Article pubs.acs.org/JAFC
Determination of Phosphine in Plant Materials: Method Optimization and Validation in Interlaboratory Comparison Tests Thomas M. Amrein,*,† Lara Ringier,† Nathalie Amstein,† Laurence Clerc,† Sabine Bernauer,† Thomas Baumgartner,† Bernard Roux,§ Thomas Stebler,§ and Markus Niederer§ †
COOP Central Laboratory, Head Quarters, Gottesackerstrasse 4, 4133 Pratteln, Switzerland State Laboratory Basel-City, P.O. Box, 4012 Basel, Switzerland
§
ABSTRACT: The optimization and validation of a method for the determination of phosphine in plant materials are described. The method is based on headspace sampling over the sample heated in 5% sulfuric acid. Critical factors such as sample amount, equilibration conditions, method of quantitation, and matrix effects are discussed, and validation data are presented. Grinding of coarse samples does not lead to lower results and is a prerequisite for standard addition experiments, which present the most reliable approach for quantitation because of notable matrix effects. Two interlaboratory comparisons showed that results varied considerably and that an uncertainty of measurement of about 50% has to be assessed. Flame photometric and mass spectrometric detection gave similar results. The proposed method is well reproducible within one laboratory, and results from the authors’ laboratories using different injection and detection techniques are very close to each other. The considerable variation in the interlaboratory comparison shows that this analysis is still challenging in practice and further proficiency testing is needed. KEYWORDS: phosphine, fumigation, headspace, GC-FPD, GC-MS, interlaboratory comparison
■
with flame photometric (FPD), mass spectrometric (MS), or thermoionic (NPD) detection allows a specific and sensitive determination.10−12 Older methods rely on derivatization of PH3 or its reaction products followed by titrimetric or photometric detection, for example, the method by Bruce et al.13 Extraction, quantitation, and matrix effects are critical aspects of each method and can strongly influence the analytical result.7 The method of the Swiss Food Manual14 refers to a published method,6 but validation data is lacking. Another issue that concerns all methods is the absence of extended validation in interlaboratory comparisons or proficiency tests using samples with incurred residues. Such tests and validation need to be done with incurred residues.7 Therefore, an improved and validated method for residues in the low μg/kg range in coarse and powdery samples was strived for, starting from the method of Amstutz et al.6 In addition, two rounds of interlaboratory comparisons were organized to gain data and knowledge about the interlaboratory variation of such methods. Thereby, the proposed method is validated and a realistic picture of the uncertainty of measurement can be drawn. In the present paper, an optimized method as well as results from two international interlaboratory comparisons are presented. To the best of our knowledge, data from such tests have not been published so far.
INTRODUCTION Phosphine (PH3) is widely used as a fumigant to control insects in stored products, and it is an important chemical alternative to the banned methyl bromide.1 Residues of phosphine in food must comply with legal limits. For example, the maximum residue limit for cereals is 0.1 mg/kg (EU and Switzerland) and 0.01 mg/kg for cereal products (Switzerland). However, the use of chemical fumigants is not allowed for organic products. Therefore, if a residue is found, the organic production and storage may be questioned or a mixing of conventional and organic commodities may be suspected. In Germany, the orientation value of 0.01 mg/kg from the Association of Organic Processors, Wholesalers and Retailers (BNN) presents a nonofficial limit for organic products,2 whereas in Switzerland enforcement authorities use a limit of 0.001 mg/kg (1 μg/kg).3 Before 2011, organic samples with residues above 0.1 μg/kg were even objected in Switzerland.4 Residues in conventional products are fairly frequent: A recent study reported traceable amounts in 31% of the conventional samples (limit of determination = 0.1 μg/kg, range = 0.1−3.7 μg/kg), but legal limits were not exceeded.5 In the same study residues were detected in 12% of the organic samples (range = 0.12−1.1 μg/ kg), which is precarious because residues were within a similar range as conventional products, although clearly less frequent. Residues in organic cereals were also reported by Amstutz et al., who found 0.3−2.5 μg/kg in rice and maize.6 For monitoring PH3 residues in the low μg/kg level, an appropriate method is needed and organic samples are of particular interest. A number of analytical methods for the determination of phosphine were published, reviewed by Desmarchelier and Ren.7 Because of the volatility of the analyte, headspace sampling over the acidified sample is a convenient analytical approach.8,9 Gas chromatography (GC) © 2014 American Chemical Society
■
MATERIALS AND METHODS
Chemicals. Sulfuric acid of reagent grade (Scharlau, Sentmenat, Spain) was used for extraction as 5% solution in deionized water. A
Received: Revised: Accepted: Published: 2049
November 1, 2013 February 10, 2014 February 15, 2014 February 24, 2014 dx.doi.org/10.1021/jf404918e | J. Agric. Food Chem. 2014, 62, 2049−2055
Journal of Agricultural and Food Chemistry
Article
Ionization was done in positive EI mode at 70 eV and 200 °C, and the masses m/z 34 and 33 were monitored. As an alternative injection technique, an automated balanced pressure system (Perkin-Elmer, Boston, MA, USA) was used with the following conditions: pressurization at 190 kPa (helium, 1 min); injection time, 0.2 min; transfer line at 80 °C (deactivated capillary, 2.0 m × 0.32 mm). Data Interpretation. All calculations were done using Microsoft Excel 2010. For interpretation of the interlaboratory comparison (round 2) the mean of all submitted results except the outlier was chosen as assigned value. The standard deviation for proficiency (σp) was derived by the Horwitz equation15 from the mean.
