METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO.
2, 2007 391
DRUG FORMULATIONS AND CLINICAL METHODS
Application of Derivative, Derivative Ratio, and Multivariate
Spectral Analysis and Thin-Layer Chomatography-Densitometry
for Determination of a Ternary Mixture Containing Drotaverine
Hydrochloride, Caffeine, and Paracetamol
FADIA H. METWALLY, YASSER S. EL-SAHARTY, MOHAMED REFAAT, and SONIA Z. EL-KHATEEB
Cairo University, Faculty of Pharmacy, Analytical Chemistry Department, El-Kasr El-Aini St, ET-11562 Cairo, Egypt
New selective, precise, and accurate methods are
described for the determination of a ternary
mixture containing drotaverine hydrochloride (I),
caffeine (II), and paracetamol (III). The first method
uses the first (D1) and third (D3) derivative
spectrophotometry at 331 and 315 nm for the
determination of (I) and (III), respectively, without
interference from (II). The second method depends
on the simultaneous use of the first derivative of
the ratio spectra (DD1) with measurement at
312.4 nm for determination of (I) using the
spectrum of 40 mg/mL (III) as a divisor or
measurement at 286.4 and 304 nm after using the
spectrum of 4 mg/mL (I) as a divisor for the
determination of (II) and (III), respectively. In the
third method, the predictive abilities of the
classical least-squares, principal component
regression, and partial least-squares were
examined for the simultaneous determination of
the ternary mixture. The last method depends on
thin-layer chromatography-densitometry after
separation of the mixture on silica gel plates using
ethyl acetate–chloroform–methanol (16 + 3 + 1,
v/v/v) as the mobile phase. The spots were
scanned at 281, 272, and 248 nm for the
determination of (I), (II), and (III), respectively.
Regression analysis showed good correlation in
the selected ranges with excellent percentage
recoveries. The chemical variables affecting the
analytical performance of the methodology were
studied and optimized. The methods showed no
significant interferences from excipients. Intraday
and interday assay precision and accuracy values
were within regulatory limits. The suggested
procedures were checked using
laboratory-prepared mixtures and were
successfully applied for the analysis of their
pharmaceutical preparations. The validity of the
proposed methods was further assessed by
Received May 11, 2006. Accepted by SW August 27, 2006.
Corresponding author's e-mail: fadiahm@yahoo.com
applying a standard addition technique. The results
obtained by applying the proposed methods were
statistically analyzed and compared with those
obtained by the manufacturer’s method.
D
rotaverine, a benzylisoquinoline derivative, is an
analog of papaverine with excellent smooth muscle
relaxant properties. It is available as the hydrochloride
and theophylline-7-acetic acid salts (1). Several methods have
been reported for the determination of drotaverine in plasma
and pharmaceutical dosage forms, including
spectrophotometry (2–4), electrochemical (5–7), and
high-performance liquid chromatography (LC; 8–12).
In the local market, drotaverine hydrochloride (I) is present
in combination with caffeine (II) and paracetamol (III) in the
drug formulation named “Petro tablets.” The analgesic
properties of (III) provide a logical addition to the action of
(I). The combined antispasmodic and analgesic action makes
Petro tablets especially suitable for the treatment of
paroxysmal pain in the hollow organs of the abdomen.
It is desirable to have an accurate and precise method for
the analytical determination of these drugs in combination.
The literature methods were used for single analysis of each
drug, so there are large demands for sensitive methods with
the ability to simultaneously determine the 3 drugs in their
ternary mixture, as no method is available for this analysis.
Using computer-controlled instrumentation, derivative
techniques and multivariate calibration methods are playing a
very important role in the multicomponent analysis of
mixtures by UV-Vis spectrophotometry (13, 14). Both
approaches are useful for the resolution of band overlapping
in quantitative analysis. Derivative techniques have proved to
be useful in the resolution of simple binary mixtures or turbid
samples (15, 16), whereas multivariate calibration has been
found to be the method of choice for more complex
mixtures (17, 18). The advantage of multicomponent analysis
using multivariate calibration is the speed of the method of
determination for the components of interest in admixture, as
the separation step can be avoided (19).
In order to avoid time-consuming procedures, attempts to
resolve overlapping spectra by using instrumental approaches
392 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007

Table 1. The concentrations of different mixtures of

drotaverine hydrochloride (I), caffeine (II), and
paracetamol (III) used in the training set

Concentration, mg/mL

Sample No. (I) (II) (III)

1 2 1.5 24
2 2 3 20
3 2 4.5 16
4 2 6 12
5 2 7.5 8
6 4 1.5 24
Figure 1. Zero-order absorption spectra of
7 4 3 20
drotaverine hydrochloride (I), 4 mg/mL (——);
8 4 4.5 16 caffeine (II), 6 mg/mL (- - -); and paracetamol (III),
9 4 6 12 10 mg/mL (– – –) in methanol.
10 6 1.5 24
11 6 3 20
12 6 4.5 16
13 8 1.5 24
14 8 3 20
15 10 1.5 24

or various chemometric methods have been made. Both

derivative and multivariate statistical analysis methods
presume that there is a linear relationship between the
absorbance values and the concentration of the components in
the mixture to be analyzed. Each of these methods needs a Figure 2. First-derivative absorption spectra of
calibration step, where the relationship between spectral data drotaverine hydrochloride (I), 4 mg/mL (——);
and concentration is deduced from a set of reference samples. caffeine (II), 6 mg/mL (- - -); and paracetamol (III),
This should be followed by a prediction step in which the 10 mg/mL (– – –) in methanol.
results of the calibration are used to determine the
concentrations of the components from the spectra of the
analytes (20).
The purpose of this study was to determine simultaneously
(I) with (II) and (III) by first (D1) and third (D3) derivative
spectrophotometry and first derivative of the ratio spectra
(DD1) and multivariate spectral analysis techniques and
thin-layer chromatography (TLC)-densitometry assays for
quality control and routine analysis.
Experimental

