antioxi-
dants, mainly derived from phenolic compounds in tea leaves,
might have a pivotal function in preventing cardiovascular disease
(Nakachi, Matsuyama, Miyake, Suganuma, & Imai, 2000), chronic
gastritis (Setiawan et al., 2001; Shibata et al., 2000) and certain
types of cancer (Fujiki et al., 2001; Jian, Xie, Lee, & Binns, 2004).
In general, majority of phenolic compounds in green tea leaves,
many of which are known as catechins, belong to flavonoids
(Lakenbrink, Lapczynski, Maiwald, & Engelhardt, 2000). The major
tea catechins are a mixture of epicatechin isomers, including ()-
epigallocatechin 3-gallate (EGCG), ()-epigallocatechin (EGC),
()-epicatechin gallate (ECG), and ()-epicatechin (EC). These cat-
echin compounds are mainly responsible for the astringent and
bitter taste of green tea brews (Zhang, Kuhr, Engelhardt, &
Lebensm, 1992). As an instability nature of catechins, hence the
rapid and simultaneous analysis is important in determining the
quality of green tea (Chen, Zhao, Chaitep, & Guo, 2009).
Many analytical approaches, including HPLC and the capillary
electrophoresis have been used for determining the levels of caf-
feine and catechins in green tea (Friedman, Levin, Choi, Kozukue,
& Kozukue, 2006; Goto, Yoshida, Kiso, & Nagashima, 1996; Horie,

http://dx.doi.org/10.1016/j.foodchem.2014.02.127
0308-8146/Ó 2014 Elsevier Ltd. All rights reserved.
352 M.-S. Lee et al. / Food Chemistry 158 (2014) 351–357
Mukai, & Kohata, 1997; Kodamatani, Shimizu, Saito, Yamazaki, &
Tanaka, 2006; Kotani, Takahashi, Hakamata, Kojima, & Kusu,
2007; Kuhr & Engelhardt, 1991; Yang, Ye, Xu, & Jiang, 2007).
Although these analytical methods usually offer a high level of
accuracy and precision, these methods are not only time-consum-
ing, expensive, and labour-intensive but also relatively less advan-
tageous (Kim, Park, & Choung, 2007) for the mass screening of
superior lines of tea plants. Nevertheless, a rapid and accurate
method for the evaluation of quantitative traits in tea leaves is in
high demand for tea breeding programs. Recently, the spectro-
scopic techniques which have been developed the equipment with
improved electronic and optical components, enhanced computers
capacity for effectively process information contained in spectra,
and powerful chemometric tools have facilitated to extend in an
increasing number of fields, thereby allowing efficient manage-
ment of spectra and chemical data obtained from the samples.
The application of rapid analytical techniques, such as near-infra-
red reflectance spectroscopy (NIRS) has many advantages com-
pared to classical standard techniques, such as chromatographic
methods. NIRS analysis is performed not only with considerable
reduction in analytical time and cost, but also with using less haz-
ardous chemicals. In addition, samples can be analysed in their
natural form without destruction, and moreover, simultaneous
analyses of acid detergent fibre (ADF), fatty acid composition, pro-
tein and oil content may be performed (Font, del Rio-Celestino, &
de Haro-Bailon, 2006). Therefore, it may be assumed that the clas-
sical chemical analysis approaches could be replaced with NIRS
which is a fast, accurate and non-destructive analytical tool. Schulz
et al. (1999) attempted to use NIRS to simultaneously predict alka-
loids and phenolic substances in green tea leaves. Recently, FT-NIR
along with multivariate calibration methods has been applied to
simultaneously analyse the content of major catechins and total
polyphenols in green tea (Chen et al., 2009; Chen, Zhao, Liu, Cai,
& Liu, 2008). However, these studies are only useful to determine
caffeine and major catechins in green tea leaves.
To date, it is known that the individual catechins have func-
tioned to enhance diverse health properties of tea, such as antiox-
idative activity (Schulz et al., 1999). It is recognized that a rapid
and simple approach is highly demanded to determine individual
catechins in tea leaves. Therefore, the objective of this study was
to investigate NIRS application to simultaneous estimate the con-
tents of nine individual catechins and caffeine in green tea leaves
for the tea plant breeding program.
2. Material and methods
2.1. Tea sample
A total of 665 samples of tea (C. sinensis L.) leaves as indigenous
tea germplasms collected from Korea, Japan, and China were
obtained from the tea research farm of Sulloc Cha R&D Center,
Jangwon Co., Seogwipo, Jeju, Republic of Korea. The first through
the third leaves of the first apical bud were harvested from April
17 to April 26, 2010. Approximately 200 g of harvested fresh leaves
were immediately processed for the production of green tea fol-
lowed by the method as reported previously (Choung & Lee, 2011).
