ArticlePDF Available

Nutrient profiles: Options for definitions for use in relation to food promotion and children's diets Final report

Authors:
Nutrient profiles: Options for
definitions for use in relation to food
promotion and children’s diets
Final report
Mike Rayner, Peter Scarborough and Lynn Stockley
British Heart Foundation Health Promotion Research
Group, Department of Public Health, University of Oxford
October 2004
2
Contents
Executive summary.................................................................................. 5
Summary................................................................................................... 7
Background................................................................................................................7
Role of the Expert Group...........................................................................................7
Methodology..............................................................................................................8
Conclusions and recommendations for future work................................................13
References................................................................................................................13
Section 1 - Review of Relevant National and International Literature
1.1 Summary...............................................................................................................16
1.2 Objectives..............................................................................................................17
1.3 Methods.................................................................................................................17
1.4 Findings.................................................................................................................18
1.4.1 Dietary guidance for children .........................................................................18
1.4.2 Relevant definitions........................................................................................20
1.4.3 Vending machines...........................................................................................21
1.4.4 School lunches................................................................................................23
1.4.5 Advertising and marketing foods to children..................................................27
1.4.6 Point of purchase schemes..............................................................................29
1.5 Discussion .............................................................................................................31
1.5.1 Correspondence between schemes identified in this review and a theoretical
approach to developing nutrient profiles .................................................................31
1.5.2 Evaluation of nutrient profiling schemes........................................................32
1.6 References.............................................................................................................33
Section 2 - A Theoretical Approach to Developing Definitions
2.1 Summary...............................................................................................................36
2.2 Introduction..........................................................................................................37
2.2.1 Age groups......................................................................................................37
2.2.2 The intended uses of definitions....................................................................38
2.2.3 Stages in developing nutrient profiles.............................................................39
2.3 Stage 1. Choice of nutrients and other food components................................39
2.3.1 Prioritising nutrients and other food components...........................................39
2.3.2 Fortification.....................................................................................................43
2.3.3 Combining criteria for nutrients and other components.................................44
2.4 Stage 2. Choice of base........................................................................................44
2.4.1 Pros and cons of different bases......................................................................45
2.4.2 Combining bases.............................................................................................46
2.5 Stage 3. Choice of model type ............................................................................50
2.5.1 Stage 3a: Thresholds models, simple scoring system or complex scoring
systems?...................................................................................................................50
2.5.2 Stage 3b: Food category specific criteria or across the board criteria? ..........55
2.6 Stage 4. Choice of numbers................................................................................57
2.7 References.............................................................................................................59
3
Section 3 - Development and Testing of Possible Definitions I
3.1 Summary...............................................................................................................60
3.2. Introduction.........................................................................................................60
3.2.1 Recommendations of the Expert Group..........................................................60
3.2.2 Consequences of the Expert Group’s recommendations................................62
3.3 Methods for testing definitions...........................................................................64
3.3.1 Development of a database of foods...............................................................65
3.3.2 Categorisation of foods in the database into Balance of Good Health food
groups and selection of indicators for food groups..................................................66
3.3.3 Development of ‘healthier’ and ‘less healthy’ indicator foods.......................67
3.4 Results...................................................................................................................69
3.4.1 Testing definitions of ‘foods high in fat, salt or sugar’...................................69
3.4.2 Testing complete models including definitions of ‘foods high in fat, salt or
sugar’, ‘intermediate foods’ and ‘healthier food choices’.......................................71
3.5 Conclusions and recommendations of the Expert Group................................74
3.6 References.............................................................................................................75
Section 4 - Development and Testing of Possible Definitions II
4.1 Summary...............................................................................................................77
4.2 Introduction..........................................................................................................78
4.3 Methods.................................................................................................................80
4.4 Results..................................................................................................................83
4.4.1 Choice of thresholds .......................................................................................83
4.4.2 Choice of energy/fat criteria ...........................................................................84
4.4.3 Choice of nutrient group and of model type..................................................84
4.4.4 Drink and fibre modifications.........................................................................85
4.5 Conclusions and recommendations....................................................................85
4.6 References.............................................................................................................87
Section 5 - Recommendations for Further Testing of Definitions
5.1 Development of databases of foods relevant to the purpose............................88
5.2 Further development of panels of indicator foods and of tests of sensitivity
and specificity.............................................................................................................88
5.3 Assessing the acceptability of schemes...............................................................89
5.4 References.............................................................................................................89
Glossary.......................................................................................................................90
Appendix 1. Literature review: Summary tables........................................................92
Appendix 2. The Coronary Prevention Group Nutrition Banding Scheme.............103
Appendix 3a. Lists of ‘Healthier’ and ‘Less healthy’ food indicators.......................106
Appendix 3b: Nutritional composition of Balance of Good Health indicators .........108
4
Appendix 4. Details of models and definitions..........................................................111
Appendix 5 - results of testing 28 definitions of ‘foods high in fat, salt or sugar’....119
Appendix 6 – Results of eight models testing definitions of ‘foods high in fat, salt or
sugar’, ‘intermediate foods’ and ‘healthier food choices’.........................................149
Appendix 7. Results of testing twelve further models and possible modifications
EXECUTIVE SUMMARY
Diet and obesity are important factors in determining risk of cancer, stroke and
coronary heart disease. Health promotion targeted solely at individuals has not been
successful in arresting the ongoing rise in prevalence of obesity, and this has led the
Food Standards Agency to consider new approaches which have the potential to
reinforce healthy eating advice, and make it easier for consumers to make healthy
choices. The Agency’s Action Plan on Food Promotions and Children’s Diets, for
example, seeks to address the imbalance in the way foods are currently promoted to
children and their carers.
Although many would argue that the overall balance of the diet is more important
than the individual foods consumed, it is the “imbalance” in the consumption of
individual foods that can contribute to health problems. In addition, dietary surveys
provide clear evidence of the areas in which children’s diets in particular need to
improve, i.e., by reducing their consumption of fat (especially saturated fat), salt and
sugar, and increasing their consumption of fruit and vegetables. It is therefore
appropriate to consider nutrient profiling, which can be defined as “the science of
categorising foods according to their nutritional composition”, to enable interventions
that differentiate between foods on this basis. Various nutrient profiling systems have
been developed throughout the world, and applied in a variety of consumer
information and regulatory contexts. However, there is often a lack of detail available
about the criteria underpinning these systems, or the scientific rationale on which they
have been based.
The purpose of this project was to develop a nutrient profiling model to support the
Agency's work to redress the current imbalance in the way foods are currently
promoted to children. This will include advice on nutrition and health claims aimed
specifically at children, and advice on the balance of TV advertising for foods during
children’s programming. The work was based on existing Government healthy eating
advice, and built on modelling work carried out in connection with the Department of
Health’s 5 A DAY initiative. The models developed focused on children aged from
11 to 16, although they are likely to be applicable to other age groups, and work
extending the principle is this way is in hand.
The work was overseen by an Expert Group, comprising nutrition scientists;
dieticians; food industry and consumer organisation representatives; and policy
makers. The research took a systematic approach to developing models, taking
account of public health recommendations, basing criteria on Guideline Daily
Amounts and Dietary Reference Values. The expert group assessed the success of
each model on its overall performance against three factors:
6
Ÿ A statistical test of accuracy, based on the model correctly classifying indicator
panels of approximately 200 “healthier” and “less healthy” foods.
Ÿ The proportion of a database of around 1000 foods, that was classified by the
model as “less healthy”, “intermediate”, or “healthier”.
Ÿ The Expert Group’s qualitative assessment of how the models categorised
approximately 100 key “example indicator foods”, representing the food groups
on which healthy eating advice (the Balance of Good Health) is based.
The report recommends one model that, with further refinement, could form the basis
of a workable system. This is a scoring model that takes account of energy, saturated
fat, non-milk extrinsic sugars, and sodium; and the degree to which these nutrients are
balanced by calcium, iron, long chain n-3 polyunsaturated fatty acids, and fruit and
vegetable content. The model therefore identifies foods high in fat, salt or sugar,
while recognising the important contribution of dairy, meat, fish, and fruit and
vegetable based products to a balanced diet. The flexibility provided by the scoring
system means that the model could be adapted to suit a range of applications.
Further work will be required to refine the model; to test it against a wider range of
foods; and to consider how it might be applied in practice and communicated to
stakeholders. The expert group was particularly keen for the model to be tested
against a wider range of foods. The report therefore recommends:
Ÿ the development of a database of food composition data with which to test the
preferred model;
Ÿ devising further panels of ‘healthier’, ‘intermediate’ and ‘less healthy’ indicator
foods, across a wider range of food groups;
Ÿ considering what modifications to the proposed model would be necessary for its
use for other age groups;
Ÿ assessing the acceptability of the model to both experts and consumers; and
considering communication and support issues for consumers, health and other
professionals, and the food industry as appropriate.
In summary, this work represents a significant step forward beyond nutrient profiling
previously used in the UK or elsewhere. The model that has been developed has
potential for use to underpin a range of interventions, including some involving
consumer information and public health messages. The approach used delivers the
opportunity to encourage product formulation and innovation with public health
benefits. Although further work on the model will be required before it is ready for
use in these contexts, the results of this work demonstrate that nutrient profiling can
be made to work in practice.
Nutrient profiles for food promotion to children: Summary
7
SUMMARY
Background
1. The type of foods eaten and the frequency with which they are consumed is
associated to varying degrees with several chronic diseases, including
cardiovascular disease, Type 2 diabetes, some types of cancer and dental caries
and also the precursors to these diseases – in particular obesity and hypertension.
The continuing high levels and, in some cases, the increasing rates of diet-related
diseases have recently been highlighted in several high-profile reports, including a
report on public health for the Prime Minister, the Secretary of State for Health
and the Chancellor of the Exchequer (Wanless, 2004).
2. It is recognised that the causes of chronic disease are complex, and actions to
tackle them must be multifactorial, but a key area for action, highlighted in many
recent reports on public health in the UK, is the promotion of foods to children
(e.g. House of Commons Health Committee, 2004).
3. The effect of food promotion on the choice of foods by children is difficult to
assess. Food choice a complex area where it is not easy to measure the effect of
single factors and the interaction between multiple determinants. In order to try
and shed some light on this, the Food Standards Agency (FSA) commissioned a
report to review all of the research carried out to date on the effects of food
promotion to children. This report concluded there is sufficient evidence that food
promotion does have an effect, particularly on children’s preferences, purchase
behaviour and consumption (Hastings et al., 2003).
4. On this basis the FSA agreed an Action Plan on Food Promotion and Children’s
Diets (Food Standards Agency, 2004). This includes developing advice and
guidelines for the food industry on reducing amounts of fat, salt and sugar in
products specifically aimed at children, and agreeing guidelines on the labelling of
these products to enable consumers to identify more easily and accurately what
are healthier options.
5. Following on from this, the FSA commissioned this work, which aims:
(a) To identify a range of options for nutrient profiles that could be used to
define ‘healthier food choices’ and ‘foods high in fat, salt or sugar’ in
relation to specific aspects of the promotion of food to children.
(b) To assess how these nutrient profiles would apply to a range of foods.
Role of the Expert Group
6. The work was overseen throughout by an expert group of nutrition scientists;
representatives from industry and a consumer organisation; and policy makers.
The membership of the expert group is given on page 15.
Nutrient profiles for food promotion to children: Summary
8
7. The role of group was to oversee the work and provide specialist input. The
direction of the work was determined on the basis of their discussions and
consensus views. The main body of the report provides a more detailed discussion
of the rationale behind the decisions taken, and the alternatives considered.
Methodology
8. The work consisted of three broad phases:
Ÿ Phase 1: Background work including a literature review to identify existing
use of nutrient profiles in relation to food promotion to children and the
development of a theoretical approach to developing nutrient profiles
(Sections 1 and 2).
Ÿ Phase 2: Discussions by the Expert Group to determine the focus of the work
(Section 2).
Ÿ Phase 3: Initial testing and subsequent refinement of definitions and models
(Sections 3 and 4).
Phase 1: Background work including a literature review to identify existing use
of nutrient profiles in relation to food promotion to children and the
development of a theoretical approach to developing nutrient profiles and
(Sections 1 and 2).
Review of existing practice (Section 1)
9. The review found the following
Ÿ Food labelling legislation in both the USA and Canada contains definitions based
on nutrient criteria.
Ÿ The majority of the work that was identified on nutrient profiling for use in
vending machine policies was from the USA, and examples of some detailed
guidance are given.
Ÿ Various approaches have been adopted to healthier school lunch provision
including: checklists for use by caterers; nutrient assessments of menus provided
over a period of time; lists of recommended foods; and the use of nutrient profiles
to identify foods which can contribute to healthier food provision.
Ÿ Over the last few years there have been various initiatives around the world to
raise awareness of the issue of marketing of foods to children. Generally speaking
these propose either banning all food advertising targeted at certain age groups of
children, or banning or restricting the marketing and advertising of ‘unhealthy’ or
‘junk’ foods. There is usually very little attempt to define what is meant by these
terms, although specific foods are often mentioned. Restrictions are most
Nutrient profiles for food promotion to children: Summary
9
frequently applied to products such as chocolate, sweets, soft drinks, snacks and
‘other similar products’.
Ÿ Nutrient criteria are used in some countries to assess eligibility for health claims,
addition of nutrients, and also in public health point of purchase schemes. Many
retailers in the United Kingdom also run healthier choice schemes, to highlight
products that meet certain nutrient criteria. These schemes are often for the
general population, rather than being intended for children. However, several UK
retailers do have ranges of foods specifically for children, and most of these
specify that the foods contain ‘controlled levels of sugar, fat and salt’, as well as
restricting the use of additives. Some manufacturers of children’s foods, and
organisations who licence characters for use on products, have adopted a similar
approach.
10. Existing schemes have the following features:
Ÿ Choice of nutrients: The nutrients selected for developing nutrient profiles almost
always include total fat and sodium. Saturated fat is often included. Sugars are
included less frequently. Energy is also sometimes included.
Ÿ Several of the schemes take account of nutrients that are eaten at lower than the
recommended intake. Fibre is the most frequently selected, followed by calcium.
Some schemes, particularly those from North America, include iron, vitamin C,
vitamin A, and protein. Many schemes make specific provisions which promote
foods containing fruits and vegetables.
Ÿ Choice of base: The most common choice of base is ‘per serving’, although ‘per
100g’ is also used frequently. An energy base (e.g. per 100kJ) is used
infrequently.
Ÿ Choice of model type: A threshold model is used most commonly, with only one
or two schemes attempting to use scoring systems. The schemes are evenly split
between those which use an ‘across the board’ or a ‘food category’ approach.
Ÿ Choice of thresholds or scoring criteria: These are sometimes based on public
health recommendations, particularly for vitamins and minerals, but for many
schemes it is not possible to identify the reasons for the choice of numbers, and it
is assumed that it is often a pragmatic decision.
A theoretical approach to developing nutrient profiles (Section 2)
11. The development of nutrient profiles necessarily involves a number of stages.
These stages can be approached in any order, and decisions at any one stage affect
decisions at others. However the most logical order is shown in the figure below:
Nutrient profiles for food promotion to children: Summary
10
Stages in developing profiles
Stage 1 Stage 2 Stage 3 Stage 4
Ÿ Choice of nutrients: There are a number of different nutrients and other food
components that could possibly be used in nutrient profiles. Rationales for choice
of nutrients and components are described in detail in Section 2.
Ÿ Choice of base: There are three basic ways of setting nutrient profiles: per 100g,
per 100kJ and per serving. The strengths and weaknesses of each method are
discussed in Section 2.
Ÿ Choice of model type: There are three different options for model types that can
be used for nutrient profiling; threshold models, simple scoring systems, and
complex scoring systems. Once the type of model is chosen it is necessary to
choose between food category specific or across the board criteria. Again there is
an in depth discussion of the advantages and disadvantages of each approach in
Section 2.
Ÿ Choice of numbers: The levels set for the thresholds of the individual nutrient
criteria (or points scored for a particular level) can be pragmatically chosen, taken
from respected sources, or linked to public health recommendations. The ways in
which this can be done are described further in Section 3.
Phase 2: Discussions by the Expert Group to determine the focus of the work
(Section 2)
12. Ages of children: The project focused on the development and testing of
definitions for ‘foods high in fat, salt or sugar’ and ‘healthier food choices’ for
children aged 11-16.1 It was noted that the principles would be applicable to
children of any age between 5 and 16 and also to adults.
1 For many of the nutrients relevant to this work, dietary recommendations are set as a proportion of
energy intake, and are therefore the same for children over 5 as for adults. However, recommendations
for iron and calcium intake are higher for 11-16 year olds because of the body’s increased requirements
for these nutrients during puberty. This age group was therefore chosen in order to ensure that the
definitions were relevant to the group with the greatest requirement for these nutrients.
Choice of
nutrients and
other food
components
Choice of
base (e.g. per
100g, per
100kJ, per
serving)
Choice of
model type Choice of
numbers (i.e.
how much is
desirable or
undesirable?)
Nutrient profiles for food promotion to children: Summary
11
13. The intended uses of definitions and models: The project concentrated on
developing and testing definitions of ‘foods high in fat, salt or sugar’ and
‘healthier food choices’ for the purpose of:
Ÿ Advice on the use of nutrition and health claims on foods aimed specifically at
children; and
Ÿ Advice on the balance of TV advertising for foods during children’s TV
programmes.
But the potential use in signposting on labels and menus would also be taken into
account.
14. The Expert Group agreed that models should be based on the following guiding
principles:
15. Choice of nutrients: Simple definitions involving only criteria for energy,
saturated fat, non-milk extrinsic (NME) sugar and sodium should be developed
and tested (for the sake of brevity here called ‘A’ nutrients), but that more
complex definitions involving these nutrients together with criteria for fruit and
vegetables, long chain n-3 polyunsaturated fatty acids (‘B’ nutrients), and calcium
and iron (‘C’ nutrients) should also be developed.
(Note: A cumulative approach was used such that definitions involving ‘B’
nutrients include ‘A’ nutrients, and definitions involving ‘C’ nutrients included
both ‘A’ and ‘B’ nutrients)
16. Fortification: Where definitions involved criteria for calcium, iron and n-3 fatty
acids then those criteria should in principle be for the levels of those nutrients
prior to any fortification, except where products had been fortified in line with
legal requirements. (However, in practice, because limited pre-fortification data
were available, model testing was based on final nutrient levels).
17. Choice of base: The base/denominators to be considered for definitions should
be: per 100g, per 100kJ, per 100g AND/OR per serving, per 100kJ AND/OR per
serving.
18. Choice of model type: Both threshold models and scoring systems should be
investigated. ‘Across-the-board’ rather than food ‘category specific’ criteria
should be investigated in the first instance.
19. Choice of numbers: The numbers used for thresholds or points scored should
bear a consistent and transparent relationship to public health recommendations.
Phase 3: Initial testing and subsequent refinement of definitions and models
(Sections 3 and 4)
Initial testing (Section 3)
20. Twenty-eight definitions of ‘foods high in fat, salt or sugar’ were developed on
the basis of the recommendations of the Expert Group. On the basis of testing the
28 definitions, eight definitions were selected for development into complete
Nutrient profiles for food promotion to children: Summary
12
models with definitions of ‘foods high in fat, salt or sugar’, ‘healthier food
choices’ and ‘intermediate foods’. The Expert Group agreed that three models
were worthy of further consideration and should be tested further and if possible
refined.
Further testing (Section 4)
21. This round of development and testing followed a more incremental approach
than the previous round, starting with a threshold model and making sequential
changes to this model. 12 models were developed on the basis of the
recommendations of the Expert Group following the initial testing, and were
tested further. Two possible modifications to the two most promising of these 12
models were also tested.
22. The models tested in this round were compared to each other when judged by
various ‘success’ criteria: including: simplicity and transparency; accuracy; and
distribution of foods between ‘high in fat, salt or sugar’ ‘intermediate’ and
‘healthier choice’ categories
23. As a result of this round of testing one model was identified for further
consideration: (Model SSCg3d). This used a simple scoring system, Group C
nutrients and a per 100g base, with a modification for drinks. (One member of the
Expert Group would have preferred a ‘per serving’ base, because of concerns that
choice of the 100g base would lead to anomalies for foods eaten in larger or
smaller portion sizes).
24. The scoring bands for foods are as follows (to a maximum of 10 points per
nutrient/food component):
Ÿ Energy: = 335kJ = 0; 335-670kJ = 1; 670-1005kJ = 2, etc.
Ÿ Saturated fat: = 1.0g = 0; 1.0-2.0g = 1; 1.0-2.0g = 2, etc.
Ÿ NME sugars: = 2.4g = 0; 2.4-4.8g = 1; 4.8-7.2g = 2, etc.
Ÿ Sodium: = 90mg = 0; 90-180mg = 1; 180-270mg = 2, etc.
Ÿ Calcium: = 105mg = 0; 105-210mg = 1; 210-315mg = 2, etc.
Ÿ Iron: = 1.5mg = 0; 1.5-3.0mg = 1; 3.0-4.5mg = 2, etc.
Ÿ n-3 fatty acids: = 0.05g = 0; 0.05-0.10g = 1; 0.10-0.15g = 2, etc.
Ÿ Fruit and vegetables: 0-30% = 0; 50% = 2; 70% = 4; 100% = 10. 2
The scoring bands for drinks are half the width of these bands. This is because
the high water content of such products means they tend to be less nutrient dense
per 100g.
Total score = A nutrients – B nutrients – C nutrients.