Apparatus
(a) Spectrophotometer.—Double-beam UV-Vis
spectrophotometer 1601 PC connected to an IBM-compatible
computer, with UVPC personal spectroscopy software
Version 3.7 (Shimadzu Corp., Kyoto, Japan). The absorption
spectra were measured using 1 cm quartz cells over the range Figure 3. Third-derivative absorption spectra of
of 200–400 nm. Data analysis was performed using drotaverine hydrochloride (I), 4 mg/mL (——);
PLS-Toolbox 2.0 running under MatlabTM, Version 5.3 caffeine (II), 6 mg/mL (- - -); and paracetamol (III),
(Eigenvector Research, Manson, WA; 21). 10 mg/mL (– – –) in methanol.
 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007 393

(b) Densitometer.—Dual wavelength, flying spot Working Standard Solutions

CS-9000 densitometer with video display and high-speed,
Four mL of the stock solutions of (I), (II), and (III) were
high-quality, parallel-head printer/plotter (Shimadzu).
transferred in 3 separate 100 mL volumetric flasks and diluted
(c) Microsyringe.—25 mL capacity, calibrated at
to the line with methanol to prepare a final concentration of
0.2 mL/unit (Hamilton Co., Reno, NV).
40, 60, and 400 mg/mL (I), (II), and (III), respectively, for D1,
(d) TLC plates.—Precoated with silica gel GF254, 20 ´
D3, and DD1 spectrophotometric methods and multivariate
20 cm, 0.25 mm thickness (E. Merck, Darmstadt, Germany).
spectral analysis methods. Stock solutions of (I), (II), and (III)
Pure Samples were used for the TLC-densitometric method.

(I), (II), and (III) were kindly supplied by Alpha Chem
Advanced Pharmaceutical Industries Co. (ACAPI Co.),
Cairo, Egypt. Their purity was found to be 99.72 ± 0.69, 99.63
± 0.58, and 100.31 ± 0.47%, respectively, according to a
reported LC method using 2 mobile phases, one for separation
of (I) and the other for separation of (II) and (III) (personal
communication, ACAPI Co., Cairo, Egypt, 2005).
Market Samples
Petro tablets (ACAPI Co., Batch No. 01101157) were
labeled to contain 40, 60, and 400 mg (I), (II), and (III)/tablet,
respectively.
Chemicals and Reagents
All chemicals were of analytical grade, and the solvents
were spectroscopic grade.
(a) Methanol and ethyl acetate.—E. Merck.
(b) Chloroform.—El-Nasr Pharmaceutical Chemicals
(Cairo, Egypt).
Stock Standard Solutions
(I), (II), and (III) stock solutions (1, 1.5, and 10 mg/mL,
respectively) were prepared by accurately weighing 100 mg
(I), 150 mg (II), and 1000 mg (III) powder into 3 separate
100 mL volumetric flasks; 50 mL methanol was added and the
solution was shaken for a few minutes and diluted to volume
with the same solvent.
Laboratory-Prepared Mixtures
(a) For D1, D3, and DD1 spectrophotometric
methods.—Accurate aliquots equivalent to 40–200 mg (I)
were transferred from its working solution (40 mg/mL) into a
series of 10 mL volumetric flasks, and portions equivalent to
60–240 mg (II) from its working solution (60 mg/mL) and
40–400 mg (III) from its working solution (400 mg/mL) were
added to the same flasks. The volumes were diluted to mark
with methanol and mixed well.
(b) For multivariate spectral analysis methods.—Into a
series of 10 mL volumetric flasks, accurately measured
aliquots of (I) equivalent to 20–80 mg were transferred from
its working solution. Aliquots of (II) and (III) equivalent to
15–60 and 120–240 mg, respectively, from their working
solutions were added, diluted to volume using methanol, and
mixed well.
(c) For TLC-densitometry.—Aliquot portions of 1, 1, 1, 2,
and 3 mL (I) from its stock solution (1 mg/mL) were
transferred into a series of 10 mL volumetric flasks. Aliquot
portions of 1, 2, 1, 1, and 2 mL (II) and 1, 1, 1.5, 1, and 1.5 mL
(III) from their stock solutions (1.5 and 10 mg/mL,
respectively) were added to the same flasks and diluted to
volume with methanol.
D1, and D3 Spectrophotometric
Accurate aliquots equivalent to 20–280 mg (I) from its
working solution (40 mg/mL) and aliquots equivalent to
40–400 mg (III) from its working solution (400 mg/mL) were

Table 2. Determination of drotaverine hydrochloride (I) by the D1 method and paracetamol (II) by the D3 method in
different laboratory-prepared mixtures

Claimed taken, mg/mL (I) D1 at l = 331 nm (III) D3 at l = 315 nm

Mixture No. (I) (II) (III) Founda, mg/mL Found, % Founda, mg/mL Found, %

1 4 6 40 4.061 101.53 39.308 98.27

2 8 24 12 7.915 98.94 11.870 98.92
3 12 18 20 12.218 101.82 20.116 100.58
4 16 12 28 15.688 98.05 27.482 98.15
5 20 6 4 19.940 99.70 3.951 98.78
b
Mean ± SD 100.01 ± 1.63 98.94 ± 0.97

a
Average of 3 determinations.
b
SD = Standard deviation.
394 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007

Table 3. Application of the standard addition technique to the analysis of drotaverine hydrochloride (I) and
paracetamol (III) in Petro tablets (Batch No. 01101157) by the D1 and D3 methods

