Determination of Three Alkyl Camphorsulfonates as Potential Genotoxic Impurities Using GC-FID and GC-MS by Analytical QbD

Abstract

Camphorsulfonic acid salts are commonly used in the manufacturing production of active pharmaceutical ingredients (APIs) and have the potential to form alkyl camphorsulfonates, which can be considered as potential genotoxic impurities (PGIs). Alkyl camphorsulfonates should be controlled using the Threshold of Toxicological Concern (TTC) when detected in APIs due to their genotoxicity. An in silico study utilizing the ICH M7 guideline was performed in order to classify the alkyl camphorsulfonates that can be produced from the reaction of camphorsulfonic acid salts with methanol, ethanol, and isopropyl alcohol, which are commonly used solvents in API manufacturing processes. Two sensitive, reproducible, and accurate analytical methods using GC-FID and GC-MS were developed using the analytical Quality By Design (QbD) approaches for the quantitation of three alkyl camphorsulfonates in APIs satisfying the control limit of PGIs according to the TTC. The detection limits of the GC-FID method were found to be between 1.5 to 1.9 ppm, and the detection limits of the GC-MS method were found to be between 0.055 to 0.102 ppm. The method was validated in terms of accuracy, linearity, precision, detection limit, quantitation limit, specificity and robustness.

1. Introduction

The formation of salts with various pharmaceutically available counter ions has been widely used for the advantages of physicochemical properties including solubility, formulation, purification, stability and other various reasons [1,2]. Approximately 50% of APIs include formed salts that can be used pharmacologically, and thus salt development is an integral part of the process development of APIs [3,4,5]. Various types of cation and anion salts have been used, and sulfonic acid salts, such as methanesulfonic acid, camphorsulfonic acid, toluenesulfonic acid, and benzenesulfonic acid, are the most commonly used anion salts [6]. The use of these sulfonic acid salts could potentially lead to the presence of small amounts of sulfonic acid esters as impurities in the final API. Alkyl sulfonates can mainly be generated by the esterification reaction of sulfonic acid with residual alcohols used in the manufacturing process of APIs such as methanol, ethanol and isopropanol. These alkyl sulfonates are known as alkylating agents and can induce genetic mutations or chromosomal aberrations [7,8,9,10,11,12]. Numerous studies have been conducted on the toxicity and analytical methods for these alkyl sulfonates as potential genotoxic impurities (PGIs).

Alkyl sulfonates should be strictly controlled with tight acceptance criteria if produced or introduced in the production of APIs due to their potentially genotoxic behavior. According to the ICH M7 guideline regarding mutagenic impurities, PGIs should be controlled below the Threshold of Toxicological Concern (TTC) in the absence of appropriate limits based on toxicity data. The limits of PGIs in APIs derived from the TTC can be calculated based on the expected maximum daily dose. The TTC value for PGIs in APIs is set at 1.5 µg/day, corresponding to a 10-5 lifetime risk of cancer [13,14].

There have been numerous studies throughout many years to develop analytical methods for the control of PGIs. Most studies have used chromatographic method such as liquid chromatography (LC), gas chromatography (GC) and ion chromatography (IC) with various detectors such as the evaporative light scattering (ELSD), charged aerosol (CAD), mass spectrometry (MS), flame ionization (FID) and UV detectors depending on the target analyte’s chemical properties. [15,16,17,18,19,20]

Since more than 5% of all APIs are prepared in the presence of sulfonic acid, many studies on sulfonic acid alkyl esters have been conducted [21,22,23,24,25]. However, these studies have only been performed for alkyl methanesulfonate, alkyl benzenesulfonate, alkyl toluenesulfonate and alkyl isethionate. Camphorsulfonic acid (CSA) is the second most used sulfonic acid salt after methanesulfonic acid, but not much research has been done on its possibly formed alkyl sulfonate and alkyl camphorsulfonate. Alkyl camphorsulfonates are potential genotoxic impurities and present structurally alerting properties like other sulfonic acid alkyl esters.

Similarly to other alkyl sulfonates, these three alkyl camphorsulfonates chosen can behave as alkylation reagents and have a possibility of alkylating DNA bases (on N-7 of guanosine and N-3 of adenine). These substances have the possibility of causing genetic mutations and present carcinogenic behaviors at even small exposure levels, and therefore must be accurately quantitated and controlled below the TTC.

In order to control these alkyl camphorsulfonates as PGIs, it is essential to develop a highly sensitive, reproducible and stable analysis method. Analytical Quality by Design (QbD) is a systematic approach to developing an optimal analytical method by applying process QbD concepts to method development, promoting method flexibility, and increasing scientific understanding of critical method parameters (CMPs) [26,27,28,29,30,31,32,33,34]. Many studies in the literature have introduced the concept of analytical QbD to develop analytical methods, but there are few studies that use analytical QbD to develop high-sensitivity GC methods to control trace amount of PGIs.

An in silico study using the quantitative structure activity relationship (Q-SAR) approach was conducted for methyl camphorsulfonate (MCS), ethyl camphorsulfonate (ECS) and isopropyl camphorsulfonate (ICS), which can mostly be produced from the reaction of camphorsulfonic acid and alcohols in API manufacturing processes [35,36]. The term “in silico” refers to the computational models that study pharmacological hypotheses using databases. The use of an in silico study in the prediction of toxicity is very useful in order to speed the rate of discovery while reducing cost and time using computer simulations.

An analytical method using GC with the FID was developed for these alkyl camphorsulfonates as PGIs. The analytical QbD concept was adopted for the method development for PGI analysis that requires high sensitivity and reproducibility to meet regulatory requirements. Since the most important point in developing the PGIs analysis method is to have high sensitivity to detect trace amounts of PGIs, the height of the each PGI peak was selected as the critical method attributes (CMAs). Based on the optimized GC-FID method by analytical QbD, an additional method using the MS was developed to obtain better detection sensitivity. Both methods using GC-FID and GC-MS were validated through analytical method validation according to the ICH Q2 (R2) guideline regarding the validation of analytical procedures in terms of linearity, accuracy, robustness, specificity, precision and system suitability [37]. The potential synthetic route and structures of MCS, ECS and ICS are shown in Figure 1.

