Content uploaded by Pamela Ebner
Author content
All content in this area was uploaded by Pamela Ebner on Apr 22, 2022
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=iijf20
International Journal of Food Sciences and Nutrition
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/iijf20
How are the processing and nutrient dimensions
of foods interconnected? an issue of hierarchy
based on three different food scores
Pamela Ebner, Kelly Frank, Aris Christodoulou & Sylvie Davidou
To cite this article: Pamela Ebner, Kelly Frank, Aris Christodoulou & Sylvie Davidou (2022):
How are the processing and nutrient dimensions of foods interconnected? an issue of hierarchy
based on three different food scores, International Journal of Food Sciences and Nutrition, DOI:
10.1080/09637486.2022.2060951
To link to this article: https://doi.org/10.1080/09637486.2022.2060951
View supplementary material
Published online: 09 Apr 2022.
Submit your article to this journal
View related articles
View Crossmark data
RESEARCH ARTICLE
How are the processing and nutrient dimensions of foods interconnected?
an issue of hierarchy based on three different food scores
Pamela Ebner, Kelly Frank, Aris Christodoulou and Sylvie Davidou
Siga, Saint-Mand
e, France
ABSTRACT
Worldwide, foods are scored with composition indices. However, processing scores are now
emerging. The objective of this study was to study the interconnectedness of the degree of
processing and composition for 28,747 industrially packaged foods (71.6% of ultra-processed
foods, UPFs) representative of retail assortments. The Nutri-score and Traffic Light Labelling
System (TLLS) were used to assess the composition, and the Siga index was used to assess the
degree of processing. On average, the more nutritionally favourable Nutri-score and TLLS groups
exhibited 56.5 and 50.0% UPFs, respectively. Among markers of ultra-processing non-additives
mostly included added fat/sugar/fibre/vitamin, animal and/or plant protein isolates, and taste
exhausters, while additives mostly included sweeteners and taste exhausters, suggesting that
markers of ultra-processing (MUP) are added to foods to improve composition scores. In conclu-
sion, both types of scores are not complementary as such but obey to a fundamental hierarchy:
processing first, then composition if necessary.
ARTICLE HISTORY
Received 5 January 2022
Revised 16 March 2022
Accepted 29 March 2022
KEYWORDS
Industrial packaged foods;
Siga index; composition
scores; ultra-processed
foods; markers of ultra-
processing; additives; ultra-
processed ingredients
Introduction
To date, the principal and preferred tools used world-
wide to help consumers to purchase healthier food are
compositional scores, such as the Health Star Rating
System in Australia since 2014 (Maganja et al. 2019),
the Traffic Light Labelling System (TLLS) in the UK
since 2014 (Mach
ın et al. 2018), the Nutri-score in
France since 2017 (Egnell et al. 2018), the Keyhole
Symbol in Nordic countries since 2009 (Mørk et al.
2017), the Nutrition Facts Panel in the USA since
1994 (Graham et al. 2015; Hastak et al. 2020) and
Nutritional Warnings in Chile since 2016, and others
for Latin American countries (Reyes et al. 2019;
Nobrega et al. 2020). These compositional scores have
been designed following a hypothetical-deductive
approach that starts with a theoretical concept, model
or algorithm (conceived from reductionist data)
(Fardet and Rock 2020). They are then tested in labo-
ratories and specific populations, but they carry the
risk of being disconnected from reality, notably that
of real foods because foods are not only sums of
nutrients (Fardet and Rock 2020); and they are gener-
ally based on an aggregate of nutrients to be favoured
or limited, neglecting the fundamental matrix effect
and therefore they may potentially confuse consumers
by suggesting that nutrients are interchangeable from
one food to another, which is not scientifically true
(Fardet and Rock 2022). Besides, not only the health
potential of a food is first determined by the quality
of its matrix –and secondarily its nutrient compos-
ition –(Fardet and Rock 2022), but also by the equi-
librium of the overall complex diet in which it is
included (Visioli et al. 2021); which may lead in the
end to a false “dichotomic classification of foods into
“healthy”and “unhealthy”products”(Visioli
et al. 2022).
Although some of them being applied only quite
recently, their efficiency nevertheless deserves to be
questioned with regard to the continuous increased
prevalence of overweight/obesity and type 2 diabetes
in these countries, especially among the youngest and/
or the poorest populations (Ezzati et al. 2017; Inoue
et al. 2018;F
ed
eration Internationale du Diab
ete 2019;
International Diabetes Federation 2019; Merema et al.
2019; Altobelli et al. 2020; Fuentes et al. 2020;
Rasmussen et al. 2020; U.S. Department of Health and
Human Services and Centre for Disease Control and
Prevention 2020; Vos et al. 2020; Wong et al. 2020;
Hemmingsson et al. 2021), and also to the concomi-
tant increased sales of ultra-processed foods (UPFs)
worldwide (Baker et al. 2020; WHO 2020). Otherwise,
CONTACT Sylvie Davidou sylvie@siga.care Siga, 5 Avenue du G
en
eral De Gaulle, 94160 Saint-Mand
e, France
Supplemental data for this article can be accessed here.
ß2022 Taylor & Francis Group, LLC
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION
https://doi.org/10.1080/09637486.2022.2060951
Dubois et al. (2021) recently concluded that “front-of-
pack nutrition labels had disappointingly modest effects
on the nutritional quality of the foods purchased in
real-life grocery shopping conditions”.
For other researchers, an increased prevalence of
chronic diseases appears more linked to the degree of
processing than to an apparently balanced content in
a few nutrient (Monteiro 2009). Besides, some studies
have shown that there was no link between the com-
position of real foods in a given “nutrient to limit”
and chronic diseases, e.g. saturated fatty acids, cheeses
and cardiovascular diseases (Chen et al. 2017) or sim-
ple sugars, fruits and type 2 diabetes (Du et al. 2017),
among others. In contrast, fruits and cheeses appear
protective, suggesting that the food matrix environ-
ment of the given nutrient, and hence the degree of
processing, may have a greater effect on health than
the nutrient itself. Thus, the real issue appears more
linked to added industrial and highly processed fats,
sugars, salt and some additives, which are discon-
nected from a complex matrix (i.e. a-matrix nutrients)
and conveyed by degraded and artificialised food
matrices such as UPFs (Fardet and Rock 2019). Thus,
the consumption of excess sugars, fats, salt and/or
additives appears to be only an effect, not the real
root cause, these latter being linked to the degree of
processing (Fardet and Rock 2022); and the apparent
“hyper-formulation”of foods with added cosmetic
compounds (effects) seems to result from the ultra-
processing of foods (cause) to compensate for the deg-
radation of food matrices (Fardet and Rock 2020).
Therefore, from a preventive perspective, a holistic
approach based on the degree of processing appears
relevant for society. New way of classifying foods
according to their degree of processing has emerged
with several proposed classifications, such as NOVA
(Moubarac et al. 2014), Siga (Davidou et al. 2020),
and others (Gonz
alez-Castell et al. 2007; Slimani et al.
2009; Asfaw 2011; Poti et al. 2015; Sanchez-Siles et al.
2019). Contrary to composition indices, they consider
food health potential more holistically because proc-
essing affects both the food matrix and composition
(Fardet and Rock 2018,2020). Notably, NOVA
defined the new concept of UPFs (Monteiro et al.
2019). Contrary to composition scores, NOVA classifi-
cation is empirico-inductive and based on qualitative
observations of different types of foods in real life,
both in societies and countries where the prevalence
of chronic diseases has dramatically increased; and it
has been suggested that UPF is an important or
explanatory variable to consider in chronic disease
prevention (FAO et al. 2019). Accumulating scientific
evidence of the deleterious effects of UPFs on health
has been well documented (Costa et al. 2018; Askari
et al. 2020; Chen et al. 2020; Elizabeth et al. 2020;
Pagliai et al. 2021), without counting all previous
studies on the influence of various processes on food
health potential (Fardet 2018) before the appearance
of the UPF concept in 2009 (Monteiro 2009); but
results might differ according to the processed food
classification system used, as reported with cardiome-
tabolic health in an elderly population (Martinez-
Perez et al. 2021).
Otherwise, while used worldwide by many aca-
demic researchers, the concept of UPFs is still dis-
cussed by some others (Gibney 2019; Sadler et al.
2021; Castro-Barquero and Estruch 2022); and some
advocate for a better understanding of biological
mechanisms involved (Tobias and Hall 2021). Finally,
one ecological study reported a positive and signifi-
cant correlation between household availability of
UPFs and obesity prevalence in nineteen European
countries (1991–2008) (Monteiro et al. 2019), whereas
another reported no association between ultra-proc-
essed food and drink consumption (% energy) and
country-level burden of high body mass index during
the 1997–2015 period in twenty two European coun-
tries (Mertens et al. 2022).
