Development and Validation of an HPLC-UV Method for the Determination Bis(2-ethylhexyl) Phthalate Ester in Alcoholic Beverages

Abstract

An HPLC method with UV detector was developed for the determination of DEHP phthalate ester in the alcoholic beverage “Ouzo”. Phthalate esters are added to plastic packaging for food and beverages to increase flexibility, transparency, strength, and longevity. When these substances come into contact with food or beverages, they can lead to the migration of phthalate residues into the product. This paper presents a two-step process involving extraction of the sample with hexane and separation of the phthalates by HPLC. The method was validated for specificity, linearity, limit of quantification, accuracy, precision, range, and ruggedness. The linear range is 0.3–1.5 mg/L DEHP, with a lower limit of quantification of 0.06 mg/L. The precision study showed acceptable RSD values, and the working range is 0.3 to 1.5 mg/L DEHP. The relative standard uncertainty of DEHP determination in Ouzo was ±8%. The results show that the in-house method is suitable, reliable, and fit-for-purpose.

1. Introduction

Phthalates esters are essential components of plastics used in various food and beverage packaging, generally called plasticizers. They are added to plastics to increase their flexibility, transparency, strength, and longevity. When these substances come into contact with food or beverages, they can lead to the migration of phthalate residues into the product [1]. The amount of migration depends on several factors, such as the type and composition of the food or beverage, the total contact time, the temperature, the concentration of phthalates in the packaging material, etc. The most commonly used phthalate is “Bis(2-ethylhexyl) benzene-1,2-dicarboxylate” (IUPAC name) which is abbreviated as “DEHP”. Studies have shown that phthalates are toxic and dangerous for the human body. For this reason, European legislation for some of the phthalates has established maximum limits in food. Specifically, for DEHP the limit is 3.5 mg/kg of food [2].

Alcoholic beverages, especially those with a high ethyl alcohol content, are prohibited from being packaged in plastic containers. However, there are times when alcoholic beverages show high concentrations of phthalates and especially DEHP. The phenomenon may be due to the plastics used in the production of the product and the raw materials. Therefore, in order to ensure the health of consumers and the quality of the product, in this study a method was developed and validated to measure the specific phthalate ester in “Ouzo” which is a national Greek anise-flavored aperitif. The alcoholic beverage “Ouzo” in the European regulation 2019/787 concerning all alcoholic beverages is defined as distilled anis. Distilled anis is anis which contains alcohol distilled in the presence of the seeds: star anise (Illicium verum Hook f.), anise (Pimpinella anisum L.), fennel (Foeniculum vulgare Mill.), or any other plant which contains the same principal aromatic constituent, provided such alcohol constitutes at least 20% of the alcoholic strength of the distilled anis. The minimum alcoholic strength by volume of distilled anis shall be 35% [3]. “Ouzo” is produced exclusively in Greece, where it is widely consumed, and it is also a product that is exported in large quantities to all markets of the world.

Up to now, in the literature, some methods have been developed for the determination of phthalates esters in various alcoholic beverages. Specifically, in 1995 Leibowitz et al. developed a method for the determination of phthalate esters in grain neutral spirits and vodka with no sample preparation or sample enrichment using liquid chromatography (LC) and gas chromatography coupled with mass spectrometry (GC/MS) [1]. However, this method cannot be used for alcoholic beverages containing sugar, such as “Ouzo”. Additionally, by Del Carlo et al. in 2008 [4] a method was developed for the determination of phthalate esters in wine using solid-phase extraction and gas chromatography–mass spectrometry, and by Russo et al. in 2014 [5] a corresponding method was developed in light alcoholic drinks and soft drinks by XAD-2 adsorbent and gas chromatography coupled with ion trap-mass spectrometry detection.

All this research has contributed to the study of plasticizers in beverages. With this study, we come to report a new method which is fast, with simple sample preparation and based on equipment that most quality control laboratories in distilleries and wineries have. In addition, with the method we propose in this study, we can determine phthalate esters in alcoholic beverages with high ethyl alcohol concentration > 35% vol and at the same time a high sugar concentration of 20–40 g/L. In contrast to the above methods which involve either low alcohol drinks or drinks without added sugar, our method has been fully validated; the uncertainty has been calculated with Nordtest technique and applied to “Ouzo” samples.

