Association between advertising patterns and ultra-processed food in small markets

This article aims to characterize the advertising appeals present in the food environment to market ultra-processed foods and to analyze the nutritional profile of these foods according to PAHO criteria and the presence of food additives. Cross-sectional study, with data audited in 20 small supermarkets in São Paulo. The INFORMAS protocol was used to classify the advertising messages. The foods were classified according to NOVA. The PAHO profile model was used to classify foods high in critical nutrients. Advertising patterns were identified by factor analysis. The association between patterns and food groups was investigated by linear regression. More than 95% of the ultraprocessed foods had at least 1 critical nutrient in excess. There was a positive association between the new brand, fun and advantageous pattern with snacks, ready-made products, dairy products and cookies; between the new brand and suggested use pattern with dairy products. The standardization of food advertising in small re-tail stores is associated with offering snacks, dairy products, ready-to-eat foods and cookies, products that exceed critical nutrients.


Introduction
The food environment is the physical, economic, political and socio-cultural context in which consumers engage with the food system to make their decisions about acquiring, preparing and consuming food 1 . Recent conceptual models describe the food environment as a critical place in the food system to implement interventions to support sustainable diets and address the global syndemic of obesity, undernutrition and climate change, emphasizing its importance for food choices 2 . The food environment is complex and multidimensional and influences eating patterns 3 . This environment can present a risk factor for obesity and an unbalanced and unhealthy diet when it poses obstacles such as cost, difficulties accessing and lack of availability of healthy food, advertising of ultra-processed products and lack of information 4. The consumer food environment is the space in which food is available, where it is possible to plan and make purchases to prepare meals. This environment includes food retailers, such as supermarkets, grocery stores, fresh food markets, markets, restaurants and bars 5 . Within these spaces, factors such as food availability, price, placement on store shelves and advertising influence food choices, food consumption and spending on food 6,7 .
Data from Brazilian population surveys 8 show a relationship between the recent increase in consumption of ultra-processed foods and obesity and other chronic non-communicable diseases 4 . Ultra-processed foods are generally high in sugars, fats and sodium, low in fiber, vitamins and minerals, and contain a range of synthetic additives. In addition, they tend to be aggressively marketed and widely available in the consumer food environment, especially in supermarkets 9 .
A study in Brazil found that shopping in supermarkets is associated with an increase in the purchase of ultra-processed foods 10 and that supermarkets are the most common food shopping choice. In supermarkets, customers are exposed to a range of marketing strategies that use different methods to appeal to customers, including packaging and labelling, product placement, promotional pricing and brand advertsing 11 .
The marketing of ultra-processed foods is one of the most important factors influencing the purchase and consumption of these foods, especially among children and adolescents 12 . The range of advertsing methods used by food retail-ers, including displays, posters and overhanging signage, promotional leaflets and flyers, and promotional islands encourage impulse buying and the purchase of unhealthy foods 13 . Advertising messages using nutrition and health claims increase consumers' perceptions of the healthiness of foods, even when the claim is not consistent with the real nutrition value of the food 14 .
According to the literature, the marketing strategies used by food retailers can either favor or hinder the purchase of healthy food 15 . Brazil's regulatory agency, ANVISA, recently issued new labelling requirements for packaged food products (RDC Resolution n. 429/2020), advancing the debate on nutrition labelling on commercial foods and beverages. However, gaps remain in the discussion of advertising on food packaging and in the consumer food environment. Further research is therefore warranted to investigate the relationship between advertising in the consumer food environment and sales of foods high in critical nutrients like salt, fat and sugar. The evidence produced by such studies can provide valuable inputs to help shape policies regulating the marketing and labelling of foods and beverages in the food environment, thus assisting consumers to make healthy food choices. To promote the creation of healthy food environments it is vital to explore the relationship between advertising claims and the nutrient content of foods available in the consumer environment. To this end, the aim of this study was to characterize the advertising claims used in the consumer food environment to sell ultra-processed foods and analyze the nutrient profile of these types of products according to the criteria adopted by the Pan American Health Organization (PAHO) and presence of food additives.

