Obesity Prevention: A Systematic Review of Setting-Based Interventions from Nordic Countries and the Netherlands

Aim Effective evidence-based interventions have an important role in obesity prevention. Our aim was to present a qualitative synthesis of setting-based health promotion interventions on obesity, from Nordic countries and the Netherlands. Methods A systematic review of the literature was completed for studies in the community, schools, and worksite, with BMI as an outcome. A descriptive analysis was completed for all full-text articles meeting the inclusion criteria. Results Thirty-three articles were identified: 7 whole of community, 3 worksite, and 23 school-based interventions. The studies were largely quasiexperimental in design (21/33), with follow-up from 4 months to 8 years. The explicit use of theory was not featured in many of the studies (20/33). No consistent direction for BMI change could be identified in the whole of community interventions (2/7 positive, 2/7 negative, and 3/7 no effect) and no effect for worksite (3/3 no effect) or many of the school-based interventions (1/23 negative, 4/23 positive, 15/23 no effect, 1/23 BMI significant increase only for control group and 3/23 no data available). Conclusions There is a need to prioritise interventions with study designs of high quality, theory, and a participatory approach, for optimal implementation and evaluation of obesity prevention interventions.


Introduction
e rise in obesity in the past several decades has been dramatic worldwide, particularly in the Western world. According to data from 2016, WHO reports that the Nordic countries and the Netherlands have similar rates for overweight and obesity (people with a BMI ≥ 25 kg/m 2 ) that vary within 4 percentage points; from Denmark with the lowest 55.4% to Iceland with the highest 59.1%. ese rates are lower than many Western countries (such as Canada, USA, Australia, New Zealand, UK, France, Spain, Greece, and the Middle East). Similar are the results for obesity (lowest for Denmark with 19.7% and highest for Norway with 23.1%). ese rates are also lower than many Western countries, as mentioned above (excluding France) [1]. However, these rates are still considered high and suitable initiatives are needed in order to reduce them.
Nordic countries and the Netherlands are highly regulated welfare states. ey are also countries in geographical proximity with similarities in their societies such as economic and social policies. erefore, these countries can apply similar initiatives and can be compared with each other. A regional focus allows for a more targeted analysis and provides results and conclusions that can benefit at the regional level [2,3]. In addition, these countries prioritise public health and have been progressive in implementing health promotion strategies, addressing the lifestyle determinants of obesity at a national level. ese strategies have included the provision of national nutrition and physical activity (PA) guidelines and associated campaigns, positive changes to school curriculum, and, in Sweden and Finland, the provision of free school lunches and financial incentives for health promotion at the worksite. Nordic countries, in recent years, have also seen a general shift of responsibility for obesity prevention interventions to the local municipality level [2]. erefore, they can provide valuable information about health promotion in relation to obesity, compared to other European countries or even serve as a model/example for the other countries.
Well planned, implemented, and evaluated setting-based interventions are paramount in measuring the success, future directions, and financial commitment of interventions for obesity prevention. Bottom-up approaches enable taking into account the needs of the intervention participants and the characteristics and resources of the context. is makes interventions more feasible to implement and more salient to the participants; these aspects increase the sustainability of desired outcomes. Research evidence supports the bottom-up approach since it can help overcome barriers of required change [4].
Multilevel approaches that involve the environment of the individual are highly significant for fighting the obesity epidemic, as environmental factors are often a root cause of obesity [5][6][7]. Integrated, multilevel approaches are needed instead of single level interventions targeting separate determinant levels [8]. ese approaches involve intervention components that create a "healthier environment," such as school curricula and built environment changes, in addition to traditional approaches such as individual counselling and screening that have a limited impact. An example of the environmental approach in childhood obesity prevention could be changing the classroom interior, to allow physical activity in all lessons, instead of only during physical exercise lessons. Monitoring this practice showed clear decrease of obesity among the school children in Finland in the school setting-with results in obesity decrease [9]. e school setting is equally important both for children and their parents, especially as this setting is where children spend a large amount of their time during the day. Schools are places where children consume one or more meals per day. ey are places where canteens, vending machines, and restaurants are often available which can negatively influence children's eating habits. In addition, children spend a lot of time sitting in school. Physical education, as well as the provision of available spaces for play and activities, can improve their PA levels. School-based interventions have provided evidence for effectiveness of childhood obesity prevention [10].
Similarly, the worksite setting is of high importance, due to the considerable amount of time most adults spend at work. ere are also opportunities to improve the worksite with exercise facilities, such as access to gyms, and with improved access and availability of healthy food provided in restaurants, canteens, or as snacks that can encourage people towards healthier habits.
Community-based interventions are also very important because they can create a healthier environment for people to live in, through parks, policies on fast food, cycling and jogging tracks, awareness campaigns, and so on. erefore, they can be very powerful for affecting diet and PA habits in a community [11].
A thorough review of community-based interventions, addressing obesity prevention in the Netherlands through an equity lens, reported that these interventions have impacted socioeconomic inequalities in health behaviour positively and negatively [3]. A recent review of lifestyle interventions implemented in European schools including five studies from Norway, Iceland, Sweden and the Netherlands revealed limited studies which reported a reduction in Body Mass Index (BMI) [12]. Moreover, results from a scoping review of 71 community-based interventions against childhood obesity in Europe revealed limited studies where BMI was measured as an outcome [13].
Furthermore, reviews of health promotion interventions implemented at the worksite, globally and in Nordic countries, have found that the majority of studies utilised the worksite as a convenient setting to implement interventions targeted at individual behaviour change, rather than use a setting-based, multilevel approach including changes to the worksite environment [14][15][16]. To the knowledge of the authors, no review has been identified which has given an overview of all setting-based obesity prevention interventions, implemented in Nordic countries and the Netherlands.
Different components that define the quality of a study such as representativeness, randomisation process, comparability of chosen intervention and control groups, attrition rate, and spillover effect/attributability to intervention also need to be considered. e quality of a study affects highly the outcome, and a low-quality study might obscure the impact of the intervention otherwise evidenced. Another important element in evaluating the quality of interventions and an integral part of designing and planning complex interventions is the use of theory. is has also been acknowledged by the British Medical Research Council and forms part of its guidance [17]. e aim of this review was to identify, synthesise, and evaluate the quality of interventions including environmental components based in the in settings from Nordic countries and the Netherlands, aimed at preventing obesity where BMI was measured and reported as an outcome.

