Successful Development by Design of Experiments of a Gas Chromatography Method for Simultaneous Analysis of Residual Solvents of Classes 1 and 2

Successful Development by Design of Experiments of a Gas Chromatography Method for Simultaneous... Abstract In this study, design of experiments (DoE) was employed to develop a single injection method using a headspace gas chromatograph with flame ionization detector for resolution of residual solvents of United States Pharmacopeia (USP) listed classes 1 and 2, against current recommendation of independent injections. G43 column (6% cyanopropylphenyl and 94% dimethylpolysiloxane) and nitrogen were used as the stationary phase and carrier gas, respectively. Initial temperature, hold time, temperature ramp and carrier gas velocity were the critical method parameters. Resolution of 1,1,1-trichloroethane and tetrahydrofuan (THF); THF and chloroform; benzene and 1,2-dichloroethane (DCE), and 1,2-dimethoxyethane and DCE were selected as critical quality attributes. These were optimized by DoE that resulted in resolution of >1.34 among various solvents. The separation of all the solvents was achieved within total run time of 77 min. A better resolution of 2.66 was observed in the case of acetonitrile and methylene chloride; there was improved Signal/Noise ratio of 8.86 for 1,1,1-trichloroethane; tailing factor for pyridine was 1.00, and the method showed acceptable repeatability of peak areas (%RSDmax = 11.53) and retention times (%RSDmax = 0.45). Thus, while system suitability criteria and validation results very well met the USP requirements, the optimized method proposed in this study proved advantageous additionally in terms of single injection; short run time, and use of nitrogen as a carrier gas instead of costly helium in the USP method. Introduction Residual solvents in pharmaceuticals are remnants of solvents, which are used during manufacture of the drug substances, excipients, or the drug products. The same do not provide therapeutic benefit, rather can be harmful or may even react with drug or its degradation products to yield unwanted interaction products. United States Pharmacopeia (USP) mentions three classes of residual solvents, i.e., Class 1 (solvents to be avoided); Class 2 (solvents to be limited) and Class 3 (solvents with low toxic potential) (1,2). These solvents ought to be monitored in drug substances and products at various levels, e.g., 2–1,500 ppm for Class 1, 50–3,880 ppm for Class 2 and 5,000 ppm or less for Class 3. Three procedures (A, B and C) each are outlined for the analysis of Classes 1 and 2 water soluble and water insoluble solvents by gas chromatograph (GC) equipped with a headspace (HS) injector. Procedure A is recommended for initial screening and identification of the solvents. If a solvent is above the specified concentration limit, then Procedure B is used to confirm its identity. Procedure C is applied for quantification purpose. International Council for Harmonisation guideline Q3C (R6) suggests the following under the head “Analytical Procedures”: “Residual solvents are typically determined using chromatographic techniques such as gas chromatography (3). Any harmonized procedures for determining levels of residual solvents as described in the pharmacopoeias should be used, if feasible. Otherwise, manufacturers would be free to select the most appropriate validated analytical procedure for a particular application. If only Class 3 solvents are present, a non-specific method such as loss on drying may be used.” Accordingly, it favors pharmacopoeial methods but it allows use of alternate validated analytical procedures, without outlining specific experimental approach. Hence, several alternate GC methods targeted to different classes of residual solvents have been reported in the literature (4–14). Among them, some of the reports only considered Class 1 solvents (11–14), while other publications mainly focused on Class 2 and Class 3 solvents (4–10). However, no common method is reported yet for residual solvents for Classes 1 and 2, which can detect most of them in a single injection. The endeavor of this study was to apply design of experiment (DoE) strategy to develop a common GC method for solvents belonging to Classes 1 and 2, giving compliance to system suitability criteria mentioned in USP. The compendium recommends use of standard mixtures of solvents, classifying them into Classes 1, 2A, 2B and 2C. The pharmacopoeia highlights that Class 2C solvents are not detected easily by GC methods. Thus, the present study was restricted to separation and detection of solvents using standard mixtures of Classes 1, 2A and 2B. Experimental Chemicals and reagents Standard mixtures of solvents of Classes 1, 2A and 2B were purchased from Sigma Aldrich (New Delhi, India). Dimethylsulfoxide (DMSO; 99.5% pure) was procured from Merck Specialties Pvt Ltd (Mumbai, Maharashtra, India). Nitrogen, hydrogen and zero air cylinders (99.9% pure) were supplied by Vikas Gases (Panchkula, Haryana, India). Ultrapure water was obtained from ELGA water purification unit (Elga, Wycombe, UK). Equipments and software The GC (Clarus 680) with TurboMatrix 40 HS injector and flame ionization detector (FID) were from PerkinElmer, Inc. (MA, USA). The GC was operated using TotalChrom workstation (version 6.3.2). The sampler was operated using TurboMatrix (version 2.5.0) software. Design expert (trial version 9.0.3.1) was used for DoE studies. Sample preparation Class 1 standard solution. About 1 mL of standard mixture of Class 1 solvents was transferred to a 100-mL volumetric flask, previously filled with 80 mL DMSO and the volume was made up with the same solvent. About 1 mL of this solution was similarly diluted to 100 mL in DMSO. Still further, 1 mL of resultant solution was diluted 10 times in DMSO. Class 2A standard solution. About 1 mL of standard mixture of Class 2A solvents was transferred to 100-mL volumetric flask, previously filled with 80 mL DMSO. Volume was made up to 100 mL with the same solvent. Class 2B standard solution. About 0.5 mL of standard mixture of Class 2B solvents was