Childhood and adolescence are considered formative years for food preferences that tend to persist into adulthood(Reference Lien, Klepp and Lytle1,Reference Sawyer, Afifi and Bearinger2) . Upon entering adolescence, peers, current food trends and the surrounding food environment gain progressively more influence on adolescents’ food choices(Reference Neumark-Sztainer, Story and Perry3). This commonly entails a shift towards less favourable eating habits associated with a lower fibre, fruit and vegetable intake, higher intake of energy-dense foods, fat and sugar and more frequent snacking. The last is presumably due to the omnipresence of convenience and highly processed food, often targeted to appeal specifically to this age group. Effective and sustainable strategies to counteract these unfavourable changes in eating habits among adolescents are thus considered vital for an individual’s healthy development and well-being in later life(Reference Sawyer, Afifi and Bearinger2). As adolescents are highly susceptible to environmental influences on food choices, choice architecture interventions (CAI) present a promising tool to promote healthier food choices and eating habits in this age group(Reference Pearson, Ball and Crawford4–Reference Hansen and Jespersen6). CAI are based on the concept that food choices are often the result of habits and patterns involving emotional (and minimal conscious) thought in decision-making(Reference Marteau, Hollands and Fletcher7). By changing favourable food options’ position and/or properties, thus increasing their salience and/or convenience within a given food environment, CAI aims to encourage unconscious choice of healthier v. less favourable food options without actually limiting overall food range available or limiting free choice(Reference Hollands, Shemilt and Marteau8,Reference Hollands, Bignardi and Johnston9) . As adolescents spend a considerable part of their day in school, where they consume one or several of their meals, school cafeterias and lunch rooms appear as ideal settings for CAI implementation and promoting the establishment of healthier food choice habits(Reference Briefel, Crepinsek and Cabili10).
We aimed to carry out a systematic review of existing literature on interventions for healthy adolescents’ food choices using CAI in a secondary school setting to evaluate the impact of such changes in respective food environments in terms of effectiveness, while differentiating between CAI types as well as respective number of implemented interventions. Additionally, we attempt to determine sustained CAI effects as reported food selection changes could result primarily due to a form of novelty effect(Reference Schomaker and Meeter11).
Based on our findings, we would like to propose possible recommendations for the successful CAI implementation for our targeted age group and setting as well as suggest future areas for associated research. Thus, this review aims to answer the question whether CAI can be an effective strategy in promoting adolescents to change existing eating patterns and nutritional choices within a secondary school setting in a lasting way and which factors might contribute to its continuous success.
Method
Selection criteria
To evaluate sustained CAI effectiveness, only quantitative studies in a secondary school setting (i.e. middle, high, or intermediate school, depending on the country and school system) were included in this review. Studies including children from primary schools or students from universities, vocational schools and/or colleges were not included unless the data also included secondary school settings and was analysed in sub-groups, allowing extraction of the secondary school setting data. A similar approach was used for studies reporting on additional educational and/or exercise programmes in combination with CAI. Studies were excluded if only the interaction effect of CAI with the additional programmes was reported. Additionally, studies with a particular focus on a specific subset of secondary school students concerning gender, ethnicities, religion, learning abilities, health and/or social status or studies aimed at specific school environments were also excluded. On account of the selected setting, students’ age ranged from 10 to 12 up to 18 years, depending on the school system of the country in which the study was conducted, corresponding to the WHO definition of adolescence(12) and encompassing the formative years for developing and consolidating major food preferences and eating habits(Reference Lien, Klepp and Lytle1,Reference Sawyer, Afifi and Bearinger2) .
Definition of terms
Favourable/healthy nutritional choices comprise all provisions of foods, beverages and meals higher in nutrient density, lower in energy, salt, sugar, cholesterol and fat as well as less processed items. Based on such choices, students should choose, purchase and consume more vegetables, fruits, less meat, fat, sweets and sugar-sweetened beverages, more whole-grain and unsweetened food and beverages and more water(Reference Gidding, Dennison and Birch13).