Methods
e review of the literature was completed systematically, guided by the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) statement [18], with guidance for the search strategy from a previous review of whole of community interventions [11]. e eligibility criteria for the studies selected was defined using PICO (Participants, Intervention, Comparison, Outcome).

Types of Participants.
Interventions targeted all age groups, living in either the Netherlands or the Nordic countries: Denmark, Finland, Iceland, Norway, and Sweden, regardless of socioeconomic status (SES). e Nordic countries were included in the review based on geographical and cultural similarities. e Netherlands were included due to the similarities of their historical welfare model to that of the Nordic countries in general [19]. Studies selected included participants that were otherwise healthy, for example, not obese or with a preexisting condition, for example, hypertension.
such as cities, villages, or regions, are commonly defined as "whole of community" interventions [21]. An environmental component was defined as any effort in the setting that did not include individual-based strategies such as counselling for individuals, web-based computer-tailored feedback, or individual counselling. Such components were, for example, school curriculum changes, infrastructure and built environment, policies, restaurants, and so on. Since all studies had at least one environmental component, they were all socioecological models [22].
No restrictions were made to length of follow-up. English language studies published in the literature up to and including April 2016 were included in the review. Hospitalbased clinical interventions or those primarily based in the primary care setting were excluded. Furthermore, worksitebased interventions were excluded if the target group was deemed too specialised and not representative of the general population of employees, for example, one professional group only. If there was more than one article referring to different follow-up points, the longest follow-up was chosen as the included article.

Types of Studies.
All intervention study designs other than purely qualitative were included.

Outcome Measures.
Interventions where the outcome was obesity or chronic disease prevention and where BMI was measured and reported as either a primary or secondary outcome were included. Studies that measured only behavioural outcomes including dietary or PA levels were excluded.