Petro tablets Founda, % Pure added, mg/mL Founda, mg/mL Recovery, %

(I) 99.12 ± 0.63 4 3.968 99.20

8 7.900 98.75
12 12.014 100.12
16 15.869 99.18
Mean ± SD 99.31 ± 0.58
(III) 99.82 ± 0.26 4 3.954 98.85
8 7.974 99.68
12 12.014 100.12
16 16.056 100.35
Mean ± SD 99.75 ± 0.66

a
Average of 4 determinations.

transferred into 2 sets of 10 mL volumetric flasks and diluted respect to a blank of methanol. The composition of the
to the line with methanol to give final concentration ranges of samples was randomly designed in order to obtain
2–28 and 4–40 mg/mL for (I) and (III), respectively. The D1 noncorrelated concentration profiles and this calibration
and D3 spectrum was recorded for each solution. Two design prepared to obey Beer’s law. Several multivariate
calibration curves were constructed relating the peak calibration models were constructed using the data obtained.
amplitude of (D1) at 331 nm for (I) and the peak amplitude of
TLC-Densitometry Linearity
(D3) at 315 nm for (III), and their corresponding
concentrations and the regression equations were then Accurate aliquots equivalent to 10–100 mg (I), 15–150 mg
computed. (II), and 25–150 mg (III) from their stock solutions (1, 1.5, and
10 mg/mL) for (I), (II), and (III), respectively, were
DD1 Method Linearity
transferred into 3 separate series of 10 mL volumetric flasks
(a) For (I).—The absorption spectra of (I) in the range of and diluted to volume with methanol. Ten mL of each series of
2–40 and 60 mg/mL (II) were divided by absorption spectrum (I), (II), and (III) was applied separately to a TLC plate (20 ´
of (III) (40 mg/mL), and the obtained ratio spectra were 20 cm) using a Hamilton microsyringe (25 mL) with a space of
differentiated with respect to wavelength. The peak 1 cm between each spot and 2 cm up from the bottom edge of
amplitudes at 312.4 nm of (I) were measured. The calibration the plate. The chromatographic chamber was equilibrated
curves representing the relation between peak amplitude and
the corresponding concentrations were constructed, and the
regression equations were derived.
(b) For (II) and (III).—The absorption spectra of (II) in
the range of 6–48 mg/mL and (III) in the range of 4–40 mg/mL
were divided by absorption spectrum of (I) (4 mg/mL), and the
obtained ratio spectra were differentiated with respect to
wavelength. The peak amplitudes at 286.4 and 304 nm (II)
and (III), respectively, were measured. The calibration curves
representing the relation between peak amplitude and the
corresponding concentrations were constructed, and the
regression equations were derived.
Multivariate Spectral Analysis Calibration Curves
Calibration solutions (training set) of (I), (II), and (III)
were prepared by diluting different volumes of their working
solutions into 10 mL volumetric flasks. The volume was Figure 4. First-derivative ratio spectra of 2–40 mg/mL
diluted to the line with methanol to reach the concentrations drotaverine hydrochloride (I) (——) and 60 mg/mL
listed in Table 1. The absorbencies of these mixtures were caffeine (II) (- - -) using 40 mg/mL paracetamol (III) as a
measured between 200–400 nm at 0.2 nm intervals with divisor.
 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007 395
Figure 5. First-derivative ratio spectra of 6–48 mg/mL
caffeine (II) (——) and 8 mg/mL paracetamol (III) (- - -)
using 4 mg/mL drotaverine hydrochloride (I) as a divisor.
with the mobile phase, ethyl acetate–chloroform–methanol
(16 + 3 + 1, v/v/v), for at least 30 min. The plate was
developed at room temperature by ascending migration of the
mobile phase, removed from the chamber and dried, and the
spots were visualized under a UV lamp at 254 nm. The spots
were scanned with the spectrodensitometer at 281, 272, and
248 nm for (I), (II), and (III), respectively. Calibration curves
relating the peak areas and its corresponding concentrations
were plotted, and the corresponding regression equations
were calculated.
Assay of Laboratory-Prepared Mixtures
The values of D1 at 331 nm for (I), D3 at 315 nm for (III),
and DD1 at 312.4 nm for (I) or at 286.4 and 304 nm for (II) and
(III), respectively, were recorded for different laboratory-
prepared mixtures containing different ratios of (I), (II), and
(III).
The developed models of multivariate spectral analysis
were used to analyze the spectra of different
laboratory-prepared mixtures. Ten mL of each laboratory-
prepared mixture was analyzed by TLC-densitometry as
described above. The concentrations of (I), (II), and (III) were
calculated from the corresponding regression equations for
each method.
Application to a Pharmaceutical Preparation (Petro
Tablets)
An accurately weighed amount of mixed and pulverized
powder from 10 tablets, equivalent to 100 mg (I), 150 mg (II),
and 1000 mg (III), was extracted well by sonication with
methanol (3 ´ 25 mL) and filtered quantitatively into a 100 mL
volumetric flask, and then the volume was diluted to the line
with the same solvent [1 mg/mL (I), 1.5 mg/mL (II), and
10 mg/mL (III)]. For the TLC-densitometric method, the
procedures were completed on the solution as described under
TLC-Densitometry Linearity. For derivative
spectrophotometric (D1 and D3), first derivative of the ratio
spectra, and multivariate spectral analysis techniques, 4 mL
extract was transferred into a 100 mL volumetric flask and
diluted to the line with methanol to obtain a final
concentration of 40 mg/mL (I), 60 mg/mL (II), and 400 mg/mL
(III); the procedure was performed as under D1 and D3
Spectrophotometric Linearity Method, and the developed
multivariate models were used to calculate the concentrations
of (I), (II), and (III).
Results and Discussion
New methods for simultaneous determination of 2 or more
compounds in the same sample, without previous chemical
separation, are always of interest. This work was devoted to
the analysis of (I) in the presence of (II) and (III), which are
available together in the form of tablets. This was achieved
using first and third derivative spectrophotometric, first
derivative of the ratio spectra, multivariate spectral analysis,
and TLC-densitometric methods.
D1 and D3 Spectrophotometric Methods
Derivative spectrophotometry is a good technique capable
of enhancing the resolution of overlapped bands (13). It can
be applied for the determination of drug mixtures by selecting
a wavelength where contribution of one compound is zero
while the other to be determined has a reasonable value.
Zero-order absorption spectra of (I), (II), and (III) showed
severe spectral overlapping (Figure 1). The first-derivative
(D1) technique was used to resolve spectral overlapping of
this mixture (Figure 2); (II) and (III) have no interference with
(I) as shown at 331 nm also. The D3 technique was applied
(Figure 3). A zero crossing point for (II) and (I) with (III) was
shown at 315 nm.
Linear correlations were obtained between peak
amplitudes and corresponding concentrations in the range of
2–28 and 4–40 mg/mL for (I) and (III), respectively. The linear
regression equations were found to be:
Y = 0.0032X + 0.0197 r = 0.9997 for (I)
Y = 0.0008X + 0.0005 r = 0.9997 for (III)
Figure 6. First-derivative ratio spectra of 4–40 mg/mL
paracetamol (III) (——) and 6 mg/mL caffeine (II) (- - -)
using 4 mg/mL drotaverine hydrochloride (I) as a divisor.
396 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007

