Systematic review of dietary trans-fat reduction interventions

Abstract Objective To systematically review published studies of interventions to reduce people’s intake of dietary trans-fatty acids (TFAs). Methods We searched online databases (CINAHL, the CRD Wider Public Health database, Cochrane Database of Systematic Reviews, Ovid®, MEDLINE®, Science Citation Index and Scopus) for studies evaluating TFA interventions between 1986 and 2017. Absolute decrease in TFA consumption (g/day) was the main outcome measure. We excluded studies reporting only on the TFA content in food products without a link to intake. We included trials, observational studies, meta-analyses and modelling studies. We conducted a narrative synthesis to interpret the data, grouping studies on a continuum ranging from interventions targeting individuals to population-wide, structural changes. Results After screening 1084 candidate papers, we included 23 papers: 12 empirical and 11 modelling studies. Multiple interventions in Denmark achieved a reduction in TFA consumption from 4.5 g/day in 1976 to 1.5 g/day in 1995 and then virtual elimination after legislation banning TFAs in manufactured food in 2004. Elsewhere, regulations mandating reformulation of food reduced TFA content by about 2.4 g/day. Worksite interventions achieved reductions averaging 1.2 g/day. Food labelling and individual dietary counselling both showed reductions of around 0.8 g/day. Conclusion Multicomponent interventions including legislation to eliminate TFAs from food products were the most effective strategy. Reformulation of food products and other multicomponent interventions also achieved useful reductions in TFA intake. By contrast, interventions targeted at individuals consistently achieved smaller reductions. Future prevention strategies should consider this effectiveness hierarchy to achieve the largest reductions in TFA consumption.


Introduction
Over two-thirds of the global burden of disability and death is attributable to noncommunicable diseases. Diseases such as cardiovascular diseases, common cancers, dementia, diabetes and respiratory disorders kill over 35 million people annually. 1 The World Health Organization (WHO) priority areas for reducing noncommunicable diseases include tobacco, alcohol, physical inactivity and poor diet. Poor diet generates a larger burden of disease than the other three risk factors combined. It accounts for an estimated 11.3 million deaths annually, compared with 2.1 million for low physical activity, 6.1 million for tobacco smoke and 3.1 million for alcohol and drug use. 1 The problem predominantly reflects an unhealthy global food environment dominated by processed foods high in sugar, salt, saturated fats and, crucially, industrial transfatty acids (TFAs). 1 TFAs are found naturally in small amounts in some meat and dairy products produced by bacterial action in the stomach of ruminant animals. 2 However, the majority of TFAs are industrial, being manufactured by partial hydrogenation of edible vegetable oils, such as palm oil, cottonseed oil, soybean oil or canola oil. 3 Industrial TFAs are then added to processed or packaged food, mainly to prolong shelf life and enhance taste and texture at a low cost. 4 Since the 1990s, research evidence has accumulated demonstrating that TFA consumption substantially increases people's risk of coronary heart disease. 2,5 It does this mainly by elevating harmful low-density lipoprotein cholesterol levels and decreasing protective high-density lipoprotein cholesterol. 2 TFAs may also increase the risk of Alzheimer's disease and certain cancers, 2 and worsen insulin sensitivity, thereby increasing the risk of type 2 diabetes. 5 A reduction in people's intake of industrial TFA is thus a WHO policy priority. 2 However, TFA intake in most countries still exceeds the WHO target of 2 g/day, mainly reflecting consumption of industrial TFAs in processed food. 6 Furthermore, even as overall TFA consumption falls, intake is likely to remain higher in poorer populations, who are more likely to eat processed food products. 7 Different strategies and policy options, targeting different groups, have been proposed to meet these targets. These can be described as upstream or downstream interventions. Downstream interventions generally target individuals and involve behavioural approaches. 8 Intermediate interventions target subgroups in worksites, schools or communities. Both downstream and intermediate interventions are dependent on the individual to respond. By contrast, upstream interventions take place at the population level and typically involve regulatory approaches, taxes or subsidies. By creating a healthier environment, they avoid any dependence on an individual response. 8 In alcohol and tobacco control policies, for instance, an effectiveness hierarchy of preventive interventions has been observed, whereby population-wide policy interventions appear to be consistently more powerful than interventions targeting individuals. 9,10 Policy interventions to remove industrial TFAs from foods have therefore been suggested as the most effective public health approach for reducing TFA intake and decreasing the burden of noncommunicable diseases. 11 Some countries have demonstrated that this is feasible. In Denmark, for example, multicomponent interventions progressively reduced the population's TFA intake and a subsequent legislative ban virtually eliminated TFAs in margarines and vegetable shortenings. 12 However, that success required substantial political will sustained over a decade. Most other countries only have achieved voluntary TFAs limits, reflecting concerns about political feasibility and generally lower levels of public pressure for change. 13 The evidence supporting the most effective policies for reducing TFA intake remains unclear. To inform future prevention strategies we conducted a systematic review of the evidence on the effectiveness of interventions to reduce people's TFA intake. We also hypothesized that an effectiveness hierarchy might exist.

