Said 2011

International Journal of Pharmaceutics 415 (2011) 102 109

Contents lists available at ScienceDirect

International Journal of Pharmaceutics

journa l h omepa g e: www.elsev ier .com

Near-in meand UK se

Mazlina irea The School of b Faculty of Pha

a r t i c l

Article history:Received 1 MaReceived in reAccepted 21 MAvailable onlin

Keywords:NIRSSpectral databParacetamolMalaysiaCounterfeit medicinesSIMCAPCAQualitative studyChemometrics

t sournd qts by d to ales of tudy odel

tamo 95%gh th

sample set, our results indicated variability in the quality of the Malaysian tablets. A database of spectrawas created and search methods were evaluated for correct identication of tablets. The approach pre-sented here can be further developed as a method for identifying substandard pharmaceutical products.

2011 Elsevier B.V. All rights reserved.

1. Introdu

Challenghealthcare unregistereistered prodcounterfeit results of smtampering, meditag (Ba

Among and adultethroughoutexample of example, a (DCA) websregistered pthe main acmarket. Thethe quality presence of

CorresponE-mail add

0378-5173/$ doi:10.1016/j.ction

es faced by the pharmaceutical enforcement bodies insystems of developing countries includes an inux ofd products, adulterated products, adulteration in reg-ucts, adulteration in food and food-supplements, andmaterials. Problems faced by regulatory agencies areuggling, illegal entry, parallel importation, diversion,

repackaging, relabeling and the use of fake hologramte, 2008; Ismail, 2009; Wertheimer, 2008).these challenges, the issues of counterfeit medicinesration in pharmaceutical formulations are escalating

the globe. The Malaysian pharmaceutical industry is anproblems that a developing country may experience. Forrecent search on the Malaysian Drug Control Authorityite (DCA, 2010) showed that there are more than 250roducts containing paracetamol as the main or one oftive pharmaceutical ingredients (APIs) available on the

main efforts of the government agency are to ensureof these products on the market and to establish the

counterfeit medicine in Malaysia. The rst survey was

ding author. Tel.: +44 20 7753 5879; fax: +44 20 7753 5964.ress: [email protected] (M. Zloh).

conducted in 1997 on cough and cold medication which revealedthat 5% of such medicines on the market were counterfeit (Cruez,2006). Another survey conducted by the Pharmaceutical Associa-tion of Malaysia (PhAMA) between 1998 and 2002 showed that 14out of 289 samples (nearly 5%) of three prescription medicines pur-chased from a total of 196 pharmacies and clinics in 6 states werefound to be counterfeit (PAM, 2006). Additionally in 2008, 5.28%of over-the-counter products were identied as fake (Bernama).These percentages indicate that the prevalence of counterfeit drugswere consistent throughout almost a decade in this country.

In addition to this consistent problem, adulteration of herbalmedicines appeared to be on the increase. From 2006 to 2008, 17products indicated for mens health, 6 products for weight loss,5 cough medicines and 4 products for joint pain were found to beadulterated with synthetic drugs in order to increase their pharma-cological effects (ASEAN, 2007). This is potentially an exceptionallydangerous problem as herbal products are highly complex andconventional single chemical entity-herbal drug combination phar-macology is poorly understood.

Near-infrared (NIR) spectroscopy has gained wide acceptancewithin the industry for various pharmaceutical applications (Blancoet al., 1998; Luypaert et al., 2007), particularly for the identicationand detection of counterfeit and substandard medicines throughthe comparison of suspected tablets with the spectra of authentictablets (Assi et al., 2008, 2009; Moffat et al., 2010; ONeil et al.,

see front matter 2011 Elsevier B.V. All rights reserved.ijpharm.2011.05.057frared spectroscopy (NIRS) and chemo paracetamol tablets: A spectral databa