99.73 ± 1.09
where Y is peak amplitude [D1 at 331 nm for (I) and D3 at

Found, %
315 nm for (III)], X is the concentration in mg/mL, and r is the

98.71
99.65
98.62
100.82
100.85
correlation coefficient.
The proposed method is valid for determination of (I) and
(III)

(III) in different laboratory-prepared mixtures with mean

Found, mg/mL

39.484
percentage recoveries of 100.01 ± 1.63 and 98.94 ± 0.94,
11.958
19.724
28.230
4.034
respectively, as shown in Table 2.
The suggested method was applied to assay (I) and (III) in
Table 4. Determination of drotaverine hydrochloride (I), caffeine (II), and paracetamol (III) in laboratory-prepared mixtures by the DD1 method

their pharmaceutical dosage form, and its validity was further

assessed by applying the standard addition technique
(Table 3).
DD1 Spectrophotometric Method
99.28 ± 0.61
Found, %

This method takes into consideration the theory of

99.35
100.31
98.91
99.07
98.78

derivative ratio spectrophotometry (22) to solve the problem

of overlapping absorption spectra of (I), (II), and (III). The
(II)

method is based on the use of first derivatives of the ratio

mg/mL

spectra. The absorption spectrum of the mixture is obtained

5.961
24.074
17.804
11.888
5.927

and divided (amplitudes at each wavelength) by the

Founda,

absorption spectrum of a standard solution of one of the

components, and the first derivative of the ratio spectrum is
obtained. In this method, careful choice of concentration of
the divisor and the wavelength is of great importance, as it
affects both sensitivity and selectivity.
(I) can be assayed in the presence of (II) and (III) by
dividing the absorption spectra of different concentrations of
100.73 ± 0.99

(I) in the range of 2–40 mg/mL by the absorption spectrum of

Found, %

99.08
100.91
101.65
101.28
100.74

40 mg/mL (III); the (II) absorption spectrum was divided by

the absorption spectrum of 40 mg/mL (III). The obtained ratio
spectra were differentiated with respect to wavelength
(I)

mg/mL

(Figure 4). By applying the DD1 technique, a zero crossing

3.963
8.073
12.198
16.205
20.148

point of (II) with (I) at 312.4 nm was found. A linear

Founda,

correlation was obtained between peak amplitude and the

corresponding concentration in the range of 2–40 mg/mL for
(I). The linear regression equations were found to be:

Y = 0.2247X + 0.4411 r = 0.9995 for (I)

where Y is the peak amplitude of the DD1 value at 312.4 nm,

X is the concentration in mg/mL, and r is the correlation
coefficient.
(III)

40
12
20
28
4

(II) and (III) could be assayed in the presence of (I) by

dividing the absorption spectra of different concentrations of
Claimed taken, mg/spot

(II) and (III) in the range of 6–48 and 4–40 mg/mL for (II) and
(III), respectively, by the absorption spectrum of 4 mg/mL (I).
The obtained ratio spectra were differentiated with respect to
(II)

6
24
18
12
6

wavelength. By applying the DD1 technique, a zero crossing

point of (III) with (II) was shown at 286.4 nm (Figure 5). Also,
Average of 3 determinations.

a zero crossing point of (II) with (III) was shown at 304 nm

(Figure 6). A linear correlation was obtained between peak
4
8
12
16
20
(I)

amplitude and the corresponding concentration in the range of

6–48 and 4–40 mg/mL for (II) and (III), respectively. The
linear regression equations were found to be:
Mean ± SD
Mixture No.

Y = 0.1935X + 0.6015 r = 0.9995 for (II)

Y = 0.0886X + 0.0402 r = 0.9992 for (III)

1
2
3
4
5

a
 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007 397

Table 5. Application of the standard addition technique to the analysis of drotaverine hydrochloride (I), caffeine (II),
and paracetamol (III) in Petro tablets (Batch No. 01101157) by the DD1 method

Petro tablets Founda, % Pure added, mg/mL Founda, mg/mL Recovery, %

(I) 99.60 ± 0.57 4 3.983 99.58

8 8.042 100.53
12 12.080 100.67
16 16.179 101.12
Mean ± SD 100.48 ± 0.65
(II) 101.05 ± 0.52 12 12.103 100.86
18 18.191 101.06
24 24.228 100.95
30 30.387 101.29
Mean ± SD 101.04 ± 0.19
(III) 98.87 ± 0.73 4 3.956 98.90
8 7.924 99.05
12 11.899 99.16
16 15.805 98.78
Mean ± SD 98.97 ± 0.17

a
Average of 4 determinations.

where Y is the peak amplitude of the DD1 value at 286.4 and

304 nm for (II) and (III), respectively, X is the concentration in
mg/mL, and r is the correlation coefficient.
Results obtained (Table 4) showed that the proposed
method is valid and applicable for simultaneous determination
of (I), (II), and (III) in different laboratory-prepared mixtures
with mean percentage recoveries of 100.73 ± 0.99, 99.28 ±
0.61, and 99.73 ± 1.09, respectively.
The validity of the DD1 spectrophotometric method was
further assessed by applying a standard addition method for
the analysis of Petro tablets (Table 5).

