Classification of reversed-phase columns based on their selectivity towards vancomycin compounds

Talanta 71 (2007) 31–37

Classification of reversed-phase columns based on their selectivity towards vancomycin compounds Erik Haghedooren a , Jos´e Diana a , B´ela Nosz´al b , Jos Hoogmartens a , Erwin Adams a,∗ a

Katholieke Universiteit Leuven, Laboratorium voor Farmaceutische Chemie en Analyse van Geneesmiddelen, O & N 2, PB 923, Herestraat 49, B-3000 Leuven, Belgium b Semmelweis University, Department of Pharmaceutical Chemistry, H˝ ogyes E.u. 9, H-1092 Budapest, Hungary Received 18 October 2005; received in revised form 24 February 2006; accepted 3 March 2006 Available online 18 April 2006

Abstract The selection of a reversed-phase liquid chromatographic (RP-LC) column with suitable selectivity for a particular separation is difficult if the brand name of the column is not known. The monographs of the European Pharmacopoeia and other official compendia for drug analysis only give a general description of the stationary phase to be used in the operating procedure of a liquid chromatographic method. A project to develop a chromatographic test procedure to characterise RP-LC C18 columns was started earlier and resulted in a fast, simple, repeatable and reproducible test procedure. Four column parameters, determined on 69 RP-LC C18 columns, allowed the characterisation and ranking of these columns. In this paper, an overview of this column classification system is given with an application on the separation of vancomycin and some of its impurities. It is shown that the column ranking system is a helpful tool in the selection of a suitable column. © 2006 Elsevier B.V. All rights reserved. Keywords: Reversed-phase liquid chromatography; Column characterisation; Chromatographic tests; Column classification; Vancomycin

1. Introduction Official compendia, like the European Pharmacopoeia (Ph. Eur.) [1] and the United States Pharmacopeia (USP) [2] prescribe many different liquid chromatographic (LC) analyses. In most cases, the use of RP-LC is prescribed with C18 stationary phases. Nowadays, an extended number of different C18 reversed-phase (RP) columns is available on the market. Since monographs of the Ph. Eur. and other official compendia for drug analysis only give a general description of the stationary phase to be used in the operating procedure of an LC method, the selection of a suitable column can be problematic. The problem also rises when a column, prescribed in literature or by compendia, is not available in the laboratory. Proper column selection is also needed during method development when an analyst wants to try columns giving different selectivity. The issue of RP-LC column selection was discussed earlier by Steffeck et al. [3] and Engelhardt and Gr¨uner [4].

∗

Corresponding author. Tel.: +32 16 3234 44; fax: +32 16 3234 48. E-mail address: [email protected] (E. Adams).

0039-9140/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2006.03.013

Dolan and Snyder characterised more then 300 columns based on an empirical equation including conditions-dependent properties of the solute and conditions-independent properties of the column. Five determined parameters (H, S* , A, B and C) can be combined to an Fs value, to evaluate the (dis)similarity of two selected columns [5–10]. Euerby et al. studied 170 columns based on 6 chromatographic parameters, defining surface coverage, hydrophobic selectivity, shape selectivity, hydrogen bonding capacity and ion-exchange capacity at pH 2.7 and 7.6. Principal component analysis led to the study of a subset of the database including only C18 phases with non-acidic silica material [11,12]. Vander Heyden et al. tested 8 chromatographic parameters on 28 columns, using chemometric techniques like Paretooptimality concept, principal component analysis and Derringer’s desirable function. Column selection based on the Pareto-optimality concept led to a subset of columns, but was found less flexible when new columns were added to the original dataset [13]. Many other papers, describing methods to characterise columns, were published to solve this issue, but only recently published articles are cited here [14–19].

