drug formulations and clinical methods chemometrics … · 2017. 2. 13. · 948 el-ragehy et al.:...

948 El-RagEhy Et al.: JouRnal of aoaC IntERnatIonal Vol. 99, no. 4, 2016

DRUG FORMULATIONS AND CLINICAL METHODS

Chemometrics Tools in Detection and Quantitation of the Main Impurities Present in Aspirin/Dipyridamole Extended-Release CapsulesNarimaN a. El-ragEhy, ali m. yEhia, Nagiba y. hassaN, and mahmoud a. TaNTawy

1

Cairo University, Faculty of Pharmacy, Analytical Chemistry Department, Kasr el Aini St, 11562 Cairo, EgyptmohamEd abdElkawyFuture University, Faculty of Pharmacy, Analytical Chemistry Department, End of 90th St, Fifth Settlement, New Cairo, Egypt

Aspirin (ASP) and dipyridamole (DIP) in combination is widely used in the prevention of secondary events after stroke and transient ischemic attack. Salicylic acid is a well-known impurity of ASP, and the DIP extended-release formulation may contain ester impurities originating from the reaction with tartaric acid. UV spectral data analysis of the active ingredients in the presence of their main impurities is presented using multivariate approaches. Four chemometric-assisted spectrophotometric methods, namely, partial least-squares, concentration residuals augmented classical least-squares (CRACLS), multivariate curve resolution (MCR) alternating least-squares (ALS), and artificial neural networks, were developed and validated. The quantitative analyses of all the proposed calibrations were compared by percentage recoveries, root mean square error of prediction, and standard error of prediction. In addition, r2 values between the pure and estimated spectral profiles were used to evaluate the qualitative analysis of CRACLS and MCR-ALS. The lowest error was obtained by the CRACLS model, whereas the best correlation was achieved using MCR-ALS. The four multivariate calibration methods could successfully be applied for the extended-release formulation analysis. The application results were also validated by analysis of the stored dosage-form solution, which showed a susceptibility of DIP esterification in the extended-release formulation. Statistical comparison between the proposed and official methods showed no significant difference.

Detection and quantitative determination of impurities in bulk drugs and pharmaceutical formulations are important regulatory requirements for new drugs (1).

Impurities may originate from different sources, including excipients and degradation products, and as a result of incomplete reactions during the synthesis (2). In addition, some pharmaceutical formulations showed certain drug–excipient incompatibility via chemical reactions (3). Therefore, comprehensive study of possible impurities, along with their detection and quantification in dosage forms, is absolutely vital.

Antiplatelet combination therapy for secondary-event prevention after stroke and transient ischemic attack is strongly supported (4). One of the commonly used antiplatelet agents is aspirin (ASP; 2–acetoxybenzoic acid). ASP interferes with the formation of thrombi by inhibition of cyclooxygenase enzyme and, therefore, reduces the risk of stroke (5, 6). Another antiplatelet agent is dipyridamole (DIP; 2,2′,2′′,2′′′-{[4,8-Di(piperidin-1-yl)pyriMido(5,4-d)pyriMidine-2,6-diyl]bis(azanetriyl)}tetraethanol). DIP acts differently by inhibiting platelet aggregation through the prevention of adenosine uptake into platelets (5, 6). Clinical evidence proved that the risk of stroke is reduced significantly using ASP plus DIP in extended-release formulation compared to either single agent (7). The extended-release formulation of DIP encapsulates a tartaric acid core. This formulation was developed to enhance the absorption under pathological or pharmaceutically induced low-acidic conditions in the stomach, as well as to enhance drug absorption from the neutral or basic intestinal environment (8). Salicylic acid (SAL) is a well-known impurity of ASP, which originates during synthesis and is considered a degradation product of ASP (9). Because alcoholic DIP is most likely to react with tartaric acid, DIP tartaric acid ester impurity (DIP-I) is probably formed in this extended formulation (8, 10). The ester has been reported as an impurity in DIP extended-release pharmaceutical formulations (11). The chemical structures of the studied drugs, along with their impurities, are presented in Figure 1.

Several analytical techniques have been reported in the literature for the determination of intact drugs in their binary mixtures, including spectrophotometry (12), spectrofluorometry (13), and LC (14, 15). Other spectrophotometric (16), spectrofluorometric (17), and chromatographic (17, 18) methods were reported for their simultaneous determination in presence of SAL. Two HPLC methods for DIP determination in the presence of its impurities, including DIP-I, have been reported (19, 20). In addition, Wang et al. (21) reported HPLC coupled with MS for the simultaneous determination of DIP and SAL in plasma.

