Chiral capillary zone electrophoresis in enantioseparation and analysis of cinacalcet impurities: Use of Quality by Design principles in method development

1 Chiral capillary zone electrophoresis in enantioseparation and analysis of 2 cinacalcet impurities: use of Quality by Design principles 3 in method development 4 5 Benedetta Pasquinia, Serena Orlandinia,*, Mercedes Villar-Navarrob, Claudia Caprinia, 6 Massimo Del Bubbac, Michal Doušad, Alessandro Giuffridae, Roberto Gottif, Sandra Furlanetto a,* 7 8 aDepartment of Chemistry “U. Schiff”, University of Florence, aVia U. Schiff 6, cVia della Lastruccia 3-13, 9 50019 Sesto Fiorentino, Florence, Italy 10 bDepartment of Analytical Chemistry, Faculty of Chemistry, University of Seville, 41012 Seville, Spain 11 dZentiva, k.s. Praha, a Sanofi Company, U Kabelovny 130, 102 37 Praha 10, Czech Republic 12 eDepartment of Chemical Sciences, University of Catania, Viale A. Doria 6, 95125 Catania, Italy 13 fDepartment of Pharmacy and Biotechnology, University of Bologna, Via Belmeloro 6, 40126 Bologna, Italy 14 15 16 17 18 19 20 21 22 23 24 25 *Corresponding authors. Tel.: +39 055 4573717 (S. Furlanetto), +39 055 4573733 (S. Orlandini) 26 E-mail addresses: [email protected] (S. Furlanetto), [email protected] S. 27 Orlandini) 2 28 Abstract 29 A capillary electrophoresis method for the simultaneous determination of the enantiomeric purity and of 30 impurities of the chiral calcimimetic drug cinacalcet hydrochloride has been developed following Quality by Design 31 principles. The scouting phase was aimed to select the separation operative mode and to identify a suitable chiral 32 selector. Among the tested cyclodextrins, (2-carboxyethyl)-β-cyclodextrin and (2-hydroxypropyl)-γ-cyclodextrin 33 (HPγCyD) showed good chiral resolving capabilities. The selected separation system was solvent-modified capillary 34 zone electrophoresis with the addition of HPγCyD and methanol. Voltage, buffer pH, methanol concentration and 35 HPγCyD concentration were investigated as critical method parameters by a multivariate strategy. Critical method 36 attributes were represented by enantioresolution and analysis time. A Box-Behnken Design allowed the contour plots to 37 be drawn and quadratic and interaction effects to be highlighted. The Method Operable Design Region (MODR) was 38 identified by applying Monte-Carlo simulations and corresponded to the multidimensional zone where both the critical 39 method attributes fulfilled the requirements with a desired probability π≥90%. The working conditions, with the MODR 40 limits, corresponded to the following: capillary length, 48.5 cm; temperature, 18 °C; voltage, 26 kV (26-27 kV); 41 background electrolyte, 150 mM phosphate buffer pH 2.70 (2.60-2.80), 3.1 mM (3.0-3.5 mM) HPγCyD; 2.00% (0.00- 42 8.40%) v/v methanol. Robustness testing was carried out by a Plackett-Burman matrix and finally a method control 43 strategy was defined. The complete separation of the analytes was obtained in about 10 min. The method was validated 44 following the International Council for Harmonisation guidelines and was applied for the analysis of a real sample of 45 cinacalcet hydrochloride tablets. 46 47 48 49 50 51 52 53 54 55 Keywords: Capillary Electrophoresis; Chiral separation; Cinacalcet; Impurities; Method Operable Design Region; 56 Quality by Design 57 3 58 1. Introduction 59 Cinacalcet hydrochloride (CIN) is the first agent of a new therapeutic class of calcimimetic compounds, which 60 act by increasing the sensitivity of calcium-sensing receptors to the extracellular calcium ions, thus lowering the release 61 of parathyroid hormone [1]. CIN is an approved medication for the treatment of secondary hyperparathyroidism in adult 62 patients with chronic kidney disease on dialysis and hypercalcemia in patients with parathyroid carcinoma [2]. It has 63 been commercialized as single R enantiomer, due to the fact that R-CIN is about 75 times more potent than its 64 corresponding S-enantiomer [3]. According to the drug product producer (Zentiva, Praha), the main potential impurities 65 of CIN which may be found in CIN bulk samples and dosage forms are the distomer S-cinacalcet (S-CIN), impurity 1 66 (I1) and impurity 2 (I2) , whose chemical structures are shown in Fig. 1. 