M. Saida,b, Simon Gibbonsa, Anthony C. Moffata, MPharmacy, University of London, 29/39 Brunswick Square, London WC1N 1AX, UKrmacy, National University of Malaysia, 43600 UKM, Bangi Selangor, Malaysia

e i n f o

rch 2011vised form 20 May 2011ay 2011e 27 May 2011

ase

a b s t r a c t

The inux of medicines from differena challenge to monitor their origin aspectra prevents the analysis of tableMalaysian pharmacies were comparespectroscopy (NIRS). Additional sampwith other actives were added to the sand compared by using multivariate mspectral database. All analysed parace1 out of 20 batches excluded from theclassied as outliers of the set. Althou/ locate / i jpharm

tric analysis of Malaysianstudy

Zloha,

ces into healthcare systems of developing countries presentsuality. The absence of a repository of reference samples ordirect comparison. A set of paracetamol tablets purchased in

similar set of sample purchased in the UK using near-infraredproducts containing ibuprofen or paracetamol in combinationas negative controls. NIR spectra of the samples were acquireding and classication algorithms (PCA/SIMCA) and stored in al samples contained the purported active ingredient with only

condence interval, while the negative controls were clearlye substandard products were not detected in the purchased

M.M. Said et al. / International Journal of Pharmaceutics 415 (2011) 102 109 103

2008). An alternative approach for drug identication by NIR is todetermine the APIs (Khan et al., 1997) or the source of the tablets(Yoon et al., 2004). Drug analysis has become easier and faster assample preparation is not required. The ability to analyse samplesintact is another advantage of the technique whereby samples canbe retained for further analysis or as evidence for forensic purposes.

Previous studies have discussed NIR analysis by quantitativeassay of intact tablets as the most obvious way to detect sub-standard preparations (Eustaquio et al., 1999; Trafford et al., 1999).These methods are based on developing calibration curves of activeingredient (Blanco and Peguero, 2010). In all these studies, theinformation about the tablets analysed were available and the anal-yses were designed based on this knowledge. However, one of thereal problems occurring in developing countries is a consequenceof medicines being sold or dispensed without original packaging;tablets are either wrapped in a piece of paper, or given in an enve-lope or a plastic bag, where the information about the tablets maybe given verbally or just hand written on the container (Goodman etal., 2007). Inical methodwhat is thedispensed m

This reseof these proproduct ideof NIR spegerprints ocomparisonence sampl

Qualitatnamely parmarkets, wof Malaysiacompared timported inusing two tsame time aacquired a the distributhe databas

2. Materia

2.1. Drug sa

A samplof paracetapurchased representeddient. Threeingredients

included into the test set as negative controls. After a year, a new setof tablets consisting of three new batches of the brands available inthe database and one additional batch of paracetamol from a differ-ent manufacturer were purchased from the Malaysian market andtheir NIR spectra were recorded. The additional spectra were addedto the database for identication purposes as test sets. Details of thesamples are listed in Table 1.

All samples were obtained by purchasing samples in pharmaciesand supermarkets since attempts to obtain the samples directlyfrom the manufacturers had either been ignored or rejected (forMalaysian samples). The Malaysian samples were supplied eitherin original blister packaging or as repacked products. All UK sampleswere dispensed in blister packs. Samples consist of uncoated tabletsof different shapes and colours.

2.2. NIR analysis

The NIR diffuse reectance analyses were carried out using atemsalyzlaceNIR hile

ns, ded ov

spea mo

ftwa

NIRftwacquisiginaonveectrare, CL ID

Suite

emo

Princ treags foe of ivari

PC mple

s andre popu

Table 1Details on the sed in

Purposes (batc

Calibration s

PCM 650 Negative con

Internal valiExternal valUnknown this scenario, the questions to be answered by analyt-s are more complex: is the correct medicine dispensed,

composition of the dispensed formulation and is theedicine counterfeit?arch intended to propose a way forward to solve someblems. We have designed a strategy that can be used forntication, screening and classication using a databasectra. The spectroscopic data reecting chemical n-f the products stored in a database will enable direct

between different products without the use of refer-es provided by manufacturers.ive analysis of the most common type of analgesic,acetamol tablets purchased in pharmacies and super-as used as an example in this work. Different brandsn paracetamol products were analysed by NIR ando a similar set produced in the UK. NIR spectra wereto the database and validation searches were conductedest sets, a set of paracetamol samples acquired at thes the training set and a new set of paracetamol samplesyear later. Chemometric analysis was used to observetion of samples in the score plot and its correlation toe search outcome.

