Aim

Our aim was to synthesize existing evidence reviews that investigate how whole systems approaches to diet and healthy weight have been implemented and evaluated nationally and internationally. This review of reviews sought to answer the following questions: 1. What models or theories (these are empirical frameworks, agent based, behaviour change models or statistical models) have been used to implement WSAs to diet, healthy weight, or obesity? 2. How have WSAs to diet, healthy weight or obesity been evaluated to date? 3. What evidence is there of the effectiveness of WSAs to diet, healthy weight, and obesity? and 4. What has been the contribution of the public and/or service users in the development of WSAs to diet, healthy weight, or obesity? Given that we were interested in a broad set of research questions about the current state of the WSA evidence base, rather than a single specific or precise question which would be typical of a standard systematic review, it was determined that a scoping review of reviews was the most appropriate methodological approach [18].

Information sources and search strategy

The following five databases were searched from 1995 to 2022 to identify eligible studies for the review: CINAHL, Scopus, PsycINFO (ProQuest), the Cochrane Library, and MEDLINE. This timeframe was deemed sufficiently wide (12 years before the Foresight report and its accompanying complex systems map of obesity [22, 23]) to include all relevant evidence. The search strategy included a combination of keywords and database-specific terms, including the medical subject headings (MeSH) that cover WSAs in promoting diet and healthy weight: diet, nutrition, malnutrition, eating habit, eating behaviour, food choice, unhealthy diet, healthy weight, unhealthy weight, obesity, obese, overweight, underweight, fat, body fat, body mass index, body weight, physically inactive, physically active, physical inactivity; whole system approach and related terms such as systems approach, system modelling, collaborative working, joint working, multiagency, multiagency working, interagency, interprofessional collaboration, interprofessional teamwork, and community wide. These terms were modified to meet the search requirements of each database. Included papers were those reporting WSAs and their implementation, to identify the levers and opportunities influencing this. In addition, reference sections of identified articles were scrutinised for additional relevant articles.

Eligibility criteria

Review papers were included if they satisfied all the following eligibility criteria: a review paper (narrative, scoping, systematic or meta-analytic); available in English language; published between 1995-March 2022; focus of the review was on the application of a WSA to diet, healthy weight, or obesity; reported the approach, theory, model or framework used to implement a WSA. Papers that did not meet these criteria were excluded from the review.

Selection of reviews

The results from the literature searches were exported into Covidence, a web-based screening and data extraction tool [24]. Duplicate articles were removed. The review team comprised of 13 members of the wider PHIRST research team. Eleven (JI, CB, NL, OF, JJ, IF, KBa, GB, WW, KBr, AW) members of the research team screened the title and abstracts to determine each article’s eligibility for full-text screening based on the eligibility criteria. The PIRg, part of the NIHR-funded PHIRST Connect team, were included in the review team (JJ, SH, KBa, GB) and participated in the title and abstract screening, as well as full-text review to identify eligible articles [25]. Reviewers met throughout the screening process to resolve conflicts and discuss uncertainties related to study selection [20]. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) chart (see Fig 1) reports the phases of article identification and selection [26].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram illustrating the search strategy.

https://doi.org/10.1371/journal.pone.0292945.g001

Full-text screening was carried out independently. A data extraction form was developed by the authors to confirm relevance and to extract study characteristics such as: author information (title, author name and year of publication), aims and objectives of the review, type of review (e.g. systematic, rapid etc.), PICO (population, intervention, comparator and outcomes as seen in systematic reviews), inclusion criteria (where appropriate), models or theories used to implement WSAs, evidence of the effectiveness of WSAs, and public/service user contribution in the development of the WSA. This form was reviewed by the research team and piloted by all reviewers before implementation, resulting in minor modifications to the standardised form. Reviewers OAF and CB met to resolve any conflicts, and to ensure consistency between reviewers and the research question/purpose [20]. OAF mapped the initial process of data extraction onto the form. GB, KB and HG reviewed and corroborated the extracted data on the form, ensuring that all relevant data was included. Articles were excluded at this phase if they were found on detailed inspection, to not meet the eligibility criteria (see Fig 1). Some challenges were experienced in applying the inclusion criteria as the term WSA was relatively new, and some studies although applying a WSA may have been referred to differently. This concern was also evident when examining what model of a WSA was implemented and subsequently evaluated. Both challenges relate to the various terms and definitions that are used for WSA, a consideration raised in previous reviews [16].

