A. Goodson, W. Summerfield and I. Cooper*
Pira International, Randalls Road, Leatherhead, KT22 7RU, UK
* To whom correspondence should be addressed. e-mail: IanC@pira.co.uk
(Received 16 November 2001; revised 6 March 2002; accepted 25 March 2002)
Bisphenol A (BPA) and bisphenol F (BPF) have been determined in a range of canned foods. Sixty-two different canned foods were purchased from retail outlets in the UK from January to November 2000 and the contents extracted and analysed by GC-MS for BPA and BPF isomers. The following canned products were analysed: fish in aqueous media, 10 samples; vegetables, 10; beverages, 11; soup, 10; desserts, five; fruit, two; infant formula, four; pasta, five; and meat products, five. BPF isomers were not detected in any of the canned foods with detection limits of 0.005 mg kg -1 for the 2,2 0 and 2,4 0 isomers and 0.01mg kg -1 for the 4,4' isomer. BPA was detected in 38 samples with a detection limit of 0.002mg kg -1. Of these, BPA was quantified in 37 canned foods at levels from 0.007 mg kg -1, with one sample of meat containing a mean level of 0.38 mg kg -1. All other samples contained <0.07 mg kg -1 BPA.
Keywords : bisphenol A, bisphenol F, canned foods, survey, migration, can coatings
Introduction
Food and beverage cans often have an internal polymeric coating to protect the food and prevent undesirable interactions between the metal from the can and the food. The polymeric coatings are usually highly cross-linked thermoset resins that can withstand typical processing conditions (1.5 h at 121°C). The most widely used lacquer types for food cans, where the food is retorted (sterilized) in the can to ensure preservation, are the following.
Of these coatings, the epoxyphenolic is most important, being used universally for both can bodies and ends for two- and three-piece constructions, although more usually for shallow draw cans in the case of two-piece. Beverage can bodies are commonly epoxyamine coated.
Bisphenol A (2,2'-bis(4-hydroxyphenyl) propane (BPA) is a starting substance used in the manufacture of most types of epoxy resins, which are then cross-linked and used to coat food cans. BPA is not normally present in PVC organosol coatings. However, if bisphenol A diglycidyl ether (BADGE) was used as an additive to scavenge hydrogen chloride in these coatings, residues of BPA, as unreacted starting material in the BADGE, may be present.
Another use of BPA is in the manufacture of plastic materials, in particular polycarbonates used in contact with foods. Directive 90/128/EEC lists BPA in the positive list with a specific migration limit of 3mg kg -1 (European Commission 1990).
Previous research has shown that migration of BPA can occur from can coatings to food simulants (Cooper et al. 1996). It was also shown that migration to the fatty food simulant olive oil was generally much lower compared with aqueous simulants, such as 10% ethanol.
BPA has been the subject of many studies exploring its potential endocrine disrupting properties, e.g. Vom Saal et al. (1998). These studies have largely given conflicting results.
BPF, a mixture of 3 isomers (2,2'-, 2,4'-, and 4,4'-dihydroxydiphenylmethane used commercially as a mixture in the ratio 15, 50 and 35%, respectively) can also be used in the manufacture of epoxy resins, but as a fully cross-linked polymer it is rarely used in food-contact materials. Residues of BPF isomers (Novolac) may arise from their use in the manufacture of Novolac glycyidyl ethers (NOGE), which are used as scavengers for hydrogen chloride in PVC organosol coatings.
Materials and methods
Samples
Three cans of each sample, all bearing the same batch number, were purchased from retail outlets in the south of England. The distribution of types of samples tested in this survey was similar across the UK. Collection of samples was weighted approximately 80% from supermarkets and included about 40% 'own brand' foods to reflect approximately consumer shopping habits.
Samples were stored sealed at room temperature. After cans were opened, the total contents of each can were homogenized and an aliquot taken for analysis. The remaining contents of each can were then frozen and stored in a freezer.
Calibration materials
BPA (>99%), and BPF isomers 2,2 0-dihydroxydiphenylmethane (95%) and 4,4 0-dihydroxydiphenylmethane (98%) were obtained from Sigma-Aldrich, Gillingham, Dorset, UK. 2,40-Dihydroxydiphenylmethane (97%) was obtained from Fluorochem Ltd, Old Glossop, Derbyshire, UK. BPA-d16 (98 atom% D) was obtained from Sigma-Aldrich and converted to BPA-d14 by dissolution in aqueous sodium hydroxide and reprecipitation by acidifying with dilute sulphuric acid. This was carried out to avoid the possibility of differing degrees of exchange of the deuterium atoms on the oxygens with hydrogen during the analysis . This BPA-d14 was used as an internal standard.
