Two multivariate calibration prediction algorithms, principal component regression (PCR) and partial least squares (PLS), were applied to the spectrophotometric multi-component analysis of a drug containing antazoline hydrochloride (AN) and naphazoline hydrochloride (NP) without any chemical separation procedure. The absorbance data matrix was obtained by measuring the absorbance in the range of 200–400 nm at 1 nm intervals. Additionally, the optimum number of components for PCR and PLS analysis were determined. A series of synthetic solutions containing different concentrations of AN and NP were used to check the prediction ability of the PCR and PLS. The proposed methods were validated for their accuracy and precision, and they were successfully applied for the assay of AN and NP in a commercial pharmaceutical product. The average %recovery values were 103.2 ± 2.3 for AN and 100.8 ± 3.3 for NP in case of PCR, and 104.9 ± 0.73 and 101.8±3.5 for AN and NP, respectively, in case of PLS. No interferences were observed from the common excipients usually used. The results of the proposed methods were compared with those obtained using the reference HPLC method, and excellent agreement was obtained.
Naphazoline hydrochloride (NP), chemically known as [2-(naphthalene-1-yl-methyl)-4,5-18 dihydro-1H-imidazole hydrochloride], is a vasoconstrictor of relatively long-lasting action that acts on the α receptors of the vascular smooth muscle (1,2). Antazoline hydrochloride (AN), chemically known as N-benzyl-N-(4,5-dihydro-1H-imidazol-2-ylmethyl) aniline hydrochloride, is another imidazoline ligand that has alpha 2-adrenoceptor antagonistic properties. It acts as a histamine H1 receptor antagonist. The two active ingredients (Figure 1) have been marketed as combination drugs in eye drops (1,2). Several methods have been reported in the literature for the simultaneous determination of NP and AN in pharmaceutical products. The most-reported methods are zero-crossing first derivative and ratio derivative spectrophotometry (3–5). The main disadvantage of these methods is their dependence on instrumental parameters such as the speed of the scan and the slit width. A direct spectrophotometric method was reported for the simultaneous determination of NP and AN based on measuring the absorbance at two wavelengths, and the concentration of each in the mixture was calculated by solving for two simultaneous equations (6). The direct method is not suitable because it is subjected to spectral overlapping. Chromatographic methods (7,8) are specific, accurate, and free of any possible interferences. However, chromatographic techniques are time-consuming and require expensive instruments.
Multivariate spectral calibrations are receiving considerable attention as standard methods for quantitative spectral analysis. Factor analysis based algorithms, such as principal component regression (PCR) and partial least squares (PLS) regression, are considered very popular methods for multivariate calibration (9). Both PCR and PLS involve extracting spectral loadings and scores through data set decomposition followed by building the model using new variables. These multivariate algorithms are powerful tools that have been successfully applied to spectroscopic data for quantitative analysis. The popularity of PCR and PLS comes from their strength in overcoming classic statistical problems, such as collinearity, band overlapping, and variable interactions (9–12). The proposed spectrophotometric and multivariate calibration methods have many advantages over some classical analytical techniques, such as liquid chromatography–ultraviolet (LC–UV) and LC–mass spectrometry (LC–MS). Spectrophotometric techniques are simple, rapid, sensitive, and low-cost, making them ideal for this type of analysis (13,14). On the other hand, coupling spectrophotometric techniques with multivariate algorithms for data analysis enables the simultaneous determination of analytes without any pre-separation procedures. These data analysis methods provide such determinations, even in the case of severe spectral overlapping. This study aimed to use PCR and PLS to develop a suitable method for simultaneous spectrophotometric determination of AN and NP in synthetic mixtures and pharmaceutical products. The results are also compared to those obtained using high performance LC (HPLC) for the same samples.
An ultraviolet–visible (UV-vis) double-beam spectrophotometer from Thermo Fisher Scientific was equipped with 1.0-cm quartz cells and connected to a personal computer. PCR and PLS analyses were carried out using the PLS Toolbox 4.0 software (Eigenvector Research, Inc.) and Matlab 7.0.4 (Math Works).
