[ Pobierz całość w formacie PDF ]
//-->ChromatographiaDOI 10.1007/s10337-014-2633-9ORIGINALRobust UHPLC Separation Method Development for Multi-APIProduct Amlodipine and Bisoprolol: The Impact of ColumnSelectionRóbert Kormány · Imre Molnár · Jeno Fekete ·˝Davy Guillarme · Szabolcs FeketeReceived: 26 September 2013 / Revised: 9 January 2014 / Accepted: 20 January 2014© Springer-Verlag Berlin Heidelberg 2014AbstractThis paper describes a new and fast ultra-highpressure liquid chromatographic separation of amlodipineand bisoprolol and all their closely related compounds, forimpurity profiling purposes. Computer-assisted methoddevelopment was applied and the impact of several state-of-the-art stationary phase column chemistries (50×2.1 mm,sub-2μm,and core–shell type materials) on the achiev-able selectivity and resolution was investigated. The workwas performed according to quality by design principlesusing design of experiment with three experimental factors;namely the gradient time (tG), temperature (T), and mobilephase pH. Thanks to modeling software, it was proved thatthe separation of all compounds was feasible on numerouscolumn chemistries within <10 min, by proper adjustmentsof variables. It was also demonstrated that the reliability ofpredictions was good, as the predicted retention times andresolutions were in good agreement with the experimen-tal ones. The final, optimized method separates 16 peaksrelated to amlodipine and bisoprolol within 7 min, ensuringbaseline resolution between all peak-pairs.KeywordsUHPLC · Method development ·Quality by design (QbD) · DryLab · Amlodipine ·BisoprololIntroductionWhen dealing with reversed-phase liquid chromatographic(RPLC) method development, computer modeling pro-grams can be employed to improve the analysis throughputas well as maximize information about method selectivity.The most successful and widespread modeling program(DryLab, Molnar-Institute, Berlin, Germany) optimizesthe Design Space mainly by measuring and visualiz-ing the effects of mobile phase conditions: gradient timeand shape, pH, ionic strength, ternary eluent composi-tion, additive concentrations, or temperature [1]. For thispurpose, the program suggests a relatively well-definednumber of experiments on a particular stationary phase;furthermore it can predict the separation inside the DesignSpace, based on changes in the mobile phase composition,mode of elution (either isocratic or gradient), temperature,pH or column parameters such as column length, internaldiameter, particle size, and flow-rate [2]. The retentionmechanism in RPLC can be explained by the solvophobictheory that gives a guidance for planning the experimentsfor RPLC method development and optimization [3]. Thetheory describes the effects on the chromatographic behav-ior of components, when varying different parameters.DryLab chromatographic optimization software is mostlybased on this theory [4], and its three-dimensional (3D)application helps to understand the peak movements andPublished in the special paper collection9th Balaton Symposiumon High-Performance Separations Methodswith guest editorAttila Felinger.R. Kormány (*)Egis Pharmaceuticals Plc, Budapest, Hungarye-mail: rkormany@gmail.comI. MolnárMolnár-Institute for Applied Chromatography, Berlin, GermanyJ. FeketeBudapest University of Technology and Economics,Budapest, HungaryD. Guillarme · S. FeketeUniversity of Geneva, Analytical Pharmaceutical Chemistry,Geneva, Switzerland13R. Kormány et al.the selectivity or resolution changes within the DesignSpace [5,6].Searching for alternative columns, while keeping thequality of a given separation is always one of the key pur-poses of method robustness testing, but finding the alterna-tive column for a given separation (column interchange-ability) is often complicated. Generally, the method isdeveloped using one given column and then, an alternativecolumn can be considered during the validation procedureunder the optimized conditions. Since the alternative col-umn probably has not the same working point (optimalconditions in a robust