Review type 1: Reviews that include studies that have evaluated the implementation of WSAs to obesity

The Bagnall et al. [16] review is a systematic review of studies that have evaluated WSAs to obesity and other complex health challenges. It used the ten features of a WSA set out by Garside et al. [15] to identify the extent to which interventions featured in articles that met their inclusion criteria could be said to be taking a WSA approach to healthy weight/obesity. The review included 65 studies of which 33 focused on obesity. Findings showed improved health outcomes such as reductions in Body Mass Index (BMI), increased parental and community awareness, community capacity building, nutrition and physical activity environment changes, and improved safety and wellbeing of community members [16]. Of the 33 obesity focused studies, 13 included all ten features outlined within Garside et al [15]. Of those 13 studies, five reported health or wellbeing outcomes, for example, consumption of sugary drinks, changes to BMI, changes to physical activity. Two reported outcomes that were associated with the social determinants of health (social economic factors), while eleven reported mainly process outcomes. The review suggests that all ten features of a WSA were not required to achieve changes in health. Improvements in nutrition and physical activity were shown in an evaluation of the Central California Regional Obesity Prevention Program [33], which met all ten WSA guidance features. BMI, parental awareness, and community capacity building was improved in an evaluation of the Romp and Chomp programme in Australia [34, 35], which met nine out of the ten WSA guidance features. BMI was also improved in two non-RCTs of Be Active Eat Well (BAEW) in Australia [36, 37] which met seven out of the ten WSA features and ‘Shape up Somerville’ an intervention in the USA [38], which met eight out of the ten guidance features. Improvements in fitness and BMI z-scores (BMI z score is a measure of how many standard deviations a child or young person’s BMI is above or below the average BMI for their age and gender) were observed in a prospective cohort study of Healthy Living Cambridge Kids in the USA, which met four of the ten WSA features [39]. Taken collectively it would appear that having all ten features of the WSA guidance is beneficial but not always necessary to lead to improvement in BMI or reduction in obesity. The success of a WSA was attributed to some key facilitators. These included the full engagement of stakeholders, good governance, trust and capacity, sufficient time to build relationships, sufficient finance, and the embedding of the WSA within broader policy so that other less obvious policy changes may also impact obesity (e.g., town planning). This meant ensuring that WSA principles were established within the objectives of individual organisations, rather than being something organisations were required to do in addition to their core work. Although initial findings are promising, a cautionary approach was advised when advocating the benefits of a WSA. Many study descriptions of what constitute a WSA, and the outcomes reported in some studies, were limited, or lacked longitudinal follow-up. Furthermore, consistency in definition, application, and thorough evaluation of WSAs was lacking.

Similar to Bagnall et al, the rapid review of WSAs to obesity prevention commissioned by Safefood published in 2021 [12] included evaluations of WSAs tackling obesity and aimed to progress understanding of WSAs to childhood obesity. It included reference to the Garside et al [15] ten features of a WSA and the 10 pillars of the Amsterdam Healthy Weight programme [40] in its assessment of the evidence. The review identified 14 studies that explored WSAs to obesity. Most interventions involved a town or city-level community intervention, with varying durations. Three of the 14 studies reported positive outcomes in health behaviours and/ or anthropometric measures such as improvements in BMI and BMI z-score. These studies included the ‘Healthy Living Cambridge Kids USA’ [39], ‘Romp and Chomp’ [34, 35] and ‘Shape Up Somerville’ [38] initiatives (similar to the Bagnall et al. study above [16]), all were deemed to have strong to moderate methodological quality. Seven studies reported mixed findings on the effectiveness of WSA; three reported negative outcomes; and, one study was on-going at the time of the review, the Amsterdam Healthy Weight Approach (AHWA) in the Netherlands but has subsequently been reported [41]. For the AHWA a logic framework was developed describing the mechanisms underpinning a WSA to childhood overweight and obesity prevention [41]. The aim of the logic framework was to inform further AHWA development, monitoring, and evaluation and to promote lessons learned and share understanding of the wider working principles of WSAs in public health to others embarking on WSAs. The Safefood [12] review identified that facilitators to implementing a WSA were: the need for strong leadership; allocating sufficient time to building relationships; community involvement and capacity building; ensuring consistency in language of WSA across sectors and allocating adequate time, resource, and financial support. Funding and evaluation commitments using existing population and health surveillance data to implement a WSA to preventing childhood obesity was also recommended.

