A simplified guide for charged aerosol detection of non-chromophoric compounds—Analytical method development and validation for the HPLC assay of aerosol particle size distribution for amikacin

Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis journal homepage: www.elsevier.com/locate/jpba

A simplified guide for charged aerosol detection of non-chromophoric compounds—Analytical method development and validation for the HPLC assay of aerosol particle size distribution for amikacin Arianne Soliven a,1 , Imad A. Haidar Ahmad a,1 , James Tam a , Nani Kadrichu a , Pete Challoner b , Robert Markovich b , Andrei Blasko a,∗ a b

Novartis Pharmaceuticals Corporation, San Carlos, CA, USA Nektar Therapeutics, San Francisco, CA, USA

a r t i c l e

i n f o

Article history: Received 3 March 2017 Received in revised form 28 April 2017 Accepted 1 May 2017 Keywords: Charged Aerosol Detection CAD Power function value Evaporator temperature CAD linearity Validation HPLC-CAD

a b s t r a c t Amikacin, an aminoglycoside antibiotic lacking a UV chromophore, was developed into a drug product for delivery by inhalation. A robust method for amikacin assay analysis and aerosol particle size distribution (aPSD) determination, with comparable performance to the conventional UV detector was developed using a charged aerosol detector (CAD). The CAD approach involved more parameters for optimization than UV detection due to its sensitivity to trace impurities, non-linear response and narrow dynamic range of signal versus concentration. Through careful selection of the power transformation function value and evaporation temperature, a wider linear dynamic range, improved signal-to-noise ratio and high repeatability were obtained. The influences of mobile phase grade and glassware binding of amikacin during sample preparation were addressed. A weighed (1/X2 ) least square regression was used for the calibration curve. The limit of quantitation (LOQ) and limit of detection (LOD) for this method were determined to be 5 g/mL and 2 g/mL, respectively. The method was validated over a concentration range of 0.05–2 mg/mL. The correlation coefficient for the peak area versus concentration was 1.00 and the y-intercept was 0.2%. The recovery accuracies of triplicate preparations at 0.05, 1.0, and 2.0 mg/mL were in the range of 100–101%. The relative standard deviation (Srel ) of six replicates at 1.0 mg/mL was 1%, and Srel of five injections at the limit of quantitation was 4%. A robust HPLC-CAD method was developed and validated for the determination of the aPSD for amikacin. The CAD method development produced a simplified procedure with minimal variability in results during: routine operation, transfer from one instrument to another, and between different analysts. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Amikacin, (2S)-4-amino-N-{(1R,2S,3S,4R,5S)-5-amino-2-[(3amino-3-deoxy-␣-d-glucopyranosyl)oxy]-4-[(6-amino-6deoxy-␣-d-glucopyranosyl)oxy]-3-hydroxycyclohexyl}-2hydroxybutanamide, a broad spectrum aminoglycoside antibiotic derived from kanamycin A, is commonly used for treating severe, hospital-acquired infections caused by Gram-negative bacteria. Due to the molecule’s lack of a UV chromophore, its analysis has always been challenging; therefore, pre- and post-column deriva-

∗ Corresponding author. E-mail address: [email protected] (A. Blasko). 1 Co-First authorship, contributed equally to the manuscript. http://dx.doi.org/10.1016/j.jpba.2017.05.013 0731-7085/© 2017 Elsevier B.V. All rights reserved.

tization and non-UV detection techniques have been developed to monitor its content in pharmaceutical formulations. Aminosugar analytical methods using pre-column [1] derivatization in liquid chromatography (LC), capillary electrophoresis (CE) [2]. and micellar electrokinetic chromatography (MEKC) or post-column [3] derivatization or complexation in LC prior to UV or fluorescence detection have been reported [4]. Both detection techniques have pros and cons; the choice of one or the other depends on the analyte, its derivatization site(s) or, in many cases, the preference or expertise of the analyst. However, due to the reactivity of multiple functional groups, post-column derivatization is recommended [5]. The drawbacks associated with derivatization techniques are that they can be time-consuming, labor intensive, and difficult to quantitate, can demonstrate a larger overall variability due to extra sample preparation steps, and the reactions can

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

69

Fig. 1. Plot of peak area as a function of concentration for power function values of 1.00 (open circles) and 1.32 (open squares) at evaporative temperature of 35 ◦ C. Both axes are on logarithmic scale to show the variability of the best fit line at low concentration.