transferred to a 10-mL volumetric flask. Volume was made up to 10 mL with DMSO. Combined standard solution. About 1 mL was taken from each of the three above-prepared standard solutions and transferred to a 20-mL crimp top glass vial. About 5 mL of water was added and the vial was crimped. This vial, containing homogenous mixture of solvents of classes 1, 2 A and 2B, was subjected to GC analysis. GC analysis The GC column used was DB-624 UI, 30 m, 0.32 mm i.d., 1.8 μm film thickness having G43 stationary phase (6% cyanopropylphenyl, 94% dimethylpolysiloxane). Nitrogen was used as a carrier gas. The flow rates of zero air and hydrogen in flame ionized detector (FID) were 450 mL/min and 45 mL/min, respectively. The optimized HS injector parameters were oven temperature, 80°C; thermostat time, 45 min; needle temperature, 90°C; transfer line temperature, 110°C; pressurize time, 3 min and injection volume, 1 mL. Method development using DoE approach Critical method parameters A risk number (R) was assigned to various method parameters that were supposed to effect resolution (Rs) among various solvents during GC analysis. Parameters like column type, initial temperature, hold time, temperature ramp, carrier gas velocity, type of carrier gas, etc. were categorized on severity and probability basis. The score for severity was decided on the basis of Rs value, i.e., 1 for minor (Rs > 1.5), 2 for major (0 < Rs < 1.5) and 3 for critical (zero or peak exchange). Probability scores of 1–4 were given on the basis of effect of method parameters on the Rs value, i.e., unlikely (1), occasional (2), probable (3) and frequent (4). Design of experiments In the case, where response of interest is influenced by several variables, like the GC study in this case, response surface designs based on the fit of a polynomial equation could be well applied. They are most relevant and useful for optimization of levels of these variables to attain the desired outcome (15, 16). Among various response surface designs, optimal designs are recommended when we need to fit a cubic or higher order model with minimum number of runs (17). In the present study, among various A-, C-, D-, E-, T-, G-, I- and V-optimal response surface designs, I-optimal design was selected as it suited better because its algorithm tends to select runs that minimize the integral of the prediction variance across the factor space (18). The Design Expert® software was accordingly used first to generate experimental conditions for the identified critical method parameters (CMPs), and subsequently for statistical analysis of results and response optimization. Before practical conduct, all the experimental runs were randomized to avoid any bias to the study. The responses were noted in terms of resolution (Rs) between two immediately resolving solvent peaks in the GC chromatogram. Critical quality attributes The critical quality attributes (CQAs) in the present case were considered to be those pairs of solvents, whose Rs values, as responses of multiple experimental runs, were associated with scores for severity of 2 and 3. Method suitability The suitability of the developed method was evaluated by USP system suitability criteria for residual solvent analysis and method repeatability. For the latter, separate injections were given under the same conditions from freshly prepared six vials containing mixture of standard solutions. Results The method parameters with their risk numbers are listed in Table I. The matrix of experimental runs proposed by I-optimal design for CMPs is given in Supplementary Table S1. The responses in terms of resolution Rs for CQAs where scores for severity were either 2 or 3 (Sections 2.5.1 and 2.5.3) are listed in Supplementary Table S2. ANOVA details for response surface cubic models employed for CQAs are given in Table II. The effect of CMPs on the CQAs was accomplished with the help of 3D surface response plots (Figure 1) and the Derringer desirability plot (Figure 2). The details of CMPs and CQAs, along with their constraints, are provided in Table III, while optimized method parameters are outlined in Table IV. The resultant chromatogram generated using an optimized method is shown in Figure 3. The system suitability data are outlined in Table V, while method repeatability data of the peak areas and the retention times are listed in Supplementary Tables S3 and S4, respectively. Table I. Risk Analysis for Different Method Parameters Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Table I. Risk Analysis for Different Method Parameters Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Table II. ANOVA Details for Response Surface Cubic Models Employed for the Solvent Pairs Identified as CQAs CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – Solvent pair A: 1,1,1-Trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane; and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table II. ANOVA Details for Response Surface Cubic Models Employed for the Solvent Pairs Identified as CQAs CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – Solvent pair A: 1,1,1-Trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane; and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table III. List of CMPs and CQAs along with Goal, Limits and Importance Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Solvent pair A: 1,1,1-trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table III. List of CMPs and CQAs along with Goal, Limits and Importance Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Solvent pair A: 1,1,1-trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table IV. Optimized Method Parameters for Residual Solvent Analysis GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 Table IV. Optimized Method Parameters for Residual Solvent Analysis GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 Table V. System Suitability Data Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Acceptance criteria (USP): Rs > 1 and S/N > 5 for all class 1 solvents except 1,1,1-trichloroethane (S/N > 3). Table V. System Suitability Data Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Acceptance