dilution of phosphine in nitrogen (74.1 mL/m3, corresponding to 102.97 μg/L) was used as reference gas (Linde, Unterschleissheim, Germany). Its shelf life is 1 year as stated by the provider. Caution: PH3 is toxic and its gas preparation has to be handled with care and appropriate equipment in a fume hood. Sample Preparation. Samples with incurred residues were obtained from different supermarkets and companies. They were raw and untreated, except of storage, milling, filling. Samples (about 250 g) were ground using standard laboratory mills, for example, a Grindomix (Retsch, Basel, Switzerland), and homogenized for 1 min using a Turbula-mixer (Willy A. Bachofen AG Maschinenfabrik, Muttenz, Switzerland). Twenty milliliter headspace vials were used for calibration and extraction of samples (BGB Analytik, Böckten, Switzerland). One gram of sample (standard amount) was weighed as such or ground into a headspace vial, and 5% sulfuric acid was added to reach the same level as in the calibration vials. Samples were analyzed in duplicates, and per vial only one injection was made. Powdery samples have to be briefly mixed with a plastic stirrer to avoid enclosure of bubbles and to facilitate mixing. The vials were immediately closed (caps with Teflon septum, Wicom Int., Maienfeld, Switzerland) and vigorously shaken by hand for 5 s to achieve good mixing and wetting before incubation in a heated agitator. If samples foam, 3−4 drops of silicon oil (Merck) can be added. Calibration. Reference gas was filled into a gastight bag with a septum (Tedlar gas sampling bag, SKC, Dorset, UK). For calibration, appropriate aliquots were taken through this septum using gastight syringes (Hamilton, Bonaduz, Switzerland) and directly added to the headspace of the closed vials containing 15 mL of 5% sulfuric acid. Afterward, the vials were immediately shaken by hand. For screening purposes an external calibration of 0.1, 0.3, 0.5, and 1.0 μg/kg may be used whereby 1, 3, 5, and 10 μL of reference gas are added (on the basis of 1 g of sample). Quantitation is done by standard addition to compensate for the different partitioning of PH3 over the solvent if sample matrix is present (see below). Thereto, the aliquots of reference gas were added to the headspace of the closed vials containing sample and 5% sulfuric acid prior to incubation. Additions should be approximately 50, 100, and 150−200% of the signal from the unspiked sample, which may be analyzed in duplicate. The content of phosphine is calculated via the regression equation, and the coefficient of determination (R2) must be ≥0.95. This criterion is somewhat loose but is a compromise between practical feasibility and accuracy and reflects the substantial variation of this analysis (see also Interlaboratory Comparisons). Instrumentation. Partitioning of phosphine was carried out under acidic conditions with heating and shaking to reach a stable equilibrium. Headspace sampling was followed by gas chromatography (GC) with either pulsed flame photometric detection (PFPD) or MS detection. Incubation of all vials (samples and calibration) was done using an MPS2-Sampler (Gerstel, Sursee, Switzerland) at the following conditions: heating at 65 °C for 20 min with shaking in intervals, 60 s at 750 rpm followed by 90 s without shaking. After incubation, 1000 μL of headspace was injected splitless (1 min) onto a PLOT column (PLOT fused silica, 25 m × 0.53 mm; Coating Poraplot Q, Agilent, Basel, Switzerland) in 2 s using a heated syringe (70 °C, Gerstel). Alternatively, only 500 μL can be injected if sensitivity allows, and headspace or splitless liners (i.d. 2 mm, Agilent) may be used. The syringe was flushed with nitrogen after each injection for 30 s. A GC system (Varian 3800, Zug, Switzerland) with FPD was used. The injector was held at 220 °C, and chromatography was performed isothermally at 50 °C for 15 min with a column head pressure of 10 kPa and hydrogen as carrier gas. A