FOOD REFORMULATION FOR NCD- PREVENTION: REGULATORY OPTIONS AND POTENTIAL BARRIERS

In the context of NCD prevention, reformulation means reducing the salt, sugar, fat, or overall calorie content of processed foods. Reformulation has the potential to be a powerful public health intervention, because it involves making changes to unhealthy foods upstream in the food supply: improving population diets without individual consumers needing to change their behaviour. However, questions remain as to which regulatory approaches will be most effective at spurring the food industry to reformulate. The prevailing view has been that governments should persuade and encourage companies to reformulate ‘voluntarily’. However, there is emerging evidence that the most effective voluntary reformulation schemes are those with a high degree of government involvement, monitoring and oversight. This suggests that mandatory reformulation — ie, the highest degree of government involvement, through legislated mandatory nutrient limits — may well be a promising approach. While voluntary reformulation has wide support, it is criticised for being weak, ineffective and open to regulatory capture. Mandatory reformulation offers the possibility of a stronger regulatory approach and a level playing field for industry, but critics are wary of its impact on free choice. Mandatory nutrient limits may also constitute technical barriers to trade. This article considers food reformulation as a policy goal, before exploring these regulatory options and the criticisms of each.


INTRODUCTION
Around the world, unhealthy diets have become a major contributor to disease, death and disability. 1 Diets characterised by high consumption of salt, sugar, fat and overall energy, and low consumption of fruits, vegetables and whole grains, are a leading risk factor for obesity as well as for diet-related noncommunicable diseases (NCDs). These include cardiovascular disease, stroke, type-2 diabetes, and certain cancers, and account for 40 per cent of all NCD deaths, or one in five deaths worldwide. 2 Over the past two decades, governments around the world have become increasingly concerned about the significant health, social and economic costs of unhealthy diets, prompting increased interest in policy interventions. 3 Governments wishing to take preventive action on diet-related NCDs can now draw upon a well-documented 'menu' of policy options. 4 However, as Hawkes and colleagues have demonstrated, the bulk of policy actions that are implemented tend to be those 'downstream' interventions primarily targeting eaters' behaviour. 5 These include food-based dietary guidelines, health promotion campaigns and nutrition labelling. Though widely implemented, 6 experts assess these behavioural interventions as 'weak' 7 and 'limited in reach and scope', 8 and their impact on shifting population diets as 'patchy' 9 and 'poor'. 10 In order to bring about population-wide dietary change (and so halt the growth of obesity and diet-related NCDs) such interventions will need to be supported by policies that target the wider food environment and the food supply itself. 11 Food reformulationchanging the composition of processed foods to make them healthieris an increasingly popular policy goal, which aims to tackle poor diets upstream, in the food supply. However, the question of how best to achieve this policy goal remains contested. The food industry and many governments have tended to prefer 'voluntary' approaches to reformulation, which can run the gamut from companies' purely voluntary actions, to coand quasi-regulatory arrangements between industry and government. Such schemes are relatively cheap to implement, but are criticised for being ineffective and open to industry capture. By contrast, the public health community has advocated a mandatory approach to reformulation, in the form of legislated upper limits on particular nutrients in particular foods. Mandatory nutrient limits offer a stronger regulatory approach, but are criticised for limiting personal and commercial freedoms, and may also function as technical barriers to trade. This article begins by setting out the background and rationale of food reformulation as a policy goal (Section II), before introducing different regulatory approaches to achieving this goal (Section III). Section IV focuses on what is described as 'voluntary' reformulation, despite featuring varying degrees and kinds of state involvement. It gathers evidence of this prevailing approach at both the national and international levels, before dealing with criticisms of the approach. In essence, critics argue that voluntary reformulation policies are ineffective and undermined by industry conflicts of interest. Instead, evidence suggests that the more effective voluntary reformulation policies are those that take on features of traditional, 'command-andcontrol' regulation, such as government oversight and specific timelines and targets. This challenges the widespread understanding of 'mandatory' and 'voluntary' reformulation as binary options. Rather, they should be seen as points on a regulatory spectrum or continuum. 