often be difficult to control. Other methods use resonance Rayleigh scattering [6], chemiluminescence [7], cyclic voltammetry [8], or even colorimetric methods based on gentamicin-induced collapse of an Au–lipid capsule [9]. For a quality control (QC)–friendly method it is always preferable to have a direct detection mode; evaporative light scattering detection (ELSD) has therefore been used for the determination of amikacin in drug products [10]. There also exist LC methods that push the UV detection limits to 191 nm [11], or MEKC methods at 200 nm [12], but these approaches are not viable for stability-indicating methods. The ion chromatography (IC) technique with pulsed amperometric detection (PAD) became more appealing to the analytical community as the instrumentation became more robust [13]. The United States Pharmacopeia (USP) monograph and the European Pharmacopeia (Ph. Eur.) monographs for amikacin use ion chromatography with electrochemical detection in integrated amperometric mode [14]. Pre-column derivatization with 2,4,6-trinitrobenzene sulfonic acid with UV detection has also been reported [15]. Existing analytical methods including the USP and Ph. Eur. methods, are still being used; however, more robust methods with better precision are needed. In the pharmaceutical industry, advances were implemented to improve existing separations and aid method development. The corona charged aerosol detector (CAD) was developed as a direct detection mode for non-chromophoric compounds as an alternative to ELSD and was first commercially released in 2004 [16]. This mode of detection is mass sensitive, in contrast to the UV concentration dependent detector, and can detect molecules with weak or no chromophores [17]. Since its development, this detection technique has been implemented in industries, such as the pharmaceutical industry, that require rugged and robust quantitative methods [18]. CAD approaches have been criticized for their limited linear working dynamic detection range [18,19]. This narrow dynamic range resulted from the difficulty in distinguishing between tiny charged spherical particulates with smaller masses and larger charged particulates with higher masses, since both follow a nonlinear signal versus charge relationship. The main disadvantage of this detector was thus its non-linear response, which complicated quantitation [20]. However, a log–log transformation of the peak area response for specific concentrations, followed by linear regression was suitable for calibration purposes [20].

The linear dynamic range of CAD detectors has been gradually extended through exploitation of a power transformation before signal output. For example, the manipulation of raw data utilizing a power transformation resulted in increased peak intensity and, decreased peak width, while peak asymmetry remained constant [21,22]. Numerous raw data sets, raised to different power functions, were shown to practically improve 1DLC, 2DLC, and postcolumn derivatization separations [22], demonstrating the benefit of embedding the power transformation into the instrumentation’s firmware [21]. Despite the foregoing, the use of caution and working within an experimentally-determined calibration curve were highly recommended by the manufacturers; extrapolation was not advised [23]. Optimization based upon an empirical approach has been recommended as a result of a critical evaluation of the use of CAD in the pharmaceutical industry [19]. CAD detection requires more parameters for optimization when compared to UV and fewer when compared to mass spectrometry (MS) detection techniques. The reluctance associated with utilizing CAD is diminished by understanding both the non-linear detection response and the sensitivity of the detector to all non-volatile compounds. Additionally, method development is tightly coupled to the volatility of the compound of interest; hence CAD method optimization is compound specific [19]. In developing a practical CAD method, one should consider that: (1) stationary phase bleeding effect is more pronounced with CAD, and (2) the column must be selected carefully to ensure that the peak of interest is not obstructed [19]. The use of LCMS grade solvents is required to achieve ultimate sensitivity as trace particulates and impurities in HPLC grade solvents contribute to a higher background signal. Any mobile phase additives (e.g. pH buffer, ion-pair reagents) must also be selected for higher volatility than the target compound(s). An alternative approach to minimize the influence of the mobile phase volatility on the response, utilizes a T-piece and second identical pump in the workflow to deliver a reverse gradient with respect to the chromatographic separation gradient. However, this involves a more complicated workflow set-up and is not practical for all laboratories [24]. Aerosol–based detectors have been reviewed in terms of operation and application in different industries [18,25]. The main advantage of CAD over other aerosol detectors is its sensitivity − which can be up to four orders of magnitude higher than ELSD [25].

70

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

Fig. 2. Response factor as a function of concentration for power function 1.00 (open circles) and power function 1.32 (open squares), Tevaporator = 35 ◦ C.