67 The effective and safe therapy with chiral drugs basically depends on the quality of the pharmaceutical dosage 68 forms, which must be controlled both in terms of drug and potential impurities content and enantiomeric purity. In 69 particular, the assessment of enantiomeric purity is required by regulatory agencies prior to the marketing of optically 70 active drugs, and in this context the efficient separation of enantiomers represents a fundamental analytical task. 71 Enantioseparation can be effectively obtained with several chromatographic techniques including HPLC, ultra-high 72 performance liquid chromatography, nano-liquid chromatography and TLC [4-7]. Anyway, for routine analysis it is 73 important to have at disposal fast and low cost analytical methods with simple sample preparation and eco-friendly 74 characteristics as small reagent consumption. CE easily fulfills these requirements and has been effectively applied 75 during all stages of drug discovery as well as for the quality control of the finished pharmaceutical products. Among the 76 well-known advantages of CE there are its inherent flexibility as well as the various available operative modes, high 77 separation efficiency, low sample and reagent consumption [8]. In pharmaceutical analysis, CE has been widely applied 78 to the determination of the main component as well as for the purity of drugs with regard to related substances and 79 stereoisomeric impurities [9]. It represents one of the major techniques not only for the achievement of analytical scale 80 enantioseparations but also for a better understanding of the fine intermolecular recognition mechanisms which take 81 place [10-13]. In addition, CE utilizes a wide spectrum of chiral selectors as buffer additives for chiral separation, 82 without the need for expensive columns as for chromatography. The most commonly used chiral selectors are 83 cyclodextrins (CyDs), due to their wide variety of cavity size, side chain, degree of substitution and charge [14]. When 84 added to plain buffers, CyDs are able to resolve enantiomeric guests due to formation of diastereomeric inclusion 85 complexes, which present different electrophoretic mobilities. When the background electrolyte (BGE) contains 86 pseudostationary phases as micelles or microemulsions, it has been recently suggested that more complex phenomena 87 may take place [15,16]. In the latter case also the monomers of surfactant can interact with the CyD, possibly involving 88 the formation of ternary complexes 1:1:1 of CyD:enantiomer:surfactant, or giving rise to the displacement of CyD 4 89 bound enantiomeric guests with surfactant tails due to competitive binding. In any case the affinity of the enantiomers 90 for the CyD can be effectively modulated. . 91 The analysis of CIN and two process-related impurities has been performed by HPLC [17], and stability 92 indicating chromatographic methods involving forced degradation studies have been recently presented [18-21]. 93 Enantioseparation of CIN and its S-enantiomer was obtained by chiral normal phase HPLC [22], by indirect reversed- 94 phase HPLC with chiral auxiliaries as derivatizing agents and by direct thin-layer chromatography [23], and by 95 polysaccharide chiral reversed-phase HPLC [24]. Recently, a LC/MS-MS method for separation and determination of 96 CIN enantiomers in rat plasma on chirobiotic V column packed with vancomycin has been presented [25]. To the best 97 of our knowledge, CE methods for the determination of I1 and I2 have not been reported in previous literature. There is 98 only one reported CE method for the chiral separation of CIN [26], which was developed by univariate experiments to 99 test the enantiopurity of CIN in tablets without considering any of the other potential impurities which can be found in 100 the pharmaceutical dosage form. 101 For the first time in the literature, the purpose of this study was to set up a CE method for the simultaneous 102 determination of CIN, its chiral impurity S-CIN and its main impurities I1 and I2 in pharmaceutical formulations, for 103 fulfilling the requirements of an effective Quality Control strategy. Method development has been carried out by 104 following Quality by Design (QbD), a risk-management approach following the recent guidelines in the 105 pharmaceutical field [27]. QbD has been recently focused in the field of separation methods [28,29], and its 106 application in this study made it possible to identify not only a single optimum point but a multidimensional zone 107 where the desired quality of analytical data for determining all the four analytes was achieved. 