The estimated impact of mandatory front-of-pack nutrition labelling policies on adult obesity prevalence and cardiovascular mortality in England: a modelling study

count: 294 Number of tables: 3 Number of figures: 2 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2

Objectives Since 2013, industry-endorsed front-of-pack traffic light labels have been implemented voluntarily on packaged food in the UK. The UK Government is now considering alternative labelling approaches which may be more effective, such as Chile’s mandatory nutrient warning labels. The primary aim of this study was to model the likely impact of implementing mandatory front-of-pack nutrition labels in England on energy intake and consequent population-level obesity, and, secondarily, cardiovascular disease (CVD) mortality. Design Microsimulation modelling analysis Setting England Model A microsimulation model (2024-2043) to estimate the impact of changing front-of- pack nutrition labels in England. The two main policy scenarios tested were mandatory implementation of (i) traffic light labels and (ii) nutrient warning labels. For each scenario, the impact of the policy through assumed changes in energy intake due to consumer behaviour change and reformulation was modelled. Main outcome measures Change in obesity prevalence (%) and CVD deaths prevented or postponed.

Compared to the baseline scenario (current voluntary implementation of traffic light labelling), mandatory implementation of traffic light labelling was estimated to reduce obesity prevalence in England by 2.28% (95% UI –4.06 to –0.96) and prevent or postpone 17000 (95% UI 4700 to 48000) CVD deaths. Mandatory implementation of nutrient warning labelling was estimated to have a larger impact; a 3.68% (95% UI –9.94 to –0.18) reduction in . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 3 obesity prevalence and the prevention/postponement of 29000 (95% UI 1200 to 110000) CVD deaths.

This work offers the first modelled estimation of the impact of introducing mandatory front-of-pack nutrition labels on health outcomes in the adult population in England. Findings suggest that mandatory implementation of nutrient warning labels would reduce rates of obesity and CVD deaths, compared to current voluntary or mandatory implementation of traffic light labelling, and should therefore be considered by the UK government. Funding: European Research Council (Grant reference: PIDS, 8031940).

microsimulation model; policy evaluation; inequalities; food labelling policies . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 4 Abbreviations BMI: Body mass index CVD: Cardiovascular disease IMD: Index of Multiple Deprivation NCD: Non-communicable diseases . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 5

1 Diet-related disease is a major cause of poor population health and social inequalities in 2 health (1). Many pre-prepared foods and non-alcoholic beverages (hereafter: food) are high in 3 calories, added sugar, salt, and/or saturated fat (2,3). Excessive consumption of these nutrients 4 increases the risk of obesity and other associated non-communicable diseases (NCD) such as 5 cardiovascular disease (CVD), and NCD mortality (4). 6 In the UK, the average adult consumes an excess of 200-300 calories per day, and nearly two-7 thirds of UK adults are living with overweight or obesity (5,6). Notably, the prevalence of 8 overweight and obesity is patterned by deprivation (14 percentage points higher in the most 9 relative to the least deprived areas), and education (12 percentage points higher for those with 10 no qualifications compared to those who are degree-level educated) (5). Therefore, there is a 11 need for equitable public health policies that improve dietary quality across the population. 12 Front-of-pack nutrition labels are an evidence-based policy tool used to help consumers make 13 healthier food choices and encourage industry to improve the nutritional profile of the 14 products they sell (7). In the UK, an industry-endorsed traffic light front-of-pack nutrition 15 label (see Figure 1.A) has been implemented voluntarily since 2013. This traffic light label 16 uses green, amber, and red colours to indicate whether a product contains low, moderate, or 17 high levels of nutrients of concern, alongside guideline daily amount (GDA) percentages for 18 each nutrient (typically per serving). However, UK consumers report that the traffic light 19 label is difficult to interpret, which may widen health inequalities (8). Additionally, less than 20 half of consumers use the label to determine product calorie content, and calorie content 21 specifically is not designated with a traffic light colour (9). It may be that simpler labels are 22 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 6 required, as most consumers typically spend no more than a few seconds examining labels 23 before making a food selection (10). 24 In July 2020, the UK Government launched a consultation considering an alternative front-of-25 pack nutrition label to the traffic light (11). In the consultation, Chile’s nutrient warning labels 26 were highlighted as a potential alternative, and the benefits of implementing mandatory front-27 of-pack labelling were discussed. 28 In 2016, Chile implemented a mandatory policy requiring packaged foods containing ‘high’ 29 amounts (as defined by thresholds set by the Ministry of Health) of calories, added sugar, 30 sodium, and/or saturated fat to display nutrient warning labels (12) (see Figure 1.B). Very 31 similar policies have since been implemented in other South American countries, including 32 Argentina, Brazil, Colombia, Mexico, Peru, and Uruguay (13,14). Mandatory nutrient 33 warnings have also been implemented further afield in Canada and Israel, and policy 34 development is under consideration in several other countries, including the US, India, and 35 South Africa (15). Evidence indicates that implementation in Chile has reduced the purchase 36 of energy (a relative 8.3% decrease, 95% CI: [5.0, 11.6]) and nutrients of concern (ranging 37 from –9.6% for saturated fat to –20.2% for sugar) (16), and has led to product reformulation 38 across all food groups, leading to reductions in energy content (-3.9%), and other labelled 39 nutrients of concern (ranging from –1.5% for saturated fat to –15% for sugar) (17). 40 Furthermore, evidence from a meta-analysis of over 100 randomised controlled trials (RCTs) 41 and quasi-experimental studies suggests that nutrient warning labels may perform better than 42 traffic light labels in terms of reducing consumers’ purchase of energy (an additional 6.4% 43 (95% CI: [0.4; 12.5] reduction) and nutrients of concern, and probability of choosing less 44 healthy products (7). Therefore, it is important to examine the potential impact of their 45 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 7 implementation in the UK on health outcomes such as adult obesity prevalence, to inform 46 policy decision-making. 47 The present study aimed to estimate the likely long-term impacts of implementing (i) 48 mandatory nutrient warning labels and (ii) mandatory traffic light labels on packaged in-store 49 foods, relative to the current voluntary implementation of traffic light labels, on energy intake 50 and consequent population-level obesity prevalence and cardiovascular mortality due to 51 change in BMI in England.52 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 8

