Experimental central composite design-based dispersive liquid-liquid microextraction for HPLC-DAD determination of diazinon in human urine samples: method development and validation

HPLC-grade methanol, acetonitrile, water, toluene and dichloromethane were purchased from Merck Chemical Co. (Darmusdat, Germany). HPLC-grade standards for diazinon, pirimiphos-methyl, azinphos-ethyl, and chlorpyrifos were purchased from Dr. Ehrenstofer GmbH (Augsburg, Germany). All other chemicals and reagents were of analytical grade, purchased from Merck Chemical Co.

Separation, identification, and quantification were carried out on a Knauer HPLC system (Smartline Series 1200, Berlin, Germany). Chromatography was run isocratically on a Nucleosil® C₁₈ analytical column (250×4.6 mm, 5 μm particle size, Perfectsil® Target). An RP-18 guard column was fitted upstream of the analytical column. The mobile phase was a mixture of acetonitrile and buffer, optimised (63:37 v/v, pH 3.2) and delivered by a Knauer 1050 HPLC pump at a flow rate of 1 mL/min. A diode array detector (K-2800, Knauer) with a wavelength range of 190–740 nm was used for detection. The system was equipped with ChromGate® software (version 3.3.2., Knauer).

We used diazinon-free urine samples provided by healthy volunteers in our laboratory. They were kept frozen at -20 °C until analysis and then thawed to room temperature. Each sample was added 10 μL of pirimiphosmethyl (internal standard, IS) (2.5 μg/mL) and vortexed at 1250xg for 10 min. Then we added 100 μL of sodium lauryl sulphate (SLS) (3 % w/v) and 100 μL of sodium chloride (NaCl) (1 %,w/v) to the glass tubes containing 1000 μL of urine. The final solution was then prepared following the DLLME procedure.

A mixture of 800 μL of methanol (disperser solvent) plus 310 μL of toluene (extraction solvent) was quickly injected to the samples with a syringe (Hamilton, NV, USA), which dispersed fine droplets of toluene to form a cloudy solution. Over just a few seconds, the analytes were extracted on toluene droplets and after centrifugation at 1250xg for 15 min, these droplets became a supernatant on the surface of the conical test tube. The supernatant phase was then completely transferred into another conical test tube and the residue dried by evaporation with nitrogen in a water bath, dissolved to a mobile phase, and then 20 μL of the sample injected into the HPLC.

To achieve maximum recovery, the selection of extraction efficiency variables was based on preliminary experiments that yielded distinct responses of eight variables on three levels of experimental designs (Table 1). Optimisation was performed on spiked samples.

Standard solutions were prepared by serially diluting the diazinon stock solution (100 μg/mL) with HPLC-grade water to 0.5, 1.0, 1.5, 2, 2.5, 3, and 4 μg/mL. The stock solution (2.5 μg/mL) of pirimiphos-methyl (IS) in methanol was prepared and stored at -20 °C. The stock and standard solutions were prepared on a daily basis and stored in the dark at 4 °C. All solutions were used on the day they were prepared.

To obtain optimal conditions, we relied on the four-factor-two-level central composite design (CCD), which is used in response surface methodology. Briefly, each numeric factor is varied over five levels: plus and minus alpha (axial points), plus and minus 1 (factorial points), and the centre point. If categorical factors are added, CCD will be duplicated for every combination of the categorical factor levels. It was also used to investigate parabolic interactions between the following parameters: volume of extraction solvent (toluene), salt percentage (NaCl), surfactant percentage (SLS), and the volume of disperser solvent (methanol). This CCD design allowed modelling the response surface by fitting a second-order polynomial with the number of experiments equal to 21 for four factorial designs at five levels and five replicated points. Table 2 shows the range of independent variables used in this study in terms of actual and coded values.

Validation included the following parameters: linearity, precision, accuracy, limits of detection and quantification, and selectivity (25). For calibration we used seven concentrations ranging from 0.5 to 4 μg/mL of diazinon. Each concentration was prepared in triplicate and analysed three times. Calibration curves were constructed by plotting the concentration of compounds versus peak area response. The linearity was evaluated with the least square regression method.

The limit of quantification (LOQ) was determined during the evaluation of the linear range of calibration curve.

The limit of detection (LOD) and LOQ were calculated as follows:

LOD=3.3σ/S, and

LOQ=10σ/S,

where σ is the standard deviation of response, and S is the slope of the calibration curve.

The method’s precision was determined by repeatability (intra-day) and intermediate precision (inter-day) and was expressed as relative standard deviation (RSD). Five replicate injections of diazinon standard solutions were prepared at concentrations ranging from 0.5 to 4 μg/mL. The intra-day variation was assessed on the same day, while inter-day precision encompassed three consecutive days. Both assessments were carried out by the same analyst.