criteria (USP): Rs > 1 and S/N > 5 for all class 1 solvents except 1,1,1-trichloroethane (S/N > 3). Figure 1. View largeDownload slide 3D surface response plots showing the effect of CMPs on selected CQAs. color changes from blue to red (Min Max) represents CQA responses. Figure 1. View largeDownload slide 3D surface response plots showing the effect of CMPs on selected CQAs. color changes from blue to red (Min Max) represents CQA responses. Figure 2. View largeDownload slide Derringer desirability plot showing the effect of (a) hold time and initial temperature with 10°C/min temperature ramp and 12 cm/s carrier gas velocity, and (b) carrier gas velocity and temperature ramp with 28°C initial temperature and 45 min hold time. Color changes from blue to red (Min Max) represent the CQA responses. Figure 2. View largeDownload slide Derringer desirability plot showing the effect of (a) hold time and initial temperature with 10°C/min temperature ramp and 12 cm/s carrier gas velocity, and (b) carrier gas velocity and temperature ramp with 28°C initial temperature and 45 min hold time. Color changes from blue to red (Min Max) represent the CQA responses. Figure 3. View largeDownload slide Zoom chromatogram showing separation of class 1, 2A and 2B residual solvents in single run using an optimized method. Figure 3. View largeDownload slide Zoom chromatogram showing separation of class 1, 2A and 2B residual solvents in single run using an optimized method. Discussion As shown from values in Table I, four parameters viz., initial temperature, hold time, temperature ramp and carrier gas velocity, had value of R >4, and hence these were selected as CMPs. The Design Expert® software, which was used to generate experimental conditions for the identified CMPs, suggested a total of 35 model points, which after inclusion of five lack-of-fit points and five replicate points, meant 45 runs in total. The lower and higher ranges for the CMPs were: initial temperature, 28°C and 45°C; hold time, 20 min and 50 min; temperature ramp, 5°C/min and 15°C/min; and carrier gas velocity, 10 cm/s and 20 cm/s. The data for all 45 experimental runs for 28 solvents showed that optimal resolution (score for severity = 1) was achieved for most of the solvents under different conditions of CMPs. The exceptions were 1,1,1-trichloroethane and tetrahydrofuran (solvent pair A); tetrahydrofuran and chloroform (solvent pair B); benzene and 1,2-dichloroethane (solvent pair C) and 1,2-dichloroethane and 1,2-dimethoxyethane (solvent pair D), whose scores for severity were either 2 or 3, as depicted from response (Rs) values in Supplementary Table S2. These solvent pairs were considered as CQAs for further optimization. Once the CQAs were identified, the next step was the determination of the effect of CMPs on them. For the same, cubic polynomial model was selected, which contained model terms for both the main and the interaction effects. Evaluation of model fitness using ANOVA demonstrated that the model was significant for all the four CQAs (Table II). In case of solvent pairs A–D, the F-values of 3.30, 8.60, 29.14 and 16.88, respectively, meant that models were significant, highlighting that there were only 2.45%, 0.05%, 0.01% and 0.01% chances that these F-values could occur due to noise. Values of ‘Prob > F’ < 0.0500 for all solvent pairs also indicated that the model terms were significant. Even the ‘lack-of-fit F-value’ of 0.57 for solvent pair A implied that the lack-of-fit was not significant relative to the pure error. The 3D surface response plots in Figure 1 were generated after validation of the models. It was shown that among the CMPs, initial temperature had influence mainly on resolution for the solvent pair B containing tetrahydrofuran and chloroform. The increase in initial temperature resulted in decrease in resolution between these two solvents. Therefore, decrease in the initial temperature was favored for their resolution. On the other hand, hold time had significant influence on separation of solvent pairs B–D. The resolution increased upon increase in hold time at a lower temperature. The hold time had little impact on the resolution at higher initial temperature; temperature ramp had little impact on solvent pair B, while it had positive impact on solvent pair C. A decrease in temperature ramp was found beneficial for resolution between all the solvents. All CQAs further showed improved resolution when carrier gas velocity was decreased to an optimum level. The combined effect of the all four CMPs on selected CQAs was also evaluated by Derringer desirability plots (19). In order to optimize Rs for solvent pairs A–D, the goals for CMPs as well as CQAs were defined as ‘maximized’/‘minimized’/‘in the range’. For all the CMPs and CQAs, importance was assigned a value of 5 on the scale of 1–5, except temperature ramp where ‘in the range’ was used as a goal (Table III). Resultant desirability plots, as shown in Figure 2, with value of 0.92 helped in optimizing the method parameters, which are listed in Table IV. The gas chromatogram in Figure 3 clearly shows that the optimized method was able to separate all the residual solvents present in a mixture of standard solutions of Classes 1, 2A and 2B. As per USP, the Rs between acetonitrile and methylene chloride should not be less than 1. Also, the Signal/Noise (S/N) ratio should be not <3 for 1,1,1-trichloroethane, and it should be not <5 for other Class 1 solvents. A shown in Table V, the minimum Rs achieved using newly developed method was 1.34, resolution between acetonitrile and methylene chloride was 2.66, while minimum S/N ratio was 6.14. Figure 3 shows that the peak shape was good for even pyridine (peak tailing 1.00) and total run time for the method optimized in this study was just 77 min against 188 min for the USP prescribed method. Hence, the optimized method fulfilled all the system suitability requirements of USP for residual solvent analysis. Finally, the experiments conducted to verify the repeatability of the method for peak areas (Supplementary Tables S3) and retention times (Supplementary