Multivariate Spectral Analysis Methods

Figure 7. RMSEC plot of a calibration set prediction
using cross-validation (PCR model).
Figure 8. RMSEC plot of a calibration set prediction
using cross-validation (PLS model).
The classical least-squares (CLS), partial least-squares
(PLS), and principal component regression (PCR) techniques
are typical full-spectrum methods in which the data are fit to
many data points. In these techniques, a calibration model is
suggested, validated, and then used for the prediction of
unknown samples.
Haaland and Thomas (17) made a comparison of the
different multivariate calibration methods for quantitative
spectral analysis. They concluded that it is difficult to
generalize about the superiority of one method over another
because the relative performance of the method is often
dependent on the particular data set being analyzed.
The first step in the simultaneous determination of the
3 analytes by multivariate calibration methods involves
constructing the calibration matrix for a ternary mixture. The
calibration set was obtained by using the absorption spectra of
a set of 15 mixtures of (I), (II), and (III) with different ratios of
each component. Initial developed models were found to have
Table 6. Results obtained for the analysis of ternary mixtures (validation set) of drotaverine hydrochloride (I), caffeine (II), and paracetamol (III) by the multivariate
spectral analysis methods

Concentration, mg/mL CLS recovery, % PCR recovery, % PLS recovery, %

Sample No. (I) (II) (III) (I) (II) (III) (I) (II) (III) (I) (II) (III)

1 2 3 20 100.35 98.54 100.31 99.49 99.95 101.15 99.99 98.96 101.64

2 2 4.5 16 100.58 99.25 100.12 97.99 98.04 101.62 97.94 101.55 101.00
3 2 6 12 99.87 99.68 99.98 98.39 99.37 100.84 98.39 99.37 100.84
4 4 1.5 24 98.62 98.86 101.57 97.82 101.54 99.74 98.08 101.52 99.75
5 4 3 20 99.86 99.97 99.21 98.66 99.42 100.23 98.65 99.43 100.23
6 4 4.5 16 96.99 101.25 101.34 98.45 97.76 101.68 98.44 99.97 101.67
7 4 6 12 97.61 97.49 99.88 97.49 99.05 101.27 100.65 99.05 101.26
8 6 1.5 24 100.00 101.60 101.67 101.71 101.43 99.13 100.43 101.43 100.91
9 6 3 20 101.85 100.42 101.85 100.33 100.42 99.43 99.48 100.65 100.15
10 6 4.5 16 101.90 98.60 100.16 99.79 99.10 100.67 100.50 99.11 100.67
11 8 1.5 24 101.40 100.86 101.99 100.32 101.66 99.64 100.46 99.30 99.65
12 8 3 20 101.43 99.75 100.74 100.00 99.34 100.26 101.28 99.35 100.26
Mean ± SD 100.04 ± 1.60 99.69 ± 1.22 100.73 ± 0.92 99.20 ± 1.27 99.76 ± 1.29 100.47 ± 0.87 99.52 ± 1.17 99.89 ± 1.03 100.67 ± 0.68
398 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007
Table 7. RMSEC and statistical parameter values of drotaverine hydrochloride (I), caffeine (II), and paracetamol (III) prediction using multivariate spectral analysis
methods

CLS PCR PLS

Item (I) (II) (III) (I) (II) (III) (I) (II) (III)

RMSEC 0.0797 0.0481 0.2530 0.0569 0.0472 0.1606 0.0504 0.0345 0.1617
2
Q 0.9986 0.9989 0.9965 0.9993 0.9990 0.9984 0.9994 0.9995 0.9984
Intercept –0.0960 0.0417 –0.5386 –0.0811 0.0494 0.5223 –0.0805 0.0242 0.3042
Slope 1.0247 0.9814 1.0376 1.0129 0.9792 0.9754 1.0159 0.9912 0.9896
Correlation coefficient (r) 0.9992 0.9996 0.9993 0.9995 0.9997 0.9993 0.9997 0.9996 0.9993
a
SE of intercept 0.0458 0.0242 0.1684 0.0360 0.0200 0.1588 0.0297 0.0243 0.1631
SE of slope 0.0090 0.0063 0.0088 0.0071 0.0052 0.0083 0.0058 0.0064 0.0085

a
SE = Standard error.