12  Section V introduces 'mandatory nutrient limits', a small but growing group of regulations whose aim is mandatory reformulation, and examines the reasons why countries may choose to introduce these. To date, the major objection to mandatory nutrient limits is that they are freedom-limiting, as I have discussed elsewhere. 13 However, as mandatory nutrient limits gain acceptability through greater use, it will be increasingly important to consider other barriers to their implementation. The precedent of tobacco control has demonstrated the importance of considering the impact of international trade rules on innovative NCD prevention policies. 14 As such, this article concludes by considering whether mandatory nutrient limits may constitute technical barriers to trade (TBTs) under World Trade Organization (WTO) rules, 15 an issue which has not been dealt with extensively in the literature. Unlike voluntary standards, mandatory technical regulations must be complied with, and non-compliant products may not enter the market. As such, mandatory regulations are more likely than voluntary standards to be prima facie considered TBTs. 16 Interest in reformulation, in all its regulatory permutations, is growing. As more and more countries implement reformulation policies, we will have greater access to evidence regarding their effectiveness. However, it is equally important to gather the regulatory evidence. Voluntary and mandatory approaches to reformulation come with different benefits, burdens, and barriers, and these should factor into the relative feasibility and acceptability of each as a policy option.

II FOOD REFORMULATION: DEFINITION AND RATIONALE
In the general context of food manufacturing, reformulation refers to changing the chemical or nutritional composition of processed foods for any reasoneg, to make them look or taste better, or because of supply chain requirements. 17 However, in the context of diet-related NCDs, reformulation means changing the composition of processed food in order to make it healthier. 18 Here, reformulation aims to:  Remove trans fatty acids (TFAs, also known as trans fats), which cause coronary heart disease;  Limit salt/sodium, overconsumption of which is linked to high blood pressure and strokes;  Limit sugar, which is linked to high blood sugar, diabetes and dental caries; and/or  Reduce the overall energy content or density of a product (measured in calories or kilojoules). Energy imbalance, ie, when an individual consumes more energy than they expend, including through the consumption of energy-dense foods, is linked to obesity.
In theory, reformulation can also involve adding food components associated with NCD prevention (eg, fruit, vegetables, whole grains, fibre). However, while there are examples of individual companies making such changes to their products, 19 there are few examples of policy or regulation to achieve this. 20 Therefore, in this paper, I use the term 'reformulation' to mean removing or reducing trans fats, saturated fats, salt, sugar or energy density.
Salt, sugar and fat are added to foods during processing and manufacture for a variety of reasons. 21 Salt, for instance, can act as a preservative, as an agent in baking, or as a flavour enhancer. Fat contributes to the 'mouth feel' and texture of baked and fried foods, enhancing their palatability. 22 In the case of very highly processed (or 'ultra-processed') foods, ingredients have undergone such a degree of refinement, extrusion and re-composition that their original tastes and textures are long gone: here, salt, sugar and fat are added back in, to make their taste acceptable. 23 Compared to whole grains, fruits, and vegetables, sugar and fat are relatively cheap ingredients (often due to mismatches between incentives for agricultural production and dietary goals) and so adding sugar and fat rather than whole grains etc increases profit margins. 24 Regardless of why they are added, the fact that they have been added provides the entry point for policy intervention: at food production, rather than at consumption.
