Choosing a front-of-package warning label for Brazil: A randomized, controlled comparison of three different label designs

Food Research International 121 (2019) 854–861

Contents lists available at ScienceDirect

Food Research International journal homepage: www.elsevier.com/locate/foodres

Choosing a front-of-package warning label for Brazil: A randomized, controlled comparison of three different label designs

T

⁎

Neha Khandpura, , Laís Amaral Maisb, Priscila de Morais Satoa, Ana Paula Bortoletto Martinsb, Carla Galvão Spinilloc, Carlos Felipe Urquizar Rojasc, Mariana Tarricone Garciad, Patrícia Constante Jaimea a

Center for Epidemiological Studies in Health and Nutrition (NUPENS), Faculty of Public Health, University of São Paulo, Brazil Brazilian Institute for Consumer's Defense (IDEC), Brazil c Research Group of Digital and Information Design, Department of Design, Federal University of Paraná, Brazil d Health Institute, São Paulo State Health Secretariat, Brazil b

A R T I C LE I N FO

A B S T R A C T

Keywords: Warning labels Nutrition label design Randomized experiment Brazil Front-of-package labels Health promotion

Introduction: Warning labels (WLs) are the newest paradigm of nutrient-based, front-of-package labels (FOP) that have been consistently shown to be more effective at informing consumer purchases than other FOP labels. Their effectiveness may be attributed to their design and the consistency in the information they communicate. The present study assessed designs effects of WLs in improving understanding and perceptions among 2419 Brazilian adults. Methods: Participants were randomly assigned to one of four study arms in an online, randomized, controlled experiment: (1) the Triangular WL ‘A lot of /Muito’ (Muito); (2) Triangular WL ‘High in/Alto em’ (Alto); (3) the Chilean WL ‘High in/Alto em’ (Chile); and (4) a control condition (no WL). Participants responded to a series of questions, while viewing images of 9 products, that assessed their understanding of the nutrient content of the product, its healthfulness, and their intentions to purchase. All aggregate differences in responses between study arms were estimated using oneway-ANOVAs. Results: WLs were superior to the list of ingredients and the nutrition facts panel in improving participant understanding and perceptions of the nutrient profile of products, particularly in helping identify nutrients in excess. Alto was significantly better than the control at identifying which of the two products contained an excess of nutrients and at improving consumer understanding of nutrient content across different indicators. Alto, followed by Muito, was also better at communicating that fewer WLs signified a healthier product. The Muito WL was significantly more visible to participants than the Chile WL. There were no differences between study arms in shifting consumer purchase intentions. Conclusions: In this study, WLs were an important addition to the current nutrition label. The triangular symbol, the white background and the signal word, ‘High in/ Alto em’ showed consistently better results than the signal word ‘A lot of/ Muito’ and the octagonal symbol. These design elements serve to make WLs more visible to consumers, capturing their attention and informing their food evaluations, and may have greater potential to influence appropriate for a Brazilian audience.

1. Introduction Front-of-package labels (FOP) are icons, symbols or logos that provide consumers with simple and easy-to-interpret information about the nutrient profile of a packaged product at points of purchase (Boon, Lichtenstein, & Wartella, 2010). Their objective is to aid consumers in the selection, purchase and consumption of healthy products. FOP have also emerged as a tool to encourage the food industry to reformulate ⁎

existing products and introduce healthier options (Kanter, Vanderlee, & Vandevijvere, 2018; Vyth, Steenhuis, Roodenburg, Brug, & Seidell, 2010). International health agencies like the World Health Organization (WHO) and the Pan American Health Organization (PAHO) recognize the importance of FOP as an important, and perhaps even a necessary, part of the cluster of measures targeting the promotion of healthy diets (PAHO, 2016; WHO, 2013). Warning labels (WLs) are the newest paradigm of nutrient-based

Corresponding author. E-mail address: [email protected] (N. Khandpur).

https://doi.org/10.1016/j.foodres.2019.01.008 Received 24 September 2018; Received in revised form 3 January 2019; Accepted 4 January 2019 Available online 07 January 2019 0963-9969/ © 2019 Published by Elsevier Ltd.

Food Research International 121 (2019) 854–861

N. Khandpur et al.

There exists limited evidence on consumer opinions of WL design features (Acton et al., 2018) and on the design effects of WLs on consumer perceptions of product healthfulness (Cabrera et al., 2017). The present study seeks to add to the evidence base by evaluating diverse consumer-based outcomes using a randomized, controlled experiment. The results from this work will inform policy decisions within Brazil and across other countries preparing to select appropriate FOP labels for their populations. This study compared improvements in consumer-based outcomes between the Triangular warning label ‘A lot of /Muito’ (Muito), Triangular warning label ‘High in/Alto em’ (Alto), and the Chilean octogonal warning label ‘High in/Alto em’ (Chile) (Fig. 1). A four-arm, online, randomized, controlled experiment was conducted to determine which of the three WL designs was best at: (i) improving understanding of the nutrient content of the product; (ii) reducing perceptions of the healthfulness of the product and; (iii) reducing purchase intentions. Participants'’ opinions about the label were also captured. The study also assessed if the presence of WLs was more effective than the current sources of nutrition information (nutrition facts panel and the list of ingredients). The study hypothesized that WLs would perform better than the current sources of nutrition information and that the black triangular WL, with a white background and the formal signal words ‘Alto em’ would demonstrate the greatest design effects.