zone) as the primary column, this“trial and error” like approach often fails at the end ofmethod development. Column databases could be helpfulfor selecting an alternative column but common station-ary phase tests are not always able to predict certain col-umn similarity for particular separations. Numerous papersdealing with stationary phase characterization procedures,developed by Snyder, Dolan, Tanaka, Euerby, and Peters-son are available and could be helpful for users, in find-ing a similar column during the method development andvalidation [7–10]. One of our previous work illustrated thatthe baseline separation of amlodipine impurities was fea-sible on nine different 50×2.1 mm columns packed withsub-2μmfully porous and core–shell particles [11]. In thatwork, the authors compared the selectivity and achievableanalysis time when selecting the condition that ensures thehighest possible resolution. Another recent study showedthat if column was not directly interchangeable, it was stillpossible to achieve very similar separations by adjustingthe chromatographic conditions [12]. The study suggestedthat the evaluation of column interchangeability should bea part of early stage method development and not of themethod validation.In this current study, our aim was to develop a fast androbust ultra-high pressure liquid chromatographic (UHPLC)method for the separation of amlodipine and bisoprolol-related impurities. Amlodipine is a long-acting calciumchannel blocker dihydropyridine and acts by relaxing thesmooth muscle in the arterial wall, decreasing total periph-eral resistance, thereby reducing blood pressure. Bisoprololbelongs to the group of beta-blockers and is used primar-ily in cardiovascular diseases. The combination of these twoactive drugs is applied for the treatment of chronic stableangina pectoris and hypertension. Previous works describedthe spectrophotometric and conventional high-performanceliquid chromatographic determination of amlodipine andbisoprolol from pharmaceutical preparations and plasma[13–15]. To the best of our knowledge, no UHPLC separa-tion of all the related impurities was reported up to now.In this study, a novel and fast UHPLC impurity profilingmethod is reported for amlodipine and bisoprolol combinedactive pharmaceutical ingredients (API), and the benefitsof computer-assisted method development is discussed.Moreover, the impact of RP stationary phase selection onthe selectivity is studied and reported in details.ExperimentalChemicalsAcetonitrile (gradient grade), phosphoric acid, and natriumdihydrogen phosphate were purchased from Merck (Darm-stadt, Germany). For the measurements, water was pre-pared freshly using ELGA Purelab UHQ water (ELGA,Lane End, UK).Amlodipine and its Ph.Eur. impurities (A, B, D, E, F, G,H) and bisoprolol and its Ph.Eur. impurities (A, G, L, R)were purchased from European Directorate for the Qualityof Medicines and HealthCare (EDQM). The structure ofthe compounds is shown in Fig.1.Preparation of SolutionsThe mobile phase used in this work was a mixture of ace-tonitrile and 30 mM phosphate buffer (pH 2.0, 2.6, and3.2).The buffers were prepared by mixing the appropriateamount of 30 mM phosphoric acid and 30 mM sodiumdihydrogen phosphate. Buffers were filtered before use onregenerated cellulose filter membrane, 0.2μmpore size(Sartorius, Goettingen, Germany).Mobile phase “A” was 30 mM phosphate buffer (pH 2.0,2.6, and 3.2) and mobile phase “B” was acetonitrile.Sample solvent was a mixture of acetonitrile:water10:90 (V:V).Representative real-life sample of amlodipine, biso-prolol, and their Ph.Eur. impurities contained 1 mg mL−1amlodipine besilate and bisoprolol fumarate and theirimpurities at 0.1 % level was prepared by spiking all theimpurities to the API solution.Chromatographic SystemUHPLC experiments were performed on a Waters AcquityUPLC system (Milford, USA) equipped with binary sol-vent delivery pump, auto sampler, photodiode array detec-tor, and empower software. This UHPLC system had 5μLinjection loop and 500 nL flow cell. The dwell volume ofthe system was measured as 125μL.The column compart-ment of the system is equipped with a CM-A column man-ager that enables the use of four columns and programma-ble switching of the mobile phase among the columns.13Robust UHPLC Separation Method DevelopmentOHNONHOOONH2NHOOONONHOONOONHNH2OOOOOClOOClOOClOOClAmlodipineNHOOONH2A-ImpANHOOONH2A-ImpBNHA-ImpDNHONHOOOOOOOHOOClOOClOOClOOClA-ImpEOHOHNA-ImpFOHOHNA-ImpGOHOHNA-ImpHOHOHNOHOHNHOOHCOOH3COOOBisoprololB-ImpAB-ImpGB-ImpLB-ImpRFig. 