Exposure to contaminants and nutritional intakes in a French vegetarian population.

The NutriNet Santé study collected, on a voluntary basis, the dietary consumption of French vegetarian populations (N = 1766, including 188 vegan individuals) from 18 to 81 years (18-77 years for the vegan). Taking advantage of the availability of contamination data generated in the context of the second French total diet study, dietary exposures of French vegetarian populations to several contaminants were estimated. Results showed that exposures to persistent organic pollutants (PCBs, PCDD/Fs for instance) was dramatically lower than those of the general French population due to the non consumption of food of animal origins. On the other hand, exposures to phytoestrogens, some mycotoxins (T2 and HT2 toxins) and some trace elements (Cd, Al, Sn, Ni) were higher in the vegetarian population compared to those of the general population. Despite some limitations of this approach (both the consumption study and the total diet study were not aimed to estimate dietary exposure of the vegetarian populations), this study showed that dietary habits can dramatically influence the exposure of some contaminants.


Introduction
Since the 70's, the number of studies focusing on vegetarian diet dramatically increased (Scopus-Analyze search results). In early 70's, about 10 scientific publications per year were dealing with vegetarian diet, nowadays every year more than 300 scientific documents deal with this topic.
However, it is very difficult to have a clear description of the dietary habits of vegetarian populations since most of the studies published were conducted in small and highly selected samples. Based on the review by Guillocheau (2013), the percentage of vegetarians in the population greatly varies depending on the countries. Whereas it is close to 2% in France, it reaches 8% in Germany and as high as 31% in India.
People choose vegetarianism for several reasons: preservation of natural resources, protection of animal welfare or possible health benefits (Fox et al. 2008). Vegetarianism led to a variety of dietary habits: with or without milk and dairy products, eggs and fish (Leblanc et al. 2000). The American Dietetic Association stated that well-planned vegetarian diets are healthy and nutritionally adequate and might be beneficial in the prevention and treatment of some illnesses. Kahleova et al. (2015) showed that type 2 diabetes is 1.6 to 2 times lower in vegetarians than in the general population.
However, vegetarian diet can also be deleterious to human health (Mariotti, 2017). It has been reported that eliminating all animal products from the diet increases the risk of certain nutritional deficiencies. Micronutrients of special concern for vegans include vitamins B-12 and D, calcium, and long-chain n-3 (omega-3) fatty acids Educating the vegetarian population on nutrient supplements is important to reach the recommended daily intakes of each nutrient and prevent pathologies caused by nutritional deficiencies (Mariotti, 2017). In some cases, iron and zinc status of vegans may also be of concern because of the limited bioavailability of these minerals in vegetable sources (Craig W.J. The NutriNet-Santé study (Hercberg et al. 2010) was launched in France in 2009 to investigate the relationships between nutrition and health, the role of various determinants (sociological, economic and cultural) of dietary patterns and nutritional status and also their interactions. The study targeted adult volunteers of 18 years of age or more. On the other hand, a total diet study was run in France (Sirot et al. 2009) that provided information on concentrations of chemicals in food as well as associated dietary exposure for the general population. However, specific subpopulations were not included in this study as well as more sensitive or more exposed populations (Akhandaf et al. 2014).
Taking advantage of the consumption data previously published (Hercberg et al., 2010) and of the contamination data generated in the context of the second total diet study carried out in France, the present study aims to explore the exposure to different chemical contaminants (persistent organic pollutants, trace elements, mycotoxins phytoestrogens, acrylamide and polycyclic aromatic hydrocarbons) and some minerals of the vegetarian population among a French population and was compared to that of the general French population.

Processing consumption data
The aim of the NutriNet-Santé web-based cohort (Hercberg et al. 2010) is to investigate the relationship between nutrition and health. It also enables to study the role of various determinants (sociological, economic and cultural) of dietary patterns and nutritional status, and their interactions.
The study targeted adult volunteers aged 18 years or more.
Dietary intakes were assessed by three non-consecutive validated web-based 24h-dietary records, randomly assigned over a 2-week period (2 weekdays and 1 weekend day) (Lassale et al., 2016 ;Lassale et al., 2015 ;Touvier et al., 2011). Participants used a dedicated interface of the study website to declare all foods and beverages consumed during a 24h-period: three main meals (breakfast, lunch, dinner)  4 (LeMoullec et al., 1996). Mean daily energy, alcohol and nutrient intakes were estimated using a published French food composition table (>3 300 items) (Arnault et al., 2013). Amounts consumed from composite dishes were estimated using French recipes validated by food and nutrition professionals. Dietary underreporting was identified on the basis of the method proposed by Black (2000).
Only individuals that self-declared to be vegetarian were considered in the present study. No definition of vegetarianism has been given to the participants. Consequently individuals were categorized based on the food consumption they declared during the NutriNet Santé study.