Bleich et al’s [28] systematic review included nine studies, given the year of publication and when WSA was not as commonly used it was no surprise that WSA was not mentioned. The authors did not make reference to a theory, model, or framework of a WSA in their assessment of the literature, but it was retained within our scoping review of reviews because it is one of only three reviews identified where evaluations of the implementation of a WSA to obesity have been included, all studies were community based and some were in response to a policy change [12, 16, 28]. All studies were from high income countries. Some encouraging findings were shown for improvement in BMI/BMI z-scores. A total of nine studies were included; five were randomized controlled trials (RCTs) and four were non–RCTs. Five studies were conducted in the community setting in combination with at least one other setting such as the home, three were conducted in the community and schools, and one study was conducted only in the community. BMI or BMI z-scores were reported in four of the nine studies. Two studies reported significant improvements in physical activity levels and one in vegetable intake. The authors concluded that evidence is moderately convincing that diet and physical activity interventions conducted in the community with a school component are more effective at preventing obesity. They also highlighted that the high level of variability across the research designs may have contributed to inconsistent findings. Notably, studies varied in their use of research design methodologies (e.g., RCTs vs quasi-experimental studies), were located in non-uniform settings, and used different intervention types to improve health (e.g., physical activity or combined diet and physical activity).

Review type 2: Reviews that draw on evidence gathered to present a model, theory, or framework on which to base a WSA approach to diet, healthy weight, and obesity in the future

Nader et al. [32] conducted a literature review of 1,438 articles that focused on obesity, prevention and interventions in pregnant women, infants, children, and adolescents. The rationale for the focus on pregnancy and children was based on intervening as early as possible. They included review articles initially, then used these to refine their search for examining the early life cycle approach to the prevention of childhood obesity. It was not clear from what was reported how many articles met the final criteria for inclusion in the review. They reported that in a Cochrane review [42] article that obesity prevention interventions may produce the largest magnitude of effect early in life, while two Institute of Medicine (IOM) reports emphasized both the need for interventions early in life [43] and the use of a ‘systems perspective’ to fill in gaps in obesity research evidence that can more effectively guide policy [44]. They conclude from their review that early intervention and a systems approach need to be combined as the next logical step in tackling obesity. The authors propose a framework comprised of eight pathways to guide the implementation of a systems approach targeting obesity in early life. They identify that local, state, and national policies influence the physical environment (1) and the social environment (2). Policy also influences the health care system (3) and the physical and social environment influence each other (4). The physical and social environments influence family practices and individual behaviour (5) as does the healthcare system (6). The healthcare system also influences the physical and social environments (7) and finally family practices and individual behaviour also feedback and influence local, state, and national policies. They go on to identify a set of goals or targets for health across pregnancy, infancy and toddlerhood and provide examples of interventions that would be considered appropriate for each of the eight system pathways identified. They also go on to provide recommendations for how such a WSA should be evaluated in the future and conclude that taking their outlined approach and evaluating it are the next logical steps in tackling the obesity epidemic.

Skinner and Foster [30] conducted a systematic review to examine the causes and consequences of obesity using systems science. Systems science incorporates Systems Dynamic Modelling, Agent Based Modelling or discrete event simulation as an approach to look at complex social interventions. A total of 21 studies were included that addressed four areas in systems science. These included: 1) translating obesity interventions to a large scale; 2) Determining the association of obesity with other health or economic outcomes; 3) Reporting the effect of geography on obesity and; 4) The effect of social networks on obesity. The authors concluded that more specific complex models are required that map obesity from childhood into adulthood. In order to incorporate a systems model to obesity in children to adulthood there is a need to ensure there is multidisciplinary involvement, and skills from social scientists, clinical scientists, public health researchers, and researchers with systems science knowledge to input into the development of mapping the system.

