Health and diet inequalities are closely related to socio-economic status(Reference Darmon and Drewnowski1). In Westernised countries, greater prevalence of non-communicable diseases has been repeatedly observed among socio-economically disadvantaged populations, with nutritional factors being important risk factors(Reference Petrovic, de Mestral and Bochud2). Studies investigating the links between dietary factors and socio-economic groups are plentiful among children(Reference Zarnowiecki, Dollman and Parletta3), but more sparse among adults. These have generally focused on dietary quality, by concentrating either on the intake of unhealthy or healthy food groups and nutrients(Reference López-Azpiazu, Sánchez-Villegas and Johansson4,Reference Méjean, Si Hassen and Lecossais5) or on dietary patterns, using e.g. a posteriori defined healthy dietary patterns or a priori scores reflecting adherence to official dietary recommendations(Reference Mullie, Clarys and Hulens6,Reference Lassale, Galan and Castetbon7) .
However, over the past decade, numerous studies have underlined the environmental damage caused by our current food systems, urging for a transition to healthy and environmentally sustainable dietary patterns(Reference Burlingame8–Reference Willett, Rockström and Loken10). In that context, the definition of sustainable diets stated by the FAO goes beyond the health and nutritional values of diets and also encompasses environmental, socio-economic and cultural components(Reference Burlingame8). Many studies have showed co-benefits of plant-based dietary patterns on planetary and human health(Reference Tilman and Clark9–Reference Laine, Huybrechts and Gunter12), since plant-based diets are generally both healthy and have fewer impacts on the environment than animal-based diets. On the other hand, other studies point out potential conflicts between the dimensions of food sustainability; the different dimensions do not necessarily align with each other(Reference Clark, Springmann and Hill13–Reference Perignon, Vieux and Soler15). For instance, in a study by Clark et al. (Reference Clark, Springmann and Hill13), sugar-sweetened beverages had among the lowest environmental impacts for the studied indicators despite being associated with higher risk of diseases, and the opposite was true for fish.
While lower education status linked to unhealthy diets, diets of disadvantaged individuals might not be systematically more impactful on the environment. For instance, in the Third French Individual and National Food Consumption (INCA 3) survey (2014–2015), French individuals with higher education level consumed more fruits and vegetables (which have generally low environmental footprints), but also more cheese, dairy products (which have high environmental footprint) or chocolate. In contrast, individuals with a lower educational level consumed more products with low environmental impacts such as soda and potatoes but also ate more high-impact products such as meat (excluding poultry)(16).
Different factors measuring socio-economic position (i.e. occupational position, household income or education attainment) may be differentially associated with sustainable diets, as they affect food choices in different ways(Reference Galobardes, Shaw and Lawlor17). More highly educated individuals are generally more aware of the recommended diets(Reference De Vriendt, Matthys and Verbeke18) as well as potentially of the impacts of food on the environment. They generally exhibit higher food and health literacy(Reference Galobardes, Shaw and Lawlor17), i.e. the ‘capacity to obtain, process and understand basic health information and services needed to make appropriate health decisions’.
Household income influences a person’s ability to afford a healthy and possibly a more sustainable diet. For example, the price of organic products, with their potential health and environmental benefits(Reference Kesse-Guyot, Lairon and Allès19), is often seen as a barrier for low-income households. Occupation type could also play a role in food intakes depending on the culture and the environment of the workplace. Yet limited evidence is available on the associations between overall diet sustainability, using composite indexes and socio-economic status(Reference Eini-Zinab, Shoaibinobarian and Ranjbar20).
Hence, the present study aimed to examine how socio-economic status, measured as income, education and occupational status, relates to sustainable dietary patterns, using an index accounting for various dimensions of diet sustainability, in a large cohort of French adults. We first investigated the association of socio-economic status with overall diet sustainability. A second objective was to more specifically examine the relation between socio-economic position and the different components of diet sustainability.