To the best of our knowledge, none of these reported methods has considered the simultaneous determination of ASP and DIP in the presence of SAL and DIP-I as impurities. Therefore, the aim of this work is to develop simple chemometric-assisted spectrophotometric methods that could be used for the quantitative and qualitative determination of ASP and DIP and their impurities. The proposed methods are partial least-squares (PLS), concentration residuals augmented classical least-squares (CRACLS), multivariate curve resolution (MCR) alternating least squares (ALS), and artificial neural networks (ANN).

Received March 23, 2016. Accepted by GL May 4, 2016.1 Corresponding author’s e-mail: [email protected]: 10.5740/jaoacint.16-0082

00948-00956.indd 948 24/06/16 1:21 PM

El-RagEhy Et al.: JouRnal of aoaC IntERnatIonal Vol. 99, no. 4, 2016 949

These algorithms were challenged for their qualitative and quantitative analysis of drugs and their impurities.

Experimental

Instrument and Software

Spectrophotometric measurements were carried out on a dual-beam UV-Vis spectrophotometer (Model No. UV-1601 PC; SHIMADZU Corp., Kyoto, Japan). UVPC personal spectroscopy software version 3.7 (SHIMADZU Corp.) was used to process absorption and derivative spectra. Scans were carried out in the range 200–500 nm at 0.1 nm intervals using 1.00 cm quartz cells. All data analysis was carried out using MATLAB® (7.0.1; MathWorks, Natick, MA; 22), the MCR-ALS GUI (free software available at http://www.mcrals.info), and Neural Network Toolbox™ implemented in MATLAB.

A GC-MS instrument (SHIMADZU GC-MS-QP2010) was used with a SHIMADZU AOC-20i autosampler. A 30 m, HP-5ms (5% Phenyl)-methylpolysiloxane column was used for separation, with a 0.253 mm id and a 0.50 μm film thickness (Agilent Technologies, Palo Alto, CA) using electron impact ionization mode.

Materials and Reagents

(a) Pure standards.—ASP and SAL were kindly supplied by ADWIC (Qaliubiya, Egypt). Their purity was checked and found to be 100.12 and 100.16%, respectively, according to official methods (9).

DIP was kindly supplied by Boehringer Ingelheim (Cairo, Egypt). Its purity was checked and found to be 100.83% according to official methods (9).

DIP-I was laboratory prepared using DIP and tartaric acid through esterification reaction (23). Details are described later in Procedures, section (a) The purity of DIP-I was checked and found to be 100.92% according to a reported HPLC method (20).

(b) Pharmaceutical dosage form.—Asasantin® SR (25 mg/200 mg), batch No. 307505, is labeled to contain 25 mg ASP and 200 mg extended-release DIP per capsule and is manufactured by Boehringer Ingelheim. Capsules are labeled to contain tartaric acid, povidone, methacrylic acid copolymer, talc, acacia, hypromellose, glycerol triacetate, dimethicone, lactose, maize starch, sucrose, gelatin, microcrystalline cellulose, and coloring agents such as titanium dioxide, iron oxide red, and iron oxide yellow.

(c) Chemicals and reagents.—All chemicals used throughout this work were of analytical grade and solvents were of HPLC grade. Methanol and 98% sulfuric acid were from Merck (Darmstadt, Germany), and tartaric acid was from ADWIC-Egypt.

Solutions

(a) Stock solutions.—Stock solutions of 100 μg/mL of ASP and DIP were prepared in methanol. Stock solutions of 20 μg/mL of SAL and DIP-I were prepared in methanol.

(b) Pharmaceutical dosage form solution.—Contents of 50 capsules of Asasantin SR were evacuated, accurately weighed, and finely powdered. Accurately weighed portions equivalent to 2.5 mg ASP and 20 mg DIP were transferred into a 100 mL beaker, sonicated in 30 mL methanol for 10 min, and filtered into a 100 mL volumetric flask. The residues were washed three times, each using 10 mL methanol, and the solutions were diluted to volume with the same solvent.

Procedures

(a) Preparation and identification of DIP-I.—Esterification procedure (23) was followed to prepare DIP-I. An amount of DIP pure form equivalent to 10−1 mol was added to 10−1 mol tartaric acid and dissolved in a sufficient amount of methanol. A few drops of sulfuric acid were added as the dehydrating agent, and reflux was carried out at 100°C for 2 h. The solution was then evaporated to dryness. The solid residue was used for GC-MS scan, in which helium was used as the carrier gas at a flow rate of 1.0 mL/min.

