Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801

Contents lists available at ScienceDirect

Spectrochimica Acta Part A: Molecular and Biomolecular

Spectroscopy
journal homepage: www.elsevier.com/locate/saa

Chemometric assisted UV spectrophotometric and RP-HPLC methods for

simultaneous determination of paracetamol, diphenhydramine, caffeine
and phenylephrine in tablet dosage form
Keerthisikha Palur a,⁎, Sreenivasa Charan Archakam a, Bharathi Koganti b
a
Department of Pharmaceutical Analysis, Sri Padmavathi School of Pharmacy, Mohan Gardens, Tiruchanoor, Andhra Pradesh 517503, India
b
Institute of Pharmaceutical Technology, Sri Padmavati Mahila Visvavidyalayam, Tirupati, Andhra Pradesh 517502, India

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

Article history: In this paper, Chemometric assisted UV spectrophotometric and RP-HPLC methods were developed and com-
Received 30 April 2020 pared for simultaneous determination of Paracetamol, Diphenhydramine Hydrochloride, Caffeine and Phenyl-
Received in revised form 16 July 2020 ephrine Hydrochloride in tablet dosage form. UV-Spectrophotometric analysis was carried out by applying two
Accepted 3 August 2020
chemometric models namely, Principal Component Regression method (PCR) and Partial Least Squares Regres-
Available online 8 August 2020
sion method (PLSR). Chromatographic method was developed and optimized by applying Response surface
Keywords:
methodology -Central Composite Design (CCD). These methods were considered first for the quantification of
Chemometrics the drugs present in the selected formulation. PCR and PLSR models were successfully validated and applied
Principal component regression for resolving the complex UV-spectra in the wavelength range of 240–320 nm with a data interval of 1 nm. In
Partial least square regression RP-HPLC method, the identified critical factors were methanol content (45–55% v/v) and flow rate
Central composite design (0.75–0.85 mL/min) and the selected responses were retention time (Rt4) of fourth eluted component and res-
Derringer's desirability olution (RS1,2) between first and second eluted components. Derringer's desirability function was used for the
optimization of the chromatographic method conditions which comprised of mobile phase consisting of
methanol‑potassium dihydrogen orthophosphate buffer (pH 3; 10 mM) (50: 50, v/v) and at a flow rate of
0.81 mL/min with a detection wavelength of 220 nm. One-way ANOVA in 95% confidence interval revealed
that there were no significant differences among the developed methods.
© 2020 Elsevier B.V. All rights reserved.

1. Introduction [4–6]. In the recent years, a prodigious growth is seen in the simulta-
neous determination of drugs using chemometric approach [7–9].
The increase in the usage of multicomponent formulations in the The interference of analytes among each other, complex overlapping
health care system has constituted a substantial challenge on pharma- of spectra, lack of resolving power and inability in using the entire spec-
ceutical analysts to develop simple, accurate, robust and reliable analyt- tral data forged the failure of conventional spectrophotometric analysis.
ical methods for quantification of the drugs present in the formulations. These problems can be vanquished by using chemometric assisted spec-
Multicomponent analysis using routine analytical methods is very cum- trophotometric methods [10,11]. These techniques extract the data
bersome [1–3]. Chemometrics complements various analytical methods from the whole spectra for the simultaneous estimation of analytes
in making the multicomponent analysis more feasible to the analyst and provide fast analysis with good accuracy and precision without tire-
some sample preparation [12,13]. One factor at a time (OFAT) approach
is generally seen in high performance liquid chromatographic (HPLC)
Abbreviations: %, percentage; μ, microns; μg, micrograms; ANOVA, analysis of methods, which has many limitations like time consumption, require-
variance; AR, analytical reagent; CCD, Central Composite Design; CF, caffeine; CV, ment of many trials, improper projection of optimal conditions, no con-
coefficient of variation; DP, diphenhydramine hydrochloride; HPLC, high performance
sideration about interactions among factors and needs further
liquid chromatography; LV, latent variables; min, minutes; mL, milliliters; mm,
millimeters; mM, millimoles; nm, nanometers; OFAT, one factor at a time; PC, principal improvement of the method in the process of drug development [14].
components; PCR, principal component regression; PE, phenylephrine hydrochloride; These limitations can be overcome by the application of chemometrics
PLSR, partial least squares regression; PRESS, Predicted Residual Error of Sum of Squares; to the HPLC method, which is a better choice than OFAT in screening
PT, paracetamol; R2, regression coefficient; RMSEC, root mean square error of and optimization and in the enhancement of robustness [15–17].
calibration; RMSECV, root mean square error of cross validation; RP, Reverse Phase; RSD,
relative standard deviation; UV, ultraviolet.
The combination of Paracetamol (PT), Diphenhydramine Hydro-
⁎ Corresponding author. chloride (DP), Caffeine (CF) and Phenylephrine Hydrochloride (PE) is
E-mail address: [email protected] (K. Palur). available in different formulations in the market which are prescribed