An isocratic HPLC system from Knauer was used in this study (Knauer model-501 LC system). The system was attached to a variable wavelength UV detector, and the data were acquired and processed using Eurochrom-2000 software.
All chemicals and reagents were of high purity and used without further purification. Distilled deionized water was used throughout this study. Acetonitrile was HPLC-grade from May Baker. Ammonium acetate (BDH) and trimethylamine (Cambrian Chemicals) were 99%. Glacial acetic acid was 100% from Kock-Light, England. The active ingredients, NP and AN, were obtained from Medilide, Italy. Both are USP reference standards.
For the solutions, 250 mg/L and 100 mg/L stock solutions of NP and AN, respectively, were prepared. These stock solutions were used for preparing both the training and the validation sets of samples. The training set consisted of 20 binary mixture solutions of both NP and AN as indicated in Table I (samples 1–20). The validation set consisted of six binary mixture solutions of the above materials and is displayed in Table I (samples 21–26). These ranges of concentrations were selected to match the levels of antazoline and naphazoline in the real sample.
A pharmaceutical eye drops commercial product containing 5.0 mg/mL AN and 0.25 mg/mL NP were purchased from a local pharmacy in the city of Irbid. Three test samples were prepared from this drug as a test set to evaluate both the PCR and PLS constructed models. The test samples are displayed in Table II.
The UV–vis absorbance values of all solutions, including the training, validation, and test sets, were recorded within the wavelength range of 200–400 nm at 1.0 nm intervals. All spectra were background corrected.
The HPLC separation was achieved using the method adopted by Sa Sa and others (7). The mobile phase, consisting of acetonitrile:water:triethylamine (40:59.75:0.25 v/v) and adjusted to pH = 4, passed through a reverse-phase (RP) C8 column (15 cm x 4.0 mm i.d., 5 µm particle size). The mixture was then filtered through cellulose nitrate membrane filters (0.45 µm x 47 mm) and degassed for 5 min by an ultrasonic shaker for 5 min before use.
Figure 2 shows the absorption spectra of NP and AN compounds, and their mixture in the spectral range of 200–500 nm.
A close examination of Figure 2 indicates that the spectra for AN and NP are completely overlapped in the 260–340 nm region and moderately overlapped in the 200–260 nm region. In the shown spectra, clear overlapping spectra of two pure compounds can be seen. This makes simultaneous determination of the related compounds in samples not possible by using the classical spectrophotometric methods. The current study focused mainly on the quantitative resolution of the binary mixtures of NP and AN by using PCR and PLS chemometric approaches without any separation step and graphical treatment.
For the application of the PCR and PLS calibration models, a concentration set of 20 binary mixtures of the two compounds in the range of 60–150 mg/L for AN and 2.0–40 mg/L for NP were prepared in water. All concentration combinations of both analytes are displayed in Table I. The absorption values of spectra of the concentration set were measured at the 302 wavelength points with the interval of ∆λ = 1.0 nm in the spectral region of 200–500 nm. The concentration set and absorption data were considered as the y-block (20 x 2) and x-block (20 x 302) for the construction of PCR and PLS chemometric calibrations.
The calibration PCR and PLS models were constructed after auto-scaling as a preprocessing step. The cross-validation method was applied through the leave-one-out procedure. To determine the number of components for each model, the RMSEC (root mean square error of calibration) values were calculated for the first 10 components using equation 1 (15) and displayed in Figure 3.
The optimum number of components was determined so that additional components cannot be counted unless they improve the RMSEC by at least 2%, in addition to a careful visual investigation of the constructed models. As a result, five components were used in the case of the PLS model while six components were employed for the PCR mode:
where ci is the reference concentration for the i-th sample and represents the estimated concentration. The resulted equation and the regression coefficients (R2) for both NP and AN of the calibration set are summarized in Table III.