Unlike policies that aim to change our consumption behaviours, reformulation seeks to change the nature of the available food. That is to say, it targets the supply, rather than the demand, side of dietary choice. Its rationale is that: (a) relatively minor changes to commonly consumed foods can have a major effect on population health; and (b) these changes can be made at the level of manufacturing, rather than food choice or behaviour. 25  modify drastically his or her habitual dietary food pattern', 28 but nevertheless benefits from an improved diet. Population-based policies 'tend to be sustainable, affect the whole population (including those who are difficult to reach), become systemic (affect default behaviours), and reverse some of the environmental drivers'. 29 And, unlike a medical approach that involves identifying and targeting only high-risk individuals, population-based prevention 'can be expected to generate much-needed health gains while entirely or very largely paying for themselves through their reduction of future health-care costs '. 30 Ideally, health policy is based on the strongest possible evidence. In the case of medical interventions, this would mean a prospective, double blind, randomised controlled trial. However, obtaining the same 'gold standard' of evidence in public health policy is much more difficult. This is especially true of nutritional epidemiology, where disentangling dietary factors from other closely related factors (eg, other health behaviours, wealth, or education levels) can be particularly challenging. 31 Advocates of reformulation therefore draw on a mix of evidence to argue that the intervention is both effective and cost-effective. This includes modelling studies, case studies, and systematic reviews. So, in a modelling study of interventions to reduce salt intake, Cobiac and colleagues found that 'Programmes to encourage the food industry to reduce salt in processed foods… are an excellent investment; they improve population health and will reduce health sector spending in the long term'. 32 In Mauritius, reformulation to reduce the saturated fat in cooking oil was associated with a significant fall in cholesterol levels over five years. 33 Reformulation to limit TFAs in Denmark, initially through voluntary measures and then using government regulation, was associated with a significant decline in cardiovascular disease. 34 And in a systematic review, Downs and colleagues found that reformulation eliminated TFAs from the food supply more effectively than labelling. 35 In a recent systematic review on the effectiveness of policies to reduce population salt consumption, Hyseni and colleagues found reformulation (along with multicomponent interventions) to be the most powerful intervention.

III REGULATORY APPROACHES TO REFORMULATION
If reformulation is the policy goal, then how is it to be achieved? Reducing salt, sugar, fat, or calories in the food supply means first influencing food manufacturers to act in a certain way. This is becauseother than in states where food production is nationalisedgovernment is not directly responsible for the food supply. Rather, it regulates the food industry, which in turn produces the food supply. Traditionally, governments achieved this through legislation and regulations: formal rules backed up by penalties for their breach. 37 Today however, many aspects of food regulation have become 'de-centred', 38 with governments seeking to facilitate or encourage, rather than coerce, a particular outcome. As in many industries, traditional command-and-control regulation in the food industry has been supplemented, and in many cases supplanted, by 'softer', arm's length forms of regulation including guidelines and standards (rather than laws) and self-regulation (rather than government regulation). 39 A review of the literature suggests that different policy approaches, involving different types and degrees of state involvement, can encourage food reformulation. These can be understood using terms from two influential taxonomies of public health policy intervention, namely Gostin's seven models, 40 and the 'Nuffield Ladder', 41 of interventions. So, governments wishing to spur reformulation by food manufacturers may adopt:  A 'pure' market approach to reformulation, ie 'do nothing' in a Nuffield sense, 42 leaving companies to respond to consumers' demands for healthier foods. This may play out at the level of individual companies (eg, multinational PepsiCo making a strategic commercial decision to decrease the salt, sugar and fat in its snack foods), 43 or across an industry (eg, Australian manufacturers significantly reduced TFAs across the food supply during the 1990s, in response to consumer demand). 44  A hybrid market and policy approach to reformulation, whereby government actions prompt a change in consumer expectations and demands. Informational policies, in the form of health promotion campaigns, education, or dietary guidelines, can help to shift consumer preferencesto which markets then respond. For example, after governments issued advice to consume low-fat dairy foods during the 1980s and 1990s, manufacturers around the world significantly increased the number and types of lowfat dairy products available for sale. 45  A policy approach with indirect effects on reformulation, such as labelling (especially interpretative front-of-pack labelling) or fiscal policies (taxes or subsidies). In these cases, while the measure does not regulate food composition as such, manufacturers reformulate in order to achieve a more favourable label or minimise costs. In the Netherlands, for example, adoption of a 'healthy choice' front-of-pack logo prompted manufacturers to reformulate their products, especially in relation to sodium and dietary fibre. 46 In Hungary, one study showed that 40 per cent of manufacturers had reformulated their products in order to avoid the 'junk food tax'. 47  A government-led, 'voluntary' approach with direct effects, exemplified by 'new governance' 48 tools such as self-regulatory, co-regulatory, or quasi-regulatory reformulation policies. Here, food composition is directly targeted, but the measures are imposed voluntarily by the food industry, with varying degrees of government involvement. In the UK during 2003-2011, industry and government worked together to develop and implement standards on salt reduction, described as 'voluntary with the threat of regulation/legislation'. 49  A policy approach with direct effects: mandatory nutrient limits, whereby governments set and enforce food composition requirements, leading to reformulation. The mandatory TFA limit in Denmark 50 is a good example (discussed further below).