https://doi.org/10.1016/j.saa.2020.118801
1386-1425/© 2020 Elsevier B.V. All rights reserved.
2 K. Palur et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801

for the effective treatment of fever associated with cold, nasal conges- solution, further dilutions were made to get a concentration of
tion, cough and in reducing migraine, sinus and allergies. Paracetamol 10 μg/mL of PT, 100 μg/mL of PE, 80 μg/mL of DP and 12 μg/mL of CF in
is an analgesic and antipyretic, which is used individually or with differ- order to make it suitable for UV spectrophotometric analysis. The con-
ent drugs in various dosage forms. Diphenhydramine Hydrochloride is centrations of the drugs were selected in their linearity ranges.
employed for the treatment of various allergies. Caffeine is used as
CNS stimulant and Phenylephrine Hydrochloride is used as a sympatho- 2.3.3. PCR and PLSR methods
mimetic agent. Vicks Action 500 advanced tablet containing 500 mg of A total of seventeen laboratory prepared mixtures of PT, DP, CF and
PT, 25 mg of DP, 5 mg of PE and 30 mg of CF was used in this study. PE in different ratios of each component were designed by using multi-
Extensive literature review divulged that there are no reported ana- level multifactor experimental design. The mixtures were prepared
lytical works on this formulation. This is the first analytical study for the from standard stock solutions. Out of these mixtures, ten were framed
simultaneous determination of PT, DP, CF and PE using chemometric into a training set and the remaining seven mixtures were used as test
assisted UV and RP-HPLC methods. In the present work, most widely set as shown in Table 1. The UV absorption spectra of all the mixtures
used Chemometric assisted UV spectrophotometric methods like partial were recorded over the wavelength range of 240–320 nm with 1 nm
least squares regression (PLSR) and principal component regression data interval. PCR and PLSR computational analysis was carried out
(PCR) were applied to determine the drugs in the selected quaternary using Unscrambler X software. Statistical analysis was performed for
formulation. RP-HPLC method was also developed and validated using the developed models and then parameters like Regression coefficient
central composite design (CCD). CCD, an experimental design which is (R2), Root mean square error of calibration (RMSEC), Root mean square
a subset of response surface methodology was used to study the interac- error of cross validation (RMSECV) were calculated.
tion effects of factors on the responses. Derringer's desirability function
which has the advantage of user flexibility was used to optimize the se- 2.4. Chemometric assisted RP-HPLC method
lected responses [18]. The results obtained from both Ultraviolet (UV)
and Reverse Phase (RP) HPLC methods were compared. 2.4.1. Chromatographic conditions
For the selection of initial chromatographic conditions, the physico-
2. Experimental chemical properties like dissociation constants, solubility, UV absorp-
tion characteristics and the chromatographic parameters of the
2.1. Instrumentation and software selected drugs in the literature were studied carefully and the condi-
tions were chosen.
UV-Spectrophotometric analysis was carried out using a Shimadzu
double beam UV spectrophotometer equipped with UV Probe software
2.4.2. Preparation of standard stock solutions and sample solution
(Model UV 1800). Chemometric assisted spectrophotometric measure-
Standard stock solutions of PT, DP, CF and PE (1000 μg/mL) were
ments were performed using the software Unscrambler X. Chromatog-
prepared in the mobile phase. Further aliquots were made from stock
raphy was performed using an isocratic Liquid Chromatograph
solutions to get a standard solution containing 10 μg/mL of each drug.
(Shimadzu) equipped with LC 20AT pump, SPD 20A UV–Vis detector
Sample enrichment technique was used in the preparation of the sam-
and a rheodyne injector and LC solutions software. Phenomenex ODS
ple solution. Twenty tablets were accurately weighed and powdered
analytical column (250 × 4.6 mm; 5 μ) was used for the separation
and an amount of tablet powder equivalent to 50 mg of PT was taken
and optimization of chromatographic data was carried out using Design
and transferred into a 10 mL volumetric flask and 47 mg of CF,
expert software.
47.5 mg of DP and 49.5 mg of PE working standards were added and
the volume was made up to the mark with methanol to get a concentra-
2.2. Materials and reagents
tion of 1000 μg/mL. The solution was kept for sonication for 20 min and
filtered and 0.1 mL of the filtrate was taken and diluted to get a concen-
Working standards of pharmaceutical grade PT, DP, CF and PE were
tration of 10 μg/mL of each drug.
procured from TCI chemicals, India. Methanol (HPLC grade), potassium
dihydrogen phosphate (AR grade), phosphoric acid, sodium hydroxide
2.4.3. Method optimization using chemometric design
(AR grade) were used for chromatographic analysis. Methanol (AR
Preliminary trials were performed by using various chromato-
grade) was used for spectrophotometric analysis. The pharmaceutical
graphic conditions like varying pH and composition of mobile phase
formulation (Vicks action advanced 500 tablet) manufactured by
and flow rate. From the trials, it was observed that two critical parame-
Procter & Gamble Hygiene & Health Care Ltd. was purchased from
ters have a major impact on the separation of the four drugs which were
local market.
methanol content (Factor A) in the range of 45–55% v/v and flow rate
(Factor B) in the range of 0.75–0.85 mL/min. A total of 13 runs were per-
2.3. Chemometric assisted UV spectrophotometric methods
formed as given in the CCD. Two responses were measured like
2.3.1. Preparation of standard stock solutions
Table 1
Standard stock solutions of PT, DP, CF and PE (1000 μg/mL) were
Training and test sets for PCR and PLSR methods.
prepared by using methanol as a solvent. The stock solutions were fur-
ther diluted to get suitable concentration range of 2–16 μg/mL of PT, Training seta Test seta
80–400 μg/mL of DP, 4–14 μg/mL of CF and 20–120 μg/mL of PE to con- Mixture PT PE DP CF Mixture PT PE DP CF
struct calibration curves.
1 8 20 160 6 11 8 40 160 16
2 4 60 320 6 12 12 120 400 10
2.3.2. Sample preparation 3 6 80 240 4 13 8 40 240 10
Sample enrichment technique was used in the preparation of the 4 10 20 80 12 14 16 60 400 10
5 8 20 160 14 15 16 120 160 8
sample solution [6]. Twenty tablets were accurately weighed and pow-
6 6 40 240 12 16 14 40 80 12
dered and an amount of tablet powder equivalent to 10 mg of PT was 7 10 80 80 8 17 10 40 80 4
taken and transferred into a 10 mL volumetric flask and 11.4 mg of CF, 8 2 100 400 4
79.5 mg of DP and 99.99 mg of PE working standards were added and 9 10 80 320 14
the volume was made up to the mark with methanol. The solution 10 4 20 240 4