2.2. Chemicals
Analytical grade methanol and ethanol were purchased from
Merck Co. (Darmstadt, Germany). The authentic standards of
caffeine and eight catechins, which were gallic acid (GA), ()-gallo-
catechin (GC), ()-epigallocatechin (EGC), (+)-catechin (C),
()-epigallocatechin gallate (EGCG), ()-epicatechin (EC),
()-gallocatechin gallate (GCG), and ()-epicatechin gallate (ECG)
were purchased from Sigma–Aldrich Chemical Co. (St. Louis, Mo,
USA). ()-Epigallocatechin 3-(300 -O-methyl) gallate (EGCG-3Me)
was obtained from Nacalai tesque Inc., (Kyoto, Japan). Phosphoric
acid was purchased from Sigma–Aldrich Chemical Co. Ultra-pure
water was supplied by a Mili-Q water purifier system from Millipore
(Bedford, MA, USA). Stock solutions for each caffeine and catechins
(500 mg/L) were prepared in 20% methanol in water (v/v) and
stored at 20 °C.
2.3. Reference analysis of caffeine and catechins contents using HPLC-
DAD
In order to determine the contents of caffeine and individual
catechins in green tea leaves, approximately 0.1 g of the powdered
sample was weighed and extracted with 20 mL of ethanol–water-
phosphoric acid (40:58:2, v/v/v) for 2 h at 25 °C (Choung & Lee,
2011). The extract was filtered with filter paper and 0.45 lm mem-
brane filter (Millex-HN, Millipore, Bedford, MA, USA), and then di-
luted with water (1:4, v/v). 20 lL of diluted solution was injected
into an Agilent 1200 series HPLC system (Wilmington, DE, USA),
equipped with a photo diode array detector (DAD) operating at
280 nm. Separations were performed using a YMC ODS AM 303 col-
umn (5 lm; 250  4.6 mm i.d., Waters, Milford, MA, USA) protected
with a Nova-Pak C18 guard insert column (Waters, Milford, MA,
USA). Column temperature was maintained at 30 °C. Mobile phases
consisted of solvent A (0.1% (v/v) phosphoric acid in filtered MilliQ
water), and solvent B (methanol). Elution was as follows:
0–25 min, isocratic at 20% B; 25–50 min, gradient from 20% to 30%
B; 50–51 min, gradient from 30% to 20% B; and 52–60 min, isocratic
at 20% B. The flow rate was 1.0 mL/min and the injection volume was
20 lL (Choung & Lee, 2011). Caffeine and catechin contents were
calculated by reporting peak areas to external standard calibration
curves. The linear standard calibration curves (r2 P 0.999⁄⁄) were
calculated by injecting solutions of the standards in 20% methanol
in the concentration range of 0.5–1000 ng/mL.
2.4. Spectra collection and pretreatment
The NIR spectra were measured on a NIRSystem model 6500
near-infrared scanning monochromator (Foss NIRSystems Inc., Sil-
ver Spring, MD, USA) in the reflectance mode. Ground samples
(about 2 g) were used for the measurement. All spectral data were
recorded as the logarithm of the reciprocal of reflectance (log 1/R)
in the wavelength range from 400 to 2500 nm at 2 nm intervals to
give a total of 1050 data points per sample. NIRS instrument con-
trol as well as all graphics and NIRS specific calculations were all
performed with the WinISI II software (Windows version 1.50, Foss
and Infrasoft International LLC, State College, PA, USA) (Lee &
Choung, 2011)
The Win ISI Score program was used to determine samples that
were spectral outliers and those to represent the entire sample set
(n = 665) before multivariate data analysis. The distance between a
sample and their spectrum center was measured as the Mahalan-
obis distance (H distance) and was used as a criterion for selecting
the representative sample sets. The Score algorithm ranks spectra
according to Global H (GH) distance from the center spectrum,
gives spectral boundaries to eliminate outliers with H > 3.0 for
the development of an accurate and robust prediction equation
(Shenk & Westerhaus, 1991b). The final number of samples for cal-
ibration and validation was variable and based on the cutoff point
of H distance, depending on the spectral and chemical variability of
samples in the population used for NIRS estimation. The samples
(n = 640) were split into two sets using the Score and Select pro-
gram in WinISI software. The distance between a sample and its
Neighbourhood H (NH) was measured as the H distance. The sam-
ples with similar spectra (H < 0.6) were randomly divided up in the
 M.-S. Lee et al. / Food Chemistry 158 (2014) 351–357
calibration and validation set in the ratio of two to one, and the
samples with H > 0.6 were randomly divided up in the calibration
and external validation set in the ratio of one to one. The calibra-
tion set (385 samples) was used to calibrate and cross-validate
the derived regression model, and the other 255 samples were
used as an external validation set to test the fit of the derived
regression models.
