2 of 15
46
into consideration cultural and dietary patterns, education level and alignment with national dietary
47
guidance.
48
FOP labels can be roughly divided into two types namely interpretative and non-interpretative
49
labels. Non-interpretative FOP labels provide information on the amount of key nutrients (e.g. fat,
50
sugar and sodium) and percent recommended intakes, with little interpretation of this information
51
(e.g. Guideline Daily Amounts [GDA] or Reference Intake [RI]logos [6]). Interpretative FOP labels
52
contain an interpretation of the nutritional quality of the products, with or without information about
53
the amount of key nutrients. Some FOP labels use colours (e.g. Multi Traffic Light [MTL] [7]) to
54
indicate whether the level of a nutrient is high (red), medium (amber) or low (green). Other
55
interpretive FOP labels provide a simple summary score of a product’s overall nutritional profile (e.g.
56
Keyhole logo [8], Choices Programme Logo [9]), a ranking (e.g. Health Star Rating [10], Nutri-Score
57
[11]) or warn for high levels of certain critical nutrients in products (e.g. Warning labels [12]).
58
Many studies have examined the different attributes of effectiveness of FOP labels, but there are
59
numerous inconsistencies in the results. This can be explained by the fact that the definition of
60
effectiveness differs from study to study. Some studies focus on consumer liking, understanding or
61
preference for FOP labels, others on food choice or actual or intended food purchase and some on
62
food intake. There are also major differences in methodologies used. However, research into the
63
effectiveness of FOP labels on consumer behavior in practice is lacking [13].
64
Studies focusing on consumers’ understanding of FOP labels and product choice generally show
65
that FOP labels appear to help consumers determine which foods are healthier and which are less
66
healthy [14,15]. Simple FOP labels such as MTL, warning labels and Nutri-Score appear to be most
67
effective [15].
68
The MTL label has been implemented in the United Kingdom [7]. It provides information on
69
energy, fat, saturated fat, sugar and salt content per 100 gram and as percentage of reference intake,
70
combined with traffic lights colours (green, amber and red) to highlight low, medium or high levels
71
of the nutrients. It also provides serving size information that is expressed in easily recognizable and
72
meaningful ways to the consumer (e.g. ¼ of a pie). Nutri-Score is an interpretive FOP label that uses
73
letters and colours to rank healthiness of products [11]. The French government adopted the Nutri-
74
Score in 2017 and since then governments of other countries such as Belgium, Switzerland and
75
Germany have also chosen to adopt the Nutri-Score [16,17]. In Latin America, the implementation of
76
warning labels is spreading. Ecuador was the first to implement a mandatory FOP label system, i.e.
77
a traffic light system [18]. They were followed by Chile which implemented mandatory warning
78
labels in 2016 [19]. Since then, Peru [20], Paraguay and Uruguay also decided to implement warning
79
labels and Mexico has recently followed suit [21]. Brazil recently reviewed mandatory nutrition
80
labelling. Anvisa, the National Health Surveillance Agency of Brazil established a working group on
81
Nutrition Labelling to identify problems in the transmission of nutritional information and
82
alternatives that could help improve the effectiveness of nutrition labelling [22]. Several FOP labels
83
were proposed to Anvisa and reviewed. In October 2020, ANVISA approved a FOP label for food
84
and beverages, which is a warning label that uses a nutrient profile based on added sugar, saturated
85
fat and sodium content per 100g or 100 ml of product [23,24].
86
This study was performed in 2019 and designed to determine which of five different types of
87
nutritional front-of-pack labels best helps Brazilian respondents identify the healthiest choice
88
between two food products, compared to a non-label control. Four of the tested FOP labels were also
89
reviewed by Anvisa. We wanted to test the robustness of the efficacy of different FOP labels by also
90
comparing products from different product categories, consumed in different serving sizes or with
91
closer nutritional profiles. As a result, wide range of products were tested.
92
2. Materials and Methods
93
2.1. Study population
94
Study participants were recruited from an existing research panel of Brazilian consumers that
95
represent general members of the public. The aim was to have a representative sample of
3 of 15
96
respondents and an even gender split. Participants were eligible for participation if they were aged
97
between 18 and 65 years.
98
2.2. Front-of-pack labels
99
Five FOP labels were tested in this study, i.e. GGALIii Nutrient Profile, IdeC, ABIA, Nutri-
100
Score and a hybrid label which was developed for this study. The first four FOP labels were
101
selected because they were proposed by different stakeholders to Anvisa, the National Health
102
Surveillance Agency of Brazil [22]. The selected labels cover three different visual expressions of
103
nutritional labelling. GGALIii Nutrient Profile and IdeC labels are warning style labels, Nutri-Score
104
is an interpretative colour coded label. The ABIA label is a traffic light system that provides
105
nutritional information. The hybrid label is an adaptation of the Evolved Nutrition Label [25] and
106
contains both nutritional information and a red colour code, as warning sign, if levels of an
107
ingredient are high. The Control Group received visual expressions of products without a FOP
108
label. This group was used as a reference.
109
This study focused on the three nutrients of concern that were initially proposed by Anvisa, i.e.
110
saturated fat, sugar and sodium [22]. Hence other nutrients were not taken into account.
111
GGALIii NP label: GGALI (Gerência-Geral de Alimentos) is Anvisa’s General Food
112
Management who prepared the "regulatory impact analysis" published by Anvisa in 2018 [22].
113
GGALI proposed two nutrient profiles. We selected the stricter one – GGALIii –. The GGALIii NP
114
label is a warning style label highlighting high levels of nutrients of concern. It is based on the
115
nutrient content per 100 g or 100 ml for food and beverages, having as reference the guidelines of
116
the World Health Organization (WHO) and Codex Alimentarius [22]. Criteria were defined for low,
117
medium and high content of free sugars, saturated fat, total fat and sodium [22]. For the GGALIii
118
NP label in this study criteria for high levels were applied. When this study was designed and
119
conducted, Anvisa had not chosen the nutrient profile or visual model. The nutrient profile that
120
was approved by ANVISA in 2020, is more lenient than the nutrient profile that we used in this
121
study. The magnifying glass visual that we used is similar to the visual approved by Anvisa [23,24].