Methods
Study population
We used data from the NutriNet-Santé study. The NutriNet-Santé study is an ongoing web-based prospective cohort study launched in 2009 in France, established to investigate the associations between nutrition and health as well as determinants of dietary behaviours. Participants are Internet-using adult volunteers recruited from the general population. At recruitment and during follow-up, lifestyle, medical history and socio-economic factors are collected via online questionnaires available on a web platform (https://etude-nutrinet-sante.fr/). Additional information is also collected regarding dietary practices and behavioural issues related to nutrition and health. The study protocol and the procedures have been fully described previously(Reference Hercberg, Castetbon and Czernichow21).
Assessment of socio-demographic variables and socio-economic position
Socio-demographic characteristics, including sex, age, marital status, presence of children in the household and residential area, were assessed using self-administered questionnaires(Reference Vergnaud, Touvier and Méjean22).
Socio-economic position included monthly income per household unit, highest educational attainment and current or last (in case of retirement or unemployment) occupational category. More specifically, in the self-administered questionnaires, participants completed their total income per month assessed from various sources (salary, rental income, family allowance or social benefits). Monthly household income was then defined by household unit, according to the definition of the National Institute of Statistics and Economic Studies (INSEE)(23), i.e. 1 household unit was allocated to the first adult in the household, 0·5 to other individuals aged ≥14 years and 0·3 to children aged below 14 years. Household income was classified into five groups (<1200 €/month; 1200–1800 €/month; 1800–2700 €/month; >2700 €/month and refuse to declare). Education level was categorised as primary level education, secondary level education, undergraduate (up to 3 years after high school diploma) and postgraduate (>3 years after high-school diploma). We created a six-category variable for occupational status based on INSEE categories’ definition: self-employed (craftsman, shopkeeper, entrepreneur and farmer); managerial staff/intellectual profession; intermediate profession; employee; manual worker and never-employed (homemaker, student and disabled)(23).
Individual data were extracted from the questionnaire completed on the date closest to the date of completion of the FFQ described below.
Organic FFQ
Dietary information was collected using a 264-item self-administered FFQ (Org-FFQ) designed to evaluate individual’s habitual diet over the preceding year. The Org-FFQ has been described elsewhere(Reference Baudry, Méjean and Allès24). In brief, the Org-FFQ was developed within the frame of the BioNutriNet project(Reference Kesse-Guyot, Lairon and Allès19) and administered in 2014. It is based on an existing previously validated semi-quantitative FFQ(Reference Kesse-Guyot, Castetbon and Touvier25). Additional questions regarding organic food consumption frequency were included. More specifically, for each item, participants were asked to provide the quantity consumed and the frequency of consumption (per day, week, month or per year). Quantity of food consumed was estimated using standard portion sizes (e.g. a slice of bread) or coloured photographs displayed to estimate the quantity consumed among seven portion sizes. Participants were also asked to provide their organic food intake frequency for each food and beverage item available as organic option, through a 5-point scale, ranging from ‘never’ to ‘always’ (intermediate modalities included ‘rarely, ‘half-of-the-time’ and ‘often’). Total food intake was obtained by multiplying the frequency of consumption by the amount of food consumed. Organic food consumption for each item was estimated by assigning the following proportions 0, 0·25, 0·5, 0·75 and 1 to the corresponding frequency modalities. The share of organic food in the diet was then obtained by dividing the total organic food intake (g/d) out of the total intake excluding water (g/d).
The sustainable diet index
We used the Sustainable Diet Index (SDI) whose construction and validation have been extensively described elsewhere(Reference Seconda, Baudry and Pointereau26). In short, the original SDI includes seven indicators categorised into four equally weighted subscores, based on the FAO’s definition of sustainable diets(Reference Burlingame8) which encompasses four pillars (nutrition, environment, economy and socio-cultural aspects).
