Correspondence to Mr Barthélémy Sarda; [email protected]
WHAT IS ALREADY KNOWN ON THIS TOPIC
Previous modelling studies on reformulation have shown that using hypothetically reformulated products would improve diet quality. However, the use of fictional products has been pointed out as a limitation, given questions regarding the actual feasibility and acceptability of such products.
WHAT THIS STUDY ADDS
Reformulation of all processed and ultra-processed food products towards healthier existing products led to an improvement of diet quality, across the population, including nutritionally at-risk populations. Using nutrient profiling models, such as the nutrient profiling system underpinning the front-of-pack nutrition label Nutri-Score, appeared relevant to guide reformulation towards healthier products.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Reformulation should be promoted, in conjunction with other policies, to improve the dietary status of the population, considering the global impact of such policy.
Background
Poor nutrition is estimated to be responsible for around a quarter of deaths related to non-communicable diseases in Europe,1 through an increased risk of cardiovascular diseases, diabetes and some types of cancer.2–4 Consequently, nutrition has been identified as key actionable lever to improve the health of the population and reduce the burden chronic diseases represents to health systems.1 5–7 To promote healthy diets, a wide range of public policies and programmes are implemented with different strategies: modifying either individual dietary behaviours with policies such as food labelling, direct consumer education or media campaigns; or improving the food environment by regulating marketing, implementing financial incentives/disincentives, increasing the availability and accessibility of healthy foods and by reformulating the existing food offer.8
The WHO indeed considers reformulation an important element in the adoption of healthy and sustainable diets for all as it does not require individuals to make conscious efforts to seek healthier alternatives.9 As a result, governments implement policy actions to promote industry reformulation—mainly nutritional reformulation—with either mandatory enforcement through regulations or stimulate voluntary efforts in the food industry through incentivising corporate pledges or commitments.10 Governmental and international institutions play an important role in steering the reformulation, by setting targets to reach and by monitoring the quality of the food offer. In France, for example, the Ministry of Agriculture and the Ministry of Health work in collaboration with different sectors of the food industry to define nutritional targets (eg, salt content in the breadmaking industry11). Despite the consensus among public authorities, it should be noted that, in the framework of commercial determinants of health,12 nutritional reformulation is criticised as it may be seen as a measure to divert the attention from other health dimensions of foods, such as degree of processing,13 or to avoid regulations, by emphasising on self-regulatory initiatives.14
To study the impact of reformulation policies, past research has used two main methods: modelling studies and a posteriori evaluations. Modelling studies project the gains induced by either future or hypothetical policies. Modelling relies on the assumption that modifying the nutritional composition of foods available (eg, a reduction of X% in a specific nutrient content in some foods) has a direct impact on the dietary intakes of the population, and ultimately its health. Modelling studies thus provide an a priori estimate of the potential effects of a given policy, and support policy-making with attainable impact assessments for informed decisions. Such modelling studies are useful to evaluate the relevance and compare different policy schemes beforehand to select the most effective one.
However, though hypothetical scenarios can provide a range of expectations for a given policy, they may fail to take into account some constraints pertaining to food reformulation, such as unintended substitution effects of nutrients,15 sensory acceptability or technological issues,15 16 due to the need for simplifying assumptions to develop reformulation models. On the other hand, studies evaluating reformulations interventions in retrospect address the aforementioned limitations of modelling studies but they are generally focused on targeted categories (eg, bread) and/or specific nutrients (eg, salt),17 which limits general inferences on the overall diet quality of the population.
An innovative approach would be to combine the two methods—modelling and empirical studies—to provide realistic estimates of potential effects of reformulation by using the existing options on the market as potential targets for reformulation. Such a method would cover the overall diet and multiple nutrient, while still considering products that are technologically feasible and organoleptically accepted by consumers.
This study, thus, models the potential changes in energy and macronutrient intake at the population level induced by industry-wide reformulation of food and beverage. Rather than considering hypothetical reformulated products, we based our projections on existing products in the French food market, capitalising on the existing variability in nutrient composition of the foods on the market. Our purpose was to estimate the magnitude of effect on macronutrient intake that may be attributable to intraproduct nutritional variability (ie, reformulation) as opposed to food group variability (ie, diet structure). To qualify the nutritional quality of products, the updated Food Standard Agency-Nutrient Profiling System score was used (uNS-NPS), which is the currently applicable form of the nutrient profiling system underlying the Nutri-Score front-of-pack labelling scheme that has been adopted in France in 2017 and in several others European countries since 2017.
Methods
Study design
In this paragraph, we provide an overview of the method we used, which is then explained in detail in the following sections.
Figure 1 summarises the different steps in the study design. First, based on detailed 24 hours dietary records collected in the NutriNet-Santé cohort among the French population (see section on Study population and Dietary data), we identified foods consumed in the population that could be reformulated by the food industry and thus change in nutritional composition (ie, processed and ultra-processed foods). Then, we matched each generic food that could be reformulated identified previously with a list of corresponding food products in the French market using the food offer data from the OpenFoodFacts database (see section Food composition data). For example, the generic ‘white sandwich bread’ in the 24 hours dietary records was matched with 567 different white sandwich breads on the market and this list of products was considered as the current food offer for that particular product. Then, the food products on the market were sorted by nutritional quality, using the uNS-NPS (see section on Assessment of the nutritional quality of foods). We then considered three scenarios of varying food offer: (1) the baseline scenario, in which all products could be selected for participants; (2) a scenario with products of more favourable nutritional profile (later referred to as improvement scenario), in which the food offer would only be composed by products of better nutritional quality and only products from this restricted food offer could be selected for participants and (3) a scenario with products of less favourable nutritional profile (later referred to as deterioration scenario), in which the food offer would only be composed by products of poorer nutritional quality and only products from this restricted food offer could be selected for participants (see the ‘Elaboration of the scenarios’ section). Finally, to compute dietary intakes at the individual level, for each food that could be reformulated and consumed by a participant, we randomly allocated an item of the food offer available in the scenario. For foods that could not be reformulated (ie, unprocessed and homemade foods), we used a generic composition in all scenarios.
Enlarge this image.Figure 1. Schematic representation of the study design. uNS-NPS, nutrient profiling system underpinning the front-of-pack nutrition label Nutri-Score.
