Occurence of as heat-induced contaminant of carrot juice for babies in a general survey of beverages Dirk W Lachenmeier, Helmut Reusch, Constanze Sproll, Kerstin Schoeberl, Thomas Kuballa

To cite this version:

Dirk W Lachenmeier, Helmut Reusch, Constanze Sproll, Kerstin Schoeberl, Thomas Kuballa. Oc- curence of benzene as heat-induced contaminant of carrot juice for babies in a general survey of bever- ages. Additives and Contaminants, 2008, 25 (10), pp.1216-1224. ￿10.1080/02652030802036230￿. ￿hal-00577446￿

HAL Id: hal-00577446 https://hal.archives-ouvertes.fr/hal-00577446 Submitted on 17 Mar 2011

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Food Additives and Contaminants

For Peer Review Only

Occurence of benzene as heat-induced contaminant of carrot juice for babies in a general survey of beverages

Journal: Food Additives and Contaminants

Manuscript ID: TFAC-2008-015.R1

Manuscript Type: Original Research Paper

Date Submitted by the 28-Feb-2008 Author:

Complete List of Authors: Lachenmeier, Dirk; Chemisches und Veterinäruntersuchungsamt Karlsruhe Reusch, Helmut; Chemisches und Veterinäruntersuchungsamt Karlsruhe Sproll, Constanze; Chemisches und Veterinäruntersuchungsamt Karlsruhe Schoeberl, Kerstin; Chemisches und Veterinäruntersuchungsamt Karlsruhe Kuballa, Thomas; Chemisches und Veterinäruntersuchungsamt Karlsruhe

Chromatography - GC/MS, Chromatography - Headspace, Market Methods/Techniques: basket survey, Survey

Additives/Contaminants: Benzene, Additives general, Metals, Process contaminants

Food Types: Baby food, Beer, Beverages, Fruit juices

http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 1 of 21 Food Additives and Contaminants

1 2 3 4 5 1 Occurrence of benzene as heat-induced contaminant of carrot 6 7 8 2 juice for babies in a general survey of beverages 9 10 11 3 12 13 14 4 Dirk W. Lachenmeier ∗∗∗, Helmut Reusch, Constanze Sproll, Kerstin Schoeberl, Thomas 15 16 For Peer Review Only Kuballa 17 5 18 19 6 20 21 7 Chemisches und Veterinäruntersuchungsamt (CVUA) Karlsruhe, Weißenburger Str. 3, 22 23 8 D-76187 Karlsruhe, 24 25 26 9 27 28 10 ∗ To whom correspondence should be addressed. e-mail: [email protected] 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

1 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 2 of 21

1 2 3 4 5 11 6 7 12 Abstract 8 9 10 13 A survey of benzene contamination of 451 beverage samples using headspace sampling 11 12 14 in combination with gas chromatography and mass spectrometry (HS-GC/MS) with a 13 14 -1 15 15 quantification limit of 0.13 µg l was conducted. Artefactual benzene formation during 16 For Peer Review Only 17 16 headspace sampling was excluded by only mildly heating at 50°C and adjustment of 18 19 20 17 sample pH to 10. The incidence of benzene contamination in soft drinks, beverages for 21 22 18 babies, alcopops and beer-mixed drinks was relatively low with average concentrations 23 24 19 below the EU drinking water limit of 1 µg l-1. Significantly higher concentrations were 25 26 27 20 only found in carrot juices, with the highest levels in carrot juices specifically intended 28 29 21 for infants. About 94% of 33 carrot juices for infants had detectable benzene levels with 30 31 -1 32 22 an average concentration of 1.86 ± 1.05 µg l . The benzene contamination of beverages 33 34 23 was significantly correlated to iron and copper concentrations, which act as catalyst in 35 36 37 24 benzene formation. The formation of benzene in carrot juices was predominantly caused 38 39 25 by a heat-induced mechanism, which explains the higher occurrence in infant carrot 40 41 26 juices that are considerably heated to exclude microbiological contamination. 42 43 44 27 45 46 28 Keywords: Benzene, , beverages, soft drinks, fruit juice, carrot juice, baby 47 48 49 29 food 50 51 52 53 54 55 56 57 58 59 60

2 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 3 of 21 Food Additives and Contaminants

1 2 3 4 5 30 Introduction 6 7 31 Benzene is one of the contaminants with the highest level of evidence for 8 9 10 32 carcinogenicity. For example, it was classified as carcinogenic to humans (Group 1) by 11 12 33 the International Agency for Research on Cancer (IARC 1987). Active as well as 13 14 15 34 , automobile exhaust, and driving or riding in automobiles are 16 For Peer Review Only 17 35 postulated as the most important pathways for benzene exposure (Wallace 1996). Since 18 19 20 36 high levels of benzene metabolites are frequently reported among children and non- 21 22 37 smoking workers without occupational exposure, Johnson et al. (2007) hypothesize that 23 24 38 there may significant sources of benzene that are hitherto unidentified. 25 26 27 39 28 29 40 Since early 1990s, concerns about benzene contamination of food have been raised. 30 31 32 41 Several sources can contribute to the occurrence of benzene in . Benzoic acid, a 33 34 42 widely used food , may decarboxylate to benzene in the presence of ascorbic 35 36 37 43 acid (Gardner and Lawrence 1993). The formation of benzene from benzoic acid is 38 39 44 influenced by the presence of transition-metal catalysts (for example, Cu(II) or Fe(III) 40 41 45 ions) and is dependent on pH, UV light or temperature (Gardner and Lawrence 1993; 42 43 44 46 McNeal et al. 1993; Chang and Ku 1993; Barshick et al. 1995). Benzene may be 45 46 47 introduced into foods through leaching from various packaging materials or storage 47 48 49 48 environments, from contamination of water supplies, or may be formed during 50 51 49 irradiation processes (Barshick et al. 1995). Another source of benzene in soft drinks 52 53 54 50 and beer is contaminated carbon dioxide used for carbonation (Long 1999; Wu et al. 55 56 51 2006). 57 58 52 59 60