The nutritional component is composed of an indicator reflecting the gap between energy needs (estimated using Schofield equation(Reference Schofield27)) and energy intake, and of the PANDiet (Diet Quality Index Based on the Probability of Adequate Nutrient Intake) index, which reflects adherence to French nutritional references(Reference De Gavelle, Huneau and Mariotti28). The environmental component includes the pReCiPe (partial ReCiPe)(Reference Kramer, Tyszler and Veer29) (a synthetic environmental score, including greenhouse gas emissions (GHGe), cumulative energy demand and land occupation, weighted by means of coefficients) and organic food consumption as a proxy for greater biodiversity(Reference Reganold and Wachter30). The socio-cultural component includes an indicator related to the type of food supply and a score related to consumption frequency of ready-made foods. In the original SDI, the economic component is composed of an indicator defined as the proportion of income devoted to food. In the present work, we elected to not include affordability (i.e. proportion of the income devoted to diet) since socio-economic markers would be highly correlated to it. Therefore, we computed a modified Sustainable Diet Index (mSDI), by removing the economic component.
A detailed description pertaining to the assessment of the different indicators included in the original SDI is given in the Supplemental material.
For indicators presumed to be ‘favourable’ for sustainability, 1 point was assigned to participants in the lowest quintile, 2 points to those in the second quintile and so on. The scoring was inverted for indicators considered as ‘detrimental’. The seven indicators were weighted so that each component had the same weight and had a range comprised between 1 and 5. The final score was obtained by summing up the points for each subscore and ranged from 4 (low sustainability) to 20 (high sustainability). Quintile values retrieved during the development of the score were used(Reference Seconda, Baudry and Pointereau26). The original scoring is presented in Supplemental Table 1.
The modified SDI ranged (without the affordability component) from 3 to 15.
Selection of the study sample and weighting procedure
Of the 37 685 NutriNet-Santé participants who responded to the Org-FFQ from June to December 2014, 29 119 were included in the analyses. Individuals with missing covariates, those with an implausible ratio of energy intake to energy requirement estimated by Schofield equations(Reference Schofield27) (using predefined cut-offs(Reference Baudry, Méjean and Allès24)), those not living in mainland France, those with unavailable data on food source supply and for the computation of the SDI were excluded from the analysis. Additionally, we also excluded unemployed and retired participants whose last occupational status could not be identified (see online Supplemental Fig. 1).
We applied a weighting procedure to all analyses in order to better fit the proportions of socio-demographic categories existing in the French population. To do so, using an iterative proportional fitting procedure, a weighting was performed separately for each sex, according to age, occupational status, education level, residential area, presence of children in the household (<18 years) and marital status, attributing a weight to each individual (SAS macro %CALMAR and the 2009 national census INSEE data).
Statistical analysis
General characteristics of the study participants are presented as mean (sd) or percentage according to quintiles of the mSDI.
Collinearity between the three socio-economic indicators was examined using the variance inflation factor. Variance inflation factor was below 4 in all regression models, suggesting no substantial collinearity and the three indicators were included simultaneously in the models.
Q–Q plots and histograms were used to visualise whether the residuals were normally distributed for all unweighted regression models conducted.
Associations of general characteristics across weighted mSDI quintiles were tested with linear contrast tests for continuous variables, Mantel–Haenszel chi-square trend tests or chi-square tests for ordinal and categorical variables, respectively.
Mean contribution of each food group to total consumption (g/d), cost (€/d) and energy intake (kcal/d) across weighted quintiles was calculated. Average diet monetary cost (€/d) distinguishing the farming system of food consumed, according to mSDI quintiles, was also computed.
For the main analysis, using the mSDI as dependent variable, normality assumptions were fulfilled and therefore linear models were fitted to estimate the effect estimates (β) and 95 % CI for the association of the three individual socio-economic indicators and the mSDI (modelled as a continuous variable), in the full sample and by sex. Models were adjusted for (sex), age, total energy intake, parental status, residential area and the two remaining socio-economic factors.
