Abstract
Objective:
The aim of present study was to establish near infrared-chemometric methods that could be effectively used for quality profiling through identification and quantification of amoxicillin (AMOX) in formulated capsule which were similar to commercial products. In order to evaluate a large number of market products easily and quickly, these methods were modeled.
Materials and Methods:
Thermo Scientific Antaris II near infrared analyzer with TQ Analyst Chemometric Software were used for the development and validation of the identification and quantification models. Several AMOX formulations were composed with four excipients microcrystalline cellulose, magnesium stearate, croscarmellose sodium and colloidal silicon dioxide. Development includes quadratic mixture formulation design, near infrared spectrum acquisition, spectral pretreatment and outlier detection. According to prescribed guidelines by International Conference on Harmonization (ICH) and European Medicine Agency (EMA) developed methods were validated in terms of specificity, accuracy, precision, linearity, and robustness.
Results:
On diffuse reflectance mode, an identification model based on discriminant analysis was successfully processed with 76 formulations; and same samples were also used for quantitative analysis using partial least square algorithm with four latent variables and 0.9937 correlation of coefficient followed by 2.17% root mean square error of calibration (RMSEC), 2.38% root mean square error of prediction (RMSEP), 2.43% root mean square error of cross-validation (RMSECV).
Conclusion:
Proposed model established a good relationship between the spectral information and AMOX identity as well as content. Resulted values show the performance of the proposed models which offers alternate choice for AMOX capsule evaluation, relative to that of well-established high-performance liquid chromatography method. Ultimately three commercial products were successfully evaluated using developed methods.
KEY WORDS: Amoxicillin, chemometric, diffuse reflectance, method development, near infrared
Quality control and assurance of the pharmaceutical formulations possess a significant role specifically when formulations are manufactured. Their quality screening and control has to be performed in order to provide safe and genuine product to the public. Even after approval, they must go for periodical evaluation. To solve the purpose thin layer chromatography, infrared spectroscopy (IR), high performance liquid chromatography (HPLC), mass spectrometry (MS), colorimetric methods, dissolution assay and visual inspection are considered as well established and pharmacopoeial recognized techniques for transitional development and routine quality estimation.[1,2] However, there are certain complications with most of these pharmacopoeial approaches such as the resource intensive and invasive nature of the analysis, consumption of organic and inorganic solvents, require sample preparation and generate hazardous waste. For these reasons, efficient analytical approaches with rapid, reliable, and non invasive nature are required primarily for routine drug analysis specifically for a large number of samples. Hence to overcome such difficulties fourier transform (FT) near infrared spectroscopy (NIRS) has already been proven as a significant analytical technology not only for pharmaceutical product quality estimation but also for detection of counterfeit or spurious drugs at a very low cost with little or no sample preparation.[1] Additionally it can be used in a broad diversity of applications such as in food and agriculture industries, raw material, and biomedical analysis.[3,4,5,6,7] Being a nondestructive approach, it is also very advantageous for particle size determination, hardness, and moisture content of the samples.[8] In short, this technique requires no large sample, produce no waste, and it minimize analysis time from hours to minutes.
Development of NIRS model requires chemometric analysis in combination with NIRS that defines the scope of NIRS procedure used for the intended purpose.[8] Chemometric like principle component analysis (PCA), partial least square (PLS), principal component regression (PCR), discriminant analysis (DA), and Soft Independent Modeling of Class Analogy (SIMCA) are some of the widely used methods.[1,9] Pretreatment of the spectrum are prerequisite for accurate and precise result thus the most common preprocessing approaches such as first and second derivative transformation, Savitzky–Golay filtering, Norris derivative filtering, standard normal variate (SNV), multiplicative scatter correction, peak ratio or normalization, and sometimes spectral subtraction or a combination are some fundamental necessity.[3,10] These are excellent for differentiation between samples and spectral differences.[11] However, they require expertise and supervision to avoid unstable calibration model.
European Medicine Agency has published guidelines for the NIRS calibration and validation, otherwise except this no harmonized regulations are established or drafted.[8] However, NIRS has already been incorporated as an established monograph in United State Pharmacopoeia, European Pharmacopoeia and Indian Pharmacopoeia (IP). In spite of its extensive acceptability in pharmaceutical industries, it lacks specific monograph of pharmaceuticals for a multivariate method of quantification.[5] And specifically this technique remains largely unexplored in India notably in drug products quality testing.
