Food contamination from epoxy resins and
organosols used as can coatings:
analysis by gradient NPLC 

Food Additives and Contaminants, 1998, v.15, n.5, 609-618 1may98

(Received 15 September 1997; revised 10 December 1997; accepted 31 December 1997)

Keywords: bisphenol-A diglycidyl ether (BADGE), epoxy resins, Bisphenol-F diglycidyl ether (BFDGE), Novolak glycidyl ether (NOGE), canned foods

Most cans for food preserves are internally coated by an epoxy resin or an organosol (PVC) stabilized with epoxy compounds (scavenger for hydrochloric acid). These coatings may release Bisphenol-A diglycidyl ether (BADGE), Novolak glycidyl ethers (NOGE), as well as oligomers and derivatives into the packed foods.

In Switzerland. law requires that BADGE is not detectable in foods at a detection limit of 20 µg/kg. Samples were confiscated if their BADGE content exceeded 100 ug/kg. NOGE is not accepted as a component for coatings in contact with foods and, in fact, the same limit was applied as for BADGE.

The Scientific Committee for Foods (SCF) of the EU recommends (SCF 1997) the application of a limit of 1 mg/kg in foods for the sum of BADGE, the two hydrolysis products (BADGE.H₂0 and BADGE.2H₂0) and the two chlorohydroxy derivatives (BADGE.HCl and BADGE.2HC1).

The analysis of BADGE and Bisphenol-F diglycidyl ether (BFDGE) from epoxy resins in packing materials and aqueous foods by reversed phase HPLC (RPLC) was reported by the groups of Paseiro Losada (Paseiro Losada et al. 1991, Simal Gandara et al. 1993) and Castle (Philo et al. 1994, Sharman et al. 1995). Cottier et al. (1997) identified migrants from organosols stabilized with BADGE.

Release of BADGE into the canned foods was shown to be a problem for oily foods, since the oil swells and extracts the coating and protects the epoxy group against hydrolysis. Concentrations were found sometimes to exceed 10mg/kg in the whole can contents (Biedermann et al. 1996). Slightly more than half of the canned foods containing oil or fat exceeded the Swiss limit. In the meantime. great efforts have been taken to improve the situation and most of the products presently sold on the Swiss market release less than 20 µg/kg of BADGE.

Global migration from can coatings (gravimetrically determined extract with water, water/ethanol, 3% acetic acid, or heptane) is typically in the range of 1-5 mg dm², the Swiss legal limit being 10 mg/dm² for cans with less than 500 ml internal volume. Expressed as concentration in the food of a typical fish can, this corresponds to 10-50 mg/kg. Can coatings normally contain 1-3% lubricant (Oldring 1996), often mineral oil products (Grob et al. 1997), which may explain a substantial part of the migrants in heptane, but not that in the more polar solvents. The question has not been answered, of what this released material really consists.

Liquid chromatograms recorded with fluorescence detection, rather selectively revealing phenolic compounds, demonstrate the frequent presence of substantial amounts of materials other than BADGE. In a previous paper we reported results from canned foods containing chlorohydroxy derivatives as well as oligomers of BADGE (Biedermann et al. 1997). In total 200 samples have been analysed. The 30 samples containing BADGE at detectable concentrations were further analysed for the chlorohydroxy compounds BADGE.HCI and BADGE.2HC1, formed by reaction of BADGE with hydrochloric acid released from organosols during heat curing. In most instances, concentrations of BADGE.HCI were in the range of 15-50% of that of BADGE, while BADGE.2HC1 concentrations were again several times lower. Some samples, however, primarily contained BADGE.2HCl (maximum of 2700 pg kg in the can content, going along with 50 µg/kg of BADGE only). The coating was brownish, suggesting an overheated organosol. For the glycidoxy-terminated dimers and trimers of BADGE, usually higher concentrations were determined than for BADGE. In an extreme sample, 80 µg/kg of BADGE came along with 1600 µg/kg of dimer and 1800 µg/kg of trimer.

