Jing-wu KangAnn Van SchepdaelEugène RoetsJos Hoogmartens

Laboratorium voorFarmaceutische Chemieen Analyse van Geneesmiddelen,Faculteit FarmaceutischeWetenschappen,Katholieke Universiteit Leuven,Van Evenstraat 4,B-3000 Leuven, Belgium

Analysis of vancomycin and related impurities bymicellar electrokinetic capillary chromatography.Method development and validation

A fast and highly selective micellar electrokinetic capillary chromatography (MEKC)method for quantitative analysis of vancomycin and related impurities is described.Among the tested surfactants, cetyltrimethylammonium chloride (CTAC) offered thebest selectivity. Another important parameter, which strongly influenced the selectivity,was buffer pH. It was found that the selectivity increased with buffer pH decreasingfrom 9 to 5. Using Tris-phosphate buffer containing CTAC, satisfactory separationcould be obtained in the pH range from 5.0 to 5.5. Excellent repeatability in terms ofmigration time and peak area could be obtained when the capillary was carefullywashed between two runs. In order to obtain optimal conditions and to evaluate themethod robustness, a central composite experimental design was carried out. Theoptimal conditions were: 44 cm length of fused-silica capillary with 50 �m ID, 120 mM

Tris-phosphate buffer (pH 5.2) containing 50 mM CTAC, –15 kV applied voltage, UVdetection at 210 nm, and a column temperature of 25�C. Under the optimal conditions,more than 20 peaks could be separated within 8 min. The method has a linearity rangefrom 0.004 to 1.2 mg/ml (concentration of vancomycin B, active component). The limitof detection (LOD) and limit of quantitation (LOQ) were 0.4 �g/mL vancomycin, equiva-lent to 0.3 �g/mL vancomycin B (0.04%) and 1.1 �g/mL vancomycin, equivalent to0.9 �g/mL vancomycin B (0.1%), respectively.

Keywords: Vancomycin / Antibiotics / Micellar electrokinetic capillary chromatography EL 4516

1 Introduction

Vancomycin, isolated from the soil microorganism Strep-tomyces orientalis, is a glycopeptide antibiotic used fortreating severe infections caused by Gram-positiveorganisms [1]. It is used in penicillin-allergic patients, andis particularly useful against penicillin- and methicillin-resistant staphylococcal infections [2]. Figure 1 repre-sents the chemical structure of vancomycin B, whichis the main compound in vancomycin. Vancomycindegrades into a crystalline degradation product (CDP-I)which has been associated with adverse effects, andwith therapeutic failure due to underdosing of vancomy-cin [2]. Aglucovancomycin B is another degradation pro-duct resulting from the loss of the disaccharide moiety in

vancomycin B. As vancomycin is a fermentation product,the presence of structurally related substances includingother glycopeptides can not be avoided.

Several methods have been used for analysis of vanco-mycin, such as thin-layer chromatography [3], liquid chro-matography (LC) [3–5], fluorescence polarization immu-noassay (FPI) [6], microbiological assay [7], and radio-immunoassay [7]. Most of these methods were used tomonitor the vancomycin concentration in blood. Onlytwo LC methods [1,5] were used for impurity testing. Cur-rently, an LC method is the official method used by theEuropean Pharmacopoeia [8] and the United States Phar-macopeia [9].

In the last decade, capillary electrophoresis (CE) hasbecome a powerful separation tool for drug analysis. Theadvantages of high resolution and short analysis timemake CE attractive for the analysis of samples with acomplex composition. To the best of our knowledge, noreport concerning vancomycin analysis with CE exists inliterature. The purpose of this paper is to present a fastand highly selective CE method for vancomycin analysis.The separation parameters were optimized and methodrobustness was evaluated by using a central composite

Correspondence: Dr. Ann Van Schepdael, Laboratorium voorFarmaceutische Chemie en Analyse van Geneesmiddelen,Faculteit Farmaceutische Wetenschappen, Katholieke Univer-siteit Leuven, Van Evenstraat 4, B-3000 Leuven, BelgiumE-mail: ann.vanschepdael@farm.kuleuven.ac.beFax: +32-16-323448

Abbreviations: CDP-I, crystalline degradation product I; CTAC,cetyltrimethylammonium chloride; PAPS, 3-(N,N-dimethylhexa-decylammonium)propanesulfonate

2588 Electrophoresis 2001, 22, 2588–2592

ª WILEY-VCH Verlag GmbH, 69451 Weinheim, 2001 0173-0835/01/1207–2588 $17.50+.50/0

Figure 1. Chemical structure of vancomycin B and CDP-I.

experimental design with three variables. Method valida-tion was carried out and three commercial samples wereanalyzed.

