Eliminating artificial trans fatty acids in Argentina: estimated effects on the burden of coronary heart disease and costs

Abstract Objective To estimate the impact of Argentine policies to reduce trans fatty acids (TFA) on coronary heart disease (CHD), disability-adjusted life years (DALYs) and associated health-care costs. Methods We estimated the baseline intake of TFA before 2004 to be 1.5% of total energy intake. We built a policy model including baseline intake of TFA, the oils and fats used to replace artificial TFAs, the clinical effect of reducing artificial TFAs and the costs and DALYs saved due to averted CHD events. To calculate the percentage of reduction of CHD, we calculated CHD risks on a population-based sample before and after implementation. The effect of the policies was modelled in three ways, based on projected changes: (i) in plasma lipid profiles; (ii) in lipid and inflammatory biomarkers; and (iii) the results of prospective cohort studies. We also estimated the present economic value of DALYs and associated health-care costs of coronary heart disease averted. Findings We estimated that projected changes in lipid profile would avert 301 deaths, 1066 acute CHD events, 5237 DALYs and 17 million United States dollars (US$) in health-care costs annually. Based on the adverse effects of TFA intake reported in prospective cohort studies, 1517 deaths, 5373 acute CHD events, 26 394 DALYs and US$ 87 million would be averted annually. Conclusion Even under the most conservative scenario, reduction of TFA intake had a substantial effect on public health. These findings will help inform decision-makers in Argentina and other countries on the potential public health and economic impact of this policy.


Introduction
Artificial trans fatty acids (TFAs) are produced during the industrial processing of vegetable oils. The main source of such TFAs is partially hydrogenated vegetable oils. 1 Consumption of TFAs alters the plasma lipid profile 2,3 in such a way that it increases the risk of coronary heart disease (CHD). 4 A 2% increase in energy intake from TFAs may increase the risk for a coronary event by up to 23%. 3 Other potential adverse effects of TFAs include systemic inflammation, endothelial dysfunction, insulin resistance and arrhythmias. 2 Based on these adverse effects, several countries have implemented policies to reduce industrial TFA consumption, 5 including nutrition guidelines, awareness programmes, voluntary or mandatory labelling of the TFA content of foods and health warning labels. Voluntary or legislated programmes to encourage industry to reformulate food products without TFAs and support the production of healthy alternatives have led to improvements in some countries. 6 Mandatory food labelling in Canada 7 and the United States of America 8 have led some manufacturers to reduce or eliminate artificial TFAs in their products. However, many food products still contain such TFAs, especially when served in restaurants, schools, cafeterias and coffee shops. 9 In Argentina, before 2004, artificial TFAs were present in most sweet or salty solid snack foods, such as biscuits. 10 Between 2004 and 2014, Argentina implemented several policies to reduce artificial TFAs. After 2004, the industry voluntarily reformulated foods by replacing approximately 40% of TFAs from partially hydrogenated vegetable oils, mainly with TFAfree sunflower oil with high-oleic acid content. 11 Regulations enforcing mandatory labelling of artificial TFAs in food were introduced in 2006. 12 With support from the Pan American Health Organization, 13,14 the Argentine Ministry of Health negotiated with industry to eliminate artificial TFAs. The country's food code was amended, 15 such that, by the end of 2014, industrially-produced TFAs in food should not exceed 2% of total fats in vegetable oils and margarines and 5% of total fats in other foods (Fig. 1). 16 Here we estimate the potential reductions in annual CHD events, disability-adjusted life years (DALYs) and associated health-care costs attributable to reductions in artificial TFAs in the diet.

Methods
The main inputs of the policy model for the analysis were: (i) the estimated baseline intake of TFAs before 2004; (ii) the types of alternative oils and fats used to replace TFAs; (iii) the effects of the improvements in plasma lipid profile on CHD risks and (iv) the health-care costs and DALYs saved due to averted fatal and nonfatal CHD events. Although our study is not a full economic evaluation, we used the CHEERS statement as a guide for reporting. 17

Baseline intake of TFAs
To identify estimates of baseline TFA intake in Argentina and the fats used to improve the dietary fat profile between 2004 and 2014, we conducted a literature search using MEDLINE, Embase, LILACS databases and official documents from the government, academia, industry and other public and private organizations. For the database searches, we used the search string "trans fat OR trans fatty acids OR partially hydrogenated oils OR partially hydrogenated fat AND Argentina".