2. Materials and Methods

2.1. Chemical and Reagents

All chemicals and solvents were of analytical reagent grade and used without further purification, Including hexane, acetonitrile, methanol, ultrapure water, isopropanol, and absolute ethanol; all these solvents are HPLC grade from Panreac and DEHP (>99% purity) from Sigma-Aldrich. They were prepared 3 independent replicates for each of the concentrations of 0.3, 0.7, 1.1, and 1.5 mg/L DEHP in matrix 40% v/v ethanol from an initial solution of 1000 mg/L DEHP in matrix methanol. Additionally, a solution of acetonitrile with 1% v/v methanol was prepared, which was used as the mobile phase in the liquid chromatograph [1].

2.2. Instrumentation

Instrumentation included the following: separating extraction funnels 50 mL, Rotary Evaporator of Steroglass model Strike 300, spherical flask of 50 mL with grinding and special thread to adapt to the rotary evaporator, pipette 500–5000 μL, volumetric flasks class A, High-Performance Liquid Chromatograph with UV detector of Agilent Technologies model 1200 series, Agilent Technologies Chemstation software with which peak heights or areas can be automatically measured and with which the data can be processed via PC, MZ-Analysentechnik GmbH column with the following characteristics: HPLC column 250 × 4.0 mm, Spherisord ODS-2 C18 5 μm, and analytical balance 0.1 mg by SHIMADZU model AUX220. The statistical analysis for the analytical responses and validation data was evaluated with Microsoft Excel 2016 and Minitab 17.

2.3. Procedure

The method includes two phases. Phase A included extraction of the sample with hexane solvent to remove interferences and move the phthalate esters in the organic phase and evaporation of the solvent.

Specifically, a pipette of 10 mL of hexane and 10 mL of sample were transferred into the separating extraction funnel. This is followed by vigorous shaking of the funnel, and the substances were allowed to settle to separate the organic from the aqueous phase completely. Through the stopcock of the separating funnel, we discarded the lower phase (aqueous) and collected the organic phase in the 50 mL spherical flask. The spherical flask is attached to the rotary evaporator and evaporated to dryness. We removed the spherical flask from the evaporator and added 10 mL of acetonitrile.

Phase B included the separation of phthalates by HPLC High Performance Liquid Chromatography and determination with a UV Detector.

Chromatographic conditions were the following flow: 1.000 mL/min, mobile phases ultrapure water and acetonitrile with 1% methanol. The column temperature was kept at 25 °C, whereas the detector’s wavelength was adjusted to 225 nm [1]. The gradient was 37.5% of ultrapure water, decreased to 0% at 10 min, kept at 0 % for 10 min, and increased back to 37.5% within 2 min.

Before starting the analysis, the UV detector response is calibrated and checked for linearity by analyzing each of the standard solutions in succession. From the integration of the peak area of each injection, a plot of concentration in mg/L versus area is constructed. A straight line with an R² coefficient > 0.999000 should be taken as a graph. All the above calculations and diagrams are conducted with the Chemstation software. Next is the analysis of the sample; the solution we prepared in part A is injected into the liquid chromatograph with a syringe through the instrument’s inlet. The result given by the liquid chromatograph is expressed in mg/L with one decimal place.

2.4. Method Validation

2.4.1. Specificity

The elution time was found, and the separation capacity for DEHP was checked. Specifically, a standard solution of 0.7 mg/L DEHP in a matrix of ethyl alcohol 40% v/v, a pure solvent solution of ethyl alcohol 40% v/v, and a real sample of Ouzo were measured following all the analytical procedure described.

Ouzo matrix effect check was conducted. The main test of the specificity of the method should be to demonstrate that the peak area (signal), at the elution time, is due solely to DEHP and not to some other substance present in the matrix of the actual sample [6]. To check the effect of the matrix of a real “ouzo” sample, the following tests were performed:

• In a real sample of Ouzo, 4 independent spikes of DEHP were made at concentrations of 0.3, 0.7, 1.1, and 1.5 mg/L and measured following the entire analytical procedure. Additionally, unspiked “ouzo” was measured. The spike was conducted by adding a small volume of a dense standard to avoid dilution of the sample.
• In pure solvent ethyl alcohol 40% v/v, 4 independent spikes of DEHP were made at concentrations of 0.3, 0.7, 1.1, and 1.5 mg/L and were measured following the entire analytical procedure. The pure solvent without spike was also measured.

2.4.2. Linearity

In the unknown samples to be analyzed, the average expected DEHP concentration is 0.7 mg/L. To check the linearity, 12 DEHP standard solutions (4 concentrations × 3 independent replicates) were measured, whose concentrations ranged from 40 to 200% of the expected [6]. Specifically, 3 independent replicates were measured for each of the concentrations: 0.3, 0.7, 1.1, and 1.5 mg/L DEHP in a 40% v/v ethyl alcohol matrix. The 12 standard solutions were measured directly on the liquid chromatograph skipping the extraction step.