3 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 4 of 21

1 2 3 4 5 53 Methods for determination of benzene in food usually apply GC/MS and different 6 7 54 sample clean-up techniques like liquid-liquid extraction (Carrillo-Carrión et al. 2007), 8 9 55 static or purge and trap headspace sampling (Page et al. 1992; Gardner and Lawrence 10 11 12 56 1993; McNeal et al. 1993; Chang and Ku 1993; Barshick et al. 1995; Fabietti et al. 13 14 57 2001; Cao et al. 2007) or headspace solid-phase dynamic extraction (Ridgway et al. 15 16 For Peer Review Only 17 58 2007). Headspace solid-phase microextraction (HS-SPME) in combination with HPLC 18 19 59 was proposed for determination of benzene in fruit juices (Clasadonte et al. 1997). The 20 21 22 60 recent finding of benzene in drinks has raised concerns over analytical methodology, 23 24 61 especially heating during the purge-and-trap step that might lead to artefactual formation 25 26 62 of benzene (Hileman 2006). 27 28 29 63 30 31 64 There are only a few surveys that are currently available about benzene in food. Page et 32 33 -1 34 65 al. (1992) determined benzene levels ranging from 0.018 to 3.83 µg kg in 97 Canadian 35 36 66 samples of various fruits, juices and drinks. McNeal et al. (1993) showed that food 37 38 39 67 without added benzoates contained lower levels than some foods and beverages 40 41 68 containing both ascorbic acid and . In a study of 60 soft drinks from 42 43 69 Italy (Fabietti et al. 2001), none of the samples were found to have benzene levels above 44 45 -1 46 70 the World Health Organization (WHO) guideline value of 10 µg l for benzene in 47 48 71 drinking water (WHO 2006). Out of 150 soft drinks sampled in the , 49 50 51 72 107 did not contain detectable levels of benzene but four contained average levels above 52 53 73 10 µg kg -1 ( 2006). In a sample of 68 Australian flavoured 54 55 -1 56 74 beverages, 38 contained trace levels of benzene ranging from 1 to 40 µg kg (Anon 57 58 75 2006). From 84 Chinese lager beers, six contained detectable benzene concentrations up 59 60 76 to 7.1 µg l-1 (Wu et al. 2006). The U.S. Food and Drug Administration (FDA) tested

4 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 5 of 21 Food Additives and Contaminants

1 2 3 4 5 77 more than 100 soft drinks and other beverages from the . Almost all of 6 7 78 which contained either no benzene or levels below 5 µg kg -1 (Meadows 2006). Cao et al. 8 9 79 (2007) assessed benzene levels in 124 Canadian beverages, and found that 40% of the 10 11 12 80 samples contained detectable benzene levels and two products had benzene levels above 13 14 81 10 µg l-1. 15 16 For Peer Review Only 17 82 18 19 83 In our study, which is the first survey of benzene in German food products, we 20 21 22 84 concentrated on product groups that were previously described as likely to contain 23 24 85 benzene such as beverages containing ascorbic and benzoic acid. We focused on 25 26 86 products sold as baby or infant food and intended for the most susceptible consumer 27 28 29 87 group (Suk et al. 2003). In addition, we included certain alcoholic beverages in the 30 31 88 survey, because ethanol in alcoholic beverages was recently evaluated by the IARC as 32 33 34 89 carcinogenic to humans (Group 1) (Baan et al. 2007; IARC 2007). The IARC evaluation 35 36 90 emphasizes the importance and necessity of avoiding and controlling carcinogenic 37 38 39 91 contaminants in alcoholic beverages, since synergistic effects between ethanol and other 40 41 92 contaminants are possible (Lachenmeier 2007). In the case of benzene, such an 42 43 93 interaction with ethanol is likely. This is because significant deviations of acute toxicity 44 45 46 94 were demonstrated for the binary mixture (Tichý et al. 2002; Rucki and Tichý 2004) and 47 48 95 both substances may be metabolized by cytochrome P450 in humans (Rana and Verma 49 50 51 96 2005; Seitz and Stickel 2007). 52 53 97 54 55 56 57 58 59 60