108 QbD has been defined as “a systematic approach to development that begins with predefined objectives and 109 emphasizes product and process understanding and process control, based on sound science and quality risk 110 management” and has the aim of improving product quality and of increasing regulatory flexibility [27]. In terms of 111 application to analytical methods, Analytical Quality by Design (AQbD) emphasizes the need to thoroughly understand 112 the analytical system by an in-depth study of critical method parameters (CMPs) based on risk assessment and 113 multivariate tools [28,29]. Such strategy enables the robust optimization of the analytical method in compliance with 114 the recommendations of US Food and Drug Administration (US FDA) [30] and with the guidelines of International 115 Council for Harmonisation (ICH) [27], and represents a great step forward with respect to the traditional quality-by- 116 testing approach. As a matter of facts, the latter furnishes data and information limited to the experiments run by the 117 operator, without studying the possible interactions between CMPs and without building a model relating the CMPs to 118 critical method attributes (CMAs). On the other hand, by AQbD it is possible to properly manage the risk and to 119 identify a probabilistic design space, which in terms of analytical concept should be better defined as the method 5 120 operable design region (MODR) [31]. The MODR corresponds to a set of experimental conditions where the desired 121 quality, measured by CMAs values, is achieved with a selected probability. Thus, flexibility is augmented and possible 122 method modifications which could be necessary in the future are facilitated. This is possible thanks to the knowledge of 123 their potential impact on the method performances, gained by the risk-based approach. 124 The pharmaceutical regulatory requirements and the advantages of QbD methodology have led to an increase in 125 the use of QbD compliant analytical methods for the determination of impurities through the use of experimental design 126 and the computation of a MODR. The use of experimental design allows the exploration of the effects and of the 127 interactions of the CMPs by the calculation of polynomial models relating the CMPs to the CMAs. On the other hand, 128 the MODR is suitable to predict optimal analytical conditions within the experimental domain, with the associated 129 probability of fulfilling the CMAs requirements. Recent examples of applications of AQbD for the analysis of 130 impurities mainly concern the development of separation methods by HPLC [32-34], supercritical fluid chromatography 131 [35] and CE in its various operative modes [36-38], with the latter technique also effectively used for the determination 132 of enantiomeric impurities [39,40]. 133 In this study, QbD scouting allowed capillary zone electrophoresis with the addition of methanol (MeOH) and 134 (2-hydroxypropyl)-γ-cyclodextrin (HPγCyD) to be selected as separation system. Response Surface Methodology 135 (RSM) [41] was carried out by using a Box-Behnken design for investigating the effects of the CMPs (voltage, buffer 136 pH, MeOH concentration and HPγCyD concentration) on the CMAs, represented by enantioresolution and analysis 137 time. The calculated models and Monte-Carlo simulations [42] led to the definition of the MODR. Validation of the 138 method was carried out according to ICH guidelines [43], and the method was applied for the analysis of a real sample 139 of CIN tablets. 140 141 2. Materials and methods 142 2.1. Chemicals and reagents 143 Reference standards of CIN and its impurities (S-CIN, I1, I2) were kindly supplied by Zentiva (Praha, Czech 144 Republic), as well as Mimpara® coated tablets labeled to contain 90 mg of CIN and coated tablets excipients, e.g. pre- 145 gelatinised starch, microcrystalline cellulose, povidone, crospovidone, magnesium stearate, colloidal anhydrous silica, 146 lactose monohydrate, hypromellose, titanium dioxide (E171), glycerol triacetate, FD&C Blue (E132), yellow iron oxide 147 (E172). 