53 Model overview 54 We built a dynamic, discrete-time, stochastic, open-cohort microsimulation model to quantify 55 the estimated effects of implementing front-of-pack nutrition labels in England; an adaptation 56 of the IMPACT NCD Model based on the IMPACT Food Policy Model (18). The model 57 simulates the life-course of individuals and their counterfactuals under alternative policy 58 scenarios. This enables the detailed simulation of diet policies and their impact on relevant 59 exposures, subsequent disease epidemiology, and mortality in a competing risk framework 60 that accounts for different lag-times between exposures and outcomes. In this case, we 61 simulated the effects of implementing mandatory front-of-pack nutrition labels (nutrient 62 warning and traffic light) on daily energy intake from packaged food, and subsequent 63 population-level obesity prevalence and CVD mortality due to change in BMI. We modelled 64 the population of England, aged 30 to 89 years, over 20 years (2024 to 2043) using a synthetic 65 population stratified by age, sex and Index of Multiple Deprivation (IMD) that captures the 66 real demographics, energy intakes, and disease epidemiology of the actual population of 67 England using available national data sources (see below and in Appendix section “Creation 68 of our synthetic population”). 69 We evaluated two main policy scenarios: 70 1. Traffic light labels are implemented as a mandatory policy 71 2. Nutrient warning labels are implemented as a mandatory policy 72 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 9 We compared each scenario with a counterfactual “no intervention” (baseline) scenario, 73 which corresponds to the current England legislation: continued voluntary implementation of 74 traffic light labels. 75 We did not model the impact of Nutri-Score, an alternative front-of-pack label which uses a 76 colour spectrum and letter grades to summarise product healthiness, as a main scenario (19), 77 This is because meta-analytic evidence suggests that it does not perform significantly 78 differently to the traffic light label in terms of reducing energy purchased (7). Instead, results 79 for Nutri-Score are presented in the Appendix (see Appendix Table 4). 80 Front-of-pack nutrition labels 81 Front-of-pack nutrition labels impact diet through (1) consumer behaviour change, and (2) 82 industry response, i.e., reformulation of the products by industry (see Figure 2). 83 Effect on consumer behaviour change 84 We assumed that the traffic light labels and nutrient warning labels would reduce energy 85 purchased from packaged food by 6.5% (95% CI: [2.0; 11.0]), 12.9% (95% CI: [8.0; 18.0]), 86 and 6% (95% CI: [1.0; 11.0]) respectively, compared to no label, based on the estimates from 87 Song et al.’s review and network meta-analysis (7). Based on the same meta-analysis, we 88 assume that nutrient warning labels will outperform traffic light labels in reducing the total 89 amount of energy purchased by 6.4% (95% CI: [0.4; 12.5]). Based on existing literature, we 90 assumed no differential policy effects by sex, age or socioeconomic position (7,20). Due to an 91 absence of evidence, we assumed both labels have a consistent effect on consumer behaviour 92 over time. 93 Effect on energy content reformulation 94 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 10 For nutrient warning labels, we assumed a 3.9% (95% CI: [12.5; 4.95]) reduction in energy 95 content of labelled packaged foods, based on evidence from Chile post-implementation (17). 96 While there is no available data specifically in relation to traffic light labelling and product 97 reformulation, evidence suggests that a small amount of reformulation does occur in response 98 to food labelling, particularly when it is implemented mandatorily (21–23). Therefore, we 99 also assumed the same 3.9% reduction in energy content of packaged foods in response to 100 mandatory traffic light labelling. 101 Label coverage 102 We assumed that all packaged products (100%) would feature a traffic light label, as under 103 mandatory implementation, this would be required by law (16). Under current voluntary 104 implementation, it is estimated that 75% of packaged products feature the label (24), so 105 mandatory implementation would yield an additional 25% coverage. For nutrient warning 106 labels, based on evidence on the proportion of products featuring a “high in” warning in 107 Chile, we assumed that 51% (95% CI: [49.0; 52.0]) of packaged foods in England would 108 feature the label (i.e., will be above threshold for warning) (25). The nutritional quality of 109 packaged food in Chile is relatively similar to the UK; the average Health Star Rating for 110 packaged food is 2.44 compared to 2.83 (scores range from 0.5 to 5, with a higher score 111 indicating better nutritional quality) (26). Moreover, an analysis of food items from the UK 112 NDNS indicated that approximately 40% of UK food items meet requirements for a red traffic 113 light label, and this figure does not include items that would be labelled due to being high in 114 energy (27). Research suggests that 32% of UK supermarket snack foods alone exceed adult 115 energy intake recommendations (3) and therefore it is reasonable to estimate that this would 116 amount to at least an additional 10% of products being labelled, consistent with the 51% 117 figure derived from Chile. 118 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 11 Estimating model uncertainty 119 We used the Monte Carlo approach (100 iterations) to estimate the uncertainty of model 120 parameters. The sources of uncertainty we considered were the uncertainty of the relative risk 121 of coronary heart disease (CHD) and stroke based on BMI, the uncertainty of mortality 122 forecasts, and the uncertainty of the policy (label) effect. We summarised the output 123 distributions by reporting the medians and 95% uncertainty intervals (UIs). 