2.4.3. Limit of Detection (LOD) and Limit of Quantification (LOQ)

A real Ouzo sample was used to determine the LOD and LOQ, whose concentration is close to the LOQ of the method (<0.1 mg/L DEHP); the real Ouzo sample was measured 10 times following the whole procedure [7,8].

2.4.4. Accuracy

Since the method was developed in-house, the accuracy tests have been conducted extensively to cover the entire expected concentration range. At the same time, it should be confirmed if there is a “proportional” and/or “standard” systematic error. For reasons of economy (time and resources) the real Ouzo sample and the spiked samples used in the estimation of the specificity of the method were used [9]. Additionally, calibration standards and spikes were prepared from a different analyte stock solution [10]. The above samples were measured, and a correlation diagram was constructed with the results (x-axis: known concentration, y-axis: results from the method being validated) [9].

2.4.5. Precision

Because in this particular method the analyst who applies it is one, and the organization is unique, a technique in which only the day factor is varied was used to check the fidelity. Eight days over a two-week period were randomly selected, and two repetitions were performed on each of the eight days. The repetitions were conducted in order to simultaneously calculate the intermediate precision and the repeatability of the method [7,11]. Additionally, for the measurements, a real sample of Ouzo was used, which has been spiked with DEHP to have a concentration of almost 1.0 mg/L (which is the midpoint of the calibration curve).

2.4.6. Range

Linearity and precision studies were used to determine the range of the method [7].

2.4.7. Ruggedness

To assess the ruggedness of the DEHP determination method in Ouzo, the factors and levels shown in

Table 1. Factors and levels studied during the evaluation of the ruggedness of the DEHP determination method in Ouzo.

Factor		Levels
		Regular Value	Deliberate Change
		1	2
A	Concentration of methanol solution in acetonitrile	1.00%	0.99%
Β	Gradient of mobile phase solvents: A: Ultrapure water B: Acetonitrile with 1% methanol	−3 min A: 37.5%, B: 62.5% −10 min A: 0%, B: 100% −20 min A: 0%, B: 100% −22 min A: 37.5%, B: 62.5%	−3 min A: 36.5%, B: 63.5% −10 min A: 1%, B: 99% −20 min A: 1%, B: 99% −22 min A: 36.5%, B: 63.5%
C	Mobile phase flow rate	1.00 mL/min	0.98 mL/min
D	Column oven temperature	25.0 °C	25.5 °C
E	Wave length	225 nm	226 nm

were studied. The order of magnitude of the deliberate changes applied to the factors was not random; they are either the upper or the lower end of the limit accuracy.

Accuracy limits are defined by the instrument manufacturer or calibration certificate. The experimental design used is the one shown in

Table 2. 7-parameter (A, B, C, D, E, F, G) two-level (1, 2) experimental design.

Experiment	Parameter
	A	B	C	D	E	F	G
1	1	1	1	1	1	1	1
2	1	1	1	2	2	2	2
3	1	2	2	1	1	2	2
4	1	2	2	2	2	1	1
5	2	1	2	1	2	1	2
6	2	1	2	2	1	2	1
7	2	2	1	1	2	2	1
8	2	2	1	2	1	1	2

, with which the main effects of 7 factors with two levels each can be studied by running only 8 experiments. Because in this particular case the 5 factors of Table 1 will be studied, factors F and G will be considered “imaginary”, as if they do not exist [7,12]. The same sample was measured twice for each experiment.

2.5. Assessment of Uncertainty

To estimate the uncertainty, we used the data that the laboratory has from a control sample, which has been measured 36 times over a period of 1 year, and the measurements are placed on a control chart to check if the process is within statistical control. In this way, it is established, in routine conditions, if the measurements we obtain from the DEHP determination method in Ouzo are reliable at the level of intermediate precision. Additionally, in order to fully cover the internal quality control of the method, the laboratory, in addition to the measurements of the above control sample, periodically determines the recovery of spiked real Ouzo samples. In this way, the accuracy of the method is systematically checked. In recovery checks, it is important that the calibration standards and spikes are prepared from a different analyte stock solution.

As long as the data are available, from the fixed control sample and the recovery, it was preferred to estimate the uncertainty according to the NORDTEST methodology, which is simpler to calculate and considered as reliable as that of EURACHEM.

To estimate the uncertainty according to NORDTEST Equation (1) was used:

$${\mathrm{u}}_{\mathrm{C}}=\sqrt{\mathrm{u}{\left({\mathrm{R}}_{\mathrm{w}}\right)}^2+{\mathrm{u}}_{\mathrm{bias},\mathrm{N}}{}^2}$$

(1)

where u_bias,N: the relative standard uncertainty by NORDTEST of the systematic errors (method precision). The recovery data were used to calculate it. u_(Rw): the relative standard uncertainty from intra-laboratory reproducibility. For its calculation, the data of the control sample were used [13,14].