2. Materials and Methods

2.1. Reagents, Materials and Standards

The reference standards for (1S)-(+)-10-camphorsulfonic acid methyl ester (MCS, Cas No. 62319-13-5, purity: 99.9%), (1S)-(+)-10-camphorsulfonic acid ethyl ester (ECS, Cas No. 154335-57-6, purity: 94.5%) and (1S)-(+)-10-camphorsulfonic acid isopropyl ester (ICS, Cas No. 247078-58-6, purity: 98.1%) were purchased from Toronto Research Chemicals (Toronto, ON, Canada). All of the reference standards were characterized by NMR, HPLC and MS. Phenyl ether was purchased from Acros Organics (Geel, Belgium). Methanol, ethanol, isopropanol and ethyl acetate were purchased from Merck (Darmstadt, Germany). Acetonitrile was purchased from Fisher Scientific (Pittsburg, PA, USA). (1S)-(+)-10-camphorsulfonic acid was purchased from TCI (CSA, Tokyo, Japan). The Amlodipine camsylate, API (CAS No.: 652969-01-2) bulk drug was obtained from Hanmifine Chemical (Shihung, South Korea).

2.2. Analytical Condition and Equipments

2.2.1. GC-FID Condition

The analyses and validation of MCS, ECS and ICS were conducted using an Agilent 7890 gas chromatograph equipped with an FID (Agilent Technologies, Santa Clara, CA, USA). Sample injection was performed using an Agilent 7693 auto-sampler (Agilent Technologies, Santa Clara, CA, USA). Mass Hunter Data Analysis Version. B.07.04.2260 was used as the data acquisition software from Agilent Technologies.

The analyses of MCS, ECS, ICS with the FID were performed using an HP-1 capillary column (25 m × 0.32 mm i.d. × 0.52 μm film thickness) from Agilent Technologies. The column oven temperature was set at the following temperature program: initial oven temperature set at 200 °C, held for 8 min, then ramped at 70 °C/min to 260 °C, held for 8 min. The flow rate of the helium carrier gas was set at 2.0 mL/min with the constant flow mode and total flow was set at 25 mL/min. The Split Inlet was used at a split ratio set at 10:1 and split flow was set at 20 mL/min. For injection, a 4 mm straight liner Inlet was used. The temperature of injection heater was set at 280 °C, septum purge flow was set at 3.0 mL/min, the pressure was 13.7 psi and the average velocity was 43.5 cm/s. The detector heater temperature was set at 300 °C, air flow was set at 400 mL/min, H₂ fuel flow was set at 30 mL/min and the make-up flow was set at 25 mL/min. Helium was used as the make-up gas. Injection volume was set at 2.0 μL using a 5 μL auto liquid sampler syringe. The syringe needle was rinsed with ethyl acetate between injections. Total run time was 16.9 min with no solvent delay time. Data acquisition rate was 50 Hz and the minimum peak width was set at 0.004 min.

2.2.2. GC-MS Condition

The analyses and validation of MCS, ECS, and ICS were conducted using an Agilent 7890 gas chromatograph equipped with a 5977A Single Quadrupole mass detector (Agilent Technologies, Santa Clara, CA, USA). Sample injection was performed using an Agilent 7693 auto-sampler (Agilent Technologies, Santa Clara, CA, USA). Mass Hunter qualitative analysis version B.07.00 from Agilent Technologies was used for the control of equipment, processing and data acquisition software.

The GC conditions are the same as the GC-FID conditions. The MS was operated using the high efficiency EI as the type of source in Selective Ion Monitoring mode (SIM). The ions with m/z = 81.0, 109.1, 151,1, 170.1 with 50 ms dwell time were selected for analysis. The temperature of source was set at 230 °C, the temperature of the quadrupole was set at 150 °C and the temperature of transfer line was set at 280 °C. The gain factor was set at 1.0. Injection volume was set at 2.0 μL using a 5 μL auto liquid sampler syringe. The syringe needle was rinsed with ethyl acetate between injections. Total run time was 16.9 min with a solvent delay time of 0.8 min.

2.3. Preparation of Standard Solution and Sample Solution

To enhance the precision of measurements, phenyl ether was used as internal standard.

2.3.1. Internal Standard Solution Preparation

A quantity of 0.1 g of phenyl ether was weighed into a 100 mL volumetric flask, then sufficiently mixed, dissolved and diluted to the volume mark with ethyl acetate. Then, 5.0 mL of this solution was added to a 100 mL volumetric flask, then sufficiently mixed and diluted to the volume mark with ethyl acetate.

2.3.2. Standard Stock Solution Preparation (STD-2)

Totals of 0.1 g of the MCS reference standard, 0.1 g of the ECS reference standard and 0.1 g of the ICS reference standard were weighed to a 100 mL volumetric flask, then sufficiently mixed, dissolved and diluted to the volume mark with ethyl acetate (STD-1). 6.0 mL of STD-1 solution was added to a 100 mL volumetric flask, then sufficiently mixed and diluted to the volume mark with ethyl acetate (STD-2).

2.3.3. Standard Solution Preparation

Quantities of 5.0 mL of the STD-2 solution and 5.0 mL of the internal standard solution were added to a 50 mL volumetric flask, then sufficiently mixed and diluted to the volume mark with ethyl acetate. This solution was sonicated for 1 min and filtered through a 0.45 µm membrane filter. The concentration of the standard solution is 150 ppm for each of the target analytes of MCS, ECS, and ICS.