124 One-way sensitivity analyses on key parameters 125 Change in nutrient warning labels coverage 126 Evidence from Chile suggests that approximately one year after initial implementation of the 127 nutrient warning label policy, reformulation resulted in a decrease in the proportion of 128 products featuring a label from 51% to 44% (95% CI: [42.0 - 45.0]) (25). Reformulation to 129 reduce nutrients of concern is consistently observed in response to the introduction of front-130 of-pack nutrition labelling policies in various countries, including Australia, Canada, the 131 Netherlands, and New Zealand, to avoid a “negative” label (e.g., a low health rating) or the 132 absence of a “positive” label (e.g., a healthy choice indicator) (28). Therefore, in this 133 sensitivity analysis we assume that coverage is 51% for the first-year post-implementation, 134 and coverage then drops to 44% thereafter. 135 Chile’s black octagon specifically (as opposed to nutrient warning labels more generally) 136 In this sensitivity analysis, we test based on evidence from Chile specifically, post-137 implementation (as opposed to meta-analytic data on nutrient warning labels in general from 138 experimental studies), which suggests an overall 8.8% (95% CI: [-7.1 to –10.5]) reduction in 139 energy purchased (16). Notably, nutrient warning labels were introduced in Chile as part of a 140 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 12 set of policies, including restrictions on food marketing to children, and therefore this 141 reduction in energy purchase may not me wholly attributable to nutrient warning label 142 implementation. 143 Lower reformulation due to traffic light labels 144 It is possible that reformulation of energy content may be lower in response to traffic light 145 labelling relative to nutrient warning labelling. This is because calories are not colour-coded 146 in traffic light labels and therefore food companies may be less inclined to reformulate energy 147 content of products. We assumed there would be a smaller 0.9% (95% CI [-3.1, 4.9]) 148 reduction in energy content, based on a meta-analysis of food labelling effects on product 149 energy reformulation (23). 150 Table 1: Summary of key model assumptions 151 Traffic light label Nutrient warning label Main assumptions Effect on energy intake -6.5% [-11%; -2%] (7) -12.9% [-18%; -8%] (outperforms the traffic light label by 6.4% [0.4; 12.5] (7) Effect on reformulation in terms of energy content -3.9% [-12.5; 4.95] (17) -3.9% [-12.5; 4.95] (17) Label coverage on packaged products 100% (currently 75% under voluntary implementation) (24) 51% [49%; 52%] (25) Sensitivity assumptions Changes in label coverage over time due to reformulation - Drops to 44% [42.0; 45.0] 4 years post-implementation (25) Chile’s black octagon nutrient warning label effectiveness on energy intake from labelled products - -8.8% [7.1.; -10.5] (16) Effect on reformulation in terms of energy content -0.9% [-3.1, 4.9] (23) - . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 13 A further detailed description of the model, input sources, and key assumptions are provided 152 in the Appendix. 153 Model engine 154 Front-of-pack nutrition labels are hypothesised to reduce energy intake, which will 155 subsequently impact the body weight of the population (i.e., BMI), and, in turn, change CVD 156 mortality risk. This pathway is described in Figure 2 and detail in Appendix (section 157 “Estimating the effect of change in energy intake upon obesity prevalence and CVD 158 mortality”). In short, the change in energy intake is calculated by subtracting intake post-159 intervention from baseline intake for each year. Changes in energy intake are then converted 160 into changes in body weight, based on principles of energy conservation, using the 161 Christiansen & Garby prediction formula (29) (detail in Appendix section “Estimating the 162 effect of change in energy intake on BMI”). The estimated change in BMI is then calculated 163 based on the estimated change in body weight, which allows us to estimate the change in 164 obesity prevalence. Next, these changes in BMI are used to estimate changes in CVD 165 mortality risk, with a 6-year lag time (30) (see details in Appendix section “Estimating the 166 effect of change in BMI upon CVD mortality”). Using this information, new mortality rates 167 and, consequently, the number of deaths projected can be estimated. 168 Model outputs 169 The model produced the change in obesity prevalence and the total number of deaths 170 prevented or postponed (DPPs) for each scenario. The equity impact of the intervention was 171 examined by calculating the ratio between the most and least deprived quintile groups (using 172 the IMD). Results are presented for English adults aged 30 to 89 years from 2024 to 2043, 173 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 14 rounded to 2 significant figures for mortality and rounded to 2 decimal places for obesity 174 prevalence. 175 Data sources 176 We constructed a synthetic population of England to simulate the population-level impact of 177 the policy scenarios. This is described in the Appendix section “Data sources used in our 178 model” and Appendix Table 1. The England population projections were derived from the 179 Office for National Statistics (ONS), and mortality trend projections were based on the CVD 180 deaths observed in England from 1981 to 2016. 181 We used generalised additive models for location, shape and scale (GAMLSS) to estimate (i) 182 BMI and (ii) energy intake distributions dependent on age, sex, and IMD. GAMLSS can 183 handle complex relationships between the response variable and its predictors and numerous 184 types of distributions (31). Trends in energy intake daily energy intakes and BMI were 185 obtained from the nationally representative National Diet and Nutrition Survey (NDNS) 186 2009-2019. These trends in energy intake and BMI observed in the last 10 years in England 187 were assumed to continue in the future. To obtain the daily energy from packaged food 188 bought from grocery retail stores, we assumed that 55% of all food and beverage expenditure 189 (including alcoholic beverages) was for at-home consumption (vs. 45% spent on restaurants 190 and other out-of-home food services) (32) and that 80% of the products purchased are 191 packaged (vs. 20% fresh) (8) (see details in Appendix section “Modelling approach and 192 scenarios”). 193 R (version 4.3.0) was used to conduct all data management and statistical analyses. We used 194 the “demography” package (33) for forecasting mortality and the “gamlss” package to fit the 195 distribution (34). For code, see https://github.com/zoecolombet/FoPLabels_code 196 197 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 15

