Human health, environmental sustainability, and food security are linked in complex and multidirectional ways. One important intersection occurs between food production and the natural environment. At present, agriculture is the largest consumer of fresh water and the second largest contributor to anthropogenic greenhouse gas emissions worldwide (including land use change).1, 2 Agriculture is also the driving force behind the disruption of global nitrogen and phosphoruscycles, with particular burden coming from increases in the quantity and intensity of livestock production.3 Planetary boundaries for biogeochemical flows (of nitrogen and phosphorus) and climate change are estimated to already be exceeded.4
Disrupted planetary systems produce effects that will continue to feed back into agricultural and food systems if left unchanged. Rising atmospheric carbon dioxide(CO2) levels, for example, are projected to reduce the nutritional value of key food crops under climate change.5 Ensuring a safe operating space for humanity,6therefore, requires shifts in what, how, and how much food is produced. Food production is driven by policy choices and consumer demand, among other factors. Developing policy that catalyses shifts to sustainable food consumption patterns, therefore, is a key way to mitigate negative effects.
The concept of long-term food security takes into account these current and projected future dynamics, positing a fundamental connection between the health of human and earth systems. Agricultural, environmental, and nutrition policy must converge to reflect this interconnection. For example, some nations are incorporating sustainability into their dietary guidance. According to the UN Food and Agricultural Organization (FAO), “sustainable diets are those diets with low environmental impacts which contribute to food and nutrition security and to healthy life for present and future generations. Sustainable diets are protective and respectful of biodiversity and ecosystems, culturally acceptable, accessible, economically fair and affordable; nutritionally adequate, safe and healthy; while optimising natural and human resources.”7 Driven both by the FAO's Decade of Nutrition and the general recognition of agriculture's effect on the environment, countries such as the Netherlands,8 Sweden,9 and Brazil10 have developed dietary guidance that promotes planetary health.
Research in context
Evidence before this study
We searched Web of Science and Google Scholar for peer-reviewed studies on dietary guidelines and sustainability, published from January, 2000, to December, 2017. We used the search terms “dietary guidelines for americans,” “dietary guidelines”, “dietary recommendations”, “sustainability”, “environment”, “sustainable diet”, and “united states”. We restricted the search to articles published in English. We found a few studies that focused on the environmental impacts of the Dietary Guidelines for Americans. The studies varied in either the types of healthy eating patterns chosen or the number environmental impacts analysed, but not both. None of these studies included all three patterns in the latest version of the Dietary Guidelines.
Added value of this study
This study developed and combined multiple datasets to assess the environmental burdens of the three healthy eating patterns put forth in the Dietary Guidelines for Americans. To our knowledge, we analysed for the first time the three dietary patterns of the Dietary Guidelines for Americans using a life cycle approach that focuses on six impacts of crucial policy importance: climate change, land use, water depletion, freshwater and marine eutrophication, and particulate matter emissions. To our knowledge, this is also the first analysis that has linked environmental impacts to the actual as-consumed foods underlying the healthy patterns.
Implications of all the available evidence
Although equally recommended as healthy, the patterns in the Dietary Guidelines for Americans might have very different implications for the environment and for human health beyond nutrition. Our findings add to the body of literature that suggests that healthy diet patterns that are higher in plant-based foods and lower in animal-based foods have environmental benefits. Given the evidence and the scale of influence of the Dietary Guidelines for Americans on food systems, incorporating sustainability into them should be prioritised. Future research should focus on improving life cycle assessment data quality for seafood and soy products, expanding system boundaries, and quantifying the net impact of shifting US diets toward these healthy patterns.