Data sources and searches
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 14 We carried out a systematic search of online databases (CINAHL, the Centre for Reviews and Dissemination Wider Public Health database, Cochrane Database of Systematic Reviews, Ovid®, MEDLINE®, Science Citation Index and Scopus) for studies of interventions to reduce people's TFA intake, published between 1985 and 24 August 2017. A combination of relevant keywords, identified from exemplar papers (including a systematic review of policies for reducing dietary TFA), 15 was used to construct the search strategy (Box 1). All identified papers were imported into a web-based data management software (Zotero, version 4.0.29, Roy Rosenzweig Center for History and New Media, Fairfax, United States of America).
One author conducted the searches and removed the duplicates. Two authors then independently screened titles and abstracts for eligibility; if they deemed papers eligible, the full text was retrieved and again screened independently. We also scanned reference lists of included studies for potential relevant papers. Any differences in screening outcomes were resolved either by consensus, or by involving the senior author.

Study selection
We included a wide range of study designs including trials, observational studies, meta-analyses and modelling studies. Modelling studies added value by allowing the evaluation of certain interventions using different scenarios where empirical data are impractical or lacking (e.g. food labelling). However, we analysed them separately from the empirical studies. To be eligible, studies had to include the effectiveness of specific interventions on dietary TFA intake and have quantitative outcomes. Only studies published in English language were included. We assessed retrieved studies by using the population, intervention, comparison and outcomes study design approach (Box 2). 14 The primary outcome was dietary TFA intake in a population, reported as g/day. For studies reporting TFA intake as a percentage of total energy intake (E%), we converted the data to g/day using the conventional formula below (1). 16 Trans fat intake

Box 1. Search strategy used in the systematic review of dietary trans-fatty acid reduction policies
The following keywords were used in a search of CINAHL, the Centre for Reviews and Dissemination Wider Public Health database, Cochrane Database of Systematic Reviews, Ovid®, MEDLINE®, Science Citation Index and Scopus: (("Trans-fats" OR "trans-fats" OR "Trans-fat" OR "Dietary trans-fat" OR "Industrial trans-fat") AND ("Reformulation" OR "Regulation" OR "Self-regulation" OR "Labelling" OR "Limits" OR "Ban" OR "Elimination" OR "Legislation" OR "Agreements" OR "Campaigns" OR "Tax" OR "Health promotion" OR "Nutrition education" OR "Marketing control")) For Scopus, additional terms were used to narrow down the search: (("Intake" OR "Consumption" OR "Composition" OR "Content" OR "Effect" OR "Effectiveness" OR "Cons-effectiveness") AND ("Public policy" OR "Nutrition policy" OR "Health policy" OR "Policies" OR "Interventions" OR "Strategies" OR "Initiatives" OR "Policy options" OR "Actions"))

Box 2. Inclusion and exclusion criteria for selecting studies for the systematic review of dietary trans-fatty acid reduction policies
Papers only reporting on the reduction of TFA content in food products without a link to intake were excluded.
Secondary outcomes included changes in clinical or physiological indicators related to noncommunicable diseases (e.g. coronary heart disease deaths) and behaviours associated with a healthy diet.

Data extraction and quality assessment
We used pre-designed and pre-piloted data extraction forms to extract data from all included studies. One author conducted the data extraction, which was independently checked by other authors for the empirical and modelling studies.
We used the National Heart, Lung and Blood Institute quality assessment tools to assess the quality of empirical studies. 17 Two authors independently assessed the methodological quality of each study as poor, fair or good. Modelling studies were independently assessed by two modelling experts using an adapted version of a published quality assessment tool. 18 Quality was reported as poor, fair or good. Any disagreements were resolved by consensus or with another author.

Data synthesis and analysis
We conducted a narrative synthesis to interpret the data, grouping the studies according to intervention type. In accordance with the McLaren continuum of structural-agentic interventions 8 and the Nuffield ladder taxonomy, 19 we defined downstream (agentic) interventions as those where the principal mechanism of action is dependent on individuals altering their consumption behaviour. Conversely, we defined upstream (structural) interventions as those creating changes that target an entire population -not a subset, however large -thus effectively eliminating individual agency. We then categorized interventions according to their position in the McLaren continuum ( Fig. 1). 8 We separately analysed multicomponent interventions. An unweighted median regression model was fitted to the TFA intake data from eight empirical studies and four modelling studies.