FOP that have been implemented in Chile (Corvalán, Reyes, Garmendia, & Uauy, 2013) and have been signed into policy in Peru and Uruguay (Ministry of Health Peru, 2017; Ministry of Health, Uruguay, 2018). WL icons are displayed on ultra-processed products (Monteiro et al., 2018) that contain an excess of sodium, sugars, or fats (and an excess of calories in the case of Chile). A separate WL is displayed for every nutrient in excess. Unlike other FOP labels that display nutrient content (like the Guideline Daily Amount label), use colors to signal nutrient level (high/medium/low, like the Multiple Traffic Light symbol), or indicate the overall nutritional quality of the product by combining content of positive (protein, and fibre) and negative nutrients (sodium, sugars, and fats) (like the Health Star Rating), WLs are only present on ultra-processed products when negative nutrients are above recommended levels (Kanter et al., 2018). WLs have consistently been shown to be more effective than other FOP labels. Compared to the Multiple Traffic Light symbol, WLs are better at: (1) capturing consumer attention; (2) improving consumer understanding; (3) reducing perceptions of healthfulness of unhealthy products; and (4) reducing the intention to purchase unhealthy products among adult Brazilian consumers (Khandpur et al., 2018; Khandpur, Swinburn, & Monteiro, 2018). Even among children, WLs have been shown to significantly discourage choices of relatively unhealthy products like wafers and juice compared to the traffic-light labels (Arrúa et al., 2017). They allow for the quickest identification of excess nutrients (Arrúa et al., 2017) and have more pronounced decreases in consumer perceptions of product healthfulness and purchase intentions when compared to summary labels like the Nutri-Score and the Health Star Rating (Ares et al., 2018). These improvements in consumer-based outcomes may be attributed to the design elements of WLs, and the simplicity and consistency in the information they communicate (Acton, Vanderlee, Roberto, & Hammond, 2018; Khandpur, Swinburn, & Monteiro, 2018). WLs are comprised of a universally recognized, colored symbol for an alert, and a signal word (Edworthy, 1996; Wogalter, 1994, Wogalter, DeJoy, & Laughery, 2005). Their symbol makes them easy to identify and might aid in information recall (Strasser, Tang, Romer, Jepson, & Cappella, 2012). The use of simple language and the absence of numbers on WLs may also promote understanding among consumers with low education levels, who find it difficult to interpret percentages and technical terms on labels (Cowburn & Stockley, 2005). The Chilean model, for instance, comprises of a black octagon with white borders, the signal phrase ‘High in/ Alto en’ in white to indicate the nutrient in excess, and the endorsement of the Ministry of Health/ 'Ministerio de Saude’ (Fig. 1, Ministerio de Salud, Chile, 2015). An alternative WL design was developed by Brazilian researchers specializing in information design and presented to the regulatory agency as a FOP option for Brazil (ANVISA, 2018). This label displays a black triangle, an internationally standardized shape for warnings (ISO, 2011; Wogalter, Silver, Leonard, & Zaikina, 2006), with white lettering, against a white square background (Fig. 1). Both the symbol and its background present a starker contrast to the colors of the packaged products on which they appear. Additionally, they are thought to be more likely to catch consumer attention when stacked together on smaller packaged products. This may be due to the fact that the triangle is a simpler geometric figure than a more complex figure like an eightsided octagon and is therefore perceived more promptly (Koffka, 2013). With regards to the signal word, preliminary results from a qualitative evaluation undertaken by the authors suggest that the Brazilian consumer may be more familiar with the phrase, ‘A lot of /Muito,’ to signal the presence of excess nutrients. However, there was some suggestion that this phrase may be positively interpreted by consumers and may stimulate perceptions of healthfulness. The more formal, ‘High in /Alto em’ was suggested as an alternative. This phrase appears on health reports and diagnostic tests in Brazil (Izar & Fonseca, 2007) and may more suitably signal the negative impact of the presence of excess nutrients.

2. Methods 2.1. Participant recruitment The study was conducted in the form of a survey and administered in Portuguese via an online platform during the first two weeks of August 2017. Recruitment was conducted by a Brazilian research firm. 2419 adults who were responsible for grocery shopping, had no ties to the food industry, were not health professionals and had not previously worked for market research firms, formed the study sample. Details of study recruitment and reimbursement may be found elsewhere (Khandpur, de Morais Sato, et al., 2018). Fig. 2 describes the final sample size. The study was approved by the ethics committee of the University of São Paulo [#68795417.6.0000.5421]. 2.2. Experimental design After obtaining consent, participants were randomly allocated to one of the four study arms. Participants saw the same images of food products and responded to the same set of questions in all the study arms. The presence and type of WL on the food products was the only aspect that differed across arms. Study arm 1 was the control condition where product images did not display any WL. In study arm 2, participants saw product images with the triangular WL with the signal word ‘A lot of/Muito’ (Muito). Study arm 3 had images of products with the triangular WL and the signal word ‘High in/Alto em’ (Alto). In study arm 4, products displayed the Chilean octogonal WL with the signal word ‘High in/Alto em’ (Chile). WLs in study arms 2, 3 and 4 targeted six nutrients. They were displayed when the product contained an excess of free sugars, saturated fat, total fat, or sodium or if the product contained any amount of trans fat or artificial sweeteners (Fig. 1). The Nutrient Profile Model of the PAHO was used to determine nutrient thresholds using the actual nutrient information displayed on the nutrition facts panel of the products. WLs were displayed on the top right corner of the products and were approximately 20% the size of the front panel. The size of the WLs and colors used in these options were informed by existing literature and variations in these two design aspects were not studied. The colour black has previously been associated with reduced perceptions of healthfulness and 10–20% is the minimum recommended size for a WL to be effectively seen (Cabrera et al., 2017). Additionally, the 855