Study design and sample
We conducted a cross-sectional study using data from advertising messages on ultra-processed foods and beverages collected in the consumer food environment. Convenience sampling was used to select a group of small grocery stores (up to three checkouts) 17 in three cities in the Metropolitan Region of Sao Paulo. Twenty grocery stores were visited (10 in Osasco-SP, three in São Paulo-SP and seven in São Bernardo do Campo-SP). Using Google Maps, we selected stores located within a 2km-radius from the researchers' place of study and residence to facilitate access to the establishments. The stores selected in Osasco and São Bernardo do Campo were located in low-income neighborhoods, while those in São Paulo were from middle to high-income neighborhoods.

Consumer food environment audit
We carried out an internal audit of the grocery stores using the previously validated AU-DITNOVA audit instrument 18 . For the purposes of this study, only the 18 ultra-processed foods considered by the AUDITNOVA were assessed. The instrument was adapted to capture information on the marketing strategies used in the stores. The researchers identified which of the 18 foods were sold in the store and recorded the advertising messages related to each food in their entirety (for example: Practicing joy is good for you! Now even more crispy! Give it try!). The messages were codified and grouped into categories with similar themes following a specific protocol described below. The researchers were previously trained to use the data collection protocols developed for the application of AUDIT-NOVA 18 . The data were collected in July and August 2018.

Study variables
The advertising messages were grouped into nine categories of advertising claims adapted from the INFORMAS (International Network for Food and Obesity/non-communicable Diseases Research, Monitoring and Action Support) Protocol -Food Promotion Module 19 . The aim of the protocol is to provide guidance on the monitoring of food marketing in different countries. The following categories were used: 1. Sensory-based characteristics (taste, texture, appearance, aroma); 2. New brand development; 3. Suggested use (e.g. great for lunchboxes); 4. Suggested users are children or whole family; 5. Emotive claims (fun, feelings, popularity); 6. Puffery (claiming to be advantageous over other products); 7. Convenience; 8. Price (promotions and gifts); 9. Health and nutrition.

Processing of nutrient content data
We collected information on ingredients and nutrients (macro and micronutrients) from the food ingredient lists and nutrient facts labels. All items on the list were included except food additives, which were analyzed separately. Information on additives was collected from the manufacturers' website. When the information was not available on the manufacturers' website, it was collected in the stores.
The macro and micronutrients were calculated per 100 grams or 100 ml. As it is not mandatory for food manufacturers to declare the free sugars content on the nutrient facts label in Brazil, based on PAHO criteria 16 , we estimated free sugars content based on the total sugars content declared on the nutrient facts label. However, as it is also not mandatory for manufacturers to declare the total sugars content on the nutrient facts label, only 41% of the products declared the sugar content. For sugar-sweetened beverages and dairy products with sugar on the ingredient list, total sugars content was therefore estimated based on carbohydrate content per 100g/ml using the definition of total sugars established in the nutrient profile model proposed by the WHO Regional Office for Europe 22 .
To determine whether the ultra-processed food was nutritionally unbalanced, we used the PAHO Nutrient Profile Model 16 for critical nutrients. The following critical nutrients were analyzed: free sugars, sodium, total fats, saturated fats and other sweeteners. We did not include trans fats because, despite the inclusion of this information on food labels being mandatory since 2003, a study reported that current legislation fails to ensure that manufacturers provide clear information about the trans fat content of packaged foods 23 . The following cutoff points set out in the PAHO Nutrient Profile Model 16 were used to classify whether a food was high in a critical nutrient: sodium ≥ 1 mg of sodium per kcal; free sugars ≥ 10% of total energy value comes from free sugars; saturated fats ≥ 10% of total energy value comes from saturated fats 16 .