In the USA, the Dietary Guidelines for Americans (latest edition 2015–20)11 are the cornerstone of nutrition policy. They have great influence on food systems in the USA and beyond, informing more than US$80 billion of government spending per year, shaping decision making in the food industry, and underpinning consumer education on healthy diets.12, 13 Every 5 years, the Dietary Guidelines are revised after a multistep process that includes evidence review and recommendations by an independent scientific committee (Dietary Guidelines Advisory Committee), public review, and synthesis into the final policy by the US Department of Agriculture and US Department of Health and Human Services. In consideration of the mounting evidence regarding the environmental effects of foods, in 2015, the Dietary Guidelines Advisory Committee included for the first time a chapter focused on food safety and sustainability.14 One of their key conclusions was that “a dietary pattern that is higher in plant-based foods, such as vegetables, fruits, whole grains, legumes, nuts, and seeds, and lower in animal-based foods is more health promoting and is associated with lesser environmental impact than is the current average US diet”.14 Despite unprecedented public support,15 this and other sustainability language were not included in the final 2015–20 Dietary Guidelines published by the US Department of Health and Human Services and the US Department of Agriculture.11 When the policy was released, the Secretaries of Health and Human Services and Agriculture announced that though sustainability was important, it was beyond the scope of the Dietary Guidelines.16
Although substantial literature has been produced on the environmental burdens of foods and diets, few studies have focused on the healthy diet patterns put forth in the Dietary Guidelines.17, 18, 19, 20, 21 These studies have included variation in either the number of impact categories considered17, 18, 19 or in the types of healthy patterns analysed.20, 21 The objective of this study is to assess multiple environmental impacts of the three distinct healthy eating patterns in the 2015–20 Dietary Guidelines. To our knowledge, this study is the first analysis of any food pattern from the Dietary Guidelines for Americans with resolution at the level of individual foods used to derive the patterns (eg, chicken, broiled or baked, no added fat) as opposed to food types (eg, chicken) or groups (eg, proteins).
This modelling study was done at the University of New Hampshire (Durham, NH, USA). We analysed the three diet patterns described by the 2015–20 Dietary Guidelines for Americans: the healthy US-style eating pattern (US pattern), the healthy Mediterranean-style eating pattern (MED pattern), and the healthy vegetarian eating pattern (VEG pattern).14 Food groups and subgroups were composed of nutrient-dense versions of 415 commonly consumed foods from the National Health and Nutrition Examination Survey (NHANES).22 The compositions of groups and subgroups were determined by the US Department of Agriculture during the Dietary Guidelines Advisory Committee process using a food pattern modelling approach.14, 23
Healthy pattern data
We extracted data from the US Department of Agriculture's publicly available documents of the three patterns and their constituent foods and consumption frequencies to tabular format using Tabula Technology software, version 1.1.1. Data entries were manually checked to ensure accuracy and completeness. Further analysis was done within R and Microsoft Excel. Foods that had a food group proportion equal to zero were removed from the dataset, resulting in 321 foods for analysis. We matched the pattern foods with their Food and Nutrient Database for Dietary Studies (FNDDS) codes (Pannucci T R, 2017, Center for Nutrition Policy and Promotion, personal communication). We then converted the pattern foods from the food pattern equivalents provided by the Dietary Guidelines for Americans11 to 100-g amounts using the Food Patterns Equivalents Database (2007–08). All foods were matched except soy protein isolate, which was assigned a conversion factor from the user guide (14·175 g/oz equivalent).24
Food environmental impact data
The environmental impacts of the pattern foods were developed using two datasets: the Food Intakes Converted to Retail Commodities Database (FICRCD)25 and a life cycle assessment dataset of raw and lightly processed foods. We first describe each dataset, and then how they were combined to assign environmental impacts to the foods in the healthy patterns (figure 1).
Figure 1. Process and datasets used to assign life cycle impacts to pattern foods
We used FICRCD to deconstruct pattern foods into their retail-level commodity ingredients according to their proportional mass in 100 g of food. FICRCD consists of 65 retail commodities and translates as-consumed foods to retail-level commodities using multiple conversion factors that account for the loss or gains of mass from preparation, cooking, other processing, and discarding of non-edible parts.