ls and methods

mples

e set used in this study consisted of 16 and 6 batchesmol 500 mg tablets (except for one batch with 650 mg)in Malaysia and the UK, respectively. Each batch was

by 20 tablets with paracetamol as the only active ingre- additional products, that contained additional active

to paracetamol or did not have paracetamol at all, were

NIRSystent Anwere pin the times w32 scarecorda meanunder

2.3. So

Theysis Sodata athe orthen cThe spsoftwaSPECTA(Gram

2.4. Ch

2.4.1. PCA

loadinpurposa mult2005).the samclusterthe scosame p

samples used for product assessment, database development and validation discus

Active pharmaceutical ingredients (APIs) Amount

amples Paracetamol 500 15

6

Paracetamol 650 mg 1 trol Paracetamol 500 mg, dihydrocodeine tartrate 7.46 mg 1

Paracetamol 200 mg, aspirin 300 mg, caffeine 45 mg 1 Ibuprofen 200 mg 1

dation Paracetamol 500 mg (internal) 3 idation Paracetamol 650 mg (external 1) 3

Paracetamol unknown (external 2) 1 6500 spectrophotometer, equipped with a Rapid Con-er (FOSS NIRSystems, Silver Springs, USA). Intact tabletsd centrally on the sample stage over the light beaminstrument. Each side of the tablet was analysed four

rotating them at 90 each time. Eight accumulations ofata spacing every 2 nm of the reectance spectra wereer the wavelength between 1100 and 2500 nm to givectrum for each sample. The experiment was conductednitored humidity and ambient temperature.

re

instrument was controlled by Vision Spectral Anal-re for Windows (FOSS NIRSystems, version 2.11) forition. Chemometric analysis was conducted by takingl averaged NIR spectra, exported in ASCII format andrted to JCAM-DX les using an in-house programme.l les were then imported into the Unscrambler v9.7AMO (Oslo). The spectral database was created using

v9.0 database module implemented in the GRAMS ID Software; Thermosher).

metric methods

ipal component analysis (PCA)tment of spectra decomposed them into scores andr variables called principal components (PC). The mainPCA was to reduce the number of variables to representate data table in a low dimensional space (Esbensen,odels were constructed individually for each batch ofs. After the rst run, the presence of outliers, groups,

trends were determined based on the observation oflots. At this stage, the outliers detected belonged to thelation but they were poorly described by the model.

this work. A set of 20 tablets were analysed for each batch.

h/es) Source Sample label

Malaysia BG A, BG B, BG C, FP, IF, MIL A, MIL B,OR, PC, PG, PR, PM, PT, UP A, UP B

UK UK bts, UK lidl, UK mrs, UK vh, UKwtr, UK gsl

Malaysia UP CUK NC1 (paramol)UK NC2 (epr)UK NC3 (ibu)Malaysia IF, PM, UP AMalaysia IF, PM, UP AMalaysia pcm unknown

104 M.M. Said et al. / International Journal of Pharmaceutics 415 (2011) 102 109

The optimum number of PCs was determined based on the totalexplained variance plot. PCA was conducted on the data with lever-age correction as the validation method and the scaling factor wasset as 1.

2.4.2. Soft ISIMCA w

Classicatiosample as auation was

2.5. Databa

2.5.1. DatabRaw NI

database ufunctions wOther inforbrand and facturer namavailable) a

2.5.2. DatabBefore e

the databasing effect thusing the aubaseline err

A databaon a correlaof the unkndatabase afof the top m(HQI) valuespectra wit

3. Results

The rstngerprintsations in NIto the searc

3.1. NIR an

All of threlative to tmost of theNIR raw spea similar spbeing differThe main dconcerningtering effecsurface struabsorbancewith one oftrol sampleparacetamo

3.2. Princip

PCA waprovide mobatches. Thnot the am

batch from the Malaysian samples, identied as sample PM, was notincluded in the 95% condence interval set, indicating signicantlydifferent properties of the tablets in that batch, which may indi-cate different excipients or manufacturing process (Fig. 2). The PCA

lot iell ee ot

cond twplesctiveed iner, it

MCA

he Pe tohe msing n of ved. Aed froget0 mgdoled inat usd innto ca waomisd.

taba

Dataspects usa-axeeterctraed sped a. The

Datactralparerst quarn algcatiothermrroga

usedm ose usted a

givethe ht matsequmanuainst

hadndependent Modeling of Class Analogy (SIMCA)as applied in the original method (Candol et al., 1999).n was carried out on all of the PC models against one

class model with the signicance level set at 95%. Eval-made by using the distance versus leverage plot.