122
This visual is also under discussion in Canada.
123
IDEC label: IDEC (Instituto Brasileiro de Defesa do Consumidor) is a civil society in Brazil.
124
They proposed a FOP label that is a warning style label and uses black triangles to inform the high
125
content of sugars, total fat, saturated fat and sodium, and the presence of trans fats and sweeteners.
126
The nutritional profile model was adapted from the Pan American Health Organization (PAHO)
127
profile model and is based on percentage of energy [22].
128
ABIA label: ABIA (Associação Brasileira das Indústrias da Alimentação), representing the
129
Brazilian food industry sector, proposed a Multiple Traffic Light (MTL) FOP label, based on the
130
United Kingdom traffic lights, which reports the absolute quantities of sugars, saturated fats and
131
sodium per serving. It uses the red, amber and green colours to indicate the high, medium and low
132
levels of each nutrient according to criteria per serving [22].
133
Nutri-Score: Nutri-Score is an interpretative, graded, colour-coded FOP label that has been
134
developed by French researchers [11]. It is based on the nutrient profiling system of the United
135
Kingdom Food Standards Agency which uses the nutrient content per 100 g for food and beverages.
136
Positive points (0-10) are allocated for energy, total sugar, saturated fat and sodium content and
137
negative points (0-5) are allocated for fruit, vegetables and nuts, fibre and protein content. Products
138
scores range from -15 (most healthy) to +40 (least healthy) [11] and are translated into five
139
categories of nutritional quality ranging from A (green) to E (red).
4 of 15
140
Hybrid label: The hybrid FOP label is a model developed for this study and is an adaptation of
141
the Evolved Nutrition Label (ENL) [25]. Criteria for sugar, saturated fat and sodium were defined
142
for small serving sizes (<60 g), medium serving sizes (60-120 g) and large serving sizes (>120 g). In
143
line with ENL, calculations were done per serving size, except for serving sizes between 60 and 120
144
g, in which case calculations were done per 100 g. The label provides quantitative nutritional
145
information per serving and uses the red colour to indicate high amounts of the nutrient in a
146
serving of the product.
147
An example of the five labels is provided in Figure 1. The specific criteria used for the ABIA,
148
GGALIii, IdeC and Hybrid labels are presented in Table S1. For the Nutri-Score label, the Nutri-
149
Score algorithm was used to calculate the score for each product [11].
150
151
Figure 1. Example of front-of-pack labels used in this study (translated from Portuguese to English)
152
2.3. Food stimuli
153
This study included 18 food items, which were presented to the respondents in sets of two. The
154
respondents were asked to indicate which of the two products they thought was healthier. The correct
155
answer was defined considering the contents of the three nutrients highlighted on the front of pack
156
label. That is, the product with the lowest sugar, saturated fat and sodium content per serving was
157
considered the healthiest choice. When a product was higher in one nutrient and lower in another,
158
the larger difference was considered most important. In a few cases, one nutrient was slightly lower
159
and one much higher. Subsequently the much higher nutrient was considered more important for
160
the classification.
161
The food sets (food stimuli) were carefully selected to test the robustness of the labels to help the
162
consumer identify the healthier option. The food stimuli differed with regard to the following
163
variables: similar products consumed in small serving size, similar products consumed in large
164
serving size, similar product but consumed in different serving sizes, and products from different
165
food categories but consumed in same eating occasion. The food categories included in the study
166
were soft cheeses, fat spreads, ice creams, lasagne, frozen meals, fermented milks & chocolate oat
167
drinks, sweet snacks, cereal bars & yoghurt and chocolate bars. The products corresponded to
168
different consumption occasions (e.g. breakfast, lunch or main meal and in-between meal snack).
169
Figure 2 shows an example of one of the stimuli as presented to the respondents. In Brazil it is not
170
mandatory to declare sugar content of food products on the packaging. For some of the products
171
used in this study, i.e. ice creams, frozen meals, the dairy alternative drink and sweet snacks, we had
172
to estimate the sugar content. Estimations were based on similar products marketed in countries
173
where sugar content is declared in the nutrition table. Table 1 summarizes the nutritional profiles of
174
the food stimuli. Other, detailed information about the food stimuli can be found in Figure S1.
175
5 of 15
176
177
Figure 2: example of stimulus used for each randomized group
178
Table 1. Nutritional profile of food stimuli
179
Per serving
Per 100 g
Serving
Sugar
Saturated Sodium
Sugar Saturated Sodium
Food category
Product
size (g)
(g)
fat (g)
(mg)
(g)
fat (g)
(mg)
Soft Cheese
Product 1*
30
NA
1,9
118
NA
6,3
393
Soft Cheese
Product 2
30
NA
5,2
239
NA
17,3
796
Fat spreads
Product 1*
10
NA
0,9
70
NA
9
700
Fat spreads
Product 2
10
NA
4,8
90
NA
48
900
Ice
cream Product
1
86 21,5 7,7 40 25 8,9 46,5
Ice cream
Product 2*
60
13
1,8
12
21,7
3
20
Lasagne Product
1
400
12 10
1280
3 2,5
320
Lasagne Product
2*
400
11,2
5,2
1440
2,8
1,3
360
Frozen meals
Product 1*
300
NA
2,9
250
NA
0,9
83,3
Frozen
meals Product
2
275 NA 9,3 1242
NA 3,4 451,6
Fermented
milk
drink
Product
1
200
32,5 0 75
16,25 0 37,5
Chocolate oat drink
Product 2*
260
12
0,6
60
4,6
0,2
23
Sweet snacks
Product 1
40
10,6
3,4
68
26,6
8,4
170
Sweet snacks
Product 2*
20
6,6
2,1
26
33
10,5
130
Cereal Bar
Product 1*
21
6,9
0,8
0
32,8
3,8
0
Yoghurt Product
2
170
20,4
4,6
160
12
2,7
94
Chocolate bars
Product 1*
16,7
8
2,6
16
47,9
15,5
100
Chocolate bars
Product 2
40
19,9
6,6
40
49,8
16,5
100
* Healthier option
180
6 of 15
181
2.4. Data collection
182
A 5-minute online survey was conducted in July 2019, using Toluna QuickSurveys. All
183
respondents were randomly allocated to one of six groups 1) no label (control), 2) ABIA label, 3)
184
GGALIii NP label, 4) IdeC label, 5) Hybrid label or 6) Nutri-Score label.