5 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 6 of 21

1 2 3 4 5 98 Materials and Methods 6 7 99 8 9 10 100 Sample collective 11 12 101 Between 2006 and 2007, 451 samples submitted to the CVUA Karlsruhe have been 13 14 15 102 analysed for benzene. As part of official food control in the German federal state Baden- 16 For Peer Review Only 17 103 Württemberg, our institute covers the district of Karlsruhe in North Baden (Germany) 18 19 104 with a population of approximately 2.7 million. The sampling has been done by local 20 21 22 105 authorities directly at food producers or retail trade. The samples have been randomly 23 24 106 selected and collected by government food inspectors, as our institute has requested only 25 26 27 107 the food group and sample number to collect but not specific brands. We had initially 28 29 108 requested for food groups “soft drinks”, “alcopops”, “beer-mixed drinks”, and 30 31 32 109 “beverages for babies/infants” that have been previously described to be susceptible to 33 34 110 benzene contamination. Due to our risk-oriented sampling approach (Roth et al. 2007), 35 36 111 we then increased the sample numbers of conspicuous food groups. For example, after 37 38 39 112 the first detection of comparably high benzene levels in carrot juices, we had 40 41 113 specifically requested the food inspectors to collect further samples of carrot juices 42 43 44 114 intended for normal consumption as well as carrot juices intended as baby food for 45 46 115 infants and young children in the sense of Commission Directive 2006/125/EC 47 48 49 116 (Commission of the European Communities, 2006). 50 51 117 52 53 118 Quantitative determination of benzene 54 55 56 119 Benzene was purchased from Sigma-Aldrich (Taufkirchen, Germany). Benzene-d6 was 57 58 120 obtained from Merck (Darmstadt, Germany). The analysis of benzene was done using 59 60 121 static headspace (HS) sampling in combination with gas chromatography and mass

6 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 7 of 21 Food Additives and Contaminants

1 2 3 4 5 122 spectrometry (GC/MS). For sample preparation, 10 ml of sample was placed in a 20 ml 6 7 123 headspace vial. After adjustment with KOH solution (10%, m/m) to pH 10, the solution 8 9 124 was spiked with 100 µl of benzene-d (100 µg l-1 in methanol) as internal standard. The 10 6 11 12 125 vials were tightly sealed and homogenized using a vortex mixer. 13 14 15 126 The HS-GC/MS system used for analysis was an Agilent model 6890 Series Plus gas 16 For Peer Review Only 17 127 chromatograph in combination with a CTC Combi PAL autosampler and an Agilent 18 19 20 128 5973N mass selective detector. Acquisition and analysis of data were performed using 21 22 129 standard software supplied by the manufacturer. The samples were incubated in the 23 24 130 agitator oven of the autosampler at 50°C for 30 min. After that, 1000 µL of sample 25 26 27 131 headspace were injected into the GC/MS system using split injection mode (split ratio 28 29 132 2:1). Substances were separated on a fused silica capillary column (Optima 624, 60 m x 30 31 -1 32 133 0.25 mm I.D., film thickness 1.4 µm). Helium with a constant flow rate of 1.0 ml min 33 34 134 was used as carrier gas. Temperature program: 35°C hold for 1 min, 10°C min -1 up to 35 36 37 135 240°C, hold for 10 min. The temperatures for the injection port, ion source, quadrupole 38 39 136 and interface were set at 250°C, 230°C, 150°C and 250°C, respectively. 40 41 42 137 To determine retention times and characteristic mass fragments, electron impact (EI) 43 44 138 mass spectra at 70 eV of the analytes were recorded by total ion monitoring. The 45 46 47 139 retention time was 11.70 min benzene and 11.65 min for benzene-d6. For quantitative 48 49 140 analysis, the chosen diagnostic mass fragments were monitored in the selected ion 50 51 52 141 monitoring (SIM) mode. Benzene: m/z 78 as target ion and m/z 77 as qualifier ion; 53 54 142 benzene-d6: m/z 84 as target ion and m/z 82 as qualifier ion. For quantification, peak 55 56 143 area ratios of the analytes to the internal standard were calculated as a function of the 57 58 59 144 concentration of the substances. 60

7 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 8 of 21

1 2 3 4 5 145 The method was validated using two authentic samples analyzed seven times. The 6 7 146 precision of the method never exceeded 6.6% (coefficient of variation) and the trueness 8 9 147 never exceeded 5.3% (as compared to spiked concentrations), indicating good assay 10 11 -1 12 148 accuracy. Determined according to DIN 32645, the limit of detection was 0.04 µg l and 13 14 149 the limit of quantitation was 0.13 µg l -1. 15 16 For Peer Review Only 17 150 18 19 151 Quantitative determination of possible benzene precursors and catalysts 20 21 22 152 A subset of samples (n=165) was analyzed for benzoic acid, ascorbic acid, copper and 23 24 153 iron, which were previously discussed to influence benzene formation in food. Benzoic 25 26 154 acid was determined using HPLC according to the German reference method (Anon 27 28 29 155 1984). L-Ascorbic acid was analyzed using a commercially available enzymatic test-kit 30 31 156 (R-Biopharm, Darmstadt, Germany). Copper and iron were analyzed using atomic 32 33 34 157 absorption spectroscopy, according to the German reference method (Anon 1993). 35 36 158 37 38 39 159 Experiment about artefactual benzene formation during headspace sampling 40 41 160 The possible formation of benzene during the headspace sampling procedure was 42 43 161 studied using a full-factorial experimental design. The experiment was conducted using 44 45 -1 -1 46 162 a model solution containing sucrose (60 g l ), (4 g l ) and ascorbic acid (0.4 47 48 163 g l-1). Three variables were studied: benzoic acid concentrations (6 levels: 1, 10, 50, 49 50 -1 51 164 100, 500 and 1000 mg l ), pH (2 levels: 3 and 10) and temperature (2 levels: 50°C and 52 53 165 80°C). 54 55 166 56 57 58 167 Experiment about heat-induced formation of benzene in carrot juice 59 60