2 The Guideline Daily Amounts used to develop this model were: Energy: 2130 kcal, 8950 kJ;
Saturated fat: 26g; NME sugars: 63g; Sodium: 2.35 g; Calcium: 690mg; Iron: 10.1mg; n-3 fatty acids:
0.4g; Fruit and vegetables: 380g.
Nutrient profiles for food promotion to children: Summary
13
25. The food is defined as a ‘healthier food choice’ if the score is 2 or less, as
‘intermediate’ if the score is 3-8, and as a ‘food high in fat, salt or sugar’ if the
score is 9 or more.
26. Although scoring systems, such as Model SSCg3d, are seemingly more complex
than threshold models they are also more accurate. Scoring systems are also
more flexible than threshold models, making them more adaptable to a variety of
purposes. Scoring systems can more easily be used for comparing foods within
categories (i.e., because foods can be compared on the basis of their points score,
as well as their overall category, thus providing greater distinction between foods
with similar nutrient content).
Conclusions and recommendations for future work
27. Nutrient profiling attempts to develop objective criteria to define, for example,
‘foods high in fat, salt or sugar’ or ‘healthier food choices’. Expert judgements
about what constitutes a ‘food high in fat, salt or sugar’ or a ‘healthier food
choice’ are based on recommended quantities of nutrients, the form in which
nutrients are present in particular foods, considerations about frequency of
consumption, interactions between food components, bio availability, and many
other issues. It is difficult to capture these considerations in a mathematical model,
however this study has identified a model which goes some way towards doing so.
This significant step forward offers the prospect of development of a scheme
which could be used in a number of ways to tackle problems with children’s diets.
It is recommended that future work tests the preferred model against experts’
assessments of a wider range of foods.
28. The specific recommendations made for future work are:
Ÿ the development of a database of food composition data with which to test the
preferred model.
Ÿ devising a panel of ‘healthier’ and ‘less healthy’ indicator foods with input
from experts in practical food and nutrition.
Ÿ considering what modifications to the proposed model would be necessary for
its use for other age groups.
Ÿ assessing the acceptability of the model to both experts and consumers.
Ÿ considering communication and support issues for consumers, health and other
professionals, and the food industry as appropriate.
29. A glossary of terms used in this report can be found on page 90.
References
Food Standards Agency (2004) Agency agrees plan for an overhaul of the way food is
promoted to children - press release, action plan and consultation document. Food
Standards Agency, London.
Hastings G, Stead M, McDermott L, Forsyth A, MacKintosh A M, Rayner M,
Godfrey C, Caraher M & Angus K (2003) Review of Research on the effects of food
Nutrient profiles for food promotion to children: Summary
14
promotion to children. Prepared for the Food Standards Agency, London, by
University of Strathclyde, Glasgow.
House of Commons Health Committee (2004) Obesity: Third report of session 2003-
4. House of Commons, London.
Wanless D (2004) Securing Good Health for the Whole Population. HM Treasury and
Department of Health, London.
Nutrient profiles for food promotion to children: Summary
15
MEMBERSHIP OF EXPERT GROUP
Danila Armstrong Department of Health
Susan Jebb Medical Research Council Human Nutrition
Research Centre, Cambridge
Luci Daniels Independent Dietician
Judy Buttriss British Nutrition Foundation
Karen Tonks British Retail Consortium nominee
Gaynor Bussell Food and Drink Federation nominee
Sue Davies Consumers Association
Observer
Ian Blair Ofcom
Nutrient profiles for food promotion to children: Section 1: Literature review
16
SECTION 1. REVIEW OF RELEVANT NATIONAL AND
INTERNATIONAL LITERATURE
1.1 Summary
The aim of the review was to identify national and international data applicable to
children aged 2- 16 years, which contains information relevant to defining ‘foods high
in fat, salt or sugar and ‘healthier food choices’.
A multi-pronged approach was used to collecting information, including: key word
searches in Medline and an extensive personal database; Internet searches; contacting
key individuals and organisations directly.
Food labelling legislation in both the USA and Canada contains relevant definitions,
and these are summarised in the review.
The majority of the work that was identified on vending machines was from the USA,
and examples of some detailed guidance are given.
Various approaches have been adopted to healthier school lunch provision including:
checklists for use by caterers; nutrient assessments of menus provided over a period
of time; lists of recommended foods; and the use of nutrient profiles to identify foods
which can contribute to healthier food provision.
Over the last few years there have been various initiatives around the world to raise
awareness of the issue of marketing of foods to children. Generally speaking these
propose either banning all food advertising targeted at certain age groups of children,
or banning or restricting the marketing and advertising of ‘unhealthy’ or ‘junk’ foods.
There is usually very little attempt to define what is meant by these terms, although
specific foods are often mentioned. Restrictions are most frequently applied to
products such as chocolate, sweets, soft drinks, snacks and ‘other similar products’.
Nutrient criteria are used in some countries to assess eligibility for health claims,
addition of nutrients, and also in public health point of purchase schemes. Many
retailers in the United Kingdom also run healthier choice schemes, where products
that meet certain nutrient criteria are indicated in some way. These schemes are often
for the general population, rather than being intended for children. However, most UK
retailers do have ranges of foods specifically for children, and most of these specify
that the foods contain ‘controlled levels of sugar, fat and salt’, as well as restricting
the use of additives. Some manufacturers of children’s foods, and organisations who
licence characters for use on products, have adopted a similar approach.
Overall, there does not appear to be an explicit and transparent basis to most of the
schemes that were identified, and there is little evidence of systematic development
and testing of schemes.
Nutrient profiles for food promotion to children: Section 1: Literature review
17
1.2 Objectives
This review was intended to inform further stages in this project. It was not
systematic or comprehensive, but provided information that has been used to inform
decisions made in later stages of this project.
The focus of the review was to identify national and international data applicable to
children aged 2-16, which contains information relevant to definitions of ‘foods high
in fat, salt or sugar and ‘healthier food choices’.
This included:
Healthy eating advice, including dietary guidelines, guideline daily
amounts (or equivalent) and food guides. Since the focus of the work is
policy in the context of the UK and EU,
This element of the review is restricted to information relating to the UK.
Information on:
definitions of ‘foods high in fat, salt or sugar and ‘healthier food
choices’
nutrient or food based profiling intended for foods or drinks that are
promoted to or formulated for consumption by children. This includes
foods or drinks: which are for available in vending machines in
schools; provided through other school or pre-school food provision;
which carry health claims which claim health benefits for children; or
are advertised during children's TV programmes.
This element of the review included searches for information from the UK,
other European and Nordic countries, North America, Australia and New
Zealand.
1.3 Methods
A multi-pronged approach was used to collecting information, including:
Key word searches in Medline and an extensive personal database. The key words
used were: ‘nutrient profile’, ‘nutrient criteria’, ‘un/healthy food/s’, ‘un/healthful
food/s’, definition, advertising, children, vending, ‘menu signposting’, school/,
‘school food service’, catering, meals. Searches were carried out for publications
over the past 10 years.
Papers were excluded which focused on public health interventions in schools, or
which defined dietary quality indices for individuals.
Nutrient profiles for food promotion to children: Section 1: Literature review
18
Internet searches a) key word b) government department websites in
Australia/New Zealand, Canada and the USA. The searches were all made
using ‘Google’, and used the same keywords as the Medline searches, except
that ‘kids’ was used in addition to ‘children’.
Key individuals who may have knowledge of the area or who work in
organisations which are known to be undertaking relevant initiatives, were
approached. This included: contacts in government agencies in the USA,
Canada, Australia, and selected European countries; academic contacts in
Europe; international organisations including the EU, WHO, FAO and
international consumer groups; advocacy groups in the USA and UK; and
members of the Expert Group for this project, which included representatives
with links to retailers, food service providers, and food manufacturers. A
‘snowballing’ approach was used, with one contact passing on the request for
information to other contacts.
1.4 Findings
1.4.1 Dietary guidance for children
1.4.1.1 UK
The Department of Health has published Dietary Reference Values (DRV) for groups
in the population (Department of Health, 1991). For children the categories for these
age groups are: 1-3 years, 4-6 years, 7-10 years, 11-14 years, and 15-18 years. The
DRV report is the main source document for vitamin, mineral, and energy
recommendations for children.
The report also contains population recommendations for a total fat intake of 35% of
food energy, and saturated fat intake of 11 % of food energy, as well as
recommendations for monounsaturated and polyunsaturated fat intakes as a
percentage of energy. It also recommends consumption of 18 grams of non-starch
polysaccharide a day. These recommendations are for adults.
The Balance of Good Health, the UK's Food Guide, is based on dietary reference
values as well as food intake data. The guide states that is not applicable to children
below five years old. Between the ages of two and five is regarded as a transition
between an infant diet and an adult diet, and recommendations for fat and fibre do not
apply to children aged less than five years old. For children aged five and over, the
adult recommendations apply.
The Dietary Reference Value report also reiterates an earlier recommendation for non
milk extrinsic (NME) sugars (Department of Health: Committee on Medical Aspects
of Food Policy, 1989) that at a population level NME sugar intake should be reduced
to 11 % of food energy. The Weaning Diet Panel of the Committee on Medical
Aspects of Food Policy considered this recommendation. Their final report concluded
that this value for NME sugars is also applicable to pre-school children, because they
are at a relatively high risk of dental caries (Department of Health, 1994).
Nutrient profiles for food promotion to children: Section 1: Literature review
19
The Committee on Medical Aspects of Food Policy also recommended that in the UK
fruit and vegetable consumption should be increased to at least five portions per
person per day on average. This is applicable to children of all ages as well as adults.
However, although quantified portion sizes have been developed for adults, they have
not yet been developed for children.
The Scientific Advisory Committee on Nutrition recently reviewed salt and health,
and made recommendations for average intakes by population sub-groups including
children (Scientific Advisory Committee on Nutrition, 2003). The relevant
recommendations are 1-3 years, 2 g/day; 4-6 years, 3 g/day; 7-10 years, 5 g/day; 11-
14 years, 6 g/day.
In terms of individual food products intended for consumption by children it should
be noted that recent regulations provide detailed nutrient profiles for cereal foods
intended for consumption by children under three years old (e.g. for relevant
legislation from one of the countries of the UK (National Assembly for Wales, 2004))
In 1998 (that is, after all of the dietary recommendations mentioned previously had
been published, apart from the Scientific Advisory Committee on Nutrition report on
salt) the Caroline Walker Trust published nutritional guidelines for children under
five in child care (Caroline Walker Trust, 1998). This was accompanied by a practical
menu planner for use by child care establishments (Caroline Walker Trust, 2000).
With the exception of salt, these guidelines provide a synthesis of dietary guidance for
children between two and five years of age. In relation to fibre, the report additionally
makes the point that recommendations for children would be expected to be
proportionate to the adult recommendations.
Finally, although the Balance of Good Health is not intended for children below five
years old, versions have been developed (for example by retailers) for children over
this age showing the types of food more likely to be consumed by children.
1.4.1.2 USA
The USA also has detailed dietary recommendations, and these have evolved in very
similar ways to those in the UK. The points made here are relevant to
recommendations for children, in particular the so-called ‘healthy eating’
recommendations for fat, fibre, salt and sugar.
As long ago as 1996 the American Heart Association made a recommendation that the
dietary guidelines for the reduction of risk of heart disease should apply to all children
over two years of age. The relevant recommendations in the USA are to reduce
saturated fat to less than 10 % of total calories, total fat to an average of no more than
30 % of total calories, and dietary cholesterol to less than 300 milligrams per day.
The USDA has developed a food pyramid (the American Food Guide) specifically for
2-6 year old children, and this can be downloaded from the USDA website. At the
time of writing this report there was a consultation in process proposing revisions to
the food intake patterns that formed the basis for the pyramid, including the guide for
Nutrient profiles for food promotion to children: Section 1: Literature review
20
children. These proposals included nutritional goals for proposed daily food intake
patterns for children of different ages between 1-18 years of age. The goals showed
an explicit graded progression in dietary recommendations between infancy and
adulthood for fat and fibre. Salt and added sugar recommendations were the same for
young children as for adults. A detailed table was found at
http://www.usda.gov/cnpp/pyramid-update/FGP docs/TABLE 3.pdf.
1.4.2 Relevant definitions
The searches included looking for information on definitions of ‘foods high in fat, salt
or sugar and ‘healthier food choices’. This section describes those definitions that
appear in legislation from the USA and Canada.
In the USA there is an extremely lengthy definition of the term ‘healthy’ or any
derivative of the term e.g. ‘healthful’, ‘healthier’, ‘healthily’, and ‘healthiness’. This
is contained in Federal Regulation 58, 1993, with the most recent amendment being
FR 63, 1998. There is a far more consumer friendly definition in the USDA’s
guidance to food labelling, which is reproduced here:
Labels in the USA also include ‘Daily Values’ (DVs) for macronutrients that are
sources of energy: fat, saturated fat, total carbohydrate (including fibre), and protein;
and for cholesterol, sodium and potassium, which do not contribute calories. DVs for
the energy-producing nutrients are based on the number of calories consumed per day.
DVs for some nutrients represent the uppermost limit that is considered desirable. The
DVs for total fat, saturated fat, cholesterol, and sodium are:
total fat: less than 65 g
saturated fat: less than 20 g
cholesterol: less than 300 mg
sodium: less than 2,400 mg
The DV’s for vitamins, minerals and protein are based on Reference Daily Intakes
(the old Recommended Daily Amounts).
“A "healthy" food must be low in fat and saturated fat and contain limited
amounts of cholesterol and sodium. In addition, if it' s a single-item food, it
must provide at least 10 percent (of the Daily Reference Value) of one or
more of vitamins A or C, iron, calcium, protein, or fiber. Exempt from this
"10-percent" rule are certain raw, canned and frozen fruits and vegetables
and certain cereal-grain products. These foods can be labeled "healthy," if
they do not contain ingredients that change the nutritional profile, and, in
the case of enriched grain products, conform to standards of identity, which
call for certain required ingredients. If it's a meal-type product, such as
frozen entrees and multi-course frozen dinners, it must provide 10 percent
of two or three of these vitamins or minerals or of protein or fiber, in
addition to meeting the other criteria. The sodium content cannot exceed
360 mg per serving for individual foods and 480 mg per serving for meal-
type products.”
Nutrient profiles for food promotion to children: Section 1: Literature review
21
Manufacturers can declare nutrients as a percentage of dietary values, except for
foods intended for children aged less than 4 years of age. These may only carry %
DVs for protein, vitamins and minerals.
The USA also has guidance for nutrient content claims, expressed as grams per
serving. For example a food that makes a low fat claim must contain less that 3g fat
per serving. The main point of relevance to this report is that ‘a little’ of a nutrient is
regarded as a synonym for ‘low’.
In contrast, in Canada there is guidance for nutrient content relating to the Daily
Values (which are similar to those in the USA) (Health Canada, 2004). The DV is
intended to help individuals assess whether a food contains ‘a lot’ or ‘a little’ of a
nutrient. Examples of specific criteria for nutrition claims included:
A food that has a % DV of 5% or less for fat, sodium or cholesterol would
be low in these three nutrients
A food that has a % DV of 10% or less for saturated + trans fats would be
low in these two nutrients
A food that has a % DV of 15% or more for calcium, vitamin A or fibre
would be high in these nutrients
There is no DV for sugars in Canada (and there does not appear to be one in the USA
either) because “there is no recognised guideline on the amount that should be
consumed by healthy populations”.
Finally, there are several populist websites on the internet that attempt to define ‘junk’
or ‘unhealthy’ food. It is worth noting that these often include the use of additives and
the degree of processing foods are subject to, as part of a definition.
1.4.3 Vending machines
France introduced amendments to its public health bill in July 2004, which could
affect vending machine provision in schools. Nutritional profiles would be used to
determine which foods should not be provided in machines. This, and other
requirements still to be developed, could take effect from September 2005.
The remainder of the work that was identified for this section comes from the USA.
The Surgeon General recommended that “healthy snacks and foods are provided in
vending machines, school stores, and other venues within the school’s control;
prohibit student access to vending machines, school stores, and other venues that
compete with healthy school meals in elementary schools and restrict access in
middle, junior, and high schools” (US Dept of Health and Human Services, 2001).
This was supported by a recent report by the Democratic Party as part of a review of
the Children's Bill, which was finally passed in May 2004 (Democratic Staff of the
Senate Committee on Agriculture & Nutrition, 2004). This report also provides a
useful overview of recent initiatives in different states of the USA to improve
availability of healthy foods in vending machines.
Nutrient profiles for food promotion to children: Section 1: Literature review
22
The Centre for Food and Justice, Urban and Environmental Policy Institute has
compiled a list of innovative school policies, including some relating to vending
machines (Center for Food and Justice, 2002). These included:The American School
Food Service Association (ASFA) produces various resources to support healthier
food provision (http://www.asfa.org/childnutrition/fsoperations). This includes a
publication on ‘Healthy school snacks and beverages: selected policies and
guidelines’, which is only available to members. ASFSA’s website also contains a
guide to assessing the healthiness of foods for vending machines, from Utah
(American Food Service Association, Date unknown). This uses a scoring system,
based on work by the Dairy Council. If a food has points, it is considered healthy; if a
food has zero or negative points, it is not considered healthy. The guidance is
reproduced below:
1. Add one point to the food for each of the following:
o Has 10% or more Daily Value for vitamin A
o Has 10% or more Daily Value for vitamin C
o Has 10% or more Daily Value for calcium
o Has 10% or more Daily Value for iron
o Has 10% (5g) or more Daily Value for protein
o Has 10% or more Daily Value for fiber
2. Subtract one point from the food for each of the following
o Has 10% or more Daily Value for total fat (based on a 30% fat
diet)
o Has 250 or greater kilocalories
1.“An individual food item sold to a pupil during morning or afternoon breaks at
elementary schools shall meet all of the following standards: (a) not more than 35 percent
of total calories shall be from fat. This subparagraph does not apply to the sale of nuts or
seeds. (b) Not more than 10 percent of its total calories shall be from saturated fats. (c)
not more than 35 percent of its total weight shall be composed of sugar. This
subparagraph does not apply to the sale of fruits or vegetables”.
2. “Regardless of the time of day, water, milk, 100 percent fruit juices, or fruit-based
drinks that are composed of no less than 50 percent fruit juice and that have no added
sweeteners are the only beverages that may be sold to pupils at an elementary school.”
California Senate Bill 19.
3. “Resolved, that effective January 2004, the only beverages authorized for sale at the
Los Angeles Unified School District before, during, and until one half hour after the end
of the school day at all sites accessible to students shall be: fruit based drinks that are
composed of no less than 50 percent fruit juices and have no added sweeteners; drinking
water; milk, including, but not limited to, chocolate milk, soy milk, rice milk, and other
similar dairy or non-dairy milk; and electrolyte replacement beverages that do not contain
more than 42 grams of added sweetener per 20 ounce serving”.
4.“Vending machines accessible to students shall not dispense sodas, drinks that contain
caffeine or a high concentration of sugar, candy, or similar products during school
hours.”
Nutrient profiles for food promotion to children: Section 1: Literature review
23
3. If the food still has any points left, then the food is considered healthy.
Foods with zero or negative points are not considered to be healthy.
4. Also, any fruit or vegetable, regardless of whether or not it fits the criteria,
is suitable for the vending machine. In addition, fruit juice should be 100%
juice or at least 75% juice.
The detailed document provides examples of foods that are healthy or unhealthy
choices using this scoring system.
Another example of guidance is from San Antonio in Texas (Health Collaborative San
Antonio, 2004). Snacks and beverages for vending machines were placed in 3
categories: healthiest, healthier, and excluded.
Snacks:
Healthiest – must meet both criteria
o 3 grams of Total Fat or fewer per serving (Nuts and seeds
exempt from restrictions.)
o 30 grams of Carbohydrates or fewer per serving (All candies
are considered unhealthy. Fruit in any form is permitted,
regardless of carbohydrate count.)
Healthier – must meet both criteria
o 5 grams of Total Fat or fewer per serving (Nuts and seeds
exempt from restrictions.)
o 30 grams of Carbohydrates or fewer per serving (All candies
are considered unhealthy. Fruit in any form is permitted,
regardless of carbohydrate count.)
Drinks:
Healthiest o Milk – Low fat (1%) or non-fat preferred, any flavor
o Water – Pure
o Juice – at least 50% fruit or vegetable juice
Healthier o Water – Flavored or vitamin enhanced
o Low-Calorie Beverage – (<50 calories per 12 oz serving)
These guidelines were used to develop lists of foods that fell into the three categories.
These lists included food items that were marginal in terms of exclusion.
1.4.4 School lunches
1.4.4.1 UK
National minimum nutritional standards for school meals were reintroduced in the UK
in 2001. The regulations are based on the main food groups shown in the Balance of
Nutrient profiles for food promotion to children: Section 1: Literature review
24
Good Health. For primary schools the emphasis is on achieving an appropriate
balance of foods from the different food groups over a period of time. For secondary
schools the regulations stipulate (roughly) that at least two items from the four main
food groups of the Balance of Good Health must be available each day. During the
consultation on the draft regulations there was extensive discussion about whether
nutrient monitoring or a checklist for use by caterers was more appropriate. The final
form of the regulations lends itself more to using a checklist. The Department for
Education and Skills provides guidance on healthy lunches for pre-school children,
primary and secondary school children. This includes a description of the national
nutritional standards, good catering practice, and advice on how the standards can be
monitored.