According to the CAI definition by the TIPPME (typology of interventions in proximal physical micro-environments)(Reference Hollands, Shemilt and Marteau8,Reference Hollands, Bignardi and Johnston9) , all studies included in this review had to change position/placement and/or properties of objects and stimuli within a given food microenvironment without limiting or restricting the students’ choices. A detailed description of CAI parameters used for this review is presented in Table 1. Studies were excluded that directly influenced students’ choice, e.g. via free food programmes or other monetary benefits. This was also applied to studies that restricted the students’ choice, e.g., by removing unfavourable food and/or beverages from the school, as well as any study where the fundamental quality (such as energy density, fat content or palatability) of the food was changed as part of the intervention. Food environments in a school setting included the school cafeteria, canteen, lunchroom, as well as vending machines or concession stands at school events. The search term ‘food’ includes food and/or beverage.
Table 1 Overview of the classification of CAI implemented in the included studies based on the TIPPME typology(Reference Hollands, Shemilt and Marteau8,Reference Hollands, Bignardi and Johnston9)
* None of the studies matching the inclusion criteria for this review implemented this intervention type.
Implemented CAI impact was evaluated by reported quantitative changes occurring in food choice and/or consumption by students during an intervention. The term ‘pre-post design’ was assigned to any study comparing baseline data prior to the implementation of CAI with that during the intervention itself. Sustainability of changes was assessed from studies that compared data from different time points during the intervention (and/or after changes were removed) with that prior to CAI implementation.
Search strategy, data extraction and synthesis
Authors AEB and EAS systematically searched the electronic databases PUBMED and Web of Science in October 2021 using the search criteria shown in Fig. 1 for subject, intervention theory, intervention target, setting and target group, respectively. No limit was placed on the publication date, with languages limited to English and German. Following the retrieval of 789 records and removal of duplicates, publications were reviewed independently by AEB and EAS at title and abstract level. Four additional records were identified through the screening of reference lists of related articles and reviews by EAS(Reference DeCosta, Moller and Frost14–Reference Bucher, Collins and Rollo18). A total of fifty-two records met the inclusion criteria, or did not meet the exclusion criteria, and were assessed for eligibility in full. Finally, based on the selection criteria stated above, AEB and EAS unanimously agreed on fourteen studies matching the selection criteria included for this qualitative synthesis. Data extraction was initially performed by EAS and verified for accuracy by AEB. Due to heterogeneity among study designs, intervention types and measured parameters, a meta-analysis could not be performed. A narrative synthesis was used to evaluate the studies included in this review, with CAI types, measured parameters, outcomes and main results summarised and tabulated to provide an overview. Studies were additionally divided by the number of interventions implemented (one or multiple) as well as the duration of intervention and time points of result reports during or after the intervention.
Fig. 1 Overview of search criteria and strategy
The quality of the study design was judged based on a procedure proposed by Sanson-Fisher et al., who rated the study design based on its ability to answer three methodological questions (1) has a change occurred (2) was the change due to the intervention (3) is the degree of change significant. Based on their assessment, a cluster randomised trial design is rated as having a strong ability for all three questions, a controlled pre–post design as having a moderate ability for all three questions and an uncontrolled pre–post design to have a moderate ability for questions 1 + 3 and a weak ability to answer question 2. Hence, we judged the overall quality of cluster randomised controlled trials as strong, controlled pre–post design as moderate and uncontrolled pre–post designs as weak-to-moderate (see Table 2).
Table 2 Overview of study design of included studies
Light grey = studies implementing one intervention only.
Medium grey = studies reporting results for different time points during or after intervention.
* Studies from this group have provoked criticism as can be seen at https://peerj.com/preprints/3137.pdf. As a result, some studies had to be retracted. However, any study cited in this review has not been retracted at the point of writing, though results should be considered with care.
† Terminology based on Sanson-Fisher(Reference Sanson-Fisher, D’Este and Carey37); the term pre–post design is used for all studies comparing baseline data with that during CAI. Study quality is also based on Sanson-Fisher (see methods for explanation).
‡ Schools were matched for demographics and then randomly assigned to control or intervention.