1Structure of Amlodipine, Bisoprolol and their impuritiesFor the initial model runs, the mobile phase flow ratewas set to 0.5 mL min−1and gradients were run from 10 to90 %B. The injection volume was set to 1μL.ColumnsAcquity columns (50×2.1 mm, 1.7μmBEH C18, BEHShield RP 18, BEH C8, BEH Phenyl, CSH C18, CSH Phe-nyl-Hexyl, CSH Fluoro-Phenyl, 1.8μmHSS C18, HSSC18 SB, HSS T3, HSS PFP, HSS CN) were purchasedfrom Waters (Milford, USA).The 50×2.1 mm, 1.7μmAeris peptide XB-C18, andkinetex columns (XB-C18, C18, C8, Phenyl-Hexyl, PFP)were purchased from Phenomenex (Torrance, USA).Hypersil columns (50×2.1 mm, 1.9μmGold, GoldC8, Gold CN) were purchased from Thermo Scientific(Waltham, USA).The 50×2.0 mm, 1.9μmTriart C18 column was pur-chased from YMC (Kyoto, Japan).Zorbax columns (50×2.1 mm, 1.8μmSB-C18,SB-C8, SB-Phenyl) were purchased from Agilent (SantaClara, CA, USA).SoftwareModeling was carried out using DryLab v.4.0 and the quan-titative robustness evaluation of generated models was per-formed with the latest DryLab Robustness Module v.1.0.(Molnár-Institute, Berlin, Germany).Results and DiscussionDesign of Experiments (DoE)The selected example describes a fast and efficient methoddevelopment for the determination of impurities and degra-dation products of combined active pharmaceutical ingredi-ents, utilizing the separation power of state-of-the-art col-umns. A general methodology is to simultaneously modelthe effect of temperature and gradient steepness on selec-tivity with a given RP column. Thanks to the current devel-opments in chromatographic modeling software products,it is now possible to model the effect of three variablessimultaneously for a given separation. In our case, gradi-ent steepness (tG), temperature (T), and mobile phase pHwere selected as model variables to create a cube resolu-tion map, showing the critical resolution of the peaks to beseparated against the three factors. Probably, these selectedvariables have the most significant effect on the selectivityand resolution for such analytes. In most cases, the reten-tion can be described as a function of gradient steepness,with the linear solvent strength theory and its temperaturedependence following a van’t Hoff type relationship. Bothrelationships can be transformed to linear dependencies.When separating ionizable compounds, strong pH-relatedchanges in retention occur for pH values within±1.5unitsof thepKavalue. Outside this range, the compound is con-sidered as mostly ionized or non-ionized, and its retentionis not significantly altered with pH. In a relatively small13R. Kormány et al.pH range—within the±1.5units of thepKavalue—thedependence of retention on the mobile phase pH can gener-ally be described using quadratic polynomials.Therefore, in the proposed final model, two variables (tGandT)were set at two levels (tG1=3 min,tG2=9 min, andT1=20 °C andT2=50 °C) while the third factor (pH) wasset at three levels (pH1=2.0, pH2=2.6, and pH3=3.2).This full factorial experimental design required 12 experi-ments (2×2×3) on a given column.Column ScreeningIn a first instance, several state-of-the-art columns were eval-uated by performing the 12 experiments and creating the cor-responding 3D resolution maps. By utilizing the benefits ofthe column manager unit and small columns (50×2.1 mm),the column screening procedure requires only 4–5 days fortesting 25 columns, since a lot of work can be automated.Based on the resolution maps, the peak movements and thechange in selectivity/resolution were assessed and the col-umns were ranked in terms of achievable resolution.Table1shows the achievable maximum