Antibiotics medicines are life saving commodities since they cure, treat, prevent and mitigate the infectious condition, thereby they are safeguarding the public health. Antibiotics like ampicillin, amoxicillin (AMOX), co-trimoxazole, gentamicin, erythromycin, and ciprofloxacin are the most counterfeited products globally.[12] Globally, AMOX is among the most prescribed drugs and produced at large scale.[13] Surprisingly it was at the top of the World Health Organization (WHO) list of 47 antibiotics in 2010 being the most counterfeit active ingredient in the world.[14] Poor quality, insignificant amount or no active pharmaceutical ingredient (API) in drug products tend to cause grievous consequences such as mortality and morbidity.[15] Such situation also accounts for drug resistance and adverse clinical outcomes such as lack of therapeutic effects, treatment failure, toxicity, and side effects.[12,16] On such grounds, monitoring and quality profiling of large numbers of pharmaceutical products by fast, efficient and inexpensive analytical method are highly demandable for drug regulatory authorities.
Dry formulations can be evaluated noninvasively on reflectance mode[1] while liquid formulations on transmission mode using NIRS.[17] Contrary to transmission mode, reflectance measurement impart the remarkable bulk chemical information with small particles size and also with lambertian (diffuse reflectors) surfaces; it is mainly because it is not predominantly affected by surface scattering or reflectance losses due to the exclusion of major portion of the specular component.[18] Many researchers have been worked on NIR-chemometric models for the qualification and quantitation of AMOX,[5,6,13,19] however qualitative identification model which can identify AMOX in capsule powder means formulations which are analogous to commercial market products; and quantitative analysis of same API in these formulations has not been reported previously so that it could be used for market products evaluation. Therefore using our developed spectral library, we designed a model based on diffuse reflectance that can be used for AMOX identification and quantification in capsule formulations which were very similar to commercial products. Adding new spectra of new AMOX product or batch will update calibration model which will facilitate an easy and direct comparison between different products without the use of reference samples in future.
Evaluation with conventional HPLC method requires sample preparation and is expensive, sample destructive, and time consuming.[20] Therefore, a NIR-chemometric method is intended to be an alternative of the reference HPLC method for its reliability and time saving features. Additionally to provide the robustness, some generic AMOX commercial capsules were evaluated as independent validation samples using the predictive model.
Materials and Methods
Materials, instruments and softwares
Amoxicillin trihydrate, magnesium stearate, and croscarmellose sodium were provided as gift samples by Ranbaxy Laboratories Ltd., India, while microcrystalline cellulose-avicel and colloidal silicon dioxide-aerosil was provided as gift samples by FMC Biopolymer Brussels, Belgium and Evonik Industries, Germany, respectively.
Amoxicillin trihydrate reference standard (RS) and cefadroxil RS were directly procured from Sigma Aldrich. HPLC grade acetonitrile (Lichrosolv), potassium dihydrogen phosphate (LiChropur), potassium hydroxide guaranteed reagent grade were procured from Merck (India). Polytetrafluoroethylene filter of 0.45 µm and nylon filter of 0.20 µm pore size from Millipore system (Millipore Inc., USA) was used throughout the HPLC analysis. All AMOX commercial samples used in this study were purchased over the counter from the open market of Northern India.
A Thermo Scientific Antaris II FT-NIR analyzer equipped with an integrating sphere coupled with indium gallium arsenide detector was employed to generate diffuse reflectance spectra, and data were collected by means of inbuilt RESULT software. A HPLC (Waters, Milford, MA, USA) equipped with Alliance 2695 separations module and 2996 photodiode array detector was used in reference analysis using octadecylsilane bonded C-18 (250 mm × 4.6 mm, 5 µm) column. All samples during analysis were weighed using TB-215D (Denver Instrument, Germany) analytical balance.
All chemometric data analysis and modeling were carried out using Chemometric Software Package TQ Analyst 7.2.0 (Thermo Fisher Scientific Inc., MA, USA), along with the R Version 3.0.3 and Statistical Software MATLAB version 7.6.0 (MathWorks, Natick, USA).
