Introduction

Non-communicable diseases (NCDs) now exceed 34 million deaths annually (65% of global deaths). NCDs, particularly cardiovascular diseases, common cancers, respiratory diseases, diabetes and dementia, are a common cause of disability.1

Four major risk factors generate the majority of NCDs. Of these, poor diet makes the biggest contribution, larger than tobacco, alcohol and physical inactivity combined.2 Globally, increased consumption of processed food containing excess amounts of sugar, salt, saturated and trans fats are compounded by low intakes of healthy foods like fruit and vegetables, whole grains, nuts, pulses and seafood.

Hence a key policy question is: how can diets be improved to prevent NCDs? Two options are often discussed: ‘downstream’ behavioural interventions focusing on individuals or, the focus of this review, ‘upstream’ policy interventions that aim to promote healthier environments.3 Worldwide, many different ‘upstream’ policy options have been proposed to promote healthier eating and prevent diet-related NCDs. These include targeting price, promotion, provision, composition, labelling, supply chain, trade and investment, or a combination thereof.4 However, these options are broad and potentially confusing, with continuing uncertainties about which interventions are most effective. To inform their decisions, policy makers need clear evidence on the potential population benefits (as well as the ease and cost of implementation).

This scoping review, therefore, collates the current evidence on policy interventions to promote healthier dietary consumption.

Methods

Search strategy

We first identified exemplar studies of dietary interventions, which were used to define and refine search terms. This process identified interventions targeting price, promotion, provision, composition, labelling, supply chain, trade and investment.4 This search strategy was piloted for price interventions to determine appropriate databases for identifying all relevant studies and was adapted for the remaining interventions.

The search strategy consisted of a combination of two sets of keywords. The first one focused on dietary components such as salt, fruit, vegetables, fat, trans fat, sugar, diet or equivalents of these terms. The second search string focused on policy interventions including taxes, subsidies, reformulation, elimination, labelling, campaigns, marketing, advertisement, workplace, schools, community, food system, food supply chain, food trade, food investment, legislation, regulation or equivalents of these terms.

A total of six databases were searched: MEDLINE; Science Citation Index, SCOPUS, Cochrane Database of Systematic Reviews, The Campbell Collaboration Library of Systematic Reviews and the CRD Wider Public Health database. These were identified during the pilot searches as the most appropriate databases to generate relevant papers on this specific topic. Reference lists of included papers were also scanned for potential eligible reviews and systematic reviews (SR).

Selection of studies and inclusion criteria

Systematic and non-SRs addressing policy interventions to improve healthy eating and to prevent NCDs, with a quantitative outcome, dating from January 2004 onwards were included. A PICOS approach (participants, interventions, comparators, outcomes and study design) was used to assess the retrieved studies (Table 1). Only reviews in English were considered.

Table 1 Inclusion/exclusion criteria
Full size table

One researcher (LH) conducted the searches and removed duplicates. All remaining titles and abstracts were screened independently by two researchers (LH and MA), based on the inclusion/exclusion criteria (Table 1). All papers deemed potentially eligible were retrieved in full-text form and again screened independently by two researchers. Discrepancies on whether to include or exclude reviews were resolved by consensus or by seeking the opinion of a senior team member (SC).

Data extraction

Both empirical studies (presenting primary data) and modelling studies (using secondary data to project health outcomes) were included. However, the results were analysed and discussed separately, recognising their different natures.

Data extraction was done by one researcher using pre-designed and pre-piloted forms (Table 2 and Supplementary file 1). Extracted data was independently verified by another researcher to ensure accuracy and completeness. Owing to logistical and time constraints, it was not possible to contact study authors regarding any unclear, missing or additional data.

Table 2 Summary tables
Full size table

Data categorisation

Evidence from the reviews was categorised using a seven-part framework, which was developed and refined from the ‘4Ps’ marketing framework:4 price, place, promotion and product. In social marketing these 4Ps are key in planning and implementing strategies and this translates well to a policy context. In the social marketing mix, product stands for the desired behaviour and what people will gain when they perform the behaviour, whereas in the policy context this refers to the actual product (for example, foods). Price refers to the cost exchanged for the promised benefits and place refers to the location where the goods are distributed. Finally, promotion refers to a persuasive communication strategy to influence change and can involve multiple channels and target groups. After extensive piloting the following categories were used:

  • Food price—policies influencing prices through taxes, subsidies or economic incentives;

  • Food promotion—advertising/marketing; particularly on children; media campaigns and health education;

  • Food provision—in specific settings: schools, communities or workplaces;

  • Food composition—reformulation or elimination;

  • Food labelling—nutrition labelling, calorie labelling in stores/restaurants;

  • Food supply chain, trade and investment—including legislation or regulation affecting production policies or supply-chain logistics;

  • Multi-component interventions (including at least two of the categories described above).