32

E. Haghedooren et al. / Talanta 71 (2007) 31–37

In our laboratory, a general test method was developed by selection of the most appropriate test methods from the literature. This column characterisation and classification system allows the selection of columns with selectivity similar to or different from a chosen reference column [20–26]. The column classification system would also allow to follow up column ageing so that analysts can check whether the characteristics of their columns have changed over time. In this work, a short overview is given of the development of our column classification system and an application is shown, using the separation of vancomycin (VM) from some of its impurities. Vancomycin was already used by Forlay-Frick et al., combining 28 different chromatographic conditions, i.e. 7 stationary phases with 4 different mobile phases to study the replacement of columns. This was evaluated based on the theoretical plate number and symmetry factor of three test compounds; benzoic acid, N,N-dimethylaniline and vancomycin [27]. Three columns out of seven had the same stationary phase as used in this article: LiChrospher, Purospher and Symmetry, but the columns had different dimensions, making comparison between the two systems troublesome. VM is a glycopeptide antibiotic used in the prophylaxis and treatment of infections caused by Gram-positive bacteria,

including methicillin-resistant and oxacillin-resistant staphylococci [28,29]. It is a branched tricyclic glycosylated nonribosomal peptide produced by the fermentation of the Actinobacteria species Amycolatopsis orientalis, formerly Nocardia orientalis, appearing as a mixture of similarly structured compounds of which many components have not yet been identified [30–33]. See Fig. 1 for the structure of known components. VM B is the main compound. Monodechlorovancomycin 2 (MDCV 2) is a side-product isolated from fermentation broth. Selective dehalogenation of VM results in monodechlorovancomycin 1 (MDCV 1) [34]. VM degrades into a crystalline degradation product (CDP-I) by hydrolytic loss of ammonia [35,36]. CDP-I exists in two isomeric forms, the major form (CDPM) and the minor form (CDPm) [36–38]. Aglucovancomycin (AGLUV) and desvancosaminylvancomycin (DESV) are other degradation products resulting from the loss of the disaccharide moiety and the vancosamine sugar, respectively [39]. The LC method that is used to analyse vancomycin on the different columns has been discussed previously [40]. The peaks that are considered in the discussion of the classification correspond to vancomycin B (VM B), monodechlorovancomycin 1, monodechlorovancomycin 2 and two isomeric forms of a crystalline degradation product, the major form and the minor form.

Fig. 1. Structure of vancomycin B and related substances.

E. Haghedooren et al. / Talanta 71 (2007) 31–37

33

2.2. Instrumentation and liquid chromatographic conditions

2. Experimental 2.1. Reagents and samples Acetonitrile HPLC grade was purchased from Biosolve LTD (Valkenswaard, The Netherlands). Dioxane HPLC grade and formic acid were purchased from Acros Organics (Geel, Belgium). Concentrated ammonia and tetrahydrofuran were from Riedel-de Ha¨en (Seelze, Germany). Water was purified in the laboratory by distillation of demineralised water. Reference substances of VM B, MDCV 1, MDCV 2, DDCV, CDPM and CDPm were obtained from Abbott Laboratories (Abbott Park, IL, USA). Vancomycin samples were prepared at a concentration of 2 mg/ml. Solutions of reference substances were prepared at a final concentration of 0.05–0.1 mg/ml. Both samples and reference substances were dissolved in mobile phase. Sample solutions slightly degraded within one day when they were kept at room temperature, but they could be used for several days when they were stored at 4 ◦ C.

The LC apparatus consisted of a L-6200 Intelligent Pump (Merck Hitachi, Darmstadt, Germany), an autosampler Model 655A-40 (Merck Hitachi) equipped with a 20 ␮l loop, a Linear UVIS 200 UV detector (Thermo Separation Products, San Jose, CA, USA) set at 280 nm and Chromperfect 4.4.23 software (Justice Laboratory Software, Fife, UK) for data acquisition. A set of 37 RP-LC C18 columns (Table 1) was investigated. The columns were kept at 35 ◦ C in a water bath heated by a Julabo EC thermostat (Julabo, Seelbach, Germany). A mobile phase consisting of dioxane–0.3 M ammonium formate, pH 1.7–water (6:5:89, v/v/v) was used for all columns at a flow rate of 1.0 ml/min. The mobile phases were degassed by sparging helium. To prepare the 0.3 M ammonium formate solution, the appropriate amount of concentrated ammonia was diluted in water and adjusted to the required pH using 10% (v/v) formic acid, before bringing to volume.

