Research Article

Consumer Perception of the Healthfulness ofUltra-processed Products Featuring DifferentFront-of-Pack Nutrition Labeling Schemes Leandro Mach�ın, BSc1; Manuel Cabrera, BSc1; Mar�ıa Rosa Curutchet, MSc2;Joseline Mart�ınez, BSc2; Ana Gim�enez, BSc1,3; Gast�on Ares, PhD1,3

1CentroRep�ubli2Instituto3SensomQu�micaConflict owith thiAddressultad deelones 91�2016 Sreservedhttp://dx

330

ABSTRACT

Objective: To examine the influence of front-of-pack nutrition information on the perception of health-fulness of ultra-processed products across 2 income levels.Design: A between-participants design was used to compare healthfulness perception of ultra-processedproducts featuring different front-of-pack nutrition information schemes (guideline daily amount system,traffic light system, and monochromatic traffic light system).Participants: A total of 300 people (aged 18–70 years, 75% female) fromMontevideo, Uruguay, partic-ipated in the study; half were middle- or high-income people and the other half were low-income people.Main Outcome Measures: Participants were shown the labels of each product and asked to rate theirperceived healthfulness and the frequency with which each product should be consumed.Analysis: Results were analyzed using analysis of variance for statistical significance (P < .05).Results: Low-income participants perceived ultra-processed products to be significantly (P < .05) morehealthful than did middle- and high-income participants. The lowest perceived healthfulness scoresfor low-income participants were obtained for products featuring the colored and monochromatic trafficlight system whereas no significant differences (P> .05) among schemes were found for middle- and high-income participants.Conclusions and Implications: Nutrition education programs aimed at increasing low-income people’sknowledge of the nutritional composition of these products and their potential negative effects on healthseem to be necessary. Although the inclusion of semidirective front-of-pack nutrition informationdecreased the perceived healthfulness of low-income people, it seemed unlikely to influence how theyperceive these products.Key Words: nutrition labeling, income, guideline daily amounts, traffic-light system, consumers, pro-cessed foods, front-of-pack (J Nutr Educ Behav. 2017;49:330-338.)

Accepted December 12, 2016. Published online February 6, 2017.

INTRODUCTION

Dietary patterns are strongly influ-enced by the environment.1 In mostindustrialized countries the food envi-ronment is characterized by the largeavailability and affordability of foodthat promotes positive energy balance

de Investigaci�on B�asica en Psicolog�a,ca, Montevideo, UruguayNacional de Alimentaci�on, Montevid

etr�a y Ciencia del Consumidor, Institu, Universidad de la Rep�ublica, Canelonf Interest Disclosure: The authors’ conflis article on www.jneb.org.for correspondence: Gast�on Ares, PhDQu�mica, Universidad de la Rep�ublica,000, Uruguay; Phone: þ(59)82292202ociety for Nutrition Education and Beh..doi.org/10.1016/j.jneb.2016.12.003

in their inhabitants.2 In particular,the increasing availability and con-sumption of ultra-processed productshas been linked to obesity and othernoncommunicable diseases.3-7

The PanAmericanHealthOrganiza-tion defined ultra-processed productsas ‘‘industrial formulations manufac-

Facultad de Psicolog�a, Universidad de la

eo, Uruguayto Polo Tecnol�ogico de Pando, Facultad dees, Uruguayct of interest disclosures can be found online

, Instituto Polo Tecnol�ogico de Pando, Fac-By Pass de Rutas 8 y 101 s/n, Pando, Can-1; E-mail: gares@fq.edu.uyavior. Published by Elsevier, Inc. All rights

Journal of Nutrition Education and Beh

tured from substances derived fromfoodsor synthesized fromorganic sour-ces.’’8 Ultra-processed products typi-cally include flavorings, colorings,sweeteners, emulsifiers, andother addi-tives in their formulation and containhigh sugar, salt, sodium, and low-quality fat content.9 These productsare often marketed as healthful andinclude references to naturalness andhealth/nutrition claims on their pack-ages and in advertising campaigns,which can create a positive healthfulimage in consumers' minds.8 For thisreason, the Pan AmericanHealthOrga-nization recently claimed the need todevelop strategies aimed at reducingconsumption of ultra-processed prod-ucts through regulation of the prod-ucts’ labeling and marketing.8,10

Commercially, the most attractivemarkets for ultra-processed products

avior � Volume 49, Number 4, 2017

Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017 Machı́n et al 331

are no longer the fully industrializedhigh-income countries but developingcountries in Asia, Eastern Europe, andLatin America.8 In particular, ultra-processed products have becomedominant in the food system of LatinAmerican countries and contribute toa significant proportion of daily energyintake.7-10 Uruguay has shown thefastest growth rate in sales of ultra-processed products in Latin America:68.4% between 2000 and 2013.8 Thissituation makes it necessary to studyhow consumers perceive these prod-ucts and to evaluate the potentialimpact of different public policies toreduce their consumption.