8 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 9 of 21 Food Additives and Contaminants

1 2 3 4 5 168 In order to evaluate possible formation mechanisms of benzene in carrot juice with a 6 7 169 minimum amount of experiments, the information gleaned from each experiment and 8 9 170 the relationships between the experiments have to be fully exploited. This was done 10 11 12 171 using a D-optimal design type (Box et al. 2005; Montgomery 2005). The D-optimal 13 14 172 algorithm was used because it chooses an ideal subset of all possible combinations and 15 16 For Peer Review Only 17 173 significantly reduces the number of required experiments compared to standard design 18 19 174 types. 20 21 22 175 The carrot juice used in this experiment was freshly prepared from organic carrots using 23 24 176 a juice extractor (centrifuge-type, Starmix, Reichenbach, Germany). The heating 25 26 177 temperature was studied at four levels (unheated, 100°C, 125°C, 150°C) and the heating 27 28 29 178 time was varied at three levels (30 min, 60 min, 120 min). The experiments were 30 31 179 conduced with 10 ml of juice directly in the headspace vials that were measured 32 33 34 180 afterwards without opening to avoid any loss of benzene. 35 36 181 37 38 39 182 Statistics 40 41 183 The experimental designs and calculations were done using the Software Package 42 43 184 Design Expert V6 (Stat-Ease Inc., Minneapolis, Minnesota, USA). The experiments 44 45 46 185 were evaluated using Analysis of Variance (ANOVA) to find the significance of 47 48 186 variables and their interactions in the models. The models were checked for consistency 49 50 51 187 by looking at the lack of fit and possible outliers. 52 53 188 The survey data were evaluated using standard statistical packages for Windows. 54 55 56 189 Statistical significance was assumed at below the 0.05 probability level. Groups of two 57 58 190 cases were compared using t-tests. One-way ANOVA was used to test whether three or 59 60 191 more cases had the same mean including the Bonferroni post hoc means comparison.

9 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 10 of 21

1 2 3 4 5 192 Pearson’s test was used to evaluate the significance of linear relations. Box and whisker 6 7 193 plots were used for visualization of data (box 25 th - 75 th percentile, line in the box: 8 9 194 median, whiskers: minimum and maximum (max. 1.5 times the length of the inner 10 11 12 195 quartiles), data points outside are outliers). 13 14 196 15 16 For Peer Review Only 17 197 Results and Discussion 18 19 20 198 Artefactual benzene formation during headspace sampling 21 22 23 199 The results of the headspace sampling experiment are shown in Figure 1. The ANOVA 24 25 200 results prove that benzoic acid concentration (p=0.0008), as well as pH (p<0.0001) and 26 27 28 201 temperature (p<0.0001) all have a significant influence on benzene formation. However, 29 30 202 benzene formation that would lead to false-positive results (that is, results above LOQ 31 32 -1 33 203 of 0.13 µg l ) only occurs at low pH values in combination with high temperatures and 34 35 204 high concentrations of benzoic acid. Our results verify the observation by Cao et al. 36 37 205 (2007) that formation of artefacts may occur at 100°C in a 0.5 h time-period, but not at 38 39 40 206 22-24°C. We suppressed artefactual benzene formation by setting the incubation 41 42 207 temperature during headspace GC to 50°C and by pH adjustment of the samples to pH 43 44 45 208 10. All samples in our market survey and formation experiments were measured under 46 47 209 these conditions. 48 49 50 210 [Insert Figure 1 about here] 51 52 211 53 54 212 Survey of benzene in German foods 55 56 57 213 The results of the determination of benzene in different soft drinks, drinks intended for 58 59 60 214 babies and infants (drinks with fruits, vegetables and/or tea), alcopops, beer-mixed

10 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 11 of 21 Food Additives and Contaminants

1 2 3 4 5 215 beverages, as well as carrot juices for the general consumer and infants are shown in 6 7 216 Table I and Figure 2. 8 9 10 217 [Insert Figure 2 & Table I about here] 11 12 13 218 14 15 16 219 Using ANOVA,For it was Peer proved that Review benzene concentrations Only in different food groups 17 18 220 showed highly significant differences (p<0.0008). The concentrations in carrot juices 19 20 221 and especially in carrot juices intended for infants were higher than those in all other 21 22 23 222 groups of beverages. The carrot juices for normal consumption contained average 24 25 223 benzene in concentrations below 1 µg l-1, whereas carrot juices intended for infants had 26 27 -1 28 224 significantly (p<0.0001) higher concentrations clearly above 1 µg l . Approximately 29 30 225 88% of all carrot juices for infants had benzene contaminations above the EU drinking 31 32 -1 33 226 water limit of 1 µg l (European Council 1998). 34 35 227 36 37 38 228 The lowest concentrations were detected in other drinks for babies and infants as well as 39 40 41 229 beer-mixed drinks. Soft drinks and alcopops contained slightly higher values. However, 42 43 230 only 2% of all samples had benzene contaminations above the EU limit, and 44 45 -1 46 231 only one sample had a very high benzene concentration of 41.8 µg l , which was also 47 48 232 above the WHO limit of 10 µg l-1. We were unable to provide an explanation for this 49 50 233 high level in the soft drink, and the levels of the analyzed precursors and catalysts were 51 52 53 234 inconspicuous. The samples from the same product charge, which were simultaneously 54 55 235 analyzed, had benzene levels below 1 µg l-1. 56 57 58 236 59 60