PFPD was used in the P-mode at 300 °C. For system maintenance regular change of the injector septum (about every 30 injections) and regular baking out of the column at its maximum temperature (250 °C) are recommended. For detection by MS (DSQII, Thermo Electron Corp., Austin, TX, USA) helium was used as carrier gas (3 mL/min), and a deactivated fused silica column (BGB, Boeckten, Switzerland) was installed after the PLOT column (2 m, 0.25 mm i.d.) ending in the transfer line (200 °C). To prevent particles from the PLOT column entering the ion source, a press-fit connection with an inner diameter of about 0.1 mm was used.
■
RESULTS AND DISCUSSION Method Optimization. The method was optimized for sensitivity, precision, and reliable quantitation. For signal optimization, the sample amount, the volume of 5% sulfuric acid, and the conditions to establish equilibrium (temperature, time, shaking) were investigated using samples with incurred residues only. Signals increased for larger sample amounts for a given volume of acid (Figure 1). However, doubling the sample
Figure 1. Influence of sample amount on PH3 signal for wheat flour and semolina. Basis: average peak area for 1 g (n = 3, 15 mL of 5% sulfuric acid).
amount did not result in doubled signals, especially for the flour. Higher sample amounts increased the solubility of phosphine in the sample phase due to stronger interaction with matrix components and changed the viscosity of the sample phase. These effects changed the activity coefficient of PH3, whereby its concentration in the gas phase did not increase to the same extent as the sample amount. The effect depended on the sample type (see slopes in Figure 1): whereas the signal for semolina almost doubled for doubled sample amounts, the increase was 4.6 times lower for flour. This is explained by their different compositions and interactions with the analyte and by their different viscosities. For larger sample amounts the partitioning strongly changed as compared to pure solvent; that is, the signal per weight decreased. Thus, the sample amount is a critical factor and needs to be carefully controlled and quantitation with external calibration is pointless, especially for more than 1 g of sample (see also below). The two samples were not related to each other and, thus, no direct conclusion can be drawn. However, the effect on viscosity and matrix in solution is expectedly smaller for semolina, which is in line with the usually larger signals in whole compared to ground samples. Small sample amounts such as 1 g are easier to mix and suspend during shaking (see below), and the difference in partitioning as 2050
dx.doi.org/10.1021/jf404918e | J. Agric. Food Chem. 2014, 62, 2049−2055
Journal of Agricultural and Food Chemistry
Article
frequent. This decrease may be due to oxidation or other reactions with matrix components. Therefore, an incubation temperature of 65 °C was chosen as a compromise. In the literature, boiling for 20 min in 10% sulfuric acid was described as release condition,17,18 whereas other methods use longer times.7 In contrast, Ren et al. found a maximum signal in the headspace after heating wheat at 45 °C for 20 min.19 Extraction and partitioning were also checked for the present method as follows: A sample of wheat flour (1 g, 15 mL of 5% sulfuric acid) was analyzed in duplicate. After cooling to room temperature, the vials were opened, flushed with a gentle stream of air for 10 s to remove PH3 from the headspace, capped again, and reanalyzed including incubation in the heated agitator. The same was done with a calibration standard (0.3 μg/kg, n = 3). The PH3 signal in the sample reached 23% of the first analysis, whereas in the calibration standard 16% was found. This shows that equilibrium is on the side of the gas phase and that partitioning is the key factor after extraction. Also, it again demonstrates the different partitioning if sample matrix is present (see Figure 4).