Study population
Detailed dietary data were retrieved from the NutriNet-Santé cohort. Briefly, NutriNet-Santé is a French ongoing web-based cohort launched in 2009, which aims to study the associations between nutrition and health as well as the determinants of eating behaviours and nutritional status. The functioning of the cohort was previously described in detail elsewhere.18 Eligibility criteria are the following: being aged 15 years or older and having access to the internet. On inclusion and each year/semester thereafter, participants are sent a set of five questionnaires related to sociodemographic and lifestyle characteristics, anthropometric measurements, physical activity, health status and dietary intakes.
Dietary data
On inclusion, and every 6 months thereafter, participants are asked to complete three non-consecutive 24 hours dietary records, in a 2-week period (2 weekdays and 1 weekend day). Participants indicate the foods and beverages consumed in a 24-hour period, as well as the quantity consumed using validated photographs, quantity (g or mL), or by indicating usual serving containers.19 Reported food items are then matched with the NutriNet-Santé food composition database, which contains more than 3000 generic food items. For the present study, we included individuals that had at least three valid 24 hours dietary records, completed within the first 6 months of participation in the cohort. Additionally, under-reporters were identified using the method proposed by Black,20 by using the basal metabolic rate and Goldberg cut-off, and were excluded from the study. Daily food consumption was estimated using a weighted average between weekdays and weekend dietary records.
Food composition data
To obtain a representative food composition database of foods that could be reformulated, the following procedure was conducted:
Using the food database from Nutrinet-Santé which contains 3065 generic food items, a list of foods that could be reformulated was determined through an elimination process. First, we considered that foods classified as unprocessed or minimally processed, according to the NOVA classification,21 or homemade could not be reformulated. The categorisation used to classify foods was realised by three trained dietitians, trained in nutritional epidemiology. To identify homemade products, names of foods and the proportion of reported branded products consumed were used, as participants to the NutriNet-Santé cohort have the possibility to indicate whether they cook an item from unprocessed or minimally processed ingredients or directly buy a ready-made alternative. We also considered that alcoholic beverages could not be reformulated as the uNS-NPS does not cover this category of product. As a result, products that could be reformulated corresponded to processed and ultra-processed foods (NOVA 3 and 4), which were unlikely to have been homemade. At the end of this process, 1517 generic food items from the Nutrinet-Santé database were identified.
Nutritional composition of products from the French food market was based on the crowd-sourced database OpenFoodFacts, extracted on the 4 November 2021. In case of missing macronutrients that are mandatory on the nutritional declaration (ie, energy, carbohydrates, sugars (defined as all monosaccharides and disaccharides present in food, but excluding polyols), total fats, saturated fats, sodium or salt, proteins), products were excluded. In case of missing fibre content which is optional according to food labelling regulation in Europe, we followed recommendations from Ispirova et al for missing values in food composition databases and used multiple imputation.22 For the products that could not be reformulated, the generic composition from the NutriNet-Santé database was used.
Generic food items identified in step 1 were matched with products from the OpenFoodFacts database, by matching product name in the Nutrinet-Santé database with category names and product names in the OpenFoodFacts database. In total, 119 073 products from the OpenFoodFacts database were matched with the initial 1517 generic products from the Nutrinet-Santé database. Additionally, in case of missing fruit, vegetable and pulse proportion in the OpenFoodFacts database, the value used by the corresponding generic food item in the Nutrinet-Santé database was imputed. Details regarding food composition databases are given in online supplemental material 1–5.
Assessment of the nutritional quality of foods
The uNS-NPS was calculated for all foods and products in both databases (ie, the generic food database from the Nutrinet-Sante cohort and the branded OpenFoodFacts database). Briefly, the uNS-NPS is a nutrient profiling system which ranks foods on a discrete scale from −17 (better nutritional quality) to 55 (worse nutritional quality).23 The system allocates points based for the content in energy (in kJ), in sugars (in g), in saturated fat (in g), in salt (In g), in proteins (in g), in fibre (in g) and based on the proportion of fruit, vegetables and pulses (in %). Adaptations have been made to specific food groups given their nutritional specificities and maintain consistency with dietary guidelines. The full method for calculating the uNS-NPS is presented in online supplemental material 6.
Elaboration of the scenarios
To assess the potential impact of widespread change in the nutritional composition of the food offer, different scenarios were constructed to generate alternative food offers, based on the uNS-NPS score computed previously. The first step in the construction of each scenario was to define which products would be available:
Baseline scenario: All matched products were considered available and thus susceptible to being consumed by the study population.
Nutritional improvement scenario: Only matched products in between the 10th and the 20th percentile of uNS-NPS score (ie, more favourable nutrient composition) for each generic food item were considered available. Of note, the lower the uNS-NPS score, the better the nutritional quality. To give an example, for canned sardines, the 10th and 20th score percentiles were respectively −1 and 1. Thus, canned sardines selected for the improvement scenario were those with a uNS-NPS score in between −1 and 1. The process was repeated for all generic food item that had matches in the OpenFoodFacts database.
Nutritional deterioration scenario: Same procedure as the nutritional improvement scenario except only products with a uNS-NPS score in between the 80th and 90th percentiles (ie, products of less favourable nutrient composition) was considered available.
Then, once the food offer was determined in each scenario (baseline, improvement, deterioration), for each individual in our study population, a food product from the available food offer in the scenario was randomly allocated for each food consumed by the individual. If food items from the Nutrinet-Santé database could not be reformulated or were not matched with products in the OpenFoodFacts database, the generic food item composition from the Nutrinet-Santé database was used. As products were randomly allocated in this step, Monte-Carlo iterations (n=300) were performed to generate uncertainty intervals.
Statistical analyses
The impact of each scenario was assessed using energy and macronutrient intakes at the population level. Macronutrient intakes were adjusted on energy intake using residual method as proposed by Willet and Stampfer.24 In this method, energy-adjusted intakes are calculated using the residual of a regression model in which nutrient intake is the dependent variable and total energy intake is the independent variable. To facilitate the interpretation, the average nutrient intake in the population is added to the residuals. The obtained values correspond, therefore, to a hypothetical nutrient intake that is uncorrelated with energy intake and is directly related to overall variation in food choice and composition.
To assess potential differential effects across the population, results were analysed depending on the simplified PNNS-GS2 score (sPNNS-GS2), which reflects adherence to French national dietary recommendations. Briefly, the Programme National Nutrition Santé-Guideline Score 2 evaluates the consumption of different food groups in regard to French dietary recommendations: consumption of healthier food groups (eg, fruits and vegetables, nuts or pulses) impact positively the overall score while consumption of food groups or elements which consumption should be limited (eg, red meat, sugary foods, salt) negatively impact the score. The score and its simplified alternative have been validated in the French context.25 The method of calculation is detailed in the original article on the score development.25 Sex-specific quintiles of sPNNS-GS2 were used to stratify the population.
