Information-based cues at point of choice to change selection and consumption of food, alcohol and tobacco products: a systematic review

To be included in the review studies were required to be randomised controlled trials, or cluster-randomised trials. Non-randomised controlled trials were only included when investigators had attempted to standardise the groups and minimise allocation bias [22].

Interventions considered eligible for this review were those which involved the comparison of the effect of an information-based cue at point of choice on food, alcohol or tobacco selection or consumption to that of a non-information-based cue condition. Interventions included those conducted in any out-of-home environment, where an individual had a range of food, alcohol or tobacco products to select from; including grocery stores, supermarkets, restaurants, bars, school canteens, and workplace cafeterias. Eligible studies were required to report unregulated selection or consumption (with or without purchasing) of food, alcohol or tobacco item(s). Unregulated refers to the behaviour of individuals not being regulated by explicit instructions or actions of the researcher.

Information-based cues were defined as those comprising any combination of words, symbols, numbers or pictures that convey information about the product or its use (including the impact of its use) [7]. The media by which the information-based cues were communicated could include: point of purchase advertising boards, display stands, banners, posters, flyers, and labelling on store equipment (e.g. shelving, shopping trolley, baskets, cafeteria trays). Standardised in-store announcements (e.g. those conveyed over a speaker system) and standardised information on screens within the out-of-home environment were also included.

Interventions were excluded if the information-based cue was placed directly on the product of interest (i.e. the consumable substance and its immediate or integral packaging), for examples studies which investigated the effect of traditional food labels (back of pack and front of pack labels), or interventions which provided warning messages on tobacco products (noting that systematic reviews are available on nutritional labelling [12, 13, 23]. and tobacco warning labels [24]). Information-based cues which provided specific nutrient or energy claims, (e.g. “low-fat”), those which provided the nutritional content of a product, and those which listed specific nutritional values (e.g. amount of fat, carbohydrate, sugar, salt or energy content, alcohol volumes) were excluded. Information-based cues provided through the Internet or television, interventions targeted or tailored to specific individuals, or those that were designed to be delivered in an interactive fashion (for example interventions delivered in real-time by humans) were excluded. Studies involving non-human participants, and those for which the information-based cue concerned food allergens (e.g. gluten free information) were excluded. No other exclusion criteria were set.

To ensure a complete yet precise search was conducted, we developed a strategy to include medical subject headings and free text terms based on the eligibility criteria. The search was initially developed for MEDLINE ((OvidSP In-Process 1946 to 25th November 2016) (Additional file 1 Table S1), and tested for sensitivity to retrieve a set of reference papers. The search strategy was then adapted for EMBASE (OvidSP 1974 to 26th November 2016), PsycINFO (OvidSP inception to 26th November 2016), Cochrane Central Register of Controlled Trials (CENTRAL 1992 to 26th November 2016) and Web of Science (inception to 25th November 2016), using each individual database thesauri and notes. There was no restriction on publication date, format or language. Reference lists of all eligible articles were searched, and we additionally conducted forward citation tracking using Google Scholar.

All abstracts were imported into a reference manager software package to facilitate selection, duplicates were removed and abstracts screened against the eligibility criteria by the lead author. Potentially relevant full texts were obtained and screened independently by two authors. Disagreements were resolved via discussion and a third author acted as arbiter where necessary.

All articles were assessed using the Cochrane Collaboration risk of bias tool [20]. The tool comprises six specific domains of potential bias: selection bias, performance bias, detection bias, attrition bias, reporting bias and any other sources of bias. The tool was applied to each included study individually by two authors and justification for each judgement of bias (low, unclear or high risk) for each domain was recorded. Any disagreements were resolved by discussion by the two authors, with a third author acting as arbiter where required. Following these guidelines [20] an overall summary ‘Risk of bias’ judgement (low, unclear or high risk) for each study was derived based on the included domains. This means an article was only considered to have low risk of bias if all domains were judged as low risk. If any one domain was judged as high risk, the overall summary was judged as high risk of bias.