Food Research International 121 (2019) 854–861

N. Khandpur et al.

Fig. 1. Warning label designs and example product images.

included the front portion of the product, the nutrition facts panel, and the list of ingredients for every product (Fig. 1). Participants had the option to zoom in on images of the products in sections one and two. In the first section participants viewed three products – a cereal bar, crackers and chocolate milk – displayed one at a time. A total of five questions were asked for each product (Table 1). These questions were designed to objectively measure consumer understanding of WLs, intentions to purchase the product, and perceptions about its nutrient content. In the second section of the survey, participants saw a pair of products from different brands but the same food category – breakfast cereals, breads and juices. Three product pairs were shown in total and participants responded to three questions per product pair. Questions in this section were designed to simulate in-store product evaluation tasks of choosing between multiple products and selecting the most preferred option to purchase. The third section showed only the stand-alone image of the WL (Muito or Alto or Chile). Participants responded to thirteen questions to determine their opinion of the WL (Table 1). The fourth and final section recorded socio-demographic characteristics of the participants. Sections one, two and four, were identical for every study arm in terms of the sequencing of questions and the images seen. The only difference was the absence of a WL on products in the control condition. Section three was only for participants in the WL arms.

Fig. 2. Study recruitment.

positioning of all WLs was chosen as the top right corner of the front panel of the products as this location has been previously shown to capture the most consumer attention (Bialkova & van Trijp, 2010; Buckley & Shepherd, 1993).

2.4. Study indicators Participant responses across all products were combined to create the following scores: Nutrition knowledge score. Correct responses to questions that asked participants about the principal nutrient associated with hypertension [correct response: sodium] and the average calories consumed by a healthy adult [correct response: 2000 kcal] were combined and converted into a percentage (0−100). Understanding of nutrient content. Correct responses to questions 2

2.3. Study procedures The survey was divided into four sections. The first two sections included images of products that were easily available in Brazilian supermarkets and were perceived as being healthy by the general public. These products also had at least one of the six nutrients in excess. All product images were created in Adobe Photoshop CC 2017 and 856

Food Research International 121 (2019) 854–861

N. Khandpur et al.

Table 1 Survey questions. Indicator

Survey question

Response scale

Section 1 - Participants see images of three single products, one at a time Purchase intentions (single (1) Imagine you were looking to buy [type of product]. Would you buy this product or product) a similar product from a different brand, for yourself or your family? Identifying nutrients in excess (2) Do you think this product contains certain nutrients in levels higher than recommended for a healthy diet? Consider 1 portion of this product.

Subjective perceptions of nutrient levels Perceptions of product healtfulness Identifying nutrients not in excess

(3) How much of these nutrients are in one portion of this product? Total fat, Sugar, Sodium, Saturated fat, Trans fat, Artificial sweeteners (4) Do you think this product is healthy? (5) Do you think this product contains certain nutrients that are not in excess or rather within recommended levels for a healthy diet? Consider 1 portion of this product.

Section 2 - Participants see images of two products at the same time Purchase intentions (product pair) (6) Imagine you were looking to buy [type of product]. Would you buy one of these products for yourself or your family?

Product pair with higher nutrient content

(7) Which of these products has a larger quantity of the following nutrients: Total fat, Sugar, Sodium, Saturated fat, Trans fat, Artificial sweeteners

Healthier product pair

(8) Which of these two products is relatively healthy?

Section 3 - Participants see the image of the label only Label opinions (9) In your opinion, how frequently should a product with this label be consumed? (10) In your opinion, in what quantities should a product with this label be consumed? (11) In your opinion, how healthy would it be to excessively consume a product displaying this label? (12) What would you do if you saw this label on a product that you usually buy for yourself or your family? (13) I think this label is difficult to read. (14) I think this label will not help me buy healthier food. (15) I would like to see this label on packages of food and drink. (16) I think this label is confusing. (17) I consider the information on this label credible and true. (18) I think this label is easy to understand. (19) This label will help me decide what products to buy. (20) The label on the product draws my attention. (21) The label on the product is not visible.

