A new matrix-matched calibration strategy for static headspace gas chromatography to enable high throughputs in pharmaceutical quality control laboratories

Dirk Jung *, Jörg-Detlef Kreher , Hans-Ullrich Kratz and Ulrike Michalik
Arevipharma GmbH, Meissner Strasse 35, 01145 Radebeul, Germany. E-mail: dirk.jung@arevipharma.com

Received 25th February 2019 , Accepted 2nd July 2019

First published on 18th July 2019


The analysis of residual solvents or generally, volatile organic substances using static headspace gas chromatography, is often accompanied by matrix effects. The standard addition method (SAM) is the calibration method of choice for circumvention of matrix effects because exactly the same sample matrix is present in both the calibration standard and the sample itself. However, the main disadvantages of the SAM are the large amounts of samples needed, the time-consuming sample preparation steps and long run times. To overcome these drawbacks by full compensation of matrix effects, a new calibration methodology was developed to combine the advantages of the external calibration and the standard addition methodologies. A comparison between the different calibration methods was performed using Bland–Altman plots and F- and t-tests. It was demonstrated that the application of the new calibration methodology leads to validatable methods in a more efficient way. A full validation according to the ICH Q2 guideline covering selectivity, linearity, limits of detection and quantification, system precision, repeatability, intermediate precision, and accuracy was successfully performed.


Introduction

Since the finalization of the guideline Q3C “Impurities: Residual Solvents” by the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) in 1970,1 tremendous efforts have been undertaken in the pharmaceutical industry to establish methods for the control of residual solvents in active pharmaceutical ingredients.2 Based on the recommendations of the ICH guideline, the U.S. Food and Drug Administration (FDA) issued its Guidance for Industry: “Residual Solvents in Drug Products Marketed in the United States”.

General methods for the identification and control of residual solvents in drug substances employing gas chromatography are defined in the European (5.4 and 2.4.24) and in the U.S. Pharmacopeia (〈467〉). Furthermore, as described in a review dealing with residual solvents published by B'Hymer,3 static headspace gas chromatography (HS-GC) is the most widely used technique for residual solvent determination.4 In order to achieve reliable and accurate results using HS-GC, attention should be paid to sample preparation and as well to the quantification method.5 Samples must be prepared in order (a) to maximize the concentration of the analyte (the residual solvents to be tested), and (b) to minimize the influence of the sample matrix (the drug substance).6

The partition coefficient (K)7 is defined as the equilibrium distribution of an analyte between the volume of the sample phase (cs) and the volume of the gas phase (cg):

image file: c9ay00400a-t1.tif

Analytes having low K values move to the gas phase easily which results in high GC responses and thus, low limits of detection can be achieved. The K value depends strongly on the sample matrix. For example, salt concentrations in aqueous preparations decrease the solubility of polar organic volatiles, lower the K value and support their transport into the gas phase. The various techniques and current analytic procedures for residual solvent testing are described and discussed by B'Hymer.3 Although the calibration methodologies are relevant besides the techniques like direct injection or headspace gas chromatography in order to suppress matrix effects, the calibration is not extensively addressed in the recent literature. The calibration approach influences directly the sequence of analysis, the sample preparation and the run times as well as the throughput in the laboratory.

The standard addition method is a generally applicable calibration technique to overcome matrix effects.8–10 Two main types of matrix effects are known. A rotational matrix effect is caused by non-analyte constituents of the test solution leading to a change of the slope of the calibration function, but not of its intercept, while a translational matrix effect is caused by concomitant substances and is therefore independent of the concentration of the analyte (also called ‘background’ or ‘baseline’ interference). Thus, this kind of matrix effect leads to a change of the intercept of a calibration function, but not its slope.11–13

The external calibration method is in general not suitable for compensation of these matrix effects similarly to the standard addition method.14,15

We established a combination of external calibration with the standard addition approach in order to gain highly accurate results by suppression of matrix effects combined with the time-saving evaluation and simplified sample preparation using external calibration. Following these new methodologies, an appropriate validation of the method including the calibration mode together with appropriate statistical treatment of data is provided.