Differences in energy and nutrient intakes between the baseline situation and modelled scenarios were tested using paired t-tests. To consider multiple testing, Bonferroni correction was applied. Food groups contributing the most to changes in nutrient intakes were described.
Then, nutrient intakes were compared with recommended intakes. Recommended intakes were derived from the reference intakes for an average adult consuming 2000 kcal/day as stated in the European food labelling regulation.26 Individual reference intakes were then calculated using an estimated individual daily energy expenditure. For fibre, as there is no reference intake in the European regulation, the reference value considered was 25 g in line with adequate intake proposed by the European Food Safety Authority (EFSA).27 To achieve recommendations, the following criteria were used: for sugar (90 g/day for an average adult), saturated fat (20 g/day for an average adult) and salt (6 g/day for an average adult), if participants consumed less than the reference intake; for protein (50 g/day for an average adult) and fibre (25 g/day for an average adult), if participants consumed more than the reference intake.
Analyses were conducted on SAS V.9.5 (SAS Institute). The significance level was set at 5% and tests were two sided.
Results
Study population
In total, detailed dietary data from 100 418 participants were extracted from the NutriNet-Santé cohort after exclusion of under-reporters. The flow chart is detailed in online supplemental material 7. Table 1 presents the characteristics of the included study sample. Participants of the study were majority women (78% female vs 22% male) and were aged 43.0 (SD=14.7) on average. Differences in characteristics were observed across quintiles of sPNNS-GS2. Participants in the first quintile (ie, showing less adherence to the dietary guidelines) were on average younger (39.6 years old for the first quintile vs 46.3 years old for the last quintile) and had a higher body mass index (24.6 kg/m² for the first quintile vs 23.1 kg/m² for the last quintile). Finally, the products that could be reformulated considered in the study represented on average 48.4% (SD=13.3%) of the total energy intake with differences across quintiles of sPNNS-GS2 (52.8% for the first quintile and 42.7% for the last quintile).
Table 1Characteristics of the population from the Nutrinet-Santé cohort (N=100 418)
| Total | Quintiles of sPNNS-GS2* | |||||
| Q1 | Q2 | Q3 | Q4 | Q5 | ||
| N | 100 418 | 19 946 | 20 637 | 20 232 | 20 002 | 19 599 |
| Age (years) | 43.0±14.7 | 39.6±14.0 | 41.4±14.4 | 43.2±14.6 | 44.7±14.7 | 46.3±14.8 |
| Sex (male/female) | 22%/78% | 22%/78% | 22%/78% | 22%/78% | 21%/79% | 23%/77% |
| Educational level | ||||||
| <High school degree | 18.0% | 19.1% | 18.0% | 17.7% | 18.0% | 17.4% |
| <2 years after high school | 17.0% | 19.0% | 17.6% | 16.9% | 15.9% | 15.5% |
| 65.0% | 61.9% | 64.4% | 65.4% | 66.1% | 67.1% | |
| Income level | ||||||
| <€1800/month/CU† | 40.6% | 46.3% | 42.4% | 40.0% | 38.7% | 35.8% |
| <€1800–€2700/month/CU† | 23.6% | 22.3% | 23.3% | 24.0% | 23.9% | 24.7% |
| ≥€2700/month/CU† | 23.8% | 20.1% | 22.2% | 24.3% | 25.2% | 27.5% |
| Did not want to answer | 12.0% | 11.3% | 12.1% | 11.7% | 12.2% | 12.0% |
| Area of residence (rural/urban) | 22%/78% | 25%/75% | 23%/77% | 22%/78% | 21%/79% | 20%/80% |
| Body mass index (kg/m²) | 24.1±4.9 | 24.6±5.3 | 24.0±4.6 | 23.8±5.0 | 23.6±4.3 | 23.1±4.2 |
| sPNNS-GS2‡ | 1.3±3.6 | −3.7±1.9 | −0.6±1 | 1.4±1 | 3.3±1 | 6.1±1.5 |
| Proportion of energy coming from products in scope | 48.4%±13.3% | 52.8%±13% | 50.7%±12.9% | 48.8%±12.8% | 46.7%±12.8% | 42.7%±12.9% |
The table is original and provided by the authors.
Reformulations scenarios.
*Q1 corresponds to individuals with the lowest sPNNS-GS2 (lowest adherence to guidelines), Q5 to individuals with the highest sPNNS-GS2 (highest adherence to guidelines).
†CU: Consumption unit as defined by the French National Institute of Statistics and Economic Studies.52
‡sPNNS-GS2: simplified Programme National Nutrition Santé-Guideline Score 2. Sex-specific cut-offs for sPNNS-GS2 quintiles were the following: men: −3.75/−1.33/0.67/2.92; women: −1.17/0.92/2.67/4.67.
Table 2 presents the potential impact at population level of widespread improvement (or deterioration) of the French food offer. Reformulation towards healthier products led to a significant reduction in energy intake (from 1884 kcal/day to 1829 kcal/day, corresponding to −55 kcal/day or −2.9%), sugars (from 90.3 g/day to 85.5 g/day, −4.8 g/day, −5.3%), saturated fat (from 31.7 g/day to 29.3 g/day, corresponding to −2.4 g/day or −7.6%) and salt (from 6.48 g/day to 5.94 g/day, corresponding to −0.54 g/day or −8.3%). Conversely, the intake of fibre increased in the improvement scenario (from 20.8 g/day to 21.8 g/day, corresponding to +1.0 g/day or +4.9%). Regarding proteins, there was no significant difference overall or minor changes at population level. Conversely, the deterioration of the food offer led to effects in the opposite direction (ie, increase in energy, sugar, saturated fat and salt and reduction of fibre intake). Between the scenario with products of better and worse nutritional profile, the difference in daily intake was 10.6% lower for sugars (−9.6 g/day), 16.3% for saturated fats (−5.1 g/day), 16.1% for salt (−1.06 g/day) and 10.3% (2.1 g/day) greater for fibre in the scenario with an offer of better nutritional quality. When investigating differential effects across the sPNNS-GS2 adherence score, improvements were greater for participants who were less adherent to dietary guidelines (ie, had a lower sPNNS-GS2) for energy, saturated fat or salt (online supplemental material 8): energy intake (−74 kcal/day in Q1 vs −40kcal/day in Q5), saturated fat intake (−2.6 g/day in Q1 vs −2.1 g/day in Q5), salt intake (−0.65 g/day in Q1 vs −0.52 g/day in Q5). Finally, the improved food offer would improve compliance with recommended intakes for saturated fat, sugar, salt and fibre (online supplemental material 9).