compared to pure solvent is smaller (see Figures 1 and 4). This allows to draw indicative values from external calibration and facilitates estimating appropriate amounts for standard additions. Figure 2 shows that the signal of phosphine increased with higher volumes of acid. This can be explained by a smaller
Figure 2. Influence of volume of 5% sulfuric acid on PH3 signal (wheat flour, 1 g; injection volume = 1 mL).
phase ratio and a lower retention of PH3 in the more diluted sample phase (i.e., increased activity coefficient).16 Both effects may have contributed to the increase of signal. Although a small headspace volume has the advantage of larger signals, an intermediate volume of 15 mL acid was chosen as a compromise. Thereby, shaking and suspending were more effective (see below), enough solvent was provided to establish equilibrium faster, and enough headspace was available for needle intrusion and sampling. Increasing temperature during incubation in the heated agitator gave larger signals (Figure 3) because the partition Figure 4. Calibration functions obtained in different matrices. Signal from sample was subtracted (1.5 g of sample, 15 mL of 5% sulfuric acid).
In addition, the influence of shaking on the signal was checked. Samples need to be intensively shaken by hand to achieve good mixing and wetting before incubation in a heated agitator. Thereto, an appropriate ratio of sample and solvent is important, for example, 1 g and 15 mL. Small sample amounts are easier to suspend and swell during shaking and are less prone to form viscous doughs. An intensive (750 rpm on the agitator) and prolonged shaking (20 min) gave best results for a given combination of sample amount, temperature, and volume of acid. Equilibrium was reached for 1 g of wheat flour in 15 mL of solvent after incubation for 20 min, whereas for solvent standards it was established already after 5 min (no data shown). The effect of a more intensive shaking becomes particularly evident for larger sample amounts and for matrices with pronounced swelling. Samples may also be incubated in a GC oven, for example, for 20 min at 65 °C and manual shaking in regular intervals (e.g., every 5 min for 1 min) if a heated agitator is not available. If a classical convection oven is used, a higher temperature (about 80 °C) is recommended because the temperature of the vials is substantially lower due to less efficient heat transfer as compared to a heated agitator or a GC oven. Nevertheless, the use of a heated agitator combined with an autosampler is recommended for practical and reproduci-
Figure 3. Influence of incubation temperature on PH3 signal: injection of hot and cooled samples (2 g of wheat flour, n = 2, 15 mL of 5% sulfuric acid).
coefficient depends exponentially on the inverse temperature.16 At about 70 °C a maximum was observed for both hot and cool injected samples. The influence of temperature was less pronounced for samples cooled before injection, and their signals were generally lower compared to immediate, hot injection. This is because partitioning toward the liquid phase is favored at lower temperatures. Nevertheless, injection after cooling has the advantage of reducing solvent load on the analytical column. Incubation at 80 °C decreased the signal in both hot and cool injected samples, and clumping was more 2051
dx.doi.org/10.1021/jf404918e | J. Agric. Food Chem. 2014, 62, 2049−2055
Journal of Agricultural and Food Chemistry
Article
Table 1. Characterization of Samples and Results of the Interlaboratory Comparison Testsa characterization of samples
results of interlaboratory comparison
1
2
3
(n = 10)
(n = 10)
(n = 3)
COOPb
State Labc
(n = 2)
(n = 2)
mean
n
18.5 0.26 0.9
9.9 (79%) 0.18 (51%) 0.9 (26%)
9 3 6
σp
Round 1 wheat flour wheat semolina rice wheat flour red lentils jasmine rice
23 (11%) 0.18 (6%) 1.03 (7%) 3.0 (13%) 2.5 (18%) 15.4 (20%)
27 (4%) 0.23 (5%) 0.9 (10%) 3.7 (14%) 2.9 (13%) 15.9 (10%)
21 22.9 0.20 0.21 0.4 1.2 Round 2 4.8 5.1 3.1 3.5 17.2 13.7
4.3 3.5 15.7
2.9 (56%) 2.5 (55%) 13.1 (47%)
12 11 12
1.1 1.0 4.0
RSD, relative standard deviation (in parentheses); σp, standard deviation for proficiency, Horwitz (only round 2); outlier excluded. bInjection with syringe (1 mL), detection by PFPD. cInjection with pressure system, detection by FPD. a