Thus the core approach to school lunches and nutrition focuses, quite understandably,
on the overall menu provided over a number of days rather than the nutrient profile of
individual food products. It is worth noting though that although the current approach
to monitoring schools meals tends towards a catering checklist, there have been very
well respected tools developed proposing nutrient based guidelines for school meals
The original work was done by the Caroline Walker Trust (CWT) who produced
nutritional guidelines for school meals (Caroline Walker Trust, 1993). This was
followed by a computerised assessment pack using the CWT guidelines, produced by
the National Heart Forum (National Heart Forum, 1995).
More recently in Scotland the Scottish Executive’s Expert Panel on School Meals
established nutrient standards for the nutritional content of school meals (Scottish
Executive, 2002) largely based on CWT guidelines but modified to take into account
Scientific Advisory Committee on Nutrition guidance on salt . As part of these
Scottish guidelines FSA Scotland prepared target nutrient specifications for
manufactured products (www.scotland.gov.uk/library5/education/niss-00.asp). These
guidelines will be monitored as part of the Her Majesty's Inspectorate of Education
programme of school inspections and by independent research in 2007. Nutritional
software (Scottish Executive, 2004) was also produced to help local authorities plan
menus and monitor nutritional intake to meet the nutrient standards.
In the catering sector more generally, menu labelling or signposting schemes have
been developed. These commonly consist of symbols attached to food items on a
menu, which meet certain criteria, often fat levels. Relatively few of these have been
described in detail in the scientific literature and even less are from the UK. Balfour
tested out different formats for labelling, and foods were judged to be high in a
nutrient if they contained more than 50 % of the Dietary Reference Value, or low in a
nutrient if they contained less than 10 % of the Dietary Reference Value (Balfour et
al 1994.). Williams evaluated the ‘ Star-struck ‘scheme. Foods were assigned one
yellow star to indicate medium fat items i.e. contained between 5 and 10 grams of
fat/100g, or two yellow stars to indicate low-fat i.e. less than 5 g fat/100g. Medium
fibre was indicated by one green star i.e. a product had to contain between 2 - 4 g
fibre/100g, and high fibre was indicated by two green stars i.e. a product contained
>4g fibre/100g (Williams & Poulter, 1991).
Finally some catering companies, including Sodexho, have used smart cards to offer
rewards to children who choose healthier items.
Nutrient profiles for food promotion to children: Section 1: Literature review
25
1.4.4.2 France
France has voluntary food based dietary guidelines for ‘community’ meals prepared
by public and private organisations (Ministere de l'Economie dFedl, 2001). These
cover school meals, hospital and prison provision. The guidelines provide advice on
the general structure and desirable frequency of provision for some food items in 20
consecutive meals. A list of acceptable frequently consumed foods, with portion sizes
is also provided. The objective of the guidelines is to decrease fat intake, particularly
saturated fat, and to increase iron, calcium, fibre and vitamin (including vitamin C)
intakes. Recently some work at the National Institute for Medical and Health
Research (INSERM) has explored the possibility of using linear programming based
on the guidelines, to evaluate and provide guidance on the foods provided, and the
nutritional value of meals in school canteens. This approach is currently under
discussion but preliminary work has indicated that this approach might provide a basis
for designing user-friendly programs which could be used in school meal planning to
generate a large number and variety of nutritionally adequate menus (Darmon et al.,
2004).
1.4.4.3 Australia
There are a wide variety of approaches to encouraging healthier menu provision for
school lunches throughout Australia. Several states operate reward schemes and
accreditation programmes, where schools need to meet a range of standards, often
including list of recommended foods and preparation methods. New South Wales
operates a canteen smart card which gives the highest points to recommended foods,
including sandwiches, muffins and approved ice-creams and juices. Fewer points are
earned for foods that should be limited in the diet, such as reduced fat hot dogs. No
points are given for sport drinks, pies and crisps.
New South Wales and the Western Australian School Canteen Association (WASCA)
both operate schemes that combine lists of recommended foods with using nutrient
profiles to a certain extent.
WASCA runs a Star Canteen Accreditation Programme. This awards schools
operating healthy canteens by using a star rating of three, four or five stars. These
stars are earned by using recommended core foods and ‘ Star Choice registered
products’. Detailed lists of recommended foods are provided. The national criteria
used to assess Star Choice Registered Products address levels of fat, salt, fibre and
sugar in foods, and some categories of products also have a criterion for calcium. The
Star Choice Products Register and Canteen Buyer's Guide lists mainly processed
foods that meet the criteria. This is limited in some ways since only products that have
been submitted are assessed. WASCA also recommends product which are indicated
by the National Heart Foundation Australia's Pick the Tick scheme (see Section 4.5).
New South Wales has published a Canteen Menu Planning Guide (New South Wales
Health, 2004), which divides foods into three categories: Green – ‘Fill the menu’;
Amber – ‘Select carefully’; and Red – ‘Occasional foods’. Occasional foods are only
intended to be consumed on two occasions per term - when there is a particular
celebration or event happening. . Figure 1 reproduces the relevant section from the
Canteen Menu Planning Guide.
Nutrient profiles for food promotion to children: Section 1: Literature review
26
Figure 1: Nutrient criteria in the New South Wales Canteen Menu Planning Guide
(New South Wales Health, 2004).
The green category includes fresh or minimally processed foods, whereas the amber
category is mainly processed foods that have some sugar, salt or fat added to them.
The canteen guide provides detailed lists of the types of foods that fall into these two
categories, but does not appear to have nutrient criteria to define them.
1.4.4.4 USA
Current USDA regulations only apply to foods served in cafeterias at the time of meal
service, and contain a ban on the sale of a category of foods called “ foods of minimal
nutritional value”. This comprises sodas, water ices, chewing gum, and candies
composed predominantly of sugar. Some states have established what appears to be
called in the USA ‘ competitive food policies’. The extent of these policies varies,
with some states using the basic USDA requirement, but barring the sale of foods of
minimal nutritional value until after the lunch period (e.g. Maryland). Other states, for
example West Virginia, have taken the basic requirements much further and regulate
the sale of all foods during the teaching day, have established nutritional guidelines
for sugar, and required that foods sold as individual items must be the same as those
foods offered through the National School Lunch Programme (Democratic Staff of
the Senate Committee on Agriculture Nutrition and Forestry, 2004). The Children’s’
Bill, which was passed in May 2004, rejected an amendment to commission the
National Academy of Sciences’ Institute of Medicine to provide recommendations on
school nutrition guidelines to the Secretary of Agriculture.
The National School Lunch Programme is administered by the USDA and as well as
being available on a means tested basis is encouraged more widely by offering
Nutrient profiles for food promotion to children: Section 1: Literature review
27
subsidies for school lunch provision which are part of the Programme. Meals must
meet Federal nutrition standards, which are based on the Dietary Guidelines for
Americans, and the school lunches must also provide at least one third of the
Recommended Daily Allowances of protein, vitamin A, vitamin C, iron, calcium, and
calories. The USDA’s Team Nutrition encourages adoption of these standards using a
variety of communication channels and support activities.
The USDA also provides childcare meal patterns for children aged 1-2 years, 3-5
years, and 6-12 years. Unlike the Caroline Walker Trust guidelines, USDA guidance
is presented in terms of types and amounts of different foods.
The American School Food Service Association recognises that most food service
directors are more familiar with what they call a ‘traditional food based meal pattern’.
However, they provide information on nutrient standard menu planning, and a list of
approved nutrient standard menu planning software programs.
Thus, in the USA, most of the information that was retrieved suggests that it is more
usual to use a menu based approach rather than a product-based approach. Some
research studies have used a more product specific approach. For example the
Pathway study, a model for lowering fat in school meals, used dietary guidance to
derive food specific guidelines for fat in different categories of food (Snyder et al.,
1999). The maximum amounts of fat (grams) that were permitted per serving were:-
Meat or meat alternative <12
Meat or meat alternative with bread, fruit or vegetables <15
Vegetables or fruit 1
Salad dressings 3
Oven-baked fries 3
Salads 3-5
Breads, pasta and rice 3
Quick breads 5
Snacks 3
Desserts 5
Milk and cheese 5
1.4.5 Advertising and marketing foods to children
Over the last few years there have been various initiatives around the world to raise
awareness of the issue of marketing of foods to children. Generally speaking these
propose either banning all food advertising targeted at certain age groups of children,
or banning or restricting the marketing and advertising of ‘unhealthy’ or ‘junk’ foods.
There is usually very little attempt to define what is meant by these terms, although
specific foods are often mentioned. For example in New Zealand, one of the groups
responding to a Ministry of Health discussion document on children's food, would
like to see control of the marketing of soft drinks, pies, sweets and chocolate.
Nutrient profiles for food promotion to children: Section 1: Literature review
28
In France an amendment to the Public Health Bill, which was adopted in July 2004
would introduce requirements for television advertising of foods and drinks. There
would be restrictions on the types of foods and drinks which could be advertised, and
any departure from this would only be agreed if a financial contribution were made.
This contribution would be used by the Institute of Health Education and Prevention
to fund information and education initiatives.
Recently the World Health Organisation published a review of international practices
in marketing of foods to children, including: television advertising, in school
marketing, sponsorship, product placement, internet marketing and sales promotions
(Hawkes, 2004).
Relevant sections from the report are summarised below:-
1. Television advertising is perhaps the most popular means of promoting food
and beverage products worldwide and consequently has been the subject of
more debate, in terms of its effects on children, than any other marketing
practice. It is also the most widely regulated; 85% of the 73 countries surveyed
had some form of regulation on television advertising to children and almost
half (44%) had specific restrictions on the timing and content of television
advertisements directed at children. Two countries and one province have
banned television advertising to children.
2. In France, an amendment to the Public Health Law to ban commercials for
high-fat and high-sugar foods during children's television has been proposed
to parliament. The International Obesity Task Force (IOTF) has proposed that
advertising of “inappropriate foods and drinks” to children be prohibited.
3. 22 of the 73 countries surveyed have either a specific clause on food
advertising contained within their existing regulations on advertising, or a
separate code covering food advertising that is distinct from regulations
pertaining to nutrition and health claims
Generally speaking, the purpose of incorporating a food clause is to prevent
advertisements that promote ‘unhealthy’ diets (15 countries), especially when
these advertisements are targeted at children (10 of the 15 countries). The
wording of the guidelines varies between countries, for instance:
o advertisements should not give the impression that sweets, soft
drinks etc. can replace a regular meal
o advertisements should in general encourage healthy diets, and
discourage unhealthy ones
o advertisements should not encourage excessive consumption.
In most countries, it is not clear how these guidelines are applied, interpreted
and enforced.
4. The review contains a list of regulations specific to food advertising in
different countries. Scanning this list indicates there are several countries that
require that when confectionery is advertised, a toothbrush and logo also has
to be shown. Restrictions are most frequently applied to products such as
Nutrient profiles for food promotion to children: Section 1: Literature review
29
chocolate, sweets, soft drinks, snacks and ‘other similar products’. Brazil
requires that when food products are advertised they shall expressly indicate if
possible the nutritional and calorific value of the product being advertised. In
the United States “representation of food products should be made so as to
encourage sound use of the product with a view towards healthy development
of the child and development of good nutritional practices”
5. In-schools marketing: In Saudi Arabia the sale of carbonated soft drinks is
either totally or partially prohibited, in the United States foods of minimal
nutritional value must not be sold in food service areas during the school lunch
period.
1.4.6 Point of purchase schemes
Nutrient profiles are or will be used in the legislation of some countries to determine
foods which are eligible for addition of nutrients or to carry a health claim (Stockley,
2003). Public health agencies and charities have developed point of purchase
programmes to identify ‘healthier’ foods, for example foods that can carry a heart
symbol, or a logo indicating that the food can contribute to eating 5 portions of fruit
and vegetables per day without compromising other healthy eating guidance. Table 1
in Appendix 1 summarises the nutrient criteria which are applied to eligibility for
health claims, addition of nutrients, and public health point of purchase schemes.
This table includes a relatively new scheme from Australia, called the Glycaemic
Index Symbol Programme. The University of Sydney, Diabetes Australia and the
Juvenile Diabetes Research foundation support the programme. To carry a GI symbol
foods must meet specific nutritional criteria and have their glycaemic index measured
using an approved method. This differs from previous schemes because a
physiological effect of carbohydrates is measured, rather than assessing sugar or
starch levels. The nutrition criteria include the product containing at least 10 grams of
carbohydrate per serving. The other criteria are different for different food categories,
and only a general summary is provided in Table 1 of Appendix 1.
As well as being used by government departments, and public health agencies,
retailers and manufacturers have developed or are developing schemes. In the United
Kingdom many supermarkets run programmes, where products that meet certain
nutrient criteria are indicated in some way. For example, Tesco has recently
announced that it will be running a traffic light labelling scheme to highlight total fat,
saturated fat, salt and sugar content of products with a red, amber or green colour
code.
However all of the schemes described above are for the general population, rather
than being intended for children. Most UK retailers do have ranges of foods
specifically for children, Waitrose has the Food Explorers range, and these foods
contain ‘controlled levels of fat, added sugar and salt’. In addition treat foods have at
least 25% less fat, added sugar or salt than a typical product. Asda has the ‘More for
Kids’ range of foods for 4-8 year old children, with requirements for levels of fat, salt
and sugar. Products also do not contain azo dyes, and additives have been taken out
where possible. Marks and Spencer’s has the ‘Loved by Kids’ range, again with
Nutrient profiles for food promotion to children: Section 1: Literature review
30
controlled amounts of fat, salt and sugar, and no artificial colours or flavours, and no
added preservatives.
Sainsbury’s launched the Blue Parrot Café range in 2001. The development of criteria
for this scheme is described in Sainsbury’s Taste of Success website. The criteria are
shown below:
meat based meals to only contain selected cuts of meat, poultry or fish
at least half a daily portion of vegetables in vegetable based meals
drinks
o fruit based (at least 10% fruit juice)
o milk based and flavoured waters only
o choice of products with artificial sweeteners and without
desserts
o fruit based desserts to contain added fruit
o all products to have controlled fat and salt (sodium) levels
a main meal product (entire meal)
o 25g fat or less (maximum of 7g saturated fat)
o 0.8g sodium
meal component, e.g. chicken nuggets, sausages, jumbo fish fingers
o 15g fat or less (maximum 5g saturated fat)
o 0.4g sodium
products for different eating occasions, e.g. lunch boxes (muffins, biscuits,
bacon and cheese bites)
o 5g fat maximum
o 0.2g sodium
additives approved by the Hyperactive Children's Support Group (although
some additives are needed, for example cured meats and sausages contain
additives essential to food safety)
o only natural flavours
o no flavour enhancers, e.g. monosodium glutamate
some products to be gluten free, wheat free, dairy free, vegetarian
include organic options were appropriate
Some manufacturers of children’s food also use nutrient criteria. For example the
Bird’s Eye ‘Captain’s Promise’ is that dishes contain less than 13 g total fat, 3.4g
saturated fat, and 0.3 g sodium per serving (Orsi, 2004).
Finally, the BBC recently announced that it would be launching new standards for
children’s licensed products (e.g. the use of BBC characters like the Tweenies,
Teletubbies and Fimbles) to address current concerns over levels of salt, sugar and
fats.
Nutrient profiles for food promotion to children: Section 1: Literature review
31
1.5 Discussion
This review describes, in fairly general terms, initiatives that are relevant to
developing nutrient profiles for foods promoted to children.
The review is useful in clarifying the overall context of initiatives, but not necessarily
in identifying the details of specific schemes. This is often because the schemes
operate on a commercial basis, or may have commercial implications, or because the
underlying detail is not seen as relevant to the consumer and so is not easily
accessible.
1.5.1 Correspondence between schemes identified in this review and a
theoretical approach to developing nutrient profiles
The next section of this report describes a theoretical approach to developing nutrient
profiles for children's foods. It suggests that the development of nutrient profiles
involves four stages. The stages are shown in Figure 1.
Fig 1. Stages in developing nutrient profiles
Stage 1 Stage 2 Stage 3 Stage 4
An attempt has been made in Table 2 of Appendix 1 to examine each of the schemes
where there was more detailed information, in order to extract information which is
relevant to each of these stages.
Choice of nutrients
The nutrients selected for developing nutrient profiles almost always include total fat
and sodium. Saturated fat is often included. Sugars are included less frequently.
Energy is also sometimes included.
Several of the schemes involving nutrient profiling include nutrients where people eat
less than the recommended intake. Fibre is probably the most frequently selected,
followed by calcium. Some schemes, particularly those from North America include
iron, vitamin C, vitamin A, and protein. Many schemes make specific provision to
promote foods containing fruits and vegetables.
Choice of
nutrients Choice of
base (e.g. per
100g, per
100kJ, per
serving)
Choice of
model type Choice of
numbers (i.e.
how much is
desirable or
undesirable?)
Nutrient profiles for food promotion to children: Section 1: Literature review
32
Choice of base
The most common choice of base is probably ‘per serving’, although ‘per hundred
grams’ is also used quite frequently. An energy base is used infrequently.
Choice of model type
A threshold model is used most commonly, with one or two schemes attempting to
use scoring. The schemes are about evenly split between those which use an across
the board or a food category approach.
Choice of numbers
These may be based on public health recommendations. This appears to be
particularly true for vitamins and minerals. With many of the schemes it is not
possible to identify the reasons for the choice of numbers, and it is assumed that it is
often a pragmatic decision.
1.5.2 Evaluation of nutrient profiling schemes
It was not part of the objectives of this review to include evaluation of nutrient
profiling schemes, although it would be desirable to do this in the future.
If this were undertaken as a future exercise, there would be two dimensions to
consider: -
a) the objective of the evaluation. These can be very variable, and include for example
consumer acceptability, consumer understanding, wider acceptability to industry and
others, and effects on consumer knowledge, attitudes, and /or behaviour.
b) the setting in which the nutrition profiling initiative is located. For example the
objectives of an evaluation would be very different in settings and initiatives as
diverse as school canteens, organisations using vending machines, promotional and
advertising campaigns, and point of choice schemes.
There are reviews that have already systematically considered evaluations of some of
aspects of these two dimensions. For example a systematic review of consumer
understanding of nutritional labelling (Cowburn and Stockley, in press) included point
of choice initiatives, many of which were based on nutrient profiling. The review
concluded that: these programmes in a supermarket setting could have a modest effect
on sales of labelled products during the time the intervention was in place; these
schemes were considered to be feasible interventions to help consumers compare
between brands; and they seemed to work best when they were highly visible and
supported by other information.
Nutrient profiles for food promotion to children: Section 1: Literature review
33
1.6 References
American Food Service Association (Date unknown) A Step-by-Step Guide to
Incorporating Healthy Foods into Your New or Existing School Vending
Machine. American Food Service Association.
Balfour D et al. (1994) Visual nutrition information for menu labelling. In Hygiene
and Nutrition in Foodservice and Catering, pp. 3.
Caroline Walker Trust (1993) Nutritional Guidelines for School Meals. Caroline
Walker Trust, London.
Caroline Walker Trust (1998) Eating Well for under 5s in child care. Caroline Walker
Trust, London.
Caroline Walker Trust (2000) CHOMP Menu Planner. Leaflet and software.
Center for Food and Justice (2002) Healthy School Food Policies: A Checklist (
Working paper). Center for Food and Justice Urban and Environmental Policy
Institute, Los Angeles.
Cowburn G & Stockley L Consumer understanding and use of nutrition labelling: a
systematic review. Public Health Nutritio, in press.
Darmon N, Darmon M & Ferguson E. (2004) The use of linear programming to test
the nutritional impact of existing food-based guidelines for school canteens in
France. National Institute for Medical and Health Research (INSERM), Paris.
Democratic Staff of the Senate Committee on Agriculture & Nutrition and Forestry
(2004) Food Choices at School: Risks to Child Nutrition and Health: Call for
Action. Washington DC.
Department of Health (1991) Dietary Reference Values for Food Energy and
Nutrients for the United Kingdom. In Committee on Medical Aspects of Food
Policy Report No 41. The Stationery Office, London.
Department of Health (1994) Weaning and the Weaning Diet. Committee on Medical
Aspects of Food Policy, London.
Department of Health: Committee on Medical Aspects of Food Policy (1989) Dietary
Sugars and Human Disease. HMSO, London.
Nutrient profiles for food promotion to children: Section 1: Literature review
34
Hawkes C (2004) Marketing Food to Children: the Global Regulatory Environment.
World Health Organisation, Geneva.
Health Canada (2004) Nutrition Labelling Toolkit for Educators. Download from
website.
Health Collaborative San Antonio (2004) Healthy Vending Guidelines: Part of the Fit
City/Fit Schools campaign in San Antonio Texas.
Ministere de l'Economie dFedl (2001) GPEMDA (Group Permenent d'Etude des
Marches de Denrees Alimentaire): Recommendation relative a la nutrition.
Recommandation J3-99 du 6 Mai 1999: Collection Marches Publics: Nouvelle
Edition.
National Assembly for Wales (2004) Welsh Statutory Instrument 2004 No. 314
(W.32) The Processed Cereal-based Foods and Baby Foods for Infants and
Young Children (Wales) Regulations 2004.
National Heart Forum (1995) School Meals Assessment Pack: a guide for those
involved in influencing, providing and monitoring school meals in secondary
schools. National Heart Forum, London.
New South Wales Health (2004): Canteen Menu Planning guide.
Orsi J (2004) Manufacturing wholesome food for children: an industry opinion. BNF
Nutrition Bulletin 29, 57-59.
Scientific Advisory Committee on Nutrition (2003) Salt and Health. The Stationery
Office, London.
Scottish Executive (2002) Hungry for Success: A Whole School Approach to School
Meals in Scotland. The Stationery Office, Edinburgh.