Results
Of the 789 records originally retrieved through the indicated search terms, only fourteen matched the inclusion criteria and were included in this review(Reference Conklin, Cranage and Lambert19–Reference Ensaff, Homer and Sahota32). Eleven out of these fourteen were conducted in the USA and one each in Denmark, Germany and Great Britain(Reference dos Santos, Nogueira and Rodrigues22,Reference Winkler, Berger and Filipiak-Pittroff30,Reference Ensaff, Homer and Sahota32) . Four(Reference Wansink, Just and Hanks20,Reference Hanks, Just and Wansink23,Reference Cullen, Chen and Dave24,Reference Hanks, Just and Smith27) of the nine US studies were conducted as part of the smarter lunchrooms movement(Reference Wansink33).
Twelve out of fourteen studies(Reference Conklin, Cranage and Lambert19,Reference Hunsberger, McGinnis and Smith21–Reference Hanks, Just and Wansink23,Reference Greene, Gabrielyan and Just25–Reference Ensaff, Homer and Sahota32) compared data for students’ food choice and/or consumption collected prior to CAI implementation with that during the intervention, with one of these also reporting on student’s post-intervention food choices(Reference Ensaff, Homer and Sahota32) (Table 2). Out of these twelve, five reported changes in food choice within a selected school between baseline and intervention only (uncontrolled pre–post design)(Reference Hunsberger, McGinnis and Smith21,Reference Hanks, Just and Wansink23,Reference Hanks, Just and Smith27,Reference Winkler, Berger and Filipiak-Pittroff30,Reference Koch, Wolf and Trent31) . From the remaining seven, five were controlled pre–post studies(Reference Conklin, Cranage and Lambert19,Reference dos Santos, Nogueira and Rodrigues22,Reference Quinn, Johnson and Podrabsky26,Reference Boehm, Read and Henderson28,Reference Ensaff, Homer and Sahota32) and two cluster-randomised controlled trials(Reference Greene, Gabrielyan and Just25,Reference Kenney, Grotmaker and Howe29) . Two more studies were cluster-randomised but did not report baseline data(Reference Wansink, Just and Hanks20,Reference Cullen, Chen and Dave24) . All CAI were implemented at school cafeterias or lunchrooms with one study catering for the control group at a university laboratory(Reference dos Santos, Nogueira and Rodrigues22). Students’ food selection was determined via sales records (Table 3)(Reference Conklin, Cranage and Lambert19,Reference Wansink, Just and Hanks20,Reference Greene, Gabrielyan and Just25,Reference Quinn, Johnson and Podrabsky26,Reference Boehm, Read and Henderson28,Reference Winkler, Berger and Filipiak-Pittroff30,Reference Ensaff, Homer and Sahota32) . Food consumption was assessed either via visual observation (either directly or using photographs; quarter waste method) and/or tray/plate waste measurements(Reference Wansink, Just and Hanks20,Reference Hanks, Just and Wansink23–Reference Hanks, Just and Smith27,Reference Kenney, Grotmaker and Howe29,Reference Koch, Wolf and Trent31) . Gross calorie and fat intake were determined by weighing selected food and beverage choices before and after service(Reference Hunsberger, McGinnis and Smith21). Frequency of school visits ranged from 1 to 17 d during the course of the CAI intervention period in eleven studies, while three used continuing sales data reports throughout the course of their respective study to calculate results(Reference Conklin, Cranage and Lambert19,Reference Boehm, Read and Henderson28,Reference Ensaff, Homer and Sahota32) .
Table 3 Overview of outcomes and results of studies included in this review
SSB = sugar-sweetend beverages.
Light grey = studies implementing one intervention only.
Medium grey = studies reporting results for different time points during or after intervention.
* Studies from this group have provoked criticism as can be seen at https://peerj.com/preprints/3137.pdf. As a result, some studies had to be retracted. However, any study cited in this review has not been retracted at the point of writing, though results should be considered with care.
† Methodology described in Getts KM, Quinn EL, Johnson DB, Otten JJ. Validity and Interrater Reliability of the Visual Quarter-Waste Method for Assessing Food Waste in Middle School and High School Cafeteria Settings. J Acad Nutr Diet. 2017 Nov;117(11):1816–1821. doi: 10·1016/j.jand.2017·05·004. Epub 2017 Jul 6. PMID: 28688883; PMCID: PMC7261231.
