Effectiveness of behavioral interventions and behavior change techniques for reducing soft drink intake in disadvantaged adolescents: A systematic review and meta‐analysis

Summary Reducing sugar‐sweetened beverage (SSB) intake is an important dietary target, especially among socioeconomically disadvantaged ethnic minority adolescents. This review and meta‐analysis evaluated the effectiveness of behavioural interventions aiming to reduce SSB intake in socioeconomically disadvantaged ethnic minority adolescents and examined which behaviour change techniques (BCTs) were most effective. A systematic search was conducted using the PRISMA criteria. Quality assessments were done using the Cochrane criteria. In a narrative synthesis, studies were divided into effective and non‐effective, and relative effectiveness ratios of individual BCTs were calculated. Pooled standardized mean differences (SMDs) and their 95% confidence intervals were estimated with random‐effects models using cluster robust methods. Twenty‐two studies were included in the qualitative synthesis. A meta‐analysis (n = 19) revealed no significant between‐group differences in reduction of SSB intake. Five self‐regulatory BCTs had an effectiveness ratio >50%: feedback, goal‐setting, action planning, self‐monitoring and problem‐solving/barrier identification. The risk of bias assessments were judged to be moderate to high risk for randomized controlled trials (RCTs) studies and low to moderate for pre–post studies. There was no indication of publication bias. In conclusion, self‐regulatory BCTs may be effective components to change SSB behaviour. However, high‐quality research is needed to evaluate the effectiveness of behavioural interventions and identify BCTs effective for reducing SSB intake among disadvantaged adolescents with ethnic minority backgrounds.