We constructed the food life cycle assessment database primarily using impact assessment results from the World Food LCA Database, version 3.1, which were provided by Quantis.26 Data gaps were filled with life cycle assessments from Ecoinvent, version 3.3, using SimaPro, version 8.4.0, and publicly available life cycle impact assessment results from Agribalyse, version 1.3. Emissions from direct land-use change using a crop-specific approach were included in the life cycle inventories of the World Food LCA Database foods and were added to Ecoinvent foods using the Direct Land Use Assessment Tool by Blonk Consultants, modified by Quantis for the World Food LCA Database. The system boundaries for the life cycle assessments were cradle to farm gate or cradle to processor gate (excluding packaging). When food life cycle assessments included multiproduct systems (eg, multiple products from a slaughterhouse), burdens were allocated to co-products on the basis of economic value with the exception of dairy products, which used biophysical bases.27
Impacts were estimated using the International Reference Life Cycle Data System method,28 which provides results for categories of burdens on ecosystems, human health, and resource use (appendix). For this analysis, we tabulated results for the full suite of 16 International Reference Life Cycle Data System impacts but focus on a subset of categories: global warming potential, marine eutrophication potential, freshwater eutrophication potential, land use, water depletion, and particulate matter or respiratory organics. Global warming potential, marine and freshwater eutrophication, land use, and water use correspond to planetary boundaries at a high or increasing risk of being exceeded, either globally or regionally.4 Particulate matter is a criteria air pollutant according to the US Environmental Protection Agency, because of its human health and environmental effects.29 Additionally, agriculture might be a significant and largely unregulated driver of particulate matter emissions in the USA.30
We matched the most closely related food life cycle assessments to FICRCD commodities for each NHANES food ingredient. Because FICRCD excludes seeds and soy protein isolate and uses a proxy for other processed soy products, we added life cycle impact assessment results for these commodities to the FICRCD list. We mapped 55 unique food life cycle assessments to 63 commodities (appendix). The mapping varied in resolution, which we classified as one of the following categories: a direct match; a match of the farm gate commodity, adjusted for processing (for some processed foods); a proxy for one food type (eg, FICRCD turkey mapped to a chicken meat life cycle assessment); or a proxy for a composite category of less commonly consumed foods (eg, FICRCD grains other than corn, oats, wheat, and rice mapped to a wheat life cycle assessment). We adjusted farm-gate commodities for processing by applying a multiplier that accounted for the farm-gate mass needed per unit of processed product, adjusted for allocation between co-products.31 In terms of origin countries, 45% of the food life cycle assessments corresponded to the USA, Canada, or a key import country for that commodity, whereas the remainder were global average (36%), or best available proxy countries (19%; appendix). Finally, NHANES foods were reconstructed and total environmental impacts were calculated, accounting for ingredient mass in the recipe to make 100 g of final food product.
Environmental impacts of diet patterns
We did impact assessments of the three US Department of Agriculture diet patterns for at the 2000 kcal intake level, which is the reference intake used in nutrition guidance on food labels (percentage daily value). For context, we also estimated impacts for at a 3000 kcal level, which corresponds to the maximum estimated energy requirement of active adults (aged >19 years).11 Life cycle environmental impacts of foods recommended (per 100 g) on a weekly basis were multiplied by 100 g amounts of foods comprising the patterns.
In terms of interpretation, we assumed a 10% difference between patterns was significant for climate change and a 30% difference was significant for eutrophication potential and particulate matter or respiratory inorganics.32 These are default estimates in life cycle assessment based on expert judgment when a more comprehensive uncertainty analysis is not possible or available.32 In the absence of guidance for water depletion and land use, we adopted the 30% cutpoint for those impacts.
To better understand the reliability of the results, we did a post-hoc analysis of the life cycle assessment data quality, focusing on the subgroups that varied between the patterns and had significant impact contributions. This assessment included three aspects. First, we analysed the resolution of the life cycle assessment data that was mapped onto food commodities (eg, direct match vs proxy for one food type). Second, we analysed the geographical representativeness of these life cycle assessments for US consumption. Finally, we assembled data quality ratings of the food life cycle assessments, which were precalculated according to Product Environmental Footprint criteria and only available for food life cycle assessments from the World Food LCA Database.26 The range of possible data quality ratings were poor, fair, good, very good, and excellent quality. Together, we interpreted the resolution, geographical representativeness, and data quality rating indicators, highlighting where data representativeness is likely to be high or in need of improvement.
Role of the funding source
There was no funding source for this study. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
We compared the composition of the three diet patterns presented in the 2015–20 Dietary Guidelines for Americans: the US pattern, MED pattern, and VEG pattern. Although all three diets are recommended as healthy patterns of eating, the composition of the patterns varied in the amount of fruit, whole grains, dairy, oils, and discretionary calories recommended, as well as the amount and types of proteinincluded (table 1).