Also, in contrast to other detectors, the CAD is compatible with a lower flow rate, due to the need to volatilize the sample [25]. Additionally, because CAD only relies on a single nebulizer, in contrast to ELSD that uses multiple nebulizers, it is less expensive and easier to maintain [20,26]. In newer detectors, the nebulizer pressure setting is optimized by the manufacturer. Practical and fundamental literature studies, together with commercial improvements, have resulted in a technique that is more user friendly and robust for use in R&D settings as well as in highly controlled, regulated laboratories [22,23,26]. Recently, Hetrick et al. evaluated the direct detection of sugars and sugar alcohols with aerosol based detectors (ELSD and CAD) [27]. However, due to the long equilibration step required for HILIC phases in comparison to reversed phase stationary phases, the method run time was long (6 min initial equilibration step, 20 min total run time). In terms of throughput, this approach is limited to 3 samples per hour or 72 samples per day. Furthermore, for the low target levels the recovery accuracy ranged from 109 to 113%, with a precision% RSD of 0.5-6.7 [27]. These results may fall outside the accuracy and sensitivity needed for a late phase routine assay method. The present study focused on developing a robust HPLC-CAD assay method that could (i) pass stringent late-phase method validation according to International Conference on Harmonization (ICH) guidelines and (ii) support routine throughput of 200 samples per day. The power function value (PFV) and evaporative temperature parameters were optimized while evaluating the response factor variability, linearity, and limit of quantitation. The developed HPLC-CAD assay method passed validation and was successfully transferred to a contract lab. 2. Experimental section 2.1. Materials Amikacin reference standard was sourced from USP (Rockville, MD, USA). LC/MS grade methanol was obtained from EMD (Billerica, MA, USA). HPLC grade water was obtained from Honeywell (Muskegon, MI, USA). Pentafluoropropionic acid (PFPA) was obtained from Oakwood Chemicals (West Columbia, SC, USA). Screened chromatography phases included Zorbax Extend C18

3.5 m, Zorbax Eclipse XDB C18 3.5 m, and Poroshell 120 bonus RP 2.7 m from Agilent Technologies (Santa Clara, CA, USA) and HALO C18 2.7 ␮m, HALO RP-amide 2.7 ␮m, HALO Phenyl-Hexyl, 2.7 m, ACE Excel SuperC18 3 m, and ACE UltraCore Super C18 2.5 m from Mac-Mod (Chadds Ford, PA, USA). 2.2. Instrumentation The initial column selection experiments were performed on a Waters 2695 HPLC system connected to a Corona CAD detector (Thermo Scientific, Chelmsford, MA, USA). The final method development chromatographic separations were performed on an Agilent 1260 HPLC system that incorporated a quaternary pump, online degasser, auto-injector, and thermostated column compartment. The column outlet was connected directly to the CAD Veo RS Charged Aerosol Detector, bypassing the detector’s divert valve. The system dwell volume was determined to be 1.15 mL. The reversed ® phase column employed was the ACE UltracoreTM SuperC18, 2.5 m, 100 × 4.6 mm column. Chromeleon 6.80 software (Thermo Scientific) was used for data acquisition and analysis. 2.3. HPLC − CAD method The amikacin reference and check standards were prepared at concentrations of 2.0 and 1.0 mg/mL, respectively, in mobile phase A. The amikacin standards at 1.0 and 0.05 mg/mL were prepared by diluting the 2.0 mg/mL reference standard with mobile phase A. The calibration curve was constructed from weighed (1/X2 ) least square regression of the 0.05, 1.0, and 2.0 mg/mL reference standard injections. Mobile phase A contained 0.1% PFPA in water, and mobile phase B contained 0.1% PFPA in methanol. A mobile phase gradient under reversed-phase conditions was applied to separate and quantitate amikacin: 0.00-1.00 min (5% B), 1.00-4.50 min (5–90% B), and 4.51-7.00 min (5% B) (re-equilibration). The column flow rate was set at 1.5 mL/min. The column temperature was set at 55 ± 2 ◦ C, and the auto-sampler temperature at 23 ± 2 ◦ C. The injection volume was 3 L, and the needle wash was neat methanol. Data collection rate was set at 5 Hz, using a filter constant of 3.6 s. CAD evaporator temperature was set at 50 ± 2 ◦ C. The nitrogen gas pressure for the nebulizer was set at the optimized, manufacturer recommended setting specific to the nebulizer.

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

71

Fig. 3. Change in the slope of response factor versus concentration plots (solid circles) and in relative standard deviation of the response factors (open circles) as a function of the power function value. 0 min (5% B), 8 min (40% B), 9 min (90% B), 9.1-13 (5% B). F = 1.0 mL/min. T = 55 ◦ C.