Individual dietary counselling
One study in 2007, rated as fair quality, targeted Aboriginal families in Canada. It investigated the effect of dietary counselling on dietary intake, including TFA, using a 24-hour recall. The 29 intervention households significantly reduced their consumption of TFA (P = 0.02) from 0.6 to 0.5 g/day over 6 months, while the 28 control households increased consumption from 0.7 to 1.3 g/day. 20

Modelling studies
We found no empirical studies investigating the sole effect of labels showing the TFA content of food products, but we included five modelling studies (two of good quality, 32,33 two fair 34,35 and one of poor quality 36 ). In the Netherlands a healthier choices logo for food packages was implemented in 2006. Replacing all packaged products with those that complied with a healthier choices logo was projected to lead to a 0.8 g/day reduction in TFA intake from 2.1 to 1.3 g/ day. 34 Another Dutch study in 2013, using a similar approach, projected a higher reduction of 1.2 g/day (from 2.2 to 1.0 g/day). 35 A British study in 2015 estimated that labelling is at best only half as effective as a total ban on TFAs in terms of health and socioeconomic benefits. Improved labelling might save approximately 3500 deaths from coronary heart disease over the period 2015-2020 (1.3%, 3500 of the total 273 000 deaths) and reduce socioeconomic inequalities by some 1500 deaths (from 20 400 to 18 900). 32 Another modelling study in 2016 investigated the cost-effectiveness of mandatory labelling in products on sale in the European Union (EU) and projected that it may prevent 0.98 million of the 1076 million disabilityadjusted life years (DALYs) attributed to coronary artery disease. However, compared with taking no action, this option incurred greater costs than it saved, in terms of health-care costs, lost productivity and implementation costs. 33 In a similar approach to the Dutch study, researchers in Brazil investigated replacing products with ones that complied with a healthy choices logo. 36 Estimated TFA intakes were 0.8 g/day (standard deviation, SD: 1.0) for typical menus, 0.1 g/ day (SD: 0.2) for the choices menus and 0.2 g/day (SD: 0.3) for energy-adjusted choices menus, i.e. the same as choices menu, but adjusted for energy of a typical menu.

Empirical studies
Two empirical studies examined the effect of reformulating industrially produced foods. A good-quality study reported the results of a TFA monitoring programme after voluntary reformulation limits for TFA content were put in place in Canada in 2005 on vegetable oils and soft margarines and other prepackaged foods (< 2% and < 5% of total fat, respectively). TFA intake, measured using 24-hour food recalls in the general population (33 030 people), fell from 4.9 g/day in 2004 to 3.4 g/day in 2008. 22 The other study, rated as fair quality, conducted two meta-analyses of North American and European data. Both analyses investigated the effect of TFA consumption on coronary heart disease risk factors (one included 13 randomized controlled trials and the other covered four prospective studies). The researchers calculated the effect of reformulating the TFA content of partially hydrogenated vegetable oils with other fats. They found higher risk reductions when reformulating oils with higher TFA content. The randomized controlled trials reported an approximately 19% risk reduction, whereas the prospective studies found a 39% reduction in coronary heart disease risk. 23

Modelling studies
One study of poor quality in 2011 modelled the effect of reformulation of foods to reduce TFA in the Netherlands. TFA intake in the population of 750 participants aged 19-30 years was projected to fall from 2.3 to 1.9 g/day after reformulating specific food groups (e.g. bread, pastry, cakes and biscuits; meat snacks and salads; fat and margarines). 37    Notes: Plot of the reported reductions in TFA consumption (g/day) of the 12 empirical studies included in this systematic review, by type of intervention. An unweighted median regression model was fitted to the data and is depicted by the grey line on the plot. We used the National Heart, Lung and Blood Institute quality assessment tools to assess the quality of empirical studies. 17 Systematic reviews Dietary trans-fatty acid reduction policies Lirije Hyseni et al.

