Facilitated column selection in pharmaceutical analyses using a simple column classification system

12
Journal of Chromatography A, 1101 (2006) 103–114 Facilitated column selection in pharmaceutical analyses using a simple column classification system ora Visky a,b , Erik Haghedooren a , Pieter Dehouck a , Zsuzsanna Kov´ acs a,b , Kristof K ´ oczi´ an a,b , B´ ela Nosz´ al b , Jos Hoogmartens a , Erwin Adams a,a Katholieke Universiteit Leuven, Laboratorium voor Farmaceutische Chemie en Analyse van Geneesmiddelen, Van Evenstraat 4, B-3000 Leuven, Belgium b Semmelweis University, Department of Pharmaceutical Chemistry, H˝ ogyes E. u. 9. H-1092 Budapest, Hungary Received 8 July 2005; received in revised form 20 September 2005; accepted 26 September 2005 Available online 19 October 2005 Abstract In this paper, the performance of a previously developed classification system applied to pharmaceutical chromatographic analyses, is investigated. The separation of seven different drug substances from their respective impurities was studied. The chromatographic procedure for acetylsalicylic acid, clindamycin hydrochloride, buflomedil hydrochloride, chloramphenicol sodium succinate, nimesulide and phenoxymethylpenicillin was performed according to the corresponding European Pharmacopoeia (Ph. Eur.) monograph. The separation of dihydrostreptomycin sulphate was performed according to the literature. It is shown that the column ranking system is a helpful tool in the selection of a suitable column in these analyses. © 2005 Elsevier B.V. All rights reserved. Keywords: Column characterisation; Chromatographic tests; Column classification 1. Introduction Nowadays, hundreds of different brands of reversed phase (RP) liquid chromatographic (LC) C18 columns are available on the market. Therefore, the selection of a suitable RP-LC C18 col- umn is difficult. This explains why analysts have much interest in the characterisation and classification of C18 columns. Many papers have been published since the 1970s, but only those, which appeared in the last 15 years, are cited here [1–19]. The aim of these studies was: (i) to examine properties of RP-LC supports like efficiency, hydrophobicity, silanol activity, ion- exchange capacity, metal impurity level and steric selectivity with simple chromatographic methods and (ii) to characterize and classify the different brands of stationary phases. Chemo- metric tools were often used to facilitate the data evaluation [4,8–10,15,16]. The availability of a good characterisation and classification system is important for several reasons. Often, one has to find a column similar to one that is described in an existing method or in a paper because the prescribed column is not available in the laboratory. Sometimes, the column that was used for method Corresponding author. Tel.: +32 16 323443; fax: +32 16 323448. E-mail address: [email protected] (E. Adams). development is simply not available anymore on the market. It is also possible that column properties differ between batches or that previous use of a column changed its properties. Many laboratories, e.g. control laboratories, do not use a new column for each separation they have to perform. In each of these cases, it is desirable to be able to identify an alternative column of similar selectivity, so that the replacement column will provide an “equivalent” separation as the original column. Since 1998, column selectivity in RP-LC has been exten- sively studied by the group of Snyder. This leads to the hydrophobic-subtraction model describing five various solute- column interactions [20–30]. Recently, a procedure to char- acterise the selectivity of RP-LC columns was presented and evaluated. The selection and comparison of equivalent columns were examined for 12 routine separations performed in five dif- ferent laboratories [26,27]. Interesting work has also been published by the group of Kaliszan. Their approach is based on quantitative structure–retention relationships (QSRR) where retention is evaluated in terms of the chemical structure of the analytes and of the physicochemical properties of both the stationary and mobile phase [31–35]. In the same time period, a simple chromatographic test pro- cedure to characterise RP-LC C18 columns was derived from a 0021-9673/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2005.09.069

Transcript of Facilitated column selection in pharmaceutical analyses using a simple column classification system

Journal of Chromatography A, 1101 (2006) 103–114

Facilitated column selection in pharmaceutical analysesusing a simple column classification system

Dora Viskya,b, Erik Haghedoorena, Pieter Dehoucka, Zsuzsanna Kovacsa,b,Kristof Kocziana,b, Bela Noszalb, Jos Hoogmartensa, Erwin Adamsa,∗

a Katholieke Universiteit Leuven, Laboratorium voor Farmaceutische Chemie en Analyse van Geneesmiddelen, Van Evenstraat 4, B-3000 Leuven, Belgiumb Semmelweis University, Department of Pharmaceutical Chemistry, Hogyes E. u. 9. H-1092 Budapest, Hungary

Received 8 July 2005; received in revised form 20 September 2005; accepted 26 September 2005Available online 19 October 2005

Abstract

In this paper, the performance of a previously developed classification system applied to pharmaceutical chromatographic analyses, is investigated.The separation of seven different drug substances from their respective impurities was studied. The chromatographic procedure for acetylsalicylicacid, clindamycin hydrochloride, buflomedil hydrochloride, chloramphenicol sodium succinate, nimesulide and phenoxymethylpenicillin wasp sulphate wasp e analys©

K

1

(tuipwasewam[

saot

et. Ittches

anylumnases,of

ide

ten-the

lute-ar-

andumns

dif-

rouptiven is

s andand

0d

erformed according to the corresponding European Pharmacopoeia (Ph. Eur.) monograph. The separation of dihydrostreptomycinerformed according to the literature. It is shown that the column ranking system is a helpful tool in the selection of a suitable column in theses.2005 Elsevier B.V. All rights reserved.

eywords: Column characterisation; Chromatographic tests; Column classification

. Introduction

Nowadays, hundreds of different brands of reversed phaseRP) liquid chromatographic (LC) C18 columns are available onhe market. Therefore, the selection of a suitable RP-LC C18 col-mn is difficult. This explains why analysts have much interest

n the characterisation and classification of C18 columns. Manyapers have been published since the 1970s, but only those,hich appeared in the last 15 years, are cited here[1–19]. Theim of these studies was: (i) to examine properties of RP-LCupports like efficiency, hydrophobicity, silanol activity, ion-xchange capacity, metal impurity level and steric selectivityith simple chromatographic methods and (ii) to characterizend classify the different brands of stationary phases. Chemo-etric tools were often used to facilitate the data evaluation

4,8–10,15,16].The availability of a good characterisation and classification

ystem is important for several reasons. Often, one has to findcolumn similar to one that is described in an existing methodr in a paper because the prescribed column is not available in

he laboratory. Sometimes, the column that was used for method

development is simply not available anymore on the markis also possible that column properties differ between baor that previous use of a column changed its properties. Mlaboratories, e.g. control laboratories, do not use a new cofor each separation they have to perform. In each of these cit is desirable to be able to identify an alternative columnsimilar selectivity, so that the replacement column will provan “equivalent” separation as the original column.

Since 1998, column selectivity in RP-LC has been exsively studied by the group of Snyder. This leads tohydrophobic-subtraction model describing five various socolumn interactions[20–30]. Recently, a procedure to chacterise the selectivity of RP-LC columns was presentedevaluated. The selection and comparison of equivalent colwere examined for 12 routine separations performed in fiveferent laboratories[26,27].

Interesting work has also been published by the gof Kaliszan. Their approach is based on quantitastructure–retention relationships (QSRR) where retentioevaluated in terms of the chemical structure of the analyteof the physicochemical properties of both the stationary

∗ Corresponding author. Tel.: +32 16 323443; fax: +32 16 323448.E-mail address: [email protected] (E. Adams).

mobile phase[31–35].In the same time period, a simple chromatographic test pro-

cedure to characterise RP-LC C18 columns was derived from a

021-9673/$ – see front matter © 2005 Elsevier B.V. All rights reserved.

oi:10.1016/j.chroma.2005.09.069

104 D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114

series of published tests by Hoogmartens and co-workers. Thesystem allows to rank C18 columns, which are each charac-terised by four parameters: the retention factor of amylbenzene,k′

amylbenzene (k′amb), the relative retention factor benzy-

lamine/phenol at pH 2.7,rk′benzylamine/phenol (rk′

ba/ph 2.7),the relative retention factor triphenylene/o-terphenyl,rk′

triphenylene/o-terphenyl(rk′tri/ter) and the retention factor of

2,2′-dipyridyl, k′2,2′-dipyridyl (k′

2,2′-d) [36–41]. This approachstarts with the selection of a reference column or 4 referenceparameter values. AF-value for a columni is calculated as:

F = (k′amb, ref − k′

amb, i)2 + (rk′

ba/pH 2.7, ref − rk′ba/pH 2.7, i)

2

+ (rk′tri/ter, ref − rk′

tri/ter, i)2 + (k′

2,2′−d, ref − k′2,2′−d, i)