was kept for sonication for 20 min and filtered. From the above stock a
Concentration values are expressed as μg/mL.
 K. Palur et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801 3

1.000

0.800

0.600
PT PE
DP
Abs.

CF

0.400

0.200

0.000
200.0 250.0 300.0 350.0 400.
nm.

Fig. 1. Overlay absorption spectra of PT, DP, CF & PE.

retention time of 4th eluted drug (Rt4) and resolution of first and second concentrations and peak area of the drugs over the concentration
eluted drugs (RS1,2). Statistical analysis was carried out by using ANOVA range of 2–18 μg/mL of PT, DP, CF and PE. The precision of the developed
and the interaction of various factors on responses was studied using method was determined by performing method precision and system
perturbation plots, contour and response surface plots. Derringer's de- precision. Accuracy was performed by standard addition method at
sirability function was used for optimization of the selected responses. 80, 100, 120% levels of the sample concentrations and their percent re-
The mobile phase consisted of methanol‑potassium dihydrogen ortho- covery values were calculated. Robustness of the optimized method was
phosphate buffer (pH 3; 10 mM) (50:50, v/v). The pH of the buffer determined by changing flow rate (±0.01 mL/min), and change in or-
was adjusted to 3 using 10% ortho phosphoric acid. Phenomenex ODS ganic phase composition (±0.5 mL).The assay values and their % RSD
analytical column was used for separation and flow rate was maintained were calculated.
at 0.81 mL/min and detection was carried out at 220 nm. Ambient tem-
perature was maintained for all the separations.
3. Results and discussion

2.4.4. Validation of developed RP-HPLC method 3.1. PCR and PLSR assisted spectrophotometric methods
The chromatographic conditions were optimized and validated for
various parameters like linearity, selectivity, precision, accuracy and ro- Due to the extensive overlapping of the absorption spectra of the
bustness. Six replicate injections of standard solution of PT, DP, CF and drugs PT, CF, DP & PE present in the formulation (Fig. 1), simultaneous
PE were given to evaluate system suitability. The chromatographic pa- determination can not be performed by conventional methods. In an at-
rameters like retention time, theoretical plates, peak area and tailing tempt to resolve the complex spectra of the selected drugs, chemomet-
factor were determined in terms of percentage RSD. The method was ric assisted UV spectrophotometric methods like PCR & PLSR were
validated for specificity to check for the presence of interference of ex- applied. PCR and PLSR methods are the most widely applied chemomet-
cipients by comparing the chromatograms of blank and sample solution. ric models for the simultaneous estimation of multicomponent formula-
Linearity was assessed by the regression line plotted between tions and especially for the drugs with extensive overlapping in their

Fig. 2. RMSECV graph for PCR. Fig. 3. RMSECV graph for PLSR.
4 K. Palur et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801

Table 2
Statistical parameters obtained by PCR &PLSR methods.