Scottish Executive (2004) A Menu Planner for School Meals in Scotland. Nutmeg
UK.
Snyder P, Anliker J, Cunningham-Sabo L, Dixon L B, Altaha J, Chamberlain A,
Davis S, Evans M, Hurley J & Weber J L (1999) The pathways model: a
model for lowering the fat in school meals. American Journal of Clinical
Nutrition 69S, 810S- 5.
Stockley L (2003) Nutrition Profiles for foods to which nutrients could be added, or
on which health claims could be made. Report prepared for the Food
Standards Agency, UK, London.
US Dept of Health and Human Services (2001) The Surgeon General’s Call To
Action To Prevent and Decrease Overweight and Obesity. US Dept of Health
and Human Services, Rockville.
Nutrient profiles for food promotion to children: Section 1: Literature review
35
Williams C & Poulter J (1991) Formative evaluation of a workplace menu labelling
scheme. Journal of Human Nutrition and Dietetics 4, 251-62.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
36
SECTION 2. A THEORETICAL APPROACH TO DEVELOPING
DEFINITIONS
2.1 Summary
This section outlines a theoretical approach to developing nutrient profiles for
definitions of ‘foods high in fat, salt or sugar’ and ‘healthier food choices’ in relation
to foods and promotions aimed at children. It sets out the range of options available
when developing definitions for ‘foods high in fat, salt or sugar’ and ‘healthier food
choices’ 3.
It notes, on the basis of the available literature (see Section 1), the options people
have usually selected when developing nutrient profiles for such descriptors.
Using this theoretical approach, and having considered the various options, the Expert
Group for this project recommended a number of different possible definitions for
‘foods high in fat, salt or sugar’ and ‘healthier food choices’ be developed.
Ages of children
They agreed that the project should focus on the development and testing of
definitions for ‘foods high in fat, salt or sugar’ and ‘healthier food choices’ for
children aged 11-16. But noted that these should be applicable to children of any age
between 5 and 16 and also to adults.
The intended uses of definitions and models
They agreed that the project should focus on developing and testing definitions of
‘foods high in fat, salt or sugar’ and ‘healthier food choices’ for the purpose of:
Advice on the use of nutrition and health claims on foods aimed specifically at
children; and
Advice on the balance of TV advertising for foods during children’s TV
programmes.
But the potential use in signposting on labels and menus would also be taken into
account.
Choice of nutrients
They agreed that simple definitions merely involving criteria for energy, saturated fat,
non-milk extrinsic (NME) sugar and sodium should be developed and tested, but that
more complex definitions involving these nutrients together with criteria for fruit and
vegetables, long chain n-3 polyunsaturated fatty acids (n-3 fatty acids), calcium and
iron should also be developed.
In addition they agreed that where definitions involved criteria for calcium, iron and
n-3 fatty acids then those criteria should be for the levels of those nutrients prior to
any discretionary fortification.
3 Throughout this report the terminology of the brief for this project i.e. ‘foods high in fat, salt or sugar’
and ‘healthier food choices’ has been used. These terms are cumbersome and in future it might be
simpler to use ‘healthy’ and ‘unhealthy’ foods or ‘healthier’ and ‘less healthy’ foods.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
37
Choice of base
They agreed that the base/denominators to be considered for definitions should be: per
100g, per 100kJ, per 100g AND/OR per serving, per 100kJ AND/OR per serving.
Choice of model type
They agreed that both threshold models and scoring systems should be investigated.
They agreed that across-the-board rather than food category specific criteria should be
investigated in the first instance.
Choice of numbers
They agreed that the numbers used for thresholds or points scored should bear a
consistent and transparent relationship to public health recommendations.
2.2 Introduction
The brief for the current project was to identify and assess options for a number of
uses of definitions of ‘foods high in fat, salt or sugar’ and ‘healthier food choices’ in
relation to the promotion of foods to children including:
§ Advice on front of label and menu signposting
§ Advice on the use of nutrition and health claims on foods aimed specifically at
children
§ Identification of healthier food and drink options for sale though vending
machines in schools
§ Advice on the balance of TV advertising for foods during children’s TV
programmes
2.2.1 Age groups
The dietary recommendations for children of different age groups are different. In
particular dietary recommendations for children aged 2-5 are quite different than for
older children. So definitions of ‘foods high in fat, salt or sugar’ and ‘healthier food
choices’ for the purposes of promoting foods to children aged 2-5 will be different
from those for purposes in relation to older children.
However, from the age of 5 dietary recommendations set on an energy basis are the
same as or similar to those for adults. This means that nutrient profiles for the
purposes of promoting food to children older than 5, and indeed for the purposes of
promoting food to adults, will be similar.
For many macronutrients (fat, saturated fat, non-milk extrinsic sugar, etc) dietary
recommendations set on an energy basis are the same for children over 5 as for adults.
This also means that when converted to a weight basis (e.g. to Guideline Daily
Amounts) then the relationship between the recommendations is always the same for
all age groups over 5.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
38
The situation is similar for micronutrients (sodium, calcium, iron etc). In general the
dietary recommendations for these nutrients are proportional to energy intake, but
there are some exceptions. Most notably, dietary recommendations for some
micronutrients such as calcium and iron are higher than would be expected for 11-16
year olds. This is due to the body’s increased need for these nutrients during puberty.
Because of the link between dietary recommendations and energy intakes, whichever
age group’s dietary recommendations are used as the basis of a nutrient profile, the
profile will be transferable to all other age groups (with the exception of under 5s).
The Expert Group agreed that the project should focus on the development and
testing of definitions for ‘foods high in fat, salt or sugar’ and ‘healthier food
choices’ for children aged 11-16, in order to be sure that definitions involving
micronutrients are relevant to the group with the greatest requirement for these
nutrients. But the Expert Group noted that these definitions are likely to be
applicable to children of any age between 5 and 16 and also to adults.
2.2.2 The intended uses of definitions
Different definitions are likely to be needed depending on the use for which they are
required. For instance: for advice on the balance of TV advertising for foods during
children’s TV programmes it might only be necessary to develop a definition of
‘foods high in fat, salt or sugar’; whereas for advice on front of label signposting a
definition of ‘foods high in fat, salt or sugar’ is unlikely to be sufficient. Moreover a
definition of ‘foods high in fat, salt or sugar’ for the two purposes need not
necessarily be the same.
Moreover definitions will have different users with different requirements. For
example, if school caterers were to be required to use nutrient profiles for deciding
how to stock vending machines, they most likely would prefer simple definitions e.g.
involving as few nutrients as possible. In contrast the children buying the products
are only going to see the end result of the school caterers using nutrient profiles, and
theoretically need nutrient profiles that are as nutritionally accurate as possible
regardless of their complexity.
For the purposes of food and menu labelling it is possible that, instead of developing
definitions of ‘foods high in fat, salt or sugar’ and ‘healthier food choices’, the
nutritional properties of foods could be signposted though defining desirable and
undesirable levels of different nutrients. Desirable and undesirable levels of different
nutrients could be visually represented by the use of colour (‘traffic lights’),
graphically (e.g. though bar charts), in words (e.g. though use of verbal descriptors
such as ‘high’ and ‘low’) or numerically (e.g. as percentages of Guideline Daily
Amounts).
For this project new ways of defining desirable and undesirable levels of particular
nutrients have not been developed. There are already two ways of doing so that
could be used i.e. the Food Standards Agency (FSA) advice for what counts as ‘a lot’
or ‘a little’ of different nutrients (Food Standards Agency, 2002; Rayner et al, 2004)
and the Coronary Prevention Group (CPG) Nutrition Banding Scheme (Coronary
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
39
Prevention Group, 1988; Appendix 2) 4. However, before doing so, both schemes
need to be revised in the light of recent thinking. For example neither scheme gives
guidance on what constitutes a food with a ‘high’ energy density.
The Expert Group for this project agreed that the project should focus on
developing and testing definitions of ‘foods high in fat, salt or sugar’ and
‘healthier food choices’ for the purpose of:
§ Advice on the use of nutrition and health claims on foods aimed
specifically at children; and
§ Advice on the balance of TV advertising for foods during children’s TV
programmes.
But the potential use in signposting on labels and menus would also be taken into
account.
It was agreed that the needs of all users of the definitions should be taken into
consideration.
2.2.3 Stages in developing nutrient profiles
Whatever the intended use of definitions of descriptors such as ‘foods high in fat, salt
or sugar’ and ‘healthier food choices’ - it is suggested that their development
necessarily involves four stages (Figure 1). These stages can be approached in any
order and decisions at any one stage affect decisions at others. However the most
logical order is shown in Figure 1.
Figure 1. Stages in developing nutrient profiles
Stage 1 Stage 2 Stage 3 Stage 4
2.3 Stage 1. Choice of nutrients and other food components
2.3.1 Prioritising nutrients and other food components
There are a number of different nutrients and other food components that could
possibly be used for nutrient profiles. Intakes of most of these nutrients and other
components have been assessed by the National Diet and Nutrition Surveys for
4 Furthermore criteria for the use of nutrition claims – such as contained within the proposed EU
Regulation on nutrition and health claims (European Commission, 2003) – define desirable levels of a
number of nutrients.
Choice of
nutrients and
other food
components
Choice of
base (e.g. per
100g, per
100kJ, per
serving)
Choice of
model type Choice of
numbers (i.e.
how much is
desirable or
undesirable?)
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
40
children and young adults (Gregory et al, 1995; Gregory & Lowe, 2000). Table 1
shows the possible range of nutrients and other components, together with
recommended intakes.
Table 1. Possible nutrients and other components for nutrient profiles
Nutrient/Food Recommended
average intakes for
11-16 year olds
Source of
dietary
recommend
-ation*
Comment Nutrients/
components
considered
priorities by
Expert
Group
Total fat 35% of food energy 1Total fat includes poly-
unsaturated fatty acids
and mono-unsaturated
fatty acids, which have
health benefits.
Definitions involving
total fat as well as
saturated fat involve
double counting.
Saturated fat 11% of food energy 1v
Trans-fatty
acids 2% of food energy 1Population average
intake currently meets
population dietary goal
Carbohydrates 50% of food energy 1Glycemic Index has
been used in some
recent nutrient profiles
but more research is
needed in healthy
subjects before the
concept can be used
more widely in public
health (Laville, 2004)
Total sugar No recommendation Relatively simple to
measure but
discriminates against
fruit and milk products
Non-milk
extrinsic
(NME) sugar
11% of food energy 1No clear definition of
NME sugar content,
and no generally agreed
method of measuring
levels in foods
v
Added sugar No recommendation Added sugar is always
extrinsic. Not all NME
sugar is added. No way
of knowing how much
sugar has been added
without info. from
manufacturers
Salt 6 g/day 3v
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
41
Nutrient/Food Estimated Average
Requirements (EARs)
or, where no EAR,
Reference Nutrient
Intakes for 11-16 year
olds
Source of
dietary
recommend
-ation*
Comments Nutrients/
components
considered
priorities by
Expert
Group
Energy 1645-2755 kcal/day ** 1This could be energy
per 100g (energy
density) or energy per
serving but not
energy per 100kJ (see
p x below).
v
Vitamin A 400-500 µg/day 1
Thiamin 0.3 mg/1000kcal 1
Riboflavin 0.9-1.0 mg/day 1
Niacin 5.5 mg/1000kcal 1
Vitamin B6 13 µg/g protein 1
Vitamin B12 1.0-1.25 µg/day 1
Folate 150 µg/day 1
Vitamin C 22-25 mg/day 1
Iron 8.7-11.4 mg/day 1Main source: meat,
fish and alternatives v
Calcium 625-750 mg/day 1Main source: milk
and dairy foods v
Phosphorus 15.6-18.6 mmol/day 1
Magnesium 230-250 mg/day 1
Potassium 3100-3500 mg/day § 1
Zinc 5.5-7.3 mg/day 1
Copper 0.8-1.0 mg/day § 1
Iodine 130-140µg/day § 1
Protein 11.7-46.1 g/day 1
Recommended
average intakes for
whole population
Fibre (Non-
Starch
Polysaccharides)
18 g/day 1
Long chain n-3
polyunsaturated
fatty acids
0.45 g/day 4Main source:
oily fish v
Fruit and
vegetables 50% increase for whole
population (equates to
approximately 400g or
5 portions a day)
2v
Fish 2 portions a week (one
of oily fish) 4
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
42
* Sources:
1. Department of Health, Dietary Reference Values for Food Energy and Nutrients for the United
Kingdom. Report on Health and Social Subjects No 41. London: HMSO, 1991;
2. Department of Health. Nutritional Aspects of Cardiovascular Disease. London: HMSO, 1994.
3. Scientific Advisory Committee on Nutrition. Salt and Health. London: The Stationery Office, 2003.
4 Scientific Advisory Committee on Nutrition. Advice on Fish Consumption: Benefits and Risks.
London: The Stationery Office, 2004
** Where a range is given this is because Source 1 gives EARs and RNIs for 11-14 and 15-18 year olds
and for males and females separately and no overall figure for 11-16 year olds or sexes combined.
§ No EARs are given, so Reference Nutrient Intakes (RNIs) are given instead.
Reductions in the intake of fat, saturated fat, NME sugar and sodium are priorities for
current Government food and health policy, all of these nutrients are therefore
obvious candidates for nutrient profiles.
The Expert Group noted that, whilst there were more difficulties in defining NME
sugar and measuring its levels in foods, than with total sugar, it is a reduction in NME
sugar intake rather than total sugar that is the Government’s priority. Moreover they
noted that including total sugar levels in nutrient profiles would tend to disadvantage
fruit and some vegetables and dairy products.
The energy density of foods is increasingly recognised as an important cause for
concern. Foods of high energy density (energy per unit weight) have been reported
to undermine normal human satiety control mechanisms, leading to over-consumption
of energy (Prentice & Jebb, 2003). The recent Select Committee Inquiry on Obesity
recommended a food labelling scheme ‘based on energy density’ (House of Commons
Health Committee, 2004).
The Expert Group felt that if energy levels (either energy density or energy per
serving) were to be used in nutrient profiles it would be less important to include total
fat levels and noted that on an international basis there was increasing uncertainty
over recommended levels of total fat.
The literature review indicated that other nutrient-based definitions in relation to the
promotion of foods to children almost always include total fat and sodium. Saturated
fat is often included, and sugars are included less frequently. Energy (including
energy density) is also sometimes included.
Previous work on nutrient profiles (O’Neill, 2004) has suggested that if they only
involve criteria for fat, saturated fat, NME sugar and sodium then they are likely to
classify foods such as fruit juices, fish, etc. in the same category as food such as high
sugar carbonated drinks and confectionery. This seems anomalous. Accordingly the
Expert Group felt that definitions of ‘foods high in fat, salt or sugar’ and ‘healthier
food choices’ should be developed which include criteria for nutrients and foods that
people are encouraged to eat more of and/or are necessary for a healthy diet.
The Expert Group noted that the Government recommended an increased
consumption of fruit and vegetables and fish, particularly oily fish. Therefore they
considered whether the definitions might involve criteria for levels of fruit and
vegetables and fish. They noted that the increased consumption of fruit and
vegetables is recommended for a variety of reasons but that the fish recommendation
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
43
is because of the high levels of n-3 fatty acids in fish and that there is also a specific
Government recommendation for n-3 fatty acids. Accordingly the Expert Group
selected fruit and vegetable levels and n-3 fatty acid levels as criteria for possible
inclusion in the nutrient profiles.
Finally the Expert Group selected two nutrients - iron and calcium – in order to
categorise foods as ‘healthier food choices’ that might otherwise be regarded as less
healthy but for a redeeming feature e.g. iron in the case of lean meat, calcium in the
case of low-fat dairy products. Other nutrients could have been chosen, such as fibre,
zinc or vitamin C, but it was felt that calcium and iron were the most appropriate
choices, as they would identify healthier food choices from the ‘milk and dairy foods’
and ‘meat, fish and alternatives’ groups of the Balance of Good Health. According to
the Balance of Good Health, approximately one quarter of the diet should consist of
foods from these categories, yet many could be labelled as ‘foods high in fat, salt or
sugar’ by a nutrient profile which only considers fat (or energy) saturated fat, NME
sugar and sodium levels.
The literature review indicates several of the schemes involving nutrient profiling
include nutrients where children often eat less than the recommended intake. Fibre is
probably the most frequently selected, followed by calcium. Some schemes,
particularly those from North America include iron, vitamin C, vitamin A, and
protein. Many schemes make specific provision to promote foods containing fruits
and vegetables.
The Expert Group agreed that the following nutrients (and other components of
foods) be used in developing definitions for possible testing:
§ energy, saturated fat, non-milk extrinsic (NME) sugar and sodium
(henceforth called ‘A’ nutrients)
§ fruit and vegetables and long chain n-3 polyunsaturated fatty acids (n-3
fatty acids) (‘B’ nutrients)
§ calcium and iron (‘C’ nutrients)
It was agreed that they should be used in the following combinations:
Group A nutrients: just ‘A’ nutrients
Group B nutrients: ‘A’ nutrients + ‘B’ nutrients
Group C nutrients: ‘A’ nutrients + ‘B’ nutrients + ‘C’ nutrients
2.3.2 Fortification
The Expert Group felt that nutrient criteria should exclude fortificants, except in the
case of mandatory fortification. However they recognised that it might not be
possible to test definitions using pre-fortification criteria because of lack of pre-
fortification data for most products.
It was agreed that where definitions involved criteria for calcium, iron and n-3
then those criteria should be for the levels of those nutrients prior to any
discretionary fortification.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
44
2.3.3 Combining criteria for nutrients and other components
When a nutrient profile is being developed it is necessary to think not only of the
individual nutrient criteria, but also how the criteria are combined. It is possible to
combine criteria with any logical operator, but in all practical circumstances a
combination of AND and OR will suffice.
Instinctively, we tend to think of healthier foods as ‘constrained’ or ‘bounded’ and
less healthy foods as ‘unconstrained’ or ‘unbounded’. This means that we
instinctively feel that definitions of ‘foods high in fat, salt or sugar’ should use the OR
operator when combining the criteria for fat, salt and sugar. And similarly we
instinctively feel that a ‘healthier food choice’ would be a food that is low in fat, salt
AND sugar.
This however, is not the only way to proceed. For example even with nutrients such
as fat, salt and sugar it might be necessary to limit either the sugar level OR the salt
content when defining a ‘healthier food choice’, so long as the fat level is limited.
And similarly when defining a ‘food high in fat, salt or sugar’ it might be appropriate
to use the AND operator.
With increasing numbers of nutrients it is generally inappropriate to link all nutrients
with just one operator. For example, although it might be felt that for a food to be a
‘healthier food choice’ it ought to be low in fat AND salt AND sugar, it would seem
inappropriate to demand that it also had to be high in calcium AND iron AND n-3
fatty acids, since very few foods would be expected to contain significant levels of
calcium and iron and n-3 fatty acids. The use of some OR operators in such a
definition would seem to be required.
The Expert Group agreed that for the purposes of this project:
‘Foods high in fat, salt or sugar’ would need to be
above defined energy OR saturated fat OR NME sugar OR sodium
levels
below defined levels of fruit and vegetables AND n-3 fatty acids
AND calcium AND iron (where the definition involved criteria for
these nutrients/components)
‘Healthier food choices’ would need to be:
below defined energy AND saturated fat AND NME sugar AND
sodium levels
above defined levels of fruit and vegetable OR n-3 fatty acids OR
calcium OR iron (where the definition involved criteria for these
nutrients/components)
2.4 Stage 2. Choice of base
There are three basic ways of setting nutrient profiles: per 100g, per 100kJ and per
serving.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
45
2.4.1 Pros and cons of different bases
Per 100g
Pros: The use of a per 100g basis for nutrient profiles is common because it is usual to
give nutrient composition data on a per 100g basis, e.g. in food composition tables
and in nutrition labelling. 100g of food is also relatively simple to conceptualise –
particularly compared to 100kJ.
Cons: A food that is high in a nutrient on a per 100g basis may supply little of that
nutrient because it is eaten in small servings and/or infrequently (e.g. fat in mustard).
Conversely a food that may be low in a nutrient per 100g may supply a lot of that
nutrient because it is eaten in large servings and/or frequently (e.g. NME sugar in soft
drinks). One of the main reasons for this is that foods vary widely in their water
content and foods that have low levels of nutrients on a per 100g basis tend both to
contain high levels of water (on a per 100g basis) and to be eaten in large servings
(e.g. beverages, stews, etc.). Of course there are other determinants of serving size
and/or frequency of consumption besides the water content of foods. These include
intensity of taste (cf. mustard) and many other properties of foods.
Per 100kJ
Pros: Using a per 100kJ basis discounts the water content of foods, because foods
that have a low water content and are therefore energy dense tend to be consumed in
relatively small amounts, and conversely foods with a high water content/low energy
density tend to be eaten in larger quantities. This is, in turn, because the amount of
food that people eat is, in general, governed by their energy needs.
One of the consequences of this is that that there is stronger correlation between
nutrient per serving and nutrient per 100kJ than between nutrient per serving and
nutrient per 100g. Using an energy basis therefore partially takes account of serving
sizes.
Because the amount of food that people eat is governed by their energy needs
recommendations for daily intake of nutrients tend to be set as a proportion of total or
food energy. Therefore using a per 100kJ basis means that per 100kJ criteria can
directly correspond with dietary recommendations unlike per 100g criteria which rely
on recommendations being converted into g per day using an average energy intake
figure.