3. Results and Discussion

3.1. Specificity

3.1.1. Elution Time and Separation Capacity of DEHP

Figure 1 shows the chromatogram of the standard solution, in which DEHP elutes at retention time of 14.3 min.

In the chromatogram, Figure 2, of the real Ouzo sample, a peak of the same shape is observed at the same time, 14.3 min. In this case it was considered unnecessary to calculate the separation capacity because the DEHP peak is clear and there are no other peaks around it.

Additionally, it is worth noting that in the pure solvent chromatogram, Figure 3, no peak is observed at the elution time of DEHP. Therefore, we conclude that satisfactory separation of the DEHP peak in Ouzo samples is achieved.

3.1.2. Ouzo Matrix Effect Check

The results from the tests are described in Section 2.4.1. Based on these, two least squares curves are constructed in the same diagram, one with the “ouzo” matrix and the other with the pure 40% ethyl alcohol solvent. The curves are shown in Figure 4.

Observing the two curves in Figure 4 it appears that they are parallel. However, the documentation of the specificity should be conducted statistically by the regression analysis for the two matrices and the statistical test of the equality of the slopes of the curves. The acceptance criterion is that the slope value of one curve is included in the 95% confidence interval of the slope of the other curve [15]. From the regression analysis for the two matrices in Excel, the confidence intervals (CI) of the slopes are obtained; for Ouzo matrix, slope b = 136.77:CI 95% = (129.45; 144.09), whereas for matrix ethyl alcohol, slope b = 138.17:CI 95% = (129.75; 146.59)

From CI it can be seen that the slope value of each curve is included in the CI of the slope of the other curve. Therefore, the slopes of the curves are statistically equal for a significance level of 95%, and it is safely concluded that the impurities of the Ouzo matrix do not affect the determination of DEHP with the specific analytical method.

3.2. Linearity

The results of the measurements of the 12 standard solutions are shown in the A’ part of

Table 3. Results of measurements to estimate the linearity of the method.

A’ part	Concentration Level	Duplicate Per Level	DEHP Concentration, mg/L	Area
	1	1	0.3	34.65
	1	2	0.3	41.72
	1	3	0.3	39.98
	2	1	0.7	96.47
	2	2	0.7	99.12
	2	3	0.7	95.86
	3	1	1.1	148.80
	3	2	1.1	153.69
	3	3	1.1	150.97
	4	1	1.5	205.14
	4	2	1.5	207.12
	4	3	1.5	203.34
B’ part	SUMMARY
	DEHP Concentration, mg/L	Mean Value Area	Standard Deviation, s	RSD,%
	0.3	38.78	3.684	9.498
	0.7	97.15	1.733	1.784
	1.1	151.15	2.450	1.621
	1.5	205.20	1.891	0.921

, while, in the B’ part of the same table, a summary is made by calculating the average value, the standard deviation and the RSD of the measurements per concentration level. The RSD values calculated in Table 3 satisfy the acceptable values predicted based on the concentration level of the analyte [16,17].

3.2.1. Statistical Data Checks

For the selection of the appropriate regression model, tests of normality (Grubbs test) and homoscedasticity (Bartlett test) of the data were performed. Furthermore, based on the fact that the matrix does not have any significant impact as confirmed in Section 3.1.2, we can confidently conclude that the simple linear regression model is appropriate [18].

3.2.2. Simple Linear Regression

Based on the data in Table 3, the diagram is constructed: area (y-axis) in terms of DEHP concentration (x-axis). Additionally, applying the least squares method gives the equation of the regression curve, as shown in Figure 5.

3.2.3. Linearity Checks (Linear Range)

Residual-concentration chart:

The plot of residuals with respect to DEHP concentration is constructed, Figure 6. The points in this plot are randomly distributed around zero and do not show any particular pattern [19]. Therefore, no deviation from linearity is seen.

Sensitivity plot:

For its construction, the response factor (area/concentration) was first calculated. In the sensitivity diagram, Figure 7, it is observed that there is no point outside the interval ± 5% from the average of the response factor [11]. This confirms the linear range of the method for the range of standard solutions used 0.3–1.5 mg/L DEHP.

3.3. Limit of Detection (LOD) and Limit of Quantification (LOQ)

The results, peak area given by the liquid chromatograph for the 10 measurements of a real Ouzo sample, and the calculation of LOD and LOQ from Equations (2) and (3), respectively, are shown in

Table 4. Accuracy of the method.