Some can coatings released epoxy Novolaks (NOGE) instead of epoxy Bisphenol-A compounds. BFDGE is the two-ring component of NOGE and was the NOGE component commonly analysed in the past (Simal Gandara et al. 1992). In contrast to the reaction between phenol and acetone (resulting in Bisphenol-A), condensation of phenol and formaldehyde may occur in the ortho as well as in the para position of the phenol. This is why BFDGE consists of three isomers and is always accompanied by compounds with three and more rings. Each class of a given ring number again exists as numerous isomers. Hence NOGE, the closely related alternative to BADGE, is a highly complex mixture.

Of 217 canned food samples analysed, 20 contained BFDGE at concentrations exceeding 20 µg/kg (Bronz et al. 1997). For some of them, the three- and fourring NOGE were analysed. Their concentrations substantially exceeded that of BFDGE. The chlorohydroxy derivatives, the hydrolysis products, and the oligomers of NOGE must also be expected in many of these samples, resulting in an exceedingly complex mixture. Only BFDGE.HCI has been analysed thus far. Concentrations were below those of BFDGE, as commonly observed in the instance of BADGE.

In the past (Biedermann et al. 1996), determinations were performed with normal phase HPLC (NPLC) and NPLC-NPLC with heart cutting. Oligomers of BADGE and NOGE with more than two rings were

analysed by a different NPLC method. NPLC was chosen because solutions of edible oils or extracts of oil-containing food samples could be injected without prior clean-up. Routine analysis requires a more simple analytical method, however. A new NPLC method with gradient elution is described here, which enables the analysis of all the compounds discussed above in one single run, together with additional compounds of the same type.

In the interest of optimizing can coatings, the window of the analytical method used for evaluation should be broad in order to provide a picture of the materials released which is as complete as possible. In the course of eliminating BADGE, it must be prevented that just other compounds with similar problems are introduced.

LC was performed on an automated LC-GC system (Dualchrom 3000, C.E. Instruments. Milan. Italy) equipped with an autosampler, two syringe pumps, and a fluorescence detector (Merck F1050). For GCMS, an UltraTrace gas chromatograph equipped with an autosampler for large volume on-column injection and a vapour exit was coupled to a mass spectrometer MD-800 (all C.E. Instruments).

LC was performed on a 25 cm x 2 mm i.d. column packed with the cyano phase GromsilCN 2 PR 5 pm (Stagroma, Wallisellen, Switzerland). Pentane of technical grade (Siegfried, Zofingen, Switzerland) and methyl-tert.-butyl ether (MTBE, Merck, Darmstadt. Germany) were redistilled.

GC involved an 8 m x 0.25 mm i.d. capillary column coated with a 0.2 pm film of SOP-50. a symmetric 50% phenyl polysiloxane obtained from W. Blum (Novartis, Basel, Switzerland). The system ahead of the early vapour exit consisted of a 1.5 m x 0.53 mm i.d. uncoated precolumn. deactivated by phenyldimethyl silylation or by coating with a 1 nm layer of OV-1701, and a 1.5 m x 0.32 mm i.d. retaining precolumn coated with PS-255 (a methyl polysiloxane. Fluka) of 0.15 pm film thickness.

Standards of BADGE and epoxy resins were obtained from Ciba (Basel, Switzerland) and Vernicolor

(Grüningen, Switzerland). The following standards are available from Fluka (Buchs, Switzerland): BADGE (15138), Bisphenol-A-bis-(chlorohydroxypropyl) ether (BADGE.2HCl, 15136) and Bisphenol-A-bis-(2,3-dihydroxypropyl) ether (15137). Mixtures of BADGE oligomers and of NOGE are available from Haase-Aschoff (Bad Kreuznach, Germany).

Sample preparation for foods. For the analysis of the edible oil phase, a 10% solution of the oil was prepared in 15% dichloromethane/pentane. Whole food samples were homogenized after a 1:1 addition of water; I ml of the resulting slurry was extracted with 5ml of 15% dichloromethane/pentane.

Analysis of empty cans. Cans were filled with acetonitrile and allowed to stand at ambient temperature for 24 h. To I ml of the acetonitrile extract, I ml of 15% dichloromethane/pentane was added and the resulting solution filled up to 25 ml with water to split the phases.

