is it possible to develop a global nutrient profile model? · diet cola potato crisps french fries...
TRANSCRIPT
Is it possible to develop a global nutrient profile model?
Mike Rayner
British Heart Foundation Health Promotion Research Group
IUNS 20th International Conference of Nutrition, Granada, Spain, 15 -20 September 2013. Parallel Symposium on Nutrition Profiling
Presentation aims
• Explain why we need nutrient profiling at all
• Outline the general principles of nutrient profiling
• Consider what we mean by a global nutrient profile model – World-wide? Regional?
– For every conceivable/more than one application?
• Suggest a practical way forward: – Developing a common nutrient profile model for
marketing restrictions in Europe
A definition
• Nutrient profiling is:
’The science of classifying or ranking foods according to their nutritional composition for reasons related to preventing disease and promoting health’
Source: WHO, Guiding Principles and
Framework Manual for the development or adaptation of nutrient profile models. in press.
WHO Manual
• Definition and guiding principles
• Module 1: Planning
• Module 2: Development and adaptation
• Module 3: Validation
• Module 4: Implementation, monitoring and evaluation
FSA/Ofcom model
Points 0 1 2 … 10
Energy (kJ) ≤335 ≤670 ≤1005 … >3350
Sat fat (g) ≤1.0 ≤2.0 ≤3.0 … >10.0
Total sugar (g) ≤4.5 ≤9 ≤13.5 … >45.0
Sodium (mg) ≤90 ≤180 ≤270 … >900
5
Protein (g)* ≤1.6 >1.6 >3.2 … >8.0
Fibre (NSP) (g) ≤0.7 >0.7 >1.4 … >3.5
Fruit, Veg & Nuts
(g)
≤40 >40 >60 … >80
Healthy/Intermediate food = 3 or less
Healthy/Intermediate drink = 0 or less
Unhealthy food = 4 or more
Unhealthy drink = 1 or more
For both food and drinks: scores are
based on the content of nutrient
in 100g.
*If food scores 11 for protein, fibre and F&V then
scores 0 for protein except if scores 5
for FV&N
Can be advertised Cannot be advertised Wholemeal and white bread
Muesli and wheat biscuit cereal
with no added sugar
Fresh fruit
Most nuts
Takeaway salads with no dressing
or croutons
Most brands of baked beans
Some brands of baked oven chips
Some brands of chicken nuggets
Fish fingers
Chicken breast
Unsweetened fruit juice
Skimmed, semi-skimmed and
whole milk
Diet cola
Potato crisps
French fries
Most breakfast cereals
Raisins and sultanas
Cheddar cheese, half and full fat
Butter and margarine
Most sausages and burgers
Cookies
Confectionary
Peanut butter
Mayonnaise, reduced and full
calorie
Most pizzas
Sweetened milkshakes
Cola and other carbonated
sweetened drinks
The ‘4 Ps’
of marketing
theory
Voluntary public
health interventions
Public health
regulation
1. Product Product reformulation Compositional
standards
2. Promotion
(advertising)
Front-of-pack labelling
Advertising regulation
Health and nutrition
claims legislation
Advertising regulation
3. Place
(availability)
Place-based
promotions
Standards for public
provision (schools,
hospitals)
4. Price Price-based based
promotions
Agricultural subsidies
Health-related taxes
Specific applications of nutrient profiling
Module 3 of the manual: validating a nutrient profile model
• ‘Subjective’ judgements using a panel of nutritionists and a set of indicator foods
• Quantitative comparisons with views of expert nutritionists
• Comparison with currently achieved diets
• Comparison with theoretical diets
• Prediction of health outcomes
US ONQI model, Nurses’ Health Study
*Adjusted for age, energy intake,
smoking, aspirin, exercise, alcohol intake,
family history of MI, family history of
cancer, history of hypertension and
history of high cholesterol and BMI
Source: Chiuve SE, Sampson L, Willett WC.
The association between a nutritional quality
index and risk of chronic disease. Am J Prev
Med. 2011 May;40(5):505-13.
