Climate change is predicted to exacerbate many of the major public health problems facing the world today( Reference Krueger, Biedrzycki and Hoverter 1 ) and has been linked to increasing anthropogenic greenhouse gas emissions (GHGE)( Reference Solomon, Plattner and Knutti 2 ). Although estimates of the environmental impact of the food system vary based on methodological considerations, in developed countries it has been estimated that between 15 and 28 % of all GHGE are due to patterns of food consumption( Reference Garnett 3 ). The food system is also thought to be the primary global driver of the increasing double burden of obesity and malnutrition( Reference Godfray, Beddington and Crute 4 ).
Currently, strategies to combat the rise of diet-related disease in the developed world focus on manipulating the food environment to increase purchasing of ‘healthy’ foods and decrease purchasing of ‘unhealthy’ foods. Examples of interventions include food labelling( Reference Rayner 5 ), food taxes( Reference Vallgårda, Holm and Jensen 6 ), advertising restrictions( Reference Scarborough, Payne and Agu 7 ) and food reformulation( Reference Wyness, Butriss and Stanner 8 ).
Similar manipulations of the food environment could be applied to promote consumption of foods associated with lower GHGE, in order to lead to more sustainable dietary patterns( Reference Jeffries 9 ). It has been suggested that many of the nutrients currently associated with health risks are also high in foods associated with high environmental impact and therefore switching to a ‘healthy’ diet may also contribute to reductions in GHGE( Reference Green, Milner and Dangour 10 ).
Recently, three reviews have been published that summarize current literature on the relationship between the health and environmental impacts of dietary patterns. Joyce et al. found that results were inconsistent, but that higher levels of plant and plant-based foods were generally associated with both positive health outcomes and diet-related GHGE( Reference Joyce, Hallett and Hannelly 11 ). Hallstrom et al. had similarly mixed results, finding that dietary changes proposed specifically for health reasons had the potential to reduce up to 50 % of diet-related GHGE( Reference Hallström, Carlsson-Kanyama and Börjesson 12 ). Auestad and Fulgoni included studies that addressed other aspects of environmental impact as well as GHGE and also found overall heterogeneity in the results of studies( Reference Auestad and Fulgoni 13 ). Thus we undertook a review that aimed to summarize the findings in the literature and combine the results to allow for an assessment of the weight of evidence regarding healthy, sustainable diets.
Objectives
The objective of the present literature review was to identify studies that have reported changes in GHGE and changes in either nutritional components or health outcomes as a result of variation in actual or modelled consumption patterns, and to use these data to answer the following research questions:
-
1. Are low-GHGE dietary patterns associated with reduced CVD and cancer incidence or mortality?
-
2. Are low-GHGE dietary patterns associated with reduced saturated fat, sugar or salt consumption?
-
3. Are low-GHGE dietary patterns associated with increased micronutrient consumption?
Additionally, we used the results to assess whether associations between GHGE and health/nutrition outcomes vary by level of meat consumption in the low-GHGE dietary patterns.
Methods
We used several sources to search peer-reviewed, published literature reporting the results of the following types of study: (i) prospective cohort studies, in which baseline diet groups are defined by dietary GHGE; (ii) cross-sectional dietary surveys (e.g. national surveys of patterns of food consumption; baseline surveys for observational or experimental studies), which use cluster analysis to identify dietary patterns with differing levels of GHGE; and (iii) cross-sectional nutritional surveys that are used to model potential diets with differing levels of GHGE. We included only studies for which quantitative data on GHGE and dietary nutritional content or health outcomes were available.
We searched the reference lists of the three previously published reviews on this topic( Reference Joyce, Hallett and Hannelly 11 – Reference Auestad and Fulgoni 13 ). We also searched the citations of these articles using Google Scholar.