185
Apart from the control group, the respondents were shown the FOP label. The FOP label was
186
briefly explained. The respondents were then shown 9 food stimuli consisting of two products and,
187
based on the information they received, they were asked which product they thought was the
188
healthier choice. The respondents in the control group were also shown the same 9 choice sets of
189
products, but without a label. Respondents were provided with four potential answers: 1) Product 1;
190
2) Product 2; 3) No difference and 4) I don’t know. Respondents were also asked to rate, on a scale
191
from 0 to 10, how useful the label was in helping them make a healthy food choice and what they
192
liked and disliked about the specific type of label that they had evaluated.
193
2.5. Statistical analyses
194
Descriptive statistics were calculated for the sociodemographic data of the participants. The
195
percentage of participants selecting the answers “Product 1”, “Product 2”, “No difference” or “Don’t
196
know” were calculated for each set of food stimuli, for each FOP label group as well as for the control
197
group. Significance testing (Z-test) was performed to test if the proportion of participants correctly
198
identifying the healthier product differed between the FOP label groups. Significance tests were
199
performed within these subgroups to test whether participants who correctly chose the healthier
200
option differed according to education level or income. The mean scores for usefulness of the labels
201
were calculated. T-tests were used to test for statistical differences between mean scores.
202
Statistics were performed with the Toluna Analytics tool. A significance testing at a 95%
203
confidence level was used.
204
3. Results
205
3.1. Respondents
206
A total of 1072 Brazilian men and women participated in the online survey. Sociodemographic
207
data are presented in Table S2. A total of 176 respondents were included in the control group, 181
208
respondents were allocated to the ABIA group, 177 respondents were allocated to the GGALIii NP
209
warning label, 181 respondents to the IdeC – triangle warning label, 178 respondents to the hybrid
210
colour code label and 179 respondents were allocated to the Nutri-Score label.
211
There were no consistent significant differences in terms of education and income levels between
212
the different groups.
213
3.2. Accuracy of choosing the healthier product
214
Table 2 summarizes for each of the labels and the control group the percentages of respondents
215
that correctly identified the healthier product for each of 9 stimuli.
216
When the soft cheese food stimuli were presented, most respondents accurately identified the
217
healthier product. Respondents who were shown the IdeC warning label performed significantly
218
worse than the control group; a third of the respondents in this group indicated that there was no
219
difference between the two products and only 51% of the respondents correctly identified the
220
healthier product.
221
Respondents who were shown the ABIA or hybrid labels were most likely to identify the
222
healthier choice from the fat spreads category. Most respondents who were shown the GGALIii or
223
IdeC warning style labels selected the least healthy product of the two (41% and 44%, respectively)
224
or indicated that there was no difference between the two products (32% and 24%, respectively). The
225
GGALIii label scored worse than the control group. Also 55% of the respondents in the control group
226
chose the less healthy product.
7 of 15
227
For the ice cream stimuli, the group of respondents who were shown the IdeC labels had most
228
difficulty choosing the healthier option; only 13% of the respondents chose the healthier product.
229
Forty percent of the respondents in the control group and 60% of the respondents from the IdeC
230
group indicated that there was no difference between the two ice cream products. The ABIA and
231
hybrid labels helped the respondents best to make the healthier choice, with 85% and 88%,
232
respectively selecting the healthier product.
233
Many respondents had difficulty identifying the healthier lasagne. Only the respondents who
234
were shown the hybrid label performed significantly better (64% correct) than the control group (48%
235
correct). The Nutri-Score label helped only 20% of respondents make the right choice, compared to
236
48% of the respondents in the control group. Seventy percent of the respondents who were shown
237
the Nutri-Score label indicated that there was no difference between the products.
238
Respondents in the control group and those shown the IdeC label found it particularly difficult
239
to choose the healthier option from the frozen meals. A total of 56% of the respondents in the control
240
group and 62% of respondents in the IdeC label group indicated that there was no difference between
241
the two frozen meals. The ABIA and hybrid label performed best.
242
In the category fermented milk and chocolate oat drinks, the ABIA, Hybrid and Nutri-Score
243
labels performed best and the GGALIii and IdeC labels performed worse with about one-third of
244
respondents choosing the less healthy option and about one-third indicating that there was no
245
difference between products.
246
There was a marked difference between labels in their ability to help choose the healthier sweet
247
snack. The majority of the respondents in the control group (56%) and those who were shown the
248
GGALIii (73%) or Idec (57%) label considered that there was no difference between the two products
249
with regard to health. The labels ABIA, Hybrid and Nutri-Score performed significantly better.
250
When respondents were shown a cereal bar and a yoghurt product, i.e. two very different
251
products, the ABIA, IdeC and hybrid labels were most successful in helping them make a healthier
252
choice. Forty-five percent of respondents who were shown the GGALIii label believed that there was
253
no difference between products and 72% of respondents who were shown the Nutri-Score label
254
selected the less healthy product.
255
There was confusion amongst the majority of the respondents for the chocolate bar category
256
(showing two different sizes of the same brand chocolate bar). Most respondents seeing the GGALIii
257
(59%), Idec (56%) or Nutri-score (63%) labels, which do not consider serving size, thought that there
258
was no difference between the two products. The ABIA and Hybrid labels, which do consider serving
259
size, resulted in the highest numbers of respondents choosing the healthier option.