11 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 12 of 21

1 2 3 4 5 237 Precursors and catalysts influencing benzene formation in beverages 6 7 238 A multivariate data analysis was conducted to gain a first overview of the dataset of 165 8 9 10 239 samples analyzed for benzene and the possible precursors and catalysts benzoic acid, 11 12 240 ascorbic acid, copper and iron. The results of the principal components analysis (PCA) 13 14 15 241 are shown in Figure 3. The correlation loadings plot shows that benzene appears to be 16 For Peer Review Only 17 242 closely correlated to copper and iron, which influence the PC1 component. Benzoic acid 18 19 20 243 and ascorbic acid are inversely correlated on the PC2 component. The results of PCA 21 22 244 were confirmed by univariate data analysis (Table II). The copper and iron 23 24 245 concentrations were significantly higher in benzene-positive samples than in benzene- 25 26 27 246 negative samples. Table III shows that the concentration of benzene has a relation to that 28 29 247 of copper and iron. However, there also appear to be other - so far unknown - influences 30 31 32 248 as the metal concentrations show a large dispersion, and some benzene-free cases had 33 34 249 higher metal concentrations than benzene-positive ones. The influence of copper and 35 36 37 250 iron was not unexpected due to the experiments in model solutions by Gardner et al. 38 39 251 (1993). Our study suggests that a metal-catalyzed formation mechanism at least 40 41 252 influences benzene formation in real food. We also can partially confirm the observation 42 43 44 253 of Gardner et al. (1993) that there may possibly be non-linear effects or confounding 45 46 254 between different constituents on the formation on benzene. For example, Gardner et al. 47 48 49 255 (1993) detected a maximum of benzene production at 1.0 nM CuSO 4, higher 50 51 256 concentrations of this metal resulted in a small decrease in benzene concentration. And 52 53 54 257 even when there was no CuSO 4 in the solution, significant amounts of benzene were 55 56 258 still produced. In the same study, it was also observed that addition of FeSO 4 in the 57 58 259 absence of CuSO led to benzene formation, but at higher concentrations of FeSO the 59 4 4 60 260 benzene production decreased. In model solutions, there appears to be a dependency

12 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 13 of 21 Food Additives and Contaminants

1 2 3 4 5 261 between the concentrations of the various constituents and benzene formation. The same 6 7 262 appears to be true for our real food matrices; however, the analysis of our data was 8 9 263 unable to detect interactions between both metals or a certain metal concentration that 10 11 12 264 leads to maximum benzene formation. Presumably, other influences (e.g. temperature) 13 14 265 may have larger influences than the metals and lead to confounding of the data. 15 16 For Peer Review Only 17 266 [Insert Figure 3 & Tables II and III about here] 18 19 20 267 In contrast to the clear influence of copper and iron, benzoic acid concentration tended 21 22 268 to be higher in benzene positive samples but not on a statistically significant level 23 24 25 269 (Table III). The PCA scores plot (Figure 3) shows that only some soft drink samples 26 27 270 containing benzoic acid were also high in benzene. But our results show that the 28 29 30 271 presence of benzoic acid was not causative for benzene formation in our samples and 31 32 272 that other mechanisms appeared to be more predominant. No relation at all could be 33 34 35 273 proven between ascorbic acid and benzene. Our results proved the inconsistency 36 37 274 detected between levels of benzene and benzoate or ascorbic acid previously 38 39 275 described by Cao et al. (2007). In contrast, earlier studies had confirmed a clear 40 41 42 276 connection between benzoic acid and benzene, which was dependent on the 43 44 277 concentration of benzoic acid in the beverages and showed large interactions with the 45 46 47 278 concentration of ascorbic acid (Gardner et al., 1993; Page et al. 1992; McNeal et al., 48 49 279 1993). Therefore, it is likely that due to the efforts of the beverage industry to avoid 50 51 52 280 benzoic acid especially in combination with ascorbic acid, the benzene formation 53 54 281 mechanism from those precursors is nowadays only of secondary relevance, but still 55 56 282 may be causative if high concentrations of benzoic acid are used as preservative. 57 58 59 283 60