We searched PubMed, Web of Science and Scopus in October 2015 using the following search strategy:
(GHG* OR greenhouse gas emission* OR environmental impact OR sustainab*) AND (diet*) AND (nutrition* OR health*)
This search was applied to the fields Title (PubMed), Article title/abstract/keywords (Scopus) and Title (Web of Science). In PubMed, we limited our search to journal articles focusing on human populations between 2000 and present. In Scopus, we limited our search to include journal articles only. We limited our results to English-language publications with medium to high relevance to the query and to those with a publication date between 2005 and present.
We searched the Food Climate Research Network (FCRN) research library (http://www.fcrn.org.uk/research-library) in May 2015 using the keyword ‘sustainable healthy diets’.
We scanned the title and abstract of all studies selected by these four searches to determine their suitability for inclusion in the present review. We included studies that:
-
1. evaluated the diet as a whole (excluding alcohol);
-
2. evaluated actual or modelled diets, including those based on dietary recommendations; and
-
3. included quantitative data expressing the GHGE and nutrient content and/or health outcomes of diets.
We excluded studies that:
-
1. were not articles published in peer-reviewed journals;
-
2. evaluated only separate food groups within a diet;
-
3. did not enable a direct matched comparison (e.g. reported GHGE for individual food groups within a diet and nutritional data for the diet as a whole); and
-
4. used a measure other than GHGE to estimate environmental impact, without giving GHGE data.
We extracted data on macro- and micronutrients that have been associated with ‘healthy’ or ‘unhealthy’ diets. Sugar, salt and saturated fat are the three nutrients identified by the WHO as important to limit in order to achieve a healthy diet associated with reduced non-communicable disease( 14 ). The WHO also recommends reductions in the consumption of trans fats, which were not included in the current analysis as trans fat reductions are generally achieved at the industrial stage rather than by changes in diet. Although many micronutrients are considered important for health, previous research has shown that meat production is responsible for a large proportion of dietary carbon emissions( Reference Caro, Davis and Bastianoni 15 ); therefore we extracted data on micronutrients that have been highlighted as potentially deficient in a low-meat diet( Reference Millward and Garnett 16 ).
We used a modified QUADAS-2 tool( Reference Whiting, Rutjes and Westwood 17 ) to assess the risk of bias in individual studies. The details of the questions developed for this tool are shown in Table 1.
Table 1 The modified QUADAS-2 tool used to identify risk of bias within individual studies and to inform the presentation of data and analysis of results
GHGE, greenhouse gas emissions.
We deemed the results inappropriate for a formal meta-analysis due to the heterogeneity of study designs and dietary scenarios that were compared. We grouped the studies based on the health outcome and/or nutrient being measured and conducted a sign test to evaluate whether each dietary comparison answered the three research questions positively or negatively.
The full protocol for this search was published on Prospero (registration number CRD42015027360).
Results
Description of included studies
Figure 1 illustrates the results of our search strategy. We screened the title and abstracts of 2027 entries from four databases (FCRN Research Library, PubMed, Web of Science and Scopus) and three reference lists. From these, we identified sixteen studies( Reference Biesbroek, Bueno-de-Mesquita and Peeters 18 – Reference Wilson, Nghiem and Ni Mhurchu 33 ) that present data linking GHGE and nutritional information or health outcomes of 100 dietary patterns and yielding data on 346 cases of effects on nutritional quality or health outcomes of reduced-GHGE diets.
Fig. 1 The search process employed in the present review (GHGE, greenhouse gas emissions; FCRN, Food Climate Research Network)
Table 2 details the studies included in the current review. Five studies used cluster analysis of actual diets and the remaining eleven studies used dietary modelling. The direction of the analysis, and the populations and markets represented by the dietary and GHGE data were identified by the QUADAS-2 tool to be potential sources of bias. Only one study was judged not to be at risk of bias, since it used aggregated data from populations and markets worldwide( Reference Tilman and Clark 21 ). All others used data from a single country (n 14) or continent (n 1), which may produce results that are not necessarily relevant in a global context.