260
Overall, the IdeC warning label was least helpful for consumers to make the healthier choice.
261
Eight out of nine times, the IdeC label for the healthier product was the same as for the less healthy
262
product, providing no guidance to the consumer. Only once, when comparing the cereal bar and
263
yoghurt, the IdeC label outperformed the control group. In that case, 70% of the respondents chose
264
the healthier option. The other warning label, GGALIii NP, performed a bit better than the IdeC label,
265
but also failed to distinguish products six out of nine times. It outperformed the control group in only
266
two out of nine cases. Nutri-Score performed reasonably well but also failed two times in guiding the
267
consumer to the healthier choice when products were given the same rating. This happened for
268
example when the serving sizes of the two products differed significantly. In case of the cereal bar
269
(21g) versus yoghurt (170g) food stimulus, a better Nutri-Score was given for the least healthy
270
product (yoghurt).
271
The hybrid label performed best, resulting in statistically significantly higher percentage of
272
correct answers as compared to the control in all cases. The ABIA label outperformed the control
273
group eight times out of nine.
8 of 15
274
Table 2: Percentage of participants correctly choosing the healthiest option within each food stimulus, by
275
randomization assignment to FOP label or control
Soft
Fat
Ice
Lasagne Frozen Milk
Sweet
Cereal
Chocolate
cheese
spreads
cream
meals
drinks
snacks
bar &
bars
yoghurt
Control
76d
37c
31d 48d,f
13 40 26c 44 c,f 39 c,f
ABIA
88a,c,d 82a,c,d,f
85 a,c,d,f 46 d,f 81 a,c,d,f 73 a,c,d 77 a,c,d,f 85 a,c,d,f 76 a,c,d,f
GGALIii
77d
24 65a,d 41f 59a,d 32 15 31
29
IdeC
51 32 13 33f 12 34 23 70 a,c,f 33
Hybrid
88 a,c,d 75 a,c,d,f 88 a,c,d,f 64a,b,c,d,f 85 a,c,d,f 78 a,c,d 79 a,c,d,f 86 a,c,d,f 72 a,c,d,f
Nutri-Score
87 a,c,d 58 a,c,d 60 a,d 20 67a,d 74 a,c,d 66 a,c,d 22
26
276
Performing significantly (p<0.05) better than a: Control; b: ABIA; c: GGALIii NP; d: IdeC; e: Hybrid; f: Nutri-
277
Score within the same category
278
279
If we look at the participants who correctly identified the healthier options, there appears to be
280
no effect of education level or income. Statistical tests showed a statistically significant effect for only
281
3 of the 36 subgroups tested (9 food stimuli x 6 FOP label groups). Low-income participants from the
282
control group scored significantly higher than high-income participants in selecting the healthier ice
283
cream. Participants with a higher income who were shown the ABIA label scored significantly higher
284
participants with a lower income in selecting the healthier milk drink. Finally, the less educated in
285
the control group scored higher than the higher educated when selecting the healthier chocolate bar.
286
3.3. Usefulness ratings
287
Respondents were asked to indicate on a scale of 0-10 how useful the label was in helping to
288
choose the healthier product. Between 89% and 92% of respondents rated their label as either very
289
useful (7-8) or extremely useful (9-10). Mean usefulness scores were 9.7, 9.6, 9.4, 9.6 and 9.8 for the
290
ABIA, GGALIii NP, IdeC, Hybrid and Nutri-Score labels, respectively, and did not differ statistically
291
significant. The IdeC and GGALIii NP warning labels had the highest proportion (7% and 5%,
292
respectively) of respondents indicating that the label was not at all useful (0-4). The proportion of
293
respondents that rated the IdeC labels as not at all useful (7%) was significantly higher than the 2%
294
of respondents rating the Nutri-Score and Hybrid label as not useful. See Figure 3 for usefulness
295
ratings of FOP labels.
296
297
Figure 3: Usefulness ratings of the five FOP labels
9 of 15
298
299
Answer to Question: : To what extent is the label below useful when helping you make healthy food choices?
300
3.4. Feedback on labels
301
To better understand how the FOP labels were perceived by the respondents, they were all asked
302
what they liked or disliked about the FOP label they were shown.
303
ABIA label: Most respondents said they liked the label, especially the use of colours and the
304
clarity and objectivity of the information provided. When asked about what they did not like, some
305
respondents indicated that they would also like information about other nutrients and energy.
306
GGALIii NP label: Respondents were pleased with the label’s simplicity, its decisive message
307
and the fact that it attracts attention. Not all respondents were satisfied with the black colour and
308
some were missing nutritional information.
309
Hybrid label: Respondents were particularly pleased with the use of the red colour as it draws
310
attention and also the clear and easy to understand information about the nutrients and serving size.
311
However, the language should be kept simpler (e.g. salt instead of sodium) and some respondents
312
want more information about other nutrients and energy.
313
IdeC: Respondents were especially pleased with the simple information that helps people make
314
a quick decision. Not all respondents were happy with the black colour and said it wasn’t noticeable,
315
and some would like more specific information about the amount of the nutrients in the products.
316
Nutri-Score: When evaluating the Nutri-Score label, the respondents indicated that they liked
317
the simple, clear message and the use of colours. Negative aspects of the Nutri-Score were the lack of
318
information about nutrient levels and the underlying model. Some respondents misinterpreted the
319
label as they believed that the colours and letters represented the presence of vitamins in the
320
products.
321
In summary, respondents prefer simple FOP labels that use colours (not black) to convey the
322
message. They would like to receive information about the amount of nutrients in the products, but
323
in simple language.
324
4. Discussion
325
FOP labels are designed to help consumers choose healthier food and drinks. This study
326
compared how well five different FOP labels helped Brazilian consumers make a healthier choice
10 of 15
327
between two food or drink products. It showed that the Hybrid label and the ABIA label,
328
performed best under the tested conditions. Both labels provide both nutritional information per
329
serving as well as concise interpretation using colours (i.e. traffic light or red light only).