Recently, a classification using the concept of
markers of ultra-processing (MUPs) has been devel-
oped to systematically identify UPFs independently of
their nutritional composition (Davidou et al. 2021).
They contain industrial ultra-processed ingredients
and additives that generally modify the colour, aroma,
texture and taste of foods (Davidou et al. 2020). Some
of these MUPs may be potentially used to replace salt,
sugar and/or fat within packaged foods. If not,
decreasing sugar or fat content will be naturally and
mechanistically accompanied by an increased propor-
tion of other sweet molecules or fatty molecules that
are in reality fat or sugar. For example, table sugar
may be replaced by maltodextrin MUP, not counted
in theory in the composition scores as sugars. Yet,
maltodextrin is a worse sugar than table sugar from a
nutritional point of view, notably with a higher gly-
caemic index (Hofman et al. 2016) and potential dele-
terious impact on microbiota (Naimi et al. 2021).
Within this context, the main objective of this
study was to characterise the degree of processing of
foods in relation with their nutritional quality in order
to determine how these two dimensions are intercon-
nected. For this, we used two different developed
French scores, i.e. the Nutri-score to assess food com-
position based on some nutrients to favour or to limit
2 P. EBNER ET AL.
(Julia et al. 2014) and the Siga score to assess the
degree of processing, including added fat, salt and/or
sugar (Davidou et al. 2020). For UPF only, a compari-
son was also realised with the TLLS based on only the
amount of four nutrients (sugar, salt, fat, saturated
fat) and energy. A total of 28,747 industrial packaged
foods were selected as representative of assortments in
French super- and hypermarkets. A secondary object-
ive was to identify which MUPs may be added to
foods to modify their compositional scores.
Materials and methods
Food selection
Baby foods, alcoholic beverages and food supplements
were not considered in this study because they are
generally reserved for specific populations and because
alcoholic beverages are not considered as foods. A
total of 28,747 packaged foods were selected as repre-
sentative of food assortments in French supermarkets
and hypermarkets and extracted in totality from the
Siga database –up to October, 2021 –according to 10
food categories (i.e. beverages (7.2% of products), bak-
ery products and pastries (5.7%), starchy foods (4.6%),
fruits and vegetables (6.9%), ready-to-eat meals
(8.4%), seafood products (4.2%), dairy products
(13.2%), salted products (12.9%), sweet products
(23.7%), and meats and eggs (12.2%)). The Siga data-
base is built from data provided from its partners
(manufacturers and retailers for 65% of all foods) and
that collected within the framework of the partnership
with the “Consumer Transparency program”of
Alkemics (i.e. a retail collaborative platform that helps
retailers and brands manage, collaborate and share
product data, for 35% of all foods). Food data are the
set of legal information found on the product packag-
ing, i.e. the list of ingredients and the fat, sugar and
salt contents indicated within the nutritional table,
which makes it possible to calculate the Siga score.
They were collected between July, 2017, and
September, 2020. Besides, manufacturers’data are rep-
resentative of the French market, including around
6,000 different food brands. Overall, 20% of food
products belong to private labels specific to retailers,
and the remaining 80% are national industrial brands
that are distributed in all different brands of French
super- and hyper-markets. The percentages of UPFs
were calculated directly from the Siga database and
the list of ingredients (Davidou et al. 2020).
Concerning the application of the Nutri-score, it
was listed directly in the French Open Food Facts
database. This open collaborative database of food
products marketed in France, licenced under the
Open Database Licence (ODBL), is the most compre-
hensive database in this regard (open source data-
base). Concerning the TLLS, it was applied directly on
foods from the Siga database, based on fat, saturated
fat, sugar and salt contents (n¼25,875 pack-
aged foods).
The Siga classification
The Siga classification and its methodology have been
previously described in detail (Davidou et al. 2020). In
this study, for a better description and analysis, the
UPF group, initially described by three technological
groups (C01, C02, C1), were subdivided into five
groups according to the number and nature of MUPs,
the presence of risk-associated additives, and the
added sugar, salt and/or fat contents, i.e. C01, C02,
C1, C2 and C3, respectively. In addition, the Siga clas-
sification distinguishes two levels of MUPs as follows:
MUP1: obtained through chemical synthesis with
the final compound being identical to natural sub-
stances; successive processes lead to both purifica-
tion or high deterioration of the ingredient matrix,
as is the case for isolated protein, starches, natural
flavouring, and yeast extract. They also include
drastic processes directly applied to foods that
highly modify their matrix, e.g. extrusion-cooking
(Monteiro et al. 2019; Davidou et al. 2020).
MUP2: obtained through artificial chemical synthe-
sis with the final compound not existing in nature;
successive processes lead to the combined purifica-
tion and substantial deterioration of the matrix, as
is the case for glucose syrup, dextrose, hydrolysed
proteins, carboxymethylcellulose, etc.
C01 contains only one MUP1 with low levels of
added salt, sugar and/or fats (according to the FSA -
Food Standard Agency - threshold) (U.K. Department
of Health et al. 2016); C02 contains only one MUP1
with high levels of added salt, sugar and/or fats
(according to the FSA threshold) (Health et al. 2016);
The medium FSA nutritional thresholds were 1.5 g
salt/100 g, 12.5 g sugars/100 g and 17.5 g fat/100 g for
foods, and 0.75 g salt/100 g, 6.25 g sugars/100 g and
8.75 g fat/100 g for beverages (Health et al. 2016).
When added sugars, fats and/or sugars were not
detected within the list of ingredients of packaged
foods, the above-mentioned thresholds were not taken
into consideration for calculations.
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION 3
C1 contains several MUP1s (from 2 to 4 MUP1s);
and C2 and C3 contain several MUP1s or at least one
MUP2 and/or at least one “at-risk”additives. The sum
of MUP1 and MUP2 is equal to total MUPs and
includes additive MUPs (A-MUPs) (according to
European law) as well as non-additive ingredient
MUPs (NA-MUPs). MUPs were also ranked in 15
subcategories as follows:
1. Fibre NA-MUPs: added fibre isolates
(non-additive).
2. Animal protein NA-MUPs: added animal protein
isolates and/or hydrolysates (non-additive).
3. Plant protein NA-MUPs: added plant protein
isolates and/or hydrolysates (non-additive).
4. Sweetener A-MUPs: added sweeteners, e.g. aspar-
tame or sorbitol.
5. Taste exhauster A-MUPs: Taste exhauster as
additives, e.g. monosodium glutamate or diso-
dium guanylate.
6. Taste exhauster NA-MUPs: Taste exhauster as a
non-additive, e.g. yeast extract.
7. Sugar NA-MUPs: added sugars as non-additives,
e.g. fructose-glucose syrup or invert sugar.
8. Fat NA-MUPs: added fats as non-additives, e.g.
refined or hydrogenated oils.
9. Other A-MUPs: all additive MUPs except taste
exhausters, sweeteners and non-trace MUPs.
10. Non-traced NA-MUPs: NA-MUPs not included
in the assessment of UPF Siga categories due to
regulation, e.g. gluten.
11. Non-traced A-MUPs: A-MUPs not included in
the assessment of UPF categories due to regula-
tion, e.g. non “at risk”preservatives and antioxi-
dant additives.
12. Process MUPs: technologically processed MUPs,
e.g. extrusion-cooking.
13. Meat NA-MUPs: meat ingredients as non-additives.
14. Vit NA-MUPs: added vitamins as non-additives.
15. OtherNA-MUPs:allotherMUPsnotlistedabove.
Beyond A-MUP and NA-MUP ingredients, only
industrial processes applied to foods and highly modify-
ing their matrix (e.g. extrusion-cooking or puffing) were
considered as MUPs (Davidou et al. 2020,2021)orindi-
cators of ultra-processing (Monteiro et al. 2019).
The Nutri-score and Traffic Light Labelling System
The Nutri-score is a five-level nutrition labelling sys-
tem, with scores ranging from A to E and green to
red, based on the nutritional value of a food product.
The score is calculated by a rating system, with a
lower score being better. There are four elements
unfavourable to the score (per 100 g): calorie, sugar,
saturated fat and salt contents; and three elements
favourable to the score: contents of fibre, proteins,
and fruits, vegetables, legumes, nuts, rapeseed, walnut
and olive oil. In the calculation of the contents of
fruits and vegetables, starchy foods (such as potatoes,
sweet potatoes, taro, cassava and tapioca) are not
taken into account. Finally, an algorithm integrates all
seven elements into a both aggregated and weighted
score according to different thresholds of
these components.
The TLLS tells whether a food has high, medium
or low amounts of fat, saturated fat, sugars and salt. It
also tells the number of calories in that particular
product. The attribution of green (G), orange (O) and
red (R) colours was assessed according to high,
medium and low levels of fat, saturated fatty acids,
total sugars and salt in foods as determined by the
FSA thresholds (Health et al. 2016). For calculations,
25,875 out of 28,747 packaged foods were selected
2,872 products being withdrawn due to their lack of
data concerning their contents in saturated fats.
