Simultaneous determination of valsartan, amlodipine besylate and hydrochlorothiazide in tablets by near infrared

Size: px

Start display at page:

Download "Simultaneous determination of valsartan, amlodipine besylate and hydrochlorothiazide in tablets by near infrared"

Marcus Thomas
6 years ago
Views:

Journl of Applied Phrmceuticl Science Vol. 7 (11), pp. 074-078, November, 2017

1 Journl of Applied Phrmceuticl Science Vol. 7 (11), pp , November, 2017 Avilble online t DOI: /JAPS ISSN Simultneous determintion of vlsrtn, mlodipine besylte nd hydrochlorothizide in tblets by ner infrred Ntn Becker 1, Gbriel R. Foresti 1, Willin R. R. Almeid 1, Krine F. Nicoren 1, Mrco F. Ferrão 2, Fbin E. B. Silv 1* 1 Progrm de Pós-Grdução em Ciêncis Frmcêutics, Cmpus Uruguin, Universidde Federl do Pmp , Uruguin, Rio Grnde do Sul, Brzil. 2 Instituto de Químic, Universidde Federl do Rio Grnde do Sul, , Porto Alegre, Rio Grnde do Sul, Brzilnd Instituto Ncionl de Ciênci e Tecnologi de Bionlític, Cmpins, SP, Brzil. ARTICLE INFO Article history: Received on: 01/09/2017 Accepted on: 29/10/2017 Avilble online: 30/11/2017 Key words: chemometrics, Exforge HCT, phrmceuticl, qulity control. ABSTRACT Intervl prtil lest-squres clibrtion (ipls) nd synergicl prtil lest-squres clibrtion (sipls) methods in combintion with ner infrred spectroscopy coupled with integrting sphere (NIRA) hve been developed for simultneous determintion of mlodipine (AML), vlsrtn (VAL) nd hydrochlorothizide (HCT) in rw mteril powder mixtures used for production commercil phrmceuticl product (tblets). Vrible selection methods (ipls nd sipls) were pplied to select spectrl rnge tht provided significnt informtion nd models with lower prediction errors. A ner-infrred reflectnce ccessory (NIRA) ws used for direct smple nlysis. Spectrl dt were cquired between cm -1 nd divided into 32 intervls. A reltive stndrd error of prediction of 1.27% for VAL, 1.92% for HCT nd 5.19% for AML ws obtined fter selection of better intervls by sipls. Results shown tht vrible selection methods ssocited to NIRA is reltively simple, free solvent nd non-destructive procedure tht could be pplied to the simultneous determintion of AML, VAL nd HCT in tblets. INTRODUCTION In the field of phrmceuticl reserch, High- Performnce Liquid Chromtogrphy (HPLC) is method-ofchoice in bulk drugs nd phrmceuticl dosge forms nlysis. Although HPLC cn be well estblished nd recognized by regultory gencies, hve some disdvntges such s lrge consumption of solvents. "Green methods" s infrred spectroscopy hs become n ttrctive lterntive becuse they provide ccurte results nd low smple preprtion time. Ner infrred using reflectnce ccessory (NIRA) hs been extensively used in qulity control of finished phrmceuticls nd monitoring processes in combintion with multivrite regression * Corresponding Author Fbin Ernestin Brcellos d Silv. Progrm de Pós-Grdução em Ciêncis Frmcêutics, Cmpus Uruguin, Universidde Federl do Pmp , Uruguin, Rio Grnde do Sul, Brzil. E-mil: unipmp.edu.br methods s Prtil Lest Squres (PLS) (Roggo et l., 2007; Alclà et l., 2008; Souz et l., 2012; Pn et l., 2014; Wood et l., 2016). Most recently pplictions hve been published showing tht spectrl region selection using suitble lgorithms such s intervl prtil lest squres (ipls) nd synergicl prtil lest squres (sipls) cn significntly improve the performnce of PLS regression techniques (Xiobo et l., 2010; Shi et l., 2012; Pintvini et l., 2015; Silv, Flores nd Prisotto, 2016). Moreover, spectrl region selection contining relevnt chemicl informtion is especilly importnt for spectr with strong bnd overlpping such s phrmceuticl formultions (contining ctive substnce, impurities nd excipients) (Silv et l., 2009; Pintvini et l., 2014). Recent ppers in the literture describe ssy methods for tblets contining triple combintion of vlsrtn (VAL), hydrochlorothizide (HCT) nd mlodipine besylte (AML) by HPLC nd ultrviolet spectroscopy (UV) (Shln et l., 2013; Drwish et l., 2014) Ntn Becker et l. This is n open ccess rticle distributed under the terms of the Cretive Commons Attribution License -NonCommercil- ShreAlikeUnported License (