Table 1 Specifications for the examined columns Column number

Name of the column

Length (mm)

Particle size (␮m)

Manufacturer/supplier

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

␮Bondapak ACE 5 Alltima Apex Basic Apex ODS II Brava BDS Discovery Hypersil BDS Hypersil ODS Kromasil (EKA) Kromasil (MN) LiChrospher Luna Nucleosil Nucleosil HD Nucleosil Nautilus OmniSpher Platinum Platinum EPS Purospher Purospher endcapped Purospher STAR Spheri-5 Spherisorb ODS2 Supelcosil LC-18 Supelcosil LC-18 DB Superspher Symmetry TracerExcel ODS A Uptisphere HDO Uptisphere ODB Validated Wakosil HG YMC-Pack Pro Zorbax Eclipse XDB Zorbax Extend Zorbax SB

250 250 250 250 250 250 250 250 250 250 250 250 150 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 150 250 250 250

10 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 3 5 5 5

Waters Advanced Chrom. Tech./Achrom Alltech Jones Chromatography/Sopachem Jones Chromatography/Sopachem Alltech Supelco ThermoQuest ThermoQuest Akzo Nobel/SerCoLab Macherey-Nagel/Filter Service Merck Phenomenex/Bester Macherey-Nagel/Filter Service Macherey-Nagel/Filter Service Macherey-Nagel/Filter Service Varian Alltech Alltech Merck Merck Merck PerkinElmer Waters Supelco Supelco Merck Waters Teknokroma/SerCoLab Interchrom/Achrom Interchrom/Achrom PerkinElmer SGE/Achrom YMC Sep. Techn./ThermoQuest Agilent Technologies Agilent Technologies Agilent Technologies

34

E. Haghedooren et al. / Talanta 71 (2007) 31–37

3. Results and discussion 3.1. Development of the column classification system The column characterisation and classification project consisted of three major parts. First, a procedure, allowing the measurement of a number of parameters, reflecting chromatographic characteristics was developed. In order to make the procedure as user friendly as possible, the number of test parameters was kept minimal. The second part intended to classify RP-LC columns with closely related characteristics. The third part consisted in performing pharmaceutical separations to check the usefulness of the developed ranking system of the columns in practice. In the first part, test methods from the literature for characterisation of RP-LC columns were critically evaluated. Properties of RP-LC stationary phases can be checked by both non-chromatographic and chromatographic methods. Carbon content, amount of metal impurities, particle size, surface area, pore size, packing density and acidity can be determined by non-chromatographic methods. However, these techniques are not readily performed and cannot be carried out on the packed column without destruction. Therefore the choice of chromatographic methods was obvious. Visky et al. made a selection of 36 test parameters from literature, testing as many as possible different properties of RP-LC columns, like column efficiency, hydrophobicity and silanol activity. Eight chromatographic methods were developed, allowing to determinate the 36 test parameters [20]. Since a general procedure needs to use repeatable and reproducible test parameters, the eight methods were examined in three different laboratories. Of the initial 36 parameters, 24 proved to be repeatable and reproducible, based on the calculation of the relative standard deviation [21]. Iv´anyi et al. applied chemometrics to diminish the number of parameters, while maintaining the classification. Principal component analysis (PCA) was found to offer the possibility to evaluate column clustering or differentiation [22]. Similar to this approach, the 24 parameters of our study were reduced to  4 parameters: the retention factor of amylbenzene, kamylbenzene  ), the relative retention factor benzylamine/phenol at pH (kamb 2.7, rkbenzylamine/phenol (rkba/ph pH 2.7 ), the relative retention factor triphenylene/o-terphenyl, rktriphenylene/o-terphenyl (rktri/o-ter )  and the retention factor of 2,2 -dipyridyl, k2,2  -dipyridyl  (k2,2  -dip ). In the second part of the project, PCA of the four parameters was at first used to distinguish groups of columns (groups Ia, Ib, IIa, IIb, IIc and III) [23]. These groups were constructed based on the different reversed phases described in the Ph. Eur. and on information provided by the column manufacturers. A more practical approach to rank RP-LC columns was introduced by using F-values:   F = (kamb,ref − kamb,i ) + (rkba/ph pH 2.7,ref − rkba/ph pH 2.7,i ) 2

The F-value of a column i equals the sum of squares of the differences between each parameter value of a chosen reference column and of the column i. The smaller the F-value, the more similar is column i to the reference column. In order to have the same weighing of each parameter in this equation, the parameters are autoscaled using formula (2) before being introduced in Eq. (1): xij − x¯ j sj

