A systematic approach for reversed phase liquid chromatographic method development of fingolimod hydrochloride via design augmentation

2.1.1 Chemicals & reagents

FIM and impurities 1–7 (Fig. 1) were synthesized at Dr. Reddy’s (Hyderabad, India) and completely characterized using NMR, Mass and IR spectroscopy. FIM capsules formulated in Dr. Reddy’s laboratories were used for the study. Tetrabutyl ammonium hydrogen sulfate (TBAHS) and ortho phosphoric acid (oPA) (85%) were obtained from MerckKGaA (Darmstadt, Germany). Acetonitrile of gradient grade was obtained from Rankem (New Delhi, India). HPLC grade water was obtained from Milli-Q water purification system (Millipore, Milford, MA, USA).

Close

Figure 1

Structure of fingolimod HCl and related compounds.

Export to PPT

2.1.2 Columns & instruments

The HPLC analytical column used was an X-Bridge C18 (150 mm × 4.6 mm, 3.5 μ) and the UHPLC column used was BEH-C18 (50 mm × 2.1 mm, 1.7 μ). The columns were manufactured by Waters (Waters Corporation, Milford, MA, USA).

LC1: Alliance 2695 separation module equipped with PDA detector (Waters) was used for specificity and development studies. LC2: Agilent 1200 series equipped with VWD (Agilent Technologies, Waldbronn, Germany) was used for validation parameters.

pH meter: Metrohm 780 (Metrohm AG, Herisaw, Switzerland).

2.1.3 Software

LC1 and LC2 were monitored with Empower 2 software (Waters). Design Expert version 8.0.7.1 (Stat-Ease Inc., Minneapolis) was used for experimental designs and interpretation. Microsoft Excel 2007 was used for analysis of validation results.

2.1.4 Chromatographic condition (HPLC)

Mobile phase A, 3 mM TBAHS and 0.02% oPA in water; mobile phase B, acetonitrile: water (9:1, v/v). Flow rate, 1.0 ml/min; column oven temperature, 60 °C; injection volume, 10 μl; and wavelength, 218 nm. Diluent for drug substance preparation is water and acetonitrile (7:3, v/v) and for drug product is 0.1% HCl in methanol: water (1:1, v/v).

Gradient program for related substances: T/%B: 0/30, 30/50, 35/80, 40/80, 45/100, 50/100. Gradient program for assay: T/%B: 0/40, 15/60, 15.1/100, 20/100.

A minimum of 10 min re-equilibration time is required between runs.

Sample concentration for related substances is 1.4 mg/ml and diluted standard of FIM for external quantification of related substances is 0.0014 mg/ml. Sample concentration for assay of FIM is 0.07 mg/ml.

For method validation, the limit for impurities 1–7 was specified as 0.15% and for unknown impurity the limit was kept as 0.10%.

2.1.5 Preparation of sample solution for drug product

The sample solution for related substances analysis was prepared from a 30 capsule composite sample. A 25 capsule equivalent mass of the capsule contents was mixed with 10 ml of diluent. The solution was sonicated for 30 min with intermittent shaking with bath temperature of 20 °C. Further the sample solution was centrifuged at 4000 rpm for 10 min. The supernatant liquid was used for related substances analysis. 1/20th concentration of related substances preparation was used for assay analysis.

2.1.6 LC-MS condition

The mobile phase A was kept as 0.1% trifluoro acetic acid. All other chromatographic conditions are as described in Section 2.1.4. Agilent 1100 series liquid chromatograph (Agilent Technologies, Inc., Santa Clara, CA, USA) coupled with an Applied Biosystems 4000 Q Trap triple quadruple mass spectrometer with Analyst 1.4 software, MDS SCIEX, USA (Thermo Scientific, Sunnyvale, CA, USA) was used for the mass identification of the degradation products. The analysis was performed in positive electro-spray/positive ionization mode with an ion source voltage of 5000 V and a source temperature of 450 °C. GS1 and GS2 were optimized to 30 and 35 psi, respectively. The curtain gas flow was 20 psi.