Statistical analysis
Exploratory factor analysis was used to determine patterns of advertising claims related to food groups. This method was chosen because we did not know which potential combinations of advertising claims to expect for each ultra-processed food group. The Kaiser-Meyer-Olkin (KMO) test was used to measure sampling adequacy and how suited the data was for factor analysis, where a value of > 0.6 indicates that sampling is adequate. The following criteria were used to determine the number of factors to retain in the exploratory factor analysis: eigenvalue > 1.0 and Cattell's scree test. Orthogonal rotation was performed using the varimax method, which adjusts the factors to find a better distribution of factor loadings, producing a simpler and more easily interpretable structure of factors. Factor loadings after rotation of > 0.20 were taken to indicate an advertising claim pattern. After extracting the factors, we calculated standardized scores for each factor and each food group, where the higher the score, the stronger the association with the food group.
The variables were analyzed using descriptive statistics (means and 95% confidence intervals and relative frequencies of the presence of critical nutrients and additives in each food group). Linear regression models were used to measure the association between advertising claims patterns and food groups, where the outcomes were the pattern scores and the exposure variables were the food groups. A correlation coefficient of > 1.0 was considered to indicate a strong positive association. Statistical analyses were performed using Stata version 15.
The purpose and procedures of the study and possible risks were explained to the store owners and each owner signed an informed consent form. The study protocol was approved by the research ethics committee (CAAE N o 69045917.5.0000.5421).

Results
A total of 202 advertising messages were identified in the 20 stores audited by this study. The ultra-processed foods that accounted for the largest share of advertising messages were sugar-sweetened beverages (28.2%), followed by baked goods and cookies and crackers (both 25.8%), reconstituted meats (16.8%), and sweet snacks (10.9%). Table 1 shows the mean content of macro and micronutrients per 100 g/ml for each ultra-processed food group. The groups that showed the highest means for energy value were savory snacks, cookies and crackers and sugary cereals, respectively, while the groups with the highest carbohydrate content were cookies and crackers, sugary cereals and savory snacks, respectively. The group with the highest mean protein content was reconstituted meats. The groups with the highest mean fat content were savory snacks and reconstituted meats, while the groups with the highest mean saturated fat content were margarines and savory snacks. The highest mean total sugar contents were found in sweet snacks and sugary cereals, while the groups with the highest mean sodium content were ready-to-eat foods and reconstituted meats. Sugary cereals had the highest mean fiber content.
The analysis of the critical nutrients showed that more than 80% of sweet snacks were high in free sugars and more than 50% were high in total and saturated fats. All ready-to-eat foods were high in sodium and more than 70% of cookies and crackers were high in total fats. The overwhelming majority of ultra-processed foods (more than 95%) were high in at least one critical nutrient ( Table 2).
The foods with the highest mean number of ingredients (excluding additives) were sugary cereals (14.9), reconstituted meats (10.8) and savory snacks (10.2), while the foods with the highest number of additives were dairy products (13.3), reconstituted meats (11.3) and fats and baked goods (approximately nine each). With regard to types of additives, 50% of dairy products contained emulsifiers and sweeteners and 66.6% contained acidulants; almost 30% of reconstituted meats contained acidulants; 85% of baked goods contained emulsifiers and 38,5% contained sweeteners; and approximately 30% of sugar-sweetened beverages contained sweeteners (Table 3).
Four patterns of advertising claims were identified: Pattern 1, which we named the "new brand, fun and advantageous pattern" and con-sisted of new brand development, suggested use (e.g. great for lunchboxes), emotive (fun, feelings, popularity), and puffery (claiming to be advantageous over other products) claims; Pattern 2, which included emotive claims and was named the "fun and popular" pattern; Pattern 3, named the "healthy and nutritious" pattern, containing health and nutrition claims; and Pattern 4, made up of new brand development and suggested use claims and named the "new brand and suggested use" pattern ( Table 4).
The results of the linear regression analysis showed a positive association between Pattern 1 -New brand, fun and advantageous and savory snacks, ready-to-eat foods, dairy products, and cookies and crackers, and between Pattern 4 -New brand and suggested use and dairy prod-ucts, and a negative association between Pattern 2 -Fun and popular and reconstituted meats. The positive associations indicate that the number of these claim patterns are higher in the respective food groups, showing that these groups are more likely to be associated with these patterns. No statistically significant associations were found between Pattern 3 -Healthy and nutritious and the food groups (Table 5).

