To provide an adequate diet to the growing world population, estimates indicate that an increase in the global food production is needed at a rate of 1·2 % per year( Reference Tilman, Balzer and Hill 1 ). At the same time, the food production system is recognised as a major threat to the environment, including climate change and depletion of the planet’s natural resources( Reference Vermeulen, Campbell and Ingram 2 ). This is partly driven by habitual consumption patterns tending towards a higher consumption of animal-based products( Reference Alexandratos and Bruinsma 3 ). It is thus an important global challenge to secure adequate diets within a sustainable food production system( Reference Godfray, Beddington and Crute 4 ). In this regard, an adequate diet implies that it meets energy requirements and provides sufficient nutrients in line with the dietary guidelines for healthy growth and ageing( Reference Maynard 5 ). Because diet is an important modifiable factor for well-being and disease prevention( 6 ), both the adequacy of nutrient intake and the observed or projected prevalence and/or occurrence of health/disease outcomes are of importance.
Shifting towards a more sustainable food consumption pattern is considered an important factor to tackle the challenge of harmonising the rapidly changing food demand for the larger and more affluent population and its supply( Reference Garnett 7 ). A recently published review suggested that a reduction of up to 50 % in diet-related greenhouse gas emissions and land use can be realised by dietary changes in areas with affluent diet( Reference Hallström, Carlsson-Kanyama and Börjesson 8 ). Especially the reduction of animal-based products is often regarded as the main option for lowering diet-related environmental impact( Reference Vermeulen, Campbell and Ingram 2 , Reference Garnett 7 , Reference Hallström, Carlsson-Kanyama and Börjesson 8 ). However, severe reductions without the inclusion of appropriate meat and/or dairy substitutes might lead to inadequacies of several nutrients (e.g. vitamin B12, Zn, Fe) across population groups( Reference Craig and Mangels 9 ). Therefore, the concept of a sustainable diet, as defined by the FAO, is briefly described as a diet that has a low impact on the planet’s resources and the environment, including respectfulness for biodiversity and animal welfare, and contributes to an adequate diet that is promoting a healthy life. Sustainable diets also feature characteristics such as cultural acceptability, accessibility, economic fairness and affordability( Reference Burlingame and Dernini 10 ). This definition highlights the connection between the health, the environmental sustainability and the food production aspects of a diet, with the dietary pattern of consumers as a common denominator. The design of those diets asks for a collaboration between nutritional and environmental sciences along with the agricultural food chain( 11 ).
The aim of the present review is to categorise and summarise the different approaches that are currently used to operationalise the health aspects of environmentally sustainable diets. Also, the relevance of these approaches for research on environmentally sustainable diets is discussed; each approach addresses a particular research question, but is built upon some assumptions that should be taken into account when using the approach. The review provides an overview of the way in which such diets have been addressed in research, particularly the relationship between health and environmental sustainability of a diet. On the basis of this overview, recommendations for future research on designing sustainable diets are given and discussed.
Methods
The literature search was performed in October 2015 and identified relevant articles through conventional keyword searching strategies, using the search terms ‘diet’ or ‘food’ and ‘climate’ or ‘greenhouse gas’ or ‘land’ or ‘sustain’, in PubMed, Scopus, Web of Knowledge and CAB Abstracts, and through bibliographies of published papers. Articles included in the review met the following five inclusion criteria: (i) English-language publication; (ii) published between 2005 and October 2015; (iii) dietary data collected for the diet as a whole at the national, household or individual level; (iv) comparison of the current diet with dietary scenarios; and (v) for results to consider the health aspect in some way. The selection of articles that met the inclusion criteria was based on information available in titles and abstracts of the articles, without restrictions on the geographical location. Given the aim of the review to categorise and summarise the different methodological approaches, some articles that inadvertently may have been missed were not expected to influence the results of the approaches identified.
Results
In the period 2005–2015, we identified forty-nine papers that studied diet as related to health and environmental sustainability.
Dietary data collected for the diet as a whole included food availability estimates at the population and household level, and actual food intake at the individual level. The food availability estimates included data on the food supply at the population level using Food Balance Sheets of the FAO or from the US Department of Agriculture, Economic Research Service( Reference Eshel and Martin 12 – Reference Jalava, Kummu and Porkka 27 ) and data on the food purchases at the household level using Household Budget Surveys( Reference Sáez-Almendros, Obrador and Bach-Faig 21 , Reference Collins and Fairchild 28 – Reference Briggs, Kehlbacher and Tiffin 32 ). Regarding individual-level food intake assessments, diet records were the most frequently used dietary survey method( Reference Capone, Iannetta and Bilali 20 , Reference Berners-Lee, Hoolohan and Cammack 33 – Reference Friel, Dangour and Garnett 49 ) with recording ranging from 2 to 14 d; followed by a single or replicated 24 h recalls( Reference Friel, Dangour and Garnett 49 – Reference Wilson, Nghiem and Mhurchu 56 ) and FFQ( Reference Monsivais, Scarborough and Lloyd 57 – Reference Biesbroek, Bueno-de-Mesquita and Peeters 60 ). The number of food items in these dietary assessments generally ranged from twenty-five to 100 in Food Balance Sheets, and from 130 in FFQ to 1314 in diet records or 24 h recalls. However, sustainability indicators (e.g. greenhouse gas emissions, land use) were available only for a limited number of foods, meaning a higher food aggregation level had been used. This food aggregation level was specified in forty-five studies, of which only seventeen studies applied a more precise level of aggregation into food items, with the number of food items ranging between seven and 