2.2.1 Precision

The repeatability of the related substance method was checked by injecting six individual preparations of FIM spiked with impurities 1–7 at 100% and 150% of specification level. Percentage RSD for the content of each impurity was calculated. Repeatability of the assay method was checked by injecting six preparations of FIM at three different concentrations (0.035, 0.07 and 0.105 mg/ml) and calculated % RSD for the content of FIM. The intermediate precision of the method was also evaluated using different analyst, on a different day with different make instrument in a laboratory located at different premise of Dr. Reddy’s.

2.2.2 Limit of detection (LOD) and limit of quantification (LOQ)

The LOD and LOQ for impurities and FIM were estimated by signal to noise ratio method (ICH Q2(R1), 2005). Precision study was also carried at the LOQ level by injecting six individual preparations of impurities and FIM and calculating the % RSD of the area. Accuracy at LOQ level was evaluated in triplicate for the impurities by spiking the impurities at the estimated LOQ level.

2.2.3 Linearity

Linearity test solutions for related substance method were prepared by diluting the impurities stock solution and FIM standard solution to the required concentrations. The solutions were prepared at six concentration levels from LOQ to 150% of specification limit for impurities. The calibration curve was drawn by plotting the peak areas of impurities and FIM versus its corresponding concentration. Linearity test solutions for assay were prepared at six concentration levels from 25% to 150% of assay concentration.

2.2.4 Accuracy

Standard addition and recovery experiments were conducted to determine accuracy of the related substance method. The accuracy was calculated by spiking the impurities in placebo based test solution. The study was carried out in triplicate at 50%, 100% and 150% of the specification level of impurities. The accuracy of the assay was evaluated in triplicate at three concentration levels, i.e. 0.035, 0.07 and 0.105 mg/ml in by spiking the drug substance to placebo based test solution. Placebo is a mixture of mannitol and magnesium stearate. The % recovery was calculated against 0.07 mg/ml of FIM reference standard preparation.

3.1.1 Selection of column & mobile phase pH

Column with C-18 stationary phase, which was reported earlier for FIM (Emotte et al., 2012) was chosen. To keep the molecule in unionized form, it is essential to keep the pH of the mobile phase at 2 units above the pKa for bases and 2 units below the pKa for acids. FIM being a basic substance; to analyze in unionized form it is essential to maintain the pH above 9.8 (pKa of fingolimod is 7.8). At basic pH, limitation of using high column temperature and dissolution of silica are obvious (Kirkland et al., 1995; Classens et al., 1996). Hence, we decided to go with alternate approach of using an acidic mobile phase and keeping the basic compound in completely ionized form. Hydrophobic nature of FIM (log P is 4.178), makes it possible to analyze in ionized form. Furthermore, the pKa of silanols in the chosen column (X-bridge) is greater than 8, which infers that the silanols on the silica are fully protonated below pH 6.0. Hence at acidic pHa positively charged basic compound does not interact strongly with silanols which may result in a good peak shape. Thus, we decided to operate in acidic pH.

3.1.2 Screening of mobile phase components

To screen the mobile phase components, a moderate polar to non-polar linear gradient program was fixed on a trial basis (T: %A/B as 0: 70/30, 60: 30/70). The other conditions employed for screening the mobile phase were: Flow rate – 1.0 ml/min, column temperature – 25 °C, wavelength 218 nm and injection volume – 10 μl. Impurity spiked solution was injected in each trial.

Most commonly used simple mobile phases were chosen for initial screening. To operate in acidic condition widely used phosphate buffer was chosen. Triethylamine is a well-known mobile phase additive to reduce the peak tailing of basic molecules (Bian et al., 2012; Ruiz et al., 2006; Reta and Carr, 1999). In addition to that we have decided to investigate tetrabutyl ammonium hydrogen sulfate (TBAHS) also. TBAHS helps in reducing the tailing of basic compound through various mechanisms and mainly by masking the residual silanol activity of a C18 stationary phase (Wells and George, 1982). A combination of TBAHS and oPA was reported for a chromatographic separation of basic drug (Fletouris et al., 1996).