For validation purposes of the constructed PCR and PLS models, samples 21–26 in Table I were used as an independent data set. Plots of the measured versus predicted values of both NP and AN using the PCR and PLS calibration models were created. The %recovery values were 98.8 ± 2.4 and 101.5 ± 2.2 for AN and NP, respectively, upon applying the PCR algorithm and when applying the PCR analysis, these values were 102.1 ± 2.2 for AN and 103.5 ± 3.1. The obtained R2, percentage recoveries, and relative standard deviations (RSD) were displayed in Table IV. From these numerical data, the recovery values were found very satisfactory for the validity of both PCR and PLS.
To assess the reliability of the current analysis method, commercial pharmaceutical eye drop was analyzed. This drug contains 5.0 mg/mL AN and 0.25 mg/mL NP. Three binary concentration combinations of both NP and AN were prepared from the eye drops product and analyzed using the above PCR and PLS models. Table II shows the concentration composition of these real samples. The results of applying both PCR and PLS to the test data are summarized in Table IV. The %recovery values for AN and NP were 103.2 ± 2.3 and 100.8 ± 3.3, respectively, in case of PCR. On the other hand, upon applying the PLS algorithm to the test samples, the %recovery values were 104.9 ± 0.73 for AN and 101.8 ± 3.5 for NP. The mean of the percent recovery and standard deviation values reflect the good prediction ability of both PCR and PLS models for the determination of NP and AN, simultaneously in a real drug using the spectroscopic data without any pre-separation step. For external validation purposes, the test set samples were analyzed HPLC and the results are summarized in Table V. The results obtained by the PCR and PLS methods were compared with those found by the HPLC procedure for using the one-way ANOVA test at 95% confidence level. Results indicated that there are no significant differences between the means obtained by the three methods (α < 0.05).
In the current work, simple UV-vis spectra were analyzed by two chemometric procedures, PCR and PLS. In this proposed method, simultaneous determination of both NP and AN in a real drug was achieved efficiently without any prior separation procedure. The average %recovery values for AN ranged from 98.8 ± 2.4 to 103.2 ± 2.2 and from 100.8 ± 3.3 to 101.5 ± 2.2 for NP when analysis by the PCR algorithm. And in case of PLS analysis the %recovery values ranged from 102.1 ± 2.2 to 104.9 ± 0.73 for the AN and from 101.8 ± 3.5 to 103.5 ± 3.1 for the NP. Analytical figures of merit were used to evaluate the quality of the proposed analysis methods throughout all stages of the study. The calculated mean percentage recovery and the relative standard deviation values of NP and AN indicated very satisfactory results using PCR and PLS methods. In addition, statistical tests indicated that no significant difference was found between the results obtained by the proposed chemometric methods and the HPLC method.
(1) Wisher, D. Martindale: The Complete Drug Reference. 37th ed. J. Med. Libr. Assoc. 2012, 100 (1), 75–76. DOI: 10.3163/1536-5050.100.1.018.
(2) Gumustas, M.; Alshana, U.; Ertas, N.; et al. Determination of Antazoline and Tetrahydrozoline in Ophthalmic Solutions by Capillary Electrophoresis and Stability-indicating HPLC Methods. J. Pharm. Biomed. Anal. 2016, 124, 390–398. DOI: 10.1016/j.jpba.2016.02.032
(3) Al-Momani, I. F. Simultaneous Determination of Naphazoline Nitrate and Antazoline Sulphate in Pharmaceutical Preparations by Derivative Ultraviolet Spectroscopy. Orient. J. Chem. 2001, 17 (1), 17–22.
(4) Souri, E.; Amanlou, M.; Farsam, H.; Afshari, A. A Rapid Derivative Spectrophotometric Method for Simultaneous Determination of Naphazoline and Antazoline in Eye Drops. Chem. Pharm. Bull. 2006, 54 (1), 119–122.
(5) Hajian, R.; Shams, N.; Kaedi, I. Application of Ratio Derivative Spectrophotometry for Simultaneous Determination of Naphazoline and Antazoline in Eye Drops. E-J. Chem. 2010, 7 (4), 1530–1538.