Cons: There are still some energy dense foods, such as mustard, which appear high in
a nutrient when measured per 100kJ, yet supply little of that nutrient because they are
eaten in small servings. Conversely there are some foods, such as some fruits and
vegetables, that have a very high water content and very small amount of energy
(either per 100g or per serving). However, if this small amount of energy is from a
small amount of fat, saturated fat or sugar (on a per serving basis), then the food will
appear to have a large amount of this nutrient when measured per 100kJ.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
46
Per serving
Pros: The use of a per serving basis for nutrient profiles recognises that people
consume irregular amounts of different foods and that foods that they eat in large
amounts supply more of the nutrients they contain than foods eaten in small amounts.
Serving size, together with frequency of consumption, are the main determinants of
how much of a nutrient a food supplies to peoples’ diets.
Cons: Serving sizes can vary considerably. For example, the serving size of milk
depends on whether it is drunk by itself or with breakfast cereal or with coffee. Also,
serving sizes partly depend on the energy needs of the consumer (e.g. children tend to
eat smaller serving sizes than adults). Serving sizes for some foods are therefore
difficult to define.
A per serving basis does not take account of frequency of consumption so some foods
(such as fat spreads) may be eaten in small servings but if eaten frequently can
contribute a lot of a nutrient to a person’s diet. But neither a per 100g nor a per
100kJ basis take account of frequency of consumption either.
The choice of which base is used for nutrient profiles could be crucial as different
bases will rank foods in different orders (e.g. Spaghetti Bolognese has less fat per
100g than mayonnaise, but more fat per serving).
The choice of which base to use becomes more important if ‘across the board’ criteria
are used than if ‘food category specific’ criteria (see Section 2.5.2) are used because
foods within categories tend to vary less in their serving sizes or their energy density
than the whole range of possible foods. Therefore a choice of base is partially
dependent on the choice of model to be used.
The literature review indicates that the most common choice of base for nutrient
profiling is probably per serving although per 100g is also used frequently. A per
100kJ base is used infrequently.
The Expert Group agreed that when developing models with their associated
definitions for testing no possible bases should be ruled out.
2.4.2 Combining bases
It is possible to combine two bases for nutrient profiles so as to gain the advantages of
both bases, and this technique can and has been used to introduce a judgement about
how much nutrient a food actually delivers to a diet. An example of this is the Codex
Alimentarius definition of ‘low in saturated fat’, which is that a food should have less
than 1.5g saturated fat per 100g and less than 10kJ saturated fat per 100kJ.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
47
The two bases which rank foods most similarly in order of their nutrient contents are
the per 100g and per 100kJ bases5. Using a per 100g/per 100kJ in combination with
a per serving basis would seem to be the most promising way of gaining the
advantages of two bases.
If two bases are used, a decision must be made about how they are combined. There
are two possibilities here, for example a definition might specify that a food meets a
per serving AND a per 100g criteria or it might specify that a food meets a per serving
OR a per 100g criteria.
Whether to combine the bases with AND or OR depends upon the bases that are being
combined.
For example Figure 2a shows the fat levels of 100 foods selected at random from the
McCance and Widdowson nutrient database, and plotted on a per 100g basis vs. a per
serving basis. Figure 2b shows the fat levels of the same foods plotted on a per
100kJ vs a per serving basis. Thresholds for fat per 100g, per 100kJ and per serving
have been added, which might be used as criteria for ‘foods high in fat’.
For both graphs, it is clear that the foods in the top right quadrant would most likely
be included in the definition of a ‘food high in fat’ and equally that the foods in the
bottom left quadrant would most likely be excluded. It is not as clear, however, for
foods in the remaining two quadrants.
The foods in the top left quadrant of Figure 2a are foods that provide a lot of fat per
serving, but not much fat per 100g. Foods which are eaten in large quantities but have
a high water content may fall into this quadrant (such as a Quarterpounder with
cheese , where the bun and the salad provide a significant amount of water). In order
to include such foods within the definition of ‘foods high in fat’, the OR operator
must be used to connect the bases. Using this operator will also mean that all the
foods in the bottom right quadrant of 2a are included i.e. foods high in fat on a per
100g basis but which are eaten in small amounts (such as chocolate éclairs, but also
nuts and some fish products). Although some apparent anomalies are inevitable, per
100g OR per serving combines the bases in the only way that includes foods with
high water content that might supply significant amounts of nutrients to the diet.
The water content is not taken into account when the fat is measured on a per 100kJ
basis and to a less extent on a per serving basis. As a result of this, the foods in the
top left quadrant of 2b are foods that are high in fat per serving, but are likely to be
high in energy (since they are not high in fat per 100kJ). Because of this, they are
likely to satisfy a consumer whilst not putting them over the recommended intake of
fat. These foods should perhaps be excluded from the definition of ‘foods high in fat’
and hence per 100kJ AND per serving should be used.
It is not as clear for foods in the bottom right quadrant of 2b, which again includes
foods such as nuts and fish products, but also high fat snacks such as éclairs and
5 Using a database of foods derived from McCance and Widdowson’s The Composition of Foods (see
Section 3 below) the correlation (r2) between fat levels measured per 100g and per 100kJ was 0.63,
whereas the correlation between the amount of fat per serving and the amount of fat per 100kJ was
0.34, and that between the amount of fat per serving and the amount of fat per 100g was 0.20.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
48
biscuits. However, there are many more foods in this quadrant than in the
corresponding quadrant of 2a, and these extra foods must all have been in the bottom
left quadrant of 2a. It is therefore wise to exclude all the foods from this quadrant, in
order to avoid these extra apparent anomalies.
Similar considerations will need to be made when drawing up definitions of ‘healthier
food choices’ based on two bases. But in this case the conclusion would be that with
a definition of a ‘food low in fat’ the food should be lower than a per 100g threshold
AND a per serving threshold, or a per 100kJ OR a per serving threshold.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
49
Figure 2a and 2b: Per 100g AND / OR per serving; Per 100kJ AND / OR per
serving
Per 100g vs per serving
0.00
5.00
10.00
15.00
20.00
25.00
30.00
35.00
40.00
45.00
50.00
0.00 10.00 20.00 30.00 40.00 50.00
Fat (g per 100g)
Fat (g per serving)
Per 100kJ vs per serving
0.00
5.00
10.00
15.00
20.00
25.00
30.00
35.00
40.00
45.00
50.00
0.00 20.00 40.00 60.00 80.00 100.00
Fat (kJ per 100kJ)
Fat (g per serving)
Plain risotto
Almond
s
Quarterpounder
with cheese
Eclairs
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
50
The Expert Group agreed that when combining per 100g and per serving criteria the
OR operator should be used for definitions of ‘foods high in fat, salt or sugar’ and the
AND operator for ‘healthier food choices’ but conversely when combining per 100kJ
and per serving criteria the AND operator should be used for definitions of ‘foods
high in fat, salt or sugar’ and the OR operator should be used for foods for ‘healthier
food choices’.
Accordingly the Expert Group agreed that when the developing models with
their associated definitions the following bases, or combinations of bases, should
be used
per 100g
per 100kJ
per 100g AND/OR per serving
per 100kJ AND/OR per serving
2.5 Stage 3. Choice of model type
There are three different options of model type that can be used for nutrient profiling;
threshold models, simple scoring systems, and complex scoring systems. Once the
type of model is chosen it is necessary to choose between food category specific or
across the board criteria.
2.5.1 Stage 3a: Thresholds models, simple scoring system or complex
scoring systems?
2.5.1.1 Pros and cons of different types of model
Threshold models
A threshold model works by setting thresholds for the level of each of the selected
nutrients. A food will be classified as a ‘food high in fat, salt or sugar’ or a ‘healthier
food choice’ depending on whether the levels of its nutrients are above or below the
thresholds.
Pros: Threshold models are relatively easy to understand. For example, with most
conceivable threshold models, it would be easy to explain to a consumer the
maximum amount of salt that a food defined as ‘a healthier food choice’ would
contain. Justification of threshold models is also clearer since, as long as the
thresholds are set in relation to justifiable targets, then the models are also justifiable.
Cons: They can often be over-simplistic, which can lead to apparent anomalies. For
example, suppose a model used to define ‘healthier food choices’ contained the
criterion ‘must contain less than 45kJ fat per 100kJ’, in line with the Coronary
Prevention Group Nutrition Banding Scheme definition of high fat foods. Then nuts,
grilled fish and tofu will all be excluded by the model.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
51
Note that with threshold models an implicit judgment is made that crossing one
threshold is equivalent to crossing another. So for example if a food is defined as ‘a
food high in fat, salt or sugar’ if it contains more then x g/100g of fat OR more then y
g/100g sodium then it is implied that x g of fat in 100g of a food is as undesirable as y
g of sodium. Similar but more complex assumptions have to be made when
developing scoring systems.
Simple scoring systems
A simple scoring system works in a similar fashion to a threshold model, except that
for each nutrient a series of thresholds are set. This creates a series of bands, and
depending on which band the food lies in for each nutrient, a certain number of points
are scored. The scores from each of the nutrient bands are added to give the food a
total score. Then a threshold score is set, which defines whether a food is ‘a food high
in fat, salt or sugar’ or a ’healthier food choice’.
Pros: A simple scoring system model is less likely to produce apparent anomalies, as,
for example, foods that just cross the threshold for a nutrient such as fat under a
threshold model will not necessarily be defined as ‘a food high in fat, salt or sugar’
using a definition based on a scoring system – particularly if their levels of nutrients
such as calcium or iron are used to lower their scores.
The scores that foods are awarded by a simple scoring system can be used for a
variety of purposes (for example, for comparison of foods in one particular food
category), and different score thresholds can be used for different purposes (for
example it might be agreed that there should be restrictions on the TV advertising of
foods that score at least 5 points, whilst there should be restrictions on the use of
health claims for foods that score at least 4 points).
Scoring systems give greater discrimination between different foods than threshold
models. Therefore in some circumstances they could more readily be used to create
incentives to industry to reformulate foods.
Cons: Scoring systems are necessarily more complex than threshold models, and their
development relies on more subjective judgements (for example, the width of the
nutrient bands, and the number of points required to be included in the definitions
such as ‘foods high in fat, salt or sugar’).
Complex scoring systems
Complex scoring systems are essentially scoring systems which use more complicated
methods than a series of bands to score foods for each nutrient. The aim of a complex
scoring system would be to rank foods by healthiness according to levels of the
selected nutrients, in a way that most accurately fits a pre-described mathematical
model.
Pros: Complex scoring systems are likely to produce the least amount of apparent
anomalies of the three types of model, as they will take the most amount of
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
52
information into consideration. For this reason, they may be the most accurate and
hence justifiable type of model.
Cons: These models could be very difficult to understand, and this may lead to
manufacturers being unable to reformulate foods to comply with them. The models
also require even more subjective judgements than simple scoring systems during
their development.
2.5.1.2 Comparing different types of model
The three types of model will produce different groups of foods that are categorised as
‘healthier food choices’ or ‘foods high in fat, salt or sugar’, even when starting with
the same nutrient criteria. This can be demonstrated with the following example.
Suppose that a definition of ‘healthier food choices’ is being developed. Two
nutrients are being considered: fat and sodium, and it is decided that the target levels
for these nutrients are less than 300mg sodium per 100g and less than 15g fat per
100g. The three following models could be developed:
Threshold model
:
A food is a ‘healthier food choice’ if it is:
<300mg sodium per 100g
AND
<15g fat per 100g
Simple scoring system:
A food is a ‘healthier food choice’ if it scores 2 points or less
Where: Sodium per 100g Fat per 100g
0-100mg = 0 points 0-5g = 0 points
100-200mg = 1 point 5-10g = 1 point
200-300mg = 2 points 10-15g = 2 points
300mg+ =3 points 15g+ = 3 points
Complex scoring system
:
A food is a ‘healthier food choice’ if (salt points)2 + (fat points)2 < 9
Where Salt points: Fat points:
x/100, where x = salt content in mg per 100g x/5, where x = fat content in g per 100g
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
53
The boundaries for foods to be included in the definition of ‘healthier food choice’
for each model type are shown in Figure 3 on the following page.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
54
Figure 3. Boundaries for foods to be included in the definition of ‘healthier food
choice’ for a threshold model, a simple scoring system model, and a complex
scoring system model
Threshold model
0
50
100
150
200
250
300
350
0 5 10 15 20
Fat (g per 100g)
Sodium (mg per 100g)
Inc / Exc Boundary
Roast Chicken
Chocolate Fudge
Cake
Beef curry
Irish Stew
Simple scoring system
0
50
100
150
200
250
300
350
0 5 10 15 20
Fat (g per 100g)
Sodium (mg per 100g)
Inc / Exc Boundary
Roast Chicken
Chocolate Fudge
Cake
Beef Curry
Irish Stew
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
55
Complex scoring system
0
50
100
150
200
250
300
350
0 5 10 15 20
Fat (g per 100g)
Sodium (mg per 100g)
Inc / Exc Boundary
Roast Chicken
Chocolate Fudge
Cake
Beef Curry
Irish Stew
The figures show that all three models would define roast chicken and beef curry as
‘healthier food choices’, but that a threshold model would also include chocolate
fudge cake and Irish stew within its definition of a ‘healthier food choice’, and a
simple scoring system would exclude Irish stew.
The literature review shows that a threshold model is used most commonly for
nutrient profiles, with one or two schemes attempting to use scoring systems.
The Expert Group agreed that in the first place threshold models should be
developed and tested but that scoring systems should also be considered.
2.5.2 Stage 3b: Food category specific criteria or across the board
criteria?
All three types of models can, in theory, involve food category specific criteria, or
across the board criteria.
2.5.2.1 Pros and cons of food category specific criteria and across
the board criteria
Food category specific criteria
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
56
There are different ways of categorising foods e.g. on the basis of how they are eaten
(e.g. ingredients, constituents of meals, whole meals) or by making assumptions about
their composition (e.g. the ‘food groups’ of the Balance of Good Health or the
categories of the Food Standards Agency’s Salt Model (FSA, 2003)).
Many existing examples of nutrient profiles involve food category specific criteria
e.g. the Australian Heart Foundation’s Pick the Tick scheme, the Swedish Green Key
Hole scheme (see Section 1).
Pros: Food category specific criteria take account of the fact that certain food
categories are higher in certain nutrients than others. Using food category specific
criteria would mean, for example, that it would be easier to ensure that a definition of
‘healthier food choices’ includes the healthier choices from each food category. This
might increase the incentive to manufacturers to reformulate foods, in order for them
to be included by the model.
Categorising foods for the purposes of developing nutrient profiles – albeit nutrient
profiles involving only one nutrient – has already been done for the purpose of
developing targets for sodium levels in foods with the Food Standards Agency’s Salt
Model.
Cons: Without using the same criteria for all food categories consumers may believe
that healthier food choices from each food category are as healthy as each other which
would not necessarily be the case.
It is not clear whether the categories used elsewhere e.g. for the Balance of Good
Health or the FSA’s Salt Model, would be appropriate for the purpose of defining
‘foods high in fat, salt or sugar’ or ‘healthier food choices’.
Furthermore the Expert Group noted that categorising foods is difficult. For example
the ‘Balance of Good Health’ does not successfully split all foods into one and only
one category: many composite foods, such as pizza, are clearly composed of foods
from more than one group, and it is unclear where some relatively simple foods such
as currant buns and baked beans belong.
Across the board criteria
Many existing examples of nutrient profiles use across the board criteria (see Section
1). For example the US Nutrition Labelling and Education Act specifies that a claim
that a food that helps reduce the risk of heart disease must be < 3g per serving of fat,
< 1g per serving of saturated fat and < 480 mg per serving of sodium regardless of
the type of food.
Pros: Across the board criteria are simple and allow consumers to make
comparisons across food categories.
Cons: Across the board criteria – particularly if they only take into consideration a
few nutrient levels - tend to generate more apparent anomalies.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
57
The literature review found that nutrient profiling schemes are evenly split between
those which use across the board criteria or food category specific criteria.
The Expert Group agreed that across the board criteria should be developed and
tested in the first instance.
2.6 Stage 4. Choice of numbers
The levels set for the thresholds of the individual nutrient criteria (or points scored for
a particular level) can be pragmatically chosen, taken from respected sources, or
linked to public health recommendations.
2.6.1 Pros and cons of different ways of choosing numbers
Pragmatically chosen levels
As long as a database of foods to test potential models is available, it is possible to set
the nutrient criteria so that a certain pre-selected proportion of foods is included
within a definition.
Pros: Pragmatically set thresholds can be set to achieve any objectives that are
required. e.g. equal levels of inclusion/exclusion for each nutrient considered
could be achieved, or the levels can be set to coincide with manufacturers’ views
of what are attainable targets.
Cons: The model would be hard to justify on public health grounds. If equal levels
of inclusion/exclusion for each nutrient is an aim when setting nutrient profiles,
then it is unlikely that the criteria would deliver equal health benefits across
nutrients.
Levels derived from respected sources
Nutrient profiles have been used around the world for a variety of purposes but mostly
when making claims about the health benefits of foods (see Section 1). The
originators are often government departments (e.g. in the process of setting criteria for
the use of health or nutrition claims), public health organisations or charities (e.g. in
developing health-related endorsement schemes such as the Swedish Green Key Hole
and the Australian Pick the Tick schemes). Some retailers and manufacturers in the
UK have developed their own schemes.
Pros: The criteria are justifiable on the basis that they have been tried, and to some
degree, tested elsewhere.
Cons: The criteria do not normally bear any transparent or logical relationship
with dietary recommendations. If a model is created which uses nutrient criteria
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
58
from different sources then it may lead to an unbalanced level of
inclusion/exclusion across nutrients, and they may not deliver equal health
benefits across nutrients.
Levels linked to public health recommendations
Threshold levels or points scored can be explicitly and logically linked to public
health recommendations in various ways.
Across the board criteria set on per 100kJ basis can be linked simply by taking
specified proportions of population dietary goals if these are set on a per energy basis
(as is generally the case with fat, saturated fat and NME sugar). This is the way the
Coronary Prevention Group Nutrition Banding scheme is linked to public health
recommendations.
Across the board criteria set on a per 100g or per serving basis can similarly be linked
to population dietary goals by first converting these into Guideline Daily Amounts
(GDAs) and then setting the levels as specified proportions of the GDAs. This was
the approach used when developing the FSA’s advice for what counts as ‘a lot’ or ‘a
little’ of certain nutrients (Rayner et al, 2004).
Food category specific criteria can be linked to public health recommendations by
deriving targets for desirable levels in food categories from modelling what needs to
be achieved to reach desirable intake levels from current intake levels (as with the
FSA Salt model).
Pros: The criteria are justifiable since they are transparently related to dietary
recommendations. If the same method of linking to the dietary recommendation is
used for each nutrient, then the same health benefits should be achieved across
each nutrient. Linking to public health recommendations is surprisingly flexible,
since any proportion of/relation to the recommendation can be used so long as the
same proportion is used for each nutrient.
Cons: If the health benefits for each nutrient are equal, then it is likely that there
will be an unbalanced inclusion/exclusion of foods across nutrients.
The Expert Group agreed that threshold levels and points scored should bear a
transparent relationship to agreed public health recommendations but that the
numbers chosen in this way should be checked against those used by respected
sources.
Nutrient profiles for food promotion to children: Section 2: Theoretical approach
59
2.7 References
Coronary Prevention Group (1988) Nutritional Labelling of Foods: A Rational
Approach to Banding. London: CPG.
European Commission (2003) Proposal for a regulation of the European Parliament
and of the Council on nutrition and health claims made on foods. Brussels, 16.7.2003
COM(2003) 424 final.
Food Standards Agency (2002) Labelling claims. London: FSA (and ‘Salt’, ‘Sugar’
and ‘Fat’ in this series.)
Food Standards Agency (2003) Salt in processed food: modelling food and intake
reductions. London: FSA.
Gregory J, Collins D, Davies P, Hughes J, Clarke P (1995) National Diet and
Nutrition Survey: children aged 1 ½ to 4 ½ years. Volume 1: Report of the diet and
nutrition survey. London: HMSO.
Gregory J, Lowe S (2000) National Diet and Nutrition Survey: young people aged 4
to 18 years. Volume 1: report of the diet and nutrition survey. London: HMSO.
House of Commons Health Committee (2004) Obesity. Third Report of Session 2003-
04. London: The Stationery Office.
Laville, M. (2004) Could glycaemic index be the basis of simple nutritional
recommendations? British Journal of Nutrition 91: 803-4
O’Neill M. (2004) Traffic lights for food? How nutrient profiling can help make
healthy choices become easy choices. London: National Consumer Council.
Prentice A, Jebb S (2003) Fast foods, energy density and obesity: a possible
mechanistic link. Obesity Reviews 4: 187-94.
Rayner M, Scarborough P, Williams C (2004) The origin of Guideline Daily Amounts
and the Food Standards Agency’s guidance on what counts as a lot and a little, Public
Health Nutrition 7 (4): 549-556.
Nutrient profiles for food promotion to children: Section 3: Development and testing I
60
SECTION 3. DEVELOPMENT AND TESTING OF
POSSIBLE DEFINTIONS 1
3.1 Summary
This section describes the development and testing of possible definitions of ‘foods
high in fat, salt or sugar’ and ‘healthier food choices’ for the purpose of advice on the
use of nutrition and health claims on foods aimed specifically at children; and advice
on the balance of TV advertising for foods during children’s TV programmes.
28 definitions of ‘foods high in fat, salt or sugar’ were developed on the basis of the
recommendations of the Expert Group using the theoretical approach described in the
previous section. On the basis of testing the 28 definitions, eight definitions were
selected for development into complete models with definitions of ‘foods high in fat,
salt or sugar’, ‘healthier food choices’ and ‘intermediate foods’.