Johnston et al. [31] conducted a review of systems science and obesity policy to provide a novel framework for analysing and rethinking population-level planning. Their aim was to demonstrate the use of a systems-based framework to assess solutions to complex health problems such as obesity. They explored the utility of the Intervention Level Framework (ILF) and identified a set of recommendations and specific interventions for decision-makers to implement ILFs to obesity for large scale behaviour change [45]. To achieve this, they reviewed twelve reports published between 2004 and 2013, these consisted of nine strategies or reports written by or for governments or health authorities in the United States and Canada, one Cochrane review of interventions to prevent childhood obesity, and two reports produced by the Institute of Medicine (IOM). They used the ILF framework to code the recommended strategies (or interventions) for tackling obesity within each of the 12 selected reports according to which level of the system they were operating: 1) Paradigm (deepest held beliefs or the ethos of an approach e.g. adopt a holistic view of health); 2) Goals (or targets) e.g. achieve a childhood obesity rate of 5%; 3) Structure (across the system) e.g. recommendations about cross-sectoral collaborations such as food producers and schools; 4) Feedback loops and delays (loop dynamics) which involves collecting and analysing data about the relationships between different factors in the system and how changes in one may influence or cause changes elsewhere e.g. evaluate the effect of a new tax on unhealthy foods; or 5) Structural elements (subsystem specific) e.g. improving food and physical activity environments and running health promotion campaigns targeted at particular groups. They also coded recommended strategies based on variables listed in the Foresight Obesity systems map [23] forming a taxonomy of 30 variables organized around four subsystems; 1) social and individual psychology; 2) food production and consumption; 3) physiology and clinical care; and 4) physical activity. They found that the majority of strategies focused on altering the determinants of energy imbalance through targeting food intake and physical activity at an individual level and that 76% of recommended strategies were Structural (i.e., ILF level 5). The findings suggest there is considerable scope for increasing the systems focus of policy design to address obesity. They concluded that the ILF provides a template to encourage systems thinking and more strategic policy design grounded in complexity science that goes beyond the focus on individual responsibility alone.

Review type 3: Reviews that refer to systems dynamic models, causal loop diagrams and agent based models

Langellier et al. [29] conducted a systematic review of complex systems approaches to diet using Agent Based Models (ABMs) and System Dynamic Models (SDM) incorporating epidemiological data. A total of twenty-seven studies were included; 22 ABM studies and five SDM studies. Factors that influenced diet in the complex system were neighbourhood (e.g., residential segregation), interpersonal (e.g., social influence) and individual level (e.g., food purchasing decisions), while food pricing, food environment, advertising, nutrition labels, and social norms influenced policy decisions. For those using SDM, several studies used longitudinal data to estimate the values of parameters related to either the population (i.e., initialized the population based on demographic and health data from the baseline observation of a cohort study) or processes under investigation. Most studies used empirical data to inform values of key parameters for each of the models, however there was variation in approaches to model calibration and validation. The authors recommended that integrating complex systems modelling like SDM with policy intervention research shows promise for preventing obesity. They suggested that the Childhood Obesity Modelling for Prevention and Community (COMPACT) study is an example of where systems approaches are linked with community obesity interventions where the model can be updated as new implementation and evaluation data become available, it is then that the intervention can be refined to get the optimal outcome [46, 47].

Sawyer et al. [27] conducted a recent umbrella systematic review of the dynamics of the complex food environment underlying dietary intake in low-income groups and produced a data informed systems map based on the 43 reviews included. Studies were extracted if they reported the economic, social, physical, and political food environment associations with dietary intake. The authors reported that the dietary intake system operated within an economic paradigm based on supply and demand, and that the system incorporates five sub-systems: 1) geographical accessibility; 2) household finances; 3) household resources; 4) individual influences; and 5) social and cultural influences. A causal loop diagram comprising 60 variables was reported. Their findings revealed how poor dietary intake in low-income groups can sustain a food environment that increases the accessibility, availability, affordability, and acceptability of unhealthy foods. They proposed that in order to reshape system dynamics that are currently driving unhealthy food environments, innovative strategies are needed to facilitate longer-term management of household finances and socially-oriented practices around healthy food production, supply, and intake. The diagram in the Sawyer et al. [27] article (they cite as Fig 2) conveyed a map of the system illustrating that disruption to the key feedback loop of ‘demand’ for unhealthy food is necessary, and that to influence this, simultaneous interventions targeting the supply of food, household finances and resources, and social, cultural, and individual influences (part of people’s ‘lived experiences’) are likely needed.