330
Most of the other studies that investigated how well different FOP labels helped consumers
331
make a healthier choice between products, compared products within the same food category and
332
with the same serving size, but with marked differences in nutrient profiles. These studies often
333
found that simple interpretative labels such as Nutri-Score or warning labels were effective in
334
helping the consumer make a healthier choice [26-29]. This makes sense, because when a label
335
clearly distinguishes between products (e.g. different score, colour or with or without a warning
336
label), the consumer can easily make a choice. However, this does not reflect the complexity
337
consumers face when shopping as this clear distinction does not always exist. Our research showed
338
that when labels do not clearly distinguish between two products, or when products from different
339
product categories or with different serving sizes are compared, these simple labels do not help the
340
consumers to make an informed choice.
341
Many factors influence how consumers process information on a FOP label and how deeply
342
this information is processed [30]. Consumers may only glance at the FOP label, process partial
343
information or process the FOP label in depth. For example, the level of nutritional knowledge
344
influences the type of information the consumer processes. Knowledgeable consumers are more
345
likely to use the more complex nutrient information on complex labels, while a less knowledgeable
346
consumer may look for calorie and color-coded information. Average consumers are more likely to
347
process the information on the FOP label in depth. Under time pressure consumers will only
348
quickly inspect the information on the FOP label and not process all available information [30].
349
Health-motivated consumers may also look more actively for nutritional information, while
350
hedonically-motivated consumers may not look at nutritional information, but more at brand
351
names [30]. So, depending on the situation, different types of FOP labels can be the most effective.
352
Our study showed that simple summary labels are effective when there is a clear distinction
353
between products, but is a quick decision cannot be made, consumers will consider the nutritional
354
information on the FOP label, when available. In those cases, interpretative labels that provide
355
nutritional information to the consumer better assist the consumer in making an informed choice.
356
This study showed that, even when the colours on the Hybrid and ABIA labels did not differ
357
between products, participants could choose the healthier option based on the nutritional
358
information presented on these labels. This suggests that the nutritional information facilitates the
359
comparison of the nutritional content of the products, allowing the consumer to make a healthier
360
choice. The two warning labels, i.e. IdeC and GGALIii NP were in most cases not sensitive enough
361
to help the consumer distinguish products based on healthiness. They did not outperform the
362
control group. Both the Idec and GGALIii labels use very strict nutrient profiles and therefore most
363
products bear the logo [22], making them less sensitive to distinguish products. The nutrient profile
364
that will be implemented in Brazil is more lenient and if we would have used this more lenient
365
profile, the number of warning labels would have been different for two food stimuli; only one of
366
two frozen meals would have carried a warning logo for sodium and the yogurt would have
367
carried no warning label for added sugar. Besides the lack of discrimination between products,
368
these labels are also very simplistic and do not contain additional nutritional information to help
369
the consumer make an informed decision, where the number of warning labels for sugars, saturated
370
fats and sodium does not differ between two foods.
371
When labels on two different products are the same, respondent interpret this as if products
372
are equally healthy, or respondents base their decision on other information that they have about
373
the product (e.g. packaging, type of product, claims on product, presence of other ingredients,
374
knowledge of the brand). For example, the vegetable-oil based spread used in this study contained
375
less saturated fat and sodium than the presented butter, so it is nutritionally, the healthier choice.
11 of 15
376
However, Brazilian Dietary guidelines promote butter consumption, not vegetable-oil based
377
spreads. Butter can therefore be seen as more natural and healthier than vegetable-oil based spreads
378
in Brazil and this is also reflected in the results. When the FOP labels indicated that the vegetable-
379
oil based spread was the healthier choice, most respondents chose that product. However, if no
380
label was shown, or if labels on butter and vegetable-oil based spreads did not differ (in case of
381
IdeC and GGALIii NP), more than 40% of participants chose butter as the healthier option.
382
A recent review of FOP schemes performed by the European Commission concluded that FOP
383
schemes providing nutritional information per 100g were better understood than portion-based
384
schemes [13]. However, more than 90% of the food categories in Brazil have regulated serving sizes
385
less than 100g / 100ml. When a nutrient profile is standard applied in 100g or 100mL distorted
386
comparisons are generated. For products consumed in serving sizes <100g or ml, the amount of
387
nutrients to calculate the FOP label is overestimated, while for products consumed in portions of
388
>100g/ml it is underestimated. As a result, some products with small serving sizes will unfairly
389
receive a warning label, while some products with large serving sizes that are high in nutrients of
390
concern receive no warning label. For example, in this study, two lasagnas with a 400g serving size
391
were compared. According to the nutritional profile criteria defined by GGALIii, based on 100g,
392
neither of the two products would receive a warning label, and with Nutri-Score both lasagnas
393
would receive a score of B. These two FOP labels would thus suggest that products are healthy,
394
despite the relatively high saturated fat and sodium contents per serving as % GDA.
395
While the GGALIii NP and IdeC warning labels were the least successful in helping
396
participants make the healthier food choice, the labels were considered by the respondents to be as
397
useful as the other FOP labels. It is important to note that the respondents did not receive any
398
feedback on how well they did. So, they were not aware of the correct answer and how often they
399
correctly identified the healthier option or mistakenly assumed there was no difference. One could
400
speculate that if they got this feedback, ratings of usefulness would be lower. In any case, the
401
ratings show that any FOP label that could help make the consumer an informed choice is
402
considered useful by consumers. For research purposes, asking this question without providing
403
feedback to the participants does not seem relevant. Feedback from the participants suggests that
404
simple FOP labels that use bright colours and contain nutritional information in simple language,
405
are liked.