None of the included studies implemented changes in sizing (Table 1), while two implemented changes in functionality(Reference Kenney, Grotmaker and Howe29,Reference Winkler, Berger and Filipiak-Pittroff30) . The intervention type most often implemented was information/labelling (12/14), followed by changes in presentation and position/proximity (8/14 each). This preference was also observed in the four studies implementing only one intervention type(Reference Conklin, Cranage and Lambert19–Reference dos Santos, Nogueira and Rodrigues22): three introduced specific labels (two with high calorie or fat labels, one ‘Dish of the Day’ label) and one offered pre-sliced apples, thus changing targeted product presentation. As a result of introducing calorie or fat content information, one study reported a significant decline in high fat content food sales(Reference Conklin, Cranage and Lambert19), while another saw a significant reduction in student’s gross calorie and fat consumption(Reference Wansink, Just and Hanks20) (Table 3). The study introducing the label ‘dish of the day’ for a vegan or vegetarian lunch alternative reported no difference in meal choice between intervention and control group(Reference dos Santos, Nogueira and Rodrigues22). The study offering pre-sliced apples in addition to whole ones also observed desired effects in apple sales as well as consumption, though for the latter changes were NS(Reference Wansink, Just and Hanks20).
The remaining ten studies in this review implemented two or more CAI simultaneously(Reference Hanks, Just and Wansink23–Reference Ensaff, Homer and Sahota32). Four out of these compared changes in fruit and vegetable selection and consumption as part of the students’ lunch meal(Reference Hanks, Just and Wansink23–Reference Quinn, Johnson and Podrabsky26), with all of these reporting significant increases in fruit selection and consumption during the intervention. Results for vegetable selection and consumption, though showing the desired effect by increasing, were not always significant in comparison to control or baseline. No opposing effects were reported. One study focussed on water consumption only and reported a significant increase within intervention schools after CAI(Reference Kenney, Grotmaker and Howe29).
The remaining five studies investigated multiple CAI for a wider range of outcome parameters, including whole grain products, salads, sweets, milk and water, either recorded separately as part of the main dish, a side serving or as part of a sandwich, or included under healthy or less healthy food items or entrée(Reference Hanks, Just and Smith27,Reference Boehm, Read and Henderson28,Reference Winkler, Berger and Filipiak-Pittroff30–Reference Ensaff, Homer and Sahota32) . Two of these recorded changes occurring during the course of the intervention by comparing students’ choice of specific food groups or consumption of school lunches at the beginning of the implementation of CAI to that of later time points(Reference Winkler, Berger and Filipiak-Pittroff30,Reference Koch, Wolf and Trent31) , while another reported on changes in food selection after interventions were removed(Reference Ensaff, Homer and Sahota32).
In the latter study(Reference Ensaff, Homer and Sahota32), changes in food environment were implemented for 6 weeks (intervention period) during the summer term 2013/14 before the baseline state of the school canteen was reinstated (post period), i.e. all interventions were removed. During the intervention period, sales of designated food items (vegetarian specials, salads and fruits) increased significantly by 3 % and stayed up by 2·2 % in comparison to baseline for the 3 weeks post period even after the removal of all CA changes. In contrast, CAI remained in place in the other two longer-term studies(Reference Winkler, Berger and Filipiak-Pittroff30,Reference Koch, Wolf and Trent31) . In the German study, sales of designated food items were observed in comparison to baseline (phase 1), right after the introduction of changes (phase 2), after 5 months (phase 3) and after 1 year (phase 4) on selective days. Sales in fruit rose by 4·1% to 7·2 % before decreasing to 2·3 % in the last phase of the study – still significantly higher than baseline sales. Though results for all other food items investigated showed significant changes in the intended direction at some phase of the study, results were less conclusive overall. For some food opposing effects to those intended were reported: e.g. sales in vegetables and sweets decreased during phase 2 as did sales for the vegetarian/vegan main dish during phases 2 and 3 before increasing again in a later phase (Table 3)(Reference Winkler, Berger and Filipiak-Pittroff30). A similar design was employed by the US-based study where consumption of different food groups as part of a school lunch menu was observed pre-, 3 months and 1 year after a redesign of the schools’ cafeteria. While non-starchy vegetable consumption increased slightly at the 3-month point, an overall decrease was reported after 1 year. This was also seen for grains and fruits consumption, while white potato and protein consumption increased overall.