Diet categories
According to the consumption habits, three categories of diets were defined: • the pesco-ovo-lacto-vegetarian diet (POLV) included the consumption of fish, seafood, eggs and milk,.
• the ovo-lacto-vegetarian diet (OLV) included the consumption of eggs and milk, • the vegan diet (V) excluded the consumption or the use of any products from animal origin.

Contamination data
Contamination data generated in the context of the second French total diet study (TDS2) were used in the present study to estimate dietary exposures. A core food approach was chosen: (1) identification of the most commonly consumed foods by the population of interest (90 % of the French total diet) and the main contributors to energy and nutrient intakes; (2) assessment of the mean intake of these foods; (3) sampling of the selected foods, preparation as consumed and analysis; (4) assessment of exposure combining consumption and contamination data (Sirot et al. 2009

Censored data and hypothesis
Concentrations below the limit of detection (LOD) or the limit of quantification (LOQ) are referred to as censored data. Lower bound (LB) and upper bound (UB) hypotheses were considered. The LB assumption corresponds to a scenario in which non-detected values are estimated to be 0 and the values detected, but not quantified, are estimated to be equal to the LOD. The UB assumption corresponds to a scenario in which non-detected values are estimated to be equal to the LOD and the values detected but not quantified are estimated to be equal to the LOQ. The LB scenario represents the minimum possible value as it underestimates concentrations below the LOD and those below the LOQ; on the other hand, the UB scenario represents the maximum possible value as it overestimates concentrations below the LOD and those below the LOQ (WHO 1995).

Estimation of the dietary exposure and the nutrient intakes
Where is the exposure (or nutrient intake) of individual i, the number of foodstuff that has been consumed by individual i, C , the amount consumed of the food k (in g/day) by the individual i, L the level of contamination (or nutrient content) in the food k, BW the individual body weight (in kg) of the individual i.

Risk characterization
The estimated exposures were compared to reference values in order to characterize the risk.
Tolerable daily intake (TDI), provisional tolerable weekly intake (PTWI) or provisional Tolerable Monthly Intake (PTMI) were considered for substances with a threshold effect. Benchmark dose limits (BMDL) were also considered (including for some compounds known to show genotoxic activity -non threshold compounds). The (BMDL 01 ), (BMDL 5 ) or (BMDL 10 ) are equivalent to an increasing of respectively 1%, 5% or 10% of the effects selected to set the health based guidance value. These health-based guidance values (HBGV) have been established by scientific instances (French, European or International). In the present study, only dietary exposures were considered. Additional routes of exposure (dermal, inhalation) were not considered. The methodology of total diet studies does not allow either to evaluate intakes and exposure in particular situations like in the case of an accidental or local contamination of food. It does not enable to characterize the risk relying to the consumption of "organic food" or imported products (Sirot et al. 2009).

Population characteristics and consumption habits
In had higher education with more than a half graduated a 2 nd or a 3 rd cycle. Socio-professional categories that were most represented in the studied population were intellectual functions (21.3%) and employees (18.9%). The most represented vegetarian diet was the OLV diet (54.9% of the population). The POLV diet was the second most represented diet (34.4%) and the vegan was the less represented (10.7%) ( Figure 2).
Twelve individuals did not indicate their bodyweight in the database and were excluded from the database and 1 754 individuals were finally included in the analysis.

Consumption
Consumption rate by food groups ranged from 0 (meat, poultry and game, offal and delicatessen meats) to 98.4% (for vegetables). Food groups that had the highest consumption rates were: , respectively) whereas the consumption of alcoholic drinks was more consumed in the general population than that in the vegetarian population (154.8 versus 62.6 g d -1 , respectively) ( Table 2).
Highest mean consumption quantities of slow-release carbohydrates were from bread and dried bread products. The mean consumption of these products (82.7 g d -1 ) was lower than that reported for the general French population (115 g d -1 ). This was also the case for potatoes. However, the percentages of consumers of cooked fruits and compotes, breakfast cereals, other cereals and chocolate were higher than those in the general French population. These food items were also more consumed by  (Table 2). Seafood and eggs were less consumed by vegetarians than by the general French population (about 1.5-2 times).
Nevertheless, vegetables and fruits were consumed in higher quantities and by a larger part of the vegetarian population in comparison to the general French population.