Search Strategy.
A thorough search of the databases Medline and Embase through the Ovid search strategy was completed for articles published until April 2016 (Table 1). Additional sources included articles sourced from reference lists of review articles, identified through the original Ovid database search strategy, and from a search of databases: health evidence reviews and the cochrane database. Other sources by snowballing included articles identified from screening references of full-text articles. One researcher essentially performed the search and screening. After duplicates were removed, records were screened by title and abstract by the selection criteria, before full-text articles were identified. Reasons were provided for why articles were excluded by full-text. Full-text articles were reviewed by all authors.
2.6. Extracted Information. Studies meeting the inclusion criteria by full-text were classified by setting and country. Data were extracted independently by two researchers. A descriptive analysis of the studies involved extracting information including study design, participants, gender as a percentage of females, mean age (SD), total follow-up, measure of SES (education), and if a theoretical base (data not shown) was used for the intervention design and implementation. Further assessment of the outcomes of each study was reviewed with information extracted including the outcomes measured, description of the study population units, response rate and loss to follow-up, randomisation used, selection process for setting or community of choice, summary of intervention implemented, and lastly the outcome related to BMI. Where information was insufficient regarding baseline data or intervention design for a particular article, additional reference articles were sourced from respective reference lists or via a search in Pubmed by study name. Lastly, some additional estimated calculations were made by the authors for the response rate, lost to follow-up, and gender, based on the information available from the articles.

Quality Assessment of
Studies. An analysis of the methodological quality of the studies was then completed using a quality assessment tool [23], previously used by the authors of a review of lifestyle interventions in the Netherlands [3,23]. e quality assessment was performed independently by two researchers. Representativeness was considered as a response rate of 60% or more in samples randomly recruited from the study population, or that the study showed otherwise to be representative of the population [3]. In the case of the whole of community interventions, we considered the participants within the community (random selection) as units to determine representativeness. For the other setting-based interventions, we considered the schools or worksites as units to determine representativeness (not the children or students). We also considered the choice of setting/community (e.g., convenience, volunteering, and participation in existing programs) in order to judge whether a sample was representative or not.
Finally, comparability was difficult to determine, especially if some but not all baseline characteristics were similar. Available data were assessed when a study included BMI in their baseline description and were deemed noncomparable if there were differences in BMI, even if there were no differences in other characteristics. In addition, we considered the baseline characteristics comparable, if the intervention and control group were matched or selected based on similar characteristics, such as SES. Table 2 shows the descriptive characteristics and assessment of all three types of setting-based interventions, with Table 3 providing a summary of key characteristics of these setting-based interventions. Out of the total number of studies, seven were whole of community interventions [37,39,42,44,45,52,56], three were worksite-based interventions [34,36,43], and 23 were school-based interventions [24-33, 35, 38, 40, 41, 46-51, 53-55].

Whole of Community
Interventions. Among the whole of community interventions, two were pre-post studies without a control group [37,44], and there was no cluster randomised design in any of them. All other studies were quasiexperimental [39,42,45,52,56]. e percentage of females ranged from 46.0% to 57.2%.
ere was no information for either education level or SES of the participants in 2 [37,44] of the 7 interventions. BMI change for adults was measured in all studies. e total time of followup varied from three to eight years, and four interventions included a cohort (same individuals followed) and crosssectional samples [37,42,44,56].  Journal of Obesity All whole of community interventions focused on risk factors for cardiovascular disease (CVD) and involved individual and environmental components in the interventions. Five studies were multicomponent studies (three components and above) [37,39,45,52,56]. Six studies were interested in diet, PA, and other risk factors [37,42,44,45,52,56]. One study was interested in PA only [39]. Two interventions had components related to worksites [39,45] and one intervention to schools and worksites [37] even though BMI changes for children were not measured in this study. Environmental components such as awareness campaigns were available in all seven studies [37,39,42,44,45,52,56], organised activities at little or no cost in six studies [37,39,42,45,52,56], food stores in four [37,44,52,56], infrastructure in two [39,45], and policy in two (smoking) [45,52]. One study focused additionally on capacity building (at least as reported strategy) [37]. Only one [39,59] of the seven whole of community interventions was explicitly theory-based and used a multilevel approach, and only one [45] mentioned a theory (data not shown).
Changes in BMI for the interventions are presented in Table 2 in a variety of ways, depending on the type of followup (wave of cross-sectional samples or cohort). Out of the seven whole of community interventions, two showed a significantly lower increase in BMI in the intervention (I) group compared to the control (C) group [39,52]. One study showed a significantly higher increase in BMI in the I group compared to the C group [45], and another showed an increase in BMI in a before-after study (comparisons among cross-sectional samples) [44].
ree studies showed no difference (one before-after study with comparisons among cross-sectional samples [37], one with cross-sectional samples and a cohort [56], and one cohort [42]). For the one cohort in this category [42], the differences for I and C groups were not tested. In this category, for the study that included cross-sectional samples and a cohort [56], there Reasons for exclusion by full-text BMI not measured or recorded as an outcome n = 24 Intervention largely individual-based n = 9 Participants do not meet the inclusion criteria n = 2 Short term results n = 2 Not the primary article of a study describing BMI outcomes n = 13 Mass-media campaign n = 1 Figure 1: Study identification, screening, and eligibility, guided by PRISMA.  Journal of Obesity 7