2.3.4. Sample Solution Separation

Approximately 2.0 g of the test specimen and 5.0 mL of the internal standard solution were weighed and added to a 50 mL volumetric flask and then sufficiently mixed, dissolved and diluted to the volume mark with ethyl acetate. This solution was sonicated for 1 min and filtered through a 0.45 µm membrane filter (concentration: 4 mg/mL).

2.4. In Silico Study

According to the ICH guidelines, the toxicity assessments of bacterial mutation assay can be computationally assessed using two correlative rules-based and statistics-based Q-SAR methodologies. The rules-based software used was Derek Nexus (version 6.2.0, built on 11 March 2022) from Lhasa Limited (Leeds, UK), and the statistics-based software used was Sarah Nexus (v.3.2.0 built on 11 March 2022) purchased from Lhasa Limited [38]. Both softwares have been validated according to the guidelines set by the Organization for Economic Cooperation and Development (OECD). For additional analysis of the results, the free software VEGA (v.1.1.5, Vegahub, Milan, Italy), country was used [39]. The in silico study was conducted for the chemical structures of MCS, ECS, ICS as well as CSA.

2.5. Method Screening

Fusion QbD Professional 9.8.1.135 (S-matrix, Eureka, CA, USA) was used to create the Design of Experiment (DoE) for the method screening and optimization study [40]. The major parameters which can affect the analysis results were screened. The screening of various types of columns was performed using the one-factor-at-a-time (OFAT) method. Then the column oven temperature, inlet temperature, flow rate of carrier gas, detector temperature and injection volume were screened to understand the interactions of each analytical parameters using the DoE and analytical QbD concepts. The DoE was conducted in two-level-resolution V type. Three center points were included in the model.

2.6. Method Optimization

The developed analytical methods were optimized using the DoE for the selected CMPs using the results of the screening study. The column oven temperature, flow rate of carrier gas and injector temperature were determined as the CMPs. The DoE was created using a response surface-central composite including three center points. These results were used to determine the method operable design region (MODR) and the proven acceptable range (PAR), which allows for flexibility in the analytical method.

2.7. Method Validation

The ICH Q2 guideline was used for the validation of the developed methods. The method validation consisted of the parameters of specificity, detection limit, quantitation limit, accuracy, precision, linearity, and robustness.

2.7.1. Determination of Specificity

Specificity was evaluated by determining the resolution between the MCS, ECS and ICS peaks from the chromatograms. The standard solution prepared was sampled and analyzed in 6 replicates. Using the chromatogram, the resolutions between MCS, ECS and ICS peaks were calculated automatically using the computerized data system for peak integration.

2.7.2. Determination of the Detection Limit and Quantitation Limit

The detection limit (DL) and quantitation limit (QL) were determined by calculating the signal-to-noise ratio obtained from the comparison of each signal (height) of analytes and noise in the given time range closest to the analyte. The signal-to-noise ratios at approximately 3:1 and 10:1 were considered acceptable for estimating the DL and QL, respectively.

2.7.3. Determination of Accuracy

Accuracy was evaluated using the recovery study of spiked amounts of MCS, ECS and ICS. Each of the analyzed accuracy solutions were prepared in triplicates at three levels of the specifications with known amounts of MCS, ECS, and ICS spiked to a single batch of the matrix sample. The recoveries of the accuracy solutions of MCS, ECS, and ICS were then calculated.

2.7.4. Determination of Precision

Precision was evaluated to prove the analytical method’s ability to obtain close data results between a series of measurements from several injections of the standard solution over a short time interval. Precision was evaluated based on the repeatability precision of the results.

2.7.5. Determination of Linearity

Linearity was evaluated by assessing the linear proportionality of the concentration of the analytes to the signals obtained over the specific range from the reporting level of genotoxic impurities (the QL concentration) to 120% of the specifications. The results were evaluated using the slope, y-intercept, and correlation coefficient of the linear regression obtained from the linearity solutions at five concentration levels within the specified range.

2.7.6. Determination of Robustness

The robustness of the proposed analytical methods was evaluated by determining the influences on the analysis results when varying method parameters such as the column oven temperature and the flow rate of the carrier gas.

3. Results

3.1. In Silico Study

Alkyl sulfonates such as alkyl methanesulfonates, alkyl toluenesulfonates and alkyl benzenesulfonates are chemicals widely known to be genotoxic. However, since the genotoxicity of alkyl camphorsulfonates are not well known, an in silico study was performed according to the ICH M7 guideline for their classification. Derek Nexus, an expert knowledge-based software, and Sarah Nexus, a statistical-based software, were used along with the VEGA Q-SAR model to correspondingly evaluate the genotoxicity of the target analytes.

CSA was found to be negative in both Derek and Sarah prediction results. MCS, ECS and ICS were all found to be positive and have structural alerts and were classified as class 3. The VEGA Q-SAR model also showed similar results. The genotoxicity predictions for the alkyl camphorsulfonates are shown in

Table 1. The genotoxicity predictions for alkyl camphorsulfonates.

Name	Derek Prediction	Sarah Prediction	Vega Prediction	Overall In Silico *	ICH M7 Class *
CSA	Negative	Negative	Negative	Negative	Class 5
MCS	Positive	Negative	Negative	Positive	Class 3
ECS	Positive	Equivocal	Positive	Positive	Class 3
ICS	Positive	Positive	Positive	Positive	Class 3

* Overall in silico result and ICH M7 class were automatically calculated by Lhasa Nexus program.

. The detailed prediction results are summarized in Table S1 of the Supporting Document.

3.2. Method Screening

The type of column used is a very important factor due to the various effects the column has on the CMAs such as resolution, peak shape, reproducibility and sensitivity. Capillary columns were chosen due to their separation ability and convenience of use, and various types of capillary columns were screened using the one-factor-at-a-time (OFAT) method. The WAX column (100% polyethylene glycol), FFAP column (acid modified polyethylene glycol), CAM (base deactivated polyethylene glycol) were excluded from the column screening process due to their low analysis temperature range. DB-624 (6% cyanopropylphenyl/94% dimethylpolysiloxane) and HP-1 (100% dimethylpolysiloxane) were both tested.