Xi,sp	Known Concentration, μi (μi = xi, sp + x0, x0 = 0.62)	Method under Validation, xi
0.3	0.92	0.94
0.7	1.32	1.35
1.1	1.75	1.70
1.5	2.12	2.14

.

$$\mathrm{LOD}=3.3\frac{\mathsf{\sigma }}{\mathrm{b}}$$

(2)

$$\mathrm{LOD}=10\frac{\mathsf{\sigma }}{\mathrm{b}}$$

(3)

where σ: the standard deviation of the sample replicates used to determine LOD and LOQ, and b: the slope of the calibration curve [8,20].

3.4. Accuracy

The measurement results for each spiked sample are shown in Table 4; the real Ouzo sample without spike was measured three times and gave an average value, x₀ = 0.62 mg/L DEHP.

Then, the correlation diagram, Figure 8, is constructed, in which the experimental values obtained from the method under validation, x_i (y-axis), are presented in terms of the corresponding known concentrations of the samples, μ_i (x-axis). Additionally, the regression line is plotted using the least squares method. Finally, with the help of Excel we obtain the regression analysis table, a part of which is shown in Figure 8.

To check the accuracy of the method, the following hypothesis tests were performed:

• For the slope b = 0.9875: H₀: b = 1 vs. H_A: b ≠ 1, to ascertain whether the value of b is statistically equal to monad, the 95% confidence interval of b should contain monad. From the regression analysis table, Figure 8 (below), CI 95% = (0.863; 1.112). Since CI includes monad we accept the null hypothesis, H₀: b = 1. Therefore, we conclude that there is no “proportional” systematic error.
• For the intercept a = 0.0315: H₀: a = 0 vs. H_A: a ≠ 0, to ascertain whether the value of a is statistically equal to zero, the 95% confidence interval of a should contain zero. From the regression analysis table, Figure 8 (below), CI 95% = (−0.167; 0.230). Since CI includes zero, we accept the null hypothesis, H0: a = 0. Therefore, we conclude that there is no “standard” systematic error [9].

3.5. Precision

Table 5. Precision test results.

Day:	1		2		3		4		5		6		7		8
DEHP mg/L:	1.11	1.05	1.09	1.06	1.04	1.01	1.07	1.09	1.05	1.03	1.10	1.06	1.10	1.14	1.04	1.06

gives the results of the precision measurements (two repetitions) for each of the eight days. Entering the data in Table 5 into the MINITAB statistical package and applying ANOVA gives Table 6, which is the analysis of variance table.

Table 6 shows in red font the values of the variance for the factor that varies (day) and for the error, which represents the repeatability. Therefore, the following results are obtained:

• Repeatability:

  −

  Standard deviation$:{ \mathrm{s}}_{\mathrm{r}}=\sqrt{{\mathrm{s}}_{\mathrm{error}}{}^2}=\sqrt{0.00061}=0.02470 \mathrm{mg}/\mathrm{L}$.

  −

  Relative standard deviation, from equation ${\mathrm{RSD}}_{\mathrm{r}}=\frac{{\mathrm{s}}_{\mathrm{r}}}{\overline{\mathrm{x}}}100$ = 2.3%, as $\overline{\mathrm{x}}$ = 1.07 mg/L the average of the measurements in Table 7.

  −

  Limit of repeatability, from equation $\mathrm{r}=2.8\times {\mathrm{s}}_{\mathrm{r}}$: r_s = 0.069 mg/L, for two repeated measurements under repeatability conditions and 95% significance level.

• Intermediate precision:

  −

  Standard deviation:

  ${\mathrm{s}}_{\mathrm{I}}=\sqrt{{\mathrm{s}}_{\mathrm{day}}{}^2+{\mathrm{s}}_{\mathrm{error}}{}^2}=\sqrt{0.00057+0.00061}=0.0344$ mg/L.

  −

  Relative standard deviation, from equation ${\mathrm{RSD}}_{\mathrm{I}}=\frac{{\mathrm{s}}_{\mathrm{I}}}{\overline{\mathrm{x}}}100$ = 3.2%, as $\overline{\mathrm{x}}$ = 1.07 mg/L the average of the measurements in Table 7.

  −

  Limit of repeatability, from equation $\mathrm{r}=2.8\times {\mathrm{s}}_{\mathrm{I}}$: r_s = 0.096 mg/L, for two repeated measurements under repeatability conditions and 95% significance level [7,11,21].