2

(1)

TheF-value of a columni equals the sum of squares of thedifferences between each parameter value of the reference col-umn and of a columni. The smaller theF-value, the more similaris columni to the reference column and the higher is columnifound in the ranking (high ranked columns). Before being intro-duced in Eq.(1), the parameters are autoscaled:

xij − xj

sj(2)

wherex is the value of parameterj on columni, x is the meanv dd

thati ity.F n.W liest ivityH ns ichht ri htinf

vins thisc umnp r wh evet peci

acid( d car . Eum ,m ma-t chail omt tm d fom or iP wee

the ranking of the columns and the selectivity in the separationof ASA was demonstrated and it was concluded that the col-umn classification system can help analysts in the selection ofa suitable RP-LC C18 column for the analysis of ASA. In themean time, a freely accessible website was constructed[41,43].In the ASA study, it was also examined whether the Ph. Eur.system suitability test (SST) was able to differentiate betweensuitable and non-suitable columns. It was found that meetingthe SST requirements does not always predict the suitability ofa column. In order to evaluate the separation on the stationaryphases, the chromatographic response function (CRF), which isa measure for the overall selectivity, was used[44]. The CRFwas calculated as:

CRF=n−1∏

i=1

fi

gi

(3)

wheren is the total number of solutes,g the interpolated peakheight, i.e. the distance between the baseline and the line con-necting the two peak tops, at the location of the valley andf thedepth of the valley, measured from the line connecting the twopeak tops[40,45]. This means that a baseline separated peak pairhas anf/g ratio of 1, a non-separated pair has a value of 0, while apartly separated peak pair has an intermediate value. The use ofthese values has been described for thin-layer chromatographicmethods[44], but they can be used in LC as well[46].

pro-c rdert dure,m tweenc o, thisp h werep r. Thes sub-s werea oft bed int ilh ulidea fromt -a estedw naryp d thes esti-g lopedt lumns

2

2

hro-m s andc td oved

ij j

alue of parameterj on all tested columns andsj is the standareviation for parameterj.

The calculation ofF-values results in a single parameters a function of four different contributions to column selectivurthermore, all values ofF are relative to a single columhile this is a convenient and useful simplification, it imp

hat all columns can be arranged in order of relative selectowever, the accuracy ofF as a measure of similar columelectivity will be decreased for two columns, each of whasF � 0. An alternative approach is to define the value ofF in

erms of the two columns being compared[23]. Since it was ountention to keep the system as simple as possible, no weigactors that would introduce complexity, were used.

Of course, one also has to check whether columns haimilar parameters give similar separations in practice. Inontext, it should be remarked that although the four colarameters are each independent of column size, the latteave an influence on the real separations performed. How

his is of minor importance, as method descriptions always sfy the column length to be used.

In a previous study, the separation of acetylsalicylicASA) and its impurities was selected as a case study anied out as prescribed by the European Pharmacopoeia (Phonograph on the columns tested earlier[40,41]. In generalonographs in the Ph. Eur. only give very general infor

ion about the stationary phase to be used in terms ofength, end-capping, base-deactivation, particle size and simes pore size and specific surface area[42]. For more recenonographs, the brand name of the column that was useethod development can be found on the Ph. Eur. websiteharmeuropa. After testing 69 columns, a relationship bet

.

g

g

illr,-

r-r.)

ne-

rnn

In this paper, the ASA study will be evaluated using aedure without principal component analysis (PCA). In oo become a general chromatographic column test proceore case studies are needed to evaluate the correlation be

olumn test parameters and separation characteristics. Saper describes six other representative separations, whicerformed on the stationary phases already tested earlieelected methods are used for impurity profiling of drugtances. They are all isocratic and reference substancesvailable in order to enable calculation of CRF. A majorityhe methods are tests for related substances as prescrihe Ph. Eur.[42] (ASA, clindamycin hydrochloride, buflomedydrochloride, chloramphenicol sodium succinate, nimesnd phenoxymethylpenicillin), while one other was chosen

he literature (dihydrostreptomycin sulphate[47]). For the 6 seprations from the Ph. Eur., a SST was available. It was thether the SST could predict the suitability of the statiohases. Also the correlation between the column ranking anelectivity in each separation was examined in order to invate whether the column ranking system based on the deve

est procedure could be used to facilitate RP-LC C18 coelection.

. Experimental

.1. Chromatographic tests and columns tested

General information concerning the test methods, the catographic conditions applied, the measured parameter

olumn properties were published earlier[36,38]. Columns thaid not meet the requirements of the monographs were rem

D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114 105

from the data set: column 8 (containing 10�m particles), 11 (amonolithic column) and 53 (containing 2�m particles). Somestationary phases, which were used in previous experiments,were no longer included in the study (columns 14, 15, 17, 21, 24,

27, 42, 48, 62 and 63) since they had been used for other separa-tions in the mean time. A mix of used and (nearly) new columnswould not allow an objective comparison and so only the lat-ter were used here (Table 1). The original column numbering,

Table 1Columns involved in this study

Column number Name of the column Length (mm) Particle size (�m) Manufacturer/supplier

1 ACE 3 C18 150 3 Advanced Chrom. Tech./Achrom2 ACE 5 C18 250 5 Advanced Chrom. Tech./Achrom3 Alltima C18 3 150 3 Alltech4 Alltima C18 5 250 5 Alltech5 Apex Basic 250 5 Jones Chromatography/Sopachem6 Apex ODS II 250 5 Jones Chromatography/Sopachem7 Aqua 150 5 Phenomenex/Bester9 Brava BDS 3 150 3 Alltech

10 Brava BDS 5 250 5 Alltech12 Discovery C18 250 5 Supelco13 Genesis C18 3 100 3 Jones Chromatography/Sopachem16 Hypersil BDS 5 250 5 ThermoQuest18 Hypersil ODS 5 250 5 ThermoQuest19 HyPURITY Elite 3 150 3 ThermoQuest, SerCoLab20 HyPURITY Elite 5 150 5 ThermoQuest, SerCoLab22 Kromasil (MN) 250 5 Macherey-Nagel/Filter Service23 Kromasil (EKA) 250 5 Akzo Nobel/SerCoLab25 LiChrospher 250 5 Merck26 Luna 150 5 Phenomenex/Bester28 Nucleosil 5 250 5 Macherey-Nagel/Filter Service233333333334444444455555555566666666

T

9 Nucleosil HD 2500 Nucleosil Nautilus 2501 OmniSpher 2502 Pecospher C18 833 Platinum C18 3 150

4 Platinum C18 5 2505 Platinum EPS C18 3 1506 Platinum EPS C18 5 2507 Prodigy 1008 Purospher 2509 Purospher endcapped 2500 Purospher STAR e 2501 SPHERI-5 2503 Spherisorb ODS2 5 2504 Supelcosil LC-18 2505 Supelcosil LC-18 DB 3 1506 Supelcosil LC-18 DB 5 2507 Superspher 2509 Symmetry 5 2500 TracerExcel ODS A-3 1501 TracerExcel ODS A-5 2502 TSKgel ODS-80TS 1504 Uptisphere 3 HDOC18 1005 Uptisphere 5 HDOC18 2506 Uptisphere 3 ODB 1007 Uptisphere 5 ODB 2508 Validated C18 2509 Wakosil C18 HG 5 – 10 1000 Wakosil C18HG 5 – 25 2501 Wakosil C18 RS 3 – 10 1004 YMC-Hydrosphere C18 1505 YMC-Pack Pro C18-3 1506 YMC-Pack Pro C18-5 1507 Zorbax Eclipse XDB-C18 2508 Zorbax Extend C18 2509 Zorbax SB-C18 250

he long columns (0.25 m) are indicated in bold, the short ones in italic. The in

5 Macherey-Nagel/Filter Service5 Macherey-Nagel/Filter Service5 Varian3 Perkin-Elmer3 Alltech

5 Alltech3 Alltech5 Alltech3 Phenomenex/Bester5 Merck5 Merck5 Merck5 Perkin-Elmer5 Waters5 Supelco3 Supelco5 Supelco4 Merck5 Waters3 Teknokroma/SerCoLab5 Teknokroma/SerCoLab5 TosoHaas/SerCoLab3 Interchrom/Achrom5 Interchrom/Achrom3 Interchrom/Achrom5 Interchrom/Achrom5 Perkin-Elmer5 SGE/Achrom5 SGE/Achrom3 SGE/Achrom5 YMC Sep. Techn./ThermoQuest3 YMC Sep. Techn./ThermoQuest5 YMC Sep. Techn./ThermoQuest5 Agilent Technologies5 Agilent Technologies5 Agilent Technologies

ternal diameter is always 4.6 mm.

