Healthy, safe and environmentally sustainable consumption and production is important for human beings and the planet. In order to mitigate climate change, we need to consume and produce in a more environmentally sustainable manner. In the long term, the consumption of unsafe and unhealthy food might cause adverse health effects, varying from diarrhoea to several types of cancer(1).
Several studies have described the relationship between education level and health-related behaviours, including dietary habits(Reference Cardel, Guo and Sims2,Reference Lopez-Azpiazu, Sánchez-Villegas and Johansson3) . According to several studies, the highly educated consume more healthy foods such as fruit and vegetables compared with less-educated ones(Reference De Irala-Estevez, Groth and Johansson4,Reference Geurts, Beukers and van Rossum5) . Little is known regarding educational differences in other aspects of the diet, such as environmental sustainability and food safety. Friedl et al.(Reference Friedl, Pack and Omann6) showed that people with low education level consumed more food that have a higher impact on the environment (e.g. meat products) compared with those with high education level. In contrast, Reynolds et al.(Reference Reynolds, Horgan and Whybrow7) have shown that greenhouse gas emissions (GHGe) of the total diet are similar between income groups, although there are differences by types of meat. The relationship between education level and food safety is rather unknown. In previous Dutch National Food Consumption Surveys, differences in food consumption patterns between high and low education groups were observed(Reference van Rossum, Fransen and Verkaik-Kloosterman8); differences in food safety, environmentally sustainable and healthy food consumption are, therefore, to be expected.
To decrease inequalities in health between education groups, insights are necessary in the underlying factors, such as healthy and safe food consumption. Healthy food consumption is important for planetary health. More and more dietary guidelines target health as well as environmental aspects(Reference Kromhout, Spaaij and de Goede9,Reference Buttriss10) . However, it is not yet known whether the environmental sustainability of diets differ for different education groups, and thus whether such guidelines can focus on the general population or should focus on specific subgroups of the population. This study aimed to describe the educational differences in healthy, environmentally sustainable and safe food consumption among adults aged 19–69 in the Netherlands.
Healthy food consumption was evaluated by the components of the Dutch Healthy Diet Index 2015 (DHD15 Index); environmentally sustainable consumption, by diet-related GHGe; and safe food consumption, by exposure to a selection of contaminants present in food. Microbiological food safety was not addressed in this study.
Methods
In the present study, data of the Dutch National Food Consumption Survey 2007–10 (DNFCS 2007–10) were used(Reference van Rossum, Fransen and Verkaik-Kloosterman8). Details of the design and methodology of DNFCS 2007–10 have been described previously(Reference van Rossum, Fransen and Verkaik-Kloosterman8). Briefly, the study population consisted of people living in the Netherlands aged 7–69 years. The sampling frame was a representative consumer panel from which sex- and age group-stratified random samples were taken. Data were collected between March 2007 and April 2010. Representativeness of the Dutch population was monitored and adjusted during recruitment, regarding age groups, region, urbanisation level and education level. In total, all data from this survey were included in the present analysis (1055 males and 1051 females aged 19–69 years). This age range was based on the age boundaries in the Dutch dietary reference values. Response rate for this age group was 70 %(Reference van Rossum, Fransen and Verkaik-Kloosterman8).
Dutch National Food Consumption Survey 2007–10 data collection
Data collection within DNFCS 2007–10 consisted of a questionnaire to obtain general information of the participants, including sociodemographic characteristics and lifestyle factors, and two non-consecutive 24-h dietary recalls. The sociodemographic characteristics included working status, income and highest obtained education level. Education level was categorised into low (primary school, lower vocational, low or intermediate general education), moderate (intermediate vocational education and higher general education) and high (higher vocational education and university). The lifestyle factors included alcohol consumption and general characteristics of the diet.