Analytical methodology
Published methods for determining BPA in milk or food simulants were not considered to have sufficient detection capability or robustness for the determination of BPA in a wide range of foods (Biles et al. 1997, Franz and Rijk 1997, Mountfort et al. 1997, Howe and Borodinski 1998, Kawamura et al. 1999). A more recent publication described the application of HPLC to the analysis of canned fruit and vegetables purchased from the Japanese market (Yoshida et al. 2001). No methodology was found in the literature for the determination of BPF in foods. Considering the possibility of other migrants from can coatings that exhibit UV fluorescence, e.g. hydrolysis and/or other reaction products of BADGE, BFDGE or NOGE (Biederman and Grob 1998, Lintschinger and Rauter 2000), there could be a significant potential for interferences with BPA and BPF using HPLC- fluorescence detection and, therefore, false-positives. Therefore, a method was developed to measure these analytes simultaneously using gas chromatography-mass spectroscopy (GC/MS), with deuterated BPA as an internal standard, in which BPA and the BPF isomers were acylated using acetic anhydride after isolation from the food. Derivatization of BPA and BPF was found to improve the peak shapes and the robustness of the method.
The entire contents of a can were well mixed using a MSE homogenizer. Twenty grams (g) of subsample were taken and an aliquot of internal standard added. Each subsample was extracted in a beaker with a mixture of 20ml n-heptane (for fat removal) and 20ml acetonitrile using an Ultra-Turrax T25 for 1min. The mixture was left to stand for 15min and was filtered using a GF/C filter and a Buchner flask. The heptane layer was removed using a pipette and this, together with the filter cake, was returned to the beaker for a second extraction with a fresh 20-ml portion of acetonitrile using the Ultra-Turrax. The second extract was filtered as before and the heptane layer removed and discarded. The two acetonitrile filtrates were combined and treated with 30 g anhydrous sodium sulphate for 1min and the solvent decanted into a 50-ml measuring cylinder. For the canned samples with a low fat content, e.g. vegetables, soup, desserts and pasta, heptane was omitted from the procedure above. The extract was then evaporated under nitrogen to approximately 5ml, diluted to 50ml with water and transferred to a 250-ml separating funnel for the derivatization step. Ten millilitres of 72% w/v potassium carbonate and 10ml methanol were added and swirled to mix. Of acetic anhydride, 10ml was added and swirled gently after reaction had subsided. The solution was left to stand for 15min with occasional swirling and then extracted with 5ml n-heptane. The heptane layer was collected and analysed by GC-MS. For the analysis of infant formulae, 10 g formula were extracted with 20ml acetonitrile by shaking vigorously in a 40-ml vial for 1 min. The layers were allowed to settle. The acetonitrile portion was then filtered through a Whatman GF/C filter. The residue was shaken with a further 10ml acetonitrile and filtered. The combined solutions were evaporated under nitrogen and derivatized as described above. For the analysis of beverages, the samples were opened and allowed to degas. A total of 50ml was measured into a 250-ml separating funnel and derivatized as described above.
Quantitative results were obtained using the GC-MS results by comparison against external standards. Stock solutions of BPA and BPF isomers at a concentration of about 1000mg l-1 were prepared separately in acetonitrile. These solutions were diluted to a concentration of 10mg l-1 with acetonitrile and further diluted to 50ml with water with addition of internal standard (0.1 ml 20mg l-1 BPA-d14 in acetonitrile). These standards were derivatized as described above. For each analytical run, a further portion of a food sample was `spiked' with a known quantity (about 0.1 mg kg -1) of BPA and BPF isomers and then processed in exactly the same fashion as described above. The results for this sample were used to estimate the method recovery for the respective batch and check that the method was `in control'. A solvent blank was also prepared and run alongside the samples.
In all cases, calibration curves were constructed by plotting µg BPA or BPF isomer against the respective peak area ratio to BPA-d14. Quantities were obtained from the graphs by interpolation and the concentrations in the food (mg kg -1) were calculated by dividing by the mass of food taken.
