The adaptation, validation, and application of a methodology for estimating the added sugar content of packaged food products when total and added sugar labels are not mandatory

Efficacy of Brazilian and Mexican front-of-package nutrition labeling systems on simulated online purchase of non-alcoholic beverages by adolescents: A randomized controlled study

Front-of-package nutrition labeling (FoPNL) has been gaining prominence as a strategy to reduce the consumption of sugar-sweetened beverages as it helps consumers to clearly identify products with excess sugars. This study aimed to evaluate the efficacy of the Brazilian and Mexican FoPNL systems on simulated purchases of non-alcoholic beverages by adolescents in a fictitious online store. A total of 437 Brazilian adolescents (15-18 years old) participated in the study and were randomized into three experimental groups: control (online store with products displayed without FoPNL), Mexican FoPNL system (products with "excess in" octagonal black warning labels based on the Mexican nutritional profile), and Brazilian FoPNL system (products with "high in" magnifying glass icons following the Brazilian nutritional profile). FoPNL systems have been applied in accordance with the legislation of each country. Participants selected a non-alcoholic beverage to be purchased from 30 options in the experimental online store. After completing the simulated purchase, they responded to questionnaires regarding their perceptions of healthfulness, harmfulness, and excess nutrient content of the beverages, as well as their socioeconomic status. Compared to the control, the Mexican FoPNL system significantly reduced the prevalence of adolescents choosing beverages in the highest tertile of free sugar density (PR: 0.74, 95 %CI: 0.58-0.94), added sugar density (PR: 0.76, 95 %CI: 0.57-0.99), and energy density (PR: 0.75, 95 %CI: 0.58-0.96). Participants in the Mexican FoPNL system group selected beverages containing 11.5 %, 10.3 %, and 7.7 % less free sugars, added sugars, and energy density, respectively, compared to the control group. Regarding total quantities, the Mexican FoPNL system led to the choice of beverages with 5.3 %, 5.0 % and 3.7 % less free sugars, added sugars, and energy, respectively, than those in the control group. The Brazilian FoPNL system resulted in a 2.3 % reduction in the calories purchased compared to the control group but was ineffective in changing the amount of free and added sugars in the simulated purchases. Additionally, the Mexican FoPNL system increased the odds and the prevalence of participants perceiving sweetened beverages as harmful to health by 82 % and 30 %, respectively, compared to the control. In conclusion, the Brazilian FoPNL system was ineffective in reducing the simulated purchase quantities and densities of free sugars, added sugars, and energy. In contrast, the Mexican FoPNL system was effective, outperforming the Brazilian system by reducing the prevalence of participants choosing beverages in the highest tertiles of free sugar density, added sugar density, and energy density, and by increasing the odds and prevalence of adolescents recognizing sweetened beverages as harmful to health.

Effectiveness of an Intervention to Prevent Ultra-Processed Foods and Added Sugar in the First Year of Life: A Multicentre Randomised Controlled Trial in Brazil

BACKGROUND

The early consumption of ultra-processed foods (UPFs) and added sugars (AS) has been linked to adverse outcomes in infancy. The objective of this study was to determine the effectiveness of a dietary counselling strategy to prevent the consumption of UPFs and AS in the first year of life.

METHODOLOGY

A multicentre randomised controlled trial was conducted with 516 mother-child pairs in three state capitals of Brazil. Mothers were randomly assigned to the control group (CG) or intervention group (IG) after childbirth. The IG received orientation based on UNICEF dietary guidelines and five monthly telephone calls to reinforce the intervention. Dietary intake was measured using food introduction questionnaires and 24-h recalls during home visits at 6 and 12 months. Between-group differences were analysed by generalised estimating equations and presented as mean difference (95% CI).

RESULTS

Children in the IG had lower UPF intake at 6 and 12 months of age (-20.69 g/day; 95% CI: -37.87 to -3.50; p = 0.018 and -32.51 g/day; 95% CI: -61.03 to -3.99; p = 0.025) and lower AS intake at 12 months of age (-4.92 g/day; 95% CI: -9.43 to -0.41; p = 0.033). The intervention also had a positive impact on the period of exclusive breastfeeding, reducing the offer of infant formula, cow's milk, and toddler milk in the first year of life.