The 24-h recalls were conducted via computer-assisted telephone interviews using GloboDiet software (©International Agency for Research on Cancer; previously called EPIC-Soft©). The GloboDiet classification consists of seventeen main food groups (including seventy-two subgroups)(Reference Slimani, Casagrande and Nicolas11). Interviewers were trained dieticians and called unannounced(Reference van Rossum, Fransen and Verkaik-Kloosterman8). During these interviews, a detailed description of all foods (including beverages) and amounts consumed (by means of household measures, by weight or volume photographed from a delivered booklet) was collected. During the interviews, height and body weight (BW) was reported. BMI was calculated by dividing the BW (in kg) by height-squared (in m2). All reported foods were matched to the codes in the Dutch Food Composition Database (NEVO-2011)(12), the so-called NEVO codes.
Healthiness of the diet
We used the Dutch dietary guidelines 2015 of the Health Council and an overall score, the DHD15 Index, to score the diet on healthiness using the food intakes of DNFCS 2007–10 (see Table 1)(Reference Kromhout, Spaaij and de Goede9,Reference Looman, Feskens and de Rijk13) . The index is a summary score based on fifteen single components, including fruit, vegetables, fish, wholegrain products, fats and oils, legumes, nuts, dairy intake, red meat and processed meat, sodium, coffee, tea, sweetened beverages, fruit juices and alcohol. As described by Looman et al.(Reference Looman, Feskens and de Rijk13), some recommendations require a minimal intake (e.g. fruit, vegetables) or maximal intake (e.g. sodium); other recommendations an optimal intake (e.g. dairy products) or a replacement (e.g. fats and oils). For each recommendation, participants can proportionally score between 0 and 10 points, depending on the type of recommendation (minimum, maximum, optimal intake or replacement). For instance, in case of a minimum intake, a score of 10 points was allocated when the consumption was higher than or equal to the minimum intake (e.g. 200 g of fruits per d); no consumption was given 0 point. In case of a maximum intake, a score of 0 point was allocated when the consumption was higher or equal to the maximum intake (e.g. 6 g of salt per d); no consumption was given 0 point. In the present study, food intake relevant to each guideline was calculated as well as DHD15 Index per participant using the average of the two 24-h recalls.
Components of the Dutch dietary guidelines 2015 and their definition in the present study(Reference Kromhout, Spaaij and de Goede9)*
NOVO, Dutch Food Composition Database.
* Health Council of the Netherlands (2015), Dutch dietary guidelines 2015(Reference Kromhout, Spaaij and de Goede9).
† The components of dietary guidelines were all used in the calculation of the Dutch Healthy Diet Index (2015, except for coffee).
‡ No data was available on distinction between filtered or unfiltered coffee. This component was therefore excluded from the present analysis.
Greenhouse gas emissions of diets
For assessing the environmental sustainability of foods consumed, indicators such as the use of energy, water and land, and GHGe were typically used. GHGe has been used as an indicator for the overall environmental impact in multiple studies(Reference Fisher, James and Sheane14–Reference Marinussen, Kramer and Pluimers16), and consists of the emission of CO2 equivalents (e.g. CO2, NO2 and CH4) along the supply chain. In the present analysis, this indicator was used to assess the environmental sustainability of food. The data and methodology has been described in Temme et al.(Reference Temme, Toxopeus and Kramer15). In summary, this was done by linking the values of GHGe per NEVO code (Blonk dataset version 2014) to the food consumption data coded with NEVO codes. The GHGe data were calculated via life cycle assessment (LCA)(Reference Temme, Toxopeus and Kramer15). All stages of a product’s life – from primary production, processing, packaging, transportation, storage, preparation and cooking – were taken into account. Food waste was included by using food group-specific percentages for avoidable and unavoidable food losses throughout the food chain, including the consumer phase(Reference Temme, Toxopeus and Kramer15). The LCA took into account the origin of foods as available on the Dutch market (e.g. share of imported foods)(Reference van de Kamp, van Dooren and Hollander17). In total, 254 food products in the Blonk database were previously extrapolated to 1595 consumed food products in the food consumption database to quantify GHGe. Extrapolation was used based on ingredient composition and similarities in the type of food or production methods. In the present analysis, the GHGe of the overall diet was calculated. In addition, the GHGe of several food groups were described.