Statistical parameters Principal component regression [PCR] Partial least squares regression [PLSR]

PT PE DP CF PT PE DP CF

Concentration range (μg/mL) 2–16 20–120 80–400 4–14 2–16 20–120 80–400 4–14
No. of factors 8 8 7 4 5 5 4 4
R2 0.9996 0.9999 0.9999 0.9991 0.9999 0.9999 0.9996 0.9991
RMSEC 0.050 0.151 0.819 0.119 0.034 0.171 1.898 0.113
RMSECV 0.410 1.233 4.718 0.285 0.433 1.316 3.793 0.279
RMSEP 0.075 0.302 1.596 0.054 0.072 0.375 1.254 0.047
Calibration set 100.04 ± 0.75 100.01 ± 0.67 99.97 ± 0.88 100.07 ± 1.13 100.94 ± 0.68 100.02 ± 0.68 100.08 ± 0.92 100.06 ± 1.09
Mean ± SD
Validation set 100.39 ± 1.97 100.80 ± 1.71 100.01 ± 1.59 99.97 ± 1.83 100.36 ± 1.82 100.62 ± 1.89 99.82 ± 1.95 99.53 ± 1.31
Mean ± SD
Assay 101.07 ± 0.68 103.60 ± 0.22 101.971 ± 0.52 102.05 ± 0.48 100.94 ± 0.68 103.81 ± 0.25 102.30 ± 0.38 102.03 ± 0.48
Mean ± SD

Table 3
Assay results obtained by PCR method.

Marketed formulation Predicted values (μg/mL) % Assay

PT PE DP CF PT PE DP CF

1 10.039 103.831 82.046 12.189 100.390 103.831 102.557 101.573

2 10.108 103.603 81.447 12.247 101.076 103.603 101.808 102.059
3 10.176 103.374 81.238 12.305 101.762 103.374 101.548 102.545
Mean 101.076 103.603 101.971 102.059
SD 0.686 0.228 0.524 0.486
%RSD 0.679 0.220 0.514 0.476

absorption spectrum. These models were found to be highly effective in PLSR model is the primary step that affect the model accuracy. The num-
determining the concentrations of drugs as they involve the measure- ber of principal components and latent variables were calculated based
ment of absorbances at many wavelengths and their ability in using on loadings and their coefficients as generated by the software. The
all the spectral data. This kind of procedure enhances the accuracy of number of PCs and LVs should be neither too high nor too low as it re-
spectral analysis and also have the advantage of selecting the most in- sults in the overfitting and underfitting of the data respectively. K-fold
formative data and removing the unnecessary ones, which further
makes the models worthy. Hence, Chemometric assisted spectrophoto-
metric methods were found to be more preferable and advantageous of
being economical, simple, rapid and sensitive when compared to con- Table 5
ventional time consuming analytical methods. Central composite design of experiments and responses.

Factor 1 Factor 2 Response 1 Response 2

3.1.1. Selection of wavelength range
Run A:A B:B Rt4 RS1,2
Selection of wavelength range plays a crucial role in the prediction of ORG.PHASE FLOW RATE
analyte concentration by PCR and PLSR models. The wavelength range 1 50 0.8 8.395 2.183
selected for the spectral analysis was 240–320 nm with 1 nm data inter- 2 50 0.8 8.432 2.197
val. The data below 240 nm was discarded due to the presence of noise 3 45 0.75 12.841 2.701
4 50 0.8707 7.001 1.941
and above 320 nm was not selected due to presence of infinitesimal 5 50 0.8 8.451 2.177
absorbance. 6 57.071 0.8 7.204 1.99
7 45 0.85 11.332 2.458
8 55 0.75 6.512 1.784
3.1.2. Selection of principal components and latent variables 9 50 0.8 8.422 2.182
PCR and PLSR calibration models were constructed by selecting the 10 42.928 0.8 15.116 3.021
optimal number of factors. PCR uses principal components (PCs) and 11 50 0.8 8.468 2.186
PLSR uses latent variables (LVs) as factors for the prediction of compo- 12 50 0.7292 8.125 2.001
13 55 0.85 6.451 1.921
nents. Selection of principal components in PCR and latent variables in

Table 4
Assay results obtained by PLSR method.