According to theories on persuasivedesign, behavior depends on 3 factors:motivation, ability, and triggers.11 Onestrategy that can be used to encouragepeople to reduce their consumption ofultra-processed products is to changetheir motivation to consume theseproducts by modifying the way inwhich theyperceive them. In this sense,the inclusion of semidirective or direc-tive front-of-pack (FOP) nutrition infor-mation, which highlights the highcontent of nutrients that have beenlinked with negative health condi-tions,12,13 can increase consumerawareness of the nutritional profile ofultra-processedproductsandtentativelyencourage them to replace these prod-ucts with healthier options.14

This FOP nutrition information iseasier to find and understand than theconventional nutrition labels used inmost countries, includingUruguay.15-18

Although different formats of FOPnutrition labels have been developed,several studies showed that the trafficlight system (TFL) provides the best resultsunder experimental conditions.12,19-21

This system uses the traffic light colorcode to classify the content of keynutrients (sugar, fat, saturated fat,and salt) as high, medium, or low.22

Income was identified as 1 of themain sociodemographic variables thatinfluence how people select andperceive food.23 Low-income peoplefrequently describe their food choicesas primarily driven by economic fac-tors, and report that nutrition informa-tion does not have a relevant role inshaping the food choices of this popu-lation.24-26 Low-income people haveless nutrition knowledge and find itdifficult to understand or use nutrition

labels for food selection comparedwithmiddle- or high-income people.27-29

However, research showed that low-income people would appreciate hav-ing simplified or easy to read andunderstand FOP nutrition informationformats such as the TFL.24,26,30,31

In this context, the aim of the cur-rent work was to examine the influ-ence of FOP nutrition informationon the perception of healthfulness ofultra-processed foods across 2 incomelevels.

MATERIALS ANDMETHODSParticipants

A total of 300 people (aged 18–70 years;75% females) from Montevideo,Uruguay, participated in the study.Half of the participants (150) were mid-dle/high income people who were re-cruited from a consumer databasecontaining 2,500 profiles maintainedby the research group that authoredthestudy.Onlyparticipantswhohadre-ported a daily income > $20 US, whichcorresponds to middle/high income inUruguay,32 were invited via e-mail toparticipate in the study. The other halfof participants (150) was composed oflow-income people and was recruitedamong beneficiaries of the nationalfood stamp program administered byInstituto Nacional de Alimentaci�on,Uruguay. Beneficiaries of this programshould have a daily income < $3 US,which can be classified as low-incomein Uruguay.32 Low-income participantswere recruited at the offices of InstitutoNacional de Alimentaci�on while theyasked for assistance related to the foodstamp program. All participants werein charge of food purchase at theirhomes at least occasionally. Participantssigned an informed consent agreementand received a small gift for participa-tion equivalent to $5 US. The studywas approved by the Ethics Committeeof the School of Chemistry of Universi-dad de la Rep�ublica, Uruguay.

Experimental DesignFront-of-pack nutrition labelingschemes. Three FOP nutrition label-ing schemeswere considered: guidelinedaily amounts (GDA),33 a TFL,22 and amonochromatic TFL (MTFL).34 Textand color interpretational aids were

considered in the TFL and MTFL.High, medium, and low nutrient con-tent were coded with red, yellow, andgreen in the TFL, whereas differentshades of black were used in the MTFLto classify nutrients ashigh (black),me-dium (gray), and low (white). The lattersystemwas considered based on the as-sociation of black color with the wordsfataland poison.35-37 Inaddition, resultsfrom a preliminary study in which 100participants rated the healthfulness ofdifferent colors showed that black wasassociated with unhealthy more thanwas red.38

The 3 schemes contained theamount of sugar, total fat, and sodiumper portion. In the TFL and MTFL theamount of each nutrient per portionwas classified into high, medium, andlow following the guidelines providedby the Food Standards Agency.22

Ultra-processed products. Twelve ultra-processed products were considered inthe study: cereal bars, yogurt, instantsoup,processedcheese,panbread,applejuice, soy juice, ham, gelatin, crackers,breakfast cereals, and diet soda. Accord-ing to their ingredients, allof theseprod-ucts can be classified as ultra-processedfollowing the definition proposed bythe Pan American Health Organiza-tion.8 The selected products were re-ported to be consumed frequently inthe country.39,40 The ingredients andnutritional composition of the productswere selected based on the character-istics of the products available in themarket (Table 1). Except for diet soda,all productshadahighcontentof at least1 target nutrient (sugar, total fat, and/orsodium).

The products were presented to thestudy participants using package im-ages. Nutrition information wasincluded on the images using the FOPlabeling schemes described in the pre-vious section. The Figure shows exam-ples of the images included in thestudy, as presented to participants.

Measures

Participants were randomly allocatedto 1 of 3 groups, 1 for each FOP nutri-tion labeling scheme (GDA, TFL, andMTFL). Therefore, each participantevaluated products featuring 1 of theFOP schemes.

Table 1. Nutritional Composition and Ingredients of Ultra-processed Productsa

Product IngredientsbPortionSize

TotalSugar (g/Portion)

TotalFat (g/Portion)

Sodium(mg/

Portion)

Apple juice Pure filtered water, concentrated apple juice, sugar,vitamin C, calcium citrate, potassium phosphate

250 mL 15 0 52

Breakfastcereals

Milled corn, sugar, malt, flavoring, high-fructose corn syrup,salt, vitamins, iron

200 mL 26 2 69

Cereal bar Glucose syrup, whole grains, sugar, wheat gluten, honey,puffed triticale, minerals, oat fiber, salt, molasses,flavorings, barley malt extract, vitamins

30 g 13 1.5 131

Crackers Flour, soybean oil, sugar, partially hydrogenated cottonseedoil, salt, leavening, high-fructose corn syrup, soy lecithin,malted barley flour, natural flavor