106 D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114

employed in our previous papers[36,38–41], was maintained.All 0.25 m columns are indicated in bold while shorter columnsare presented in italics.

2.2. Samples and reagents

Buflomedil hydrochloride, 4-(pyrrolidin-1-yl)-1-(2-hy-droxy-4,6-dimethoxyphenyl)-butan-1-one (impurity A) and4-(pyrrolidin-1-yl)-1-(4-hydroxy-4,6-dimethoxyphenyl)-butan-1-one (impurity B) were obtained from the Ph. Eur.laboratory (Strasbourg, France). Chloramphenicol andchloramphenicol disodium dissuccinate were purchasedfrom the Ph. Eur. and chloramphenicol sodium succinatefrom Pharm-Inter (Brussels, Belgium). Nimesulide,N-(2,4-dinitro-6-phenoxyphenyl)methanesulphonamide (impurityA), N-(2-phenoxyphenyl)methanesulphonamide (impurityB), 2-phenoxyaniline (impurity C), 4-nitro-2-phenoxyaniline(impurity D) andN,N-bis(methylsulphonyl)-2-phenoxyaniline(impurity E) were obtained from the Ph. Eur. Clindamycin wasfrom Alpha Pharma (Zwevegem, Belgium), while lincomycin,lincomycin B, 7-epilincomyin and 7-epiclindamycin wereobtained from Pharmacia Upjohn (Kalamazoo, Michigan,USA). Phenoxymethylpenicillin (pen V) was obtained fromBiochemie (Kundl, Austria), benzylpenicillin (sodium salt)and a mixture of two penilloic acids (sodium salt) from Gist-Brocades (Delft, the Netherlands), phenoxyacetic acid fromA illin( atew lphaw eda ottL cins

ce-t LCg asef siumd mon er-m erseB diums illedt

2

CA,U –visd atoryS asa d bi

m-p perfT cordi m-

marised below:

Analysis of ASA: This separation has been described already indetail in previous papers[40,41].Analysis of clindamycin: The mobile phase consisted of ace-tonitrile: 6.8 g/l potassium dihydrogen phosphate adjusted topH 7.5 with a 250 g/l solution of potassium hydroxide (40:60,v/v). The column was maintained at 30◦C and the detec-tion wavelength was 210 nm. The clindamycin hydrochlo-ride sample contained clindamycin (88.6% m/m), lincomycin(5.0%), lincomycin B (0.9%), 7-epilincomyin (3.0%) and 7-epiclindamycin (2.5%). 15 mg of this sample was dissolved in10 ml of mobile phase.Analysis of buflomedil: The mobile phase consisted of acetoni-trile: 9.25 g/l potassium dihydrogen phosphate adjusted to pH2.5 with phosphoric acid (35:65, v/v). The column was main-tained at 40◦C and the detection wavelength was 210 nm. Thesample contained buflomedil (92% m/m), impurity A (3%) andimpurity B (5%) and 30 mg of this sample was dissolved in10 ml of mobile phase.Analysis of chloramphenicol sodium succinate: The mobilephase consisted of methanol: 20 g/l phosphoric acid–water(40:5:55, v/v). The column was maintained at 25◦C and detec-tion was performed at 275 nm. The sample contained chloram-phenicol sodium succinate (93% m/m), chloramphenicol (3%)and chloramphenicol disodium disuccinate (4%). 2.5 mg of this

heni-and

ni-o pHnedined),).

ase.n-ium

l ofheos-ro-col-atxys-ihy-pto-d incond

M:55,in-Theiumlpe-

d in

cros (Geel, Belgium) and 4-hydroxy-phenoxymethylpenic4-OH pen V) from the Ph. Eur. Dihydrostreptomycin sulphas a house standard and dihydrodeoxystreptomycin suas prepared according to[48]. Streptidine was preparccording to[49], hydroxystreptidine sulphate was from Abbaboratories (North Chicago, IL, USA) and streptomyulphate from VMD (Arendonk, Belgium).

All solvents and reagents were of Ph. Eur. quality. Aonitrile (Biosolve, Valkenswaard, the Netherlands) was ofrade, other chemicals of AR grade. Methanol was purch

rom Fisher Chemicals (Leicestershire, England), potasihydrogen phosphate from Fluka (Buchs, Switzerland), amium dihydrogen phosphate from Ferak Laborat (Berlin, Gany) and phosphoric acid from Janssen Chimica (Beelgium). Sodium octanesulfonate was from Acros and soulphate from Merck (Darmstadt, Germany). Water was distwice before use.

.3. Chromatographic conditions

Analyses were carried out using a Varian (Walnut Creek,SA) 9010 LC pump, a 9100 autosampler and a 9050 UVetector with ChromPerfect 4.4.0 software (Justice Laboroftware, Fife, UK) for data acquisition. The laboratory wir-conditioned and the column temperature was maintaine

mmersion in a water bath at 25◦C.The separations of ASA, clindamycin, buflomedil, chlora

henicol sodium succinate, nimesulide and pen V wereormed according to the respective Ph. Eur. monographs[47].he separation of dihydrostreptomycin was carried out ac

ng to the literature[42]. Chromatographic conditions are su

te

d

-

,

y

-

-

sample was dissolved in 10 ml of mobile phase. Chlorampcol succinate is a mixture of chloramphenicol-1 succinatechloramphenicol-3 succinate.Analysis of nimesulide: The mobile phase consisted of acetotrile: 1.15 g/l ammonium dihydrogen phosphate adjusted t7.0 with ammonia (35:65, v/v). The column was maintaiat 25◦C and detection was at 230 nm. The sample contanimesulide (29% m/m), impurity A (3%), impurity B (11%impurity C (21%), impurity D (23%) and impurity E (13%10 mg of this sample was dissolved in 10 ml of mobile phAnalysis of dihydrostreptomycin: The mobile phase cosisted of an aqueous solution containing 4 g/l of sodsulphate: 1.5 g/l of sodium octanesulfonate; 120 ml/acetonitrile–50 ml/l of 0.2 M phosphate buffer pH 3.0. Tlatter was prepared by mixing a 0.2 M solution of phphoric acid and a 0.2 M solution of potassium dihydgen phosphate until a pH of 3.0 was achieved. Theumn was maintained at 45◦C and detection was done205 nm. The sample contained streptidine (2%), hydrotreptidine sulphate (3%), streptomycin sulphate (5%), ddrostreptomycin sulphate (86%) and dihydrodeoxystremycin sulphate (4%). 5.6 mg of this sample was dissolve5.0 ml of water. The sample was prepared on every seday.Analysis of pen V: The mobile phase consisted of 0.5phosphate buffer solution pH 3.5, methanol–water (5:40v/v). The flow rate was 1.0 ml/min. The column was matained at 25◦C and the detection wavelength was 225 nm.sample consisted of pen V (48.9%), benzylpenicillin sod(30%), phenoxyacetic acid (1%), sodium phenoxymethynilloates (20%) and 4-OH pen V (0.1%). It was dissolve

D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114 107

the dissolution mixture as described in the Ph. Eur. monograph[42] to obtain a 1 mg/ml solution.

The chromatographic conditions prescribed by the mono-graphs may be adjusted when necessary to reach the SST limits.However, as the aim of this study was to compare different typesof RP-LC C18 columns, the mobile phase composition wasnever changed nor were the other conditions. Columns wereequilibrated for 30-90 min dependent on their length. Heliumwas used to degas the mobile phases. A 1 ml/min flow rate wasemployed and 20�l were injected. Uracil was used to determinethe dead volume. The chromatograms were recorded three times.Chromatographic parameters were calculated with ChromPer-fect 4.4.0. software. Nomenclature of Ph. Eur. was used. Theresolution (Rs) was calculated as described in the Ph. Eur.[42].