(b) Construction of calibration models.—Spectral characteristics of the studied components were examined by scanning different concentrations of each from 200 to 500 nm against methanol as a blank. The normalized spectra were calculated to check the linearity of each component. The calibration set was designed using 25 training samples, containing different concentrations of ASP, DIP, SAL, and DIP-I in the ranges of 5–15, 10–20, 1–3, and 2–4 μg/mL, respectively. The solutions were prepared by mixing different aliquots from their respective stock standard solutions in 10 mLvolumetric flasks and then diluted to volume with methanol. The spectra of the prepared solutions were recorded over the range 200–500 nm. The spectra were digitized each 1.0 nm over the range 220.0–440.0 nm, and 221 experimental points were used in the calculations. The data points of spectra were transferred to MATLAB for subsequent data analysis, and multivariate calibration models were constructed. All the spectral data were mean centered before calibration to construct PLS, CRACLS, MCR-ALS, and ANN models. Optimal regressions were systemically studied for each model. In the PLS model calibration, the number of latent variables (LVs) was optimized by leave-one-out cross-validation. The optimum number of LVs

Figure 1. Chemical structures of the studied components. M.wt., molecular weight.

00948-00956.indd 949 24/06/16 1:21 PM


corresponding to the lowest root mean square error of calibration (RMSEC) value was 10. To optimize the CRACLS calibration model, repetitive approximation of absorptivity matrix continued for several iterations. The number of iterations used was eight. Applied constraints were the critical optimization parameter in MCR-ALS calibration; non-negativity constraint (non-negativity least-squares) was applied to the concentration profi le with equality constraint (equal than) to obtain satisfactory parameters, with the least number of iterations being 11. The ANN type used in this work is the feed-forward model that was trained with the Levenberg-Marquardt back propagation of errors learning algorithm. Back propagation is considered the most traditional nonlinear feed-forward multilayer network that can capture and denote complex input and output relationships and, frequently, can be used comparatively for building the regression model ( 24 ). The numbers of neurons in input and output layers were 221 and 4, respectively. The number of neurons in the hidden layer was optimized ( 2 , 4 , 6 , 8 , and 10 ). Lastly, the optimum was using four neurons in the hidden layer with pureline transfer function, a learning rate of 0.1, and 20 epochs.

(c) Validation of calibration models .—Fifteen independent validation samples were prepared by mixing different aliquots from their respective stock standard solutions in 10 mL volumetric fl asks and then diluting to volume with methanol. The developed parameters for PLS, CRACLS, MCR-ALS, and ANN models were used for the determination of the drugs, along with their impurities, in the external validation set.

(d) Application to pharmaceutical dosage form .— Aliquots of 1.0 mL were transferred from the prepared solutions to 10 mL volumetric fl asks and diluted with methanol for direct determination of DIP. However, the standard addition technique was applied for the accurate determination of ASP, whereby the transferred aliquots were fortifi ed with successive addition (0.25, 0.5, and 0.75 mL) of ASP stock standard solution. The general procedure described in Procedures, section b was followed to determine the concentrations of each drug in the prepared dosage form solutions.

Results and Discussion

Analysis of DIP-I

DIP-I was prepared as in Procedures , section (a) The formation of DIP-I was confi rmed by detection of its peak by a reported HPLC method ( 20 ) using an Inertsil ® octadecilsilane-2

(150 mm × 4.6 mm ) 5 μm column (GL Sciences Inc. USA) with mobile phase containing a mixture of phosphate buffer pH 7.0 and methanol, with a gradient program (time per percentage methanol) set at 00/50, 4/50, 25/95, 28/95, 30/50, and 35/50, a fl ow rate of 1.0 mL/min, and a detection wavelength at 295 nm. Complete esterifi cation was confi rmed after 2 h by the disappearance of a DIP chromatographic peak. The mass spectrum of DIP-I (C 28 H 44 N 8 O 9 ; Figure 2 ) shows its molecular ion peak at m/z 636.