PRINCIPAL CONCLUSIONS

The dietary counselling strategy was effective at preventing the early consumption of UPFs and AS in the first year of life. Future research should focus on social and cultural barriers to improve adherence to infant feeding interventions.

Nutrition classification schemes for plant-based meat analogues: Drivers to assess nutritional quality and identity profile

Changes in dietary patterns promoted by the emergence of alternative food systems are becoming increasingly common. The decrease in the consumption of animal-derived products promoted exponential growth in plant-based product demand and, consequently, the availability of several meat analogues for this consumer market. Plant-based meat analogues (PBMAs) were developed to mimic the physical and sensory characteristics of meats and their derivatives. Therefore, the composition of these products has been studied in some countries as an attempt to evaluate their nutritional quality in comparison with that of traditional meat products. The main aim of this study was to employ different Nutrition Classification Schemes (NCSs) to assess the nutritional quality of plant-based meat and to discuss the application of one or more NCSs in defining the identity and quality profile of these foods. Five NCSs were used: three nutrient-based (Nutri-Score; Nutrient Profiling Model (NPM) from Brazil; NPM from PAHO); one food-based (NOVA classification); and one hybrid (Plant-Based Nutrient Profile Model). The nutritional composition and ingredients were collected from labels of 349 PBMAs; 117 were classified as burgers, and 182 products employed soy as the main protein ingredient. The use of different NCSs is strategic for PBMAs' nutritional quality evaluation, and the Nutri-Score was able to show the effectiveness of differentiating products as having poor nutritional quality. In this way, the employment of NPM from Brazil is recommended as a driver for PBMAs choices, especially due to the excellent agreement between the Nutri-Score and NPM from Brazil for burgers.

Food additives and PAHO's nutrient profile model as contributors' elements to the identification of ultra-processed food products

The NOVA classification system categorizes foods according to the extent and purpose of industrial processing. Ultra-processed food products (UPF) are frequently composed of excessive amounts of sugars, salt, oils, and fats, and cosmetic additives designed to make them palatable and/or appealing. We aimed to describe the presence of critical nutrients in excess and cosmetic additives in packaged foods and beverages and to evaluate the proportion of UPF that can be correctly identified through the presence of critical nutrients in excess or the presence of cosmetic additives in food products. A total of 9851 items available in Brazilian supermarkets containing lists of ingredients and nutrition facts panels were analyzed. Cosmetic additives and critical nutrients in excess, according to Pan American Health Organization (PAHO)'s nutrient profile model, were assessed. All food items were categorized into the four NOVA classification groups. Relative frequencies of items with at least one critical nutrient in excess and one type of cosmetic additive were estimated. For UPF, 82.1% had some cosmetic additive, and 98.8% had some cosmetic additive or a nutrient in excess. This combined criterion allowed the identification of 100.0% of sweet cookies, salted biscuits, margarine, cakes and sweet pies, chocolate, dairy beverages, and ice cream. Combining the presence of cosmetic additives and the PAHO's nutrient profile model contributes to the identification of UPF.

Enabling Informed Decision Making in the Absence of Detailed Nutrition Labels: A Model to Estimate the Added Sugar Content of Foods

Obesity and diabetes have emerged as an increasing threat to public health, and the consumption of added sugar can contribute to their development. Though nutritional content information can positively influence consumption behavior, added sugar is not currently required to be disclosed in all countries. However, a growing proportion of the world's population has access to mobile devices, which allow for the development of digital solutions to support health-related decisions and behaviors. To test whether advances in computational science can be leveraged to develop an accurate and scalable model to estimate the added sugar content of foods based on their nutrient profile, we collected comprehensive nutritional information, including information on added sugar content, for 69,769 foods. Eighty percent of this data was used to train a gradient boosted tree model to estimate added sugar content, while 20% of it was held out to assess the predictive accuracy of the model. The performance of the resulting model showed 93.25% explained variance per default portion size (84.32% per 100 kcal). The mean absolute error of the estimate was 0.84 g per default portion size (0.81 g per 100 kcal). This model can therefore be used to deliver accurate estimates of added sugar through digital devices in countries where the information is not disclosed on packaged foods, thus enabling consumers to be aware of the added sugar content of a wide variety of foods.