RQ1. What models, theories, or frameworks have been used to implement WSAs to obesity?

The models, theories, or frameworks that have been used within reviews are summarised in Table 2 below. These include systems science that incorporate mathematical simulations, such as Agent Based Models and Systems Dynamics Models for mapping the complex system with key features of the system and feedback loops to show how system components links together [30]. According to Skinner and Foster [30] there is a need for multidisciplinary involvement in mapping a system and more specific models that represent diet and healthy weight in children. In Langellier et al.’s [29] systematic review of complex systems approaches, twenty-seven studies highlighted that neighbourhood- (e.g., residential segregation), interpersonal- (e.g., social influence) and individual-level (e.g., food purchasing decisions) factors influenced diet, while food pricing, food environment, advertising, nutrition labels, and social norms influenced policy decisions. Encouragingly there were some longitudinal studies reported in their review. The most recent review was conducted by Sawyer et al. [27] and involved 43 studies that described economic supply and demand factors as the main components underpinning the obesity system. A total of 60 variables were reported and a systems map that could be applied to areas of low income. Four A‘s was described as influencing unhealthy food choice: accessibility; availability; affordability; and acceptability of unhealthy foods. Encouraged in Sawyer et al.’s [27] review was the incorporation of lived experience of the participants, a key feature of public health research in recent years, and that will likely only be more prominent in the design and evaluation of public health research into obesity going forwards [25]. In Bagnall et al.’s [16] review they described articles based on the extent to which they included Garside et al.’s [15] key features of a WSA, although not a theory, this list of features can provide a useful checklist for those developing and implementing WSAs. Thirteen of the 33 studies that were focussed on WSAs to obesity included some of the key features. Findings suggest that all ten features were not required for implementation of a WSA to show effectiveness. Key WSA facilitators according to Bagnall and colleagues were full engagement of stakeholders, good governance, trust and capacity, sufficient time to build relationships, sufficient finance, and the embedding of the WSA within broader policy so that other less obvious policy changes may also impact obesity. In the review conducted by Safefood [12] they described the Amsterdam Healthy Weight Approach (AHWA), and a AHWA logic framework [41] that has shown merit in WSA development, monitoring and evaluation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Identifying the various models and features extracted from the review articles to describe or support a systems approach to diet, healthy weight, and obesity.

https://doi.org/10.1371/journal.pone.0292945.t002

Because the models, theories and frameworks identified and extracted at the review level and included in Table 2 above were either an outcome of the review process or typically used to assess the included literature we also considered the primary studies within the three reviews that focused on the implementation and evaluation studies of WSAs [12, 16, 28]. In Table 3 below we present data extracted from each of the primary studies on which review it came from and the quality rating provided for the study by that review (as applicable; some studies are in more than one review; some in only one); the model, theory or framework (if any) used to develop and implement the WSA (RQ1), the study designs used to evaluate the WSA (RQ2), what the reviews said about the effectiveness of the WSA (RQ3) and any information identified about whether and how the public or end-users were involved in developing, implementing or evaluating the WSA (RQ4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Table presenting primary studies included across both Bagnall et al. [16], Safefood [12], and Bleich et al. [28] reviews arranged by review ranked study quality (highest quality first).