The following system set-up procedure was performed on each analysis day. The column was replaced with a zero-dead volume connector, and 100% water was pumped through the system for at least 30 min at approximately 2.0 mL/min, until a stable baseline was achieved. The column was then installed and heated to 55 ± 2 ◦ C and 100% mobile phase B was pumped through the system at 2.0 mL/min flow rate for at least 60 min until a stable baseline was achieved. The initial gradient conditions were then set. After each sequence, the column at 55 ± 2 ◦ C was washed with 100% methanol at 1.0 mL/min for 60 min. During the optimization process, the following two gradients were used:

(1)

0–1 min(5%B), 4.5 min(95%B), 4.51-7.00(5%B). ◦

F = 1.5 mL/min.T = 55 C.V inj = 3 L.

S1 = So PFV + k

(1)

where, So is the raw detector signal, S1 is the transformed detector signal, PFV is the power function value (0.67 − 2.00), and k is a constant. 3. Results and discussion

0 min(5%B), 8 min(40%B), 9 min(90%B), 9.1–13(5%B). F = 1.0 mL/min.T = 55 ◦ C,V inj = 2 L

The raw signal current depends on the number of charges on a particle and is not linear with respect to sample concentration, since it depends on the surface area rather than the volume of the particle. The application of a built-in power function enhances the apparent linearity of the detector over a wider concentration range, a key requirement for the calibration curve (Eq. 1).

(2)

A solution of 0.1% (v/v) PFPA in water, used as mobile phase A in the final optimized HPLC-CAD method, was selected as the sample diluent. The adsorptive loss of aminoglycosides in general has been reported [28] and can be mitigated by protonating the amino groups via acidic pH adjustment. 2.4. CAD: power function value and evaporator temperature optimization The CAD detector works by nebulizing the column effluent stream into small droplets by means of the Venturi effect, with nitrogen as the carrier gas, followed by evaporation of the mobile phase [19]. A charged nitrogen gas (over a corona discharge) collides with the resulting particulate stream and charges the particulates. An ion trap removes the excess charged nitrogen gas, and the charged particles are detected by an electrometer. The signal is mathematically transformed into a chromatographic response [19–22].

3.1. CAD power function value and evaporator temperature optimization By plotting the amikacin peak area as a function of concentration (Fig. 1), the CAD signal experienced a clear deviation from linearity at a power function value of 1.00. In Fig. 1, both axes are scaled logarithmically for an easier visual assessment of deviations from linearity, particularly at both low and high concentrations. This was used only for the optimization of the PFV, which involved an exhaustive empirical study of values ranging from 1.00-1.70 (data not shown). As shown in Fig. 1, the effect of applying an appropriate power function value of 1.32 considerably improved the linearity of the signal. The deviation from the linear regression line at low concentration was minimized, while the correlation coefficient was significantly improved. Due to the large number of chromatographic conditions and permutations, each of which affected the optimal PFV to achieve a linear signal, plotting the peak area as a function of concentration was not practical for use in method development. Additionally, the correlation coefficient of the linear regression lacked sensitivity at high concentrations. An alternative method was utilized to assess the linearity of the signal by evaluating the response factor (ratio of peak area/concentration) rather than the peak area as a function of concentration. The data obtained for Fig. 1 were replotted in Fig. 2 to show the variability of the response factor as a function of concen-

72

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

Fig. 4. Change in the slope of response factor versus concentration plots (solid circles) and in relative standard deviation of the response factors (open circles) as a function of the evaporator temperature. Conditions, refer to Fig. 3.

Fig. 5. Detector response as a function of retention time overlay of the 0.05, 0.1, 0.2, 0.5, 0.8, and 1 mg/mL of amikacin peak. Gradient conditions: 0–1 min (5% B), 4.5 min (95% B), 4.51-7.00 (5% B). F = 1.5 mL/min. T = 55 ◦ C. Vinj = 3 L.

tration. As shown in Fig. 2, a PFV value of 1.32 resulted in a more precise response factor across a wider concentration range, and the slope of this plot approached zero. To select the optimal PFV, the slope and relative standard deviation (Srel ) of the response factor as a function of concentration were evaluated across a range of PFVs as shown in Fig. 3. This type of representation was instrumental in determining the optimal PFV. Using the same approach that was used to select the optimum PFV, the evaporator temperature was assessed as shown in Fig. 4. An optimal post-nebulization droplet evaporator temperature of 40 ◦ C was selected based on the slope value and the Srel . It is important to note that the optimum PFV and evaporator temperature values are dependent on the separation conditions, such as, for example, flow rate, mobile phase composition, mobile phase additives, temperature, and gradient slope. This was evident when further optimization of the gradient conditions was desired to improve the peak shape as discussed in the next section.