Marketed formulation Predicted values (μg/mL) % Assay

PT PE DP CF PT PE DP CF

1 10.025 104.066 82.155 12.186 100.255 104.066 102.694 101.548

2 10.094 103.814 81.847 12.244 100.943 103.814 102.308 102.034
3 10.163 103.561 81.538 12.302 101.631 103.561 101.923 102.520
Mean 100.943 103.814 102.308 102.034
SD 0.688 0.252 0.385 0.486
%RSD 0.682 0.243 0.377 0.476
 K. Palur et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801 5

Table 6 to predict the concentrations of compounds in the test set. The differ-
Statistical parameters obtained from ANOVA for CCD. ences between the actual and predicted concentrations were calculated
Statistical parameter obtained from ANOVA Rt4 RS1,2 by using statistical terms like Root Mean Square Error of Cross Valida-
P value <0.0001 <0.0001
tion (RMSECV), Correlation coefficient (R2), Predicted Residual Error
R2 0.9998 0.9991 of Sum of Squares (PRESS) and Root Mean Square Error of calibration
Adjusted R2 0.9997 0.9984 (RMSEC). RMSECV is the important parameter, which determines the
% CV 0.519 0.609 error when test set was predicted by calibration set. The same process
Adequate Precision 271.1298 133.0104
was repeated for all the k-folds. RMSECV values were calculated at
PRESS 0.0910 0.0078
each given no. of PCs in PCR and LVs in PLSR. The number of PCs in
PCR and LVs in PLSR were chosen based on the obtained RMSECV values.
The model with the least RMSECV values was chosen for the simulta-
neous determination of the drugs PT, DP, CF and PE. For PCR method,
cross validation method is used to select the optimum number of PCs the optimum number of PCs for PT, DP, CF and PE were found to be 8,
and LVs. In this method, the entire data set was randomly divided into 7, 4 and 8 respectively as shown in Fig. 2. For PLSR method, the optimum
three k-folds. One of the k-fold is considered as test set and remaining number of LVs for PT, DP, CF and PE were found to be 5, 4, 4, and 5 re-
two were framed into the training set. Then, the models were applied spectively as shown in Fig. 3.

Fig. 4. Perturbation plots corresponding to (A) Retention time; (B) Resolution.

Fig. 5. Contour plots corresponding to (A) Retention time; (B) Resolution.

6 K. Palur et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801

Fig. 6. Response surface plots corresponding to (A) Retention time; (B) Resolution.

3.1.3. Statistical analysis and predicted values, which was reported as percentage recovery)
From the results obtained by the applied PCR & PLSR Chemometric and RMSEP (Root Mean Square Error of Prediction) values which re-
models as shown in Table 2, it was observed that the actual and pre- vealed that both the developed models were found to be accurate. The
dicted values showed good correlation at the selected PCs and LVs, calibrated and validated PCR and PLSR models were applied to predict
which indicated the accuracy of the developed models. The statistical the concentrations of PT, DP, CF and PE in the sample solution and
parameters like mean, SD, % RSD, RMSEC, RMSECV obtained for calibra- assay results were presented in Tables 3 and 4. The assay results ob-
tion set were found to be within the acceptance criteria for both PCR and tained for PT, DP, CF and PE in the marketed formulation by PCR and
PLSR models. Hence, this calibration set was optimized to predict the PLSR models were found to be within the acceptance criteria.
concentrations of drugs in the marketed formulation. External valida-
tion was performed to the test set containing seven standard mixture 3.2. Chemometric assisted RP-HPLC method development and validation
solutions, which were not used in the calibration set to determine the
predictive ability of PCR & PLSR models. The predictive ability of PCR & 3.2.1. Data analysis
PLSR models was defined by accuracy (the difference between actual In HPLC, the separation quality is impacted by the selection of chro-
matographic parameters. From the literature survey and preliminary
trials, factors selection was made. CCD was applied to study the interac-
tion between the factors and to optimize the chromatographic method
with limited number of experiments. The design was executed using
methanol content in the range of 45–55% v/v and flow rate in the
range of 0.75 to 0.85 mL/min. The number of runs were performed as
given by the CCD. The retention time of fourth eluted component (Rt4,
DP) and resolution between first and second eluted components
(RS1,2,PE & PT)) were chosen as responses which had a profound effect
with the changes in factors. Table 5 summarizes the executed runs and
the obtained responses. The quadratic CCD model was suggested and
the ANOVA was generated for the design.
The mathematical expression obtained by the quadratic model for
estimating the responses in terms of coded factors were shown below.

Rt 4 ¼ 8:43366−2:799 A −0:3949B þ 0:3620AB

þ 1:3438A2 −0:4547B2

Table 7
Comparison of experimental and predictive values under optimum condition.