(2)

where xij is the value of parameter j on column i, x¯ j the mean of parameter j on all tested columns and sj is the standard deviation on the mean parameter value. With this F-value, a ranking of all columns is obtained, indicating how close columns are to the selected reference column. Low F-values correspond to high ranking [24]. The third part of the project consisted in performing pharmaceutical analyses to check the performance of the column classification system in real separations. Dehouck et al. carried out the separation of acetylsalicylic acid (aspirin) according to the Ph. Eur. monograph. The system suitability test (SST) prescribed by the Ph. Eur. to distinguish between suitable and non-suitable columns was also evaluated. It was concluded that this SST could not always predict the suitability of the column to separate all the aspirin components. Alternatively, the suitability was investigated by calculation of the chromatographic response function (CRF). The CRF value is equal to 1 if all the peaks are baseline separated, and equal to 0 if two peaks are coeluted. Partial separations lead to intermediate values. The column ranking approach starts with the choice of a reference column or of reference parameters. For each column, the F-value versus this reference column was calculated and the columns were ranked according to their F-value, starting with the smallest one. It was concluded that the chance of selecting a suitable column clearly increased with a smaller F-value. All columns with an F < 2 gave baseline separation for all peaks (CRF = 1) while this number decreased to 43% for columns with 2 < F < 6 and to 18% for columns with F > 6. The column ranking system is freely accessible on our website: http:// www.pharm.kuleuven.be/pharmchem/columnclassification. This database contains already more than 50 types of RP-LC C18 columns and new types are being characterised in our laboratory at this moment. Analysts can either classify all columns from the database with regard to a freely chosen reference column from the list or they can fill in four parameter values, determined on their own column [24,25]. In order to further investigate the possibilities of this column classification system, its performance towards the separation of vancomycin compounds is examined here. 3.2. Column classification system based on vancomycin analyses

2

    + (k2,2  -dip,ref − k2,2 -dip,i ) + (rktri/o-ter,ref − rktri/o-ter,i ) 2

2

(1)

Each of the 37 columns in Table 1 was used for the separation of vancomycin and its impurities. The suitability of the stationary phases for this separation was examined by calculating the CRF, which is a measure for the overall selectivity and which is

E. Haghedooren et al. / Talanta 71 (2007) 31–37

35

Fig. 2. Typical chromatogram of: (A) VM commercial sample (2 mg/ml) and (B) a mixture of VM and some of its potential impurities (1 mg/ml) analysed during columns investigation. Column: Symmetry, maintained at 30 ◦ C with dioxane–0.3 M ammonium formate, pH 1.7–water (6:5:89, v/v/v) as mobile phase. Detection: UV at 280 nm; flow rate: 1.0 ml/min; injection volume: 20 ␮l.

calculated as follows: CRF =

n−1  i=1

fi gi

(3)

where n is the total number of peaks, g the interpolated peak height between two peaks, i.e. the distance between the baseline and the line connecting the two peak tops, at the location of the valley, and f is the depth of the valley, measured from the line connecting the two peak tops [41,42]. This means that a baseline-separated peak pair has an f/g ratio of 1, a coeluting pair has a value of 0, while a partly separated peak pair has an intermediate value. The CRF was calculated based on the following peaks: unknown, CDPm, MDCV 1, MDCV 2, VM B and CDPM (Fig. 2). From the separations, it was observed that complete baseline separation for all peaks (CRF = 1) could only be obtained on two columns, Nucleosil HD (column 15) and Spheri-5 (column 23). Some columns did not separate all peaks and therefore had a CRF of 0. However, many columns showed a CRF above 0.91. This is mainly due to partially coelution of CDPm (peak 2 in Fig. 2) and MDCV 1 (peak 3 in Fig. 2). Fig. 3 shows an example of a separation with CRF = 0.50, 0.91 and 1, respectively. As can be observed, a CRF of 0.91 still gives a very acceptable separation. Therefore, the criterion of a good separation was set at a CRF of more than 0.90. For each test parameter, the average of all columns, matching the criterion of CRF > 0.90, was calculated to obtain a ‘virtual suitable column’. With these four values as reference, a ranking was obtained based on the F-values (Table 2), calculated from the measured parameter values, as discussed above. It was observed that all columns with F < 2 are suitable for the analysis of vancomycin and its impurities (CRF > 0.90). For columns with F > 2 the probability of separation of impurities clearly decreases. When F is between 2 and 6, 6 out of 10 (60.0%) columns gave a CRF