Experimental

Materials and methods

Chemicals used are as follows: oxycodone hydrochloride (manufactured at Arevipharma GmbH, batch DP090-RI030), ethanol (Merck, art.-no. 1.00983), ethyl formate (Merck, art.-no. 8.00891), benzene (Merck, art.-no. 1.01779), toluene (Merck, art.-no. 1.08325), and dimethylformamide (Merck, art.-no. 1.10983).

The concentrations given in mg L−1 refer to the concentration of the residual solvent to be tested in dimethylformamide (solvent for the analytical test). The content of each individual residual solvent in the test sample (drug substance) is given in ppm.

Stock solution 1. Approximately 50 mg (±0.1 mg) of benzene is weighed into a 50 mL volumetric flask, which was previously filled with 30 mL of dimethylformamide. Afterwards, the flask is filled up with dimethyl-formamide to the graduation.
Solution 1. 2.5 mL of stock solution 1 is added into a 50 mL volumetric flask which is filled thereafter with dimethylformamide to the graduation.
Stock solution 2. Approximately 500 mg of ethanol, 250 mg of ethyl formate and 45 mg of toluene are weighed (±0.1 mg) into a 25 mL volumetric flask, which was previously filled with 10 mL of dimethylformamide. Afterwards, 2.0 mL of solution 1 is added and the flask is filled with dimethylformamide to the graduation.
External standard/system suitability test solution. 2.0 mL of stock solution 2 is added into a 20 mL volumetric flask which is filled thereafter with dimethylformamide to the graduation. The dilution ratio is 0.0004 for benzene and 0.1 for ethanol, ethyl formate and toluene, respectively.
Spiked sample. Approximately 100 mg of a test sample to be tested is weighed (±0.1 mg) into a 10 mL headspace vial and 500 μL of external standard is added (standard addition). Immediately, the headspace vial is sealed tightly with a Teflon-coated butyl septum and an aluminium flange cap.

The concentrations of the solvents which are added to the reference sample correspond to about 10[thin space (1/6-em)]000 ppm of ethanol, 5000 ppm of ethyl formate, 900 ppm of toluene and 2 ppm of benzene, all in relation to the weighed amount of oxycodone hydrochloride in the test sample.

Un-spiked sample. Approximately 100 mg of the sample used for the preparation of reference Sample 1 is weighed (±0.1 mg) into a 10 mL headspace vial. 500 μL of dimethylformamide is added, and the headspace vial is sealed tightly with a Teflon-coated butyl septum and an aluminium flange cap.
Test sample. Approximately 100 mg of each sample to be tested is weighed (±0.1 mg) into a 10 mL headspace vial. 500 μL of dimethylformamide is added and the vial is sealed immediately and tightly.

Chromatographic conditions

Gas chromatographic parameters. Gas chromatographic parameters are as follows: gas chromatograph, e.g. Agilent 6890; DB-624, 30 m length, 0.32 mm inner diameter, 1.8 μm thickness of film; carrier gas: nitrogen; flow rate: 1.0 mL min−1 (constant flow); injector temperature: 250 °C; split flow rate: 5.0 mL min−1; FID/250 °C; hydrogen: 40 mL min−1; air: 450 mL min−1; make-up gas (nitrogen): 25 mL min−1; oven temperature: 50 °C (8 min isothermal) to 240 °C (10 K min−1).
Headspace parameters. Headspace parameters are as follows: headspace autosampler, e.g. Agilent G1888; sample temperature: 130 °C; needle temperature: 200 °C; transfer temperature: 220 °C; equilibration time: 30 min; pressurization time: 0.25 min; loop fill time: 0.20 min; loop equilibration time: 0.05 min; injection time: 0.50 min; injections volume: 1 mL.
System suitability test. The resolution (RS) of the peaks of ethanol and ethyl formate in the standard solution has to be at least 1.5, according to the formula:
image file: c9ay00400a-t2.tif
tR1: retention time of ethanol [min], tR2: retention time of ethyl formate [min], W1,h/2: peak width of ethanol at half height [min], W2,h/2: peak width of ethyl formate at half height [min].