LC-analysis. Eighty pl of the dichloromethane/ pentane phase were injected into the LC. LC-pump 1 contained 20% MTBE/pentane, and pump 2 50% 1-propanol/20% MTBE/30% pentane (the MTBE concentration was kept constant in order to ensure constant detector response). The flow rate was 400µl/min; 100% pump 1 (9 min), then 2%/min gradient up to 30% pump 2 and 5%/min to 60% pump 2 (5 min). The column was reconditioned for 5 min. The total analysis time was 38 min. Detection occurred at 225/295 nm. Related to the can content, the detection limit for BADGE and BFDGE was 10µg/kg. It was 300 µg/kg for the complex mixture of NOGE.

Quantitation. Owing to the lack of pure standards for most of the compounds analysed, quantitation was based on the assumption that all BADGE- and NOGE-type components (chlorohydroxy compounds, oligomers) had the same response on the fluorescence detector. This assumption was checked and found to be correct within 10% for BADGE, BADGE.2HC1 and BFDGE. For the other compounds it was observed that they had identical maxima in the fluorescence spectra.

Also because of the lack of reference materials, no detailed testing of extraction efficiency could be performed. It was, however, observed that re-extraction of samples showed less than 10% of the materials observed in the first extraction.

GC-MS analysis. Large volume on-column injection was applied in order to obtain sufficient sensitivity for GC-MS of LC fractions. The mobile phase was evaporated to 200 µl; 100-150 pl were injected using concurrent eluent evaporation with the on-column interface (Grob, 1991). The injection rate by the autosampler was 3µl/s. Helium was provided at 110 kPa. During transfer, the column was at 80°C; it was then programmed at 15°C/min to 200°C, at 7°C/min to 280°C and at 20°C/min to 320°C (10min, removing triglycerides). The vapour exit was closed 40s after starting the injection.

Standards, extracts, or LC-fractions were acetylated in order to demonstrate the presence of hydroxy and phenol groups. In a second step, they were trifluoroacetylated for the determination of epoxy groups. Acetylation was conducted under conditions derivatizing hydroxy and phenol functions, but leaving epoxy groups unaffected. From extracts or LC-fractions, the solvent was evaporated. Twenty pl each of pyridine and acetic anhydride were added and the mixture kept at ambient temperature for 5 min. Then the reagent was removed by a stream of nitrogen. The reaction with trifluoroacetic anhydride (TFAA) opens the epoxide to form a bis-trifluoroacetate. To the (usually) acetylated and dried sample, 50-100µl of TFAA were added and the sample allowed to react for 15 min at ambient temperature. Then TFAA was removed by a stream of nitrogen.

The reaction conditions were optimized using a mixture of BADGE and Bisphenol-A. As shown in figure 1, acetylation removed Bisphenol-A without affecting BADGE and its glycidoxy-terminated dimer (present as an impurity of BADGE). Bisphenol-A-diacetate was detected by GC-MS, but not in the liquid chromatogram, because it is not fluorescent under the conditions used. As shown in the bottom chromatogram, trifluoroacetylation also removed BADGE and its dimer from the liquid chromatogram. Again, the reaction product was found in GCMS, but apparently is no longer fluorescent.

Figure 1. Acetylation for the determination of phenol and hydroxy groups: trifluoroacetylation for the detection of epoxy compounds. NPLC-FD chromatograms showing a test mixture with BADGE and Bisphenol-A.

The top chromatogram in figure 2 was obtained from a Bisphenol-A epoxy resin (Araldit GT 7071, Ciba. Basel, Switzerland). Peaks were identified by RPLCMS as epoxidized compounds with up to seven Bisphenol-A BADGE units (Biedermann et al. 1997). The centre chromatogram shows an epoxy Novolak (NOGE, Araldit EPN 1179. Ciba). Also from RPLC-MS and transferring fractions of identified peaks from RPLC to NPLC (Bronz et al. 1997), the peaks were assigned to the groups of isomers of given number of aromatic rings.

The bottom chromatogram of figure 2 shows the three compounds which have largely determined the selection of the gradient. The sample consisted of a can extract, to which extra virgin olive oil and BADGE.2HC1 have been added. The peak labelled as `olive oil peak' is an unidentified component in extra virgin olive oils. Using 2-propanol/pentane gradients, all three compounds were co-eluted.