The accuracy of the method was tested with five replicates of three samples containing different diazinon concentrations, and the measurements were compared with its actual concentration. The accuracy was expressed with recovery percentage.

Selectivity was evaluated by comparing chromatograms of different batches of urine spiked with diazinon, IS, tramadol, azinphos-ethyl, pirimiphos-methyl, and chlorpyrifos.

For regression analysis and diagram plotting for the experimental results we used the Design Expert v. 7.01 software (Stat-Ease Inc., Minneapolis, MN, USA).

This project was approved by the Ethics Committee of the Legal Medicine Research Centre.

Sonication and pH had negative effects on maximum recovery (p>0.05). Other parameters had a positive effect (p<0.05) and were selected for further optimisation (Table 3). Methanol (as disperser solvent) and toluene (as extraction solvent) had positive effects on all variables and were used in the experiments.

The actual and statistically predicted diazinon recoveries for experiments are shown in Table 4, while Figure 1 shows the relationship between the two. The mathematical model was as follows:

Figure 1

(Recovery)^0.81=+22.87-1.01*A-3.76*B+2.98*C+2.97*D-2.64*AB+2.58*AC-3.35*AD-1.56*BC+1.09*A²-1.79*C²+1.01*D²

This equation represents the relationship that diazinon recovery (R) has with the volume of the disperser solvent (A), salt concentration (B), surfactant concentration (C), the volume of extraction solvent (D), and their combinations (AB, AC, AD, and BC). Table 5 shows the results of the analysis of variance (ANOVA) for the CCD model and the significance of each coefficient determined by F-values (variation of data about mean value) and P-values (probability). The model turned out to be highly predictive of the experimental results. Extraction solvent volume had a high linear and quadratic effect on response. In addition, the interaction effects of combined variables were significant. The correctness of the model was also ensured by multiple correlation coefficient (R²), which was 0.9872 and showed high prediction of the actual value and excellent response, with 0.85 % of the total variation. The predicted R² (0.8624) was in reasonable agreement with the adjusted R² (0.9716). Furthermore, the coefficient of variance (CV=4.76 %) was low, which indicates significant precision and reliability of the experimental data.

For precision to be adequate, the signal-to-noise ratio should be >4. With our model it was 30.73, indicating that it could be used to evaluate experiments.

Figure 2 shows a 3D response surface diagram of the effects of two factors on diazinon recovery. Figure 2A shows significant interactions of extraction and disperser solvent volume with diazinon recovery (p<0.0002). Diazinon recovery increased with the increase in extraction solvent volume (from 300 to 450 μL) and disperser solvent volume (from 400 to 800 μL). Significant increase in diazinon recovery was noted when the extraction solvent volume reached 310 μL and disperser solvent 800 μL.

Figure 2

Figure 2B shows that the interactions between diazinon recovery and disperser solvent volume and surfactant concentrations were significant (p<0.0001) at maximum surfactant concentration of 3 % and maximum disperser solvent volume.

Figure 2C, in turn, shows that diazinon recovery also had significant interactions with disperser solvent volume when it reached its maximum volume of 800 μL and when salt percentage was at its lowest (p<0.001).

Standard calibration curves for diazinon were linear with the concentration range of 0.5–4 μg/mL, yielding a regression equation of Y=0.254X+0.006 with a correlation coefficient of 0.993. This is generally considered evidence of an acceptable fit and good linearity over the concentration range.

The method yielded LOD and LOQ of 0.15 μg/mL and 0.45 μg/mL, respectively, and its precision met the acceptance criteria (Table 6). The intra- and inter- day RSD values did not exceed 5 % (bias interval between 3.0 and 5.0 %), which indicates that the method is accurate, reliable, and reproducible.

Table 6 also shows that the recovery percentages comply with the acceptance criteria (25).

The specificity of the method was tested with peak purity on blank and spiked urine samples. Blank samples showed no interference when diazinon and IS were added. Under optimised conditions, the separation of diazinon and pirimiphos-methyl was complete (Figure 3).

Figure 3

The applicability of the proposed DLLME-HPLC-DAD method was evaluated in undiluted urine samples collected from patients poisoned with diazinon who were receiving hospital treatment (Sanandaj, Iran). Relative diazinon recoveries were determined at the spiking level of 0.5, 1, and 3 μg/mL. The results of six replicate experiments of each sample were in the range of 75–95.6 %. Therefore, the proposed method can be applied for determining diazinon in human urine samples.

Table 7 summarises a comparison of the proposed DLLME-HPLC-DAD method with other methods and shows that its LOD, R², and recovery are well within acceptable ranges.