7-point Likert scale ‘I would certainly not buy’ - ‘I would definitely buy’ Choice of multiple response options: Total fat Sugar Sodium Saturated fat Trans fat Artificial sweeteners Or the response option: None of these nutrients are in excess 7-point Likert scale ‘None’ - ‘A lot’ 7-point Likert scale ‘Not at all healthy’ - ‘Very healthy’ Choice of multiple response options: Total fat Sugar Sodium Saturated fat Trans fat Artificial sweeteners Or the response option: None of these nutrients are in excess Product A Product B Both products Neither product Product A has more of this nutrient Product B has more of this nutrient Both products have high levels of this nutrient Both products have low levels of this nutrient Product A is healthier Product B is healthier 7-point Likert scale ‘Never’ – ‘Always’ 7-point Likert scale ‘In small quantities’ – ‘In large quantities’ 7-point Likert scale ‘Not at all healthy’ – ‘Very healthy’ 7-point Likert scale ‘I would not buy it’ – ‘I would continue buying it’ 7-point Likert scale ‘Totally disagree’ – ‘Totally agree’

while Nutrient perceptions (low nutrients), captured responses to nutrients that did not require WL. Responses ranged from 1 (none) to 7 (a lot). Similarly, responses across all 3 products for question 4 were combined to create an average Perceptions of product healthfulness response that ranged from 1 (not at all healthy) to 7 (very healthy). Purchase intentions. Responses to question 1 were averaged across all three single products to create an overall response, Purchase intentions (single product), that ranged from 1 (certainly not buy) to 7 (certainly buy). For product pairs, responses to question 6 were recoded to indicate purchase of the unhealthy product, purchase of the healthy product and purchase of neither product. The average number of participants that selected one of the three response options, across the three product pairs, was calculated to create a summary indicator Purchase intentions (product pair). Label opinions. Summary responses were created for perceived label

were added across all nutrients and products and converted into a percentage (0–100) to reflect participant ability in Identifying nutrients in excess. Correct responses to question 5 were similarly combined to create the Identifying nutrients not in excess indicator. For product pairs, correct responses to question 7 were added across all nutrients and products and converted into a percentage (0–100) to assess participant ability to identify the Product pair with higher nutrient content. The same approach was used with responses to question 8 to create the Healthier product pair indicator. An indicator that assessed overall ability to identify elevated nutrient content was created by combining scores from the above indicators. Nutrient and product perceptions. Participant perceptions of nutrient levels captured in question 3 were combined across products to create two separate indicators. The Nutrient perceptions (high nutrients), captured mean responses to nutrients that were in excess and required WL 857

Food Research International 121 (2019) 854–861

N. Khandpur et al.

Table 2 Study demographics. Indicators

Age, mean yrs. (SD) BMI, mean kgs/m2 (SD)

Total sample

Study arms

n = 2419

Control n = 604

TAO n = 607

THI n = 607

CHI n = 601

p-Value

38.10 (13.35) 28.22 (10.39)

38.22 (13.42) 27.92 (9.75)

37.92 (12.17) 28.66 (11.98)

38.11 (13.47) 28.56 (10.98)

38.16 (13.37) 27.73 (8.51)

0.982 0.313

50.41 49.59

50.41 49.59

50.92 49.08

0.997

3.46 78.25 18.29

5.11 77.27 17.63

3.99 79.53 16.47

0.540

17.63 48.76 33.61

17.46 48.76 33.77

17.64 44.59 37.77

0.698

46.62

44.48

45.26

0.900

15.49

14.33

14.64

0.947

19.60

17.96

22.46

0.258

33.28

29.32

28.62

0.296

59.47 40.53

61.29 38.71

61.73 38.27

0.412

4.42a

4.38

4.48a

0.002

4.63a

4.59a

4.67a

< 0.001

57.74

57.01

59.06

0.345

Sex, % Female 50.52 50.33 Male 49.48 49.67 Education, % Primary or less 3.97 3.31 Secondary 79.00 80.96 Tertiary 17.03 15.73 SES, % Low 17.36 16.72 Medium 47.25 46.85 High 35.39 36.42 Live with children < 15 yrs. of age, % Yes 45.51 45.70 Currently diagnosed with CVD or cholesterol, % Yes 14.76 14.57 Currently diagnosed with diabetes or hypertension, % Yes 20.17 20.70 Currently dieting, % Yes 30.26 29.80 Responsible for grocery shopping, % Solely responsible 59.98 57.45 Co-responsible 40.02 42.55 Frequency of reading labels 1 Never – 7 Always 4.36 4.15a Frequency that labels influence food choices 1 Never – 7 Always 4.55 4.31a Nutrition knowledge score 0–100 58.53 60.34

Comparing study arms

Responses with the same superscript are significantly different from each other based on Tukey HSD post-hoc tests.

effects on behaviour (question 9–12), understanding (question 16, 18), helpfulness (question 14, 15, 17, 19) and visibility (question 13, 20, 21). Statements were reverse coded where required.

as participants in the label arms were cued by the presence of the WLs into seeking out the information they displayed. Controlling for these variables did not change the results.

2.5. Statistical analysis

3.2. Understanding of nutrient content

All aggregate differences in continuous mean responses between control and label conditions for the indicators described above were estimated using t-tests and oneway-ANOVAs. Exploratory analyses were also conducted using linear regression models to assess if sex, education or socio-economic status (SES) modified the effect of the WL designs on the study outcomes. All significant omnibus tests, with p < .05, were followed by posthoc Tukey HSD tests to assess differences between study arms. Stata v.14 was used to analyse the data.