Figure 2. Gradient-NPLC-FD chronratogrants of a Bisphenol-A and a Novolak epoxy resin, as tivell as of the critical separation between BADGE.2HCI, a fluorescing component typically found in extra virgin olive oil, and the cyclo-dimer of BADGE.

Extracts of cured epoxy Bisphenol-A resins were frequently found to contain a component producing a predominant LC signal shortly before the glycidoxy-terminated dimer. Corresponding chromatograms are shown for an empty tube (for mustard) and a can in figure 3. Recovered from LC, this component produced a peak in GC-MS with a spectrum corresponding to that shown by Biedermann et al. (1997) and then assigned to the phenolglycidoxy-terminated dimer of BADGE. There were two arguments casting doubt on this interpretation. This compound seemed to be left over from the polymerization process and it could not be explained why it was not integrated into the polymer through its phenol or epoxy group as the other, similar compounds. Secondly, it was eluted from the LC column before the glycidoxy-terminated dimer. Since BisphenolA eluted far after BADGE (2-propanol/MTBE/ pentane gradient), the phenol-glycidoxy-terminated dimer of BADGE would also have been expected to be eluted after the glycidoxy-terminated dimer.

Figure 3. Extracts from empti', unused cans and a tube; concentrations correspond to those of extracts from canned foods in the following figures.

These disturbing aspects are both removed if a cyclic structure is assumed, the epoxy group having reacted with the phenol on the other end of the molecule. This would eliminate the two reactive groups and leave just two hydroxyl groups. The structure of the acetate is shown in figure 4. Although formed from Bisphenol A and BADGE, the cyclic dimer is formally the cyclo-Bisphenol-A monoglycidyl ether dimer. Since ring formation does not change the molecular weight, the mass spectrum does not distinguish between the linear and the cyclic molecule. The cyclic structure explains the low reactivity (polymerization is conducted under conditions avoiding reaction with hydroxyls) and renders the LC-retention time plausible. The fact that the compound forms two peaks with an approximate 1:1 ratio suggests two stereo isomers. The two hydroxy groups can, in fact, be in cis or trans positions to each other, depending on the side from which the phenol attacks the epoxy group.

Figure 4. GC-MS and structure of the acetylated cyclodimer of BADGE.

Further confirmation of the structure was obtained from the two-step acylation procedure outlined in the experimental section. The top chromatogram of figure 5 shows the (strongly diluted) extract from an unused can, with a predominant, split peak of the compound of interest. Acetylation shifted the peak to a much earlier retention time. The second chromatogram shows partial acetylation of the component, resulting in double peaks for the mono- and the diacetates. For the earlier eluted pair of peaks, the MS confirmed the presence of two acetyl groups (figure 4). Reaction with TFAA had no effect on the peaks of interest, ruling out the presence of an epoxy group. The chromatogram also shows that about a quarter of the material observed by the small peaks in the upper chromatogram could be acetylated, i.e. that merely a minor part of the material contained hydroxyl or phenol groups. Additional peaks, particularly at the rear of the chromatogram, probably represent higher molecular weight components which were moved into the window analysed by acetylation.

Figure 5. Confirmation of the structure of the cyclodimer by acylation.

Figure 3 shows selected chromatograms of extracts from new, empty cans and a tube (obtained using acetonitrile and re-extraction with dichloromethane/ heptane). The chromatogram at the bottom shows an example of what was found in about 30% of all cans and extracts from canned foods: there is hardly any material fluorescing under the conditions used. The largest peak, the cyclo-dimer, would have reached about 50 µg/kg when fully transferred to the food filled into the can. The two other chromatograms show more substantial amounts of released material. They differ from each other as they also do from other extracts. No peak other than the cyclo-dimer has been identified so far.

Figure 6 shows the extract of a new, empty can as such, as well as after acylation. Acetylation reduced the retention time of most peaks. The most important exceptions are marked by arrows and represent compounds without hydroxy or phenol groups. Trifluoroacetylation left these peaks unaffected, indicating also the absence of epoxy groups. It did, however, remove nearly half of the other major peaks, suggesting the presence of numerous components with epoxy groups.