On the basis of the testing of these 28 definitions and the eight complete models the
Expert Group agreed that three models were worthy of further consideration and
should be tested further and if possible refined i.e.
§ Model TAg (Threshold model, Group A nutrients (energy, saturated fat,
non-milk extrinsic (NME) sugar and sodium) per 100g base)
§ Model TCg (Threshold model, Group C nutrients (energy, saturated fat,
NME sugar, sodium, fruit and vegetables, n-3 fatty acids, calcium and
iron), per 100g base)
§ Model SSCg (Scoring system, Group C nutrients, per 100g base) 6
Model TAg was chosen because it is relatively simple but the Expert Group
noted that it is not particularly accurate. Model TCg was chosen because,
although slightly more complex than Model TAg, it is more accurate. Similarly
Model SSCg was chosen because, although this specific model was no more
accurate than Model TCg, scoring systems were generally found to improve the
accuracy of models.
3.2. Introduction
3.2.1 Recommendations of the Expert Group
In Section 2 it was noted that, for the purposes of developing definitions of ‘foods
high in fat, salt or sugar’ and ‘healthier food choices’ for testing, the Expert Group
agreed that:
Ages of children
The project should focus on the development and testing of definitions for ‘foods high
in fat, salt or sugar’ and ‘healthier food choices’ for children aged 11-16. But noted
6 For the numbers see Appendix 6
Nutrient profiles for food promotion to children: Section 3: Development and testing I
61
that these should be applicable to children of any age between 5 and 16 and also to
adults.
The intended uses of definitions and models
The project should focus on developing and testing definitions of ‘foods high in fat,
salt or sugar’ and ‘healthier food choices’ for the purpose of:
Advice on the use of nutrition and health claims on foods aimed specifically at
children; and
Advice on the balance of TV advertising for foods during children’s TV
programmes.
Choice of nutrients
The nutrients (and other components of foods) for the criteria should be:
Energy, saturated fat, non-milk extrinsic (NME) sugar and sodium (‘A
nutrients)
Fruit and vegetables and long chain n-3 polyunsaturated fatty acids (n-3 fatty
acids) (‘B’ nutrients)
Calcium and iron (‘C’ nutrients)
That these nutrients be used in the following combinations:
Group A: just ‘A’ nutrients
Group B: ‘A’ nutrients + ‘B’ nutrients
Group C: ‘A’ nutrients + ‘B’ nutrients + ‘C’ nutrients
That for the purposes of this project:
Choice of base
That the following bases should be used:
§ per 100g
§ per 100kJ
§ per 100g AND/OR per serving
§ per 100kJ AND/OR per serving
That when combining per 100g and per serving criteria the OR operator should be
used for definitions of ‘foods high in fat, salt or sugar’ and the AND operator for
‘healthier food choices’ but conversely when combining per 100kJ and per serving
‘Foods high in fat, salt or sugar’
would need to be
above defined energy OR saturated fat OR NME sugar OR
sodium levels
below defined levels of fruit and vegetables AND n-3 fatty acids
AND calcium AND iron (where the definition involved criteria
for these nutrients/components)
‘Healthier food choices’ would need to be:
below defined energy AND saturated fat AND NME sugar AND
sodium levels
above defined levels of fruit and vegetable OR n-3 fatty acids OR
calcium OR iron (where the definition involved criteria for these
nutrients/components)
Nutrient profiles for food promotion to children: Section 3: Development and testing I
62
criteria the AND operator should be used for definitions of ‘foods high in fat, salt or
sugar’ and the OR operator should be used for foods for ‘healthier food choices’.
Choice of model
That in the first place threshold models should be developed and tested but that
scoring systems should also be considered. That across the board criteria should be
tested in the first instance rather than food category specific criteria
Choice of numbers
That the levels set for the thresholds of the individual nutrient criteria (or points
scored for a particular level) should bear a transparent relationship to agreed public
health recommendations but that the numbers chosen in this way should be checked
against those used by respected sources.
3.2.2 Consequences of the Expert Group’s recommendations
In view of the Expert Group’s recommendations it was necessary to select agreed
population dietary goals and develop Guideline Daily Amounts (GDAs) for selected
nutrients and other food components for children aged 11-16. The population dietary
goals and the GDAs for this age-group are given in Table 2.
Table 2. Population dietary goals and Guideline Daily Amounts for children
aged 11 – 16
Nutrient Source of
population
dietary goal
Population dietary goal GDA
Energy 1 (a) 2130 kcal 2130 kcal
Saturated fat 1 (b) 11% food energy 26 g
NME sugar 1 (c) 11% food energy 63 g
Sodium 2 (d) 2.35 g/day 2.35 g
Fruit and
vegetables 3 (e) 380 g/day 380 g
Calcium 1 (f) 690 mg/day 690 mg
Iron 1 (g) 10 mg/day 10 mg
Long chain n-3
polyunsaturated
fatty acids
4 (h) 0.4 g/day 0.4 g
Sources:
1. Department of Health, Dietary Reference Values for Food Energy and Nutrients for the United
Kingdom. Report on Health and Social Subjects No 41. London: HMSO, 1991;
2. Department of Health. Nutritional Aspects of Cardiovascular Disease. London: HMSO, 1994.
3. Scientific Advisory Committee on Nutrition. Salt and Health. London: The Stationery Office, 2003.
4. Scientific Advisory Committee on Nutrition. Advice on Fish Consumption: Benefits and Risks.
London: The Stationery Office, 2004
Notes:
(a) Weighted average of 2220 kcal/day (Estimated Average Requirement (EAR) for boys aged 11-14);
1845 kcal/day (EAR for girls aged 11-14); 2755 kcal/day (EAR for boys aged 15-18); 2110 kcal/day
(EAR for girls aged 15-18)
Nutrient profiles for food promotion to children: Section 3: Development and testing I
63
(b) GDA calculated using EARs for energy as in note (a) and a conversion factor of 9 kcal for 1 g fat.
(c) GDAs calculated using EARs for energy as in note (a) and a conversion factor of 3.75 kcal for 1 g
sugar.
(d) The Scientific Advisory Committee on Nutrition (SACN) recommends; 6g/day for children aged 11
and over.
(e) The Committee on Medical Aspects of Food and Nutrition Policy (COMA) recommends a 50%
increase in consumption of fruit and vegetables. This has been translated into 5 servings a day or 400g.
Here this amount in g has been adjusted to take account of the lower energy needs of children.
(f) Weighted average of: 750 mg/day (EAR for boys aged 11-14); 625 mg/day (EAR for girls aged 11-
14); 750 mg/day (EAR for boys aged 15-18); 625 mg/day (EAR for girls aged 15-18).
(g) Weighted average of 8.7 mg/day (EAR for boys aged 11-14); 11.4 mg/day (EAR for girls aged 11-14);
8.7 mg/day (EAR for boys aged 15-18); 11.4 mg/day (EAR for boys aged 15-18).
(h) SACN recommends that the average ‘population’ intake should be 0.45 g/day. Here this amount in g
has been adjusted to take account of the lower energy needs of children.
In establishing standard relations to dietary recommendations the three types of
nutrients were treated differently.
For criteria for ‘A’ nutrients (energy, saturated fat, NME sugar and sodium) set per
100g, the advice of the Food Standards Agency (FSA) on what counts as ‘a lot’ or ‘a
little’ was used as the starting point for thresholds (Food Standards Agency, 2002). ‘A
lot’ (according to the FSA) is 20% of the GDA and ‘a little’ is 3.3% (Rayner et al,
2004). For a per 100kJ base the Coronary Prevention Group (CPG) Nutrition
Banding Scheme definitions of ‘high’ and ‘medium low’ was used as the starting
point (Coronary Prevention Group, 1988). ‘High’ (according to the CPG) is more
than 150% of the population dietary goal, ‘medium low’ is less than 100% of the
population dietary goal (see Appendix 2 for details of the scheme).
It was clear from previous work that these choices of numbers would not be optimum
for definitions of ‘foods high in fat, salt or sugar’ or ‘healthier food choices’ (and see
results below) so other standard proportions of GDAs and population dietary goals
were also tested. It is also important to note that ‘A’ nutrients include energy.
Neither the FSA nor the CPG give guidance on energy. In addition the CPG
definitions are set on per 100kJ basis. This meant that definitions using just a per
100kJ basis did not include energy as a criterion.
The proportion of GDAs used as a threshold was kept the same for each ‘A’ nutrient,
so that the health benefits suggested by the models would be the same for each
nutrient. The same principle was also applied to the ‘C’ nutrients (i.e. calcium and
iron). Calcium and iron, however, were included in order to identify foods which
contain a high enough level of these nutrients to ‘counterbalance’ high amounts of ‘A’
nutrients, and it was considered that this level should be set at a higher proportion of
the GDA than for the ‘A’ nutrients.
In order to decide which proportions of the GDAs were appropriate starting points for
definitions involving per serving criteria for ‘A’ nutrients and for all criteria involving
‘C’ nutrients distribution graphs of the foods from the Balance of Good Health
subsets of the database (see Section 3.2 below) were produced for each of the ‘A’ and
‘C’ nutrients, measured per 100g, per 100kJ and per serving. From studying these
graphs, it was possible to estimate how many foods from each subset would be
classified as ‘healthier food choices’ or ‘foods high in fat, salt or sugar’ depending on
Nutrient profiles for food promotion to children: Section 3: Development and testing I
64
where the thresholds were set and to make an initial judgement about where levels
should be set.
For example, by using these distribution graphs, it was judged that one third of the
GDA was a good starting point for thresholds for ‘A’ nutrients measured per serving
for definitions of ‘foods high in fat, salt or sugar’ (and one sixth for ‘healthier food
choices’ thresholds). For ‘C’ nutrients, 30% of the GDA per 100g, 6% per 100kJ,
and 50% per serving were considered to be good starting point for thresholds for both
definitions of ‘foods high in fat, salt or sugar’ and of ‘healthier food choices’.
In essence the ‘B’ nutrients were selected in order to identify specific foods (i.e. oily
fish, fruit and vegetables) and prevent them from being classified as ‘foods high in fat,
salt or sugar’. Because of this, the thresholds were set at a level that would include
these foods and exclude all others. Therefore the starting point used for thresholds
for definitions of both ‘foods high in fat, salt or sugar’ and of ‘healthier food choices’
for n-3 fatty acids was 25% of the GDA per 100g; 50% of the GDA per serving and
7.5 % of the GDA per 100kJ. Similarly the starting point used for fruit and vegetable
thresholds was 25% of the GDA per 100g.
The Department of Health’s advice is to eat at least five portions of fruit and
vegetables a day. This is equivalent to the 400g per day recommended by COMA.
25% of the GDA per 100g therefore works out at 100g per 100g. Given the relatively
crude methods of testing fruit and vegetable criteria (see below) no attempt was made
to develop fruit and vegetable criteria on a per serving or per 100kJ basis.
Since there were three agreed options for combinations of nutrients, four agreed
options for bases or combination of bases and two possibilities for type of model
(threshold and scoring system) the process of developing different models with their
associated definitions for testing was initiated by generating 24 (3 x 4 x 2) different
definitions for ‘foods high in fat, salt or sugar’.
For the four definitions involving Group A nutrients two different choices of numbers
were tested. So in total 28 different definitions of ‘foods high in fat, salt or sugar’
were tested.
On the basis of the results of this testing eight definitions were selected for
development into complete models with definitions of ‘foods high in fat, salt or
sugar’, ‘intermediate foods’ and ‘healthier food choices’.
The details of each of these models, and the abbreviations used to describe them in
this section are shown in Appendix 4.
3.3 Methods for testing definitions
The method used for testing definitions are based on previous work in developing
criteria for use of the 5 A DAY logo on composite foods containing fruit and
vegetables (Stockley et al, 2003) and on other earlier work (O’Neill, 2004).
Nutrient profiles for food promotion to children: Section 3: Development and testing I
65
3.3.1 Development of a database of foods
For the purpose of this project a database of foods using McCance and Widdowson’s
The Composition of Foods Sixth Summary Edition (Roe et al, 2003) was developed.
This contains nutrient composition data for 1,235 foods but no information about
serving sizes, NME sugars, fruit and vegetable levels in composite foods or n-3 fatty
acids. Furthermore for many foods there is more than one entry. There is also often
a duplication for the cooked and uncooked versions of a food.
In order to render the McCance and Widdowson database more suitable for testing
definitions, firstly 205 foods were removed that were inedible (e.g. raw meat, raw fish
and uncooked pasta, some raw vegetables such as aubergine but not carrots, and raw
inedible ingredients such as flour and sunflower oil but not butter). This left 1,030
foods.
Secondly, for each food a serving size was calculated. Information about serving
sizes was gathered from the following sources: the MAFF guide to food portion sizes
(940 foods) (Ministry of Agriculture, Fisheries and Food, 1993); the Department of
Health’s guide to a portion of fruit or vegetables (62 fruits and vegetables for which
the MAFF guide provides no details) 7; Kellogg’s website (one food); Tesco’s website
(14 foods); Sainsbury’s website (eight foods); and NutritionData.com (four rare
foods: chicken skin, turkey skin, pheasant and pumpkin).
All of these serving sizes were for adults, not children. It is unlikely that this would
substantially alter the results, as the GDA for energy for children aged 11-16 (Table
2) is about 95% of the GDA for energy for adults.
Thirdly the NME sugar content for each food was calculated based on the FSA’s
methods for calculating the NME sugar content of foods (Buss et al, 1994). For raw
and boiled fruits and vegetables (including nuts and potatoes): none of the sugars
were counted as NME sugar. For canned, stewed, grilled, fried, baked, dried or
preserved fruit or vegetables: half of all sugars were counted as NME sugar. For
foods (excluding drinks) estimated to contain at least one portion of fruit or
vegetables (i.e. fresh equivalent weight of 80g) per serving of the food: half of all
sugars minus lactose were counted as NME sugar. (Foods were categorised as
containing more or less than one portion of fruit and vegetables using data from an
unpublished report from Leatherhead Food RA (Angus, 2003)). For all other foods
(including fruit juices): all of the sugars minus lactose were counted as NME sugar.
Fourthly the percentage (by weight) of fruit and vegetables (excluding potatoes) for
each food in the database was estimated. As there was no ingredient composition
data for the vast majority of foods, each food was evaluated and assigned one of five
values: 0% (for foods consisting of approximately 0-20% fruit or vegetables
(excluding potatoes)); 30% (approximately 21-40%); 50% (approximately 41-60%);
70% (approximately 61-80%); and, 100% (approximately 81-100%).
0% foods consist of breads, cereals, potatoes, milk, most dairy products, meat, fish,
nuts, most fatty and sugary foods, and some composite foods. 30% foods consist of
7 Department of Health website - http://www.doh.gov.uk/fiveaday/portions.htm (20/01/04)
Nutrient profiles for food promotion to children: Section 3: Development and testing I
66
some yoghurts, carbonated fruit juice drinks, jams, fruit pies, vegetable pasties,
lasagne, takeaway burgers, etc. 50% foods consist of sauces, fruit crumbles,
chutnies, etc. 70% foods consist of vegetable based dishes such as cauliflower
cheese, vegetable soup or hummus. 100% foods consist of all fruit and vegetables
(however they are cooked), beans and fruit juices.
Fifthly the n-3 fatty acid composition data was obtained from the 7th supplement of
the 5th edition of the McCance and Widdowson series (Ministry of Agriculture,
Fisheries and Food, 1998). This provides data on long chain n-3 polyunsaturated
fatty acids for 522 different foods. For the purpose of this project long chain n-3
polyunsaturated fatty acids were defined as in the recent report of the Scientific
Advisory Committee on Nutrition (i.e. C20:5, C22:5 and C22:6 fatty acids) (Scientific
Advisory Committee on Nutrition, 2004). Data on n-3 fatty acid composition was
transferred to the existing database by relating the foods from the 7th supplement to
their closest equivalents in the existing database (e.g. ‘Haddock, raw’ from the 7th
supplement was related to ‘Haddock, steamed’). Of the 522 foods from the 7th
supplement, 34 did not have an obvious equivalent in the existing database, and were
hence excluded. This meant that 542 foods in the existing database had no equivalent
foods in the 7th supplement for data to be transferred. These 542 foods were assumed
to have 0g of n-3 fatty acids per 100g, per 100kJ and per serving, except for all fish or
fish products, which were recorded as ‘not sure’.
In testing definitions of ‘foods high in fat, salt or sugar’ or complete models with
definitions of ‘foods high in fat, salt or sugar’, ‘intermediate foods’ and ‘healthier
food choices’ firstly the numbers of foods that were categorised by the definitions
were examined. It should be noted that since the modified McCance and Widdowson
database bears no relation to current or desirable food purchasing or food intake
patterns it is difficult to interpret the absolute numbers of foods classified by a
definition or a complete model. This would be improved by developing more
sophisticated databases for testing (see Section 5).
However, comparison between models suggests which models are most likely to be
useful to consumers in different circumstances, e.g. consumers looking for the
healthiest possible foods would be best served by models with fewer foods classified
as ‘healthier food choices’; but if they are just looking for generally healthier foods
then they would be best served by models with more foods classified as ‘healthier
food choices’. Furthermore it is suggested that models which put very large numbers
of foods into one category would seem to be less than optimal.
3.3.2 Categorisation of foods in the database into Balance of Good Health
food groups and selection of indicators for food groups.
Using a Food Classification List (Gatenby et al, 1994) developed for use with the
Balance of Good Health all foods in the database were classified into six categories:
the five food groups of the Balance of Good Health and composite foods.
For each of the five food groups of the Balance of Good Health eight indicator foods
were selected – representative of the variation of nutrient profiles for the foods in the
Nutrient profiles for food promotion to children: Section 3: Development and testing I
67
group. These indicator foods are listed below, as they appear in the results tables
(with complete McCance and Widdowson titles in brackets, if different):
Bread, other cereals and potatoes:
Roast potatoes (old potatoes, roast in blended oil); fresh pasta (pasta,
plain, fresh, cooked); wholemeal rolls; cream crackers; sugar puffs; corn
flakes; oven chips (oven chips, frozen, baked); currant buns
Milk and dairy foods:
Half fat creme fraiche; semi-skimmed milk (semi-skimmed milk,
average); whole milk (whole milk, average); low fat yoghurt (yoghurt,
low fat, fruit); greek yoghurt (greek yoghurt, sheep); cheddar;
camembert; cottage cheese (cottage cheese, plain)
Fruit and vegetables
Canned pineapple (pineapple, canned in juice); orange juice (orange
juice, unsweetened); peaches (peaches, raw); currants; grilled tomato;
avocado (avocado, average); celery (celery, raw); lettuce (lettuce, raw)
Meat, fish and alternatives:
Roast lamb (lamb, leg, lean only, roast); chicken leg (chicken, leg
quarter, roasted, meat and skin); lentils (lentils, red, split, dried, boiled
in unsalted water); cod (cod, baked); mackerel (mackerel, grilled);
beefsteak (beef, rump steak, lean and fat, fried); walnuts; boiled egg
(egg, chicken, boiled)
Foods containing fat, foods containing sugar
Choc chip cookies; Mars bar; cola; jam doughnuts; olive oil; butter; low
fat spread (fat spread (20-25% fat), polyunsaturated); crisps (potato
crisps)
The classification of each of these indicators was assessed for each of the 28
definitions of ‘foods high in fat, salt or sugar’ and the eight complete models with
definitions of ‘foods high in fat, salt or sugar’ and ‘healthier food choices’.
3.3.3 Development of ‘healthier’ and ‘less healthy’ indicator foods
A set of ‘healthier’ and ‘less healthy’ indicator foods had previously been developed
for the purpose of testing definitions of ‘unhealthy’ food (O’Neill, 2004). The aim
was to develop two lists of foods: a list of foods that would be subjectively regarded,
by most people, as ‘healthier’ and another list which would be regarded as ‘less
healthy’. In order to identify the indicator foods, the foods in the modified McCance
and Widdowson database were categorised by two researchers into three categories –
‘healthier’, ‘less healthy’ and ‘neither healthy nor unhealthy’. The categorisation was
done solely on the basis of the names of the foods and prior knowledge, and without
any further information such as the precise details of the nutrient composition.
Nutrient profiles for food promotion to children: Section 3: Development and testing I
68
The two researchers agreed on 143 ‘healthier’ and 98 ‘less healthy’ indicator foods,
and that 630 foods were ‘neither healthy nor unhealthy’. Inter-rater reliability was
determined in the standard way and was ‘good’ (? = 0.69) (Altman, 1991).
Foods that were considered similar enough to others in the list to be duplicates (for
example, ‘Apples, eating, average, raw, peeled’ was considered a duplicate of
‘Apples, eating, average, raw’) were then removed from the agreed lists of ‘healthier’
and ‘less healthy’ indicator foods. The researchers then discussed all disputed
categorisations and agreement was reached on all of them, leaving 128 ‘healthier’
indicators and 89 ‘less healthy’ indicators. Finally, nine foods were removed for
which there was insufficient compositional data in the database to test the models.
This left 124 ‘healthy’ indicators and 84 ‘unhealthy’ indicators. (See Appendix 3 for
lists of ‘healthier’ and ‘less healthy’ indicator foods.)
For the current project, the average percentage of ‘healthier’ and ‘less healthy’
indicators misclassified by the 28 definitions for ‘foods high in fat, salt or sugar’
definitions was calculated.