Optimum condition Responses Experimental Predictive Percentage error

MeOH -50% (v/v) Rt4 8.395 8.328 0.804

Mobile phase pH-3 RS1,2 2.183 2.175 0.460
Flow rate-0.81 mL/min
Fig. 7. Desirability function graph.
 K. Palur et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801 7

Table 8 This plot reveals that Rt4 is highly sensitive to organic phase content
Validation and Assay results of chromatographic method. than flow rate. In the contour plot of RS1,2 as shown in Fig. 5B, the con-
Method parameter PT PE CF DP tour with yellow in the lower left corner indicates the contour where
Linearity 2–18 μg/mL 2–18 μg/mL 2–18 μg/mL 2–18 μg/mL
RS1,2 is highest and from the plot, it was also revealed that RS1,2 is sen-
Slope 49,978 38,173 82,382 77,111 sitive to both organic phase content and flow rate. The 3D response sur-
Intercept 29,870 15,991 54,369 148,257 face plot of Rt4 indicates that there is a curvilinear decrease in Rt4 with
R2 0.9990 0.9990 0.9998 0.9992 increase in methanol content and a slight decrease with increase in flow
Method precision (Mean 100.27 101.65 101.88 101.51
rate. The response surface plot of factors on response RS1,2 reveals that
± SD) ± 1.38 ± 0.77 ± 0.38 ± 1.60
Accuracy (Mean ± SD) 100.85 102.91 ± 101.93 ± 101.72 the response RS1,2 can be optimized by using low flow rate and low or-
± 0.59 0.65 0.42 ± 0.31 ganic phase content.
Assay (Mean ± SD) 100.53 103.16 ± 101.43 ± 102.02 ±
± 0.28 0.54 0.41 0.75
3.2.2. Optimization of the design
In the present study, two criteria were identified for optimization of
Rt4 and RS1,2. The criteria selected for the optimization of responses is to
RS1;2 ¼ 2:1850−0:3640A −0:0239B þ 0:0950AB minimize Rt4 in order to reduce the analysis time and to maximize the
þ 0:1547A2 −0:1126B2 RS1,2 so as to achieve a good separation efficiency. After the selection
of criteria, the optimization process was carried out. Derringer's desir-
Note: * indicates significant terms (P value <0.005). ability function was used for the optimization of the responses. The re-
The design was significant as its P value was less than 0.05. All the sponse surface obtained for the desirability function is shown in Fig. 7.
statistical parameters like R2, adjusted R2, adequate precision and % CV The maximum desirability value (D = 0.531) was achieved for metha-
were found to be within limits which indicate the fitness and signifi- nol concentration of 50% v/v and a flow rate of 0.81 mL/min. These coor-
cance of the selected model and were presented in Table 6. The qua- dinates were optimized and validated and further used for analysis of
dratic equation revealed that, for the responses Rt4 and RS1,2, all the marketed formulation. The predicted values and the experimental
terms A, B, AB, A2 and B2 were found to be significant. The negative values have a close agreement with each other for the optimized
sign and largest coefficient of factor A indicated that changing the con- model, which indicates the good efficiency of the design employed.
centration of methanol content from low to high levels showed a The results were shown in Table 7.
rapid decrease in Rt4. Further, at high levels of methanol content, an in-
crease in flow rate showed a marginal decrease in Rt4. The interaction 3.2.3. Validation of RP-HPLC method
term AB indicated a synergistic interaction between factor A and B on The optimized method was validated according to ICH guidelines.
Rt4. From the study, the results revealed that both methanol content The method was found to be linear within the concentration range of
and flow rate should be set at the highest level chosen for the study in 2–18 μg/mL for all the drugs and their correlation coefficients were
order to decrease Rt4. In case of RS1,2, the negative sign of factors A 0.9990 for PT, 0.9992 for DP, 0.9998 for CF and 0.9990 for PE. Both sys-
and B indicated that decreasing the concentration of methanol content tem precision and method precision were performed and their %RSD
and flow rate showed an increase in resolution. The positive sign of values were found to be less than 2. Accuracy was performed by using
the squared term A2 indicates a positive correlation to RS1,2, while sim- standard addition method and % recovery values were found to be sat-
ilar increase in flow rate results in a decrease of RS1,2. The positive sign isfactory as shown in Table 8. From the CCD, it was revealed that small
of the interaction term AB indicated a little synergistic interaction be- changes in Factor A (±0.5%) and Factor B (±0.01 mL/min) does not
tween factor A and B on RS1,2. The quadratic equation for RS1,2 revealed deter the responses like Rt4 and RS1,2 while remaining parameters like
that, both the factors A & B should be set at the lowest levels chosen for theoretical plates, capacity factor were found to be unaffected. This indi-
the study to optimize RS1,2. cates that the method is robust. The optimized method was applied to
From the perturbation plot of Rt4 (Fig. 4A) it was observed that the analyze the content of each drug in the marketed formulation and the
factor A showed a steep slope for Rt4 which indicated that the response assay results were shown in Table 8 and assay chromatogram was
is sensitive to that factor. Factor B showed a relatively flat line with less shown in Fig. 8. The assay results obtained for PT, DP, CF and PE in the
curvature, which indicated that changes in factor B had a lesser impact marketed formulation by RP-HPLC method were found to be within
on Rt4. From the perturbation plot of RS1,2 (Fig. 4B) it was noticed that the acceptance criteria.
RS1,2 increased with a decrease in factor A and small changes were ob-
served with changes in factor B. Interaction effects were also studied 4. Conclusion
by using contour and response surface plots as shown in Figs. 5 & 6, re-
spectively. In the contour plot of Rt4 (Fig. 5A), the blue color in the lower Analytical methods with good accuracy, precision and capable of
right corner of the plot indicates the contour where Rt4 is minimum. performing analysis of multicomponent formulations within a very

Fig. 8. Assay chromatogram.