2 Materials and methods

2.1 Apparatus

CE was performed on Spectraphoresis 1000 equipment(Thermo Separation Products, Fremont, CA, USA) con-trolled by PC 1000 software (Version 3.0.1) which wasoperated under IBM OS/2 (Version 1.2). The untreatedfused-silica capillary (375 �m OD, 50 �m ID) with 44 cmlength (36 cm to detection window) was from PolymicroTechnologies (Phoenix, AZ, USA). Buffer pH was mea-sured with a Consort C-831 pH meter (Turnhout, Bel-gium).

2.2 Chemicals and reagents

Cetyltrimethylammonium chloride (CTAC) (analyticalgrade), polyoxyethylene 23 lauryl ether (Brij 35), sodiumdodecyl sulfate (SDS) (analytical grade), sodium cholate,and tris(hydroxymethyl)aminomethane (Tris) (analyticalgrade) were from Acros Organics (Geel, Belgium). 3-N,N-Dimethylhexadecylammonium)propanesulfonate (PAPS)was from Fluka (Fluka Chemie, Buchs, Switzerland).Phosphoric acid (analytical grade) was from Riedel-deHaën (Seelze, Germany). Acetonitrile (HPLC grade) wasfrom Biosolve (The Netherlands). A vancomycin samplewas supplied by the European Pharmacopoeia (Stras-bourg, France). CDP-I and aglucovancomycin B wereprepared according to literature [10].

2.3 Procedure

Before use, a new capillary was filled with 0.1 M NaOHand left for 2 h. Capillary temperature was thermostatedat 25�C. Detection was performed by UV at 210 nm. In thefinal method, the polarity was reversed and the anodewas at the detector side. The samples were hydrodyna-mically injected for 4 s�0.75 psi (1 psi = 6894.76 Pa).Between two runs, the capillary was flushed with 0.1 M

NaOH solution, 0.1 M HCl solution, water, acetonitrile,and run buffer for 2 min each. All solutions and sampleswere prepared with Milli-Q water (Millipore, Milford, MA,USA) and filtered through 0.2 �m nylon filters (Alltech,Lokeren, Belgium) prior to use. Tris-phosphate bufferwas prepared by dissolving Tris at the concentration men-tioned, followed by adjustment of the pH with 1.5 M phos-phoric acid.

2.4 Software

Experimental design, optimization, and method robust-ness evaluation were performed using Modde 4.0 soft-ware (Umetri, Umea, Sweden).

3 Results and discussion

3.1 Method development

Initially, capillary zone electrophoresis with different typesof buffers at different pH values, was tried out. However,poor separation selectivity was obtained. No improve-ment was achieved after adding individually to the buffer�- or �-cyclodextrin or organic solvents. Therefore, micel-lar electrokinetic capillary chromatography (MEKC) wasconsidered for further investigation. Several commonlyused surfactants, i.e., SDS, CTAC, and sodium cholate,as well as the zwitterionic surfactant PAPS and nonionicsurfactant Brij 35 were tried out individually at differentconcentrations and different pH values (from pH 5 to 9).However, only CTAC gave satisfactory selectivity andwas used for further study. The concentration range ofCTAC from 40 to 60 mM was found to be suitable for opti-mization. Buffer pH was another important parameter thatstrongly influenced the separation. The effect of buffer pHon separation was investigated in the range from pH 5 topH 9. It was found that the selectivity increased withdecreasing pH. No pH lower than 5 was tried becausevancomycin is unstable at pH values below 5 [1]. The pHrange from 5.0 to 5.8 was useful for further optimization ofthe separation conditions. In the present study, Tris-phos-phate buffer was used rather than sodium phosphate buf-fer due to the fact that the former buffer generated lowerJoule heating and yielded higher peak efficiency.

Electrophoresis 2001, 22, 2588–2592 Analysis of vancomycin by MEKC 2589

CEan

dCEC

In the process of method development, it was found thatthe repeatability of migration time and the efficiencystarted to become poor after several runs. This shouldbe attributed to the adsorption of the vancomycin ontothe capillary wall because of the electrostatic interactionbetween the positively charged vancomycin and thenegatively charged capillary wall. Although the capillarywall had been dynamically coated by positively chargedCTAC molecules, it seemed that this coating was notstrong enough to minimize adsorption. A thorough clean-ing procedure was necessary, in order to restore the capil-lary wall. In fact, the peak shape remained poor even after10 min washing with 0.1 M NaOH solution or 0.1 M HClsolution. Finally, a multiple washing step was found to beeffective to clean the capillary wall. This consisted of suc-cessive washes with 0.1 M NaOH, 0.1 M HCl, water, acet-onitrile, and run buffer for 2 min each between two runs.