Table 2Daily energy and nutrient intake after nutritional improvement or deterioration of the food offer
| Total (N=100 418) | |||||
| Nutrient | Baseline | Improvement Mean (95% UI) | Δ I-B* | Deterioration Mean (95% UI) | Δ D-B† |
| Energy (in kcal day) | 1884b (1881–1888) | 1829a (1825–1834) | −2.9% | 1938c (1935–1942) | 2.9% |
| Carbohydrates (in g/day) | 199b (198-199) | 194a (193-194) | −2.8% | 204c (204-205) | 2.4% |
| Sugars (in g/day) | 90.3b (90.1–90.5) | 85.5a (85.3–85.7) | −5.3% | 95.1c (94.9–95.3) | 5.3% |
| Lipids (in g/day) | 78.2b (78.0–78.4) | 74.5a (74.4–74.6) | −4.7% | 81.8c (81.7–81.9) | 4.7% |
| Saturated fats (in g/day) | 31.7b (31.6–31.8) | 29.3a (29.3–29.4) | −7.6% | 34.4c (34.4–34.5) | 8.6% |
| Salt (in g/day) | 6.48b (6.45–6.49) | 5.94a (5.92–5.97) | −8.3% | 6.98c (6.95–7.00) | 7.8% |
| Proteins (in g/day) | 76.5b (76.3–76.6) | 77.2a (77.1–77.3) | 0.9% | 76.5b (76.3–76.7) | −0.1% |
| Fibres (in g/day) | 20.8b (20.7–20.9) | 21.8a (21.8–21.9) | 4.9% | 19.7c (19.6–19.7) | −5.4% |
Superscript letters indicate the presence or absence of significant difference between the scenarios (paired t-tests with Bonferroni correction).
The table is original and provided by the authors.
*Δ I-B: Relative changes between the baseline situation versus the improvement scenario.
†Δ D-B: Relative changes between the baseline situation versus the deterioration scenario.
Food group contributions
When investigating the groups contributing most to the improvement of the nutrient intakes between baseline and the improvement scenario, we observed the following (table 3): (1) for energy: bread/rusks (14.4%), dairy products (14.1%), sugary products (13.8%) and prepared dishes (13.4%); (2) for sugars: sugary products (29.9%), sweet bakery products (13.4%), dairy products (13.4%) and sugar-sweetened beverages (12.4%); (3) for saturated fat: sweet bakery products (19.8%), dairy products (16.8%), prepared dishes (14.5%) and processed meat (12.7%); (4) for salt: bread /rusks (25.8%), prepared dishes (13.7%), vegetable preparations and soups (11.2%) and meat and meat products (10.8%) ; (5) for fibre: bread/rusks (48.6%), sweet bakery products (12.1%), breakfast cereals (7.7%) and fruit and vegetable preparations (7.4%).
Table 3Food groups contributing to the dietary changes in the improvement versus baseline scenario
| Food groups | Energy | Sugars | Saturated fat | Salt | Fibre |
| Appetisers | 2.1% | 2.0% | |||
| Meat and meat products | 9.9% | 12.7% | 10.8% | ||
| Meat | 3.7% | ||||
| Processed meats | 9.9% | 12.7% | 7.1% | ||
| Cheeses | 4.6% | 2.5% | 10.4% | ||
| Dairy products | 14.1% | 13.4% | 16.8% | ||
| Yoghurts and quarks | 3.4% | 6.8% | 3.8% | ||
| Dairy desserts | 10.7% | 6.6% | 13.0% | ||
| Sugary products | 13.8% | 29.9% | 8.3% | 6.4% | |
| Confectionary and sweets | 7.6% | 18.2% | |||
| Chocolate and chocolate products | 6.7% | 6.4% | |||
| Ice creams and sorbets | 6.2% | 5.0% | 8.3% | ||
| Sweet bakery products | 6.8% | 13.4% | 19.8% | 4.7% | 12.1% |
| Biscuits | 3.7% | 10.2% | 14.3% | 2.3% | 7.4% |
| Pastries | 3.1% | 3.2% | 5.5% | 2.4% | 4.7% |
| Potatoes and tubers | 3.6% | ||||
| Bread, rusks | 14.4% | 7.0% | 3.1% | 25.8% | 48.6% |
| Breakfast cereals | 4.5% | 3.4% | 7.7% | ||
| Prepared dishes | 13.4% | 14.5% | 13.7% | 5.5% | |
| Ready-to-eat meals | 7.8% | 6.5% | 8.8% | 5.5% | |
| Pizzas | 5.6% | 8.0% | 4.9% | ||
| Fruit and vegetable preparations | 2.0% | 5.9% | 3.5% | 11.2% | 7.4% |
| Fruits preparations | 5.9% | 2.0% | |||
| Vegetable preparations | 4.2% | 5.4% | |||
| Soups and broths | 2.0% | 3.5% | 7.0% | ||
| Seasonings | 2.9% | 2.0% | 2.5% | 3.7% | |
| Dressings and sauces | 2.9% | 2.0% | 3.7% | ||
| Fats and oils | 2.5% | ||||
| Beverages | 4.4% | 12.4% | |||
| Sweetened beverages (other than fruit juices) | 4.4% | 12.4% |
Only foods groups contributing at least 2% in one nutrient of concern or energy were presented in the table above. Thus, the total for each nutrient is different from 100%. The table is original and provided by the authors.
Discussion
The present study investigated the potential impact of industry-wide food reformulation on energy and nutrient intakes at population level. The modelled scenarios assessed the effects of substituting the existing food offer with an offer of higher nutritional quality (or lower), constituted by already existing healthier products (or less healthy) and without modifying the structure of food choices at the individual level. Assessment of healthiness was based on the uNS-NPS score, which is the current nutrient profiling system used in France for front-of-pack labelling.
Our study used a different approach from the other modelling studies published to date limiting direct comparisons with the current literature. Rather than projecting modification of the composition based on guidelines or policies (ie, a fixed relative reduction of nutrients on all products of one or several categories), our approach was to set as reformulation targets existing products that would be classified as healthier on the food market. This method overcomes several limitations of the modelling studies based on a priori approaches to reformulation: consideration of the interdependence of nutrients when reformulating,15 technological feasibility and to some extent organoleptic acceptability.15 16 Thus, this study provides a realistic estimate of the impact of reformulation by using the current variability of the food offer.