Fig. 3. Chromatogram of the test mixture on: (A) Nucleosil Nautilus with CRF = 0.50, (B) Brava BDS with CRF = 0.91 and (C) Nucleosil HD with CRF = 1. The chromatographic conditions were as in Fig. 2. The peak numbering of Fig. 2 was maintained.

value above 0.90, while for F-values above 6, only 1 out of 8 (12.5%) showed an appropriate selectivity. Although suitable columns can be found among the low-ranked columns as well, the probability has clearly decreased. So, the column classification system is a helpful tool in finding a suitable stationary phase, since any column having an F-value lower than 2.0 appeared to

36

E. Haghedooren et al. / Talanta 71 (2007) 31–37

Table 2    Column ranking according to the F-values, calculated vs. a ‘virtual ideal column’ with kamb = 5.78, rkba/ph pH 2.7 = 0.09, k2,2  -dip = 9.76, rktri/o-ter = 1.46  kamb

rkba/ph pH 2.7

 k2,2  -dip

rktri/o-ter

Reference column Virtual suitable column

5.78

0.09

9.76

1.46

Column Validated C18 Nucleosil HD Tracerexcel 5 Uptispher ODB5 Zorbax Extend C18 Wakosil HG 5 25 Superspher Symmetry Kromasil NM YMC-Pack-Pro C18-3 Zorbax Eclipse XDB Purospher Star Uptispher HDO5 Alltima 5 Kromasil EKA ACE C18-5 Zorbax SB-C18 OmniSpher Luna 5

0.34 0.54 0.52 0.70 0.62 0.14 0.75 0.81 0.64 0.36 0.54 1.05 0.39 1.26 1.37 −0.31 −0.02 1.42 0.36

−0.424 −0.349 −0.429 −0.418 −0.434 −0.413 −0.231 −0.563 −0.445 −0.526 −0.413 −0.461 −0.354 −0.402 −0.386 −0.343 −0.434 −0.365 −0.617

−0.10 −0.48 −0.36 −0.28 −0.62 −0.65 −0.27 −0.34 −0.23 −0.57 −0.58 0.51 −0.86 0.41 −0.22 −0.99 −0.27 −0.18 −0.41

Supelcosil LC-18 DB 5 Nucleosil NM Hypersil BDS Brava BDS 5 Purospher endcapped Discovery Purospher Bondapak Spherisorb ODS2 Spheri

−0.64 −0.40 −0.88 −1.20 1.53 −1.19 −0.34 −1.50 0.25 1.16

0.097 −0.188 −0.080 −0.172 −0.472 −0.338 −0.778 −0.316 0.569 0.451

Platinum 5 LiChrospher Nucleosil HD Nautilus Platinum EPS 5 Hypersil ODS Supelcosil LC18 Apex Basic Apex ODS

−1.79 0.78 −1.08 −2.38 −0.88 −0.54 −1.70 −0.69

0.236 0.059 −0.708 1.368 2.473 3.294 −0.778 3.294

be suitable for the separation of vancomycin and some of its impurities. It should be noted that the column ranking does not distinguish between “good” and “bad” columns. A column with a higher F-value only indicates that it has properties different from that of the reference column. For another application, using other reference values, such a column may be ranked high. Even a column that ranks low for many applications may be of great value for some other, specific applications. 4. Conclusions This paper discusses the performance of 37 RP-LC C18 columns for the separation of vancomycin and some of its impurities. The columns were characterised by a set of four chromatographic test parameters. To determine whether a sep-

F-value

CRF

−0.44 −0.27 −0.61 −0.61 −0.14 −0.44 0.33 0.20 0.50 −0.82 −0.91 0.12 −0.91 −0.23 0.20 −0.01 −1.25 0.67 −1.55

0.075 0.150 0.253 0.295 0.318 0.401 0.440 0.459 0.594 0.626 0.732 0.977 1.024 1.055 1.130 1.283 1.296 1.840 1.984

0.93 1 0.97 0.94 0.92 0.94 0.94 0.96 0.93 0.94 0.93 0.97 0.97 0.96 0.94 0.97 0.95 0.94 0.93