se creation and search

ase developmentR spectra were imported directly into Spectral IDsing NSAS format. The database has a smart-converthich automatically converted the NSAS le to .SPC les.mation computed together with the spectra was theproprietary name, batch number, expiry date, manu-e and address, sample origin, other excipients (where

nd sample description.

ase searchvery search, the unknown spectra and other spectra ine were set to be baseline corrected to reduce the scatter-at highly contaminate NIR spectra. This was achievedto baseline correction algorithm which removes linearor of positive going peak data (Grams).se search was conducted on the whole spectrum basedtion algorithm; whereby the least squares dot productsown spectra were compared with the spectra in theter being centralised to their respective means. A hit listatches was provided together with hit quality indexs. Low HQI values measured between the unknownh the spectra in the library indicating a good match.

and discussions

part of this work was focussed on obtaining spectral of each set of samples and observing the product vari-R raw spectra and PCA. This information was correlatedh outcomes in the database.

alysis

e tablets contained 7590% mass fraction paracetamolhe tablet weight and hence it was expected to that the

peaks NIR spectrum are coming from the paracetamol.ctral analyses indicated that all of the samples producedectroscopic ngerprint of paracetamol (Fig. 1A) whileent compared to the negative control samples (Fig. 1B).ifference among the spectra was the absorbance shift

the whole spectral range which corresponded to scat-ts and physical properties (e.g. particle size, shape orcture) of the samples. Moreover, a clear difference of

over the whole range of wavelengths could be observed the Malaysian samples, PM. One of the negative con-s NC1 had some degree of similarity with the otherl spectra due to similar main active composition.

al component analysis (PCA)

s employed to validate spectral interpretations andre information on the chemical differences betweenese results indicated that the type of ingredients butount of active ingredient inuence classications. One

score pwere wwith th

Theples anthe sammain aincludHowev

3.3. SI

All tdistancto be tPCA. Utinctioobservscattertered tthe 65

Canincludand thance anfactor ithe datcompraffecte

3.4. Da

3.4.1. All

sider itof the Xnanomthe speaveragcombinsearch

3.4.2. Spe

to comvalue, least srelatioclassi

Furto inteoptionspectrudatabapresenspectravalue, perfec

Subferent out agspectran Fig. 3 showed that all of the negative control samplesxcluded while the 650 mg sample was well clusteredher samples.dence ellipse has expanded to accommodate new sam-o negative controls samples were clustered apart from. However NC1, which had 500 mg of paracetamol as the

ingredient in addition to other active ingredients was the 95% ellipse together with the paracetamol sample.s cluster was separated from the rest of the group.

classications

C models were projected into the PC space based on theirwards the sample PM as class model. PM was selectedodel class due its distinct character in NIR analysis andthe distance versus leverage plot in Fig. 4, a clear dis-ariability between the Malaysian and UK samples wasll Malaysian samples clustered well within batches butom 0 to 120 leverage ranges while the UK samples clus-her within only 15 to 35 leverage ranges. Additionally,

batch was also well separated from the other batches. et al., have demonstrated that the number of PCs

classication models is an important factor for SIMCAing pre-processed data decreases the within-class vari-creases the between-class variance [20]. Taking the rstonsideration, we have used the raw NIR spectra, as all ofs sufciently described by only one and two PC. Despiteing the second factor, our results did not appear to be

se construction

base libraryra were imported into a paracetamol database to con-ge and reliability for classication. In the database, alls of the spectra represented the spectral wavelength ins between the range 1100 and 2500 and the Y-axes werel absorbance values. For each brand of paracetamol, 20ectral proles with 700 individual points in each weres a multi-le to enhance the accuracy of the spectral

spectra were saved in 32-bit resolution.

base search searches of several paracetamol tablets were conducted

seven different algorithms; euclidian distance, absolutederivative absolute value, least square, rst derivativee, correlation and rst derivative correlation. The cor-orithm proved to be the most reliable in the spectraln (supplementary material Figure S1).ore, the NIR spectra of examined tablets were usedte the spectral database by comparison of spectra. The

was the spectrum search method, where a wholef an unknown sample was compared to the spectra in theing correlation algorithms. The outcome of a search wass a database hit list and a correlation value between then as a hit quality index (HQI) value. The lower the HQIigher similarity between spectra would be observed. Ach gave an HQI value of 0.ent searches for several paracetamol samples from dif-facturers and different batches (5 brands) were carried

the spectral database that in addition of paracetamol NIR spectra of other pharmaceutical formulations;


Fig. 1. Mean Nspectra (NC1,

amoxicillin(3 brands), tures (2 bracorrect typeturer of thein the datab

These reclassicatio

The valuother produthan 75% of

Table 2Proposed cut-degree of simthe database.