Research
Trans fats in Argentina Adolfo Rubinstein et al.
Because TFAs cannot be replaced on a 1:1 basis with other specific fatty acids, the unit of replacement was partially hydrogenated vegetable oils (comprised of various fatty acids, including TFAs). Thus, we evaluated both the total partially hydrogenated vegetable oils consumed and the usual proportion of TFAs in partially hydrogenated vegetable oils during 2004-2014. Our search was complemented by a consensus meeting of local experts and decision-makers including officials from the Ministry of Health, epidemiologists, nutritionists, cardiologists and food engineers closely involved with the oils' and fats' suppliers of TFA replacements. They identified key estimates for the model, including the baseline intake of TFA, the proportion of TFA from ruminants and the replacement fats used by industry. Our central estimate of baseline TFA consumption in 2004 was 1.5% of total energy intake, with a lower limit of 1% 18 and an upper limit of 3%. 14 The most common replacement oil was sunflower oil with high-oleic acid content (base case estimate 42.0%; range: 33.6-50.4), followed by interesterified fats (18.0%; range: 14.4-21.6) and beef tallow (12:0% range: 9.6-14.4; Table 1).

Changes in lipid profile
Improvements in the plasma lipid profile were expected to result in improvements in CHD risks. We assessed the relevant changes in plasma lipid profiles and other biomarkers of CHD risk based on meta-analyses of controlled dietary feeding trials. 2,4 These estimates were used to drive projections of CHD risks, as outlined below.

CHD risk
To estimate reductions in CHD risk in the national population, we adapted a cardiovascular risk calculator, based on the Framingham risk equation and AS-SIGN scores. 30 We used individual level data on CHD risk factors from a national prospective cohort study. 31 The study collected baseline data in 2011-2012 on age, gender, smoking, systolic blood pressure, diabetes, left ventricular hypertrophy and the ratio total cholesterol (TC)/high-density lipoprotein cholesterol (HDL-C). We combined these results with demographic data for Argentina using the 2010 census to create a national CHD risk profile. 32 According to the consensus of our expert panel, between 2004 and 2014, most of the partially hydrogenated vegetable oils in the diet were replaced by healthier fats. In 2011-2012, when the prospective cohort study took place, 75% had been replaced. We used the observed TC/HDL-C ratio for each person in the 2011-2012 cohort study to calculate the expected TC/HDL-C ratios in 2004 and 2014. This calculation was based on the estimated baseline intake of TFA, the established effects of TFA on the TC/HDL-C ratio and the types and percentages of different fats/oils used by industry for replacements. Accord-ing to the distributions of TC/HDL-C and other risk factors in the population in 2004 and 2014, we calculated the difference in the CHD risk between both years.
Three alternative scenarios were analysed: (i) the effects of improved dietary fat profile on the ratio of TC/HDL-C and the relation of this ratio to the incidence of CHD (scenario 1); 33 (ii) the CHD risk reduction through changes in other biomarkers such as apolipoprotein (apo) B, ApoA1, lipoprotein (a), triglycerides and C-reactive protein (scenario 2); and (iii) the reported relation of TFA intake, substituted for carbohydrate intake, with the incidence of CHD in a pooled analysis of prospective studies and attributed to several pathophysiological effects of TFA (scenario 3). 2

Mortality from CHD
We estimated the annual number of deaths caused by CHD using national mortality statistics for 2010. We included deaths coded according to the International Classification of Diseases (ICD-10) as I20-I25. We also assumed that 80% of the sudden deaths (ICD-10 code R96) were due to CHD. 34,35 We increased the number of CHD deaths by 21.5%, to account for the underreporting of CHD as a cause of death. 36 The difference in CHD risk predicted by the cardiovascular risk calculator was calibrated to the annual mortality from CHD. We assumed that the reduction in CHD deaths was proportional to the difference in estimated CHD risk. We also assumed that the difference in 10 year-CHD risk was equally distributed in each year of the decade 2004-2014, by age and sex.

