TABLE 1.
Key features of the identified studies published since January 2015 that used mathematical optimization relating to human diets and ≥1 aspect of environmental sustainability1
| Study (ordered alphabetically by first author surname) | Key variables to reduce/increase or which imposed constraints | Main findings | Comments |
|---|---|---|---|
| Barré et al. 2018 (11) | Reducing each of diet-related environmental impacts (GHGEs, marine eutrophication, atmospheric acidification); agricultural coproduction; nutritional adequacy; nutrient bioavailability; acceptability; affordability | “Decreasing meat content was strictly needed to achieve more sustainable diets for French adults,” but the reduction was reduced by 6% when nutrient bioavailability and 25% when coproduction links were taken into account (e.g., coproduction included both milk and meat from dairy cattle) Food costs were lower in the more sustainable diets |
A study strength in terms of realism was modeling around “departing the least from the mean observed French diet”; other strengths were the unique focus on both “nutrient bioavailability” and “coproduction”; and considering multiple environmental impacts and affordability |
| Gephart et al. 2016 (12) | Reducing each of: GHGEs (carbon footprint), nitrogen release (nitrogen footprint), water use (blue and green water footprint), and land use (land footprint); adequate nutrients | “We find that diets for the minimized footprints tend to be similar for the four footprints, suggesting there are generally synergies, rather than trade-offs, among low footprint diets. Plant-based food and seafood (fish and other aquatic foods) commonly appear in minimized diets” “Livestock products rarely appear in minimized diets, suggesting these foods tend to be less efficient from an environmental perspective, even when nutrient content is considered” “The results' emphasis on seafood is complicated by the environmental impacts of aquaculture versus capture fisheries, increasing in aquaculture, and shifting compositions of aquaculture feeds” |
A major strength of this US study was the use of 4 environmental indicators (although they were given equal weighting which may be a problematic simplification). The issues around seafood use (aquaculture vs. capture fisheries) are also well discussed. But only 19 nutrients were considered in this study. Also for some analyses only requirements for calories, protein, carbohydrate, and fiber were involved. The authors acknowledged that “there are other aspects of environmental change not covered with these footprints including antibiotic and pesticide use, animal welfare (when applicable), biodiversity, GMO's, industrial pollution, disease risk, etc.” Also “the negative impacts of overfishing and bycatch in capture fisheries are not considered here” |
| Green et al. 2015 (13), with the health aspects covered further in reference (14) | Reducing GHGE; nutritional adequacy; reducing the loss of consumer welfare (reflected by the deviation between the current and the optimized diets, and the use of own-price elasticities) | If “average diets among UK adults conformed to WHO recommendations, their associated GHG emissions would be reduced by 17%. Further GHG emission reductions of around 40% could be achieved by making realistic modifications to diets so that they contain fewer animal products and processed snacks and more fruit, vegetables and cereals. However, our models show that reducing emissions beyond 40% through dietary changes alone will be unlikely without radically changing current consumption patterns and potentially reducing the nutritional quality of diets” “Our dietary optimisations show that emissions reductions can be achieved by reducing consumption of animal products, switching to meats and dairy products with lower associated emissions (e.g., pork, chicken and milk), reducing consumption of savoury snacks, switching to fruits and vegetables with lower emissions” |
A novel strength was how consumer welfare was considered. Also the use of a complete life-cycle analysis of emissions specific to the UK (where possible). It noted the problem of the “fact that consumption of unhealthy foods is more likely to be under-reported than consumption of healthy foods means that we are likely to have underestimated the unhealthiness of the UK diet” In the associated publication the health aspects were quantified: “Our model suggests that it would save almost 7 million years of life lost prematurely in the UK over the next 30 years and increase average life expectancy by over 8 months” (14) |
| Horgan et al. 2016 (15) | Reducing GHGEs; achieving dietary recommendations | “The healthy diets and sustainable diets produced a 15 and 27% reduction in greenhouse gas emissions respectively” “Sodium proved the most difficult nutrient recommendation to meet” “While the majority had to make more substantive dietary changes to the amount of any food currently eaten and addition of new foods to their diet, for only a very small proportion of the sample foods needed to be removed from their diet.” The optimized diets involved “reductions in sweet foods (biscuits, cakes and desserts), processed meats, alcohol and white bread”. Also “total meat consumption, especially beef and lamb, decreased more to achieve sustainable diet than just a healthy diet” |