https://doi.org/10.1371/journal.pone.0292945.t003

Notably, those developing, implementing, and evaluating WSAs to obesity, healthy weight and/or diet have drawn on very different frameworks to guide their activities than those typically referenced in reviews of WSA evidence. Several papers reference use of socioecological theory to support them in considering appropriate interventions at the individual, family, community/school, and population/policy level [39, 48–50]. Several do not cite a model as such but describe having used some form of community capacity building or participatory research approach [36–38, 49]. Verbestel et al. [51] reports on the same WSA as De Henauw et al. [48] but reports use of the Intervention Mapping approach for the WSA development [52]. Singh et al [53] also report applying Intervention Mapping to the development of their WSA. Two studies refer to having taken a social marketing approach [49, 54]. Bell et al. [49] also make reference to a modelling system set out in a document authored by the National Health and Medical Research Council in Australia. EPODE was identified by three studies as having been drawn on for WSA development and implementation [40, 55, 56]. Additional models, theories and frameworks mentioned include the Foresight obesity map [55]; the ANGELO framework [57]; The CCROPP model [33]; an adaptation of the RE-AIM framework and the KTA framework [50]; and the 10 pillars of a WSA ]. Finally, Gadsby et al. [56] cite use of the best practice principles for community-based obesity prevention developed in Australia, the World Health Organisation Good Practice Appraisal Tool, and the PESTEL framework (distinguishing political, economic, sociocultural, technological, and physical and legal environments) in addition to EPODE. Four studies did not report any use of a model, theory or framework in the development or implementation of their WSA [34, 35, 58, 59]. Three studies included only within the Bleich et al review [28] were identified as not qualifying as a WSA (see Table 3) so no models are reported in relation to these.

RQ2. How have WSAs to obesity been evaluated to date?

The study designs from the three reviews which focus on evaluation of the implementation of a WSA to obesity by Bagnall et al. [16]; Safefood [12] and Bleich et al [28]are included in Table 3. Examples of both process and outcome-based evaluations have been reported. Bagnall et al. [16] present their process analysis synthesis across all included studies making it difficult to ascertain evidence that relates specifically to obesity. Based on their summary table however it is likely that the process analysis for obesity studies is based on eight mixed methods evaluations [33, 34, 50, 54, 64–67] and seven qualitative studies [68–74]. Quality assessment of these studies is summarised in terms of numbers of studies meeting or not meeting quality criteria, but specific studies are not referenced in the text description covering this assessment. Many studies did not meet several quality criteria or did not provide sufficient information for an assessment to be made. The Safefood review [12] and the Bleich et al review [28] focused only on studies reporting quantitative outcome data. Although Bleich et al [28] engaged in risk of bias assessment and strength of evidence assessment for their nine included studies, they did not report this information by study. Instead, it was reported as an aggregate by setting of the intervention. All studies had either a high (n = 4) or moderate risk of bias (n = 5). Most studies were identified as providing insufficient strength of evidence (n = 6). Three studies based in both community and school settings were identified as providing moderate strength of evidence.

Outcome evaluations methods applied across these three reviews include: Two cluster randomised controlled trials [Sallis et al [61] (included in Bleich et al [28] only; quality rating not aligned with paper)]; Singh et al [53] (included in Bleich et al only; quality rating not aligned with paper)].

Seven non-randomised controlled trials [Johnson et al. [36] (included in Bagnall et al. [16] only; rated as good quality]; Sanigorski et al. [37] (included in Bagnall et al. [16] only; rated as good quality); Economos et al. [38] (included in all three reviews; rated as moderate to good quality by Bagnall et al. [16] and moderate quality bordering on strong by Safefood, [12]); Hoelsher et al. [60] (included in Bagnall et al. [16] only; rated as moderate quality); Bell et al. [49] (included in Safefood [12] only; rated as moderate quality); Verbestel et al. [51] (included in Safefood [12] only; rated as strong); Five mixed methods evaluations [DH, [54] (included in Bagnall et al. [16] only; rated as moderate quality); de Groot et al. [34] (included in Bagnall et al. [16] only; rated as moderate to poor quality); de Silva-Sanigorski et al. [35] (included in Bagnall and Safefood only; rated as moderate to poor quality by Bagnall et al. [16] and moderate quality by Safefood [12]); Amed et al. [50] (included in Bagnall et al. [16] only; rated as poor quality); Gadsby et al. [56] (included in Safefood [12] only; rated as poor quality); Two prospective cohort studies (Chomitz et al. [39] (Bagnall and Safefood only; rated as moderate quality by Bagnall et al. [16] and strong by Safefood [12]); Vinck et al. [58] (Safefood [12] only; rated as moderate quality bordering on strong); A prospective cohort study with embedded non-randomized controlled trial (de Henauw et al. [48] (included in Safefood [12] only; rated as moderate quality bordering on strong); Two pre-post designs [Atalla et al. [59] (included in Safefood [12] only; rated as moderate quality); Mead et al. [75] (included in Safefood [12] only; rated as moderate quality); One pre-post design with matched comparators from existing national survey data (Raine et al. [57] (included in Safefood [12] only; rated as moderate quality bordering on strong); an evaluation of unclear design [Schwarte et al. [33] (included in Bagnall et al. [16] only; unclear quality); And one nested case control study [Copeland et al. [55] (included in both Bagnall and Safefood reviews; rated as moderate to poor quality by Bagnall et al. [16] and as low quality by Safefood [12].