30 g 2 9 251

Diet soda Carbonated water, colorings, phosphoric acid, citric acid,preservative, sweeteners, natural flavorings, caffeine

200 mL 0 0 20

Gelatin Gelatin, fumaric acid, sodium citrate, flavorings, salt,disodium phosphate, sweeteners, colorings

100 g 12.5 0 89

Ham Pork, water, sugar, salt, sodium lactate, corn syrup,modified starch, dextrose, sodium phosphate, sodiumdiacetate, sodium erythorbate, sodium nitrate, flavorings

30 g 0.6 3 1,000

Instant soup Vegetable powders, potato starch, maltodextrin, vegetableoil, dried vegetables, yeast extract, salt, flavorings,lactose, milk protein, colorings, wheat flour, whey powder

250 mL 0.9 0.5 700

Pan bread Flour, water, sugar, beef tallow, yeast, salt, emulsifiers,preservatives, flour improvers

54 g 2 1.5 254

Processedcheese

Milk, whey, milk protein concentrate, milk fat, salt, sodiumcitrate, whey protein concentrate, sodium phosphate,sorbic acid, cheese cultures, colorings

28 g 0.3 8.1 433

Soy juice Soymilk (filtered water, whole soybeans), sugar, salt,carrageenan, flavorings, calcium carbonate, vitamins

200 mL 15 0.6 34

Yogurt Pasteurized milk, live active yogurt cultures, sugar, modifiedcorn starch, high-fructose corn syrup, gelatin, flavorings,colorings

30 g 17 0 150

aUltra-processed products are defined as ‘‘industrial formulations manufactured from substances derived from foods or syn-thesized from organic sources8; bThe ingredients were not presented to participants and are listed only to clarify the classifi-cation of products as ultra-processed.

332 Machı́n et al Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017

Participants were asked to imaginethat they were purchasing foods at thesupermarket. Then the 12 productswere presented 1 at a time, followingan experimental design balanced forpresentationandcarry-over effects (Wil-liams Latin square).41 For each product,participants were asked to indicate howhealthful it was, using a 10-point scale(1¼ not healthful; 10¼ very healthful)and the frequency with which partici-pants should consume it, using a 5-point scale (1 ¼ I should avoid it;2 ¼ occasionally; 3 ¼ several times amonth; 4 ¼ several times a week;5¼ several times aday).After evaluatingthe products, participants completed ashort sociodemographic questionnaire.The experimental procedure was pre-testedwith 5 low-income and 5middle-

to high-income participants to checkunderstanding. No changes to theexperimental procedure were necessaryto improve the understandability ofthe questionnaire.

Data Analysis

ANOVAwas used to analyze healthful-ness and recommended consumptionfrequency scores. Labeling scheme,product, income, and their interactionswere considered fixed sources of varia-tion. When differences were signifi-cant, Tukey's test was used for post hoccomparison of means. Data were alsoanalyzed separately for low- and mid-dle- to high-income participants,considering product, labeling scheme,

and their interaction as fixed sourcesand variation. A 5% significance levelwas considered (P < .05). All analyseswere performed using R software (RFoundation for Statistical Computing,Vienna, Austria).

RESULTS

The distribution of age, gender, andeducational level of participants isshown in Table 2. There were no signif-icant differences in age and genderdistribution of low-income andmiddle-or high-income participants (P ¼ .20and .66, respectively), whereas educa-tional level was significantly affectedby income (P < .001). In addition, nosignificant differences were noted in

Figure. Three sample products considered in the study with different front-of-pack nutrition information schemes. The (A) guidelinedaily amounts system for corn flakes, translated from Spanish, reads: ‘‘Each portion (30 g) contains: 17 g sugars, 0 g total fat, and150 mg sodium.’’ The (B) traffic light system for Toni yogurt with peach pieces translates as: ‘‘Each portion (200 mL) contains: 26 gsugars (with the red indicating that this is a high amount), 2 g total fat (with the green indicating that this is a low amount), and 69 mgof sodium (low).’’ The, (C) monochromatic traffic light system for Fundo Los Alerces cheddar cheese translates as ‘‘Each portion(28 g) contains: 0.3 g sugar (with the white indicating that this is a low amount), 8.1 g total fat (with the black indicating this is ahigh amount), and 433 mg sodium (high).’’

Table 2. Sociodemographic Characteristics (%) of Participants

SociodemographicCharacteristic

Total(n ¼ 300)

Low Incomea

(n ¼ 150)Middle/High Incomeb

(n ¼ 150)

GenderFemale 75 78 71Male 25 22 29

Age, y18–30 39 37 4131–45 23 25 2146–70 38 38 38

Educational levelComplete/incompleteprimary school

13 26 0

Incomplete secondaryschool

18 33 2

Complete secondaryschool

25 32 18

Technical education 9 9 10University degree 35 0 70

aParticipants were recruited from the consumer database of the research groupthat authored the study and were people who had a daily income > $20 US; bPar-ticipants were recruited among beneficiaries of 1 of the national food stamp pro-grams administered by Instituto Nacional de Alimentaci�on, Uruguay, and had adaily income < $3 US.

Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017 Machı́n et al 333

the gender, age, and educational leveldistribution of the 3 groups of partici-pants who evaluated products featuringdifferent FOP nutrition labelingschemes for both low-income (P > .49)and middle- and high-income partici-pants (P> .58).