Table S4) for all solvent peaks confirmed the repeatability of the method. The %RSD values for the peak areas and retention times were <15 and <1, respectively. Conclusion A single injection method for simultaneous analysis of residual solvents of Classes 1 and 2 (except 2C) was successfully developed by employing the DOE approach. The effect of GC parameters on the separation was systematically studied through I-optimal design, and also by taking help of 3D surface response and Derringer desirability plots. The optimized method proposed in this study carries the following significant advantages over USP/EP method or any other method reported in the literature: Separation of all solvents in a single injection with total run time of 77 min. Better resolution of 2.66 in the case of acetonitrile and methylene chloride. Improved S/N ratio of 8.86 for 1,1,1-trichloroethane. Improvement in peak tailing of pyridine (tailing factor = 1.00). Acceptable repeatability of peak areas (%RSDmax = 11.53) and retention times (%RSDmax = 0.45). Use of nitrogen as a carrier gas, instead of costly helium. Supplementary data Supplementary material is available at Journal of Chromatographic Science online. References 1 General chapter . <467> Residual Solvents, in: USP38, The United States Pharmacopeial Convention, Rockville (USA), 2015 . 2 Countryman , S. , Kelly , K. , Harder , B. ; USP method 467: organic volatile impurities testing ; LCGC North America , ( 2005 ); 23 ( SUPPL. 1 ): 65 . 3 Impurities: Guideline for Residual Solvents , Q3C (R6), International Council for Harmonisation (ICH), IFPMA, Geneva (Switzerland), 2016 . 4 Panovska , A.P. , Acevska , J. , Stefkov , G. , Brezovska , K. , Petkovska , R. , Dimitrovska , A. ; Optimization of HS–GC–FID–MS Method for Residual Solvent Profiling in Active Pharmaceutical Ingredients Using DoE ; Journal of Chromatographic Science , ( 2016 ); 54 ( 2 ): 103 – 111 . Google Scholar PubMed 5 Grodowska , K. , Parczewski , A. ; Grouping of residual solvents present in pharmaceuticals using experimental planning and chemometric methods ; Acta Poloniae Pharmaceutica - Drug Research , ( 2013 ); 70 ( 2 ): 191 – 203 . 6 Andrews-Ashley , A.J. , Belsky , J.L. , Joyce , H.R. , Pan , C. , Pappa , H.N. , Wahab , S.Z. ; Static headspace GC–MS detection of residual solvents—a possible simultaneous identification and quantitation method ; LCGC North America , ( 2012 ); 30 ( 1 ): 1 . 7 Wang , C.-Y. , Li , L.-X. , Xie , T.-T. , Lin , J.-F. , Chu , N.-Q. , Tang , L.-C. , et al. . ; Simultaneous determination of the residual contents of 21 harmful organic solvents in leather and leather products by GC-MS/MS coupled with ultrasonic extraction ; Journal of Chinese Mass Spectrometry Society , ( 2016 ); 37 ( 2 ): 121 – 133 . 8 Puranik , S.B. , Sanjay Pai , P.N. , Rao , G.K. ; Determination of organic volatile impurities in herbal formulations and extracts by capillary gas chromatography ; International Journal of Applied Research in Natural Products , ( 2009 ); 2 ( 1 ): 33 – 46 . 9 Cheng , C. , Liu , S. , Mueller , B.J. , Yan , Z. ; A generic static headspace gas chromatography method for determination of residual solvents in drug substance ; Journal of Chromatography A , ( 2010 ); 1217 : 6413 – 6421 . Google Scholar CrossRef Search ADS PubMed 10 Camarasu , C.C. , Mezei-Szuts , M. , Varga , G.B. ; Residual solvents determination in pharmaceutical products by GC-HS and GC-MS-SPME ; Journal of Pharmaceutical and Biomedical Analysis , ( 1998 ); 18 : 623 – 638 . Google Scholar CrossRef Search ADS PubMed 11 Heydari , R. , Azizi , S. ; Rapid Monitoring and Determination of Class 1 Residual Solvents in Pharmaceuticals Using Dispersive Liquid–Liquid Microextraction and Gas Chromatography–Mass Spectrometry ; Journal of Chromatographic Science , ( 2015 ); 53 ( 6 ): 1020 – 1025 . Google Scholar CrossRef Search ADS PubMed 12 Farajzadeh , M.A. , Goushjuii , L. , Djozan , D. , Mohammadi , J.K. ; Dispersive liquid–liquid microextraction combined with gas chromatography for extraction and determination of class 1 residual solvents in pharmaceuticals ; Journal of Separation Science , ( 2015 ); 35 ( 8 ): 1027 – 1035 . Google Scholar CrossRef Search ADS 13 Pérez Pavón , J.L. , Del Nogal Sánchez , M. , Fernández Laespada , M.E. , García Pinto , C. , Moreno Cordero , B. ; Analysis of class 1 residual solvents in pharmaceuticals using headspace-programmed temperature vaporization-fast gas chromatography-mass spectrometry ; Journal of Chromatography A , ( 2007 ); 1141 ( 1 ): 123 – 130 . Google Scholar CrossRef Search ADS PubMed 14 Lakatos , M. ; Measurement of residual solvents in a drug substance by a purge-and-trap method ; Journal of Pharmaceutical and Biomedical Analysis , ( 2008 ); 47 ( 4-5 ): 954 – 957 . Google Scholar CrossRef Search ADS PubMed 15 Myers , R.H. , Khuri , A.I. , Carter , W.H. ; Response surface methodology: 1966–1988 ; Technometrics , ( 1989 ); 31 : 137 – 157 . 16 Candiotim , L.V. , De Zan , M.M. , Cámara , M.S. , Goicoechea , H.C. ; Experimental design and multiple response optimization. Using the desirability function in analytical methods development ; Talanta , ( 2014 ); 124 : 123 – 138 . Google Scholar CrossRef Search ADS PubMed 17 Hardin , R.H. , Sloane , N.J.A. ; A new approach to the construction of optimal designs ; Journal of Statistical Planning and Inference , ( 1993 ); 37 : 339 – 369 . Google Scholar CrossRef Search ADS 18 Goos , P. , Jones , B. ; Optimal design of experiments: a case study approach , 1st ed. Wiley , West Sussex (UK) , ( 2011 ). Google Scholar CrossRef Search ADS 19 Derringer , G.C. ; A balancing act: optimizing a product’s properties ; Quality Progress , ( 1994 ); 27 : 51 – 58 . © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Chromatographic Science Oxford University Press

Successful Development by Design of Experiments of a Gas Chromatography Method for Simultaneous Analysis of Residual Solvents of Classes 1 and 2

Loading next page...
 
/lp/ou_press/successful-development-by-design-of-experiments-of-a-gas-UfvfvtIBEt
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
0021-9665
eISSN
1945-239X
D.O.I.