Due to their distribution, the individual subscores (namely environment, nutritional and socio-cultural subscores) were treated as ‘count-like’ discrete variables and Poisson regressions were constructed to assess the associations between socio-economic indicators and individual subscores. In doing so, we sought to assess whether any of the socio-economic factors were most strongly related to any of the individual subscores.
Statistical significance was set at P−value < 0·05. Data treatment and statistical analyses were performed using SAS (version 9.4, SAS Institute Inc.) and R (version 3.6.2).
Results
Sample description
Table 1 displays the characteristics of the study sample across weighted mSDI quintiles. Individuals in Q5 (with a more sustainable diet), compared to individuals in Q1, were more often females, aged 65 years or older and less likely to have a child at home. No association was found between residential area and the mSDI. The proportion of individuals with primary education level was highest in Q2 while the proportion of postgraduate participants increased with increasing mSDI. The highest percentage of low-income individuals was found in Q2 whereas participants unwilling to provide their income were the most numerous in Q5. Q1 and Q2 showed the highest proportion of manual workers and Q5 the highest of intermediate profession.
Characteristics of participants according to mSDI quintiles (weighted data)*
mSDI, modified sustainable diet index.
* Values are means (sd) or percent, as appropriate. All P-values < 0·0001, except residential area (P-value = 0·84). P-values refer to P-trend estimated using linear contrast tests for continuous variables, Mantel–Haenszel chi-square trend tests and chi-square tests for ordinal and categorical variables, respectively.
Diet-related characteristics across mSDI quintiles
Daily food intake, diet monetary cost and energy intake, according to mSDI quintiles are shown in Fig. 1. Total food intake ranged from 3151 (1639) g/d (Q1) to 3600 (844) g/d (Q5). Diet monetary cost ranged from 7·43 (4·14) €/d in Q1 to 8·10 (2·38) €/d in Q5. Meat was the most important contributor to total diet monetary cost in Q1, whereas intake of fruit and vegetables was the most important contributor in Q5. Total energy intake decreased along with higher mSDI. Participants in Q5 showed a higher contribution to energy intake from fruit and vegetables than those in Q1 (approximately 20 % for Q5 v. < 10 % for Q1) and a threefold lower contribution to energy intake from meat. Unsurprisingly, diet monetary cost dedicated to organic food increased with increasing mSDI (see online Supplemental Fig. 2).
Average food intake (g/d), diet monetary cost (€/d) and energy intake (kcal/d), according to mSDI quintiles (weighted data)* Abbreviation: mSDI, modified Sustainable Diet Index. All P-values for linear contrast < 0·0001. *Data are unadjusted.
Associations between socio-economic factors and mSDI
Figure 2 presents the association between the three measures of socio-economic factors and the mSDI. Individuals with monthly income comprised between 1800 and 2700 € and those with an income below <1200 € did not have significantly different mSDI compared to high-income individuals (>2700 €/month) (whole sample). The mSDI for individuals with intermediate income (1200–1800 €/month) was lower than for high-income individuals. Some differences were observed across sex, in particular, low-income males showed mSDI higher than the reference (high-income individuals) while the opposite held for females. Regarding education, lower educational level was overall associated with lower mSDI (whole sample and both sexes). Of note, no difference was detected when comparing low-educated to postgraduate females. Employees and manual workers (full sample, males and females) had a lower mSDI score than intermediate professionals. Females who never worked had a mSDI score lower than intermediate professionals whereas the opposite was observed for males. Compared to the reference, managerial staff did not have different mSDI scores.
Association between socio-economic factors and mSDI, linear regression coefficients and 95 % CI (weighted data)* Abbreviation: mSDI, modified Sustainable Diet Index. *Multivariable linear regression for the association between the mSDI and each socio-economic factor. Multivariable-adjusted models include (sex), age, total energy intake, parental status, residential area and socio-economic factors. The socio-economic factors are mutually adjusted for the remaining factors.