Table 1. Recommended weekly amounts for healthy patterns at 2000 kcal per day11
|Healthy US-style||Healthy Mediterranean-style||Healthy vegetarian|
|Red or orange||5·5||5·5||5·5|
|Beans and peas*||1·5||1·5||1·5|
|Protein foods, oz||39·0||45·5||24·0|
|Nuts and seeds||4·5||4·5||7·0|
|Beans and peas*||..||..||6·0|
|Limit on calories for other uses†||1890||1820||2030|
All food or beverage recommendations are amounts per week. Vegetables, fruits, and dairy are measured in equivalent amounts in cups and protein and grains are equivalent amount in oz. This might be less than a measured cup or oz if the food is concentrated or low in water content, or more than a measured cup or oz if the food is airy or contains a large amount of water.11 The two subcategories of meats, poultry, eggs and nuts, seeds, soy products in the Dietary Guidelines were disaggregated here into individual categories using publicly available data.23
* Beans and peas recommendations are included in both the vegetable and protein groups for the healthy vegetarian pattern only.
† Includes calories for added sugars, solid fats, alcohol, additional refined starches, and amounts in excess of the recommended values for the other food groups.
In the life cycle impact assessment of the three diet patterns at the 2000 kcal level, the US diet pattern and MED diet pattern had similar results for all impacts except freshwater eutrophication potential; the US pattern had a 31% lower freshwater eutrophication potential than the MED pattern (difference of 6·6 g phosphorusequivalent per week; figure 2, table 2). This difference in freshwater eutrophication potential was almost entirely due to increased seafood recommendations for the MED pattern; the MED pattern's impact from seafood was 6·8 g phosphorus equivalent per week greater than that of the US pattern (appendix), some of which was offset by decreases in dairy and discretionary calorie intake. Notably, although the seafood recommendation in the MED pattern is greater than that for the US pattern, the other protein food recommendations are the same—ie, the extra seafood does not replace other protein sources but adds to them.
The diet pattern with the highest impact in each category is 100% and impacts of other diet patterns are relative to it. MED=healthy Mediterranean-style. US=healthy US-style. VEG=healthy vegetarian.
Table 2. Environmental impacts of the weekly per capita patterns in the Dietary Guidelines for Americans
|Climate change (kg CO2equivalent)||Land use (kg carbon deficit)||Water depletion (m3 water equivalent)||Freshwater eutrophication (g phosphorus equivalent)||Marine eutrophication (g nitrogen equivalent)||Particulate matter or respiratory inorganics (g PM2·5equivalent)|
|2000 kcal per day|
|3000 kcal per day|
The VEG diet pattern had 42–84% lower impacts than the US and MED patterns, with the exception of water depletion (figure 2, table 2); water depletion was the only impact that was similar across all three patterns (maximum 16% difference between the patterns). Furthermore, most impacts of the VEG pattern at 3000 kcal per day were still lower than those of the MED and US patterns at 2000 kcal per day (table 2). The 3000 kcal VEG pattern had lower climate change (16·5 kg CO2 equivalent vs24·7 for MED and 24·8 for US 2000-kcal patterns), freshwater eutrophication (4·7 g phosphorus equivalent vs 21·1 and 14·5), marine eutrophication (104 g nitrogenequivalent vs 224 and 177), and particulate matter (8·3 g PM2·5 equivalent vs 14·3 and 13·9) impacts than the 2000 kcal MED and US patterns. Results for water depletion and land use were similar (<30% difference) across the 3000 kcal VEG and 2000 kcal MED and US patterns. Life cycle impact assessment results of the three diet patterns for all sixteen impact categories are provided in the appendix.
In the analysis of individual food group contributions to each impact, the protein group had the largest proportion contribution across all impacts except water depletion in the US and MED patterns (figure 3A,B). By contrast, protein was the top contributor for only one impact, land use, in the VEG pattern (figure 3C). Protein was a much smaller contributor to all impacts in the VEG pattern, both absolutely and relatively, ranging from 7% of water depletion to 31% of land use. The protein groups of the US and MED patterns had absolute freshwater eutrophication impacts 3·5–5·5 times greater than the entire VEG pattern (11·8 g phosphorus equivalent for the US and 18·6 g for the MED pattern protein impacts vs 3·4 g for the whole VEG diet; appendix). In addition to different protein sources (eggs, nuts, seeds, soy, and beans and peas), the VEG pattern has a much lower protein group recommendation of 24·0 oz equivalents of protein per week than the US pattern recommendation of 39·0 oz and the MED pattern recommendation of 45·5 oz.