13 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 14 of 21

1 2 3 4 5 284 Metal catalyzed and heat-induced formation of benzene in carrot juices 6 7 285 None of the carrot juices in our study contained benzoic acid. However, both groups of 8 9 10 286 carrot juices contained higher concentrations of the investigated metals than the soft 11 12 287 drinks (p<0.0001 for copper and iron). Therefore, a metal catalyzed formation of 13 14 15 288 benzene from so far unknown aromatic precursors may be at least partially causative in 16 For Peer Review Only 17 289 the case of carrot juice. 18 19 20 290 21 22 23 291 Due to the differences of benzene concentration in carrot juices for general consumption 24 25 292 and for infants described above, the question of formation mechanism in this special 26 27 28 293 kind of vegetable juice arose. A metal-catalyzed mechanism alone could not explain the 29 30 294 higher concentrations in carrot juices for infants, since both groups of carrot juices had 31 32 33 295 the same levels of metals (p=0.5245 for iron, and p=0.0900 for copper). 34 35 296 36 37 297 Our working theory was that infant juices were heated more extensively, which led to 38 39 40 298 heat-induced formation of benzene similar to the well-characterised heat-induced 41 42 299 formation of other food contaminants like acrylamide or furan (Wenzl et al. 2007). 43 44 45 300 The rationale for the extended heating of carrot juice for infants is to minimize the 46 47 301 microbiological risk. A typical sterilisation process of carrot juices for infants involves 48 49 50 302 autoclaving at 122°C for 45 min at 2.2 bar, whereas normal carrot juices are typically 51 52 303 heated at 120°C for only 5 min at normal pressure. 53 54 304 55 56 57 305 The results of our experiments about heat-induced formation of benzene in carrot juice 58 59 306 (Figure 4) show that the formation in fact significantly depends on temperature 60

14 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 15 of 21 Food Additives and Contaminants

1 2 3 4 5 307 (p=0.0007) and heating time (p=0.0020), which supports our theory. The formation of 6 7 308 benzene in carrot juices appears to be predominantly heat-induced. It is so far unknown 8 9 309 if the metals play a role in this heat-induced mechanism. Further research is needed to 10 11 12 310 determine if elimination of the metals could minimize the benzene formation during the 13 14 311 sterilisation step. 15 16 For Peer Review Only 17 312 [Insert figure 4 about here] 18 19 20 313 21 22 23 314 Food toxicological evaluation 24 25 26 315 Besides a few exceptions, benzene exposure of the consumer due to soft drinks and 27 28 316 alcoholic beverages on the German market appears to be very low and nearly negligible 29 30 31 317 in consideration of the exposition to benzene from other sources. According to Council 32 33 318 Regulation (EEC) No 315/93 laying down Community procedures for contaminants in 34 35 319 food (Council of the European Communities 1993), no food containing a contaminant in 36 37 38 320 an amount unacceptable from the viewpoint and in particular at a 39 40 321 toxicological level shall be placed on the market. Furthermore, contaminant levels shall 41 42 43 322 be kept as low as reasonably can be achieved by following good practices. 44 45 46 323 47 48 324 It is also reasonable to apply the EU drinking water limit of 1 µg l-1 as basis to evaluate 49 50 51 325 other beverages. According to our results, it is possible to manufacture beverages with 52 53 326 benzene contamination below 0.5 µg l-1. Therefore we think that soft drinks with 54 55 -1 56 327 benzene contamination above 1 µg l are in offence against Council Regulation (EEC) 57 58 328 No 315/93 and should be removed from the market. 59 60 329

15 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 16 of 21

1 2 3 4 5 330 Conclusion 6 7 331 In general, our study confirms the established view that food is not an important 8 9 10 332 pathway for benzene exposure (Wallace 1996). However, some food groups seem to 11 12 333 have a higher risk of benzene contamination than others, and a systematic study of 13 14 15 334 exposure to benzene in the complete human food chain is missing so far. In our study, 16 For Peer Review Only 17 335 carrot juices formed the food group with the highest susceptibility to benzene 18 19 20 336 contamination due to heat-induced formation. In the nutrition of babies and infants, 21 22 337 carrot juices belong to the standard program as a complimentary food for infants from 23 24 338 four months, when breast milk or infant formulas alone can no longer meet the baby's 25 26 27 339 growing nutritional requirements. For this reason, significant consumption of carrot 28 29 340 juice per kg of bodyweight is possible in infants. It would be certainly interesting to 30 31 32 341 study if the exposure to benzene containing baby food may provide an answer to the 33 34 342 elevated levels of benzene biomarkers in children, whose origin was so far unexplained. 35 36 37 343 (Johnson et al. 2007). As toxicology does not allow establishment of safe benzene levels 38 39 344 above zero, any exposure at all to benzene and benzene-containing products should be 40 41 345 avoided (Mehlman 2002). This is especially true in the case of exposure of children to 42 43 44 346 benzene. Since there are two infant carrot juices without benzene contamination in our 45 46 347 survey, it appears to be technologically possible to manufacture such juices without 47 48 49 348 benzene formation. The manufacturers of carrot juice should optimize their processes 50 51 349 with priority. 52 53 54 350 55 56 351 Further research concerning precursors of heat-induced benzene formation is necessary 57 58 352 and food surveillance should introduce carrot and other vegetable-containing infant 59 60 353 products into market surveys.