2.5. Data processing
The regression models for NIRS prediction were developed
using the Global program in WinISI software with modified partial
least-squares (MPLS) regression using wavelengths of the entire
visible (400–1100 nm) and near-infrared (1100–2500 nm) regions
at every 2 nm. Besides MPLS, regression methods, such as partial
least-squares (PLS), principal component regression (PCR), and
multiple linear regression (MLR) were tested to develop calibration
for caffeine and individual catechins contents in tea leaves. Various
mathematical treatments using the raw optical spectrum (log 1/R),
or first or second derivatives of the 1/R data, were applied for cal-
ibration model development. For example, in 1,4,4,1, the first num-
ber indicates the order of the derivative (one is the first derivative
of log 1/R), the second number is the gap in data points over which
the derivative was calculated, and the third and fourth numbers
represent the number of data points used in first and second smoo-
things, respectively (Shenk & Westerhaus, 1991a). The application
of the first and second-derivative algorithm to the raw spectra (log
1/R) induced a clear separation between peaks, which overlapped
in the raw spectra (Lee & Choung, 2011). In addition, diverse scat-
ter correction methods (none, SNVD, Weighted MSC, Detrend only)
were evaluated for the calibration. Standard normal variate trans-
formation (SNV) only takes each spectrum and scales it so it has a
standard deviation of 1.0. Detrend only indicates that the linear
and curvature of each spectrum are removed. Standard normal var-
iate transformation detrend (SNVD) or SNV and detrend presents
that the SNV and detrend calculations are combined into one cor-
rection. Weighted multiplicative scatter correction (MSC) indicates
that a correction for the mean and standard deviation is used at
each wavelength. Application of diverse scatter correction trans-
formation to raw spectral data reduces differences in spectra
related to physical characteristics such as particle size and path
length of samples (Barnes, Dhanoa, & Lister, 1989; Shenk &
Westerhaus, 1991c). Calculated calibration statistics included the
standard error of calibration (SEC), the coefficient of determination
in calibration (R2), and the standard error of cross-validation
(SECV). The performances of the different regression models
obtained in the calibration were determined from cross-validation
as an internal validation method. Internal cross-validation was
used to avoid overfitting of the regression models by selecting
the minimum number of PLS terms in each model (Shenk &
Westerhaus, 1996). The best predicted regression models for each
chemical component were selected on the basis of minimizing
SECV and increasing R2 (Windham, Mertens, & Barton, 1989).
Two passes to eliminate outliers were set by two outlier detection
methods, t and H statistics (Mahalanobis distance) in WinISI
software. The t statistics identified outliers having residuals from
reference analysis of >2.5 times the SEC. Outliers indicated that
their reference values were in doubt and that the samples were
in different populations due to atypical spectra (Williams &
Sobering, 1996). The ratio (SD/SECV) of the standard deviation of
reference data (SD) to SECV, designated RSC, was calculated as a
criterion for evaluating the performance of calibrations (Williams
& Sobering, 1996). After calibration, the developed regression
models allowed for accurate analysis of many other samples by
prediction of data based on the spectra. In addition to the internal
cross-validation, the external validations of calibration models
353
were tested for the prediction capacity on the basis of the standard
error of prediction (SEP) and the coefficient of determination in
prediction (r2). The ratio of SD for the external validation samples
to the corrected SEP (designated RSP) was also used as a criterion
to evaluate the accuracy of the regression models. The validation
sample set allowed the NIRS regression model to be validated for
prediction accuracy, using random samples not included in the cal-
ibration sample set (Williams & Sobering, 1996). The regression
models selected for caffeine and catechins in tea leaves were mon-
itored with the Monitor program in WinISI software, using the val-
idation sample set (n = 255).
3. Results and discussion
3.1. Spectroscopic analysis
The first-derivative spectra and mean standard deviation (SD)
spectrum of tea leaves samples are shown in Fig. 1. The first-deriv-
ative was calculated from the log (1/R) spectra at gaps of four data
points (2 nm) and a smoothing over segments of four data points
(1,4,4,1) with scatter correction (Weight MSC). The main different
absorption bands in the mean SD spectrum at near-infrared
regions (1100–2500 nm) were observed at 1896 nm [related to
phenolic OH and C@O stretching second overtone (COOH)], 1390,
2310, 2320 and 2460 nm (related to CH2), 1418, 1958 and
2380 nm (related to ROH), 1480 nm (related to NH), 1690 and
2256 nm (related to CH3), 2026 nm (related to CONH), and 2116
and 2150 nm (related to HC@CH). Here, several high standard devi-
ation bands were closely connected to various functional groups
(phenolic OH, CH2, COOH, ROH, CH3, CONH, and HC@CH). Those
bands were used for NIRS calibration for caffeine and catechins.
The information for the functional groups in the spectrum was
determined using WinISI software.
Compared to the references, absorption bands at 1690 nm (CAH
stretching first overtone, –CH3), 2250 nm (CAH stretching plus
CAH deformation, CH3), 2326 nm (all CAH stretching plus CAH
deformation, CH2), 2422 nm (CAH stretching plus CAC stretching,
CH2) and 2050 nm (NAH asymmetric stretching, CONH2) were
associated with caffeine. Also, the different absorption bands at
1410 nm (OAH stretching first overtone, ROH), 1900–1940 nm
(OAH stretching first overtone, phenolic-OH) and 2130–2150 nm
@CAH stretching plus C@C stretching, HC@CH) were associated
with polyphenols, of which absorption bands were appearing at
similar wavelengths to these spectrum (Kenjiro et al., 1998).