Table 2 Details of the sixteen studies included in the present review, including aspects identified as presenting a potential risk of bias, and the results of each study
GHGE, greenhouse gas emissions; EPIC, European Prospective Investigation into Cancer and Nutrition; NZANS, New Zealand Adult Nutrition Survey; LCA, life cycle analysis; Vit, vitamin.
* A ‘+’ sign indicates that the study found a positive association between the ‘healthiness’ and GHGE of a diet; a ‘–’ sign indicates that the study found a negative association (a non-directional, neutral relationship between GHGE and ‘healthiness’ was also included in this). The number of +/−signs indicates the number of diets, not the strength of the association.
† A positive association was considered to be an association between reduced GHGE and reduced quantity of these macronutrients in the diet.
‡ A positive association was considered to be an association between reduced GHGE and increased quantity of these micronutrients in the diet.
§ A positive association was considered to be an association between reduced GHGE and reduced number of/risk of deaths, as represented by each diet.
Six studies were also judged to be at risk of further bias due to incomplete information regarding the source of GHGE data (n 4) or to GHGE data sourced from a market that was not geographically congruent with the source of dietary data (n 2). Finally, the direction of the studies – whether they chose or grouped diets according to GHGE, nutritional indices or other measures – may introduce bias. The majority of studies in the review used nutritional criteria to group or model diets (n 12); others used GHGE (n 1), price (n 1) or a combination of criteria (n 3).
In total, the results of the review yield data on 334 dietary links between GHGE and health measures, from 100 dietary scenarios reported in sixteen studies. The number of diets evaluated in each study ranged from two to sixteen.
Table 2 also summarizes the results of each study. Five studies reported on health outcomes, thirteen studies reported on macronutrient content of diets and six studies reported on micronutrient content of diets. Overall, 64 % (n 214) of dietary patterns showed an association between reduced GHGE from diets and worse health indicators.
Strength of associations between greenhouse gas emissions of diets and health indicators
The units representing both GHGE and health indicators were heterogeneous between studies, as was the reporting of data for a baseline and/or contemporary diet. This meant that the studies are not directly comparable in terms of reporting the strength of association between GHGE and health indicators of dietary patterns, only the direction. Therefore we represent the ‘strength’ of associations reported in different studies by acknowledging the number of dietary patterns that corresponded to a positive or negative association between reduced GHGE and improved health indicators (Table 2).
Association between greenhouse gas emissions of diets and health outcomes
Five studies reported the GHGE and health outcomes of diets( Reference Biesbroek, Bueno-de-Mesquita and Peeters 18 , Reference Soret, Mejia and Batech 20 , Reference Tilman and Clark 21 , Reference Briggs, Kehlbacher and Tiffin 24 , Reference Milner, Green and Dangour 28 ). Of these, two were based on reduced meat and/or dairy consumption and found consistent association between reduced GHGE and positive health outcomes( Reference Soret, Mejia and Batech 20 , Reference Tilman and Clark 21 ). However, the remaining three studies( Reference Biesbroek, Bueno-de-Mesquita and Peeters 18 , Reference Briggs, Kehlbacher and Tiffin 24 , Reference Milner, Green and Dangour 28 ) looked at diets based on cost, emissions and a combination of nutrition and GHGE, respectively, and found either an inconsistent relationship between GHGE and health outcomes or that low-GHGE diets were associated with increased risk from non-communicable disease. Of the twenty-five dietary scenarios compared, twelve showed positive health outcomes for the lower-GHGE diets (see online supplementary material, Table S1).