10.1093/chromsci/bmy026
Publisher site
See Article on Publisher Site

Abstract

Abstract In this study, design of experiments (DoE) was employed to develop a single injection method using a headspace gas chromatograph with flame ionization detector for resolution of residual solvents of United States Pharmacopeia (USP) listed classes 1 and 2, against current recommendation of independent injections. G43 column (6% cyanopropylphenyl and 94% dimethylpolysiloxane) and nitrogen were used as the stationary phase and carrier gas, respectively. Initial temperature, hold time, temperature ramp and carrier gas velocity were the critical method parameters. Resolution of 1,1,1-trichloroethane and tetrahydrofuan (THF); THF and chloroform; benzene and 1,2-dichloroethane (DCE), and 1,2-dimethoxyethane and DCE were selected as critical quality attributes. These were optimized by DoE that resulted in resolution of >1.34 among various solvents. The separation of all the solvents was achieved within total run time of 77 min. A better resolution of 2.66 was observed in the case of acetonitrile and methylene chloride; there was improved Signal/Noise ratio of 8.86 for 1,1,1-trichloroethane; tailing factor for pyridine was 1.00, and the method showed acceptable repeatability of peak areas (%RSDmax = 11.53) and retention times (%RSDmax = 0.45). Thus, while system suitability criteria and validation results very well met the USP requirements, the optimized method proposed in this study proved advantageous additionally in terms of single injection; short run time, and use of nitrogen as a carrier gas instead of costly helium in the USP method. Introduction Residual solvents in pharmaceuticals are remnants of solvents, which are used during manufacture of the drug substances, excipients, or the drug products. The same do not provide therapeutic benefit, rather can be harmful or may even react with drug or its degradation products to yield unwanted interaction products. United States Pharmacopeia (USP) mentions three classes of residual solvents, i.e., Class 1 (solvents to be avoided); Class 2 (solvents to be limited) and Class 3 (solvents with low toxic potential) (1,2). These solvents ought to be monitored in drug substances and products at various levels, e.g., 2–1,500 ppm for Class 1, 50–3,880 ppm for Class 2 and 5,000 ppm or less for Class 3. Three procedures (A, B and C) each are outlined for the analysis of Classes 1 and 2 water soluble and water insoluble solvents by gas chromatograph (GC) equipped with a headspace (HS) injector. Procedure A is recommended for initial screening and identification of the solvents. If a solvent is above the specified concentration limit, then Procedure B is used to confirm its identity. Procedure C is applied for quantification purpose. International Council for Harmonisation guideline Q3C (R6) suggests the following under the head “Analytical Procedures”: “Residual solvents are typically determined using chromatographic techniques such as gas chromatography (3). Any harmonized procedures for determining levels of residual solvents as described in the pharmacopoeias should be used, if feasible. Otherwise, manufacturers would be free to select the most appropriate validated analytical procedure for a particular application. If only Class 3 solvents are present, a non-specific method such as loss on drying may be used.” Accordingly, it favors pharmacopoeial methods but it allows use of alternate validated analytical procedures, without outlining specific experimental approach. Hence, several alternate GC methods targeted to different classes of residual solvents have been reported in the literature (4–14). Among them, some of the reports only considered Class 1 solvents (11–14), while other publications mainly focused on Class 2 and Class 3 solvents (4–10). However, no common method is reported yet for residual solvents for Classes 1 and 2, which can detect most of them in a single injection. The endeavor of this study was to apply design of experiment (DoE) strategy to develop a common GC method for solvents belonging to Classes 1 and 2, giving compliance to system suitability criteria mentioned in USP. The compendium recommends use of standard mixtures of solvents, classifying them into Classes 1, 2A, 2B and 2C. The pharmacopoeia highlights that Class 2C solvents are not detected easily by GC methods. Thus, the present study was restricted to separation and detection of solvents using standard mixtures of Classes 1, 2A and 2B. Experimental Chemicals and reagents Standard mixtures of solvents of Classes 1, 2A and 2B were purchased from Sigma Aldrich (New Delhi, India). Dimethylsulfoxide (DMSO; 99.5% pure) was procured from Merck Specialties Pvt Ltd (Mumbai, Maharashtra, India). Nitrogen, hydrogen and zero air cylinders (99.9% pure) were supplied by Vikas Gases (Panchkula, Haryana, India). Ultrapure water was obtained from ELGA water purification unit (Elga, Wycombe, UK). Equipments and software The GC (Clarus 680) with TurboMatrix 40 HS injector and flame ionization detector (FID) were from PerkinElmer, Inc. (MA, USA). The GC was operated using TotalChrom workstation (version 6.3.2). The sampler was operated using TurboMatrix (version 2.5.0) software. Design expert (trial version 9.0.3.1) was used for DoE studies. Sample preparation Class 1 standard solution. About 1 mL of standard mixture of Class 1 solvents was transferred to a 100-mL volumetric flask, previously filled with 80 mL DMSO and the volume was made up with the same solvent. About 1 mL of this solution was similarly diluted to 100 mL in DMSO. Still further, 1 mL of resultant solution was diluted 10 times in DMSO. Class 2A standard solution. About 1 mL of standard mixture of Class 2A solvents was transferred to 100-mL volumetric flask, previously filled with 80 mL DMSO. Volume was made up to 100 mL with the same solvent. Class 2B standard solution. About 0.5 mL of standard mixture of Class 2B solvents was transferred to a 10-mL volumetric flask. Volume was made up to 10 mL with DMSO. Combined standard solution. About 1 mL was taken from each of the three above-prepared standard solutions and transferred to a 20-mL crimp top glass vial. About 5 mL of water was added and the vial was crimped. This vial, containing homogenous mixture of solvents of classes 1, 2 A and 2B, was subjected to GC analysis. GC analysis The GC column used was DB-624 UI, 30 m, 0.32 mm i.d., 1.8 μm film thickness having G43 stationary phase (6% cyanopropylphenyl, 94% dimethylpolysiloxane). Nitrogen was used as a carrier gas. The flow rates of zero air and hydrogen in flame ionized detector (FID) were 450 mL/min and 45 mL/min, respectively. The