3. Results and discussion

3.1. Column selectivity in separations of pharmaceuticals

Seven different representative separations, from which six aredescribed by the Ph. Eur., were performed on 56 RP-LC C18 sta-tionary phases. As the Ph. Eur. monographs do not prescribe thebrand names of the stationary phases to be used, analysts haveto select a suitable column themselves. Once a chromatogra-pher has selected a column, it has to be checked for compliancew umr hicha ter-m ASA,c iums hosec r thes itablea alsog

alityo terd CRFc it cann es thea nce,w

3c

usseda er nicer ups.I CA.F tiono ft alues.I ch oft tually

Table 2Column ranking according to theF-values, relative to the mean parameter values(k′

amb: 6.02,rk′ba/ph 2.7: 0.066,rk′

tri/ter: 1.42,k′2,2′-d: 9.27) for the separation of

acetylsalicylic acid (ASA)

No. Column F-values CRF

51 Tracerexcel 5 0.067 1.0058 Validated C18 0.068 1.0057 Uptispher ODB5 0.083 1.0029 Nucleosil HD 0.089 1.0013 Genesis C18-3 0.108 0.7450 Tracerexcel 3 0.155 1.0056 Uptispher ODB3 0.193 0.9468 Zorbax Extend C18 0.249 1.0060 Wakosil HG 5 25 0.358 1.0065 YMC-Pack-Pro C18-3 0.374 1.0067 Zorbax Eclipse XDB 0.433 1.0052 TSKgel ODS-80TS 0.460 0.8459 Wakosil HG5 10 0.476 0.6654 Uptispher HDO3 0.496 1.0049 Symmetry 0.524 1.00

7 Aqua 5 0.569 1.0066 YMC-Pack-Pro C18-5 0.585 1.0061 Wakosil RS 3 10 0.689 0.8947 Superspher 0.708 1.0055 Uptispher HDO5 0.776 1.0022 Kromasil NM 0.932 1.0037 Prodigy 3 1.052 0.81

3 Alltima 3 1.057 0.001 ACE C18-3 1.076 0.934 Alltima 5 1.136 1.00

23 Kromasil EKA 1.155 1.0069 Zorbax SB-C18 1.164 1.0040 Purospher Star 1.300 1.00

2 ACE C18-5 1.613 1.0026 Luna 5 1.630 1.00

31 OmniSpher 2.163 1.0032 Pecosphere 2.224 0.5045 Supelcosil LC-18 DB 3 2.293 0.8664 YMCHydrospherCW 2.468 1.0046 Supelcosil LC-18 DB 5 2.508 0.7116 Hypersil BDS 3.295 0.9128 Nucleosil NM 3.581 1.00

9 Brava BDS 3 3.981 0.6310 Brava BDS 5 4.272 0.6739 Purospherendcapped 4.626 1.0012 Discovery 5.183 0.9619 HyPurity Elite 3 5.202 0.8020 HyPurity Elite 5 5.335 0.66

38 Purospher 6.114 0.0033 Platinum 3 6.822 0.4034 Platinum 5 7.150 0.2441 Spheri 7.356 0.9643 Spherisorb ODS2 7.680 1.0025 LiChrospher 8.821 1.0030 Nucleosil HD Nautilus 9.412 0.0035 Platinum EPS 3 17.284 0.0936 Platinum EPS 5 18.004 0.4218 Hypersil ODS 19.832 0.88

5 Apex Basic 22.611 0.0044 Supelcosil LC18 26.100 0.00

6 Apex ODS 26.492 0.67

The long columns (0.25 m) are indicated in bold, the short ones in italic. Theinternal diameter is always 4.6 mm.

ith the SST requirements. SST usually asks for a minimesolution between two compounds, often compounds wre difficult to separate (“critical pair”). SST values were deined whenever a SST was prescribed by the Ph. Eur. (

lindamycin, buflomedil, chloramphenicol, nimesulide, soduccinate and pen V). According to the monographs, only tolumns, which comply with the SST, can be employed foeparation. We will check whether all these columns are sund if other columns (not compliant with the SST) maybeive a good separation.

The CRF was introduced as a criterion to evaluate the quf the separations[40]. It has to be noted that this parameoes not take into account the peak shape. Although theould be used to evaluate the separations in this study,ever be prescribed in practice as a SST, as the CRF requirvailability of the potential impurities as a reference substahich is often not the case in daily practice.

.2. Separation of acetylsalicylic acid and its relatedompounds

As mentioned above, this separation has been disclready in previous papers[40,41]. The interpretation of thesults was intrinsically based on a PCA plot. Althoughesults were obtained, it was difficult to define distinctive gron this paper, the data will be reanalysed without the use of Pirst, all columns yielding a CRF equal to 1 for the separaf ASA and its impurities, were selected (seeTable 2). Each o

hese columns was also characterised by the 4 parameter vn order to define a virtual ideal column, mean values for eahe 4 parameters were calculated after checking for even

108 D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114

Table 3Ratios of suitable columns for the different separations, for the group of highranked columns (F < 2), the intermediate group (2 <F < 6) and the group of lowranked columns (F > 6)

Separation F < 2 2 <F < 6 F > 6

Acetylsalicylic acida 16/16 3/7 2/11Clindamycina 16/16 5/8 4/10Buflomedil 30/30 15/15 5/11Chloramphenicol 31/31 14/15 7/10Nimesulideb 16/16 5/5 0/1Dihydrostreptomycin 32/32 12/14 6/10Phenoxymethylpenicillin 29/33 9/13 3/10

a Only 0.25 m columns.b Only short columns.

outlying values using a Grubbs test (α = 0.05). So, columns 25and 43 were removed from the data set to determine the 4 meanparameter values of the columns with CRF = 1. Next, theF-values of all columns were calculated using the mean parametervalues of the virtual column as reference. The ranking obtained,based on theseF-values, is shown inTable 2(more detailed datacan be provided on request). The columns withF < 2 were con-sidered as high ranked, columns withF > 6 as low ranked andcolumns with 2 <F < 6 as intermediate. The Ph. Eur. prescribesthe use of a 0.25 m column. All 0.25 m columns (16/16, indi-cated in bold inTable 1) with F < 2 show a CRF = 1, i.e. theyshow a baseline separation of ASA and all related substances.the group of low ranked columns only 2 out of 11 (18%) 0.25 mcolumns show a CRF = 1, while in the intermediate group this isthree out of seven (43%) (Table 3). These results are similar tothose obtained previously and further discussion of the analysiof ASA can be found in[40,41].

3.3. Separation of clindamycin hydrochloride and itsrelated compounds

A typical chromatogram is presented inFig. 1. The SST ofthe Ph. Eur. monograph for clindamycin mentions only relativeretention times of the impurities compared to the main com-pound. Problems with the use of relative retention times haveb as

F mn:P

not considered as sufficient to decide on the quality of a separa-tion.

As described above, the parameter values of a virtual idealcolumn were calculated using the columns with a CRF = 1 forthe separation of clindamycin and its impurities. Here, columns18, 25 and 44 were found to be outliers. The average of eachchromatographic parameter was calculated for the remainingcolumns with a CRF = 1 and used as reference values to calculatetheF-values. The ranking according to theseF-values is shownin Table 4and an overview is given inTable 3. As for ASA,the Ph. Eur. prescribes a 0.25 m column. All high ranked 0.25 mcolumns withF < 2 show baseline separation (16/16), comparedto 63% (5/8) of the intermediate (2 <F < 6) 0.25 m columns and40% (4/10) of the low ranked (F > 6) 0.25 m columns. This showsthat the column ranking can help analysts with their columnselection for the analysis of clindamycin. Analysts preferablyshould select a 0.25 m column withF < 2 (relative to the meanparameter values) in order to obtain a baseline separation ofclindamycin and its related substances.

As the Ph. Eur. allows a deviation of the prescribed columnlength by±70%, 0.15 m columns may be selected for this anal-ysis too. However, this study, as well as the previous one, showsthat these 0.15 m columns often give somewhat less good sep-arations, while 0.25 m columns of the same type show overallbaseline separations. Examples are column pairs 1–2, 54–55,56–57 and 59–60 (Table 4). This suggests that the allowed col-u

efer-e n ofT ond-i ), thec bef oft roups orts aH clin-d edc ni y twoh ndo 5 and1 aram-e sucha ilar,h umnc ns.

3

eP urityB lfilt , 38,5 hesec and2 Theo ty

een discussed elsewhere[50], and therefore this parameter w

ig. 1. Separation of clindamycin hydrochloride and its impurities. Coluurospher Star (No. 40). Other conditions: see Section2.