Equipment

Gas chromatographs and head space samplers from Agilent with the software Chromeleon 6.80® were used.

Results and discussion

External calibration

The most widely used quantification method is the external calibration. The major advantage is that the necessary samples and references are very easy to prepare and a multitude of samples can be analysed very quickly. The analyte concentration is calculated with respect to the established response using separately prepared standards. Using this method, it is assumed that the response is the same for samples and standards while matrix effects or instrumental drift are not accounted for. This means that signals arise only from the respective analyte. To circumvent these drawbacks an internal standard should be used. A typical sequence of analysis (employing three-point calibration) could for instance be the following:

Solvent; System Suitability Solution; Standard Solution (level 1); Standard Solution (level 2); Standard Solution (level 3); Solvent; Test Sample 1; further Test Samples; System Suitability Solution.

By employing HS-GC, the responses of samples and standards are generally not the same. To overcome this limitation, response factors, which are determined during method development or by the use of an internal standard, are used for quantification. The calculation could for instance be the following:

image file: c9ay00400a-t3.tif
RF: individual response factor, AISTD: the peak area of the internal standard, aSTD: amount of standard used for the preparation of the standard solution, ASTD: the peak area of each individual residual solvent in the standard solution.
image file: c9ay00400a-t4.tif
C: content of each individual residual solvent in the test sample, ATS: peak area of each residual solvent in a test sample, RF: individual response factor, DF: dilution factor (correction of sample and standard preparation), AISTD, TS: the peak area of the internal standard in the test sample, EWTS: weight-in quantity of the test sample.

Standard addition

External calibration in static headspace analysis is not very accurate, even if no recovery standards are used. To compensate these matrix signal suppressions, the standard addition method can be applied. A typical sequence of analysis (employing standard addition) could for instance be the following:

Solvent; System Suitability Solution; Test Sample 1 (un-spiked); Test Sample 1 (spiked); Test Sample 2 (un-spiked); Test Sample 2 (spiked); further un-spiked and spiked Test Samples; System Suitability Solution.

The standard addition method is an applicable calibration technique, to overcome matrix effects. To determine the content of the analyte, the sample is split into two portions and a known amount of a standard solution of analyte is added to one portion. The concentration of the analyte is determined for both the spiked and un-spiked portions. The use of several spiking concentrations in order to check for linearity is not necessary in routine analysis. During the validation it should be demonstrated that the analytical calibration is truly linear.

Following the standard addition approach, the unknown concentration of the analyte in the sample is derived by extrapolation.

The content of each solvent is determined by the standard addition method.

image file: c9ay00400a-t5.tif
C: content of each individual residual solvent in the test sample, AU: peak area of each individual residual solvent (un-spiked sample), AB: peak area of each individual residual solvent (solvent, blank), AS: peak area of each individual residual solvent (spiked sample), EWU: weight-in quantity (un-spiked sample), EWS: weight-in quantity (spiked sample), SASol: spiked amount of each solvent.

The new matrix-matched addition procedure

Analyzing a multitude of samples with the same substance according to the standard addition approach would require far more work for the analyst than is reasonable for obtaining external calibration. Each sample needs to be analyzed un-spiked and spiked with the residual solvents to be tested. In order to enable high throughputs of samples with the same substance as required in routine quality control laboratories, we use a combination of external calibration with the standard addition approach. The first sample of the analytical sequence is spiked with the analytes in order to compensate matrix effects, while the contents of the following samples are calculated by external calibration with reference to the difference of the response obtained from the spiked and from the un-spiked first sample preparation.