Income influenced perception ofhealthfulness and modulated the in-fluence of FOP nutrition informationschemes, as evidenced by the signifi-cant income main effect and the sig-nificant interaction between incomeand FOP scheme (Table 3). Low-income participants rated the health-fulness of all evaluated productssignificantly higher than did middle-and high-income participants regard-less of the FOP scheme included onthe labels. As shown in Table 4, low-income participants regarded all prod-ucts as healthful whereas middle- tohigh-income participants rated themin the unhealthful to moderatelyhealthful part of the scale.

The influence of FOP nutrition in-formation depended on participants’

Table 3. Effect of Evaluated Variables on Perceived Healthfulness andRecommended Consumption Frequency Ratings (n ¼ 300 UruguayanParticipants)

IndependentVariable

Healthfulness(F value)

Recommended ConsumptionFrequency (F value)

Product 27.2* 25.0*FOP scheme 29.6* 7.6*Income 1,410.5* 1,839.0*Product x FOP scheme 1.5, ns 0.9, nsProduct x income 5.8* 4.9*Income x FOP scheme 10.1* 1.1, nsProduct x FOP

scheme* income1.2, ns 2.1*

FOP indicates front-of-package; ns, not significant.*P < .05.Note: F values from ANOVA determined considering the product, FOP scheme, in-come, and their interactions as fixed sources of variation are shown.

334 Machı́n et al Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017

income level. As shown in Table 5, thehealthfulness perception of middle- tohigh-income participants dependedonly on the product andwas not signif-icantly affected by the FOP scheme.Onthe contrary, the FOP schemehad a sig-nificant influence on the healthfulnessperception of low-income participants.For this group, the influence of FOPnutrition schemes depended on theproduct being considered, as indicatedby the significant interaction between

Table 4. Mean (SD) Healthfulness Scores*Low- and Middle- to High-Income

Product

Low Income

GDATraffic Light

SystemM

Diet soda 5.8 (2.5)c,A,B 5.1 (2.4)d,B

Crackers 7.5 (2.5)b,A 6.4 (2.5)c,B

Slicedcheese

7.7 (2.5)b,A 7.5 (2.3)a,b,A

Apple juice 8.0 (2.3)a,b,A 7.9 (2.2)a,A

Pan bread 8.1 (2.2)a,b,A 7.3 (2.2)a,b,B

Ham 8.1 (2.4)a,b,A 6.9 (2.4)b,c,B

Cereal bar 8.2 (2.5)a,b,A 6.9 (2.4)b,c,B

Instantsoup

8.2 (2.3)a,b,A 8.2 (2.4)a,A

Breakfastcereal

8.4 (1.9)a,b,A 7.5 (2.1)a,b,B

Gelatin 8.5 (2.1)a,A 7.6 (2.1)a,b,B

Soy juice 8.5 (2.2)a,A 7.7 (2.1)a,b,A,B

Yogurt 8.5 (2.0)a,A 7.3 (2.2)a,b,B

GDA indicates guideline daily amount.*Perceived healthfulness was evaluated uNotes: Mean values within a column with d(P< .05). Means with different uppercaseTukey’s test (P < .05).

the product and the FOP scheme(Table 5).

For low-income participants, prod-ucts featuring the MTFL system wereperceivedas significantly lesshealthfulthan those featuring the GDA system,except for diet soda and apple juice(Table 4). However, products featuringthe TFL system received significantlylower healthfulness ratings withrespect to those featuring theGDA sys-tem, except for diet soda, sliced cheese,

for Ultra-processed Products Featuring DiffereUruguayan Participants

(n ¼ 150) Midd

onochromatic TrafficLight System GDA

TrafS

6.4 (2.5)a,b,A 3.6 (2.1)d,A 3.6 (6.2 (2.3)b,B 3.8 (2.1)d,A 3.0 (6.7 (2.0)a,b,B 3.7 (2.0)d,A 3.6 (

7.2 (2.3)a,b,A 4.9 (2.2)b,c,A 4.8 (7.2 (2.5)a,b,B 5.2 (2.2)b,A 4.8 (7.0 (2.4)a,b,B 3.8 (2.2)d,A 4.2 (7.0 (2.5)a,b,B 5.3 (2.2)a,b,A 5.4 (7.3 (2.4)a,b,B 4.2 (2.3)c,d,A 4.3 (

6.9 (2.4)a,b,B 6.1 (1.8)a,A 5.6 (

7.6 (2.5)a,B 5.5 (2.4)a,b,A 5.3 (7.1 (2.5)a,b,B 5.5 (2.0)a,b,A 5.0 (7.5 (2.4)a,B 5.5 (2.0)a,b,A 5.1 (

sing a 10-point scale (1 ¼ not healthful; 10 ¼ifferent lowercase superscripts were significsuperscripts within a row for an income level

apple juice, instant soup, and soy juice.In the great majority of products, nosignificant differences were found be-tween the TFL and MTFL systems.

The perceived recommended con-sumption frequency was also signifi-cantly influenced by income (Table 3).As shown in Table 6, low-income par-ticipants rated the recommended con-sumption frequency of all productssignificantly higher than did middle-and high-income participants. Low-income participants thought that theproducts should be eaten morefrequently than several times a month,whereas middle-income participantsthought that these products should beeaten occasionally or several times permonth (Table 6).