Indeed, the purpose of this work was to assess if the current differences in nutritional composition observed on the food market could lead to significantly different nutrient intakes and to quantify the magnitude of effect. On the other hand, magnitude of effects found in other modelling studies tends to be greater,28 suggesting that more ambitious policies could be effective. Indeed, a recent systematic review found interventions on sodium intake could reduce the daily sodium intake from 0.48 g of salt/day to 4.55 g of salt/day when interventions are applied to all processed foods, with average salt reduction in foods ranging from 10% to 60%.28 Of note, the interventions take place in different countries, thus nutritional context, and the results are rarely adjusted for energy which complicates comparisons with our results.
With regard to nutrient intake, the improvement scenario showed a reduction in saturated fat, sugars and salt intake at the population level, with relative reductions from 5.3% to 8.3% depending on the nutrient. These nutrients are considered by the EFSA of public health importance as they are consumed in excessive amounts in Europe.29 The relevance of reducing such intake through reformulation has been demonstrated in the past to improve health status.17 30–32 As shown in online supplemental material 9, fewer individuals exceeded their daily recommended intake for saturated fat, sugars and salt when the nutritional quality of the food offer was improved. Interestingly, the French Agency for Food, Environmental and Occupational Health and Safety (ANSES) published in 2021 a report on the impact of setting reformulation thresholds, based on nutritional data collected on the French food market,33 and the second Individual and National Survey on Food Consumption (INCA2). The methodology presented similarities with the present study, such as basing the modelled scenario on existing products or excluding homemade products. On the other hand, nutrients were analysed independently from one another, therefore, not considering potential substitution of nutrients, and the classification of food was less detailed, leading to potentially greater differences between products of the same groups. It was found that if products with a content in a specific nutrient above the median of its category was reduced to the median content, it would lead to a reduction in sugar intake by 2.2 g/day, in saturated fat intake by 1.6 g/day, in salt by 0.23 g/day and an increase in fibre intake by 0.6 g/day. These results are of similar magnitude as the one estimated in the present study, though slightly smaller. An explanation could be that in our study products were improved to the level of the 20% products of better nutritional quality whereas in the ANSES study they were improved to the level of the median. As such, the nutritional quality of improved products in our study was overall better, thus the magnitude of effect was greater.
It should be pointed out that among the plurality of possible strategies to improve the nutritional status of the population, reformulation appears to have a smaller effect than modifications of dietary patterns. We indeed found that the increase of fibre intake in the improvement scenario would be of limited impact (1 g/day), when compared with the average 5–10 g needed at the population level to reach the recommended amounts of 25–30 g of fibres/day. Shifting consumption from refined to whole-grain cereal products may prove a more effective approach.34 Nonetheless, reformulation requires little to no efforts from consumers as the diet structure remains unchanged, and thus may be more acceptable, while more radical changes such as consuming different food groups—which may be more effective—necessitate a conscious decision to eat more healthfully. Furthermore, relatively small changes affecting the entire population may greatly impact the population health in long term. A study found that a reduction by only 5% in caloric intake or a similar increase in caloric expenditure may limit weight gain in most people.35 Additionally, a Cochrane systematic review in 2013 found that a modest and prolonged reduced salt intake an important impact on systolic blood pressure in both hypertensive and normotensive individuals, thereby reducing cardiovascular risk in the population.36 As a result, different policies to improve population’s diet quality should not be seen as mutually exclusive and may all contribute to improve the diet quality at population level, by both improving the food environment (reformulation, change of serving size…) and the demand (nutrition campaigns, front-of-pack labels…). The WHO and the Word Cancer Research Fund indeed argue in favour of a combination of interventions at different levels—including reformulation—to improve the population nutritional status, and thus health outcomes.37–39
To put into perspective the food groups contributing to the nutritionally improved diet, groups identified in this study were important contributors to nutrient intakes within diets in the third French INCA (INCA 3),40 even though relative importance among them may vary. For example, for salt intake, main contributors in INCA 3 were breads (22.5% of the daily salt intake), prepared dishes (19.4%), dressings and sauces (12.2%), soups (9.5%), processed meats (7.8%) and cheeses (5.5%)40; and reformulation of foods in these groups significantly contributed to reduced salt intake in our study (ie, bread, prepared dishes, vegetable preparations and soups, meat and meat products, and cheeses). This phenomenon is unsurprising as reformulation of foods greatly contributing to a specific nutrient is expected to be the most impactful. This is why many governments tackle the reduction of salt intake through the reduction of salt in prepared foods high in salt (eg, processed meats, cheeses or soups) or heavily consumed foods (eg, bread), through mandatory measures,9 or voluntary commitments.41 42 In turn, some foods in our study appeared to have played a minimal role in the improvement of diet after reformulation even though they were an important contributor to some nutrients at baseline. The rather low importance of such foods may be explained by the limited variability across products of the nutrient in question. For example, cheeses are the first contributor to saturated fat intake in the French population,40 but their high content in saturated fat is rather stable across products of the same type, due to product specifications and legal definitions. The lack of effect observed in this case may be explained by the choice in the study not to alter dietary choices and by the fact that we used a detailed classification of products. In such cases, only substitution for different products (eg, from a hard-cheese to a fresh cheese) could decrease the saturated fat intake.
To obtain results as estimated in this study, food manufacturers would be required to engage in active reformulation policies. In order to steer the food industry, public authorities rely on the formulation of reformulation targets, usually nutrient specific. The uNS-NPS score could be a tool to complement and encourage reformulation policies. On the one hand, it could be an indicator to monitor the overall nutritional quality of the offer. In the analysis, we observed that the median uNS-NPS was on average lower by −2.6 points (SD=2.3) when comparing products from the improved and baseline offer, indicating a food offer of better nutritional value leading to more nutritionally adequate intakes. Interestingly, protein intake, which is not considered of concern in Europe,29 was only marginally affected in all scenarios, showing that the score rather discriminated products based on nutrients that should either be limited (ie, saturated fat, sugars and salt) or increased (ie, fibre). On the other hand, the use of the uNS-NPS in the form of a front-of-pack label would provide food manufacturers the opportunity to signal to consumers their improvement efforts. Indeed, the Nutri-Score -a 5-colour label based on the uNS-NPS- has been shown to be able to discriminate foods, including foods of varying nutritional quality in the same category.23 A study in the Netherlands also found that a reduction by 1 FSA-NPSm—the previous version of the nutrient profiling system underlying Nutri-Score—through limiting the saturated fat, sugars or salt content, would lead to a more favourable Nutri-Score in a large number of food groups.43 Thus, in an environment where the alignment with nutrient targets remains optional, food manufacturers would be encouraged to steer towards them through the opportunity to gain an advantage over competitors by displaying a better Nutri-Score. However, it is worth noting that nutritional reformulation does not systematically translate into a more favourable Nutri-Score, highlighting the need for other complementary policies aiming at promoting the improvement of the food offer. In turn, few studies looked at the actual impact of the implementation of the Nutri-Score. A study in Belgium found over a 1-year period the nutritional quality of breakfast cereals improved—with significantly fewer sugars (−5%) and sodium (–20%).44 Nonetheless, as pointed out by Vandevijvere and Vanderlee,45 the actual impact of front-of-pack labelling schemes is insufficiently investigated, even though the effectiveness of such policies is greater when mandatory to use—which is not the case for Nutri-Score—aligned with other policies and monitored.