Ethyl acetate was used as the diluent in order obtain sufficient resolution between the solvent peak and the target analytes. All of the target analytes have high boiling points and, therefore, a relatively faster eluting solvent was chosen rather than a slowly eluting solvent. The reasons for choosing ethyl acetate were due to the fact that all three analytes chosen for this study presented sufficient solubility in this solvent, the solvent being a GC-MS compatible, and its faster elution time before the target analytes.

The peak height of each analyte and the resolution between adjacent peaks were selected as the critical method attributes (CMAs) since the most critical analytical target profiles (ATPs) in the analysis of PGIs is detection sensitivity and specificity due to their low specification limits. A risk assessment was performed in order to determine the critical method parameters (CMPs) expected to have the greatest impacts on the CMAs. Through the risk assessment, the flow rate of carrier gas, inlet temperature, detector temperature, column oven temperature and sample injection volume were chosen as the CMPs.

The Fusion QbD software was used to create an experimental plan using a two-level-resolution V type including three center points. A total of 19 experiments were performed. The Design of Experiments (DoE) is tabulated in

Table 2. Design of Experiment for screening.

Run	Flow Rate (mL/min)	Oven Temp (°C)	Inlet Temp (°C)	Detector Temp (°C)	Inject Vol (µL)
1	1.5	170	340	330	2.2
2	1.5	230	340	330	1.8
3	1.5	170	180	330	1.8
4	2.5	170	340	270	2.2
5	2.5	170	180	330	2.2
6	2.5	170	180	270	1.8
7	2.5	230	180	330	1.8
8	2.5	230	340	270	1.8
9	2.0	200	280	300	2.0
10	2.0	200	280	300	2.0
11	2.5	230	340	330	2.2
12	1.5	230	340	270	2.2
13	2.5	230	180	270	2.2
14	1.5	230	180	270	1.8
15	1.5	230	180	330	2.2
16	2.0	200	280	300	2.0
17	1.5	170	340	270	1.8
18	1.5	170	180	270	2.2
19	2.5	170	340	330	1.8

and the experimental results of each CMA are tabulated in

Table 3. Response corresponding DoE.

Run	R1	R2	R3	R4	R5
1	2.06	1.75	0.46	14.34	6.74
2	4.15	4.26	0.59	4.28	1.98
3	1.57	1.32	1.14	14.08	6.61
4	2.75	2.25	0.83	13.46	6.19
5	2.72	2.19	2.00	12.92	5.95
6	2.32	1.89	1.68	13.31	6.21
7	5.28	5.39	3.57	3.93	1.97
8	5.43	5.67	1.37	4.05	1.94
9	4.20	3.93	3.09	8.08	3.94
10	4.23	3.96	3.14	8.15	3.97
11	6.39	6.60	1.64	3.91	1.87
12	4.48	4.67	0.65	4.03	1.84
13	5.89	5.90	3.85	3.73	1.87
14	3.60	3.72	2.01	4.05	2.02
15	4.45	4.60	2.47	4.04	2.02
16	4.27	4.00	3.12	8.28	4.03
17	1.68	1.40	0.38	14.61	6.85
18	1.92	1.65	1.42	14.04	6.65
19	2.28	1.84	0.72	13.56	6.20

R1 = peak height of MCS, R2 = peak height of ECS, R3 = peak height of ICS, R4 = resolution between MCS and ECS, R5 = resolution between ECS and ICS.

. The Analysis of Variance (ANOVA) statistical test was performed in order to determine the statistical significance of each response, shown in

Table 4. Statistical significance for each response.

		R2	Adj. R2	Sum of Squares	Degree of Freedom	Mean Square	F-Ratio	p-Value
MCS	Height	0.9995	0.9987	38.8914	10	3.8891	1334	<0.0001
ECS	Height	0.9998	0.9995	51.1069	10	5.1107	3658	<0.0001
	Resolution	0.9999	0.9998	385.4958	7	55.0708	15098	<0.0001
ICS	Height	0.9999	0.9997	14.8934	12	1.2411	3349	<0.0001
	Resolution	0.9997	0.9995	81.2602	6	13.5434	5475.8411	<0.0001

. The p-values of the responses for each of the PGIs were found to be below 0.05 and, thus, were proven to have statistical significance.

The model term ranking Pareto charts represent the effects and interactions of each of the parameters based on the CMAs established by the model term effect and the cumulative percentage. The model term ranking Pareto charts are shown in Figure 2. The Pareto charts show that the column oven temperature and/or the flow rate had the greatest effect on the CMAs, such as the peak height of each of the target analytes, the resolution between each of the target analytes and the retention time of ICS. The selected responses were shown to not be affected or partially affected by the detector temperature and injection volume. As a result, the column oven temperature, flow rate of the carrier gas and inlet temperature were selected as the CMPs for the method optimization study.

3.3. Method Optimization by Analytical QbD

The analytical method evaluated through the screening process was optimized using the analytical QbD and DoE approaches. Optimization designs are typically used to define the experiment variable level settings used to predict the best overall performance, and the experiment variable ranges used to predict the acceptable overall mean performance and acceptable robustness. Similarly to the screening process, the CMAs were chosen to be the peak height of each of the target analytes, resolution between each of the target analytes, and the retention time of ICS. The CMPs were chosen based on the screening results to be the column oven temperature, flow rate of the carrier gas and inlet temperature, which were shown to have the greatest effects on the CMAs.

The experiments were designed using a response surface-central composite full type including three center points by a DoE using the Fusion QbD software. The Design of Experiments (DoE) is tabulated in

Table 5. Design of Experiment for Optimization.