The FOP nutrition schemes signifi-cantly influenced the perceived recom-mendedconsumption frequencyof theproducts only for low-income partici-pants (Table 5). In the case oflow-income participants, some of theproducts featuring semidirective FOPnutrition labeling schemes receivedsignificantly lower recommended con-sumption frequency scores. As showninTable6, theperceived recommendedconsumption frequency was signifi-cantly lower when labels included theTFL system, compared with when

nt Front-of-Package Nutrition Labels for

le/High Income (n ¼ 150)

fic Lightystem

Monochromatic TrafficLight System

2.0)d,e,A 3.2 (2.3)d,A

1.8)e,A 3.0 (2.0)d,A

2.0)d,e,A 3.5 (1.8)c,d,A

1.9)a,b,c,A 4.3 (2.0)b,c,A

1.9)a,b,c,A 5.5 (1.6)a,A

1.9)c,d,A 4.3 (1.9)b,c,A

2.0)a,A 4.8 (2.1)a,b,A

2.1)b,c,d,A 4.3 (2.4)b,c,A

1.8)a,A 5.4 (2.1)a,A

2.4)a,A 4.6 (2.2)a,b,A

1.7)a,b,c,A 4.6 (1.9)a,b,A

2.1)a,b,A 4.9 (2.0)a,b,A

very healthful).antly different according to Tukey’s testwere significantly different according to

Table 5. Effect of Evaluated Variables on Recommended Consumption FrequencyRatings From Low-Income and Middle- to High-Income UruguayanParticipants

IndependentVariable

Healthfulness (F Values)

RecommendedConsumption Frequency

(F Values)

Low-Income(n ¼ 150)

Middle/High-Income (n¼ 150)

Low-Income(n ¼ 150)

Middle/High-Income (n¼ 150)

Product 11.27* 32.2* 7.8* 11.72*FOP scheme 5.47* 1.1, ns 22.64* 1.90, nsProduct x FOPscheme

1.82* 1.25, ns 1.43, ns 0.77, ns

FOP indicates front-of-package; ns, not significant.*P < .05.Note: F values from ANOVA determined separately for each income group consid-ering product, FOP scheme, and their interaction as fixed sources of variation areshown.

Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017 Machı́n et al 335

they included the GDA system for dietsoda, crackers, ham, cereal bars, andyogurt. For the rest of the products nosignificant differences were found.The MTFL system had an influence ontheperceived recommended consump-tion frequency similar to that of thecolored system. The monochromaticsystem had a significant influence on

Table 6. Mean (SD) Recommended ConsNutrition Labels for Low- and Mid

Product

Low Income (n

GDATraffic Light

SystemM

Diet soda 3.8 (1.4)c,A 2.9 (1.5)d,B

Crackers 4.4 (1.0)b,A 3.7 (1.3)c,B

Slicedcheese

4.4 (1.0)b,A 4.3 (1.1)a,A,B

Apple juice 4.2 (1.2)a,A 4.2 (1.0)a,b,A

Pan bread 4.5 (1.0)a,b,A 4.3 (0.9)a,A

Ham 4.3 (0.9)b,A 3.8 (1.0)b,c,B

Cereal bar 4.5 (1.0)a,b,A 3.7 (1.3)c,B

Instantsoup

4.3 (1.2)b,A 4.2 (0.9)a,b,A

Breakfastcereal

4.5 (0.9)a,b,A 4.1 (0.9)a,b,A,B

Gelatin 4.5 (0.8)a,b,A 4.2 (0.8)a,b,A

Soy juice 4.3 (1.2)b,A 4.1 (1.2)a,b,A

Yogurt 4.8 (1.0)a,A 4.3 (1.0)a,B

FOP indicates front-of-package; GDA, gu* Perceived recommended consumption fr3 ¼ several times per month; 4 ¼ severalNotes: Mean values within a column with d(P< .05). Means with different uppercaseTukey’s test (P < .05).

the perceived recommended consump-tion frequency of diet soda, crackers,sliced cheese, ham, breakfast cereals,and yogurt (Table 6).

DISCUSSION

Results evidenced large differences be-tween low- and middle- to high-

umption Frequency Scores* for Ultra-procesdle- to High-Income Uruguayans

¼ 150) Middle

onochromatic TrafficLight System GDA

Tra

3.0 (1.5)c,B 2.2 (1.2)e,A 2.53.6 (1.5)b,B 2.1 (1.0)e,A 1.93.8 (1.3)a,b,B 2.5 (1.4)b,c,d,e,A 2.4

4.2 (1.1)a,A 2.8 (1.4)a,b,c,A 2.94.2 (1.0)a,A 3.2 (1.3)a,A 3.23.9 (1.2)a,b,B 2.2 (1.3)d,e,A 2.74.1 (1.3)a,b,A,B 2.7 (1.2)a,b,c,d,A 3.04.0 (1.2)a,b,A 2.2 (1.3)c,d,e,A 2.6