Chemometric Methods

Spectrophotometry is widely applicable for the determination of different drugs. It offers an inexpensive and simple tool for the simultaneous determination of pharmaceutical formulations. However, simple univariate spectrophotometric methods fail to resolve spectral overlap in complex systems. Concerning the studied components ( Figure 3 ), shows a severe overlap of the ASP and SAL spectra with the DIP and DIP-I spectra, in addition to spectral similarity between DIP and its impurity. These severely overlapped spectra (between the drugs and their impurities) prohibit the use of ordinary spectrophotometric techniques. Because multivariate data analysis is always superior in fi nding a solution to complicated data, chemometrics tools were applied in this work. PLS, CRACLS, MCR-ALS, and ANN were used for quantifi cation of the cited components. Moreover, CRACLS and MCR-ALS were challenged for their qualitative data analyses.

Absorbance data from the range 220.0–440.0 nm for 25 mixtures was used to optimize and calibrate the proposed models. Wavelengths less than 220.0 nm were excluded because of high noise infl uence, whereas wavelengths greater than 440.0 nm were excluded because DIP and DIP-I show almost the same absorbance characteristics in this range and, therefore, are less informative. Therefore, absorbance values obtained outside the specifi ed range would have resulted in a signifi cant noise level within the calibration matrix, thus affecting precision.

PLS .— PLS is one of the most popular techniques to build multivariate calibration models ( 25 ). The PLS algorithm takes into account the information of responses and concentrations simultaneously. In this study, the cross-validation method of leaving out one sample at a time was used to select the optimum number of factors ( Figure 4 ). Here, 10 LVs were optimum because this number provided the least signifi cant prediction error. The high number of LVs is attributed to a large number of training samples.

Figure 2. The mass spectrum of DIP-I showing its molecular ion peak at m/z 636.

00948-00956.indd 950 24/06/16 1:21 PM


CRACLS .— CRACLS ( 26 ) is an alternative method that estimates absorptivity using both data and concentration matrixes by a process of repetitive approximation. The initial calculated absorptivity is used to predict the concentration matrix, and then error is calculated. One vector at a time, error is augmented to the original concentration and then error is considered a new component. Calculation of absorptivity matrix was repeated using the augmented concentration matrix. This iterative approach was continued until no further improvement was seen in prediction. In this study, eight iterations were required to improve model prediction.

MCR-ALS .—MCR is a factor analysis–derived method that assumes a bilinear model ( 27 ). In MCR, the measured spectra data matrix is decomposed into the concentration and spectral-profi le matrixes of the pure components in the samples, and then error is calculated. Repetitive estimations of concentrations from spectral profi les, and vice versa, by MCR were optimized by the ALS procedure. Because data matrix decomposition has no unique solution, the number of possible solutions could be minimized by applying constraints like unimodality, closure, equality, or non-negativity. To start the optimization of ALS, a method based on “simple-to-use interactive self-modeling analysis” ( 28 ) was applied to obtain a preliminary estimation of the spectral-profi le matrix that was used to calculate the unconstrained least-squares solution for the concentration profi le.

In this work, non-negative least-squares as a non-negativity constraint was applied to the concentration profi le, which necessitated the concentration to be equal to or greater than zero. In addition, equality constraint (lower than) was applied to the concentration profi le using the selectivity matrix. This allows for some departures of the updated values in the concentration profi le from the defi ned values in the selectivity matrix. The ALS optimization process ended upon reaching a defi nite convergence criterion (0.1%). The convergence was achieved after 11 iterations. The quality of MCR-ALS, fi t to the matrix of experimental data, was calculated by the percentage of lack of fi t and variance percentage (R 2 ). The calculated percentage of lack of fi t and R 2 were 0.64429 and 99.9958, respectively, which were satisfactory enough to assist the quality of the proposed model.

ANN .— ANN is a kind of imitated intelligence method that comprises a huge number of simple, highly interconnected nodes or artifi cial neurons that imitate the real biological nervous system in fi nding the relationship between inputs and outputs. Typically, ANN encompasses three layers (input, hidden, and output) with transfer functions ( 24 ). A total of 221 neurons were used in the input layer, corresponding to number of spectral data points used, and four neurons were used in the output layer, corresponding to the number of the components to be determined in each sample. The number of neurons in the hidden layer should be adjusted on a trial-and-error basis ( Figure 5 ). The fi gure shows a signifi cant decrease in RMSEC values from 2 to 4 hidden neurons, whereas the decrease was not signifi cant upon further increasing the number of hidden

Figure 3. Normalized spectra of ASP (solid line), DIP (dashed and dotted line), SAL (dashed line), and DIP-I (dotted line), using methanol as a blank.

Figure 4. RMSEC plot of the cross-validation results of the calibration set as a function of the number of principal components used to construct the PLS calibration .

