Chemicals and reagents

The analyzed substance: amlodipine besylate working standard (Lek, Ljubljana) and impurities: 3-ethyl 5-methyl 2-[(2-aminoethoxy)methyl]-4-(2-chlorophenyl)-6-methylpyridine-3,5-dicarboxylate) CRS European Pharmacopoeia reference standard (amlodipine impurity D) Sigma-Aldrich, Germany; diethyl-amlodipine CRS European Pharmacopoeia reference standard (amlodipine impurity E) Sigma-Aldrich, Germany; dimethyl-amlodipine European Pharmacopoeia reference standard (amlodipine impurity F) Sigma-Aldrich, Germany. The mobile phase and the solvents were prepared from acetonitrile (Fisher Scientific, England), ammonium acetate (Lachner, Chech Republic), glacial acetic acid (Lachner, Chech Republic), and HPLC grade water. Vazotal® 5 mg tablets (Hemofarm, Serbia) were used for analysis. All the reagents utilized in this study were of the analytical grade.

Chromatographic conditions

The experiments were performed on chromatographic system Agilent 1200 series consisting of HPLC Pump, Quaternary Pump, ALS Autosampler, and UV/VIS DAD detector. Agilent ChemStation software was used for data collection. The analytical column was ZORBAX NH₂ column (250 × 4.6 mm, 5 µm particle size) Agilent Technologies, USA. Throughout the complete experimental procedure, the following instrumental chromatographic conditions were maintained: flow rate of the mobile phase 1 mL min⁻¹, column temperature 30 °C and UV detection at 230 nm. The experimental plan was created according to Box–Behnken design for optimization and fractional factorial design (FFD 2⁴⁻¹) for robustness. The prepared mobile phases were filtered through a 0.45 µm nylon membrane filter Whatman, England.

Mobile phase

The mobile phase consisted of acetonitrile and water phase (with added ammonium acetate and glacial acetic acid), where the amount of organic solvent, ammonium acetate concentration in the water phase and pH of the water phase were varied according to the experimental plan. The mobile phase under optimal chromatographic conditions was as follows: acetonitrile–ammonium acetate (50 mmol L⁻¹) in water, pH 4.0 adjusted with glacial acetic acid (90.5:9.5, v/v).

Standard solutions

Stock solutions for the method optimization and robustness testing contained 100 μg mL⁻¹ of amlodipine besylate, and 10 μg mL⁻¹ of all impurities in the mixture of acetonitrile–ammonium acetate (50 mmol L⁻¹) in water, adjusted with glacial acetic acid to pH 4.0 (90:10, v/v). Placebo mixture (siliconized microcrystalline cellulose, pregelatinized starch and magnesium stearate – was donated by the pharmaceutical industry Hemofarm d.o.o. Banja Luka, Republic of Srpska) for selectivity estimation was prepared in a concentration ratio corresponding to the content in the pharmaceutical dosage form (Vazotal® 5 mg tablets). A standard solution containing 100 μg mL⁻¹ of amlodipine besylate and 10 μg mL⁻¹ of each impurity was used to prove the selectivity. For linearity estimation five solutions containing amlodipine besylate (from 50 to 200 μg mL⁻¹) and its impurities D, E and F (from LOQ to 15.0 μg mL⁻¹) were prepared in the mobile phase. The accuracy estimation is performed using three series of three solutions containing placebo, amlodipine besylate in concentrations 111.2 μg mL⁻¹ (80%), 139 μg mL⁻¹ (100%) and 166.8 μg mL⁻¹ (120%), as well as its impurities D, E and F in concentrations 8.0, 10.0 and 12.0 μg mL⁻¹ respectively. All solutions were prepared in the optimal mobile phase. The precision estimation was performed on real samples using Vazotal® 5 mg tablets which contain 6.934 mg of amlodipine besylate per tablet. Six samples contained 100 μg mL⁻¹ of amlodipine besylate were prepared in mobile phase and spiked with impurities (D, E and F) in concentration of 10 μg mL⁻¹.

Solutions for estimating limit of detection (LOD) and limit of quantification (LOQ) were prepared individually for each compound (impurities D, E and F) in concentrations 1.0, 0.5, 0.25, 0.10, 0.05, 0.025, 0.01 μg mL⁻¹. All solutions were prepared in mobile phase.