Chemical food safety
Chemical food safety deals with a wide range of substances present in food, including pesticides, food additives and contaminants. Contaminants are substances unintentionally present in food due to food processing (e.g. acrylamide, polycyclic aromatic hydrocarbons and 3-monochloropropane-1,2-diol (3-MCPD)), environmental contamination (e.g. dioxins, lead and cadmium), fungi (mycotoxins) or naturally (e.g. nitrate and arsenic). Based on current dietary patterns, the possible risks to public health are more frequently calculated for contaminants than for substances added by humans during food production or processing(Reference Mengelers, de Wit and Boon18). The use of the latter category of substances such as food additives, pesticides and veterinary drugs is legally regulated; these substances are only permitted if their addition might not constitute any risk to public health.
Food safety was evaluated in relation to education level for a selection of contaminants (see Table 2). For some of these contaminants, a potential health risk based on prior exposure assessments performed in the Netherlands could not be excluded(Reference Boon, Te Biesebeek and van Donkersgoed19–Reference Sprong, de Wit-Bos and te Biesebeek21). Furthermore, for the selected contaminants, concentration data are readably available. Per contaminant, the food products that may contain the contaminant are also described in Table 2.
Overview of chemical compounds and food products in which they may occur
3-MCPD, 3-monochloropropane-1,2-diol; DON, deoxynivalenol; OTA, ochratoxin A.
First, the average daily exposure to the different contaminants was calculated. Concentrations of aflatoxin B1, ochratoxin A (OTA), deoxynivalenol (DON), nitrate and acrylamide were obtained from Boon et al. and EFSA(Reference Boon, Baars and van Klaveren22–26). The concentration data of methylmercury was obtained from RIKILT and RIVM (2015) and EFSA(27,28) , of lead from Boon et al. and EFSA(Reference Boon, Te Biesebeek and Sioen29,30) and of 3-MCPD from Boon and Te Biesebeek and EFSA(Reference Boon and Te Biesebeek31,32) . The mean middle-bound concentrations (samples with an analysed level below the limit of detection or quantification assumed to contain the contaminant at half the relevant limit value) per food product were used. The analysed foods were subsequently matched – unweighted – to the relevant products or subgroups (seventy-two in total) of the GloboDiet classification. For instance, a concentration of 0·5 µg/kg OTA was assigned to biscuits (generic: subgroup biscuits), and a concentration of 10·7 µg/kg OTA was assigned to dried apricot (specific: product).
As differences in exposure to contaminants between education groups are only relevant if exposures result in potential health risks, the calculated exposures were compared with the relevant health-based guidance values (HBGV), or a margin of exposure (MOE) was calculated. HBGV is the maximum intake per unit of time, usually per day or week (such as the tolerable daily or weekly intake). The calculated exposure must be higher than HBGV for a potential health risk. MOE was calculated by dividing the lower limits of benchmark doses (BMDL) by the calculated exposure. BMDL represents doses in toxicity studies in which a percentage (e.g. 1, 5 and 10 %) increase in an adverse effect is observed. BMDL cannot be viewed as the maximum acceptable intake and is, therefore, evaluated via the calculation of MOE. For a potential health risk, MOE must exceed a minimum value, which can vary between 1 and 10 000, depending on the nature of the critical endpoint on which BMDL is based. HBGV or BMDL used in this study are listed in Table 3, including the minimum value of MOE for a negligible health risk.
Health-based guidance values and BMDL of various contaminantsa, including the minimum margin of exposure (MOE) for a negligible health risk, if relevant
3-MCPD, 3-monochloropropane-1,2-diol; ADI, acceptable daily intake; BMDL, lower limit of the benchmark dose; BMDL10, lower limit of the 95 % CI of the estimated dose with a 10 % additional risk; BW, body weight; DON, deoxynivalenol; OTA, ochratoxin A; TDI, tolerable daily intake; TWI, tolerable weekly intake.
* For a comparison with calculated intakes per d, these health-based guidance values were divided by 7.
† The minimum value of MOE of 1 for lead is related to a very low potential health risk.
Data analyses
In order to calculate the differences in healthy, environmentally sustainable and safe food consumption by education level, the mean consumption of components of the Dutch dietary guidelines 2015, and the mean emissions of CO2 equivalents and mean exposure to contaminants over the two consumption days from the 24-h recalls were calculated per participant. For the contaminants, the mean exposure was divided by the self-reported BW of the participant in kilograms, as both HBGV and BMDL are expressed per kg BW (Table 3).
