Professional Documents
Culture Documents
Cecchini2015 PDF
Cecchini2015 PDF
Cecchini2015 PDF
12364
Public Health/Behaviour
201
© 2015 World Obesity 17, 201–210, March 2016
202 Impact of food labelling M. Cecchini & L. Warin obesity reviews
and associated chronic diseases (10). For example, EU, the program’, ‘Traffic light label’, ‘Public health nutrition’,
USA, the UK and Chile are all discussing or implementing ‘Health logo’ ‘pick the tick’, ‘Guideline Daily Amounts’, ‘the
new legislation on food labelling (7). In particular, food la- Heart Symbol’, ‘the Choices logo’, ‘The Choices program in
bels are regarded as a possible tool to empower consumers the Netherlands’, ‘The Keyhole program in Sweden’, ‘Pro-
and to facilitate healthier food choices (11). Previous studies gram less salt is healthier’, ‘Green check marks’, ‘EU regula-
in the field suggest that food labelling would be associated tion food labelling’, ‘EU nutrition labelling program’ and
with healthier eating habits (12–20) and would be a cost- ‘EU nutritional labelling program’. The search strategy was
effective intervention (21). However, the majority of studies developed on a previous systematic review of interventions
are based on a relatively small sample, and in some cases, re- to promote healthy diet and physical activity (26). We identi-
sults appear mixed. Previous systematic reviews have fied all the papers on food labelling included in the review,
mainly focused on consumers understanding of food label- and then, we extracted all the keywords from the identified pa-
ling schemes (22,23) and did not attempt to quantify the ef- pers. The retrieved keywords were searched individually or in
fects of food labelling in changing food choices. combination using the standard Boolean operators.
This systematic review and meta-analysis enrich the liter- A full version of the articles meeting the inclusion criteria
ature by providing a quantitative assessment of the impact was obtained and analysed. Papers were divided into three
of food labelling. In particular, this analysis aims at categories: GDA, traffic light and other food labelling, ac-
assessing the effectiveness of food labelling schemes in in- cording to the type of food labelling investigated. If a study
creasing the selection of healthier products and in reducing evaluated multiple types of labelling, results for each type
calorie intake/choice. The secondary objective is to deter- were considered as an independent study. A researcher ex-
mine whether food labels’ format influences choices and tracted all the information on the characteristics of the study
consumption. This analysis considers three types of food la- including the geographical area where the study was con-
belling schemes: traffic light, Guideline Daily Amount ducted, the food studied and the target population as well
(GDA) and other types of food labelling (e.g. front-of-pack as the quantitative data needed to carry out the meta-
logos). Traffic light and GDA are among the most widely analysis (i.e. sample sizes, means and standard deviations
adopted nutrition labelling schemes (7) and are central to of control and intervention groups). In some cases, missing
the ongoing policy (24) and academic debate (25). dimensions, usually standard deviations, had to be calcu-
lated by using standard approaches (27).
This analysis focuses on two outcomes. The first outcome
Methodology
is the number of people that, following the implementation
This review was restricted to peer-reviewed studies that of food labelling, switches to a healthier product. The out-
were designed to evaluate the impact of food labelling in come was analysed as a percentage by dividing the number
terms of likelihood of selection of either a healthier option of people choosing the healthier option by the total number
or calorie choice/intake. Studies had to be randomized and of people in the group. The second outcome is change in cal-
should include a control population with either a orie intake/choice following the introduction of a food la-
‘before/after’ design or a ‘case/control’ design. The inclusion belling scheme. For some studies, food quantities (e.g.
of randomized studies minimizes the influence of confound- grammes of pizza or cereals) or food nutrients (e.g.