8 K. Palur et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 243 (2020) 118801

Table 9 caffeine in pharmaceuticals by chemometric methods, Spectrochim. Acta A 70

One way ANOVA results for the developed methods. (2008) 491–499, https://doi.org/10.1016/j.saa.2007.07.033.
[2] M. Heba, Mohamed, a study of selective spectrophotometric methods for simulta-
Drug Results PCR PLSR RP-HPLC neous determination of Itopride hydrochloride and Rabeprazole sodium binary mix-
ture: resolving sever overlapping spectra, Spectrochim. Acta A Mol. Biomol.
PT Mean Assay 101.073 100.940 100.537 Spectrosc. 136 (2015) 1308–1315, https://doi.org/10.1016/j.saa.2014.10.018.
± SD ± 0.685 ± 0.690 ± 0.266 [3] Erdal Dinc, Dumitru Baleanu, Giuseppina Ioele, Michele de Luca, Gaetano Ragno,
F-value 0.69 Multivariate analysis of paracetamol, propiphenazone, caffeine and thiamine in qua-
P-value 0.537 ternary mixtures by PCR, PLS and ANN calibrations applied on wavelet transform
PE Mean Assay 103.600 103.810 103.160 data, J. Pharm. Biomed. Anal. 48 (2008) 1471–1475, https://doi.org/10.1016/j.jpba.
± SD ± 0.230 ± 0.250 ± 0.537 2008.09.035.
F-value 2.45 [4] Alaa El-Gindy, Samy Emara, Heba Shaaban, Development and validation of
P-value 0.166 chemometrics-assisted spectrophotometric and liquid chromatographic methods
DP Mean Assay 101.963 102.013 102.023 for the simultaneous determination of two multicomponent mixtures containing
± SD ± 0.524 ± 0.635 ± 0.746 bronchodilator drugs, J. Pharm. Biomed. Anal. 43 (2007) 973–982, https://doi.org/
F-value 0.01 10.1016/j.jpba.2006.09.020.
P-value 0.993 [5] Alaa El-Gindy, Fawzy El-Yazby, Ahmed Mostafa, Moustafa M. Maher, HPLC and che-
mometric methods for the simultaneous determination of cyproheptadine hydro-
CF Mean Assay 102.057 102.030 101.430
chloride, multivitamins, and sorbic acid, J. Pharm. Biomed. Anal. 35 (2004)
± SD ± 0.485 ± 0.490 ± 0.416
703–713, https://doi.org/10.1016/j.jpba.2004.02.027.
F-value 1.74
[6] Bassam M. Ayoub, Development and validation of simple spectrophotometric and
P-value 0.253 chemometric methods for simultaneous determination of empagliflozin and met-
formin: applied to recently approved pharmaceutical formulation, Spectrochim.
Acta A Mol. Biomol. Spectrosc. 168 (2016) 118–122, https://doi.org/10.1016/j.saa.
2016.06.010.
less span of time is always an interest of analysts. This can be achieved [7] F. Belal, F. Ibrahim, Z.A. Sheribah, H. Alaa, New spectrophotometric/chemometric
assisted methods for the simultaneous determination of imatinib, gemifloxacin,
when analytical methods are coupled with chemometric tools. The as- nalbuphine and naproxen in pharmaceutical formulations and human urine,
sistance of chemometrics to the spectroscopic and chromatographic Spectrochim. Acta A Mol. Biomol. Spectrosc. 198 (2018) 51–60, https://doi.org/10.
methods makes them powerful and efficient in the simultaneous deter- 1016/j.saa.2018.02.048.
[8] Mahsa Khalili, Mahmoud Reza Sohrabi, Vahid Mirzabeygi, Nahid Torabi Ziaratgahi,
mination of drugs in multicomponent formulations. Chemometric simultaneous determination of Sofosbuvir and Ledipasvir in pharma-
A comparative study of the use of chemometric assisted UV spectro- ceutical dosage form, Spectrochim. Acta A Mol. Biomol. Spectrosc. 194 (2018)
photometric & RP-HPLC methods were performed for the determina- 141–151, https://doi.org/10.1016/j.saa.2018.01.011.
[9] Veena D. Singh, Sanjay J. Daharwal, Development and validation of multivariate cal-
tion of PT, CF, DP & PE in tablet dosage form by using one way ANOVA
ibration methods for simultaneous estimation of Paracetamol, Enalapril maleate and
as shown in Table 9. The calculated F-value and P-values revealed that hydrochlorothiazide in pharmaceutical dosage form, Spectrochim. Acta A Mol.
there are no significant differences among the three methods and all Biomol. Spectrosc. 171 (2017) 369–375, https://doi.org/10.1016/j.saa.2016.08.028.