Comparing Latin American nutrient profile models using data from packaged foods with child-directed marketing within the Brazilian food supply

OBJECTIVES

This study aimed to examine and compare the extent to which different nutrient profile models (NPMs) from Latin America (LA) identify packaged foods and beverages with child-directed marketing sold in Brazil as being high in nutrients associated to the risk of non-communicable diseases (NCDs).

MATERIALS AND METHODS

In this cross-sectional study, we evaluated 3,464 foods found in the five largest Brazilian supermarkets. Child-directed marketing was coded using the International Network for Food and Obesity/NCDs Research, Monitoring and Action Support (INFORMAS) protocol. Differences in medians of sugar, saturated fats, and sodium per 100 kcal in foods, with the presence and absence of child-directed marketing, were tested using the Mann-Whitney test. We compared six NPMs in LA and examined to what extent they targeted these products using prevalence ratios. Analyses were performed overall and by the degree of food processing according to the Nova food classification.

RESULTS

We found 1,054 packages with child-directed marketing. Among these, candies, cakes and pies, sauces and creams, and sugar-sweetened beverages were significantly higher in sugar, saturated fat, and sodium per 100 kcal than products that are not targeted at children (p < 0.05). Compared with PAHO and the Mexico models, the Brazilian NPMs would allow three times more ultra-processed foods to omit warnings for sodium (p < 0.05). The Uruguayan NPM also flagged fewer ultra-processed foods high in sodium (p < 0.05). The Brazilian model also allows four times more sugar-sweetened beverages and six times more dairy drinks to omit warnings for sugar than the Mexico and PAHO models. In comparison to all other NPMs, the Brazilian model showed the worst performance in identifying baked goods as high in sodium. Chile, Uruguay, and Peru models would also target significantly less sugar-sweetened beverages and high in at least one critical nutrient than PAHO and Mexico models.

CONCLUSION

Compared with other NPMs in LA, the NPM criteria adopted in Brazil are more permissive and less likely to inform consumers of the poor nutritional quality of ultra-processed foods and beverages with child-directed marketing.

From the most to the least flexible nutritional profile: Classification of foods marketed in Brazil according to the Brazilian and Mexican models

Nutrient profiling is the science of classifying or ranking foods according to their nutritional composition, for reasons related to disease prevention and health promotion. To be effective, policies such as front-of-pack nutrition labeling (FoPNL) must have an adequate nutritional profile model, since it will determine which products will be eligible to receive a FoPNL. This study aimed to determine the percentage of packaged food and drink products available in Brazil that would be subject to FoPNL under two different legislations: Brazilian and Mexican. This is a cross-sectional study in which we collected information on food products (photos of the ingredients list, the front label, the barcode, and the nutrition facts table) from one of the largest stores of a supermarket chain in the city of Belo Horizonte-MG, Brazil, from March to May 2021 (~6 months after the publication of the Brazilian legislation about FoPNL and a year and a half before the legislation came into force). The products were classified in relation to the BNPM (added sugars, saturated fats, and sodium) and the MNPM (energy, free sugars, saturated fats, trans fats, sodium, non-sugar sweeteners, and caffeine). A total of 3384 products were collected and, after applying the exclusion criteria, 3,335 products were evaluated. Of these, 2,901 would be eligible to receive FoPNL in Brazil and 2,914 would be eligible to receive FoPNL in Mexico. According to the BNPM, 56.7% (95% CI 54.9; 58.5%) of the products were “high in” critical nutrients, 27.1% (95% CI 25.5; 28.7%) of the products in added sugars, 26.7% (95% CI 25.2; 28.4%) of the products in saturated fats, and 21.4% (95% CI 19.9; 22.9%) of the products in sodium. As for the MNPM, 96.8% (95% CI 96.1; 97.4%) of them were “high in” up to five critical nutrients and up to two warning rectangles (caffeine and non-sugar sweeteners), 45.8% (95% CI 44.0; 47.6%) of them in free sugars, 43.7% (95% CI 41.9; 45.5%) of them in saturated fats, and 47.9% (95% CI 46.1; 49.7%) of them in sodium. We concluded that the eligibility to receive FoPNL by BNPM and MNPM was relatively similar between products; however, almost all products would have at least one FoPNL and/or warning rectangles according to Mexican legislation, and nearly half of them would have at least one FoPNL, considering BNPM. The MNPM is much more restrictive than the BNPM. The Nutrient Profile Model (NPM) that regulates FoPNL, and other health policies, must be carefully defined to ensure that foods are properly classified according to their healthiness.

