Chemometric technique for the optimization of chromatographic system: Simultaneous HPLC determination of Rosuvastatin, Telmisartan, Ezetimibe and Atorvastatin used in combined cardiovascular therapy

2.1 Apparatus

Chromatographic measurements were made on a Shimadzu (Tokyo, Japan) model which consisted of a LC10AD and LC10 ADvp solvent delivery module, SPD 10A UV–Visible detector, a Rheodyne injector (model 7125, USA) valve fitted with a 20 μl loop, and UV detector (SPD-10A). The system was controlled through a system controller (SCL-10A) and a personal computer using a Shimadzu chromatographic software (LC Solution, Release 1.11SP1) installed on it. The mobile phase was degassed using Branson sonicator (Branson Ultrasonics Corporation, USA). Absorbance spectra were recorded using an UV–Visible spectrophotometer (Model UV-1601PC, Japan) employing quartz cell of 1.00 cm of path length.

2.2 Softwares

Experimental design, data analysis and desirability function calculations were performed by using Design-Expert® trial version 7.0.0. (Stat-Ease Inc., Minneapolis). The rest of the calculations for the analysis were performed by use of Micro soft Excel 2007 software (Microsoft, USA).

2.3 Chemicals and reagents

Working standards of rosuvastatin, ezetimibe, telmisartan, atorvastatin and amlodipine (IS) were donated by M/S. Pharma analytical Lab., Puducherry, India. Acetonitrile (MeCN) and methanol (MeOH) were of HPLC grade and dipotassium hydrogen phosphate and phosphoric acid were of analytical-reagent grade supplied by M/S SD Fine Chemicals, Mumbai, India. The HPLC grade water was prepared by using Milli-Q Academic, Millipore, Bangalore, India. The pharmaceuticals Rosavel EZ, (RS-10 mg with EZ-10 mg), Avas-EZ (AT-10 mg with EZ-10 mg) and Lipisar 20 tablet (TL-20 mg with AT-10 mg) were purchased from Sun pharmaceuticals (J&K, India) Micro Labs Limited (Baddi, India) and INTAS pharmaceuticals Ltd. (Ahamedabad, India) respectively.

2.4 Standard solutions

Stock standard solutions of RS, TL, EZ and AT (1 mg/ml) were prepared in mobile phase. The prepared stock solution was stored at 4 °C protected from light. Working standard solutions were freshly obtained by diluting the stock standard solutions with mobile phase during the analysis day. Calibration curves reporting peak area ratios of RS, TL, EZ, and AT to that of the IS versus drug concentrations were established in the range of 0.5–5 μg/ml for RS, EZ, AT and 1–10 μg/ml for TL in the presence of amlodipine (2.5 μg/ml) as internal standard. The standard solution prepared for the optimization procedure constituted RS, TL, EZ, AT and IS at 10.0, 10.0, 10.0, 10.0 and 5 μg/ml, respectively.

2.5 Sample preparation

Twenty tablets were weighed and finely powdered. An amount of pharmaceutical products powder equivalent to 10 mg of RS with 10 mg of EZ, 10 mg of AT with 10 mg of EZ, and 10 mg of AT with 20 mg of TL were accurately weighed and transferred into a 50 ml volumetric flask; suitable quantity of IS was added followed by 25 ml of mobile phase. This mixture was subjected to sonication for 10 min for complete extraction of drugs and the solution was made up to the mark with a mobile phase to obtain a concentration of RS, TL, EZ, AT and IS as 2.5, 5.0, 2.5, 2.5 and 2.5 μg/ml, respectively. The solution was centrifuged at 4000 rpm for 10 min; the clear supernatant was collected and filtered through a 0.2 μm membrane filter (Gelman Science, India) and 20 μl of this solution was injected for HPLC analysis.

2.6 Chromatographic procedure

Chromatographic separations were carried out on a Phenomenex® C18 analytical column (150 mm × 4.6 mm i.d., 5 μm) connected with a Phenomenex® C18 guard cadridge (4 mm × 3 mm i.d., 5 μm). The mobile phase consisted of MeOH–MeCN–dipotassium hydrogen phosphate buffer (pH 3.0), adjusted with 10% phosphoric acid. Wavelength of 239 nm was selected for detection. An injection volume of the sample was 20 μl. The HPLC system was used in an air conditioned laboratory atmosphere (20 ± 2 °C).

2.7 Validation

Validation studies were conducted using the optimized assay conditions based on the principles of validation described in the ICH guidelines “Text on Validation of Analytical Procedures”(International Conference on Harmonization, Q2A, 1995) and “Q2B, Validation of Analytical Procedures: Methodology” (International Conference on Harmonization, Q2B, 1997). Key analytical parameters, including, specificity, accuracy, precision, linearity, detection limit and quantitation limit were evaluated. For specificity study, placebo containing starch, lactose monohydrate, aerosil, hydroxypropyl methylcellulose, titanium dioxide and magnesium stearate was used. The calibration curve was tested using one-way ANOVA at 5% significance level (Thanikachalam Sivakumar et al., 2007). Calibration curves were constructed in a low region of 0.05–1.0% of the target analyte concentration for the limit of detection and quantification (Crowther, 2001). Also, robustness of the proposed method was assessed with respect to small alterations in the MeCN concentration (34.27 ± 0.5%), the pH value (3.0 ± 0.2) and the buffer concentration (20 ± 2.0 mM).