Methods
Capsule content formulation
Capsule formulation was comprised AMOX trihydrate equivalent to AMOX used as an API in a range of approximately 50–110% of 250 mg AMOX. This range was grounded on to have more formulations and maximum variability. Total content was fixed at 400 mg based on commercial capsule size available in the market for 250 mg AMOX that is one (hard gelatin capsule size). Label claim with 100% of API, in this case, was 250 mg AMOX making this a high-dose capsule. Four common excipients for AMOX capsule were microcrystalline cellulose, magnesium stearate, croscarmellose sodium, and colloidal silicon dioxide and their selection was based on a short survey of the United States National Library of Medicine portal where various AMOX capsules belong to different companies including the Indian Companies with their ingredient were mentioned,[21] and their added quantity was derived from available monographs.[22]
AMOX and added excipients at different quantity were formulated by quadratic mixture modeling[23] and blended. A NIR preliminary study was accomplished to determine the optimum number of scans for each sample and number of latent variables. This study proclaimed 32 scans and for maximum variability it entailed six latent variables (LVs). Therefore as per the ASTM guidelines for sample set during method development which define 6× (number of LVs + 1) sample for calibration and 4× (number of LVs) for validation set were opted out.[5] Samples were formulated accurately, providing inconsiderable variability to create a stable calibration model. A total of 78 compositions were prepared by weighing appropriate amount of ingredients on highly sensitive analytical balance stored in 10-mL scintillation vials and ordered mixing were done manually. Samples were formulated accurately, providing inconsiderable variability to create a stable calibration model.
Near infrared spectroscopy data acquisition
This study was accomplished in three stages; model development with optimization, validation and model applicability on real samples. Method development with optimization was determined with a set of 50 calibration standards known as a training set and validated with 28 validation standards known as test set containing the same original AMOX range as in calibration samples.
A placebo of all excipient in equal amount was also formulated, and the total content of each formulation was divided into three aliquots. Simultaneously spectra were recorded in triplicate using integrating sphere in diffuse reflectance mode at 8 cm−1 resolution over the spectral range of 4000–10,000 cm−1. To avoid the error during NIRS, each spectrum was acquired after shaking or whirling of the sample vial for mixture homogeneity as per recommendations,[18] with 32 scans corresponding to measurement time on few seconds. Hence, 711 spectra were procured totally on account of 78 formulations including placebo. Figure 1 shows the pure NIR raw spectrum of AMOX and a similar molecule cefadroxil and all excipients except colloidal silicon dioxide.
Figure 1.
Full range raw near infrared spectroscopy mean spectra (a) amoxicillin active pharmaceutical ingredient, (b) microcrystalline cellulose, (c) magnesium stearate, (d) croscarmellose sodium, (e) placebo, (f) cefadroxil reference standard
Before developing chemometric model, a standard analysis of variance between two randomly selected calibration standards was checked for feasibility, and a high value of F-ratio showed sufficient variation between the samples thus allowed us to continue for method development. Every time before scanning, powdered sample in the vial was rotated and not tapped, as tapping can cause particles segmentation which may give greater density at the bottom of the powder sample and hence increase the reflectance.
Reference Method
IP prescribes IR and HPLC for identification and assay respectively for AMOX trihydrate product.[2] Thus being highly sensitive; HPLC was preferred for both identification and assay. The peak homogeneity of each chromatogram was expressed in terms of peak purity values. Resulted assay values were used as reference values in NIR-chemometric model development.
Chemometric method development for identification
On the ground of mahalanobis distance, kennard-stone algorithm[24] was used for selecting 50 calibration standards out of 78 formulations as shown in Figure 2 using the prospectr package in R Software (The Comprehensive R Archive Network).[25] For developing AMOX identification model DA algorithm was applied on 50 standards, a mean spectrum of pure AMOX API and six mean spectra of placebo. However, three outliers were observed, and the model was optimized after removing two outliers except placebo, as it was included by choice to make the model more accurate.
Figure 2.