Key strengths were the consideration of individual diets (n = 1491 in a UK national diet survey) and then minimizing the changes they need to make. The authors acknowledged the following limitations: GHGEs were not for the full life cycle (e.g., in processing composite foods and waste disposal); the sustainability of fishing was not considered; and the monetary cost of the dietary changes was also not included. |
| Kramer et al. 2017 (16) | Reducing an integrated impact covering: GHGEs, fossil energy use and land occupation; nutrient composition; a metric for popularity (weight of food consumed) | “Reducing meat is the most effective option for lowering the environmental impact of diets in all age–gender groups. Reducing alcoholic and non-alcoholic beverages is another option. Leaving out fish and dairy products are not” “[T]he preferred environmental savings come almost exclusively from consuming less meat, especially beef” “Our results indicate that alcoholic and non-alcoholic beverages are responsible for 14–16% of the total environmental impact of the current diet in the Netherlands” |
A key strength of this study was the “cradle-to-grave life cycle analyses” for a large number of products (n = 207). The authors also minimized the distance relative to the current diet. However, a limitation was that they ignored “important global impacts such as overfishing and water stress.” Also diet affordability was not considered. They argued that “fish cannot be replaced, primarily because it is the exclusive dietary source of EPA and DHA” |
| Milner et al. 2017 (17) | Reducing GHGEs and water per person for irrigation (blue water footprint); achieving nutritional guidelines; acceptability (minimized deviation from existing patterns) | For this study of India: “The optimised diets had up to 30% lower blue water footprints and generally contained lower amounts of wheat, dairy, and poultry, and increased amounts of legumes. In the 2050 scenario, adoption of these diets would on average result in 6800 life-years gained per 100 000 total population (95% CI 1600–13 100) over 40 years. The dietary changes were accompanied by reductions in greenhouse gas emissions. The magnitude of the health and environmental effects varied between dietary patterns” GHGE changes ranged from reductions of 2% to 36% |
This appears to be the first such study to particularly focus on the problem of limited freshwater availability for agriculture. It is also a relatively uncommon study in that it covers a low-income country. The health metrics are relatively advanced (life-years gained based on 9 diet-related diseases) and the Monte Carlo simulations improve the considerations around uncertainty Study limitations included the following: “Our scenarios were based on projections of population growth only and did not account for other potential drivers with more complex and uncertain effects, such as climate change and aquifer depletion. We also did not account for non-food crop production, current dietary trends, temporal variation in water availability, or the effects of socioeconomic differences between dietary patterns” |
| Perignon et al. 2016 (18) | Reducing GHGEs; achieving nutritional adequacy; acceptability |
“30% GHGE reduction could be achieved in a nutritionally adequate diet by increasing fruits and vegetables while maintaining intake of meat/fish/eggs at approximately 100 g/d, mainly by substituting ruminant and deli meats by fish products” “Higher GHGE reductions either impaired nutritional quality, even when macronutrient recommendations were imposed, or required non-trivial dietary shifts compromising acceptability to reach nutritional adequacy” “Imposing the nutritional constraints of the ADEQ [all nutrient recommendations] scenario slightly increased the cost of the diet. High GHGE reductions (≥50%) decreased diet cost” “The maximal GHGE reduction achievable from the observed level, while respecting all the nutritional recommendations, was 69.7% for women and 74.0% for men” “High GHGE reductions resulted in the elimination of some food groups, namely Dairy and MFE [meat/fish/eggs]” |
Strengths of this study were that it minimized the departure from observed diets in the French diet and included a relatively large number of foods (n = 402). It also had food costs as an outcome measure (which increased slightly for the healthier diet but decreased with higher GHGE reductions) Limitations the authors acknowledged included not covering water footprints and the sustainability concerns around fisheries |
| Song et al. 2017 (19) | Reducing the carbon footprint; meeting DRIs for adults; acceptability (avoiding marked deviations from current diets) | “The theoretical optimal diet reduced daily footprints by 46%, but this diet was unrealistic due to limited food diversity.” Constrained by acceptability, the optimal diet reduced the daily carbon footprints by 7–28% for men and by 5–26% for women. Seven of 8 scenarios showed that reductions in meat consumption resulted in greater reductions in GHGE. “However, dramatic reductions in meat consumption may produce smaller reductions in emissions, as the consumption of other ingredients increases to compensate for the nutrients in meat. A trade-off between poultry and other meats (beef, pork, and lamb) is usually observed” | Strengths of this work were that it was for a country outside of the high-income group (i.e., China). There was also good consideration of uncertainty (around 443 variables). The lack of China-specific carbon footprint data was a limitation. Nevertheless, the data were sourced from an LCA database derived from 1237 reviewed LCA studies of food carbon footprints along with food supply chains including crop cultivation, breeding, industrial processes, transportation, and storage |