Journal of Obesity
Test for comparisons between (I) and (C): NA Journal of Obesity 11        16 Journal of Obesity Journal of Obesity 17         Mean BMI before-after (I) 18

NS)
Jansen et al. [38] ( e Netherlands) "Lekker fit!" (enjoy being fit!)  Singh et al. [53] ( e Netherlands) "Dutch obesity intervention in teenagers (DOiT)"   24 Journal of Obesity Naul et al. [48] ( e Netherlands) "Healthy children article) for further information on baseline data and design. b Does not apply for mean age when all children are at the same grade (same age). c Mean years of education or SES for schools where we refer to the years of education of the parents and not of the children or adolescents. d High refers to high education as defined as university education [34,42,43,65], or 13 years or more of education [56,57]. e Data presented only for BMI changes and not obesity prevalence changes. Data are presented by gender and intervention only if total data by intervention group are not available. were cross-sectional samples taken for both I and C groups. Only the I group cohort for the study showed a decrease in BMI over time [56].

Worksite-Based Interventions.
All three worksite-based interventions were quasiexperimental and measured BMI change for adults. e percentage of female participants in two of the studies was 37.4% to 50.7%, whilst one intervention included 100% male drivers [36]. For one [36] of the 3 interventions, there was no information for either education level or SES of the participants. e duration of the total follow-up time varied from one to two years, and all were cohorts. Two interventions focused on CVD risk factors [34,36] and one focused on body mass composition [43]. Two of the interventions had an individual component [36,43] and one involved the families of the participants (educational sessions) [36]. One study was a multicomponent study [43] and one had only one component [36]. All three studies were interested in diet and PA [34,36,43]. Environmental components such as awareness campaigns were available in two studies [34,43], canteens in two [34,43], and organised activities in two studies [36,43]. One [43,60] of the interventions was theory-based and built upon a multilevel approach, including setting and individual levels (data not shown). All three worksite-based interventions showed no differences in the changes for the I and C groups. One of the interventions showed a decrease in BMI in the I and C groups (not significant) but the differences for the I and C groups were not tested in this study [34]. In another study, there was a decrease in BMI in the I group but not in the C group [43].
Ten [31-33, 35, 38, 41, 46, 48, 49, 53] of the 23 schoolbased interventions were explicitly theory-based using a multilevel approach: like intervention mapping and whole school participation (data not shown). Most of them did not show any differences in BMI changes between the I and C groups [25, 26, 30-33, 35, 38, 40, 41, 46, 50, 51, 53, 55]. In one study, there were only differences for one of the two schools [28] (different intervention for each school); in another study, there were no differences for 2nd grade children [30] (however, no test for comparing the I and C groups was available in this study). One study showed a significantly higher increase in BMI in the I group compared to the C group [27]. Four studies showed a significantly lower increase in BMI in the I group compared to the C group [28,30,35,48,54]. Of these studies, one showed these differences only for females [35], one only for one of the two schools studied [28], and one only for 3rd grade children [30].
Finally, two studies showed a decrease in BMI in the I group [54] or in the I group for males but not for females [24]. Of these studies, Ask et al.'s study was the only study where there were no significant changes for the I group but there were for the C group [24] (however, no test for comparing the I and C groups was available in this study).
ree studies did not have any available data as BMI mean or percent changes but changes in obesity rates. erefore, they were excluded from comparisons [29,47,49]. Table 4 presents the quality assessment results of the included studies  with the use of supporting articles, where required [57,59,[65][66][67][68]. None of the studies fulfilled all five criteria.