Morbidity from CHD
Total acute CHD events -fatal and non-fatal acute myocardial infarctions and acute coronary syndrome -were estimated from national data on CHD deaths based upon sex-specific 28 day-CHD case-fatality rate for acute myocardial infarctions in southern Latin America (38% in women; 44% in men). 36 For acute coronary syndrome, we used one-third of the case-fatality rate of acute myocardial infarctions, according to local sources. 21 All values were calibrated by age-sex hospital casefatality rate in Argentina, obtained from the national hospital discharge registry for the public sector. 37

Calculation of DALYs
We calculated DALY using individual equations for years of life lost (YLL)     Notes: Discount rates of 5% (range 0-10) were used. 28 We used the average conversion rate during 2012, which was US$ 1 to 4.55 Argentine dollars. 29 and years of life with disability (YLD) according to the Global Burden of Disease Study. 38 Briefly, YLL were calculated from national health statistics as the difference between local life expectancy and age at death. YLD is the product of disability weight and length of survival with disability for CHD events. Disability weights for acute myocardial infarctions and acute coronary syndrome were considered equal. 20 Survival length was estimated using the software DISMOD II (World Health Organization (WHO), Geneva, Switzerland). 39 Finally, DALYs were reported with discounting at a 5% rate.

Costs
Cost inputs for the model were costs of acute CHD events, their follow-up and programmatic costs. A micro-costing approach was undertaken considering a health system perspective. Identification of resources related to CHD events, quantities and utilization rates were obtained from secondary local sources [22][23][24][25][26][27] and unit costs were derived from public, social security, and private tariffs of local health insurance institutions.
Costs of annual management of non-fatal CHD were calculated from the individual's age at the episode to the average Argentine life expectancy, by age and gender, and discounted at a 5% annual rate. 28 Finally, costs borne by the Ministry of Health for the implementation of annual surveillance and monitoring of the compliance of the industry with the regulations were also estimated, and included costs of personnel, food analysis, and onsite training at food companies (Daniel Ferrante, Ministry of Health, personal communication, 2013).
All costs were converted to United States dollars, corresponding to the exchange rate of 2012. 29

Sensitivity analyses
To evaluate parameter uncertainty, we performed sensitivity analyses according to established guidelines. 40 A deterministic sensitivity analysis was first performed to evaluate the uncertainty related to specific parameters and their relative importance, depicted in a tornado analysis (Fig. 2). Ranges used for the parameters were extracted from the published literature or expert opinions. To assess global uncertainty, a probabilistic sensitivity analysis was performed, incorporating the main parameters and their distributions. Uncertainty in results was reported using 95% confidence intervals (CI) based on 1000 Monte Carlo simulations. All model inputs including TFA-related, epidemiological and costs parameters are shown in Table 1.

Results
Mortality, case-fatality and acute coronary events per 100 000 population are shown in Table 2. Based on an estimated 24 875 deaths from CHD in 2010, we estimated 83 830 CHD acute events in Argentina in people older than 34 years old. The results reported here assume a baseline consumption of 1.5% of total energy intake as TFA in 2004.
Based on the most conservative scenario of TFA replacements only influencing CHD events through changes in the TC/HDL-C ratio (scenario 1), we estimated 301 CHD deaths, 572 acute myocardial infarctions, 1066 acute CHD events and 5237 DALYs averted after 2014, compared with the expected events if the policy had not been implemented (Table 3). In addition, more than US$ 17 million would be saved annually due to averted acute CHD events and lower costs of chronic treatment and follow-up.
When effects of TFA on CHD were calculated considering additional effects on other biomarkers (scenario 2), under the central estimate of 1.5% energy intake of TFA, a total of 3109 acute CHD events, 15 271 DALYs, and more than US$ 50 million in costs will be averted after 2014. If the effects of TFA on CHD were based on observed relationships with clinical events reported in prospective cohort studies (scenario 3), which may more fully account for the various effects of TFA, 1517 CHD deaths, 2884 acute myocardial infarctions, 5373 acute CHD events and 26 394 DALYs were averted, resulting in estimated savings of USD 87 million ( Table 3). The proportion of events averted by the artificial TFA reduction policy in 2014 ranged from 1.26% (scenario 1) to 6.35% (scenario 3) of total CDH events ( Table 3). The estimated reductions in CHD were sensitive to the assumed baseline TFA intake in 2004 (Fig. 2).