Classicatio

IIIIII IVIR spectra in reectance mode of (A) 22 batches of paracetamol samples and (B) reprNC2 and NC3).

(4 brands), mefenamic acid (2 brands), Ginkgo bilobaEurycoma longifolia (3 brands) and Chinese herbal mix-nds). This conrmed that it was possible to identify the

of medicines and furthermore to suggest a manufac- examined samples (if the similar product was presentase (supplementary material Figure S1 and Table 1).sults were used to propose the cut-off values for then of the analysed sample into four types (Table 2).es in Table 2 are suitable to be used as a reference forcts with a similar composition of samples (where more

the tablet mass is the active pharmaceutical ingredient)

off points for pharmaceutical formulation classication based on theilarities and differences between unknown spectrum and spectra in

n type Search outcomes Cut-off values (HQI)

Match (same batch)


Fig. 2. PCA overview of raw NIR reectance spectra of 16 Malaysian paracetamol samples. The ellipse represents 95% condence interval for the two PCs (PC1 = 97%, PC2 = 2%).

The UP B samples were classied (type II) to one of its co-batchproduct in the database UP A and type III to another batch UP C (hit#13). This was because this set of sample had a different dosage(650 mg). This showed that the database was capable of identifyingproducts of

The HQI value for products of hit #20 in both hit lists wereclassied in ambiguity between the similar or different class ofmedicines from the rest of the samples. This corresponded wellwith the PCA observation whereby samples PM in Table 3A was

ed in

Fig. 3. PCA ovfor the two PC different dose. excluderview of NIR reectance spectra of Malaysian and UK paracetamol samples, and the negs (PC1 = 92%, PC2 = 5%). the 95% condence interval ellipses in PCA and sam-ative control samples. The ellipse represents 95% condence interval


Fig. 4. The distance versus leverage plot showing the relative distance of the samples to sample PM (class model). All of the UK samples were clustered together in the regionbetween 15 and 35 leverage scales.

ple distance plot (Figs. 2 and 4) while sample paramol in Table 3Bwas one of the negative control (NCI) samples, namely consistingof 500 mg p

3.5. Interna

Internalreproducibiof samples after a yeabrand to thket (externawas also su

The spesearched agafter the daducted. Be

serviced once and the processing software (VISION, Foss) has beenupgraded to a newer version.

PCA avaof thl in t

(pcmplecetaally, ectratabasafter

specalso f anatify

Table 3The search ou(HQI: 0.0009),

Hit #

A 1 2 3 4 5 6 7 8 9

10 11 12 13 1415 16 1718 1920aracetamol and 7.46 mg of dihydrocodeine tartrate.

l/external validation

and external validations were conducted to show thelity and reliability of the database search. Three setsfrom the initial sample set were re-analysed (internal)r. Additionally, a set of three new samples of similare spectra in the database were obtained from the mar-l 1). A set of new brand of paracetamol (pcm unknown)bjected to the database search (external 2).ctra of all of the above samples were acquired andainst the existing paracetamol database, 12 monthstabase being created and the initial study being con-tween these periods, the NIR instrument had been

Thespectra

All intervasampleThis saas para

Idethe spthe daations causedtion is time oto identcome of samples (A) BG B and (B) UP B has classied the new spectra as samples of a s respectively.

Hit quality Samples ID. Hit #

BBG C 1

0.0038 BG A 2 0.0229 FP 3 0.0247 UP A 4 0.0329 UK lidl 5 0.036 UK bts 6 0.0385 PT 7 0.0471 PC 8 0.0509 PR 9 0.0513 UK vh 10 0.0514 UK gsl 11 0.0538 UK mrs 12 0.0553 OR 13 0.0604 UP C 14 0.0633 UK wtr 15 0.0643 MIL A 16 0.0667 PG 17 0.0669 MIL B 18 0.0781 IF 19 0.161 PM 20 distribution of the validation spectra among the otherilable in the database can be viewed in Fig. 5.e validation samples fell within the 95% condencehe distribution range of the available spectra. The newm unknown) was however excluded from the ellipse.

was actually paracetamol with a different dosage formmol granules in capsules.the internal validation should give a perfect match to

in the database as similar samples were used to teste. However, product degradation and instrument vari-

servicing were maybe two of the important factors thattra variations between these samples. External valida-affected by product variation, instrument variation, andlysis. Despite these drawbacks, the database managedall of the internal and external samples to their simi-imilar brand/source to BG C (HQI = 0); BG A (HQI = 0.0038) and UP A

Hit quality Samples ID.