Only the last two involve efforts by government to influence food composition directly. As such, they represent a paradigm shift from the weaker, indirect or behavioural policy approaches described above, which have been criticised as ineffective. These newer approaches, targeting the food supply directly, are the focus of this article. In the nutrition policy literature, the distinction between the two strategies is often expressed as 'voluntary' versus 'mandatory'. However, it is important to emphasise that 'voluntary reformulation' in this (policy) context does not refer to the purely voluntary, commercial decisions of food manufacturers. Rather, voluntary reformulation is used to describe a range of policy approaches with varying degrees of government oversight, stopping short of laws or regulations. Indeed, given the heterogeneity of the approaches used, the main purpose of the term 'voluntary' seems to be to distinguish them from traditional, command-and-control regulation. A more accurate though unwieldyformulation would be 'government action to encourage private sector action'. 51  An Australian taskforce in the late 1980s drew on similar evidence from Norway, whose approach was 'the most comprehensive adopted by any developed country, and quite unique'. 55 In contrast to other countries, the taskforce wrote, 'Norway's policy does not emphasize individual behaviour as much as it proposed to change dietary habits by changing the process of food production and distribution in order to reduce the availability of what is termed unhealthy food"'. 56 Norway had achieved this 'through liaison with and education of primary and secondary food industries'. 57 During the 1990s, the idea that achieving a healthier food supply depended on first securing the agreement and cooperation of the food industry became the default assumption. For instance, in a 1995 article that elsewhere called for government action on obesity, James wrote: And even in their 'new frontier' article on the emerging role of law in obesity prevention (often cited as the first American journal article on the subject), Mello and colleagues nevertheless judged persuasion and encouragement to be more effective than regulation, when it came to changing the food supply: [T]he success of government regulation of the food industry will probably fall short of what industry could accomplish alone if it were strongly motivated to do so. Efforts to encourage self-regulation and corporate responsibility could go far toward improving the healthfulness of foods sold. 59 As described below, 'persuade and encourage' has achieved policy salience at both the national and international levels. The DPAS has since been replaced by the WHO's Global Action Plan for the Prevention and Control of Noncommunicable Diseases 2013-2020, 77 but 'persuade and encourage' remains an important tenet. One of the overarching principles of the GAP is multisectoral action: 'coordinated multistakeholder engagement for health both at government level and at the level of a wide range of actors … and partnership with relevant civil society and private sector entities'. 78 Objective 3 of the GAP is 'to reduce modifiable risk factors for noncommunicable diseases and underlying social determinants through creation of health-promoting environments', including the promotion of a healthy diet. WHO is agnostic as to how that might be achieved: member states are urged to 'develop guidelines, recommendations or policy measures that engage different relevant sectors, such as food producers and processors' to reduce salt, sugar and overall calories, and to replace trans and saturated fats with unsaturated fats. 79

D Voluntary Reformulation: Criticism and Barriers
Although voluntary reformulation schemes come in a wide array of regulatory forms, there is an emerging literature arguing that these forms are not created equal. In particular, there is growing evidence that a reformulation scheme's degree of voluntariness is inversely correlated to its effectiveness. Persuading and encouraging, or even partnership with the food industry, is not sufficient. Rather, certain structuring conditions need to be met in order for partnerships with the food industry to be effective for public health. These conditions include concrete timelines and targets, transparency, incentives for compliance, and consequences for noncompliancein short, higher degrees of government involvement. the arrangement 'undermined its potential as a public health policy tool and hindered its ability to deliver at a population level'. 