A strength of the study was setting reformulation targets on existing products rather than theoretical products aimed to address recurring limitations of modelling studies. Then, dietary data were collected thanks to the NutriNet-Santé cohort, for more than 100 000 individuals, with more than 3000 unique food and beverage items. This detail in the dietary records allowed us to propose reformulation for the same type of product as the one consumed. Additionally, dietary data were then matched with a large branded food composition database, with 119 073 products from the French market, providing important data regarding the variability of composition on the market. Finally, homemade foods were identified and thus excluded from the potential products to reformulate. Indeed, improvement of the nutritional quality of these foods depends on cooking practices at home rather than modification of the food offer. However, it could be argued that even some foods prepared at home may rely on processed products in the scope for reformulation (eg, use of a store-bought pie crust). As we did not include these products, the effect observed would be underestimated.
Nonetheless some limitations should be acknowledged. First, participation in the NutriNet-Santé cohort is voluntary, leading to a sample with more women, with a higher educational level and a higher socioeconomic situation,46 47 and could have led to healthier dietary patterns in comparison with the general population. It could be thus expected that the effect should be at least equal, if not greater, in the general population. Then, as the nutrient intake was based on data collected on nutritional declaration, we were unable to distinguish intrinsic sugar (fructose, lactose) and added sugar consumption. Given that reformulation targeted processed and ultra-processed products, we estimated that sugars removed through reformulation would be mostly added sugars. Indeed, when looking at the groups contributing to a lower sugar intake, only dairy products and fruit preparations were susceptible to contain intrinsic sugars and, in these categories, the modification of sweetened products (sweetened yoghurts, chocolate mousses and compotes) were important contributors. Then, impact on micronutrient intake was not assessed, even though some of them are considered of public health importance in either the entire population (eg, potassium) or in specific subpopulations (eg, calcium, iron), due to the lack of information available on nutrition labels. As only intraproduct substitution was considered and micronutrient intake is rather influenced by the consumption of specific food groups, it was hypothesised that it would be marginally affected, as observed for calcium in another modelling study.48 Additionally, higher protein intake has been linked with higher iron intake,49 50 and proteins are considered in the uNS-NPS algorithm as a proxy for micronutrients,23 particularly calcium and iron. Given that protein intake was stable in all scenarios, this element supports our hypothesis that micronutrient intake would be marginally affected. From a feasibility point of view, modification of the food offer as modelled in this study are unlikely as they require all food manufacturers to actively reformulate their products when reformulation is still optional for most nutrients and for most products. Nonetheless, our purpose was rather to estimate the overall potential for improvement by substituting for similar products of better nutritional quality. Finally, actual modification of consumer behaviour was not taken into account in the modelling and substitution effects for cost or individual preference reasons were not considered.
In conclusion, this study showed that widespread reformulation towards healthier products within the current boundaries of existing products using nutrient profile models could lead to substantial improvements in intakes of all nutrients of concern. As such, the use of the uNS-NPS score was found relevant in the effort to identify healthier products, in the context of guiding food manufacturers in the nutritional improvement of their products. Reformulation, by affecting the food offer, represents a great opportunity to improve diets of the entire population,51 but several nutrients were insufficiently modified when compared with dietary recommendations. Further improvement could be thus achieved with complementary policies, such as better information to consumers, to help them select healthier products, or change consumer practices during food purchase and preparation.
We thank Nathalie Druesne-Pecollo, PhD (operational coordinator); Younes Esseddik, Selim Aloui, Thi Hong Van Duong, Régis Gatibelza, Jagatjit Mohinder, Rizvane Mougamadou, and Aladi Timera (computer scientists); Fabien Szabo de Edelenyi, PhD, Julien Allegre, Nathalie Arnault, Laurent Bourhis, Nicolas Dechamp (data-manager/statisticians); Cédric Agaësse, Alexandre De Sa, Rebecca Lutchia (dietitians); Paola Yvroud, MD (health event validator), and Maria Gomes (Nutrinaute support) for their technical contribution to the NutriNet-Santé study. We thank all the volunteers of the NutriNet-Santé cohort.
Data availability statement
Data described in the manuscript, code book, and analytic code will be made available upon request pending application and approval. Researchers from public institutions can submit a collaboration request including information on the institution and a brief description of the project to [email protected] requests will be reviewed by the steering committee of the NutriNet-Santé study. A financial contribution may be requested. If the collaboration is accepted, a data access agreement will be necessary and appropriate authorizations from the competent administrative authorities may be needed. In accordance with existing regulations, no personal data will be accessible.
Ethics statements
Patient consent for publication
Consent obtained directly from patient(s).
Ethics approval
This study involves human participants and the NutriNet-Santé study is registered on ClinicalTrials.gov as NCT03335644 and is conducted according to the Declaration of Helsinki guidelines and is approved by the Institutional Review Board of the French Institute for Health and Medical Research (IRB Inserm) and the 'Commission Nationale de l’Informatique et des Libertés' (CNIL no 908450/no 909216). Participants gave informed consent to participate in the study before taking part.