Mean consumption and emission levels were used as dependent variables in ANOVA to test on statistical significance between education groups. Education level was used as the independent variable. All statistical analyses were performed with SAS 9.3 (SAS Institute). A weighting factor was used to correct for small deviances in sociodemographic characteristics (e.g. region, level of urbanisation), season and day of the week(Reference van Rossum, Fransen and Verkaik-Kloosterman8). Since men and women have different energy intakes, the statistical analyses were performed separately for men and women(Reference van Rossum, Fransen and Verkaik-Kloosterman8). It was assumed that a P-value <0·05 is statistically significant.
Results
On average, low-, moderate- and high-educated men were 45, 43 and 46 years old, respectively (P = 0·002). Low-, moderate- and high-educated women were aged on average 49, 40 and 43 years, respectively (P < 0·0001). The BMI of men did not differ between education groups (26 kg/m2; P > 0·05). For women, the mean BMI of low, moderate and high education groups was 27, 26 and 25 kg/m2, respectively (P = 0·0007) (see Table 4). The mean energy intake for men was 2687, 2638 and 2504 kcal for low, moderate and high education groups, respectively (P = 0·008). For women, the corresponding figures are 1915, 2001 and 1933 kcal, respectively (P > 0·05) (see Table 4).
General characteristics (income, working status, age, BMI, intake of energy, proteins, fats and carbohydrates) for men and women aged 19–69 years by education level (weighted for sociodemographic factors, n 2106, DNFCS 2007–10)
Significantly different from the low education group with *P < 0·05, idem with **P < 0·01, idem with ***P < 0·001.
† Education groups.
Healthiness of the diet
Table 5 shows the results of healthy food consumption. For both men and women, the high education group consumed on average more vegetables and fruit than the low education group. Particularly, the consumption of fruit was approximately a quarter more in high than in low education group. In contrast, the low education group consumed significantly more meat and meat products than the high education group (men 148 v. 115 g, P < 0·0001; women 93 v. 79 g, P = 0·02). In line with this, the consumption of red meat was higher in low than high education group. Finally, salt consumption was lower in high-educated men compared with low-educated men (2995 v. 3174 mg, P = 0·03); salt consumption of moderate-educated women was higher compared with low-educated women (2466 v. 2330 mg, P = 0·02). Altogether, for both men and women, the high education group had a higher overall DHD15 Index score compared with the low education group (men 59 v. 53 points, P < 0·0001; women 69 v. 64 points, P = 0·0002).
Components of Dutch dietary guidelines 2015 (in g/d) for men and women aged 19–69 years by education level (weighted for sociodemographic factors, season, day of the week, per age-sex group, n 2106, DNFCS 2007–10)
Significantly different from the low education group with *P < 0·05, idem with **P < 0·01, idem with ***P < 0·001.
† Education groups.
Some educational differences were observed in men or women only. Among men, the consumption of wholegrain products was higher in the moderate and high education group compared with the low education group (114 and 113 v. 99 g, P = 0·02). Low-educated men consumed significantly more processed meat and sugar-containing beverages compared with high-educated men (processed meat 69 v. 48 g, P < 0·0001; sugar-containing beverages 344 v. 265 g, P = 0·04). Among women, the consumption of cereals and cereal products was significantly higher in moderate-educated than low-educated women (190 v. 167 g, P = 0·0004). Moreover, the consumption of non-alcoholic beverages and tea was significantly higher in high-educated compared with low-educated women (non-alcoholic beverages 1990 v. 1802 g, P = 0·009; tea 390 v. 283 g, P = 0·01).
Greenhouse gas emissions of the diet
The overall GHGe and the GHGe of food groups contributing most to total GHGe are shown in Table 6. The food groups contributing most are mainly animal-based products, including meat products, dairy products, fish and eggs. Beside these food groups, some plant-based food groups contribute to total GHGe, including cereal products, vegetables and (non-)alcoholic beverages. Overall, the GHGe for both men and women did not differ between education groups. However, the sources of GHGe are different between education groups. The GHGe through the consumption of vegetables and fruiting vegetables was approximately a quarter higher in high compared with low education group. Moreover, the GHGe via the consumption of fruit juices was about 33 % higher in high-educated men and 40 % in high-educated women compared with the low education groups. The GHGe of meat consumption did not differ between high and low education groups.