3.2. Reference analysis of caffeine and catechin contents using HPLC-
DAD
Reference analyses were followed by the previous HPLC-DAD
methods to determine caffeine and nine individual catechins
(Choung & Lee, 2011). Gallic acid (GA), ()-gallocatechin (GC),
()-epigallocatechin (EGC), (+)-catechin (C), caffeine, ()-epigallo-
catechin gallate (EGCG), ()-epicatechin (EC), ()-gallocatechin
gallate (GCG), ()-epigallocatechin 3-(300 -O-methyl) gallate
(EGCG-3Me), and ()-epicatechin gallate (ECG) in tea leaves were
detected within 51 min with base-line resolution.
Numerous studies have reported on the variation of individual
catechins and caffeine contents among genotypes (El-Shahawi,
Hamza, Bahaffi, Al-Sibaai, & Abduljabbar, 2012; Schulz et al.,
1999; Zuo, Chen, & Deng, 2002). The descriptive statistics including
average, standard deviation (SD), and range for nine individual cat-
echin and caffeine contents of the tea samples in the calibration
and validation sets are shown in Tables 1 and 2. Large variation
data allow to be applying for the approach to wide ranges of
samples.
354 M.-S. Lee et al. / Food Chemistry 158 (2014) 351–357

Fig. 1. First derivate (1,4,4,1 + Weighted MSC) spectra and standard deviation (SD) of spectrum of green tea leaves.

Table 1 to those in the calibration sample set (Table 2). Also, the observed
Descriptive statistics for the contents (mg/g DW) of caffeine and nine individual range values were larger than other results reported (Chen et al.,
catechins in green tea leaves used in the calibration sample set.
2009; Schulz et al., 1999).
Calibration set n Average Range SD
Caffeine 375 15.8 4.6–35.9 8.6 3.3. Regression model
GA 226 0.13 0.02–0.89 0.13
GC 309 1.2 0.3–2.9 0.8 3.3.1. Calibration
EGC 380 24.4 1.0–59.8 15.1
C 310 1.2 0.1–2.8 0.8
The regression model is built for each compound using the sam-
EGCG 374 42.6 5.6–143.9 24.5 ples in the calibration set (n = 385), the NIR spectra of these sam-
EC 374 6.8 2.0–15.2 3.6 ples, and the reference value of the compound concentration
GCG 376 0.91 0.08–3.28 0.55 determined by HPLC-DAD. Table 3 showed the statistics of calibra-
EGCG-3Me 246 0.46 0.07–2.60 0.45
tions and cross-validations for caffeine and individual catechins
ECG 371 9.3 1.9–26.6 5.6
Total catechins 377 86.8 22.1–206.8 48.1 contents in green tea leaves. The MPLS regression model of the
whole vis–NIR spectral range (400–2500 nm) had the best accu-
Gallic acid (GA); ()-Gallocatechin (GC); ()-Epigallocatechin (EGC); (+)-Catechin
racy for caffeine and individual catechins in the various regression
(C); ()-Epigallocatechin gallate (EGCG); ()-Epicatechin (EC); ()-Gallocatechin
gallate (GCG); ()-Epigallocatechin 3-(300 -O-methyl) gallate (EGCG-3Me); ()-Epi- methods and wavelength ranges.
catechin gallate (ECG). Optimum wavelengths for NIR analysis have generally relied on
empirical calibrations to predict qualitative constituents for agri-
cultural products due to the broad array of chemical compounds
Table 2 present in the samples, which lead to extensively overlapped and
Descriptive statistics for the contents (mg/g DW) of caffeine and nine individual perturbed NIR absorption bands. The regression models for caf-
catechins in green tea leaves used in the validation sample set.
feine, GA, GC, EGC, EC, GCG, EGCG-3Me, ECG and total catechins
Validation set n Average Range SD using mathematical treatment 1,4,4,1 were selected with higher
Caffeine 250 16.4 4.6–35.4 8.9 RSC (SD/SECV) values as the selection criteria of models, rather
GA 158 0.13 0.02–0.75 0.11 than the different mathematical treatments. However, the regres-
GC 213 1.2 0.4–2.9 0.8 sion models for C and EGCG using mathematical treatment
EGC 251 25.5 1.0–53.8 15.2
2,8,6,1 were selected with high RSC values. The reliable regression
C 220 1.2 0.1–2.8 0.8
EGCG 234 42.8 11.5–124.6 23.8 models for caffeine, EGC, EC, GCG, ECG, and total catechins con-
EC 249 7.1 1.8–14.9 3.7 tents using mathematical treatment 1,4,4,1, and C and EGCG con-
GCG 243 0.91 0.11–2.63 0.50 tents using mathematical treatment 2,8,6,1 had high values of R2
EGCG-3Me 167 0.45 0.07–1.80 0.35 (>0.91) and RSC (>2.55), indicating a close relationship between
ECG 249 9.7 1.9–25.7 5.8
reference values and NIRS estimated values. The regression models
Total catechins 254 91.7 30.6–199.9 50.2
for caffeine, EGC, EGCG, EC, ECG, and total catechins as major con-
Gallic acid (GA); ()-Gallocatechin (GC); ()-Epigallocatechin (EGC); (+)-Catechin stituents in tea leaves were accurate with high R2 (>0.95) and low
(C); ()-Epigallocatechin gallate (EGCG); ()-Epicatechin (EC); ()-Gallocatechin
SEC values. However, the models for GA, GC and EGCG-3Me con-
gallate (GCG); ()-Epigallocatechin 3-(300 -O-methyl) gallate (EGCG-3Me); ()-Epi-
catechin gallate (ECG). tents using mathematical treatment 1,4,4,1 were not as accurate
(R2 < 0.90). These results may result from the narrow range and/
or standard deviation within samples used for the calibration.
