Measurement of Bromate in Bread by Liquid Chromatography with Post-Column Flow Reactor Detection

Transcription

1 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, FOOD COMPOSITION AND ADDITIVES Measurement of Bromate in Bread by Liquid Chromatography with Post-Column Flow Reactor Detection PVM 2:1999 Abstract HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, 2000 Peer Laboratory This method is suitable for the determination of bromate residues in a variety of baked goods. The peer-verified method trial was performed on white bread, multigrain bread, and coffee cake spiked with known levels of potassium bromate. The analytical portion is extracted with deionized water to remove bromate from the bulk of the baked product. The aqueous extract is carried through a series of steps to remove co-extractives that would interfere with the liquid chromatography (LC) in the determinative step or hasten the deterioration of the LC column. The extract is filtered before passing it through a reversed-phase solid-phase extraction (SPE) column and a cation-exchange column in the silver form to remove lipids and chloride, respectively. Ultrafiltration is then used to remove proteins with molecular weights of > daltons. Finally, a cation-exchange column in the sodium form is used to remove silver ions from the extract. The determinative step uses LC with a reversed-phase column and an ion-pairing agent in the mobile phase. Detection is based on the post-column reaction of bromate with o-dianisidine to form an oxidation product that is quantitated spectrophotometrically at 450 nm. Overall agreement between the submitting and peer laboratories was quite good. For bromate levels of ppb, overall mean recoveries were 76.9 and 78.8% for the submitting and peer laboratories, respectively. The standard deviations were higher for the results of the peer laboratory, probably because of the generally higher level of baseline noise present in the chromatograms. The results demonstrate that the method provides adequate accuracy with low-fat as well as high-fat foods. Bromate at levels as low as 5 ppb (ng/g) can be detected with the method. Method Authors KATSUICHI HIMATA, MASAAKI NODA, SUSUMU ANDO, and YUJI YAMADA Yamazaki Baking Co., Ltd., Central Laboratory, , Chitose, Sumida-ku, Tokyo 130, Japan, Tel.: ; Fax: Submitting Laboratory Yamazaki Baking Co., Ltd., Central Laboratory, , Chitose, Sumida-ku, Tokyo 130, Japan U.S. Food and Drug Administration, Office of Pre-Market Approval, Division of Product Manufacture and Use, HFS-248, Washington, DC Acknowledgment of Reviewers The authors would like to thank Charles R. Warner and Daniel H. Daniels of the U.S. Food and Drug Administration, for participating in this study, and providing invaluable suggestions about LC with the oxidimetric post-column reaction system and the design of the experiment for the peer-verified method trial. 1 Summary of Results of Verification Study Results obtained by the submitting and peer laboratories are summarized in Table Matrixes The method is applicable to bread and rolls. Three types of fortified bread were analyzed by the submitting and peer laboratories: white bread, multigrain bread, and coffee cake. 1.2 Number of Samples The submitting and peer laboratories each analyzed 27 test samples (3 types of bread fortified at 3 of the following levels: 0, 10, 11, 20, 22, 50, and 52 ng bromate/g) without blind duplicates for a total of 27 data points. The samples were prepared in Tokyo and split into separate portions. A set of samples was shipped to the peer laboratory as unknowns with only a letter designation. Special precautions were taken to preserve the samples at 30 C. A schedule for analysis was established so that the samples were analyzed by the peer and submitting laboratories on the same days to minimize the effects of sample decomposition that occurs during storage of the ground sample. 1.3 Accuracy and Recoveries For the submitting laboratory, individual recoveries ranged from 66 to 85% with a mean recovery of 76.9%. For the peer laboratory, individual recoveries ranged from 49 to 103% with a mean recovery of 78.8%.