Association between greenhouse gas emissions of diets and macronutrient content
Thirteen studies reported the GHGE and macronutrient content of diets( Reference Scarborough, Appleby and Mizdrak 19 , Reference Vieux, Soler and Touazi 22 – Reference Wilson, Nghiem and Ni Mhurchu 33 ). Six studies, five of which were modelled based on health or meat content( Reference Hallström, Röös and Börjesson 25 , Reference Hendrie, Ridoutt and Wiedmann 26 , Reference Temme, Bakker and Brosens 30 – Reference Wilson, Nghiem and Ni Mhurchu 33 ) and one on tax scenarios( Reference Briggs, Kehlbacher and Tiffin 24 ), found that diets with lower GHGE also had a consistently lower saturated fat content. While no study found a consistent relationship between lower GHGE and lower salt or sugar content of the diets that they evaluated, four studies found that lower-GHGE diets were consistently associated with elevated levels of salt( Reference Biesbroek, Bueno-de-Mesquita and Peeters 18 , Reference van Dooren, Marinussen and Blonk 32 ) and sugar( Reference Scarborough, Appleby and Mizdrak 19 , Reference Vieux, Soler and Touazi 22 , Reference van Dooren, Marinussen and Blonk 32 ). A further four studies found that while some diets with lower GHGE had a reduced sugar or salt content, other diets within the same study were high in one or both of these nutrients( Reference Vieux, Soler and Touazi 22 , Reference Berners-Lee, Hoolohan and Cammack 23 , Reference Hoolohan, Berners-Lee and McKinstry-West 27 – Reference Meier and Christen 29 , Reference Wilson, Nghiem and Ni Mhurchu 33 ). Of 151 dietary comparisons across all studies, seventy-nine showed reduced levels of nutrients to limit for lower-GHGE diets, including twenty-seven out of forty-one for saturated fat, thirty-five out of fifty-five for salt, and seventeen out of fifty-five for sugar (see online supplementary material, Table S2).
Association between greenhouse gas emissions of diets and micronutrient content
Six studies reported the GHGE and micronutrient content of diets. Three studies found reduced micronutrient content in lower-GHGE diets for every micronutrient considered( Reference Biesbroek, Bueno-de-Mesquita and Peeters 18 , Reference Briggs, Kehlbacher and Tiffin 24 , Reference Hallström, Röös and Börjesson 25 ). Three studies( Reference Vieux, Soler and Touazi 22 , Reference Milner, Green and Dangour 28 , Reference Wilson, Nghiem and Ni Mhurchu 33 ) found that in some lower-GHGE diets, Fe, Zn, vitamin A (thirty-three only) and vitamin B12 content were higher than in a higher-GHGE diet. Overall, of 158 dietary comparisons, 129 showed lower levels of essential micronutrients in lower-GHGE diets (see online supplementary material, Table S3).
Discussion
The current review presents an overview of published quantitative data that indicate whether there is an association between the health and environmental impact of actual and modelled dietary patterns.
We found highly inconsistent results regarding links between reduced GHGE and reduced content of nutrients to limit. In the cases of salt and saturated fat, the majority of dietary patterns found a reduction in levels of these nutrients in diets with reduced GHGE (twenty-nine of forty-three diets for salt, twenty-seven of forty-one diets for saturated fat). This may be due to the reduction in meat consumption in lower-GHGE dietary patterns. Of the twelve studies that reported salt and saturated fat content of diets, eight analysed diets with reduced levels of meat and dairy. However, the majority of dietary patterns that reported sugar intake showed increased sugar in lower-GHGE diets (thirty-eight of fifty-five diets). The reasons for this are unclear.
We also found an inconsistent relationship between reduced GHGE and positive health outcomes. Overall mortality was associated with four higher-GHGE diets and four lower-GHGE diets. Cancer was similarly inconsistent, with increased incidence in five reduced-GHGE diets and decreased incidence in seven reduced-GHGE diets. CVD risk was assessed by only two studies, and within these this disease was higher in low-GHGE diets in four out of five scenarios.