| INTRODUCTION
Adolescent overweight and obesity is a major public health priority because of its link to adverse short-and long-term outcomes. 1 Compounding the seriousness of the adolescent overweight and obesity trend are stark disparities by ethnic minority status (i.e., residents with immigrant origins), with minority adolescents with low socioeconomic status (SES, e.g., parental education, occupation status or household income) being disproportionately affected by obesity. 2 In developed countries, higher overweight and obesity rates are reported among socioeconomically disadvantaged ethnic minority adolescents than in advantaged ethnic majority adolescents. 3 Given these findings, the increased risk of morbidities during adolescence and adulthood and the anticipated increase in socioeconomic disadvantage and ethnic diversity, effective prevention efforts are warranted to circumvent obesity-related burden in these adolescents.
Disparities in sugar-sweetened beverage (SSB) intake, defined as nondiet, non-alcoholic and non-dairy cold or warm drink with added sugars including sodas, diet drinks, fruit juices (excluding 100% fruit juice), flavoured juice drinks, sports drinks, energy drinks and sweetened tea and coffee, 4 have been suggested to be a contributing factor to the obesity epidemic. 5 A high intake of SSBs has been shown to be associated with adolescent obesity. 6 While health behaviors such as SSB intake are ultimately based on individual decisions, socio-ecological models suggest that such behaviors are shaped and influenced by the social and physical contexts. 7 Therefore, it is widely acknowledged that the mechanisms underlying the observed differences in SSB intake are multifactorial, involving a complex web of interacting individual, sociocultural, and environmental factors. 8 Ethnic patterning in SSB intake might be explained by the multiple socioeconomic disadvantages that underserved adolescents face, including growing up in low-educated households, interacting with peers who encourage obesogenic-related dietary behaviors and living in segregated deprived areas characterized by a high density of obesogenic food outlets. 9 The inverse relationship between individual, family and area-level indicators of SES is well established; with disadvantaged adolescents exhibiting poor dietary behaviors than their advantaged ethnic majority peers. 10 These findings are attributed to the fact that these adolescents are less healthconscious, have limited nutritional knowledge due to individual or parental education level, and lack financial resources, which can restrict food choices. 11 Within the social and physical home environments, parental SSB intake, permissive food-related parenting styles and home accessibility and availability of SSBs have been linked to a high SSB intake among disadvantaged minority adolescents. 12 Next to the family context, peers can also influence individual SSB intake. Seen from the group norm and socialization perspectives, 13,14 peers help to create norms concerning a behavior (e.g., what SSB behaviors are appropriate and socially acceptable). Given their developmental stage, peer influences are influential during middle adolescence (14-16 years of age). 7 Because this developmental stage is marked by a heightened need for peer approval and identity, it is assumed that peers could influence individual SSB intake possibly via negative peer modelling and conformity to pro-SSB social norms. 7 Underserved adolescents, residing or attending schools located in low-income, high ethnic minority and socially deprived neighborhoods, are found to have fewer health-promoting social norms and are often characterized by a high density of obesogenic food environments. 15,16 Neighborhood-level variables such as SSB availability, accessibility, affordability and disproportionate advertising and targeted marketing of SSBs in deprived areas have been linked to higher individual SSB intake and purchasing. [15][16][17] Because adolescents spend majority of their time in schools, school correlates of SSB intake have also been explored. School food environment and the obesogenic environment around schools in deprived neighborhoods have been shown to influence individual SSB behavior through increased availability, affordability, convenience and in-store price promotion of SSBs. 15,16,18 Public health initiatives such as education campaigns, nutrition guidelines and school policies restricting sales of SSBs have been undertaken to reduce SSB intake. 19 Although important, initiatives that rely on educating or informing individuals are often not sufficient enough to engender behavior change, 20 given that humans are not rational and biased when it comes to their behaviors and recognizing the mindless nature of dietary behaviors. 21 Additionally, these initiatives address only one context and are often not theory driven.
As a result, they are more likely to lead to intervention-generated inequalities, 22 because evidence shows that disadvantaged groups with poor literacy skills are less receptive to interventions that rely on educating or informing individuals. 20,23 Recently, there has been a growing appreciation for using theory-driven behavioral interventions defined as a coordinated set of activities designed to change specified behaviour patterns. 24 Theory-based interventions not only provide a framework for designing interventions but also facilitate an understanding of the underlying mechanisms that drive behavior change. 24 Additionally, they can be used as a guide to identify behaviour change techniques (BCTs) described as the 'active ingredients of an intervention designed to change behavior'. 25 To date, more than 93 BCTs have been identified. 26 Previous systematic reviews and meta-analyses of behavioral interventions targeting adolescent SSB intake suggest that although these interventions lead to small reductions in SSB intake, more evidence regarding their effectiveness is needed. 4,[27][28][29] Some limitations of existing reviews and meta-analyses are that they have included adolescents from the general population or have examined the effectiveness of behavioral interventions within specific settings such as preventive eHealth interventions in school settings. 27  subjectively or objectively at baseline and at post-intervention, in 12to 18-year-old disadvantaged ethnic minority adolescents (Table 1).
To identify studies missed by the electronic search, one reviewer (S. S.) screened the reference lists of selected studies and of relevant reviews to retrieve studies that met the eligibility criteria. Where information was needed, authors were contacted to provide the missing data. Studies identified through the electronic search were exported to Endnote for removal of duplicates. Both reviewers (S. S. and E. G.), independently screened the abstracts, titles and full texts of retrieved studies. Interrater agreement of record screening and inclusion of studies between reviewers was 98%. Disagreements on eligibility were resolved through discussion. One reviewer (S. S.) extracted data of all included studies using an adapted Cochrane spreadsheet form.
Two reviewers (S. S. and E. G.) assessed the risk of bias of RCTs using the Cochrane risk of bias tool (RoB) 31 in Review Manager (Version 5.3 Cochrane, London, UK). One reviewer (S. S.) assessed the quality of non-RCTs using the Newcastle-Ottawa Scale (NOS). 32 For the latter risk of bias assessment, a maximum of nine points was allocated for the least risk of bias in three domains: (a) selection of bias Primary outcome measures -Reduction in SSB consumption quantified as difference between pre-and post-intervention and where possible follow up. Quantity of SSB consumed (ml of SSB consumed per day per week) and frequency of SSB consumed (e.g., percentage of participants consuming a given quantity of SSB), purchases of SSB, energy intake from SSB (e.g., total energy intake in kcal/day or per person) will be included -Studies using different types of SSBs. the outcomes could be reported separately (e.g., soft drinks and sports drinks) or collectively (i.e., all types of SSBs) Secondary outcome measures -Subjective changes in knowledge/attitude/beliefs related to SSB intake

Exclusion criteria
The following studies were excluded: -Animal studies -Studies conducted in adults -Studies reporting only baseline data -Case-control. Cohort. cross-sectional and longitudinal studies -Systematic reviews and meta-analyses (to maintain consistency and because it is difficult to interpret results from previous meta-analyses pooling estimates from RCTs using different control groups. We decided to exclude them from the current study -Meetings/congress reports Abbreviation: PICO, participant, intervention, control/comparison and outcomes; RCT, randomized controlled trials; SSB, sugar-sweetened beverage. (max 4 points); (b) comparability of the study (max 2 points); and (c) ascertainment of exposure and outcomes (3 points). The scores were subsequently categorized as: high (0-3 points), moderate (4-6 points) or low (7-9 points).