Discussion
To our knowledge, this is the first systematic review focusing on setting-based interventions on obesity prevention in Nordic countries and the Netherlands, which includes all age groups and types of settings. Results for BMI change showed no consistent direction for whole of community interventions (2/7 positive, 2/7 negative, and 3/7 no effect), no effect for worksite-based interventions (3/3), and no effect for many of the school-based interventions (1/23 negative, 3/23 positive, 15/23 no effect, 1/23 BMI significant increase for control group only, and 3/23 no data available). A quality appraisal showed that many studies poorly fulfilled criteria related to representativeness (25/33) and randomisation (20/33) or had no available information on attributability of the intervention (25/33). However, for comparability of baseline data (20/33) and attrition rates (20/33), evaluation was better. eoretical constructs are very important in research in general, for illustrating the associations between variables, the change process, and so on thus helping understand the interventions' mechanisms [69]. Especially in the field of behaviour change which is relevant to healthcare, ample work has been done in developing theories to guide studies. In this area, theory is very much needed and its value has often been underrecognised [70].
ere are examples of interventions in obesity that showed results by using theory constructs' [71,72].
Participatory interventions are also very important because they engage with people whose life-world and meaningful actions are under study. e target group should play a key role in planning, implementing, and adjusting the interventions. When constructing an intervention with a participatory dimension, it entails the mobilisation of people, feeling of empowerment, and self-efficacy. In the long run, this creates a better opportunity for sustainable solutions.
e participatory research methods are geared towards planning and conducting the research process consequently, which means that the aim of the inquiry and the research questions develops out of the convergence of two perspectives: of science and of practice [73][74][75]. Research in the field of obesity has shown that participatory approaches are beneficial [76].
is systematic review illustrates that the studies in general used theory more as a background understanding, than for guiding the interventions, or for the discussion and interpretation of the results, and implications for further research. Several studies (18/33) did not use theory explicitly [24-27, 30, 34, 36, 37, 42, 44, 45, 47, 50-52, 54-56] and others used theories as the basis for choice of study design or approach (3/33) [35,39,43]. Some of the studies (7/33) [31,32,38,41,48,49,53] built on multicomponent and multilevel interventions such as intervention mapping (IM), socioecological theories or models, and models for individual behaviour; for example, theory of planned behaviour (TPB), social cognitive learning theory (SCT), and health belief model (HBM). Especially in the whole of community and schoolbased interventions, the background and theoretical frameworks were linked to theories of empowerment, participation, and whole-school participatory tailored approaches (5/33) [28,29,33,40,46]. Across the studies, there existed "a light way" of using theory. eoretical framework was more often simply referred to, rather than something that was used concretely in problem formulation and as a basic structure in the studies. e lack of theoretical development and use in the field is an important finding. is can point to a greater need for the application of theory more specifically in future settingbased studies, thereby fulfilling the need for developing common standards and concrete theoretical basis for planning, implementing, and evaluating interventions within this field.