ing factors and increases the strength and reliability of the grammes of fat) had to be converted into calories using a
pooled results. Studies that did not focus on consumers or nutrient database (28). Included studies investigated a
that took marketing or psychological perspectives were ex- broad range of products with different calorie content. To
cluded. For example, we excluded studies that investigated adjust for this heterogeneity, the change in calorie intake
the type of labelling format that looks more appealing to in the intervention group was scaled to a percentage by
consumers. In addition, the review protocol excluded stud- reporting it to the calorie intake of the control group. A sec-
ies that evaluated menu labelling as well as studies that ond set of analyses investigated the two same outcomes in
did not provide the requested quantitative data to feed the terms of, respectively, standardized mean difference (SMD)
meta-analysis. Finally, studies had to be in English or French and absolute number of calories. Details of the methodol-
and had to be published between January 2008 and April ogy and results can be found in Appendix 2. Data were
2015 when the final search took place. analysed with STATA 13 (StataCorp, College Station, TX,
The search was conducted using the following databases: USA). Forest plots were generated, and overall estimates
Pubmed, Biomed, Science Direct, Sage Database, Google of the pooled relation and 95% confidence intervals (CI)
Scholar, EBSCO Host Database Academic. Searched key- were calculated with the use of fixed-effect and random-
words included were the following: ‘Food labelling’, ‘traffic- effect models. Heterogeneity across studies was tested with
light labelling’, ‘traffic-light nutrition labelling’, ‘Nutrition la- the I2 statistics (29). In case of low heterogeneity (30), we
belling’, ‘Nutrition claims labelling’, ‘Nutrition claims regula- carried out a graphical assessment of the potential publica-
tion’, ‘EU regulation food labelling’, ‘EU nutrition labelling tion bias through a funnel plot (31).
Figure 1 Flow diagram of the literature search and filtering results of the systematic review.
Author, year Country/ Type of study Type of food Type of food Population Outcome Authors’
geographical and setting labelling system studied of interest conclusions
area studied
Borgmeier and Hamburg, Experiment in (i) simple ‘healthy choice’ tick; 78 foods from different 420 adult subjects from Food choice, Different food label formats
2012 (appetite) controlled setting front-of-package nutrition on front-of-pack labels can increase
label in the USA knowledge, but the Smart Choices
symbol has little impact on behaviour.
Watson et al., 2014 Australia Randomized Seven different front-of-pack Nine pairs of commonly 4357 grocery shoppers Label use and The number of correct responses
online trial labelling schemes comprising purchased food products in Australia understanding in the control group was significantly
variants of the traffic light less than in other schemes (P < 0.05).
labelling scheme and Even without an accompanying
the percentage daily intake education campaign, front-of-pack
scheme, and a star labels can assist shoppers
rating scheme to make healthy choices.
Roberto et al., USA Randomized (i) no label; (ii) traffic light; US brand name products 703 US consumers Food choice Overall, those in the traffic
2012 (AJPM) online trial (iii) traffic light and protein/fiber selected from eight recruited through light condition performed better
traffic light; (iv) facts up front; categories an online database than those in the facts up front
(v) facts up front and info on conditions on measures of nutrition
‘nutrients to encourage’ knowledge and label perceptions.
Gaigi et al., 2015 Marseille, Experiment in Logo ‘Le Choix Vita+’ Dairy products, cooked 2083 French consumers Purchase There was no main impact of the
France real-world setting meals and fresh snacks logo on food purchase, probably
because of several factors such as
lack of visibility of the logo on
supermarket shelves and clients’
low socio-economic group.
Spanos et al., 2015 Australia Experiment in Serving size labels Pizza 104 female undergraduate Food intake The nature of the information provided
controlled setting students at an Australian influences how much participants eat.
university Participants had lower consumption
obesity reviews
4 servings’ compared to
was labelled ‘contains
most effective labelling scheme. GDA becomes the second
‘contains 2 servings’.
influential or useful.
most effective intervention.
Findings of the analyses on the effectiveness of food
conclusions
Food choice,
consumption
of absolute number of calories, are presented in Annex 2.