To select aqueous phase components and organic modifier, we decided to use the traditional approach of Changing One Single (or separate) factor at a Time (COST), while keeping other factors constant. This approach, is also called as One Factor at a Time (OFAT) is convenient for a limited number of predetermined experiments. The outcome of COST is given in Table 1.

Table 1 shows clearly that, symmetric peak can be achieved with TBAHS mobile phase (Trial 2). Almost similar result was obtained with phosphate buffer and Triethylamine mobile phase. In the trials 1 and 3, separation between imp-3 and 4 was not achieved. Hence trial-2 was repeated with methanol as organic phase (trial-4). In trial-4, FIM was observed as asymmetric peak and impurities 3–5 were merged together. Hence, we decided to work further with TBAHS and oPA as aqueous mobile phase components and acetonitrile as organic modifier.

3.1.3 Selection of factors & responses for method optimization

The optimization of the method has been carried out through DoE. Factors and the levels to be optimized through DoE are given in Table 2. Rationale for choosing the factors and their range is discussed below.

In trial-2 (Table 1), tailing factor of FIM is below 1.0. The slight fronting of peak may be an indication of TBAHS more than the required amount. Hence we decided to study TBAHS from high (10 mM) to low (1 mM) concentration. TBAHS is generally used in combination with other phosphate or acetate buffer (Bielejewska and Gold, 2005). pH of an aqueous solution of TBAHS (10 mM) with or without oPA (0.1%) is about 2.2. Without combining the phosphate or acetate buffer, TBAHS alone was used as mobile phase additive for chromatographic elution of basic compound (Plum et al., 2003). Hence, we decided to investigate that whether oPA is really required to include in mobile phase and it was studied from zero level.

Beneficial effect of higher column temperature on resolution of closely eluting peaks has been demonstrated adequately (Gant et al., 1979; Lestremau et al., 2006). To improve the resolution of closely eluting pair Imp-3 and 4, we have decided to study the higher column temperature. Based on column capacity to withstand higher temperature at acidic pH and LC instrument capability, it was decided to study to the maximum of 65 °C.

From COST trials it was understood that resolution between imp-3 and 4 and tailing factor of FIM were critical tasks to achieve. In addition to that, monitoring the resolution between FIM and imp-3 is considered important, as they are closely eluting peaks. Hence, these three responses were chosen for method optimization.

3.1.4 Execution of DoE

Gradient program, flow rate, injection volume and wavelength used for COST trials (Section 3.1.2) were adopted for DoE runs. A full factorial design with center points in 4 replicates was executed for three factors (Table 3).

Standard run, run order adopted and the measured outcome (responses) is given in Table 3A. Statistically significant curvature (P < 0.05) (Table 4) was observed for all three responses, which suggest that a quadratic model may be a better fit. Hence the design was augmented by including six axial points. Axial points together with the factorial and center points constitute a central composite design (CCD). Outcome of the augmented runs is given in Table 3B.

3.1.5 Data interpretation

Classical statistical tools (Suresh et al., 2012) were employed to validate each model and reported in Table 4. In the present study, the adjusted R² was >0.90 which revealed that the experimental data show a good fit with the polynomial equations. The adequate precision value greater than 4 is desirable. In this study, the value was found to be in the range of 16.4–39 and hence the model is significant for the optimization process. A model can be considered reasonably reproducible if the coefficient of variation (C.V.) is less than 10% and in this study all the models meet the requirement.