bility reasons. Heating and shaking of all vials prior to injection presents an improvement for quantitation as compared to other methods relying on external calibration where partition of phosphine between headspace and liquid phase is not taken into account.17,18 Linearity was checked using five concentration levels (n = 1) in 5% sulfuric acid in the range of 0.19−1.95 μg/kg plus a blank. Mandel’s test confirmed a linear calibration function (R2 = 0.98). Linearity was also checked in the range from 0.1 to 10 μg/kg (total of 10 levels, n = 1), and again a linear function was found (R2 = 0.99). On the basis of 1 g of sample and 15 mL of acid, a limit of quantitation of 0.1 μg/kg was easily achieved without matrix, which corresponds to 0.1 to about 0.4 μg/kg in real samples. This is sufficient to check samples for compliance to residue limits, especially organic samples. In cereal samples, concentrations as low as 0.2 μg/kg were determined by standard addition in our laboratory. The precision, that is, RSD, of results obtained by standard addition accounted for 16% (wheat flour, n = 4, mean = 3.5 μg/kg). This is similar for results obtained with all types of samples analyzed by the two authors’ laboratories (COOP and State Lab) in the interlaboratory comparison studies (Table 1). Thus, the method is well reproducible provided that some experience and appropriate equipment are present. Figure 4 shows that the signal for a given amount of PH3 was largest in the absence of any sample matrix (5% sulfuric acid only). The effect of sample type (Figure 4) and sample amount (Figures 1 and 5) can vary considerably. Similar effects were also reported in the literature.9,17 As a consequence, quantitation by external calibration in 5% sulfuric acid mostly leads to an underestimation of phosphine contents and comparability of results may be limited. External calibration using a generic matrix devoid of residues only partially overcomes this problem because of varying effects in different matrices. Therefore, results obtained by external calibration are indicative and not quantitative, especially when large sample amounts are used. Furthermore, recovery experiments and partly also standard additions suffer from the fact that added PH3 remains largely in the gas phase and poorly sorbs to the matrix.7 In contrast, incurred residues need to be extracted from the matrix into the liquid phase (diffusion controlled) from where they partition to the gas phase. Standard additions compensate at least for part of these effects, for example, partition between liquid and gas phase, as can be seen from the lower signals from matrices as compared to solvent (Figure 4). Therefore, recoveries in the classical analytical sense cannot be truly determined for incurred residues of phosphine. The key
Figure 5. Standard addition experiments with ground cereal risotto using different sample amounts (15 mL of 5% sulfuric acid). Results for different sample amounts: 0.5 g, 0.93 μg/kg; 1.0 g, 0.89 μg/kg; 1.5 g, 0.97 μg/kg.
point is to reach equilibrium under reproducible conditions. A reasonably good correlation in standard addition serves as an indicator for reproducible equilibrium. Recoveries from 50 to 90% were reported, and sample type and amount both affected recovery.17,20 In this study recoveries from as low as 20 to 140% were found, depending on matrix and sample amount, showing the limited significance of recoveries. Time is another factor affecting results obtained in screening with external calibration. This was checked with two sets of samples of ground rice (1 g, 15 mL of acid, n = 10 each), of which the first set was directly incubated and analyzed and the other set was left to stand for 2 h before incubation. A significant decrease of 17% of the signal (t test: p < 0.0001) was observed for samples left to stand for >2 h. Therefore, long series of samples should be avoided or samples should be prepared batchwise to prevent standing times of >2 h before incubation. The only feasible way to obtain reliable quantitative data is to analyze samples using the standard addition approach. This is preferably done after screening the samples against an external calibration in 5% sulfuric acid to check for the presence of PH3. In the performance of standard addition experiments, the sample amount is not critical as long as it is within a low range (0.5−1.5 g) as can be seen in Figure 5: very similar results were obtained for the different sample amounts analyzed because the matrix effect was compensated (see slopes in Figure 5). 2052
dx.doi.org/10.1021/jf404918e | J. Agric. Food Chem. 2014, 62, 2049−2055
Journal of Agricultural and Food Chemistry
Article
Figure 6. Concentrations of phosphine in whole and ground samples, determined by standard addition. Error bars are ±25% (n = 2, 1 g of sample, 15 mL of 5% sulfuric acid).
Sample homogeneity is critical for standard addition experiments: For heterogeneous samples (e.g., cereal mixtures or Muesli) and for large particles (e.g., cereal grains) the variation of residue per “unit” makes it sometimes impossible to obtain a standard addition with an acceptable R2. The limited sample amount (usually