Purchasing
of interest
intentions
Outcome
desired
effects. Homogeneity across studies is also confirmed by
low levels of the I2 statistics. Seven out of the 12 included
interventions show that calorie intake/choice decreases once
1017/1002 consumers
of the UK
studied
Chocolate and
Type of food
cereal bars
(Fig. 4). The studies were plotted with the estimated effect
food labels
colours on
studied
Randomized
Randomized
online trial
online trial
Canada
Discussion
area
Table 1. (Continued)
UK
Figure 2 Effect of food labels in selecting/purchasing a healthier option. CI, confidence interval; D + L, random-effects estimate–DerSimonian and Laird
method; GDA, Guideline Daily Amount; I V, fixed-effects estimate–inverse variance method; WMD, weighted mean difference.
labelling scheme in steering consumers’ choices. Conversely, consumers would have only a limited understanding of
our findings show a less clear picture in terms of whether the information provided by food labels (37). Findings
food labelling schemes affect calorie choice or consumption. of our study support this hypothesis by showing that
Overall, food labels would have the potential to decrease interpretive nutrition labels, as traffic light systems, are
calorie choice/intake. However, single studies report large more effective in helping consumers in choosing healthier
CIs, suggesting that different individuals respond to the products.
introduction of food labels with a wide range of behaviours, Another reason may also explain why significant positive
indifference included. effects on food choice are not followed by a decrease in cal-
Previous systematic reviews in the field focused on con- orie choice/intake of a similar magnitude. Calorie content is
sumers understanding of food labelling schemes (22,23) only one of the multiple dimensions affecting the healthiness
and on whether menu labels would affect selection and of products. For example, consuming a product with re-
consumption of calories in restaurants (35,36). This duced salt content is healthier but does not reduce calorie
study enriches the literature by investigating the effective- intake. Again, substituting unhealthy nutrients (e.g. trans
ness of food labels in increasing the selection of healthier fats) with healthier options (e.g. polyunsaturated fats) may
products and in modifying calorie choice/intake. Our not produce any significant effect on calorie intake. The
results suggest that menu labels and food labels may so-called ‘halo effect’ may also play a key role. Consumers
have similar effects on consumers’ eating behaviours. tend to generalize to a whole food product a piece of infor-
Previous literature showed that this may happen because mation found on the label (38). This especially applies in the
Figure 3 Effect of food labels on calorie intake/choice (percentage of control group). CI, confidence interval; D + L, random-effects estimate–DerSimonian
and Laird method; ES, effect size; GDA, Guideline Daily Amount; I V, fixed-effects estimate–inverse variance method; WMD, weighted mean difference.
Findings of this study have strong policy implications. confounders and preclude analyses by population subgroups.
Many countries are in the process of debating, developing or Second, there is a need of a higher number of studies carried
implementing new food labelling schemes (7,41). Results of out in the ‘real world’ as opposed to laboratory settings. These
our work support the implementation of food labelling as a two issues could be addressed with a closer collaboration be-
key tool to tackle unhealthy diet and obesity. Food labels em- tween researchers and retailers, e.g. by carrying out large-scale
power consumers by providing nutrition information (23). randomized trials entailing the introduction of innovative la-
Food labels may also prompt the industry to produce healthier bels in a random group of stores. The linkage of sales data
food through nutrient reformulation (38,42). In fact, product with ‘loyalty programmes’ data would provide a solution to
reformulation may actually be one of the main mechanisms the two aforementioned issues. Finally, researchers should
through which food labels could impact consumers’ diets standardize study protocols and, in particular, outcome defini-
(42) in the short term as changing people’s behaviour is more tions. A number of studies had to be discarded because results
complex and may require longer than changing the environ- were reported in non-standard units of measure (e.g. 43–45)
ment. It is up to policymakers to decide the trade-off between or studied unusual outcomes (e.g. 46,47).