For an experimental design with three factors, the model including linear, quadratic, and cross terms can be expressed as

The magnitudes of the coefficients in the regression equation (Table 4) were used as the basis for judging statistical significance and illustrating the relative effects of linear, quadratic and cross product interactions between the parameters. By using regression equations and the response surface graphs (Fig. 2), conclusions were drawn as follows: (1) Resolution between imp-3 and 4 (R1), which was the critical task to achieve, increases with increase in column temperature. TBAHS concentration should be kept low to improve the R1. Phosphoric acid has positive effect on R1. Hence it should be included in the mobile phase. (2) USP tailing factor (R2) decreases with increase in TBAHS concentration and column temperature has no effect on peak tailing. Inclusion of oPA will increase the R2. Hence concentration of oPA should be limited in the mobile phase. (3) Low TBAHS is preferred for improved resolution between FIM and Imp-3. Increase in column temperature may decrease the resolution.

Close

Figure 2

Response surface graphs. Each response is plotted at two temperature, (i) 30 °C and (ii) 60 °C (method condition). Red surface on the graph indicates the maximum response. Response decreases toward blue region.

Export to PPT

3.1.6 Fixing the responses for method optimization

Minimum tailing and maximum resolution are desired for a chromatographic method. To optimize the method lower and upper values of the responses are defined and given in Table 5. Rationale for fixing the responses is discussed below:

In our experience, a method with resolution between two low level impurities as 1.0 (United States Pharmacopeia, 2011) is good enough for accurate quantification. Hence lower limit kept for resolution between imp-3 and 4 was 1.0.

From Section 3.1.5 it was understood that TBAHS concentration should be kept low to increase the resolution responses (R1 & R3). At low TBAHS concentration tailing factor of FIM would be high. Achieving resolution between the imp-3 and 4 is more important than a perfect symmetrical peak and if tailing factor is high, external quantification method can be adopted for accurate quantification. Hence, an arguably accepted limit for tailing factor was kept.

While proceeding with high tailing factor, quantification of a small impurity (imp-3) which elutes immediately after a big asymmetric peak (FIM) will be accurate, only if it separates very well. Hence the desired resolution is at least 3.0.

Derringer’s desirability function (Parajo et al., 1992) was used to optimize the method for three responses with three different targets. Weights equal to 1 was given for all three responses. Highly desirable regions (Desirability = 1.0) were examined to select the point for normal operation. Contour plot of desirability and design space is given in Fig. 3. From the design space, components of the mobile phase were chosen as 3 mM TBAHS in 0.02% oPA and the column temperature as 60 °C.

Close

Figure 3

A – Design space (overlay plot of effects): yellow region indicates design space. In gray region, responses are below the desired level. Within design space, desirability plot helps to choose the region at which desired responses are at maximum level. The responses and factors at method condition are shown in the flag. Orthophosphoric acid (F(B)) is kept as 0.02%. B – Desirability plot. Red region is highly desirable. Desirability decreases toward blue region. Desirability plot can be used as a tool to choose the region at which desired responses are at maximum level in the design space. Refer Table 5 for criteria and target set for design space and desirability plot.

Export to PPT

3.1.7 Retention window

To verify that whether the predefined gradient program is a hidden factor in influencing the response, the responses measured are viewed through a RT window (RTW) (Fig. 4); a plot of retention time versus responses studied. Observation of increasing or decreasing trend of responses with respect to RT will be a clear indication of influence of gradient program.

Close

Figure 4

Retention window; a plot of responses versus retention time. No correlation between retention time and response is observed in this scatter plot.

Export to PPT

From Fig. 4 it was understood that the responses are not arranged in a linear fashion and no correlation exists between RT and responses. High and low responses are observed within one minute window. For example, in the window of 24–25 min the tailing factor ranges from 0.74 to 3.88 (RTW2). In the window of 25 to 26 min resolution between imp-3 and imp-4 is in the range of 0–1.34 (RTW1) and resolution between FIM and imp-3 ranges from 1.64 to 3.16 (RTW3). Each run experiences same ratio of aqueous and organic mobile phases at a particular RT and hence the variation observed in the responses is concluded as predominant effect of factors studied.