The Safefood Review [12] also included two articles focussed on the Amsterdam Healthy Weight Approach [40, 76] and presents data on changes seen at a whole-city level on rates of obesity that present reasons to be optimistic. Also, in a recent article by UNICEF [77] the positive outcomes of the AHWA are described. Full peer-reviewed publication of findings are still to emerge however, and until they are, attributing causality to the programme is cautioned against.

RQ3. What evidence is there of the effectiveness of WSAs to diet, healthy weight, and obesity?

Only four articles evaluating the implementation of WSAs to diet, healthy weight and obesity were identified as being of good quality across the Safefood [12] and Bagnall et al. [16]reviews. All four of these articles report a positive effect on outcome variables related to healthy weight, diet, and obesity. Chomitz et al. [39], who report on the ‘Healthy Living Cambridge’ WSA, is reported to have found increases in physical fitness and decreases in BMI and BMI z score by Safefood and to have had positive effects on fitness and obesity by Bagnall. There are apparently conflicting reports about the direction of outcomes by Safefood and Bagnall on the impact of the ‘Be Active Eat Well’ WSA reported in Sanigorski et al. [37]. Safefood report increased BMI z scores whilst Bagnall et al. report positive effects on BMI and health behaviour. Closer inspection of the actual article reveals that children in the intervention arm gained less weight, had lower increases in waist circumference and lower increases in BMI z score and waist to height ratio than did controls but that there was no overall significant difference between groups in the incidence of overweight or obesity [37]. Bleich et al [28] reports a desirable effect of the WSA on BMI and BMI z score from baseline. The Johnson et al. paper [36] which was only included in the Bagnall review also reported on the ‘Be Active Eat Well’ WSA. Positive effects on BMI and health behaviour were identified. The ‘Shape Up Somerville’ WSA reported on by Economos et al. [38] and included in all three of the reviews was identified as having had a positive effect on (i.e., decrease in) BMI for participants. All four of these WSAs were identified as including at least some of the NICE guidance features for a WSA (at least 4/10 in the case of ‘Healthy Living Cambridge Kids’ (Chomitz et al., [39]) as reported by Bagnall et al. [16]. In addition, they were reported as having been developed by drawing on either socioecological theory or by using some kind of capacity building or community participatory approach.

Ten articles across the Bagnall et al and Safefood reviews were identified as being of moderate quality. The best ratings for studies included in the Bleich et al [28] review was moderate risk of bias and moderate strength of evidence, so we include commentary on the three WSA studies included only in their review here too [53, 61, 78]. These included two articles focused on the European Commission funded ‘IDEFICS’ WSA (De Henauw et al. [48]; Verbestel et al. [51]), two articles focussed on the ‘Romp and Chomp’ WSA {De Silva-Sanigorski et al. [35b]; De Groot et al. [34]), and two articles focussed on the UK ‘Change4Life’ WSA (Copeland et al. [55]; Department of Health, [54]). These thirteen studies identified mixed results on outcomes relevant to obesity, diet, and healthy weight. A number showed mainly negative or non-significant findings. The two ‘IDEFICS’ papers report increases in BMI z scores, fatness, body fat mass and waist to hip ratio (though it should be noted the target population were children who are still growing) (De Henauw et al. [48]) and decreases in physical activity and increases in sedentary time (Verbestel et al. [51]). In the ‘OPAL’ WSA, it was reported that there were no significant changes over time on multiple anthropometric measurements or BMI z score, multiple health behaviours including fruit and vegetable intake, nor health related quality of life (Bell et al. [49]). The ‘Change4Life’ WSA was reported to have had mixed effects on health and wellbeing outcomes with no reduction in obesity prevalence (Copeland et al. [55]), but positive effects on parental health behaviour and awareness (Department of Health [54]). Raine et al. [57] reported no significant changes in chronic disease risk, BMI, or blood pressure, Vinck et al. [58] reported no significant change in BMI. Chang et al [78] reported no significant change in prevalence of obesity and Singh et al [53] reported no significant differences between groups on change in BMI.