The environmental subscore was higher for individuals with income below 2700 €/month than for the reference category (>2700 €/month) while it was lower for individuals with low-educational levels compared to postgraduate individuals (Table 2). Employees, manual workers and managerial staff had an environment subscore that was lower than intermediate professionals. Individuals with low-income or low-education level had lower socio-cultural subscore compared to those with high income or postgraduate. Never-employed, manual workers and employees had a lower socio-cultural subscore than intermediate professionals while this subscore was higher for self-employed individuals. Only individuals with intermediate income (1800–2700 €/month) had a significant lower nutritional subscore compared to high-income levels. Individuals with low educational level (primary and secondary) had lower nutritional subscore in comparison to postgraduate participants. Self-employed and employees had a lower nutrition subscore than participants belonging to the intermediate profession category while no significant associations were observed for other occupational categories. Supplemental Table 2 shows the relationships between socio-economic factors and individual subscores according to sex.
Association between socio-economic factors and individual subscores, Poisson regression coefficients and 95 % CI (weighted data)*
* Multivariable Poisson regression for the association between the individual component and each socio-economic factor. Multivariable-adjusted models include sex, age, total energy intake, parental status, residential area and socio-economic factors. The socio-economic factors are mutually adjusted for the remaining factors.
Bold denotes significance.
Discussion
Summary of the findings
We assessed the relationships between socio-economic status using three indicators and sustainable dietary patterns expressed by the mSDI (the original SDI without the economic dimension). By construction, a more sustainable diet was associated with lower energy intake, lower intakes of animal-sourced foods and higher intakes of fruit and vegetables. A higher mSDI was also linked to a slightly higher monetary diet cost in the crude analysis. After adjusting for multiple factors including total energy intake, we observed a mSDI gradient across educational level. Manual workers and employees had also lower mSDI than intermediate professionals.
Comparison with studies on overall socio-economic status in relation to diet sustainability as a whole
To our knowledge, studies examining the association between various socio-economic factors and sustainable diets are scarce, and a comparison with those of the literature proves difficult.
However, certain of our findings can be compared in light of socio-economic determinants of dietary patterns since plant-based diets are generally nutritionally and environmentally better(Reference Tilman and Clark9–Reference Laine, Huybrechts and Gunter12). Socio-economic factors associated with dietary patterns have indeed been extensively described(Reference Darmon and Drewnowski1,Reference Mullie, Clarys and Hulens6) .
A study has been recently conducted in a representative sample of Iranian households with a comparable research question as ours(Reference Eini-Zinab, Shoaibinobarian and Ranjbar20), i.e. examining socio-economic status, using a composite indicator, in relation to overall diet sustainability (assessed based on dietary carbon footprint, dietary water footprint, dietary costs and dietary quality index). Interestingly, the most privileged households had a less sustainable diet, due to their high energy and food intake, especially of animal-based foods, in contrast to our findings. A hypothesis advanced by the authors is that in a middle-income country such as Iran, urbanised and affluent populations might be experiencing the nutrition transition(Reference Popkin31) faster than the rest of the population. Apart from the different country context, dietary data were based on household expenses and socio-economic status was assessed as a whole by deriving, through principal component analysis, an aggregated socio-economic index including different socio-economic indicators, limiting the comparison with our work.
Comparison with studies on education
We observed a graded association between education level and a more sustainable diet. In particular, individuals with low-education had a lower nutritional subscore. While limited data is available regarding sustainable diets, the latter observation is consistent with studies reporting that a higher level of educational attainment is associated with a better adherence to different diet quality indices(Reference Mullie, Clarys and Hulens6,Reference Lassale, Galan and Castetbon7,Reference Backholer, Spencer and Gearon32) . Similarly, in the French nationally representative ESTEBAN survey, individuals with higher levels of education reported a diet more in line with the recommendations(33). In particular, higher intake of fruit and vegetables and wholegrain products have been consistently associated with higher socio-economic status(Reference Darmon and Drewnowski1,34) .