Discussion
Our findings show an association between different types of advertising claims and specific groups of ultra-processed foods. Savory snacks, ready-to-eat foods, dairy products and cookies and crackers were associated with a wider variety of advertising claims (Pattern 1 -New brand, fun and advantageous), illustrating that retailers and food manufacturers use a diverse range of strategies to promote these products. In addition, more than 95% of the foods available in the stores assessed by this study were nutritionally unbalanced, exceeding the limits set by the PAHO for at least one critical nutrient and illustrating the mass exposure of consumers to unhealthy foods. As mentioned above, the stores assessed by the present study located in Osasco and São Bernardo do Campo were located in low-income neighborhoods. A study in Brazil conducted in Jundiaí-SP showed that the concentration of small food retailers is higher in low-income neighborhoods with low levels of education located in urban peripheries 24 . A nationwide study conducted by the Interministerial Chamber of Food and Nutrition Security (CAISAN) in 2018 showed a predominance of "mixed food retailers" (which sell both healthy and unhealthy food) over other types of retailers. This classification of retailer proposed by CAISAN includes grocery stores with similar characteristics to those of our sample 5 .
A study by Machado et al. undertaken in 2018 using data from the 2008/2009 National Household Budget Survey reported that the traditional retailer purchasing pattern (markets and street vendors, small grocery stores, small farmers, butchers) resulted in a lower share of ultra-processed foods in food purchases than the supermarket pattern 10 . However, while the study showed that purchasing food in small retailers was associated with the purchase of healthier food, it did not assess food marketing. According to conceptual models of the food environ- Table 2. Relative frequency of critical nutrients in the ultra-processed food groups using advertising claims in the consumer food environment according to the PAHO Nutrient Profile Model (n = 202). São Paulo, 2018.   ment, food marketing exerts a strong influence on food choices and should be assessed in combination with food availability rather than as an isolated factor 2,3 . Our findings show that small food retailers are characterized by advertising claims patterns that prioritize products with low nutritional value, showing that the advertising practices used by these establishments should be monitored and evaluated.