Total catechins represent the sum of individual catechins con- The regression model for GC had a relatively high R2 (0.86) (similar
tents without the caffeine and GA contents. Each reference value to GCG) and low SECV, although the low 1  VR value observed in
for caffeine and catechins in a validation sample set was similar the cross-validation indicated that it was not a strong model. In
 M.-S. Lee et al. / Food Chemistry 158 (2014) 351–357 355

Table 3
Equation development statistics using MPLS and scatter correction for the NIRS prediction of caffeine and nine individual catechins contents in green tea leaves.

Constituent Math treatment Scatter collection na Calibration Cross-validation

b 2 c
SEC R 1  VRd SECV e
RSCf
Caffeine 1,4,4,1 Weighted MSC 375 1.300 0.97 0.96 1.639 5.26
GA 1,4,4,1 Weighted MSC 226 0.042 0.89 0.78 0.063 2.03
GC 1,4,4,1 Detrend only 309 0.278 0.86 0.72 0.454 1.66
EGC 1,4,4,1 Weighted MSC 380 2.946 0.96 0.94 3.569 4.22
C 2,8,6,1 Standard MSC 310 0.239 0.92 0.88 0.295 2.80
EGCG 2,8,6,1 None 374 3.589 0.98 0.96 4.798 5.10
EC 1,4,4,1 SNV and detrend 374 0.769 0.95 0.93 0.929 3.85
GCG 1,4,4,1 None 376 0.163 0.91 0.85 0.215 2.55
EGCG-3Me 1,4,4,1 SNV only 246 0.239 0.71 0.50 0.515 0.86
ECG 1,4,4,1 SNV and detrend 371 1.103 0.96 0.94 1.422 3.95
Total catechins 1,4,4,1 Weighted MSC 377 7.058 0.98 0.97 9.003 5.34

Gallic acid (GA); ()-Gallocatechin (GC); ()-Epigallocatechin (EGC); (+)-Catechin (C); ()-Epigallocatechin gallate (EGCG); ()-Epicatechin (EC); ()-Gallocatechin gallate
(GCG); ()-Epigallocatechin 3-(300 -O-methyl) gallate (EGCG-3Me); ()-Epicatechin gallate (ECG).
a
Samples used to develop the models.
b
SEC, standard error of calibration.
c
R2, coefficient of determination of calibration.
d
1  VR, one minus the ratio of unexplained variance divided by variance.
e
SECV, standard error of cross-validation.
f
RSC, SD/SECV, the ratio of SD (standard deviation of reference data) to SECV in the calibration set.

this study, the 1  VR statistic for calibration was an important fac- Table 4
tor alongside R2. Even models having high R2 values did not have a Validation statistics for the contents of caffeine and nine individual catechins in green
tea leaves.
good correlation between reference values and NIRS estimated
values if they did not have a high 1  VR value, close to 1.0, as Constituent Averagea SDb Biasc r2 d
SEP(C)e RSPf
an explained variation. Therefore, the best calibration models for Caffeine 16.442 8.903 0.010 0.97 1.538 5.79
caffeine, EGC, EGCG, EC, ECG and total catechins were developed GA 0.125 0.110 0.006 0.85 0.045 2.44
using the mathematical approach compared to the visible and GC 1.217 0.745 0.023 0.78 0.374 1.99
EGC 25.492 15.148 0.135 0.95 3.333 4.54
near-infrared segment (400–2500 nm). C 1.148 0.799 0.025 0.91 0.250 3.20
EGCG 42.835 23.783 0.091 0.97 4.313 5.51
EC 7.079 3.705 0.030 0.95 0.848 4.37
GCG 0.912 0.501 0.008 0.85 0.200 2.51
3.3.2. External validation
EGCG-3Me 0.448 0.348 0.018 0.58 0.256 1.36
The regression model built and cross-validated in the calibra- ECG 9.680 5.748 0.095 0.94 1.419 4.05
tion step is now validated using the samples in the external set Total catechins 91.690 50.190 0.975 0.97 9.463 5.30
(n = 255), which were not included in the calibration process. To
Gallic acid (GA); ()-Gallocatechin (GC); ()-Epigallocatechin (EGC); (+)-Catechin
perform this external validation, the regression model obtained (C); ()-Epigallocatechin gallate (EGCG); ()-Epicatechin (EC); ()-Gallocatechin
in calibration step is used to predict the concentration of caffeine gallate (GCG); ()-Epigallocatechin 3-(300 -O-methyl) gallate (EGCG-3Me); ()-Epi-
and individual catechins in each sample of the external validation catechin gallate (ECG).