Among the 25,875 packaged foods, we selected those
with the best nutritional profile whatever the nutrient
considered, i.e. 4 G (4 greens), 3G1O (3 greens, 1
orange), 3G1R (3 greens, one red), 2G2O (2 greens, 2
oranges), 2G1O1R (2 greens, 1 orange, one red),
2G2R (2 greens, 2 reds) and 1G3O (1 green 3
oranges) (n¼13,741 products), and we calculated the
percentages of non UPF and UPF according to Siga
index within each one of these categories.
Statistical and machine learning analyses
Most results are presented as distributions according
to different criteria for each Nutri-score and TLLS
category: UPF and non UPF percentages; number of
A-MUPs and NA-MUPs by food; food percentages
containing MUPs leading to the modification of the
Nutri-score (e.g. added fibre, proteins …); percentages
of foods with medium or high fat, sugar and salt con-
tents (FSA threshold for foods and drinks); percen-
tages of food with added fat or sugar and with added
fat, sugar and/or salt (n>1 added); and percentages
of food with medium and high FSA thresholds for fat,
sugar and/or salt (n¼[1–3]). The average number of
A-MUPs/food versus NA-MUPs/food in each Nutri-
score category was evaluated for significance with
Student’st-test at p<.05 (SPAD9.1 software,
Coheris#, Suresnes, France).
4 P. EBNER ET AL.
Decision tree analysis was applied to the “13 MUP
sub-categories (excluding Non-traced NA/A-MUPs)
20,595 UPFs”matrix to define rules for foods in
regard to belong to the five Nutri-score categories.
Among the machine learning algorithms, the chi-
square automatic interaction detector (CHAID) was
chosen, instead of classification and regression trees
(CARTs) and C4.5 algorithms, because it is more
adapted to an exploratory study with a big sample. As
recommended for a large sample, a total of 80% of
the foods were used for the learning sample
(n¼16,476), and the remaining 20% were used for
the test sample (model validation). In the final ana-
lysis, “<0.5”indicates the absence of a MUP and “
0.5”indicates the presence of a MUP. Decision tree
analysis was performed on a PC computer with the
SPAD9.1 software (Coheris#, Suresnes, France).
Results
Product characterisation
Concerning food categories, starchy foods, fruits, veg-
etables, and ready-to-eat meals exhibited the highest
percentages with Nutri-scores of A or B (55%),
while dairy products, salted and sweet products, and
meats and eggs had the lowest percentages (18%)
(Supplemental Figure 1(a)). For the Siga score, starchy
foods, fruits, vegetables, seafood, dairy products, and
beverages exhibited the highest percentages of foods
scored by Siga A or B (44%), while bakery prod-
ucts, pastries, ready-to-eat meals, salted and sweet
products, and meats and eggs obtained the lowest per-
centage (22%) (Supplemental Figure 1(b)). In add-
ition, according to the Siga score, 72% of the 28,747
food items were UPFs (Figure 1(a)), and with Nutri-
score 70% were scored C, D or E (Figure 1(b)).
Numbers of UPFs and MUPs in the different Nutri-
score, Siga and Traffic Light Labelling
System categories
The percentage of UPFs increased from Nutri-score A
to score C and stabilised at 79–82% (Figure 2(a)). In the
first two Nutri-score categories, there were an average of
approximately 56% UPFs, with 41% in the Nutri-score
Agroup(Figure 2(a)). Considering the Siga score in
more detail, the number of UPFs scoring C1-3 increased
from Nutri-score A to Nutri-score D and decreased to
4,083 products in the Nutri-score E group
(Supplemental Figure 2(a)), which corresponded to an
increase in the percentage of C1-3 UPFs from 25 to
73% (percentages not shown). Conversely, the number
of foods scored A according to Siga decreased from
Nutri-score categories A to E (Supplemental Figure 2(a))
(i.e. from 41 to 4% of food products, percentages not
shown). Otherwise, the intermediary foods scored B
according to Siga did not follow a clear tendency
regarding Nutri-score categories from A to E
(Supplemental Figure 2(a)), with 18% in the Nutri-score
A group, 11% in the Nutri-score B and C group, 21%
in the Nutri-score D group, and 13% in the Nutri-score
E group (percentages not shown). When considering the
distribution of foods according to Nutri-score in each
Siga category the number of foods scored Nutri-score A
decreased from Siga category A to C1–C3 (i.e. from 50
to 7% of food products, percentages not shown)
(Supplemental Figure 2(b)). A reverse trend was
observed for foods scored Nutri-score B-E for which
percentages increased from 6–18%inSigacategoryAto
14–31% in Siga category C1-C3 (Supplemental
Figure 2(b)).
Concerning the TLLS, the percentages of UPFs
increased from the more nutritionally balanced cat-
egory 4 G (28%) to the moderately nutritionally bal-
anced 1G3O category (80%), with the exception of the
2G2R category (46%) that exhibited a largely lower
percentage of UPF than the 2G1O1R category (80%)
a) b)
16
14
23
28
19
0
5
10
15
20
25
30
ABCDE
% products
Nutri-score
13 15 15
57
0
10
20
30
40
50
60
ABC01-2C1-C3
% products
Siga score
Figure 1. (a–b) Distribution of foods by (a) Siga score and (b) Nutri score (n¼28,747 products).
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION 5
(Figure 2(b)). On average, 50% of foods presenting
4 G or 3 G were UPFs, and 59% of foods presenting at
least 1 G and 0 R were UPFs.
Otherwise, the number of A-MUPs and NA-MUPs
by food increased by at least 2.4-fold from Nutri-score
category A to category E, with three distinct profile
41
72 79 75 82
59
28 21 25 18
0
20
40
60
80
100
120
ABCDE
% UPF % non UPF
a)
b)
69
26
15 12 9
42
11
3
25
12
9
11
13
9
7
17
29
17
25
22
15
21
32
44
62
55
24
65
0
10
20
30
40
50
60
70
80
90
100
4G 3G1O 3G1R 2G2O 2G1O1R 2G2R 1G3O
A B C01-2 C1-C3
Figure 2. (a–b) (a) Percentages of UPFs and non-UPFs by Nutri-score categories, and (b) Percentages of UPFs (C01-2 and C1-C3
Siga categories) and non-UPFs (A and B Siga categories) by Traffic Light System categories, including foods with 4 “green”low lev-
els (4 G, n¼1033 foods), 3 “green”low levels and 1 “orange”medium level (3G1O, n¼3906 foods), 3 “green”low levels and 1
“red”high level (3G1R, n¼2461 foods), 2 “green”low levels and 2 “orange”medium levels (2G2O, n¼3906 foods), 2 “green”low
levels, 1 “orange”medium level and 1 “red”high level (2G1O1R, n¼1712 foods), 2 “green”low levels and 2 “red”high levels
(2G2R, n¼430 foods), and 1 “green”low level and 3 “orange”medium levels (1G3O, n¼2618 foods). (Green, red and orange
only refer to the colors used by the Traffic Light System categories and do not appear in the figure)
6 P. EBNER ET AL.
groups, i.e. Nutri-score A, B/C and D/E (Figure 3).
Overall, there were significantly more NA-MUPs than
A-MUPs by food in all Nutri-score categories.
The nature of MUPs in the different Nutri-
score categories
Considering all foods and only MUPs than can poten-
tially improve the Nutri-score rate, added animal and
plant protein NA-MUPs were the most frequent, espe-
cially added plant protein, which was present in 6–8%
of all products regardless of the Nutri-score category
(Figure 4). Added isolated fiber was found in approxi-
mately 4% of Nutri-score A and B foods, with cat-
egory B having the highest percentage of 5%.
Sweetener A-MUPs (especially in light beverages) and
taste exhauster NA-MUPs were especially added in
foods scored B. Except for animal and plant protein
NA-MUPs, other MUPs were present in relatively few
foods within the less nutritionally balanced Nutri-
score E, i.e. 0.4–2.5% (Figure 4, mainly in sweet prod-
ucts, salted products, meats & eggs, beverages, dairy
products, and bakery products & pastries, results not
shown). Animal protein NA-MUPs were particularly
added in the meats & eggs, ready-to-eat meals, sweet
0.6
1.4
1.6
2.1
2.1
1.0
2.1
2.2
2.4 2.4
0
1
2
3
EDCB
A
A-MUP/food
NA-MUP/food
*
****
Figure 3. Average number of A-MUPs and NA-MUPs by food for each Nutri-score category (n¼28,747 products). An asterisk indi-
cates that the average NA-MUP/food is significantly higher than the average A-MUP/food in each Nutri-score category (Student’st-
test, p<.05).