391 food items( Reference Eshel and Martin 12 , Reference Baroni, Cenci and Tettamanti 13 , Reference Gerbens-Leenes and Nonhebel 16 , Reference Vieux, Darmon and Touazi 35 – Reference Van Dooren and Aiking 41 , Reference Arnoult, Jones and Tranter 43 , Reference Thompson, Gower and Darmon 45 , Reference Tyszler, Kramer and Blonk 52 – Reference Temme, Toxopeus and Kramer 54 , Reference Wilson, Nghiem and Mhurchu 56 , Reference Biesbroek, Bueno-de-Mesquita and Peeters 60 ). For two studies, it was specified that this covered 71 % of the total food weight intake (including all solid foods and excluding foods typically consumed as beverages, such as milk, juices and other drinks) and 66 % of total energy intake from all foods and beverages( Reference Vieux, Soler and Touazi 37 , Reference Masset, Vieux and Verger 38 ). In most studies, food items without a sustainability value were assigned a value from a similar food item within the same food group to cover the total food consumption. Sustainability was mainly operationalised by greenhouse gas emissions( Reference Eshel and Martin 12 , Reference Westhoek, Lesschen and Rood 15 , Reference Sáez-Almendros, Obrador and Bach-Faig 21 , Reference Heller and Keoleian 25 , Reference Heller and Keoleian 26 , Reference Friel, Barosh and Lawrence 29 – Reference Masset, Vieux and Verger 38 , Reference van Dooren, Marinussen and Blonk 40 – Reference Germani, Vitiello and Giusti 42 , Reference Macdiarmid, Kyle and Horgan 44 – Reference Milner, Green and Dangour 48 , Reference Carvalho, César and Fisberg 51 – Reference Monsivais, Scarborough and Lloyd 57 , Reference Biesbroek, Bueno-de-Mesquita and Peeters 60 ), followed by land use( Reference Stehfest, Bouwman and van Vuuren 14 – Reference Gerbens-Leenes and Nonhebel 16 , Reference van Dooren, Marinussen and Blonk 40 , Reference Van Dooren and Aiking 41 , Reference Arnoult, Jones and Tranter 43 , Reference Temme, van der Voet and Thissen 50 , Reference Tyszler, Kramer and Blonk 52 , Reference Biesbroek, Bueno-de-Mesquita and Peeters 60 ) and other sustainability indicators including livestock production, biodiversity and use of the planet’s resources( Reference Eshel and Martin 12 – Reference Westhoek, Lesschen and Rood 15 , Reference Buzby, Wells and Vocke 17 – Reference Pairotti, Cerutti and Martini 30 , Reference Roos, Karlsson and Witthoft 39 , Reference Germani, Vitiello and Giusti 42 , Reference Friel, Dangour and Garnett 49 , Reference Tyszler, Kramer and Blonk 52 , Reference Meier and Christen 58 , Reference Meier, Christen and Semler 59 ), which is partially biased towards the search terms used to define sustainability.
Approaches for operationalising the health aspect could be categorised into three main categories (Fig. 1 and Table 1): simple approaches focusing on a single nutritional aspect (A); approaches capturing the complexity of the diet (B); and approaches evaluating the health impact (C). More specifically, the simple approach refers to food item replacements. Three approaches were identified to capture the complexity of the diet: dietary guidelines (B1); dietary quality scores (B2); and diet modelling techniques (B3). For diet-related health impact, one approach was identified. Studies generally did not address policy options to achieve dietary changes, the time dimension for environmental effects to occur (except for direct greenhouse gas emissions) or the robustness of alternative dietary options in different socio-economic and ecological contexts.
Conceptual overview showing the approaches used to consider the health aspect in environmentally sustainable diets when using population-, household- or individual-level food intake assessment
Approaches used to consider the health aspect in research on environmentally sustainable diets
Simple approach: food item replacements (A)
Food item replacement is a ready-to-use and illustrative approach that addresses the question ‘What would be the change in environmental sustainability when replacing a particular food item or food group in the diet by a more environmentally sustainable alternative food item or food group?’ Ten studies used this approach and replacement of food items was food weight-based( Reference Temme, van der Voet and Thissen 50 ), protein-based( Reference Stehfest, Bouwman and van Vuuren 14 ) or energy-based( Reference Eshel and Martin 12 , Reference Baroni, Cenci and Tettamanti 13 , Reference Westhoek, Lesschen and Rood 15 , Reference Collins and Fairchild 28 , Reference Berners-Lee, Hoolohan and Cammack 33 – Reference Werner, Flysjo and Tholstrup 36 ) (Table 2). To develop a more environmentally sustainable diet, all studies focused on a replacement of the animal-based products in the diet, varying from a shift to a moderate reduction or a total elimination of these products. In some replacement diets, total meat consumption was kept constant, shifting the consumption from higher carbon-intensive meats (i.e. beef and lamb) to less carbon-intensive meats (i.e. pork and poultry)( Reference Eshel and Martin 12 , Reference Hoolohan, Berners-Lee and McKinstry-West 34 ). More commonly used replacement diets were those in which the total meat consumption was moderately reduced( Reference Stehfest, Bouwman and van Vuuren 14 , Reference Westhoek, Lesschen and Rood 15 , Reference Collins and Fairchild 28 , Reference Hoolohan, Berners-Lee and McKinstry-West 34 , Reference Vieux, Darmon and Touazi 35 , Reference Temme, van der Voet and Thissen 50 ) or completely eliminated( Reference Eshel and Martin 12 – Reference Stehfest, Bouwman and van Vuuren 14 , Reference Collins and Fairchild 28 , Reference Berners-Lee, Hoolohan and Cammack 33 , Reference Hoolohan, Berners-Lee and McKinstry-West 34 , Reference Werner, Flysjo and Tholstrup 36 , Reference Temme, van der Voet and Thissen 50 ); the former decreasing the meat intake by keeping the same types of meat in the diet and the latter being vegetarian or vegan options depending on their dairy content. In these replacement diets, meat (and dairy) substitutes can include either a single food group (e.g. dairy or fruit/vegetables, cereals, etc.)( Reference Westhoek, Lesschen and Rood 15 , Reference Berners-Lee, Hoolohan and Cammack 33 , Reference Vieux, Darmon and Touazi 35 , Reference Werner, Flysjo and Tholstrup 36 ) or a combination of different food groups (e.g. pasta, rice, pulses, cereals, breads, salads, fruit and vegetables, dairy, eggs, nuts and seeds, etc.)