Discussion
This systematic review focussed on the target group of healthy adolescents in a secondary school setting while evaluating the impact of implemented CAI depending on type and number of implemented interventions as well as possible sustained effects.
We identified fourteen papers reporting results of one or more CAI according to the TIPPME typology(Reference Hollands, Shemilt and Marteau8,Reference Hollands, Bignardi and Johnston9) , indicating the promotion of healthier eating for our relevant target group in a secondary school setting. The majority of studies included in this review (10/14) implemented more than one CAI at the same time (Table 3). However, multiple changes in food environment do not seem to be essential for success, as three of the four studies implementing only one CAI reported significant changes in food choices for their respective targeted food items in the intended direction(Reference Conklin, Cranage and Lambert19–Reference Hunsberger, McGinnis and Smith21). Due to the limited number of papers reporting results for the implementation of a single change in the cafeterias’ food environment as well as intervention types implemented, no conclusions can be drawn so far with regards to the effectiveness of one intervention type compared to another, or how several interventions might interact for this setting. Both intervention types implemented alone – labelling or presentation – produced significant changes for at least some of the measured parameters without causing any opposite effects. However, while concurrent calorie as well as nutrient content labelling at point-of-purchase did decrease the selection of high-fat entrées in one study, the selection of high-calorie entrées was not affected by that change(Reference Conklin, Cranage and Lambert19). In contrast, by adding only a calorie label to different lunch menus in a similar approach, researchers reported a decrease in gross calorie intake and an associated significant reduction in total fat intake(Reference Hunsberger, McGinnis and Smith21). These results suggest that fewer, but well-aimed, labelling approaches might actually prove more effective than more extensive ones, and that any change in the students’ food environment should be clearly targeted and based on desired aims.
In addition, the impact of one or multiple CAI might be enhanced by assessing the reasons for not choosing certain healthier food items prior to implementing any changes in the food microenvironment. One study reported that one reason why whole apples were only chosen rarely by students as part of their lunch menu was the difficulty in eating them, especially for students with braces(Reference Wansink, Just and Hanks20). By providing pre-sliced apples, a significant increase in apple sales was observed though consumption did not increase to the same extent. Another study focussing on increasing students’ water consumption from existing water fountains in the lunch room achieved its goal by providing information about the safety and benefit of the provided tap water in addition to offering free, reusable drinking cups as part of its CAI, thus specifically addressing concerns and reservations within the student population(Reference Kenney, Grotmaker and Howe29). However, any kind of prior questioning of the student population towards their food preferences could bias them, as adolescents do have the tendency to comply with what they think is expected of them(Reference Klein, Gould and Corey34) thus leading to a form of social desirability bias. Previous studies have shown that students are in general aware about what healthier food choices are and what they are expected to choose(Reference Herbert, Clemow and Pbert35).
Taken together, these results suggest that any CAI should be implemented with special emphasis towards the aims regarding the selected target food group, as even single well-chosen and placed changes can promote effective positive changes in food choices. Future research should be designed to investigate the impact of other intervention types when implemented alone to determine parameters that foster a successful implementation for each type independently. Students’ opinions and attitudes should also be further investigated.
Another area of future interest should be the determination of lasting effects of any CAI implemented, as some of the immediate changes in food choices could primarily be the result of some form of novelty effect(Reference Schomaker and Meeter11). Only three out of the fourteen papers meeting the inclusion criteria for this review were designed to investigate the sustained impact of the changes implemented on the food environment within the investigated student population(Reference Winkler, Berger and Filipiak-Pittroff30–Reference Ensaff, Homer and Sahota32). The major difference between these three studies was the removal of all intervention measures in the UK-based study at the end of the intervention period(Reference Ensaff, Homer and Sahota32), whereas interventions remained in place in the other two studies conducted in Germany and the USA, respectively(Reference Winkler, Berger and Filipiak-Pittroff30,Reference Koch, Wolf and Trent31) . Results from these two studies indicate that permanent changes in the food microenvironment do not necessarily guarantee continuous positive effects. Though changes overall were encouraging in the study conducted in Germany(Reference Winkler, Berger and Filipiak-Pittroff30), they were less clear cut, with some of the food groups investigated showing opposing effects in sales in comparison to control at certain time points over the course of the intervention year. Similar results were also reported for the US-based study(Reference Koch, Wolf and Trent31) where consumption of non-starchy vegetables as well as fruits and grains as part of the students’ school lunch decreased after 1 year of CAI, despite showing promising results after 3 months. Possible explanations offered by the authors were, amongst others, seasonal changes in food preferences by the students, changes and general challenges within the organisational set-up as well as presentation and promotion of certain foods offered in the course of the school year, combined with some habituation effect towards the no longer ‘new’ changes.