Determination of calibration/training set (blue solid circles) using kennard-stone algorithm for 50 samples out of 78
Chemometric method development for quantitation
Same 50 calibration standards were utilized for quantitative analysis using two well known factor analysis based multivariate techniques that are PCR and PLS. Each spectrum was pretreated with various data processing techniques. Initially a calibration curve between reference value and NIR predicted values was plotted using six factors and resulted in 0.9861 of correlation coefficient with root mean square error of calibration (RMSEC) of 3.68% which are not justifiable. Like identification model, there were three spectral leverages found in this evaluation as done by the Chauvenet test[26] as shown in Figure 3, where except placebo two other outliers were removed.
Figure 3.
Outlier detection plot for calibration/training set. High leverage value shows outlier
For applying chemometric on spectral data for creating the calibration; spectral pretreatment methods and wave number range of interest must be selected. Being the multivariate method, the entire range found to be rich in information was considered. However, during optimization of the model, regions of interest were selected accordingly.
Several chemometric models were developed, as mentioned in electronic supplementary material (ESM) data Table S1, to demonstrate the ability of NIRS. Model accomplished covering multivariate algorithm such as PLS and PCR followed by leave one out cross validations. Based on the P value of intercept and slope along with minimum RMSEC and root mean square error of prediction (RMSEP), many models were filtered out and ultimately a PLS model finally opted based on latent variable, root mean square error of cross-validation (RMSECV) value. Therefore the method, in this case, used a PLS algorithm with SNV as scattering correction. It is feasible to identify from second derivative spectra, two regions of high correlation, especially located from 5064 cm−1 to 5253 and 8573 cm−1 to 8674 cm−1 (ESM Figure S1). The first region corresponds to the fundamental combination band of OH stretching and CH bending while the other region corresponds to a second overtone of C-H bending. Thus, this model was not only based on purely mathematical calculation but was an artifact corresponds to the notable spectral region.
Table S1.
Linear regression parameter with P value and RMSE values of the calibration models for optimization
Figure S1.
Region of interest (a) and (b) in the second-derivative spectra of calibration/training set
Chemometric method validation for identification
For validation, samples must be independent and cannot be included in the training set during development. Samples used as independent validation set were divided into two sets; internal validation set or test set, and external validation set.
The test set was comprised 28 samples. And since ampicillin classifies as penicillin; and cefadroxil belongs to cephalosporin family which is much similar to penicillin class, therefore, external validation set included the commercial product of each ampicillin and cefadroxil along with three AMOX commercial products. In addition, for the specificity of the model, cefadroxil RS was included in the external validation set as the negative control to design a robust model.
Chemometric method validation for quantitative analysis
There were 28 samples used as the internal validation sample. In accordance with the recommended validation parameters by EMA and ICH guidelines like specificity, linearity, accuracy, precision, and robustness; demonstration for quantitative analysis were covered using five independent formulations and three commercial real capsules as external validation set.
Results and Discussion
Identification model development and validation
The detection and the subsequent identification of AMOX relied heavily on the overtone region of the NIRS spectra. Using the DA algorithm mean spectrums of three aliquots for each sample in triplicate were calculated and distribution model generated by estimating the variance at each data point of spectrum in the range of analysis. Identification model was optimized on 48 samples training set. Principle component (PC) 1 to PC2 accounted for 87.71% of the total variance in the data, with PC3 and PC4 accounting for the 10.87% and up to PC6 data describe the 99.8% variability.
Specificity of the model was evaluated with cefadroxil RS as a negative control. The result suggests that the proposed model was well efficient to differentiate AMOX from cefadroxil and ampicillin also as depicted in Figure 4.
Figure 4.
Identification of amoxicillin sample with respect to distance to amoxicillin-based on discriminant analysis algorithm
Quantitative model development and validation
NIRS quantitative method was established to explore the quantity of AMOX in AMOX trihydrate capsule formulations. A quantitative model was optimized on 48 samples training set using PLS with latent variable up to four which shows maximum spectral information as shown in Figure 5a and b. However, before removal of outliers LVs were six. As a result, four LVs a good correlation coefficient was found to be 0.9937 with RMSEC 2.17%. In this context, Figure 6a shows calibration curve and Figure 6b shows the residual for the training set. The accepted PLS models were eventually used to predict the AMOX in 28 samples test set. The accuracy of the NIRS prediction equation was then evaluated using linear regression analysis between NIRS-predicted values and those acquired by the reference HPLC method. Therefore, RMSEP of 2.38% was estimated which indicate a high degree of correlation in this method. This correlation between the spectra and reference values was examined using established PLS method and was optimized by cross-validation. With cross-validation, each sample was eliminated one at a time from the training set, then a new calibration was executed, and a predicted score was determined for the removed sample. This mechanism was repeated until every sample had been omitted once using the leave one out method. In this case to determine the model over fitting, minimum predicted residual sum of squares (PRESS) and RMSECV of 2.43% were estimated with factor four.