Figure 6. Extract of an empty can as such as it ^,ell as after acylation.

Figure 7 shows extracts from canned fish in oil. The bottom chromatogram contains a predominant peak of the cyclo-dimer, suggesting the presence of a Bisphenol-A resin. No BADGE or oligomers are visible, presumably because they fully reacted with a hardener. The centre chromatogram is similar. However, this sample contained 35 µg/kg of BADGE (small peak in the early part), as well as the oligomers of BADGE. The composition is similar to a type 7 Bisphenol-A epoxy resin (not shown). Compared with that shown in figure 2 (top chromatogram), the cyclodimer is enriched about 100-fold, suggesting that the epoxy compounds were largely, but not fully removed by condensation, whereas the less reactive cyclodimer was left behind. Some of it might have been formed during the curing process only.

Figure 7. Extracts of canned foods. Practically all peaks are assumed to be related to components from the coatings of the cans. In the early part of the chromatograms, there is a system peak apparently resulting from the start of the gradient.

The top chromatogram shows a totally different pattern of contamination, i.e. from a NOGE-containing coating. It resembles the epoxy Novolak shown in figure 2, with the larger molecular weight components being more prominent. The total concentration in the packed food of the NOGE compounds with a molecular weight up to 1000 (including the six-ring structures) approached 20 000 µg/kg. BFDGE was present at a concentration of about 7 µg/kg, but is not visible at the high attenuation shown. This two-ring NOGE has probably been almost completely removed by evaporation during stoving. It demonstrates that the analysis of BFDGE alone may yield a totally misleading picture.

Figure 8. Extract of a canned tuna sample. The fraction collected (marked in the top chromatogram) was acylated in order to characterize the poorly resolved material eluted after the cyclo-dimer.

Figure 8 shows an extract from canned tuna in oil with the peak pattern characteristic of a well cured Bisphenol-A resin; the sample contained about 5 µg,/ kg of BADGE and no significant amounts of epoxy oligomers. From the LC of the extract, a fraction was recovered (marked in the chromatogram) which included the cyclo-dimer (1600 µg/kg) as well as the material up to the retention time of the BADGE decamer (almost 3000 µg/kg). From the comparison with many other samples of tuna in oil it is assumed that practically all the material detected was released from the can. After acetylation, the cyclo-dimer markedly changed retention time. The complex mixture of the other material was also affected: many peaks were eluted earlier than before, but shifts were smaller than for the dimer with two hydroxy groups. Trifluoroacetylation had some minor effects, ruling out the presence of much epoxidized material. The material behaved similarly to that shown in figure 5 and resulted in similar peak patterns. This confirms its origin from the coating of the can.

In July 1997, 30 samples of canned fish products and two samples of meat products were collected from Swiss food stores. Concentrations of the compounds analysed are shown in table 1. In no sample did BADGE exceed 100 pg/kg; four samples slightly exceeded the Swiss legal limit (not detectable with a detection limit at 20µg/kg). This was (once more) a demonstration that it is technically feasible to pack even the most problematic products in cans releasing virtually no BADGE.

Results for BFDGE were similar: one sample slightly exceeded 100 µg/kg, an additional one was between 20 and 100 µg/kg. However, as mentioned above, BFDGE is an analytical standard rather than a technical product. The material really used is a complex mixture of NOGE. Considering NOGE with more than two rings, i.e. components hardly analysed thus far, results were, in fact, far worse: in 6 of the 32 samples, concentrations exceeded 1000 µg/kg. A chromatogram of the most extreme sample is shown in figure 7. Concentrations listed under `unknown' refer to peaks eluted shortly before the three-ring NOGE, always appearing together with NOGE, but of still unknown identity.