*
Alternatives to nutrient profiling
• Lists of healthy and unhealthy foods
• Classification systems based on level of processing
Whoa! Croissants, muffins, doughnuts, sweet rolls, crackers made with transfats, sweetened breakfast cereals
Missing from this list Crumpets, currant buns, crackers made without transfats, muesli.
Classifies foods into Ultra-processed Processed Ingredients Foods
Classifies foods into: Ultra-processed: Processed: Ingredients: Foods:
N= 382 Food (1) Ingredient (1) Processed (1) Ultra-processed (1)
Food
(2) 19%
Ingredient
(2) 10%
Processed
(2) 16%
Ultra-
processed
(2)
55%
Two researchers were asked to use the system to classify 382 Tesco foods
N= 382 Food (1) Ingredient (1) Processed (1) Ultra-processed (1)
Food
(2) 49 1 5 0
Ingredient
(2) 5 40 5 2
Processed
(2) 13 0 48 5
Ultra-
processed
(2)
5 0 15 192
Agreement 86%, expected agreement 36%, Kappa 0.78
Two researchers were asked to use the system to classify 382 Tesco foods
% of foods, ingredients, processed and ultra -processed that are unhealthy (according to FSA/Ofcom)
0102030405060708090
100
Unhealthy
Not unhealthy
WHO Manual • Definition and guiding principles
• Module 1: Planning
• Module 2: Development and adaptation
• Module 3: Validation
• Module 4: Implementation, monitoring and evaluation
Field testing workshops: South Africa, 14-15 April 2011 Canada, 7-9 June 2011 Thailand, 13-15 June 2011 Norway, 6-7 December 2011 Slovenia, 12-13 April 2012 United Arab Emirates, 8-10 May 2012
Lessons learnt from field testing the manual
• It’s easier to adapt an existing model than to develop a new model from scratch
Module 2b of the manual: adapting an existing nutrient profile model
• Choose an existing model
• Run that model against a sample of say 200 foods from your country
• Compare the model’s classifications against those of a few other models
• Modify model e.g. by creating exemptions, adding categories, changing thresholds/points awarded, etc.
FSA/Ofcom → FSANZ model
• Additional category – Beverages
– ‘Edible oils, edible oil spreads, butter and margarine, and cheese and processed cheese with a calcium level of more than 320 mg/100 g’
– All other foods
• Other minor changes e.g. – 0 points for < 5.0 g/100g of sugar rather than 4.5
– Fibre points for all foods but not drinks
Uses of the FSA/Ofcom model around the world
• Regulating tv advertising of foods to children (in UK, Ireland and New Zealand)
• Voluntary restrictions on the marketing of foods to children in general (McDonalds)
• Regulating health claims (in Australia/New Zealand and South Africa)
• Food labelling (in Australia/New Zealand)
• Monitoring food marketing spend (in US by Rudd Centre, Yale University), advertising exposure (in UK by Newcastle University)
• Setting targets for /scoring a food company’s product portfolio (in UK by Pepsico and by Access To Nutrition Index )
• Consumer education (in Australia (Food Switch ‘app’ for mobile phones))
Lessons learnt from field testing the manual
• It’s easier to adapt an existing model than to develop a new model from scratch
• A model developed for say the UK can be adapted for quite a different county say South Africa
• Some countries do not have the resources to develop their own models and need help
• Some countries want a model for a specific application and some want a model for lots of different applications.