3.2 Method optimization and robustnessevaluation

In order to obtain optimal separation conditions and toevaluate the method robustness, a central compositeexperimental design was carried out. This experimentaldesign included three variables (factors), i.e., buffer pH(from pH 5.0 to 5.8), concentration of CTAC (from 40 to60 mM) and Tris (from 100 to 140 mM). The choice of thehigh and low levels resulted from the preliminary work. Afused-silica capillary with a length of 44 cm / 36 cm wasselected to fit the capillary cartridge. An applied voltageof –15 kV was found to give a good compromise betweenseparation selectivity and the analysis time. The capillarytemperature was kept at 25�C. The number of peaks sepa-rated from the main peak was used as the response in theexperimental design. For a central composite experimen-tal design with 3 variables, 17 experiments including3 cen-tral points (2k + 2k + 3, where k is the number of studiedvariables) [11] are needed. These experiments were per-formed in a random order to avoid system errors. Theexperimental data were fitted with multiple linear regres-sion (MLR) modeling [11]. For an experimental design withthree variables, the model can be expressed as:

y = b0 + b1x1 + b2x2 + b3x3 + b12x1x2 + b13x1x3 + b23x2x3 +b11x1

2 + b22x22 + b33x3

2 + E

where b is the regression coefficient and E is the experi-mental error, the square term of each variable describesthe nonlinear effect on the response, and the cross termof the two different variables describes the effect of theirinteraction on the response. The regression coefficientplot with respect to the vancomycin separation is shownin Fig. 2. The 95% confidence interval is expressed interms of error bar over the coefficient bar. The positive or

negative coefficient in the chart indicates that the corre-sponding variable shows a positive or negative effect onthe response, respectively. It can be seen from Fig. 2 thatbuffer pH shows a significant but negative influence onseparation. This means that the selectivity increases withdecreasing the buffer pH. The concentration of CTAC hasa positive, nonlinear effect on the separation.

The response surface plots constructed by plotting thenumber of peaks as a function of the most important vari-ables are shown in Fig. 3. The top region of the responsesurface corresponds to the optimal conditions, estab-lished as follows: 44 cm length of fused-silica capillarywith 50 �m ID, 120 mM Tris-phosphate buffer (pH 5.2) con-taining 50 mM CTAC, –15 kV applied voltage, UV detection

Figure 2. Regression coefficient plot showing the effectof the investigated variables and the interactions betweenthem, on the number of peaks separated. pH, buffer pH;CT, CTAC concentration; Tr, Tris concentration.

Figure 3. Response surface plot showing the depend-ence of the number of peaks separated on buffer pH andconcentration of CTAC.

2590 J.-wu Kang et al. Electrophoresis 2001, 22, 2588–2592

Figure 4. Electropherogram for a commercial sample ofvancomycin under optimized conditions: capillary, 44 cmtotal length, 50 �m ID; background electrolyte, 120 mM

Tris-phosphate buffer containing 50 mM CTAC (pH 5.2);– 15 kV (61 �A); 25�C; UV detection at 210 nm; sampleconcentration 1.0 mg/mL vancomycin in water, equiva-lent to 0.8 mg/mL vancomycin B. Peak identification:1, CDP- I; 2, vancomycin B; 3, aglucovancomycin B.

at 210 nm and a column temperature of 25�C. Robustnessis an important aspect of an analytical method used fordrug analysis. For a chromatographic method, it is a mea-sure of the method’s capability to produce similar separa-tions when the parameters change to a small extent.Besides for method optimization, the results of the experi-mental design can also be used to evaluate the methodrobustness [12]. It can be seen from Fig. 3 that separationremains good when buffer pH and CTAC concentration(two important parameters) change a little around the opti-mal conditions, i.e., pH from 5.0 to 5.4, CTAC concentra-tion from 45 to 57 mM. This means that the present methodis robust in this range. Figure 4 shows a typical electro-pherogram for a commercial sample. The peaks corre-sponding to the impurities CDP-I and aglucovancomycinB were identified with the compounds prepared in house.