3.2. Chromatographic gradient optimization Columns of a wide range of selectivity were chosen based on the hydrophobic subtraction model [29,30]. The suitable column was chosen based on amikacin peak width and asymmetry as well as resolution between amikacin and kanamycin. On the selected column, the peaks obtained with gradient 1 (Experimental section) were wide and showed unacceptable peak distortion at higher concentrations. In addition, unacceptably high peak area variability was observed at low concentrations. The method was improved by increasing the flow rate from 1.0 mL/min to 1.5 mL/min and increasing the steepness of the gradient profile. These changes resulted in improved peak width, peak shape, and signal-to-noise ratio. As a result, a determination of optimum PFV and evaporator temperature had to be repeated at the modified gradient conditions; the corresponding optimum PFV and evaporator temperature were 1.39 and 50 ◦ C, respectively. Under these optimized conditions, the variability in peak areas from 16 injections at each concentration

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

73

Fig. 6. Sequence of blank injections (1–5) showing decrease in baseline noise and undesirable peaks due to inadequate system washing before the start of the sequence. The numbers inside the vertical axis indicate the order of the blank injections. Insert: baseline with HPLC grade methanol (6) versus MS grade methanol (7) in mobile phase B.

Table 1 Validation parameters, acceptance criteria and resultsa . Validation Parameter

Methodology

Acceptance criteria

Results

Accuracy

Mean recovery at 3 concentration levels (0.05, 1.0, and 2.0 mg/mL amikacin), 3 replicates per level

Mean recovery at 0.05 mg/mL level: 90%−110%% Mean recovery at 1.0 and 2.0 mg/mL levels: 95–105%a Srel ≤ 10%

0.05 mg/mL level: 100% 1.0 mg/mL level: 101% 2.0 mg/mL level: 100%

Srel for recovery (n = 9)

Srel : 1%

Precision-Repeatability

Srel (n = 6 samples at 100% nominal concentration of 1.0 mg/mL)

Srel ≤ 3%

Srel : 1%

Intermediate precision

Srel (n = 12, (6 × 2)) per Kojima design

Srel ≤ 3%

Srel : 1%

Specificity

Record chromatograms and compare visually: diluent, placebo solution, stressed sample, Kanamycin sample, extract from material that may come in contact with sample.

No interference with amikacin peak larger than 10% of reporting limit concentration of 0.05 mg/mL.

No interference with amikacin peak larger than 10% of reporting limit concentration of 0.05 mg/mL.

Linearity

Amikacin solution (n = 6) from 0.05 to 2.0 mg/mL: Correlation coefficient (R), y-intercept, and Residual standard deviation

R ≥ 0.99 y-intercept ≤ 3%a Residual standard deviation ≤ 5%a

R: 1.00 y-intercept: 0% Residual standard deviation: 1%

Range

Report quantitation range from precision, accuracy and linearity.

Report as mg/mL and as % of 1.0 mg/mL nominal concentration

Range: 0.05–2.0 mg/mL amikacin or 5–200% of 1.0 mg/mL

Reporting limit

Determine from limit of quantitation, accuracy, and linearity

Report as mg/mL and as % of 1.0 mg/mL nominal concentration

Reporting limit: 0.05 mg/mL or 5% of 1.0 mg/mL

Limit of quantitation (LOQ)

Calculate Srel of the amikacin peak area response of LOQ test solution (5 g/mL), n = 5 injections

Srel ≤ 15%

LOQ: Srel = 4%

Limit of detection (LOD)

Calculate as limit of quantitation divided by 3

Report

LOD: 2 g/mL

a

Relative to 100% concentration of 1.0 mg/mL.

ranging from 0.05 to 1.0 mg/mL and 6-L injection volume, was less than 1.5%. The high end of the concentration range in the final method was extended to 2.0 mg/mL by reducing the injection volume to 3 L (to maintain same column load) with no impact on variability. The overlay of the chromatograms at the different concentration levels showed minimal peak distortion as illustrated in Fig. 5.

3.2.1. CAD operational improvements In addition to optimizing flow rate and gradient profile, additional approaches were utilized to reduce baseline noise including (i) adequate wash steps using a 100% aqueous wash (without column) to efficiently purge the system of any residual inorganic salts, and (ii) a 100% organic wash with mobile phase B to purge the system of any organic residue prior to initiating system suitability and an injection sequence. It was necessary to perform an exten-

74

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

Table 2 Analysis of variance (ANOVA) results. ANOVA Results Source of variation

Sum of square

Degrees of freedom

Mean of square

Between Groups Within Groups Total

0.00064272

5

0.000128544

0.00055564

6

9.26067E-05

0.00119836

11

Overall mean: 1.0133 mg/mL; Overall standard deviation: 0.01044; Overall Srel : 1%.