0.0009 UP A0.0095 PT0.0131 PC0.0148 BG A0.0149 UK bts0.0156 UK lidl0.0157 PR0.0209 UK vh0.0216 BG C0.0255 FP0.0267 UK gsl0.027 PG0.0287 UP C0.0313 MIL B0.0324 UK wtr0.0326 UK mrs0.035 MIL A0.0392 IF0.046 OR0.1783 Paramol (NCI)


Fig. 5. PC scor ctanc(valid-int) and

Table 4The hit-qualityparacetamol athe database h

Samples

UPAIF PM

lar samplesfrom the savalues are l

4. Conclus

In conclproven to bscreening pples accordvariability. months proallowed obvariability fto the specfully classicut-off valuusing eithetechniques.

Buildingprocess, hoysis will becost-effectiprehensive of data for tication, de plot and 95% condence interval ellipse to describe the distribution of NIR ree external (valid-ext) validation samples, and unknown sample (pcm unknown). index (HQI) for internal and external validations of three brands ofvailable in the database. The analysis was conducted 12 months afterad been created.

Internal validation (HQI) External validation (HQI)

0.0023 0.01090.0163 0.03360.0008 0.0003

in the database and classied them as samples comingme source or of similar brand names (type 2). The HQIisted in Table 4.

ion

usion, the NIR spectral database of paracetamol hase reliable for identication and for a quick productrocedure. The database also managed to classify sam-ing to their type of dose, dosage forms and productThe internal and external validation conducted over 12ved that the database search was reproducible. PCAservation of differences in intra-batch and inter-batchor different products. Although some degree of changestra were observed after a year, the database success-ed the tablets accordingly using the pre-determinede. The possible ambiguity issues could be resolved byr chemometric analysis or other analytical chemistry

a database of medicines on the market can be a tediouswever once it has been established, drug spectral anal-

a relatively simple task, less time consuming and ave procedure. Furthermore, the existence of a com-spectral database will be of a benet as a repositoryfurther chemometric analysis and also for drug iden-rug quality surveillance and as a potential method

of counterfcase wheremanufactur

Current with other herbal mixtypes of spspectromettigated areadulterated

Appendix A

Supplemthe online v

References

ASEAN, 2007.SingaporeMalaysia,

Assi, S., Watt, tablets by

Assi, S., Watt, packaging

Bate, R., 2008.Bernama, V., 2

Kuala LumBlanco, M., Co

spectroscoBlanco, M., Peg

out a referCandol, A., D

IdenticatJ. Pharm. B

Candol, A., Mwhich conlations. An

Cruez, A.F., 200Lumpur.e spectra of samples used to create library (pcm database), internal

eit and adulterated drug-screening, particularly in the the original samples are difcult to obtain from theers.and future work will be the expansion of the databasetypes (and dosage forms) of drugs including complextures. The database is also being enriched with otherectra such as data from NMR spectroscopy and massry. Other applications of the database currently inves-

its use in drug quality control and the detection of

and counterfeit products.

. Supplementary data

entary data associated with this article can be found, inersion, at doi:10.1016/j.ijpharm.2011.05.057.

ASEAN Post-Marketing Alert (PMA) System The Malaysian and Experience. Joint report by: National Pharmaceutical Control Bureau,and Health Sciences Authority, Singapore.R.A., Moffat, A.C., 2008. Assay of ciprooxacin in intact and powderednear-infrared spectroscopy. J. Pharm. Pharmacol. 60, 19.R.A., Moffat, A.C., 2009. Identication of tablets through their blister