148 Boric acid, 86.1% phosphoric acid, acetic acid, MeOH (HPLC grade), metformin hydrochloride (MET), n- 149 butanol, ethanol, acetonitrile, urea, all the CyDs tested with their degree of substitution (D.S.) in brackets, i.e. α- 150 cyclodextrin, γ-cyclodextrin, methyl-β-cyclodextrin (D.S. 1.5-2.1), heptakis (2,6-di-O-methyl-β-cyclodextrin) (D.S. 6 151 7.0), heptakis (2,3,6-tri-O-methyl-β-cyclodextrin) (D.S. 3.0), (2-hydroxyethyl)-β-cyclodextrin (D.S. 0.7), (2- 152 hydroxypropyl)-α-cyclodextrin (D.S. 0.6), (2-hydroxypropyl)-β-cyclodextrin (D.S. 0.6), HPγCyD (D.S. 0.6), sulfated- 153 β-cyclodextrin sodium salt (D.S. 12–15), (2-carboxyethyl)-β-cyclodextrin sodium salt (CEβCyD) (D.S. 3.0) and all the 154 other chemicals used were from Sigma-Aldrich (St. Louis, MO, USA). Water used for the preparation of the solutions 155 and running buffers was purified by Elix and Simplicity 185 systems (Millipore, Billerica, MA, USA). 156 157 2.2. Solutions and sample preparation 158 Standard stock solutions of CIN (10 mg ml-1), of the impurities (1 mg ml-1) and of the internal standard MET (1 159 mg ml-1) were prepared in ethanol and stored at 4 °C for a week. Working standard solutions were daily prepared in 160 water by adequate dilution. Running buffer solutions were prepared by adjusting the pH value of a proper volume of the 161 corresponding 0.5 M acid or mixture of acids by NaOH and by dilution up to the desired concentration. Britton- 162 Robinson buffers were prepared from a mixture of acetic acid, phosphoric acid and boric acid. Additives (CyDs, 163 organic solvents) were directly added to the plain buffers in order to obtain the final BGE. For sample preparation, 164 twenty Mimpara® tablets were weighed, crushed and powdered. The equivalent of 100 mg CIN was accurately weighed 165 and transferred to a 10 mL beaker and dissolved in ethanol. The suspension was stirred and sonicated for 10 min. One 166 milliliter of the mixture was centrifuged and 50 µL of the surnatant was diluted in a vial up to 500 µL by addition of 30 167 µL of MET stock solution and 420 µL of water. The final test concentration of CIN was about 1 mg mL-1 and MET 168 concentration was 0.06 mg mL-1. 169 170 2.3. Capillary electrophoresis apparatus and analyses 171 The CE experiments were performed on a HP 3DCE system (Agilent Technologies, Waldbronn, Germany) 172 equipped with an on-column UV-visible DAD and an air thermostating system. The CE instrument was driven and the 173 data were collected by the software 3DCE ChemStation Rev.09.01 (Agilent Technologies). Uncoated fused silica 174 capillaries (50 µm inner diameter, 365 µm outer diameter, total length 48.5 cm, effective length 40.0 cm) were used for 175 running the analyses and were purchased from Unifibre (Settimo Milanese, Italy). The detection wavelength was 220 176 nm and the samples were hydrodynamically injected (50 mbar for 3 s), followed by a BGE plug at 50 mbar for 3 s. Each 177 new capillary was flushed with 1 M NaOH, 0.1 M NaOH and water (5 min each). Between injections, the capillary was 178 washed with methanol, 0.1M NaOH, water for 1 min each, and then with the proper BGE for 3 min. The working 179 conditions (with the interval corresponding to the MODR) were: temperature, 18 °C; voltage, 26 kV (26-27 kV); BGE, 180 150 mM BGE phosphate buffer, pH 2.70 (2.60-2.80), 3.1 mM (3.0-3.5 mM) HPγCyD concentration, 2.00% (0.00-8.40 181 %) v/v MeOH concentration. 7 182 183 2.4. Calculations and softwares 184 In order to obtain the calibration graphs, the peak corrected area (area/migration time) ratios were plotted against 185 the corresponding analyte concentration. Two samples for each of five different concentration values of the compounds 186 were analyzed with the concentration of MET (internal standard) set at 0.06 mg mL-1. The CIN regression curve was 187 calculated in the range 0.6-1.2 mg mL-1, corresponding to the range 60-120% with respect to the test concentration of 1 188 mg mL-1. The regression curves for the impurities were from the respective limit of quantitation (LOQ) to 1% with 189 respect to the test concentration of the main compound: I1, 0.0005-0.0100 mg mL-1; I2, 0.0010-0.0100 mg mL-1; S-CIN 190 0.0010-0.0100 mg mL-1. 