In

s

mn length variation of 70% may be too large.It is clear that the selection of the reference column or r

nce values is of major importance. If a high ranked columable 4is chosen as a reference column (or if the corresp

ng parameter values are proposed in a method descriptionolumn ranking is similar. However, suitable columns canound among the low ranked columns as well. As only 40%hese low ranked columns are suitable, columns from this ghould better not be reported as suitable. Pharmeuropa repypersil ODS (column 18) as suitable for the separation ofamycin[51]. This column belongs to the group of low rankolumns in the column ranking (Table 4). When this colums used as a reference column in the ranking system, onligh ranked columns (F < 2, columns 6 and 44) are found anly one of these shows a baseline separation (CRF = 0.9.00, respectively). Indeed, the differences between the pter values of these low ranked columns are larger and ifcolumn is selected as reference column, only a few sim

igh ranked columns will be found. This shows that the collassification system is not omnipotent and has its limitatio

.4. Separation of buflomedil and its related compounds

A typical chromatogram is shown inFig. 2. The SST of thh. Eur. asks for a resolution between buflomedil and imp(Rsbuf) of more than 5.0. The following columns did not fu

his criterion: columns 5, 9, 13, 19, 20, 25, 26, 30, 32, 372, 54, 56, 59, 61, 64, 65, 66 and 68. Although none of tolumns was suitable according to the SST, only columns 55 did not give baseline separation of all peaks (CRF < 1).ther columns gave a Rsbuf below 5, but buflomedil and impuri

D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114 109

Table 4Column ranking according to theF-values, relative to the mean parameter values(k′

amb: 5.84,rk′ba/ph 2.7: 0.066,rk′

tri/ter: 1.43,k′2,2′-d: 9.03) for the separation of

clindamycin

No. Column F-values CRF

29 Nucleosil HD 0.063 1.0058 Validated C18 0.070 1.0051 Tracerexcel 5 0.087 1.0056 Uptispher ODB3 0.118 0.9857 Uptispher ODB5 0.156 1.0013 Genesis C18-3 0.182 1.0068 Zorbax Extend C18 0.206 1.0060 Wakosil HG 5 25 0.235 1.0050 Tracerexcel 3 0.250 1.0059 Wakosil HG 5 10 0.335 0.9665 YMC-Pack-Pro C18-3 0.348 1.0067 Zorbax Eclipse XDB 0.453 1.0052 TSKgel ODS-80TS 0.483 1.0054 Uptispher HDO3 0.505 0.9849 Symmetry 0.533 1.00

7 Aqua 5 0.612 1.0066 YMC-Pack-Pro C18-5 0.621 1.0047 Superspher 0.702 1.0055 Uptispher HDO5 0.722 1.00

1 ACE C18-3 0.822 0.9861 Wakosil RS 3 10 0.885 1.0022 Kromasil NM 0.889 1.0037 Prodigy 3 1.124 1.0069 Zorbax SB-C18 1.140 1.00

2 ACE C18-5 1.296 1.003 Alltima 3 1.303 1.00

23 Kromasil EKA 1.305 1.004 Alltima 5 1.397 1.00

40 Purospher Star 1.496 1.0026 Luna 5 1.699 1.0045 Supelcosil LC-18 DB 3 1.965 0.93

46 Supelcosil LC-18 DB 5 2.182 0.9831 OmniSpher 2.286 1.0064 YMC Hydrospher C18 2.299 1.0032 Pecosphere 2.488 0.8316 Hypersil BDS 2.839 1.0028 Nucleosil NM 3.445 1.00

9 Brava BDS 3 3.520 0.9610 Brava BDS 5 3.749 0.9712 Discovery 4.602 1.0019 HyPurity Elite 3 4.612 0.9820 HyPurity Elite 5 4.736 0.9539 Purospher endcapped 4.925 1.0038 Purospher 5.811 0.89

33 Platinum 3 6.362 0.9734 Platinum 5 6.718 0.8341 Spheri 7.837 0.9343 Spherisorb ODS2 7.904 0.9430 Nucleosil HD Nautilus 8.742 1.0025 LiChrospher 9.071 1.0035 Platinum EPS 3 16.463 0.0036 Platinum EPS 5 17.206 0.0018 Hypersil ODS 19.902 1.00

5 Apex Basic 21.696 0.0044 Supelcosil LC18 26.198 1.00

6 Apex ODS 26.568 0.95

The long columns (0.25 m) are indicated in bold, the short ones in italic. Theinternal diameter is always 4.6 mm.

Fig. 2. Separation of buflomedil hydrochloride and its related compounds. Col-umn: Zorbax Eclipse (No. 67). Other conditions: see Section2.

B were still baseline separated. Columns 6, 18, 41 and 44, whichcomplied with the SST, did not give overall baseline separation.Although these columns gave sufficient separation of buflomediland impurity B, they did not separate buflomedil and impurityA.

This shows that a simple SST as prescribed by the Ph. Eur. isnot always providing the right information. Therefore, it waschecked whether the column classification could help in theselection of a suitable column. The same procedure as above wasapplied. From the columns giving baseline separation, columns35, 36 and 43 were removed since they were found to be outliers.The 4 mean parameter values of the remaining columns wereused as reference values in the column ranking. Results are pre-sented inTable 5and summarized inTable 3. All columns withF < 6 showed a CRF = 1.Fig. 2provides a typical chromatogramon column 67, withF = 0.577. In the group of the low rankedcolumns, only 45% (5/11) showed a CRF of 1. More columns(broaderF-range) give complete baseline separations comparedto the separation of clindamycin, which can be explained bythe fact that only three compounds have to be separated. Thisalso results in less influence of column length. It can be con-cluded that the column ranking system can help analysts withtheir column selection for this analysis too.

3.5. Separation of chloramphenicol sodium succinate andits related compounds

w l andc m thet andc thisr s hadt 5, 36a para-t col-1s elines

sis.T l col-u on the

A typical chromatogram can be seen inFig. 3. To complyith the SST, the peaks corresponding to chloramphenicohloramphenicol disuccinate must be clearly separated frowo peaks corresponding to chloramphenicol-1 succinatehloramphenicol-3 succinate. Only column 5 did not fulfilequirement. However, based on the CRF values, 4 columno be ranked as “not suitable” for this analysis: columns 5, 3nd 38. Although columns 35, 36 and 38 showed a clear se

ion in the SST, the peaks corresponding to chlorampheniuccinate and chloramphenicol-3 succinate were not baseparated.

Only 4 out of 56 columns were not suitable for this analyo calculate the mean parameter values of the virtual ideamn, columns 6, 18, 30, 41, 43 and 44 were deleted based

110 D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114

Table 5Column ranking according to theF-values, relative to the mean parameter values(k′

amb: 5.30,rk′ba/ph 2.7: 0.078,rk′

tri/ter: 1.43,k′2,2′-d: 8.64) for the separation of

buflomedil

No. Column F-values CRF

56 Uptispher ODB3 0.030 1.0060 Wakosil HG 5 25 0.080 1.0059 Wakosil HG5 10 0.158 1.0058 Validated C18 0.182 1.0029 Nucleosil HD 0.205 1.0051 Tracerexcel 5 0.259 1.0065 YMC-Pack-Pro C18-3 0.377 1.0068 Zorbax Extend C18 0.378 1.00

1 ACE C18-3 0.426 1.0057 Uptispher ODB5 0.467 1.0013 Genesis C18-3 0.495 1.0054 Uptispher HDO3 0.517 1.0052 TSKgel ODS-80TS 0.543 1.0067 Zorbax Eclipse XDB 0.577 1.0050 Tracerexcel 3 0.597 1.0055 Uptispher HDO5 0.670 1.00

7 Aqua 5 0.697 1.002 ACE C18-5 0.745 1.00

66 YMC-Pack-Pro C18-5 0.775 1.0049 Symmetry 0.929 1.0069 Zorbax SB-C18 0.972 1.0047 Superspher 1.016 1.0045 Supelcosil LC-18 DB 3 1.169 1.0022 Kromasil NM 1.174 1.0037 Prodigy 3 1.261 1.0046 Supelcosil LC-18 DB 5 1.340 1.0061 Wakosil RS 3 10 1.407 1.0064 YMC Hydrospher C18 1.754 1.0026 Luna 5 1.784 1.0016 Hypersil BDS 1.871 1.00

23 Kromasil EKA 2.068 1.003 Alltima 3 2.089 1.00

40 Purospher Star 2.238 1.004 Alltima 5 2.244 1.009 Brava BDS 3 2.377 1.00

10 Brava BDS 5 2.552 1.0032 Pecosphere 3.011 1.0031 OmniSpher 3.093 1.0028 Nucleosil NM 3.235 1.0012 Discovery 3.344 1.0019 HyPurity Elite 3 3.385 1.0020 HyPurity Elite 5 3.528 1.0033 Platinum 3 4.873 1.0034 Platinum 5 5.235 1.0038 Purospher 5.613 1.00

39 Purospher endcapped 6.098 1.0030 Nucleosil HD Nautilus 7.740 1.0043 Spherisorb ODS2 8.270 1.0041 Spheri 8.830 0.0025 LiChrospher 9.878 0.0035 Platinum EPS 3 14.416 1.0036 Platinum EPS 5 15.187 1.0018 Hypersil ODS 19.170 0.00

5 Apex Basic 20.370 0.9444 Supelcosil LC18 25.554 0.00

6 Apex ODS 25.819 0.00

The long columns (0.25 m) are indicated in bold, the short ones in italic. Theinternal diameter is always 4.6 mm.