Software

Experimental plan and data analysis, according to Box–Behnken design and FFD 2⁴⁻¹ were created in Design-Expert 7.0.0 software (Stat-Ease Inc., Minneapolis, MN, USA). Three-dimensional response surface was obtained by Design-Expert 7.0.0 software. Method optimization, Derringer's desirability function and global optimum have been studied and defined in Design-Expert 7.0.0 software. The values of partition coefficient (log P), distribution coefficient (log D) and the value of pKa of the analyzed compounds were estimated in MarvinSketch 15.4.13 (ChemAxon Kft, Budapest, Hungary).

Preliminary experiments and identification of critical parameters of the chromatographic system

In this study, good separation of all analyzed compounds in a short time of the analysis was set as a goal. Before performing preliminary analyzes, pKa, log P and log D values were determined based on the chemical structure of tested compounds. For this analysis, MarvinSketch® program was used. This procedure is very important in HILIC chromatography in order to quickly make the best selection of the stationary phase, the mobile phase composition and the mobile phase pH value. By studying these values it can be ensured that all tested compounds during the analysis will be either 100% in ionic or 100% in molecular form.

Based on the obtained pKa values of the analyzed compounds (AB pKa = 9.40; impurity D, impurity E and impurity F pKa = 9.45) the most suitable pH value of the mobile phase was selected. The selected pH value of 4.0 will ensure that AB and impurities D, E and F are ∼100% dissociated (ionic form).

By analyzing the obtained log P and log D values for AB (log P = 1.64; log D = –1.17) and its impuritiy D (log P = 2.66; log D = –0.37), impuritiy E (log P = 1.99; log D = –1.04) and impuritiy F (log P = 1.28; log D = –1.75), the order of elution in HILIC system should be impurity D → impurity E → AB → impurity F. Elution order is followed by the pattern of the increasing hydrophilicity. This means, impurity F has the lowest log P and log D values and as most hydrophilic compounds will be elueted last in HILIC system.

Preliminary tests were performed on several different HILIC columns: silica, cyano, amino and diol. The amino column was selected as the most suitable stationary phase for further analysis. In preliminary studies, acetonitrile was selected as the organic component of the mobile phase, the content of which varied in the range of 80–96%. Ammonium acetate was selected as a buffer and its concentration in the aqueous phase was tested in the range of 10 – 70 mmol L⁻¹. Since AB, as well as its impurities D, E and F are of an alkaline character, it is necessary to perform the experiments at low pH values in order to maintain them in ionized form (as cations). In this sense, the pH of the aqueous phase was adjusted with glacial acetic acid in the range of 3.5–5.0, and then analyzes were performed. The results obtained during the preliminary research made it possible to select a set of critical responses of the chromatographic system. These responses that will be monitored in the further method development are:

1. Rs _impD,impE – the resolution between impurity D and impurity E;

2. Rs _impE,impAB – the resolution between impurity D and amlodipine besylate;

3. Rs _AB,impF – the resolution between amlodipine besylate and impurity F;

4. As _AB – the asymmetry factor of amlodipine besylate chromatographic peak;

5. tR _impF – the retention time of impurity F (analysis time).

In the second step of the HILIC method development, through the performance of preliminary experiments, the critical parameters of the chromatographic system were identified. The ones that will be monitored during optimization phase of the method are:

1. acetonitrile content in mobile phase

2. ammonium acetate concentration in water phase (buffer concentration), and

3. pH value of water phase.

Their influence on the monitored responses of system proved to be very significant. With small changes in all three selected factors, large changes in all monitored system responses are caused (resolution factors, the asymmetry factor of AB and analysis time). In these studies, the behavior of examined responses to changes in selected factors was concluded – by increasing the acetonitrile content in mobile phase and pH value of water phase, and decreasing the ammonium acetate concentration in water phase leads to better resolution of all tested compounds, better asymmetry factor of AB, and longer analysis time.

During the preliminary studies, the influence of column temperature (25–40 °C), mobile phase flow (0.5–1.5 mL min⁻¹), and wavelength of absorption maximum (190–400 nm) for all four tested compounds were also examined. It was concluded that increasing the value of mobile phase flow and column temperature, leads to irrelevant shortening of the analysis time. These factors do not affect the critical pairs resolution, nor the asymmetry factor of AB. Given the quite predictable influence of these factors on the HILIC system behavior, they were defined as uncritical and were maintained at a constant level in the continuation of the test. It was concluded that the best values of all monitored system responses were obtained at a column temperature of 30 °C, a mobile phase flow rate of 1 mL min⁻¹, and a determination wavelength of 230 nm.

Optimization – multi-criteria decision making

In the HILIC method optimization phase, for creating the Box–Behnken design, the intervals of the examined factors were selected and the final ranges were:

• 89–92% for acetonitrile content in the mobile phase,

• 4.0 – 5.0 for pH value of water phase, and

• 40 – 60 mM for ammonium acetate concentration in water phase.