Comparison of two front-of-pack nutrition labels for Brazilian consumers using a smartphone app in a real-world grocery store: A pilot randomized controlled study

One of the suggestions for improving the understanding of food labels is implementing front-of-pack nutrition labeling (FoPNL), where nutritional information is objectively made available to consumers. Scientific data on the best FoPNL model to be adopted for the Brazilian population is still emerging, especially in real-world purchase situations. This study aims to evaluate/compare the proposed Brazilian and Mexican FoPNL systems, on different outcome measures, using an application, in dairy foods available in a supermarket aisle. This pilot randomized controlled trial in a real-world purchase situation was conducted in June/July 2021. A total of 230 participants were randomly allocated to one of the three study arms (Mexican and Brazilian FoPNL systems or control—nutritional information table and ingredients list). Using a smartphone, the participants scanned a product barcode and received the allocated FoPNL (with information about excessive added sugars, sodium, and/or saturated fat content) or the control. After, they answered questions related to our primary outcome (decision to buy or not to buy a product) and secondary outcomes (perceived healthiness, facilitation of a quick purchase decision, and identification of excess nutrients). The Mexican FoPNL system performed better in the primary outcome (3.74 ± 1.34) and “facilitation of a quick purchase decision” (3.59 ± 1.31), compared to the control (3.28 ± 1.45;p = 0.043 and 3.11 ± 1.42; p = 0.029). The Mexican FoPNL system performed better in supporting consumers to identify dairy foods, among the selected sample in this study, high in added sugars than the control (82.2% and 63.5% of correct answers, respectively; p = 0.009). For saturated fats, the Brazilian FoPNL resulted in 93.1% of correct answers against 48.2% for the control and 58.9% for the Mexican system (p ≤ 0.001). The Mexican FoPNL system facilitated consumer decision-making on when to buy or not to buy a selected dairy product and in helping to quickly decide which dairy products to buy, among the selected sample in this study, compared to the control. Considering the right answers of critical nutrients in excess or not, both models of FoPNL, delivered by a smartphone app, performed well.

University Students' Knowledge and Perceptions About Concepts, Recommendations, and Health Effects of Added Sugars

It is recommended to limit added sugars to below 10% of the daily energy intake, as excessive consumption has been associated with several chronic non-communicable diseases. This exploratory qualitative study used focus groups to investigate the knowledge and perception of Brazilian university students about added sugars concepts, consumption recommendations, and health effects. Focus groups were led by a moderator using a semi-structured discussion guide. The focus groups were recorded, transcribed verbatim, and subjected to thematic analysis. Five focus groups were conducted with a total of 32 participants (50% women, mean age 23 years). Participants could not distinguish added sugars from sugars naturally present in foods and were unaware of the health impacts associated with excessive added sugar consumption, except for the risk of diabetes. Although most participants reported limiting sugar consumption, they had no knowledge of official consumption recommendations. Given that current public policy agendas aim to reduce added sugar intake, there is a need to strengthen strategies for disseminating information on added sugar concepts, recommendations, health effects and how to identify them in the foods products.