critical resolu-tion (Rs,crit) on all the 25 columns, when operating them attheir own optimal working point.In this study, we also compared the selectivity of thecolumns based on the snyder–dolan hydrophobicity sub-traction (SDHS) database that is available in the columnmatch tool of DryLab. This model takes into account thehydrophobicity (H), hydrogen bond basicity (B), ionicinteractions at two pH (C(2.8) andC(7.0)),hydrogen bondacidity (A), and steric selectivity (S). The degree of selec-tivity similarity can be obtained on the basis ofH, B, C, A,andSvalues. The resulting similarity factors (Fs) were alsoreported in Table1,when available. Fs < 3 means excellentsimilarity of selectivity between the compared columns;between 3 < Fs < 5, the selectivity similarity is moderate,and between 5 < Fs < 10, there is a questionable but stillfair comparability of selectivity. As shown in Table1,thisSDHS-based ranking sometimes resulted in unexpectedresults. As an example, the Hypersil Gold C8 column that isthe third most similar (Fs=6.3) to the reference BEH C18phase gave completely different working point. This col-umn has to be operated atT=42 °C, to reach the highestpossible resolution while the BEH C18 requires low operat-ing temperature (T=13.5 °C). Moreover, the critical peakpair was ImpD and ImpF on the Hypersil Gold C8 while itwas the ImpG–ImpH pair on the BEH C18 Phase. On thecontrary, the Kinetex PFP phase appears as the most differ-ent stationary phase on the basis of its Fs value (Fs=81.6).However, its working point was found to be very close tothe BEH C18 material and possesses the same critical peakpair. To conclude on the SDHS-based column comparisonapproach, it gives some useful idea for selecting a similaror diverse stationary phase in terms of interaction mecha-nisms but does not give information about the achievableresolution and analysis time when separating a specificcomplex mixture. The other disadvantage of the SDHSapproach is that the database is not regularly updated andit does not include data on several popular state-of-the-artstationary phases.Our 12 experiments based approach seems to be amore reliable procedure when comparing the achievableanalysis time, resolution, and working point. By applying50×2.1 mm columns, it takes approximately only 2–3 hof experimental work for one given column. The advantageof this column screening approach is that the suitability ofa column—for a given application—can be evaluated atthe very early stage of the method development. In addi-tion, the column interchangeability can also be estimatedduring the method development. Based on our experience,it appears that most of the columns can provide sufficientresolution within an acceptable analysis time, by adjustingproperly the chromatographic conditions. In this example,only one column among the 25 ones tested failed to achieveRs,crit> 1.5 (see table1).To conclude on our column screening approach, a prom-ising method development strategy consists in performinginitial runs and building up 3D models using different col-umns at the early phase of method development.Finding the Optimal ConditionsFor the mixture of compounds, the highest resolution couldbe performed on the Acquity CSH C18 material. Therefore,this column was selected for the final method (Table1).Itis also worth mentioning that this column also provided thehighest peak capacity (P=201 with a 10 min long gradient).First, the criteria for the minimum required resolutionwere set. The impurities have to be separated from (a)each other, (b) the APIs, and (c) other possible disturbingcompounds such as the fumaric acid and benzenesulfonicacid. For the baseline separation of the critical peak pairs,the value ofRs,critshould be higher than 1.5. But consider-ing that impurities are present in small concentrations (at~0.1 %), and have to be separated from the APIs at highconcentration, theRs,crit> 1.5 might not be enough. In thiscase, it is better to selectRs,crit> 2.5 as criteria. Figure2shows the 3D resolution map obtained with the AcquityCSH C18 material. Red