83 The persuade and encourage approaches rest on a belief that, given the right information, support or incentives, the food industry will be sufficiently motivated to change its products. To many public health observers, this belief is at best optimistic, and at worst deluded. They argue that the making, selling and advertising of highly processed foods, high in sugar, salt and fat, is not incidental to the industry's activitiesrather, it is the very basis of its profitability. 84 Accordingly, schemes that involve governments partnering with industry to improve public health are doomed to fail, because the interests of these two stakeholders are fundamentally not aligned. 85 In such cases, self-, co-or quasi-regulation is simply a smokescreen for inaction or, worse, a way of blocking or co-opting government regulation. 86 Moreover, as Scrinis and Monteiro have noted recently, at an ecological level, voluntary reformulation efforts do nothing to challenge unhealthy food consumption patterns, and may in fact legitimise them. 87 On the other hand, cases of successful voluntary reformulation are characterized by a high degree of government oversight and corporate accountability. The UK's 2003-2011 efforts on salt reduction mentioned above are widely hailed as a successful example of voluntary reformulation, in terms of achieving targets. Though voluntary, companies' actions to reduce salt took place against a backdrop of strong scientific leadership, repeated monitoring and evaluation of population salt intake, a salt reduction strategy with progressively lower targets, ministerial support including the threat of regulation, clear nutrition labelling, and a high profile consumer awareness campaign. 88 The success of the UK's voluntary salt reductions is attributed to all of these components, acting in concert. 89 Drawing on this case study and others, as well as frameworks derived from regulatory theory, Reeve and Magnusson have recently proposed a systematic framework under which governments can become progressively more involved as and when voluntary schemes prove ineffective. 90 They describe this as 'regulatory scaffolding', and suggest three domains in which governments can operate to make selfregulatory schemes more robust. These are:  Regulatory content (eg, specifying goals and targets, clarifying definitions, specific terms and conditions);  Regulatory processes (eg, regulating the administration, monitoring, and review of selfregulatory processes; requiring independent audits); and  Enforcement (eg, building in incentives for compliance and penalties for noncompliance). 91

V MANDATORY NUTRIENT LIMITS
A What Are They?
If the way to improve voluntary schemes is, in effect, to make them less voluntarythen what of mandatory reformulation? Though less popular than voluntary approaches, an increasing number of jurisdictions are using mandatory regulation to spur reformulation. This regulation does not tend to specify the desired process (ie, reformulation) but rather the desired outcome: upper limits on particular nutrients in particular foods. Manufacturers must comply with these limits if they want their food to remain on the market. These mandatory nutrient limits 92 now exist in 24 jurisdictions around the world, and more jurisdictions are adopting them. 93 Table 1 lists mandatory nutrient limits with jurisdictions where they are currently in force.  Mandatory nutrient limits are traditional, command-and-control government regulations, usually implemented as national food standards (though some take the form of legislation or other rule types, such as Royal Decrees). As shown in Table 1, existing mandatory nutrient 91 Reeve and Magnusson, 'Food Reformulation and the (Neo)-Liberal State', above n 90. This work builds on Ayres and Braithwaite's 'responsive regulation' concept, which described how the State could step back from command-and-control regulation, but nevertheless promote effective outcomes by scaling up its involvement in response to non-compliance: Ian Ayres and John Braithwaite, limits apply to fats (primarily trans fats) and salt. At a technical level, the regulations tend to follow a familiar structure, specifying:  The regulated food (eg, bread, oil, tomato paste, and may also specify that imported foods as well as domestic products are included);  The nutrient to be limited (eg, sodium, industrially produced TFAs); and  The limit to be imposed (eg, 2 per cent).