completeness of information and facility of understanding. In conclusion, from the evidence produced by this system-
This analysis suggests that interpretive nutrition labels may atic review and meta-analysis, it appears that food labelling
be marginally more effective in steering consumers’ choices. schemes would have a statistically significant effect in
This investigation has three main strengths. First, previ- steering consumers’ choice towards healthier products. In-
ous literature focused on analysing whether consumers un- terpretive nutrition labels, as traffic light schemes, may be
derstand labelling. This study, instead, investigates the more effective than other approaches. Food labels could
effects of labelling on selecting foodstuff. Thus, it quantifies also help consumers in choosing/consuming foodstuff with
how much food labels can modify diets and, eventually, lower calorie content, but the available evidence is currently
obesity prevalence. Second, this study goes beyond a quali- too limited to produce statistically significant results.
tative analysis of the literature by pooling together quantita-
tive evidence in a meta-analysis. Therefore, the results of
this work can be used to compare the effectiveness of food Conflict of interest statement
labels with other labelling policies (e.g. menu labelling) No conflict of interest statement.
and, more broadly, with other policies to tackle overweight
and obesity (e.g. counselling and mass media campaigns).
Third, this study compares the marginal effects of different Acknowledgements
labelling schemes producing new evidence about the ex-
The opinions expressed and arguments employed herein are
pected effectiveness of GDA labels, traffic light labels and
solely those of the authors and do not necessarily reflect the
other types of food labels.
official views of the OECD or of its member countries.
The main potential limitation of this study lies with the
quality of the studies included in the meta-analysis. The ma-
jority of studies are based on a relatively small sample, par- Supporting information
ticularly by considering that the total number of
participants is often divided into groups to test different Additional Supporting Information may be found in the
food labelling schemes. Too small samples may not have online version of this article, http://dx.doi.org/10.1111/
enough statistical power to fully account for possible con- obr.12364
founders (e.g. socio-demographic conditions and price).
Small samples may also explain most of the cross-study Appendix 1. List of studies retrieved in the systematic
(analysis on the selection of healthy options) and within- review but discarded.
study (analysis on calorie intake/choice) heterogeneity that Appendix 2. Additional results.
we found in the retrieved studies. The inevitable assump- Figure A1. Effect of food labels in selecting/purchasing a
tions we had to make cluster together slightly different la- healthier option.
belling formats could be, instead, a limitation for the Figure A2. Effect of food labels on calorie intake/choice
analyses by type of labelling scheme. For example, this (absolute calories).
study groups together standard GDA labels with their
‘coloured’ variant in which levels of nutrients are empha-
sized by a traffic light approach. It cannot be excluded that References
variants of the same scheme have different effects.
1. World Health Organization. Global and Regional Food
Future research on food labelling should focus on address- Consumption Patterns and Trends, 2015. [WWW document].
ing three key deficiencies. The first concerns the small samples URL http://www.who.int/nutrition/topics/3_foodconsumption/en/
of the studies that do not allow for full consideration of (accessed 21 October 2015).
2. Brunello G, Michaud PC, Sanz-de-Galdeano A. The Rise in Obesity 22. Cowburn G, Stockley L. Consumer understanding and use of nutri-
Across the Atlantic: An Economic Perspective. IZA Discussion Papers tion labelling: a systematic review. Public Health Nutr 2014; 8: 21–28.
3529. Institute for the Study of Labor (IZA): Bonn, 2008. 23. Campos S, Doxey J, Hammond D. Nutrition labels on pre-
3. Snowdon C. Institute of Economic Affairs. The fat lie, 2014. [WWW packaged foods: a systematic review. Public Health Nutr 2011;
document]. URL http://www.iea.org.uk/sites/default/files/in-the-media/ 14: 1496–1506.
files/Briefing_The%20Fat%20Lie.pdf (accessed 21 October 2015). 24. European Commission. Monitoring the Activities of the EU
4. Monteiro CA, Moubarac JC, Cannon G, Ng SW, Popkin B. Platform on Diet, Physical Activity and Health – Draft Annual
Ultra-processed products are becoming dominant in the global food Report 2015. EC: Brussels, 2015.