3.3 Method validation and analysis ofcommercial samples

As no chemical reference substance was available, thePh. Eur. sample was used for method validation. The con-centration of vancomycin B (active component) wasdetermined by normalization based on the assumptionthat all the components have the same UV absorption at210 nm. The peak area linearity of vancomycin B wasexamined in a range of 0.04–1.2 mg/mL (6 points), eachpoint being determined in triplicate. Limit of quantitation(LOQ) and limit of detection (LOD) corresponding to sig-nal-to-noise ratios of 10 and 3, respectively, were alsodetermined with the same sample. The validation dataare summarized in Table 1. Three commercial samples

were analyzed and the contents of vancomycin B aregiven in Table 2. Compared with the LC methods men-tioned before, this CE method has higher selectivity andshorter analysis time. For example, sample 1 shows thepresence of 15 peaks separated from the main peak andhaving a normalized percentage above 0.05%, within ananalysis time of 25 min, with the Ph. Eur. LC method.

Table 1. Validation data for vancomycin B

Parameter

Intraday repeatability (n = 6):a)

Migration time RSD 0.8%Corrected areab) RSD 1.9%

Interday repeatability (n = 6 days):a)

Migration time RSD 1.6%Corrected areab) RSD 4.4%

Linearityc), y = 105974x – 442;r = 0.9998, Sy,x = 1066

LOQ (S/N = 10), RSD = 7.9% (n = 6) 1.1 �g/mL (0.1%)LOD (S/N = 3) 0.4 �g/mL (0.04%)

Injection volume was 7.5 nL for 4 s injectiona) Vancomycin concentration: 1 mg/mL water, equivalent

to 0.8 mg/mL vancomycin Bb) Corrected area = peak area divided by migration timec) Injection for 4 s; y = corrected area, x = concentration

in mg/mL, Sy,x = standard error of y-estimate; numberof concentration points = 6; each concentration wasinjected in triplicate.

Table 2. Content of vancomycin B in three commercialsamplesa)

% Content (RSD %)b)

Sample 1 Sample 2 Sample 3

Vancomycin B 80.3 (0.5) 79.0 (0.6) 73.1 (0.5)Nc) = 20 Nc) = 19 Nc) = 15

a) Conditions as in Fig. 4; each sample was injected intriplicate and the average value is given.

b) Corrected peak areas were employed to calculate the% content by normalization based on the assumptionthat all the components have the same response factorfor UV detection at 210 nm.

c) Number of peaks separated from the main peak, andhaving a normalized percentage above 0.05%.

In conclusion, a highly selective MEKC method for vanco-mycin analysis has been developed. The method wasestablished to be robust and is superior to the reportedLC methods. Therefore, it can be used for quantitativeanalysis of vancomycin samples and for quality control.

Reveived March 6, 2001

Electrophoresis 2001, 22, 2588–2592 Analysis of vancomycin by MEKC 2591

4 References

[1] Harris, C. M., Kopecka, H., Harris, T. M., J. Am. Chem. Soc.1983, 105, 6915–6922.

[2] Somerville, A. L., Wright, D. H., Rotschafer, J. C., Pharma-cotherapy 1999, 19, 702–707.

[3] Thomas, A. H., Newland, P., J. Chromatogr. 1987, 410, 373–382.

[4] Inxnan, E. L., J. Chromatogr. 1987, 410, 363–372.[5] Bauchet, J., Pussard, E., Garaud, J. J., J. Chromatogr. 1987,

414, 472–476.[6] Morse, G. D., Nairn, D. K., Bertino, J. S., Walshe, J. J., Ther.

Drug Monit. 1987, 9, 212–215.

[7] Pfaller, M. A., Krogstad, D. J., Granich, G. G., Murray, P. R.,J. Clin. Microb. 1984, 20, 311–316.

[8] European Pharmacopoeia, 3rd edition, Council of Europe,Strasbourg, France, 1997.

[9] United States Pharmacopeia 24, United States Pharmaco-peial Convention, Rockville, MD 1999.

[10] Marshall, F. J., J. Med. Chem. 1965, 8, 18–22.

[11] Morgan, E., Burton, K. W., Church, P. A., in: Massart, D. L.,Brereton, R. G., Dessy, R. E., Hopke, P. K., Spiegelman, C.H., Wegscheider, W. (Eds.), Chemometrics Tutorials, Else-vier, Amsterdam 1990, pp.104–123.

[12] De Beer, J. O., Hoogmartens, J., J. Pharm. Biomed. Anal.1993, 11, 1239–1250.

2592 J.-wu Kang et al. Electrophoresis 2001, 22, 2588–2592