2 348 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, 2000 Table 1. Summary of results of peer-verified method trial Laboratory Bromate added, ppb Mean recovery, % a Standard deviation, % Submitting 10, , , Peer 10, , , a n = Precision Repeatability (RSD r ) = 6.04% (submitting laboratory) and 14.93% (peer laboratory). Reproducibility (RSD R ) = 8.93% (submitting and peer laboratories combined). 2 Safety Precautions Observe common safety procedures for operating the centrifuge. o-dianisidine dihydrochloride is chemically related to benzidine and should be considered carcinogenic. In case of contact with the eyes, rinse immediately with plenty of water and seek medical advice. 3 Scope An LC method is described for the determination of bromate in bread when present at levels from 10 to 50 ppb. 4 References (1) Warner, C.R., Daniels, D.H., Joe, F.L., Jr, & Diachenko, G.W. (1996) Food Addit. Contam. 13, (2) Himata, K., Noda, M., Ando, S., & Yamada, Y. (1997) Food Addit. Contam. 14, Definitions For the purposes of this peer-verified method, the following definition applies: Bromate content of bread. The bromate content is expressed in nanograms per gram of product. 6 Principle Deionized water is used to extract bromate from bread samples. The extract is treated in sequence with a C 18 solid-phase extraction (SPE) column to remove lipids, a cation-exchange column with the silver cation to remove chloride, and an ultrafiltration membrane to remove proteins. Further treatment of the extract with the sodium form of propylsulfonic acid in an ion-exchange column is used to remove the silver that leached from the silver column. Bromate is separated by reversed-phase ion-pair LC and detected at 450 nm with a post-column flow reactor as the oxidation product from the reaction with o-dianisidine in acidic solution (1). A calibration curve is used for quantitation. 7 Standards Stock bromate standard solution mg bromate/l. Weigh mg potassium bromate and dissolve in deionized water. Dilute to volume with water in 100 ml volumetric flask and mix well. Store at room temperature Intermediate bromate standard solution. 100 ng/ml. Dilute 1.0 ml stock bromate standard solution to 100 ml with deionized water and mix well. Dilute 1.0 ml of this solution to 100 ml with deionized water and mix well. Store at room temperature Working bromate standard solutions. 1.0, 2.0, 4.0, and 10.0 ng bromate/ml. Dilute 1.0, 2.0, 4.0, and 10.0 ml intermediate standard solution to 100 ml with deionized water in separate volumetric flasks. Mix well. Store at room temperature. 8 Reagents Reagent grade chemicals. Potassium bromate, potassium bromide (infrared grade or ACS reagent grade), glacial acetic acid, and o-dianisidine dihydrochloride Methanol and water. LC grade Nitric acid. 70% doubly distilled from Vycor (heat-resistant glass). The stock bottle must be protected from light while in storage. Nitric acid that is not water clear should not be used because the post-column reagent will have excessive color Tetrabutylammonium hydroxide solution. 40% (w/v) aqueous solution. Methanolic tetrabutylammonium hydroxide should not be used because some products have excessive levels of bromide that interfere with the ion-interaction chroma-

3 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, Figure 1. Sample preparation procedure for bromate in bread. Figure 2. Typical liquid chromatogram of bromate in bromate standard solution and in bread sample analyzed on a 5 m C 18 column, mm id.

4 350 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, 2000 Table 2. Results from submitting laboratory for bromate found a Fortification level, ppb Sample White bread ND, ND, ND Multigrain bread ND, ND, ND Coffee cake ND, ND, ND a ND = none detected. tography. (Note: It is very important to establish from the manufacturer that the tetrabutylammonium hydroxide has not been contaminated with bromate. Some products have excessive levels of bromate that cause severe baseline disturbances.) LC mobile phase. Add ca 800 ml LC grade water to 100 ml methanol in 1 L beaker. Add 2.0 g glacial acetic acid, then add 19 g 40% tetrabutylammonium hydroxide solution. Adjust ph of mobile phase to with tetrabutylammonium hydroxide solution or acetic acid. Dilute to 1 L with LC grade water and mix well. Filter solution through 0.45 µm membrane filter. Degas mobile phase under vacuum with ultrasonification Post-column reagent. Add 90 g (or 60 ml) 70% nitric acid to 1000 ml glass stopper graduated cylinder containing ca 700 ml LC grade water. Dissolve 10 g potassium bromide in this solution. Add 15 ml LC grade methanol to 0.5 g o-dianisidine dihydrochloride contained in 30 ml screw cap glass bottle. Grind crystals gently against bottom of glass bottle with flat-tipped stirring rod to expedite dissolution of o-dianisidine. Decant supernatant liquid into 1000 ml graduated cylinder containing nitric acid solution. Stopper graduated cylinder and mix contents. Repeat solubilization of o-dianisidine with additional 15 ml LC grade methanol, and decant supernatant liquid into 1000 ml graduated cylinder. Repeat this operation until all the o-dianisidine is dissolved. Add remainder of 200 ml LC grade methanol to 1000 ml graduated cylinder. Dilute reagent to 1000 ml mark with LC grade water and mix carefully. The reagent can be stored for several weeks at room temperature under ordinary room illumination. 9 Apparatus Stirrer. Magnetic stirrer and stirring bar Centrifuge. Refrigerated centrifuge, capable of holding 50 ml centrifuge tubes and delivering rpm Centrifuge tubes. 50 ml, 26 mm id, plastic, with cap C 18 SPE column. Cartridge equipped with Luer-Lock fittings and 900 mg silica-based C 18 with ca 8% carbon loading. Condition with 10 ml methanol followed by 10 ml water Cation-exchange column. Bond Elut PRS (propylsulfonic acid), 500 mg in 3 ml column. Condition with 10 ml methanol followed by 10 ml water Cation-exchange column in the silver form. Dionex OnGuard-Ag or equivalent. A strong cation-exchange resin in the silver form in a cartridge with Luer-Lock fittings. The total capacity of the packing should be ca meq/cartridge. Condition with 10 ml water Ultrafiltration cartridge. Millipore Molcut L, or equivalent, MW cutoff Disk filters µm 25 mm id nylon membrane filter. Table 3. Results from peer laboratory for bromate found a Fortification level, ppb Sample White bread ND, ND, ND Multigrain bread ND, ND, ND Coffee cake ND, ND, ND a ND = none detected.