Figure 5. Plot of RMSEC against number of neurons in the hidden layer during optimization of the ANN calibration model.

00948-00956.indd 951 24/06/16 1:21 PM


neurons. The optimum number of hidden neurons is four with purelin-purelin transfer function. Other parameters were also set, such as a 0.1 learning rate and 20 epochs.

To test the predictive ability of the developed PLS, CRACLS, MCR-ALS, and ANN models, they were challenged with the spectra of the validation set. The mean of recoveries and RSDs were calculated for each component (Table 1). Regression parameters of the validation sets were calculated, along with root mean square error of prediction (RMSEP) and standard error of prediction (SEP; Table 2). The close values of RMSEP and SEP indicated no case of overfitting in the proposed calibration models.

The CRACLS and MCR-ALS models provided a qualitative meaning in their algorithms; therefore, spectral profiles of drugs and their impurities could be estimated. The resemblance was observed with the pure spectra (Figure 6), and r2 values were calculated that indicated good correlation between the estimated and pure spectra for each component. However, a lower correlation for the impurities was obtained, especially for the CRACLS model.

Table 1. Predicted percentage recoveries of validation sets

Mix No.Actual concn,

μg/mL PLS CRACLS MCR-ALS ANN

ASP

1 10.0 100.05 99.85 100.94 100.18

2 10.0 99.91 100.09 98.11 99.95

3 5.0 98.98 99.83 100.94 99.49

4 7.5 99.75 99.96 99.47 99.52

5 5.0 100.10 100.20 101.72 100.26

6 15.0 100.06 100.06 102.66 100.10

7 15.0 99.89 99.98 100.28 99.93

8 10.0 100.21 100.01 101.10 100.26

9 7.5 100.08 99.87 98.80 100.31

10 15.0 99.99 99.92 103.05 100.02

11 7.5 100.03 99.83 99.68 99.81

12 12.5 99.86 100.09 100.16 99.71

13 12.5 100.09 99.93 98.75 100.23

14 10.0 99.89 99.93 101.59 99.80

15 15.0 99.91 100.02 101.53 99.99

Mean ± RSD 99.92 ± 0.286

99.97 ± 0.109

100.59 ± 1.434

99.97 ± 0.263

DIP

1 15.0 102.43 102.17 100.88 101.72

2 10.0 100.14 99.81 98.92 100.43

3 12.5 101.32 100.32 102.21 101.18

4 10.0 100.52 99.92 100.63 100.18

5 20.0 99.79 99.69 101.21 99.81

6 20.0 99.80 99.80 98.28 99.58

7 15.0 100.30 100.03 101.03 100.27

8 12.5 99.94 100.21 99.21 100.02

9 20.0 99.77 100.02 99.99 99.70

10 12.5 100.42 100.49 97.70 100.13

11 17.5 99.98 100.35 98.62 99.71

12 17.5 100.41 99.87 102.62 100.58

13 15.0 99.78 100.31 100.22 100.23

14 20.0 99.90 99.92 103.79 99.75

15 17.5 100.11 101.61 101.03 101.78

Mean ± RSD 100.31 ± 0.714

100.30 ± 0.692

100.42 ± 1.687

100.34 ± 0.703

SAL

1 2.0 100.26 100.15 97.67 100.47

2 1.5 100.69 100.45 101.61 103.01

3 1.0 99.51 99.86 97.65 100.66

4 3.0 100.09 99.93 101.43 99.35

5 3.0 100.35 100.16 101.43 100.72

6 2.0 100.41 100.24 97.67 101.33

7 1.5 100.35 100.15 101.61 100.01

8 3.0 100.10 100.02 101.43 100.62

Mix No.Actual concn,

μg/mL PLS CRACLS MCR-ALS ANN

9 1.5 99.61 100.17 101.61 98.01

10 2.5 100.28 100.38 97.60 99.45

11 2.5 100.26 99.89 97.60 98.66

12 2.0 99.88 99.96 97.67 98.65

13 3.0 99.76 99.94 101.43 100.97

14 2.5 99.90 99.88 97.60 99.82

15 3.0 99.74 100.09 101.43 99.66

Mean ± RSD 100.08 ± 0.335

100.08 ± 0.182

99.70 ± 2.001

100.09 ± 1.243

DIP-I

1 3.0 99.52 99.48 98.63 99.06

2 2.0 99.02 99.43 99.97 98.89

3 4.0 96.94 98.78 99.89 98.34

4 4.0 99.69 100.04 99.89 98.71

5 3.0 99.59 102.54 98.63 99.68

6 2.5 103.00 101.80 104.30 105.16