3
2
6
1
0
1
10
5
1
8
4
1
4
20
4
8
7
12
3
19
4
11
8
222
23
2
77
1
10
16
0
5
10
15
20
25
Fiber NA-MUP Animal protein NA-MUP Plant protein NA-MUP Sweetener A-MUP Taste exhauster A-MUP Taste exhauster NA-MUP Contains at least one the six
MUP
A B C D E
Figure 4. Percentages of foods containing MUPs that affect the Nutri-score and of those containing at least one of these six MUPs
for each Nutri-score category (n¼28,747 products).
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION 7
products, and bakery products & pastries categories
and to a lesser extent in the salted products, starchy
foods, dairy products, and seafood categories (results
not shown). Plant protein NA-MUPs were particularly
added in sweet products, dairy products, meats &
eggs, and bakery products & pastries categories and to
a lesser extent in ready-to-eat meals, seafood, starch
foods and salted products (results not shown). When
considering food products containing at least one of
these six MUPs (i.e. 1, Figure 4), we identified per-
centages of 10, 20, 19, 23 and 16% for Nutri-score cat-
egories A, B, C, D and E, respectively; approximately
one of six foods scored A and B (15% on average)
contained at least one of these MUPs versus 19% on
average of foods scored C–D.
Considering now only UPFs, other A-MUPs were
particularly present in Nutri-score C, D and E, i.e.
from 71.9 to 87.6% of UPFs, and to a lesser extent in
Nutri-score A and B, i.e. from 53.7 to 61.3% (Table
1). UPF percentages containing other A-MUPs
increased from Nutri-score A to E. Besides Other A-
MUPs and Other NA-MUPs, among Nutri-score cate-
gories B, C, D and E, MUPs the most present in
UPFs were Fat NA-MUPs, Sugar NA-MUPs, Plant
protein NA-MUPs and Animal protein NA-MUPs
(between 7 and 56%) while in Nutri-score category A
Process MUPs was more present than Animal NA-
MUPs (Table 1). Among MUPs than can potentially
improve the Nutri-score rate, Fibre NA-MUPs was
less and less present in UPFs from Nutri-score A to
E, and Taste exhauster NA-MUPs was less and less
present in UPFs from Nutri-score B to E. Finally, in
the Nutri-score category B, Taste exhauster NA-MUPs
and Sweetener A-MUPs were present in more UPF
(i.e. higher percentages) than in other Nutri-score cat-
egories A, C, D and E.
Added fat, sugar and/or salt contents in foods in
the different Nutri-score categories
The percentages of foods containing added fats or
sugars increased from Nutri-scores A to E, but there
was no clear difference between scores C and D
(Figure 5). More specifically, 26 and 53% of foods
scored A and B, respectively, contain added fat, while
26 and 48% had added sugar, respectively. A total of
29 and 57% of foods scored A and B had at least two
added ingredients: salt and fat, fat and sugar or sugar
and fat.
Decision tree analysis
Decision tree analysis allowed the identification of the
most discriminant rules (i.e. the nature of the MUPs)
for UPFs belonging to the five Nutri-score categories.
The first criterion separating UPFs according to
Nutri-score categories was Other A-MUPs, which
were more present in categories C–E than in catego-
ries A-B (Figure 6), as also observed in Table 1. Then,
among UPFs without Other A-MUPs, Process MUPs
became the most discriminating criterion for Nutri-
score categories, notably for Nutri-score category A,
and among foods with no Process MUPs, Fat MUPs,
Other NA-MUPs and sugar NA-MUPs became the
most discriminating criterion for Nutri-score catego-
ries, especially for categories B, C and D. Among
foods with Other A-MUPs, a strong co-occurrence of
Sugar NA-MUPs, Animal protein NA-MUPs and
Plant protein NA-MUPs was found for Nutri-score
category D. Except those co-occurrences, there was no
other strong MUP co-occurrences specific of a Nutri-
score category.
Table 1. Main MUPs by Nutri-score category and UPF percentages containing these MUPs
a
.
Nutri-score A (n¼1852)
b
Nutri-score B (n¼2883) Nutri-score C (n¼5172) Nutri-score D (n¼6110) Nutri-score E (n¼4578)
MUP % MUP % MUP % MUP % MUP %
Other NA-MUPs 60.0 Other NA-MUPs 71.6 Other A-MUPs 71.9 Other A-MUPs 77.8 Other A-MUPs 87.6
Other A-MUPs 53.7 Other A-MUPs 61.3 Other NA-MUPs 61.3 Other NA-MUPs 65.7 Other NA-MUPs 76.6
Fat NA-MUPs 47.9 Fat NA-MUPs 56.0 Fat NA-MUPs 45.6 Fat NA-MUPs 46.0 Sugar NA-MUPs 46.1
Sugar NA-MUPs 17.6 Sugar NA-MUPs 24.9 Sugar NA-MUPs 34.9 Sugar NA-MUPs 42.9 Fat NA-MUPs 34.1
Plant protein NA-MUPs 14.0 Plant protein NA-MUPs 11.3 Animal protein NA-MUPs 10.1 Animal protein NA-MUPs 14.6 Animal protein NA-MUPs 8.6
Process MUPs 8.8 Animal protein NA-MUPs 7.4 Plant protein NA-MUPs 8.5 Plant protein NA-MUPs 10.7 Plant protein NA-MUPs 7.6
Fibre NA-MUPs 6.9 Fibre NA-MUPs 6.4 Fibre NA-MUPs 5.2 Fibre NA-MUPs 4.9 Fibre NA-MUPs 3.0
Vit NA-MUPs 5.9 Vit NA-MUPs 6.2 Vit NA-MUPs 4.1 Taste exhauster A-MUPs 3.0 Vit NA-MUPs 2.3
Animal protein NA-MUPs 5.0 Taste exhauster NA-MUPs 5.5 Taste exhauster NA-MUPs 3.8 Process MUPs 3.0 Process MUPs 1.6
Taste exhauster NA-MUPs 3.5 Sweetener A-MUPs 5.4 Process MUPs 3.0 Sweetener A-MUPs 2.8 Taste exhauster A-MUPs 1.2
Sweetener A-MUPs 3.3 Process MUPs 2.0 Taste exhauster A-MUPs 2.0 Vit NA-MUPs 2.7 Sweetener A-MUPs 1.2
Taste exhauster A-MUPs 1.1 Taste exhauster A-MUPs 1.5 Sweetener A-MUPs 1.5 Taste exhauster NA-MUPs 2.1 Taste exhauster NA-MUPs 0.5
Meat NA-MUPs 0.2 Meat NA-MUPs 0.1 Meat NA-MUPs 0.3 Meat NA-MUPs 0.5 Meat NA-MUPs 0.3
a
For the description of MUP categories, see in Methods section;
b
Number of UPFs.
MUPs: Markers of Ultra-Processing; UPF: Ultra-Processed Foods.
8 P. EBNER ET AL.
26
53 54 54
69
26
48
62 60
83
29
57
60 59
80
0
10
20
30
40
50
60
70
80
90
EDCBA
Food percentage (%)
Added fat (MUP & non MUP)
Added sugars (MUP & non MUP)
Added fat, sugar and/or salt (n > 1)
Figure 5. Percentages of foods in each Nutri-score category with added fat or sugar and with added fat, sugar and/or salt (n>1
added) (n¼28,747 products).
Figure 6. Decision tree analysis for a food belonging to a Nutri-score category (a, b, c, d and e on the figure) according to differ-
ent MUP criteria. “NA-MUP”are non-additive markers of ultra-processing; “A-MUP”are additive markers of ultra-processing; for
“other NA-MUP”see Method section.
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION 9
Discussion
The main results of this study showed that more than
50% of foods with Nutri-scores A-B or with nutrition-
ally balanced TLLS categories (i.e. 4 G/3G1O/3G1R)
were UPFs, and that more than 70% were UPFs with
Nutri-score B, C, D and E, and with 2G2O, 2G2R and
1G30 TLLS categories. For Nutri-score, this was due
to the presence of both A-MUPs and NA-MUPs, with
1.1 (for category E) to 1.7 (for category A) times
more NA-MUPs than A-MUPs in the five Nutri-score
categories, also indicating that the NA-MUP/A-MUP
ratio is more important in nutritionally balanced cate-
gories. This is in accordance with the results of an
Australian study comparing the Health Star Rating
with the NOVA classification, which revealed that 3
out of 4 UPFs displayed at least 2.5 “health”stars
(Dickie et al. 2018). The authors concluded that this
“is misrepresenting the healthiness of new packaged
food products and creating a risk for behavioural
nutrition”(Dickie et al. 2018). A similar conclusion
can be also made with the TLLS, so that all these
three nutrient-based scores highlight at least one out
of two nutritionally balanced packaged food whereas
they are UPFs. It is very likely that the same may be
true for other nutritional labelling systems such as the
Nordic Keyhole (Nordic Council of Ministers 2010)
or The Nutrient Warning Labels in Latin America
(Ministerio de Salud - Gobierno de Chile 2021), that
are also based on similar nutrients to encourage and/
or to limit. Another study comparing 2,036 food items
of the food composition table of the French Nutrinet-
Sant
e cohort according to NOVA classification and
the Nutri-score found that on average, 41% of foods,
including fresh and packaged foods, scored A and B
are UPFs (Srour et al. 2019).