Fig. 3. Separation of chloramphenicol sodium succinate and its related com-pounds. Column: TSKgel ODS-80TS (No. 52). Other conditions: see Section2.

Grubbs test. The obtained values were used as references andthe columns were ranked according to theirF-value. Resultsare presented inTable 6and they are summarized inTable 3.All columns withF < 2 showed a complete baseline separation(CRF = 1). 93% (14/15) of the columns with 2 <F < 6 showed aCRF of 1, while this percentage is decreasing to 70% (7/10) forthe low ranked columns (F > 6).

3.6. Separation of nimesulide and related compounds

A typical chromatogram can be seen inFig. 4. The Ph. Eur.prescribes a 0.125 m column for this analysis and allows theadaptation of the stationary phase column length by±70%.Therefore, no 0.25 m column may be selected and only the shortcolumns in our study (presented in italics inTable 1) can betaken into account. The SST of the Ph. Eur. prescribes a res-olution of more than 2.0 between impurity C and impurity D.According to this requirement all short columns are suitable forthis analysis. However, short column 35 did not give an overallbaseline separation of all compounds although it complied withthe SST. Moreover, the selectivity of this column was differentas the elution order of the last two peaks, corresponding to impu-rities D and E, was reversed. As the SST only takes into accountthe resolution between impurities C and D, this peak sequence

F PackP

ig. 4. Separation of nimesulide and its related compounds. Column: YMCro 3 (No. 65). Other conditions: see Section2.

D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114 111

Table 6Column ranking according to theF-values, relative to the mean parameter values(k′

amb: 5.39,rk′ba/ph 2.7: 0.082,rk′

tri/ter: 1.42,k′2,2′-d: 8.83) for the separation of

chloramphenicol

No. Column F-values CRF

56 Uptispher ODB3 0.013 1.0060 Wakosil HG 5 25 0.108 1.0058 Validated C18 0.125 1.0029 Nucleosil HD 0.181 1.0051 Tracerexcel 5 0.195 1.0059 Wakosil HG 5 10 0.203 1.0065 YMC-Pack-Pro C18-3 0.340 1.0057 Uptispher ODB5 0.372 1.0068 Zorbax Extend C18 0.381 1.0013 Genesis C18-3 0.398 1.0054 Uptispher HDO3 0.439 1.0052 TSKgel ODS-80TS 0.460 1.0050 Tracerexcel 3 0.486 1.0067 Zorbax Eclipse XDB 0.507 1.00

1 ACE C18-3 0.545 1.007 Aqua 5 0.596 1.00

55 Uptispher HDO5 0.646 1.0066 YMC-Pack-Pro C18-5 0.699 1.0069 Zorbax SB-C18 0.896 1.00

2 ACE C18-5 0.901 1.0049 Symmetry 0.917 1.0047 Superspher 0.996 1.0037 Prodigy 3 1.142 1.0022 Kromasil NM 1.191 1.0061 Wakosil RS 3 10 1.225 1.0045 Supelcosil LC-18 DB 3 1.285 1.0046 Supelcosil LC-18 DB 5 1.444 1.0026 Luna 5 1.663 1.0064 YMC Hydrospher C18 1.768 1.00

3 Alltima 3 1.896 1.0023 Kromasil EKA 1.970 1.00

4 Alltima 5 2.044 1.0040 PurospherStar 2.086 1.0016 Hypersil BDS 2.089 1.00

9 Brava BDS 3 2.579 1.0032 Pecosphere 2.740 1.0010 Brava BDS 5 2.802 1.0031 OmniSpher 3.029 1.0028 Nucleosil NM 3.245 1.0012 Discovery 3.652 1.0019 HyPurity Elite 3 3.706 1.0020 HyPurity Elite 5 3.862 1.0033 Platinum 3 4.999 1.0034 Platinum 5 5.328 1.0038 Purospher 5.815 0.6639 Purospher endcapped 5.912 1.00

43 Spherisorb ODS2 7.984 1.0030 Nucleosil HD Nautilus 8.186 1.0041 Spheri 8.396 1.0025 LiChrospher 9.656 1.0035 Platinum EPS 3 14.773 0.9236 Platinum EPS 5 15.520 0.9418 Hypersil ODS 18.800 1.00

5 Apex Basic 21.005 0.0044 Supelcosil LC18 25.146 1.00

6 Apex ODS 25.416 1.00

The long columns (0.25 m) are indicated in bold, the short ones in italic. Theinternal diameter is always 4.6 mm.

Table 7Column ranking according to theF-values, relative to the mean parameter values(k′

amb: 5.29,rk′ba/ph 2.7: 0.079,rk′

tri/ter: 1.44,k′2,2′-d: 8.87) for the separation of

nimesulide

No. Column F-values CRF

56 Uptispher ODB3 0.028 1.0060 Wakosil HG 5 25 0.129 1.0058 Validated C18 0.157 1.0029 NucleosilHD 0.217 0.8759 Wakosil HG5 10 0.221 1.0051 Tracerexcel 5 0.289 1.0068 Zorbax Extend C18 0.400 0.9465 YMC-Pack-Pro C18-3 0.456 1.00

1 ACE C18-3 0.466 1.0057 Uptispher ODB5 0.487 1.0013 Genesis C18-3 0.521 1.0054 Uptispher HDO3 0.571 1.0052 TSKgel ODS-80TS 0.601 1.0050 Tracerexcel 3 0.624 1.0067 Zorbax Eclipse XDB 0.667 1.00

7 Aqua 5 0.758 1.002 ACEC 18-5 0.795 1.00

55 Uptispher HDO5 0.796 1.0066 YMC-Pack-Pro C18-5 0.869 1.0049 Symmetry 0.882 1.0047 Superspher 0.942 1.0069 Zorbax SB-C18 1.045 1.0022 Kromasil NM 1.078 1.0045 Supelcosil LC-18 DB 3 1.254 1.0037 Prodigy 3 1.371 1.0046 Supelcosil LC-18 DB 5 1.431 1.0067 Wakosil RS3 10 1.436 1.0016 Hypersil BDS 1.879 1.0064 YMC Hydrospher C18 1.880 1.0026 Luna 5 1.914 1.00

3 Alltima 3 1.975 1.00

23 Kromasil EKA 2.008 1.0040 PurospherStar 2.081 1.00

4 Alltima 5 2.137 1.009 Brava BDS 3 2.488 1.00

10 Brava BDS 5 2.602 1.0032 Pecosphere 2.863 1.0028 Nucleosil NM 2.965 1.0031 OmniSpher 2.983 1.0012 Discovery 3.457 1.0019 HyPurity Elite 3 3.475 1.0020 HyPurity Elite 5 3.609 1.0033 Platinum 3 4.978 1.0034 Platinum 5 5.280 1.0038 Purospher 5.329 1.0039 Purospher endcapped 5.813 1.00

30 Nucleosil HD Nautilus 7.600 *1,0043 Spherisorb ODS2 7.882 1.0041 Spheri 8.509 1.0025 LiChrospher 9.413 1.0035 Platinum EPS 3 14.160 *0.5436 Platinum EPS 5 14.901 *0.4818 Hypersil ODS 18.840 1.00

5 Apex Basic 20.023 0.0044 Supelcosil LC18 25.166 1.00

6 Apex ODS 25.442 1.00

The long columns (0.25 m) are indicated in bold, the short ones in italic. Theinternal diameter is always 4.6 mm.

112 D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114

problem cannot be detected in the SST. This may lead to wrongpeak identification and shows the limitation of a SST based onone critical peak pair alone[50]. The 0.25 m columns 30 and 36showed a similar selectivity problem. Although column 30 gavea baseline separation of all peaks, the elution order of the lasttwo peaks was reversed.

After removing outlying columns (6, 18, 41, 43, 44), meanparameter values of the group of columns with CRF = 1 for thisseparation were taken as reference values. Results are presentedin Table 7andTable 3. All short columns withF < 6 showedbaseline separation for all peaks while columns with changedselectivity were among the low ranked columns (F > 6), indi-cating again the relationship between the column classificationsystem and column selectivity in real separations.

Pharmeuropa reports two suitable columns for this separa-tion, a Lichrospher 100 RP-18 and a Capcell Pak C-18[52].However, none of these columns was available in this study.

3.7. Separation of dihydrostreptomycin sulphate and itsrelated compounds

A representative chromatogram is shown inFig. 5. This sep-aration was performed according to the literature[47].