7 4.0 99.87 100.12 99.89 101.34

8 2.5 101.34 98.56 104.30 102.88

9 3.5 100.71 100.17 99.79 100.33

10 3.5 97.25 98.20 99.79 97.08

11 3.0 98.61 97.83 98.63 97.09

12 4.0 99.10 100.15 99.89 96.45

13 3.5 99.89 98.79 99.79 101.70

14 4.0 102.01 100.53 99.89 101.16

15 2.0 99.52 100.63 99.97 98.06

Mean ± RSD 99.74 ± 1.601

99.80 ± 1.296

100.22 ± 1.730

99.73 ± 2.396

Table 1. (continued )

00948-00956.indd 952 24/06/16 1:21 PM


Table 2. Regression parameters of the validation sets calculated for each proposed model

Component Model Slope Intercept LOD, μg/mL LOQ, μg/mL r2 RMSEP SEP

ASP PLS 1.0010 −0.0153 — — 1.0000 0.017624 0.017396

CRACLS 1.0004 −0.0064 — — 1.0000 0.009114 0.009084

MCR-ALS 1.0251 −0.1869 — — 0.9990 0.191550 0.181825

ANN 1.0009 −0.0105 — — 1.0000 0.021235 0.021929

DIP PLS 0.9902 0.1934 — — 0.9996 0.109829 0.106090

CRACLS 0.9977 0.0808 — — 0.9995 0.116360 0.110936

MCR-ALS 1.0218 −0.2612 — — 0.9971 0.292069 0.290523

ANN 0.9920 0.1709 — — 0.9995 0.120568 0.115417

SAL PLS 0.9999 0.0021 0.021 0.064 0.9999 0.006453 0.006397

CRACLS 0.9992 0.0034 0.012 0.037 1.0000 0.003976 0.003736

MCR-ALS 1.0198 −0.0486 0.140 0.424 0.9980 0.043859 0.045233

ANN 0.9991 0.0032 0.079 0.239 0.9994 0.023064 0.023840

DIP-I PLS 0.9837 0.0419 0.176 0.533 0.9973 0.053060 0.053797

CRACLS 0.9940 0.0123 0.130 0.394 0.9985 0.038761 0.039403

MCR-ALS 0.9799 0.0681 0.143 0.433 0.9982 0.043490 0.044901

ANN 0.9766 0.0639 0.243 0.737 0.9949 0.072457 0.074000

Figure 6. Pure spectra (solid line) and extracted spectra by CRACLS (dotted line) and MCR-ALS (dashed line) for the four components with the corresponding r2 values (shown as R) of (a) ASP, (b) DIP, (c) SAL, and (d) DIP-I.

00948-00956.indd 953 24/06/16 1:21 PM


Application of the Proposed Methods to an Assay of the Pharmaceutical Dosage Form

The proposed methods were applied to the determination of intact drugs in their combined dosage form; the mean recoveries and RSDs are shown in Table 3. The table shows the analysis of freshly prepared and stored dosage form solutions. Good recoveries were obtained for ASP in both conditions, whereas DIP shows a lower recovery in the stored dosage form, which indicates the possibility of its esterification during 1 week of storage at room temperature. This stored dosage form solution was compared to a synthetic mixture of drugs and their impurities by TLC using toluene–methanol–ethyl acetate (2 + 3 + 5, v/v/v) as the mobile phase. The chromatograms were scanned at 275 nm. Retardation factor values were 0.19, 0.40, 0.56, and 0.77 for ASP, SAL, DIP-I, and DIP, respectively. Traces of DIP-I were detected in the stored dosage form solution (Figure 7).

Table 3. Determination of ASP and DIP in Asasantin SR capsules

Fresh solutiona Stored solutionb

Model ASPc DIP ASPc DIP

PLS 100.40 ± 1.213

99.75 ± 0.494 101.37 ± 1.245 84.75 ± 1.486

CRACLS 99.92 ± 1.345

100.13 ± 2.100 101.22 ± 1.193 83.92 ± 1.938

MCR-ALS 98.83 ± 1.737

98.52 ± 1.205 98.71 ± 1.831 84.20 ± 0.968

ANN 99.89 ± 0.313

98.46 ± 1.345 101.01 ± 0.931 84.38 ± 1.649

a Average of five determinations of fresh capsule dosage form solution (values are mean percentage recovery ± RSD).

b Average of three determinations of capsule dosage form solution stored for 1 week at room temperature (values are mean percentage recovery ± RSD).

c Recoveries obtained by standard addition technique.