a
set using the NIR spectrum of each sample. The deviation of the Calculated average of validation sample set using calibration model.
b
SD, standard deviation of average.
calculated value by the regression model from the reference value c
Bias, average difference between reference and NIRS values.
(determined by HPLC-DAD) is used to evaluate the performance of d 2
r , coefficient of determination of validation.
the regression model. The statistics of external validation for caf- e
SEP(C), the corrected standard error of prediction.
f
feine and individual catechins in tea leaves were shown in Table 4 RSP, SD/SEP(C), the ratio of SD of reference data to SEP(C) in the external vali-
dation set.
which includes r2, SEP(C) (the corrected standard error of predic-
tion), and RSP [SD/SEP(C)] values, which were factors used to eval-
uate the reliability of the calibration model.
With consideration of the low SEP(C) and high r2 and RSP val- Fig. 2 represented the relationship between the reference con-
ues, an accurate prediction can be monitored with the reliability centration determined by HPLC-DAD (X-axis) and the concentra-
of the established calibration models. The r2 and RSP values for tion obtained using the regression model built in the calibration
GC and EGCG-3Me were relatively low; suggesting a poor correla- step (Y-axis) for caffeine, EGCG, and total catechins. These results
tion between reference values and NIRS estimated values, similar demonstrated the accurate predictive capabilities of the calibration
to the calibration models developed in this study. The predictive models for caffeine, GA, EGC, C, EGCG, EC, GCG, ECG, and total cat-
models for caffeine, EGC, EGCG, EC, ECG, and total catechins had echins using NIRS method in tea leaves and further application for
high r2 (0.97, 0.95, 0.97, 0.95, 0.94 and 0.97, respectively) and routine analysis using these calibration models showed a high reli-
RSP values (5.79, 4.54, 5.51, 4.37, 4.05 and 5.30, respectively), ability for application for tea breeding. However, the reliability of
indicating a good correlation between reference values and NIRS the calibration models for the determination of GC and EGCG-
predicted values and the high accuracy of the calibration models. 3Me was considerably low and could not be considered adequate
In contrast, those for GC and EGCG-3Me had low values of r2 and for use in routine for screening large tea breeding populations.
RSP (below 0.8 and 2.0, respectively), indicating they are unsuit- Determination of caffeine and catechins (GA, EGC, C, EGCG, EC,
able for screening purposes. The GC and EGCG-3Me were minor GCG, ECG and total catechins) in tea leaves can be reliably deter-
constituents in tea leaves and calibrations for these compounds mined using NIRS analysis. However, the future availability of tea
did not show sufficient accuracy. Future studies require develop- samples covering a wide range of reference values is required to
ment of a new model for GC, and EGCG-3Me using more diverse obtain an accurate prediction of GC and EGCG-3Me. Although NIRS
tea germplasms then the current study. is a practical method, this result suggests that a wide range of
356 M.-S. Lee et al. / Food Chemistry 158 (2014) 351–357

Fig. 2. Scatter plots of caffeine (A), ()-epigallocatechin gallate (EGCG; (B)), and total catechins (C) content in green tea leaves determined by HPLC-DAD (X-axis) vs by NIR
(Y-axis) for validation sample sets.

contents in minor catechins including GC and EGCG-3Me, derived
from environmentally and genetically diverse tea populations
should provide much more robust and reliable calibration models
in green tea leaves. There are several reports on the estimation of
caffeine and major catechins in green tea using by NIRS methods
(Chen, Zhao, Huang, Zhang, & Liu, 2006; Chen, Zhao, Zhang, &
Wang, 2006; Luypaert, Zhang, & Massart, 2003; Sinija & Mishra,
2009). However, this study is the first report for the estimation
of nine individual catechins of tea leaves simultaneously with a
NIRS calibration model developed. The development of these NIR
 M.-S. Lee et al. / Food Chemistry 158 (2014) 351–357
regression models represents only an initial step. Therefore, the
regression models should be further updated, expanded, and
improved with more samples from different environments and
germplasm populations.
References
Balentine, D. A., Wiseman, S. A., & Bouwens, L. C. M. (1997). The chemistry of tea
flavonoids. Critical Reviews in Food Science and Nutrition, 37, 693–704.
Barnes, R. J., Dhanoa, M. S., & Lister, S. J. (1989). Standard normal variate
transformation and detrending of near-infrared diffuse reflectance spectra.
