Levels of lead in foods from the first French total diet study on infants and toddlers

Food Chemistry 237 (2017) 849–856

Contents lists available at ScienceDirect

Food Chemistry journal homepage: www.elsevier.com/locate/foodchem

Analytical Methods

Levels of lead in foods from the first French total diet study on infants and toddlers Thierry Guérin a,⇑, Emilie Le Calvez b, Julie Zinck a, Nawel Bemrah c, Véronique Sirot c, Jean-Charles Leblanc d, Rachida Chekri a, Marion Hulin c, Laurent Noël e a

Université Paris-Est, ANSES, Laboratory for Food Safety, F-94701 Maisons-Alfort, France Cofrac, Paris, France ANSES, Risk Assessment Department, Maisons-Alfort, France d Office of Food Safety, Agriculture and Consumer Protection Department, Food and Agricultural Organization of the United Nations (FAO), Rome, Italy e The French Directorate General for Food, Ministry of Agriculture, Agro-16 Food and Forestry, Paris, France b c

a r t i c l e

i n f o

Article history: Received 23 February 2017 Received in revised form 2 June 2017 Accepted 6 June 2017 Available online 7 June 2017 Chemical compounds studied in this article: lead (PubChem CID: 5352425) Keywords: Lead ICP-MS Method validation Foods Total diet study Infants and toddlers

a b s t r a c t Infants and toddlers are highly vulnerable to exposure to lead due to its higher absorption in small children than in adults. This study describes the optimisation and validation of a very sensitive method for the determination of low levels of lead in foods mostly consumed by infants and toddlers. This method, based on inductively coupled plasma-mass spectrometry with a programmable temperature cyclonic spray chamber, attained a limit of quantification (LOQ) of 0.6 or 0.9 mg Pb kg 1 for a liquid or a solid sample, that was improved by a factor 5.6–8.3 compared to the previous method (LOQ: 5 mg kg 1). The analytical method was then applied to 291 food samples from the first French total diet study on infants and toddlers. Lead was detected in most samples at relatively low concentrations (range 0.0–16 mg kg 1). The highest lead concentrations were mainly found in processed food products (e.g. products containing chocolate). Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Infants and toddlers are highly vulnerable to exposure to toxic trace elements due to their immature renal system and their low tolerance of toxic trace elements such as lead (Pb) (ANSES, 2013; Codex Alimentarius Commission, 2003; EFSA, 2010; Kazi et al., 2010; Tuzen & Soylak, 2007). Lead is absorbed more efficiently in children than in adults and it accumulates in soft tissues and, over time, in bones. The half-lives of lead in blood and bone are approximately 30 days and 10–30 years, respectively, and excretion is primarily in urine and faeces. The European Food Safety Authority (EFSA) identified developmental neurotoxicity in young children and cardiovascular effects and nephrotoxicity in adults as being the critical effects for risk assessment (EFSA, 2010), and derived benchmark dose lower confidence limits (BMDLs) based on the effects related to blood lead levels. EFSA concluded that in infants,

⇑ Corresponding author. E-mail address: [email protected] (T. Guérin). http://dx.doi.org/10.1016/j.foodchem.2017.06.043 0308-8146/Ó 2017 Elsevier Ltd. All rights reserved.

in children from 1 to 7 years of age and in pregnant women, the possibility of a health risk due to lead cannot be excluded and that there is reason for concern given the current levels of exposure to lead and its adverse effects, particularly on neurodevelopment. More recently, the French Agency for Food, Environmental and Occupational Health & Safety (ANSES) concluded that studies on the link between exposure to lead and effects on the central nervous system were sufficiently robust to conclude that lead has adverse effects (ANSES, 2013). Therefore, protecting infants and children against the potential risk of adverse effects of lead is beneficial for the entire population. Human exposure occurs mainly via food and water, as well as via air, dust and soil. Lead in milk products and cereal-based products are of particular concern and are closely monitored by international organisations because these food groups are considered as the first solid foods introduced in the diet of infants and children (Codex Alimentarius Commission, 2003; EFSA, 2010, 2012). The primary techniques for analysing lead in food samples are based on atomic absorption spectrometry (AAS), atomic emission

850

T. Guérin et al. / Food Chemistry 237 (2017) 849–856

spectrometry (AES) and mass spectrometry (MS) after digestion of organic material with concentrated acids. Currently, inductively coupled plasma-mass spectrometry (ICP-MS) is the most sensitive method, with a median limit of detection (LOD) of 3 mg Pb kg 1 followed by non-specified AAS and graphite furnace atomic absorption spectrometry (GFAAS) with median LODs of 6 mg kg 1 and 10 mg kg 1, respectively (EFSA, 2010). In 2011, ANSES started the first French total diet study on infants (<12 months old) and toddlers (from 12 up to 36 months old) to estimate dietary exposure to a wide range of chemical substances, including essential and non-essential elements (Hulin et al., 2014). Generally, for accurate dietary exposure assessments, the LOD and the limit of quantification (LOQ) should be as low as technically possible, because most foods do not contain detectable residues, and the value assigned to those samples will affect the estimated dietary exposures (WHO/IPCS, 2009, chap. 6). Although the LOD (2.5 mg Pb kg 1) of the ICP-MS method used for the second French total diet study (Millour, Noël, Chekri, Vastel, & Guérin, 2011) is in the median range indicated in the EFSA opinion (EFSA, 2010), a preliminary exposure assessment indicated that such analytical limits may lead to greater uncertainty in risk assessments due to the processing of left-censored data. Therefore, to prevent this bias as much as possible, reverse theoretical calculations have been performed to estimate the lowest attainable analytical limits so as to be able to determine a potential exposure risk for the population in the case of non-detection or nonquantification in food samples analysed (Hulin et al., 2014). For lead, to attain an exposure value (with an upper bound estimation) of between 10 and 50% of the lowest BMDL derived by EFSA in the case of non-quantification in all samples, analytical limits should be at least about 10 times lower than the current method’s LOQ (i.e. 2.5 mg Pb kg 1). The ICP-MS method mentioned above was first re-evaluated on an Agilent 7700x ICP-MS using a conventional spray chamber (Scott-type double-pass water cooled) and a quartz concentric nebuliser (400 mL min 1), to improve sensitivity. The LOD estimated from the intersection between the acceptance and tolerance limits of the accuracy profile (AFNOR, 2010; Mermet & Granier, 2012) was improved by a factor of two, i.e. 1.25 mg Pb kg 1 (Chevallier, Chekri, Zinck, Guérin, & Noël, 2015). However, this value is still too high for assessing exposure to lead in foods consumed by infants and toddlers because only 34% of food samples could be quantified (unpublished data). Therefore, the primary objective of this study was to develop and validate, according to the accuracy profile procedure, a more sensitive ICP-MS method after closed-vessel microwave digestion to lower the percentage of left-censored data (data < LOQ or < LOD). Secondly, we determined the level of lead in 291 samples of food mostly consumed by French infants and toddlers to assess, for the first time in France, the dietary lead exposure of these populations.