Whole of Community Interventions.
e only available review on whole of community interventions on obesity [11] showed surprisingly that there are no available studies focusing primarily on obesity for adult populations. All of the whole of community interventions in our review had a broader scope, often targeting other risk factors for type 2 diabetes and CVD, beyond BMI. In addition, only one study [37] included a component on schools even though there were no measurements of change in risk factors for this young population. Although obesity is part of the causal pathway for these diseases, studies that do not primarily focus on obesity may risk having a weaker impact on BMI changes, due to the multiplicity of efforts directed towards other risk factors. Not only do other factors such as cholesterol, lipids, and blood glucose require a different timeframe to show meaningful changes, addressing alcohol and tobacco consumption also requires different strategies other than diet and PA.
Unfortunately, the vast majority of reviews on obesity and related risk factors are on children. We found only one systematic review [77] from 2003 on community-based interventions and type 2 diabetes that showed no significant improvements for BMI after intervention. Half of the whole of community interventions showed no change, and this did not seem to be related to a short duration of followup or to the use of a nonmulticomponent design. Other factors such as lack of complete implementation can explain the reduced impact in these studies. For two of these [37,56], there was no information on implementation, and for one [42], full implementation was not achieved (problems with implementing PA due to lack of facilities and resistance to jogging) [42]. In addition, there was no information on implementation in the two studies that showed a favourable effect for BMI [39,52]. However, in one of the studies [52], the authors considered that the reason for their success was the use of an integrative approach (participation of many actors such as organizations, sponsors, municipalities, etc.,) and the length of follow-up (five years).
Methodological issues can also reduce the strength of these studies and result in poor outcomes. One major issue is the use of a control group; without one, it is not possible to judge whether the changes are due to the project or due to other changes in that population. For the whole of community interventions, five were quasiexperimental. In addition, two of those used a cohort design, whereas the other three [42,44,56] also used cross-sectional samples, which increase the variance compared to following the same individuals over time. Randomisation is another concern, and it was not used in any of the whole of community interventions. However, randomisation in this setting is often difficult to implement. A review that analysed the effect of randomisation on the heterogeneity of studies showed that randomisation does not introduce a serious bias [20]. However, in all of the studies, there was an effort to choose a control group with similar characteristics to the intervention group, such as SES or a similar or neighbouring community. Apparently, this was not enough to create comparable groups, since 2 studies [45,52] had comparable characteristics at baseline and 2 studies did not [39,42]. Finally, attributability is a very important concern in health promotion studies due to possible 'contamination' or other interventions implemented during the study period. Surprisingly, most of the studies including whole of community did not discuss this possibility even though it can significantly affect the outcomes.

Worksite-Based Interventions.
We were only able to detect three studies, where most of those excluded were studies that used individual counselling. e included studies focused largely on CVD risk factors, similarly to the whole of community interventions. It is very difficult to draw any conclusions from this limited number of studies. However, the results identify a clear need for more obesity prevention studies with an environmental component, implemented in the worksite setting. ere were no changes in BMI observed in any of these interventions (3/3), and this might be partially due to relatively short follow-ups (one to two years) and no multicomponent studies, which usually reflect a serious effort for change in the environment. For example, in only one of the studies [36], support from the family was considered. On the other hand, all studies had a control group, who were cohorts, showed mostly attributability and comparability even though there was no randomisation process employed in any of them. It has to be noted that one of the studies [43] made an effort to use randomisation; however, there was enormous resistance from the different worksites to be randomly selected as controls (out of 128 randomly selected worksites, 12 were finally studied). is is an example showing that methodological weakness can arise despite an effort from the researcher's side.
A review by Maes et al. [78] which included two of our studies [36,43] categorised these studies as "moderate quality," whereas many others were considered of "weak quality." e quality appraisal included important criteria such as prior analysis of the needs of the worksite, integration of the activities in the management practices and daily working life of the enterprise, and theory-based intervention development. e authors also argued that the inconclusive effects on BMI change were mostly due to the lack of sufficient and high-quality studies. However, another review [14] showed a moderate effect on BMI. None of the above reviews can be directly comparable to ours because they also included studies with an individual component (e.g., dietician counselling) which might have resulted in a positive effect for BMI, at least for the intervention duration.

School-Based Interventions.
Most of the studies showed no effect for BMI (15/23). A meta-analysis by Waters [20] showed that interventions in schools were effective for BMI, particularly for children 6-12 years old, and these were similar to the whole of community interventions for children [11]. However, the authors mentioned that there should be caution due to the small study bias and heterogeneity in the studies. Interestingly, nine out of 23 of the school-based interventions in our review were small studies (maximum three schools as I or C group) [24,25,28,29,46,50,51,54,55]. An interesting observation was that all of the single component interventions were in schools, but one [54] showed no changes in BMI (6/7), suggesting that such an approach is very limited for changing outcomes such as BMI. From the reasons described for initiation of these studies, it seems that in most cases, they had a very narrow scope related to the specific needs of the school. e duration of the intervention however did not seem to be related to BMI changes.
In some of the studies [24,30,35], there was no change in BMI (3/23) in the I group, and this is difficult to interpret. It would be tempting to consider this result as successful by suggesting that it stops the further BMI increase; however, there are other possibilities which actually do not reflect a successful programme: (1) there is no true effect or (2) ceiling effect because of normal weight of participants at baseline [35].
As for methodological issues, all school-based interventions were cohorts and had a control group but only nine were randomised studies. It is somehow reassuring that 15 studies showed comparability for baseline characteristics between I and C but that does not mean that studies should not use randomisation as a regular process. However, as mentioned previously, the process of randomisation is sometimes inhibited, as in one case where the schools participated only if they were considered as I schools [33]. Unfortunately, as for the other types of interventions, attributability was rarely discussed.