−0.95 0.80 −0.93 −1.07 0.96 −1.49 0.26 −0.72 1.94 1.94

−0.61 0.63 0.20 −0.06 0.84 −0.06 1.65 −1.25 0.16 −0.40

2.055 2.207 2.484 3.464 3.620 4.366 4.386 5.014 5.179 5.456

0.92 0.96 0 0.91 0.88 0.95 0 0.19 0.93 1

−0.33 1.94 −0.98 −0.07 1.94 1.94 −0.66 1.94

−1.21 1.10 1.86 1.57 −0.87 −0.35 3.44 −0.48

6.066 6.258 7.371 13.595 13.989 18.019 18.229 18.388

0 0.96 0.5 0.39 0 0.67 0 0

aration was good or poor, the overall selectivity was evaluated using the Chromatographic Response Function. Average values of the four test parameters were calculated for columns showing a CRF > 0.90. Using these values as a reference for a “suitable” column, the F-values were calculated and a ranking was obtained. All columns with an F-value below 2 were found suitable. An F-value between 2 and 6 still offered a 60% possibility to find a good column. Above 6, the chances to find a suitable column were very low. These findings are in accordance with the conclusions drawn from a previous experiment with aspirin. In the future, more complex case studies (also using gradient elution) will have to be performed to examine the correlation between the column test parameters and their separation characteristics. The final aim of the project is to provide a simple column test procedure with accompanying limits, which can help to predict the suitability of a column for real separations.

E. Haghedooren et al. / Talanta 71 (2007) 31–37

Acknowledgments The authors thank the manufacturers and the suppliers for the gift of columns. E. Haghedooren enjoys a grant of the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen). E. Adams is a post-doctoral fellow of the Fund for Scientific Research (FWO)-Flanders, Belgium. Financial support to this project is given by a Research Grant of the Fund for Scientific Research-Flanders (Belgium). References [1] European Pharmacopoeia, fifth ed., Council of Europe, Strasbourg, France, 2005, p. 2671. [2] United States Pharmacopeia, vol. 28, The United States Pharmacopeial Convention, Rockville, MD, USA, 2005, p. 2013. [3] R.J. Steffeck, S.L. Woo, R.J. Weigand, J.M. Anderson, LC–GC 13 (1995) 720. [4] H. Engelhardt, R. Gr¨uner, Inter. Lab. (1999) 34. [5] L.R. Snyder, A. Maule, A. Heebsh, R. Cuellar, S. Paulson, J. Carrano, L. Wrisley, C.C. Chan, N. Pearson, J.W. Dolan, J.J. Gilroy, J. Chromatogr. A 1057 (2004) 49. [6] J.W. Dolan, A. Maule, D. Bingley, L. Wrisley, C.C. Chan, M. Angod, C. Lunte, R. Krisko, J.M. Winston, B.A. Homeier, D.V. McCalley, L.R. Snyder, J. Chromatogr. A 1057 (2004) 59. [7] L.R. Snyder, J.W. Dolan, P.W. Carr, J. Chromatogr. A 1060 (2004) 77. [8] D.H. Marchand, K. Croes, J.W. Dolan, L.R. Snyder, J. Chromatogr. A 1062 (2005) 57. [9] D.H. Marchand, K. Croes, J.W. Dolan, L.R. Snyder, R.A. Henry, K.M.R. Kallury, S. Waite, P.W. Carr, J. Chromatogr. A 1062 (2005) 65. [10] J. Pellett, P. Lukulay, Y. Mao, W. Bowen, R. Reed, M. Ma, R.C. Munger, J.W. Dolan, L. Wrisley, K. Medwid, J. Chromatogr. A 1101 (2006) 122. [11] M.R. Euerby, P. Petersson, J. Chromatogr. A 994 (2003) 13. [12] M.R. Euerby, P. Petersson, J. Chromatogr. A 1088 (2005) 1. [13] E. Van Gyseghem, M. Jimidar, R. Sneyers, D. Redlich, E. Verhoeven, D.L. Massart, Y. Vander Heyden, J. Chromatogr. A 1042 (2004) 69. [14] M. Kele, G. Guiochon, J. Chromatogr. A 830 (1999) 41. [15] W. Verstraeten, J. de Zeeuw, J. Crombeen, N. Vonk, Int. Lab. (March) (2000) 20. [16] U.D. Neue, K. Van Tran, P.C. Iraneta, B.A. Alden, J. Sep. Sci. 26 (2003) 174. [17] K. Le Mapihan, J. Vial, A. Jardy, J. Chromatogr. A 1061 (2004) 149. [18] R. Kaliszan, M.A. van Straten, M. Markuszewski, C.A. Cramers, H.A. Claessens, J. Chromatogr. A 855 (1999) 455.