2. Materials and methods 2.1. Reagents and gas All solutions were prepared with analytical reagent grade chemicals and ultrapure water (18.2 MO cm) obtained by purifying distilled water with the Milli-QTM PLUS system associated with an Elix 5 pre-system (Millipore S.A., St Quentin-en-Yvelines, France). Nitric acid: Suprapur HNO3 (67% v/v) and Rectapur HNO3 (54% v/ v) was purchased from VWR (Fontenay-sous-Bois, France). Standard solution: Standard stock solution containing 1000 mg L 1 lead was purchased from Analytika (Prague, Czech Republic) and was used to prepare calibration standards. Working standards were

prepared daily in 6% HNO3 (67%, v/v) (a) and were used without further purification. Tuning solution: a 10 mg L 1 multi-element solution (Agilent Technologies, Courtaboeuf, France) was used to prepare a tuning solution containing several elements such as lithium (Li), yttrium (Y), thallium (Tl), cobalt (Co), cerium (Ce), covering a wide mass range. Factor P/A solution: a 2.5 mg L 1 to 20 mg L 1 factor P/A multi-element solution (Agilent Technologies) was used to obtain a linear response of the detector between pulsed and analogical modes. Internal standard solutions: 1000 mg L 1 standard stock solutions of bismuth (Bi) was purchased from Analytika. An internal standard solution was added to all samples, calibration standards and blanks at the same concentration (2 mg Bi L 1), to obtain information on changes in sensitivity. Certified reference materials (CRMs): SRM 1548a (Typical diet) from the National Institute of Standards and Technology, BCR 063R (Skim milk powder) from the Community Bureau of Reference and DOLT-4 (Fish liver) from the National Research Council Canada were purchased from Courtage Analyses Services (MontSaint-Aignan, France) and from LGC Standards (Molsheim, France). Ultrapure grade carrier (argon (Ar), 99.9995% pure) was supplied by Linde Gas (Montereau-Fault-Yonne, France).

2.2. Selection and preparation of samples The establishment of the sampling plan and the preparation of food samples are described elsewhere (Hulin et al., 2014). Briefly, the results of a survey on food consumption of children under 3 years of age were used to determine the food list (Fantino & Gourmet, 2008). It includes individual, consecutive 3-day weighted food records for 705 children. Based on this consumption data, the most consumed food in terms of quantity and/or percentage of consumers in the targeted population was first selected. Then, foods that are known to be the main contributors to the exposure to at least one substance of interest were added. From this selection, 291 food items were defined for trace elements, including 219 infant foods and 72 common foods or bottled water. Foods were sampled in the same region and prepared as consumed by the same operator (i.e. peeled, cooked, etc.) based on a result of a national study on parents’ food preparation/cooking practices carried out by ANSES (Hulin et al., 2014). For foods that have to be diluted, namely infant and follow-on formulae, a ‘‘reference water” (a brand of mineral water in a glass bottle) was used. Products were diluted according to manufacturer’s indications. Therefore, the results are expressed for ready-to-eat products and included the lead content in the formula powder and the ‘‘reference water” used for reconstituting the product as consumed. To evaluate lead content contributed by the ‘‘reference water, a composite sample made of sub-samples coming from every batches of the bottles used to dilute the products was analysed. Cereal-based food and chocolate powder were not diluted with infant formulae and milk, respectively, in order to distinguish contamination due to each product separately. For these products, and to compare with other food groups prepared as consumed, the ‘‘diluted concentration” was calculated from the concentration in the powdered form and the mean dilution factor obtained from the consumption survey. For each of the 291 samples, 12 sub-samples of equal weight of the same food items were bought every month during the one-year sampling programme (2011–2012), prepared and homogenised by a single cryo-grinding in liquid nitrogen to obtain an individual composite sample (Hulin et al., 2014). These individual composite samples were grouped into 36 food groups representative of the French child food habits with 11 food groups containing only infant foods and 25 food groups of common foods. The samples were stored in high-density polyethylene bottles with screw caps, frozen and sent to the laboratory for analysis.