Four indicators captured different dimensions of this construct (Table 3). Among single products, there were significant differences in the mean scores reflecting participant ability in Identifying nutrients in excess across study arms. Participants in the Alto WL arm correctly identified the most nutrients in excess while those in the control arm identified the least. Post-hoc tests indicated that the control arm was significantly different from all label arms, but no significant differences were seen between label arms. There were no differences across any of the study arms for the indicator capturing participants' ability to identify nutrients that were not in excess (Identifying nutrients not in excess). Among product pairs, there were significant differences across study arms in participant ability to correctly identify the product with the higher nutrient content. Participants in the Alto study arm performed best on this indicator and had significantly greater correct responses compared to the control. No other comparisons were significant. Exploratory sub-group analysis showed that men had higher correct scores with the Alto label than women (42.19 vs 38.56, p-value = .05), while women scored higher in the Chile WL arm compared to men (41.83 vs 36.74, p-value < .01). No other differences were seen. Significant differences were also seen for participants' ability to identify the overall healthier product between two options (Healthier product pair). Participants in the Alto study arm scored the highest on this indicator suggesting improved ability to detect the healthier product, with scores significantly higher in comparison with the control.

3. Results 3.1. Participant characteristics Table 2 presents the descriptive characteristics of the study sample. The sample had almost equal representation from men and women, mean age was 38.1 years ( ± 13.3), nearly 80% of the sample had secondary or higher education and about 18% of the sample was low SES. As anticipated, there were no statistical differences across the four study arms for sex, age, education, body mass index (BMI), SES, shopping responsibility, current diagnosis of chronic disease or nutrition knowledge. There were significant differences between study arms in participants' responses to the frequency of reading labels and the frequency with which labels influenced purchase decisions. Since these questions were asked after the product-related tasks, it is likely that the sequencing of the questions may have resulted in the differences seen, 858

Food Research International 121 (2019) 854–861

N. Khandpur et al.

Table 3 Study outcomes across arms reported as means and standard deviations. Indicators

Understanding of nutrient content Identifying nutrients in excess 0–100 Identifying nutrients not in excess 0–100 Product pair with higher nutrient content 0–100 Healthier product pair 0–100

Study arms Control

Triangular WL ‘A lot of’

Triangular WL ‘High in’

Chilean WL ‘High in’

F-statistic p-Value

27.36 (20.86)a

35.67 (28.07)a

39.05 (29.54)a

35.51 (27.33)a

31.29 (25.16)

28.05 (23.36)

28.43 (23.70)

28.59 (25.37)

21.01 < 0.001 2.23 0.08 3.73 0.01 7.06 < 0.001

38.55 (23.83)

40.36 (24.43)

a

39.33 (22.93)

68.26 (26.86)a

73.47 (27.30)a

75.01 (26.87)a

71.38 (27.11)

3.40 (1.14)a

3.76 (1.35)a

3.87 (1.41)a

3.82 (1.37)a

a

a

36.12 (20.54)

Nutrient and product perceptions High nutrients 1 none – 7 a lot Low nutrients 1 none – 7 a lot Perceptions of healthfulness 1 not healthy – 7 very healthy

2.61 (1.07)

Purchase intentions Single product 1 certainly not buy – 7 certainly buy Label opinions Change eating behaviour 1 positively – 7 negatively Improve understanding 1 disagree – 7 agree Is helpful 1 disagree – 7 agree Is visible 1 disagree – 7 agree

a

a

2.78 (1.21)

2.86 (1.15)

3.90 (1.34)a

3.54 (1.53)a

3.42 (1.48)a

3.50 (1.53)a

4.61 (1.49)a

4.12 (1.69)a

4.06 (1.73)a

4.15 (1.72)a

13.72 < 0.001

–

1.86 (0.97)

1.95 (1.01)

1.98 (1.08)

–

5.82 (1.35)

5.92 (1.24)

5.96 (1.27)

–

5.65 (1.20)

5.69 (1.16)

5.74 (1.17)

5.83 (1.11)

5.70 (1.17)a

2.16 0.11 1.89 0.15 0.76 0.46 4.03 0.01

–

5.88 (1.07)

a

2.82 (1.14)

15.97 < 0.001 5.59 < 0.001 12.31 < 0.001

Responses with the same superscript are significantly different from each other (p-value < .05) based on Tukey HSD post-hoc tests.

3.4. Purchase intentions

Significant differences were also found among participants in the Muito arm compared to the control. There were no differences between control and the Chile WL (Table 3). Analysis using the indicator that combined scores for nutrients in excess and healthier product identification, reflected a similar trend – Alto had the highest cumulative scores followed by Muito, Chile and the control group, with significantly different comparisons between all label arms and the control (p-value < .01 for all comparisons) and between Alto and Chile (p-value = .04).

There were significant differences in the mean responses across study arms for the purchase intentions of single products. Participants in the Alto arm had the lowest responses indicating a lower desire to purchase the product while those in the control condition had the highest mean responses. Post-hoc tests demonstrated that the control was significantly different from all label arms, but no significant differences were seen between label arms (Table 3). For product pairs, there were no significant differences in purchase intentions across any study arm (p-value = .55) (Fig. 3).