Table 8. Quantitative determination of drotaverine hydrochloride (I), caffeine (II), and paracetamol (III) in Petro tablets (Batch No. 01101157) by the proposed
multivariate spectral analysis methods

CLS found, %a ± SD PCR founda, % ± SD PLS founda, % ± SD

(I) (II) (III) (I) (II) (III) (I) (II) (III)

99.96 ± 0.71 99.87 ± 0.52 100.93 ± 0.68 99.20 ± 0.73 99.75 ± 0.61 100.47 ± 0.54 99.52 ± 0.66 99.98 ± 0.43 100.67 ± 0.62

a
Average of 3 determinations.
METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007 399
Table 9. Results of the standard addition technique for the determination of drotaverine hydrochloride (I), caffeine (II), and paracetamol (III) in Petro tablets by the
proposed multivariate spectral analysis methods (claimed taken at 3, 3, and 10 mg/mL for (I), (II), and (III), respectively)

Claimed taken, mg/mL Pure added, mg/mL Recovery of pure added, %

CLS PCR PLS

(I) (II) (III) (I) (II) (III) (I) (II) (III) (I) (II) (III) (I) (II) (III)

3 3 10 2 1.5 8 98.91 99.21 100.34 99.07 99.76 99.89 99.82 99.54 100.24
3 3 10 3 3 10 99.26 100.35 100.42 99.64 99.52 100.37 98.65 100.47 99.71
3 3 10 4 4.5 12 98.14 99.38 100.66 100.09 99.10 101.23 99.32 98.69 100.43
Mean ± SD 98.77 ± 0.57 99.65 ± 0.62 100.47 ± 0.17 99.60 ± 0.51 99.46 ± 0.33 100.50 ± 0.68 99.26 ± 0.59 99.57 ± 0.89 100.13 ± 0.37

Table 10. Determination of drotaverine hydrochloride (I), caffeine (II), and paracetamol (III) in laboratory-prepared mixtures by the TLC-densitometry method

Claimed taken, mg/spot (I) (II) (III)

400 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007

Mixture No. (I) (II) (III) Founda, mg/spot Found, % Founda, mg/spot Found, % Founda, mg/spot Found, %

1 1 1.5 10 0.992 99.20 1.499 99.93 10.101 101.01

2 1 3 10 0.993 99.30 3.006 100.20 10.091 100.91
3 1 1.5 15 1.002 100.20 1.498 99.87 15.197 101.31
4 2 1.5 10 1.997 99.85 1.505 100.33 9.987 99.87
5 3 3 15 2.994 99.80 3.006 100.20 15.086 100.57
Mean ± SD 99.67 ± 0.41 100.11 ± 0.20 100.73 ± 0.55

a
Average of 3 determinations.
 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007 401

high spectral residuals in the rejoin from 200 to 220 nm, so
this region was rejected.
For CLS, the prediction matrix was calculated using the
algorithm described by Kramer (23). PCR and PLS are based
on factor analysis. Calibration is then performed using a
subset of these factors instead of the original variables. Both
PCR and PLS have the signal averaging advantage of
full-spectrum techniques such as CLS, while they do not need
the prior knowledge of the characteristics of all interfering
substances (24, 25).
Another advantage of PCR and PLS is the transformation
of the numerous and probably correlated original variables
into a smaller number of orthogonal variables (latent vectors).
The first latent vectors contain the information carried in the
manifest variables, where the last ones represent noise, which
can be filtered away and will not be considered in the
modeling (25).
PCR is simply principal component analysis followed by a
regression step. PLS is related to PCR in that spectral
decomposition is also performed, but in a different way. In
PCR, the spectra are decomposed on the basis of the
maximum variance between spectral data and without using
information about concentration. PLS differs from PCR in
that it uses both spectral and concentration data in the
modeling (19).
Selection of the optimum number of factors for PLS and
PCR techniques is a very important step before constructing
the models because, if the number of factors retained is more
than required, more noise will be added to the data. On the
other hand, if the number retained is too small, meaningful
data that could be necessary for the calibration may be
discarded.
To select the number of factors for the PLS and PCR
methods, a cross-validation method was used in which one
sample is left out in turn (23). Given a set of 15 calibration
samples, the PLS and PCR calibrations were performed on
14 samples; then the concentration of the sample left out
during calibration was predicted. This process was repeated
15 times, until each calibration sample had been left out once.
The predicted concentrations were then compared with the
known concentrations using RMSEC (root mean square error
of calibration). The optimum number of factors was selected
according to the method of Haaland and Thomas (20, 26).
The maximum number of factors used to calculate the
optimum RMSEC was selected to be 8 (half the number of
samples + 1; 18).
Figures 7 and 8 show the RMSEC plot of the
cross-validation results of the training set as a function of the
number of factors used to construct the calibration. The
optimum number of factors used for building PLS and PCR
models was found to be 3.
To validate the prediction ability of the suggested models,
they were used to predict the concentration of (I), (II), and
(III) in laboratory-prepared mixtures containing different
ratios of the components and satisfactory results were
obtained (Table 6). Another diagnostic test was performed by
plotting the concentration residuals against the predicated
concentrations. The residuals appear randomly distributed
around zero, indicating adequate models.
The RMSEC was used as a diagnostic test for examining
the error in the predicted concentrations. Another parameter,

Table 11. Application of the standard addition technique to the analysis of drotaverine hydrochloride (I), caffeine (II),
and paracetamol (III) in Petro tablets (Batch No. 01101157) by the TLC-densitometry method

Petro tablets Founda, % Pure added, mg/spot Founda, mg/spot Recovery, %

(I) 98.12 ± 0.48 1 0.982 98.20

2 1.988 99.40
3 3.003 100.10
4 3.989 99.73
Mean ± SD 99.63 ± 0.82
(II) 99.74 ± 0.61 1.5 1.497 99.81
3 2.942 98.07
4.5 4.488 99.73
6 6.007 100.12
Mean ± SD 99.43 ± 0.92
(III) 100.02 ± 0.38 2.5 2.513 100.52
5 5.037 100.74
7.5 7.498 99.97
10 9.821 98.21
Mean ± SD 99.86 ± 1.15

a
Average of 4 determinations.
402 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007

99.72 ± 0.69 99.63 ± 0.58 100.31 ± 0.47

0.221
(III)

—
—
Q2, was calculated that determined the variation in the

5
Manufacturer’s method
prediction samples. Table 7 shows the RMSEC and Q2 values
that were obtained; they are satisfactory and indicate good
Table 12. Statistical comparison of the results obtained by applying the proposed D1, D3, DD1, and TLC-densitometry methods and the LC instrument

predicative abilities of the developed models. Plotting the

0.336
(II)