Applied Spectroscopy, 43, 772–777.
Chen, Q., Zhao, J., Chaitep, S., & Guo, Z. (2009). Simultaneous analysis of main
catechins contents in green tea (Camellia sinensis (L.)) by Fourier transform near
infrared reflectance (FT-NIR) spectroscopy. Food Chemistry, 113, 1272–1277.
Chen, Q., Zhao, J., Huang, X., Zhang, H., & Liu, M. (2006). Simultaneous determination
of total polyphenols and caffeine contents of green tea by near-infrared
reflectance spectroscopy. Microchemical Journal, 83, 42–47.
Chen, Q., Zhao, J., Liu, M., Cai, J., & Liu, J. (2008). Determination of total polyphenols
content in green tea using FT-NIR spectroscopy and different PLS algorithms.
Journal of Pharmaceutical and Biomedical Analysis, 46, 568–573.
Chen, Q., Zhao, J., Zhang, H., & Wang, X. (2006). Feasibility study on qualitative and
quantitative analysis in tea by near infrared spectroscopy with multivariate
calibration. Analytica Chimica Acta, 572, 77–84.
Choung, M. G., & Lee, M.-S. (2011). Optimal extraction conditions for simultaneous
determination of catechins and caffeine in green tea leaves. Food Science and
Biotechnology, 20, 327–333.
Dreosti, I. E., Wargovich, M. J., & Yang, C. S. (1997). Inhibition of carcinogenesis by
tea: The evidence from experimental studies. Critical Reviews in Food Science and
Nutrition, 37, 761–770.
El-Shahawi, M. S., Hamza, A., Bahaffi, S. O., Al-Sibaai, A. A., & Abduljabbar, T. N.
(2012). Analysis of some selected catechins and caffeine in green tea by high
performance liquid chromatography. Food Chemistry, 134, 2268–2275.
Font, R., del Rio-Celestino, M., & de Haro-Bailon, A. (2006). The use of near-infrared
spectroscopy (NIRS) in the study of seed quality components in plant breeding
programs. Industrial Crops and Products, 24, 307–313.
Friedman, M., Levin, C. E., Choi, S. H., Kozukue, E., & Kozukue, N. (2006). HPLC
analysis of catechins, theaflavins, and alkaloids in commercial teas and green
tea dietary supplements: Comparison of water and 80% ethanol/water extracts.
Journal of Food Science, 71, C328–C337.
Fujiki, H., Suganuma, M., Okabe, S., Sueoka, E., Sueoka, N., & Fujimoto, N. (2001).
Cancer prevention with green tea and monitoring by a new biomarker, hnRNP
B1. Mutation Research/Fundamental and Molecular Mechanisms of Mutagenesis,
480–481, 299–304.
Goto, T., Yoshida, Y., Kiso, M., & Nagashima, H. (1996). Simultaneous analysis of
individual catechins and caffeine in green tea. Journal of Chromatography A, 749,
295–299.
Graham, H. N. (1992). Green tea composition, consumption, and polyphenol
chemistry. Preventive Medicine, 21, 334–350.
Hollman, P. C. H., Tijburg, L. B. M., & Yang, C. S. (1997). Bioavailability of flavonoids
from tea. Critical Reviews in Food Science and Nutrition, 37, 719–738.
Horie, H., Mukai, T., & Kohata, K. (1997). Simultaneous determination of
qualitatively important components in green tea infusions using capillary
electrophoresis. Journal of Chromatography A, 758, 332–335.
Jian, L., Xie, L. P., Lee, A. H., & Binns, C. W. (2004). Protective effect of green tea
against prostate cancer: A case-control study in southeast China. International
Journal of Cancer, 108, 130–135.
Kenjiro, I., Takayanagi, H., Anan, T., Iwamoto, M., Uoaumi, J., Nishinari, K., et al.
(1998). Determination of the content of total nitrogen, caffeine, total free amino
acids, theanine and tannin of sencha and maccha by near infrared reflectance
spectroscopy. Bulletin of the National Institute of Vegetable and Tea Science, B2,
47–90.
Kim, K. S., Park, S. H., & Choung, M. G. (2007). Nondestructive determination of oil
content and fatty acid composition in perilla seeds by near-infrared
spectroscopy. Journal of Agricultural and Food Chemistry, 55, 1679–1685.
Kodamatani, H., Shimizu, H., Saito, K., Yamazaki, S., & Tanaka, Y. (2006). High
performance liquid chromatography of aromatic compounds with
photochemical decomposition and tris (2,20 -bipyridine) ruthenium(III)
chemiluminescence detection. Journal of Chromatography A, 1102, 200–205.
357
Kotani, A., Takahashi, K., Hakamata, H., Kojima, S., & Kusu, F. (2007). Attomole
catechins determination by capillary liquid chromatography with
electrochemical detection. Analytical Sciences, 23, 157–163.
Kuhr, S., & Engelhardt, U. H. (1991). Determination of flavanols, theogallin, gallic
acid and caffeine in tea using HPLC. European Food Research and Technology, 192,
526–529.