Handling editor John Lee
Twitter @BernardSrour
Contributors BSarda, EK-G, PG, SH, MT and CJ: designed the study and collected the data; BSarda: performed statistical analyses and wrote the manuscript; CJ: supervised data analysis and paper writing; and BSarda, EK-G, BSrour, MD-T, MF, LKF, PG, SH, MT and CJ: contributed to the data interpretation, revised each draft for important intellectual content, and read and approved the final manuscript. CJ is the guarantor of the data, accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
1 Afshin A, Sur PJ, Fay KA. Health effects of dietary risks in 195 countries, 1990–2017: a systematic analysis for the global burden of disease study 2017. Lancet 2019; 393: 1958–72: S0140-6736(19)30041-8. doi:10.1016/S0140-6736(19)30041-8
2 World cancer research fund. Ernährung 2007; 1: 464–9. doi:10.1007/s12082-007-0105-4
3 Steyn NP, Mann J, Bennett PH, et al. Diet, nutrition and the prevention of type 2 diabetes. Public Health Nutr 2004; 7: 147–65. doi:10.1079/phn2003586
4 Mozaffarian D, Clarke R. Quantitative effects on cardiovascular risk factors and coronary heart disease risk of replacing partially hydrogenated vegetable oils with other fats and oils. Eur J Clin Nutr 2009; 63 Suppl 2: S22–33. doi:10.1038/sj.ejcn.1602976
5 Aburto NJ, Ziolkovska A, Hooper L, et al. Effect of lower sodium intake on health: systematic review and meta-analyses. BMJ 2013; 346: f1326. doi:10.1136/bmj.f1326
6 Hooper L, Martin N, Abdelhamid A, et al. Reduction in saturated fat intake for cardiovascular disease. Cochrane Database Syst Rev 2015: CD011737. doi:10.1002/14651858.CD011737
7 Reynolds A, Mann J, Cummings J, et al. Carbohydrate quality and human health: a series of systematic reviews and meta-analyses. The Lancet 2019; 393: 434–45. doi:10.1016/S0140-6736(18)31809-9
8 Yi SS, Lee M, Russo R, et al. “Dietary policies and programs: moving beyond Efficacy and into ‘“real-world”’ settings”. Health Equity 2021; 5: 194–202. doi:10.1089/heq.2020.0050
9 World Health Organization. Reformulation of food and beverage products for healthier diets: policy brief. Geneva, Available: https://www.who.int/publications/i/item/9789240039919
10 Hyseni L, Atkinson M, Bromley H, et al. The effects of policy actions to improve population dietary patterns and prevent diet-related non-communicable diseases: Scoping review. Eur J Clin Nutr 2017; 71: 694–711. doi:10.1038/ejcn.2016.234
11 Ministère de l’Agriculture et de la Souveraineté alimentaire [Internet]. Réduction de plus de 20% des teneurs en sel dans le pain: grâce à un engagement collectif de la filière, Available: https://agriculture.gouv.fr/reduction-de-plus-de-20-des-teneurs-en-sel-dans-le-pain-grace-un-engagement-collectif-de-la-filiere
12 de Lacy-Vawdon C, Livingstone C. Defining the commercial determinants of health: a systematic review. BMC Public Health 2020; 20: 1022. doi:10.1186/s12889-020-09126-1
13 Scrinis G, Monteiro CA. Ultra-processed foods and the limits of product reformulation. Public Health Nutr 2018; 21: 247–52. doi:10.1017/S1368980017001392
14 Madureira Lima J, Galea S. Corporate practices and health: a framework and mechanisms. Global Health 2018; 14. doi:10.1186/s12992-018-0336-y
15 Buttriss JL. Food Reformulation: the challenges to the food industry. Proc Nutr Soc 2013; 72: 61–9. doi:10.1017/S0029665112002868
16 Erickson S, Carr J. The technological challenges of reducing the sugar content of foods. Nutrition Bulletin 2020; 45: 309–14. doi:10.1111/nbu.12454
17 Gressier M, Swinburn B, Frost G, et al. What is the impact of food reformulation on individuals’ behaviour, nutrient intakes and health status? A systematic review of empirical evidence. Obes Rev 2021; 22: e13139. doi:10.1111/obr.13139
18 Hercberg S, Castetbon K, Czernichow S, et al. The Nutrinet-Santé study: a web-based prospective study on the relationship between nutrition and health and determinants of dietary patterns and nutritional status. BMC Public Health 2010; 10: 242. doi:10.1186/1471-2458-10-242
19 Le Moullec N, Deheeger M, Preziosi P, et al. Validation Du Manuel-photos Utilisé pour L’Enquête Alimentaire de L’Étude SU.VI.MAX. Cah Nutr Diet 1996; 31: 158–64.
20 Black AE. Critical evaluation of energy intake using the Goldberg cut-off for energy intake: basal metabolic rate. A practical guide to its calculation, use and limitations. Int J Obes Relat Metab Disord 2000; 24: 1119–30. doi:10.1038/sj.ijo.0801376
21 Monteiro CA, Cannon G, Moubarac JC, et al. The UN decade of nutrition, the NOVA food classification and the trouble with ultra-processing. Public Health Nutr 2018; 21: 5–17. doi:10.1017/S1368980017000234
22 Ispirova G, Eftimov T, Seljak BK. Evaluating missing value imputation methods for food composition databases. Food Chem Toxicol 2020; 141: 111368. doi:10.1016/j.fct.2020.111368
23 Scientific Committee of the Nutri-Score. Update report from the scientific committee of the Nutri-score 2022 [Internet]. 2022. Available: https://www.santepubliquefrance.fr/determinants-de-sante/nutrition-et-activite-physique/articles/nutri-score
24 Willett W, Stampfer MJ. Total energy intake: implications for Epidemiological analyses. Am J Epidemiol 1986; 124: 17–27. doi:10.1093/oxfordjournals.aje.a114366
25 Chaltiel D, Adjibade M, Deschamps V, et al. Programme national nutrition Santé–guidelines score 2 (PNNS-Gs2): development and validation of a diet quality score reflecting the 2017 French dietary guidelines. Br J Nutr 2019; 122: 331–42. doi:10.1017/S0007114519001181
26 European Union. Regulation (EU) no 1169/2011 of the European Parliament and of the Council of 25 October 2011 on the provision of food information to consumers [Internet]. OJ L, 32011R1169. 2011. Available: http://data.europa.eu/eli/reg/2011/1169/oj/eng
27 EFSA [Internet]. Dietary reference values | DRV finder. 2023. Available: https://multimedia.efsa.europa.eu/drvs/index.htm
28 Federici C, Detzel P, Petracca F, et al. The impact of food Reformulation on nutrient intakes and health, a systematic review of Modelling studies. BMC Nutr 2019; 5: 2. doi:10.1186/s40795-018-0263-6