In the process of the method optimization, Box–Behnken design was used (Table 1). That enabled a clear definition of individual and joint influences of the examined factors on the observed system responses. Response surface methodology was used in the optimization process – enables definition of quadratic response models that accurately describes the response for all values of the chromatographic conditions in the experimental region. To calculate quadrate regression model coefficients, each design variable must be studied at three distinct levels at least, and, consequently, a Box–Behnken design was used in this optimization study [9].

From the obtained results (Table 1) it can be seen that the system is quite sensitive to small changes in acetonitrile content, but also to other two examined factors – this is a specific characteristic of HILIC system. The monitored responses of the system show that during optimization process, the values for all three resolution factors (Rs) vary from a value of 0.15–0.34 (significant overlap of the critical chromatographic pair: Rs _impD,impE , Rs _impE,AB and Rs _AB,impF ) to values significantly greater than 1.2 (very good separation of the critical chromatographic pair). The asymmetry factor AB values vary in the range 0.971–5.022, which indicates a high level of all three factors influence on the asymmetry of this chromatographic peak (normal as value is 1.0 < As < 1.2). The retention time of impurity F varies in the range of 3.16–11.26 min. This response is monitored due to the duration of the analysis, but also in the optimization process, it shows how sensitive the HILIC system to the slightest changes of all three examined factors is.

Using DoE in the method optimization process, in order to clearly define how each individual factor and their two-factor interactions affect the observed system responses, statistical analysis yielded ANOVA test parameters for a model describing the chromatographic behavior of AB and its impurities D, E and F. The obtained coefficients are shown in Table 2.

The obtained results for the ANOVA test parameters show that a very high value was calculated for the mathematical model for coefficients of determination R² and for adjusted R². The lowest values obtained are R² = 0.8978 and R² adj = 0.8839, which indicates a good correlation between experimentally obtained and model-calculated values of responses (Rs _impD,impE, Rs_impE,AB , Rs _AB,impF , As _AB, tR _impF ). In this study, the adjusted R² was well within acceptable limits of R² ˃ 0.80 which revealed that the experimental data were a good fit for the quadratic model equations [9, 12]. Also, the model reliability was additionally confirmed with small RSD values (RSD < 1.0%) [22].

Based on the obtained results (Table 2), the factors and factor interactions that are relevant for each examined response are precisely defined (the one that meets the requirement that the p-value is <0.05).

A – acetonitrile content in the mobile phase; B – pH value of water phase; C – ammonium acetate concentration in water phase (buffer concentration).

Based on the coefficients of the quadratic response model, as well as their signs (Table 2), it is possible to estimate which factors and which factor interactions have an impact on the examined responses. From the obtained results of the ANOVA test, factors A, B and C have a significant influence on the response of Rs _impD,impE (Eq. (2)), as well as factor interactions AB and AC – by increasing acetonitrile content and pH value of water phase, and by decreasing buffer concentration in the mobile phase, the resolution between critical pair of impurities D and E increaeses. Since there is a significant influence of buffers in the two-factor interaction AC, additional analysis is necessary in order to find the best value for this response.

The response of Rs _impE,AB (Eq. (3)) and the response of Rs _AB,impF (Eq. (4)) are significantly influenced by factor A, and factor interactions AC – by increasing the content of acetonitrile and buffer concentration, increases the resolution between the chromatogram pair of impurities E and AB. Also, the resolution between the critical pair AB and impurity F increases.

On tR _impF response (Eq. (5)), factors A and C, and factor interaction AC have a significant effect – by increasing the content of acetonitrile and buffer concentration, the retention time of impurity F prolongs.

On the As _AB response (Eq. (6)) factor C and factor interaction AC have a significant effect – by decreasing the content of acetonitrile and buffer concentration, As _AB response is increasing. This change in the As _AB response is the opposite of all other responses, and because of that research is required to find the global optimum.

Factor interactions A², B², C² have a significant effect – by increasing the content of acetonitrile, the pH-value of the buffer aqueous solution increases while lowering the buffer concentration achieves better symmetry of the AB peak.

Quadrate members of factors A², B², C² (Eqs. (2)–(6)) indicate the presence of a nonlinear relationship between A, B and C, and the analyzed system responses. These factors with the (+) sign indicate that this change has a concave shape (extreme value is the minimum of function), while factors with the (–) sign indicate that this change has a convex shape (extreme value is the maximum of function). Quadrate members of factors A² and C² with the (–) sign indicate that the response values will decrease with increasing the content of acetonitrile and buffer concentration to a certain threshold, after which the response value starts to increase. Quadrate member of factors B² with the (+) sign indicates that the response value will increase with decreasing the pH value of water phase to a certain threshold, after which the response value starts to decrease.