851

T. Guérin et al. / Food Chemistry 237 (2017) 849–856

2.3. Sample digestion procedure Fresh samples were digested using a Multiwave 3000 microwave digestion system (Anton-Paar, Courtaboeuf, France), equipped with a rotor for eight 80 mL quartz vessels (operating pressure, 80 bar). The sample digestion procedure was performed according to the NF EN 13805 standard (AFNOR, 2014). Before use, quartz vessels were decontaminated with 4 mL of HNO3 (54%, v/v) (a) in the microwave digestion system, then rinsed with ultrapure water and dried in a 40 °C oven. Dietary samples of 0.3– 1.3 g were weighed precisely in quartz digestion vessels and wetoxidised with 3 mL ultrapure water and 3 mL HNO3 (67%, v/v) (a) in the microwave digestion system. One randomly selected vessel was filled with reagents only and taken through the entire procedure as a blank. The digestion procedure was previously optimised (Noël, Leblanc, & Guérin, 2003). After cooling at room temperature, sample solutions were quantitatively transferred into 50 mL polyethylene flasks. Then, 100 mL of the internal standard solution (e) (1 mg L 1) was added to a final concentration of 2 mg L 1; the digested samples were filled with ultrapure water to the final volume before ICP-MS analysis. 2.4. Instrumentation ICP-MS measurements were performed using a 7700 Series  (Agilent Technologies) equipped with a third-generation Octopole Reaction System (ORS3). A Quartz concentric nebuliser (MicroMist, Courtage Analyses Services) at 200 mL min 1 and a temperature controlled Teflon cyclonic spray chamber (IsoMist, Courtage Analyses Services), which can control the chamber wall temperature between 10 °C and 60 °C, were used as a sample introduction system. With a peristaltic pump, the sample solutions were siphoned from tubes arranged on a CETAC ASX-500 Series auto-sampler (CETAC Technologies, Omaha, Nebraska, USA). To reduce the analysis time, an integrated sample introduction system (ISIS) (Nagaoka & Wilbur, 2003) was used (Agilent Technologies). Further details of instrument settings are given in Table 1. Torch position, ion lenses, gases output, resolution axis (10% of peak height, m ± 0.05 a.m.u.) and background (<20 counts per second (cps)) were optimised daily with the tuning solution (c) (1 mg L 1) to carry out a short-term stability test of the instrument,

Table 1 7700 Series ICP-MS operating parameters. Operating conditions Nebuliser Spray chamber Temperature of spray chamber Sampling cone Skimmer cone RF power Reflected power

Quartz concentric at 200 mL min 1 Temperature controlled Teflon cyclonic 35–45 °C Nickel, 1.0 mm orifice Nickel, 0.75 mm orifice 1400–1500 W <10 W

Standard mode Plasma gas flow Nebuliser gas flow Auxiliary gas flow Expansion stage Intermediate stage Analyser stage Octopole bias Quadrupole bias

15 L min 1 0.95–1.05 L min 1 0.99 L min 1 2.0 mbar 2.0  10 4–3.0  10 1.0  10 6–2.0  10 8V 3V

Acquisition parameters Mass range Number of channels Dwell time Number of sweeps Total acquisition time

206–209 a.m.u. 500 100 ms 500 14 s

4 6

mbar mbar

to maximise ion signals and to minimise interference effects from oxide levels (CeO+/Ce+ <2.5%) and doubly charged ions (Ce2+/Ce+ <4%). Linearity response in pulsed and analogical modes (P/A factor determination) was verified daily using 200-fold diluted P/A tuning solutions (d). The determination of lead using ICP-MS was performed in standard mode on 206Pb, 207Pb, 208Pb with 209Bi as an internal standard. 2.5. Analytical quality assurance Several internal quality controls (IQC) were used throughout for the analysis of real samples, as described briefly below. Concentrations were calculated after subtraction of blank values if the IQC were satisfactory. When acceptance criteria were not met, the results were discarded and the samples were re-analysed. The external calibration was considered satisfactory when the determination coefficient (r2) was 0.995. A reagent blank was prepared for each run of samples digestion/extraction and then further analysed in the same conditions as the samples in order to assess the cross-contamination. In all cases, the reagent blanks levels were <2 LOQ (with a maximum of 0.022 mg L 1). A control standard (0.25 mg L 1) was also analysed every 6 to 10 samples to monitor instrumental drift. Of the analysed standards, 97% fell in the confidence interval (CI) between 80% and 120%. For the internal standard added to samples and standard solutions to monitor instrumental drift and matrix effects, 96% of the recoveries fell between 80% and 120%. We randomly selected 10% of samples to analyse in duplicate (n = 29), all results fell between 80% and 120% of the CI. Finally, three CRMs were included in each analytical sequence to check the trueness (Table 2); all values were comprised within the 95% confidence interval of the CRMs and all the Z-score values were considered acceptable (<±2). 2.6. Calculations and statistical methods The optimised method was validated according to the accuracy profile procedure (AFNOR, 2010; Mermet & Granier, 2012), based on tolerance intervals to select the best calibration function and to determine the validated concentration ranges. An accuracy profile summarises every validation element in a single plot, giving a graphical representation of the error risk for each concentration in the validated range. This procedure gives a precise estimation of the accuracy of the analytical method and provides a realistic LOQ according to Mermet, Granier, and Fichet (2012). The accuracy expresses the total error, including the systematic error (trueness) and the random error (repeatability and intermediate precision) for each concentration level. The validity domain can be defined between the LOQ and the upper tested concentration, as the bexpectation limits are comprised between the acceptance limits (k). Tolerance limits (b-expectation tolerance interval) were calculated for the mean bias at each concentration level. The LOQ is deduced from the intersection between acceptance and tolerance limits. Data were subjected to linear regression and analysis of variance using Microsoft Excel software. Concentrations are expressed in micrograms of lead per kilogram of fresh weight

Table 2 Results of trueness in certified reference materials (mg Pb kg

1 2 3

1

dry mass, n = 4).