Due to the heterogeneous nature of the included articles in terms of intervention, study design and participant characteristics, the results were synthesised narratively. The included articles reported outcome data in various ways; therefore, we have not attempted to standardise outcomes, but present data as stated by authors in each included article.

Seven studies were conducted in the United States of America [25‐31], three in the Netherlands [32‐34], two in Switerland [35] and one in France [36]. Three of the studies targeted selection by children [26, 31, 35], one targeted purchases by university students [28], three studies were conducted in the workplace [32, 33, 35] and one in bars [34]. The remaining studies were conducted in locations used by both adults and children [25, 27, 29, 30], although the most likely consumers would be adults.

Twelve of the 13 studies targeted food intake, eight studies included a measure of fruit and vegetable selection or consumption, of which six solely targeted fruit and vegetables [25, 26, 29‐31, 36]. Four of the studies aimed to increase the selection of healthier food and beverage items or reduce selection of unhealthy options [27, 28, 32, 33]. Two studies targeted selection of both healthy and unhealthy snacks [35]. One study targeted alcohol sales [34], and we did not identify any studies targeting tobacco products.

The review does not permit firm conclusions to be made for the use of information-based cues at the point of choice to change selection and consumption of food, alcohol or tobacco products, nor to reliably estimate the likely effect sizes of these interventions. The reported results, however, clearly indicate the potential beneficial effect that these interventions have, with ten out of the 13 included interventions reporting significant intervention effects for some or all targeted products. Interventions were typically successful in increasing selection of fruit and vegetables; although studies aiming to increase the selection of other healthier products and decrease selection of unhealthy products provide more mixed results. Importantly, we identified a significant gap in the literature concerning these interventions; no included studies targeted tobacco products, and only one targeted alcohol products. In characterising the evidence base, the types of information-based cues employed, as well as the intervention deigns and outcomes were highly variable across studies. Furthermore, suboptimal conduct and reporting of the included studies raises concerns over risk of bias within individual studies and the quality of the overall evidence.

The interventions identified within this review fall within a broader set of physical micro-environment interventions which aim to change the proximal environment to change behaviour, and which have been described in the TIPPME intervention typology [7]. More specifically we focused on interventions classed as “information” within that typology, being those that comprise cues containing any combination of words, symbols, numbers or pictures that convey information about a product or its use [7]. Despite this broad classification, the information-based cues in this review varied considerably. In order to better characterise this evidence base, with a more descriptive level of detail we have therefore, attempted to further categorise these interventions into i) Attention cues, ii) Educational cues and iii) Implicit cues. Six of the interventions are categorised as ‘Attention’ cues, where the cues attempted to draw the consumer’s attention to the target products [25, 26, 29‐31, 36]. Four studies are categorised as ‘Educational cues’, in that they focused on providing educational information about the products, in this case these studies promoted products as the healthier option [27, 28, 32, 33]. Three studies are categorised as ‘Implicit cues’, acting as primes to drive selection of the product [34, 35]. In these interventions, the link between the cue and the target product is not made explicit to the potential consumer. Instead, the cue is intended to induce or influence less conscious behavioural responses via the activation of, for example, semantic relationships or associative processes [34, 35]. This basic categorisation is emergent and based on observing where interventions share common characteristics, and it primarily serves as a descriptive function. It shares similarities with the conceptual grouping used by authors of other reviews [37]. As more research is generated in this area, development of more consistent terminology or a suitably detailed classification system will enable easier grouping of interventions within further evidence synthesis, and allow the integration of more complex theoretical or conceptual ideas.