2 Becker et l. / Journl of Applied Phrmceuticl Science 7 (11); 2017: However, only one pper cn be founding reporting simultneous determintion of this phrmceuticl product involving multivrite clibrtion (Drwish et l., 2013). Thus, the min objective of this work ws to investigte the vilbility of NIRA spectroscopy ssocited to ipls nd sipls lgorithms for the obtined regression models. These models were used to predicting VAL, HCT nd AML in powder mixtures nd commercil tblets. EXPERIMENTAL Smples VAL, HCT nd AML bulk drugs commercilly cquired were used to prepre synthetic smples. Thirty six formultions (synthetic smples) contining VAL (261 to 500 mg g -1 rnge), HCT (20 to 84 mg g -1 rnge), AML (11 to 50 mg g -1 rnge) nd diluents mgnesium sterte, tlc nd microcrystlline cellulose (2.5:5:92.5, respectively) were prepred in lbortory. Commercil tblets (Exforge HCT, Novrtis Phrm, Switzerlnd) lbeled to contin 160 mg VAL, 12.5 mg HCT nd 5 mg AML were purchsed from locl drugstore. VAL, HCT, AML bulk drugs nd commercil tblets were previously nlyzed by HPLC (El- Gizwy, 2012). NIR spectroscopy All spectr were recorded from 10,000 cm -1 to 4,000 cm -1 using Frontier Optic NIR spectrometer (Perkin Elmer, United Sttes). The spectrophotometer ws set to record n verge spectrum from 16 scns t 4 cm -1 resolution. A nerinfrred reflectnce ccessory NIRA (Perkin Elmer, USA) ws used for direct smple nlysis. Dt processing nd chemometrics For PLS multivrite clibrtion models, the PLS Toolbox 2.0 (Eigenvector Technologies, Mnson, USA) nd MATLAB softwre 7.10 (The Mth Works, Ntick, USA) were used. The spectrl dt were preprocessed with either multiplictive sctter correction (MSC), utosclling (A), men centering (MC), MC followed by MSC or A followed by MSC. Leve-one-out cross-vlidtion procedure ws pplied in order to detect outliers nd select the finl PLS model. The clibrtion set ws constructed with 26 synthetic smples nd the prediction set ws constructed using 10 synthetic smples (Kennrd-Stone lgorithm - MATLAB 7.10, 2010). Rel smple (Exforge HCT, Novrtis Phrm, Switzerlnd) ws included in prediction set since it is unique phrmceuticl formultion vilble in mrket. The full spectrum ws divided into 8, 16 nd 32 intervls nd combintion of 2 or 3 subintervls ws used for vrible selection (itoolbox for MATLAB, USA). The optimum number of ltent vribles (LV) ws chosen ccording to root men squre error of cross vlidtion (RMSECV). The ipls nd sipls routines generte grphicl informtion indicting the RMSECV vlues used in ech intervl. In this cse ws selected the intervl (or combined intervl) thn presented the minor RMSECV vlue. Resulting prediction models were vlidted bsed on LV used in the model, root men squre error of prediction (RMSEP), reltive stndrd error of prediction (RSEP%), intercept, slope nd correltion coefficient of clibrtion (r² cl ) nd vlidtion (r² vl ). Root men squre error of clibrtion (RMSEC) nd RMSEP were used to evlute the prediction bility mong different PLS models (F-test α= 0.5% ). The results obtined by proposed models were compred to the intervl llowed by Brzilin Phrmcopoei (90 110% declred vlue) considering ech seprte ctive substnce since there is no monogrph vilble for coformulted tblets studied (ANVISA, 2010). RESULTS AND DISCUSSION Full-spectrum PLS models Full-spectrum PLS models of VAL, HCT nd AML were obtined with five, four nd seven LV, respectively (results shown in Tble 1). MC followed by MSC preprocessing showed the lower RMSEP nd RMSECV vlues for full spectrum PLS models of VAL, HCT nd AML. Vlsrtn PLS models In Tble 1 re shown selected intervls, LV, totl number of vribles (TV), RMSECV nd RMSEP obtined to fullspectrum PLS, ipls nd sipls models. Intervl number 32 for ipls model with spectrum subdivided into 32 intervls produced better result ccording to lower RMSECV vlue (using 4 LV nd 188 TV). In generl, ipls models showed RMSEP vlues lower thn those obtined with full-spectrum model. However, overfitting demonstrted tht this lgorithm ws not ble to produce dequte models for this dt set. Synergicl PLS ws implemented to develop PLS models for ll possible combintions (2 or 3 subintervls). The sipls models obtined by full-spectrum division into 8 nd 16 intervls showed high RMSEP vlues ttributed to the commercil smple error (dt not shown). These results clerly show tht the representtivity of the clibrtion set is essentil to the success of model (Ebrhimi-Njfbdi et l., 2012). The chievement of low prediction errors on prediction set is possible s consequence of the fct tht the clibrtion set be representtive of phrmceuticl formultion (quntittive composition of synthetic smples very similr to the rel smple). Nevertheless, spectrl region selection using suitble lgorithms cn be reducing these prediction errors ttributed to composition of clibrtion nd prediction set. The best results for sipls models were obtined whit the full-spectrum divided into 32 intervls nd combined 3 subintervls. In generl wy, the combintion of intervls 21 nd 30 by sipls lgorithm reduced RMSEP vlues [sipls32(2) model]. Intervl 19 inclusion [model sipls32(3) with intervls numbers 19, 21 nd 30] modified significntly the qulity of model. This model showed no overfitting between RMSECV nd RMSEP vlues, smll prediction errors nd reduced vrible numbers (560 TV compred to 6000 TV used in full-spectrum model). The selected intervls included the regions of