4.0 (1.2)a,b,B 3.2 (1.0)a,A 3.2

4.2 (0.8)a,A 2.8 (1.4)a,b,A 3.14.2 (1.2)a,A 3.1 (1.2)a,A 2.84.3 (1.2)a,B 3.1 (1.3)a,A 3.0

ideline daily amount.equency was evaluated using a 5-point scaltimes per week; 5 ¼ several times per day).ifferent lowercase superscripts were significsuperscripts within a row for an income level

income people regarding how theyperceived the healthfulness of ultra-processed products. Low-income partic-ipants perceived most ultra-processedproducts considered in the study to behealthful and appropriate for frequentconsumption. These results were inagreement with previous studies thatsuggested that low-income people usu-ally have low nutrition knowledge andtend to select low-cost products, whichare usually calorie-dense and sati-ating.23-28,30,42-45 On the contrary,middle- to high-income participantswere more critical about the healthful-ness of ultra-processed products. Thissuggests that the low-income popula-tion with low educational levels maybe more vulnerable to marketing strate-giesusedby foodcompanies topromoteultra-processed products, which oftencreate the false impression that theseproducts are healthful by includinghealth claims and pictures of naturalfoodsonpackages andpromotionalma-terial, and/or on advertising.8

Simplified strategies to classify prod-ucts as healthful or unhealthful werereported to be frequent among low-incomemothers inUruguay,whichsug-gests that semidirective or directive FOP

sed Products Featuring Different FOP

/High Income (n ¼ 150)

ffic LightSystem

Monochromatic TrafficLight System

(1.5)d,e,A 2.2 (1.5)f,A

(1.1)f,A 2.3 (1.1)e,f,A

(1.3)e,A 2.5 (1.4)d,e,f,A

(1.2)a,b,c,d,e,A 2.7 (1.3)c,d,e,f,A

(1.2)a,b,A 3.4 (1.1)a,b,A

(1.3)b,c,d,e,A 2.7 (1.2)c,d,e,A

(1.2)a,b,c,d,A 3.0 (1.3)a,b,c,d,A

(1.3)c,d,e,A 2.7 (1.2)c,d,e,A

(1.1)a,A 3.4 (1.1)a,A

(1.2)a,b,A 2.8 (1.3)b,c,d,e,A

(1.3)a,b,c,d,e,A 3.0 (1.3)a,b,c,d,A

(1.1)a,b,c,A 3.3 (1.3)a,b,c,A

e (1¼ I should avoid it; 2¼ occasionally;

antly different according to Tukey’s testwere significantly different according to

336 Machı́n et al Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017

nutrition labelsonultra-processedprod-ucts can modify their perception ofhealthfulness.26 In the current work,ultra-processed products featuring 2 for-mats of the TFL received significantlylower healthfulness ratings than didthose featuring the GDA system forlow-income participants. This resultsuggests that semidirective FOP nutri-tion labeling encouraged them to thinkthat the products were less appropriatefor frequent consumption, which is inagreement with previous studies thatshowed that low-income people werepositive about simplified nutrition in-formation schemes.30,31 In this sense, arecent qualitative study conducted inUruguay showed that the TFL was thepreferred system for low-incomemothers when selecting food for theirchildren.26Onthecontrary, theTFLsys-tem did not significantly modify thehealthfulness perception of middle- tohigh-income participants, who werealready critical of the ultra-processedproducts considered in the currentstudy.

Regarding the comparison of thetraditional TFL system and the mono-chromatic alternative, the researchersfound no large differences. However,in 2 of the 11 products that containeda high content of at least 1 nutrient,the MTFL system appeared to decreasethe healthfulness and perceived rec-ommended consumption frequencyratings of low-income participantscompared with those for the coloredsystem. This might be because thecolored TFL highlights nutrients withlow content using a green color,whichwas reported to convey health associa-tions and may increase the perceptionof healthfulness.35-38 In contrast, themonochromatic scheme used white, aneutral color, to indicate low nutrientcontent, and black to indicate highnutrient content, which conveysnegative associations.35-37 Therefore,although red is associated withdanger, monochromatic systemsmay be more efficient in conveyingunhealthful associations in productswith a low content of at least 1 keynutrient. Black is used in the warningsigns recently introduced in Chile toflag products high in calories,saturated fat, sugar, and/or sodium.46

The limitations of the current workshould be acknowledged. Participantswere not representative of the Uru-guayan population. Instead, they were

recruited to represent specific segmentsof the population with different in-come levels, which limits the generaliz-ability of the findings. In addition, alimited number of labels of ultra-processed products were includedinstead of a wide range of labels foreach type of product.

IMPLICATIONS FORRESEARCH ANDPRACTICE

Public policies aimed at modifying thefood environment are necessary toencourage healthier food habits andcope with the global obesitypandemic.1 The Pan American HealthOrganization has acknowledged theneed to develop strategies to regulatethe marketing and commercializationof ultra-processed products and toencourage consumers to substitutethese products with culinary prepara-tions made from unprocessed or mini-mally processed foods.8 Results fromthe currentwork suggest that these stra-tegies should consider the characteris-tics of specific target populations. Theperception of the healthfulness ofultra-processed products was stronglyinfluenced by income, which suggeststhat specific strategies should be imple-mented for low-income people. Consi-dering that this consumer segmentperceived ultra-processed products tobe healthful, specific nutrition educa-tionprograms aimed at increasing theirknowledge of the nutritional composi-tion of these products and their poten-tial negative effects on health seem tobe necessary.