Greenhouse gas emission (in kg CO2 equivalents per d) for contributing food groups and for the overall diet for men and women aged 19–69 years by education level (weighted for demographic factors, season, day of the week, per age-sex group, n 2106, DNFCS 2007–10)
Significantly different from the low education group with *P < 0·05, idem with **P < 0·01, idem with ***P < 0·001.
† Education groups.
‡ Milk including milk beverages.
Also for GHGe, some educational differences were observed in men or women only. Among men, the GHGe of the consumption of soft drinks was higher in low-educated compared with high-educated men (0·16 v. 0·10, P = 0·0001). Among women, the GHGe via the consumption of eggs and cereals and cereal products was higher in moderate-educated compared with low-educated women (eggs 0·04 v. 0·03, P = 0·0007; cereals and cereal products 0·20 v. 0·17, P = 0·0002). In addition, the GHGe via fish consumption was also higher in high than low education group (women 0·10 v. 0·08 in kg CO2 equivalents/d, P = 0·03).
Exposure to contaminants
The results in Table 7 show that the mean intake of 3-MCPD was significantly higher in low-educated compared with high-educated men (0·49 v. 0·39 µg/kg BW/d, P = 0·002). For women, the mean exposure to methylmercury was significantly higher in high-educated compared with low-educated women (0·13 v. 0·11 µg/kg BW/d, P = 0·002). Moreover, high-educated women had also a higher intake of lead (0·40 v. 0·32 µg/kg BW/d, P < 0·0001), aflatoxin B1 (0·0005 v. 0·0003 µg/kg BW/d, P = 0·003), DON (0·06 v. 0·05 µg/kg BW/d, P = 0·01) and OTA (0·06 v. 0·05 µg/kg BW /d, P < ·0001) compared with low-educated women. The mean intake of nitrate was higher in low-educated than in moderate-educated women (1·48 v. 1·33 mg/kg BW/d).
Exposure to contaminants for men and women aged 19–69 years by education level (weighted for sociodemographic factors, season, day of the week, per age-sex group, n 2106, DNFCS 2007–10)†
3-MCPD, 3-monochloropropane-1,2-diol; BW, body weight; DON, deoxynivalenol; OTA, ochratoxin A.
Significantly different from the low education group with *P < 0·05, idem with **P < 0·01, idem with ***P < 0·001.
† Low education level used as the reference group.
‡ Education groups.
Compared with the relevant health limits, the mean intake of acrylamide and aflatoxin B1 of all education groups resulted in margins of exposure that are lower than the minimal level above which the health risk is negligible. For the other contaminants, the mean intakes were either lower than the relevant HBGV or resulted in margins of exposure that are sufficiently high in all education groups (see Table 3)(33).
Discussion
This is the first study that simultaneously describes differences in healthy, environmentally sustainable and safe food consumption across education groups in the same population. We expected differences in food consumption patterns between high and low education groups, and therefore we expected differences in food safety, environmental sustainability and healthy food consumption in high compared with low education groups. The results showed educational differences in several indicators of healthy and environmentally sustainable food consumption. Differences in education level are both favourable and unfavourable in the domains of healthy and environmentally sustainable food consumption. Overall, the high-educated group showed higher adherence to the Dutch dietary guidelines compared with the low-educated group. The high education group consumed more fruits and vegetables and less meat and fats than the low education group. In addition, no differences were found between the GHGe of high and low education groups. Regarding contaminant exposure, among men, the mean intake of 3-MCPD was estimated to be lower in high compared with low education group. Among women, the mean intakes of methylmercury, lead, aflatoxin B1, DON and OTA were estimated to be higher in high compared with low education group. The mean intakes in all education groups were lower than the relevant HBGV or resulted in margins of exposure that are sufficiently high, except for acrylamide and aflatoxin B1.