16 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 17 of 21 Food Additives and Contaminants

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

17 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 18 of 21

1 2 3 4 5 354 References 6 355 7 356 Anon. 1984. Bestimmung von Konservierungsstoffen in fettarmen Lebensmitteln. 8 357 Amtliche Sammlung von Untersuchungsverfahren L00.00-9:1-4. 9 10 358 Anon. 1993. Bestimmung von Eisen, Kupfer, Mangan und Zink mit der 11 12 359 Atomabsorptionsspektrometrie (AAS) in der Flamme. Amtliche Sammlung von 13 360 Untersuchungsverfahren L00.00-19/2:1-2. 14 15 361 Anon. 2006. Benzene in flavoured beverages. World Food Regulation Review 16:3. 16 For Peer Review Only 17 362 Baan R, Straif K, Grosse Y, Secretan B, El Ghissassi F, Bouvard V, Altieri A, Cogliano 18 19 363 V, WHO International Agency for Research on Cancer Monograph Working Group. 20 364 2007. Carcinogenicity of alcoholic beverages. Lancet Oncology 8:292-293. 21 22 365 Barshick SA, Smith SM, Buchanan MV, Guerin MR. 1995. Determination of benzene 23 366 content in food using a novel blender purge and trap GC/MS method. Journal of 24 367 Food Composition and Analysis 8:244-257. 25 26 27 368 Box GEP, Hunter JS, Hunter WG. 2005. Statistics for Experimenters. John Wiley & 28 369 Sons, Hoboken, , USA 29 30 370 Cao XL, Casey V, Seaman S, Tague B, Becalski A. 2007. Determination of benzene in 31 371 soft drinks and other beverages by isotope dilution headspace gas 32 33 372 chromatography/mass spectrometry. Journal of AOAC International 90:479-484. 34 35 373 Carrillo-Carrión C, Lucena R, Cárdenas S, Valcárcel M. 2007. Liquid-liquid 36 374 extraction/headspace/gas chromatographic/mass spectrometric determination of 37 375 benzene, toluene, ethylbenzene, ( o-, m- and p-)xylene and styrene in olive oil using 38 376 surfactant-coated carbon nanotubes as extractant. Journal of Chromatography A 39 40 377 1171:1-7. 41 42 378 Chang PC, Ku K. 1993. Studies on benzene formation in beverages. Journal of Food 43 379 and Drug Analysis 1:385-393. 44 45 380 Clasadonte MT, Giuffrida R, Zerbo A, Ciraolo L. 1997. Determination of benzene and 46 47 381 toluene in fruit juice by solid phase microextraction (SPME) and HPLC. Rassegna 48 382 Chimica 49:345-352. 49 50 383 Commission of the European Communities. 2006. Commission Directive 2006/125/EC 51 384 on processed cereal-based foods and baby foods for infants and young children. 52 385 Official Journal of the European Communities L339:16-35. 53 54 55 386 Council of the European Communities. 1993. Council Regulation (EEC) No 315/93 56 387 laying down Community procedures for contaminants in food. Official Journal of the 57 388 European Communities L37:1-3. 58 59 389 European Council. 1998. Council Directive 98/83/EC on the quality of water intended 60 390 for human consumption. Official Journal of the European Communities L330:32-54.

18 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 19 of 21 Food Additives and Contaminants

1 2 3 4 5 391 Fabietti F, Delise M, Piccioli Bocca A. 2001. Investigation into the benzene and toluene 6 392 content of soft drinks. Food Control 12:505-509. 7 8 393 Food Standards Agency. 2006. Food Survey Information Sheet No. 6/06. Survey of 9 394 benzene in soft drinks. Food Standards Agency, London, UK 10 11 12 395 Gardner LK, Lawrence GD. 1993. Benzene production from decarboxylation of benzoic 13 396 acid in the presence of ascorbic acid and a transition-metal catalyst. Journal of 14 397 Agricultural and 41:693-695. 15 16 398 Hileman B. For2006. Dispute Peer over benzene Review in drinks. Chemical Only & Engineering News 17 399 84:10. 18 19 20 400 IARC. 1987. Benzene. In: IARC Monographs on the Evaluation of Carcinogenic Risks 21 401 to Humans, Supplement 7. Overall evaluations of carcinogenicity: An updating of 22 402 IARC Monographs Volumes 1 to 42. International Agency for Research on Cancer, 23 403 Lyon, France 24 25 26 404 IARC. 2007. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, 27 405 Vol. 96, Alcoholic Beverage Consumption and Ethyl Carbamate (Urethane). in press, 28 406 Lyon, France 29 30 407 Johnson ES, Langård S, Lin YS. 2007. A critique of benzene exposure in the general 31 408 population. Science of the Total Environment 374:183-198. 32 33 34 409 Lachenmeier DW. 2007. Consequences of IARC re-evaluation of alcoholic beverage 35 410 consumption and ethyl carbamate on food control. Deutsche Lebensmittel-Rundschau 36 411 103:307-311. 37 38 412 Long DG. 1999. From cobalt to chloropropanol: de tribulationibus aptis cerevisiis 39 40 413 imbibendis. Journal of the Institute of Brewing 105:79-84. 41 42 414 McNeal TP, Nyman PJ, Diachenko GW, Hollifield HC. 1993. Survey of Benzene in 43 415 Foods by Using Headspace Concentration Techniques and Capillary Gas- 44 416 Chromatography. Journal of AOAC International 76:1213-1219. 45 46 417 Meadows M. 2006. Benzene in beverages. FDA Consumer 40:9-10. 47 48 49 418 Mehlman MA. 2002. Carcinogenic effects of benzene: Cesare Maltoni's contributions. 50 419 Annals of the New York Academy of Sciences 982:137-148. 51 52 420 Montgomery DC. 2005. Design and Analysis of Experiments. John Wiley & Sons, 53 421 Hoboken, New Jersey, USA 54 55 56 422 Page BD, Conacher HBS, Weber D, Lacroix G. 1992. A Survey of Benzene in Fruits 57 423 and Retail Fruit Juices, Fruit Drinks, and Soft Drinks. Journal of AOAC International 58 424 75:334-340. 59 60