(A) Healthy US-style pattern; (B) healthy Mediterranean-style pattern; and (C) healthy vegetarian pattern. Contributions are given as a proportion of overall impact of all food groups (100%).
The proportional impact contributions of dairy, grains, oils, and discretionary calories were higher in the VEG pattern than in the US or MED patterns, largely due to differences in the amount and composition of the protein group in the VEG pattern versus the others. The dairy group was the largest individual contributor to the climate change (42%) and particulate matter (32%) impacts of the VEG pattern; dairy had a less dominant but still substantial role for these impacts in the other patterns. Grains, oils, and discretionary calories were generally low contributors to the overall impacts of patterns.
We analysed the individual contributions of subgroups of the protein group to the impact of the healthy US-style diet pattern relative to their proportion contribution to the overall recommended protein serving by the Dietary Guidelines for Americans (figure 4). Several subgroups had disproportionately large contributions to impacts relative to their serving contribution in the patterns. Seafood, for example, accounted for 21% of protein group servings, but 66% of freshwater eutrophication, 48% of marine eutrophication, and 41% of water depletion impacts (figure 4). Nuts and seeds were also a disproportionately large contributor to water depletion, accounting for 12% of the protein group's servings and 20% of the water depletion impact. At 35% of servings, red meats drove 69% of climate change, 60% of land use, and 67% of particulate matter impacts.
Figure 4. Protein subgroup contributions to servings and environmental impacts in the healthy US-style pattern
Contributions are given as a proportion of overall servings or impact of protein (100%).
Focusing on water depletion, fruits and vegetables had the highest group contributions in all patterns (figure 3). The only difference in serving recommendations for fruits and vegetables across the patterns is a higher fruit recommendation for the MED pattern (17·5 cups equivalent) than for the other two patterns (14·0 cups equivalent). The fruit group's impact was 0·22 m3 water depletion per week for the MED pattern compared with 0·18 m3 for the US and VEG patterns. Within the vegetable group, the starchy and red subgroups account for 60% of servings and 80% of water depletion of that group in all three patterns (appendix).
The data quality assessment focused on fruit, vegetable, and protein food subgroups (appendix). All life cycle assessments with data quality ratings available (65%) were rated good or higher; none were rated as fair or poor. Most FICRCD commodities (65%) were directly matched to life cycle assessment data, and almost half of life cycle assessments (44%) were geographically representative of North America (USA, Canada, and Mexico). Combining all of these factors, commodities with notably high quality include tree nuts, beef, pork, chicken, apples, oranges, bananas, and tomatoes. Each of these commodities account for large proportions of servings in their subgroups and are directly matched to life cycle assessment data, with good or better data quality ratings, and with origins in the USA or a key importer. At the other end of the spectrum, commodities with lower-quality data include seafood and soy protein isolate. The life cycle assessment data for seafood were a farm-gate match, adjusted for processing for a French aquacultured trout. Although trout accounted for only 2% of the seafood subgroup, finfish more broadly comprised 70% of seafood (appendix). For soy protein isolate, the life cycle assessment data were a farm-gate match for US soybeans, adjusted for processing yield. Soy protein isolate accounted for 83% of the soy subgroup servings.
Our analysis shows that the three dietary patterns equally recommended as nutritious by the Dietary Guidelines for Americans11 might have starkly different implications for the environment and other aspects of human health (eg, because of poorer air quality from increases in particulate matter). Of particular note is the variation in recommended quantities and types of protein foods across the patterns and the different impacts that are attributed to them. Despite having about half of the recommended protein food servings of the MED pattern, the VEG pattern still met protein requirements by a substantial margin; the VEG pattern included 155% of the recommended dietary allowance (RDA) for protein for a female aged 19–50 years versus the 194% of the RDA that the MED pattern provides.14 Given the magnitude of impact differences between the healthy diet patterns that are more (US and MED) and less (VEG) reliant on animal-based protein, considering sustainability in the evidence-base for the Dietary Guidelines for Americans protein food recommendations could reduce these disparities. Another area to consider is that the composition of food groups in the US Department of Agriculture patterns are based on healthy versions of commonly consumed foods. Menus or individuals' diets that align with these healthy patterns, but do not follow the distribution of commonly consumed foods, might result in very different impacts. For example, the protein subgroup meats, poultry, and eggs has an aggregate recommendation. Choosing a different distribution of foods than is modelled in the US Department of Agriculture food patterns (eg, more eggs, less red meat) might influence the overall impact markedly, given the influence of protein foods on the majority of impacts for the patterns. Future research should focus on the extent to which different distributions of foods within food groups and subgroups would influence environmental burdens.