A typical sequence of analysis could for instance be the following:

Solvent; System Suitability Solution; Test Sample 1 (spiked); Test Sample 1 (un-spiked); Test Sample 2; further Test Samples; System Suitability Solution.

Calculation of the contents of each residual solvent:

image file: c9ay00400a-t6.tif
ANRL1: normalized peak area of each residual solvent (spiked sample) corrected by the weight-in quantity of the test substance used, ARL1: peak area of each residual solvent (spiked sample), ANRL2: normalized peak area of each residual solvent (un-spiked sample) corrected by the weight-in quantity of the test substance used, ARL2: peak area of each residual solvent (spiked sample), ABL: peak area of each residual solvent (solvent, blank), EWRL1: weight-in quantity of the test substance (spiked sample), EWRL2: weight-in quantity of the test substance (un-spiked sample)
AR = ANRL1 − ANRL2
AR: normalized reference peak area for each residual solvent
image file: c9ay00400a-t7.tif
C: content of each individual residual solvent in the test substance, AT: peak area of each residual solvent (test sample), SASol: spiked amount of each solvent, DF: dilution factor, EWTS: weight-in quantity of the substance to be tested in the test sample.

Consequently, the new matrix-matched addition procedure combines the advantages of the standard addition procedure (matrix effect suppression) with the simplicity of the external calibration procedure. The concentration of the analytes in the test samples is calculated according to the external calibration approach, while the response of the reference peak area is calculated following the standard addition approach.

Validation of the new matrix-matched addition procedure

The method validation relies on the determination of the overall method performance parameters including the calibration model. Therefore, the key aspects are identifying and removing significant sources of errors and uncertainties. The calibration model was used for the determination of residual solvents in oxycodone hydrochloride. It was demonstrated that residual solvents with high limitations like ethanol (5000 ppm) and with very low limitations like benzene (2 ppm) can be determined with high accuracy. The method validation was done by evaluating specificity, limits of detection and quantification, linearity, accuracy, repeatability, and method precision of residual solvents as directed in the International Conference on Harmonization (ICH) guideline Q2B “Validation of Analytical Procedures: Methodology”.16
Selectivity. The peaks of the residual solvents (ethanol, ethyl formate, benzene and toluene) to be tested are separated from the peak of the solvent dimethylformamide and from each other.
Linearity. The linear relation between the concentration and peak area was evaluated for ethanol, ethyl formate, benzene and toluene.

Eleven spiked solutions of ethanol were prepared in the range from 10 to 20[thin space (1/6-em)]021 ppm. For ethyl formate and toluene twelve spiked solutions were prepared in the range between 2 and 10[thin space (1/6-em)]123 ppm; and 2 and 1813, respectively. For benzene eight spiked solutions were prepared in the range between 0.24 and 4.02 ppm.

The linearity was evaluated in the presence of oxycodone hydrochloride, each sample was prepared and analyzed in triplicate. The regression lines were obtained from the responses (peak areas) of the residual solvents versus their concentrations in ppm.

Therefore, approximately 100.0 mg (±0.1 mg) of solvent-free oxycodone hydrochloride was weighed into a 10 mL headspace vial and 500 μL of spiked solution was added.

The linearity has been confirmed for ethanol in the range from 10 ppm to 20[thin space (1/6-em)]021 ppm, for ethyl formate from 2 ppm to 10[thin space (1/6-em)]123 ppm, for benzene from 0.24 ppm to 4.02 ppm and for toluene from 2 ppm to 1813 ppm, with respect to the amount of oxycodone hydrochloride in the test sample. The summary of the assessment of linearity is given in Table 1.