Figure 5.
Scatter Score plot for (a) partial least square-1 partial least square-2 and partial least square-3 (b) partial least square-4 partial least square-5 partial least square-6 obtained for amoxicillin data set
Figure 6.
(a) Partial least square correlation plot (b) Partial least square residual plot for the training set
The proposed quantitative model was validated using parameters usually recommended such as accuracy, precision, linearity, and robustness in accordance with the International Conference on Harmonization (ICH) and EMA guidelines.
Accuracy
Accuracy of the proposed model was obtained by performing a comparison of NIR predicted data with the reference HPLC data that was shown to be accurate conducted at low, medium, and high AMOX amount which is approximately 70%, 90%, and 110%, respectively, of the label claim. Table 1 show the trueness of the predicted values with a reference method. In paired t-test with eight degrees of freedom, the obtained experimental t-stat value was smaller than the critical t-value; and high P - value than 0.05 alpha value shows that there is no significant difference between the proposed and referenced result with a 95% confidence level. These data show that the percentage difference between the validated NIR method and reference method is insignificant for assay and that there is a good correlation between the two methods with respect to AMOX. As a result, the proposed NIR method can be accounted as a suitable alternative to the reference HPLC method for the evaluation of AMOX capsules. Moreover, there is no official procedure for validating NIR accuracy as such; thus it may be evaluated through RMSEC, RMSEP, and RMSECV which also known as figure of merit as shown in Table 2.
Table 1.
Accuracy evaluation between proposed and reference method
Table 2.
Validation parameters for evaluating the proposed NIR model
Precision
Repeatability was evaluated by inter-day precision and was performed with the same analyst on three independent validation samples and applying the model to the same formulation samples three times on three consecutive days. In this study, the precision was estimated by three determinations I, II, and III at 80%, 90%, and 110%, respectively, in triplicate as mentioned in Table 2. Less than 2% relative standard deviation shows that the proposed method may be considered as precise in relation to HPLC method.
Linearity
Linearity correlation between the predicted values from selected PLS model and the reference value of three determinations in triplicate as 70%, 90%, and 110% was evaluated. Linear regression analysis was performed between the two used methods and the compatibility of results was estimated by general equation:
y = bx + a
where y is the NIR predicted value, x is reference value, a is intercept and b is the slope. For AMOX determination, the regression equation was y = 0.958x + 4.58, being the confidence interval for the slope (0.882; 1.035) and for the intercept (−2.35; 11.51) included 1 and 0, respectively, as mentioned in Table 2. Therefore within the AMOX range of the training set the proposed model for AMOX allowed a suitable linearity with r = 0.994 for all sample set when it is compared to the reference method as shown in Figure 7.
Figure 7.
Linearity plot for the calibration (◯) and validation (+) samples, solid line shows the data fit
Robustness
Robustness of the proposed method was assessed through the results of three determinations in triplicates of a 100% label claim AMOX sample at three different temperatures as shown in Table 2. Percent RSD was less than two and result were unaffected by small and deliberate temperature variations and found no significant difference between the predicted values and the reference result, upon paired t-test P - value greater than 0.05 shows reliability and a good agreement between both analytical methods.
Model assessment on real sample
The identification and determination of the content of three generic samples were carried out by applying the HPLC method reported in AMOX monograph of IP. NIR proposed identification and quantitative model was applied on same commercial products and obtained results were finally compared with the result estimated by the reference method. For identification based on DA; all three AMOX commercial sample found close in distance to AMOX formulated samples while others similar momolecules like cefadroxil and ampicillin commercial products were clearly distinguished as shown in Figure 3. In quantitative analysis, content assay estimation from proposed method and reference method was compared by applying paired t-test which shows NIR method to be an alternative and effective method as mentioned in Table 3.
Table 3.