( Reference Eshel and Martin 12 – Reference Stehfest, Bouwman and van Vuuren 14 , Reference Collins and Fairchild 28 , Reference Berners-Lee, Hoolohan and Cammack 33 , Reference Hoolohan, Berners-Lee and McKinstry-West 34 , Reference Werner, Flysjo and Tholstrup 36 , Reference Temme, van der Voet and Thissen 50 ). However, simple replacement is seldom possible in practice, not only because physiological feedback loops interfere with the total amount of food eaten and/or energy intake; but also due to behavioural feedback loops that affect food choices, nutrient composition and/or energy density of the diet as a whole. Food item replacement is thus likely to modify the dietary pattern as a whole. For example, decreasing meat consumption and replacing it by plant-based substitutes might be beneficial for the environmental sustainability aspect of the diet, but raises concerns about the health aspect, in particular the intake of micronutrients that are largely derived from animal-based products (e.g. vitamin B12, vitamin D, Fe, Zn, Se). Also, from a consumer perspective, questions have been raised about the acceptability of replacing meat, because meat is usually an embedded food item in a consumer’s habitual dietary pattern. Nevertheless, nowadays, a substantial number of consumers belong to the segment of meat reducers or flexitarians, showing the feasibility of adopting a lower-level meat consumption( Reference de Bakker and Dagevos 61 ). In particular, potential change strategies incorporate the inclusion of meatless days with or without meat substitutes; the promotion of a smaller portion of meat; and, if possible, a combination of using sustainably produced (meat) products and/or a larger portion of plant-based products (i.e. fruits and vegetables)( Reference de Bakker and Dagevos 61 – Reference Verain, Dagevos and Antonides 63 ).
Food item replacement for the development of environmentally sustainable diets based on current diet
GHGE, greenhouse gas emissions.
* Food aggregation level: the number of food items or groups (depending on author’s terminology) for which environmental sustainability data of food intake was available.
† The theoretical diets were based on the current diet adjusted for the Danish Dietary Guidelines: six omnivorous diets with various quantities for dairy; one vegetarian diet with no cheese and meat products; and one vegan diet with no milk products, meat products and fish.
‡ Preferably plant-based meat substitutes that might reasonably be considered to be healthy alternatives, i.e. pasta, rice, pulses, cereals, breads, salads, vegetables, fruits, nuts and seeds.
§ Replacement with plant-based products that have a similar use to the reference food and therefore assumed to be consumed in similar amounts: liquid dairy foods were replaced by similar soya-based foods; meat products and cheese used as sandwich filling by a variety of other sandwich fillings/toppings; meat products in hot meals by a variety of meat replacers (e.g. vegetarian meat substitutes, egg dishes, pulses or tofu/tempeh); and soft cheese used as snack by popcorn.
|| The nutritional composition of each alternative diet was evaluated against the Nordic Nutrition Recommendations 2004 for macronutrients (protein, carbohydrate, added sugar, fat, saturated fat, mono- and polyunsaturated fat, and alcohol) and micronutrients (including dietary fibre, vitamins A, D, E, C, B12, B6, thiamin, riboflavin, niacin and folate, and minerals Mg, Fe, Zn, P, Ca, iodine and Se).
Apart from changing the dietary composition, just proportionally reducing food intake has been shown to lead to less energy while keeping the same overall nutrient density, as applied in one study( Reference Vieux, Darmon and Touazi 35 ). A shortage of energy is not a common problem in Western countries where overconsumption is contributing to overweight, obesity and related diseases( 64 ). However, adequate micronutrient intake is still a major challenge in these Western-oriented diets due to their non-optimal composition( Reference Mensink, Fletcher and Gurinovic 65 ) and micronutrient intake is often neglected in the nutritional evaluation of the ‘less meat’ diets.
Approaches capturing the complexity of the diet (B)
Dietary guidelines (B1)
Dietary guidelines are considered a descriptive approach that addresses the question ‘What would be the change in environmental sustainability when dietary guidelines are met?’ Seventeen studies used this approach to compare current diets with the recommendations for a healthy diet with regard to their health and environmental sustainability aspects (Table 3). Dietary recommendations initially provided dietary guidance with the aim to promote health and well-being, and to prevent diet-related conditions and chronic diseases( 6 ), without considering the environmental sustainability of these diets – until recently( 66 , Reference Reynolds, Buckley and Weinstein 67 ). The design of the recommended diet (e.g. the inclusion of food groups and the quantification of portion sizes) is highly dependent on the dietary guidelines used. However, when studying recommended diets in relation to environmental sustainability, the contribution of the following food groups was usually captured by the various recommended diets: bread, pasta, cereals and potatoes; fruit and vegetables; milk and milk products; meat, fish and egg products; legumes, nuts and seeds; fats and oils; and sugar, whereas alcohol was included only in the Mediterranean diets. Two studies additionally included the guidelines on total energy intake (and macronutrient composition)( Reference Gerbens-Leenes and Nonhebel 16 , Reference Germani, Vitiello and Giusti 42 ) and nine studies constructed multiple recommended diets standardised for energy intake (and protein intake)( Reference Tukker, Goldbohm and de Koning 18 , Reference Wolf, Pérez-Domínguez and Rueda-Cantuche 19 , Reference Saxe, Larsen and Mogensen 22 – Reference Heller and Keoleian 26 , Reference Meier and Christen 58 , Reference Meier, Christen and Semler 59 ); however, only one study focused on guidelines for total energy intake and macronutrient intake to design the recommended diet( Reference Jalava, Kummu and Porkka 27 ). None of these studies explicitly addressed the advice on lowering salt intake, while this, in turn, might have an impact on food production, processing and consumption, hence on environmental sustainability. This is because salt possesses certain crucial technological functions in food processing and preservation, and an important sensory function( Reference Hutton 68 ). Additionally, when using the approach of dietary recommendations, the food aggregation level was quantified at a high level of food aggregation (about twenty food groups) which allowed for a rough estimation of the environmental sustainability for a broader range of indicators, not only including greenhouse gas emissions but also the use of natural resources such as land, water, phosphorus and primary energy.