Table 1 Assessment of linearity
Parameter Ethanol Ethyl formate Benzene Toluene
Coefficient of correlation 1.000 1.000 0.999 1.000
Normal distribution of residuals Complies Complies Complies Complies
y-intercept −0.2910 0.0833 −0.0064 0.1005
Slope of the regression line 0.0181 0.0252 0.0815 0.0387
Residual sum of squares 141.2223 29.5401 0.0005 4.6244


Limits of detection and quantification. The limits of detection and quantification of each residual solvent were calculated from the calibration regression line based on the test for linearity using the four lowest concentration levels. The limits of detection and quantification are given in Table 2.
Table 2 Limits of detection (LoD) and quantification (LoQ); given in ppm with reference to the test sample
Ethanol Ethyl formate Benzene Toluene
LoD [ppm] 7 2 0.1 9
LoQ [ppm] 20 6 0.4 26


System precision. To determine the precision of the system, a reference sample was prepared and analyzed six times. The peak areas of the components were evaluated and averaged. The relative standard deviation (RSD) is evaluated. The RSD of six replicates amounts to 0.92% for ethanol, 0.66% for ethyl formate, 0.54% for benzene, and to 0.72% for toluene. A RSD of not more than 1% for the peak areas of all residual solvents confirms the high degree of system precision.
Repeatability. The repeatability was proven by preparing and analyzing six spiked samples of oxycodone hydrochloride with the residual solvents to be tested. The test sample was spiked with ethanol, ethyl formate, benzene and toluene; the concentrations of the components were calculated and averaged. The relative standard deviation (RSD) is evaluated.

The RSD of six sample preparations amounts to 1.21% for ethanol, 1.45% for ethyl formate, 2.69% for benzene, and to 1.14% for toluene. A RSD of not more than 3% for the concentrations of six sample preparations of all residual solvents confirms the high degree of repeatability.

Intermediate precision. The intermediate precision of the method was proven by investigations with variation of time, analyst and equipment. The test sample was spiked with ethanol, ethyl formate, benzene and toluene to achieve adequate contents of these residual solvents for quantification. Each analyst carried out six replicated determinations from six sample preparations. The differences of the average concentrations of both series amount to 15 ppm for ethanol, 8 ppm for ethyl formate, 0.1 ppm for benzene, and to 4 ppm for toluene, respectively. These insignificant differences confirm the excellent intermediate precision.
Accuracy. The accuracy of the method was evaluated in terms of recovery for ethanol in the range from 50 ppm to 20[thin space (1/6-em)]021 ppm, for ethyl formate in the range from 52 ppm to 10[thin space (1/6-em)]123 ppm, for benzene in the range from 0.61 ppm to 4.02 ppm and for toluene in the range from 51 ppm to 1813 ppm, with respect to the amount of oxycodone hydrochloride in the test sample. The accuracy was determined in triplicate (three preparations).
image file: c9ay00400a-t8.tif
CR: recovery in %, MC: measured concentration of each individual residual solvent in the test substance in ppm, TC: theoretical concentration of each individual residual solvent in the test substance in ppm.

The mean recoveries were calculated to be 97.9% for ethanol, 101.5% for ethyl formate, 95.1% for benzene and 102.6% for toluene (Tables 3–6) showing a highly accurate method.

Table 3 Assessment of accuracy: theoretical (TC) and measured concentration (MC) as well as the calculated recovery (CR) for ethanol
TC [ppm] MC [ppm] CR [%]
20[thin space (1/6-em)]021.2 19[thin space (1/6-em)]808.1 98.9
20[thin space (1/6-em)]021.2 20[thin space (1/6-em)]402.5 101.9
20[thin space (1/6-em)]021.2 20[thin space (1/6-em)]536.3 102.6
10[thin space (1/6-em)]010.6 10[thin space (1/6-em)]143.9 101.3
10[thin space (1/6-em)]010.6 10[thin space (1/6-em)]066.9 100.6
10[thin space (1/6-em)]010.6 9990.0 99.8
5005.3 5103.0 102.0
5005.3 5059.7 101.1
5005.3 4867.3 97.2
2002.1 2079.3 103.9
2002.1 2043.2 102.1
2002.1 2075.4 103.7
1026.4 1018.6 99.2
1026.4 1022.5 99.6
1026.4 1019.7 99.3
513.2 502.3 97.9
513.2 487.0 94.9
513.2 494.8 96.4
410.6 386.6 94.2
410.6 382.2 93.1
410.6 389.1 94.8
256.6 233.3 90.9
256.6 239.1 93.2
256.6 239.3 93.2
49.9 45.9 92.0
49.9 48.5 97.1
49.9 46.5 93.2