GC/MS conditions
The following GC/MS conditions were used: column, HP-5MS, 30m X 0.25 mm, 0.25 microns film; carrier gas, helium at 5 psi; injector, splitless 270°C; oven programme, 90°C hold for 2 min; ramp to 250°C at 5°Cmin-1; ramp to 300°C at 10°Cmin-1; hold 2min; ions, BPA diacetyl m/z 228, 213; BPF diacetyl isomers m/z 200; BPA-d14 diacetyl m/z 224; retention times, BPA diacetyl 32.5min; 2,2 0-BPF diacetyl, 26.6min; 2,4 0-BPF diacetyl, 28.8min; 4,4 0- BPF diacetyl, 30.9min; and BPA-d14 diacetyl, 32.4 min.
Analytical performance
Limits of detection for BPF isomers and BPA were defined by calculating concentrations equivalent to three times the signal:noise on analysis. The detection limits were 0.002mg kg -1 for BPA, 0.005 mg kg -1 for the 2,2 0 and 2,4 0 isomers of BPF, and 0.01mg kg -1 for the 4,4 0 isomer of BPF. Limits of quantification were calculated from the concentration of BPA or BPF giving a signal equal to 10 times the signal:noise on analysis.
The results were corrected for recovery using values obtained for control samples. The calibration graphs had correlation coefficients of 0.995 or better. The mean relative recoveries obtained for BPA were in the range 81-103%. For BPF isomers, the mean relative recoveries obtained (all three isomers combined) were in the range 61-128%. The wider spread of recoveries for BPF isomers is probably caused by the internal standard BPA-d14 not compensating so well for extraction efficiencies in the procedure compared with BPA. In all cases where BPA was quantified (50:007mg kg -1), the peak identity was confirmed by checking also for the m/z ion at 213.
Table 1.
BPF isomers Matrix spiked BPA (combined value) Tuna Added (mg kg -1) 0.08 0.133 Reported (mg kg -1) 0.09 0.120 Found/added (%) 113 90 Baked beans Added (mg kg -1) 0.11 0.2 Reported (mg kg -1) 0.1 0.24 Found/added (%) 91 120
Table 2.
Replicate BPA (mg kg -1) 1 0.011 2 0.011 3 0.010 4 0.010 5 0.011 6 0.010 Mean 0.011 Sr repeatability SD (mg kg -1) 0.0005 Repeatability (mg kg -1) 0.0016
The following procedures were used.
A solvent blank was processed and analysed alongside the samples for each analytical run. In all cases, no peaks were detected with similar retention times to the analytes.
After homogenization of the entire contents of the can, duplicate subsamples were taken from about half of the samples for analysis and single subsamples for the remainder.
Good agreement was obtained between replicates.
Duplicate controls consisting of a subsample fortified with about 0.1mg kg -1 BPA and BPF isomers were run with each batch of samples and recoveries calculated. Mean recoveries were within an acceptable range.
To ensure that results obtained were of acceptable accuracy, samples of baked beans and canned fish were `spiked' with known levels of BPA and BPF isomers by an independent laboratory (CSL, York) and delivered together with the unfortified samples to Pira International for `blind' analysis. Quantitative results were within 90-120% of the values for levels of BPA and BPF added by CSL (table 1).
An estimate of the repeatability (95% probability level) of the analytical method was made by conducting six replicate analyses on the homogenized contents of a can of salmon (table 2).
Table 3. Results
Pack size Pack Bisphenol A (g) type Country (mg kg -1) Vegetables Baked beans in tomato sauce 220 EO;3PC UK 0.014; 0.011 Sliced green beans in salted water 300 3PC Holland 0.037; 0.036 Processed peas in water 199 3PC UK 0.016; 0.016 Sliced carrots in salted water 180 3PC UK 0.010; 0.011 Baby carrots in sugared salt water 195 3PC Belgium 0.041; 0.042 Chopped tomatoes with garlic 400 EO;3PC Italy 0.027 Baked beans in tomato sauce 415 EO;3PC UK 0.009 New potatoes in water, salt added 300 EO;3PC Belgium 0.048 Peeled plum tomatoes in tomato juice 400 3PC Italy 0.026 Sweetcorn in water 200 EO;3PC Canada 0.016 Desserts Evaporated milk 170 3PC not stated 0.014; 0.011 Creamed rice 425 2PC UK 0.007; 0.013 Custard 425 2PC UK n/d Creamed rice 213 3PC UK 0.010 Chocolate sponge pudding 313 3PC UK n/d; n/d Miscellaneous Infant formula 900 EO;3PC UK n/d; n/d Infant formula 450 EO;3PC UK n/d; n/d Infant formula 450 EO;3PC not stated n/d Infant formula 300 EO;3PC France n/d Fruit pieces in juice 415 EO;3PC Greece 0.019 Fruit cocktail in syrup 411 EO;3PC Italy 0.038; 0.037 Spaghetti rings in tomato sauce 212 EO;3PC UK 0.041; 0.038 Pasta shapes in tomato sauce 400 EO;3PC UK 0.009; <0.007 Spaghetti in tomato sauce 400 EO;3PC UK 0.009 Spaghetti in tomato sauce 410 3PC UK n/d Pasta shapes in tomato sauce 213 EO;3PC UK 0.011 Chicken and ham in white sauce 418 3PC Holland 0.052; 0.053 Ham 200 3PC Holland 0.354; 0.422; 0.377 Hot dogs 400 3PC Holland 0.033; 0.021 Chopped pork and ham 200 EO;3PC Denmark 0.016; 0.017 Corned beef 200 EO;3PC Brazil 0.059; 0.068; 0.070 2,2-Bisphenol F not detected LOD=0.005 mg kg -1. 2,4-Bisphenol F not detected LOD=0.005 mg kg -1. 4,4-Bisphenol F not detected LOD=0.01 mg kg -1. Bisphenol A, LOD=0.002 mg kg -1, LOQ=0.007 mg kg -1. EO, easy open; 3PC, three-piece; 2PC, two-piece; n/d, not detected.