5 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, Table 4. Recoveries (%) and statistics of results Fortification level, ppb Laboratory Sample Mean Submitting White bread Multigrain bread Coffee cake Mean Mean of all fortifications a 76.9 Standard deviation RSD r, % Peer White bread Multigrain bread Coffee cake Mean Mean of all fortifications b 78.8 Standard deviation RSD r, % Mean of all fortifications of white and multigrain bread c 81.3 Standard deviation RSD R, % Mean of all fortifications d 77.8 Standard deviation RSD R, % a For submitting laboratory; n = 18. b For peer laboratory; n = 18. c For both laboratories; n = 24. d For both laboratories; n = LC system. LC pump for mobile phase capable of 5000 psi and 10 ml/min flow rate, and valve injector or autosampler capable of 250 µl injections Post-column reactor. Pump with poly ether ether ketone or other inert polymer; reaction coil, for example, Waters RXN 1000 COIL, or a tee-shaped mixer equipped with 3 ports for 1.6 mm tubing and the confluent stream exiting through the third port into Teflon tubing (9 m 0.53 mm id) woven with a crochet stitch, reaction coil incubator capable of heating at 60 C; UV/Vis detector capable of measuring a wavelength of 450 nm; and strip chart recorder or electronic integrator Guard column cm 4.6 mm id, with packing identical to that used in analytical column Analytical column. 5 µm porous silica microparticulate C 18 packing in 4.6 mm id 25 cm steel column. The packing should be stable when used with aqueous organic solvents containing ion pairing reagents. 10 Sample Preparation 10.1 Homogenization Slice loaf of bread, and cut each slice into small pieces, ca 2 2 2cm 3. Homogenize all the pieces in blender Storage Store bread and/or bread sample frozen to prevent deterioration and change in composition. 11 Controls Preparation Accurately weigh 10.0 g homogenized bread samples (white bread, whole grain bread, or coffee cake) into separate plastic bags (Ziploc, medium size: mm). Fortify bread samples with bromate standard solution, keeping concentrations within operating range of standard curve at 10, 20, and 50 ppb in bread, and immediately freeze samples at 30 C. (Bromate in comminuted bread samples is not stable at ambient temperature.)

6 352 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, 2000 ) Figure 3. Linear regression plot of bromate recovered vs bromate added to white bread. 12 Procedure 12.1 Extraction Accurately weigh 10 g bread sample to nearest 0.1 g into 200 ml glass beaker, and add 50 ml deionized water. Mix well using magnetic stirrer for 30 min Cleanup Centrifuge portion of sample solution at rpm (5 C) for 15 min. If centrifuge will not attain rpm, use highest rpm available and increase time to obtain enough filtrate for cleanup procedure (10 ml). (Note: To remove fat from bread extract, a refrigerated centrifuge at 5 C is preferred.) Filter supernatant through filter paper (Whatman No. 2, or equivalent). Follow sequence of cleanup procedures shown in Figure 1. Attach plastic syringe containing ml filtrate to 0.45 µm nylon syringe filter. Discard first few drops of sample solution passing through syringe filter. Attach syringe and filter to preconditioned C 18 SPE cartridge and cation-exchange column with the silver cation in tandem. Discard first 3 ml sample solution passing through this setup. Pass remaining solution through this setup, and collect ml in ultrafiltration cup. Close with screw-top cap and pressurize with air using 20 ml syringe. Collect at least 3 ml filtrate in reservoir (it may take 1 2 h). Pressurize repeatedly to obtain enough filtrate for the following cation exchange: Pass filtrate through preconditioned PRS cartridge. Discard first 1 ml sample solution passing through PRS cartridge. Collect remaining eluate for LC analysis. Analyze eluate within 24 h LC Analysis Using autosampler or sample loop injection valve, inject 250 µl sample solution, and elute with mobile phase isocratically at 1.0 ml/min. Pump post-column reagent at 0.3 ml/min, and measure oxidation product of o-dianisidine at 450 nm. Stable retention times and peak areas or heights within 3% for consecutive injections of 10 ng/ml standard bromate solutions are indicative of an equilibrated and properly operating system. After the LC system is equilibrated, inject working bromate standard solutions and sample solutions. Wait 30 min after bromate peak appears in chromatogram before next injection. Distribute working bromate standard solutions throughout each run to ensure adequate quantitation of samples. After each day s work, wash column with methanol water mixture with composition identical to that of mobile phase. Pass