the methods can be applied for the analysis of the chosen multicompo- [10] Shereen M. Tawakkol, Mohamed B. El-Zeiny, A. Hemdan, Full spectrum and selected
spectrum based chemometric methods for the simultaneous determination of
nent formulation. Chemometric assisted UV spectrophotometric cinnarizine and dimenhydrinate in laboratory prepared mixtures and pharmaceuti-
methods like PCR & PLSR were found to be economical and simple in cal dosage form, Spectrochim. Acta A Mol. Biomol. Spectrosc. 173 (2017) 892–896,
comparison with RP-HPLC, as they do not require expensive materials https://doi.org/10.1016/j.saa.2016.10.055.
[11] J. Ghasemi, A. Niazi, Simultaneous determination of cobalt and nickel. Comparison of
and equipment even though RP-HPLC is more specific. prediction ability of PCR and PLS using original, first and second derivative spectra,
Microchem. J. 68 (2001) 1–11, https://doi.org/10.1016/S0026-265X(00)00159-4.
CRediT authorship contribution statement [12] Edward V. Thomas, David M. Haaland, Comparison of multivariate calibration
methods for quantitative spectral analysis, Anal. Chem. 62 (1990) 1091–1099,
https://doi.org/10.1021/ac00209a024.
Keerthisikha Palur: Conceptualization, Methodology, Formal analy- [13] Kevin N. Andrew, Paul J. Worsfold, Comparison of multivariate calibration tech-
sis, Writing - original draft. Sreenivasa Charan Archakam: Writing - re- niques for the quantification of model process streams using diode-array spectro-
view & editing, Visualization, Resources, Validation. Bharathi Koganti: photometry, Analyst. 119 (1994) 1541–1546, https://doi.org/10.1039/
AN9941901541.
Supervision, Conceptualization, Writing - review & editing, Validation. [14] Prafulla Kumar Sahu, R. Nageswara Rao, Teresa Cecchi, Suryakanta Swain, Chandra
Sekhar Patro, Jagadeesh Panda, An overview of experimental designs in HPLC
Declaration of competing interest method development and validation, J. Pharm. Biomed. Anal. 147 (2018)
590–611, https://doi.org/10.1016/j.jpba.2017.05.006.
[15] Mangmang Sang, Du Guangyan, Jia Hao, Linlin Wang, Erwei Liua, Yi Zhang, Tao
The authors declare that they have no known competing financial Wang, Xiumei Gao, Lifeng Han, Modeling and optimizing inhibitory activities of
interests or personal relationships that could have appeared to influ- Nelumbinis folium extract on xanthine oxidase using response surface methodol-
ence the work reported in this paper. ogy, J. Pharm. Biomed. Anal. 139 (2017) 37–43, https://doi.org/10.1016/j.jpba.
2017.02.048.
[16] P. Giriraj, T. Sivakkumar, A rapid - chemometrics assisted RP-HPLC method with
Acknowledgement PDA detection for the simultaneous estimation of ofloxacin and nimorazole in phar-
maceutical formulation, J. Liq. Chromatogr. R T. 38 (2014) 904–910, https://doi.org/
10.1080/10826076.2014.991870.
The authors thank Smt. P. Sulochana, chairperson, Sri.P.Praneeth, di- [17] V. Sree Janardhanan, R. Manavalan, K. Valliappan, Chemometric technique for the
rector and Dr. D. Ranganayakulu, principal of Sri Padmavathi School of optimization of chromatographic system: simultaneous HPLC determination of
Pharmacy, Tiruchanoor, AP, India for providing necessary infrastructure Rosuvastatin, Telmisartan, Ezetimibe and atorvastatin used in combined cardiovas-
cular therapy, Arab. J. Chem. 9 (2016) 51378–51387, https://doi.org/10.1016/j.
and facilities to carry out this research work.
arabjc.2012.03.001.
[18] T. Sivakumar, R. Manavalan, C. Muralidharan, K. Valliappan, Multi-criteria decision
References making approach and experimental design as chemometric tools to optimize
HPLC separation of domperidone and pantoprazole, J. Pharm. Biomed. Anal. 43
[1] M.R. Khoshayand, H. Abdollahi, M. Shariatpanahi, A. Saadatfard, A. Mohammadi, Si- (2007) 1842–1848, https://doi.org/10.1016/j.jpba.2006.12.007.
multaneous spectrophotometric determination of paracetamol, ibuprofen and