Dietary guidelines in relation to the environmental sustainability for a descriptive analysis on environmentally sustainable diets
USDA, US Department of Agriculture; ERS, Economic Research Service; USDHHS, US Department of Heath and Human Services; GHGE, greenhouse gas emissions.
* When using food-based dietary guidelines, the contribution of the following food groups was usually captured by the various recommended diets: bread, pasta, cereals and potatoes; fruit and vegetables; milk and milk products; meat and meat products, fish and eggs; legumes, nuts and seeds; fats and oils; and sugar; while alcohol was only included in the Mediterranean diets.
† Food aggregation level: the number of food groups, categories or commodities (depending on author’s terminology) for which environmental sustainability data of food intake was available.
‡ In addition, dietary scenarios such as a healthy diet with no meat and a healthy diet with less meat were investigated, in which the meat products were replaced by pulses and oil crops.
§ Additional food groups included in the recommended diet were the non-core foods; for example, snack foods, processed meats, sugar, tea, coffee and miscellaneous, alcohol, and saturated fats and oils. In addition, dietary scenarios such as the current diet with minimal non-core foods and the foundation recommended diet were also investigated. The former scenario contained similar foods and quantities as the current diet with minimal inclusion of energy-dense processed non-core foods, thus excluding processed meat, snack foods, confectionery, soft drinks, saturated fats and oils, and alcohol; and the latter was derived from the recommended diet consistent with Australian Dietary Guidelines, however including only core foods in similar amounts to the recommended diet, while meeting minimum nutrient and energy requirements for the population. All scenarios were evaluated on macro- and micronutrient intakes: energy, carbohydrate, protein, total and saturated fat, dietary fibre, vitamin A, folate, Ca, Mg, Zn and K.
|| The recommended diet was focused only on meeting the guidelines for the intake of fruits and vegetables, total and whole grains, and dairy.
¶ Two German dietary recommendations: D-A-C-H (official recommendation of the German Nutrition Society (DGE)) and UGB (alternative recommendations by the Federation for Independent Health Consultation with less meat, but more legumes and vegetables). The lacto-ovo-vegetarian dietary patterns adopted from USDA/USDHHS guidelines excluded the food groups on meat products and fish products, and included an additional food group for nuts and seeds and a separate food group for legumes. The vegan one additionally excluded the food groups on butter, high- and low-fat dairy products, and egg products, and included an additional food group for vegan soya drink products.
** The recommended diet has an energy intake of 8368 kJ/d (2000 kcal/d) with a macronutrient share of 55–60 % of energy from carbohydrates, 10–12 % of energy from proteins and 30 % of energy from fats.
†† The aggregated environmental impact includes eight environmental impact categories: abiotic depletion, global warming, ozone layer depletion, human toxicity, eco-toxicity, phytochemical oxidation, acidification and eutrophication, all expressed as the relative changes in impact per dietary scenario to status quo diet 2003. This aggregated environmental impact and the global warming were given in absolute numbers and relative to the status quo diet.
Most studies have found that the recommended diet might have a lower environmental impact than the current diet, and thus a shift in the direction of the recommended diet might have beneficial impacts on both health and environmental sustainability. However, it is still open to debate whether the recommended diet might be the ideal solution for health and environmental sustainability combined.
Dietary quality scores (B2)
A dietary quality score (e.g. a diet score( Reference Wirt and Collins 69 ) or nutrient profile( Reference Drewnowski and Fulgoni 70 , Reference Van Kernebeek, Oosting and Feskens 71 )) is a summary measure of adherence to a set of dietary guidelines for nutrients and/or food groups. Using this score can be regarded as an application of the dietary guidelines with the aim to identify whether different diets and/or groups of the population are consuming a diet that is close to the dietary guidelines. Seven studies used this score to address the question ‘How is dietary quality – as assessed by a score – related to environmental sustainability?’ (Table 4). In these studies, this approach was applied merely for descriptive purposes as the aim was to compare nutritional quality of the diet by a score( Reference Roos, Karlsson and Witthoft 39 – Reference Van Dooren and Aiking 41 , Reference Carvalho, César and Fisberg 51 ) or by population strata( Reference Vieux, Soler and Touazi 37 , Reference Masset, Vieux and Verger 38 , Reference Monsivais, Scarborough and Lloyd 57 ), and subsequently to assess the environmental sustainability of the different diets or population strata. Out of these seven studies, three studies directly investigated the combination of a healthy and an environmentally sustainable diet by applying a dietary quality score and a sustainability score( Reference Masset, Vieux and Verger 38 , Reference van Dooren, Marinussen and Blonk 40 , Reference Van Dooren and Aiking 41 ). This sustainability score was either calculated with a composite score including diet-related greenhouse gas emissions and land use( Reference van Dooren, Marinussen and Blonk 40 , Reference Van Dooren and Aiking 41 ) or based on strata for the diet-related greenhouse gas emissions( Reference Masset, Vieux and Verger 38 , Reference Temme, Toxopeus and Kramer 54 ). For example, Masset et al.( Reference Masset, Vieux and Verger 38 ) identified the ‘more sustainable’ diets by applying both a diet score and a sustainability score, dividing the population into strata of nutritional quality and strata of greenhouse gas emissions in order to describe the diets that were ranked high on both the health and the sustainability aspects of the diet.