Therefore, when it is claimed that food compos-
ition labels make food choices healthier for consumers
(Mach
ın et al. 2018; Anabtawi et al. 2020; Defago
et al. 2020; De Temmerman et al. 2021), it is not sure
they really discourage consumer purchases of poten-
tially unhealthy industrially packaged UPFs. This is a
real concern because these reductionist compositional
scores may encourage consumers to buy a significant
amount of foods that present a false “health halo”,
and that may lead to the adoption of a deleterious
diet rather than a preventive diet.
Added cosmetic ingredients and/or additives are
not free from health consequences (Fardet and Rock
2020; Naimi et al. 2021). Besides, our study showed
that Process MUPs concerned near 9% of UPFs in
Nutri-score category A. Therefore, as result of proc-
esses directly applied to foods, apparently nutritionally
balanced UPFs may be less satiating and more hyper-
glycaemic (notably due to drastic flaking, extrusion-
cooking or puffing), such as in ready-to-eat breakfast
cereals for children (Foster-Powell and Miller 1995;
Fardet 2016; Fardet et al. 2018), and composition
scores do not take into account the important food
“matrix effect”(Fardet 2017; Fardet and Rock 2020).
In this latter case, cereals are not far from becoming
free glucose during the digestive process because the
starch matrix is unstructured, highly gelatinised and
degraded by enzymatic or thermomechanical dextrini-
zation, being readily bioavailable in humans. Thus,
the average glycaemic indices of corn flakes, rice puffs,
puffed wheat and extruded-cooked wheat were
reported to be 81, 87, 74 and 84%, respectively, while
those of minimally processed muesli and porridge
made from rolled oat are on average 57 and 58,
respectively (Foster-Powell et al. 2002).
Therefore, these scores indirectly encourage con-
sumers to purchase packaged foods, of which at least
50% could be associated with increased risks of several
chronic diseases when consumed in excess (Askari
et al. 2020; Lane et al. 2020; Pagliai et al. 2021), and
also to exclude nearly 22% of real foods, rated Nutri-
score D and E (Ebner et al. 2019)–and that can be,
either naturally or by their final composition, fatty,
sweet and/or salty (i.e. virgin oils, traditional cheeses
and biscuits …)–to the benefit of their ultra-proc-
essed counterparts. Although presented as only reflect-
ing nutritional balance in a particular food category,
which is partially true, because, first, nutritional bal-
ance is at least made at the level of the meal and,
second, no food is nutritionally balance –hence the
recommendation to “eat varied”,a“nutritionally bal-
anced”UPF remains an UPF, even if enriched with
fibre, protein isolates, sweeteners, vitamins, minerals
and/or other MUPs. Therefore, these results strongly
suggest an issue of hierarchy, i.e. first separating real
foods from UPFs, then assessing composition in proc-
essed foods and UPFs, as proposed by the holistico-
reductionist Siga algorithm based on a hierarchical
decision tree including first degree of processing, then
“matrix”effect, nutritional criteria, the number of
MUPs and the presence of “at risk”additives
(Davidou et al. 2020). Therefore, consumers need to
be informed of such a hierarchy to make an enlight-
ened choice.
In this way, our results also showed that 15% of
foods with Nutri-scores A or B contain at least one
added MUP that improves the score, i.e. fibre, animal
and/or plant protein isolates, sweeteners, or taste
exhausters (both A-MUP and NA-MUP). This may
10 P. EBNER ET AL.
lead to reformulation of a non-UPF within a UPF, e.g.
through using MUPs such as protein/fibre isolates or
sweeteners and taste enhancers that may replace sug-
ars and fat known to penalise the nutrient profiles. Is
there any reason to use such a MUP rather than
wholegrain flour, legumes, or honey? Adding artificial
sweeteners to decrease the simple sugar content of
sweetened beverages has been questioned with regard
to human health, notably in regard to obesity, type 2
diabetes and some cancers (Schernhammer et al. 2012;
Ara
ujo et al. 2014; Palmn€
as et al. 2014; Soffritti et al.
2014; Suez et al. 2014; Olivier et al. 2015; Imamura
et al. 2016; Mandrioli et al. 2016). In addition, we lack
perspective on the impacts of the massive use of pro-
tein isolates in UPFs on human health on a long
term. For example, the frequent addition of gluten
isolates has been questioned towards non-celiac glu-
ten/wheat sensitivity (Fardet 2015; Kucek et al. 2015).
Thus, while total calorie consumption can be
decreased and micronutrient/fibre consumption
improved, e.g. with micronutrient/fibre-enriched
UPFs, this will not necessarily prevent having a
chronic disease if too many UPF calories are con-
sumed (Fardet and Rock 2022). This indicates that the
food matrix and quality of calories (holistic approach)
are more crucial in the prevention of chronic diseases
than a strategy based on a quantitative reductionist
approach of foods focussing on total calories, fibre
and micronutrients (Fardet and Rock 2020).
Another negative consequence of compositional
scores is unfairly stigmatising real foods that have no
convincing link with some chronic diseases, e.g.
cheeses (Alegria-Lertxundi et al. 2014; Hjerpsted and
Tholstrup 2016; Chen et al. 2017; Dekker et al. 2019)
or butter (Pimpin et al. 2016), which are rich in nat-
ural saturated fats. This suggests that the food matrix
environment in which saturated fat are embedded
matters, and that natural saturated fats are not the
same for health than artificially added saturated fats.
Otherwise, except for breast milk as sustenance for
children from birth to 2–3 years, no food is com-
pletely nutritionally balanced. Finally, this strongly
questions the interest of scoring real foods with com-
positional scores, whatever they are.
Besides, among the five Nutri-score categories, the
B category appears the most questionable: compared
to foods with scores of C or D, foods in the B cat-
egory show similar profiles in regard to UPF percent-
age, and A-MUPs and NA-MUPs per food. Therefore,
the presentation of Nutri-score by itself, particularly
on foods with a score of B, appears problematic for
the consumer, creating a false “health halo”for
reformulated UPFs, while they might be only refor-
mulated foods initially scored C and D with
added MUPs.
Concerning medium and high FSA levels for fat,
sugar and salt and foods containing added fat and
sugars, we observe in this study that 26% of foods
with Nutri-score of A exhibited added fat and sugars,
and approximately 50% exhibited added fat and sugars
among foods scored B. Notably, added Sugar NA-
MUPs (e.g. maltodextrins/dextrins) are considered
hidden sugars, significantly improving the daily con-
sumption of free sugar, which should not exceed 10%
of total calories, ideally remaining below 5% (WHO
2015). Thus, considering that extrusion-cooking may
transform cereals into high-GI foods (carbohydrates
with rapid assimilation), which are then supplemented
with Sugar NA-MUPs, this illustrates that choosing
foods with favourable compositional scores is not at
all a guarantee to consume less sugar and not to pre-
vent exceeding a sugar intake of 10% calories daily.
Certainly, the presence of fibre and protein rebalances
the overall nutrient profile, but if the food remains
ultra-processed, this will not improve its health poten-
tial. In addition, some industrially added “refined diet-
ary fibre”have been pointed out as being potentially
detrimental to digestive health (Singh and Vijay-
Kumar 2020).
On the other hand, a surprising result showed that
there was less UPFs in the 2G2R TLLS category (46%)
–although being theoretically less nutritionally bal-
anced –than in the 2G1O1R category (80%). Actually,
this was mainly due to the presence of numerous non
UPF fatty products in the 2G1O1R category, such as
nuts, oleaginous seeds, and oils; which is problematic
for the consumer who might reject these real foods,
i.e. non-ultra-processed, because of a “red traffic
light”colour.
Otherwise, beyond Other A-MUPs, numerous
NA-MUPs were also present in each Nutri-score
categories. This may be of concern for industrially
processed organic foods: indeed, although the num-
ber of additives is restricted to 48 in France, NA-
MUPs are not restricted, probably explaining why
approximately 27% of food sales in organic stores
are UPFs when including fresh foods (Desquilbet
et al. 2018), and 48% when considering only
organic industrially processed foods, with around
three-fold more NA-MUPs used than A-MUPs
(Davidou et al. 2021). Therefore, it is not sufficient
to consider only A-MUPs to reduce UPF percen-
tages in Nutri-score categories A and B.