198 Maintaining current voluntary traffic light labelling would result in obesity prevalence of 199 28.03% (95% UI 27.74 - 28.30) by 2043. 200 The implementation of mandatory traffic light labelling in England was estimated to reduce 201 obesity prevalence by 1.49 percentage points (absolute; 95% UI –2.44 to -0.76; Table 2) in 202 the next 20 years when only considering consumer behaviour change (i.e., change in energy 203 intake). Reformulation of the energy content of the packaged products sold was estimated to 204 lower obesity prevalence by 0.66 percentage points (95% UI –2.79 to 0.00; Table 2). 205 Combining these factors would result in a decrease of 2.28 percentage points in obesity 206 prevalence among adults (95% UI –4.06 to –0.96; Table 2). 207 Implementing mandatory nutrient warning labels on packaged products was estimated to have 208 a larger impact and reduce obesity prevalence by 2.31 percentage points (95% UI –6.79 to –209 0.02; Table 2) when only considering consumer behaviour change. Reformulation of the 210 energy content of the packaged products sold was estimated to lower obesity prevalence by 211 0.96 percentage points (95% UI –6.10 to 0; Table 2). Combining these factors would result in 212 a decrease of 3.68 percentage points in obesity prevalence among adults (95% UI -9.94 to –213 0.18; Table 2). 214 Maintaining current voluntary implementation of traffic light labelling in England, the current 215 cardiovascular mortality trends were estimated to result in approximately 1,900,000 deaths 216 (95% UI 1,100,000 – 3,300,000) in English adults by 2043. 217 218 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 16 Implementing traffic light labelling mandatorily would prevent or postpone approximately 219 7300 deaths (95% UI 2500 to 21000; Table 2) attributable to BMI-related CVD, based on 220 consumer behaviour change alone. Reformulation was estimated to avert 2500 deaths (95% 221 UI 0 to 17000; Table 2). Combined, this would result in 17000 deaths (95% UI 4700 to 222 48000; Table 2) prevented or postponed. 223 Again, implementing mandatory nutrient warning labels was estimated to have a larger 224 impact, resulting in the prevention or postponement of an estimated 14300 (95% UI 240 to 225 54000) deaths based on consumer behaviour change, 4300 deaths (95% UI 0 to 42000; Table 226 2) based on reformulation, and 29000 deaths (95% UI 1200 to 110000; Table 2) based on the 227 two combined. 228 229 230 231 232 233 234 235 236 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 17 Table 2: Estimated change in obesity prevalence and CVD mortality due to change in BMI in 237 adults in England (2024–43), according to different front-of-pack labelling implementation 238 scenarios 239 Change in prevalence of obesity (%) CVD deaths prevented or postponed* Consumer behaviour change Traffic light labelling (mandatory) -1.49 (- 2.44, -0.76) 7300 (2500, 21000) Nutrient warning labelling (mandatory) -2.31 (-6.79, -0.02) 14300 (240, 54000) Reformulation Traffic light labelling (mandatory) -0.66 (-2.79, 0) 2500 (0, 17000) Nutrient warning labelling (mandatory) -0.96 (-6.10, 0) 4300 (0, 42000) Combined Traffic light labelling (mandatory) -2.28 (-4.06, -0.96) 17000 (4700, 48000) Nutrient warning labelling (mandatory) -3.68 (-9.94, -0.18) 29000 (1200, 110000) *Results from 2024 to 2043. 240 241 242 243 244 245 246 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 18 Table 3: Estimated change in obesity prevalence and CVD mortality due to change in BMI in 247 adults in England (2024–43), according to IMD quintile groups and different front-of-pack 248 labelling implementation scenarios 249 Prevalence of obesity, percentage points CVD deaths Predicted obesity prevalence CVD deaths predicted Current voluntary traffic light labelling Q1 (most deprived) 32.53 (32.00, 33.04) 470,000 (270,000 – 830,000) Q5 (least deprived) 24.29 (23.55, 24.85) 290,000 (170,000 – 500, 000) Predicted change in obesity prevalence CVD deaths prevented or postponed Mandatory traffic light labelling – consumer behaviour change Q1 -1.46 (-2.24, -0.71) 2000 (240, 5500) Q5 -1.48 (-2.46, -0.75) 1000 (0, 4500) Mandatory traffic light labelling - reformulation Q1 -0.66 (-2.85, 0) 500 (0, 6500) Q5 -0.65 (-2.73, 0) 250 (0, 2000) Mandatory traffic light labelling - combined Q1 -2.14 (-3.96, -0.91) 4000 (740, 14000) Q5 -2.28 (-4.08, -0.93) 2500 (500, 8000) Mandatory nutrient warning labelling – consumer behaviour change Q1 -2.25 (-6.25, -0.01) 3500 (0, 13000) Q5 -2.31 (-6.81, -0.03) 2000 (0, 8800) Mandatory nutrient warning labelling - reformulation Q1 -0.90 (-5.58, 0) 1000 (0, 12000) Q5 -1.05 (-6.20, 0) 500 (0, 5500) Mandatory nutrient warning labelling - combined Q1 -3.61 (-9.58, -0.19) 7500 (0, 30000) Q5 -3.59 (-9.80, -0.20) 4500 (0, 18000) 250 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 19 The introduction of either front-of-package label as a mandatory policy is estimated to reduce 251 obesity prevalence and relative CVD deaths to a similar extent across socioeconomic 252 deprivation levels (see Table 3). 253 See Appendix Table 3 for sensitivity analysis results relating to nutrient warning label 254 coverage, Chile’s nutrient warning label specifically, and traffic light label reformulation. 255 Briefly, nutrient warning labels with reduced coverage, and Chile’s warning label specifically 256 still outperformed traffic light labels. Traffic light labels saw a notable decrease in 257 performance using the more conservative reformulation estimate. See Appendix Table 4 for 258