Discussion
Given the estimated 84 000 annual CHD events in Argentina, at an annual incidence rate of almost 500 cases per 100 000 adults older than 34 years old, the current policy of near elimination of industrial TFA might avert between 1.3% and 6.35% of CHD events each year. The decrease would save between US$ 17 million and US$ 87 million in management of CHD complications and follow-up. Even in the most conservative scenario, the reduction of TFA intake has a substantial public health impact.
Although there is limited information about the distribution of TFA intakes in subpopulations in most countries, it is likely that many subgroups, particularly low-income populations, could have mean TFA intakes considerably higher than the population mean. 41 There might be subpopulations that consume more industrially processed foods and fast foods with high-TFA content. Legislative strategies to ban artificial TFAs from foods have been more successful than labelling or education as shown in Austria, Denmark, Iceland, Sweden, Switzerland and USA. 9,[41][42][43] In Denmark, the ban on artificial TFAs is thought to have played some part in the decrease of CHD. 11 WHO has identified removal of artificial TFAs from the food supply as an intervention with favourable return of invested money to reduce the economic impact of noncommunicable diseases in low-and middle-income countries. 44 However, most such countries have not yet included the restriction of TFAs' intake as a policy. Governments have been concerned about the feasibility, achievability and public health effect of removing them from the food supply. Thus, little is known about the potential effects on the reduction of CHD burden and cost savings that could be attributable to the implementation of TFAreduction policies in these countries. Some middle-income countries such as Brazil, 5 Costa Rica, 5 India 45 and Mexico 5 are following the Argentine example and are introducing policy and surveillance systems to monitor the content of TFA in foods.
A study modelling a legislative intervention to reduce artificial TFA to 0.5% of total energy intake in the United Kingdom of Great Britain and

Trans fats in Argentina
Adolfo Rubinstein et al.
Northern Ireland, estimated that approximately 2700 deaths annually would be prevented, saving the equivalent of approximately 235 million pounds sterling a year. 46 Another modelling study estimated a similar potential impact of this policy in Ireland. 47 Unlike these studies, our model is based on individual data on CHD risk from an Argentine population-based sample, calibrated with national statistics, as well as with local data on dietary fat profiles. Moreover, our study is modelling the impact of a policy that is being implemented.
Potential limitations of this study should be considered. First, to calculate CHD risk in Argentina we used a cardiovascular risk calculator. 30 The calculator is based on equations developed a couple of decades ago when the CHD incidence was higher. This could overestimate absolute risk in light of secular trends towards lower CHD risk. 48 On the other hand, these risk equations are widely validated for predicting CHD risk. Overestimation would not likely influence our estimates of proportional risk reduction, since relative risks were calibrated with Argentine absolute risks. Second, we used the global percentage estimates to adjust for underreporting of mortality from CHD. Third, costs of food reformulation by industry were not considered, based on our health system perspective. Yet, potential incremental costs for industry to reduce artificial TFA may be at least partly offset by higher pricing or sales due to marketing advantages. 11 In the USA, switching to newer frying oils that were free of TFA was cost neutral. 38 Fourth, we did not have precise data on baseline TFA, the level of which would influence results. Conversely, our nutritional inputs, particularly those related to the TFA baseline intake before 2004, and the partially hydrogenated vegetable oils' replacements used by the industry thereafter, were obtained after a thorough literature search for sources of TFA in Argentina. This information was reviewed by experts to reach consensus on information gaps to derive a reason-  In conclusion, our findings suggest that artificial TFA reduction interventions, as an example of a nutritional policy aimed to reach the overall population, have beneficial impact on the total burden of CHD in Argentina. These findings will help inform decision-makers in both Argentina and other countries on the potential public health and economic impact of this policy. ■