Table 4 Assessment of accuracy: theoretical (TC) and measured concentration (MC) as well as the calculated recovery (CR) for ethyl formate
TC [ppm] MC [ppm] CR [%]
10[thin space (1/6-em)]122.9 10[thin space (1/6-em)]037.1 99.2
10[thin space (1/6-em)]122.9 10[thin space (1/6-em)]216.4 100.9
10[thin space (1/6-em)]122.9 10[thin space (1/6-em)]253.9 101.3
5061.4 5129.0 101.3
5061.4 5061.4 100.0
5061.4 5066.1 100.1
2530.7 2537.4 100.3
2530.7 2504.8 99.0
2530.7 2493.8 98.5
1012.3 1040.2 102.8
1012.3 1029.6 101.7
1012.3 1047.9 103.5
503.2 521.2 103.6
503.2 536.5 106.6
503.2 527.9 104.9
251.6 260.5 103.5
251.6 259.0 103.0
251.6 261.6 104.0
201.3 205.5 102.1
201.3 206.5 102.6
201.3 210.3 104.5
125.8 126.0 100.2
125.8 129.8 103.2
125.8 128.9 102.4
51.8 51.0 98.5
51.8 49.1 94.8
51.8 50.1 96.8


Table 5 Assessment of accuracy: theoretical (TC) and measured concentration (MC) as well as the calculated recovery (CR) for toluene
TC [ppm] MC [ppm] CR [%]
1812.8 1801.4 99.4
1812.8 1843.9 101.7
1812.8 1863.1 102.8
906.4 917.2 101.2
906.4 911.5 100.6
906.4 904.7 99.8
453.2 468.2 103.3
453.2 455.3 100.5
453.2 452.2 99.8
181.3 190.9 105.3
181.3 187.1 103.2
181.3 189.5 104.5
313.4 327.3 104.4
313.4 326.6 104.2
313.4 327.6 104.5
156.7 162.8 103.9
156.7 159.4 101.7
156.7 162.5 103.7
125.4 129.7 103.5
125.4 128.9 102.8
125.4 130.5 104.1
78.4 78.5 100.2
78.4 80.6 102.9
78.4 81.6 104.1
50.9 51.8 101.9
50.9 52.7 103.7
50.9 52.3 102.8


Table 6 Assessment of accuracy: theoretical (TC) and measured concentration (MC) as well as the calculated recovery (CR) for benzene
TC [ppm] MC [ppm] CR [%]
4.02 3.92 97.6
4.02 3.99 99.4
4.02 4.08 101.4
2.01 2.04 101.6
2.01 2.03 100.9
2.01 2.03 100.9
1.00 0.98 97.5
1.00 0.97 96.3
1.00 0.93 93.1
0.61 0.50 82.5
0.61 0.49 81.6
0.61 0.54 88.8


Comparison of the different calibration methods. The accuracy of the new matrix-matched addition procedure is compared with the external calibration. A comparison with the standard addition method is not shown as the matrix effect is also suppressed and the accuracy is comparable. In order to evaluate the significance of the differences between both calibration methods, the Bland–Altman plot (BA plot) is used.17,18 In the following BA plots the differences between the recoveries (test for accuracy) of both calibration methods are plotted against the averages of the recoveries of both calibrations. The blue points show the data, the horizontal lines represent the mean difference, and the dotted lines represent the limits of agreement, which are defined as the mean difference plus and minus 1.96 times the standard deviation of the differences. The BA plot analysis is a simple way to evaluate a bias between two methods (Fig. 1).
image file: c9ay00400a-f1.tif
Fig. 1 Bland–Altman plot of the differences of recoveries for the solvent ethanol between the new matrix-matched addition procedure (data from Table 3) and external calibration versus the mean of the recoveries of both calibration methods.