system. Obes Rev 2013; 14(Suppl. 2): 21–28. 25. Crosetto P, Muller L, Ruffieux B. Helping consumers with a
5. Cecchini M. Use of seemingly unrelated equations to assess front-of-pack label: number or colours? Experimental comparison
changes in dietary behaviours during the UK’s economic crisis. between guideline daily amount and traffic light in a diet-building
Lancet 2014; 384(Suppl. 2): S8. exercise. GAEL working paper 2015–10. Laboratoire d’Economie
6. Ezzati M, Riboli E. Behavioral and dietary risk factors for Appliquée de Grenoble: Grenoble, 2015.
noncommunicable diseases. N Engl J Med 2013; 369: 954–964. 26. World Health Organization. Interventions on Diet and Physical
7. Organization for Economic Co-operation and Development. Activity: What Works: Evidence Tables, 2009. [WWW document].
Obesity Update. OECD Publishing: Paris, 2014. URL http://www.who.int/dietphysicalactivity/evidence-tables-WW.pdf
8. Organization for Economic Co-operation and Development. (accessed 21 October 2015).
Obesity and the Economics of Prevention: Fit not Fat. OECD Pub- 27. Cochrane Community. Cochrane Handbook for Systematic
lishing: Paris, 2010. Reviews of Interventions, 2011. [WWW document]. URL http://
9. Lim SS, Vos T, Flaxman AD et al. A comparative risk assessment community.cochrane.org/handbook (accessed 21 October 2015).
of burden of disease and injury attributable to 67 risk factors and 28. Nutrient Data Laboratory, USDA National Nutrient Database
risk factor clusters in 21 regions, 1990–2010: a systematic analysis for Standard Reference. 2015. [WWW document]. URL http://ndb.
for the Global Burden of Disease Study 2010. Lancet 2012; 380: nal.usda.gov/ (accessed 21 October 2015).
2224–2260. 29. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring
10. WHO. Discussion Paper, Prevention and Control of NCDs: inconsistency in meta-analysis. BMJ 2003; 327: 557–560.
Priorities for Investment, 2011. [WWW document]. URL http:// 30. Sterne JA, Sutton AJ, Ioannidis JP et al. Recommendations for
www.who.int/nmh/publications/who_bestbuys_to_prevent_ncds.pdf examining and interpreting funnel plot asymmetry in meta-analyses
(accessed 21 October 2015). of randomised controlled trials. BMJ 2011; 343: 1–8.
11. Méjean C, Macouillard P, Péneau S, Hercberg S, Castetbon K. 31. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-
Consumer acceptability and understanding of front-of-pack nutri- analysis detected by a simple, graphical test. BMJ 1997; 315: 629–634.
tion labels. J Hum Nutr Diet 2013; 26: 494–503. 32. Méjean C, Macouillard P, Péneau S, Lassale C, Hercberg S,
12. Borgmeier I, Westenhoefer J. Impact of different food label for- Castetbon K. Association of perception of front-of-pack labels with di-
mats on healthiness evaluation and food choice of consumers: a etary, lifestyle and health characteristics. PloS ONE 2014; 9: 1–11.
randomized-controlled study. BMC Public Health 2009; 9: 184. 33. Menard C, Dumas C, Gillot N et al. The French OQALI survey
13. Crockett RA, Jebb SA, Hankins M, Marteau TM. The impact on dairy products: comparison of nutrient contents and other nutri-
of nutritional labels and socioeconomic status on energy intake. An tion information on labels among types of brands. J Hum Nutr Diet
experimental field study. Appetite 2014; 81: 12–19. 2012; 25: 323–333.
14. Roberto CA, Shivaram M, Martinez O, Boles C, Harris JL, 34. Van Wezemael L, Caputo V, Nayga RM, Chryssochoidis G,
Brownell KD. The Smart Choices front-of-package nutrition label. Verbeke W. European consumer preferences for beef with nutrition
Influence on perceptions and intake of cereal. Appetite 2012; 58: and health claims: a multi-country investigation using discrete
651–657. choice experiments. Food Policy 2014; 44: 167–176.