The RSD_r and RSD_I values satisfy the acceptable values predicted based on the concentration level of DEHP$\left(\overline{\mathrm{x}}=1.07\frac{\mathrm{mg}}{\mathrm{L}}\right)$ [16]. Additionally, it is worth noting that there does not seem to be any variation in the measurements between the different analysis days. This is proven statistically, for a significance level of 95%, by the p-value = 0.082 > 0.05, which is shown in Table 6.

Table 6. Analysis of variance table for estimation of method precision.

ANOVA: DEHP mg/L versus Day
Factor    Type    Levels           Values
day   random   8   1; 2; 3; 4; 5; 6; 7; 8
Analysis of Variance for DEHP mg/L
Source            DF               SS                 MS               F            P
day                    7         0.0122750   0.0017536   2.86     0.082
Error                8          0.0049000   0.0006125
Total               15         0.0171750
S = 0.0247487  R-Sq = 71.47  R-Sq(adj) = 46.51%
Expected Mean
Square for Each
Term (using
Variance   Error     unrestricted
Source    component    term      model)
1 day       0.00057    2    (2) + 2 (1)
2 Error        0.00061           (2)

Table 6. Analysis of variance table for estimation of method precision.

ANOVA: DEHP mg/L versus Day
Factor    Type    Levels           Values
day   random   8   1; 2; 3; 4; 5; 6; 7; 8
Analysis of Variance for DEHP mg/L
Source            DF               SS                 MS               F            P
day                    7         0.0122750   0.0017536   2.86     0.082
Error                8          0.0049000   0.0006125
Total               15         0.0171750
S = 0.0247487  R-Sq = 71.47  R-Sq(adj) = 46.51%
Expected Mean
Square for Each
Term (using
Variance   Error     unrestricted
Source    component    term      model)
1 day       0.00057    2    (2) + 2 (1)
2 Error        0.00061           (2)

3.6. Range

In Section 3.2 reliable linearity was found in the concentration range of 0.3–1.5 mg/L DEHP. Additionally, in Section 3.4 the accuracy of the method was studied for spike concentrations of 0.3, 0.7, 1.0, and 1.5 mg/L in a real Ouzo sample; the results showed that there is no “proportional” and/or “standard” systematic error. Furthermore, the precision studies (Section 3.5) were performed on a real Ouzo sample, with a concentration of 1.0 mg/L DEHP. The results of the precision study showed that the relative standard deviations of repeatability and intermediate precision are reliable [7]. Therefore, we can safely define the working range of the method as the range 0.3–1.5 mg/L DEHP. In this region, there is reliable linearity, precision, and accuracy.

3.7. Ruggedness

The results of ruggedness tests are shown in Table 7; the same sample was measured twice for each experiment. By entering the data of Table 7 into the MINITAB statistical package and applying “General Linear Model”,

Table 8. Analysis of variance table for the ruggedness of the method.

Analysis of Variance
Source   DF   Adj SS   Adj MS  F-Value  p-Value
A       1       0.003025    0.003025     5.48        0.041
B        1       0.000025    0.000025     0.05        0.836
C           1       0.000100    0.000100     0.18        0.680
D       1       0.000625    0.000625     1.13        0.313
E        1       0.000400    0.000400     0.72        0.415
Error      10     0.005525    0.000553
Lack-of-Fit  2       0.000625    0.000313     0.51        0.619
Pure Error   8       0.004900    0.000613
Total     15      0.009700

is obtained, which is the analysis of variance table. Table 8 shows the p-values, which for all four factors (B–E) are greater than 0.05. Therefore, it is found that factor B, C, D, and E do not affect the measurement for 95% significance level. However, factor A (concentration of methanol solution in acetonitrile) has p-value = 0.041 < 0.05. So, it is found that the deliberate change in factor A affects the measurement for a significance level of 95% [12].

Table 7. Experimental design of five-parameter two-level factors, to evaluate the ruggedness of method (according to Taguchi experimental design, seven-parameter two-level factors).

Experiment	Factors							mg/L DEHP
	A	B	C	D	E	F	G
1	1	1	1	1	1	1	1	1.07	1.03
2	1	1	1	2	2	2	2	1.00	1.02
3	1	2	2	1	1	2	2	1.03	1.03
4	1	2	2	2	2	1	1	1.00	1.05
5	2	1	2	1	2	1	2	1.07	1.04
6	2	1	2	2	1	2	1	1.04	1.06
7	2	2	1	1	2	2	1	1.05	1.07
8	2	2	1	2	1	1	2	1.09	1.03

Table 7. Experimental design of five-parameter two-level factors, to evaluate the ruggedness of method (according to Taguchi experimental design, seven-parameter two-level factors).