The shown Bland–Altman plots clearly indicate that there is constant variability for the solvent ethanol between the results for accuracy calculated according to the new matrix-matched addition procedure and according to external calibration. For the other solvents, viz. ethyl formate, benzene, and toluene, a proportional constant error is obvious (Fig. 2).


image file: c9ay00400a-f2.tif
Fig. 2 Bland–Altman plot of the differences of recoveries for the solvent ethyl formate between the new matrix-matched addition procedure (data from Table 4) and external calibration versus the mean of the recoveries of both calibration methods.

In addition, the F- test and the t-tests are evaluated for the recoveries obtained for both calibration approaches and the results are presented in Table 7. These tests show in addition to the Bland–Altman plots that both sets of recoveries are dissimilar by comparing their standard deviations and means respectively. The differences are significant (Fig. 3).19

Table 7 Comparison of the recoveries for the solvents to be tested applying the new matrix-matched addition procedure (data from Tables 3–6) and external calibration by means of F- and t-tests
Ethanol Ethyl formate Benzene Toluene
F value found 1.115 0.943 0.743 0.780
F critical value 1.955 0.511 0.336 0.511
t value found −13.317 −198.606 −33.353 −305.031
t critical value 1.706 1.706 1.796 1.706



image file: c9ay00400a-f3.tif
Fig. 3 Bland–Altman plot of the differences of recoveries for the solvent benzene between the new matrix-matched addition procedure (data from Table 6) and external calibration versus the mean of the recoveries of both calibration methods.

The F- and the t-test as well the presented Bland–Altman plots show that the matrix effect prevents an accurate calibration by means of external calibration compared to the matrix-matched addition procedure (Fig. 4).


image file: c9ay00400a-f4.tif
Fig. 4 Bland–Altman plot of the differences of recoveries for the solvent toluene between the new matrix-matched addition procedure (data from Table 5) and external calibration versus the mean of the recoveries of both calibration methods.

An accurate quantification is ensured by the standard addition method. However, in this case, the time of analysis for a sequence of several samples is very long compared to the external calibration or to the matrix-matched addition procedure.

Besides the accuracy, the time for analysis is of high interest for quality control laboratories within the pharmaceutical industry. The processing times following external calibration, standard addition and the matrix-matched addition procedure are contrasted in Table 8. For the presented residual solvent determination in Oxycodone HCl, the duration of one analysis is 28 minutes. Under the conditions as described in Table 8, the period of analysis for e.g. 10 test batches of Oxycodone HCl is 12.1 hours following the external calibration method, 20.5 hours following the standard addition method, and 14 hours according to the matrix-matched addition procedure.

Table 8 Comparison of the period for analysis following the three different calibration approaches (d: duration of one analysis, n: number of samples in one sequence)
Calibration method Sequence of analysis Period of analysis
External calibration Solvent d(2n + 6)
System suitability solution standard solution (3 times)
Solvent
n test samples (2 times)
System suitability solution
Standard addition Solvent d(4n + 4)
System suitability solution
Solvent
n test samples (un-spiked 2 times, spiked 2 times)
System suitability solution
Matrix-matched addition procedure Solvent d(2n + 10)
System suitability solution
First test sample (un-spiked 3 times, spiked 3 times, for calibration)
Solvent
n test samples (2 times)
System suitability solution