3.1.8 Modification of gradient program

With optimized mobile phase and column temperature, a solution of FIM spiked with impurities was injected in the predefined gradient program (T: %A/B – 0: 70/30, 60: 30/70).

From Fig. 5 it was evident that all the impurities except impurity 7 were eluted before 30 min (slice A) and impurity 7 was eluted at the edge of the gradient program (slice B). Further it was understood that this gradient program is not sufficient enough to elute highly non-polar impurities, which may found in crude or degraded samples.

Close

Figure 5

Chromatogram obtained with DoE optimized mobile phase and column temperature (Section 2.1.4) in the predefined gradient program (T/A:B: 0/70:30, 60/30:70). The chromatogram is sliced into two. Slice A represents the critical phase (T/A:B: 0/70:30, 30/50:50); where the gradient program is not to be changed. Modification of gradient program shall be done in slice B. Mobile phase A:B ratio with respect to time is given in the x-axis.

Export to PPT

As the method was optimized for critical separation, it is important to modify the gradient without disturbing DoE outcome. Hence it was decided to maintain the same ramp of the gradient for 30 min and modifying the program after 30 min.

Here, the calculated R is 0.66 and %B at 30 min is 50. The calculated ramp is represented in Fig. 5 at every 15 min intervals. The gradient program was modified for the run after 30 min, with few trials. A chromatogram obtained with the modified gradient program (finalized method conditions, refer Section 2.1.4) is given in Fig. 6. Finalized method ensured the elution of imp-7 quickly and secured the place for elution of highly non-polar impurities.

Close

Figure 6

Chromatogram obtained with finalized method conditions (refer Section 2.1.3) for related substances. Critical separation is shown in zoom.

Export to PPT

A rapid gradient program was proposed for assay method and ensured that no peak is left merged with FIM. Sample concentration for assay analysis is twenty times less than related substances analysis and hence the observed tailing factor is around 1.1. Assay method chromatogram of sample spiked with impurities is given in Fig. 7.

Close

Figure 7

Chromatogram obtained with finalized method conditions (refer Section 2.1.4) for assay.

Export to PPT

3.1.9 Specificity & mass balance study

Forced degradation was conducted on FIM sample to ensure the specificity of the method. Each degraded sample was injected in both related substances program and assay program. Condition adopted for degradation and the mass balance is tabulated in Table 6.

From Table 6 it is evident that FIM is very stable with respect to temperature, light (ICH Q1B, 1996) and water hydrolysis and almost stable to acid hydrolysis. The molecule is sensitive to base hydrolysis and oxidation. Proposed degradation pathway is given in Fig. 8.

Close

Figure 8

Proposed degradation pathway for oxidation and base degradation.

Export to PPT

A mixture of methanol: water (8:2, v/v) was used for base degradation, as the molecule was getting precipitated when using diluent as per method condition (water:acetonitrile, 30:70). The major impurity observed at the RRT 1.25 in the base degraded sample was subjected to LCMS analysis. m/z of the impurity was found to be 350.4. Reaction of aralkyl amine with carbon dioxide (CO₂) is well known (Wright and Moore, 1948). On treating with sodium hydroxide (NaOH), the HCl (counter ion of fingolimod base) was getting neutralized and the free amine was readily available for reaction with CO₂ and forms carbamate anion (Jackson et al., 2011). Formation of carbamate anion is possible even with atmospheric CO₂ (Wright and Moore, 1948) and it was observed in this study. Based on this, structure hitherto possible and Mass spectra are given in Fig. 9. Two major impurities formed in peroxide degraded sample were identified by LCMS analysis. m/z [M + H]⁺ of the impurities were 340.2 and 322.0. Formation of unstable hydro peroxide and ketones is possible during oxidation of alkyl aromatics (Rao and Awasthi, 2007) and in general benzylic protons are highly reactive (Morrison and Boyd, 1992). Based on these two facts structure hitherto possible are proposed in Fig. 9 along with their mass spectra.