37

[19] T. Baczek, R. Kaliszan, K. Novotna, P. Jandera, J. Chromatogr. A 1075 (2005) 109. [20] D. Visky, Y. Vander Heyden, T. Iv´anyi, P. Baten, J. De Beer, B. Nosz´al, E. Roets, D.L. Massart, J. Hoogmartens, Pharmeuropa 14 (2002) 288. [21] D. Visky, Y. Vander Heyden, T. Iv´anyi, P. Baten, J. De Beer, Zs. Kov´acs, B. Nosz´al, E. Roets, D.L. Massart, J. Hoogmartens, J. Chromatogr. A 977 (2002) 39. [22] T. Iv´anyi, Y. Vander Heyden, D. Visky, P. Baten, J. De Beer, I. L´az´ar, D.L. Massart, E. Roets, J. Hoogmartens, J. Chromatogr. A 954 (2002) 99. [23] D. Visky, Y. Vander Heyden, T. Iv´anyi, P. Baten, J. De Beer, Z. Kov´acs, B. Nosz´al, P. Dehouck, E. Roets, D.L. Massart, J. Hoogmartens, J. Chromatogr. A 1012 (2003) 11. [24] P. Dehouck, D. Visky, G. Van den Bergh, E. Haghedooren, E. Adams, A. Kerner, Y. Vander Heyden, D.L. Massart, Zs. Kov´acs, B. Nosz´al, J. Hoogmartens, LC–GC Eur. 17 (2004) 592. [25] P. Dehouck, D. Visky, Y. Vander Heyden, E. Adams, Z. Kov´acs, B. Nosz´al, D.L. Massart, J. Hoogmartens, J. Chromatogr. A 1025 (2004) 189– 200. [26] D. Visky, E. Haghedooren, P. Dehouck, Zs. Kov´acs, K. K´oczi´an, B. Nosz´al, J. Hoogmartens, E. Adams, J. Chromatogr. A 1101 (2006) 103–114. [27] P. Forlay-Frick, J. Fekete, K. H´eberger, Anal. Chim. Acta 536 (2005) 71. [28] P.F. Smith, W.P. Petros, M.P. Soucie, K.R. Copeland, Ther. Drug Monit. 20 (1998) 231. [29] M. Adamczyk, E.M. Brate, E.G. Chiappetta, S. Ginsburg, E. Hoffman, C. Klein, M.M. Perkowitz, S.D. Regr, P.P. Chou, A.G. Costantino, Ther. Drug Monit. 20 (1998) 191. [30] G.K. Best, N.H. Best, N.N. Durham, Antimicrob. Agents Chemother. 4 (1968) 115. [31] R.L. Kirchmeier, R.P. Upton, Anal. Chem. 50 (1978) 349. [32] J.R. Uhl, J.P. Anhalt, Ther. Drug Monit. 1 (1979) 75. [33] E.L. Inman, J. Chromatogr. 410 (1987) 363. [34] C.M. Harris, R. Kannan, H. Kopecka, T.M. Harris, J. Am. Chem. Soc. 107 (1985) 6652. [35] F.J. Marshall, J. Med. Chem. 8 (1965) 18. [36] C.M. Harris, H. Kopecka, T.M. Harris, J. Am. Chem. Soc. 105 (1983) 6915. [37] M.P. Williamson, D.H. Williams, J. Am. Chem. Soc. 103 (1981) 6580. [38] D.W. Armstrong, Y. Tang, S. Chen, Y. Ahou, C. Bagwill, J.R. Chen, Anal. Chem. 66 (1994) 1473. [39] R. Nagarajan, A.A. Schabel, J. Chem. Soc. Chem. Commun. (1988) 1306. [40] J. Diana, D. Visky, E. Roets, J. Hoogmartens, J. Chromatogr. A 996 (2003) 115. [41] S.L. Morgan, S.N. Deming, J. Chromatogr. 112 (1975) 267. [42] W. Naidong, S. Hua, E. Roets, J. Hoogmartens, J. Planar Chromatogr. 5 (1992) 92.