The total GHGe did not differ between education groups, However, the contributing food groups differed between high and low education groups due to different food consumption patterns. These results are in line with Reynolds et al.(Reference Reynolds, Horgan and Whybrow7). In the present study, the consumption of fruit, vegetables and fish was higher in high compared with low education group. Therefore, the GHGe of these food groups was higher in high education group. In contrast, the consumption of meat was lower in the high education group. The GHGe due to half-and-half minced meat, pork meat and processed meat consumption was significantly lower in high-educated compared with low-educated men. For women, the GHGe of processed meat consumption was significantly lower in high-educated. In this way, the overall effects on GHGe are diminished. For food safety, differences in the intake of contaminants could also be explained by differences in food consumption patterns. High-educated men had a lower consumption of margarines than low-educated men. As margarines is one of the main contributors to the intake of 3-MCPD, the mean intake of this contaminant was estimated to be lower in high compared with low education group. The consumption of fruits and vegetables was significantly higher in high-educated compared with low-educated women. Fruits and vegetables contribute both to the intake of lead; therefore, the mean lead intake was estimated to be higher in high compared with low education group.
Previous research has shown that the high education group consume more fruit and vegetables than the low education group(Reference De Irala-Estevez, Groth and Johansson4). In line with the present analysis, Darmon and Drewnoski(Reference Darmon and Drewnowski34) have found that the high education group consume more fish (Denmark, the Netherlands and France), whereas the low education group consume more fats (Denmark, the Netherlands). A study by Hulshof et al. has shown that the high education group consume less potatoes and meat than the low education group (the Netherlands)(Reference Hulshof, Brussaard and Kruizinga35,Reference Geurts, Beukers and Buurma-Rethans36) .
With respect to environmentally sustainable food consumption, GHGe was used as an indicator. Insufficient data were available on water use and energy expenditures as well as other environmental aspects(Reference Aldaya, Chapagain and Hoekstra37). Additional research is needed to estimate the impact on, for example, water use and energy expenditures and how this may affect the results. Data was available on land use(Reference Marinussen, Kramer and Pluimers16); however, previous studies have shown that GHGe and land use are highly correlated and lead to similar conclusions(Reference Temme, Van Der Voet and Thissen38). Also in other studies, GHGe was often used as an indicator for environmental sustainability(Reference Jones, Hoey and Blesh39).
In relation to safe food consumption, only indicative intake estimates were calculated to obtain mean intake levels of contaminants for the different education groups. These mean intakes were estimated by linking the concentration data and food consumption data to food subgroups, and thus ignoring the variations in contamination levels within these food groups. However, all contaminants examined in this study exert their possible adverse effects on health over a longer period of time, from several years up to life-long. For this type of assessments, mean concentrations are usually used because it may be assumed that fluctuations in concentrations will level out in the long run. Personal preferences for certain (brands of) foods containing higher mean levels of contaminants were not considered in this study.
Previous research studying the intakes of contaminants via food in the Netherlands in more detail based their conclusions of food safety on the whole population’s intake distribution(Reference Boon, Te Biesebeek and van Donkersgoed19–Reference Sprong, de Wit-Bos and te Biesebeek21,27,Reference Boon and Te Biesebeek31) . The mean intake estimates of different contaminants in the present analysis showed a similar trend compared with these studies. The intakes of aflatoxin B1 and acrylamide resulted in insufficiently large margins of exposure in all education groups (see Table 7). The percentages of individuals who did not exceed the MOE of 10 000 in aflatoxin B1 were 73, 80 and 82 % for low-, moderate- and high-educated men, respectively, and 73, 78 and 84 % for low-, moderate- and high-educated women, respectively. For acrylamide, the corresponding percentages ranged from 98 to 99 % in all education groups, both men and women. For these two contaminants, a possible health risk could not be excluded. For other contaminants, the mean intakes of all education groups were below HBGV or resulted in insufficiently high MOE (see Table 7). However, based on the mean intakes, it was not possible to conclude if there is a public health concern for these contaminants. For that, the whole exposure distribution should be considered. For lead and OTA, a possible health concern could not be excluded in previous studies at the upper part of exposure distribution(Reference Boon, Te Biesebeek and van Donkersgoed19,Reference Sprong, de Wit-Bos and te Biesebeek21) .