Table 3 Linearity results from 0.05 to 2.0 mg/mL amikacin. Validation Parameter

Acceptance Criteria

Results

Correlation coefficient (R) % y-intercept Residual Std. Deviation

≥0.99 ≤3% ≤5%

1.00 0 (0.2) % 1 (1.1) %

Validation Parameter

Results

Slope Slope standard deviation y-intercept y-intercept standard deviation

4.04523322 0.02529347 0.00837266 0.02804524

sive wash procedure of both the HPLC system and column due to the sensitivity of the CAD, in order to minimize baseline noise and extraneous peaks. Fig. 6 shows the reduction in the noise and the gradual disappearance of the contaminant peaks and baseline shift following blank injections to a system that had not been properly cleaned. The methanol used for mobile phase B was switched from HPLC grade to mass spectrometry (MS) grade methanol. The improvement in using MS grade compared to HPLC grade methanol is shown in the inset in Fig. 6. The baseline of a blank injection with MS grade methanol showed lower noise and decreased number of contaminant peaks. 3.3. HPLC-CAD method validation Table 1 contains a comprehensive list of the HPLC-CAD method validation results. Accuracy was demonstrated by spiking an amikacin standard into simulated formulation sample solutions in triplicate at concentrations of 0.05, 1.0, and 2.0 mg/mL. The observed average recoveries at each spike level were 100, 101, and 100%, respectively. Overall relative standard deviation (Srel ) of the spiked recovery results was 1%. Precision (repeatability) was demonstrated by the analysis of 6 replicate preparations of a sample solution at 1.0 mg/mL. The analysis of these solutions yielded an Srel of 1%. Precision (intermediate precision) was demonstrated by a matrix “Kojima design”, where the same homogenous formulation solution was analyzed in duplicate, following changes in analysts, HPLC columns, and HPLC systems. Each analysis was performed on effectively different “days”, defined as using an HPLC system only after it had been disconnected and reconnected to the chromatography data system. Thus, the six sets of duplicate results were obtained using a unique combination of analyst, HPLC column, HPLC system, and experimental day. Analysis of variance (ANOVA) results are summarized in Table 2; an overall Srel of 1% was achieved and showed no significant statistical difference in the results. Linearity of the method was demonstrated by analyzing six standard solutions serially diluted to concentrations covering a range of 0.05–2.0 mg/mL; the results are reported in Table 3. The regression correlation coefficient was 1.00, and the y-intercept was not significantly different from zero. The limit of quantitation (LOQ) of the method was demonstrated by five replicate injections of a

solution prepared at 0.005 g/mL, one tenth of the reported method limit. The replicate LOQ injections gave a Srel of 4%. Robustness of the chromatographic conditions was demonstrated by intentionally varying the chromatography parameters (i.e., flow rate, column temperature, mobile phase modifier concentration, initial mobile phase composition, end mobile phase composition, CAD evaporator temperature, column lots, HPLC systems) and demonstrating that each intentional variation, within the operating range of the method, did not impact the ability of the method to meet the defined system suitability criteria. As reported in Table 4, the standard and sample solution stability robustness was shown by analyzing the solutions, stored at refrigerated (2 − 8 ◦ C) and room temperature conditions with no protection from ambient light, against freshly prepared amikacin standard solutions. The standard solutions were shown to be stable for at least 7 days, and the sample solution was shown to be stable for at least 6 days. 3.3.1. Precautions for method transfer: implementation of dwell volume To minimize the impact on the chromatographic separation when transferring the method from one instrument to another, the HPLC system dwell volume was included in the gradient program setup. This approach reduces the variability in retention time, peak width, signal-to-noise, and resolution of amikacin from kanamycin. The final gradient conditions also incorporated a 1.0 min isocratic hold of the initial condition. Effectively, this allows the method to be carried out on HPLC systems with 2.65 mL dwell volumes or less, with simple changes detailed below to the gradient program step times. On the rare instances when a system has a dwell volume larger than 2.65 mL, a delayed injection technique where the gradient program is started prior to sample injection should be employed, where the gradient program is modified according to Eq. 2: tmod ified = A −

tG F

(2)

In Eq. 2, tmodified is the time modified to adjust to gradient, tG is the gradient time on the instrument used, F is the flow rate, and A is a constant determined using Eq. (3). A = toriginal +

tG F

(3)

where, toriginal is the original time specified in the gradient method, and tG is the dwell time of the instrument used for method development. 4. Conclusion An assay method for amikacin has been developed and validated, with adequate throughput (7 min/sample, approximately 200 samples/day) to support amikacin aPSD testing. While postcolumn derivatization may have been the method of choice for analyzing amikacin and related aminoglycosides, new detection technologies have emerged and supplanted this technique. Though compendial methods exist to quantitate amikacin by anion exchange separation and electrochemical detection, they lack the linear dynamic range illustrated in the present study. We have shown with careful optimization of method parameters associated with the current generation CAD, a robust method intended for routine use that performs nearly on par with that of the conventional UV absorbance detector has been successfully developed. Moreover, the CAD method is much easier to operate and transfer from one instrument to another than previous CAD-based methods, and advantageously, showed minimal variability in results.