relating to Nutri Score. As expected, results for Nutri Score were very similar to those 259 for traffic light labelling. 260 261 262 263 264 265 266 267 268 269 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 20

270 This work offers the first modelled estimation of the impact of changing front-of-pack 271 nutrition label policy on obesity prevalence and CVD mortality in the adult population in 272 England. Our findings indicate that, in place of current voluntary traffic light labelling, the 273

of mandatory nutrient warning labels would reduce obesity prevalence and CVD 274 deaths substantially more than making traffic light labels mandatory, with no differential 275 effects on health inequalities. 276 Our findings are largely consistent with the existing limited evidence in this area. One 277 previous study modelled the impact of nutrient warning labels in Mexico (35). The study 278 estimated a mean caloric reduction of 36.8 kcal/day/person, and, 5 years post-implementation, 279 1.3 million fewer cases of obesity (5% reduction). A handful of studies have modelled the 280 impact of traffic light labelling on NCD mortality. One study modelling impact in Canada 281 (36) estimated that 11715 deaths per year due to diet-related NCDs, and 10490 deaths per 282 year due to energy intake alone would be prevented. However, this was contingent on 283 Canadians using the traffic light labelling to avoid foods labelled with red lights. Another 284 study estimated the impact of Nutri-Couleurs (traffic light label) across 27 EU nations and 285 found no significant effect on NCD mortality (37). However, the effect estimate for change in 286 energy intake was derived from a large-scale randomised controlled trial in French 287 supermarkets which only covered four product types (bread, ready meals, fresh catering, and 288 pastries) (38), as opposed to the use of meta-analytic evidence in the present research. 289 Although the current research provides important insights into the likely impact of changing 290 front-of-pack nutrition label policy in England, there are limitations to be acknowledged. We 291 assumed that reductions in energy intake would be in response to labelled products, which 292 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 21 may be an overestimate for traffic light labels as not all products would feature a “red” 293 indicator. We also assumed that energy intake trends from NDNS will continue, but it is 294 possible that COVID-19 and/or the cost-of-living crisis may result in long-term changes. Our 295

will also underestimate total policy benefits as we did not include changes in 296 childhood obesity in our model. 297 It is also important to acknowledge that the present research underestimates the impact of the 298 labelling policies on total CVD mortality as due to model design we do not model effects of 299 policies due to changes in intake of nutrients of concern (salt, sugar, saturated fat) and instead 300 model change via energy intake and reductions to BMI. Excess intake of salt, sugar, and 301 saturated fat is associated with CVD risk (39). Evidence suggests that labelling policies 302 decrease the purchase of nutrients of concern, especially nutrient warning labels relative to 303 traffic light labels, so impacts on CVD mortality are likely to be particularly underestimated 304 for nutrient warning labels (7,20). 305 We did not model a scenario where nutrient warning labels are implemented voluntarily, as 306 there are no examples of such implementation. Moreover, the current evidence suggest that 307 voluntary, industry-endorsed initiatives in the context of front-of-package labelling are likely 308 to be ineffective for several reasons, such as industry manipulation of label design, 309 noncompliance (particularly as nutrient warning labels are known to deter purchase of 310 labelled products), and a lack of independent target setting, monitoring, and enforcement 311 (40,41)Finally, while nutrient warning labels appear effective in reducing purchase and intake 312 of energy and nutrients of concern, it may be that alternative/additional labels are required to 313 encourage consumers to select health protective food options (i.e., those that contain nutrients 314 that the population do not consume enough of, e.g., fiber, vitamin D). 315 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 22 Several assumptions in our model were constrained by a lack of available evidence and these 316 areas might benefit from further research. Firstly, there was no available data on how the 317 effect of the label on consumer behaviour change may change over time. Theoretically, if 318 people become habituated to front of pack labels, then the effect may decrease, or conversely, 319 if nutrient literacy and awareness strengthen over time then the effect may increase (8). 320 Secondly, there was no available data on compensatory effects from intake of fresh food in 321 place of packaged food, or intake from out-of-home eating. Thirdly, although there is some 322 self-report evidence to suggest that age, education, and ethnicity may impact understanding 323 of, and therefore response to traffic light labels (8), there was no consistent evidence that 324 demographic factors moderate the effect of labels on product choice (7,20). 325 The World Health Organization (WHO) does not at present recommend the use of any 326 specific labelling scheme but encourages research institutions and member states to continue 327 analysing information to inform decisions (42). This new modelled evidence supports the use 328 of nutrient warning labels to reduce population-level obesity. While such labels are gaining 329 global popularity, the UK and Europe are yet to adopt this policy approach. It is 330 recommended that the UK Government replaces its current voluntary traffic light labelling 331 system with mandatory nutrient warning labelling to reduce rates of obesity and related CVD 332 deaths. 333

334 Mandatory implementation of nutrient warning labels appears to be the most favorable policy 335 option for the UK government to substantially reduce rates of obesity, compared to current 336 voluntary or mandatory implementation of traffic light labelling.337 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 23 Declarations Data sharing ONS and NDNS data are available online. The “demography” package for R software has been used for forecasting mortality and the “gamlss” package has been used to fit the distribution. Syntax for the generation of derived variables and for the analysis used in this study are available publicly: https://github.com/zoecolombet/FoPLabels_code Funding Salaries for ZC and ER were fully and part-funded, respectively, by the European Research Council under the European Union’s Horizon 2020 research and innovation programme (Grant reference: PIDS, 803194). ER and RE are funded by the National Institute for Health and Care Research (NIHR) Oxford Health Biomedical Research Centre (BRC) (Grant

NIHR203316). Role of the funding source The funder played no role in the study design, data collection, data analysis, data interpretation, writing of the paper, or the decision to submit this work for publication. Competing interest statement All authors have completed the Unified Competing Interest Form and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work. Transparency declaration . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 24 The lead author (R.E) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained. Copyright statement The Corresponding Author (R.E.) has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive licence on a worldwide basis to the BMJ Publishing Group Ltd to permit this article to be published in BMJ editions and any other BMJPGL products and sublicences such use and exploit all subsidiary rights, as set out in our licence. Ethical approval Ethical approval was not required for this study. Authors’ contributions ZC, RE, ER, MO'F, and EP designed the study. ZC and RE directly accessed and verified the underlying data reported in this article. ZC and RE developed the model. CK, MO'F, and ER supervised ZC and RE. RE and ZC did the analysis and drafted the manuscript. All authors contributed to the data interpretation and revised each draft for important intellectual content. All authors had final responsibility for the decision to submit for publication. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 25