color represents the regions insidethe Design Space where the resolution criteria is fulfilled,while blue colors indicate co-elutions (Rs=0). There arefour robust spaces that meet the criteria (Fig.2b).At lowpH (pH < 2.5), and at low temperature (below 30 °C) orat high temperature (above 40 °C) the resolution betweenfumaric acid and bisoprolol-ImpA was the lowest one,while at higher mobile phase pH (pH > 2.5) and at low13Robust UHPLC Separation Method DevelopmentTable 1List of columns used in the study, the conditions where the highest critical resolution can be reached, the critical peak-pairs, selectivitysimilarity (Fs), and the average of retention time relative errorsColumnsAcquity BEH C18Acquity BEH Shield RP 18Acquity BEH C8Acquity BEH PhenylAcquity CSH C18Acquity CSH Phenyl-HexylAcquity CSH Fluoro-PhenylTriart C18Acquity HSS C18Acquity HSS C18 SBAcquity HSS T3Acquity HSS PFPAcquity HSS CNHypersil goldHypersil gold C8Hypersil gold CNZorbax SB-C18Zorbax SB-C8Zorbax SB-PhenylAeris peptide XB-C18Kinetex XB-C18Kinetex C18Kinetex C8Kinetex Phenyl-HexylKinetex PFPpH2.12.02.52.03.02.13.03.02.12.02.02.03.03.02.72.92.22.82.03.02.22.52.42.22.4T(°C)13.538.333.029.313.513.513.513.524.030.031.519.513.541.342.027.829.313.513.515.013.520.313.533.816.5tG(min)8.19.89.89.89.82.92.77.49.89.89.89.87.99.89.89.09.86.18.99.89.89.89.89.89.8Rs,crit2.542.162.272.323.131.921.222.492.502.042.161.581.952.722.551.672.132.031.522.502.242.382.522.222.44Critical peak pairImpG–ImpHImpB–ImpGImpD–ImpFImpG–ImpBImpD–ImpFImpD–ImpFImpD–ImpFImpD–ImpFImpG–ImpHImpD–ImpFImpG–ImpHImpD–ImpFImpD–ImpFImpD–ImpFImpD–ImpFImpG–ImpBImpG–ImpHImpD–ImpFImpD–ImpFImpG–ImpHImpD–ImpFImpD–ImpFImpD–ImpFImpD–ImpFImpG–ImpHFs0.0–8.027.7–––––––––20.56.3–53.652.6–––4.0––81.6Average of retentiontime relative errors (%)0.23−0.850.410.880.60−0.55−0.790.57−1.95−0.37−0.27−0.15−0.94−0.10−0.240.56−0.361.09−2.420.350.81−0.540.14−0.281.67Difference (min): Predicted Retention Time−Experimental Retention Time% error: [(Predicted Retention Time−Experimental Retention Time)/Experimental Retention Time]×100temperature (<30 °C), bisoprolol and bisoprolol-ImpGwere considered as the critical peak pair. Furthermore, asteeper gradient decreases the resolution between biso-prolol and its impurity-G. Taking all these observationsinto account, the best working point is located into therobust space at high pH (pH > 2.5) and at high temperature(T > 40 °C). The final conditions were set astG=10 minstarting from 10 %B up to 90 %B (slope=8.0 %B min−1),column temperatureT=45 °C and mobile phase pH 3.0.Please note that the selected 10 min long gradient is outsidethe 3- and 9-min calibrated model, but the accuracy of theextrapolation is valid in this range [1]. Moreover, the reli-ability of the model was verified (see later).Simulated Robustness TestingThe reliability of DryLab’s new simulated robustness test-ing feature was recently reported [12]. Similarly to thisprevious work, the robustness of the optimized method wasalso assessed by the built-in robustness module. Beside thethree model variables (tG,T,pH), the flow rate, as well asinitial, and final compositions of the mobile phase rep-resent the investigated factors in the built in model. Theeffect of these six factors was evaluated at three levels.The modeled deviations from the nominal values werethe following: the gradient time was set to 9.9, 10, and10.1 min; temperature was set to 44, 45, and 46 °C; mobilephase pH was set to 2.9, 3.0, and 3.1; flow rate was setto 0.495, 0.500, and 0.505 mL min−1; initial mobile phasecomposition was set to 9.5, 10, and 10.5 %B and its finalcomposition was set to 89.5, 90, and 90.5 %B. Then, the729 experiments (36) were simulated in <1 min, thanks tothe software. A criterion ofRs,crit> 1.5 was considered.Figure3ashows the results of the experiments expressedin frequency as a function of criticalRs.As shown, themost frequent resolution value wasRs,crit=2.55 (20 condi-tions provided thisRsvalue), while the lowest predictedresolution wasRs,crit=2.21. Therefore, the method can13
[ Pobierz całość w formacie PDF ]