Table 4 Robustness evaluationMethodology. Methodology

Results

Robustness – System suitability test

Change in

Met all system suitability tests

Robustness Stability of reference (RS) and aerosol test solutions (TS)

column flow rate (±0.1 mL/min) column temperature (±2 ◦ C) change of column lot ±1% mobile phase B composition mobile phase modifier concentration (±0.02% v/v pentafluoropropionic acid) 6. Composition (±1% mobile phase B) 7. Evaporative temperature (±2 ◦ C) 8. Change of HPLC system 1. 2. 3. 4. 5.

2–8 ◦ C

Change of assay of amikacin Calculate % relative change in concentration by comparing to the initial (T0) concentrations. Relative change in concentration of reference solutions: ≤3% over specified time RS 0.05 mg/mL standard RS 1.0 mg/mL standard RS 2.0 mg/mL standard Relative change in concentration of aerosol test solutions: TS2: 5% ≤ Level < 10%: ≤5% TS1: Level ≥ 10%: ≤3% TS1 Level corresponds to% concentration of 1.0 mg/mL TS2

room temperature

1 day −2

2 days −1

7 days 0

1 day −3

2 days −1

7 days −2

0

3

0

0

1

1

1

−2

−2

−3

−2

−2

2–8 o C 1 day −1

6 days 1

room temperature 1 day −1

6 days 1

−1

0

2

2

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

Parameter

75

76

A. Soliven et al. / Journal of Pharmaceutical and Biomedical Analysis 143 (2017) 68–76

References [1] M. Abdel-Tawab Korany, R.S. Haggag, M.A. Ragab, O.A. Elmallah, Liquid chromatographic determination of amikacin sulphate after pre-column derivatization, J. Chromatogr. Sci. 52 (2014) 837–847. [2] S. Oguri, T. Fujiyoshi, Y. Miki, In-capillary derivatization with 1-methoxycarbonylindolizine-3,5-dicarbaldehyde for high-performance capillary electrophoresis, Analyst (Cambridge U.K.) 121 (1996) 1683–1688. [3] C.-Z. Yu, Y.-Z. He, G.-N. Fu, H.-Y. Xie, W.-E. Gan, Determination of kanamycin A, amikacin and tobramycin residues in milk by capillary zone electrophoresis with post-column derivatization and laser-induced fluorescence detection, J. Chromatogr. B 877 (2009) 333–338. [4] B. Blanchaert, S. Huang, K. Wach, E. Adams, A. Van Schepdael, Assay development for aminoglycosides by HPLC with direct UV detection, J. Chromatogr. Sci. 55 (2017) 197–204. [5] J.R.M.I.S. Krull, R. Mhatre, M.E. Szulc, J.T. Stults, J.H. Bourell, Detection and identification in biochromatography, in: E.D. Katz (Ed.), High Performance Liquid Chromatography: Principles and Methods in Biotechnology, Wiley, New York, 1996, pp. 186–188. [6] W. Han, F. Qu, J. Shang, W. Lu, J. Yang, Q. Ma, Determination of amikacin in different pharmaceutical formulations using a resonance rayleigh scattering method with pontamine sky blue, Curr. Pharm. Anal. 10 (2014) 105–111. [7] C. Yang, Z. Zhang, J. Wang, New luminol chemiluminescence reaction using diperiodatoargentate as oxidate for the determination of amikacin sulfate, Luminescence 25 (2010) 36–42. [8] X.L. Wang, Z.Y. Yu, K. Jiao, Voltammetric studies on the interaction of amikacin with methyl blue and its analytical application, Chin. Chem. Lett. 18 (2007) 94–96. [9] Q. Liu, H. Mu, C. Sun, J. Duan, Highly specific determination of gentamicin by induced collapse of Au-lipid capsules, RSC Adv. 6 (2016) 14483–14489. [10] E.G. Galanakis, N.C. Megoulas, P. Solich, M.A. Koupparis, Development and validation of a novel LC non-derivatization method for the determination of amikacin in pharmaceuticals based on evaporative light scattering detection, J. Pharm. Biomed. Anal. 40 (2006) 1114–1120. [11] D.A. Bhatt, L.M. Prajapati, A.K. Joshi, M.L. Kharod, Development and validation of spectrophotometry method for simultaneous estimation of cefepime hydrochloride and amikacin sulphate, World J. Pharm. Res. 4 (2015) 1482–1491. [12] H.-H. Yeh, S.-J. Lin, J.-Y. Ko, C.-A. Chou, S.-H. Chen, Rapid and selective micellar electrokinetic chromatography for simultaneous determination of amikacin, kanamycin A, and tobramycin with UV detection and application in drug formulations, Electrophoresis 26 (2005) 947–953. [13] N.H. Zawilla, B. Li, J. Hoogmartens, E. Adams, Improved reversed-phase liquid chromatographic method combined with pulsed electrochemical detection for the analysis of amikacin, J. Pharm. Biomed. Anal. 43 (2007) 168–173. [14] U.S.P. Convention, United States Pharmacopeia and National Formulary, USP, Rockville, MD, 2016.