Data analysis

We used a narrative synthesis to summarise the evidence of policy interventions to promote healthy eating patterns.55 First, the data identified and included in this review were extracted in tables and grouped based on their type of intervention. The different types of interventions included are discussed under data categorisation and are adapted from the 4P’s marketing framework. Most interventions have subcategories, for example, food provision, which can be done in schools, communities or at workplaces. Within these categories the data were grouped according to the subcategories. Furthermore, as this review is about healthy eating, multiple nutrients are included. So another layer was added for these nutrients and the results were presented under their specific (sub-) categories for each of the nutrients. This structure was then used to present the results.

All the evidence for each sub-category was reviewed and analysed according to study design and the number of studies included investigating the same intervention and outcome. The key findings are presented in the results section and all the studies are then summarised in the tables.

As the study designs and evidence per nutrition varies greatly no quantitative comparison will be made between the interventions. They will each be presented separately.

Results

About 1805 candidate publications were screened and 197 papers were retrieved for full-text screening. In all 139 were excluded as they had no quantitative outcomes or outcomes different to our inclusion criteria. We identified 58 systematic and non-SRs for inclusion (Figure 1). Eight reviews provided information on more than one category.

Figure 1
figure 1

PRISMA flow diagram.

Full size image

Food price

Taxes

One SR estimated a mean price (Table 2a) elasticity of −0.8 to −1.0 for sugar-sweetened beverages (SSBs). Thus, a 10% price increase is associated with an 8–10% reduction in SSB consumption.16 Furthermore, Thow et al.22 reported a 5–48% reduction in SSB consumption proportional to the modelled taxes applied (5–30%).

However, taxing unhealthy food may cause substitution effects. For instance, a fat tax might reduce saturated fat content but increase salt content and product consumption.23 Four modelling studies included in Thow et al.22 investigated these substitution effects. A 5–20% SSB tax could reduce calorie intake of SSBs by 10–48% in adults and by 5–8% in children, however consumption of milk, tea, coffee and low-calorie beverages may increase correspondingly.22 Finally, Welsh et al.7 included a modelling study suggesting that a US penny-per-ounce tax might reduce SSB consumption by 15% among adults and subsequently prevent ~95 000 coronary heart events, ~8000 strokes and ~26 000 premature deaths.

Several reviews included at least one modelling study investigating a ‘fat tax’.9, 10, 19, 22 All suggested that changes in fat consumption would be minor, unless the tax was substantial. Four modelling studies in Thow et al.22 suggested that a small fat tax (5–17%) might reduce (saturated) fat consumption by 0–3% and induced substitution with lower-fat options. Increasing the price of salty foods by 40% was suggested to reduce sodium consumption by ~6%.22

Mozaffarian et al.20 included a modelling study investigating a combination of taxing less healthy foods and using the $580 million revenue raised to subsidise fruit and vegetables. This might alter consumption to prevent ~6000 CVD and cancer deaths annually.

Subsidies

Powell et al. included six modelling studies reporting −0.49 mean price elasticity for fruit and vegetables, suggesting a 20% subsidy on fruit and vegetables might increase consumption by 10%,18 while a 10% subsidy might increase consumption by 5%.22 A continuing price reduction on fruit and vegetables might prevent ~6700 cases of CHD and 2950 cases of ischaemic stroke.21

Food promotion

Marketing and advertising

Four randomised controlled trials (RCTs) in Gregori et al.25 (Table 2b) indicated that exposure to food advertisements compared with non-food advertisements increased calorie intake and snacks consumed. The effect appeared largest in obese children. Two cross-sectional studies by Boyland and Halford26 supported these findings. Conversely, banning food advertisements on television might reduce the US baseline rate of childhood obesity by 2.5–6.5%.27

Nutrition education

RCTs reviewed in two SRs suggested that nutrition education focused on fruit and vegetables could increase their intake by 0.4–1.4 servings per day.28, 30 Two additional RCTs reported a net increase of 0.4 servings per day.6

A meta-analysis of RCTs found that e-learning interventions focused on dietary intake increased fruit and vegetable intake by 0.24 servings per day; reduced total fat (−0.8 g), saturated fat (−0.2 g) and daily energy intake (4 kcal).29

Media campaigns

Three reviews focused on fruit and vegetables.31, 32, 33 Benefits were modest at best.