—
—
known concentrations against the predicted concentrations

5
was also performed to evaluate the predicative abilities of the
models.
0.476 The multivariate spectral analysis methods were applied
—
—
(I)

successfully to the analysis of (I), (II), and (III) in Petro tablets

5

(Table 8). The validity of the suggested method was further

assessed by applying the standard addition technique
(Table 9).
99.70 ± 1.42 99.65 ± 1.40 99.78 ± 1.14

TLC-Densitometry Method
0.966
5.882
1.300
(III)

TLC-densitometry overcomes the problem of overlapping

TLC-densitometry method

absorption spectra of a mixture of drugs by separating the

components on TLC plates and determining each ingredient
manufacturer’s method for the analysis of pure drotaverine hydrochloride (I), caffeine (II), and paracetamol (III)

0.030
5.833
1.960

by scanning the corresponding chromatogram. The TLC-UV

(II)

densitometric method has the advantage of simultaneously

6

determining the active ingredients in multicomponent dosage

forms (27).
In this work, different solvent systems were tried, and
0.029
4.235
2.016
(I)

complete separation of the mixture was achieved using ethyl

6

acetate–chloroform–methanol (16 + 3 + 1, v/v/v). The

proposed technique is based on the difference of Rf values of
(I) (Rf = 0.26), (II) (0.51), and (III) (0.68).
The separated spots of (I), (II), and (III) were scanned on
100.13 ± 1.10

the same plate at 281, 272, and 248 nm, respectively. A linear
0.339
5.475
1.210
(III)

relation was obtained between peak area and concentration in

Proposed methods

the range of 1–10, 1.5–15, and 2.5–15 mg/spot for (I), (II) and
6

(III), respectively. The linear regression equations were found

Values in parentheses correspond to the theoretical values of t and F at P = 0.05.

to be:
DD1 method

100.09 ± 0.93

0.957
2.574
0.865
(II)

Y = 2.0652X + 0.0867 r = 0.9997 for (I)

6

Y = 3.03X + 4.5418 r = 0.9999 for (II)

99.20 ± 0.70

1.235
1.029
0.490

Y = 0.7529X + 1.626 r = 0.9997 for (III)

(I)

where Y is the area under the peak, X is the concentration in

mg/spot, and r is the correlation coefficient.
Results in Table 10 show that the method is valid and
99.92 ± 0.99

applicable for simultaneous determination of (I), (II), and (III)

(II) by D3

0.803
4.434
0.980

in different laboratory-prepared mixtures with mean

Derivative method

percentage recoveries of 99.67 ± 0.41, 100.11 ± 0.20, and

6

100.73 ± 0.55, respectively.

The proposed method has been applied to assay (I), (II),
99.60 ± 0.91

and (III) in Petro tablets. The validity of the suggested

(I) by D1

0.242
1.739
0.828

procedure was further assessed by applying the standard

n = Number of trials.
6

addition technique (Table 11).

A statistical comparison of the results obtained by the
proposed methods and the LC instrument manufacturer’s
Mean ± SD

b
Variance

method for pure drugs is shown in Tables 12 and 13. The

b
t (2.262)
Variable

F (6.26)

calculated values of t and F are less than the tabulated values,

a

which reveals that there is no significant difference with

n

b
Table 13. Statistical analysis of the results obtained by applying the proposed multivariate spectral analysis and the LC instrument manufacturer’s methods for
the analysis of pure drotaverine hydrochloride (I), caffeine (II), and paracetamol (III)

Proposed methods Manufacturer’s method

CLS PCR PLS

Variable (I) (II) (III) (I) (II) (III) (I) (II) (III) (I) (II) (III)

Mean ± SD 100.04 ± 1.60 99.69 ± 1.22 100.73 ± 0.92 99.20 ± 1.27 99.76 ± 1.29 100.47 ± 0.87 99.52 ± 1.17 99.89 ± 1.03 100.67 ± 0.68 99.72 ± 0.69 99.63 ± 0.58 100.31 ± 0.47
n 12 12 12 12 12 12 12 12 12 5 5 5
Variance 2.560 1.488 0.846 1.613 1.664 0.757 1.369 1.061 0.462 0.476 0.336 0.221
a
t (2.131) 0.405 0.099 0.911 0.812 0.203 0.365 0.336 0.671 1.016 — — —
a
F (5.94) 5.378 4.429 3.828 3.389 4.952 3.425 2.876 3.158 2.090 — — —

a
The values in parentheses correspond to the theoretical values of t and F at P = 0.05.

Table 14. Validation of the results obtained by applying the proposed D1, D3, DD1, and TLC-densitometry methods for the analysis of drotaverine hydrochloride (I),
caffeine (II), and paracetamol (III)

Derivative method DD1 method TLC-densitometry method

Parameter (I) by D1 (III) by D3 (I) (II) (III) (I) (II) (III)