| Statistical analysis
The primary outcome was a reduction in SSB intake. Due to the variability in the measures used for SSB intake, a hierarchy was developed whereby the most empirically supported method in each study was preferred. Daily SSB servings were prioritized above volumetric measures or SSB consumed in the past month, because the latter would be prone to the wrong estimation and recall bias, respectively. Effect estimates, reported either separately for each type of SSB (e.g., soda and fruit juice) or collectively (all SSB types categorized into one group) were extracted. [33][34][35][36] Only studies that reported SSB intake at pre-intervention and immediately at post-intervention were included because most studies did not assess long-term effectiveness. In studies with multiple intervention arms, each intervention arm was compared with the control arm at each time point. 37 In total, 15 studies reported mean SSB intake, [35][36][37][38][39][40][41][42][43][44][45][46][47][48][49] and four studies reported proportions of participants consuming a given quantity of SSBs. 33,34,50,51 Three studies were excluded because the change in SSB intake from baseline was not reported. 52 Publication bias was assessed by observation of funnel plots and using Eggers regression asymmetry test. 60,61 An asymmetric plot and p < 0.05 were considered to be suggestive of publication bias.
A sensitivity analysis was performed using fixed effects instead of random effects in the main meta-analysis. All statistical analyses were conducted in R (v3.5.1, Vienna Austria), using the metafor and robust packages.

| Narrative synthesis
Owning to the heterogeneity of BCTs included, a narrative synthesis was performed to identify most commonly used BCTs and types of ( Figure 1). Three studies were excluded from the meta-analysis because only baseline SSB intake was reported. [52][53][54] The characteristics of the included studies are summarized in Quasi-experimental (pre-post with a control group) Setting:

School-based
Provider: Students intervention:  46 The Netherlands To pilot the effects of balance IT, a self-regulation game on dietary intake and PA among secondary vocational students.

weeks
cluster-RCT (pre-post with a control group) Setting:

School-based
Provider: Game intervention:  39 and one lab-based. 54 The total number of intervention arms was 33 with the majority of studies being one-arm interventions (n = 18; range 1-6 arms). Intervention duration ranged from 2 weeks to 9 months and follow up from Fourteen studies [33][34][35][36][37][38]40,42,43,45,48,49,52,53 reported type of beverages included in its definition of SSBs. 4 Eight studies did not report the type of beverages included but reported only measuring SSBs. 36,39,41,44,47,50,51,54 Majority of the studies had a general SSB category without specifying the impact of the intervention per SSB type.

weeks
Most of studies only measured whether or not the intervention was effective in reducing SSB consumption.

| Risk of bias of included studies
The risk of bias of RCTs is summarized in Figures 2 and 3. There was a low risk of bias for attrition bias (n = 10, 67%), reporting bias (n = 13, 87%) and other bias (n = 8, 53%). However, for most studies, the risk associated with blinding of assessors (n = 15, 100%) was high. In addition, majority of studies were deemed to have an unclear risk of bias due to insufficient descriptions related to randomization (n = 10, 67%), allocation concealment (n = 12, 80%) and blinding of participants and research staff (n = 5, 33%). Overall, the risk of the included RCTs was deemed moderate to high risk of bias.
The risk of bias of non-RCTs was evaluated using the Newcastle-Ottawa tool and is presented in Figure 4. Four studies were judged to have a 'moderate' risk of bias, two were rated as having a 'low' risk of bias and one 'high' risk of bias.