Food provision in schools, communities or workplaces

Schools

A cross-sectional and longitudinal study included in Levy et al.6 (Table 2c) investigated the effect of eliminating access to SSBs on consumption and found reductions of 0.16 servings per day and 4%, respectively. Another time-series study found a 35% decline in consumption in schools after SSBs were removed from cafeteria vending machines. However, pupils obtained SSBs from home or non-cafeteria vending machines instead.6 A modelling study by Patel and Cabana36 suggested replacing SSBs with water might reduce intake by 235 kcal per day among 2–19 year olds.

Jaime and Lock38 reviewed the effect of a free piece of fruit in schools in three RCTs. Fruit and vegetable intake increased by 0.30–0.44 servings per day. However, benefits were not necessarily sustained long term.38 Fat intake was targeted by one RCT and two non-randomised studies in the same review, reporting decreases in saturated fat (0.9–5.2% of energy intake) and total fat (2.0–10.9%).38 A meta-analysis of school interventions targeting fruit and vegetable intake found a higher effect when fruit juice was included (0.32 portions) compared with no fruit juice included (0.25 portions).37

Communities

No reviews were found for community interventions and the effect on dietary intake or health outcomes.

Workplaces

Thompson and Ravia30 reviewed RCTs of work-based interventions and reported 0.8 servings per day increase in fruit and vegetable intake. A meta-analysis of 36 RCTs also found a significant increase in daily fruit and vegetable consumption.35 Five RCTs in Mhurchu et al.34 focused on total fat intake as a percentage of energy and found reductions in both the intervention (−2.2 to −9.1%) and control groups (+1.3 to −1.8%).

Food composition

Trans fats

Downs et al.42 included two pre–post-test studies evaluating (Table 2d) the effectiveness of interventions aimed to reduce trans fatty acids (TFAs) in food. In Denmark average iTFA intake decreased from 4.5 g per day in 1976 to 1.5 g per day in 1995 and reduced to virtually zero after the ban in 2005.34 In The Netherlands, voluntary reformulation of TFAs decreased daily energy intake from 1.0 to 0.8% (a 20% reduction).42

Mozaffarian and Clarke43 conducted two meta-analyses with 13 RCTs and 4 prospective studies to calculate the effect of reduced TFAs in partially hydrogenated vegetable oils by means of reformulation. The predicted risk reductions were greatest when replacing at least half of the iTFA in partially hydrogenated vegetable oils, with soybean oil or canola oil.43

Food labelling

Menu labelling

Two SRs evaluated the effect of menu labelling on dietary intake (Table 2e). Swartz et al.46 included two non-blind RCTs that reported no significant impact of menu labelling on calorie consumption. However, when combining the labels and labels/information categories vs no label, the labelled condition consumed 180 fewer calories.46 Similar results were found by a meta-analysis of 17 studies conducted by Sinclair et al.45

Nutrition labelling

Observational studies in Campos et al.47 reported an association between nutrition label use and fat consumption. Furthermore, experimental research included showed that participants tended to consume more energy reduced foods in terms of weight, but the overall energy intake was significantly lower.47 The theoretical concern that labelling might lead to overconsumption of foods perceived as ‘less unhealthy’ and appeared to be unfounded in a cross-sectional study by Hawley et al.49

Food supply chain, trade and investment

Trade and market liberalisation

Evidence for the effect of trade (Table 2f) market liberalisation on dietary intake is contested. Traill suggested only ‘modest’ effects,52 whereas a study included in Friel et al.51 argues that liberalisation of Foreign Direct Investment through trade agreements with the USA significantly increased soft drink consumption within low- and middle-income countries.