[15] K. Rick Lung, K.R. Kassal, J.S. Green, P.K. Hovsepian, Catalytic precolumn derivatization of amikacin, J. Pharm. Biomed. Anal. 16 (1998) 905–910. [16] B. Zhang, X. Li, B. Yan, Advances in HPLC detection-towards universal detection, Anal. Bioanal. Chem. 390 (2008) 299–301. [17] C. Crafts, B. Bailey, P. Gamache, X. Liu, I. Acworth, Comprehensive approaches for measurement of active pharmaceutical ingredients, counter-ions, and excipients using HPLC with charged aerosol detection, in: Applications of Ion Chromatography for Pharmaceutical and Biological Products, John Wiley & Sons Inc., 2012, pp. 221–236. [18] L.-E. Magnusson, D.S. Risley, J.A. Koropchak, Aerosol-based detectors for liquid chromatography, J. Chromatogr. A 1421 (2015) 68–81. [19] S. Almeling, D. Ilko, U. Holzgrabe, Charged aerosol detection in pharmaceutical analysis, J. Pharm. Biomed. Anal. 69 (2012) 50–63. [20] N. Vervoort, D. Daemen, G. Toeroek, Performance evaluation of evaporative light scattering detection and charged aerosol detection in reversed phase liquid chromatography, J. Chromatogr. A 1189 (2008) 92–100. [21] P.K. Dasgupta, Y. Chen, C.A. Serrano, G. Guiochon, H. Liu, J.N. Fairchild, R.A. Shalliker, Black box linearization for greater linear dynamic range: the effect of power transforms on the representation of data, Anal. Chem. 82 (2010) 10143–10150. [22] R.A. Shalliker, P.G. Stevenson, D. Shock, M. Mnatsakanyan, P.K. Dasgupta, G. Guiochon, Application of power functions to chromatographic data for the enhancement of signal to noise ratios and separation resolution, J. Chromatogr. A 1217 (2010) 5693–5699. [23] M. Swartz, M. Emanuele, A. Awad, A. Grenier, D. Hartley, An Overview of Corona Charged Aerosol Detection in Pharmaceutical Analysis, 2009. [24] T. Gorecki, F. Lynen, R. Szucs, P. Sandra, Universal response in liquid chromatography using charged aerosol detection, Anal. Chem. 78 (2006) 3186–3192. [25] T. Vehovec, A. Obreza, Review of operating principle and applications of the charged aerosol detector, J. Chromatogr. A 1217 (2010) 1549–1556. [26] J.P. Hutchinson, J. Li, W. Farrell, E. Groeber, R. Szucs, G. Dicinoski, P.R. Haddad, Comparison of the response of four aerosol detectors used with ultra high pressure liquid chromatography, J. Chromatogr. A 1218 (2011) 1646–1655. [27] E.M. Hetrick, T.T. Kramer, D.S. Risley, Evaluation of a hydrophilic interaction liquid chromatography (HILIC) design space for sugars and sugar alcohols, J. Chromatogr. A 1489 (2017) 65–74. [28] Dionex, Determination of Tobramycin and Impurities Using HPAE-PAD, Dionex Application Note Number 61, 2016. [29] I.A.H. Ahmad, Necessary analytical skills and knowledge for identifying, understanding, and performing HPLC troubleshooting, Chromatographia (2017) 1–26. [30] L. Snyder, J. Dolan, P. Carr, The hydrophobic-subtraction model of reversed-phase column selectivity, J. Chromatogr. A 1060 (2004) 77–116.