These results illustrate the need for further (long-term) investigations into the potential factors influencing food choice and eating habits in a school environment, despite the practical challenges of such investigations in a working environment. Some information in that respect was generated as part of the US-based study as students’ attitudes towards certain factors in the cafeterias were also investigated by attitude scales. While attitudes were highest at the 3-month post-redesign time point, most of them eventually showed no change from baseline after one year. These findings underline that food environments are not static and that the needs of certain target populations are constantly changing with respect to that environment, making it necessary to adjust interventions accordingly to achieve continuous effects. Such adjustments might even include the removal of all intervention measures after a CAI as reported by the UK-based study. Overall positive results in increasing favourable food choices continued, even after the removal of the intervention measures after 6 weeks. Though sales for favourable food items did decline after the removal of CAI in comparison to sales during the intervention period, they were still significantly higher than at baseline.
Taken together, more long-term studies preferably with higher evidence-level designs are needed, which also investigate other external factors affecting adolescent’s food choice. Comparable conclusions are also drawn by Metcalf et al(Reference Metcalf, Ellison and Hamdi36) calling for more research with regard to the impact of individual interventions as well as other factors influencing food choice during school meals. This group investigated nudging across all school forms and also highlighted the difference between food choice and actual food consumption by students. Though a positive association between CAI and food choice is described in their synthesis, conclusive data and methodology for the investigation of consumption of the chosen foods are often lacking. This should also be considered for future long-term studies, especially with regard to the sustainability not only of CAI but also regarding the overall sustainability of meal offers at schools.
Despite the promising findings, one also has to consider the limitations of this review. Only two (though large) databases were searched for this review, yielding the fourteen studies included in this synthesis. Of these, only four can be rated as having a strong quality according to Sanson-Fisher(Reference Sanson-Fisher, D’Este and Carey37), i.e. they were cluster randomized controlled trials, while the remaining ten would be rated as moderate or moderate-to-weak due to controlled or uncontrolled pre–post study design (Table 2). Moreover, only a limited number of studies matched the inclusion criteria, to begin with, and of those, heterogeneity in study design, outcomes, reporting and target foods as well as CAI implemented further restricted our ability to quantify and assess the overall effects of the different interventions. On the other hand, this review is the first to focus on the target group of students in a secondary school setting rather than certain specific food groups, thus providing a more general overview of how to successfully promote changes in this particular age group.
Conclusion
This review found encouraging evidence for the effectiveness of CAI on food selection and consumption of healthy adolescents in a secondary school setting, with all except one study reporting overall positive changes in the desired direction. Neither the number nor types of CAI implemented seemed to be a decisive factor for success. In contrast, long-term studies suggest the needs and demands of the target group should be considered more closely, and CAI should be adjusted over time for the respective food items they are promoting. However, for more differentiated conclusions, more cluster-randomized controlled trials, i.e. designs with a higher evidence level, are needed aimed at investigating not only the influence of CAI types and numbers but also other external factors on adolescent’s food choice over the course of at least one school year.
Acknowledgements
Acknowledgements: The authors would like to thank Linda D. Kleis for her assistance during the preparatory work for this publication. Authorship: E.A.S. and A.E.B. conceived the project and conducted the systematic search. E.A.S. wrote the manuscript, A.E.B. supervised the work. All authors provided critical input to work. Ethics of human subject participation: Not applicable; no separate ethical disclosure is required for a systematic review.
Conflict of interest:
Not applicable.
Financial support:
Not applicable.
Open access