a
LOD 0.54 mg/mL 0.68 mg/mL 0.92 mg/mL 1.06 mg/mL 1.10 mg/mL 0.20 mg/spot 0.21 mg/spot 0.27 mg/spot
LOQb 1.81 mg/mL 2.25 mg/mL 1.57 mg/mL 3.54 mg/mL 3.68 mg/mL 0.68 mg/spot 0.71 mg/spot 0.88 mg/spot
Range 2–28 mg/mL 4–40 mg/mL 2–40 mg/mL 6–48 mg/mL 4–40 mg/mL 1–10 mg/spot 1.5–15 mg/spot 2.5–15 mg/spot
Slope 0.0032 0.0008 0.2247 0.1935 0.0886 2.0652 3.0300 0.7529
Intercept 0.0197 0.0005 0.4411 0.6015 0.0402 0.0867 4.5418 1.6260
Mean ± SD 99.60 ± 0.91 99.92 ± 0.99 99.20 ± 0.70 100.09 ± 0.93 100.13 ± 1.10 99.70 ± 1.42 99.65 ± 1.40 99.78 ± 1.14
Correlation coefficient (r) 0.9997 0.9997 0.9995 0.9995 0.9992 0.9997 0.9999 0.9997
RSD, %c 0.72–1.41 0.57–0.82 0.81–0.75 1.07–0.91 0.96–1.03 0.97–1.64 1.47–1.36 0.88–1.06
RSD, %d 0.89–1.32 0.42–0.99 0.71–0.90 1.02–1.17 1.27–1.14 1.24–1.57 1.18–1.52 1.35–0.92
a
LOD (limit of detection) = 3´ SD of calibration/slope.
b
LOQ (limit of quantitation) = 10´ SD of the calibration/slope.
c,d
Intraday and interday (n = 4) relative standard deviations of samples of concentrations [12 and 24 mg/mL (I), (II), and (III), respectively, for D1, D3, and DD1 methods and 3.8 mg/spot (I), (II), and (III)
respectively, for the TLC-densitometry method].
METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007 403
404 METWALLY ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 90, NO. 2, 2007
respect to accuracy and precision between the proposed
methods and the manufacturer’s procedure.
The results of assay validation of the derivative, derivative
ratio, and TLC-densitometry methods show that the methods
are accurate, precise, specific, and rugged over the specified
ranges (Table 14).
Conclusions
The aim of this work was to develop simple methods for
the simultaneous determination of (I), (II), and (III). The
proposed derivative method failed in simultaneous
determination of the 3 drugs because only (I) and (III) were
determined. The derivative ratio, TLC-densitometry, and
chemometric methods could be applied for the simultaneous
determination of (I), (II), and (III) either in their pure powder
form or in their combined tablet preparation. The results
demonstrate the usefulness of the methods, which are simple,
safe, sensitive, precise, accurate, inexpensive, and
nonpolluting. Therefore, the proposed methods can be used in
routine and quality control analysis of (I), (II), and (III) in
pharmaceutical preparations containing them.
References
(1) O’ Neil, J.M. (2001) Merck Index, 13th Ed., Merck & Co.,
Rahway, NJ, pp 609–661
(2) Knaub, V.A., & Kartashov, V.A. (1989) Farmatsiya 38,
46–49
(3) Daabees, H.G. (2000) Anal. Lett. 33, 639–656
(4) Geher, J., & Szabo, E. (1988) J. Pharm. Biomed. Anal. 6,
757–764
(5) Fulop, M., Kaloy, K., & Toth, A. (1986) Magy. Kem. Foly.
92, 468–473
(6) Fulop, M., Toth, A., & Kaloy, K. (1986) Magy. Kem. Foly.
92, 474–479
(7) Issa, Y.M., Ibrahim, H., & Abu-Shawish, H.M. (2005)
Microchim. Acta 150, 47–54
(8) Girgis, E.H. (1993) J. Pharm. Sci. 82, 503–505
(9) Bolaji, O.O., Onyeji, C.O., Ogungbamila, F.O., Ogunbona,
F.A., & Ogunlana, E.O. (1993) J. Chromatogr. B 622, 93–97
(10) Mezei, J., Kuttel, S., Szentmiklosi, P., Marton, S., & Racz, I.
(1984) J. Pharm. Sci. 73, 1489–1491
(11) Lalla, J.K., Shah, M.U., Jain, M.B., & Sharma, A.H. (1993)
J. Pharm. Biomed. Anal. 11, 385–388
(12) Dyderski, S., Grzeskowiak, E., Drobnik, L., Szalek, E.,
Balcerkiewicz, M., & Dubai, V. (2004) Arzneim. Forsch. 54,
298–302
(13) O’Haver, T.C., & Green, G.L. (1976) Anal. Chem. 48,
312–318
(14) Geladi, P., & Kowalski, B.R. (1986) Anal. Chim. Acta 185,
1–17
(15) Salinas, F., Berzas Nevado, J.J., & Espinosa Mansilla, A.
(1989) Analyst 114, 1141–1145
(16) Espinosa Mansilla, A., Berzas Nevado, J.J., & Salinas, F.
(1992) J. AOAC Int. 75, 678–684
(17) Haaland, D.M., & Thomas, E.Y. (1990) Anal. Chem. 62,
1091–1099
(18) Espinosa Mansilla, A., Muñoz de la Peña, A., Salinas, F., &
Martinez-Galera, M. (1993) Anal. Chim. Acta 276, 141–149
(19) Durán-Merás, I., MuZoz de la PeZa, A., Espinosa Mansilla,
A., & Salinas, F. (1993) Analyst 118, 807–813
(20) Haaland, D.M., & Thomas, E.Y. (1988) Anal. Chem. 60,
1193–1198
(21) Wise, B.M., & Gallagher, N.B. (1998) PLS-Toolbox 2.0 for
use with MatlabTM, Eigenvector Research, Inc., Manson, WA
(22) Salinas, F., Berzas Nevado, J.J., & Espinosa Mansilla, A.
(1990) Talanta 37, 347–351
(23) Kramer, R. (1998) Chemometric Techniques for Quantitative
Analysis, Marcel Dekker Inc., New York, NY
(24) Haaland, D.M., & Easterling, R.G. (1980) Appl. Spectrosc.
34, 539–544
(25) Jouran-Rimbaud, D., Walczak, B., Massart, D.L., Last, I.R.,
& Prebble, K.A. (1995) Anal. Chim. Acta 304, 285–295
(26) Haaland, D.M., & Thomas, E.Y. (1988) Anal. Chem. 60,
1202–1208
(27) Bebawy, L.I., & EL-Kousy, N.M. (1999) J. Pharm. Biomed.
Anal. 20, 663–670