7 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, ca 20 ml 100% methanol through column, and store column with 100% methanol. Before each run, pass mobile phase through column for ca 30 min. Flush column with methanol prior to storage for extended period of time (e.g., over 1 weekend). 13 Calculations Measure peak responses (heights or areas) of bromate peaks in chromatograms of working standards (Figure 2). Perform linear least-squares analysis on relationship between standard concentration (ng/ml) and peak area (height) to obtain calibration line. Use 4 working bromate standard solutions (1.0, 2.0, 4.0, and 10.0 ng/ml) for calibration line; do not use zero (0, 0). The correlation coefficient (r) of the calibration line should be >0.99. For each sample injection, calculate bromate concentration, C, in ng/ml, in sample solution from sample peak response and calibration line: Bromate, ppb = C 5 10 W where W = sample weight (g) and 5 = dilution factor. 14 Test Results Report 14.1 Recoveries Recoveries were calculated by found bromate divided by the fortification level multiplied by 100. The mean recovery obtained by the submitting laboratory was 76.9%, and that obtained by the peer laboratory was 78.8%, for a combined mean of 77.8% over a fortification range of ppb bromate. The results from the submitting and peer laboratories are shown in Tables 2 4. Values obtained with this method show good correlations between the submitting laboratory and the peer laboratory for each of the 3 types of baked goods. The results were linear (white bread, R 2 = 0.996; multigrain bread, R 2 = 0.985; coffee cake, R 2 = 0.988). The regression plots are shown in Figures 3, 4, and 5, respectively Repeatability Three types of bread fortified with 6 levels of bromate ranging from 10 to 52 ppb were used to measure repeatability. The results are shown in Table 4. Overall RSD r values for bromate were 6.04% in the submitting laboratory and 14.93% in the peer laboratory. Figure 4. Linear regression plot of bromate recovered vs bromate added to multigrain bread.

8 354 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, 2000 Figure 5. Linear regression plot of bromate recovered vs bromate added to coffee cake Reproducibility Reproducibility (and repeatability) can be fully evaluated only through a collaborative study performed as prescribed by AOAC INTERNATIONAL. Because there are fewer data points in a peer-verified method trial, statistical results are less definitive. Examination of the data suggests that the results for white bread and multigrain bread are very similar, whereas the recoveries from coffee cake are somewhat lower than the recoveries from the other 2 types of bread. For this reason, only the white and multigrain bread results were expressed as percent recoveries, and the data from both laboratories were combined to give an estimate of reproducibility. When data from the submitting laboratory and the peer laboratory were combined, the RSD R value for the recoveries from white bread and multigrain bread was 8.93%. These results are shown in Table Linearity Peak height response was linear in the concentration range from 1.0 to 10.0 ng/ml for bromate, with an R 2 of Limit of Detection The solution detection limit was 0.4 ng/ml. The method detection limit was estimated to correspond to ca 3 ppb in bread for bread samples diluted 1:5 with deionized water in the extraction procedure. The detection limit was calculated as the concentration of bromate which would result in a peak height equivalent to 3 times the standard deviation of 30 background points of baseline (ref. 2, p. 815) Limit of Quantitation Based on a calculation of 10 times the standard deviation of background noise, the limit of quantitation was estimated to correspond to ca 10 ppb bromate in bread Specificity The chromatographic separation with a selective detector provides excellent selectivity. The following oxidants do not interfere with selectivity: hypochlorite, chlorite, iodate, chlorate, hydrogen peroxide, and nitrite (an oxidant under acidic conditions). Peak identity can be established by standard addition. The variety of baked goods used in this study illustrates the ruggedness of the method in the presence of many different food ingredients.

9 HIMATA ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 2, Quality Assurance The chromatographic separation should yield a column efficiency of ca 5000 theoretical plates for bromate (1.0 ng/ml, 250 µl injection), based on the tangent method. 16 User Comments (1) The analytical procedure is clearly described. (2) All analyses proceeded smoothly and there were no problems with clogged filters or SPEs that might be expected with food matrixes such as bread and coffee cake. (3) In the peer laboratory, the reaction coil consisted of 9mof in. id Teflon tubing woven with a crochet stitch. This seemed to be as satisfactory as the commercially manufactured reaction coil used by the submitting laboratory. However, the larger standard deviations exhibited by the data from the peer laboratory may suggest that thermal control is better with commercial equipment.