Table 4. Results
Pack size Pack Bisphenol A (g) type Country (mg kg -1) Fish Sardines in tomato sauce 120 EO;2PC Portugal 0.014; 0.014 Red salmon 213 2PC USA 0.011; 0.011; 0.01; 0.01; 0.011; 0.01 Tuna chunks in brine 185 2PC Seychelles 0.026; 0.026 Sardines in tomato sauce 120 EO;2PC Portugal 0.032 Red salmon 105 EO;2PC Canada 0.018; 0.013 Pilchards in tomato sauce 305 3PC South Africa 0.018; 0.013 Red salmon 212 2PC USA 0.018 Tuna chunks in brine 185 2PC Thailand 0.031 Skipjack tuna in mayonnaise with sweetcorn 185 2PC Thailand 0.041; 0.044 Mackerel fillets in tomato sauce 125 EO;2PC Denmark n/d Beverages Lemon & lime flavoured soft drink 330 EO;2PC not stated n/d; n/d Orange soft drink 330 EO;2PC UK n/d; n/d Orange soft drink 330 EO;2PC UK n/d; n/d Soft drink with vegetable extracts 330 EO;2PC UK n/d; n/d Strong ale 500 EO;2PC UK n/d; n/d Dry cider 440 EO;2PC UK n/d Best bitter 440 EO;2PC UK n/d Strong continental lager 440 EO;2PC UK n/d Super strength lager 440 EO;2PC UK n/d Cola 330 EO;2PC UK n/d Stout 275 EO;2PC UK <0.007 Soup Tomato soup 425 3PC UK 0.020; 0.021 Minestrone soup 295 3PC UK n/d; n/d Country vegetable soup 400 EO;3PC UK n/d Cream of tomato soup 425 3PC UK n/d; n/d Cream of tomato soup 392 3PC UK n/d; n/d Cream of chicken soup 400 EO;3PC UK 0.008; <0.007 Tomato soup 400 3PC UK n/d Chicken & Vegetable 400 EO;3PC UK n/d Chicken and white wine soup 295 3PC UK 0.021 Cream of chicken soup 290 EO;3PC UK 0.010 2,2-Bisphenol F not detected LOD=0.005 mg kg -1. 2,4-Bisphenol F not detected LOD=0.005 mg kg -1. 4,4-Bisphenol F not detected LOD= 0:01 mg kg -1. Bisphenol A , LOD=0.002 mg kg -1, LOQ=0.007 mg kg -1. EO, easy open; 3PC, three-piece; 2PC, two-piece; n/d, not detected.
Figure 1. GC/MS trace showing the separation of bisphenol A and bisphenol F, diacetyl derivatives in a calibration standard.
Figure 2. Overlay GC/MS traces of ham extract and spiked ham extract.
Results and discussion
All results are presented in tables 3 and 4, and example GC/MS traces are given in figures 1 and 2. The results have been corrected for recovery obtained from the control samples. The calibration graphs had correlation coefficients of 0.995 or better. The mean relative recoveries obtained for BPA were in the range 81-103%. For BPF isomers the mean relative recoveries obtained were in the range 61-128%.
In all cases where BPA was quantified (50:007mg kg -1), the peak identity was confirmed by checking also for the m/z ion at 213.