Category of cue, as outlined above, may be important in determining which cue has the greater effect on selection and consumption. With the exception of the study conducted by Achabal et al. [25], the cues which used brightly coloured images appeared to have a significant effect on increasing selection of fruit and vegetables. These cues were positive in nature and potentially acted to build on previous knowledge regarding the health benefits of fruit and vegetables. Notably, the study by Achabal et al was carried out prior to the introduction of the five a day message in the United States. It was not until 1991 that The National Cancer Institute and the Produce for Better Health Foundation created the 5 A Day for Better Health Program [38]. It is possible that when this study was conducted people were less informed regarding the health benefits of fruit and vegetables. The four studies which provided educational, health promoting cues appeared to be the least effective on changing selection and consumption [27, 28, 32, 33]. These cues may have been less effective in relation to the ‘Attention’ cues, as they generally focused on low-fat products. It has been suggested that different psychological and social processes maybe involved in reducing fat consumption, that is restricting a behaviour, compared to increasing or starting a behaviour, for example increasing fruit and vegetable intake [39].

Degree of exposure to the information-based cue may be important to their observed effect, and should be considered. The study conducted by Folta et al. which investigated the use of recorded, morning audio-announcements promoting bean dishes found no overall evidence that the announcements influenced the selection of these dishes [38]. Notably, however, a sub-group analysis of matched school pairs suggested that in the school where the announcement was played the most often, children were 2.5 times more likely to select the bean dishes than children in the comparison schools (OR 2.49, 1.74–3.53, absolute figures are not presented by the authors). The extent of exposure may also be a consideration for the other studies within this review. For example in both studies carried out by Payne and colleagues [29, 30], consumers were exposed to the intervention for the entire time they were in the supermarket. In addition, children in the study conducted by Hanks et al [26] were exposed to the vegetable characters for the entire time they were queuing for their lunch. These studies may have provided greater exposure to the intervention as compared to other studies included in the review. Indeed, Achabal et al suggest a potential explanation for no observable intervention effect in their study may have been because shoppers missed the signs due to their positioning [25].

The majority of previous reviews that focus on environmental interventions within physical micro-environment have solely examined dietary interventions. This review is novel in examining the effect of information-based cues on food, alcohol and tobacco products. As such, it was able to identity a clear gap in the evidence regarding both alcohol and tobacco products. Only one study concerning alcohol was included, and we did not identify any studies which attempted to change selection or consumption of tobacco products.

Our results are consistent with previous related reviews on nutritional environmental interventions. A review of workplace interventions found more than half (59%) of included studies reported significant effects on behaviour to increase fruit and vegetable intake, increase selection of healthier options and reduce calorie intake [40]. A review of obesity related interventions in grocery stores and supermarkets found they were generally effective in stimulating purchasing and consumption of healthier foods [41]. Within both reviews, the majority of included interventions contained multiple components which the authors acknowledged precluded isolating the independent impact of included intervention components. We purposefully restricted our review to interventions comprising only information-based cues to avoid this problem.

In line with previous reviews, interventions within supermarkets [41] and vending machines [42, 43] appeared to be effective in influencing food selection. In relation to supermarkets, two of the three included studies observed significant increase in sales of fruit and vegetables [29, 30]. It has been proposed that supermarkets are effective locations for environmental interventions as people make a large percentage of unplanned purchases in store [44]. This would likely similarly apply to vending machines which are designed to cater to convenient, small scale purchasing.

This review is, to our knowledge novel in its focus, and describes a body of evidence that had not previously been synthesised. While intervention characteristics varied across included studies, precluding quantitative synthesis, we have both described studies in detail and attempted to further categorise interventions to enhance understanding of the existing evidence. Importantly, the review was conducted in line with the Cochrane Handbook for Systematic Reviews [20] and the PRISMA guidelines [19]. We developed an inclusive search strategy, encompassing multiple databases to capture as many relevant articles as possible, although we cannot exclude the possibility that some eligible articles were missed. The current review is limited by the suboptimal reporting of methods and outcome data of included studies, making evidence synthesis difficult, and disallowing firm conclusions about probable effects of these interventions. Future primary research studies would benefit from adhering to available guidelines for reporting standards [45, 46].