19 http://mc.manuscriptcentral.com/tfac Email: [email protected] Food Additives and Contaminants Page 20 of 21

1 2 3 4 5 425 Rana SVS, Verma Y. 2005. Biochemical toxicity of benzene. Journal of Environmental 6 426 Biology 26:157-168. 7 8 427 Ridgway K, Lalljie SP, Smith RM. 2007. Use of in-tube sorptive extraction techniques 9 428 for determination of benzene, toluene, ethylbenzene and xylenes in soft drinks. 10 429 Journal of Chromatography A 1174:20-26. 11 12 13 430 Roth M, Hartmann S, Renner R, Hörtig W. 2007. Risikoorientiertes Probenmanagement 14 431 in Baden-Württemberg. Deutsche Lebensmittel-Rundschau 103:45-52. 15 16 432 Rucki M, TichýFor M. 2004. Peer Acute toxicity Review of binary mixture benzene-ethanolOnly and 17 433 partition coefficient K(ow) of benzene and ethanol. Central European Journal of 18 19 434 Public Health 12 Suppl:S77-S79. 20 21 435 Seitz HK, Stickel F. 2007. Molecular mechanisms of alcohol-mediated carcinogenesis. 22 436 Nature Reviews Cancer 7:599-612. 23 24 437 Suk WA, Murray K, Avakian MD. 2003. Environmental hazards to children's health in 25 26 438 the modern world. Mutation Research 544:235-242. 27 28 439 Tichý M, Borek-Dohalský V, Rucki M, Reitmajer J, Feltl L. 2002. Risk assessment of 29 440 mixtures: possibility of prediction of interaction between chemicals. International 30 441 Archives of Occupational and Environmental Health 75 Suppl:S133-S136. 31 32 442 Wallace L. 1996. Environmental exposure to benzene: An update. Environmental 33 34 443 Health Perspectives 104:1129-1136. 35 36 444 Wenzl T, Lachenmeier DW, Gökmen V. 2007. Analysis of heat-induced contaminants 37 445 (acrylamide, chloropropanols and furan) in carbohydrate-rich food. Analytical and 38 446 Bioanalytical Chemistry 389:119-137. 39 40 41 447 WHO. 2006. Guidelines for drinking-water quality. World Health Organization, 42 448 Geneva, Switzerland 43 44 449 Wu QJ, Lin H, Fan W, Dong JJ, Chen HL. 2006. Investigation into benzene, 45 450 trihalomethanes and in Chinese lager beers. Journal of the Institute of 46 451 Brewing 112:291-294. 47 48 452 49 453 50 51 52 53 54 55 56 57 58 59 60

20 http://mc.manuscriptcentral.com/tfac Email: [email protected] Page 21 of 21 Food Additives and Contaminants

1 2 3 4 5 454 Figure Legends 6 455 7 456 Figure 1. Artefactual benzene formation during headspace analysis. The experiment was 8 9 10 457 conduced using a model solution containing sucrose, citric acid, ascorbic acid and 11 12 458 different amounts of benzoic acid at 2 different pH values and 2 different temperatures. 13 14 15 459 Significant artefactual benzene formation was only detected at pH 3, 80°C and benzoic 16 For Peer Review Only -1 17 460 acid concentration above 10 mg l . 18 19 461 20 21 22 462 Figure 2. Box-plots of the benzene concentrations in different product groups. Carrot 23 24 463 juices and especially those intended for infants contained significantly higher benzene 25 26 27 464 concentrations than all other product groups. 28 29 465 30 31 466 Figure 3. PCA scores and correlation loadings plot of a dataset of 165 alcohol-free 32 33 34 467 drinks analyzed for benzene, benzoic acid, ascorbic acid, copper and iron. Three 35 36 468 categories are marked in the scores plot (soft drinks, carrot juices, infant carrot juices). 37 38 39 469 The correlation loadings show that benzene is highly correlated to both iron and copper 40 41 470 concentrations. 42 43 44 471 45 46 472 Figure 4. Heat-induced formation of benzene in carrot juice. Significant amounts of 47 48 473 benzene are formed if the carrot juice is heated at temperatures higher than 100°C for 49 50 51 474 over 30 min. Freshly pressed organic carrots without any additives were used for this 52 53 475 experiment. 54 55 56 57 58 59 60

21 http://mc.manuscriptcentral.com/tfac Email: [email protected]