3.3. Nutrient and product perceptions 3.5. Label opinions

Perception of nutrient content for nutrients that displayed a WL (Nutrient perceptions - high nutrients) were significantly different between study arms, with Alto having the highest mean perception of nutrient levels and the control the lowest. Participants in the control arm had significantly lower perceptions than all label arms. There were no differences between WLs. The presence of a WL on the product significantly increased participant perceptions of the level of even those nutrients that were not in excess and did not display a WL (Nutrient perceptions - low nutrients). Participants in the Alto and the Chile arms had significantly higher perceptions of nutrient levels compared to the control arm. The control was not different from the Muito arm and no differences were seen between WL arms. Perceptions of healthfulness among participants were significantly different across study arms. All participants in label arms rated products as less healthy, with Alto having the lowest ratings. Post-hoc analysis revealed that responses in the control arm were significantly higher compared to responses across all label arms. There were no differences between label arms for this indictor.

Participants in the Muito, Alto and the Chile WL arms had similar opinions on the effects of all WL designs in improving purchase and eating behavior (Table 3). All WLs were equally helpful, credible, and easy to understand. Compared to the Chile WL, participants in the Muito WL had significantly higher responses on the visibility of the label but there were no differences between Muito and Alto and between Alto and Chile arms. 4. Discussion This study aimed at measuring improvements in consumer understanding, perceptions, and intentions for different WL designs, among a large sample of participants who had likely never been exposed to WLs before. The presence of a WL was a significant improvement over current sources of nutrition information and helped improve participant understanding and perceptions of the nutrient profile of the product. This further validates the benefits of including simple, consistent, non-technical nutrition information on the front of food packages, in 859

Food Research International 121 (2019) 854–861

N. Khandpur et al.

Fig. 3. Purchase intentions, product pair.

making their food choices (Ares, Varela, et al., 2018). The Muito label, in this present study, was significantly more visible to participants than the Chile WL. The Muito WL (like the Alto WL) had a solid white background that served as a contrast to the black triangle. Contrasting backgrounds have been rated favorably in terms of improving noticeability and readability in previous studies (Acton et al., 2018). The only other quantitative evaluations of different WL designs have been published by researchers in Chile and Uruguay. In Chile, 15 different symbols (6 of which were octagon alert symbols, 6 were symbols of a hand symbolizing ‘stop’, among other designs), with different colors (10 red, 6 black) and messages (2 had the phrase ‘Attention/ Atención while 8 had messages about diabetes, in addition to ‘High in/ Alto en’) were evaluated within a sample of 1300 adult women. The two best performing designs were further assessed, and the current Chilean label was found to be most effective in improving comprehension and purchase intentions and had better visibility than a black-and-white symbol of a hand (Corvalán et al., 2013). Uruguayan researchers have directly compared the triangle shape to the octagon in a conjoint experiment (Cabrera et al., 2017). Although the designs were not directly comparable to those in the present study, they found that an octagon WL shape was marginally better at lowering perceptions of product healthfulness than the triangle. In another of their experiments. Cabrera et al., found that the octagon had a quicker reaction time among participants classifying food pictures as healthful or unhealthful, and therefore a stronger implicit association with unhealthfulness than the triangle (Cabrera et al., 2017). The comparisons used in the Uruguayan study did not have the additional features of a contrasting white background, or the Ministry of Health endorsement as in this study, and solely tested shape effects, which might explain the differences seen between this study and theirs. They did, however, find that the phrase ‘High in/ Alto en’ had the lowest perceptions of healthfulness, much like in this present study. In sum, the results of this study are consistent with study hypothesis and suggest the benefits of certain design elements over others. The triangular symbol, the contrasting white background and the signal word ‘High in/ Alto em’ consistently performed better than the octagonal symbol and the signal word ‘A lot of/ Muito’ in helping participants identify the nutrients in excess, in identifying the healthier product in a comparison task, and in increasing their perceptions of the general nutrient levels in the product. The white contrasting background also served to make the triangular WL more visible.

conjunction to nutrition facts panels and lists of ingredients. Among the different types of WLs compared in the study, the Alto WL consistently performed better across all indicators however, differences were small in some cases and not all comparisons reached statistical significance. The results suggest that Alto WL was significantly better than the control at identifying which of the two products contained an excess of nutrients and the Alto WL seemed particularly helpful at communicating this information to men, a subgroup that has historically reported a low use of nutrition labels (Guthrie, Fox, Cleveland, & Welsh, 1995; Kiszko, Martinez, Abrams, & Elbel, 2014). There was also some evidence that the Chilean label appealed to women. Since the signal words are the same on both labels, this may be attributed to a difference in preference for the symbol. It may also indicate a greater preference for the white background among men or the combined elements of the Alto design in capturing their attention. These sub-group differences may need to be replicated and better explored in future studies, since they were only captured for understanding of nutrient content and not confirmed across other indicators. The Alto WL followed by the Muito WL was also better at indicating that fewer WLs, and therefore fewer nutrients in excess, indicated a healthier product. All WL designs effectively highlighted that certain nutrients were in excess. In general, the results suggest that the triangle with ‘High in/ Alto em’ communicated excess nutrient levels and alerted participants to their relative unhealthfulness most effectively. For nutrients that displayed a WL, all labels increased participant perceptions of elevated content. The Alto and Chile labels led to perceptions of high levels of even those nutrients that were not in excess and did not require a WL. The Alto and Chile labels may increase perceptions of the overall unhealthfulness of these products, reinforcing the message communicated for individual nutrients. WLs also significantly decreased participant intention to purchase the products compared to just the list of ingredients and the nutrition facts panels. While the Alto arm elicited the lowest intention to purchase, differences were not significant between WLs. In terms of shifts in purchase, this study was unable to demonstrate differences between label arms and the control, suggesting that WLs may serve to better inform consumers but may not necessarily change their purchase decisions. Changes in purchase decisions will need to be tested in real world settings. Participants in this study had high evaluations on the usefulness, ease of comprehension and the credibility of the information for all WL designs. WLs clearly had participant support and were deemed helpful health promotion tools. Similar results have been reported in Uruguayan samples where only the Chile WL was evaluated (Ares et al., 2018). Participants in this study had positive opinions on the label which they regarded as easy to understand and identify. They were also reportedly willing to take the nutritional warnings into account when