Run	Flow Rate (mL/min)	Oven Temp (°C)	Inlet Temp (°C)
1	2	220	280
2	2	200	300
3	2.2	220	300
4	1.8	180	300
5	1.8	200	280
6	2.2	180	260
7	2	200	260
8	2	180	280
9	1.8	220	300
10	2	200	280
11	2	200	280
12	2	200	280
13	2.2	180	300
14	1.8	220	260
15	2.2	200	280
16	2.2	220	260
17	1.8	180	260

and the experimental results of each CMA are tabulated in

Table 6. Response Corresponding DoE.

Run	R1	R2	R3	R4	R5	R6
1	3.22	4.11	2.68	4.91	2.43	2.95
2	2.74	3.35	2.23	8.21	3.96	4.60
3	3.53	4.58	2.67	5.16	2.53	2.75
4	1.75	1.98	3.19	12.32	4.92	8.46
5	2.46	3.04	2.06	8.37	4.04	4.96
6	1.99	2.24	1.68	11.81	5.50	7.58
7	2.69	3.31	2.43	8.47	4.09	4.61
8	1.84	2.12	2.04	12.09	6.36	8.10
9	3.09	4.03	2.07	5.46	2.65	3.17
10	2.60	3.21	2.15	8.27	4.00	4.60
11	2.75	3.34	2.25	8.60	4.11	4.60
12	2.69	3.28	2.29	8.36	4.10	4.60
13	1.98	2.25	1.46	12.01	5.54	7.57
14	2.90	3.74	2.18	5.22	2.58	3.17
15	2.71	3.24	2.38	8.05	3.86	4.03
16	3.73	4.70	3.23	5.55	2.72	2.76
17	1.74	1.97	3.94	12.23	4.84	8.46

R1 = peak height of MCS, R2 = peak height of ECS, R3 = peak height of ICS, R4 = resolution between MCS and ECS, R5 = resolution between ECS and ICS, R6 = retention time of ICS.

. The models created were statistically validated using general regression statistics and the ANOVA statistical test. As shown in

Table 7. Statistical significance for each response.

	R2	Adj. R2	Sum of Squares	Degree of Freedom	Mean Square	F-Ratio	p-Value
R1	0.9954	0.9908	5.7025	8	0.7128	215	<0.0001
R2	0.9934	0.9868	11.9727	8	1.4966	150	<0.0001
R3	0.9923	0.9824	5.7238	9	0.636	100	<0.0001
R4	0.9943	0.9935	117.0834	2	58.5417	1,224	<0.0001
R5	0.9971	0.9948	23.4002	7	3.3429	434	<0.0001
R6	1.0000	1.0000	9076.33	8	1,134.54	672,190	<0.0001

R1 = peak height of MCS, R2 = peak height of ECS, R3 = peak height of ICS, R4 = resolution between MCS and ECS, R5 = resolution between ECS and ICS, R6 = retention time of ICS.

, the p-values for each of the responses are less than 0.0001, which is substantially smaller than the acceptance criteria of 0.05.

The response surface methodology (RSM) analysis was reviewed using 3D response surface plots to identify underlying interactions between selected parameters (Figure 3). The inlet temperature, selected as one of the CMPs, was shown to have no significant effect on the response and tendency of the CMA in all of the results. The 3D response surface plots at 260 °C and 300 °C are summarized in Figures S1 and S2 of the Supporting Document, respectively.

The flow rate was shown to have little effect on (a) the peak height of MCS and (b) the peak height of ECS, and the respective peak heights showed a decreasing linear trend with a decrease in column oven temperature. The response surface plots of the peak height of ICS (c) showed a curved relationship between the flow rate and column oven temperature. Both CMPs demonstrated a more varied interaction with the change in peak height. The DoE was shown to have a maximum point when the column oven temperature and flow rate is low. The 3D response surface plot for the peak height of ICS (c) was shown to have a saddle shape with the minimum point of the peak height being when the column oven temperature was low and the flow rate was high, or when the column oven temperature was high and the flow rate was low. The resolution between MCS and ECS (d) was shown to have little effect on the flow rate and tended to increase linearly with a decrease in column oven temperature. The resolution of ECS and ICS (e) showed an increase as the oven temperature decreased and had a maximum point when the flow rate was around 2.0 mL/min and decreased at both ends. The retention time of ICS (f) showed a tendency to decrease with an increase in column oven temperature rather than the effect of flow rate. Overall, the flow rate was shown to affect the established CMA only to a certain extent, but the column oven temperature was shown to have a significant effect. The method operable design region (MODR) and proven acceptable ranges (PARs) are shown in Figure 4.

3.4. Design Space

The optimal region was set by superimposing all responses of interest. The acceptance criteria for each of the responses were set to satisfy the purpose of the ATP. The acceptance criteria of all peak heights were set to 2 or greater, in order to ensure sufficient sensitivity in order to detect trace amounts of each of the PGIs. To ensure that each peak is sufficiently separated, and inter-peak interference is minimized, the acceptance criteria for the degree of separation were set at 7 or greater for MCS and ECS and 3 or greater for ECS and ICS. The retention time of the last eluted peak, ICS, was set to 7 min or less in order to shorten the analysis time. The shaded contour areas in Figure 4 indicate the unacceptable area and the unshaded areas indicate acceptable areas. The unshaded area depicts the parameter values that can be represented as a design space with analytical method flexibility. The proven acceptable range (PAR) was defined up to the set center point of the parameter range within a predefined MODR based on a linear combination of parameters. The resolution between ECS and ICS and the peak height of ICS had the most influence on establishing the MODR and PAR.

3.5. Apply MS

Based on developed GC-FID analytical method using analytical QbD, a GC-MS analytical method was developed and compared by introducing the mass-selective detector (MS) instead of the flame ionization detector (FID) under the same analytical conditions. The specifications of the PGIs of interest were set using the concept of TTC and sufficient sensitivity was obtained depending on the target API’s maximum daily dose. Generally, the MS operated under SIM mode presents a significantly higher sensitivity as opposed to the FID, and therefore, the MS was chosen as the detector to be used alongside the GC.