Figure 7. TLC chromatograms of (a) a synthetic mixture of the four components and (b) a stored dosage form solution showing traces of DIP-I, using silica gel as a stationery phase, toluene–methanol–ethyl acetate (2 + 3 + 5, v/v/v) as a mobile phase, and UV detection at 275 nm.

00948-00956.indd 954 24/06/16 1:21 PM


Statistical Analysis

The results obtained for analysis by the proposed calibrations were statistically compared with the official titrimetric methods (9) for ASP, DIP, and SAL and with the reported HPLC method (20) for DIP-I. One-way analysis of variance was used to perform the F-test at P = 0.05. The calculated t and F values were less than the tabulated ones, which indicates that there is no significant difference with regard to accuracy and precision (Table 4).

Conclusions

Advancement of chemometrics provides a large number of valuable tools that facilitate resolving complex spectral data in the study of drugs and their main impurities. The proposed calibration models were simple and accurate for the determination of ASP, DIP, SAL, and DIP-I. A lower prediction error was noticed for CRACLS compared to other models. Moreover, CRACLS, along with MCR-ALS, provided suitable tools for qualitative analyses. The extracted spectral profiles of the studied components showed good correlation with their corresponding scanned ones, specifically for MCR-ALS. All calibrations were successfully applied to the analysis of an extended-release pharmaceutical formulation without any interference from additives. However, DIP showed some instability in the stored dosage form solution.

References

(1) Görög, S. (2000) Identification and Determination of Impurities in Drugs, Vol. 4, Elsevier Science, New York, NY

(2) Nageswara Rao, R., & Nagaraju, V. (2003) J. Pharm. Biomed. Anal. 33, 335–377. doi:10.1016/S0731-7085(03)00293-0

(3) Li, M. (2012) Organic Chemistry of Drug Degradation, Royal Society of Chemistry, London, United Kingdom, pp 150–164

(4) Gebel, J.M. (2005) J. Am. Coll. Cardiol. 46, 752–755. doi:10.1016/j.jacc.2005.04.058

(5) Brunton, L., Lazo, J., & Parker, K. (2006) Goodman & Gilman’s The Pharmacological Basis of Therapeutics, McGraw-Hill, New York, NY

(6) Moffat, A., Osselton, M., & Widdop, B. (2004) Clarke’s Analysis of Drugs and Poisons, Pharmaceutical Press, London, United Kingdom

(7) Diener, H.C., Cunha, L., Forbes, C., Sivenius, J., Smets, P., & Lowenthal, A. (1996) J. Neurol. Sci. 143, 1–13. doi:10.1016/S0022-510X(96)00308-5

(8) Chien, Y.W., Cabana, B.E., & Mares, S.E. (1982) Novel Drug Delivery Systems: Fundamental, Development Concepts and Biomedical Assessments, Marcel Dekker, Inc., New York, NY, pp 126–132

(9) United States Pharmacopeia (2007) 30th Ed., Vol. 2, United States Pharmacopeial Convention, Inc., Rockville, MD

(10) Electronic Medicines Compendium, ASASANTIN® Retard, https://www.medicines.org.uk/emc/medicine/273 (accessed on March 18, 2016)

(11) Patil, A.V., Mohanty, B.B., Ravinder, K., Bakshi, G., Singh, G., Bhushan, I., & Mohan, M.S. (2008) Dipyridamole pharmaceutical compositions, European Patent EP 1894561 A1, Dr Reddy’s Laboratories, Ltd, Telangana, India

(12) Umapathi, P. (1994) Int. J. Pharm. 108, 11–19. doi:10.1016/ 0378-5173(94)90411-1

(13) Umapathi, P., Parimoo, P., Thomas, S.K., & Agarwal, V. (1997) J. Pharm. Biomed. Anal. 15, 1703–1708. doi:10.1016/ S0731-7085(96)01918-8

(14) Rajput, A.P., & Sonanis, M.C. (2011) Int. J. Pharm. Pharm. Sci. 3, 156–160

(15) Prakash, K., Kalakuntla, R.R., & Sama, J.R. (2011) Afr. J. Pharm. Pharmacol. 5, 244–251

(16) El-Yazbi, F.A., Abdine, H.H., Shaalan, R.A., & Korany, E.A. (1998) Spectrosc. Lett. 31, 1403–1414