Dietary quality scores, as an application of the dietary guidelines, in relation to environmental sustainability for a descriptive analysis on environmentally sustainable diets
GHGE, greenhouse gas emissions.
* Diet scores are used to subdivide the population into groups of nutritional quality (e.g. Vieux et al. ( Reference Vieux, Soler and Touazi 37 ), created four classes of nutritional quality in which a high-nutritional-quality diet was defined as having a Mean Adequacy Ratio score above the median, a Mean Excess Ratio score below the median and an Energy Density score below the median; Monsivais et al. ( Reference Monsivais, Scarborough and Lloyd 57 ), quintiles of DASH scores; Masset et al. ( Reference Masset, Vieux and Verger 38 ), two groups by median split of PANDiet score). Diet scores are also used for comparison of different dietary scenarios (e.g. Carvalho et al. ( Reference Carvalho, César and Fisberg 51 ), moderate meat consumption pattern with excessive meat consumption patterns (having a red and processed meat intake higher than the World Cancer Research Fund maximum recommended intake of red and processed meat of 500 g/week (≈71·4 g/d)); Röös et al. ( Reference Vanham, Mekonnen and Hoekstra 24 ), current diet with the Swedish Nordic recommended diet and the low-carbohydrate/high-fat diet applying energy-equivalent scenarios; Van Dooren et al. ( Reference van Dooren, Marinussen and Blonk 40 ) and Van Dooren and Aiking( Reference Van Dooren and Aiking 41 ), current diet with recommended Dutch diet, semi-vegetarian, traditional vegetarian, vegan, Mediterranean, New Nordic Diet, historical Low Lands and optimised Low Lands diets).
† Food aggregation level: the number of food items or commodities (depending on author’s terminology) for which environmental sustainability data of food intake was available.
‡ GHGE median cut-off point to define a lower- v. a higher-carbon diet, and then in combination with the higher-quality diet (PANDiet above median) the more sustainable diet in this populations has been identified.
§ GHGE and land use are incorporated into a composite sustainability score that is used for the comparison of different dietary scenarios.
While this approach expresses the health aspect of the diet in one overall score, the interpretation is limited by score-related limitations such as the inclusion of a selected number of dietary components, arbitrary penalties for unmet criteria and the failure of the overall score to accentuate specific shortages/deficiencies. However, although such scores summarise pre-existing knowledge of diet–disease relationships, they are considered as less detailed indicators to assess dietary quality, which might result in misclassification of diets and hence weakened associations.
Diet modelling techniques (B3)
Integrating the health aspect into environmental sciences in a more advanced way involves the application of mathematical modelling techniques, which allows for the design of optimised diets on multiple diet-related factors. Eight studies used mathematical modelling techniques including quadratic modelling( Reference Jalava, Kummu and Porkka 27 , Reference Arnoult, Jones and Tranter 43 ), smooth non-linear programming( Reference Green, Milner and Dangour 46 ) and linear programming( Reference Macdiarmid, Kyle and Horgan 44 , Reference Thompson, Gower and Darmon 45 , Reference Tyszler, Kramer and Blonk 52 , Reference van Dooren, Tyszler and Kramer 53 , Reference Wilson, Nghiem and Mhurchu 56 ) to address the question ‘What would be the food composition of a diet when aiming at the optimisation of multiple diet-related factors?’ (Table 5). These studies all aim at optimising the food composition of the diet based on objectives for health and environmental sustainability while minimising the deviation from the habitual food composition of the current diet, regardless of the modelling techniques and mathematical assumptions.
Diet modelling using mathematical programming techniques for the design of optimised diets for health and environmental sustainability
INCA 2, Individual and National Study on Food Consumption; ENIDE, Spanish National Diet Survey; GHGE, greenhouse gas emissions.
* Food aggregation level: the number of food items or groups (depending on author’s terminology) for which environmental sustainability data of food intake was available.
† Additional diet models were optimised to meet nutrient requirements and: (i) minimise costs; (ii) minimise costs and GHGE; (iii) be relatively healthy, Mediterranean- and Asian-style; and (iv) include ‘more familiar New Zealand meals’.
‡ Diet was initially optimised in view of dietary recommendations only, thereafter additional diet models were optimised using quadratic programming to meet nutritional constraints along with a forced reduction on the animal-based products, in particular including limits on the protein intake from all animal products and from meat, starting from a limit to 50 % and 16·7 %, respectively, and gradually reducing these to zero.
§ Additional diet models were optimised to (i) meet nutritional constraints only, and along with forced reductions on animal-based products (ii) excluding meat, (iii) excluding meat and fish, and (iv) excluding meat, fish, dairy and eggs.
|| Food quantity limits (i.e. upper and/or lower bounds) were set for each group to give standard usable portion sizes (i.e. in whole units or in units in which it is sold).