Dietary exposure and risk characterization
Trace elements and minerals

Arsenic (As)
The mean exposure to total As in the vegetarian population ranged from 0.56 µg. ) based on cancer effect. The margins of exposure (MOEs) were lower than that calculated in the context of the second total diet study performed in France (Arnich et al., 2010). Arsenic was already of health concern for the general population in France and consequently also is for the vegetarian population.
These exposures were higher than those previously reported by Clarke et al. (2003) Clarke et al., 2003).

Cadmium (Cd)
The mean cadmium exposure for the vegetarian population was 1.12 µg.
The mean aluminum exposure for the vegetarian population was estimated at 0.39 mg. UB: 2.7%) and fruits for the UB hypothesis (LB: 0.5 %, UB: 11.0 %). Only total mercury concentration data were available, and no hypothesis for speciation was considered to characterize the risk due to inorganic and organic mercury. However, the exposure of total mercury in the vegetarian population was in the same order of magnitude than those of the general population showing that dietary exposure to mercury is of concern for general population. Consequently, the risk due to mercury in  ) but 95 th percentile value was below the security limit.

Copper (Cu)
Mean Cu intake of the vegetarian population was estimated at 1.59 mg.d -1 . Intake at the 95 th percentile was 3.0 mg.d -1 . Main copper contributors were coffee (19.8%) dried fruits and oil seeds (10.0%).
These intakes were 20% to 25% lower than those for the general population (Kalonji et al., 2015).
Recommended dietary intakes being 1.5 and 2 mg.d -1 (Coudray 2001) these were adequate and below the security limit (5 mg.d -1 ). The vegetarian population in the United-Kingdom had a mean copper intake of 1.4 mg.d -1 according to Clarke et al. (2003).

Sodium (Na)
Mean Na intake of the vegetarian population was estimated at 1.7 g.d -1 . At the 95 th percentile, intake was 2.9 g.d -1 . Main sodium contributors were bread and dry baking (28.5%) and condiments and sauces (10.0%). Intakes in the present study were 40% lower than that for the general population and lower than recommendations (2.36 g.d -1 and 3.14 g.d  . These intakes were close to those already reported for the general population (Kalonji et al. 2015).
Main iron contributors in the vegetarian population were coffee (19.8%) and dried fruits and oil seeds (10.0%).

Summary for trace elements and minerals
Considering the vegetarian population, the risk cannot be excluded for several TEM; this is the case for Cd, Al, Ni, Mg, Cu, Na and Fe. Studied vegetarian population was more exposed to cadmium and aluminum than the general French population. On the other hand, lower intakes have been calculated for the vegetarian population than those for the general French population for Mg, Cu, Na and Fe. Mg and Fe intakes were below RDA values but copper and sodium intakes were close to those values. In the present study, percentage of individuals exceeding the respective recommended values for Cd, Al, Ni and Mg were higher than those of the general population (between 2.5 and 10 times higher). For As and Fe, the risk could not be excluded for both contaminants. The risk could be ruled out for the studied vegetarian population for Ca. However, Ca intakes were below RDA values.
Compared to the general population, it appeared that some exceedances of the health based guidance values were for the same substances except for Ca for which intakes were lower in vegetarian populations (no exceeding) and Ni for which the health based guidance value was dramatically decreased between the second total diet study published in 2009 and the present study.

Persistent organic pollutants
Version preprint ), PBDEs are of no concern for the vegetarian population. The same conclusion was drawn when the risk was characterized using BMDL 10 set by EFSA based on the effects on neurodevelopment for the following four congeners: BDE-47, -53, -153 and -209 (EFSA 2011, data not shown).

Summary for POPs
All exposure results to persistent organic pollutants were below those of the general population. There was a factor of ten between vegan and non-vegan exposure values to furans, dioxins and dl-PCBs on the one hand and to ndl-PCBs on the other hand. This was due to the strong contribution of fish and mollusks to the exposures of these compounds that are not consumed by the vegan population. No exceedance has been identified in the vegan subpopulation whereas within the general population, these substances were considered as "of-concern".