From all of the above, it is concluded that the most significant influence on this chromatographic system has the acetonitrile content in the mobile phase (factor A), followed by the buffer concentration (factor C), as well as their interactions (AC). The pH value of the water phase (factor B) had the least influence on the examined responses of the chromatographic system, of the three most significant factors selected. However, this factor influence is also very important for the Rs _impD , _impE , tR _impF and As _AB responses. Respectively, as the value of factor B increases, the value of the Rs _impD,impE , and tR _impF responses increases, and the As _AB response value decreases. This means that a better resolution of the critical pair of chromatographic peaks of impurities D and E is provided, but the analysis duration is extended and the asymmetry factor value for AB is reduced.

Since the system is quite complex and that there are significant two-factor interactions between all three examined factors, 3D graphs were constructed. In order to describe the analyzed system better, the response surface methodology was used. Based on the obtained results of the ANOVA test, 3D graphs, that represent the dependence of the examined system responses from the observed factors, were constructed. These graphs represent a very convenient way to visually study the behavior of analytes in the examined system. The obtained graphs are shown in Fig. 2.

View Full Size

3-D graph for critical responses in function of the two critical parameters of chromatographic system (content of acetonitrile in the mobile phase [%] and concentration of ammonium acetate in water phase [mmol L⁻¹] or pH-value of water phase); A) Rs _impD,impE – resolution factor between impurity D and impurity E; B) Rs _impE,AB – resolution factor between impurity E and amlodipine besylate; C) Rs _AB,impF – resolution factor between amlodipine besylate and impurity F; D) As _AB – factor asymmetry of amlodipine besylate; E) tR _impF – retention time of impurity F

Citation: Acta Chromatographica 34, 1; 10.1556/1326.2020.00875

• Download Figure
• Download figure as PowerPoint slide

From the obtained 3D-graphs for each observed response, it can be concluded that significant two-factor interactions are present. The best Rs _impD,impE , Rs _impE,AB , Rs _AB,impF responses are obtained at acetonitrile content values of 90.5–91.5%, buffer concentration of 45–50 mM and pH value of 4.0. Increasing or decreasing the values of these factors decreases the examined resolutions, but also decreases the value of the As _AB response. Increasing the acetonitrile content and the buffer concentration in the mobile phase prolongs the analysis time (tR _impF ).

In order to define optimal chromatographic conditions for pharmaceutical analysis of test compounds, it is necessary to monitor multiple responses that have opposite goals simultaneously, i.e., to provide good resolution of three critical pairs (Rs _impD,impE , Rs _impE,AB , Rs _AB,impF ) for relatively fast analysis time, but also a good AB asymmetry peak.

Combinations of examined factors must be such that the analyzed responses have satisfactory values. For simultaneous optimization of all examined factors, and in order to obtain a global optimum (space in which all set goals are met), the method of multicriteria optimization was applied – Derringer's desirability function. The acceptable ranges for examined factors, also for the selected answers, sets the goals to be achieved by Derringer's desirability function in order to achieve the global optimum, and are given in Table 3.

By the selected values of the weighting coefficients, the preferred response function shape for each of the observed system responses is defined. Afterward, the significance coefficients values, which are used to calculate the global optimum, are determined.

As chromatographic system observed responses optimization as a goal, certain ranges are set, and shown in Table 3. All values within the set ranges are considered acceptable. The weighting coefficients are set to s = t = 1, which means that the desired answers function will linearly reach the value in the defined range. Value for coefficient of significance: for resolution factors (Rs _impD,impE , Rs _impE,AB , Rs _AB,impF ) the highest coefficient of significance was set and was 5 (priority of the first place), for answer As _AB the coefficient of significance was set to 4 (priority of the second place), and for tR _impF the significance coefficient was 2 (not priority, but still significant). Value of coefficient significance for responses was set to 3 (important to keep in range). All stated values are of pi > 1. This means that in order to achieve the global optimum (D) all factors and all answers are significant. The assigned value of the coefficient of significance, the importance of reaching the individual optimal value of the answer (di) in achieving the global optimum (D) was emphasized [23, 24].