CRM

Certified value1

Observed value2

Z-score3

Skim milk powder, BCR 063R Typical diet, SRM 1548a Fish liver, DOLT-4

18.5 ± 2.7 44 ± 9 160 ± 40

21.0 ± 2.2 48 ± 2 146 ± 36

0.93 0.44 - 0.35

±expanded uncertainty U. ±standard deviation. Calculated from procedure 1.2 of Jorhem, Engman, and Schroder (2001).

852

T. Guérin et al. / Food Chemistry 237 (2017) 849–856

(mg kg 1 f.w.). As recommended for Total Diet Studies (Moy & Vannoort, 2013), the left-censored data (LC) for non-detected and non-quantified samples were replaced by zero and LOD values, respectively, in a lower bound (LB) scenario, and were replaced by LOD or LOQ, respectively, in an upper bound (UB) scenario.

3. Results and discussion 3.1. Optimisation of analytical conditions For this study, the conventional spray chamber (Scott-type double-pass water cooled at 3 °C) and nebuliser (quartz concentric (MicroMist) 400 mL min 1) of the Agilent 7700x ICP-MS were replaced with a quartz concentric nebuliser (MicroMist) at 200 mL min 1 and a temperature controlled Teflon cyclonic spray chamber (IsoMist) which controls the chamber wall temperature between -10 °C and 60 °C. The standard range of lead was also reduced from 0–10 to 0–1 mg L 1 to improve the accuracy and precision at low concentration levels. The calibration curve was plotted on six points, including the calibration blank (0–0.05–0.10–0. 25–0.50–1.00 mg L 1). The temperature of the spray chamber was first tested from 3 °C to 53 °C, at increments of 10 °C, and after stabilisation for 15 min at the target temperature. The nebuliser gas flow was tested between 0.80 and 1.10 L min 1 at increments of 0.05 L min 1. The signal was then measured for each step with a blank solution and a standard solution of 1 mg L 1 to calculate the signal-to-background ratio (SBR = cps of the standard solution/cps of the blank). The SBRs obtained were compared with those obtained in standard conditions (using the conventional Scott-type double-pass water cooled at 3 °C and a nebuliser at 400 mL min 1). These preliminary results indicated a significant increase in the SBR by a factor of about 2.5 with a spray chamber temperature of 43 °C and a nebuliser gas flow of 1.05 L min 1. A complementary experiment carried out between 35 and 50 °C at increments of 5 °C indicated that the highest SBR (n = 3 at different days) was observed for temperature settings of between 35 and 45 °C and a flow rate of between 0.95 and 1.05 L min 1. These operating conditions were optimised daily for subsequent experiments. Linearity and the LOD and LOQ values were then estimated. Linearity experiments (n = 20) were performed by using standard solutions performed on different days. Statistical tests, based on analysis of variance of least-square regressions (according to Fisher tests), indicated that the linear regression model was acceptable and no deviation from the regression model was observed in the defined range. The LOD and LOQ were defined respectively as three and six times the standard deviation of the average of 21 reagent blanks analysed on several different days after correction for a typical sample weight (0.6 g) and a final volume of 50 mL. The results indicated a LOQ of about 0.015 mg Pb L 1 corresponding to 1.3 mg Pb kg 1. To check the specificity of the method (lack of interference), recoveries of spiked standard solutions in the defined calibration range were measured in several matrices (honey, infant cereals, preparation of vegetables, chicken and rice, dairy desserts <3 years – milk and fruit, semi-skimmed milk yoghurt, mineral water, orange juice, cucumber, chicken, beef steak). Spiking (n = 10) was done before digesting the samples, which underwent the complete analytical procedure. Added concentrations varied between 0.06 and 1 mg Pb L 1 when the sample concentration was close to the LOQ. One non-spiked sample and the spiked sample were analysed and the mean content was calculated to verify if spiked amount was detected and recovered. The line of regression was tested against the line of unity (slope = 1, intercept = 0) by simultaneously testing the hypotheses of slope different from 1 and intercept dif-