076 Becker et l. / Journl of Applied Phrmceuticl Science 7 (11); 2017: 074-078 cm -1 (intervl 19), 6077-6243 cm -1 (intervl 21) nd 4374-4561 cm -1 (intervl 30).

Indeed, the spectrl region selected in this intervl (intervl 19) showed contin relevnt informtion to model constructed.

3 076 Becker et l. / Journl of Applied Phrmceuticl Science 7 (11); 2017: cm -1 (intervl 19), cm -1 (intervl 21) nd cm -1 (intervl 30). Intervl 19 corresponds to N-H stretch vibrtions presented in VAL molecule, but bsent in excipients/diluents of prediction set smples (rel nd synthetic smples). Indeed, the spectrl region selected in this intervl (intervl 19) showed contin relevnt informtion to model constructed. For the clibrtion set, the slope nd correltion coefficient with vlues close to unit nd the intercept with vlues close to zero were considered dequte (y = 0.999x ; r² cl = 0.995). The sipls32(3) model showed lower reltive stndrd error of prediction (RSEP = 1.33%) nd correltion coefficient close to unity (r² vl = 0.996), suggesting tht the method used is ccurte. Tble 1: Results to ipls nd sipls clibrtion models nd full-spectrum PLS model for the VAL. Model TV Intervls LV b RMSECV, RMSEP, VAL VAL PLS 6000 All ipls ipls ipls sipls8(2) nd sipls8(3) , 4 nd sipls16(2) nd sipls16(3) , 8 nd SiPLS32 (2) nd sipls32(3)* , 21 nd TV: Totl number of vribles; b LV: ltent vribles; * selected model Intervl PLS plots nd LV used for ech constructed model, RMSECV vlues for ech intervl selected nd the RMSECV vlues for the full-spectrum model (dotted line) re shown in Figure 1. with 32 intervls [ipls32 (30)] resulted in low RMSECV nd RMSEP (when compred with the other ipls models), s showed in Tble 2. The selected intervl included cm -1 region tht corresponds to N-H stretch vibrtions. The sipls32 (3) model showed lower RMSEP vlue when the intervls 19, 24 nd 30 were combined. This wy, it ws possible to find spectrl region to HCT determintion with reduced vrible numbers (560 TV compred to 6000 TV used in full-spectrum model). For the clibrtion set, the slope nd correltion coefficient with vlues close to unit nd intercept with vlues close to zero were considered dequte (y = 0.994x ; r² cl = 0.995). The sipls32 (3) model showed lower reltive stndrd error of prediction (RSEP = 1.30%) nd correltion coefficient close to unity (r² vl = 0.997). Tble 2: Results to ipls nd sipls clibrtion models nd full-spectrum PLS model for the HCT. Model TV Intervls LV b HCT HCT RMSECV, RMSEP, PLS 6000 All ipls ipls ipls sipls8(2) nd sipls8(3) , 6 nd sipls16(2) nd sipls16(3) , 15 nd sipls32(2) nd sipls32(3)* , 24 nd TV: Totl number of vribles; b LV: ltent vribles; *selected model Intervl PLS plots nd LV used for ech constructed model, RMSECV vlues for ech intervl selected nd the RMSECV vlues for the full-spectrum model (dotted line) re shown in Figure 2. Fig. 2: Cross-vlidted prediction errors (RMSECV) vlues to full-spectrum model nd intervl models (brs) to HCT determintion using PLS nd ipls Fig. 1: Cross-vlidted prediction errors (RMSECV) vlues to full-spectrum model nd intervl models (brs) to VAL determintion using PLS nd ipls Hydrochlorothizide PLS models In Tble 2 re shown selected intervls, LV, TV, RMSECV nd RMSEP obtined to full-spectrum PLS, ipls nd sipls models. The developed model using the intervl 30 for ipls Amlodipine besilte PLS models In Tble 3 re shown selected intervls, LV, TV, RMSECV nd RMSEP obtined to full-spectrum PLS, ipls nd sipls models. The full-spectrum divided into eight intervls produced minor RMSECV vlues for ipls s much s sipls. Intervl PLS plots nd LV used for ech constructed model, RMSECV vlues for ech intervl selected nd the RMSECV vlues for the full-spectrum model (dotted line) re shown in Figure 3.