Our findings reinforce previous research on greenhouse gas emissions20 and land carrying capacity21 that suggest that the VEG pattern recommended in the Dietary Guidelines for Americans has sustainability benefits compared with the US pattern. Notably, the weekly per capita CO2 equivalent emissions of the US pattern from our analysis were essentially identical to results calculated previously,20 despite differences in methods used. Others have also analysed the environmental impacts of shifts from average US diets to the Dietary Guidelines for Americans (2010 and 2015), with mixed results on net environmental benefits or burdens resulting from the shift.17, 18, 19 Because these analyses focused solely on a generic or US-style pattern, future research should include all three healthy patterns to explore how the different patterns influence multiple environmental outcomes. For previous studies that included water use, increasing consumption (and resultant waste) of fruits and vegetables in a shift to the healthy US-style pattern were substantial drivers of net increases in consumptive water use.17, 19 Combined with our findings that fruits and vegetables are major contributors to water depletion across the three healthy patterns that we assessed, these results suggest that meeting fruit and vegetable needs for healthy diets in the USA is a sustainability challenge. Strategies are needed to increase water-use efficiency and decrease spatial concentration of production, because some major production centres (eg, California's Central Valley) are in water-scarce areas that are increasingly vulnerable under climate change.
Our analysis has several limitations. Although a strength of our approach was relying primarily on the World Food LCA Database, which was developed with consistent and rigorous methods,27 we supplemented these data to find more appropriate matches for some foods. Our final dataset aggregated life cycle assessments with a variety of methods and temporal and geographical scopes. Differences in results, therefore, are not entirely explained by differences between products but also by variability of the underlying data. Additionally, not having access to life cycle inventories of the foods or quantitative uncertainty estimates precluded us from being able to statistically test for significant differences between diet patterns. For our data quality assessment, two areas in particular could be improved to support the evidence-based Dietary Guidelines for Americans and sustainable diet research: seafood and processed soy products. Life cycle assessments of the most commonly consumed seafood products in the USA, done using standardised methods according to best practices33 and incorporated into databases such as Ecoinvent, are urgently needed. Regarding soy products, intensive processing for products like soy protein isolate might substantially increase life cycle impacts in some categories, such as climate change.34 Life cycle assessments of processed soy products are needed to fully understand the extent to which this might influence the impact of the soy protein subgroup, particularly in the VEG pattern.
Additional limitations are spatial resolution and system boundaries. Even when life cycle assessment and consumption data were well aligned at the country level, regional impacts (eg, water depletion and eutrophication) probably vary greatly across large countries. This level of resolution is not accounted for in our analysis, mirroring the broader challenge of increasing spatial specificity for impact assessment of foods.35 With system boundaries, our results do not include full life cycle impacts, which would include additional variables such as food waste along the supply chain, packaging, distribution, and food preparation. Given the knowledge that variables like food waste might be worse in healthy diets,36expanded boundaries will be important for future studies that compare the impact of current diets to the healthy patterns recommended by the Dietary Guidelines for Americans.
To promote long-term food security, policies must consider outcomes related to human health and nutritional status, as well as environmental, social, and economic determinants of human wellbeing into the future. This endeavour is no small challenge; new interdisciplinary and transdisciplinary processes, dialogues, and collaborations are needed to develop sustainable dietary guidance.
Alterations in intestinal microbiota composition are associated with several chronic ...
Nutritional genomics promises to make diets truly personal. When Charlotte ...
It’s been a big week on the longevity front: First ...
A new research study on protein overfeeding is causing a ...
Low carbohydrate diets are again in the spotlight and have ...
Through experience, the nematode worm Caenorhabditis elegans learns to distinguish ...
Our bodies are rife with genes and the majority of ...
Many diets and nutritional advice are circulating, often based on ...