T A B L E 2 (Continued)
Author   Figures 5 and 6) revealed three studies that were identified as outliers. 33,34,50 To check if results would change after removing the three influential studies, we reran the primary meta-analysis.
The percentage effectiveness ratio of the BCTs is presented in Figure 9. In total, 13 BCTs had an effectiveness ratio >50%. The BCTs with the highest effectiveness ratio were feedback on behavior (n = 8/ 9, 78%), goal-setting of behavior (n = 9/12, 75%), instruction on how to perform a behavior (n = 6/8, 75%), health consequences (n = 12/16, 75%), followed by action planning including implementation intention (n = 8/11, 73%), self-monitoring of behavior (n = 8/11, 73%) and problem-solving and/or barrier identification (n = 9/13, 69%). Overall, our meta-analysis showed that behavioral interventions were not effective in reducing SSB intake, which is in line with another systematic review and meta-analysis. 27 In their meta-analysis, Champion et al. 27 included four studies and found no evidence found no evidence of effectiveness of interventions on reducing SSB intake.
However, in their review, Champion et al. 27 included adolescents from the general population, used only one SSB measure, and included only school-based eHealth interventions. Further, in contrast to our metaanalysis, the meta-analyses of two previous reviews 4, 28 reported that behavioral interventions lead to a small-sized reduction in SSB intake, but with considerable heterogeneity.
There are several possible explanations for the mixed findings.
First, previous reviews 4,27,28 included adolescents from the general population, whereas we primarily included disadvantaged minority adolescents, given that SSB intake is more common in these groups than in the general adolescent population. 3 Evidence suggests that one-size-fits-all interventions that are delivered in the same way to advantaged and disadvantaged groups may unintentionally lead to intervention-generated inequalities. 22 White et al. 65  and sweetened coffee and tea, 4 we also included effect estimates reported either individually for each type of SSB (e.g., soft drinks and energy drinks) or collectively (e.g., all types of SSB grouped into one category). Finally, in contrast to the review by Champion et al., 27 we included behavioral interventions implemented at the household, neighborhood, school, store and community levels.
In the narrative review, we compared individual BCTs in effective and non-effective interventions to identify BCTs that were associated with intervention effectiveness. Thirteen BCTs had an effectiveness ratio >50% of which five included the self-regulatory BCTs derived from the control theory. 67 These included feedback on behavior (n = 8/9, 78%), goal-setting of behavior (n = 9/12, 75%), action planning including implementation intention (n = 8/11, n = 73%), self-monitoring of behavior (n = 8/11, 73%) and problem solving and/or barrier identification (n = 9/13, 69% information about health consequences (n = 12/16, 75%) were also found to be effective. An example of the BCT provision of instruction is to provide participants specific instructions on how they can reduce their SSB intake. The effectiveness of the BCT providing instruction is similar to another review supporting the efficacy of providing instructions for improving dietary outcomes in adults. 72 The BCT health consequences can be used, for example, to provide factual information on health consequences related to prolonged intake of SSBs. In a comparable review, this BCT was commonly used in educational/behavioral school-based interventions. 29 This review has several strengths. To the best of our knowledge, no review to date has examined the effectiveness of behavioral interventions targeting SSB intake in socioeconomically disadvantaged ethnic minority adolescents. Further, the novelty of this review is that we are the first to explore which BCTs are associated with effective interventions in this diverse sample of adolescents. This is another strength, just as the robust method applied to code the BCTs, using the most comprehensive taxonomy available.
Although the findings of this review and meta-analysis add new knowledge to the scientific literature, there are several limitations that need to be addressed. First, our meta-analysis was based on a small number of studies, and significant heterogeneity was present between the studies. Second, high or unclear risk existed regarding blinding of assessors, randomization allocation and concealment of participants/research staff, and overall, we deemed the risk of bias to be moderate to high for RCTs and low to moderate for non-RCTs.
Third, given the insufficient quantity of studies, it was not possible to conduct proper moderator analyses. Instead, we conducted exploratory analyses to assess the impact of potential moderators. However, due to the considerable amount of heterogeneity observed, the findings of these analyses should be interpreted with caution. Fourth, given that different BCTs and theoretical frameworks were used in included studies, it was not possible to examine, for instance, which specific BCTs and BCT combinations were effective in reducing SSB intake. Fifth, examining the long-term maintenance of the effects of behavioral interventions on SSB intake was not possible in the metaanalysis because most studies had short follow up. Sixth, coding of BCTs was problematic due to inconsistencies in reporting of BCTs. In this review, a stringent coding strategy was used. A BCT was coded as present, only if it was sufficiently described or the authors referred to a BCT taxonomy. The problematic issue related to coding of BCTs has been identified in previous reviews. 63,64 Further, there was an overrepresentation of studies conducted in the USA, which limits generalizability to other contexts (e.g., built environment) and continents This will not only ease the process of identifying active BCTs but will also aid in the development of possibly effective and replicable behavioral interventions.
Further, this review demonstrates several BCTs in reducing SSB intake in underserved adolescents. Based on the narrative synthesis, individual BCTs with greater effectiveness were those that targeted participants' self-regulatory skills such as feedback on behavior, goalsetting, self-monitoring, action planning including implementation intention and problem-solving/barrier identification. Overall, these findings suggest that providing individuals personalized feedback which they can translate into behavior change as well as enhancing their self-regulatory capacity to make them more autonomous about their own behaviors may be more effective rather than offering them generic behavioral interventions. Therefore, in order to effectively facilitate behavior change, researchers should consider incorporating skill-building and self-regulatory BCTs not only to curb SSB intake but also to avoid unintended consequences (i.e., increased intake of sports drinks or energy drinks). Possible ways to curb SSB intake and avoid unintended consequences could be, for example, to promote con-