¶ For France, acceptability constraints on food quantity for each food item included a minimum value equal to the 5th percentile of consumption observed in the population (non-consumers included) and a maximum value equal to the 95th percentile of consumption observed in the population (non-consumers excluded), to ensure that the number of daily portions is acceptable to consumers. For Spain and Sweden, bounds were based on food popularity including minimum portion sizes. Food popularity (that is related to cultural preferences) was based on the current consumption as observed in the dietary surveys and expressed as the percentage of the populating consuming a particular food item. This resulted in the following acceptability constraints: (i) amounts consumed in a particular food group should at least be 60–80 % of the habitual consumption; (ii) popular foods (eaten by at least 50 % of the population) could be increased by up to four times, but not decreased by 30 % of the habitual consumption; (iii) unpopular foods (eaten by less than 25 % of the population) were limited to no more than twice the habitual consumption; and (iv) other foods could be increased up to three times the habitual consumption.
** Penalty score: any change in serving size as compared with the current diet contributes to an arbitrary penalty score with a penalty contribution that is food- and direction-dependent.
†† Diet was optimised in view of dietary recommendations only using quadratic programming; the environmental impact was not considered during the modelling, but estimated afterwards for the optimised diet model.
‡‡ Diet was initially optimised in view of dietary recommendations only using smooth non-linear programming; thereafter additional diet models were optimised in view of environmental concerns, in particular a gradual reduction by 10 % of GHGE.
In diet modelling, nutritional constraints are used to ensure nutritional adequacy and are built upon the physiological nutrient requirements, often with the addition of a few food-based dietary guidelines (e.g. on fruit and vegetables, and fish). Additional constraints are added to the model to derive diets that are acceptable to consumers; these acceptability constraints are based on habitual food preferences and therefore intend to minimise the deviation from the current diet. More specifically, constraints on the food quantity force the model to choose for standard usable portion sizes, and force the model to either select food items that would not have been selected because of high environmental sustainability or low nutritional values, or restrict the maximum quantity of food items that would have been selected otherwise( Reference Macdiarmid, Kyle and Horgan 44 , Reference Thompson, Gower and Darmon 45 , Reference Wilson, Nghiem and Mhurchu 56 ). Instead, constraints on food popularity force the model to minimise the deviations from the current diet( Reference Jalava, Kummu and Porkka 27 , Reference Thompson, Gower and Darmon 45 , Reference Tyszler, Kramer and Blonk 52 , Reference van Dooren, Tyszler and Kramer 53 ), whereby popularity is based on either the percentage of the population consuming a particular food item( Reference Thompson, Gower and Darmon 45 ) or an arbitrary penalty score for any change from the current diet( Reference Jalava, Kummu and Porkka 27 , Reference Tyszler, Kramer and Blonk 52 , Reference van Dooren, Tyszler and Kramer 53 ).
All these modelling techniques describe the optimised diet output in the format of a list of food items that can be consumed in a specified quantity, and it has been demonstrated that from such a list a seven-day-week menu based on three meals per day and in-between snacks can be created while still maintaining dietary preferences (e.g. traditional meal compositions such as milk and breakfast cereals, meat and vegetables and potatoes, etc.)( Reference Macdiarmid, Kyle and Horgan 44 , Reference Macdiarmid, Kyle and Horgan 72 ). However, the output of the diet model is highly dependent on the availability of an appropriate database, thus bridging dietary composition data with diet-related environmental sustainability data. Also, the acceptability constraints have a major influence on the output of the diet model, resulting in a sub-optimised, but more realistic diet in accordance with the current diet.
An approach evaluating the diet-related health impact: diet-related health impact analyses (C)
Diet-related health impact analysis in environmental sciences addresses the question ‘What would be the change in health impact based on nutrient adequacy and/or health/disease outcomes when individuals adopt a more environmentally sustainable diet?’ Seven studies quantified the diet-related health impact of diets differing in environmental sustainability, either directly by observing nutrient adequacy or chronic disease risk as outcomes( Reference Temme, Toxopeus and Kramer 54 , Reference Biesbroek, Bueno-de-Mesquita and Peeters 60 ) or indirectly by modelling the expected health impact( Reference Scarborough, Allender and Clarke 31 , Reference Briggs, Kehlbacher and Tiffin 32 , Reference Aston, Smith and Powles 47 – Reference Friel, Dangour and Garnett 49 ) (Table 6). The direct approach was used by one cross-sectional survey that assessed nutrient adequacy using data from the Dutch National Food Consumption Survey including 3819 subjects aged 7–69 years( Reference Temme, Toxopeus and Kramer 54 ) and by one prospective cohort study that investigated total mortality risk using data from the EPIC-NL (European Prospective Investigation into Cancer and Nutrition–Netherlands cohort) including 35 057 adults with a median follow-up of 16 years( Reference Biesbroek, Bueno-de-Mesquita and Peeters 60 ). For the indirect approach, five studies did not actually observe nutrient adequacy or risk reductions as outcomes, but they modelled the expected diet-related health impact of the more environmentally sustainable diet based on risk ratios obtained from meta-analysis on diet–disease associations( Reference Scarborough, Allender and Clarke 31 , Reference Briggs, Kehlbacher and Tiffin 32 , Reference Aston, Smith and Powles 47 – Reference Friel, Dangour and Garnett 49 ).
Diet-related health impact analyses of environmentally sustainable diets
EPIC-NL, European Prospective Investigation into Cancer and Nutrition–Netherlands cohort; CO2e, CO2 equivalents; GHGE, greenhouse gas emissions; DALY, disability-adjusted life years; YLL, years of life lost.
* Food aggregation level: the number of food items or groups (depending on author’s terminology) for which environmental sustainability data of food intake was available.
† The analysis of the health impact (i.e. mortality survival analysis) was based on data from 35 057 subjects included in EPIC-NL, a prospective cohort study with a median follow-up of 15·9 years. The main aim was to investigate the relationship between diet-related sustainability and mortality outcomes either by population stratification for the environmental indicators (e.g. GHGE and land use) or by meat-substitution scenarios.