• It is possible to develop regional (rather than country-specific) models
Outlining a process for developing a regional nutrient profile model for one
application
• Draft of 4th March 2013
• 119 models indentified,
54 met the inclusion
criteria
• 12: food labelling;
11: school food provision;
9: marketing restrictions
• Only 19 of the included
models have been
validated in any way
WHO Catalogue of Nutrient Profile Models
Models for regulating the marketing of food to children
Country Responsible agency Expressly for marketing restrictions
EJCN Paper
EHN report
Brazil National Health Surveillance Agency (ANVISA)
x x
Denmark Danish Food and Drink Federation
x x x
Korea Korea Food and Drug Institution x
UK Food Standards Agency (FSA) x x x
United States Center for Science in the Public Interest
x x
Multi-national Pepsico x x
Multi-national Disney x x
Europe EU Pledge Nutrition Criteria x x x
N&S America PAHO x
Sweden Swedish National Food Administration (Swedish keyhole)
x
Healthy Unhealthy
FSA/Ofcom model Nordic keyhole
Comparing nutrient profile model: strictness and agreement
50%
allowed
25%
allowed
Percentage
agreement :
3/4 = 75%
Kappa = 0.5
European Heart Network study to compare different nutrient profile models: methods
• Sample of 336 foods and drinks advertised to children in 2008 (weighted by number of advertisements)
• 14 commercial television channels most popular with children aged 4-16 years in the UK
• Television recordings of one week day and one weekend day every month (from 6am until 10pm) during the study period (January – December 2008).
• Original dataset 458 different food advertisements broadcast on a total of 18,888 occasions
• Individual foods, meals and brands treated specially • Nutritional data supplemented with data for equivalent
foods from the UK nutrient databank
European Heart Network study: models
• Models already used to regulate marketing of foods to children: FSA/Ofcom, Danish, EUPNC
• Models proposed for use in regulating marketing of foods to children: Keyhole, Choices International, Norwegian
• http://www.ehnheart.org/publications/publications/publication/705-ehn-research-on-nutrient-profile-model.html
Model
Kca
l/ k
J
Ad
de
d S
uga
r
Tota
l su
gar
Fat
Satu
rate
d F
at
Tran
s fa
t
Ch
ole
ste
rol
Sod
ium
/ Sa
lt
Pro
tein
Fib
re
Vit
amin
s/M
ine
rals
Tota
l Nu
trie
nts
Cat
ego
rie
s
Bas
e
EUPNC y y y y y y y y 8 8* Combination
UK y y y y y y 6 2 /100g
Keyhole y y y y y y y 7 25* Combination
Norwegian y y y y y 5 8 /100g
Danish y y 2 10 /100g
Choices International y y y y y y 6 9* Combination
Totals
5
2
5
4
5
1
0
5
2
3
1
European Heart Network study: characteristics of six models
* With subcategories
Model
Number of foods
Percentage
of foods approved
Number of foods,
weighted by ad frequency
Percentage of foods, weighted
by ad frequency
EU PNC 86 26 3758 32
UK 134 40 5571 47
Keyhole 16 5 642 5
Norwegian 151 45 4936 42
Danish 124 37 3828 33
Choices International 21 6 724 6
European Heart Network study: overall strictness
EHN Study: overall agreement between the six models
EUP
NC
UK
Keyh
ole
No
rweg
ian
Dan
ish
Ch
oic
es
Inte
rnat
ion
al
EUPNC 0.3949
*
0.0451 0.2657
*
0.2144
*
0.0287
UK 0.1206 0.3074
*
0.4933
**
0.0967
Keyhole 0.1479 0.2138
*
0.0070
Norwegian 0.5342
**
0.0385
Danish
0.0923
Choices International
P
e
r
c
e
n
t
a
g
e
Practical differences between models
• Subjective aspects to models:
– Categorising foods for models with many categories
• Information about the foods:
– Standard serving sizes
– Compositional information
Conclusions: Is a global model possible
Global Regional National
All applications
x x
(?)
One application
x
Spare slides
Links to original and modified versions of the
FSA/Ofcom model • FSA Ofcom model
• http://www.publichealth.ox.ac.uk/bhfhprg/publicationsandreports/acad-publications/bhfhprgpublished/nutrientprofilemodel
• Nutrient Profiling Calculator (FSANZ model) • http://www.foodstandards.gov.au/industry/claims/pages/nutrientp
rofilingcalculator/Default.aspx
• Proposed South African model for regulating health claims • http://www.doh.gov.za/docs/npc_nwu/NPC_NWU.html