191 MODDE 10 software [44] was used for generating the Box-Behnken design used for the RSM, to perform the 192 related data treatment and to draw the risk of failure maps by Monte-Carlo simulations. Nemrod-W software [45] was 193 used for generating the Plackett-Burman design used for robustness testing. The runs of the experimental plans were 194 carried out in a randomized order with a test solution containing 1 mg mL-1 CIN and 0.0100 mg mL-1 CIN impurities 195 (1% with respect to the main compound). 196 197 3. Results and discussion 198 The analytical target profile of the method was defined as the accurate simultaneous determination of the main 199 compound CIN and its impurities, including the chiral impurity S-CIN, in a short analysis time. The final outcome of 200 the method should be the ability to determine the impurities at a concentration equal or lower than 0.1% with respect to 201 CIN, in order to find its application in the routine quality control of pharmaceutical dosage forms. 202 203 3.1. Method scouting and critical method attributes 204 Prior knowledge is the fundamental keystone of analytical development strategies, as it can effectively address 205 the preliminary experiments in order to approach the analytical target. In this context, preliminary experiments of the 206 scouting phase were carried out to select the separation operative mode and to identify a suitable chiral selector in order 207 to reach an adequate selectivity. In these experiments, the concentration value for CIN did not correspond to the final 208 test concentration value (1 mg mL-1), but it was kept low as 0.004 mg mL-1, as well as for S-CIN, in order to obtain 209 clear indications on the resulting electrophoretic pattern and thus on the capability of the tested chiral selectors to 210 achieve enantioresolution. 8 211 CIN and its impurities are pH sensitive compounds with an amino group in their structure (CIN pKa=8.4), thus 212 possessing basic properties and presenting a positive electrophoretic mobility at acidic pH values. Hence, a set of acidic 213 conditions was evaluated using as BGE different types of buffer at different concentration values in the range 10-150 214 mM and pH values in the range 2.50-3.50: phosphate, phosphate/acetate, acetate/borate and Britton-Robinson buffers. 215 In order to achieve enantioseparation, several types of CyDs, mentioned in Sec. 2.1, were evaluated as chiral selectors at 216 two concentration levels (10-30 mM). The best results in terms of analysis time and selectivity were obtained using 150 217 mM phosphate buffer, with only CEβCyD and HPγCyD leading to the enantioseparation among the tested CyDs. 218 Considering that the chosen operative mode should be able to separate the enantiomers also at the test concentration 219 value of the compounds, it was evaluated if the use of a double cyclodextrin system could improve the separation, using 220 CEβCyD and HPγCyD at different combinations of concentration values, ranging from 1.5 to 10.0 mM. The observed 221 effect on the electrophoretic pattern was some improvement in enantioseparation together with a significant increase in 222 analysis time, as shown in the representative electropherograms reported in Supplementary Fig. S1. Anyway, the 223 increase in enantioseparation was not deemed as sufficient for selecting this separation system, considering that a 224 favorable balance between resolution and analysis time should be achieved. 225 The selection between CEβCyD and HPγCyD was based on the finding that they caused an opposite enantiomer 226 migration order, being R/S and S/R when using CECyD and HPγCyD, respectively. As a matter of facts, adding 227 HPγCyD to the BGE the main compound CIN corresponded to the last migrating peak, thus allowing the detection and 228 determination of S-CIN avoiding any interference due to a possible peak tailing of the main drug. The addition of 229 organic solvent modifiers (MeOH, acetonitrile, n-butanol and urea) was evaluated in the range 5-10% v/v, showing the 230 best results in terms of selectivity when MeOH was added to the BGE. 231 Taking into account the obtained results, the