Close

Figure 9

Mass spectra and chromatograms of base & oxidation degradation. A – Mass spectra and a possible structure of the impurity (carbamate anion of FIM). B – Chromatogram of base degradation. C – Mass spectra of oxidation degradation product-1. D – Mass spectra of oxidation degradation product-2. E – Chromatogram of oxidation degradation. (For LC-MS condition refer Section 2.1.6).

Export to PPT

The FIM samples analyzed did not have the base degraded impurity and oxidized impurities. Further, under accelerated stability conditions (40 ± 2 °C, 75 ± 5% RH for 6 months) the molecule is very stable and no degradation impurity is observed. Hence further attempts were not given for isolation or characterization of these degradants. Mass balance close to 99% (Table 6) shows the stability indicating power of the method. Homogeneity of the FIM peak was ensured with Photo Diode Array detector.

3.1.10 Chromatographic conditions for UHPLC

Mobile phase (3 mM TBAHS in 0.02% oPA) and column temperature (60 °C) as optimized through DoE were used for UHPLC method. The HPLC parameters used for DoE study: Gradient program (T/%A:%B: 0/70:30, 60/30:70), flow rate (1.0 ml/min) and injection volume (10 μl) were scaled to operate the method in UHPLC system. The elution was carried out in a BEH-C18 column (Waters). The dimension of the UHPLC column and formula used for scaling the method to UHPLC is given in Table 7.

A chromatogram obtained with UHPLC method is given in Fig. 10, which is much resembled to HPLC chromatogram of Fig. 5. Interested UHPLC users can further modify the gradient program as described in Section 3.1.8.

Close

Figure 10

Chromatogram obtained with UHPLC. Flow rate: 0.44 ml/min, Injection volume: 0.7 μl, Gradient program: T/A/B: 0/70/30, 10/30/70. For mobile phase and other chromatographic conditions refer Section 2.1.4.

Export to PPT

3.2.1 Precision

The % RSD for the content of impurities is within 2% for each specified impurity during repeatability and intermediate precision studies. % RSD for assay content during repeatability and intermediate precision studies is <0.4% for drug substance and <1.0% for drug product. The result indicates that method is precise.

3.2.2 Limit of detection & limit of quantification

LOD for Imp-1, 2, 3, 4, 5, 6, 7 and FIM is 0.005%, 0.004%, 0.005%, 0.005%, 0.005%, 0.005%, 0.005%, 0.003% and 0.004% respectively. LOQ for Imp-1, 2, 3, 4, 5, 6, 7 and FIM is 0.020%, 0.016%, 0.021%, 0.020%, 0.020%, 0.021%, 0.012% and 0.016% respectively. The LOD, LOQ values are with respect to sample concentration of 1.4 mg/ml and injection volume of 10 μl% RSD for the area at LOQ level is below 10% for all impurities. Recovery for impurities ranges from 89% to 96%.

3.2.3 Linearity

The correlation coefficient and coefficient of determination (r²) for impurities and FIM is ⩾0.999. In related substances method, % Y-intercept value for each impurity or FIM is within ±3% of response at 100% concentration level. For assay, the correlation coefficient and r² of FIM is 0.999 and % Y-intercept value is 0.3% of response at 100% concentration level. To assess the appropriateness of using a linear regression model to fit the data, residual plots were produced. The points in these plots were randomly distributed around the horizontal zero axis suggesting that the linear model gives a good fit of the data.

3.2.4 Accuracy

Individual and average recovery in three preparations and at three concentrations for impurities is within 100 ± 10%. For assay of FIM the recovery is within 100 ± 0.5% for drug substance and within 100 ± 2% for drug product.

3.2.5 Robustness

Design space (Fig. 3) of the method indicates robustness in terms of column temperature, and concentration of mobile additives (TBAHS and oPA). Mobile phase flow rate, which is another parameter normally performed during robustness checking (ICH Q2(R1), 2005) was also studied at ±0.2 ml/min and ensured that the system meets the suitability criteria (Table 5).