In this study, only a selected number of contaminants were taken into account. Due to the differences in food consumption patterns in low compared with high education group, the intakes of other chemicals were less likely to differ between education groups. However, no data was directly available for these analyses. If food consumption differences between educational groups will also result in differences in safe food consumption, it needs further research.
DNFCS 2007–10 represents the consumption behaviours of adults aged 19–69 in the Netherlands. A weight factor was used to correct for small deviances in representativeness for the Dutch population. Food consumption was assessed by two 24-h recalls per participant, and on average, energy intake was underreported. The proportion of low reporters on energy intake was 17 %, whereas the proportion of high reporters was 1·5 %(Reference van Rossum, Fransen and Verkaik-Kloosterman8). This was not taken into account. Furthermore, the energy intake of highly educated men was lower compared with low-educated men. In the present analysis, the food consumption data was not adjusted for energy intake. Energy intake might explain some of the differences found between the education groups. Nevertheless, the aim of this study was to describe the differences in healthy, environmentally sustainable and food safety between education groups. Further research is needed to examine the factors that explain these differences.
We used the mean intake of two 24-h recalls as a measure of dietary intake, which may be subject to day-to-day variation. On the group level, the within-person variation tends to be cancelled out, and only the precision of mean intake estimates is affected. With the sample size of over 2000 men and women in DNFCS 2007–10, relevant differences can be observed.
To decrease health inequalities between education groups, insights are necessary in different aspects of food consumption (e.g. healthy, environmentally sustainable and food safety). Besides education level, other factors such as lifestyle factors (e.g. smoking), obesity and price of the diet might play a role in these inequalities(Reference Peeters, Barendregt and Willekens40,Reference Darmon and Drewnowski41) .
Both the databases on food safety (concentrations used) and environmental sustainability were based on rough estimations. It was possible that the exposure to food contaminants was overestimated (by using extrapolation) and underestimated for environmental sustainability. The database for environmental sustainability includes uncertainties about shares and amounts of fertilisers and variability in energy inputs during processing, which may underestimate environmental sustainability. However, these uncertainties and variabilities relate to the nature of the data affecting food safety values and environmental variables, so that results by population groups are equally subjected to bias. Therefore, comparison between population groups is possible. Future researches should reduce uncertainties and include variability in dietary model estimates.
Overall, this is the first study to provide an insight into educational differences in healthy, environmentally sustainable and safe food consumption. The consumption patterns differed by education groups, resulting in a more healthy but equally environmentally sustainable diet among high compared with low education group. Exposure to food contaminants differed between education groups, but were not above safety levels, except for acrylamide and aflatoxin B1. For these substances, a health risk could not be excluded for all education groups. The results suggest that healthy, environmental sustainability and safe food consumption should be considered in policy measures and addressed by other researchers. Hence, the insights of this study may be useful in drafting policy measures focusing on improving healthy, safe and sustainable diets for all.
Acknowledgements
Acknowledgements: We would like to thank the reviewers for their intellectual contribution. Financial support: This research received no specific grant from any funding agency, commercial or not-for-profit sectors. Conflict of interest: None. Authorship: L.M.B. designed the study together with C.T.M.R. and M.C.O. L.M.B. carried out the analyses and wrote the article. C.T.M.R. supported and reviewed the article multiple times. M.C.O. reviewed the article with respect to healthy food consumption. E.H.M.T. played a role in the design of the sustainability aspect. E.H.M.T. reviewed the article on its content and especially sustainability. P.E.B. reviewed the article on its content and food safety. She played a role in the design and methodology for food safety. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki. The Medical Ethical Review Committee of Utrecht University confirmed the study was not subject to the Medical Research Involving Human Subjects Act (WMO) of the Netherlands (METC protocol 12-259/C). Medical ethical review was thus not needed. Written informed consent was obtained from all subjects before a face-to-face interview, but not for subjects interviewed by telephone. For the latter interviews, written informed consent was not necessary according to Dutch regulations at the time of data collection.