According to the CRF values, only columns 5, 30, 33, 34,35 and 36 have to be ranked as “not suitable”. As for the otherseparations, all columns with a CRF = 1 were tested for outliers( alueswA ion,wa

3c

S lpeni-c nt,c . Note takeni 38,

F lumn:D

Table 8Column ranking according to theF-values, relative to the mean parameter values(k′

amb: 5.51,rk′ba/ph 2.7: 0.076,rk′

tri/ter: 1.44,k′2,2′-d: 8.91) for the separation of

dihydrostreptomycin

No. Column F-values CRF

56 Uptispher ODB3 0.034 1.0058 Validated C18 0.103 1.0029 Nucleosil HD 0.122 1.0060 Wakosil HG 5 25 0.144 1.0051 Tracerexcel 5 0.193 1.0059 Wakosil HG 5 10 0.240 1.0068 Zorbax Extend C18 0.284 1.0057 Uptispher ODB5 0.339 1.0013 Genesis C18-3 0.368 1.0065 YMC-Pack-Pro C18-3 0.405 1.0050 Tracerexcel 3 0.462 1.0054 Uptispher HDO3 0.546 1.0052 TSKgel ODS-80TS 0.555 1.00

1 ACE C18-3 0.561 1.0067 Zorbax Eclipse XDB 0.573 1.0049 Symmetry 0.699 1.00

7 Aqua 5 0.704 1.0055 Uptispher HDO5 0.753 1.0066 YMC-Pack-Pro C18-5 0.774 1.0047 Superspher 0.788 1.00

2 ACE C18-5 0.944 1.0022 Kromasil NM 0.946 1.0069 Zorbax SB-C18 1.094 1.0061 Wakosil RS 3 10 1.225 1.0037 Prodigy 3 1.286 1.0045 Supelcosil LC-18 DB 3 1.493 1.0023 Kromasil EKA 1.680 1.0046 Supelcosil LC-18 DB 5 1.690 1.00

3 Alltima 3 1.691 1.0040 Purospher Star 1.817 1.00

4 Alltima 5 1.820 1.0026 Luna 5 1.854 1.00

64 YMC Hydrospher C18 2.064 1.0016 Hypersil BDS 2.196 1.0031 OmniSpher 2.639 1.0032 Pecosphere 2.714 1.00

9 Brava BDS 3 2.870 1.0010 Brava BDS 5 3.005 1.0028 Nucleosil NM 3.102 1.0012 Discovery 3.858 1.0019 HyPurity Elite 3 3.864 1.0020 HyPurity Elite 5 3.990 1.0039 Purospher endcapped 5.406 1.0038 Purospher 5.438 1.0033 Platinum 3 5.517 0.6734 Platinum 5 5.842 0.41

43 Spherisorb ODS2 7.847 1.0030 Nucleosil HD Nautilus 7.944 0.0041 Spheri 8.221 1.0025 LiChrospher 9.211 1.0035 Platinum EPS 3 14.929 0.0036 Platinum EPS 5 15.668 0.0018 Hypersil ODS 19.199 1.00

5 Apex Basic 20.525 0.0044 Supelcosil LC18 25.466 1.00

6 Apex ODS 25.785 1.00

The long columns (0.25 m) are indicated in bold, the short ones in italic. Theinternal diameter is always 4.6 mm.

columns 6, 18, 41, 43 and 44) and the mean parameter vere taken as reference values. The result is shown inTable 8.ll columns withF < 2 showed an overall baseline separathile for the lower ranked columns 86% (12/14) for 2 <F < 6nd 60% (6/10) forF > 6 were found suitable (Table 3).

.8. Separation of phenoxymethylpenicillin and its relatedompounds

A representative chromatogram is shown inFig. 6. TheST prescribes a resolution between pen V and benzyillin (RspenV) of more than 6.0. According to this requiremeolumns 5, 32, 34, 35, 36 and 38 are ranked as not suitablexactly the same result is obtained if the CRF values are

nto account (Table 9). Columns 5, 25, 30, 32, 33, 34, 35, 36,

ig. 5. Separation of dihydrostreptomycin sulphate and its impurities. Coiscovery (No. 12). Other conditions: see Section2.

D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114 113

Table 9Column ranking according to theF-values, relative to the mean parameter values(k′

amb: 5.58,rk′ba/ph 2.7: 0.071,rk′

tri/ter: 1.43,k′2,2′-d: 8.66) for the separation of

penicillin V

No. Column F-values CRF

56 Uptispher ODB3 0.052 1.0029 Nucleosil HD 0.091 1.0060 Wakosil HG 5 25 0.100 1.0058 Validated C18 0.124 1.0051 Tracerexcel 5 0.141 1.0059 Wakosil HG 5 10 0.179 0.7468 Zorbax Extend C18 0.231 0.5357 Uptispher ODB5 0.287 1.0065 YMC-Pack-Pro C18-3 0.310 1.0013 Genesis C18-3 0.310 1.0050 Tracerexcel 3 0.401 1.0067 Zorbax Eclipse XDB 0.460 1.0052 TSKgel ODS-80TS 0.489 1.0054 Uptispher HDO3 0.491 1.00

1 ACE C18-3 0.547 1.0055 Uptispher HDO5 0.614 1.00

7 Aqua 5 0.635 1.0066 YMC-Pack-Pro C18-5 0.650 1.0049 Symmetry 0.697 1.0047 Superspher 0.838 1.00

2 ACE C18-5 0.933 1.0022 Kromasil NM 1.013 1.0069 Zorbax SB-C18 1.042 0.0067 Wakosil RS 3 10 1.145 1.0037 Prodigy 3 1.154 1.0045 Supelcosil LC-18 DB 3 1.489 1.0023 Kromasil EKA 1.660 1.0046 Supelcosil LC-18 DB 5 1.689 0.8026 Luna 5 1.704 1.00

3 Alltima 3 1.747 1.004 Alltima 5 1.863 1.00

40 PurospherStar 1.922 1.0064 YMC Hydrospher C18 1.981 1.00

16 Hypersil BDS 2.285 1.0031 OmniSpher 2.667 1.00

9 Brava BDS 3 2.860 1.0032 Pecosphere 2.867 0.3410 Brava BDS 5 3.069 1.0028 Nucleosil NM 3.449 1.0012 Discovery 3.844 1.0019 HyPurity Elite 3 3.870 1.0020 HyPurity Elite 5 4.001 1.0033 Platinum 3 5.585 0.8739 Purospher endcapped 5.609 1.0038 Purospher 5.753 0.0034 Platinum 5 5.988 0.61

30 Nucleosil HD Nautilus 8.159 0.0043 Spherisorb ODS2 8.321 1.0041 Spheri 8.554 0.2025 LiChrospher 9.695 0.7335 Platinum EPS 3 15.482 0.8636 Platinum EPS 5 16.259 0.8818 Hypersil ODS 19.812 1.00

5 Apex Basic 20.992 0.0044 Supelcosil LC18 26.159 0.72

6 Apex ODS 26.479 1.00

The long columns (0.25 m) are indicated in bold, the short ones in italic. Theinternal diameter is always 4.6 mm.

Fig. 6. Separation of phenoxymethylpenicillin and its impurities. Column:Uptishere 3 HDO (No. 54). Other conditions: see Section2.

41, 44, 46, 59, 68 and 69 give a CRF < 1. The CRF criterion isstricter, as nine more columns, estimated suitable according tothe SST, do not give baseline separation. Again, the SST doesnot always provide sufficient information.

For this separation, the column classification does not dis-criminate perfectly either. When mean parameter values ofcolumns with CRF = 1 (columns 6, 18, 43 and 45 were foundto be outliers) are used as reference values, 12% of the highranked columns (4/33) do not give a baseline separation, com-pared to 31% of the intermediate columns (4/13) and 70% of thelow ranked columns (7/10). Although the column ranking helpsto exclude most of the unsuitable columns, it does not guaranteethat all high ranked columns are fully suitable for this separation(Tables 9 and 3, overview).

4. Conclusion

This paper discusses column selection in seven separationsof pharmaceuticals. The separations were selected based onisocratic elution and availability of reference substances. Theresults obtained for all performed separations are reported. Itwas shown that the SST as prescribed by the Ph. Eur. gives valu-able information about the suitability of a particular column fora particular separation, but this information is not always suffi-cient. This is because most often it is restricted to the resolutiono ail-a wasu

mns.E h ared .B de.T theirs tionse tions,1 cev olumnl ed ina dc ted

f one pair of components, which is due to the limited avbility of reference substances. In this study, the CRF valuesed to evaluate the separations.