‡ In the meat-substitution scenario, the replacement of meat was compensated by means of food weight and the plant-based meat-substitutes were potatoes, total vegetables, total fruit/nuts/seeds, pasta/rice/couscous, cheese, milk-based desserts or fish, representing acceptable alternatives for meat because these foods are consumed in significant amounts in the Dutch diet and can replace meat in a hot meal. The reduction in all-cause mortality risk and environmental impact was estimated separately per meat-substitution option and for an option with no replacement.
§ Potential impact fraction was calculated as the difference between current aggregate risk and aggregate risk under counterfactual divided by current aggregate risk, and represents the proportion/percentage of disease in the population that can be attributed to the current diet and therefore could potentially have been avoided under the counterfactual diet.
|| The DIETRON model included the intake of total energy, fruit, vegetables, fibre, total fat, mono- and polyunsaturated, saturated and trans-fatty acids, dietary cholesterol and salt as dietary input to estimate the link between food consumption and mortality using age- and sex-specific relative risk estimates from meta-analyses.
¶ A 30 % decrease in livestock production is assumed to result in a reduction of equal size in the consumption of animal-based products, and thus a decrease in the dietary intake of saturated fat.
This approach of linking diet-related health/disease outcomes to environmental sustainability might be considered as suitable evidence to influence food choices and food production, since nutrient adequacy and diet-related health/disease outcomes are predictive for the future healthiness of dietary change. The healthiness of food products has been recognised as an important determinant of food choice, apart from taste and price, whereas sustainability motives are currently not considered substantial influential factors( Reference Verain, Dagevos and Antonides 63 , Reference Roininen, Lähteenmäki and Tuorila 73 – Reference Grunert, Hieke and Wills 75 ).
Methodological considerations
The design of optimised sustainable diets should take into account certain methodological considerations as presented below. First, the current diet needs to be linked to health and environmental sustainability, whereby this link depends on the assessment method of the current diet. Second, the indicators of ‘health’ and ‘environmental sustainability’ must be well defined to support the design of sustainable diets. Third, sustainable diets incorporate more than only health and environmental sustainability, and thus future steps have to be taken to identify the social, ethical( Reference Coff, Korthals and Barling 76 ) and economic( Reference Oosterveer and Sonnenfeld 77 ) indicators related to a sustainable diet, such as the cultural acceptance of a diet, the biodiversity, animal health and welfare, the production of economically fair products that are accessible and affordable for people at all times, etc.
Food availability or food intake – how to connect health with environmental sustainability?
The assessment of the current diet can be based on either food availability related to food production and expenditure, or actual food intake closely related to food consumption and thus the health aspect of the diet. The main questions related to designing sustainable diets are ‘How to connect health with environmental sustainability?’ and ‘What is the influence of the assessment method?’
The quantification of diet-related environmental sustainability should preferably be based on food availability estimates rather than on actual food intake data. The reason for this is that food availability estimates represent the food supply/production or food expenditure/purchases at the national or the household level and thus include food losses at production level and food wastages at consumption level. For example, data on the per capita food supply obtained from the Food Balance Sheets of the FAO reflect the quantity of food products that are produced, used for trade, adjusted for stock changes and non-nutritional use, and expressed in primary equivalents (primary food commodities) per capita per day( 78 ); whereas data on the household’s consumption expenditure obtained from Household Budget Surveys reflect the quantity of food products that enters the households( Reference Trichopoulou 79 ). However, food availability estimates have little connection to the individual dietary pattern and thereby its diet–health relationship, as noticed in the limited health evaluation of the whole diet in studies using population or household measurement level.
In contrast, an individual’s diet that is obtained from individual-level food intake assessment methods enables a strong connection with individuals’ diet-related factors (e.g. age, sex, socio-economic status) and corresponding health aspects (e.g. nutrient adequacy and/or diet-related health/disease outcomes)( Reference Thompson, Subar and Coulston 80 ), but has a less strong connection with the estimation of environmental sustainability (e.g. indicators are typically available only for primary food commodities up to the regional distribution centre). When using individual-level food intake assessment, some underlying methodological issues should be taken into account for assessing the health aspect of a diet at population level, in particular the representativeness of the individual’s diet and the sample’s representativeness for the population( Reference Thompson, Subar and Coulston 80 ). National survey methods, such as diet records and 24 h recalls, are suitable methods to assess the intake of an unlimited number of food items consumed by an individual over one or more days, with portion sizes and preparation practices; hereby describing habitual intakes at population level, but not linking this with diet-related health/disease outcomes within individuals. An FFQ that focuses on ranking individuals according to their usual food intake by capturing the intake of food items over a designated time period (e.g. usually varying from the last month to the last year) from a finite list has been commonly used to assess the association between dietary intake and health/disease outcomes in large epidemiological studies. When aiming at estimating the environmental sustainability related to food consumption, the answer to the question which dietary assessment method to use depends on the desired link with health and the desired level of food aggregation, which is not yet available for sustainability indicators on the level of (all) individual food items.
In short, this discrepancy in measurement/aggregation level forms a methodological barrier in connecting both health and environmental sustainability aspects of a diet. Based on the literature review, when aiming to design sustainable diets, dietary data collected at the individual level might be considered the preferred measurement level. The main reason for the selection of this measurement level is the possibility for monitoring health in terms of foods and nutrients, without directly hampering the linkage with environmental sustainability indicators. Foods are the common denominator regardless of the higher aggregation level of sustainability indicators and their conversion into primary commodities( Reference Herforth, Frongillo and Sassi 81 ).