In testing the eight complete models a measure of the specificity of the model (the
average percentage of correct classifications of ‘healthier’ and ‘less healthy’
indicators) and a measure of the sensitivity of the model (100 – the average
percentage misclassification of ‘healthier’ and ‘less healthy’ indicators) were
calculated.8
A high level of both specificity and sensitivity is needed if a definition or model can
be said to be accurate. This is because high specificity can be achieved by reducing
sensitivity and vice versa. For example a definition of a ‘food high in fat, salt or
sugar’ which was highly specific – classifying 100% of the ‘less healthy’ indicators
correctly – could easily be devised just by having very low thresholds for energy,
saturated fat, NME sugars and sodium. But such a definition would be very
insensitive because it would thereby classify a large proportion of the ‘healthier’
indicators as ‘foods high in fat, salt or sugar.’
In testing the eight models we therefore assessed their overall sensitivity and
specificity with an ‘accuracy score’. This was defined as the average percentage of
‘healthier’ and ‘less healthy’ indicators correctly classified (a measure of the
specificity of the model) minus the average of the percentage of ‘healthier’ indicators
classified as ‘less healthy’ and of the percentage of ‘less healthy’ indicators classified
as ‘healthier’ (a measure of its sensitivity).
It should be noted that our measures of specificity, sensitivity and overall accuracy are
relatively crude and that the panels of ‘healthier’ and ‘less healthy’ indicators used
here could be further refined. (For example many nutritionists would not agree with
8 Sensitivity is defined as ‘the proportion of positives that are correctly identified by a test’ specificity
is defined as ‘the proportion of negatives that are correctly identified by a test’ (Altman. 1991).
Normally sensitivity and specificity are applied when there is just one test for just one condition and
where the prevalence of that condition is known (e.g. when screening for a disease in a population).
The eight models, in effect, involve two tests one for ‘healthiness’ and one for ‘unhealthiness’.
Moreover we have no idea of the true numbers of ‘healthy’ and ‘unhealthy’ foods in the database. So
our tests of specificity and sensitivity are somewhat different to those usually used.
Nutrient profiles for food promotion to children: Section 3: Development and testing I
69
the selections because, for instance the ‘healthier’ indicators include no dairy products
and no meats.) This situation would be improved by developing more sophisticated
panels of indicators (see Section 5).
It should also be noted that our specificity, sensitivity and overall accuracy scores take
no account of how models or definitions classify foods in the database that have not
been designated as ‘healthier’ or ‘less healthy’ indicators and no confidence limits to
scores have been calculated.
Nevertheless, the scores probably provide a better indication of the accuracy of the
models than inspecting how foods (such as the representative of the food groups of
the Balance of Good Health) are classified. For example, suppose one model were to
classify olive oil as a ‘food high in fat, salt or sugar’ and another model classified it as
a ‘healthier food choice’. Which model would be more accurate? It would be hard
to say because it is difficult to be definitive about the health benefits of olive oil.
This is true of many foods of the selected representatives of the food groups of the
Balance of Good Health, for example cheddar cheese, currant buns and boiled eggs.
Making a judgement about the accuracy of models, based on inspecting the way a
model classifies particular foods involves making subjective judgements. The use of
specificity, sensitivity and accuracy scores reduces that subjectivity to some degree.
This being said inspection of the ways models classify foods is sometimes useful for
detecting how models might be improved (see Section 3.4 below).
3.4 Results
3.4.1 Testing definitions of ‘foods high in fat, salt or sugar’
Appendix 5 gives the results of testing each of the 28 different definitions of ‘foods
high in fat, salt or sugar’. Tables 3 and 4 summarise the results of the Tables in
Appendix 5.
Table 3. Percentage of foods classified as ‘foods high in fat, salt or sugar’ by 28
definitions
Group A nutrients %Group B nutrients %Group C nutrients %
Threshold models
per 100g TAg 1 & 2 47 & 60 TBg 55TCg 47
per 100g/serving TAg/s 48 TBg/s 51TCg/s 40
per 100kJ TAkJ 1 & 2 70 & 46 TBkJ 59TCkJ 50
per 100kJ/serving TAkJ/s 41 TBkJ/s 54TCkJ/s 42
Simple scoring systems
per 100g SSAg 1 & 238 &
47 SSBg 64SSCg 57
per 100g/serving SSAg/s 1 & 245 &
39 SSBg/s 49SSCg/s 53
per 100kJ SSAkJ 61 SSBkJ 47SSCkJ 42
per 100kJ/serving SSAkJ/s 48 SSBkJ/s 50SSCkJ/s 50
Nutrient profiles for food promotion to children: Section 3: Development and testing I
70
Table 4. Average percentage of indicator foods misclassified by 28 definitions of
‘foods high in fat, salt or sugar’
Group A nutrients %Group B nutrients %Group C nutrients %
Threshold models
per 100g TAg 1& 213 & 10 TBg 7TCg 6
per 100g/serving TAg/s 10 TBg/s 8TCg/s 14
per 100kJ TAkJ 1 & 2 13 &
22 TBkJ 6TCkJ 7
per 100kJ/serving TAkJ/s 14 TBkJ/s 4TCkJ/s 7
Simple scoring systems
per 100g SSAg 1 & 2 13 & 9 SSBg 5SSCg 8
per 100g/serving SSAg/s 1 & 2 10 & 9 SSBg/s 5SSCg/s 7
per 100kJ SSAkJ 10 SSBkJ 10SSCkJ 9
per 100kJ/serving SSAkJ/s 13 SSBkJ/s 3SSCkJ/s 3
The definitions that were selected for developing into complete models are
highlighted in bold in both Tables 3 and 4.
3.4.1.1 Choice of nutrients
Table 3 shows that choice of nutrients makes little difference to the percentage of the
foods in the database which are classified as ‘foods high in fat, salt or sugar’. The
definitions developed classify between 40% and 70% of the foods in the database as
‘foods high in fat, salt or sugar’, but this would seem to be a function of the choice of
numbers rather than the choice of nutrients (or base or model type).
Table 4 shows that choice of nutrients does tend to affect the extent to which
definitions misclassify indicator foods. In general definitions involving ‘B’ nutrients
and ‘C’ nutrients misclassify fewer indicator foods than definitions involving just ‘A’
nutrients.
This was to be expected. It had previously been observed that definitions of ‘foods
high in fat, salt or sugar’ involving criteria for ‘A’ nutrients are likely to put some
foods with high levels of desirable micronutrients into the same (‘less healthy’)
category as foods with low levels of such nutrients. For example Definition TAg/2 –
a threshold model based on FSA guidance for what counts as ‘a lot’ or ‘a little’ -
classifies Mars bars, cheddar, crisps and currants as ‘foods high in fat, salt or sugar’.
In contrast Definition TCg – a threshold model on a similar per 100g basis to
Definition TAg/2 but with additional criteria for ‘B’ and ‘C’ nutrients’ - classifies
Mars bars and crisps as ‘foods high in fat, salt or sugar’ but does not include cheddar
and currants.
3.4.1.2 Choice of base
Table 3 shows that choice of base makes little difference to the percentage of foods in
the database that are classified as ‘foods high in fat, salt or sugar’.
Nutrient profiles for food promotion to children: Section 3: Development and testing I
71
Table 4 shows that choice of base also seems to make little difference to the extent to
which definitions misclassify indicator foods.
Even using combinations of bases (per 100g and per serving; or per 100kJ and per
serving) seems to make little difference to the extent to which definitions misclassify
indicator foods. This can be seen by comparing pairs of definitions in Table 4 where
the nutrient group is the same, and the type of model is the same, but the first member
of each pair uses a single base (either per 100g or per 100kJ) and the second member
of each pair uses a combination of bases (either per 100g and per serving; or per
100kJ per serving).
This was somewhat unexpected since when nutrients are treated individually the base
has a substantial effect on whether a nutrient level is described as high/a lot or low/a
little. For example, according to FSA advice raw chicory contains ‘a little’ saturated
fat. But the CPG scheme categorises raw chicory as ‘high’ in saturated fat.
3.4.1.3 Choice of model
Table 3 shows that choice of model (whether threshold model or scoring system)
makes little difference to the percentage of foods in the database which are classified
as ‘foods high in fat, salt or sugar’.
However, Table 4 shows that the use of a scoring system leads to fewer
misclassifications of ‘healthier’ and ‘less healthy’ indicators. For example, if pairs of
definitions in Table 4, where the nutrient group and the base are the same, are
compared 9 then in nine cases out of 12 the scoring system gives fewer
misclassifications than the threshold model.
Definitions SSBkJ/s and SSCkJ/s have the fewest misclassifications (both 3%), and
are both scoring systems. However there are some threshold models where
misclassifications are low – e.g. Definition TBkJ/s which misclassifies only 4% of
indicators.
3.4.2 Testing complete models including definitions of ‘foods high in fat,
salt or sugar’, ‘intermediate foods’ and ‘healthier food choices’
On the basis of the testing of the 28 definitions, eight were selected for development
into complete models and for further analysis. Both the simplest definition (TAg:
Threshold model, Group A nutrients, per 100g base) that had been tested and the most
complex (SSCkJ/s: Scoring system, Group C nutrients, per 100kJ/serving base) were
selected. At least two definitions for each nutrient group were selected. Finally
definitions that misclassified the fewest ‘healthier’ and ‘less healthy’ food indicators
were selected.10
9 Where there were two definitions using the same nutrient group, the same base and the same type of
model the definition with the fewest misclassifications was used for the comparison.
10 With the wisdom of hindsight it would have been better to have selected TAg1 (rather than TAg2)
and SSCg (rather than SSCg/s) (See Section 3).
Nutrient profiles for food promotion to children: Section 3: Development and testing I
72
Appendix 6 gives the results of testing each of the eight models developed. Table 5
summarises the results of the Tables in Appendix 6:
Table 5. Results of testing eight different models
%
‘healthier
food
choices’ %
intermediate
%
‘high in fat,
salt or
sugar’
Sensitivity (%
correct
classifications)
Specificity (100 -
% misclass-
ifications)
Overall
accuracy
score *
TAg 11 42 47 66 96 62
TAkJ 19 11 70 83 90 73
TBg/s 36 13 51 92 96 88
TBkJ/s 40 7 53 95 99 94
TCg 21 34 45 81 96 76
SSBg 25 14 61 88 95 83
SSCg/s 31 17 52 88 96 83
SSCkJ/s 41 18 41 92 99 91
* Note that the accuracy score is not a percentage but the maximum accuracy score a
model could theoretically achieve is 100.
Table 5 shows that the eight models classify quite different numbers of foods as
‘foods high in fat, salt or sugar’ (range 41% - 70%), ‘healthier food choices’ (range
11% - 41%) and intermediate foods (range 7% to 42%). As explained earlier, it is
difficult to know what the ideal percentages should be, but it seems likely that the
extremes i.e. very small percentages (e.g. 7% for ‘intermediate foods’ with Model
TBkJ/s) or very large percentages (e.g. 70% for ‘foods high in fat, salt or sugar’ with
Model TAkJ) are inappropriate. This is particularly the case where three definitions
rather than just one definition are required of a model.
Using this reasoning Model TAg (Threshold model, Group A nutrients, per 100g base
and drawing on the FSA’s advice on what counts as ‘a lot’ and a ‘little’) would seem
to classify too few foods as ‘healthier food choices’ and Model TAkJ (Threshold
model, Group A nutrients, per 100kJ base and drawing on the CPG Nutrition Banding
Scheme) would seem to classify too many foods as ‘foods high in fat, salt or sugar’.
Table 5 shows that different models have different degrees of sensitivity and
specificity with respect to the ‘healthier’ and ‘less healthy’ food indicators. The
overall accuracy scores range from 62 (Model TAg) to 94 (Model TBkJ).
Nutrient profiles for food promotion to children: Section 3: Development and testing I
73
Figure. 4 Accuracy score plotted against % of
foods classified as 'intermediate' foods for 8
complete models
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50
% intermediate foods
Score
SSCkJ/s
TAkJ
Figure 4 shows that there is a strong, inverse correlation between the number of
intermediate foods classified by a model and accuracy scores. It might be expected
that the fewer foods a model classifies as ‘foods high in fat, salt or sugar’ and
‘healthier food choices’ the greater the accuracy. However, some models (e.g .
Model SSCkJ/s) have higher accuracy scores that might be expected on the basis of
the number of foods they classify as ‘intermediate’ foods and some have worse (e.g.
Model TAkJ).
Table 5 shows that some threshold models such as TCg (Threshold model, Group C
nutrients, per 100g base) classify foods reasonably well. It has previously been noted
(in Section 2.5.1) that threshold models are easier to understand than scoring systems.
However, the threshold models that were developed (even when involving ‘B’ and
‘C’ nutrients) still led to some misclassifications of ‘healthier’ and ‘less healthy
indicators. For example the TCg model classified bread and nuts as ‘foods high in
fat, salt or sugar’.
Table 5 also shows that scoring systems such as SSCkJ/s (Scoring system, Group C
nutrients, per 100kJ/per serving base) can be better than threshold models at
classifying foods. SSCkJ/s only misclassifies two of the ‘healthier’ and ‘less healthy’
indicator foods: porridge as a ‘food high in fat, salt or sugar’ and crispie cakes as
‘healthier’. However, as previously noted, scoring systems are more difficult to
understand than threshold models which may limit their use in some circumstances.
Nutrient profiles for food promotion to children: Section 3: Development and testing I
74
3.5 Conclusions and recommendations of the Expert Group
The Expert Group agreed that from this initial testing of 28 definitions of ‘foods high
in fat, salt or sugar’ and eight complete models with definitions of ‘foods high in fat,
salt or sugar’, ‘intermediate foods’ and ‘healthier food choices’ provided a useful
basis for further work. They identified three models (and their associated definitions)
for further consideration and possible refinement. The three models were:
§ Model TAg (Threshold model, Group A nutrients (energy, saturated fat,
NME sugar and sodium) per 100g base
§ Model TCg (Threshold model, Group C nutrients (energy, saturated fat,
NME sugar, sodium, fruit and vegetables, n-3 fatty acids, calcium and
iron), per 100g base)
§ Model SSCg (Scoring system, Group C nutrients, per 100g base)
The Expert Group noted that the initial testing of definitions described in this section
indicated that base or denominator seemed to make little difference to the accuracy of
models. Therefore the Expert Group felt that only models using a per 100g base
should be explored further on the basis that serving sizes are difficult to define and
that 100g is used as the basis of many other nutrient criteria (e.g. for criteria for the
use of nutrition claims in the proposed EU Regulation of nutrition and health claims
(European Commission, 2003)).
The initial testing found that using ‘A’ nutrients (energy, saturated fat, NME sugars
and sodium) alone led to models with relatively low accuracy, but the Expert Group
still thought that models using ‘A’ nutrients alone deserved further investigation on
the grounds that they were simpler than models involving ‘B’ and ‘C’ nutrients. The
Expert Group felt that models using ‘A’ nutrients alone might be made more accurate
though moving the thresholds for these nutrients.
On the other hand the Expert Group agreed that involving ‘B’ and ‘C’ nutrients does
improve the accuracy of models and so further testing of models involving C Group
nutrients was warranted.
The Expert Group also revisited the question of the choice of an energy criterion
rather than a total fat criterion. It was felt that the decision to calculate the energy
criterion for different models on the same basis as the criteria for saturated fat, NME
sugar and sodium criteria (i.e. as standard percentages of the GDA) might have led to
the energy thresholds being too high. Two possible solutions were recommended.
Firstly the energy criterion could be unlinked from the rest of the ‘A’ nutrients and set
at a different level. Secondly the energy criterion could be replaced with a total fat
criterion, which could be calculated on the same basis as the criteria for the other ‘A’
nutrients.
The Expert Group noted that scoring systems are more accurate than threshold
models, but more complex. The Expert Group also noted that scoring systems are
more flexible than threshold models. For example scoring systems can be used, not
just for the purposes of defining ‘foods high in fat, salt or sugar’ etc., but to compare
the ‘healthiness’ of foods within categories, whereas threshold models cannot be used
Nutrient profiles for food promotion to children: Section 3: Development and testing I
75
in this way. For these reasons it was agreed that scoring systems as well as threshold
models deserved further testing.
The Expert Group also decided that any new models should be developed using an
incremental approach. It agreed that this should be done by using the TAg1 model as
a base model. TAg1 was felt to be the simplest model (as well as being in line with
other government advice i.e. the FSA guidance on what counts as ‘a lot’ or ‘a little’).
The Expert Group felt that further models could then be developed from this base
model, but the development of each further model should only involve a single
change from the previous model. In this way, the direct affects of that change might
be assessed.
The Expert Group also noted that the methods for testing definitions could be
improved. They suggested that in further testing the range of indicator foods for the
five food groups of the Balance of Good Health should be extended.
3.6 References
Altman D G (1991). Practical Statistics for Medical Research. London: Chapman and
Hall.
Angus F (2003) Assessment of portions for dried and processed fruits and vegetables
to assist in establishing criteria for the 5-a-day message. Confidential Project Report
for the Department of Health, London
Buss DH, Lewis J, Smithers G (1994). Non-milk extrinsic sugars. J Hum Nutr
Dietetics 7: 87.
Coronary Prevention Group (1988) Nutritional Labelling of Foods: A Rational
Approach to Banding. London: CPG.
European Commission (2003) Proposal for a regulation of the European Parliament
and of the Council on nutrition and health claims made on foods. Brussels, 16.7.2003
COM(2003) 424 final.
Food Standards Agency (2002) Labelling claims. London: FSA (and ‘Salt’, ‘Sugar’
and ‘Fat’ in this series.)
Gatenby S, Hunt P, Rayner M (1994). Balance of Good Health, food classification
list: extended version. Health Education Authority: London.
Ministry of Agriculture, Fisheries and Food (1993) Food portion sizes, second edition.
London: HMSO.
Ministry of Agriculture, Fisheries and Food (1998). Fatty acids. Seventh supplement
to the fifth edition of McCance and Widdowson’s The Composition of Foods.
London: Royal Society of Chemistry.
Nutrient profiles for food promotion to children: Section 3: Development and testing I
76
O’Neill M. (2004) Traffic lights for food? How nutrient profiling can make healthy
choices the easy choices. London: National Consumer Council.
Rayner M, Scarborough P, Williams C (2004) The origin of Guideline Daily Amounts
and the Food Standards Agency’s guidance on what counts as a lot and a little, Public
Health Nutrition 7 (4): 549-556.
Roe M, Finglas P, Church S (2003) McCance and Widdowson’s The Composition of
Foods. Sixth Summary Edition. London: Royal Society of Chemistry.
Stockley L, Rayner M, Scarborough P (2003). Five a day: Development of
compositional criteria for composite foods. Report prepared for the Department of
Health, London.
Nutrient profiles for food promotion to children: Section 4: Development and testing II
77
SECTION 4. DEVELOPMENT AND TESTING OF
POSSIBLE DEFINTIONS II.
4.1 Summary
This section describes the development and testing of 12 further models with
definitions of ‘foods high in fat, salt or sugar’ and ‘healthier food choices’. The 12
models were developed on the basis of the recommendations of the Expert Group
following the previous round of testing. It also describes the development and testing
of two possible modifications to these 12 models.
This new round of development and testing followed a more incremental approach than
the previous round. The development and testing of models started with a threshold
model – Model TAg1 - involving criteria for energy, saturated fat, NME sugar and
sodium (and drawing on the FSA’s advice as to what counts as ‘a lot’ and ‘a little’ of
those nutrients) and made sequential changes to that model.
The models tested in this round were judged by four ‘success’ criteria: simplicity and
transparency; accuracy; distribution of foods distribution of foods between ‘high in fat,
salt or sugar’ ‘intermediate’ and ‘healthier choice’ categories; and concurrence with
government advice.
As a result of this round of testing one model was identified for further consideration i.e.:
Model SSCg3d (Scoring system, Group C nutrients, a per 100g base, modification
for drinks)
The scoring bands for foods are as follows (to a maximum of 10 points per
nutrient/food component):
Energy: = 335kJ = 0; 335-670kJ = 1; 670-1005kJ = 2, etc.
Saturated fat: = 1.0g = 0; 1.0-2.0g = 1; 1.0-2.0g = 2, etc.
NME sugars: = 2.4g = 0; 2.4-4.8g = 1; 4.8-7.2g = 2, etc.
Sodium: = 90mg = 0; 90-180mg = 1; 180-270mg = 2, etc.
Calcium: = 105mg = 0; 105-210mg = 1; 210-315mg = 2, etc.
Iron: = 1.5mg = 0; 1.5-3.0mg = 1; 3.0-4.5mg = 2, etc.
n-3 fatty acids: = 0.05mg = 0; 0.05-0.10mg = 1; 0.10-0.15mg = 2, etc.
Fruit and vegetables: 0-30% = 0; 50% = 2; 70% = 4; 100% = 10.
The scoring bands for drinks are half the width of these bands.
Total score = A nutrients – B nutrients – C nutrients.
The food is defined as a ‘healthier food choice’ if the score is 2 or less, , as
‘intermediate’ if the score is 3-8, and as a ‘food high in fat, salt or sugar’ if the
score is 9 or more.
Although scoring systems, such as Model SSCg3d, are seemingly more complex than
threshold models they are also more accurate. Scoring systems are also more flexible
Nutrient profiles for food promotion to children: Section 4: Development and testing II
78
than threshold models so making them more adaptable to a variety of purposes.
Scoring systems can more easily be used for comparing foods within categories.