1 . Gilmore A B, F abb ri A, Ba um F, B e rt scher A, Bon dy K, C hang H J, et al . D efi ning and con ceptual i s in g t he c ommercia l dete rmi nant s o f hea l t h . The L anc e t. 2023 A p r;40 1 (10383):119 4 – 213. 2 . Huang Y, Burgoine T, Es sma n M, Th ei s DRZ , Bishop TR P, Adams J . Moni tor i ng the Nutri ent Compo siti on of F o od Pr epa red O u t -of -Hom e in the U ni ted Kingdom: D a ta ba s e De velop men t and Ca se S tudy . JM IR Publ i c Healt h S urveill . 2 022 Sep 8;8 (9) :e39033. 3 . Marty L , Eva ns R, Sh een F, Humph rey s G, Jone s A, B oy land E , et al . The en e r g y an d nutr i t i o nal con t e n t of s na ck s sold at sup ermark et s a nd coff ee shop s in t he UK. J Huma n N u tri tion Di et . 2 021 Dec;3 4 (6):1035 –41 . 4 . Anand SS, Haw ke s C , De S ouz a R J, Men te A, Dehg ha n M, Nugen t R, e t al. F o od Co nsumpt ion a nd its Impac t on Cardi ova s c ul ar Di s ea se: Importanc e of So l ut i on s F ocu s ed on t h e Glo balized F ood Sy s tem . Jou rna l o f th e Americ an Coll eg e of C a r diology . 2015 O c t ;6 6(14 ):1590– 614 . 5 . Bak er C . Re se arch Br i e fi ng - Ob es ity s t a t i stic s [In tern et ]. H ou s e o f Common s L ibra r ; 2023 [c it ed 2024 Apr 17]. Avai lable fr om: http s : / / re s earc hbri e fings .fil e s.p arliame n t . uk/documen ts / S N 0 3336 /SN03336 .p df 6 . Teds tone A, T arge tt V, Mac kinla y B, Owt r am G, Coul ton V , Mo r g an K, e t al . Ca l orie reduc t io n : The scope a nd ambi t i o n fo r action [In t e r ne t] . P ublic H e al th Engl and; 20 18 [ci ted 2024 Apr 17] . Avai lable f rom: http s : / / a sse ts.p ubli shing. s ervic e.gov .uk / governmen t/uploa ds/ s y s tem /uplo ad s/a t t a chmen t_ da ta /fil e/800675 /C alo r i e s_E videnc e _Docu m e nt.pd f 7 . Song J, Brown MK, Tan M, M a cGre gor G A , W e b ster J, C am pbell NR C, e t a l. Impa c t of co lo r -c oded and warning nutr i tion lab el ling sch e me s : A s y s tem atic revie w and n etwork me t a - a nalysi s. A re s G, edito r. PL oS M ed. 2021 Oc t 5 ;18(10 ):e10 0 3765. 8 . Os ma n M, J enkins S . Con sum er r e s p on se s to food l a belling : A rapid e vide nc e revi e w [In tern et ]. Food Standards Agenc y; 20 21 [ci t e d 202 4 Apr 17 ]. Avail able from: http s : / / w ww .food.gov. uk/re sea rc h/ c on su mer-re spon s e s - to -f ood-labe lli ng-a - r api d -evid en c e- revie w 9 . Food Standards Agenc y. E a ting Well Cho osing Be tt er Trac ke r Surve y Wave 8 202 2 [In tern et ]. 2022 . Avail able from: h ttp s ://ww w.food. gov .uk/re s e arch /ewc b-2022 -r esu lts 1 0. Sa njari SS , Ja hn S , Boz t ug Y. Dual -pr oc e ss theory a nd c on s ume r re spo ns e to fron t- of -pa cka ge nutriti on lab el forma t s . N utri tion R ev iews. 20 17 Nov 1;7 5 (11):871 –82 . 1 1. GOV.UK. F r ont -o f -pac k nutr ition la belling in the UK: bu ilding on s u cce ss [ Int erne t ]. 2020. Avai lable f rom: h t tp s: //ww w.go v.uk/gov ernmen t/ c on s u l t a t i on s / fr ont-o f- pac k-nu triti on- labe lling-i n-the -uk -building - on- s uc c es s . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 26 1 2. Rey e s M, G a rmen dia ML, O l ivare s S, Aqu e veque C , Zac aría s I, Co r v alá n C . Develo p me nt of the Chile an fro nt -o f -p a cka ge fo od warning la bel. BM C Public H ealth . 20 19 Dec ;1 9(1) :9 06. 1 3. Tai llie L S, Hal l MG, G ó mez L F, H ig gin s I, B erc holz M , Mu r uk utl a N, e t al . De s i gning a n Eff ec t i ve Front -o f- Pack age Wa r ni ng L abel for F o od and Drink s High in Added Suga r, S odium , or S atu rat ed Fa t in Col ombi a: An Online Experime n t . N utri ent s . 2020 O c t 13;12(10 ):312 4 . 1 4. White M , Barqu e r a S . Mex ico A do p ts Fo o d Warning L abel s, Why Now ? Heal th S y ste ms & Refo rm. 202 0 Dec ;6(1 ):e 1752063 . 1 5. UN C Glob al Fo o d Re se arch P rog r a m. F ro nt-o f - P ack age (F O P ) F ood L abelli ng: Empow ering con s ume r s and pr omo tin g heal t hy diet s [ Inte rne t]. Avail abl e fr om: http s : / / w ww .global fo odr e sea r c hp rogra m.org/wp - con t e n t /upl oa d s/2022/10 /FO P_Fa c t s he e t_H S R _upda te .p df 1 6. Tai llie L S, Be r c holz M , Popk in B , Reb oll edo N, Rey e s M, Corvalá n MC . Dec rea ses in pu r c has e s of energy , sodium, sugar, a nd s atu r a te d f at 3 ye ars a fter i mplem enta tion o f th e Chil e a n food labe ling a nd marke t i ng law : An in ter rupt d time s erie s a nal ysi s. P LOS M edic ine . 