5. Study limitations and strengths In an attempt to compare the effects of the WL against the nutrition information currently available to the Brazilian consumer, images of 860

Food Research International 121 (2019) 854–861

N. Khandpur et al.

the list of ingredients and the nutrition facts panels were included, along with the front panel of the products. This resulted in 2–3 panel images per product. Images therefore had to be appropriately adjusted to fit the average computer screen and participants were provided with a zoom tool to focus in on all information. This unfortunately may have resulted in fewer participants being able to see all WLs in detail at first glance and meant more effort on the part of the participant in seeking out this information. The start of the survey did orient participants on how to use the zoom feature, but it is unlikely that all participants did so for all products. This may have compromised the power of the study to some extent and resulted in lower scores as seen in Table 3. Having said that, there were still clear trends and some differences between study arms, although not all of them reached significance. The question that asked to identifying nutrients that were not in excess was perhaps too confusing or complicated for the online audience, which may have also been a reason for the low scores. This study has a number of strengths. It is one of few randomized controlled trials to capture design effects of different WLs, across a variety of products, in a large sample of Brazilian adults. Given that all participants saw the WLs for the very first time with no orientation on their use, interpretation, or function, the results highlight the health promotion potential of WLs, their intuitive interpretability and their positive influence on consumer understanding, perceptions, and opinions.