A total of four ions were selected through a standard solution analysis using GC-MS operated under scan mode for analyzing all esters and internal standards simultaneously. The selected ions had m/z values of 81.0, 109.1, 151,1, and 170.1. The fragments at m/z values of 81.0, 109.1 and 151.1 were identified as MCS, ECS and ICS. These m/z values were considered the most suitable for the detection of MCS, ECS and ICS. The fragment at the m/z value of 170.1 was identified as the internal standard (diphenyl ether) and the fragment at the m/z value of 170.1 was not found in any of the signals of MCS, ECS and ICS. The same split ratio was used for both the MS and FID methods in order to set the method parameters similarly for both methods. Sufficient sensitivity was obtained for the MS method with a larger split ratio. However, unlike typical impurities, genotoxic impurities require the most optimal sensitivity due to their low acceptance criteria and therefore, the split ratio was not altered. There were also no carry-over problems observed even when using a small split ratio with the MS. As shown in Figure 5 and Figure 6, the results of the GC-FID analytical method (Figure 5) were almost the same as the results obtained from the GC-MS analytical method (Figure 6). As a result of the change in detector from the FID to MS, the detection limits (DL) for each of the target analytes were found to be between 1.5 to 1.9 ppm initially and improved to be between 0.055 to 0.102 ppm. The quantitation limits (QL) for each of the target analytes were found to be between 4.9 to 6.4 ppm initially and improved to be between 0.185 to 0.340 ppm. Thus, the DL and QL showed approximately 18 to 27 times improved limits with a change in the detector.

Both the GC-FID and GC-MS analytical methods demonstrate sufficient detection limits capable of controlling the specifications for the PGIs of amlodipine (daily dose = 10 mg) set at less than 150 ppm, calculated using the TTC and ICH M7 guideline. For APIs, including amlodipine, with PGI specifications set at greater than 10 ppm or with a maximum daily dose set at 150 mg or lower, using the FID instead of the MS is a much more cost-effective analytical method with sufficient sensitivity.

3.6. Applicability of the Method to Real Sample

In order to verify that the developed analytical method for the target PGIs is applicable to APIs, the API amlodipine camsylate (Cas No. 652969-01-2) that uses CSA as a salt, was selected as a matrix API. Amlodipine camsylate was chosen as the API of interest because alcohols including methanol and ethanol are used in the manufacturing process and thus, has potential to have residual alkyl camphorsulfonates in the final product. The matrix API amlodipine camsylate was provided by Hanmi Fine Chemical located in South Korea.

Amlodipine camsylate was also chosen as the matrix API as there was no residual alkyl camphorsulfonates detected in the sample when analyzed using the two analytical methods using GC-FID and GC-MS developed by QbD, and thus, the experiment was carried out by spiking the matrix sample with the specified limits of each of the three alkyl camphorsulfonates calculated based on the TTC. Each of the experiments were carried out three times, and the recovery and repeatability (precision) were determined for each of the results. Recovery and repeatability are two representative parameters in the ICH Q2 guideline regarding the validation of analytical procedures that identifies whether the sample API has an effect on the analytical method. Recovery was found to be between 99.9 to 103.2% when using the FID, and between 89.4 to 99.8% when using the MS. Precision (%RSD) was found to be between 0.64 to 2.11% when using the FID, and 0.83 to 2.47% when using the MS. No issues or interferences were shown when applying the analytical method developed to the API amlodipine camsylate, with accuracy and precision being confirmed for all of the alkyl camphorsulfonates. Therefore, applicability of the method to real samples was confirmed for the GC-FID and GC-MS analytical methods. All results are presented in

Table 8. Applicability of the method on the API Amlodipine camsylate.

		Present Amount in API (P)	Theoretical Spiked Amount (T)	Actual Detection Amount (A)	Recovery(%) ((A − P) /T × 100)	Precision (%RSD)	Effect of API
FID	MCS	0 ppm	120 ppm	120.5 ppm	100.4%
				122.1 ppm	101.8%	0.76%	None
				122.1 ppm	101.8%
		0 ppm	150 ppm	152.4 ppm	101.6%	2.11%	None
				152.5 ppm	101.7%
				146.9 ppm	97.9%
		0 ppm	180 ppm	179.4 ppm	99.7%
				179.7 ppm	99.8%	0.73%	None
				177.3 ppm	98.5%
	ECS			123.5 ppm	102.9%
		0 ppm	120 ppm	118.6 ppm	98.8%	2.06%	None
				120.4 ppm	100.3%
		0 ppm	150 ppm	150.5 ppm	100.4%	2.10%	None
				152.6 ppm	101.7%
				146.4 ppm	97.6%
				177.9 ppm	98.9%
		0 ppm	180 ppm	179.9 ppm	99.9%	1.17%	None
				175.7 ppm	97.6%
	ICS			122.6 ppm	102.2%
		0 ppm	120 ppm	121.6 ppm	101.4%	1.11%	None
				120.0 ppm	100.0%
		0 ppm	150 ppm	155.7 ppm	103.8%	0.64%	None
				153.7 ppm	102.5%
				154.8 ppm	103.2%
				186.9 ppm	103.8%
		0 ppm	180 ppm	18.3 ppm	101.8%	0.96%	None
				185.3 ppm	102.9%
MS				114.5 ppm	95.4%
		0 ppm	120 ppm	110.1 ppm	91.8%	4.06%	None
				105.5 ppm	87.9%
	MCS	0 ppm	150 ppm	135.8 ppm	90.5%	0.83%	None
				134.8 ppm	89.9%
				133.6 ppm	89.1%
				168.2 ppm	93.5%
		0 ppm	180 ppm	167.0 ppm	92.8%	0.35%	None
				167.6 ppm	93.1%
				117.0 ppm	97.5%
		0 ppm	120 ppm	111.0 ppm	92.5%	4.94%	None
				106.0 ppm	88.3%
	ECS	0 ppm	150 ppm	135.4 ppm	90.3%	0.93%	None
				133.9 ppm	89.3%
				132.9 ppm	88.6%
				167.2 ppm	92.9%
		0 ppm	180 ppm	166.0 ppm	92.2%	0.36%	None
				166.3 ppm	92.4%
				139.9 ppm	116.6%
		0 ppm	120 ppm	132.6 ppm	110.5%	5.79%	None
				124.6 ppm	103.8%
	ICS	0 ppm	150 ppm	153.3 ppm	102.2%	2.47%	None
				149.7 ppm	99.8%
				145.9 ppm	97.3%
				178.2 ppm	99.0%
		0 ppm	180 ppm	174.8 ppm	97.1%	1.35%	None
				173.6 ppm	96.5%

.