Table 1. Concentrations (µg/kg) of coating components in canned foods from the Swiss market in July 1997. Empty fields: concentrations below the detection limit of about 100 µg/kg

					                Epoxy Novolak               Epoxy Bis-A compounds   .
Can 	Food content 		BADGE  BFDGE  Unknown  3-Ring  4-Ring  5-Ring  DiBA  TriBA  CycloDi  Unknown
2p 	Tuna in oil 		<20    20     600      500     400     600     
3p ea d Tuna in oil 		<20   <20     						    1400     3000
2p ea 	Anchovies in olive oil	<20   <20
2p ea 	Anchovies in olive oil 	<20   <20
2p ea 	Sardines in oil 	 30   110     1700     1000    900     700
3p ea 	Meat loaf 		<20    70		180     80			    <280
2p 	Tuna in olive oil 	<20   <20						     700     1400
3p ea d Tuna in olive oil 	<20   <20						     630     2300
2p ea 	Anchovies in olive oil 	<20   <20 
2p ea	Anchovies in olive oil 	<20   <20 
2p nb d Tuna in olive oil 	<20   <20 						    1000     2000
2p ea 	Sardines in sauce 	<20   <20 						     650
2p ea d Sardines in olive oil 	<20   <20 
2p ea 	Tuna in salt water 	<20   <20 
2p ea 	Tuna in oil 		<20   <20     1400      740    240    1700
2p 	Tuna in oil 		<20   <20 
2p 	Tuna in oil 		<20   <20      800      800   2700    3400
2p 	Tuna in oil 		<20   <20      350     2000   7000    5400
2p ea d Fish pâté 		<20   <20						     300
2p ea d Salad with tuna 	<20   <20
3p 	Corned beef 		<20   <20						    <125
2p ea 	Mackerel in oil 	<20   <20						    1600     1400
2p ea 	Sardines in olive oil 	<20   <20						    1550     1800
2p ea 	Anchovies in olive oil 	<20   <20
2p ea 	Anchovies in olive oil 	<20   <20
2p ea d Sardines in olive oil 	 35   <20					200  400     800
2p ea d Herring with sauce 	 35   <20						     600
3p ea d Tuna in olive oil 	 30   <20						    1350     1800
2p 	Tuna in oil 		<20   <20		100    500     600
2p 	Tuna in oil 		<20   <20						    1600     2800
2p 	Tuna in oil 		<20   <20      340					     300      900
2v 	Tuna in oil 		<20   <20					200

2p. 3p, Two or three-piece can. 
ea. Can with easy-open lid. 
d. Decorated.

Only one sample contained noticeable concentrations of epoxy-terminated BADGE dimers and trimers. As suggested earlier (Biedermann et al. 1997), absence of BADGE is correlated with the absence of corresponding oligomers owing to rather complete removal by the curing process. In 11 samples, the cyclic dimer exceeded 500µg/kg; in six (20%) it also exceeded 1000 µg/kg. Nothing is known about the toxicity of this compound. It was accompanied by similar concentrations of a complex mixture of unknown compounds eluted later.

About a third of the samples analysed by the method described were virtually free of fluorescent contaminants (detection limit of 5-304g/kg). Since six-ring NOGE and heptamers of BADGE were within the window analysed, probably all phenolic components with a molecular weight up to 1000 were included (i.e. all those considered as potentially harmful). So far, these cans seem to be clean and it just remains to hope that in future all cans will be alike.

Among the compounds detected by fluorescence, the cyclo-dimer of BADGE was predominant. The coating industry will have to find out at what stage it is formed and whether it can be avoided. More work should, furthermore, be invested into the characterization of the complex material eluted after the cyclodimer and often present in large quantities. Its fluorescence suggests phenolic structures of the type of Bisphenol-A. Acylation experiments indicated the presence of few hydroxy or phenol groups and the almost complete absence of epoxy functions. Finally, there seems to be a variety of materials occasionally present in coatings, the identity of which is also unknown.

The acceptability of NOGE in coatings in contact with foods must be clarified. Some laboratories regularly analyse BFDGE. However, BFDGE was often just a minor compound of the product applied and can, furthermore, be removed by evaporation rather easily. Concentrations of BFDGE thus easily provide an unrealistically positive picture. One of the samples

analysed illustrated this: 7µg/kg of BFDGE was accompanied by about 20 000 µg/kg of three- to sixring NOGE. NOGE is similar to BADGE in structure and is hardly an acceptable replacement for the latter.