Becker et l. / Journl of Applied Phrmceuticl Science 7 (11); 2017: 074-078 077 vrible selection techniques produced models with better predictive cpbility compred to full-spectrum PLS models.

4 Becker et l. / Journl of Applied Phrmceuticl Science 7 (11); 2017: vrible selection techniques produced models with better predictive cpbility compred to full-spectrum PLS models. Selection of intervls by synergy demonstrted to be most suitble for quntifiction of these nlytes nd to provide n overll perspective of the significnt informtion in different spectrl regions. Fig. 3: Cross-vlidted prediction errors (RMSECV) vlues to full-spectrum model nd intervl models (brs) to AML determintion using PLS nd ipls The sipls8 (3) model showed lower RMSEP vlue when the intervls 5, 7 nd 8 were combined. For this model, results showed good correltion between reference nd predicted vlues (r² vl = 0.985) nd no overfitting. Selected intervls include N-H stretch vibrtions presented in AML molecule, but bsent in excipients/diluents of prediction set smples (rel nd synthetic smples). Tble 3: Results to ipls nd sipls clibrtion models nd full-spectrum PLS model for the AML. Model TV Intervls LV b AML AML RMSECV, RMSEP, PLS 6000 All ipls ipls ipls sipls8(2) nd sipls8(3)* , 7 nd sipls16(2) nd sipls16(3) , 11 nd sipls32(2) nd sipls32(3) , 29 nd TV: Totl number of vribles; b LV: ltent vribles; * selected model. For the clibrtion set, the slope nd correltion coefficient with vlues close to unit nd the intercept with vlues close to zero were considered dequte (y = 0.999x ; r² cl = 0.990). The sipls8(3) model showed lower reltive stndrd error of prediction (RSEP = 5.33%) nd correltion coefficient close to unity. The selection of 3 intervls for the spectrum divided into 8 (TV = 2250) ws necessry to build model with pproprite predictive bility for AML, unlike the other two phrmceuticls (VAL nd HCT) in which the spectrum hd been divided into 32 intervls nd 3 were selected (TV = 560). In this cse, lrger spectrl rnge llowed the construction model with dequte predictive cpbility. It is worth highlighting tht the wvelengths selected for the AML model included those selected for VAL nd HCT. However, the use of sipls lgorithm for quntifiction of AML not produced prediction errors in the sme order of mgnitude s for VAL nd HCT. CONCLUSION Using sipls lgorithm ssocited with NIRA dt it ws possible to build models for fst, simultneous, free solvent nd non-destructive nlysis of VAL, HCT nd AML in tblets. The ACKNOWLEDGMENTS The uthors would like to thnk Conselho Ncionl de Desenvolvimento Tecnológico (CNPq) nd Coordenção de Aperfeiçomento de Pessol de Nível Superior (CAPES). CONFLICT OF INTERESTS REFERENCES The uthors declre no conflict of interest. Alclà M, León J, Ropero J, Blnco M, Romñch RJ. Anlysis of low content drug tblets by trnsmission ner infrred spectroscopy: Selection of clibrtion rnges ccording to multivrite detection nd quntittion limits of PLS models. J Phrm Sci, 2008; 97(12): ANVISA Frmcopei Brsileir 5 edição. Brsíli. Brsil. Drwish HW, Hssn SA, Slem M Y, El-Zeny BA. Comprtive study between derivtive spectrophotometry nd multivrite clibrtion s nlyticl tools pplied for the simultneous quntittion of Amlodipine, Vlsrtn nd Hydrochlorothizide. Spectrochim Act A Mol Biomol Spectrosc, 2013; 113: Drwish H W, Hssn S A, Slem M Y, El-Zeny B A. Different pproches in Prtil Lest Squres nd Artificil Neurl Network models pplied for the nlysis of ternry mixture of Amlodipine, Vlsrtn nd Hydrochlorothizide. Spectrochim Act A Mol Biomol Spectrosc, 2014; 122: Ebrhimi-Njfbdi H, Lerdi R, Oliveri P, Chir Csolino M, Jlli-Hervi M, Lnteri S. Detection of ddition of brley to coffee using ner infrred spectroscopy nd chemometric techniques. Tlnt, 2012; 99: El-Gizwy S M. Development nd Vlidtion of HPLC Method for Simultneous Determintion of Amlodipine, Vlsrtn, Hydrochlorothizide in Dosge Form nd Spiked Humn Plsm. Am J Anlyt Chem, 2012; 3(6): Pn D, Crull G, Yin S, Grosso J. Low level drug product API form nlysis Avlide tblet NIR quntittive method development nd robustness chllenges. J Phrm Biomed Anl, 2014; 89: Pintvini M S, Pontes F L D, Cerqueir L B, Perlt-Zmor P G, Pontrolo R. Simultneous spectrophotometric determintion of pyrntel pmote nd febntel in phrmceuticl preprtions using prtil lest-squres regression. J Anl Chem, 2014; 69(10): Pintvini M S, Pontes F L D, Weiss L X, Sen M M, Pontrolo R. Comprison between Ultrviolet nd Infrred Spectroscopies for the Simultneous Multivrite Determintion of Pyrntel nd Prziquntel. J Brz Chem Soc, 2015; 26(7): Roggo Y, Chlus P, Murer L, Lem-Mrtinez C, Edmond A, Jent N. A review of ner infrred spectroscopy nd chemometrics in phrmceuticl technologies. J Phrm Biomed Anl, 2007; 44: Shln R A, Bell T S, El Yzbi F A, Elonsy S M. Vlidted stbility-indicting HPLC-DAD method of nlysis for the ntihypertensive triple mixture of mlodipine besylte, vlsrtn nd hydrochlorothizide in their tblets. Arb J Chem, 2013; 10(1):S1381- S1394. Shi J, Zou X, Zho J, Mel H, Wng K, Wng X, Chen H. Determintion of totl flvonoids content in fresh Ginkgo bilob lef with different colors using ner infrred spectroscopy. Spectrochim Act A Mol Biomol Spectrosc, 2012; 94:

5 078 Becker et l. / Journl of Applied Phrmceuticl Science 7 (11); 2017: Silv F E B, Ferrão M F, Prisotto G, Müller E I, Flores E M M. Simultneous determintion of sulphmethoxzole nd trimethoprim in powder mixtures by ttenuted totl reflection-fourier trnsform infrred nd multivrite clibrtion. J Phrm Biom Anl, 2009; 49(3): Silv F E B, Flores, É M M, Prisotto G. Green method by diffuse reflectnce infrred spectroscopy nd spectrl region selection for the quntifiction of sulphmethoxzole nd trimethoprim in phrmceuticl formultions. An Acd Brs Cien, 2016; 88:1 15. Souz J A L, Albuquerque M M, Grngeiro S, Pimentel M F, de Sntn D P, Simões S S. Quntifiction of cptopril disulphide s degrdtion product in cptopril tblets using ner infrred spectroscopy nd chemometrics. Vib Spectrosc, 2012; 62:35 41 Wood C, Alwti A, Hlsey S, Gough T, Brown E, Kelly A,Prdkr A. Ner infr red spectroscopy s multivrite process nlyticl tool for predicting phrmceuticl co-crystl concentrtion. J Phrm Biomed Anl, 2016; 129: Xiobo Z, Jiewen Z, Povey MJW, Holmes M, Hnpin M. Vribles selection methods in ner-infrred spectroscopy. Anl Chim Act, 2010; 667(1): How to cite this rticle: Becker N, Foresti GR, Almeid WR, Nicoren KF, Ferrão MF, Silv FEB. Simultneous determintion of vlsrtn, mlodipine besylte nd hydrochlorothizide in tblets by ner infrred. J App Phrm Sci, 2017; 7 (11):

631 ALLOWANES AND TOLERANCES

631 ALLOWANES AND TOLERANCES ANSI Fits. Tbles 3 through 9 inclusive show series of stndrd types nd clsses of fits on unilterl hole bsis, such tht the fit produced by mting prts in ny one clss will produce