3.7. Analytical Method Validation and Robustness Test

The analytical methods developed using GC-FID and GC-MS were validated for the purpose of analyzing the contents of PGIs according to the ICH guideline Q2 Validation of analytical procedures. The analytical methods were evaluated based on the parameters of the detection limit, quantitation limit, linearity, precision, accuracy and specificity. A robustness test was conducted to determine the ability of the analytical method to remain unaffected by small, deliberate variations in method parameters. The matrix sample amlodipine camsylate was used for the validation study. The analytical method validation and robustness test results are summarized in

Table 9. Validation results summary.

Validation Parameters		FID			MS
		MCS	ECS	ICS	MCS	ECS	ICS
DL QL	DL	1.5 ppm	1.5 ppm	1.9 ppm	0.055 ppm	0.069 ppm	0.102 ppm
	QL	4.9 ppm	5.1 ppm	6.4 ppm	0.185 ppm	0.232 ppm	0.340 ppm
Linearity	R2	0.99953	0.99988	0.99983	0.99532	0.99412	0.99370
	Slope	0.00501	0.00544	0.00494	0.00534	0.00547	0.00391
	Y-intercept	0.01200	0.00192	0.01838	0.07947	0.09033	0.06289
Accuracy	Low conc. (10 ppm)	95.3%	102.6%	96.4%	94.5%	94.9%	93.9%
	Mid conc. (120 ppm)	101.3%	100.7%	101.2%	91.7%	92.8%	110.3%
	Mid conc. (150 ppm)	100.4%	99.9%	103.2%	89.8%	89.4%	99.8%
	High conc. (180 ppm)	99.3%	98.8%	102.9%	93.1%	92.5%	97.5%
Precision Acceptance criteria (%RSD ≤ 10)	Low conc. (10 ppm)	1.14%	2.50%	5.58%	3.16%	4.92%	1.73%
	Low conc. (30 ppm)	7.79%	6.97%	6.64%	3.14%	3.36%	2.49%
	Mid conc. (150 ppm)	2.27%	3.22%	8.83%	2.16%	2.21%	2.47%
Specificity	Resolution		9.51	4.53		10.21	4.78

and

Table 10. Robustness of the analytical methods.

Condition	Parameter	FID			MS
		MCS	ECS	ICS	MCS	ECS	ICS
1	Precision (%RSD)	2.27%	3.22%	8.83%	2.16%	2.21%	2.47%
	Precision (SDV)	0.01558	0.02304	0.05539	0.01652	0.01706	0.01348
	Resolution		9.51	4.53		10.21	4.78
	Tailing factor	0.90	0.91	0.99	0.97	1.00	0.95
2	Precision (%RSD)	2.07%	1.20%	2.73%	0.97%	1.16%	1.40%
	Precision (SDV)	0.01410	0.00855	0.01604	0.00733	0.00893	0.00804
	Resolution		9.07	4.43		9.97	4.66
	Tailing factor	0.88	0.90	1.00	0.96	0.97	0.97
3	Precision (%RSD)	1.84%	1.63%	9.06%	3.03%	3.49%	3.83%
	Precision (SDV)	0.01249	0.01153	0.05802	0.02195	0.02570	0.02168
	Resolution		9.12	4.33		9.59	4.49
	Tailing factor	0.87	0.94	1.02	0.97	0.98	0.99
4	Precision (%RSD)	2.27%	1.28%	1.29%	2.16%	2.70%	2.38%
	Precision (SDV)	0.01526	0.00909	0.00827	0.01635	0.02080	0.01496
	Resolution		10.03	4.68		10.56	4.93
	Tailing factor	0.90	0.93	1.04	1.02	1.00	0.94
5	Precision (%RSD)	1.24%	0.96%	1.68%	2.91%	3.49%	3.86%
	Precision (SDV)	0.00833	0.00677	0.01120	0.02184	0.02707	0.02405
	Resolution		7.90	3.80		8.85	4.15
	Tailing factor	0.85	0.90	0.97	0.90	0.97	0.95

Condition 1: flow rate 2.0 mL/min, oven temp 200 °C (Normal condition); Condition 2: flow rate 1.8 mL/min, oven temp 200 °C; Condition 3: flow rate 2.2 mL/min, oven temp 200 °C; Condition 4: flow rate 2.0 mL/min, oven temp 195 °C, Condition 5: flow rate 2.0 mL/min, oven temp 205 °C.

.