In thecurrentwork, theTFLwasableto modify how low-income peopleperceived the healthfulness of ultra-processedproducts, and led to a changein the frequency with which theythought each product should beconsumed. However, further researchshouldbeconducted toexamine the in-fluenceof theTFLonconsumers' abilityto select the most healthful alternativeamong a set of products, as well as onconsumer food choices. In addition,comparison of the TFL system anddirective FOP nutrition labelingschemes, such as the Chilean system,46

remains necessary to guide the devel-opment of public policies aimed atreducing the consumption of ultra-processed foods, in terms of their influ-

ence on consumers’ healthfulnessperception and food choice across in-come levels. Finally, in the currentwork, the color scheme used in theTFL influenced the perception ofhealthfulness, which suggests that theinfluence of color on the efficacy ofsemidirective and directive FOP nutri-tion labels deserves further explora-tion.

ACKNOWLEDGMENTS

The authors are indebted to Comisi�onSectorial de Investigaci�on Cient�ıfica,Universidad de la Rep�ublica, Uruguay,for financial support, and to the staffof Instituto Nacional de Alimentaci�on,Uruguay, for their help with imple-menting the study.

REFERENCES

1. Story M, Kaphingst KM, Robinson-O’Brien R, Glanz K. Creating healthyfood and eating environments: policyand environmental approaches. AnnuRev Pub Health. 2008;29:253-272.

2. Eaton SB, Konner M, Shostak M.Stone agers in the fast lane: chronicdegenerative diseases in evolutionaryperspective. Am J Med. 1998;84:739-749.

3. Popkin BM, Adair LS, Ng SW. Globalnutrition transition and the pandemic ofobesity in developing countries. NutrRev. 2012;70:3-21.

4. Juul F, Hemmingsson E. Trends in con-sumption of ultra-processed foods andobesity in Sweden between 1960 and2010. Pub Health Nutr. 2015;18:3096-3107.

5. Rauber F, Campagnolo PD,Hoffman DJ, Vitolo MR. Consump-tion of ultraprocessed food productsand its effects on children’s lipid pro-files: a longitudinal study. Nutr MetabCardiovasc Dis. 2015;25:116-122.

6. Moubarac J-C,Martins AP, Claro RM,Levy RB, Cannon G, Monteiro CA.Consumption of ultra-processed foodsand likely impact on human health: ev-idence from Canada. Pub Health Nutr.2013;16:2240-2248.

7. Moodie R, Stuckler D, Monteiro C,et al. Profits and pandemics: preven-tion of harmful effects of tobacco,alcohol, and ultraprocessed food anddrink industries. Lancet. 2013;381:670-679.

Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017 Machı́n et al 337

8. Pan American Health Organization.Ultra-processed Food and Drink Productsin Latin America: Trends, Impact onObesity, Policy Implications. Washing-ton, DC: Pan American Health Organi-zation; 2015.

9. Mart�nez Steele E, Baraldi LG, et al. Ul-tra-processed foods and added sugars inthe US diet: evidence from a nationallyrepresentative cross-sectional study.BMJ Open. 2016;6:e009892.

10. Pan American HealthOrganization. PanAmerican Health Organization NutrientProfile Model. Washington, DC: PanAmerican Health Organization; 2016.

11. Fogg B. A behavior model for persua-sive design. http://bjfogg.com/fbm_files/page4_1.pdf. Accessed January 9,2017.

12. Hawley KL, Roberto CA, Bragg MA,Liu PJ, Schwartz MB, Brownell KD.The science on front-of-package foodlabel. Pub Health Nutr. 2012;16:430-439.

13. Hodgkins C, Barnett J, Wasowicz-Kirylo G, et al. Understanding howconsumers categorise nutritional labels:a consumer derived typology forfront-of-pack nutrition labelling.Appe-tite. 2012;59:806-817.

14. Cheftel JC. Food and nutrition labellingin the European Union. Food Chem.2005;93:531-550.

15. Ares G, Gim�enez A, Bruzzone F, et al.Attentional capture and understandingof nutrition labelling: a study based onresponse times. Int J Food Sci Nutr.2012;63:679-688.

16. Feunekes GIJ, Gortemakers IA,Willems AA, Lion R, van denKommer M. Front-of-pack nutritionlabelling: testing effectiveness ofdifferent nutrition labelling formatsfront-of-pack in four European coun-tries. Appetite. 2008;50:57-70.

17. Jones G, Richardson M. An objectiveexamination of consumer perceptionof nutrition information based onhealthiness ratings and eye movements.Pub Health Nutr. 2007;10:238-244.

18. Roberto CA, Shivaram M,Martinez O, Boles C, Harris JL,Brownell KD. The Smart Choicesfront-of-package nutrition label: influ-ence on perceptions and intake ofcereal. Appetite. 2012;58:651-657.

19. Drescher LS, Roosen J, Marette S. Theeffects of traffic light labels and involve-ment on consumer choices for food andfinancial products. Int J Cons Stud. 2014;28:217-227.

20. Kelly B,Hughes C,ChapmanK, Louie J,Dixon H, Crawford J. Consumer testingof the acceptability and effectiveness offront-of-pack food labelling systems forthe Australian grocery market. HealthProm Int. 2009;24:120-129.

21. Borgmeier I, Westenhoefer J. Impact ofdifferent food label formats on healthi-ness evaluation and food choice of con-sumers: a randomised-controlled study.BMC Pub Health. 2009;9:184.