406
Grunert et al hypothesized that consumers’ liking for FOP labels is guided by three
407
considerations: 1) consumers like simplicity, 2) when provided with simplified information
408
consumers still want to know what it stands for and how the simplified message (e.g. warning- or
409
health logo) has been derived, and 3) nutrition information can create a consumer resistance when
410
they feel pushed to make choices that they do not want to take [31].
411
This is also confirmed by a recent study conducted by Talati et al [32] who investigated
412
consumer perception of five FOP labels, i.e Health Star Rating, MTL, Nutri-Score, RI and a warning
413
label. The coloured FOP labels MTL and Nutri-Score stood out and were most liked by consumers
414
in all countries. Although the most simplified FOP labels, Nutri-Score and warning labels, were
415
easy to understand, they were perceived as providing insufficient information and the least trusted.
416
The RI label was perceived as the most confusing but scored high on trust. Overall, the MTL label,
417
which combines nutrient-specific information and a summary interpretation using colour, was most
418
liked and trusted in this study.
419
A strength of the current study was that it really tested the robustness of five FOP labels that
420
differed not only in visual expression and the amount of information provided, but also in the
421
underlying nutrient profile. Unlike other studies that mostly tested products within the same food
422
category and with the same serving size, this study was designed to compare how well these five
423
FOP labels enabled consumers to choose between products that differ in nutritional composition,
12 of 15
424
serving size and/or food category as consumers face in real life. Another strength is that a control
425
group has been included and that the effectiveness of the FOP labels in helping the consumer to
426
choose the healthier product could therefore be compared with a reference group that was not
427
given a FOP label. This provides insight into whether the presence of a specific FOP label is of
428
added value for a consumer when making an informed choice.
429
Participants were also asked to select the healthier product. This demonstrates how effective
430
the FOP label is in helping the consumer make a choice and whether it fits its purpose. In other
431
studies consumers were asked which product they would buy [33-37], but this may be influenced
432
by factors such as familiarity and liking of the product and cost of the product [31]. Other studies
433
only asked which FOP label is preferred [32,38]. As demonstrated by our study, all FOP labels were
434
rated as very useful, irrespective of their efficacy in helping the consumer choose the healthier
435
option. Only asking for preference is thus not very useful.
436
This study also has some limitations. Participants were a representative sample of the Brazilian
437
population. Therefore, we also included participants with a lower education level, who may have
438
had difficulty understanding the information on the FOP labels. However, socio-economic status
439
and level of education was similar between the six FOP label groups and therefore we did not
440
expect this to affect the outcomes of the study. This was confirmed by statistical subgroup analyses
441
showing that participants who correctly identified the healthier options, did not differ with respect
442
to level of education or income. Another limitation is that we did not ask the participants if they
443
were colour-blind. Thus, it is possible that participants with colour-blindness were included, which
444
may have adversely affected the ability to understand the colour-coded labels. However, none of
445
the participants who were shown the Abia or Nutri-Score labels, voluntarily reported being colour-
446
blind and thus unable to interpret the labels.
447
The FOP labels that we tested in this online survey were selected because they were under
448
consideration by Anvisa, the National Health Surveillance Agency of Brazil, at the time we
449
designed this study. Anvisa proposed in its preliminary report on the regulatory impact analysis on
450
nutrition labelling [22] to focus only on the three nutrients of concern, sugar, saturated fat and
451
sodium. We therefore decided to only use the content of these three nutrients to inform the different
452
FOP labels (with exception of Nutri-Score). Focusing on just these three nutrients of concern is a
453
limitation to assessing the healthiness of a product.
454
Brazilian regulation do not require the sugar content of food products to be stated on the
455
packaging. For some of the products used in this study (ice creams, frozen meals, dairy alternative
456
drink, sweet snacks), we had to estimate the sugar content. These estimates were unlikely to deviate
457
very much from the actual sugar content and were used for all FOP labels.
458
This study was conducted online using pictures of actual products. It does therefore not reflect
459
a real-life situation in which participants can examine packaging and other information, such as the
460
nutrition table on the back, to make an informed choice. Finkelstein et al [39] attempted to mimic a
461
real-life situation by asking the participants (n=147) to purchase their weekly groceries in an online
462
grocery store with 3343 foods and 832 beverages. Participants had only access to back-of-pack
463
Nutrition Information Tables or were also shown an MTL label or Nutri-Score label. Both the MTL
464
and Nutri-Score FOP labels improved the dietary quality of the purchases as compared to the
465
control group. The Nutri-Score label performed best in improving overall diet quality, but unlike
466
Nutri-Score, the MTL label reduced calories. Thus, FOP labels had added value when purchasing
467
products, even in the presence of a Nutrition Information Table.
468
5. Conclusions
469
In conclusion, this study showed that the Hybrid and the ABIA FOP labels, two interpretative
470
labels that use colours and provide nutritional information per serving, were best suited to help
13 of 15
471
Brazilian consumers choose the healthier product. The other three labels are based on per 100g and
472
therefore will not always discriminate enough to help consumers. The ABIA and Hybrid labels
473
outperformed the other FOP labels when serving sizes differed significantly or when deeper
474
consideration of nutritional information was needed to make an informed decision.
475
Author Contributions:
476
Conceptualization, Chantal Goenee and Els M. de Groene; Data curation, Lucia Juliano; Formal analysis, Lucia
477
Juliano; Funding acquisition, Els M. de Groene; Investigation, Lucia Juliano; Methodology, Chantal Goenee,
478
Lucia Juliano, Els M. de Groene and Fernanda de Oliveira Martins; Project administration, Chantal Goenee and
479
Fernanda de Oliveira Martins; Resources, Chantal Goenee and Fernanda de Oliveira Martins; Supervision, Els
480
M. de Groene; Visualization, Wendy A.M. Blom, Lucia Juliano and Fernanda de Oliveira Martins; Writing –
481
original draft, Wendy A.M. Blom; Writing – review & editing, Chantal Goenee, Lucia Juliano, Els M. de Groene
482
and Fernanda de Oliveira Martins. All authors have read and agreed to the published version of the manuscript.