optimized HS injector parameters were oven temperature, 80°C; thermostat time, 45 min; needle temperature, 90°C; transfer line temperature, 110°C; pressurize time, 3 min and injection volume, 1 mL. Method development using DoE approach Critical method parameters A risk number (R) was assigned to various method parameters that were supposed to effect resolution (Rs) among various solvents during GC analysis. Parameters like column type, initial temperature, hold time, temperature ramp, carrier gas velocity, type of carrier gas, etc. were categorized on severity and probability basis. The score for severity was decided on the basis of Rs value, i.e., 1 for minor (Rs > 1.5), 2 for major (0 < Rs < 1.5) and 3 for critical (zero or peak exchange). Probability scores of 1–4 were given on the basis of effect of method parameters on the Rs value, i.e., unlikely (1), occasional (2), probable (3) and frequent (4). Design of experiments In the case, where response of interest is influenced by several variables, like the GC study in this case, response surface designs based on the fit of a polynomial equation could be well applied. They are most relevant and useful for optimization of levels of these variables to attain the desired outcome (15, 16). Among various response surface designs, optimal designs are recommended when we need to fit a cubic or higher order model with minimum number of runs (17). In the present study, among various A-, C-, D-, E-, T-, G-, I- and V-optimal response surface designs, I-optimal design was selected as it suited better because its algorithm tends to select runs that minimize the integral of the prediction variance across the factor space (18). The Design Expert® software was accordingly used first to generate experimental conditions for the identified critical method parameters (CMPs), and subsequently for statistical analysis of results and response optimization. Before practical conduct, all the experimental runs were randomized to avoid any bias to the study. The responses were noted in terms of resolution (Rs) between two immediately resolving solvent peaks in the GC chromatogram. Critical quality attributes The critical quality attributes (CQAs) in the present case were considered to be those pairs of solvents, whose Rs values, as responses of multiple experimental runs, were associated with scores for severity of 2 and 3. Method suitability The suitability of the developed method was evaluated by USP system suitability criteria for residual solvent analysis and method repeatability. For the latter, separate injections were given under the same conditions from freshly prepared six vials containing mixture of standard solutions. Results The method parameters with their risk numbers are listed in Table I. The matrix of experimental runs proposed by I-optimal design for CMPs is given in Supplementary Table S1. The responses in terms of resolution Rs for CQAs where scores for severity were either 2 or 3 (Sections 2.5.1 and 2.5.3) are listed in Supplementary Table S2. ANOVA details for response surface cubic models employed for CQAs are given in Table II. The effect of CMPs on the CQAs was accomplished with the help of 3D surface response plots (Figure 1) and the Derringer desirability plot (Figure 2). The details of CMPs and CQAs, along with their constraints, are provided in Table III, while optimized method parameters are outlined in Table IV. The resultant chromatogram generated using an optimized method is shown in Figure 3. The system suitability data are outlined in Table V, while method repeatability data of the peak areas and the retention times are listed in Supplementary Tables S3 and S4, respectively. Table I. Risk Analysis for Different Method Parameters Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Table I. Risk Analysis for Different Method Parameters Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Method parameter Severity (S) Probability (P) Risk (R) = (S∗P) Initial temperature 3 4 12 Hold time 3 3 9 Temperature ramp 3 2 6 Carrier gas velocity 3 2 6 Split ratio 2 1 2 Final temperature 1 1 1 Carrier gas type 1 1 1 Detector temperature 1 1 1 Injector temperature 1 1 1 Table II. ANOVA Details for Response Surface Cubic Models Employed for the Solvent Pairs Identified as CQAs CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – Solvent pair A: 1,1,1-Trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane; and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table II. ANOVA Details for Response Surface Cubic Models Employed for the Solvent Pairs Identified as CQAs CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – CQA Model Lack-of-fit Pure error Some of squares Solvent pair A 35.37 1.15 2.00 Solvent pair B 25.89 0.89 0.000 Solvent pair C 13.20 0.13 0.000 Solvent pair D 62.97 1.10 0.000 Degree of freedom Solvent pair A 34 5 5 Solvent pair B 34 5 5 Solvent pair C 34 5 5 Solvent pair D 34 5 5 Mean square Solvent pair A 1.04 0.23 0.40 Solvent pair B 0.76 0.18 0.000 Solvent pair C 0.39 0.027 0.000 Solvent pair D 1.85 0.22 0.000 F-value Solvent pair A 3.30 0.57 – Solvent pair B 8.60 – – Solvent pair C 29.14 – – Solvent pair D 16.88 – – P-value (Prob > F) Solvent pair A 0.0245 0.7209 – Solvent pair B 0.0005 – – Solvent pair C <0.0001 – – Solvent pair D <0.0001 – – Solvent pair A: 1,1,1-Trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane; and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table III. List of CMPs and CQAs along with Goal, Limits and Importance Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Solvent pair A: 1,1,1-trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table III. List of CMPs and CQAs along with Goal, Limits and Importance Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Goal Lower limit Upper limit Importance CMPs  Initial temperature, °C Minimize 28 45 5  Hold time, min Maximize 20 50 5  Temperature ramp, °C In the range 5 15 –  Carrier gas velocity, cm/s Minimize 10 20 5 CQAs, Constraint Rs > 1  Solvent pair A Maximize Rs 0 3.5 5  Solvent pair B Maximize Rs 0 2 5  Solvent pair C Maximize Rs 0 2 5  Solvent pair D Maximize Rs 0 3.8 5 Solvent pair A: 1,1,1-trichloroethane and tetrahydrofuran; Solvent pair B: tetrahydrofuran and chloroform; Solvent pair C: benzene and 1,2-dichloroethane and Solvent pair D: 1,2-dichloroethane and 1,2-dimethoxyethane. Table IV. Optimized Method Parameters for Residual Solvent Analysis GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 Table IV. Optimized Method Parameters for Residual Solvent Analysis GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 GC parameters Value Temperature Programming 28°C initial temperature (hold for 45 min); then temperature