29 Turck D, Bohn T, Castenmiller J, et al. Scientific advice related to nutrient profiling for the development of harmonised mandatory front-of-pack nutrition labelling and the setting of nutrient profiles for restricting nutrition and health claims on foods. EFS2 2022; 20. doi:10.2903/j.efsa.2022.7259
30 Wyness LA, Butriss JL, Stanner SA. Reducing the population’s sodium intake: the UK food standards agency’s salt reduction programme. Public Health Nutr 2012; 15: 254–61. doi:10.1017/S1368980011000966
31 Leroy P, Réquillart V, Soler LG, et al. An assessment of the potential health impacts of food Reformulation. Eur J Clin Nutr 2016; 70: 694–9. doi:10.1038/ejcn.2015.201
32 Grieger JA, Johnson BJ, Wycherley TP, et al. Evaluation of simulation models that estimate the effect of dietary strategies on nutritional intake: A systematic review. J Nutr 2017; 147: 908–31. doi:10.3945/jn.116.245027
33 ANSES. Simulation de Seuils de Reformulation par famille D’Aliments Transformés et impact sur LES Apports en sucres, Acides Gras Saturés, SEL et fibres de la population Française [Internet]. 2021. Available: https://www.anses.fr/fr/system/files/UOA2019SA0122Ra.pdf
34 Suhr J, Vuholm S, Iversen KN, et al. Wholegrain rye, but not wholegrain wheat, LOWERS body weight and fat mass compared with refined wheat: a 6-week randomized study. Eur J Clin Nutr 2017; 71: 959–67. doi:10.1038/ejcn.2017.12
35 Hill JO, Wyatt HR, Peters JC. The importance of energy balance. Eur Endocrinol 2013; 9: 111–5. doi:10.17925/EE.2013.09.02.111
36 He FJ, Li J, Macgregor GA. Effect of longer term modest salt reduction on blood pressure: Cochrane systematic review and meta-analysis of randomised trials. BMJ 2013; 346: bmj.f1325. doi:10.1136/bmj.f1325
37 World Health Organization. SHAKE the salt habit [Internet]. 2017. Available: https://www.who.int/publications-detail-redirect/WHO-NMH-PND-16.4
38 World Cancer Research Fund International. WCRF International [ NOURISHING framework ]. 2024. Available: https://www.wcrf.org/policy/policy-databases/nourishing-framework/
39 World Health Organization. Report of the commission on ending childhood obesity [Internet]. Geneva, 2016. Available: https://iris.who.int/handle/10665/204176
40 ANSES. Etude Individuelle Nationale des Consommations Alimentaires 3 (INCA 3)- Avis de L’ANSES- rapport D’Expertise collective [ Internet ]. 2017. Available: https://www.anses.fr/fr/content/inca-3-evolution-des-habitudes-et-modes-de-consommation-de-nouveaux-enjeux-en-mati%C3%A8re-de
41 French Ministry of Agriculture. France’s National food programme - the regions in action - 2019-2023 [ Internet ]. 2019. Available: https://agriculture.gouv.fr/programme-national-pour-lalimentation-2019-2023-territoires-en-action
42 Public Health England. Salt reduction targets for 2024 [Internet]. 2020. Available: https://assets.publishing.service.gov.uk/media/5f5618c8d3bf7f4d75de6ff1/2024_salt_reduction_targets_070920-FINAL-1.pdf
43 Ter Borg S, Steenbergen E, Milder IEJ, et al. Evaluation of nutri-score in relation to dietary guidelines and food reformulation in the Netherlands. Nutrients 2021; 13: 4536. doi:10.3390/nu13124536
44 Vermote M, Bonnewyn S, Matthys C, et al. Nutritional content, labelling and marketing of breakfast cereals on the Belgian market and their Reformulation in anticipation of the implementation of the nutri-score front-of-pack labelling system. Nutrients 2020; 12: 884. doi:10.3390/nu12040884
45 Vandevijvere S, Vanderlee L. Labelling, and taxation policies on the nutritional quality of the food supply. Curr Nutr Rep 2019; 8: 240–9. doi:10.1007/s13668-019-00289-x
46 Andreeva VA, Salanave B, Castetbon K, et al. Comparison of the sociodemographic characteristics of the large nutrinet-santé E-cohort with French census data: the issue of volunteer bias revisited. J Epidemiol Community Health 2015; 69: 893–8. doi:10.1136/jech-2014-205263
47 Andreeva VA, Deschamps V, Salanave B, et al. Comparison of dietary intakes between a large online cohort study (etude Nutrinet-Santé) and a nationally representative cross-sectional study (etude Nationale nutrition Santé) in France: addressing the issue of Generalizability in E-epidemiology. Am J Epidemiol 2016; 184: 660–9. doi:10.1093/aje/kww016
48 Gressier M, Privet L, Mathias KC, et al. Modeled dietary impact of industry-wide food and beverage reformulations in the United States and France. Am J Clin Nutr 2017; 106: 225–32. doi:10.3945/ajcn.117.153361
49 Galan P, Yoon H-C, Preziosi P, et al. Determining factors in the iron status of adult women in the SU.VI.MAX study. Eur J Clin Nutr 1998; 52: 383–8. doi:10.1038/sj.ejcn.1600561
50 Pynaert I, Delanghe J, Temmerman M, et al. Iron intake in relation to diet and iron status of young adult women. Ann Nutr Metab 2007; 51: 172–81. doi:10.1159/000103278
51 Griffith R, O’Connell M, Smith K. The importance of product reformulation versus consumer choice in improving diet quality. Economica 2017; 84: 34–53. doi:10.1111/ecca.12192
52 INSEE. Consumption unit: definition [Internet]. 2021. Available: https://www.insee.fr/en/metadonnees/definition/c1802
© 2024 Author(s) (or their employer(s)) 2024. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ. http://creativecommons.org/licenses/by-nc/4.0/ This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/ . Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Kesse-Guyot, Emmanuelle 1 ; Srour, Bernard 1
; Deschasaux-Tanguy, Mélanie 1 ; Fialon, Morgane 1 ; Fezeu, Léopold K 1 ; Galan, Pilar 1 ; Hercberg, Serge 2 ; Touvier, Mathilde 1 ; Julia, Chantal 2 1 Université Sorbonne Paris Nord and Université Paris Cité, INSERM, INRAE, CNAM, Nutritional Epidemiology Research Team (EREN), Center of Research in Epidemiology and StatisticS (CRESS), 74 rue Marcel Cachin, Bobigny, France
2 Université Sorbonne Paris Nord and Université Paris Cité, INSERM, INRAE, CNAM, Nutritional Epidemiology Research Team (EREN), Center of Research in Epidemiology and StatisticS (CRESS), 74 rue Marcel Cachin, Bobigny, France; Public health Department, Hôpital Avicenne, Assistance Publique-Hôpitaux de Paris (AP-HP), Bobigny, France