(17) Hammud, H.H., El Yazbib, F.A., Mahrousc, M.E., Sonjib, G.M., & Sonjib, N.M. (2008) Open Spectrosc. J. 2, 19–28. doi:10.2174/1874383800802010019

(18) Zhou, L., Ping, Q., & Yang, L. (2003) Chin. J. Pharm. Anal. 23, 199–201

Table 4. Statistical comparison between the results obtained by the proposed methods and the official methods of analysis of ASP, DIP, and SAL and a reported HPLC method of DIP-I

Parameter PLS CRACLS MCR-ALS ANN

Official or reported methoda

ASPMean

recovery, %99.92 99.97 100.59 99.97 100.12

SD 0.286 0.109 1.442 0.263 0.727n 15 15 15 15 6Variance 0.082 0.012 2.079 0.069 0.528Student’s

t-test (2.09)b0.936 0.810 0.743 0.717

F (2.52)b 2.022DIP

Mean recovery, %

100.31 100.30 100.42 100.34 100.83


t-test (2.09)b1.655 1.718 0.573 1.579

F (2.52)b 0.354SAL

Mean recovery, %

100.08 100.08 99.70 100.09 100.16


t-test (2.09)b0.317 0.332 0.544 0.123

F (2.52)b 0.343DIP-I

Mean recovery, %

99.74 99.80 100.22 99.73 100.92


t-test (2.09)b1.783 2.074 0.978 1.202

F (2.52)b 0.710a For ASP, add 50.0 mL 0.5 N NaOH to 1.5 g ASP, boil for 10 min, and

titrate the excess NaOH with 0.5 N H2SO4 using phenolphthalein as indicator (9). For DIP, dissolve 450 mg DIP in 50 mL glacial acetic acid, stir for 30 min, add 75 mL acetone, and titrate with 0.1 N perchloric acid with potentiometric end point detection (9). For SAL, dissolve 500 mg SAL in 25 mL diluted neutral alcohol and titrate with 0.1 N NaOH using phenolphthalein as indicator (9). For DIP-I, use the HPLC method, which utilizes an Inertsil ODS-2 (150 mm × 4.6 mm) 5 μm column with mobile phase containing a gradient mixture of phosphate buffer pH 7.0 and methanol, a flow rate 1.0 mL/min, and a detection wavelength at 295 nm (20).

b The values in parentheses represent the corresponding tabulated values of t and F at P = 0.05.

00948-00956.indd 955 24/06/16 1:21 PM


(19) Venkata Subbaiah, B., Sree Ganesh, K.K., Vamsi krishna, G., Vyas, K., Vasu Dev, R., & Subramanyam Reddy, K. (2012) J. Pharm. Biomed. Anal. 61, 256–264. doi:10.1016/ j.jpba.2011.11.028

(20) Vaghela, B.K., Rao, S.S., & Reddy, P.S. (2012) Int. J. Pharm. Pharm. Sci. 4, 615–622

(21) Wang, N., Xu, F., Zhang, Z., Yang, C., Sun, X., & Li, J. (2008) Biomed. Chromatogr. 22, 149–156. doi:10.1002/bmc.909

(22) MATLAB® 7.0.1 (2004) MathWorks, Inc., Natick, MA(23) Otera, J., & Nishikido, J. (2010) Esterification: Methods,

Reactions, and Applications, 2nd Ed., John Wiley & Sons, Hoboken, NJ

(24) Martínez Galera, M., Gil García, M.D., & Goicoechea, H.C. (2007) Trends Analyt. Chem. 26, 1032–1042

(25) Riahi, S., Hadiloo, F., Milani, S.M.R., Davarkhah, N., Ganjali, M.R., Norouzi, P., & Seyfi, P. (2011) Drug Test. Anal. 3, 319–324. doi:10.1002/dta.235

(26) Shehata, M.A., Ashour, A., Hassan, N.Y., Fayed, A.S., & El-Zeany, B.A. (2004) Anal. Chim. Acta 519, 23–30. doi:10.1016/j.aca.2004.05.011

(27) Antunes, M.C., Simao, J.E.J., Duarte, A.C., & Tauler, R. (2002) Analyst (Lond.) 127, 809–817. doi:10.1039/b200243b

(28) Windig, W., & Guilment, J. (1991) Anal. Chem. 63, 1425–1432. doi:10.1021/ac00014a016

00948-00956.indd 956 24/06/16 1:21 PM

drug formulations and clinical methods chemometrics … · 2017. 2. 13. · 948 el-ragehy et al.:...

Documents