The defined limits, weighting coefficients and significance coefficients for the examined factors and the observed system responses were processed in the Design Expert® software program and thus the combination of factors that gives the global optimum (D = 1) was determined. From many combinations of factors in which the global optimum was achieved (D = 1), the values of these factors were expressed as decimal numbers (i.e., 90.62% acetonitrile, pH = 3.65, buffer concentration 47.49 mM et cetera). However, a global optimum study was performed with the integer values of the examined factors (predicted decimal numbers are rounded to the nearest integer value). Base on the parameters of the ANOVA test, 3D graph and Derringer's desirability function, the global optimum (D = 1) was approved and the optimal chromatographic conditions were selected: 90.5% (V/V) acetonitrile, pH = 4.0, and ammonium acetate buffer concentration 50 mM. The predicted response values corresponding to the latter value of D were: Rs _impD,impE =3.52, Rs _impE,AB = 3.15, Rs _AB,impF = 3.79, tR _impF = 8.24 min, As _AB = 1.04 (Fig. 3). The response surface obtained for the global desirability function (global optimum) and the optimal chromatographic conditions are presented in Fig. 4.

View Full Size

The graphical presentation of the optimal mobile phase composition, as well as the predicted responses and corresponding adopted constraints

Citation: Acta Chromatographica 34, 1; 10.1556/1326.2020.00875

• Download Figure
• Download figure as PowerPoint slide

View Full Size

A) The 2D-graph of Desirability space in function of two critical parameters of the chromatographic process (content of acetonitrile in the mobile phase [%] and pH-value in the water phase; B) the experimentally obtained chromatogram under the optimal conditions

Citation: Acta Chromatographica 34, 1; 10.1556/1326.2020.00875

• Download Figure
• Download figure as PowerPoint slide

Validation of the HILIC method for testing amlodipine and its impurities

After defining the optimal conditions, the robustness of the method was evaluated. The influence of four factors on five system responses was investigated, and a FFD 2⁴⁻¹ to create the experiments was chosen. The experimental design is shown in Table 4. In this phase, the variations of acetonitrile content in the mobile phase and aqueous phase pH-value were followed, as well as ammonium acetate concentration in the aqueous phase and the mobile phase flow rate. The mobile phase flow rate was chosen as the fourth factor (factor D) to determine, in addition to three significant factors, its influence on the robustness of the method. This is especially important to follow changes in resolution factors.

Factor intervals were chosen to correspond to the expected factors variation (acetonitrile content ±0.5%, the aqueous phase pH-value ± 0.2 pH units, ammonium acetate concentration in the aqueous phase ±2 mmol L⁻¹ and mobile phase flow rate ±0.1 mL min⁻¹).

The obtained results for the critical pair resolution factor (Rs _impD,impE , Rs _impE,AB , Rs _AB,impF ) show that in all experiments, Rs > 1.2 values were obtained, as well as that the As _AB values are satisfactory (1.0–1.2). This confirms that a small change in all examined factors does not significantly affect the separation between these analyzed compounds. Special attention should be paid to the tR _impF response whose value varies from 6.99 to 10.80 min, ie with small changes in the examined factors there are significant changes in the duration of the analysis. From the obtained results for robustness assessment (Table 4) it can be concluded that all values of the monitored system responses are satisfactory and acceptable.

The next phase of the method robustness testing was a statistical interpretation of factors influence using a t-test and Dong's algorithm. Based on the data obtained for tracked system responses, the effect of four investigated factors was estimated. The results obtained for responses (Table 5) – statistical estimation of effects shows that all responses behave robustly within the examined range, ie all effects values have a lower value than the E _critical (Rs _impD,impE = 2.737; Rs _impE,AB = 0.643; Rs _AB,impF = 0.949; As _AB = 0.371; tR _impF = 4.670) for the probability 0.05 [25].

From the results obtained for the robustness assessment (Tables 4 and 5), it can be concluded that all the values of the tracked responses are satisfactory and acceptable, which confirms that the method is robust.

The selectivity of the method was confirmed by comparing the chromatograms of tested compounds with the placebo chromatogram (Fig. 5). It can be seen that at the retention times corresponding to AB and impurities D, E and F, there are no chromatographic peaks of substances that interfere, and that originate from placebo. It can be concluded that the method is selective.

View Full Size

Chromatogram that confirms the selectivity of the HILIC method

Citation: Acta Chromatographica 34, 1; 10.1556/1326.2020.00875

• Download Figure
• Download figure as PowerPoint slide

Linearity, accuracy and precision testing, as well as the determination of LOD and LOQ for the impurities D, E and F and amlodipine were performed and obtained results are presented in Table 6. Finally, the developed method was applied for the analysis of Vazotal® 5 mg tablets containing 6.934 mg AB. The method proved to be sensitive, specific, accurate, precise and applicable for the analysis of AB tablets.