Future perspectives
In a complex field that has emerged from different scientific disciplines, clear definitions of ‘health’ and ‘environmental sustainability’ are essential. Health can be defined on the basis of nutrients and foods; the former using dietary reference values related to physiological needs for healthy growing and ageing( 82 ), and the latter using food-based dietary guidelines related to health/disease outcomes( 83 ). A further issue in this is that nutrient-based and food-based dietary guidelines differ between countries and that they are based on population averages with average energy requirements, whereas physiological nutrient needs vary considerably because of body size, physical activity and phase of the life cycle. Expressing nutritional requirements and intakes in terms of nutrient densities might be helpful to independently address food composition and energy intake( Reference Backstrand 84 ). However, when designing an optimised sustainable diet, both facets of nutritional health should be taken into account; i.e. the essential nutrients that are consumed in insufficient amounts or in excess at population level (nutrient adequacy), and the important acceptable foods for maintaining nutrient intake and promoting health (food-based dietary guidelines).
With regard to environmental sustainability, the quantification of this is still in its infancy and driven by present know-how and available measurement equipment. This often results in focusing on the environmental impact of greenhouse gas emissions and land use, while omitting the broader perspective that also includes natural resource use and biodiversity, among others. Because this emphasis on greenhouse gas emissions and land use was included specifically in our search terms, this may have influenced the number of papers within the five approaches identified, but the range of approaches is likely to be covered. Also, the environmental assessment is often restricted to the system boundaries of the life cycle assessment, which in theory cycles from farmer production to final consumption and disposal, but in practice usually stops at the distribution centre or even at the farm gate; thus many studies do address food availability on the basis of food production and/or food purchase data, i.e. addressing food that is produced and/or entering the households, thereby ignoring inedible parts and food waste( Reference Finnveden, Hauschild and Ekvall 85 ). Future research is therefore needed to develop quantitative methods for assessing the full picture of diet-related environmental sustainability indicators.
Conclusions
In operationalising the health aspect of an environmentally sustainable diet, the first priority will be to define which research questions to address and the second will be to ascertain an appropriate match in the measurement level of health and environmental sustainability. The research questions determine whether to apply a descriptive or an analytical outline. The descriptive outline refers to the comparison of different diets based on dietary guidelines, dietary quality scores and diet-related health-impact analysis, while the analytical outline refers to the design of alternative diets based on food item replacement and diet modelling techniques. Therefore, in the context of operationalising the health aspect when designing sustainable diets, diet modelling might be considered the preferred approach since it captures the complexity of the diet as a whole. Hence, there is a need for individual-level dietary data related to the food consumption with regard to food and nutrient intakes. It is important to recognise that the concept of sustainable diets is used across multiple fields and not only includes food and nutrition as such, but also the environment, agriculture, animal sciences, social and economic sciences, which need to be taken into account when designing sustainable diets for the future.
An avenue for future research in designing sustainable diets: the SHARP diet
In the context of developing a future vision for designing optimised sustainable diets, the broader concept of sustainable diets as defined by the FAO( Reference Burlingame and Dernini 10 ) should be considered when aiming at diet optimisation in a multidimensional way. We, therefore, propose the concept of a diet that is SHARP: environmentally Sustainable (S), Healthy (H), Affordable (A; accessible for consumers yet also supporting the agriculture food sector), Reliable (R; stable in its supply and safe) and Preferable (P; consistent with cultural norms and food preferences). This SHARP diet would be in line with the wider definition of sustainability by including its social, ecological and economic dimensions. This requires further exploration of mapping these diet-related dimensions into objectives/constraints for the diet model that aims at an optimised sustainable diet for all diet-related sustainability perspectives.
Diet modelling might be the preferred approach to analyse current and design future diets as multiple diet-related aspects (e.g. health, environmental sustainability, affordability, accessibility and acceptability) can be taken into account simultaneously. The output of the diet model (i.e. food list with specified quantities) is highly dependent on the constraints included and the diet-related sustainability data available. As different parameter settings for these constraints might have major effects, the robustness of such diet models needs attention, especially with respect to the trade-off between conflicting objectives and exploring adaptiveness to future changes in environmental sustainability options (e.g. improved food production processes), food consumption patterns (e.g. innovative new food products) and/or other diet-related factors (e.g. accessibility and affordability). A major challenge with analysing potential trade-offs to identify preferred scenarios is, however, to fully understand the interaction across all indicators of a sustainable diet within the different socio-economic and environmental contexts( Reference Lock, Smith and Dangour 86 ). Importantly, the output of the diet model should not be viewed as achieving one optimum, but rather a set of preferred dietary options dependent on the optimisation aims of the different stakeholders (e.g. consumers, agricultural sectors, food industries and politicians).
Acknowledgements
Acknowledgements: The authors acknowledge Professor Toon van Hooijdonk (Food Quality and Design, Wageningen University; FrieslandCampina, Amersfoort, The Netherlands) for interesting academic partners and food industries in joint research into healthy, sustainable and consumer-friendly diets. Financial support: Funding was obtained from the Dutch Dairy Association (NZO) and the European Union’s H2020 Programme (grant agreement number 633692, SUSFANS); preparatory work was funded by the Graduate School VLAG, Wageningen. The funders had no role in the design, analysis or writing of this article. Conflict of interest: The authors have no personal or financial conflicts of interest. Authorship: P.v.V., G.J.H. and J.M.J.M.S. initiated the topic of the review. P.v.V. supervised the conduct of the review by E.M. and A.K. and the draft of the manuscript by E.M. The draft was critically evaluated and refined by P.v.V. and A.K. G.J.H. and J.M.J.M.S. evaluated the manuscript for internal coherency. All authors read and approved the final submission of the paper. Ethics of human subject participation: Not applicable.