Monetary subsidies and taxes

Thow et al.21 included two modelling studies; one suggested that increasing value-added tax in the UK on the main sources of saturated fat might reduce consumption and thereby decrease CHD deaths by 1.8–2.6%. The other suggested that halving value-added tax on fruit and vegetables could decrease consumption of sugar (−6.5%), fat (−2.5%) and saturated fat (−3.6%).21

Reformulation

Livingstone et al.44 reviewed reformulation in the supply chain by altering the diet of cows and investigated the effect on milk and dairy products. A double-blinded randomised cross-over intervention study found a 16% decrease in saturated fats and increase in monounsaturated fats (+10%) and polyunsaturated fats (+7.5%). A meta-analysis of RCTs suggested that replacing 10% of dietary saturated fat intake with polyunsaturated fats might reduce CHD risk by 10%.44

Multi-component interventions

Trans fats

Voluntary limits combined with mandatory labelling (Table 2g) were associated with a 30% reduction in dietary trans fat intake in the general population, in an interrupted time-series study in Downs et al.42

Fruit and vegetables

Fifteen intervention studies in Blanchette and Burg54 combined provision and promotion activities, and found an increase in fruit and vegetable intake of 2.54 servings per day. Conversely, single interventions rarely increased intake by more than 0.5 servings.38, 39, 40

Salt

No reviews have evaluated the effects of multi-component interventions on salt intake. However, the UK salt programme has been evaluated and a 1.5 g per day (15%) reduction was observed between 2001 and 2011. This was achieved by sustained and progressive industry reformulation involving close monitoring and political pressure to ensure compliance, reinforced by media campaigns and traffic light food labelling.56, 57

Discussion

Main findings

Our rapid scoping review suggests that ‘upstream’ interventions such as price interventions appear to be consistently effective in improving healthy eating.9, 16, 18, 19, 20, 21, 25 Multi-component interventions and reformulation also appear effective in promoting healthy diets. However, the effectiveness of the remaining policy interventions (labelling, restrictions on the provision or marketing of unhealthy food) generally demonstrated smaller effects and less certain long-term benefits.29, 36

Specific findings

Food price

Taxes and subsidies appear to be consistently effective at reducing consumption of sugary drinks,16, 22 and increasing the consumption of fruit and vegetables.18, 21, 22 Changes in consumption demonstrated price elasticity, thus the larger the taxes or subsidies, the greater the effect.16, 22 A fat tax, however, appeared to have smaller impacts on consumption,9, 19, 22 and potential substitution effects need to be anticipated and negated.22, 23

Food promotion

Junk food advertisements significantly increase total calorie intake. Banning advertising to children therefore appears to offer an effective intervention to decrease intake and potentially reduce obesity levels.25, 26 Health promotion campaigns31, 32, 33 and nutrition education6, 28, 29, 30 showed small effect sizes and benefits may diminish over time. More studies are needed to determine long-term effects.

Food provision

Most school-based interventions focused on fruit and vegetables37, 38 or access to SSBs.6, 36 Overall, these interventions seemed to be effective in the short term but effect sizes are generally small and longer-term effects were not routinely investigated.

The majority of workplace interventions focused on fruit and vegetable intake,30, 35 with a few focusing on fat intake,34 and effects were generally modest.

Food composition

Mandatory reformulation to eliminate trans fat was very effective in Denmark, reducing consumption far more than voluntary reformulation. This is in line with a modelling study comparing voluntary with mandatory salt reformulation policies.58 Furthermore, Cappuccio et al.59 suggested that nation-wide reformulation policies might reduce salt consumption by 1 g per day.

Food labelling

Although menu labelling appears weak or ineffective, food labelling may help consumers to choose and purchase healthier options,46 while also exerting pressure on manufacturers to reformulate.60 Furthermore, although 50% of EU consumers might use food labels; many report difficulties in interpreting the labels.47, 61 However, the current evidence is based on diverse methodologies and remains limited.62

Food supply chain, trade and investment

Food supply chain interventions appear intuitively powerful, but are under-researched. Interventions could take place at many levels of the food supply chain, with positive (or negative) effects on food availability, price, quality or marketing.51, 52, 53 Targeted reformulation, subsidies and taxes in the food supply chain could have beneficial effects on diet-related health outcomes.44

Multi-component interventions

Multi-component interventions appear to be highly effective at reducing the dietary intake of salt, industrial trans fats and increasing fruit and vegetable consumption, thus improving health outcomes at population level.56, 57 However, salt intake remains very high in most countries; often double the World Health Organisation recommended maximum levels of 5 g per day.63

Likewise, industrial trans fat levels have been slashed in Denmark, the USA and parts of Western Europe but remain high elsewhere.42 Furthermore, the reviews included only considered trans fat and fruit and vegetable intake. It is, therefore, not clear whether multi-component interventions might be equally effective for other dietary components such as sugar and total fat.