selected operative mode was capillary zone electrophoresis using as 232 plain buffer 150 mM phosphate with the addition of HPγCyD as chiral selector and methanol as organic modifier. In 233 these conditions the migration order of the analytes was: MET (internal standard), I1, I2, S-CIN and CIN. Resolution Rs1 234 between MET and I1, resolution Rs2 between I1 and I2 and resolution Rs3 between I2/S-CIN did not represent critical 235 analytical issues, hence the CMAs were selected as the resolution Rs4 between S-CIN and CIN enantiomers and analysis 236 time (t). The CMA requirements were chosen as Rs4≥0.5, corresponding to a baseline separation taking into account the 237 different concentration of the main compound and the S-enantiomer, and t≤11min, in order to obtain the separation in a 238 reasonable time. 239 240 3.2. Risk assessment and critical method parameters 9 11 1 22 2 33 3 44 4 241 In this study, the Ishikawa fishbone diagram [46] shown in Fig. 2 was used to formalize the risk assessment and 242 to point out the risk factors associated with the performances of the CE analysis. The different sources of factors were 243 represented by injection, separation/detection, capillary and BGE. The CNX tool [47] was employed for further 244 classifying the Ishikawa parameters in parameters which should be controlled (C), potential noise parameters (N) and 245 parameters which should be experimented to determine acceptable ranges (X). The results of the experiments of the 246 scouting phase made it possible to fix injection, detection, capillary factors and BGE factors such as type and 247 concentration of buffer, type of organic modifier, type of CyD and its degree of substitution. Instead, voltage among the 248 separation factors and other BGE characteristics needed to be further risk managed and in-depth studied by 249 experimental design in order to enhance the knowledge of their effects on the CMAs. 250 The fixed parameters were 150 mM phosphate buffer, 18 ºC capillary temperature, 3 s injection time at 50 mbar, 251 48.5 cm capillary length and detection wavelength at 220 nm. The selected CMPs for the subsequent multivariate study 252 were represented by voltage (V), buffer pH (pH), HPγCyD concentration (CyD conc) and MeOH concentration (MeOH 253 conc). 254 255 3.3. Response Surface Methodology and method operable design region 256 RSM was employed for achieving a predictive model which adequately represents the variation of the CMAs 257 inside the selected experimental domain of the four CMPs, which is reported in Supplementary Table S1. The selected 258 range for pH and CyD conc was rather narrow, according to the experimental findings from the scouting phase which 259 clearly indicated the advantages of keeping low levels of pH and low levels of CyD conc for obtaining a reasonable 260 analysis time. A Box-Behnken design was selected for estimating the second order polynomial equation relating the 261 CMPs and the CMAs and including linear, quadratic and interaction effects: 262 y=β0+β1x1+β2x2+β3x3+β4x4+β x 2+β x 2+β x 2+β x 2+β12x1x2+β13x1x3+β14x1x4+β23x2x3+β24x2x4+β34x3x4+ε 263 where y represents the experimental response, xi the independent evaluated factors, β0 the intercept, βi the true 264 coefficients and ε the experimental error. 265 Table 1 shows the 27-runs experimental plan and the measured responses including three center points for 266 estimating the experimental variance. In this three-level design, the k variables are varied two at a time by 22 designs, 267 while maintaining the remaining (k-2) variables fixed at their middle level and are very efficient designs in terms of 268 required runs [41]. The responses Rs4 and t were reverse transformed (Y-1) and some of the coefficients without 269 significant effect among the interaction and quadratic terms were removed in order to ameliorate the goodness of fitting 270 (coefficient of determination R2) and prediction (coefficient of goodness of prediction Q2) of the models, calculated by 271 multiple linear regression. In terms of ANOVA the refined model for Rs4 was both valid and significant, while a lack of 16 429 [24] M. Douša, J. Břicháč, Chiral chromatography studies of chemical behavior of cinacalcet on polysaccharide 430 chiral reversed-phase HPLC stationary phases, J. AOAC Int. 95 (2012) 1639-1645. 