Lakenbrink, C., Lapczynski, S., Maiwald, B., & Engelhardt, U. H. (2000). Flavonoids
and other polyphenols in consumer brews of tea and other caffeinated
beverages. Journal of Agricultural and Food Chemistry, 48, 2848–2852.
Lee, J. H., & Choung, M.-G. (2011). Nondestructive determination of herbicide-
resistant genetically modified soybean seeds using near-infrared reflectance
spectroscopy. Food Chemistry, 126, 368–373.
Lee, M.-J., Prabhu, S., Meng, X., Li, C., & Yang, C. S. (2000). An improved method for
the determination of green and black tea polyphenols in biomatrices by high-
performance liquid chromatography with coulometric array detection.
Analytical Biochemistry, 279, 164–169.
Luypaert, J., Zhang, M. H., & Massart, D. L. (2003). Feasibility study for the use of
near infrared spectroscopy in the qualitative and quantitative analysis of green
tea, Camellia sinensis (L.). Analytica Chimica Acta, 478, 303–312.
Mahanta, P. K. (1996). Time for tea. CEMTECH, 26, 38–42.
Nakachi, K., Matsuyama, S., Miyake, S., Suganuma, M., & Imai, K. (2000). Preventive
effects of drinking green tea on cancer and cardiovascular disease:
Epidemiological evidence for multiple targeting prevention. BioFactors, 13,
49–54.
Schulz, H., Engelhardt, U. H., Wengent, A., Drews, H. H., & Lapczynski, S. (1999).
Application of near-infrared reflectance spectroscopy to the simultaneous
prediction of alkaloids and phenolic substance in green tea leaves. Journal of
Agricultural and Food Chemistry, 47, 5064–5067.
Setiawan, V. W., Zhang, Z. F., Yu, G. P., Lu, Q. Y., Li, Y. L., & Lu, M. L. (2001). Protective
effect of green tea on the risks of chronic gastritis and stomach cancer.
International Journal of Cancer, 92, 600–604.
Shenk, J. S., & Westerhaus, M. O. (1991a). New standardization and calibration
procedures for NIRS analytical systems. Crop Science, 31, 1694–1696.
Shenk, J. S., & Westerhaus, M. O. (1991b). Population definition, sample selection,
and calibration procedures for near infrared reflectance spectroscopy. Crop
Science, 31, 469–474.
Shenk, J. S., & Westerhaus, M. O. (1991c). Populations structuring of near infrared
spectra and modified partial least squares regression. Crop Science, 31,
1548–1555.
Shenk, J. S., & Westerhaus, M. O. (1996). Calibration the ISI way. In A. M. C. Davis & P.
C. Williams (Eds.), In near infrared spectroscopy: The future waves (pp. 198–202).
Chichester, UK: NIR Publications.
Shibata, K., Moriyama, M., Fukushima, T., Kaetsu, A., Miyazaki, M., & Une, H. (2000).
Green tea consumption and chronic atrophic gastritis: A cross-sectional study
in a green tea production village. Journal of Epidemiology, 10, 310–316.
Sinija, V. R., & Mishra, H. N. (2009). FT-NIR spectroscopy for caffeine estimation in
instant green tea powder and granules. LWT – Food Science and Technology, 42,
998–1002.
Williams, P. C., & Sobering, D. C. (1996). How do we do it: A brief summary of the
methods we use in developing near infrared calibration. In A. M. C. Davis & P. C.
Williams (Eds.), In near infrared spectroscopy: The future waves (pp. 185–188).
Chichester, UK: NIR Publications.
Windham, W. R., Mertens, D. R., & Barton, F. E. (1989). Protocol for NIRS calibration:
Sample selection and equation development and validation. In G. C. Marten
(Ed.), Near infrared reflectance spectroscopy (NIRS): Analysis of forage quality
(pp. 643). Washington, DC: USDA Agricultural Handbook.
Wiseman, S. A., Balentine, D. A., & Frei, B. (1997). Antioxidants in tea. Critical Reviews
in Food Science and Nutrition, 37, 705–718.
Yang, C. S., & Wang, Z. Y. (1993). Tea and cancer. Journal of the National Cancer
Institute, 85, 1038–1049.
Yang, X. R., Ye, C. X., Xu, J. K., & Jiang, Y. M. (2007). Simultaneous analysis of purine
alkaloids and catechins in Camellia sinensis, Camellia ptilophyll and Camellia
assamica var. kucha by HPLC. Food Chemistry, 100, 1132–1136.
Zhang, D., Kuhr, S., Engelhardt, U. H., & Lebensm, Z. (1992). Influence of catechins
and theaflavins on astringent taste of black tea brews. European Food Research
and Technology, 195, 108–111.
Zuo, Y., Chen, H., & Deng, Y. (2002). Simultaneous determination of catechins,
caffeine and gallic acids in green, Oolong, Black and pu-erh teas using HPLC
with a photodiode array detector. Talanta, 57, 307–316.