Strengths of this analysis

This scoping review used two independent researchers to screen all the papers generated through the searches. Discrepancies were resolved through consensus or consulting a third senior researcher. Furthermore, a systematic approach was used based on the seven-item interventions framework.

Food supply chain, trade and investment is an underexplored area; however there is some evidence to suggest that globalisation and liberalisation of markets can influence dietary patterns and NCDs. Multi-component interventions are as suggested effective to reduce dietary intake and NCDs. Countries have implemented several policies at the same time and it is important to quantify the impact compared with single policies.

Limitations of this analysis

This scoping review has several limitations. First, we only searched for reviews from the last 10 years. However, we anticipate this captured majority of relevant reviews and thus provide an excellent overview of current food policy interventions. Second, only reviews with an abstract available in English were included. Third, this scoping review focussed on systematic and non-SRs only, thereby missing areas lacking such reviews. However, we explicitly included both systematic and non-SRs to explore under-researched areas. Fourth, some effect measures were too unclear to include in the tables because it was not possible to go back to individual studies. Furthermore, as our inclusion criteria were only focused on dietary intake (consumption) and health outcomes; studies considering any other precursors to dietary behaviour such as purchasing behaviour, awareness, knowledge, preferences and availability were not presented. Also, most studies included were conducted in high-income or middle-income countries; generalisation of the results to low-income settings might be problematic. Fifth, whereas the effects of price interventions can be quantified and occurred immediately, nutrition education and health promotion campaigns have long-term effects that cannot be so easily measured. Finally, systematic and non-SRs included many different study methodologies. Outputs from RCTs, observational cohorts, natural experiments and modelling studies are all potentially valid, but may not be directly comparable and should therefore be interpreted with caution. We have, therefore, routinely presented such results separately.

Future research

The logical next step is to conduct a fully resourced SR that includes all relevant primary studies. This could explore whether effectiveness varies in different ethnic or socioeconomic sub-groups, as socioeconomic deprivation is often associated with increased intake of foods high in saturated fat and sugar, poor nutrition and poorer health. Downstream interventions (such as advice to individuals) may widen the inequalities gap, whereas population-based policy interventions may reduce inequalities. Finally, the few analyses of multi-component interventions suggest that they may be more effective than single interventions. Further research is needed in all these areas.

Conclusions

Diverse policy interventions exist to improve healthy eating and thus prevent NCD. The evidence base varies for each category and nutrient. The evidence base for price interventions appears to be comprehensive and included studies mostly on taxing SSBs, followed by taxing dietary fat. Studies investigating the effect of subsidies mainly focused on fruit and vegetable intake. Taxes consistently decrease SSB consumption, whereas subsidies increase fruit and vegetable intake. Only a few reviews were included for food composition. The Denmark trans fat story suggests that voluntary reformulation can be effective to start with; however, a legislative ban is able to then essentially eliminate industrial trans fat consumption. Mandatory reformulation is thus consistently more effective than voluntary reformulation. Health promotion campaigns and nutrition education benefits appear modest and effects usually reduce over time. Limiting the marketing of junk food and sugary drinks appears to be effective and, thus, a ripe target for regulation.

Workplace interventions appeared to be modestly effective at increasing fruit and vegetable intake. School interventions showed a modest effect of reducing SSB consumption and increase fruit and vegetable consumption in schools. The effects for food labelling were mixed. Labelling may help pressure reformulation by manufacturers and inform consumer choice. However, much depends on individual awareness, knowledge and ability to interpret labels correctly. Trade and investment factors controlling the food supply chain may be powerful but are currently under-researched. Finally, multi-component interventions appear to be more effective than single interventions. This might be predicted, having previously been reported in tobacco control, where comprehensive, coordinated programs are superior to single interventions.64