Two of the ten articles reported mixed findings within their data. Hoelshcer et al. [60] reported mixed effects for obesity prevalence and Mead et al. [75] reported increases in health knowledge and behaviour and healthy intentions, and decreases in unhealthy food consumption but no significant change to BMI was reported. The study by Sallis et al [61] evaluating the M-SPAN WSA, showed a desirable intervention effect for change in BMI, but only in boys. The one WSA reported on by moderate quality studies that had fully positive results was the ‘Romp and Chomp’ WSA. De Silva-Sanigorski et al. [35] report that it increased vegetable servings per day, increased physical activity and led to decreases in BMI in 3.5-year-olds. Parental awareness and community capacity building were also identified as positive outcomes in addition to positive effects on BMI by Groot et al. [34]. The WSA showing the most positive effects from these moderate quality studies (the ‘Romp and Chomp’ WSA) was not reported to have been developed or implemented using a theory, framework, or model. However, Bagnall et al. [16] identified the ‘Romp and Chomp’ programme as including 9/10 of the WSA NICE guidance features.

Four studies identified to be of poorer quality also report varied results. The Safefood review [12] reported on Atalla et al. [59] which identified increases in physical activity, increases in recommended fruit and vegetable intake and decrease in BMI among those with overweight or obesity. They also report on a study by Gadbsy et al. [56] with unclear reporting but including a non-significant decrease in BMI. Bagnall et al. [16] reported that the LIVE 5-2-1-0 WSA showed early indications of improvement in community awareness and action to promote healthy behaviours, but this data were not from robust outcome indicators. They also report on the ‘Central California Regional Obesity Prevention Program’ (CCROPP), which showed positive effects on nutrition and physical activity environments.

The Safefood review [12] also highlighted work on the Amsterdam Healthy Weight programme [40, 76] that was showing some promising early indicators of success, but an outcome evaluation was yet to be published at the time of the review, so caution was suggested in identifying this as an effective WSA at that time. Taken together the evidence for the effectiveness of WSAs in tackling population level overweight and obesity remains limited. Notably, the strongest study designs tend to suggest that WSAs are effective. However, more robust evidence with data captured over longer timeframes is still needed. The data also appear to suggest that adherence to all ten WSA NICE guidance features is not necessary for effective WSA development and impact, but may be helpful. So far there is no particular evidence that use of a specific guiding model, theory or framework is needed to support development of an effective WSA. It is noted however that participatory and capacity building approaches have been cited by several of the WSA studies where positive outcomes were achieved and we address the question of public involvement in WSA development and implementation next.

RQ4. What has been the contribution of the public and/or service users in the development of WSAs?

The contribution of public and service user involvement in the development of WSA approaches has not been reported widely in the eight articles included in our review of reviews. This is with the exception of lived experience of food environments being identified in a systems map in the Sawyer et al. [27] review article and the importance of community involvement being highlighted in Bagnall et al. [16]. Consideration however of the primary studies that were included across the Bagnall et al. review [16] and the Safefood review [12] suggest that public involvement has played a significant role in the development, implementation, and evaluation of WSAs to diet, healthy weight, and obesity (see Table 3 above). For example, the four highest quality studies with positive outcomes reported across these two reviews explicitly state having either used a community-based participatory research approach (Chomitz et al., [39] ‘Shape Up Somerville’; Economos et al., [38] ‘Healthy Living Cambridge’) or took a capacity building approach, whereby communities were supported to create their own mix of solutions and interventions (Sanigorski et al. [37]; Johnson et al. [36]. Other poorer quality studies have also included WSAs that were developed with extensive community consultation and stakeholder engagement (e.g., De Silva-Sanigorski et al. [35]; De Groot et al. [34]; ‘Romp and Chomp’ which reported positive effects). Whilst public, end-user and stakeholder engagement cannot guarantee the success of a WSA for tackling diet, healthy weight and obesity, the evidence from these reviews suggests that it is an important component of getting the approach accepted by communities in which change is to take place [25].