AMOX market products evaluation by NIR proposed method
Conclusion
The present analysis supports the points raised in several research articles that an NIRS in combination with chemometric analysis can perform equal to or often with very small error than the reference method. Present NIR-chemometric models implied as good correlation with the reference HPLC method. Moreover, validation result verified that developed methods were as accurate as reference analytical technique. Thus, it has been shown a feasible alternative to HPLC for the identification and assay of AMOX capsules. These models emphasize the importance of NIRS and chemometric analysis because it is fast, nondestructive and can be employed to analyze solid sample with minimal or no sample preparation.
Pharmaceutical regulatory authorities are expected to provide alternate and quick techniques for routine analysis specifically for a large number of samples and therefore in the rapidly growing area of analytical method development NIRS have been introduced as a challenging field. The proposed model will be used for identifying and assaying a large number of AMOX products quickly and may be utilized for quality profiling of spurious and substandard medicines. It takes less than two minutes to analyze a sample once the calibration model has been set up. We hope this research will stimulate the quality assessment study of other antibiotics and other category of drug products in India and in other countries as well.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
Acknowledgement
The authors would like to gratefully acknowledge Dr. Narendra Kumar, Department of Mathematics and Mohammad Wajid, Department of Electronics and Communication, Jaypee University of Information Technology-Solan, India for statistical and Matlab support.
References
- 1.Martino R, Malet-Martino M, Gilard V, Balayssac S. Counterfeit drugs: Analytical techniques for their identification. Anal Bioanal Chem. 2010;398:77–92. doi: 10.1007/s00216-010-3748-y. [DOI] [PubMed] [Google Scholar]
- 2.6th ed. Ghaziabad: The Indian Pharmacopoeia Commission; 2010. Indian Pharmacopoeia Commission. Indian Pharmacopoeia. [Google Scholar]
- 3.Hart JR, Norris KH. Determination of the moisture content of seeds by near-infrared spectrophotometry of their methanol extracts. Cereal Chem. 1962;39:94–9. [Google Scholar]
- 4.Jamrógiewicz M. Application of the near-infrared spectroscopy in the pharmaceutical technology. J Pharm Biomed Anal. 2012;66:1–10. doi: 10.1016/j.jpba.2012.03.009. [DOI] [PubMed] [Google Scholar]
- 5.Silva MA, Ferreira MH, Braga JW, Sena MM. Development and analytical validation of a multivariate calibration method for determination of amoxicillin in suspension formulations by near infrared spectroscopy. Talanta. 2012;89:342–51. doi: 10.1016/j.talanta.2011.12.039. [DOI] [PubMed] [Google Scholar]
- 6.Qu N, Zhu M, Mi H, Dou Y, Ren Y. Nondestructive determination of compound amoxicillin powder by NIR spectroscopy with the aid of chemometrics. Spectrochim Acta A Mol Biomol Spectrosc. 2008;70:1146–51. doi: 10.1016/j.saa.2007.10.036. [DOI] [PubMed] [Google Scholar]
- 7.Blanco M, Coello J, Iturriaga H, Maspoch S, de la Pezuela C. Near-infrared spectroscopy in the pharmaceutical industry. Analyst. 1998;123:135R–50R. doi: 10.1039/a802531b. [DOI] [PubMed] [Google Scholar]
- 8.London, UK: European Medicine Agency; 2014. EMA. Guideline on the Use of Near Infrared Spectroscopy by the Pharmaceutical Industry and the Data Requirements for New Submissions and Variations; pp. 1–28. [Google Scholar]
- 9.Roggo Y, Chalus P, Maurer L, Lema-Martinez C, Edmond A, Jent N. A review of near infrared spectroscopy and chemometrics in pharmaceutical technologies. J Pharm Biomed Anal. 2007;44:683–700. doi: 10.1016/j.jpba.2007.03.023. [DOI] [PubMed] [Google Scholar]