15. Watson WL, Kelly B, Hector D et al. Can front-of-pack label- 35. Sinclair SE, Cooper M, Mansfield ED. The influence of menu
ling schemes guide healthier food choices? Australian shoppers’ labeling on calories selected or consumed: a systematic review and
responses to seven labelling formats. Appetite 2014; 72: 90–97. meta-analysis. J Acad Nutr Diet 2014; 114: 1375–1388.
16. Roberto CA, Bragg MA, Schwartz MB et al. Facts up front ver- 36. Swartz JJ, Braxton D, Viera AJ. Calorie menu labeling on
sus traffic light food labels a randomized controlled trial. Am J Prev quick-service restaurant menus: an updated systematic review of
Med 2012; 43: 134–141. the literature. Int J Behav Nutr Phys Act 2011; 8: 135.
17. Gaigi H, Raffin S, Maillot M, Adrover L, Ruffieux B, Darmon N. 37. Temple NJ, Fraser J. Food labels: a critical assessment. Nutrition
Expérimentation d’un fléchage nutritionnel dans deux supermarchés à 2014; 30: 257–260.
Marseille “Le Choix Vita+”. Cah Nutr Diet 2015; 50: 1–9. 38. Emrich TE, Qi Y, Cohen JE, Lou WY, L’Abbe ML. Front-of-
18. Spanos S, Kenda AS, Vartanian LR. Can serving-size labels re- pack symbols are not a reliable indicator of products with healthier
duce the portion-size effect? A pilot study. Eat Behav 2015; 16: 40–42. nutrient profiles. Appetite 2015; 84: 148–153.
19. Wong CL, Mendoza J, Henson SJ, Qi Y, Lou W, L’abbé MR. 39. Ruopeng A. Beverage consumption in relation to discretionary
Consumer attitudes and understanding of cholesterol-lowering food intake and diet quality among US adults, 2003 to 2012. J Acad
claims on food: randomized mock-package experiments with plant Nutr Diet 2015, http://dx.doi.org/10.1016/j.jand.2015.08.009
sterol and oat fibre claims. Eur J Clin Nutr 2014; 68: 946–952. (accessed 03 December 2015).
20. Vasiljevic M, Pechey R, Marteau TM. Making food labels so- 40. Cecchini M, Sassi F. Preventing obesity in the USA: impact on health
cial: the impact of colour of nutritional labels and injunctive service utilization and costs. Pharmacoeconomics 2015; 33: 765–776.
norms on perceptions and choice of snack foods. Appetite 2015; 41. Córvelan C, Reyes M, Garmendia ML, Uauy R. Structural
91: 56–63. responses to the obesity of non-communicable diseases epidemic:
21. Cecchini M, Sassi F, Lauer JA, Lee YY, Guajardo-Barron V, the Chilean law of food labelling and advertising. Obes Rev 2013;
Chisholm D. Tackling of unhealthy diets, physical inactivity, and 14(Suppl. 2): 79–87.
obesity: health effects and cost-effectiveness. Lancet 2011; 376: 42. Hawkes C, Smith TG, Jewell J. Smart food policies for obesity
1775–1784. prevention. Lancet 2015; 385: 2410–2421.
43. Aschemann-Witzel J, Grunert KG, Van Trijp HCM et al. consumption in restrained and unrestrained eaters. Appetite
Effects of nutrition label format and product assortment on the 2014; 82: 1–7.
healthfulness of food choice. Appetite 2013; 71: 63–74. 46. Lahti-Koski M, Helakorpi S, Olli M, Vartiainen E, Puska P.
44. Goodman S, Hammond D, Hanning R, Sheeshka J. The Awareness and use of the heart symbol by Finnish consumers.
impact of adding front-of-package sodium content labels to grocery Public Health Nutr 2012; 15: 476–482.
products: an experimental study. Public Health Nutr 2013; 16: 47. Graham DJ, Laska MN. Nutrition label use partially mediates
383–391. the relationship between attitude toward healthy eating and overall
45. Cavanagh KV, Kruja B, Forestell CA. The effect of brand dietary quality among college students. J Acad Nutr Diet 2012;
and caloric information on flavor perception and food 112: 414–418.