Type of intervention
could be saved due to reductions in direct health-care costs and in indirect costs linked to informal care, and these outweighed the costs of implementing this policy. 33

Empirical studies
Analysing TFA g/purchase not TFA g/ day, one good-quality study investigated the effect of regulations on TFA content of food sold in New York city. Instituted in 2007, the policy allowed takeaway food restaurants to sell only products with a ≤ 0.5 g TFA content per serving. The estimated mean TFA intake per purchase, based on purchase receipts matched to products, decreased from 2.9 g in 2007 (6969 purchases) to 0.5 g in 2009 (7885 purchases). 24 Another study of good quality analysed the impact of a upper limit of 0.5 g/ serving (commonly referred to as a TFA ban) on TFAs in all food-service establishments in the USA in 2016 and found a 4.5% reduction in cardiovascular disease mortality, from 13 per 100 000 persons between 2010 and 2013. 25 Finally, a study of good quality investigated the effect of TFA restrictions in restaurants in certain New York state counties implemented in 2007, comparing hospital admissions for cardiovascular events in counties with and without restrictions. Three or more years after the restrictions were implemented there were significantly fewer cardiovascular events in the intervention population of 8.4 million adults compared with the reference population of 3.3 million, after adjusting for temporal trends. These changes applied when analysing myocardial infarction and stroke events combined (change of −6.2%; 95% confidence interval, CI: −9.2% to −3.2%) and myocardial infarction alone (−7.8%; 95% CI: −12.7% to −2.8%). 26

Modelling studies
Five modelling studies, all of good quality, modelled the effect on health outcomes of legislation affecting TFA consumption. 32,33,[39][40][41] One British study projected that a ban on TFAs in sit-down and takeaway restaurants for the period 2015-2020 would save approximately 1800 (0.7%) coronary heart disease deaths annually in the population. A total ban of TFAs in all products might achieve some 7200 (2.6%) fewer deaths from coronary heart disease and reduce socioeconomic inequalities in coronary heart disease mortality by approximately 3000 deaths. 32 Two other British studies found that legislation to achieve a 0.5% reduction in the total energy derived from dietary TFA could prevent 2700 39 or 3700 cardiovascular disease deaths annually. 40 Another British study modelled the effect of eliminating TFA from industrially produced food and estimated this could prevent around 1700 coronary heart disease deaths and gain some 15 000 life years annually over a decade. Eliminating both natural-and industrial-derived TFAs may prevent some 3300 coronary heart disease deaths annually and gain approximately 30 000 life years. 41 Finally, a study in 2016 modelled the cost-effectiveness of different EU policy options to reduce the TFA intake of the population. Placing legal limits on the TFA content in industrial foods was projected to avoid 3.73 DALYs and save € 51 billion from health-care costs and lost productivity in EU Member States. 33

Empirical studies
Four fair-quality empirical studies investigated the combined effect of more than one policy on TFA consumption. [27][28][29][30] One cross-sectional study evaluated the effect of both a public health education programme and voluntary reformulation of industrially produced TFAs in soybean oil in Costa Rica. Using 3-day food records the study found a 1.7 g/day reduction in TFA intake among adolescents from 4.5 g/day (276 people) in 1996 to 2.8 g/day (133 people) in 2006. 27 Two other studies included both labelling and voluntary reformulation limits in Canada. Estimating TFA intake from breastmilk samples, they reported reductions in TFA intake from 2.0 in 2009 to 0.7 g/day in 2011 (639 people) 29 and from 4.0 to 2.2 g/day (87 people). 28 An American cross-sectional population-based study analysed blood samples before and after introduction of food labelling and voluntary limits on TFA in restaurants. A 58% reduction in TFA levels was found, from 93.1 µmol/L (229 samples) in 2000 to 39.0 µmol/L (292 samples) in 2009, suggesting a substantial reduction in dietary intake of TFAs over the period. 30 From 1993 onwards, Denmark progressively implemented multiple interventions to reduce dietary TFA intake, including food labelling and voluntary agreements with industry. 38 One study found that TFA intake dropped from 4.5 g/day to 1.5 g/day from 1976 to 1995. Intake from margarines and shortenings virtually reduced to zero after legislation to ban industrially produced TFA in 2004. 31

Modelling studies
One modelling study of fair quality used different scenarios to model the effect of food labelling and reformulation of food products on health outcomes in Argentina. Based on changes in lipid profile, the combined intervention was projected to avert 301 deaths, 1066 acute coronary heart disease events, 5237 DALYs and 17 million United States dollars in health-care costs annually in the adult population. 42