(6) Hman, S. O. Direct Determination of Antazoline and Naphazoline in Mixtures. Drug Dev. Ind. Pharm. 1987, 13 (7), 1257–1265. DOI: 10.3109/03639048709116233
(7) Sa Sa, S. I.; Al-Momani, I. F.; Jalal, I. M. Determination of Naphazoline Nitrate and Antazoline Sulphate in Pharmaceutical Combinations by Reversed-phase HPLC. Anal. Lett. 1990, 23 (6), 953–971. DOI: 10.1080/00032719008053438
(8) Ali, A.; Farooq, U.; Ahmed, M.; et al. Stability Indicating UHPLC-PDA Assay for Simultaneous Determination of Antazoline Hydrochloride and Naphazoline Hydrochloride in Ophthalmic Formulations. Acta Chim. Slov. 2017, 64, 332–341. DOI: 10.17344/acsi.2017.3166.
(9) Martens, H.; Naes, T. Multivariate Calibration. Wiley, Chichester, 1989.
(10) Boyce, M. C. Determination of Additives in Food by Capillary Electrophoresis. Electrophoresis 2001, 22 (8), 1447–1459. doi.org/10.1002/15222683(200105)22:8<1447::AID-ELPS1447>3.0.CO;2-%23.
(11) Deforce, D. L.; Ryniers, F. P. K.; Van den Eeckhout, E.; et al. Analysis of DNA Adducts in DNA Hydrolysates by Capillary Zone Electrophoresis and Capillary Zone electrophoresis−Electrospray Mass Spectrometry. Anal. Chem. 1996, 68, 3575–3584. DOI: 10.1021/ac9600013.
(12) Geldart, S. E.; Brown, P. R. Analysis of Nucleotides by Capillary Electrophoresis. J. Chromatogr. A. 1998, 828, 317–336. DOI: 10.1016/s0021-9673(98)00633-5.
(13) Aktaş, A. H.; Kitiş, F. Spectrophotometric Simultaneous Determination of Caffeine and Paracetamol in Commercial Pharmaceutical by Principal Component Regression, Partial Least Squares and Artificial Neural Networks Chemometric Methods. Croat. Chem. Acta. 2014, 87 (1), 69–74. DOI: 10.5562/cca2214
(14) Khoshayand, M. R.; Abdollahi, H.; Shariatpanahi, M.; et al. Simultaneous Spectrophotometric Determination of Paracetamol, Ibuprofen and Caffeine in Pharmaceuticals by Chemometric Methods. Spectrochim. Acta Part A, 2008, 70, 491–499. doi.org/10.1016/j.saa.2007.07.033
(15) Despagne, F.; Massart, D. L.; de Noord, O. E. Optimization of Partial-Least-Squares Calibration Models by Simulation of Instrumental Perturbations. Anal. Chem. 1997, 69, 3391–3399. DOI: 10.1021/ac970228d
Safwan M. Obeidat and Idrees F. Al Momani are with the Department of Chemistry at Yarmouk University, in Irbid, Jordan. Direct correspondence to: Safwan@yu.edu.jo and Safobeidat@yahoo.com. ●
Emerging Leader Highlights Innovations in Machine Learning, Chemometrics at SciX Awards Session
October 23rd 2024Five invited speakers joined Joseph Smith, the 2024 Emerging Leader in Molecular Spectroscopy, on stage to speak about trends in hyperspectral imaging, FT-IR, surface enhanced Raman spectroscopy (SERS), and more during the conference in Raleigh.
Advancing Forensic Science with Chemometrics: New Tools for Objective Evidence Analysis
October 22nd 2024A review by researchers from Curtin University comprehensively explores how chemometrics can revolutionize forensic science by offering objective and statistically validated methods to interpret evidence. The chemometrics approach seeks to enhance the accuracy and reliability of forensic analyses, mitigating human bias and improving courtroom confidence in forensic conclusions.