Experiment	Factors							mg/L DEHP
	A	B	C	D	E	F	G
1	1	1	1	1	1	1	1	1.07	1.03
2	1	1	1	2	2	2	2	1.00	1.02
3	1	2	2	1	1	2	2	1.03	1.03
4	1	2	2	2	2	1	1	1.00	1.05
5	2	1	2	1	2	1	2	1.07	1.04
6	2	1	2	2	1	2	1	1.04	1.06
7	2	2	1	1	2	2	1	1.05	1.07
8	2	2	1	2	1	1	2	1.09	1.03

The solution, with a concentration of 1.00% methanol in acetonitrile, is prepared in a 500 mL volumetric flask by transferring 5.000 mL of methanol and bringing the flask to volume with pure acetonitrile. The transfer of 5.000 mL of methanol is conducted with a 0.500–5.000 mL pipette, which has an accuracy, according to the manufacturer, of ±0.5%, which means 5.000 ± 0.025 mL. For the preparation of the solution of the deliberate change (0.99%), the same pipette was used, transferring 4.950 mL of methanol. Note that the deliberate change is not exactly the lower limit of accuracy, because the pipette adjustment scale does not allow transfer of 4.975 mL. This particular pipette was internally tested by the laboratory and found to be ±0.2% accurate. Therefore, the deliberate change is very large relative to the precision limits of the pipette, and we can consider that such a magnitude of variation is unlikely to occur by chance. However, since this factor has a significant effect on the measurement, the tolerance limits should be indicated in the method instruction and the control method. Finally, it should be noted that, in addition to the accuracy of the pipette, the accuracy of the 500 mL volumetric flask and the deviation of the solution temperature from 20 °C affect the overall accuracy of the preparation of the 1.00% methanol in acetonitrile solution; however, the two last parameters can be considered negligible.

3.8. Assessment of Uncertainty

3.8.1. Calculation u_bias,N

For this calculation, the data from the recovery tests were used. Over a period of one year, 23 recovery measurements were made on a spiked Ouzo sample; the recovery results are shown in

Table 9. DEHP recovery measurements in Ouzo, in the context of internal quality control of the accuracy of the method.

Measurement	1	2	3	4	5	6	7	8	9	10	11	12
recovery Ri%	93	94	93	91	95	91	92	97	95	92	91	91
Measurement	13	14	15	16	17	18	19	20	21	22	23
recovery Ri%	93	94	91	90	94	91	92	92	93	94	91

.

According to Equation (4), the uncertainty of the accuracy was factored into the uncertainty of the spike concentration, i.e., the uncertainty of the volume of the flask that prepared the spiked sample, the uncertainty of the pipette that transferred the concentrated DEHP standard, and the uncertainty from the purity of the concentrated standard. Therefore, the relative standard uncertainty of systematic errors (from recovery) by NORDTEST, u_bias,N, is calculated from Equation (4) [14]:

$${\mathrm{u}}_{\mathrm{bias},\mathrm{N}}=\sqrt{{\mathrm{u}}_{\mathrm{conc}}{}^2+{\mathrm{u}}_{\mathrm{vol}}{}^2+\left(\frac{{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{R}}_{\mathrm{i}}-100\right)}^2}{\mathrm{n}}\right)}$$

(4)

u_conc: A certified DEHP reference standard was used for spike, for which the certificate gave purity = 99.90%, and uncertainty a = ±0.05% for a significance level of 95%. In these cases, assuming a normal distribution, the standard uncertainty is given by Equation (5):

$$\mathrm{u}\left(\mathrm{purity}\right)=\frac{\mathrm{a}}{\mathrm{k}}$$

(5)

where k is the coverage coefficient, which for a 95% significance level is k = 1.96 (more commonly k = 2). According to Equation (5) the standard uncertainty from standard purity is:

u(purity) = 0.00025

and the corresponding relative standard uncertainty is:

$${\mathrm{u}}_{\mathrm{conc}}=\frac{\mathrm{u}\left(\mathrm{purity}\right)}{\mathrm{purity}}100=0.025025\%$$

u_vol: A volumetric flask (class A) V_bottle = 100 mL was used for the spike of the sample, which according to the manufacturer has a tolerance of a = ±0.08 mL. The standard uncertainty due to the tolerance, assuming a quadratic distribution, is:

$$\frac{0.08}{\sqrt{3}}=0.04618802 \mathrm{mL}$$

However, according to the manufacturer, the bottle is calibrated at 20 °C, while the laboratory temperature is 20 ± 4 °C. The uncertainty from this source can be calculated from the estimate of the temperature range and the volume expansion coefficient. The volume expansion of the liquid is significantly greater than that of the flask, so only the volume expansion needs to be considered [13]. The volume expansion coefficient for a 40% v/v alcoholic beverage is 6.5 × 10⁻⁴ °C⁻¹ [22]. Therefore, the change in volume of liquid contained in the bottle is:

±(100 × 4 × 6.5 × 10⁻⁴) = 0.26 mL

the standard uncertainty of this source, assuming a quadratic distribution, is:

$$\frac{0.26}{\sqrt{3}}=0.15011107 \mathrm{mL}$$

If the two standard uncertainties are combined, the standard uncertainty of the bottle volume is obtained:

$$\mathrm{u}\left({\mathrm{V}}_{\mathrm{bottle}}\right)=\sqrt{{0.04618802}^2+{0.15011107}^2}=0.157056253$$

and the corresponding relative standard uncertainty is:

$$\frac{\mathrm{u}\left({\mathrm{V}}_{\mathrm{bottle}}\right)}{{\mathrm{V}}_{\mathrm{bottle}}}=0.00157056$$

Furthermore, a pipette of 0.010 mL to 0.100 mL was used to transfer the concentrated DEHP standard, and in particular V_pipette = 0.050 mL were transferred, which according to the manufacturer have a tolerance of a = ±0.001 mL. Thus, the standard uncertainty of the pipette volume, assuming a quadratic distribution, is:

$$\mathrm{u}\left({\mathrm{V}}_{\mathrm{pipette}}\right)=\frac{0.001}{\sqrt{3}}=0.00057735 \mathrm{mL}$$

and the corresponding relative standard uncertainty, is:

$$\frac{\mathrm{u}\left({\mathrm{V}}_{\mathrm{pipette}}\right)}{{\mathrm{V}}_{\mathrm{pipette}}}=0.001154701$$

In summary, the relative volume uncertainty, u_vol is:

$${\mathrm{u}}_{\mathrm{vol}}=100\sqrt{{\left(\frac{\mathrm{u}\left({\mathrm{V}}_{\mathrm{bottle}}\right)}{{\mathrm{V}}_{\mathrm{bottle}}}\right)}^2+{\left(\frac{\mathrm{u}\left({\mathrm{V}}_{\mathrm{pipette}}\right)}{{\mathrm{V}}_{\mathrm{pipette}}}\right)}^2}=1.16533257\%$$

Substituting into Equation (4), the u_conc, u_vol values calculated above, R_i from Table 9 and n = 23, the relative standard uncertainty of systematic errors (from recovery) by NORDTEST is [14]:

u_bias,N = 7.67052066%

3.8.2. Calculation u(R_w)

The uncertainty from intra-laboratory reproductivity, u(R_w), was calculated using the data collected from the control sample measurements (a total of 36 measurements over one year on the same control sample). From these data, $\overline{\mathrm{x}}$ = 0.9258 mg/L DEHP and s = 0.017948 mg/L DEHP. Therefore, the relative standard uncertainty from the intra-laboratory reproductivity, u(R_w), will be equal to the relative standard deviation, expressed as a % [14]:

u(R_w) = 1.9386144%

3.8.3. Calculation of Combined Relative Uncertainty by NORDTEST

Combining the uncertainties, u_bias,N and u(R_w), calculated above, Equation (1) gives the uncertainty according to NORDTEST [14]:

u_c= 7.9% ή 0.079

In summary, we can say that the relative standard uncertainty of the method is ±8%, and the extended uncertainty, for a significance level of 95%, is ±16%.

4. Conclusions

The present study validates a method for determining DEHP in alcoholic beverages (Ouzo) using HPLC. The validation process includes testing for accuracy, precision, specificity, linearity, limit of quantification, range, and ruggedness. Each validation characteristic was tested for compliance with acceptance criteria found in the literature, and all necessary statistical tests were performed. The impurities in the Ouzo matrix did not affect the determination of DEHP using the validated method. The lower limit of quantification was determined to be 0.06 mg/L DEHP. The precision study involved duplicate measurements on eight different days, and RSD values for repeatability and intermediate precision were calculated using ANOVA analysis. The resulting RSD values met acceptable standards based on the expected DEHP concentration level. A fractional Taguchi experimental design was applied to assess ruggedness, and the relative standard uncertainty of DEHP determination in Ouzo was found to be ±8%. These results confirm that the laboratory’s in-house developed method is reliable and suitable for its intended purpose. Future research will focus on expanding the method to other phthalates and assessing the uncertainty of DEHP determination using the NORDTEST technique. Additionally, the method’s performance will be periodically assessed through interlaboratory tests.