NOGE is not acceptable as a component for coatings in contact with foods, neither by Swiss regulations, nor by those of many other European countries. It is produced in steps analogous to Bisphenol-A (Novolak) and BADGE (NOGE) and it seems unjustified to argue that NOGE is a polymer of accepted monomers (formaldehyde, phenol, and epichlorohydrin). In organosols, NOGE is an additive just like BADGE and has never been legalized as such.

The method still does not enable detection of other, non-fluorescing compounds possibly released from can coatings, such as catalysts, accelerators, epoxidized edible oils, amino resins, acrylic resins, various esters, waxes, and lubricants.

BIEDERMANN, M., GROB. K., BRONZ, M., CURCIO, R., HUBER, M., and LOPEZ-FABAL, F., 1996. Bisphenol-A-Diglycidyl Ether (BADGE) in edible-oil-containing canned foods: determination by LC-LC-fluorescence detection. Mitteilungen aus dem Gebiete der Lebensmitteluntersuchung und Hygiene. 87, 547-558.

BIEDERMANN, M., BRONZ. M., GROB, K., GFELLER, H., and SCHMID, J. P., 1997, BADGE and its accompanying compounds in canned oily foods: further results. Mitteilungen aus dem Gebiete der Lebensmitteluntersuchung und Hygiene. 88, 277-292.

BRONZ, M., BIEDERMANN, M., GROB, K., GFELLER, H., and SCHMID, J. P., 1997, Diglycidyl ethers of Bisphenol-F and Novolak in canned oily foods. Mitteilungen aus dem Gebiete der Lebensmittelntersuchung und Hygiene, 88, 525-539.

COTTIER, S., RIQUET, A. M., FEIGENBAUM, A., POLLET, B., LAPIERRE, C.. and MORTREUIL, P., 1997, Identification of potential migrants from a vinylic organosol varnish by GC-MS and LC-MS. Journal of Chromatography, 771, 366-373.

GROB, K., 1991, On-line LC-GC (Heidelberg: Hüthig).

GROB, K.. HUBER, M.. BODERIUS., U., and BRONZ, M., 1997, Mineral oil material in canned foods. Food Additives and Contaminants, 14,83-88.

OLDRING, P. (ed.), 1996. (Vaterborne & Solvent Based Surface Coatings Resins and Their Applications, Volume II, Waterhorne & Solvent Based Epoxides and their End User Applications (Wimbledon: SITA Technology Ltd).

PASEIRO LOSADA, P., LOPEZ MAHIA, P., VAZQUEZ ODERIZ, L., SIMAL LOZANO, J., and SIMAL GANDARA, J., 1991, Sensitive and rapid RP LC-fluorescence method for determining BADGE in aqueous-based food simulants. Journal of' the Association of Official Analytical Chemists, 74, 925-928.

PHILO, M. R., JICKELLS. S. M., DAMANT, A. P., and CASTLE, L., 1994, Stability of plastic monomers in food-simulating liquids under EU Migration Test conditions. Journal of Agricultural and Food Chemistry, 42, 1497-1501.

SCIENTIFIC COMMITTEE FOR FOODS (SCF), 1997, Clarification and Explanation of the SCF's Opinion of 7 June 1996 on Badge (expressed on 13 June 1997).

SHARMAN, M., HONEYBONE, C. A., JICKELLS, S. L.. and CASTLE, L., 1995, Detection of residues of the epoxy adhesive component BADGE in microwave susceptors and its migration into food. Food Additives and Contaminants, 12, 779-787.

SIMAL GANDARA, J., PAZ ABUIN, S., LOPEZ MARIA, P., PASEIRO LOSADA, P., and SIMAL LOZANO, J., 1992. Identification of RP-HPLC peaks of bisphenol F and of bisphenol F diglycidyl ether and its hydrolysis products by thermospray MS and GCMS. Chromatographia, 34, 67-72.

SIMAL GANDARA, J., LOPEZ MARIA, P., PASEIRO LOSADA, P., SIMAL LOZANO, J., and PAZ ABUIN, S., 1993, Overall migration and specific migration of BADGE monomer and m-xylenediamine hardener from an optimized epoxy-amine formulation into water-based food simulants. Food Additives and Contaminants, 10, 555-565.