by near-infrared spectroscopy. J. Pharm. Pharmacol. 61, 41. Making a Killing. The AEI Press, Washington, DC.008. New Bill to Help Curb Counterfeit Medicines. New Straits Time,pur, 22 September.ello, J., Iturriaga, H., Maspoch, S., de la Pezuela, C., 1998. Near-infraredpy in the pharmaceutical industry. Analyst 123, 135R150R.uero, A., 2010. Analysis of pharmaceuticals by NIR spectroscopy with-ence method. Trends Anal. Chem. 29, 11271136.e Maesschalck, R., Massart, D.L., Hailey, P.A., Harrington, A.C.E., 1999.ion of pharmaceutical excipients using NIR spectroscopy and SIMCA.iomed. Anal. 19, 923935.assart, D.L., Heuerding, S., 1997. Investigation of sources of variancetribute to NIR-spectroscopic measurement of pharmaceutical formu-al. Chim. Acta 345, 185196.6. Viaga not Working? It may be Counterfeit. New Straits Time, Kuala


DCA, 2010. Ministry of Health Malaysia, Drug Control Authority, http://www.bpfk.gov.my/Search/Search product.asp [online] (accessed 9.01.10).

Esbensen, K., 2005. Multivariate Data AnalysisIn Practice, Fifth ed. CAMO SoftwareAS, Norway.

Eustaquio, A., Blanco, M., Jee, R.D., Moffat, A.C., 1999. Determination of paracetamolin intact tablets by use of near infrared transmittance spectroscopy. Anal. Chim.Acta 383, 283290.

Goodman, C., Kachur, S.P., Abdulla, S., Bloland, P., Mills, A., 2007. Drug shop regulationand malaria treatment in Tanzania why do shops break the rules, and does itmatter. Health Policy Plann. 22, 393403.

Grams, Grams ID User Guide. Grams Suite Help. ThermoScientic.Ismail, M., 2009. Prevalence of counterfeit medicines in Malaysia, Pharmacy

enforcement perspective (Presentation). http://portal.bpfk.gov.my/aeimages/File/Seminar../Chapter 12 Counterfeit.pdf [online] (accessed 25.10.10).

Khan, P.R., Jee, R.D., Watt, R.A., Moffat, A.C., 1997. The identication of active drugsin tablets using near infrared spectroscopy. Pharm. Sci. 3, 447453.

Luypaert, J., Massart, D.L., Heyden, Y.V., 2007. Near-infrared spectroscopy applica-tions in pharmaceutical analysis. Talanta 72, 865883.

Moffat, A.C., Assi, S., Watt, R.A., 2010. Identifying counterfeit medicines using nearinfrared spectroscopy. J. Near Infrared Spectrosc. 18, 115.

ONeil, A.J., Jee, R.D., Lee, G., Charvill, A., Moffat, A.C., 2008. Use of a portable nearinfrared spectrometer for the authentication of tablets and the detection ofcounterfeit versions. J. Near Infrared Spectrosc. 16, 327333.

PAM, 2006. Counterfeit medicines: recommendation on measures to combat coun-terfeit medicines in Malaysia. Pharmaceutical Association Malaysia [online].http://www.phama.org.my/pdf document/CounterfeitMed PositionPapervvnal March15 pdf.pdf (accessed 3.09.07).

Trafford, A.D., Jee, R.D., Moffat, A.C., Graham, P., 1999. A rapid quantitative assayof intact paracetamol tablets by reectance near-infrared spectroscopy. Analyst124, 163167.

Wertheimer, A.I., 2008. Identifying and combating counterfeit drugs. Expert Rev.Clin. Pharmacol. 1, 333336.

Yoon, W.L., Jee, R.D., Charvill, A., Lee, G., Moffat, A.C., 2004. Application of near-infrared spectroscopy to the determination of the sites of manufacture ofproprietary products. J. Pharm. Biomed. Anal. 34, 933944.

Near-infrared spectroscopy (NIRS) and chemometric analysis of Malaysian and UK paracetamol tablets: A spectral database study1 Introduction2 Materials and methods2.1 Drug samples2.2 NIR analysis2.3 Software2.4 Chemometric methods2.4.1 Principal component analysis (PCA)2.4.2 Soft Independent Modeling of Class Analogy (SIMCA)

2.5 Database creation and search2.5.1 Database development2.5.2 Database search

3 Results and discussions3.1 NIR analysis3.2 Principal component analysis (PCA)3.3 SIMCA classifications3.4 Database construction3.4.1 Database library3.4.2 Database search

3.5 Internal/external validation

4 ConclusionAppendix A Supplementary dataAppendix A Supplementary data

Said 2011

Documents

Transcript of Said 2011