ferent from 0, using Student’s t-test; Tobserved < Tcritical value. The slope and intercept of this regression line were not significantly different from 1 and 0, respectively. Therefore the specificity of the method was considered acceptable. The recovery rates ranged from 88% in semi-skimmed milk yoghurt to 112% in mineral water. Finally, trueness was also assessed on different CRMs in these operating conditions (Table 2). All values were within the 95% confidence interval of the CRMs and all the Z-scores were much lower than ±2 and considered acceptable. 3.2. Method validation Obtaining the best estimates of dietary exposure depends essentially on the quality of the concentration data. The latter should be obtained using validated methods that — where possible — suit the purpose of the assessment (WHO/IPCS, 2009, chap. 6). To validate the method, an accuracy profile was built as per standard NF V03-110: 2010 (AFNOR, 2010). It includes a range of concentration levels comprised between the LOD (half the LOQ estimated at 0.0075 mg Pb L 1) and the maximum of the external calibration curve (validity domain). Five validation series were repeated on five different days over two months by two operators. A validation series consisted in analysing five concentration levels and six calibration standards to establish the response function in duplicate. Level 1 (LOD estimated) and level 2 (LOQ estimated) correspond to spiked water because food matrices may contain lead. Other concentration levels corresponded to the three undiluted CRMs presented in Table 2. The probability b was set at 85% for lead, meaning that the risk of results falling outside these limits was below 15% on average. The acceptance limits (k) were set to ±45% based on our experience (corresponding to an intermediate precision coefficient of variation estimated at 15% at k = 3 (p = 0.99)). The accuracy profile obtained for 208Pb is presented in Fig. 1. Except for the first concentration level, the other bexpectation tolerance intervals fell within the acceptability limits. The trueness bias ranged from 13% to 7%. The estimated repeatability coefficient of variation CVr varied from 6.7% (207Pb) to 22.0% (206Pb) in the validity domain. The intermediate precision coefficient of variation CVR obtained varied from 7.7% (207Pb) to 22.3% (206Pb). The LOQ deduced from the intersection of the acceptability and tolerance limits of the accuracy profile corresponded to the first level of concentration validated (level 2). The LOQ obtained under robust conditions was similar to that previously estimated from reagent blanks (LOQ of 0.015 mg L 1). Compared to the previous method (LOD of 2.5 mg kg 1 and LOQ of 5 mg kg 1; Millour et al., 2011), the LOD/LOQ for lead were improved by a factor 5.6 to 8.3 (LOQ of 0.9 mg kg 1 and 0.6 mg kg 1 using a sample weight, which was increased from 0.6 g (in Millour et al., 2011) to 0.8 g for a solid sample and to 1.3 g for a liquid sample, respectively, in this study and using a similar final volume of 50 mL in both studies). The sensitivity attained using this method was better or comparable to those attained with high-resolution ICP-MS methods (Frazzoli & Bocca, 2008; FSA, 2006; Pandelova, Lopez, Michalke, & Schramm, 2012) and enhanced by a factor 3 to 100 compared to previous Q-ICP-MS assessments (CarbonellBarrachina et al., 2012; EFSA, 2010, 2012). Therefore, the validity domain of the analytical method was valid in the range of 0.015 to 1.0 mg L 1. For routine analysis, the CVR is set at 17% over the entire validity domain and the expanded uncertainty (U, k = 2) is U = 0.34  M, for n = 1. 3.3. Lead content in foodstuffs The concentrations of lead are given by food group in Table 3. Of the 291 food samples analysed (including ‘‘reference water”), 81%

853

T. Guérin et al. / Food Chemistry 237 (2017) 849–856

Fig. 1. Accuracy profile of

Table 3 Levels of lead in foods ‘‘as consumed” by French infants and toddlers (in mg Pb kg Food category

a

N

ND

a

1

208

Pb (b = 85%; k = ±45%).

fresh weight). Lower Bound (LB)

Upper Bound (UB)

Mean ± SD

Min-Max

Mean ± SD

Min-Max

Infant food Milk-based beverage Cereal-based foodb Milk-based dessert Fruit juice Growing-up milk Soup, puree Fruit puree Vegetable-based ready-to-eat meal Meat/fish-based ready-to-eat meal Infant formula Follow-on formula

8 17 6 4 9 11 30 27 45 28 34

0 0 0 0 2 0 0 5 3 4 8

1.11 ± 0.21 1.40 ± 1.95 1.20 ± 0.41 3.72 ± 2.31 0.37 ± 0.34 4.03 ± 3.12 2.15 ± 2.08 3.13 ± 2.89 3.43 ± 2.01 1.01 ± 0.68 0.68 ± 0.63

0.80–1.34 0.27–7.70 0.50–1.75 0.30–5.22 0.00–1.00 1.53–12.7 0.50–11.9 0.00–10.1 0.00–9.48 0.00–2.31 0.00–2.22

1.11 ± 0.21 1.40 ± 1.95 1.27 ± 0.28 3.79 ± 2.16 0.60 ± 0.22 4.03 ± 3.12 2.19 ± 2.05 3.25 ± 2.77 3.46 ± 1.96 1.08 ± 0.58 0.82 ± 0.51

0.80–1.34 0.27–7.70 0.90–1.75 0.60–5.22 0.30–1.00 1.53–12.7 0.90–11.9 0.50–10.1 0.50–9.48 0.30–2.31 0.30–2.22

Common food Other hot beverages Butter Sweet and savoury biscuits & bars Non-alcoholic beverages Delicatessen meats Compotes and cooked fruit Bottled water Dairy-based desserts Cheese Fruit Milk Vegetables excluding potatoes Eggs and egg products Bread and dried bread products Mixed dishes Fish Potatoes and potato products Pasta Rice and wheat products Soups and broths Sugars and sugar derivatives Ultra-fresh dairy products Meat Croissant-like pastries Poultry and game

1 1 1 6 2 2 12 2 1 6 3 8 1 2 1 3 3 1 2 1 1 5 2 2 2

0 0 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1.44 2.16 9.59 2.91 ± 2.74 3.40 ± 0.40 3.24 ± 1.18 0.51 ± 0.83 4.80 ± 2.86 6.39 0.95 ± 0.42 1.11 ± 0.32 6.29 ± 5.33 1.44 4.67 ± 3.13 6.42 2.97 ± 0.86 2.51 ± 0.56 2.21 5.08 ± 2.44 4.62 0.50 1.16 ± 0.41 2.56 ± 0.74 8.23 ± 3.80 2.43 ± 1.32

– – – 0.72–6.73 3.11–3.68 2.40–4.07 0.00–2.28 2.77–6.82 – 0.50–1.63 0.83–1.46 1.95–16.1 – 2.46–6.88 – 2.12–3.83 2.03–3.13 – 3.35–6.80 – – 0.78–1.73 2.03–3.08 5.54–10.9 1.50–3.36

1.44 2.16 9.59 2.91 ± 2.74 3.40 ± 0.40 3.24 ± 1.18 0.73 ± 0.70 4.80 ± 2.86 6.39 1.08 ± 0.28 1.11 ± 0.32 6.29 ± 5.33 1.44 4.67 ± 3.13 6.42 2.97 ± 0.86 2.51 ± 0.56 2.21 5.08 ± 2.44 4.62 0.90 1.16 ± 0.41 2.56 ± 0.74 8.23 ± 3.80 2.43 ± 1.32