431 [25] R. Nageswara Rao, S. Bompelli, LC-MS/MS determination of cinacalcet enantiomers in rat plasma on 432 Chirobiotic V column in polar ionic mode: application to a pharmacokinetic study, Biomed. Chromatogr. 28 433 (2014) 1846-1853. 434 [26] P. Ginterová, J. Znaleziona, R. Knob, M. Douša, J. Petr, J. Ševčik, Enantiomeric purity control of R-cinacalcet 435 in pharmaceutical product by capillary electrophoresis, Chem. Pap. 70 (2016) 1024-1030. 436 [27] ICH Harmonised Tripartite Guideline. Pharmaceutical Development Q8(R2) (2009) International Conference 437 on Harmonisation of technical requirements for registration of pharmaceuticals for human use. 438 [28] E. Rozet, P. Lebrun, P. Hubert, B. Debrus, B. Boulanger, Design Spaces for analytical methods, Trends Anal. 439 Chem. 42 (2013) 157-167. 440 [29] S. Orlandini, S. Pinzauti, S. Furlanetto, Application of quality by design to the development of analytical 441 separation methods, Anal. Bioanal. Chem. 405 (2013) 443-450. 442 [30] Guidance for Industry. PAT - A Framework for Innovative Pharmaceutical Development, Manufacturing, and 443 Quality Assurance (2004) U.S Food and Drug Administration. 444 [31] A. Dispas, H.T. Avohou, P. Lebrun, P. Hubert, C. Hubert, “Quality by Design” approach for the analysis of 445 impurities in pharmaceutical drug product and drug substances, Trends Anal. Chem. (2017) doi 446 10.1016/j.trac.2017.10.028. 447 [32] A.H. Schmidt, I. Molnár, Using an innovative Quality-by-Design approach for the development of a stability 448 indicating UHPLC method for ebastine in the API and pharmaceutical formulations, J. Pharm. Biomed. Anal. 449 78-79 (2013) 65-74. 450 [33] C. Boussès, L. Ferey, E. Vedrines, K. Gaudin, Using an innovative combination of quality-by-design and green 451 analytical chemistry approaches for the development of a stability indicating UHPLC method in pharmaceutical 452 products, J. Pharm. Biomed. Anal. 115 (2015) 114-122. 453 [34] J. Terzić, I. Popović, A. Stajić, A. Tumpa, B. Jančić-Stojanović, Application of Analytical Quality by Design 454 concept for bilastine and degradation impurities determination by hydrophilic interaction liquid 455 chromatographic method, J. Pharm. Biomed. Anal. 125 (2016) 385-393. 456 [35] B. Andri, P. Lebrun, A. Dispas, R. Klinkenberg, B. Streel, E. Ziemons, R.D. Marini, Ph. Hubert, Optimization 457 and validation of a fast supercritical fluid chromatography method for the quantitative determination of vitamin 458 D3 and its related impurities, J. Chromatogr. A 1491 (2017) 171-181. 17 459 [36] S. Orlandini, B. Pasquini, M. Stocchero, S. Pinzauti, S. Furlanetto, An integrated quality by design and 460 mixture-process variable approach in the development of a capillary electrophoresis method for the analysis of 461 almotriptan and its impurities, J. Chromatogr. A 1339 (2014) 200-209. 462 [37] S. Orlandini, B. Pasquini, C. Caprini, M. Del Bubba, L. Squarcialupi, V. Colotta, S. Furlanetto, A 463 comprehensive strategy in the development of a cyclodextrin-modified microemulsion electrokinetic 464 chromatographic method for the assay of diclofenac and its impurities: Mixture-process variable experiments 465 and quality by design, J. Chromatogr. A 1466 (2016) 189-198. 466 [38] B. Pasquini, S. Orlandini, C. Caprini, M. Del Bubba, M. Innocenti, G. Brusotti, S. Furlanetto, Cyclodextrin- 467 and solvent-modified micellar electrokinetic chromatography for the determination of captopril, 468 hydrochlorothiazide and their impurities: A Quality by Design approach, Talanta 160 (2016) 332-339. 469 [39] S. Orlandini, B. Pasquini, M. Del Bubba, S. Pinzauti, S. Furlanetto, Quality by design in the chiral separation 470 strategy for the determination of enantiomeric impurities: Development of a capillary electrophoresis method 471 based on dual cyclodextrin systems for the analysis of levosulpiride, J. Chromatogr. A 1380 (2015) 177-185. 472 [40] S. Orlandini, B. Pasquini, C. Caprini, M. Del Bubba, M. Douša, S. Pinzauti, S. Furlanetto, Enantioseparation 473 and impurity determination of ambrisentan using cyclodextrin-modified micellar electrokinetic 474 chromatography: Visualizing the design space within quality by design framework, J. Chromatogr. A 1467 475 (2016) 363-371. 