4.2 Introduction
On the basis of the Expert Group’s recommendations 12 new models were developed
for further testing. All of the models were based on Model TAg1 (in this Section
renamed as TAg6). Model TAg6 is a simple threshold model involving ‘A’ nutrients
(energy, saturated fat, NME sugar and sodium) where the criteria are set on a per 100g
basis. The thresholds for Model TAg6 were based on the FSA’s advice for what counts
as ‘a lot’ or ‘a little’ of saturated fat, NME sugars and sodium. ‘A lot’ is 20% of the
GDA and ‘a little’ is 3.3%.
TAg models
Six different TAg models were developed involving three changes to the basic TAg6
model.
Firstly the thresholds for saturated fat, NME sugars and sodium were changed. This
meant that the thresholds for these nutrients were set in one of two ways:
a. On the basis of the FSA guidance on what counts as ‘a lot’ or ‘a little’
i.e. 20% (one fifth) and one 3.3% (one thirtieth) of the GDAs
respectively (Models TAg6, TAg7 and TAg8).
b. At 15% (three twentieths) and 7.5% (three fortieths) of the GDAs
(Models TAg3, TAg4 and TAg5).
It had already been noted that Model TAg6 defined too many foods as ‘intermediate’
foods (see above) so Model TAg3 was developed to ensure a more even distribution of
foods between ‘foods high in fat, salt or sugar’, ‘intermediate foods’ and ‘healthier food
choices’. Hence, the high thresholds were set lower (15% of GDA compared to 20%)
and the low thresholds were set higher (7.5% of GDA compared to 3.3%). 15% of the
GDA had already been tested in model TCg (see Section 3), and the new low thresholds
were set at half of the high thresholds.
Secondly the thresholds for energy were changed. This meant that two ways of setting
the energy threshold were tested i.e.:
a. Using the original method of calculating the energy criterion i.e. on the
same basis as that of the other A nutrients (Models TAg6 and TAg3).
b. At 525kJ per 100g and at 1030kJ per 100g. These new thresholds were
chosen as the Expert Group agreed that the energy thresholds previously
tested may have been too high. The new low threshold corresponds to
the average energy density of the diet (excluding drinks) of women (16-
64 years) who consumed no more than 35% energy from fat and at least
400g of fruit and vegetables a day. This figure was derived from the
National Diet and Nutrition Survey and has been used as an example of
a healthy Western diet (Prentice and Jebb, 2003). The high threshold
was set twice as high as the low threshold (Models TAg7 and TAg4).
Nutrient profiles for food promotion to children: Section 4: Development and testing II
79
Thirdly, the energy criterion was substituted with a fat criterion calculated on the same
basis as the other ‘A’ nutrients (Models TAg8 and TAg5). The GDA for fat for
children aged 11-16, calculated in the same way as for the other nutrients, is 83g.
TCg models
Three TAg models (i.e. Models TAg 3, TAg 4 and TAg 5) were then developed into
TCg models (i.e. Models TCg 3, TCg 4 and TCg 5) by the addition of criteria for ‘B’
nutrients i.e. n-3 fatty acids and fruit and vegetables and ‘C’ nutrients i.e. calcium and
iron. The thresholds for these nutrients were kept the same for all three models.
The thresholds for calcium and iron were set at 30% of the GDA. This amount had been
tested in the SSCg and SSCg/s models, and was similar to the amount in the original
TCg model. The threshold for n-3 fatty acids was set at 25% of the GDA, and the
threshold for fruit and vegetable content was set at 80-100%. These are the same levels
as tested in the original TCg model.
SSCg models
The three TCg models were then developed into SSCg models (i.e. Models SSCg3,
SSCg4 and SSCg5) by devising scoring bands and score thresholds based on the TCg
thresholds.
For each of these models, the scoring bands started at 50% of the low threshold from
the TCg models and were 50% of the low threshold wide. The maximum score for
each nutrient was set at 10. The fruit and vegetable content score was set so that a food
consisting of between 80% and 100% fruit or vegetables would score 10 points.
The maximum scores for each nutrient was set at 10 (rather than 5 or 6 as had been
tested previously) in order to develop greater sensitivity when dealing with foods that
contain a large amount of a particular nutrient. Maximum scores were necessary,
however, in order to prevent one nutrient dominating the score of some foods (for
example, if there were no maximum score for n-3 fatty acids, then canned pilchards
would score 51 points for this nutrient alone).
The score thresholds were similar to the scores that would have been attained by a food
that contained the threshold level of each nutrient in the equivalent TCg model.
In this way the scoring bands, maximum scores for individual nutrients and other
components the score thresholds for each SSCg model were set so that the three SSCg
models bore the closest possible resemblance to the corresponding TCg models.
Two modifications of two of the models
In the light of testing the 12 models derived in an incremental way from Model TAg 6
(see below) one threshold model (Model TCg3) and one scoring system model (Model
SSCg3) were then modified in two ways. Firstly different thresholds (or scoring
Nutrient profiles for food promotion to children: Section 4: Development and testing II
80
bands) were set for drinks (Models TCg3d and SSCg3d) and secondly criteria for fibre
(Non Starch Polysaccharide) were introduced (Models TCg3f and SSCg3f).
A drink was defined to be any liquid food, excluding oils, soups, condiments (vinegar,
salad cream etc.) and dressings. The alternative thresholds and scoring bands were set
at exactly half the level for other foods, making the thresholds twice as stringent. Note
that criteria for nutrition claims such as ‘low fat’ and ‘low salt’ for drinks are often set
at half those for solid foods when they are set on a per 100g basis (e.g. European
Commission, 2003) reflecting the fact that foods that have a high water content, such as
drinks, are often eaten in large servings (see Section 2).
A GDA for fibre for children aged 11-16, calculated in the same way as for the other
nutrients, is 17g. Thresholds and scoring bands for fibre were introduced at the same
level as for calcium and iron on the basis that the reason for introducing a fibre criterion
was principally to ensure that healthier options from the ‘Bread, other cereals and
potatoes’ group of the Balance of Good Health were classified as ‘healthier food
choices’ in the same way as the criteria for calcium and iron function for foods from the
‘Milk and dairy foods’ and the ‘Meat, fish and alternatives’ groups.
In total, the Expert Group’s recommendations led to the generation of 16 new models:
12 new models based on Model TAg6 (2 x 3 TAg models, 3 TCg models and 3 SSCg
models), 2 modifications of Model TGg3 and 2 of Model SSg3.
4.3 Methods
The methods for this round of testing were the same as for the previous round (See
Section 3). However in view of the Expert Group’s recommendation with respect to
the Balance of Good Health group indicators and as a result of suggestions from
members of the Expert Group, further example indicator foods were included in the
tabulated results.
These example indicators were included to provide a better indication of how each
model categorised a range of foods in each of the food groups on which healthy eating
advice (i.e., the Balance of Good Health) is based. This allowed the Expert Group to
check that each model categorised some ‘healthier’ foods in each of the important food
groups. The foods chosen are typically minimally processed, to make it easier to
estimate the likely nutritional contribution to the diet of each of them. It should also be
noted that the nutrient content of some foods will be subject to significant variation
depending on the recipe, or the way in which they have been prepared (e.g.roast
potatoes, currant buns etc.) Appendix 3b gives the nutrient content listed in McCance
and Widdowson for each of the example indicators.
McCance and Widdowson listing Abbreviated term for results
Bread, other cereals and potatoes (19)
Chips, French Fries, retail French fries’
Corn flakes
Nutrient profiles for food promotion to children: Section 4: Development and testing II
81
Cream crackers
Crispbread, rye ‘Crispbread’
Crunchy nut corn flakes
Currant buns
Malt bread, fruited ‘Malt loaf’
New potatoes, boiled in unsalted water ‘Boiled potatoes’
Old potatoes, roast in blended oil ‘Roast potatoes’
Oven chips, frozen, baked ‘Oven chips’
Pasta, plain, fresh, cooked ‘Fresh pasta’
Potato croquettes, fried in blended oil ‘Potato croquettes’
Sugar puffs
Weetabix
White bread, sliced ‘White bread’
White rice, easy cook, boiled ‘Boiled rice’
White rice, fried ‘Fried rice’
Wholemeal bread, average ‘Wholemeal bread’
Wholemeal rolls
Milk and dairy foods (14)
Camembert
Cheddar cheese ‘Cheddar’
Cheddar type, half fat ‘Half fat cheddar’
Cottage cheese, plain ‘Cottage cheese’
Cream, fresh, single ‘Single cream’
Crème fraiche, half fat ‘Half fat crème fraiche’
Fromage frais, plain Fromage frais’
Fromage frais, virtually fat free, natural ‘Low fat fromage frais’
Greek yoghurt, sheep ‘Greek yoghurt’
Semi-skimmed milk, average Semi-skimmed milk’
Skimmed milk, average ‘Skimmed milk’
Whole milk, average ‘Whole milk’
Yoghurt, low fat, fruit ‘Low fat fruit yoghurt’
Yoghurt, low fat, plain ‘Low fat yoghurt’
Fruit and vegetables (8)
Currants
Orange juice, unsweetened ‘Orange juice’
Peaches, raw ‘Peaches’
Pineapple, canned in juice ‘Canned pineapple’
Avocado, average ‘Avocado’
Celery, raw ‘Celery’
Lettuce, average, raw ‘Lettuce’
Tomatoes, grilled ‘Grilled tomatoes’
Meat, fish and alternatives (20)
Bacon rashers, streaky, fried ‘Streaky bacon’
Beef, mince, extra lean, stewed ‘Lean mince’
Nutrient profiles for food promotion to children: Section 4: Development and testing II
82
Beef, rump steak, lean only, grilled ‘Grilled steak’
Beef, rump steak, lean and fat, fried ‘Fried steak’
Beef, topside, roasted, well done, lean ‘Roast beef’
Chicken nuggets, takeaway ‘Chicken nuggets’
Chicken, breast, grilled without skin, meat only ‘Chicken breast’
Chicken, leg quarter, roasted, meat and skin ‘Chicken leg’
Cod, baked ‘Cod’
Eggs, chicken, boiled ‘Boiled egg’
Eggs, chicken, fried in vegetable oil ‘Fried egg’
Fish fingers, cod, grilled ‘Fish fingers’
Ham
Lamb, leg, lean only, roasted ‘Roast lamb’
Lentils, red, split, dried, boiled in unsalted water ‘Lentils’
Mackerel, grilled ‘Mackerel’
Pork sausages, chilled, fried ‘Sausages’
Tofu, soya bean, steamed ‘Tofu’
Tuna, canned in oil, drained ‘Tinned tuna’
Walnuts
Foods high in fat, foods high in sugar (17)
Chocolate chip cookies Choc chip cookies’
Cola
Cola, diet ‘Diet cola’
Digestive biscuits, plain Digestives’
Doughnuts, jam ‘Jam doughnuts’
Jaffa cakes
Kit kat
Mars bar
Milky way
Sponge cake
Butter
Fat spread (20-25% fat), polyunsaturated ‘Very low fat spread’
Fat spread (60% fat), polyunsaturated ‘Low fat spread’
Margarine, soft, polyunsaturated ‘Margarine’
Olive oil
Potato crisps ‘Crisps’
Potato crisps, low fat ‘Low fat crisps’
In addition, a selection of composite foods was also added to the list.
Composite foods (15)
Baked beans, canned in tomato sauce, reheated ‘Baked beans’
Celery, boiled in salted water ‘Boiled celery’
Chilli con carne
Cottage/Shepherd’s pie, chilled/frozen, reheated ‘Shepherd’s pie’
Crumble, fruit ‘Fruit crumble’
Fruit pie, pastry top and bottom ‘Fruit pie’
Lasagne
Nutrient profiles for food promotion to children: Section 4: Development and testing II
83
Mushrooms, common, fried in corn oil ‘Fried mushrooms’
Peanuts, roasted and salted ‘Salted peanuts’
Rhubarb, stewed with sugar ‘Stewed rhubarb’
Sardines, canned in tomato sauce ‘Tinned sardines’
Soya, non-dairy alternative to milk, unsweetened ‘Soya milk’
Strawberries, canned in syrup ‘Canned strawberries’
Mayonnaise
Mayonnaise, reduced calorie ‘Low fat mayo’
4.4 Results
Appendix 7 gives the results of testing each of the 12 new models and 4 modifications
Table 6 summarises the results of the Tables in Appendix 7.
Table 6. Results of testing 12 new models and 4 modifications.
Modle
%
‘healthier
food
choices’ %
intermediate
%
‘high in fat,
salt or
sugar’
Sensitivity (%
correct
classifications)
Specificity (100 -
% misclass-
ifications)
Overall
accuracy
score
TAg3 22 18 60 85 93 79
TAg4 20 19 61 83 93 76
TAg5 23 16 61 86 92 79
TAg6 11 42 47 66 95 62
TAg7 15 34 51 76 95 71
TAg8 16 35 49 77 95 72
TCg3 30 26 44 83 96 79
TCg4 26 30 44 81 96 77
TCg5 29 27 45 83 96 80
TCg3d 28 27 45 83 96 79
TCg3f 32 25 43 82 96 78
SSCg3 39 24 37 88 98 86
SSCg4 36 24 40 89 97 87
SSCg5 39 23 38 88 97 85
SSCg3d 36 26 38 89 98 87
SSCg3f 40 25 35 90 98 88
4.4.1 Choice of thresholds
The low threshold of 3.3% of the GDA and the high threshold of 20% of the GDA
(used in Models TAg6, TAg7 and TAg8) were not tested past the TAg stage. This is
because the other thresholds (7.5% and 15% of the GDA) used in Models TAg3, TAg4
and TAg5 produced more accurate models. For example, the 7.5%/15% threshold
models all attained higher accuracy scores than their associated 3.3%/20% models.
Nutrient profiles for food promotion to children: Section 4: Development and testing II
84
This improved accuracy was because Models TAg3, TAg4 and TAg5 were more
sensitive than Models TAg6, TAg7 and TAg8, i.e. they were more likely to classify
‘healthier’ and ‘less healthy’ indicators correctly. This is partly because Models TAg3,
TAg4 and TAg5 classified more foods overall as either ‘foods high in fat, salt or sugar’
and ‘healthier food choices’ rather than ‘intermediate foods’ compared with Models
TAg6, TAg7 and TAg8.
As the incremental development of models used in this round of developing and testing
models involved introducing criteria for ‘B’ and ‘C’ nutrients whilst leaving the ‘A’
nutrient thresholds at the same level, foods could only be re-categorised to a lower
(healthier) level than before by moving from TAg to TCg models. Therefore, it
seemed more appropriate to continue with Models TAg3, TAg4 and TAg5, where a
higher proportion of all foods in the database are categorised as ‘foods high in fat, salt
or sugar’.
4.4.2 Choice of energy/fat criteria
The two different energy criteria and the fat criterion made little difference to either the
accuracy scores or the distribution of foods between categories for the models in which
they were used. This was surprising as the difference between the two different sets of
energy thresholds was noticeable (5.9%/11.7% of the GDA compared with 7.5%/15%
of the GDA), and the introduction of total fat would be expected to change the results
of the models.
Inspection of the way that individual foods are categorised by the three models for each
energy/fat criterion does not help choose between models either. For example
Appendix 7 shows that Model TAg5 (with the fat criterion) is the only TAg model that
categorises fresh pasta as ‘healthier’. However, it is also the only TAg model that
categorises single cream as ‘healthier’. Similarly Model SSCg5 (with the fat criterion)
is the only SSCg model to categorise all fruit and vegetables as ‘healthier’, and potato
croquettes as ‘less healthy’, but again is the only SSCg model to categorise single
cream as ‘healthier’.
4.4.3 Choice of nutrient group and of model type
The accuracy scores for the models are higher for TCg models than for TAg models
(though only very slightly) and higher again for SSCg models. This again should have
been expected, as the added complexity of the TCg and, particularly, SSCg models is
designed to improve their accuracy.
Inspection of the way that individual foods are categorised by models indicates that
both TCg and SSCg models classify foods more appropriately than TAg models. For
example Appendix 7 shows that all but one of the TCg models and all of the SSCg
models classify mackerel as a ‘healthier food choice’ but four of the TAg models
classify mackerel as an ‘intermediate food’ and two as a ‘food high in fat, salt or sugar’.
Nutrient profiles for food promotion to children: Section 4: Development and testing II
85
4.4.4 Drink and fibre modifications
The two modifications of Models TCg3 and SSCg3 were introduced because inspection
of the way that TCg models and SSCg models classify foods suggests that they might
have a tendency to misclassify drinks and high fibre foods. For example Appendix 7
shows that all three SSCg models classify cola as an ‘intermediate food’ and all three
TCG models classify wholemeal bread as ‘a food high in fat, salt or sugar’.
Introducing different criteria for drinks made little difference to the accuracy scores or
the proportions of foods that were defined as ‘foods high in fat, salt or sugar’ and
‘healthier food choices’ by, Models TCg3 and SSCg3. This was to be expected as only
94 (9%) of the foods in the database are classified as drinks.
However, the modification obviously does make a difference to the way individual
drinks are categorised. For example Appendix 7 shows that Model TCg3d classifies
whole milk and orange juice as ‘intermediate foods’ rather than ‘healthier food
choices’. Similarly Model SSCg3d model classifies whole milk and drinking chocolate
as ‘intermediate foods’ rather than ‘healthier food choices’ and cola as a ‘food high in
fat, salt or sugar’ rather than an ’intermediate food’.
The addition of fibre criteria to Models TCg3 and SSCg 3 also seemed to make little
difference to their accuracy scores and the proportions of foods that were defined as
‘foods high in fat, salt or sugar’ and ‘healthier food choices’. This is at least partly
because only 46 (4%) of the foods in the database contained at least 30% of the GDA
for fibre, when measured per 100g. This in turn may suggest that the threshold and
scoring bands for fibre were set too high. But then again only 65 (6%) and 92 (8%) of
the foods in the database contain equivalent levels of calcium and iron, respectively.
Appendix 7 shows that the fibre modification to TCg3 led to some breakfast cereals
(e.g. Weetabix) along with some nuts (e.g. hazelnuts and almonds) being reclassified as
‘healthier food choices’ rather than ‘intermediate foods’. However it also leads to
crisps, low fat crisps, salted peanuts and peanut butter being reclassified as
‘intermediate foods’ rather than ‘foods high in fat, salt or sugar’. The fibre modification
to SSCg3f had less affect, only leading to a reclassification of dates as ‘healthier food
choices’ rather than ‘intermediate foods’ and the reclassification of some cereals such
as sugar puffs as ‘intermediate foods’ rather than ‘foods high in fat, salt or sugar’.
Neither of the fibre modifications made any difference to the classification of brown or
wholemeal bread.
4.5 Conclusions and recommendations
The twelve further models developed compare favourably with the eight models tested
in the previous round. For example all three new TCg models have higher accuracy
scores than the TCg model tested previously, whilst leading to a similar distribution of
foods between ‘foods high in fat, salt or sugar’, ‘intermediate foods’ and ‘healthier food
choices’. Although two models tested previously have accuracy scores in the 90s
(higher than for all the 12 new models in this round) they both utilise a per 100kJ and a
per serving base and the Expert Group considered that the base for definitions should
be per 100g. Both these models also classify few foods in the ‘intermediate’ category.
Nutrient profiles for food promotion to children: Section 4: Development and testing II
86
Table 7 shows how the 12 models tested in this round compare to each other when
judged by four ‘success’ criteria: simplicity and transparency; accuracy; distribution;
and concurrence with other government advice. The table suggests that the best models
are either SSCg or TCg models.
As mentioned above, the additional testing did not provide evidence to help chose
which of the three energy/fat criteria is the most appropriate. However, in theory an
energy density criterion is more justifiable than a total fat criterion for the reasons
outlined in Section 3.5 and from the testing there seems no reason why this energy
density criterion should not be linked to the GDA for energy in the same way as the
saturated fat, NME sugar and sodium criteria. On this basis Models TCg3 and SSCg3
would seem most worthy of further consideration.
Introducing different criteria for drinks appears to make a slight improvement to the
way foods are classified by Models TCg3 and SSCg3, without an appreciable increase
in complexity. It is generally recognised that nutrient profiles set on a per 100g basis
need to be different for foods and drinks. But if this modification is to be accepted,
then the problems inherent in food category specific criteria (as laid out in Section
2.5.2) must also be accepted. It may also be hard to justify introducing different criteria
for drinks but not other food categories (e.g. complete meals).
Introducing additional fibre criteria does not appear to improve the accuracy of Models
TCg3 and SSCg3. It fails to recategorise brown and wholemeal bread and yet it does
recategorise crisps and salted peanuts. A lowering of the threshold and scoring bands
(so as to recategorise brown and wholemeal bread), would also result in other foods
from the ‘foods high in fat, foods high in sugar’ group also being recategorised (such as
biscuits and trail mix).
Of the two Models SSCg3 and TCg3, Model SSCg3 is more accurate than Model
TCg3, but is also more complex. However Model SSCg3 has the added advantage
inherent in scoring systems of flexibility. That is different score thresholds can be used
for different purposes (e.g. advice on the use of nutrition and health claims or for advice
on the balance of TV advertising for foods during children’s TV programmes) and
foods within categories can be directly compared.
The Expert Group therefore recommends one model for further consideration i.e.
Model SSCg3d (Scoring system, Group C nutrients, per 100g base,
modification for drinks)
Nutrient profiles for food promotion to children: Section 4: Development and testing II
87
Table 7. Results of testing developed models against four success criteria.
Accuracy Concurrence with other
government advice
Model Simplicity &
transparency Sensitivity Specificity
Distribution
5 A
DAY ‘a lot’
and ‘a
little’
Oily
fish
TAg3