| Tyszler et al. 2016 (20) | Reducing GHGEs, fossil energy use, land occupation (all combined in an integrated score); nutrient requirements; popularity of food products | “We show, by using linear programming, that it is possible to reach 30% reduction in the environmental impact with a diet which is relatively similar to the current one and could be more likely to be accepted” “Removing meat and fish from the diet reduces the environmental impact by about 21%. A healthy vegan diet reaches 30% environmental impact reduction” “The optimal solution still contains 30% of the amount of meat quantity in the Closest healthy diet, whereas amounts of dairy (liquid and cheese), fish, and egg almost remain constant. The reduced meat consumption is responsible for 60% of the reduction in environmental impact” “Reductions in meat (beef, chicken) and beverages (beer, wine) are making the most important contributions” |
As above for Kramer et al, which utilized some of the same data, a key strength of this study was the “cradle-to-grave life cycle analyses” for a large number (n = 207) of food products. Multiple other environmental indicators were also used An acknowledged limitation was that the results were only for 1 population group (Dutch women aged 31–50 y). |
| van Dooren et al. 2015 (21) | Reducing GHGEs; achieving dietary requirements; maximizing acceptability; reducing costs | “A diet of 63 popular and low priced basic products was found to deliver all required nutrients at an adequate level for both male and female adults. This plant-based, carbohydrate and fiber-rich diet consists mainly of wholegrain bread, potatoes, muesli, open-field vegetables and fruits. The climate impact of this diet is very low (1.59 kg CO2eq/day) compared to the average Dutch diet. By constraining costs, a low carbon diet of €2.59/day is possible. Conclusions: A two-person diet consisting of 63 products and costing €37 per week can simultaneously be healthy and yet have half the average climate impact” | A strength was including food costs in the optimization, as this informs feasibility for low-income groups. Also the large number of nutrients and indicators used (n = 33) and including 206 food products. Acceptability was maximized by “maximizing the most consumed food products based on weight and minimizing absolute change in portions”. The study also showed outcomes by fossil energy use and land use (but these were not used as constraints) |
| Verger et al. 2018 (22) | Reducing water use and land use effects (including biodiversity); adherence to the Mediterranean diet pyramid |
Preliminary results for Tunisia: “Using the first model based on the current dietary habits and practices (individual level), we found that the main dietary changes needed to satisfy all the nutrient recommendations were the increases of fruits and dairy products, and decreases of meat and starchy foods. Nevertheless, these changes increased the environmental impacts of the diets. In a scenario where environmental indicators were limited to their observed levels, the dietary changes needed were still the decreases of meat and starchy foods but also lower increases of fruits and dairy products in favor of vegetables” | This publication describes a detailed framework for considering sustainable food systems for nutrition and health in the Mediterranean region. The results presented for Tunisia (a middle-income country) are preliminary but they add to the data outside of the high-income country grouping. Some of the authors have a long track record for considering many aspects of healthy diets and sustainability |
| Vieux et al. 2018 (23) | Reducing GHGE; achieving nutritional adequacy | “Diet sustainability can be improved by substituting food items from the sugar/fat/alcohol food group with fruit, vegetables, and starches, and country-specific changes in consumption of animal-based products” “Although an increase in plant-based products was needed in every country, the shifts in animal-based products were not homogeneous across countries or gender,” e.g., meat consumption increased for French men—albeit poultry and pork. Also dairy consumption increased in some populations “A maximal GHGE decrease of 62–78% was theoretically achievable while still ensuring nutritionally adequacy, but at a strong risk of compromising the cultural acceptability of the diets” |
A strength was data from 5 countries (France, the UK, Italy, Finland, and Sweden) along with minimizing deviation from observed diets for men and women The authors noted the limitation of just considering GHGEs and not “eutrophication, water footprint, land use or biodiversity indicators” |
1GHG, greenhouse gas; GHGE, greenhouse gas emission; LCA, life-cycle assessment.