2 024 Se p 27;21 ( 9 ): e100446 3. 1 7. Quin tiliano Sca r pe lli D, Pi n heir o Ferna ndes AC , R od riguez Osia c L , Pizar r o Q uev e do T. Change s in Nutr i e nt D e cla rati on af t e r the Fo od L abe li ng and Adverti s i ng L aw i n Chile : A Longi t u dinal Approac h . N u t rien t s. 2020 Aug 8;1 2(8 ):2 371. 1 8. Pea r s on- Stu tta rd J, B ando sz P , Rehm C D , P enalv o J , Whit sel L, Gazia no T, et a l . Re duc ing US ca r di ova scul ar di se ase burd en an d di spa ritie s t h roug h na tiona l an d targ e t e d diet ary po licie s : A model ling s tudy. P LoS Me d. 2017 Jun ;14( 6):e100 2311. 1 9. Sa nt é publi que Fr anc e. NU T RI - S C O RE Q u e stion s & A n s we rs Eng lis h ver sion [ Int er net ]. 2024 . Avai lable f rom: http s : / / w ww .s a n t e pub lique franc e. fr / c on tent / dow nlo ad/15026 3 / fil e /Q R_ scien ti fi que _ techni qu e _EN_12 052020.p df 2 0. Croke r H, Pa cke r J, Ru ss ell S J, S tan sfi eld C, Vin er RM . Fron t of p a ck n ut ri tiona l lab e lling s c hem es : a sy stema tic r e view and me t a -analy s i s o f rec ent evide nce rel ating t o obj ective ly mea s u red con s ump tion and purc h a s i ng. J Human N utriti on Di et. 202 0 Aug;33 (4) :5 18–37 . 2 1. Gande rat s - F u ent e s M, Morga n S . Fron t -o f-P ack age Nutri tion L abel ing and I ts Imp a c t on Fo od Indu str y P ractic es : A Sys tema tic Revie w of th e E viden ce. N utri ent s. 2023 Jun 5 ;15 (11):2630 . 2 2. Vand ev ijve r e S, Van derl ee L . Ef fec t of Fo r mulation, Lab el ling, and Ta xation P olici e s on the Nutr i tiona l Qu ality o f t he Fo od S upply . C urr N u tr Rep . 2 019 Se p;8 (3) :240–9 . 2 3. Sha nggua n S, Af sh in A, S hulkin M, M a W, Mar s den D, Smi t h J, et al . A Me ta -An a lys is o f Food La beling Eff ect s on C on sumer Die t Beha v iors and Ind u stry Pr ac t i ce s. Americ an Jou rnal of Preve n t i ve Medi cin e. 2019 F eb ;56(2 ):30 0 –14. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 27 2 4. House of Common s Heal th C om mittee . C hildho od obe si ty— brave an d bold actio n [In t e r n e t]. 2015 . Avail able from: http s : / / p ublica t i on s .p a r l iamen t.uk / pa /c m 201516/cmsel e ct/ c mh eal th/465 /465. p df 2 5. Rey e s M, Smith Ta illie L, Pop kin B, Kan ter R, Vande vijve r e S, Co rva lán C . Cha nge s in t h e amoun t of nut rien t of p a cka ged fo ods and bever ag es a f ter t he initi al impleme n t a t i on of t he Chil e an L aw of Fo od La b elling and Adv er t i sing : A non ex periment al pro sp ec t iv e study. War eha m N J, e di tor. PL oS Med. 2 020 Jul 2 8;17(7 ):e100 3220 . 2 6. Dunf ord EK, Ni Mhu rc hu C, Hua ng L, Va n devijv ere S, Swi nbur n B, Prav st I , et a l . A co mparison of the he a lthin e ss of pack aged f o od s a nd b eve r a ge s f r o m 12 c ountr i e s u sing the H e alth S t a r Ra t i ng nutrien t pro filing s y stem, 2013–20 18 . Ob es i t y Revi ews . 2019 N ov ;20( S2):10 7 –15. 2 7. Dick en SJ, B at t e r h a m RL, Brow n A. Nutr i e nts or pr oce ssing? An an al ysi s o f f o o d a nd drink item s from th e UK Na tional D ie t and Nu tr i t i on Survey bas ed on n utrien t c onte nt, the N OV A cl assi fic atio n and f r ont o f pack age tra ffic light la belli ng. B r J Nu t r . 2 024 May 14;131(9 ) : 1619 –32. 2 8. Rober to C A, Ng S W , G a ndera t s -Fu e nte s M, Hamm ond D, B arqu era S , Jaur egui A, et al. The I n f l ue nc e of Fr ont -o f-P a ck a ge N ut r i t io n L a b el i n g on C on su m e r B eh a vi o r a n d P r o d u ct Refo rmulati on. An nu Rev N ut r. 202 1 O c t 11;41(1) :5 29–50. 2 9. Chri st i a n s e n E, G arby L. Pr edic tio n of bo dy we ight cha nges c au se d by c hange s i n energy bala nce . Eur J Cli n Inve s t. 2002 N o v;32 (11) :826 –3 0. 3 0. Emergi ng Risk F acto r s Colla bo r a tion , Wo r mse r D, Kap tog e S, Di A n gela ntoni o E, Wood AM , Penn e ll s L, e t al. S e par at e and combin ed as s oc ia tion s o f body -ma ss ind ex a nd abd omin al adip o s ity with c ardiova sc ula r di sea s e: co l labora tive anal y s i s of 5 8 pro s p e ctive s tu die s. L anc et. 2011 Mar 26;377(9 771) :1085– 95. 