The major exception to this structure is the regulation of trans fat in the USA. In 2015, industrially produced TFAs, or partially hydrogenated oils, were removed from the US Food and Drug Administration's (FDA's) list of food additives 'generally recognised as safe' (GRAS). GRAS ingredients are not regarded as food additives for the purposes of US food law, and so may be added without prior approval. 95 Removal of GRAS status means that TFAs may no longer be added to food unless a particular use has received pre-marketing approval from the FDA. 96 Obtaining such approval would involve demonstrating 'a reasonable certainty of no harm of the proposed use(s)' 97 , where harm includes chronic harms such as adverse effects on cholesterol and increased risk of coronary heart disease. 98 As such, the removal of GRAS status functions as a de facto mandatory nutrient limit.
In 2004, Denmark was the first government to regulate an upper permissible limit, of 2 per cent, on artificially produced TFAs. It has since been followed by nine other jurisdictions. In 2009, The Netherlands introduced government regulation on the permissible amount of salt (sodium) in bread and flour. Eight other jurisdictions followed its example, and in 2013 both South Africa and Argentina introduced mandatory sodium limits on a wide range of foodstuffs. Public health experts are now advocating mandatory limits on sugar, 99 though so far no jurisdiction has implemented these. While their number is still relatively low, mandatory nutrient limits are on the rise, both by kind of nutrient and also by total number: see Figure 1  1

International Momentum on Diet-Related NCDs
International momentum on diet-related NCDs was an important ingredient in South Africa's decision to introduce its mandatory salt limits in a range of commonly consumed foods. In the lead up to the UN's Political Declaration of the High-Level Meeting of the General Assembly on the Prevention and Control of NCDs (2011), the South African health minister wanted to take clear, decisive action, and considered that regulation would send a strong message to the food industry. 100 2

The 'Level Playing Field'
A reason frequently cited for regulation as opposed to voluntary compliance is that regulation creates a level playing field for industry players 101the same rules apply to all firms, and none will be forced to assess whether compliance will lend a competitive advantage or disadvantage. For this reason, some firms actually prefer regulation to self-regulatory schemes. 100 Charlton, Webster and Kowal, above n 68. 101

Food Reformulation for NCD-Prevention: Regulatory Options and Potential Barriers The Success of Voluntary Reformulation Measures
A surprising factor in Denmark's decision to introduce a mandatory TFA limit in 2003 was the success of the voluntary approach that was used in the first instance. As mentioned above, there is a prevailing belief that mandatory regulation is used by governments only as a final or extreme option, or even as a threat to industry if it does not take adequate voluntary action. In the case of Denmark, voluntary measures had, over ten years, achieved very significant reductions in TFA across the food supply. However, certain subgroups of the population were still consuming high TFA diets (see below). This persuaded the government that regulation was necessary. 102 So in Denmark, the success of voluntary reformulation actually paved the way for widespread acceptance of regulation. The technology needed to reformulate had already been developed, industry buy-in and cooperation had already been secured, and consumers were already educated on the need for reformulation. 103 4 The Need to Protect Vulnerable Sub-groups In Denmark, voluntary reformulation successfully and significantly reduced the average quantity of TFA across the food supply. However, there was evidence that certain pockets of the population were still consuming very high TFA diets. Moreover, these tended to be subgroups already at greater risk of cardiovascular disease due to socioeconomic and behavioural factors. Labelling and public health awareness campaigns were deemed an ineffective way of reaching these sub-groups, as well as being impractical in the case of unpackaged takeaway foods. 104 Accordingly, the government decided to regulate an upper permissible limit of 2 per cent on industrially produced TFA in oils and fats. In Australia, voluntary reformulation has achieved similarly successful drops in average TFA consumption, 105 but, as in Denmark, vulnerable sub-groups remain. 106 5

Policy Transfer and Trade Harmonisation