Our findings are therefore in line with previous works indicating that individuals with highest education levels are generally better informed about nutritional(Reference De Vriendt, Matthys and Verbeke18) and health recommendations(Reference Galobardes, Shaw and Lawlor17) and more concerned about environmental issues(Reference Franzen and Meyer35), and in turn might be more likely to adopt healthy(Reference Hercberg, Castetbon and Czernichow21) and pro-environmental eating behaviours. In studies conducted within the NutriNet-Santé cohort, we observed that highly educated organic food consumers were also those the most concerned by ethics and environmental issues and those the most able to recognise official organic labels(Reference Baudry, Touvier and Allès36,Reference Baudry, Péneau and Allès37) .
Comparison with studies on income
Concerning income, total energy intake, animal-based food demand and a gross domestic product are positively correlated; high-income countries generally displaying a higher consumption of meat and processed food(Reference Clark, Hill and Tilman38). Conversely, in developed nations, within countries, unhealthy diets characterised by energy-dense and nutrient poor foods prevail in low-income category populations(Reference Friel, Hattersley and Ford39). Interestingly, in our study, individuals belonging to the lowest income categories did not systematically show a less sustainable diet. It was even the opposite for low-income (<1200 €/month) males in comparison to the highest income group, in spite of the adjustment for age. The strongest negative association with income was nevertheless observed among the potentially vulnerable group of low-income females. Participants with an intermediate income (1200–1800 €/month) had a lower mSDI than their more privileged counterparts. It would appear that it is possible to have a relatively low income and a sustainable diet, indicating that household income should not be too low but not necessarily very high. Of note, our models were multi-adjusted, including for remaining socio-economic factors (i.e. occupational status and education level), which may have attenuated the associations with income. In general, education level appears to be the strongest socio-economic predictor of diet quality(Reference Galobardes, Shaw and Lawlor17). For example, in a study conducted in the NutriNet-Santé cohort(Reference Méjean, Si Hassen and Lecossais5), education level modulated the relationships between income and animal food consumption and appeared to be the strongest predictor of animal food consumption. A study conducted in the United Kingdom based on household purchase data is also worth mentioning(Reference Wrieden, Halligan and Goffe40). The authors reported that sustainable diets (defined as GHGe, cost and land use being below the median and a Diet Quality Index being above the median) were more prevalent in households belonging to the lowest income quintile than the rest of the sample (22 % v. 16·6 %).
With regard solely to diet quality in an Australian study, a positive relationship was found between income and diet quality assessed using the Dietary Guideline Index(Reference Backholer, Spencer and Gearon32) for both males and females while no evident gradient was observed within our study. A possible explanation for this discrepancy between the two studies could be, among other factors, the use of different scores to assess diet quality. We use the PANDiet which is a score reflecting adherence to the French nutrient-based recommendations while the Dietary Guideline Index reflects adherence to food-based recommendations. The general public may be less aware of reference values of vitamins and minerals used for the computation of the PANDiet.
With regard to environment, in our study, low-income individuals had higher scores (i.e. lower environmental impacts) than those in the highest income category. A study conducted in Lima also reported higher income associated with higher GHGe(Reference Vázquez-Rowe, Larrea-Gallegos and Villanueva-Rey41). A study conducted in thirty countries indicated that high-income individuals are more concerned about environmental issues(Reference Franzen and Meyer35) than their counterparts. However, in a typology derived from conventional and organic food intake in the NutriNet-Santé study, we observed that individuals with the highest income were not those with the healthiest diets and greater organic food consumption(Reference Baudry, Touvier and Allès36).
Comparison with studies on occupational status
Regarding occupational status belonging to intellectual professions/managerial staff categories was not linked to a more sustainable diet when compared to intermediate professions. Nonetheless, employees, females who have never been employed and manual workers, had a less sustainable diet.
In a study conducted in the NutriNet-Santé cohort(Reference Méjean, Si Hassen and Lecossais5), in respect of animal food intake, an association was observed between occupation and dairy product consumption. Manual workers ate more cream desserts than managerial staff but were less often yoghurt consumers, suggesting that a disparity may exist in the choice of healthy v. unhealthy products rather than on the type of food in the association. In the present work, manual workers and employees had lower diet sustainability, even after adjustment for other socio-economic factors, suggesting a role of occupation beyond merely education and income.