476 [41] G.A. Lewis, D. Mathieu, R. Phan-Tan-Luu, Pharmaceutical Experimental Design, Marcel Dekker, New York, 477 1999. 478 [42] M.A. Herrador, A.G. Asuero, A.G. Gonzalez, Estimation of the uncertainty of indirect measurements from the 479 propagation of distributions by using the Monte-Carlo method: an overview, Chemom. Intell. Lab. Syst. 79 480 (2005) 115-122. 481 [43] ICH Harmonised Tripartite Guideline. Validation of Analytical Procedures: Text and Methodology Q2(R1) 482 (2005) International Conference on Harmonisation of technical requirements for registration of 483 pharmaceuticals for human use. 484 [44] MODDE v. 10, MKS Umetrics AB, Sweden. 485 [45] D. Mathieu, J. Nony, R. Phan-Tan-Luu, NEMROD-W, LPRAI sarl, Marseille. 486 [46] K. Ishikawa, What is total quality control? The Japanese Way, Prentice-Hall, Englewood Cliffs, 1985. 487 [47] P. Borman, M. Chatfield, P. Nethercote, D. Thompson, K. Truman, The application of Quality by Design to 488 analytical methods, Pharm. Tech. 31 (2007) 142-152. 18 489 [48] L. Eriksson, E. Johansson, N. Kettaneh-Wold, C. Wikström, S. Wold, Design of Experiments – Principles and 490 Applications, MKS Umetrics AB, Umeå, Sweden, 2008. 491 [49] S.A.C. Wren, R.C. Rowe, Theoretical aspects of chiral separation in capillary electrophoresis: II. The role of 492 organic solvent, J. Chromatogr. A 609 (1992) 363-367. 493 [50] A. Van Eeckhaut, M.R. Detaevernier, J. Crommen, Y. Michotte, Differential effects of organic modifiers on 494 the enantioseparation of dimetindene maleate with carboxymethyl--cyclodextrin in capillary electrophoresis, 495 J. Sep. Sci. 27 (2004) 21–27. 496 [51] S. Fanali, Enantioselective determination by capillary electrophoresis with cyclodextrins as chiral selectors, J. 497 Chromatogr. A 875 (2000) 89-122. 498 [52] L. Nompari, S. Orlandini, B. Pasquini, C. Campa, M. Rovini, M. Del Bubba, S. Furlanetto, Quality by design 499 approach in the development of an ultra-high-performance liquid chromatography method for Bexsero 500 meningococcal group B vaccine, Talanta 178 (2018) 552-562. 19 501 Figure legends 502 503 Fig. 1. Molecular structures of the compounds. 504 Fig. 2. Fishbone diagram for risk assessment. The factors evaluated by experimental design are in bold type. C- 505 controlled parameters; N-noise parameters; X-experimented parameters. 506 Fig. 3. Contour plots for Rs4 (a) and t (b) obtained by plotting CyD conc vs. pH at different MeOH conc values (0.0, 5.0 507 and 10.0% v/v). Voltage was set at a constant value of 27 kV. 508 Fig. 4. Sweet spot plots obtained by plotting CyD conc vs. pH at different MeOH conc values (0.0, 5.0 and 10.0% v/v). 509 Voltage was set at a constant value of 27 kV. Green: areas where both the CMAs fulfill the requirements; brilliant blue: 510 areas where only one CMA fulfill its requirement. 511 Fig. 5. Probability map with the MODR colored in green and identified as the zone where the risk of obtaining Rs4<0.5 512 or t>11 min is ≤10%. 513 Fig. 6. Electropherogram obtained with the working points conditions. Sample: CIN, 1 mg mL-1; CIN impurities, 0.01 514 mg mL-1; MET, 0.06 mg mL-1. Detection wavelength: 220 nm. Experimental conditions: capillary length, 48.5 cm; 515 temperature, 18 °C; voltage, 26 kV; BGE, 150 mM phosphate buffer pH 2.70, 3.1 mM HPγCyD, 2.00% v/v MeOH. Figure F3C NH NH F 3 C HCl HCl Cinacalcet hydrochloride (CIN) (R)-N-[1-(1-naphthyl)ethyl]-3-[3-(trifluoromethyl) phenyl]propan-1-amine hydrochloride CAS 364782-34-3 S-cinacalcet (S-CIN) (S)-N-[1-(1-naphthyl)ethyl]-3-[3-(trifluoromethyl)phenyl] propan-1-amine hydrochloride CAS 694495-47-1 H 3 C Impurity 1 (I1) H 3 C Impurity 2 (I2) NH HCl (R)-(+)-1-(1-naphthyl)ethylamine CAS 3886-70-2 N-[3-(3-methylphenyl)propyl]-N-[(1R)1-(1-naphthyl) ethyl]amine hydrochloride CAS 253337-60-9 Figure Polarity-C Detector Settings-N Injection Type-C Time-X Pressure or Voltage-X Buffer plug-N Voltage-X Temperature-X Short- or long-end-C Degree of substitution-C Cyclodextrin Type-X Method performances Coating-C Concentration-X Type-X Organic modifier Concentration-X Type-X Diameter-C Length-X Concentration-X pH-X Aqueous buffer Separation Detection Injection Capillary BGE Figure Figure (a) Rs4 (b) t Fig. 3, B. Pasquini et al. Figure Fig. 4, B. Pasquini et al. Figure Fig. 5, B. Pasquini et al.