ramp at 10°C/min to 240°C final temperature (hold for 10 min) Carrier gas velocity 12 cm/s Injector temperature 140°C Detector temperature 250°C Split ratio Splitless for 2 min then 5:1 Table V. System Suitability Data Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Acceptance criteria (USP): Rs > 1 and S/N > 5 for all class 1 solvents except 1,1,1-trichloroethane (S/N > 3). Table V. System Suitability Data Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Solvent Rs S/N ratio Tailing factor Methanol – 29.53 0.94 1,1-Dichloroethene 13.93 52.68 1.02 Acetonitrile 6.01 15.17 1.27 Methylene chloride 2.66 92.43 0.97 trans 1,2-Dichloroethene 5.70 157.90 1.08 Hexane 6.60 590.88 0.99 Nitromethane 12.20 13.03 0.95 cis 1,2-Dichloroethene 3.21 321.95 1.04 1,1,1-Trichloroethane 4.81 8.86 1.19 Tetrahydrofuran 2.39 72.43 1.38 Chloroform 1.79 7.97 1.08 Cyclohexane 3.76 4429.13 0.90 Carbontetrachloride 2.82 6.14 1.14 Benzene 5.58 23.34 0.93 1,2-Dichloroethane 1.34 9.37 0.90 1,2-Dimethoxyethane 3.58 32.04 2.10 Trichloroethylene 14.78 62.29 0.98 Methylcyclohexane 4.58 6474.27 0.95 1,4-Dioxane 5.91 11.91 1.30 Pyridine 13.79 80.57 1.00 Toluene 7.30 3470.18 1.00 Methylbutylketone 17.45 25.66 1.17 Chlorobenzene 16.01 719.69 1.00 Ethylebenzene 2.52 1681.43 1.00 m- and p-Xylene 2.48 2165.31 1.00 o-Xylene 8.32 4628.21 1.00 Cumene 7.55 361.93 1.07 Tetralin 54.57 144.84 1.03 Acceptance criteria (USP): Rs > 1 and S/N > 5 for all class 1 solvents except 1,1,1-trichloroethane (S/N > 3). Figure 1. View largeDownload slide 3D surface response plots showing the effect of CMPs on selected CQAs. color changes from blue to red (Min Max) represents CQA responses. Figure 1. View largeDownload slide 3D surface response plots showing the effect of CMPs on selected CQAs. color changes from blue to red (Min Max) represents CQA responses. Figure 2. View largeDownload slide Derringer desirability plot showing the effect of (a) hold time and initial temperature with 10°C/min temperature ramp and 12 cm/s carrier gas velocity, and (b) carrier gas velocity and temperature ramp with 28°C initial temperature and 45 min hold time. Color changes from blue to red (Min Max) represent the CQA responses. Figure 2. View largeDownload slide Derringer desirability plot showing the effect of (a) hold time and initial temperature with 10°C/min temperature ramp and 12 cm/s carrier gas velocity, and (b) carrier gas velocity and temperature ramp with 28°C initial temperature and 45 min hold time. Color changes from blue to red (Min Max) represent the CQA responses. Figure 3. View largeDownload slide Zoom chromatogram showing separation of class 1, 2A and 2B residual solvents in single run using an optimized method. Figure 3. View largeDownload slide Zoom chromatogram showing separation of class 1, 2A and 2B residual solvents in single run using an optimized method. Discussion As shown from values in Table I, four parameters viz., initial temperature, hold time, temperature ramp and carrier gas velocity, had value of R >4, and hence these were selected as CMPs. The Design Expert® software, which was used to generate experimental conditions for the identified CMPs, suggested a total of 35 model points, which after inclusion of five lack-of-fit points and five replicate points, meant 45 runs in total. The lower and higher ranges for the CMPs were: initial temperature, 28°C and 45°C; hold time, 20 min and 50 min; temperature ramp, 5°C/min and 15°C/min; and carrier gas velocity, 10 cm/s and 20 cm/s. The data for all 45 experimental runs for 28 solvents showed that optimal resolution (score for severity = 1) was achieved for most of the solvents under different conditions of CMPs. The exceptions were 1,1,1-trichloroethane and tetrahydrofuran (solvent pair A); tetrahydrofuran and chloroform (solvent pair B); benzene and 1,2-dichloroethane (solvent pair C) and 1,2-dichloroethane and 1,2-dimethoxyethane (solvent pair D), whose scores for severity were either 2 or 3, as depicted from response (Rs) values in Supplementary Table S2. These solvent pairs were considered as CQAs for further optimization. Once the CQAs were identified, the next step was the determination of the effect of CMPs on them. For the same, cubic polynomial model was selected, which contained model terms for both the main and the interaction effects. Evaluation of model fitness using ANOVA demonstrated that the model was significant for all the four CQAs (Table II). In case of solvent pairs A–D, the F-values of 3.30, 8.60, 29.14 and 16.88, respectively, meant that models were significant, highlighting that there were only 2.45%, 0.05%, 0.01% and 0.01% chances that these F-values could occur due to noise. Values of ‘Prob > F’ < 0.0500 for all solvent pairs also indicated that the model terms were significant. Even the ‘lack-of-fit F-value’ of 0.57 for solvent pair A implied that the lack-of-fit was not significant relative to the pure error. The 3D surface response plots in Figure 1 were generated after validation of the models. It was shown that among the CMPs, initial temperature had influence mainly on resolution for the solvent pair B containing tetrahydrofuran and chloroform. The increase in initial temperature resulted in decrease in resolution between these two solvents. Therefore, decrease in the initial temperature was favored for their resolution. On the other hand, hold time had significant influence on separation of solvent pairs B–D. The resolution increased upon increase in hold time at a lower temperature. The hold time had little impact on the resolution at higher initial temperature; temperature ramp had little impact on solvent pair B, while it had positive impact on solvent pair C. A decrease in temperature ramp was found beneficial for resolution between all the solvents. All CQAs further showed improved resolution when carrier gas velocity was decreased to an optimum level. The combined effect of the all four CMPs on selected CQAs was also evaluated by Derringer desirability plots (19). In order to optimize Rs for solvent pairs A–D, the goals for CMPs as well as CQAs were defined as ‘maximized’/‘minimized’/‘in the range’. For all the CMPs and CQAs, importance was assigned a value of 5 on the scale of 1–5, except temperature ramp where ‘in the range’ was used as a goal (Table III). Resultant desirability plots, as shown in Figure 2, with value of 0.92 helped in optimizing the method parameters, which are listed in Table IV. The gas chromatogram in Figure 3 clearly shows that the optimized method was able to separate all the residual solvents present in a mixture of standard solutions of Classes 1, 2A and 2B. As per USP, the Rs between acetonitrile and methylene chloride