– – – 0.72–6.73 3.11–3.68 2.40–4.07 0.30–2.28 2.77–6.82 – 0.90–1.63 0.83–1.46 1.95–16.1 – 2.46–6.88 – 2.12–3.83 2.03–3.13 – 3.35–6.80 – – 0.78–1.73 2.03–3.08 5.54–10.9 1.50–3.36

Non-detected data. Concentrations of chocolate powder and instant cereal for babies indicated in this table correspond to concentrations after dilution. They were calculated from the concentration in the dry powder and the mean dilution factor used in the consumption study. The mean LB/UB concentrations of the reference water used for diluting infant formula was 0.0/0.5 mg Pb kg 1. b

854

T. Guérin et al. / Food Chemistry 237 (2017) 849–856

of lead levels were quantified, 9% were below the LOQ of 0.9 mg kg 1 for solid samples or 0.6 mg kg 1 for liquid samples and 10% below the LOD (0.5 mg kg 1 for solid samples and 0.3 mg kg 1 for liquid samples). Mean lower-bound (LB) and upper-bound (UB) concentrations were generally quite similar or equal to given the low LOD and LOQ values attained in this study. Lead was detected in most samples at relatively low concentrations (mean LB/UB 2.2/2.4 mg kg 1, range 0.0/0.3–16 mg kg 1). The highest mean levels were found in common food: ‘‘sweet and savoury biscuits and bars” (9.6 mg kg 1; n = 1 in dry chocolate bis cuits),‘‘croissant-like pastries” (8.2 mg kg 1; n = 2 in chocolate croissants (10.9 mg kg 1) and brioche cake and bread (5.5 mg kg 1)), ‘‘cheese” and ‘‘mixed dishes” (6.4 mg kg 1; n = 1 in processed cheese and ravioli-type stuffed pasta) and ‘‘vegetables (excluding potatoes)” (6.3 mg kg 1; n = 8). The highest individual concentrations were found in ‘‘vegetables” (spinach and carrots with concentrations of 16.1 and 11.4 mg kg 1, respectively) for common food and in ‘‘soup, puree” and ‘‘fruit puree” (12.7 and 11.9 mg kg 1, respectively, for a sample of vegetable puree and an apple and prune puree) for infant food. When considering non-diluted products, cereal with sweet biscuit and cocoa (brand 2) had the highest concentrations, with 26.6 mg kg 1 followed by infant cereal with cocoa (19.7 mg kg 1), infant cereal with vegetables (18.8 mg kg 1) and infant cereal with cocoa (brand 2) (16.0 mg kg 1) (Fig. 2). Notably, foods containing chocolate generally had higher concentrations of lead compared with other foods, these results being consistent with the lead levels in chocolate previously reported in the range of 30 to 55 mg kg 1 (Güldas, Dagdelen, & Biricik, 2008; EFSA 2012). Other food groups contained less than 5.2 mg kg 1 on average, with the lowest average UB levels found in ‘‘Growing-up milk” (0.6 mg kg 1; n = 9) and the maximum level of 5.1 mg kg 1 found in the food groups ‘‘rice and wheat products”. Specific limits for lead in infant and follow-on formulae of 10 mg kg 1 for those marketed as liquid and of 50 mg kg 1 for those marketed as powder have been set in EC Regulation 1881/2006 as amended (EC, 2006). All the formula samples analysed in this survey showed values that were several times below this limit. For other foods used as ingredients in weaning foods, limits ranged from 50 to 300 mg kg 1 (EC,, 2006). The highest lead concentration of 16.6 mg kg 1 was found in a spinach sample. All the concentra-

tions were also well below the limit of 200 mg kg 1 for lead in cereals and pulses (EC,, 2006). However, it should be noted that these analyses were performed on pooled samples, therefore we cannot exclude the possibility one or more sub-samples exceeding the limit if they had been analysed separately. The average LB/UB level of 2.1/2.2 mg Pb kg 1 found in infant foods analysed (n = 219) was similar (EFSA, 2010) and 2–40 times lower on average compared with levels reported by previous European studies (Carbonell-Barrachina et al., 2012; EFSA, 2010, 2012; Frazzoli & Bocca, 2008; FSA, 2003, 2006; Ljung, Palm, Grandér, & Vahter, 2011; Pandelova et al., 2012). An adjusted LB/UB mean in ready-to-consume infant formulae (n = 423) of 2.0/4.7 mg kg 1 and in other infant foods (n = 947) of 8.2/20.3 mg kg 1 was estimated in the EFSA opinion (EFSA, 2010), whereas another EFSA study indicated an LB/UB mean in food for infants and small children (n = 2065; 51% of LC data) of 8–16 mg kg 1 (EFSA, 2012). Another study estimated that the mean lead content determined in the most consumed infant foods in Europe (n = 11 including infant formulae and solid foods and beverages) ranged from 8.2 mg kg 1 in the ‘‘starting” milk-based infant formulae to 30.5 mg kg 1 in ‘‘starting” soy-based infant formulae (Pandelova et al., 2012). In surveys conducted in the UK in 2006 and 2003, the mean concentrations were 6 and 8 mg kg 1, respectively, in which all samples were analysed ‘as sold’ (i.e. formulae and dried meals were not reconstituted prior to analyses) (FSA, 2003, 2006). In Italy, lead levels ranged from 8.9 to 15.9 mg kg 1 in three cow’s milk and from <1.85 (LOQ) to 12.1 mg kg 1 in six infant formula samples (Frazzoli & Bocca, 2008). In Sweden, lead levels ranged from 0.8 to 1.7 mg kg 1 in infant formula and from 1.1 to 13 mg kg 1 in infant foods intended for infants during their first 6 months of life (Ljung et al., 2011). The highest levels in Europe were found in a Spanish study, with mean lead levels ranging from <39 mg kg 1 (LOD) to 121 mg kg 1 in 40 commercial Spanish infant rice/cereals and infant foods (Carbonell-Barrachina et al., 2012). Moreover, the lead content in the seven samples of pure infant rice from Spain was significantly (p < 0.01) higher (134 mg kg 1) than those from other countries (China (n = 14), USA (n = 5) and UK (n = 5)), whose contents were <39 mg kg 1 (LOD) (CarbonellBarrachina et al., 2012). The authors concluded that it will be necessary to identify the source and reduce the levels of lead in