- 10.Kemper MS, Magnuson EJ, Lowry SR, McCarthy WJ, Aksornkoae N, Watts DC, et al. Use of FT-NIR transmission spectroscopy for the quantitative analysis of an active ingredient in a translucent pharmaceutical topical gel formulation. AAPS PharmSci. 2001;3:E23. doi: 10.1208/ps030323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Anderson CA, Drennen JK, Ciurczak EW. Handbook of Near-Infrared Analysis. Boca Raton, FL, London and New York: CRC Press; 2008. Pharmaceutical applications of near-infrared spectroscopy; pp. 585–611. [Google Scholar]
- 12.Kelesidis T, Kelesidis I, Rafailidis PI, Falagas ME. Counterfeit or substandard antimicrobial drugs: A review of the scientific evidence. J Antimicrob Chemother. 2007;60:214–36. doi: 10.1093/jac/dkm109. [DOI] [PubMed] [Google Scholar]
- 13.Bittner LK, Heigl N, Petter CH, Noisternig MF, Griesser UJ, Bonn GK, et al. Near-infrared reflection spectroscopy (NIRS) as a successful tool for simultaneous identification and particle size determination of amoxicillin trihydrate. J Pharm Biomed Anal. 2011;54:1059–64. doi: 10.1016/j.jpba.2010.12.019. [DOI] [PubMed] [Google Scholar]
- 14.Delepierre A, Gayot A, Carpentier A. Update on counterfeit antibiotics worldwide; public health risks. Med Mal Infect. 2012;42:247–55. doi: 10.1016/j.medmal.2012.04.007. [DOI] [PubMed] [Google Scholar]
- 15.Bate R, Tren R, Mooney L, Hess K, Mitra B, Debroy B, et al. Pilot study of essential drug quality in two major cities in India. PLoS One. 2009;4:e6003. doi: 10.1371/journal.pone.0006003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Newton PN, Green MD, Fernández FM, Day NP, White NJ. Counterfeit anti-infective drugs. Lancet Infect Dis. 2006;6:602–13. doi: 10.1016/S1473-3099(06)70581-3. [DOI] [PubMed] [Google Scholar]
- 17.Jerry WJ, Lois W. 2nd ed. Boca Raton, Forida: CRC Press; 2012. Practical Guide and Spectral Atlas for Interpretive Near-Infrared Spectroscopy; p. 309. [Google Scholar]
- 18.Burns DA, Ciurczak EW, editors. Handbook of Near-Infrared Analysis. 3rd ed. Boca Raton, FL, London and New York: CRC Press; 2008. p. 814. [Google Scholar]
- 19.Parisotto G, Ferrão MF, Furtado JC, Molz RF. Determination of amoxicillin content in powdered pharmaceutical formulations using DRIFTS and PLS. Journal is Brazilian Journal of Pharmaceutical Sciences. 2005, 2007;43:89–96. [Google Scholar]
- 20.Gowen AA, O'Donnell CP, Cullen PJ, Bell SE. Recent applications of Chemical Imaging to pharmaceutical process monitoring and quality control. Eur J Pharm Biopharm. 2008;69:10–22. doi: 10.1016/j.ejpb.2007.10.013. [DOI] [PubMed] [Google Scholar]
- 21.DailyMed. Amoxicillin Capsule. U.S. National Library of Medicine. 2014. [Last cited on 2014 Feb 02]. Available from: http://www.dailymed.nlm.nih.gov/dailymed/search.cfm?query=amoxicillin .
- 22.Rowe RC, Sheskey PJ, Quinn ME, editors. Handbook of Pharmaceutical Excipients. 6th ed. London and Chicago: Pharmaceutical Press; 2009. p. 888. [Google Scholar]
- 23.Gabrielsson J, Lindberg NO, Lundstedt T. Multivariate methods in pharmaceutical applications. [Last accessed on 2015 Jan 10];J Chemom. 2002 16:141–60. Available from: http://www.dio.wiley.com/10/1002/cem’ by http://onlinelibrary.wiley.com/doi/10.1002/cem.697/pdf . [Google Scholar]
- 24.Kennard RW, Stone LA. Computer aided design of experiments. Technometrics. 1969;11:137–48. [Google Scholar]
- 25.Stevens A, Leornardo RL. An Introduction to the Prospectr Package. R Package Vignette, Report No.: R Package Version 0.1.3. 2014. [Last cited on 2011 Apr 20]. Available from: http://www.cran.r.project.org/web/packages/prospectr/vignettes/prospectr-intro.pdf .
- 26.Lin L, Sherman PD. Cleaning Data the Chauvenet Way. SouthEast SAS Users Group. Hilton Head, SC: Institute for Advanced Analytics; 2007. [Google Scholar]