Self-employed (a group which includes varied profiles, including business managers and farmers) had a more sustainable diet than intermediate professions. Since this category gathers highly heterogeneous groups, conclusions can be hardly drawn and any interpretation must be made with cautious.
Socio-cultural aspects
Individuals with lower socio-economic status had a lower socio-cultural subscore, which included the main food source supply and the consumption frequency of ready-made products.
This seems in accordance with data from the French INCA 3 survey pertaining to behavioural purchase. Indeed, higher level of education (more than 3 years after high school)(16) was related to more frequent purchase in local markets, shops and short commercialisation channels(16).
In a study also conducted within the NutriNet-Santé cohort on >60 000 individuals, belonging to the lowest socio-economic groups (i.e. those belonging to the lowest education, income groups and to the category of female manual and office workers) was associated with a longer time spent on the meal preparation(Reference Méjean, Si Hassen and Gojard42), in accordance with another work(Reference Plessz and Gojard43) showing that commitment to meal preparation is generally higher among low-class individuals. However, in the same study(Reference Méjean, Si Hassen and Gojard42), it was observed that, compared to managerial staff, female manual and office workers used fewer raw materials or fresh produce to prepare their meals(Reference Méjean, Si Hassen and Gojard42), which seems rather consistent with our results.
Study limitations and strengths
Some limitations of the present work should be stressed. Firstly, dietary data were self-reported and thus subject to measurement errors. In addition, study participants are adult volunteers enrolled in a cohort focusing on nutrition and health and therefore more likely overall to be prone to exhibit heathier dietary habits than the general population(Reference Andreeva, Deschamps and Salanave44). The NutriNet-Santé cohort study thus comprises more females and more educated people than the French population. Some population segments, such as highly disadvantaged males are therefore not represented in the cohort. However, to limit these potential biases, we performed a weighting procedure so as to improve the representativeness of the sample compared to the French population. Furthermore, the score was built so that each subcomponent had the same weight, as in the FAO definition. Nonetheless, some subcomponents were composed of very detailed indicators while other indicators were simpler. It would therefore be interesting in the future to include other indicators, such as fair trade, or fine-tune some others, such as the consumption of ready-made products, in order to better take into account the potential conflicts between the different components of sustainability.
A strength of the study is its large sample size, which enabled us to access a wide range and contrasted eating habits and socio-economic profiles and to conduct stratified analyses. In addition, to assess diet sustainability, we used an index based on a validated score(Reference Seconda, Baudry and Pointereau26), which has been associated with long-term health outcomes(Reference Seconda, Baudry and Allès45) and encompassing multiple indicators, including three environmental indicators.
Final considerations
In the present work, educational level appears to be a strong driver of sustainable diets, independent of household income or occupation. Manual workers and employees appeared also to be ‘at-risk’ individuals. In their recent study exploring many drivers of a new dietary transitions for twenty-five countries across the world, Pais et al.(Reference Pais, Marques and Fuinhas46) reported that education along with price changes could promote dietary transition towards a more sustainable and healthy future.
It is important to note that apart from a reduction of animal-based food consumption, not overconsuming is a major lever for reducing diet-related GHGe(Reference Perignon, Vieux and Soler15). We observed herein that individuals with a more sustainable diet had a slightly greater diet monetary cost, due most likely to their high proportion of organic food. The cost of organic food was, however, somewhat offset by a lower energy intake and lower intake of animal-based foods, leading to only a slight increase of diet monetary cost. Lower diet quality may be explained in part by food cost(Reference Darmon and Drewnowski47), high energy-dense food being generally cheaper than healthy products. Cost of healthy and environmentally friendly diets might be higher than unhealthy diets. For instance, higher adherence to food-based dietary guidelines has been associated with greater diet monetary cost(Reference Vandevijvere, Seck and Pedroni48).