There are few data from previous surveys of food or drink for BPA with which to compare the results. However, levels of BPA in canned vegetables were similar to those reported previously (Brotons et al. 1995, Yoshida et al. 2001).
Conclusions
Sixty-two different canned foods were purchased from retail outlets from January to November 2000 and the contents were extracted and analysed by GCMS for BPA and BPF isomers. BPF isomers were not detected in any of the canned foods with detection limits of 0.005mg kg -1 for the 2,2 0 and 2,4 0 isomers and 0.01mg kg -1 for the 4,4 0 isomer. BPA was detected in 38 samples with a detection limit of 0.002 mg kg -1. Of these, BPA was quantified in 37 samples with one sample of canned meat containing a mean level of 0.38mg kg -1. The result for this one sample probably reflected the use, at that time, of BPA as a cross-linking agent in the resin used to coat the can. All other samples contained <0.07 mg kg -1 BPA.
The GC/MS analytical method developed in this study has been shown to be reliable and have low limits of detection for BPA and BPF and it can be applied to a wide range of food types.
After completion of this survey, the UK Food Standards Agency submitted the results for consideration by the independent Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment (COT). In their statement, the COT acknowledged the uncertainties that existed in the scientific understanding of potential endocrine effects of BPA. Nevertheless, on present evidence, it concluded that the levels of BPA identified in canned foods analysed in this survey are unlikely to be of concern to health, and that there is no reason for consumers to change their source of foodstuffs as a result of these findings.
Acknowledgements
The authors thank the Food Standards Agency, UK, for funding this work.
References
Biederman, M., and Grob, K., 1998, Food contaminants from epoxy resins and organosols used as can coatings: analysis by gradient NPLC. Food Additives and Contaminant s, 15, 609-618.
Biles, J. E., McNeal, T. P., and Begley, T.H., 1997, Determination of bisphenol A migrating from epoxy can coatings to infant formula liquid concentrates. Journal of Agricultural Food Chemistry, 45, 4697-4700.
Brotons, J. A., Olea-Serrano, M. F., Villalobos, M. Pedraza, V., and Olea, N., 1995, Xenoestrogens released from lacquer coatings in food cans. Environmental Health Perspectives, 103, 608-612.
Cooper, I., Bristow, A., Tice, P. A., and O'Brien, A. P., 1996 Methods of Analysis to Test for Migration from Coatings on Metal Containers. Final Report Ministry of Agriculture, Fisheries and Food, UK, Project FS 2217.
European Commission, 1990, Commission Directive 90/128/EEC of 23 February 1990 relating to plastics materials and articles intended to come into contact with foodstuffs. Official Journal of the European Communities, L75/10, 19 March 1990.
Franz, R., and Rijk, R., 1997, Development of methods of analysis for monomers and other starting substances with SML and/or QM limits in Directives 90/128/EEC and 92/39/EEC. Standards , Measurements and Testing Programme. Report EUR 17610 EN.
Howe, S. R and L. Borodinski, L., 1998, Potential exposure to bisphenol A from food contact use of polycarbonate resins. Food Additives and Contaminant s, 3, 370-376.
Kawamura, Y., Sano, H., and Yamada, T., 1999, Migration of bisphenol A from can coatings to drinks. Journal of the Food Hygienic Society of Japan, 40, 158-165.
Lintschinger , J., and Rauter, W., 2000, Simultaneous determination of bisphenol A diglycidyl ether, bisphenol F diglycidyl ether and their hydrolysis and chlorohydroxy derivatives in canned foods. European Food Research Technology, 211, 211- 217.
Mountfort, K., Kelly, J., Jickells, S. M., and Castle, L., 1997, Investigations into the potential degradation of polycarbonate baby bottles during sterilisation with subsequent release of bisphenol A. Food Additives and Contaminants, 14, 737-740.
Vom Saal, F. S., Cooke, P. S. Buchanan, D. L., Palanza, P., Thayer, K. A. Nagel, S. C., Parmigiani, S., and Welshons, W. V., 1998, A physiologically based approach to the study of bisphenol A and other estrogenic chemicals on the size of reproductive organs, daily sperm production and behaviour. Toxicology and Industrial Health, 14, 239-260.
Yoshida, T., Horie M., Hoshino, Y., and Nakazawa, H., 2001, Determination of bisphenol A in canned vegetables and fruit by high performance liquid chromatography. Food Additives and Contaminant s, 18, 69-75.
|
If you have come to this page from an outside location click here to get back to mindfully.org |