3.1 Data analysis and optimization design

The experimental design approach to HPLC method development relies on two stages of experimentation: screening and optimization. The purpose of the screening is to identify the factors that had significant effect on the responses and to investigate the curvature term using Factorial design with center points. Factorial design consisting of eight factorial runs along with six other experiments at the center of the design points was carried out to estimate the experimental error. The design experiments were produced in the random order. Before starting an optimization procedure, ANOVA was generated for 2^k factorial design which shows that curvature is significant for the variable (k₅) since p-value is less than 0.05. This implies that a quadratic model should be considered to model the separation process (Ting et al., 2009). In order to obtain second order predictive model, central composite design (CCD) is employed, which is a design type under RSM. CCD is chosen due to its flexibility and can be applied to optimize an HPLC separation by gaining better understanding of factor's main and interaction effects. (Wang et al., 2006a,b) The selection of key factors examined for optimization was based on preliminary experiments and prior knowledge from the literature The factors selected for optimization process were MeCN concentration (A), buffer molarity (B) and flow rate (C). The capacity factor for the first eluted peak (k₁), the resolution of the critical separated peak, RS and TL (Rs_2,3) and the capacity factor of the last peak, AT (k₅), were selected as responses. All experiments were conducted in a randomized order to minimize the effects of uncontrolled variables that may introduce a bias on the measurements. Replicates (n = 6) of the central points were performed to estimate the experimental error. Table 1, summarizes the conducted experiments and responses. The quadratic mathematical model for three independent factors is given in Eq. (1):

The interaction term with the largest absolute coefficients among the fitted models is BC (+0.24) of k₅ model can be seen in Table 2. The positive interaction between B and C is statistically significant (<0.0001) for k₅. The study reveals that changing the fraction of MeCN from low to high results in a rapid decline in the k₅ of AT both at the low and high levels of buffer molarity. Further at low level of factor A, an increase in the buffer molarity results in a marginal decrease in k₅. This may be due to reduced silanol effects as a result of higher buffer molarity used. Therefore, when the MeCN concentration is set at its lowest level, the buffer concentration has to be at its highest level to shorten k₅. Especially this interaction is synergistic, as it led to a decrease in k₅.

In Fig. 2 perturbation plots are presented for predicted models in order to gain a better understanding of the investigated procedure. This type of plots shows the effect of an independent factor on a specific response, with all other factors held constant at a reference point (Sivakumar et al., 2007b). A steepest slope or curvature indicates sensitiveness of the response to a specific factor. Fig. 2 (c) shows that MeCN (factor A) had the most important effect on capacity factor k₅ followed by factor C and then B. In Fig. 2a k₁ values increased as the levels of buffer concentration (factors B) increased while in Fig. 2a and b k₁ and Rs_2,3 values decreased as the levels of flow rate (factors C) increased.

Close

Figure 2

Perturbation plots showing the effect of each of the independent variables on k₁, Rs_2,3 and k₅. Where A is the concentration of acetonitrile, B the buffer molarity and C the mobile phase flow rate.

Export to PPT

Response surfaces plots for k₁, Rs_2,3 and k₅ are illustrated in Fig. 3 (% acetonitrile concentration is plotted against the flow rate with buffer concentration held at constant at the center value for Rs_2,3 plot, for k₁ and k₅, % acetonitrile concentration is plotted against the buffer concentration with flow rate held at constant). Analysis of the perturbation plots and response plots of optimization models revealed that factors A and C had the significant effect on separation of the analytes, whereas factor B, i.e. the buffer molarity, is of little significance.

Close

Figure 3

In figure (a) and (c), response surfaces related to percentage acetonitrile concentration (A) and Buffer molarity(B): Flow rate was kept constant and in figure (b), response surfaces related to percentage acetonitrile concentration (A) and Flow rate (C): Buffer molarity was kept constant (a) capacity factor of the first peak (k₁), (b) resolution of the critical pair (Rs_2,3), (c) capacity factor of the last peak (k₅).

Export to PPT

3.2 Derringer's desirability function

In the present study, the identified criteria for the optimization were: resolution between the critical peaks, capacity factors k₁ and k₅. Derringer's desirability function was used to optimize three responses with different targets (Derringer and Suich, 1980). The Derringer's desirability function, D, is defined as the geometric mean, weighted, or otherwise, of the individual desirability functions. The expression that defines the Derringer's desirability function is:

Close

Figure 4

Bar graph showing individual desirability values of various objective responses and their association as a geometric mean (D) corresponding to formulation samples.