Fig. 2. Upper-bound levels of lead in infant cereal-based food before being diluted (n = 17; value ± SD).

T. Guérin et al. / Food Chemistry 237 (2017) 849–856

Spanish foods. Still in Spain, mean lead levels in 91 infant cereals ranged from <1.6 (LOD) and 1027 mg kg 1 (mean 15.4 mg kg 1 compared with 10 mg kg 1 in this study; n = 17; range 3.8–27 mg kg 1) (Hernandez-Martinez & Navarro-Blasco, 2012). Added ingredients (milk, cocoa, fruit and honey) to cereal-based cause lead enrichment, as observed in this study (Fig. 2). Previously, Roca de Togores, Farré, and Frigola (1999) measured 29 samples of infant cereal commercialised in Spain, distinguished on the basis of incorporation of infant formula as a major ingredient, and concluded that there was a pronounced enrichment of lead in milk-added infant cereals (range 53.5–598 mg kg 1) in contrast to milk-free infant cereals (range 36.1–306 mg kg 1). Outside Europe, the mean levels of 48 tinned infant food samples of different brands and on 75 samples of different types of commercially available milk purchased in India, ranged from 1.7 to 3.35 mg Pb kg 1 and from 39.5 to 77.7 mg kg 1, respectively (Tripathi, Raghunath, Sastry, & Krishnamoorthy, 1999). Finally, lead concentrations in two Pakistani studies ranged from 28.7 to 119 mg kg 1 dry weight in 17 infant formulae samples (Kazi et al., 2009) and ranged from 52.5 to 90.6 mg kg 1 dry weight in 8 infant foods (Kazi et al., 2010). In the New Zealand TDS, mean lead concentration was estimated at 0.9 mg kg 1 with a maximum value at 2 mg kg 1 for infant and follow-on formula (n = 8; 5 < LOD) (MAF, 2011), which is consistent with our results (Table 3). Finally, several studies showed higher lead levels in cows’ milkbased infant formula than in human breast milk (EFSA, 2010; Ljung et al., 2011; Pandelova et al., 2012), except one study (Tripathi et al., 1999). According to EFSA, infant formula may contain up to three times the lead content of breast milk (EFSA, 2010). Therefore, controlling the lead content in infant formulae appears to be an essential requirement. Similarly, EFSA recommends to continue working on further reducing exposure to lead, from both dietary and non-dietary sources (EFSA, 2010). 4. Conclusions Using a sample introduction system with a temperature programmable cyclonic spray chamber, a sensitive ICP-MS method was developed and validated to attain low LOD values, close to 10 times lower than currently recommended levels for determining very low levels of lead in infant foods. Lead was quantified in 236 out of 291 infant foods (81% quantified and 90% detected) consumed by French infants and toddlers. Most samples contained relatively low concentrations of lead (range LB/UB 0.0/0.3– 16 mg kg 1). However, before being diluted, chocolate-containing foods generally had higher concentrations of lead compared to other foodstuffs. These results will be used shortly with other toxic trace element results and individual consumption data to assess the dietary exposure of infants and toddlers, and health risks. An additional study to determine the lead levels in French breast milk is currently underway, using this very sensitive ICP-MS method. Conflict of interest statement The authors state that no competing financial interests exist. Acknowledgments The authors would like to thank the Ministry of Food, Agriculture and Fisheries, the Ministry of Health and the Ministry of Ecology and Sustainable Development for their financial contribution.