483
Funding: This research was funded by Unilever.
484
Acknowledgments: We thank the volunteers who participated in this online survey and all other staff at
485
Harris Interactive, Toluna and Unilever who helped with the study or provided input to this manuscript.
486
Conflicts of Interest: The authors, W.A.M.B, C.G., E.d.G. and F.d.O.M. were all employed by Unilever, the
487
sponsor of this study, when the study was performed. L.J. declares no conflict of interest.
488
References
489
1.
Cowburn, G.; Stockley, L. Consumer understanding and use of nutrition labelling: a systematic
490
review.
Public health nutrition 2005,
8, 21-28, doi:10.1079/phn2005666.
491
2. World
Health
Organization.
Guiding principles and framework manual for front-of-pack labelling for
492
promoting healthy diets; World Health Organization: Geneva, 2019.
493
3.
Kanter, R.; Vanderlee, L.; Vandevijvere, S. Front-of-package nutrition labelling policy: global progress
494
and future directions.
Public health nutrition 2018,
21, 1399-1408, doi:10.1017/S1368980018000010.
495
4.
Labonté, M.-È.; Poon, T.; Gladanac, B.; Ahmed, M.; Franco-Arellano, B.; Rayner, M.; L Abbé, M.R.
496
Nutrient Profile Models with Applications in Government-Led Nutrition Policies Aimed at Health
497
Promotion and Noncommunicable Disease Prevention: A Systematic Review.
Advances in Nutrition
498
2018,
9, 741-788, doi:10.1093/advances/nmy045.
499
5.
van der Bend, D.L.M.; Lissner, L. Differences and Similarities between Front-of-Pack Nutrition Labels
500
in Europe: A Comparison of Functional and Visual Aspects.
Nutrients 2019,
11,
501
doi:10.3390/nu11030626.
502
6.
The Food and Drink Federation. Food labelling. Availabe online:
503
https://www.gdalabel.org.uk/gda/reference-intakes.html (accessed on
504
7.
Food Standards Agency. Front of Pack nutrition labelling guidance. Availabe online:
505
https://www.gov.uk/government/publications/front-of-pack-nutrition-labelling-guidance (accessed
506
on 25 February 2020).
507
8.
Swedish National Food Agency. The Keyhole. Availabe online:
508
https://www.livsmedelsverket.se/en/food-and-content/labelling/nyckelhalet (accessed on 25 February
509
2020).
510
9.
Choices International Foundation. Choices programme. Availabe online:
511
https://www.choicesprogramme.org (accessed on 25 February 2020).
512
10.
Australian Government of Health. Front-of-pack labelling updates. Availabe online:
513
https://foodregulation.gov.au/internet/fr/publishing.nsf/Content/front-of-pack-labelling-1 (accessed
514
on 25 February 2020).
14 of 15
515
11.
Julia, C.H., S. Development of a new front-of-pack nutrition label in France: the five-colour Nutri-
516
Score.
Public Health Panorama 2017,
3, 712 - 725.
517
12.
Reyes, M.; Garmendia, M.L.; Olivares, S.; Aqueveque, C.; Zacarias, I.; Corvalan, C. Development of
518
the Chilean front-of-package food warning label.
BMC public health 2019,
19, 906, doi:10.1186/s12889-
519
019-7118-1.
520
13.
Storksdieck genannt Bonsmann, S.M., G. ; Cirolo, E.; van Bavel, R.; Wollgast, J. Front-of-pack nutrition
521
labelling schemes: a comprehensive review. Publications Office of the European Union: Luxembourg,
522
2020; doi:10.2760/436998.
523
14.
Cecchini, M.; Warin, L. Impact of food labelling systems on food choices and eating behaviours: a
524
systematic review and meta-analysis of randomized studies.
Obesity Reviews 2016,
17, 201-210,
525
doi:10.1111/obr.12364.
526
15.
Temple, N.J. Front-of-package food labels: A narrative review.
Appetite 2020,
144, 104485,
527
doi:https://doi.org/10.1016/j.appet.2019.104485.
528
16.
Federal public service Health Food Chain Safety and Environment Belgium. Nutri-Score. Availabe
529
online: https://www.health.belgium.be/nl/de-nutri-score-0 (accessed on 3 March 2020).
530
17.
Bundesministerium für Ernährung und Landwirtschaft. Ergebnis der Verbraucherbeteiligung liegt
531
vor: Bundesministerin Julia Klöckner wird Nutri-Score(R) einfüren. Availabe online:
532
https://www.bmel.de/SharedDocs/Pressemitteilungen/2019/197-eNWK.html (accessed on
533
18.
Cox Vásconez, A.E. RESOLUCIÓN No. 14 511. Gobierno Nacional de la Republica del Ecuador.
534
Ministerio de Industrias Y Productividad. Subsecretaria de la Calidad. Quito, Ecuador, 2014.
535
19.
Carreño, I. Chile s Black STOP Sign for Foods High in Fat, Salt or Sugar.
European Journal of Risk
536
Regulation 2015,
6, 622-628, doi:10.1017/S1867299X0000516X.
537
20.
El Presidente de la Republica Peru. Aprueban Manual de Advertencias Publicitarias en el marco de lo
538
establecido en la Ley Nº 30021, Ley de promoción de la alimentación saludable para niños, niñas y
539
adolescentes, y su Reglamento aprobado por Decreto Supremo Nº 017-2017-SA. El Peruano: Peru,
540
2018.
541
21.
Rojo Sánchez, A.G.N.B., J.A. Modificación a la Norma Oficial Mexicana NOM-051-SCFI/SSA1-2010,
542
Especificaciones generales de etiquetado para alimentos y bebidas no alcohólicas preenvasados-
543
Información comercial y sanitaria, publicada el 5 de abril de 2010. Secretaria De Economia Gobierno
544
Mexicano, Ed. Mexico City, 2020; Vol. NOM-051-SCFI/SSA1-2010.
545
22.