4.4.
Limitations. Some major limitations of the articles included in our review are (1) very low reporting or discussion on attributability and SES, which is known to be associated with the effectiveness of a health promotion intervention, (2) unclear describing of results in some studies and some missing tests (comparison of changes in BMI between I and C which affected the way results are presented and possibly interpreted), and (3) reasons for choosing a region, municipality, worksite, or school such as practical (easy, already part of a project), with a very narrow scope in school-based interventions or initiation by local authorities which limit representativeness. Especially in the case of school-based interventions, representativeness was largely not fulfilled. However, in whole of community interventions, five were considered representative, showing most likely a more careful design view 'heaviness' of these projects.
All of the whole community studies included awareness campaigns but few components such as infrastructure and policies. For the schools, 19 studies made changes to the curriculum; however, fewer interventions incorporated: improved infrastructure, policies, and school-wide or community level strategies. ese results show a seeming lack in creating a healthier environment through broader and vaster changes, which is considered a major component of a health promotion study. In addition, very few studies implemented or at least described a capacity building process. Across the included studies, there was "a light way" of using theory and theoretical frameworks; these were mostly only referred to, without describing if and how they were used to guide the studies, select tools or interpret the findings.
One limitation of our study was not being able to organise the studies by length of follow-up or type of intervention components. is was due to the heterogeneity of studies. However, if we had decided to restrict our studies based on follow-up time, we would have to exclude a significant amount of studies that provide valuable information. In addition, grouping based on follow-up would have created many subgroup categories especially for the community and worksite-based interventions.

Conclusion
is review has provided an overview of obesity prevention interventions in seven whole of community, three worksite and 23 school-based interventions, implemented in communities of Nordic countries and the Netherlands, where BMI was reported as an outcome. is review was unable to demonstrate associations with BMI outcomes among these settings. However, it is very difficult to distinguish whether these results are due to the heterogeneity of the study designs, or due to poor quality in terms of design or implementation. In addition, initiation of a project especially in the school setting was often motivated by a very narrow scope related to the schools' needs, and not by an effort to test comprehensive strategies for obesity prevention.
ere is a need to prioritise interventions that include study designs of high quality, the use of theoretical constructs to guide the studies, and a participatory approach for optimal implementation and evaluation of obesity outcomes. Use of theory at all levels of an intervention as well as promoting participatory approaches have been acknowledged to improve the effectiveness of different types of interventions, including obesity. Some suggested criteria for 'good' theory in the area of behavioural change are: clarity of theoretical concepts, being explanatory, describing causality, testability, and so on. [79].
However, the future development of a published guideline on the complete, precise reporting of theory on different levels (process, implementation, and evaluation), even if possible in the obesity field, can help improve interventions.
ere are no explicit guidelines for settingbased interventions compared to clinical trials and interventions on diet and PA that focus on the individual. is leads to studies with no standard of quality based on defined criteria.
Guidelines should be created with an emphasis of criteria that can affect the study quality such as the ones we used in our review. An example of such criteria is a minimum period of follow-up based on evidence that shows how long it needs for obesity interventions to show a change in BMI, not factors such as political agendas. Another example is the consistent use of representative samples and randomisation as much as possible, instead of convenient and very small samples, especially for schools and worksites. Overall, it seems that presently there is not a serious commitment in preventing obesity through setting-based interventions and in particular in worksite interventions.
Commitment to further, more advanced research of settings-based interventions in these countries remains of vital importance, to secure and direct future investment for obesity prevention interventions.

Conflicts of Interest
e authors declare that they have no conflicts of interest.