Consumers' Response to Sugar Label Formats in Packaged Foods: A Multi-Methods Study in Brazil

Providing information about the sugar content of packaged foods on product labels is an important strategy to lower consumers' sugar intake. This study assessed the effect of exposure to different sugar labels on consumers' understanding of the sugar content of foods and their food choices. In the first phase, five focus groups were conducted with a convenience sample of Brazilian adults to explore their perceptions about food labelling in general and sugar labelling in particular. Based on the qualitative results, four sugar label formats were developed and subsequently tested in a five-arm study on 1,277 adults via a randomised controlled online survey. The formats were: (i) no sugar information—control, (ii) total and added sugar content displayed in the Nutrition Information Panel (NIP), (iii) a front-of-package (FoP) octagonal warning for “high-in-sugar” products, (iv) a FoP magnifying glass warning for “high-in-sugar” products, and (v) a “high-in-sugar” warning text embedded on the NIP. Participants from the focus groups reported being confused about the meaning of “sugar” and “added sugar” on food labels and indicated that more interpretive labels, such as the FoP warnings, would help them choose products with low sugar content. In the experiment, all intervention sugar label formats improved participants' understanding of the sugar content of the tested food products, with the FoP warnings (iii and iv) showing the best results. While non-significant differences among label conditions were observed for food choices, the FoP octagonal warning prompted participants to choose high-in-sugar products less often. Given current public policy agendas aiming to reduce added sugar intake, there is a need to strengthen food labelling policies and nutrition disclosure policies that target the display of added sugar and build consumer awareness in using these tools to avoid high-in-sugar products.

A Machine Learning Approach to Predict the Added-Sugar Content of Packaged Foods

BACKGROUND

Dietary guidelines recommend limiting the intake of added sugars. However, despite the public health importance, most countries have not mandated the labeling of added-sugar content on packaged foods and beverages, making it difficult for consumers to avoid products with added sugar, and limiting the ability of policymakers to identify priority products for intervention.

OBJECTIVE

The aim was to develop a machine learning approach for the prediction of added-sugar content in packaged products using available nutrient, ingredient, and food category information.

METHODS

The added-sugar prediction algorithm was developed using k-nearest neighbors (KNN) and packaged food information from the US Label Insight dataset (n = 70,522). A synthetic dataset of Australian packaged products (n = 500) was used to assess validity and generalization. Performance metrics included the coefficient of determination (R2), mean absolute error (MAE), and Spearman rank correlation (ρ). To benchmark the KNN approach, the KNN approach was compared with an existing added-sugar prediction approach that relies on a series of manual steps.

RESULTS

Compared with the existing added-sugar prediction approach, the KNN approach was similarly apt at explaining variation in added-sugar content (R2 = 0.96 vs. 0.97, respectively) and ranking products from highest to lowest in added-sugar content (ρ = 0.91 vs. 0.93, respectively), while less apt at minimizing absolute deviations between predicted and true values (MAE = 1.68 g vs. 1.26 g per 100 g or 100 mL, respectively).

CONCLUSIONS

KNN can be used to predict added-sugar content in packaged products with a high degree of validity. Being automated, KNN can easily be applied to large datasets. Such predicted added-sugar levels can be used to monitor the food supply and inform interventions aimed at reducing added-sugar intake.

An Innovative Machine Learning Approach to Predict the Dietary Fiber Content of Packaged Foods

Underconsumption of dietary fiber is prevalent worldwide and is associated with multiple adverse health conditions. Despite the importance of fiber, the labeling of fiber content on packaged foods and beverages is voluntary in most countries, making it challenging for consumers and policy makers to monitor fiber consumption. Here, we developed a machine learning approach for automated and systematic prediction of fiber content using nutrient information commonly available on packaged products. An Australian packaged food dataset with known fiber content information was divided into training (n = 8986) and test datasets (n = 2455). Utilization of a k-nearest neighbors machine learning algorithm explained a greater proportion of variance in fiber content than an existing manual fiber prediction approach (R² = 0.84 vs. R² = 0.68). Our findings highlight the opportunity to use machine learning to efficiently predict the fiber content of packaged products on a large scale.