Aflatoxins
For the lower bound hypothesis (LB), mean exposure and exposure at the 95 th percentile to the sum of the aflatoxins were 0 ng. . Consequently, the exposure of the vegetarian population calculated in the present study is considered of no concern.
Trichotecens Natural trichotecens are classified in different groups according to their chemical structure (groups A to C). Groups A and B are most commonly found in food. In group A, T-2 toxin and HT-2 toxin have been selected for the present study. In group B, nivalenol (NIV) and deoxynivalenol (DON) were studied. In  mycotoxins are therefore considered of concern. Main T2 and HT2 toxins contributors were pasta (LB: 32.5%, UB: 20.7%), bread and dried bread products (LB: 26.9%, UB: 27.6%) and rice and wheat products (LB: 15.2%, UB: 13.7%).

Deoxynivalenol (DON)
Mean exposure of the vegetarian population to deoxynivalenol was estimated at 285 ng. With regards to phytoestrogens, exposures of the vegetarian population were far above those reported for the general population (10 to 300 times higher). Overall exposure was mainly due to isoflavone and to a smaller extent to coumestrol and equol. Consumption rate of "food for particular uses (energy drinks, meal substitutes, etc.) " was 50 % instead of 12 % in the general French population. The quantity of food consumed from that category was 6 times higher in the vegetarian population compared to that for the general French population. "Food for particular uses" category was the main contributor of phytoestrogens exposure in vegetarian population. Clarke et al (2003) reported an average algycine daily intake of 10.5 mg.d -1 for a vegetarian population in the United-Kingdom.
Margin of exposure for PAH4 was calculated using the BMDL 10 of 0.34 mg.kg bw -1 .d -1 (set by EFSA based on carcinogenicity (2008c), and was above 100 000 (table 10). Exposure of PAH11 was below the virtually safe dose set up by RIVM (Baars et al., 2001). Consequently, considering PAH4 and PAH11 and their respective health-based guidance values, dietary exposure to these compounds is not of concern for the studied vegetarian population.

Conclusion
Exposures to pesticides were not calculated in the present study. The reason being that available contamination data were generated not taking into account the organic or non-organic origins of the food items. Indeed, the vegetarian population of the Nutrinet-Santé study seemed to consume a large part of organic food (Baudry et al., 2015;Baudry et al., 2016) that could be less contaminated than conventional food. For example, a first study among participants of the BioNutrinet-Santé study highlighted that vegetarian or vegan participants had a share of organic food of nearly 50%. The study Version preprint Comment citer ce document : Fleury, S. (Co-premier auteur) For some contaminants considered in the present study, some differences of the exposure appeared between the vegetarian and the French general populations. Exposures to phytoestrogens, some mycotoxins (T2 and HT2 toxins) and some TEM (Cd, Al, Sn, Ni) were higher in the vegetarian population compared to those in the general population.
Among the vegetarian population, main contributor to phytoestrogens appeared to be "foods for particular nutritional uses". These products are consumed by a larger part of vegetarian population (50% instead of 12% within the general population) and in higher amount: six times more than the general French population. Concerning exposure to isoflavones, the risk could not be excluded.
Mycotoxins showed a high exposure to T2-HT2 toxin and patuline. One of the main contributors to T2 and HT2 toxins exposure was the "rice & wheat" food category. Food items in this category were consumed on average in quantities two times higher in the vegetarian population than in the general population. For vegetarians, exposure to patuline was mostly due to fruits, cooked fruits and compotes and non-alcoholic flavoured drinks. Consumption level of these contributors are higher in the vegetarian population. Higher exposure to tin, nickel, cadmium, as well as to aluminum was demonstrated in the vegetarian population compared to the general population. High consumption of fruits and vegetables (cooked, raw and dry) in the vegetarian population appeared to be the reason for this observation.
The vegetarian population studied in the present study was, however, less exposed than the general population to some contaminants. This was true for several persistent organic pollutants; PAHs; Comment citer ce document : Fleury, S. (Co-premier auteur) to the general population. The lower consumption of coffee, cheese, fish lowered Cu, Ca and As exposures. Lastly, exposure to some substances did not appear to be dramatically different between vegetarian and non-vegetarian populations. This is the case for Pb, Hg and some mycotoxins (fumonisins, OTA) that showed exposure similar to those previously reported for the general population.
In conclusion, despite some limitations due to the way the consumption data were gathered and to the fact that the contamination data generated in the context of the study not specifically focusing on the vegetarian populations, the present study highlights that the exposure pattern of the vegetarian population to contaminants is dramatically different than that reported for the general population.
Comment citer ce document :