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION 11
Conclusions and perspectives
The result of this study showed that industrial proc-
essed foods that scored high with compositional
scores, such as the Nutri-score and the TLLS, may be
added with MUPs, i.e. mainly NA-MUP such as pro-
tein/fibre isolates, and some ultra-processed carbohy-
drates not counted in these scores, e.g. isolated
starches, extruded-cooked semolina/flours (Process
MUP) or maltodextrins. In the end, this creates a false
“health halo”for the consumer, without counting the
presence of numerous A-MUPs. Only assessing reduc-
tionist compositional scores appears to lead industries
to reformulate their products without changing the
real issue, i.e. degraded, hyper-attractive and artificial
ultra-processed matrices (Scrinis 2013,2016; Scrinis
and Monteiro 2018; Fardet and Rock 2020).
Therefore, as regards with the issue of the score
hierarchy for consumer food choices, the main con-
clusion of this study is that it is necessary to apply
first holistic processing scores, then, if necessary,
reductionist compositional scores for processed foods
and UPFs added with sugars, salt and/or fat, rather
than the contrary, to avoid an undue valorisation of
UPFs among consumers. However, isolated fibre and
protein isolates, when they originate from food crack-
ing and/or when they have undergone significant
changes in their initial matrix, should not be part of
composition score-related algorithms. This is what the
Siga score and its algorithm - based on a hierarchical
decision tree - have considered. Altogether, this will
avoid nutritional drifts in composition scores and an
undue stigmatisation of real foods which historically
belongs to preventive diets such as virgin oils or
cheeses in the Mediterranean diet.
We therefore propose that the degree of processing
is the first indicator of the health food potential, not
food composition, which only creates more confusion
for the consumer because such scores are not fitted
with the reality of what we eat. Regarding the high
rate of premature mortality worldwide due to
unhealthy diets (approximately 11 million per year,
i.e. 24% of total deaths in 2017) (Wang et al. 2019),
there is a critical need to adopt stronger food policies
based on stronger health indicators because the inclu-
sion of one food score on packaging will not be suffi-
cient to encourage people to eat healthy in the long
term. Therefore, these measures should be comple-
mented by diet education at young ages when tastes
and eating habits are forming, notably after weaning
from breastmilk and adopting adult diets (Black and
Hurley 2013).
Acknowledgements
The expert committee of Siga is acknowledged for the paper
proofreading.
Disclosure statement
There are no conflict of interest with industry to declare.
The Siga methodology was used as a tool for characterising
the degree of processing, with a private database. This base
has been built from publicly available information from
industry or retailors themselves. All authors are employees
of the Siga society.
This study received no specific funding except that of
the Siga society.
Funding
The author(s) reported there is no funding associated with
the work featured in this article.
ORCID
Sylvie Davidou http://orcid.org/0000-0001-7264-3839
Data availability statement
The data that support the findings of this study are avail-
able from the corresponding author, [SD], upon reasonable
request due to privacy restrictions.
References
Alegria-Lertxundi I, Pablo AR, Arroyo-Izaga M. 2014.
Cheese consumption and prevalence of overweight and
obesity in a Basque adult population: a cross-sectional
study. Int J Food Sci Nutr. 65(1):21–27.
Altobelli E, Angeletti PM, Profeta VF, Petrocelli R. 2020.
Lifestyle risk factors for type 2 diabetes mellitus and
national diabetes care systems in European countries.
Nutrients. 12(9):2806.
Anabtawi O, Swift JA, Hemmings S, Gertson L, Raaff C.
2020. Perceived healthiness of food items and the traffic
light front of pack nutrition labelling: choice-based con-
joint analysis and cross-sectional survey. J Hum. Nutr
Diet. 33:487–495.
Ara
ujo JR, Martel F, Keating E. 2014. Exposure to non-
nutritive sweeteners during pregnancy and lactation:
Impact in programming of metabolic diseases in the pro-
geny later in life. Reprod Toxicol. 49:196–201.
Asfaw A. 2011. Does consumption of processed foods
explain disparities in the body weight of individuals? The
case of Guatemala. Health Econ. 20(2):184–195.
Askari M, Heshmati J, Shahinfar H, Tripathi N, Daneshzad
E. 2020. Ultra-processed food and the risk of overweight
and obesity: a systematic review and meta-analysis of
observational studies. Int J Obes (Lond)). 44(10):
2080–2091.
12 P. EBNER ET AL.
Baker P, Machado P, Santos T, Sievert K, Backholer K,
Hadjikakou M, Russell C, Huse O, Bell C, Scrinis G,
et al. 2020. Ultra-processed foods and the nutrition tran-
sition: Global, regional and national trends, food systems
transformations and political economy drivers. Obes Rev.
21(12):e13126.
Castro-Barquero S, Estruch R. 2022. Ultra-processed food
consumption and disease: the jury is still out. Eur Heart
J. 43(3):225–227.
Chen G-C, Wang Y, Tong X, Szeto IMY, Smit G, Li Z-N,
Qin L-Q. 2017. Cheese consumption and risk of cardio-
vascular disease: a meta-analysis of prospective studies.
Eur J Nutr. 56(8):2565–2575.
Chen X, Zhang Z, Yang H, Qiu P, Wang H, Wang F, Zhao
Q, Fang J, Nie J. 2020. Consumption of ultra-processed
foods and health outcomes: a systematic review of epi-
demiological studies. Nutr J. 19(1):86.
Costa CS, Del-Ponte B, Assunc¸~
ao MCF, Santos IS. 2018.
Consumption of ultra-processed foods and body fat dur-
ing childhood and adolescence: a systematic review.
Public Health Nutr. 21(1):148–159.
Davidou S, Christodoulou A, Fardet A, Frank K. 2020. The
holistico-reductionist Siga classification according to the
degree of food processing: an evaluation of ultra-proc-
essed foods in French supermarkets . Food Funct. 11(3):
2026–2039.
Davidou S, Christodoulou A, Frank K, Fardet A. 2021. A
study of ultra-processing marker profiles in 22,028 pack-
aged ultra-processed foods using the Siga classification. J
Food Comp Anal. 99:103848.
Davidou S, Frank K, Christodoulou A, Fardet A. 2021.
Organic food retailing: to what extent are foods processed
and do they contain markers of ultra-processing? Int J
Food Sci Nutr. 73:172–183.
De Temmerman J, Heeremans E, Slabbinck H, Vermeir I.
2021. The Impact of the Nutri-Score Nutrition Label on
Perceived Healthiness and Purchase Intentions. Appetite.
157:104995.
Defago D, Geng JF, Molina O, Santa Mar
ıa D. 2020. Can
Traffic Light nutritional labels induce healthier consumer
choices? Experimental evidence from a developing coun-
try. Int J Consum Stud. 44(2):151–161. Ahead of print:.
Dekker LH, Vinke PC, Riphagen IJ, Minovi
c I, Eggersdorfer
ML, van den Heuvel EGHM, Schurgers LJ, Kema IP,
Bakker SJL, Navis G. 2019. Cheese and Healthy Diet:
Associations With Incident Cardio-Metabolic Diseases
and All-Cause Mortality in the General Population. Front
Nutr. 6:185.
Desquilbet M, Maign
e E, Monier-Dilhan S. 2018. Organic
Food Retailing and the Conventionalisation Debate. Ecol
Econs. 150:194–203.
Dickie S, Woods JL, Lawrence M. 2018. Analysing the use
of the Australian Health Star Rating system by level of
food processing. Int J Behav Nutr Phys Act. 15(1):128.
Du HD, Li LM, Bennett D, Guo Y, Turnbull I, Yang L,
Bragg F, Bian Z, Chen YP, Chen JS, China Kadoorie
Biobank study, et al. 2017. Fresh fruit consumption in
relation to incident diabetes and diabetic vascular compli-
cations: A 7-y prospective study of 0.5 million Chinese
adults. PLoS Med. 14(4):e1002279.
Dubois P, Albuquerque P, Allais O, Bonnet C, Bertail P,
Combris P, Lahlou S, Rigal N, Ruffieux B, Chandon P.
2021. Effects of front-of-pack labels on the nutritional
quality of supermarket food purchases: evidence from a
large-scale randomized controlled trial. J of the Acad
Mark Sci. 49(1):119–138.
Ebner P, Davidou S, Christodoulou A, Frank K.
Compl
ementarit
e entre l’indice Siga et le Nutri-Score
pour refl
eter le potentiel sant
e des aliments. Proceedings
of the JFN (Journ
ees Francophone de Nutrition); 2019.
Rennes: France.
Egnell M, Ducrot P, Touvier M, All
es B, Hercberg S, Kesse-
Guyot E, Julia C. 2018. Objective understanding of Nutri-
Score Front-Of-Package nutrition label according to indi-
vidual characteristics of subjects: Comparisons with other
format labels. PLOS One. 13(8):e0202095.
Elizabeth L, Machado P, Zin€
ocker M, Baker P, Lawrence M.
2020. Ultra-Processed Foods and Health Outcomes: A
Narrative Review. Nutrients. 12(7):1955.