However, we did find that decreased micronutrient content of diets was more frequently associated with reduced GHGE than increases in micronutrient content, with far more cases in which lower-GHGE diets were associated with decreases in key micronutrients (n 129) than with increases in micronutrients (n 29). In particular, no study reported an increase in dietary Ca (thirty-nine showed decreased Ca), and only one study reported an increase in dietary vitamin B12 (twenty showed decreased vitamin B12), with reduced-GHGE diets. Similarly to the negative associations discussed between key macronutrients, this is likely to be due to the decreased consumption of animal and particularly dairy products in diets with lower GHGE. Of the seven studies that reported micronutrient content of diets, three analysed diets with reduced levels of meat and dairy.
Fe was the only micronutrient for which dietary associations between an increase in micronutrient level and a reduction of GHGE were very slightly predominant (twenty of thirty-eight diets), but the reasons for this are unclear as the studies for which this is the case did not report Fe content of separate food groups( Reference Milner, Green and Dangour 28 , Reference Temme, Bakker and Brosens 30 , Reference Wilson, Nghiem and Ni Mhurchu 33 ). Zn levels, in contrast, were lower in reduced-GHGE diets in twenty-five of twenty-nine scenarios.
These results clearly indicate micronutrients as a key area for concern in advocating reduced-GHGE diets. Health problems related to micronutrient deficiency are currently increasing worldwide( Reference Black 34 ) and future research and policy should take measures to ensure that environmentally focused dietary recommendations do not neglect micronutrient recommendations.
For all associations, there were no clear differences between studies using cluster analysis and those using dietary modelling; instead, both study designs yield comparably heterogeneous results.
There are two main limitations of the present review. The first is the unrepresentative nature of published data. All studies use dietary data from countries in which the majority of the population is likely to follow a Western diet, although one study did include several ‘Asian’ diets( Reference Wilson, Nghiem and Ni Mhurchu 33 ). The majority of studies also model diets based on their perceived ‘healthiness’ according to nutritional criteria, thus biasing the data towards healthier dietary patterns. Future studies could eliminate the nature of this bias by modelling diets based on known patterns of dietary change resulting from health interventions.
Three review articles that ask similar a question to the current study have been published recently( Reference Joyce, Hallett and Hannelly 11 – Reference Auestad and Fulgoni 13 ). Both Joyce et al.( Reference Joyce, Hallett and Hannelly 11 ) and Hallstrom et al.( Reference Hallström, Carlsson-Kanyama and Börjesson 12 ) reviewed the evidence for the environmental impacts of dietary change, finding an inconsistency in associations between health and GHGE of dietary patterns and heterogeneity in study designs. Auestad and Fulgoni( Reference Auestad and Fulgoni 13 ) reviewed the literature linking dietary patterns to all aspects of the environmental impact of diets, and to economics, and found that heterogeneity in study designs and data sources prevented useful data assimilation. In the current study we asked whether diets with reduced carbon emissions are conducive to good health. Similarly to Auestad and Fulgoni, we also found that heterogeneity of data sources meant we could not assimilate data in order to assess the strength of associations, but we did find that results were highly inconsistent even when grouped by study design. We also found inconsistencies between markers of the ‘healthiness’ of diets and reduced GHGE.
We presented and discussed the results of our review based on the health measures considered in each study: health outcomes, nutrients to limit and essential micronutrients. This revealed that while lower-GHGE patterns show no consistent relationship with reduced nutrients to limit or positive health outcomes, many do show a correlation with reductions in micronutrient intake and with elevated sugar levels. We conclude that the maintenance of sufficient micronutrients in the diet, and the control of sugar levels, should be key concerns for policy makers advocating environmentally sustainable diets.
Acknowledgements
Financial support: This review was funded by the Department for the Environment, Food and Rural Affairs (grant number FO0318) and the Biotechnology and Biological Sciences Research Council as a contribution to the European Commission SUSFOOD funding call (proposal number 114: SUSDIET). The funders had no role in the design, analysis or writing of this article. Conflict of interest: None. Authorship: P.S. and L.C. formulated the research questions and study design. C.L.R.P. carried out the study, analysed the data and wrote the article. Ethics of human subject participation: Not applicable.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1368980016000495