Conclusions

The demonstrated calibration strategy offers the unique advantage for the determination of residual solvents by compensating matrix effects in an accurate manner. Furthermore, time consuming sample preparation, cumbersome evaluation (sample by sample) and long run times relating to the standard addition approach can be circumvented in order to increase the operational capacity in pharmaceutical quality control laboratories. The validation study including residual solvents with high (1% for ethanol) and low (2 ppm for benzene) specification limits provides evidence for the successful implementation of the calibration strategy with high precision and accuracy. The key performance indicators, accuracy and time, are compared in Table 9. It is demonstrated that the matrix-matched addition procedure combines high accuracy with high analysis speed compared to external calibration and standard addition. All further requirements for method validation besides a high degree of accuracy are of course fulfilled.
Table 9 Comparison of the key performance indicators of the three different calibration approaches
Calibration method Accuracy Analysis speed
External calibration Low High
Standard addition High Low
Matrix-matched addition procedure High High


Conflicts of interest

There are no conflicts to declare.

Notes and references

  1. International Conference on Harmonisation of Technical Requirements for Registrations of Pharmaceuticals for Human Use (ICH), Harmonised Tripartite Guideline on Impurities: Residual Solvents (Q3C (R6)), 2016 Search PubMed .
  2. S. Klick and A. Sköld, J. Pharm. Biomed. Anal., 2004, 36, 401 CrossRef CAS .
  3. C. B'Hymer, Pharm. Res., 2003, 20, 337 CrossRef .
  4. N. H. Snow and G. C. Slack, Trends Anal. Chem., 2002, 21, 608 CrossRef CAS .
  5. G. C. Slack, N. H. Snow and D. Kou, Extraction of volatile organic compounds from solids and liquids, in Sample Preparation Techniques in Analytical Chemistry, ed. S. Mitra, John Wiley & Sons, Canada, 2003, p. 183 Search PubMed .
  6. M. Markelov and J. P. Guzowski Jr, Anal. Chim. Acta, 2004, 276, 235 CrossRef .
  7. M. Markelov and O. A. Bershevits, Anal. Chim. Acta, 2001, 432, 213 CrossRef CAS .
  8. M. Gergov, T. Nenonen, I. Ojanperä and R. A. Ketola, J. Anal. Toxicol., 2015, 39, 359 CrossRef CAS .
  9. D. L. Massart, B. G. M. Vandeginste, L. M. C. Buydens, S. De Jong, P. J. Lewi and J. Smeyers-Verbeke, Standard addition method, in Handbook of Chemometrics and Qualimetrics, Part A, ed. B. G. M. Vandeginste and S. C. Rutan, Elsevier, Amsterdam, The Netherlands, 1997, p. 207 Search PubMed .
  10. I. G. Zenkevich and I. O. Klimova, J. Anal. Chem., 2006, 61, 967 CrossRef CAS .
  11. S. L. R. Ellison and M. Thompson, Analyst, 2008, 133, 992 RSC .
  12. M. Bader, J. Chem. Educ., 1980, 57, 703 CrossRef CAS .
  13. R. Otero, G. Carrera, J. F. Dulsat, J. L. Fábregas and J. Claramunt, J. Chromatogr. A, 2004, 1057, 193 CrossRef CAS .
  14. K. Danzer, M. Otto and L. A. Currie, Pure Appl. Chem., 2004, 76, 1215 CAS .
  15. K. Danzer, M. Otto and L. A. Currie, Pure Appl. Chem., 1998, 70, 993 CAS .
  16. International Conference on Harmonisation of Technical Requirements for Registrations of Pharmaceuticals for Human Use (ICH), Harmonised Tripartite Guideline on Validation of Analytical Procedures: Text and Methodology (Q2(R1)), 1996 Search PubMed .
  17. D. Giavarina, Biochem. Med., 2015, 2, 141 CrossRef .
  18. J. M. Bland and D. G. Altman, Stat. Methods Med. Res., 1999, 8, 135 CrossRef CAS .
  19. S. Belouafa, F. Habti, S. Benhar, B. Belafkih, S. Tayane, S. Hamdouch, A. Bennamara and A. Abourriche, International Journal of Metrology and Quality Engineering, 2017, 8, 9 CrossRef .

This journal is © The Royal Society of Chemistry 2019