Ezzati M, Bentham J, Di Cesare M, Bilano V, Bixby H,
Zhou B, Stevens GA, Riley LM, Taddei C, Hajifathalian
K, et al. 2017. Worldwide trends in body-mass index,
underweight, overweight, and obesity from 1975 to 2016:
a pooled analysis of 2416 population-based measurement
studies in 128.9 million children, adolescents, and adults.
Lancet. 390:2627–2642.
Fardet A, Lakhssassi S, Briffaz A. 2018. Beyond nutrient-
based food indices: a data mining approach to search for
a quantitative holistic index reflecting the degree of food
processing and including physicochemical properties .
Food Funct. 9(1):561–572.
Fardet A, Rock E. 2018. Perspective: Reductionist Nutrition
Research Has Meaning Only within the Framework of
Holistic and Ethical Thinking. Adv Nutr. 9(6):655–670.
Fardet A, Rock E. 2019. Ultra-processed foods: a new holis-
tic paradigm? Trends Food Sci Technol. 93:174–184.
Fardet A, Rock E. 2020. Exclusive reductionism, chronic
diseases and nutritional confusion: degree of processing
as a lever for improving public health. Crit Rev Food Sci
Nutr. 62(10): 2784–2799. doi:10.1080/10408398.2020.
1858751.
Fardet A, Rock E. 2022. Chronic diseases are first associated
with the degradation and artificialization of food matrices
rather than with food composition: calorie quality mat-
ters more than calorie quantity. Eur J Nutr. doi:10.1007/
s00394-021-02786-8 .
Fardet A. 2015. Wheat-based foods and non celiac gluten/
wheat sensitivity: is drastic processing the main key issue?
Med Hypotheses. 85(6):934–939.
Fardet A. 2016. Minimally processed foods are more satiat-
ing and less hyperglycemic than ultra-processed foods: a
preliminary study with 98 ready-to-eat foods. Food
Funct. 7(5):2338–2346.
Fardet A. 2017. L’effet matrice des aliments, un nouveau
concept. Prat Nutr. 13(52):37–40.
Fardet A. 2018. Chapter 3 - Characterization of the degree
of food processing in relation with its health potential
and effects. Adv Food Nutr Res. 85:79–121.
F
ed
eration Internationale du Diab
ete. 2019. Atlas du diab
ete
de la FID - 2019. 9
eme
ed. Brussels, Belgium:
International Diabetes Federation.
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION 13
Foster-Powell K, Holt SH, Brand-Miller JC. 2002.
International table of glycemic index and glycemic load
values: 2002. Am J Clin Nutr. 76(1):5–56.
Foster-Powell K, Miller JB. 1995. International tables of gly-
cemic index. Am J Clin Nutr. 62(4):871S–890S.
Fuentes S, Mandereau-Bruno L, Regnault N, Bernillon P,
Bonaldi C, Cosson E, Fosse-Edorh S. 2020. Is the type 2
diabetes epidemic plateauing in France? A nationwide
population-based study. Diabetes Metab. 46(6):472–479.
Gibney MJ. 2019. Ultra-Processed Foods: Definitions and
Policy Issues. Curr Dev Nutr. 3(2):nzy077.
Gonz
alez-Castell D, Gonz
alez-Coss
ıo T, Barquera S, Rivera
JA. 2007. Contribution of processed foods to the energy,
macronutrient and fiber intakes of Mexican children aged
1 to 4 years. Salud Publica Mex. 49:345–356.
Graham DJ, Heidrick C, Hodgin K. 2015. Nutrition Label
Viewing during a Food-Selection Task: Front-of-Package
Labels vs Nutrition Facts Labels. J Acad Nutr Diet.
115(10):1636–1646.
Hastak M, Mitra A, Ringold DJ. 2020. Do consumers view
the nutrition facts panel when making healthfulness
assessments of food products? Antecedents and conse-
quences. J Consum Aff. 54(2):395–416. Ahead of print:
https://doi.org/10.1111/joca.12301.
Hemmingsson E, Ekblom O, Kallings LV, Andersson G,
Wallin P, Soderling J, Blom V, Ekblom B, Ekblom-Bak E.
2021. Prevalence and time trends of overweight, obesity
and severe obesity in 447,925 Swedish adults, 1995–2017.
Scand J Public Health. 49(4):377–383.
Hjerpsted J, Tholstrup T. 2016. Cheese and Cardiovascular
Disease Risk: A Review of the Evidence and Discussion
of Possible Mechanisms. Crit Rev Food Sci Nutr. 56(8):
1389–1403.
Hofman DL, Van Buul VJ, Brouns FJPH. 2016. Nutrition,
Health, and Regulatory Aspects of Digestible
Maltodextrins. Crit Rev Food Sci Nutr. 56(12):2091–2100.
Imamura F, O’Connor L, Ye Z, Mursu J, Hayashino Y,
Bhupathiraju SN, Forouhi NG. 2016. Consumption of
sugar sweetened beverages, artificially sweetened bever-
ages, and fruit juice and incidence of type 2 diabetes: sys-
tematic review, meta-analysis, and estimation of
population attributable fraction. Br J Sports Med. 50(8):
496–504.
Inoue Y, Qin B, Poti J, Sokol R, Gordon-Larsen P. 2018.
Epidemiology of Obesity in Adults: Latest Trends. Curr
Obes Rep. 7(4):276–288.
International Diabetes Federation. 2019. Australia: Diabetes
report 2010 - 2045 International Diabetes Federation;
Julia C, Touvier M, Mejean C, Ducrot P, Peneau S,
Hercberg S, Kesse-Guyot E. 2014. Development and val-
idation of an individual dietary index based on the
British Food Standard Agency nutrient profiling system
in a French context. J Nutr. 144(12):2009–2017.
Kucek LK, Veenstra LD, Amnuaycheewa P, Sorrells ME.
2015. A Grounded Guide to Gluten: How Modern
Genotypes and Processing Impact Wheat Sensitivity.
Compr Rev Food Sci Food Saf. 14(3):285–302.
Lane MM, Davis JA, Beattie S, G
omez-Donoso C,
Loughman A, O’Neil A, Jacka F, Berk M, Page R, Marx
W, et al. 2020. Ultraprocessed food and chronic
noncommunicable diseases: A systematic review and
meta-analysis of 43 observational studies. Obes Rev. 22:
e13146.
Mach
ın L, Aschemann-Witzel J, Curutchet MR, Gim
enez A,
Ares G. 2018. Traffic Light System Can Increase
Healthfulness Perception: Implications for Policy Making.
J Nutr Educ Behav. 50(7):668–674.
Maganja D, Buckett K, Stevens C, Flynn E. 2019. Consumer
choice and the role of front-of-pack labelling: the Health
Star Rating system. Public Health Res Pract. 29(1):
2911909.
Mandrioli D, Kearns CE, Bero LA. 2016. Relationship
between Research Outcomes and Risk of Bias, Study
Sponsorship, and Author Financial Conflicts of Interest
in Reviews of the Effects of Artificially Sweetened
Beverages on Weight Outcomes: A Systematic Review of
Reviews. PLoS One. 11(9):e0162198.
Martinez-Perez C, San-Cristobal R, Guallar-Castillon P,
Mart
ınez-Gonz
alez M
A, Salas-Salvad
o J, Corella D,
Casta~
ner O, Martinez JA, Alonso-G
omez
AM, W€
arnberg
J, et al. 2021. Use of Different Food Classification
Systems to Assess the Association between Ultra-
Processed Food Consumption and Cardiometabolic
Health in an Elderly Population with Metabolic
Syndrome (PREDIMED-Plus Cohort). Nutrients. 13(7):
2471.
Black MM, Hurley KM. 2013. How to help children develop
healthy eating habits. In: Tremblay RE BM, Peters RDeV,
editor editors. Encyclopedia on Early Childhood
Development. (2nd ed). Faith MS. Qu
ebec, Canada:
Montr
eal.
Merema M, O’Connell E, Joyce S, Woods J, Sullivan D.
2019. Trends in body mass index and obesity prevalence
in Western Australian adults, 2002 to 2015. Health
Promot J Austr. 30(1):60–65.
Mertens E, Colizzi C, Pe~
nalvo JL. 2022. Ultra-processed
food consumption in adults across Europe. Eur J Nutr.
61(3):1521–1539. Online ahead of print:
Ministerio de Salud - Gobierno de Chile. 2021.: https://
www.minsal.cl/ley-de-alimentos-manual-etiquetado-nutri-
cional/. Accessed 21 December
Monteiro CA. 2009. Nutrition and health. The issue is not
food, nor nutrients, so much as processing. Public Health
Nutr. 12(5):729–731.
Monteiro CA, Cannon G, Levy RB, Moubarac J-C, Louzada
MLC, Rauber F, Khandpur N, Cediel G, Neri D,
Martinez-Steele E, et al. 2019. Ultra-processed foods:
what they are and how to identify them? Public Health
Nutr. 22(5):936–941.