All separations have been performed on RP-LC C18 coluach column is characterised by 4 column parameters, whicetermined by chromatography and allow to calculateF-valuesased on theseF-values, a ranking of the columns was mahe relationship between the ranking of the columns andeparation performance was investigated. For all separaxamined, it was observed that in six out of seven separa00% of these high ranked columns (F < 2 versus the referenalues) gave baseline separations when the prescribed cength was taken, while this percentage gradually decreasn intermediate group (2 <F < 6) and in a group of low rankeolumns (F > 6). The column length has not been incorpora

114 D. Visky et al. / J. Chromatogr. A 1101 (2006) 103–114

in the calculation of theF-value to keep the approach as simpleas possible and since this parameter is always specified in themethod description. Although theF-value does not allow exactcharacterisation of columns, it is an interesting tool to facilitatethe selection of a suitable RP-LC C18 column.

Of course, the selection of the reference column or referenceparameters is very important. Some of the low ranked columnsgive complete separation as well. If one of these columns isselected as reference column, the corresponding column rank-ing gives much less valuable information. Verification of theperformance of different columns during method developmenttherefore is very important. When needed, the column classifi-cation system can also help to identify columns with differentselectivity, which is useful during method development.

In the future, more types of RP-LC C18 columns will be char-acterized and added to the database. It is the intention to performalso more complex separations which require gradient elution ora specific column selectivity, in order to further investigate therelationship between column ranking and column selectivity.

Acknowledgements

The authors thank the manufacturers and the suppliers for thegift of columns. D. Visky, Zs. Kovacs and K. Koczian thank theMinistry of the Flemish Community for financial support.

E. Haghedooren enjoys a grant of the Institute for the Promo-t ders(

tificR

R

ya,

Lab.

991)

ber

hro-

[ phia

[ 97)

[[ 997)

[ 74.[[ 01.[ arch

[ 003)

[19] M.R. Euerby, P. Petersson, J. Chromatogr. A 994 (2003) 13.[20] N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, R.G. Wolcott, P.W.

Carr, J. Chromatogr. A 961 (2002) 171.[21] N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, P.W. Carr, J. Chro-

matogr. A 961 (2002) 195.[22] N.S. Wilson, J.W. Dolan, L.R. Snyder, P.W. Carr, L.C. Sander, J. Chro-

matogr. A 961 (2002) 217.[23] J.J. Gilroy, J.W. Dolan, L.R. Snyder, J. Chromatogr. A 1000 (2003) 757.[24] J.J. Gilroy, J.W. Dolan, P.W. Carr, L.R. Snyder, J. Chromatogr. A 1026

(2004) 77.[25] N.S. Wilson, J. Gilroy, J.W. Dolan, L.R. Snyder, J. Chromatogr. A 1026

(2004) 91.[26] 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.

[27] 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.

[28] L.R. Snyder, J.W. Dolan, P.W. Carr, J. Chromatogr. A 1060 (2004) 77.[29] D.H. Marchand, K. Croes, J.W. Dolan, L.R. Snyder, J. Chromatogr. A

1062 (2005) 57.[30] 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.

[31] R. Kaliszan, M.A. van Straten, M. Markuszewski, C.A. Cramers, H.A.Claessens, J. Chromatogr. A 855 (1999) 455.

[32] T. Baczek, R. Kaliszan, J. Chromatogr. A 962 (2002) 41.[33] J. Jiskra, H.A. Claessens, C.A. Cramers, R. Kaliszan, J. Chromatogr. A

977 (2002) 193.[34] T. Baczek, R. Kaliszan, J. Chromatogr. A 987 (2003) 29.[35] T. Baczek, R. Kaliszan, K. Novotna, P. Jandera, J. Chromatogr. A 1075

[ B.2002)

[2002)

[ ,gr. A

[ ,s, J.

[ .004)

[ ams,

[ ourg,

[ .[ atogr.

[[ nein,

[ 1999)

[ ato,

[[ t,

[[

ion of Innovation through Science and Technology in FlanIWT-Vlaanderen).

E. Adams is a post-doctoral fellow of the Fund for Scienesearch (FWO), Flanders, Belgium.

eferences

[1] K. Kimata, K. Iwaguchi, S. Onishi, K. Jinno, R. Eksteen, K. HosoM. Araki, N. Tanaka, J. Chromatogr. Sci. 27 (1989) 721.

[2] W. Jost, R. Gasteier, G. Schwinn, M. Tueylue, R.E. Majors, Int.(May 1990) 46.

[3] H. Engelhardt, M. Jungheim, Chromatographia 29 (1990) 59.[4] S.J. Schmitz, H. Zwanziger, H. Engelhardt, J. Chromatogr. 544 (1

381.[5] S.V. Galushko, Chromatographia 36 (1993) 39.[6] D.V. McCalley, J. Chromatogr. 636 (1993) 213.[7] H.A. Claessens, E.A. Vermeer, C.A. Cramers, LC-GC Int. (Novem

1993) 692.[8] T. Hamoir, F. Cuesta Sanchez, B. Bourguignon, D.L. Massart, J. C

matogr. Sci. 32 (1994) 488.[9] B.A. Olsen, G.R. Sullivan, J. Chromatogr. A 692 (1995) 147.10] E. Cruz, M.R. Euerby, C.M. Johnson, C.A. Hackett, Chromatogra

44 (1997) 151.11] H. Engelhardt, M. Arangio, T. Lobert, LC–GC Int. (December 19

803.12] W. Eymann, Chromatographia 45 (1997) 235.13] A. Sandi, A. Bede, L. Szepesi, G. Rippel, Chromatographia 45 (1

206.14] G. Wieland, K. Cabrera, W. Eymann, LC-GC Int. (February 1998)15] D.V. McCalley, R.G. Brereton, J. Chromatogr. A. 828 (1998) 407.16] U.D. Neue, B.A. Alden, T.H. Walter, J. Chromatogr. A. 849 (1999) 117] W. Verstraeten, J. de Zeeuw, J. Crombeen, N. Vonk, Int. Lab. (M

2000) 20.18] U.D. Neue, K. Van Tran, P.C. Iraneta, B.A. Alden, J. Sep. Sci. 26 (2

174.

(2005) 109.36] D. Visky, Y. Vander Heyden, T. Ivanyi, P. Baten, J. De Beer,

Noszal, E. Roets, D.L. Massart, J. Hoogmartens, Pharmeuropa 14 (288.

37] T. Ivanyi, Y. Vander Heyden, D. Visky, P. Baten, J. De Beer, I. Lazar,D.L. Massart, E. Roets, J. Hoogmartens, J. Chromatogr. A 954 (99.

38] D. Visky, Y. Vander Heyden, T. Ivanyi, P. Baten, J. De Beer, Zs. KovacsB. Noszal, E. Roets, D.L. Massart, J. Hoogmartens, J. Chromato977 (2002) 39.

39] D. Visky, Y. Vander Heyden, T. Ivanyi, P. Baten, J. De Beer, Z. KovacsB. Noszal, P. Dehouck, E. Roets, D.L. Massart, J. HoogmartenChromatogr. A 1012 (2003) 11.

40] P. Dehouck, D. Visky, Y. Vander Heyden, E. Adams, Z. Kovacs, BNoszal, D.L. Massart, J. Hoogmartens, J. Chromatogr. A 1025 (2189.

41] P. Dehouck, D. Visky, G. Van den Bergh, E. Haghedooren, E. AdA. Kerner, Y. Vander Heyden, D.L. Massart, Zs. Kovacs, B. Noszal, J.Hoogmartens, LC-GC Eur. 17 (2004) 592.

42] European Pharmacopoeia, Fifth ed., Council of Europe, StrasbFrance, 2005.

43] www.pharm.kuleuven.ac.be/pharmchem/columnclassification, 200444] W. Naidong, S. Hua, E. Roets, J. Hoogmartens, J. Planar Chrom

5 (1992) 92.45] S.L. Morgan, S.N. Deming, J. Chromatogr. 112 (1975) 267.46] E. Smet, L. Stalens, Y. Vander Heyden, R.G. Baeyers, H.Y. Aboul-E

G. Van der Weken, A.M. Garcıa-Campana, Chirality 13 (2001) 556.47] E. Adams, E. Roets, J. Hoogmartens, J. Pharm. Biomed. Anal. 21 (

715.48] H. Ikeda, K. Shiroyanagi, M. Katayama, H. Ikeda, I. Fujimaki, T. S

J. Sugayama, Proc. Jpn. Acad. 32 (1956) 48.49] M. Bodanszky, Acta. Chim. Hung. 5 (1955) 97.50] P. Dehouck, D. Visky, Z. Kovacs, B. Noszal, E. Adams, D.L. Massar

J. Hoogmartens, LC-GC Eur. 16 (2003) 764.51] Pharmeuropa 12 (2000) 361.52] Pharmeuropa 10 (1998) 608.