3.7.1. Limits of Detection and Quantification

The detection limit refers to the minimum amount of the target analyte that the analytical method is able to detect. The quantitation limit refers to the minimum amount of the target analyte that the analytical method is able to precisely and accurately quantitate. The quantitation limit is a validation parameter for the quantitative testing of samples containing trace amounts of analyte, or in particular for the determination of impurities and decomposition products. For the determination of the detection limit and quantitation limit, the signal-to-noise ratio of a reference solution containing a low concentration of 30 ppm of each of the target analytes was analyzed. The signal-to-noise ratio of 3:1 was considered for the determination of the detection limit and 10:1 was considered for the determination of the quantitation limit. When obtained using GC-FID, the detection limits were found to be 1.5 ppm, 1.5 ppm and 1.9 ppm, and the quantification limits were found to be 4.9 ppm, 5.1 ppm, 6.4 ppm for MCS, ECS and ICS, respectively. When obtained using GC-MS, the detection limits were found to be 0.055 ppm, 0.069 ppm, 0.102 ppm, and the quantification limits were found to be 0.185 ppm, 0.232 ppm, 0.340 ppm for MCS, ECS and ICS, respectively. As expected, the analytical method for the MS operated under SIM mode had much better sensitivity than the analytical method for the FID. The baseline noise was determined to be significantly better when operated using SIM mode. In addition, injection precision was confirmed through six repeated injections at a low concentration near the quantification limit. All results satisfied the acceptance criteria with the relative standard deviation (%RSD) being below 10.0%. Both of the analytical methods were confirmed to have suitable detection sensitivity for each purpose.

3.7.2. Linearity and Range

Linearity refers to the ability of the analytical method to detect a linear response of the reference standard with respect to its relative concentration within a specific range. Linearity was evaluated at five concentration levels at the QL concentration to 180 ppm (120% of the specified limit). The range was chosen to be from the range of 30 ppm to 150 ppm for MCS, ECS, and ICS.

Linearity was evaluated as coefficient of determination (R²) and all of the results using both FID and MS were not less than 0.99. The overall linearity was better established in the method using FID than MS.

3.7.3. Precision and Accuracy

Precision refers to the degree of dispersion between a series of measurements of a uniform sample analyzed using the same analytical method. Accuracy refers to the ability of the analytical method to obtain results close to a standard or known true value. Precision was evaluated by determining the %RSD of the peak area ratios between MCS, ECS, ICS and the internal standard obtained from six replicate injection of the standard solution. Precision was evaluated at the concentration levels of 10 ppm, 30 ppm and 150 ppm. All results satisfied the acceptance criteria, with %RSD values below 10.0 %.

Accuracy was evaluated by determining the recovery of the sample solutions spiked with known concentrations of MCS, ECS, and ICS. The spiked sample solutions were prepared in triplicate at three levels using the drug substance, amlodipine camsylate as the matrix sample. The amount of each target analyte was quantitated using the developed analytical method and the recoveries for each of the solutions were calculated.

Accuracy was evaluated at 10 ppm, 120 ppm, 150 ppm and 180 ppm. All results satisfied the acceptance criteria, with the recovery being within 100 ± 15% for all target analytes at each concentration level.

3.7.4. Specificity

Specificity refers to the ability of the method to quantitate the amount of each target analyte without any interferences when deliberately spiked with impurities or degradants of the sample. The resolution between each of the target analytes for the standard solution were determined to evaluate the specificity of the analytical method. The resolution between MCS and ECS was determined to be 9.51 and 4.53 between ECS and ICS obtained using the GC-FID analytical method. The resolution between MCS and ECS was determined to be 10.21 and 4.78 between ECS and ICS obtained using the GC-MS analytical method. All resolution values between adjacent peaks were not less than 3.0, meaning that all peaks were well separated from one another.

3.7.5. Robustness

Robustness of the analytical method refers to the ability of the analytical method to remain unchanged when the conditions are intentionally changed. Robustness is a validation parameter indicating how reliable the test method is during normal use. From the established analytical method, the flow rate of the carrier gas was changed from 1.8 mL/min to 2.2 mL/min, and the column oven temperature was changed from 195 °C to 205 °C. All other analytical parameters except for the flow rate and column oven temperature were fixed at their original set values in order to verify the effect of flow rate and column oven temperature on the analysis results. To determine the effects of the changes in the results, precision of altered analytical method was evaluated using the %RSD and standard deviation of the peak area ratios between MCS, ECS, ICS and the internal standard obtained from six replicate injections of the standard solution. The resolution between adjacent peaks and tailing factor were also evaluated for each peak. For the analytical method using GC-FID, the obtained precision values satisfied the acceptance criteria with %RSD being below 10.0%. The resolution for robustness conditions 1 to 5 were found to be between 7.90 and 10.03 between the peaks of MCS and ECS, and between 3.80 and 4.68 between the peaks of ECS and ICS. All results were not less than 3.0. The tailing factor was found to be between 0.85 and 1.04. For the analytical method using GC-MS, the obtained precision values also satisfied the acceptance criteria with %RSD being below 10.0%. The resolution for robustness conditions 1 to 5 were found to be between 8.85 and 10.56 between the peaks of MCS and ECS, and between 4.15 and 4.93 between the peaks of ECS and ICS. All results were not less than 3.0. The tailing factor was found to be between 0.90 and 1.02. The analytical results for both the GC-FID and GC-MS were determined to be robust even with changes in flow rate and column oven temperature.

4. Conclusions

The three alkyl camphorsulfonates were identified as PGIs through an in silico study according to the ICH M7 guideline. An analytical method for three alkyl camphorsulfonates using GC was developed by analytical QbD demonstrating high sensitivity, reproducibility and a short time of analysis. The concept of analytical QbD was effective to develop the analytical method within a design space to satisfy the needs of high sensitivity and specificity as CMAs when analyzing PGIs. The developed PGI analysis methods have been validated according to the ICH Q2 guideline. The GC-FID method showed better results in linearity and recovery, and the GC-MS method showed better results in detection sensitivity. Both the GC-FID and GC-MS methods can be used to effectively carry out routine analysis in the quality control of PGIs in the pharmaceutical industry, and the developed analytical methods can be applied to APIs containing camphorsulfonic acid salt to control PGIs. The limitations to this study may be the fact that only helium was used as the carrier gas. There may be additional testing required in order to check the effects of carrier gases other than helium on the analytical method’s capabilities. Moreover, additional studies for the development of an analytical method for other alkyl camphorsulfonate PGIs other than methyl, ethyl, and isopropyl camphorsulfonate is needed.