Monteiro CA, Moubarac J-C, Levy RB, Canella DS, Louzada
MLdC, Cannon G. 2018. Household availability of ultra-
processed foods and obesity in nineteen European coun-
tries. Public Health Nutr. 21(1):18–26.
Monteiro CA, Cannon G, Lawrence M, Louzada MLdC,
Machado PP, 2019. Ultra-processed foods, diet quality,
and health using the NOVA classification system. Rome,
Italy: FAO.
Mørk T, Grunert KG, Fenger M, Juhl HJ, Tsalis G. 2017.
An analysis of the effects of a campaign supporting use
of a health symbol on food sales and shopping behaviour
of consumers. BMC Public Health. 17(1):239.
14 P. EBNER ET AL.
Moubarac J-C, Parra DC, Cannon G, Monteiro CA. 2014.
Food Classification Systems Based on Food Processing:
Significance and Implications for Policies and Actions: A
Systematic Literature Review and Assessment. Curr Obes
Rep. 3(2):256–272.
Naimi S, Viennois E, Gewirtz AT, Chassaing B. 2021.
Direct impact of commonly used dietary emulsifiers on
human gut microbiota. Microbiome. 9(1):66.
Nobrega L, Ares G, Deliza R. 2020. Are nutritional warn-
ings more efficient than claims in shaping consumers’
healthfulness perception? Food Qual Pref. 79:103749.
Nordic Council of Ministers. 2010. The Keyhole: Healthy
choices made easy - Partnership, Synergies, Activities,
Future. Copenhagen, Denmark: Nordic Council of
Ministers
Olivier B, Serge AH, Catherine A, Jacques B, Murielle B,
Marie-Chantal CL, Sybil C, Jean-Philippe G, Sabine H,
Esther K, et al. 2015. Review of the nutritional benefits
and risks related to intense sweeteners. Arch Public
Health. 73:41.
Pagliai G, Dinu M, Madarena MP, Bonaccio M, Iacoviello
L, Sofi F. 2021. Consumption of ultra-processed foods
and health status: a systematic review and meta-analysis.
Br J Nutr. 125(3):308–318.
Palmn€
as MS, Cowan TE, Bomhof MR, Su J, Reimer RA,
Vogel HJ, Hittel DS, Shearer J. 2014. Low-dose aspartame
consumption differentially affects gut microbiota-host
metabolic interactions in the diet-induced obese rat.
PLoS One. 9(10):e109841.
Pimpin L, Wu JHY, Haskelberg H, Del Gobbo L,
Mozaffarian D. 2016. Is Butter Back? A Systematic
Review and Meta-Analysis of Butter Consumption and
Risk of Cardiovascular Disease, Diabetes, and Total
Mortality. Plos One. 11(6):e0158118.
Poti JM, Mendez MA, Ng SW, Popkin BM. 2015. Is the
degree of food processing and convenience linked with
the nutritional quality of foods purchased by US house-
holds? Am J Clin Nutr. 101(6):1251–1262.
Rasmussen M, Damsgaard MT, Morgen CS, Kierkegaard L,
Toftager M, Rosenwein SV, Krolner RF, Due P, Holstein
BE. 2020. Trends in social inequality in overweight and
obesity among adolescents in Denmark 1998-2018. Int J
Public Health. 65(5):607–616.
Reyes M, Garmendia ML, Olivares S, Aqueveque C,
Zacar
ıas I, Corval
an C. 2019. Development of the
Chilean front-of-package food warning label. BMC Public
Health. 19(1):906.
Sadler CR, Grassby T, Hart K, Raats M, Sokolovi
cM,
Timotijevic L. 2021. Processed food classification:
Conceptualisation and challenges. Trends Food Sci
Technol. 112:149–162.
Sanchez-Siles LM, Michel F, Rom
an S, Bernal MJ, Philipsen
B, Haro JF, Bodenstab S, Siegrist M. 2019. The Food
Naturalness Index (FNI): An integrative tool to measure
the degree of food naturalness. Trends Food Sci Technol.
91:681–690.
Schernhammer ES, Bertrand KA, Birmann BM, Sampson L,
Willett WC, Feskanich D. 2012. Consumption of artificial
sweetener- and sugar-containing soda and risk of
lymphoma and leukemia in men and women. Am J Clin
Nutr. 96(6):1419–1428.
Scrinis G, Monteiro CA. 2018. Ultra-processed foods and
the limits of product reformulation. Public Health Nutr.
21(1):247–252.
Scrinis G. 2016. Reformulation, fortification and functionali-
zation: Big Food corporations’nutritional engineering
and marketing strategies. J Peas Stu. 43(1):17–37.
Scrinis G. 2013. Nutritionism - The Science and Politics of
Dietary Advice. New York (NY): Columbia University
Press.
Singh V, Vijay-Kumar M. 2020. Beneficial and detrimental
effects of processed dietary fibers on intestinal and liver
health: health benefits of refined dietary fibers need to be
redefined!. Gastroenterol Rep (Oxf)). 8(2):85–89.
Slimani N, Deharveng G, Southgate DA, Biessy C, Chaj
es
V, van Bakel MM, Boutron-Ruault MC, McTaggart A,
Grioni S, Verkaik-Kloosterman J, et al. 2009.
Contribution of highly industrially processed foods to the
nutrient intakes and patterns of middle-aged populations
in the European Prospective Investigation into Cancer
and Nutrition study. Eur J Clin Nutr. 63 Suppl 4(Suppl
4):S206–S225.
Soffritti M, Padovani M, Tibaldi E, Falcioni L, Manservisi F,
Belpoggi F. 2014. The carcinogenic effects of aspartame:
The urgent need for regulatory re-evaluation. Am J Ind
Med. 57(4):383–397.
Srour B, Fezeu LK, Kesse-Guyot E, All
es B, M
ejean C,
Andrianasolo RM, Chazelas E, Deschasaux M, Hercberg
S, Galan P, et al. 2019. Ultra-processed food intake and
risk of cardiovascular disease: prospective cohort study
(NutriNet-Sant
e). BMJ. 365:l1451.
Suez J, Korem T, Zeevi D, Zilberman-Schapira G, Thaiss
CA, Maza O, Israeli D, Zmora N, Gilad S, Weinberger A,
et al. 2014. Artificial sweeteners induce glucose intoler-
ance by altering the gut microbiota. Nature. 514(7521):
181–186.
Tobias DK, Hall KD. 2021. Eliminate or reformulate ultra-
processed foods? Biological mechanisms matter. Cell
Metab. 33(12):2314–2315.
U.K. Department of Health, Food Standard Agency,
Llywodraeth Cymru Welsh Government, Food Standard
Scotland. 2016. Guide to creating a front of pack (FoP)
nutrition label for pre-packed products sold through
retail outlets. Report.
U.S. Department of Health and Human Services, Center for
Disease Control and Prevention. 2020. National Diabetes
Statistics Report - Estimates of Diabetes and Its Burden
in the United States.
Visioli F, Franco M, Mart
ınez-Gonz
alez M
A. 2021. Front of
package labels and olive oil: a call for caution. Eur J Clin
Nutr. doi:10.1038/s41430-021-00989-0.
Visioli F, Marangoni F, Poli A, Ghiselli A, Martini D. 2022.
Nutrition and health or nutrients and health? Int J Food
Sci Nutr. 73(2):141–148.
Vos T, Lim SS, Abbafati C, Abbas KM, Abbasi M,
Abbasifard M, Abbasi-Kangevari M, Abbastabar H, Abd-
Allah F, Abdelalim A, et al. 2020. Global burden of 369
diseases and injuries in 204 countries and territories,
INTERNATIONAL JOURNAL OF FOOD SCIENCES AND NUTRITION 15
1990–2019: a systematic analysis for the Global Burden
of Disease Study 2019. Lancet. 396(10258):1204–1222.
Wang DD, Li Y, Afshin A, Springmann M, Mozaffarian D,
Stampfer MJ, Hu FB, Murray CJL, Willett WC. 2019.
Global Improvement in Dietary Quality Could Lead to
Substantial Reduction in Premature Death. J Nutr.
149(6):1065–1074.
WHO 2015. Sugars intake for adults and children -
Guideline. Department of Nutrition for Health and
Development, World Health Organization. 20, Avenue
Appia, CH-1211 Geneva 27, Switzerland.
WHO 2020. The double burden of malnutrition: priority
actions on ending childhood obesity. Report.
Wong MCS, Huang JJ, Wang JX, Chan PSF, Lok V, Chen
X, Leung C, Wang HHX, Lao XQ, Zheng ZJ. 2020.
Global, regional and time-trend prevalence of central
obesity: a systematic review and meta-analysis of 13.2
million subjects. Eur J Epidemiol. 35(7):673–683.
16 P. EBNER ET AL.