Finally, there is some evidence that the existence of mandatory nutrient limits in some jurisdictions has led to their adoption in other jurisdictions. It is very likely that the introduction of a mandatory TFA limit in Denmark has led to the diffusion of this regulation to other jurisdictions. 107 Following the Danish TFA regulations, several countries have introduced regulations in almost identical formatie, they use the same upper limit of 2 per cent, and apply it to fats and oils. These include four more EU countries, plus Switzerland and Norway, and the EU is now considering whether to adopt a mandatory limit under European law. 108 In this regard, one of the considerations is that, 'By setting an EU wide harmonised legal limit, the approach would also minimise, or even abolish, the risk of national regulatory choices (further) fragmenting the single market'. 109 In South Africa, there was also some evidence of policy transfer from nutrient to nutrient: the salt limit was introduced under the same legislation that was used to regulate TFA, and so regulators already had a template for how it might be done. 110 C

Mandatory Nutrient Limits and International Trade Rules
Until recent years, the idea of using mandatory limits for NCD prevention has been given short shrift in the literature, mentioned only in passing as part of a taxonomy of policy approaches to unhealthy diets. Dismissed as overly paternalistic and intrusive, these measures tended to be seen as an extreme, 'far-reaching', 111 and 'contentious' 112 option. For instance, Resnik drew on health ethics frameworks to argue in 2010 that trans fat limits could open the door to excessive government control over consumers' freedom to choose their own diet. 113 Beyond such political/philosophical objections, otherperhaps more practicalbarriers to the implementation of mandatory nutrient limits have rarely been considered. However, as reformulation has increasingly become part of the policy landscape in recent years, this view is being re-evaluated. 114 Having dealt with the 'freedom-limiting' charges against mandatory nutrient limits elsewhere, 115 here I briefly examine one of the more practical objections. This is that mandatory nutrient limits may constitute a TBT under WTO rules. Just as the Western diet has spread around the world, NCD prevention policies are also subject to globalisation. Today, national jurisdictions are less likely to be innovating on NCD policy in isolation, but rather are looking to the guidance of the WHO in its normative role. And most recently, the WHO has recommended that countries implement mandatory limits on trans fat, and consider mandatory limits for salt. 116 With this global diffusion of NCD prevention policies, the need to consider international trade rules and their impact on public health is becoming more and more important, as further highlighted by recent challenges to innovative tobacco control regulations under international trade rules. 117 TBTs are defined in the Agreement on Technical Barriers to Trade, an international treaty administered by the WTO. The aim of the Agreement is to ensure that 'technical regulations and standards, including packaging, marking and labelling requirements… do not create unnecessary obstacles to international trade'. 118 In order to achieve this purpose, the Agreement specifies at Article 2.2 that: technical regulations shall not be more trade-restrictive than necessary to fulfil a legitimate objective, taking account of the risks non-fulfilment would create. Such legitimate objectives are, inter alia: national security requirements; the prevention of deceptive practices; protection of human health or safety, animal or plant life or health, or the environment. In assessing such risks, relevant elements of consideration are, inter alia: available scientific and technical information, related processing technology or intended end-uses of products. 119 The main difference between technical regulations and standards is that the former are mandatory and the latter are voluntary. 120 While goods that fail to comply with voluntary standards may still be sold (although they may experience adverse commercial consequences), failing to comply with technical regulations is a barrier to market entry. 121 As such, the widely held view is that that technical regulations (such as mandatory nutrient limits) are prima facie TBTs, while voluntary standards (such as measures to encourage voluntary reformulation) are not. 122 However, despite the existence of 26 mandatory nutrient limits, there has been only one challenge on the basis that it would constitute a trade barrier. That case was not brought under WTO law, but in the European Commission (EC).