Discussion
Our systematic review suggests that multicomponent interventions achieve the biggest reductions in TFA consumption across an entire population, as demonstrated in Denmark, Canada and Costa Rica. Systematic reformulation of products containing TFAs can also help, as observed in Canada and the USA. Interventions targeting individuals typically achieved smaller reductions in TFA consumption.
Several studies, in separate countries, investigated multicomponent interventions; all of them included food reformulation, labelling and voluntary limits on TFA content of industrial food. In Denmark, a progressive series of interventions finally led to a legislative ban on TFA. This package achieved the largest observed reduction in TFA intake in the population over the period from 1976 to 2005 (4.5 g/day). 12,31 The USA is now emulating this successful strategy. 43 Substantial, but smaller benefits were achieved by multi-intervention strategies lacking a legislative component in Costa Rica 27 and Canada. 28,29 Multicomponent strategies including upstream policies, such as price regulations or legislation, consistently achieved greater reductions in TFA intake than single interventions, particularly when these were downstream approaches focused on individuals.
Legislation to regulate TFA content in food achieved a 2.4 g/day reduction in intake of TFA in the city of New York. 24 This success has now been extended Dietary trans-fatty acid reduction policies Lirije Hyseni et al.
nationwide by the United States Food and Drug Administration ruling in June 2015 stating that partially hydrogenated oils are no longer generally recognized as safe for use in food. 44 Following the Danish exemplar, several other European countries have subsequently introduced legislation, setting an upper limit for TFAs of 2 g per 100 g in fat or oil. 13 However, other countries still rely on voluntary reformulation, which is less effective. 13,15 Legislation is routinely opposed by the food industry, fearful of decreased profits or the additional costs of reformulating products. 12,13,15 However, the evidence is that such legislation has generally had minimal financial impact on the food industry. 12,31,45 Several evaluation studies have reported reductions in TFA content of margarines and industrially produced foods, particularly using mandatory reformulation. 24,46 An early concern that manufacturers would substitute saturated fats such as palm oil for TFA has been dismissed. Both Canadian and American studies of reformulation have reported reductions in both TFA and saturated fats, and increases in unsaturated fat, 47 the preferred replacements for TFA. 13 Food labelling has the potential to help consumers make more informed decisions 48 and can also put pressure on the food industry to reformulate products. 49 We found no empirical studies of the effects of food labelling alone. However, several modelling studies estimated reductions in TFA consumption ranging from 0.3 g/day to 1.2 g/day. 32,34,35 In contrast, most studies involving food labelling have examined the public's understanding of labels, use of labels, food purchase and purchase-related behaviour rather than quantifying TFA intake. 50 The diverse labelling systems and claims currently generated by the industry may confuse many consumers, highlighting a need for package labels that are easier to understand. 48 Dietary counselling interventions in different settings, such as communities, worksites, schools and homes, were sparse. One study in the home 20 and another in worksites 21 both suggested that dietary counselling at an individual level could achieve TFA reductions of 0.8 g/day 20 and 1.2 g/day. 21 In practice, individuals may struggle to adhere and comply with dietary advice due to competing priorities or financial constraints. 51 We found no studies of mass media campaigns focusing on TFA. Media campaigns can achieve modest beneficial behaviour changes in nutrition, physical activity and tobacco and alcohol use -in motivated individuals. However, the overall population benefits tend to be more modest. 52 Taxation has been shown to be a potentially powerful tool for reducing consumption of tobacco, alcohol or sugar sweetened beverages. However, we found no studies of taxation focusing specifically on TFAs. 13 Our systematic review has several strengths. We did the screening, extraction and analysis in duplicate, and all the included studies were subjected to a rigorous quality appraisal. Furthermore, modelling studies were included, but considered separately, in recognition of their additional uncertainties when compared with empirical evidence. Our review also had limitations. We were unable to conduct a meta-analysis due to the profound heterogeneity of the studies and because several studies included multiple interventions. We only included studies in English. The primary outcome of this study was dietary intake and we excluded studies considering other components of dietary behaviour or changes in product content after reformulation. Generalization of the results should be cautious because countries will vary in their baseline TFA intake. Furthermore, we did not contact authors for missing data. Publication bias also remains possible, potentially overestimating the true effect of some interventions. Finally, some intervention benefits may have been overestimated due to underlying secular trends among the public towards lower TFA consumption.
In conclusion, our results suggest an effectiveness hierarchy similar to those previously demonstrated in salt, tobacco and alcohol control interventions. 9,10,53 Multicomponent interventions including a legislative ban on products appear the most effective strategy to reduce TFA intake. By contrast, more downstream interventions targeting individuals in domestic or work settings appear consistently less effective. Future prevention strategies might consider this effectiveness hierarchy to achieve the largest reductions in the consumption of TFAs or other harmful nutrients. ■

Systematic reviews
Dietary trans-fatty acid reduction policies Lirije Hyseni et al.

Systematic reviews
Dietary trans-fatty acid reduction policies Lirije Hyseni et al.