3 1. Sta sinopou l o s MD , Rig by RA, Hell e r G Z , V oudouri s V , Ba sti ani F D. Flex ible Regr e ssi on a nd Smoo t hi ng: U s in g G AM L SS in R. N ew Yor k: Cha pman an d Ha ll/ CR C; 2017. 571 p. 3 2. Stati s ta. Sh a r e o f hou s ehold f ood and dr i nk expe nditure in the Uni ted K ingdom ( U K) from 2000 to 1s t quar te r 2020, b y at-h ome a nd ou t- of -home c on s u mpt i on. Avai l able f rom: http s : / / w ww .s t ati s ta.com / st a t i stic s /941699/in-h ome - v er s u s- out -o f-ho me -f ood -a nd-drink - s p en di n g - u nit ed - k i n g dom- uk / 3 3. Hyndman R, Booth H , Ti ckle L, Ma indona ld J. Pac kag e ‘demog r a phy’ fo r R [I nte rne t] . [c i t e d 2022 Nov 9]. A v ail a ble from: h ttp s : / / g ithub .co m/robjhy ndman/d emography 3 4. Sta sinopou l o s M, Rigby R, Vou douri s V, A kantzilio tou C, En ea M , Kio se D, e t al. Pac k age ‘gamls s’ : Gene raliz ed Addi t iv e Model s for L oca tio n Sca le and Shap e [ Inte r ne t ] . 2024 [c it ed 2024 Apr 2 4]. Avai lable f rom: h t tp s: //cran . r-projec t .or g / w eb/p ack age s/ g aml s s/gaml ss .pdf 3 5. Bas to -A b reu A, To rr e s-Alv are z R, R eye s-S ánc hez F, G onzále z- Mor a l e s R, C anto - Osorio F , C olc hero MA, e t al . P redic ting obe sity r educ tion a f ter imple m enting w arning l ab el s in Mex ico : A modeli ng st u dy . Clé m ent K, edi tor . PL oS M ed . 202 0 Jul 28;17(7) :e1 003221. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint 28 3 6. La bonté ME, E mr i ch TE , Sc arbo rough P , Rayner M, L ’Ab bé MR . T r a ffi c lig ht lab el li ng c ould preve nt mo r tal i ty from nonc ommunic abl e di sea s e s in C a nada : A s c ena rio mode lli ng s tudy. Vadiv elo o MK, edit or. PLoS ON E . 2019 D ec 27;14(12 ): e022697 5. 3 7. Deva ux M, Alde a A, L ero uge A, Vui k S, C e cc hini M. E s t abli s hin g an E U -w id e fr ont -o f -p a ck nutriti on lab el : Rev iew o f optio ns a nd mode l- ba s e d eva lu atio n. Ob e sity Rev iew s. 20 24 Jun;25 (6): e1 3719. 3 8. Allai s O, Albu qu erqu e P , Bonn e t C , Duboi s P . Év aluation Expé r im e nta tion Lo go s Nutriti onnel s Rappor t pour le FF AS [ Int er n e t]. 2017. Av ailab le from: http s : / / s ant e.g ouv. fr / IM G /pdf / r appor t_ f inal _group e_ trai temen t_ e valua t i on_log os .pdf 3 9. Bowe n KJ, Sul liva n VK, Kri s - E t h e r t on P M, Pet er s e n KS. Nut ritio n and C ardiova scu la r Di s ea se—an Updat e. Cu rr A the ro s c ler R ep. 2 018 Jan 3 0;20(2) :8. 4 0. La verty AA, Ky pr i demo s C, Se f e rid i P, Va mo s E P, Pea r son -Stu tta r d J , Collin s B, e t a l. Quan tifying the imp act o f th e P ublic He alth R e spon si bility Dea l o n s alt int ak e, c ardi ova s c ul ar dise a se and gast ric c anc er bur den s : in terrupt ed t i me se r i e s a nd mic r o simul a tion st u dy . J Epi de miol Communi t y Heal th. 20 19 Se p ;73(9 ):881– 7. 4 1. Knai C, P ett icrew M, D ougla s N, Durand MA, E a s tmur e E , N o l t e E, e t a l. The Publi c Health Res pon sibility Dea l : Us i ng a S yste ms-L ev el A na ly sis t o U nde rst and th e L ac k of Im pa ct on Alc ohol , Food, Phy sica l Ac tiv ity, an d Wo r k plac e H ealth S ub -Sy ste ms. IJER PH. 2018 Dec 17; 15 ( 1 2):2895 . 4 2. World He alth O rgani sa tion. Guidi ng prin c iples an d fr ame work ma nual for fron t-o f-p a ck labell ing for pr omoting he althy di e ts [I nte rne t] . 2 019 [cited 2 024 Jul 1 5 ]. Avai l able from: http s : / / w ww .who.int /public a tion s/ m /it e m/guidin gprincip le s-la b elling -p r om ot i ng - healthy di et . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint Figure 1. Front-of-pack nutrition labels examples A. Traffic light label (UK) B. Nutrient warning label – black octagons (Chile) In English, the labels would read (left to right): [HIGH IN] SUGAR, CALORIES, SATURATED FAT, SODIUM [Ministry of Health]. 29 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint Figure 2. Logic diagram of the impact of front-of-pack labelling on obesity prevalence and cardiovascular disease (CVD) mortality. Abbreviation: BMI: Body Mass Index 30 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 16, 2024. ; https://doi.org/10.1101/2024.10.14.24315283doi: medRxiv preprint