As noted above, Denmark was the first country to introduce a mandatory TFA limit. The limit was introduced in 2003, and in 2004 the EC claimed that the regulation was a breach of Denmark's obligations under the terms of the EC Treaty relating to the free movement of goods within the EU. Articles 28-30 of the EC Treaty prohibit unjustified restriction on intra-EU trade, the 'most restrictive' of which is 'a ban on the marketing of a specific product or substance'. 123 The EU took initial steps toward prosecution in the European Court of Justice. 124 Denmark refused to change its legislation, relying on Article 36 of the EC Treaty: trade barriers may be justified if they protect human health. 125 In such cases: The Member State imposing a national ban on a product/substance has to show that the measure is necessary and, where appropriate, that the marketing of the products in question poses a serious risk to public health and that those rules are in conformity with the principle of proportionality [and] … that the stated aim cannot be achieved by any other means that has a less restrictive effect on intra-EU trade. 126 In making this case, Denmark was able to present the EC with epidemiological and other evidence, and also received support from a European consumer organization. 127 The EC dropped the case in 2007.
In 2006, a Canadian national taskforce on options for TFA reduction was advised that a mandatory TFA limit 'would not conflict with Canada's international obligations under World Trade Organization agreements, in particular the Agreement on Technical Barriers to Trade', on the grounds that it would be necessary to achieve the legitimate objective of protecting public health. 128 Two factors provided confidence to the task force: first, the advice of leading scientific organizations such as the WHO, and secondly, 'the fact that other jurisdictions have adopted measures to limit the consumption of trans fats'. 129 Ten years on, the global landscape is changing fast, as increasing numbers of countries choose to impose mandatory TFA limits (see Figure 1, above). Perhaps most notable here is the USA's recent decision to remove TFA from its list of GRAS ingredients. In light of this development, a recent European report suggests that in fact not having a mandatory TFA limit may pose a barrier to future trade with the USA. 130 There have not yet been any test cases in relation to the trade restrictiveness of mandatory salt limits (bearing in mind that the first comprehensive salt standards, in South Africa and Argentina, were introduced in 2013 and are only now being phased in). However, it is unlikely that the lessons from TFA can be applied to salt in a straightforward manner, due to notable differences between the two nutrients. In essence, TFA is an artificial substance, added at the level of manufacturing and only found in fats and oils, which carries clear risk, and no benefit, to human health. In food law terms, this is analogous to adulteration (ie, contamination with a harmful substance). 131 By contrast, salt is a naturally occurring substance, with a more complex relationship to human health, found in products throughout the food supply but also added in cooking and at the table. Countries wishing to justify the imposition of a trade barrier in the form of a mandatory nutrient limit will need to show technical, scientific, statistic and nutritional data, 132 and the evidentiary picture for salt is far more complex than that for TFA. 133

VI CONCLUSION
Halting, and ultimately reversing, the impact of unhealthy diets on human health will require strong, upstream, population-based policy responses. Reformulating commonly consumed processed foods is a policy goal that holds great promise in this regard, and reformulation schemes around the world are on the rise. But the regulatory question remains: how best to achieve this goal? Regulatory options for encouraging food reformulation run the gamut from pure, market-based approaches, to command-and-control regulation in the form of mandatory nutrient limits. Until recently, the prevailing wisdom has been that governments should seek to persuade and encourage the food industry to reformulate, an approach favoured by governments and the food industry, and encouraged by the WHO. However, measures to encourage voluntary reformulation have had mixed success. The most successful have been those with a high degree of government involvementthat is, voluntary schemes with features of mandatory regulation. Alongside evidence of the growing use of mandatory nutrient limits, this is suggestive of a trend. This in turn points to the need for greater technical understanding of mandatory limits, to enable policy learning: for instance, the major barriers and drivers faced by countries seeking to implement them. This paper has sought to add to the evidence base by providing regulatory insights on both voluntary and mandatory reformulation policies. This regulatory evidence will supplement the emerging base on the effectiveness of reformulation for public health, and ultimately guide policy makers seeking to design approaches to reformulation.