Ultra-processed food groups* High in free sugars (%) High in total fats (%) High in saturated fats (%)
Manufactures use a diverse range of marketing strategies in the consumer food environment to increase sales and influence consumers at the time of purchase 6,7 . Our results show that savory snacks, dairy products, ready-to-eat foods, and cookies and crackers used the widest range of advertising claims, including new brand development, suggested use, emotive claims and advantageous over other products. In addition, these foods were high in critical nutrients. This is particularly worrying given the increasing consumption of these foods among children 25 and adolescents 26 . Mass marketing in grocery stores combined with the growing availability of ultra-processed foods has contributed to this problem, as the eating behaviors of this age group are influenced 4 by advertising strategies used to promote these types of foods in other media such as television 27 .
A study exploring the marketing and quality of foods sold in supermarket chains in Florianópolis assessed 16 online and printed food promotion leaflets advertising 1,786 foods. The findings showed that the proportion of ultra-processed foods advertised on the leaflets was significantly higher than the proportion of fresh or minimally processed foods (p < 0.001) 28 . Although the present study did not assess advertising strategies for fresh or minimally processed foods, it is known that, despite their more limited availability 7 , these foods contribute to an increase in consumption of healthy foods in environments such grocery stores and supermarkets, especially those located in low-income neighborhoods 29 .
Our findings show that dairy products (or milk drinks) had the highest number of ingredients and additives. Studies show that food additives are associated with inflammation, metabolic changes and cancer, and can lead to changes in microbiota composition, intestinal barrier and other gastrointestinal disorders 22,21 .
Dairy products accounted for the largest share of claims in Pattern 4. "new brand and suggested use". The Suggested use claim was characterized by the keywords "any time", "any place", "to wake up", "eat together at the table" "great for lunchboxes". In other words, expressions that suggest the ideal time or place to eat or drink the product, including the school environment, or that encourage impulsive eating at any time and in any place. This type of advertising is designed to stimulate overeating and is reflected in an increase in consumption of these foods among Brazilian adolescents. According to the 2017-2018 Household Budget Survey 8 , adoles-cents drink nine times more milk drinks than older people and ultra-processed foods account for 27% the total daily calorie intake among adolescents, compared to only 15.1% in the ≥ 60 years age group.
In an attempt to reduce the consumption of ultra-processed foods in Brazil, ANVISA issued new regulations in October 2020 30 governing front-of-package labelling. This type of regulation plays an important role in promoting healthy food choices as food labels provide consumers with information on the nutrient content of foods, potentially reducing intentions to purchase ultra-processed foods due to improved perceptions of the healthiness of these products 31 . The new regulation means customers are able to identify foods with high levels of critical nutrients from the warning labels on the front of the packaging. However, based on the INFOR-MAS protocols designed to help monitor and benchmark food environments, Brazil also needs to push forward measures to regulate the quality of food sold by food retailers and the advertising strategies used in these environments, which in this study were associated with food groups made up of nutritionally unbalanced foods.
The patterns of advertising claims found in this study show that food retailers and manufacturers use a broad range of different strategies to increase sales and stimulate consumption. These strategies can influence consumers' perceptions of the healthiness of products, leading them to buy certain foods because they believe they are healthy, advantageous or nutritionally superior to others 14 .
Ultra-processed food advertising is a barrier to healthy eating 4 . In this regard, the findings of this study provide a deeper understanding of the consumer food environment as a barrier to healthy eating, especially in small retailers, which are present in the majority of Brazilian municipalities and play a fundamental role in supplying food in areas located in urban peripheries with lower socioeconomic status 5 . In addition, the findings highlight the need to broaden discussions regarding the monitoring of food availability and marketing in food retailers. With new reg-ulations in Brazil stipulating that food and drink products high in sodium, fats or sugar must be labelled with front-of-pack nutritional warnings due to come into force next year, the country could see an increase in these types of marketing strategies in the food environment, as witnessed in Mexico when similar regulations were introduced 32 .
One of the strengths of this study is that it characterized consumer food environment variables (presence of advertising claims, nutrient profile, and main ingredients and additives), while other studies have focused on specific aspects such as labels, prices or food availability. Our findings therefore provide a more detailed insight into the consumer food environment, especially in low-income areas. Limitations include the fact that the sample was selected using convenience sampling and is not representative of small food retailers. However, the increase in market concentration in the hands of large supermarket chains affects and homogenizes the sales practices and marketing strategies adopted by small food retailers 7 , which are similar across different regions of Brazil.
In conclusion, our findings show that a range of different advertising strategies are used in the aisles and on the shelves of small retailers in the Metropolitan Region of Sao Paulo to promote nutritionally unbalanced ultra-processed foods containing various types of food additives. The "new brand, fun and advantageous" pattern was particularly associated with savory snacks, ready-to-eat foods, dairy products and cookies, and all items (100%) in these groups were high in at least one critical nutrient. In addition, the average number of additives contained in dairy products was 13. The findings show that specific advertising patterns were used in the stores to promote certain types of ultra-processed foods. The type of advertising claims used, ranging from the launch of new products to fun, emotive, health and well-being claims, influence food choices in these environments and encourage the consumption of unhealthy foods, underlining the need to develop policies and actions to regulate the consumer food environment in Brazil.

Collaborations
CHK Batista contributed substantially to study conception, planning and data analysis and interpretation; FHM Leite contributed substantially to data analysis and revising the manuscript critically for important intellectual content; CA Borges contributed to data analysis and interpretation and revising the manuscript. All authors approved the final version to be published.