should not be less than 1. Also, the Signal/Noise (S/N) ratio should be not <3 for 1,1,1-trichloroethane, and it should be not <5 for other Class 1 solvents. A shown in Table V, the minimum Rs achieved using newly developed method was 1.34, resolution between acetonitrile and methylene chloride was 2.66, while minimum S/N ratio was 6.14. Figure 3 shows that the peak shape was good for even pyridine (peak tailing 1.00) and total run time for the method optimized in this study was just 77 min against 188 min for the USP prescribed method. Hence, the optimized method fulfilled all the system suitability requirements of USP for residual solvent analysis. Finally, the experiments conducted to verify the repeatability of the method for peak areas (Supplementary Tables S3) and retention times (Supplementary Table S4) for all solvent peaks confirmed the repeatability of the method. The %RSD values for the peak areas and retention times were <15 and <1, respectively. Conclusion A single injection method for simultaneous analysis of residual solvents of Classes 1 and 2 (except 2C) was successfully developed by employing the DOE approach. The effect of GC parameters on the separation was systematically studied through I-optimal design, and also by taking help of 3D surface response and Derringer desirability plots. The optimized method proposed in this study carries the following significant advantages over USP/EP method or any other method reported in the literature: Separation of all solvents in a single injection with total run time of 77 min. Better resolution of 2.66 in the case of acetonitrile and methylene chloride. Improved S/N ratio of 8.86 for 1,1,1-trichloroethane. Improvement in peak tailing of pyridine (tailing factor = 1.00). Acceptable repeatability of peak areas (%RSDmax = 11.53) and retention times (%RSDmax = 0.45). Use of nitrogen as a carrier gas, instead of costly helium. Supplementary data Supplementary material is available at Journal of Chromatographic Science online. References 1 General chapter . <467> Residual Solvents, in: USP38, The United States Pharmacopeial Convention, Rockville (USA), 2015 . 2 Countryman , S. , Kelly , K. , Harder , B. ; USP method 467: organic volatile impurities testing ; LCGC North America , ( 2005 ); 23 ( SUPPL. 1 ): 65 . 3 Impurities: Guideline for Residual Solvents , Q3C (R6), International Council for Harmonisation (ICH), IFPMA, Geneva (Switzerland), 2016 . 4 Panovska , A.P. , Acevska , J. , Stefkov , G. , Brezovska , K. , Petkovska , R. , Dimitrovska , A. ; Optimization of HS–GC–FID–MS Method for Residual Solvent Profiling in Active Pharmaceutical Ingredients Using DoE ; Journal of Chromatographic Science , ( 2016 ); 54 ( 2 ): 103 – 111 . Google Scholar PubMed 5 Grodowska , K. , Parczewski , A. ; Grouping of residual solvents present in pharmaceuticals using experimental planning and chemometric methods ; Acta Poloniae Pharmaceutica - Drug Research , ( 2013 ); 70 ( 2 ): 191 – 203 . 6 Andrews-Ashley , A.J. , Belsky , J.L. , Joyce , H.R. , Pan , C. , Pappa , H.N. , Wahab , S.Z. ; Static headspace GC–MS detection of residual solvents—a possible simultaneous identification and quantitation method ; LCGC North America , ( 2012 ); 30 ( 1 ): 1 . 7 Wang , C.-Y. , Li , L.-X. , Xie , T.-T. , Lin , J.-F. , Chu , N.-Q. , Tang , L.-C. , et al. . ; Simultaneous determination of the residual contents of 21 harmful organic solvents in leather and leather products by GC-MS/MS coupled with ultrasonic extraction ; Journal of Chinese Mass Spectrometry Society , ( 2016 ); 37 ( 2 ): 121 – 133 . 8 Puranik , S.B. , Sanjay Pai , P.N. , Rao , G.K. ; Determination of organic volatile impurities in herbal formulations and extracts by capillary gas chromatography ; International Journal of Applied Research in Natural Products , ( 2009 ); 2 ( 1 ): 33 – 46 . 9 Cheng , C. , Liu , S. , Mueller , B.J. , Yan , Z. ; A generic static headspace gas chromatography method for determination of residual solvents in drug substance ; Journal of Chromatography A , ( 2010 ); 1217 : 6413 – 6421 . Google Scholar CrossRef Search ADS PubMed 10 Camarasu , C.C. , Mezei-Szuts , M. , Varga , G.B. ; Residual solvents determination in pharmaceutical products by GC-HS and GC-MS-SPME ; Journal of Pharmaceutical and Biomedical Analysis , ( 1998 ); 18 : 623 – 638 . Google Scholar CrossRef Search ADS PubMed 11 Heydari , R. , Azizi , S. ; Rapid Monitoring and Determination of Class 1 Residual Solvents in Pharmaceuticals Using Dispersive Liquid–Liquid Microextraction and Gas Chromatography–Mass Spectrometry ; Journal of Chromatographic Science , ( 2015 ); 53 ( 6 ): 1020 – 1025 . Google Scholar CrossRef Search ADS PubMed 12 Farajzadeh , M.A. , Goushjuii , L. , Djozan , D. , Mohammadi , J.K. ; Dispersive liquid–liquid microextraction combined with gas chromatography for extraction and determination of class 1 residual solvents in pharmaceuticals ; Journal of Separation Science , ( 2015 ); 35 ( 8 ): 1027 – 1035 . Google Scholar CrossRef Search ADS 13 Pérez Pavón , J.L. , Del Nogal Sánchez , M. , Fernández Laespada , M.E. , García Pinto , C. , Moreno Cordero , B. ; Analysis of class 1 residual solvents in pharmaceuticals using headspace-programmed temperature vaporization-fast gas chromatography-mass spectrometry ; Journal of Chromatography A , ( 2007 ); 1141 ( 1 ): 123 – 130 . Google Scholar CrossRef Search ADS PubMed 14 Lakatos , M. ; Measurement of residual solvents in a drug substance by a purge-and-trap method ; Journal of Pharmaceutical and Biomedical Analysis , ( 2008 ); 47 ( 4-5 ): 954 – 957 . Google Scholar CrossRef Search ADS PubMed 15 Myers , R.H. , Khuri , A.I. , Carter , W.H. ; Response surface methodology: 1966–1988 ; Technometrics , ( 1989 ); 31 : 137 – 157 . 16 Candiotim , L.V. , De Zan , M.M. , Cámara , M.S. , Goicoechea , H.C. ; Experimental design and multiple response optimization. Using the desirability function in analytical methods development ; Talanta , ( 2014 ); 124 : 123 – 138 . Google Scholar CrossRef Search ADS PubMed 17 Hardin , R.H. , Sloane , N.J.A. ; A new approach to the construction of optimal designs ; Journal of Statistical Planning and Inference , ( 1993 ); 37 : 339 – 369 . Google Scholar CrossRef Search ADS 18 Goos , P. , Jones , B. ; Optimal design of experiments: a case study approach , 1st ed. Wiley , West Sussex (UK) , ( 2011 ). Google Scholar CrossRef Search ADS 19 Derringer , G.C. ; A balancing act: optimizing a product’s properties ; Quality Progress , ( 1994 ); 27 : 51 – 58 . © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Journal of Chromatographic ScienceOxford University Press

Published: Apr 4, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off