Export to PPT

To substantiate the flexibility of the optimization strategy and to search for an optimum experimental condition for analyzing plasma samples, criteria II was established by varying the response goals and their importance values (Table 3). For instance, high value of k₁ has to be selected for the separation of IS from the initial disturbances of plasma components. Therefore, k₁ was targeted at 1.4 and high importance value of 5 was assigned. Following the response goals above, the optimization procedure was carried out for which optimal conditions II with the maximum desirability value (D = 0.837) were MeCN concentration of 33.26%, buffer molarity of 19.34 mM and flow rate of 2.00 ml/min. In order to investigate the predictability of the proposed model, the agreement between experimental and predicted responses for both the predicted optimums I and II are shown in Table 4. The Percentage of prediction error was calculated by Eq. (3). The prediction efficiency of the model confirmed by performing the experiment under the optimal conditions I and II is also presented Fig. 5B and F, respectively. This approach offers flexibility to the chromatographer to slide the k₁ values depending upon the environment of the analyte under consideration.

Close

Figure 5

Chromatograms corresponding to (A) a placebo solution; (B) a synthetic mixture of IS (4.8 μg/ml), RS (10.2 μg/ml), TL (10.01 μg/ml), EZ (9.98 μg/ml) and AT (9.97 μg/ml); (C) a real sample of Rosavel-EZ containing IS (2.6 μg/ml), RS (2.5 μg/ml) and EZ (2.52 μg/ml); (D) a real sample of Lipisar 20 tablets containing IS (2.49 μg/ml), TL (4.97 μg/ml) and AT (2.50 μg/ml); (E) a real sample of Avas-EZ containing, IS (2.50 μg/ml), EZ (2.48 μg/ml) and AT (2.5 μg/ml) under optimum assay conditions I for formulation and (F) a synthetic mixture of IS (4.9 μg/ml), RS (9.94 μg/ml), TL (9.91 μg/ml), EZ (10.1 μg/ml) and AT (9.92 μg/ml) under optimum assay conditions II for plasma.

Export to PPT

3.3 Assay method validation

The present study was to check method's validation for specificity, linearity, accuracy, intra/inter-day precision, and robustness. The method specificity was assessed by comparing the chromatograms of standard drugs with those of placebo solutions obtained from the most commonly used excipients in pharmaceutical formulation, which included lactose monohydrate, aerosil, starch, hydroxy methylcellulose, magnesium stearate and titanium dioxide, There were no excipient peaks co- eluted with analyte and IS, indicating that the optimized assay method is selective and specific in relation to the excipients used in this study. All placebo chromatograms showed no interference peaks (Fig. 5). An excellent linearity was established at five levels in the range of 0.5–1.0 μg/ml for IS, RS, EZ and AT and 1.0–10 μg/mL for TL with R² of more than 0.997 for all the analytes. The slope and intercept of the calibration curve were 0.375 and 0.0003 for RS, 0.614 and 0.043 for TL, 0.382 and 0.002 for EZ and 0.335 and 0.0023 for AT, respectively. Since the correlation coefficients are not good indicators of linearity performance of an analytical procedure (Danzer and Currie, 1998) a one way ANOVA was performed. For all the analytes, the calculated F-value (Fcalc) was found to be less than the theoretical F-value (Fcrit) at 5% significance level, indicating that there was no significance difference between replicate determinations for each concentration level. The LOD and LOQ were estimated as 0.88 and 2.66 ng/ml for RS, 0.50 and 1.51 ng/mL for TL, 1.00 and 3.02 ng/ml for EZ, and 0.75 and 2.27 ng/mL for AT, respectively. Accuracy (n = 9), assessed by spike recovery, were found to be 99.72%, 99.85%, 99.78% and 99.66% for RS, TL, EZ and AT, respectively, which were within acceptable ranges of 100 ± 2% (Kleinschmidt, 2005). The intra and inter-assay precision (n = 6) was confirmed since, the % C.V. were well within the target criterion of ⩽2 and ⩽3, respectively (Kleinschmidt, 2005). Robustness study reveals that small changes did not alter the retention times, retention factor and resolutions more than 4% and therefore it would be concluded that the method conditions are robust.

3.4 Application of the method

The proposed RP-HPLC method was applied to the quantitative analysis of real samples, three commercial tablet products Rosavel EZ tablet (RS-10 mg with EZ-10 mg), Avas-EZ tablet (AT-10 mg with EZ-10 mg) and Lipisar 20 tablet (TL-20 mg with AT-10 mg) were assayed by the proposed HPLC method. Representative chromatograms are presented in Fig. 5. The results achieved when analyzing Rosavel EZ tablets were, 9.98 (0.97) mg of RS and 10.01 (1.11) mg of EZ; Avas- EZ tablets were, 10.02 (1.38) mg of AT and 10.02 (1.05) mg of EZ and Lipisar 20 tablet were, 20.06 (0.44) mg of TL and 10.08 (1.38) mg of AT with the values within parenthesis being the % C.V. of the six replicates. Good agreement was found between the assay results and the label claim of the product. The % C.V. for tablets were <2, indicating the precision of the analytical methodology. The mean recoveries for each analyte were also tested for significance to establish whether the recovery means are different from the label claim of the tablets by employing Student's t-test.