Agência Nacional de Vigilancia Sanitária (Anvisa).
Relatório Preliminar de Análise de Impacto Regulatório
546
sobre Rotulagem Nutricional. Gerência-Geral de Alimentos; Brasília, 2018.
547
23.
Didier, D. Resolução de diretoria colegiada – RDC nº 429, de 8 de outubro de 2020 – ANVISA.
548
ANVISA: 2020.
549
24.
Colegiada, M.d.S.A.N.d.V.S.D. Instrução normativa-in nº 75, de 8 de outubro de 2020. Estabelece os
550
requisitos técnicos para declaração da rotulagem nutricional nos alimentos embalados. Health, M.o.,
551
Ed. Diário oficial da união: 2020; Vol. 195, p 113.
552
25.
Egnell, M.; Kesse-Guyot, E.; Galan, P.; Touvier, M.; Rayner, M.; Jewell, J.; Breda, J.; Hercberg, S.; Julia,
553
C. Impact of Front-of-Pack Nutrition Labels on Portion Size Selection: An Experimental Study in a
554
French Cohort.
Nutrients 2018,
10, doi:10.3390/nu10091268.
555
26.
Egnell, M.; Ducrot, P.; Touvier, M.; Alles, B.; Hercberg, S.; Kesse-Guyot, E.; Julia, C. Objective
556
understanding of Nutri-Score Front-Of-Package nutrition label according to individual characteristics
15 of 15
557
of subjects: Comparisons with other format labels.
PloS one 2018,
13, e0202095,
558
doi:10.1371/journal.pone.0202095.
559
27.
Egnell, M.; Talati, Z.; Gombaud, M.; Galan, P.; Hercberg, S.; Pettigrew, S.; Julia, C. Consumers
560
Responses to Front-of-Pack Nutrition Labelling: Results from a Sample from The Netherlands.
561
Nutrients 2019,
11, doi:10.3390/nu11081817.
562
28.
Galan, P.; Egnell, M.; Salas-Salvado, J.; Babio, N.; Pettigrew, S.; Hercberg, S.; Julia, C. Understanding
563
of different front-of-package labels by the Spanish population: Results of a comparative study.
564
Endocrinologia, diabetes y nutricion 2019, 10.1016/j.endinu.2019.03.013, doi:10.1016/j.endinu.2019.03.013.
565
29.
Arrua, A.; Machin, L.; Curutchet, M.R.; Martinez, J.; Antunez, L.; Alcaire, F.; Gimenez, A.; Ares, G.
566
Warnings as a directive front-of-pack nutrition labelling scheme: comparison with the Guideline
567
Daily Amount and traffic-light systems.
Public health nutrition 2017,
20, 2308-2317,
568
doi:10.1017/S1368980017000866.
569
30.
Sanjari, S.S.; Jahn, S.; Boztug, Y. Dual-process theory and consumer response to front-of-package
570
nutrition label formats.
Nutrition reviews 2017,
75, 871-882, doi:10.1093/nutrit/nux043.
571
31.
Grunert, K.G.; Wills, J.M. A review of European research on consumer response to nutrition
572
information on food labels.
Journal of Public Health 2007,
15, 385-399, doi:10.1007/s10389-007-0101-9.
573
32.
Talati, Z.; Egnell, M.; Hercberg, S.; Julia, C.; Pettigrew, S. Consumers Perceptions of Five Front-of-
574
Package Nutrition Labels: An Experimental Study Across 12 Countries.
Nutrients 2019,
11, 1934,
575
doi:10.3390/nu11081934.
576
33.
Talati, Z.; Egnell, M.; Hercberg, S.; Julia, C.; Pettigrew, S. Food Choice Under Five Front-of-Package
577
Nutrition Label Conditions: An Experimental Study Across 12 Countries.
American journal of public
578
health 2019, 10.2105/ajph.2019.305319, e1-e6, doi:10.2105/ajph.2019.305319.
579
34.
Mora-Garcia, C.A.; Tobar, L.F.; Young, J.C. The Effect of Randomly Providing Nutri-Score
580
Information on Actual Purchases in Colombia.
Nutrients 2019,
11, doi:10.3390/nu11030491.
581
35.
Billich, N.; Blake, M.R.; Backholer, K.; Cobcroft, M.; Li, V.; Peeters, A. The effect of sugar-sweetened
582
beverage front-of-pack labels on drink selection, health knowledge and awareness: An online
583
randomised controlled trial.
Appetite 2018,
128, 233-241, doi:10.1016/j.appet.2018.05.149.
584
36.
Machin, L.; Arrua, A.; Gimenez, A.; Curutchet, M.R.; Martinez, J.; Ares, G. Can nutritional
585
information modify purchase of ultra-processed products? Results from a simulated online shopping
586
experiment.
Public health nutrition 2018,
21, 49-57, doi:10.1017/S1368980017001185.
587
37.
Acton, R.B.; Jones, A.C.; Kirkpatrick, S.I.; Roberto, C.A.; Hammond, D. Taxes and front-of-package
588
labels improve the healthiness of beverage and snack purchases: a randomized experimental
589
marketplace.
The international journal of behavioral nutrition and physical activity 2019,
16, 46,
590
doi:10.1186/s12966-019-0799-0.
591
38.
Dana, L.M.; Chapman, K.; Talati, Z.; Kelly, B.; Dixon, H.; Miller, C.; Pettigrew, S. Consumers Views
592
on the Importance of Specific Front-of-Pack Nutrition Information: A Latent Profile Analysis.
593
Nutrients 2019,
11, doi:10.3390/nu11051158.
594
39.
Finkelstein, E.A.; Ang, F.J.L.; Doble, B.; Wong, W.H.M.; van Dam, R.M. A Randomized Controlled
595
Trial Evaluating the Relative Effectiveness of the Multiple Traffic Light and Nutri-Score Front of
596
Package Nutrition Labels.
Nutrients 2019,
11, doi:10.3390/nu11092236.
597