855

References AFNOR (2010). Analyse des produits agricoles et alimentaires – Protocole de caractérisation en vue de la validation d’une méthode d’analyse quantitative par construction du profil d’exactitude. AFNOR NF V03–110. Saint Denis, France: Association Française de Normalisation. AFNOR (2014). Produits alimentaires – Dosage des éléments traces – Digestion Sous Pression. AFNOR NF EN 13805. Saint Denis, France: Association Française de Normalisation. ANSES (2013). Avis de l’ANSES et rapport d’expertise collective relatifs aux expositions au plomb: effets sur la santé associés à des plombémies inférieures à 100 lg/L. Maisons-Alfort, 31p. . (Accessed 03.01.17). Carbonell-Barrachina, A. A., Ramirez-Gandolfo, A., Wu, X., Norton, G. J., Burlo, F., Deacon, C., & Meharg, A. A. (2012). Essential and toxic elements in infant foods from Spain, UK, China and USA. Journal of Environmental Monitoring, 14, 2447–2455. Chevallier, E., Chekri, R., Zinck, J., Guérin, T., & Noël, L. (2015). Simultaneous determination of 31 elements in foodstuffs by ICP-MS after closed-vessel microwave digestion: Method validation based on the accuracy profile. Journal of Food Composition and Analysis, 41, 35–41. Codex Alimentarius Commission (2003). Report of the 35th session of the codex committee on food additives and contaminants (Arusha, Tanzania). EC (2006). Commission Regulation (EC) No 1881/2006 of 19 December 2006 setting maximum levels for certain contaminants in foodstuffs. Official Journal of the European Union, L364, 0005–0024. EFSA (European Food Safety Authority) (2010). Scientific Opinion on Lead in Food. EFSA Panel on Contaminants in the Food Chain (CONTAM). EFSA Journal, 8(4), 1–147 [1570]. EFSA (European Food Safety Authority) (2012). Scientific report of EFSA. Lead dietary exposure in the European population. EFSA Journal, 10(7), 1–59 [2831]. Fantino, M., & Gourmet, E. (2008). Nutrient intakes in 2005 by non-breast fed French children of less than 36 months. Archives de Pédiatrie, 15(4), 446–455. Frazzoli, C., & Bocca, B. (2008). Validation, uncertainty estimation and application of a sector field ICP MS-based method for As, Cd and lead in cow’s milk and infant formulas. Microchimica Acta, 162(1–2), 43–50. FSA (Food Standards Agency), (2003). Multi-element survey of infant foods. Food Survey Information Sheet No. 42/03, 17 pp. FSA (Food Standards Agency), (2006). Survey of metals in weaning foods and formulae for infants. Food Survey Information Sheet No. 17/06, 38 pp. Güldas, M., Dagdelen, A. F., & Biricik, G. F. (2008). Determination and comparison of some trace elements in different chocolate types produced in Turkey. Journal of Food, Agriculture and Environment, 6, 90–94. Hernandez-Martinez, R., & Navarro-Blasco, I. (2012). Estimation of dietary intake and content of lead and cadmium in infant cereals marketed in Spain. Food Control, 26, 6–14. Hulin, M., Bemrah, N., Nougadère, A., Volatier, J. L., Sirot, V., & Leblanc, J. C. (2014). Assessment of infant exposure to food chemicals: The French Total Diet Study design. Food Additives and Contaminants, Part A, 31(7), 1226–1239. Jorhem, L., Engman, J., & Schroder, T. (2001). Evaluation of results derived from the analysis of Certified Reference Materials: A user-friendly approach based on simplicity. Fresenius’ Journal of Analytical Chemistry, 370(2/3), 178–182. Kazi, T. G., Jalbani, N., Baig, J. A., Afridi, H. I., Kandhro, G. A., Arain, M. B., ... Shah, A. Q. (2009). Determination of toxic elements in infant formulae by using electrothermal atomic absorption spectrometer. Food Chemical and Toxicology, 47, 1425–1429. Kazi, T. G., Jalbani, N., Baig, J. A., Arain, M. B., Afridi, H. I., Jamali, M. K., ... Memon, A. N. (2010). Evaluation of toxic elements in baby foods commercially available in Pakistan. Food Chemistry, 119, 1313–1317. Ljung, K., Palm, B., Grandér, M., & Vahter, M. (2011). High concentrations of essential and toxic elements in infant formula and infant foods – A matter of concern. Food Chemistry, 127, 943–951. MAF (Ministry of Agriculture and Forestry) (2011). 2009 New Zealand total diet study – Agricultural compound residues, selected contaminant and nutrient elements. Report, 1–178. . (Accessed 03.01.17). Mermet, J. M., & Granier, G. (2012). Potential of accuracy profile for method validation in inductively coupled plasma spectrochemistry. Spectrochimica Acta Part B, 76, 214–220. Mermet, J. M., Granier, G., & Fichet, P. (2012). A logical way through the limits of quantitation in inductively coupled plasma spectrochemistry. Spectrochimica Acta Part B, 76, 221–225. Millour, S., Noël, L., Chekri, R., Vastel, C., & Guérin, T. (2011). Simultaneous analysis of 21 elements in foodstuffs by ICP-MS after closed-vessel microwave digestion: Method validation. Journal of Food Composition and Analysis, 24, 111–120. Moy, G., & Vannoort, R. (2013). Total diet studies. New York, Heidelberg, Dordrecht, London: Springer. Nagaoka, S., & Wilbur, M. S. (2003). Precision and accuracy of online autodilution using ISIS. Courtaboeuf, France: Agilent Technologies. Noël, L., Leblanc, J. C., & Guérin, T. (2003). Determination of several elements in duplicate meals from catering establishment using closed vessels microwave

856

T. Guérin et al. / Food Chemistry 237 (2017) 849–856

digestion with inductively coupled plasma mass spectrometry detection: estimation of daily dietary intake. Food Additives and Contaminants, 20, 44–56. Pandelova, M., Lopez, W. L., Michalke, B., & Schramm, K. W. (2012). Ca, Cd, Cu, Fe, Hg, Mn, Ni, lead, Se, and Zn contents in baby foods from the EU market: Comparison of assessed infant intakes with the present safety limits for minerals and trace elements. Journal of Food Composition and Analysis, 27, 120–127. Roca de Togores, M., Farré, R., & Frigola, A. M. (1999). Cadmium and lead in infant cereals – Electrothermal-atomic absorption spectroscopic determination. The Science of the Total Environment, 234, 197–201.

Tripathi, R. M., Raghunath, R., Sastry, V. N., & Krishnamoorthy, T. M. (1999). Daily intake of heavy metals by infants through milk and milk products. The Science of the Total Environment, 227, 229–235. Tuzen, M., & Soylak, M. (2007). Evaluation of trace element contents in canned foods marketed from Turkey. Food Chemistry, 102, 1089–1095. WHO/IPCS (World Health Organization/International Programme on Chemical Safety) (2009). Principles and methods for the risk assessment of chemicals in food, international programme on chemical safety, environmental health criteria 240. Dietary exposure assessment of chemicals in food. . (Accessed 03.01.17).