A shift towards sustainable eating patterns is urgent and requires systemic changes, including fiscal policies, appropriate governance as well as public awareness raising(Reference Garnett, Mathewson and Angelides49). The latter alone (information and awareness) might be indeed insufficient, as the transition cannot be left only to the individual, multi-level actions are therefore needed for the transition towards sustainable diets(Reference Garnett, Mathewson and Angelides49). Constraints and representations of sustainable practices of low social class groups may differ from those of ‘dominant’ groups(Reference Brocard, Saujot and Brimont50), and therefore require targeted strategies. For low social class individuals, sustainable strategies such as, for example seasonal product consumption, waste avoidance or frugal consumption may be preferentially chosen rather than consumption of organic vegetarian products(Reference Brocard, Saujot and Brimont50). Finally, interventions aiming to foster pro-health or pro-environmental behaviours would need careful attention as their effects can be sometimes divergent (e.g. sugary foods)(Reference Garnett, Mathewson and Angelides49).
Conclusions
To conclude, while diet, in its nutritional quality dimension, was already a well-documented marker of social inequality, our findings suggest that lower social group individuals, especially manual workers and low educated individuals, have also a less sustainable diet, leading us to advocate for the development and deployment of large nutritional literacy-adapted education programmes combined with appropriate food policy measures, to ensure healthy and environmentally sustainable diets at affordable cost for all. These results go further and illustrate clear associations between socio-economic position and diet sustainability.
Acknowledgements
Acknowledgements: We thank Thi Hong Van Duong, Régis Gatibelza, Jagatjit Mohinder and Aladi Timera (computer scientists); Fabien Szabo de Edelenyi, Ph.D. (data management supervisor); Julien Allegre, Nathalie Arnault, Laurent Bourhis, Nicolas Dechamp (data-manager/statisticians); Maria Gomes, Mirette Foham (Nutrinautes support), Cédric Agaesse, Rébecca Lutchia, Alexandre De Sa (dietitians); Younes Esseddik (IT supervisor) and Nathalie Druesne-Pecollo, Ph.D. (operational coordinator) for their technical contribution to the NutriNet-Santé study. We warmly thank all the volunteers of the NutriNet-Santé cohort. Authorship: J.B., B.L., S.H., D.L., P.P. and E.K.G. conducted the research. J.B., B.A., B.L., M.T., S.H., D.L., P.P. and E.K.G. contributed to the study design and led the underlying process of data acquisition. B.A., A.R., J.B.R., M.T., S.H., D.L., S.P. and E.K.G. contributed to the interpretation of data. J.B. performed statistical analyses and drafted the manuscript. All authors critically revised the manuscript and approved the final version to be published. J.B. had full access to all data and takes responsibility for the integrity of the data and the accuracy of the analyses. Ethics of human subject participation: The NutriNet-Santé study is carried out in accordance with the Declaration of Helsinki principles and was approved by the Institutional Review Board of the French Institute for Health and Medical Research (IRB INSERM 0000388FWA00005831) and the National Committee for Information Technology and Freedom (CNIL 908,450/909,216). Participants gave electronic informed consent. The NutriNet-Santé study protocol is registered in ClinicalTrials.gov as NCT03335644.
Financial support:
The NutriNet-Santé study is funded by the French Ministry of Health and Social Affairs, Santé Publique France, the French National Institute of Health and Medical Research, the National Research Institute for Agriculture, Food and the Environment, the French National Conservatory of Arts and Crafts and Sorbonne Paris Nord University. The BioNutriNet project was supported by the French National Research Agency in the context of the 2013 Programme de Recherche Systèmes Alimentaires Durables (ANR-13-ALID-0001). The funders had no role in the study design, data collection, analysis, interpretation of data, preparation of the manuscript and decision to submit the paper.
Conflicts of interest:
There are no conflicts of interest.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980022002208
Open access