22. Food Standards Agency. Front-of-pack traffic light signpost labelling.Technical guidance. Issue 2. http://tna.europarchive.org/20100111121008/http://www.food.gov.uk/multimedia/pdfs/frontofpackguidance2.pdf. Ac-cessed January 9, 2017.

23. Hough G, Sosa M. Food choice in lowincome populations—A review. FoodQual Prefer. 2015;40:334-342.

24. McArthur L, Chamberlain V,Howard AB. Behaviors, attitudes, andknowledge of low-income consumersregarding nutrition labels. J HealthCare Poor Underserved. 2001;12:415-428.

25. Burns C, Cook K, Mavoa H. Role ofexpendable income and price in foodchoice by low income families. Appe-tite. 2013;71:209-217.

26. Mach�n L, Gim�enez A, Curutchet MR,Mart�nez J, Ares G. Motives underlyingfood choice for children and perceptionof nutritional information among low-incomemothers in aLatinAmericancoun-try. J Nutr Educ Behav. 2016;48:478-485.

27. Morton JF, Guthrie JF. Diet-relatedknowledge, attitudes, and practices oflow-income individuals with children inthe household. Fam Econ Nutr Rev. 1997;10:2-14.

28. Brunt AR, Shafer E. Helping con-sumers with low literacy interpret thenew food label. J Nutr Educ Behav.1997;29:224A-225A.

29. Besler HT, Buyuktuncer Z, Uyar MF.Consumer understanding and use offood and nutrition labeling in Turkey.J Nutr Educ Behav. 2012;44:584-591.

30. Le�on-Fl�andez K, Prieto-Castillo L,Royo-Bordonaba MA. Sem�aforo nu-tricional: conocimiento, percepci�on yutilizaci�on entre los consumidores deMadrid, Espa~na. Rev Esp Nutr HumDiet. 2015;19:97-104.

31. Gregori D, Ballali S, V€ogele C, et al.What is the value given by consumersto nutritional label information? Re-sults from a large investigation in Eu-rope. J Am Coll Nutr. 2015;34:120-125.

32. PNUD Uruguay. Clases Medias enUruguay, Entre la Consolidaci�on y la Vul-nerabilidad. Montevideo, Uruguay:PNUD Uruguay; 2014.

33. Wiseman M. The COMA report: die-tary reference values for food energyand nutrients for the United Kingdom.Br Food J. 1992;94:7-9.

34. Ant�unez L, Gim�enez A, Maiche A,Ares G. Influence of interpretation aidson attentional capture, visual processing,and understanding of front-of-packagenutrition labels. J Nutr Educ Behav.2015;47:292-299.

35. Braun CC, Silver NC. Interaction ofwarning label features: determiningthe contributions of three warningcharacteristics. In: Proceedings of the Hu-man Factors and Ergonomics Society, 39thannual meeting. Santa Monica, CA: Hu-man Factors Society; 1995:984-988.

36. Chapanis A. Hazards associated withthree signal words and four colours onwarning signs. Ergon. 1994;37:265-275.

37. Wogalter MS, Magurno AB,Carter AW, Swindell JA,Vigilante WJ, Daurity JG. Hazard asso-ciations of warning header compo-nents. In: Proceedings of the HumanFactors and Ergonomics Society, 39thannual meeting. Santa Monica, CA: Hu-man Factors Society; 1995:979-983.

38. Cabrera M. La Influencia del Color en laPercepci�on de Saludable de Alimentos.Montevideo, Uruguay: Facultad dePsicolog�a; 2016.

39. Bove MI, Cerruti F. Los Alimentos yBebidas en los Hogares: ¿Un Factor deProtecci�on o de Riesgo para la Saludy el Bienestar de los Uruguayos? En-cuesta Nacional de Gastos e Ingresosde los Hogares 2005-2006. Montevi-deo, Uruguay: Instituto Nacionalde Estad�stica; 2008.

40. Ares G, Curutchet MR, Ferre Z,Gim�enezA,RossiM.Resultados de la En-cuesta sobreAlimentaci�on y Bienestar.Mon-tevideo, Uruguay: Facultad de CienciasSociales, Universidad de la Rep�ublica;2016.

41. MacFie HJH, Bratchell N,Greenhoff K, Vallis LV. Designs tobalance the effect of order of presenta-tion and first-order carry-over effectsin hall tests. J Sens Stud. 1989;4:129-148.

42. Drewnoski A. Obesity and the foodenvironment: dietary energy densityand diet costs. Am J Prev Med. 2004;27:154S-162S.

338 Machı́n et al Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017

43. Drenowski A, Darmon N. The eco-nomics of obesity: dietary energy den-sity and energy cost. Am J Clin Nutr.2005;82:265S-273S.

44. Appelhans BM, Waring ME,Schneider KL, Pagoto SL. Food prepa-

ration supplies predict children’s familymeal and home-prepared dinner con-sumption in low-income households.Appetite. 2014;76:1-8.

45. Cort�es DE, Mill�an-Ferro A,Schneider K, Vega RR, Caballero AE.

Food purchasing selection among low-income, Spanish-speaking Latinos. AmJ Prev Med. 2013;44:S267-S273.

46. Ministerio de Salud.Decreto N�umero 13,de 2015. Santiago de Chile: Ministeriode Salud; 2015.

Journal of Nutrition Education and Behavior � Volume 49, Number 4, 2017 Machı́n et al 338.e1

CONFLICT OF INTEREST

The authors have not stated any con-flict of interest.