Ares, G., Aschemann-Witzel, J., Curutchet, M. R., Antúnez, L., Moratorio, X., & Bove, I. (2018 Aug). (2018b). A citizen perspective on nutritional warnings as front-of-pack labels: Insights for the design of accompanying policy measures. Public Health Nutrition, 29, 1–12. Ares, G., Varela, F., Machin, L., Antúnez, L., Giménez, A., Curutchet, M. R., & AschemannWitzel, J. (2018). Comparative performance of three interpretative front-of-pack nutrition labelling schemes: Insights for policy making. Food Quality and Preference, 68, 215–225. Arrúa, A., Curutchet, M. R., Rey, N., Barreto, P., Golovchenko, N., Sellanes, A., ... Ares, G. (2017). Impact of front-of-pack nutrition information and label design on children's choice of two snack foods: Comparison of warnings and the traffic-light system. Appetite, 116, 139–146. Arrúa, A., Machín, L., Curutchet, M. R., Martínez, J., Antúnez, L., Alcaire, F., ... Ares, G. (2017). Warnings as a directive front-of-pack nutrition labelling scheme: Comparison with the Guideline Daily Amount and traffic-light systems. Public Health Nutrition, 20(13), 2308–2317. Bialkova, S., & van Trijp, H. (2010). What determines consumer attention to nutrition labels? Food Quality and Preference, 21(8), 1042–1051. Boon, C. S., Lichtenstein, A. H., & Wartella, E. A. (Eds.). (2010). Front-of-package nutrition rating systems and symbols: Phase I report. National Academies Press. Buckley, P., & Shepherd, R. (1993). Ergonomic factors: The clarity of food labels. British Food Journal, 95(8), 18–21. Cabrera, M., Machín, L., Arrúa, A., Antúnez, L., Curutchet, M. R., Giménez, A., & Ares, G. (2017). Nutrition warnings as front-of-pack labels: Influence of design features on healthfulness perception and attentional capture. Public Health Nutrition, 20(18), 3360–3371. Corvalán, C., Reyes, M., Garmendia, M. L., & Uauy, R. (2013). Structural responses to the obesity and non-communicable diseases epidemic: The Chilean Law of Food Labeling and Advertising. Obesity Reviews, 14, 79–87. Cowburn, G., & Stockley, L. (2005). Consumer understanding and use of nutrition labelling: A systematic review. Public Health Nutrition, 8(1), 21–28. Edworthy, J. (Ed.). (1996). Warning design: A research prospective. CRC Press. Guthrie, J. F., Fox, J. J., Cleveland, L. E., & Welsh, S. (1995). Who uses nutrition labeling, and what effects does label use have on diet quality? Journal of Nutrition Education, 27(4), 163–172. ISO 3864 (2011). Graphical symbols - Safety colours and safety signs - Part 1: Design principles for safety signs and safety markings. Izar, M. C. O., & Fonseca, F. A. H. (2007). Como diagnosticar e tratar dislipidemias. Revista Brasileira de Medicina, 64(12), 140–160. Kanter, R., Vanderlee, L., & Vandevijvere, S. (2018). Front-of-package nutrition labelling policy: Global progress and future directions. Public Health Nutrition, 21(8), 1399–1408. Khandpur, N., de Morais Sato, P., Mais, L. A., Bortoletto Martins, A. P., Spinillo, C. G., Garcia, M. T., ... Jaime, P. C. (2018). Are front-of-package warning labels more effective at communicating nutrition information than traffic-light labels? A randomized controlled experiment in a Brazilian sample. Nutrients, 10(6), 688. Khandpur, N., Swinburn, B., & Monteiro, C. A. (2018). Nutrient-based warning labels may help in the pursuit of healthy diets. Obesity, 26(11), 1670–1671. Kiszko, K. M., Martinez, O. D., Abrams, C., & Elbel, B. (2014). The influence of calorie labeling on food orders and consumption: A review of the literature. Journal of Community Health, 39(6), 1248–1269. Koffka, K. (2013). Principles of Gestalt psychology. Routledge. Ministerio de Salud, Chile (2015). Decreto número 13, de 2015. Santiago: Ministerio de Salud. http://www.minsal.cl/wp-content/uploads/2017/05/Informe-evaluaci%C3% B3n-implementaci%C3%B3n-Ley-20606-Enero-2017.pdf. Ministry of Health, Peru. https://busquedas.elperuano.pe/normaslegales/decretosupremo-que-aprueba-el-reglamento-del-decreto-legisl-decreto-supremo-n-015-2017produce-1574550-4/ (Accessed August 1st, 2018). Ministry of Health, Uruguay. https://medios.presidencia.gub.uy/legal/2018/decretos/08/ cons_min_705.pdf (Accessed August 31st, 2018). Monteiro, C. A., Cannon, G., Moubarac, J. C., Levy, R. B., Louzada, M. L. C., & Jaime, P. C. (2018). The UN Decade of Nutrition, the NOVA food classification and the trouble with ultra-processing. Public Health Nutrition, 21(1), 5–17. Pan American Health Organization. Nutrient profile model. (2016). http://iris.paho.org/ xmlui/bitstream/handle/123456789/18621/9789275118733_eng.pdf (Accessed August 2018). Strasser, A. A., Tang, K. Z., Romer, D., Jepson, C., & Cappella, J. N. (2012). Graphic warning labels in cigarette advertisements: Recall and viewing patterns. American Journal of Preventive Medicine, 43(1), 41–47. Vyth, E. L., Steenhuis, I. H., Roodenburg, A. J., Brug, J., & Seidell, J. C. (2010). Front-ofpack nutrition label stimulates healthier product development: A quantitative analysis. International Journal of Behavioral Nutrition and Physical Activity, 7(1), 65. Wogalter, M. S. (1994). Factors influencing the effectiveness of warnings. Wogalter, M. S., DeJoy, D., & Laughery, K. R. (Eds.). (2005). Warnings and risk communication. CRC Press. Wogalter, M. S., Silver, N. C., Leonard, S. D., & Zaikina, H. (2006). Warning symbols. Handbook of warnings. 159–176. World Health Organization. Global action plan for the prevention and control of noncommunicable diseases 2013–2020. (2013). http://africahealthforum.afro.who.int/ IMG/pdf/global_action_plan_for_the_prevention_and_control_of_ncds_2013-2020.pdf (Accessed August 2018).

6. Conclusion This is one of the few studies to empirically examine the influence of different WL designs in improving consumer understanding and perceptions. WLs were superior to the list of ingredients and the nutrition facts panel in improving participant understanding and perceptions of the nutrient profile of products. Results suggest that a black triangular symbol against a white background with the expression ‘High in/ Alto em’ are important design elements to consider when developing WLs as they serve to alert the consumer, enhance visibility of WLs, and make it more likely that consumers will read the information presented. The Alto WL is likely to have marginally greater influence as an information and health promotion tool than other WL designs for a Brazilian population. Regulatory agencies may want to consider including these design elements to improve label effectiveness as they formulate policies on FOP labelling. Funding The data collection for this study was financed by the Bloomberg Philanthropies (BRAZIL-IO-05). The funding body had no role to play in the design of the study, data collection, analysis, and interpretation or in writing the manuscript. Dr. Khandpur was supported by FAPESP fellowship 2016/13669-4, while Dr. Sato was supported by FAPESP fellowship 2017/05651-0. Declarations of interest None. References Acton, R. B., Vanderlee, L., Roberto, C. A., & Hammond, D. (2018). Consumer perceptions of specific design characteristics for front-of-package nutrition labels. Health Education Research, 33(2), 167–174. ANVISA (2018). http://portal.anvisa.gov.br/documents/33880/2977862/An%C3% A1lise+de+Impacto+Regulat%C3%B3rio+sobre+Rotulagem+Nutricional_vers %C3%A3o+final+3.pdf/2c094688-aeee-441d-a7f1-218336995337, Accessed date: 21 July 2018.

861