Effectiveness of Lifestyle Interventions for Prevention of Harmful Weight Gain among Adolescents from Ethnic Minorities: A Systematic Review

The escalating obesity among adolescents is of major concern, especially among those from an ethnic minority background. The adolescent period offers a key opportunity for the implementation of positive lifestyle behaviours as children transition to adulthood. The objective of this review was to examine the effectiveness of lifestyle interventions for adolescents and their impact in ethnic and racial minorities for the prevention of overweight and obesity. Seven electronic databases were searched from 2005 until March 2019 for randomized controlled trials of lifestyle programs conducted in this population. The main outcome was change in Body Mass Index (BMI) z-score (kg/m2) or change in BMI and secondary outcomes were changes in physical activity and diet. Thirty studies met the inclusion criteria. Seven studies reported and/or conducted subgroup analysis to determine if ethnic/racial group affected weight change. None demonstrated an overall decrease in BMI z-score. However, six of the seven demonstrated changes in secondary measures such as fruit and vegetable intake and screen time. Results did not differ by ethnic/racial group for primary and secondary outcomes. Overweight and obesity prevention among adolescents from ethnic minorities is an area that needs further research. There is a lack of interventions that include analyses of effectiveness in ethnic minorities.


Introduction
The World Health Organisation recognizes the role of primary prevention of obesity in childhood and adolescence is critical to halt the early onset of chronic diseases in adulthood [1]. Rates are alarmingly high in many countries, with data showing nearly 1 in 5 children and adolescents aged 6-19 years suffer from obesity in the United States [2], and almost a third aged 5-17 years are overweight or obese in Australia [3]. Not only is the quality of life compromised in adulthood, but increasing healthcare costs of resultant non-communicable diseases place a greater financial burden on our healthcare systems [4,5].
Adolescence is a complex life stage as adequate nutrition is vital for growth and development; but poor diet quality may have detrimental effects leading to obesity and other risk factors for chronic disease [6]. The transition from childhood to adulthood involves the emergence of a sense of autonomy that might precipitate resistance to perceived authoritarian healthy lifestyle programs [7]. Moreover, adolescents present with the unhealthiest diets of any age group [8,9] and many do not meet national guidelines for physical activity [10,11]. While there have been numerous programs developed and tested, mostly in schools, they have not always been designed for inclusion of ethnic minorities [6].
Furthermore, the burden of obesity is not equally shared by all sectors of society, with ethnic minority groups most affected [6,12]. This has been shown to be an international phenomenon, with increased obesity rates concentrated in ethnic minorities in several developed countries. For example, in the USA, Hispanics (25.8%) and non-Hispanic blacks (22.0%) are reported to have higher obesity prevalence than non-Hispanic whites (14.1%) among children and adolescents aged 2 to 19 years of age [13]. In the UK, children from black and Asian backgrounds had higher proportions of overweight and obesity, and were three times more likely to present with obesogenic lifestyles than their white counterparts [14]. Adolescents in Australia from the Middle East, North Africa and Oceania regions are exhibiting greater prevalence of overweight and obesity than those from English speaking countries [12,15].
Moreover, U.S. minority groups are most effected by elevated rates of overweight and obesity, as they may also transition away from healthy weight at a younger age when compared to the White population. A multi-ethnic study drawing on data from the U.S. National Health and Nutrition Examination Survey spanning early childhood to late adulthood demonstrated obesity disparities to be evident by two years of age especially for African American females and Mexican American males and females. Disparities in rates of obesity and early transitions to obesity are concerning, given the research demonstrating the difficulties in returning to a normal body weight once an individual has overweight or obesity and inequalities become exacerbated if left unaddressed. This necessitates the additional research needed to prevent obesity during this critical age epoch [16].
To our knowledge, no systematic review focusing on the effectiveness of lifestyle interventions for the prevention of harmful weight gain leading to overweight and obesity in ethnic and racial at-risk populations has been published. Therefore, the aim of this review is to (i) systematically examine the effectiveness of randomized controlled trials (RCTs) of lifestyle interventions for the prevention of overweight and obesity in adolescents and determine how many targeted or included racial and ethnic minorities; and (ii) review the impact of such interventions for adolescents from racial and ethnic minority groups. For the purpose of this review, U.S. definitions of racial and ethnic minorities were used, with American Indians/Alaskan Natives and non-Hispanic Blacks to be considered a racial minority and Hispanics to be an ethnic minority. Race is defined on the basis of physical differences that groups or cultures consider, whereas ethnicity encompasses shared cultural norms such as beliefs, practices, language and ancestry [17].

Materials and Methods
This systematic review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [18] and is registered with PROSPERO (registration number: CRD42018092825).

Identification of Studies
Seven databases were used for a systematic literature search; Medline, PsycINFO, Eric, CINAHL, Web of Science, Embase and Cochrane Central Register of Controlled Trials. The databases were searched from January 2005 until March 2019. Articles published prior to the year 2005 were excluded due to the changes in technology, lifestyle, diet and physical activity that have occurred since then. The reference lists of articles located from the search were also hand searched for any additional missed papers. The search terms were "adolescents", "lifestyle", "prevention", "intervention", "weight loss", "weight changes", "diet" and "physical activity", and synonyms of these were included. The terms were broad and combined and truncated to encompass all studies that may fit the population, intervention, study design and outcome criteria, therefore lowering the chance of exclusion of any studies that may be eligible. A complete search strategy used in the electronic database Medline is shown in Table S1.

Quality Assessment
The Cochrane assessment tool was used to assess the risk of bias at an individual level in RCTs [30]. The tool focused on biases in selection, attrition, detection and reporting. Two independent reviewers evaluated each study for biases as either low, medium or high risk. Conflicting decisions of risk rating were resolved by a third-party reviewer.

Grading of Recommendations Assessment, Development and Evaluation Assessment
The quality of the body of evidence was assessed by two reviewers using the Grading of Recommendations Assessment, Development and Evaluation system (GRADE) [31]. Five categories were assessed: limitations in study design and implementation; directness of evidence regarding study populations, study design and outcomes measured; inconsistency of results; precision of outcomes; and the probability of publication bias.

Study Selection
The search yielded 11,366 records including eight additional articles from hand searches of paper's reference lists. Duplications were removed, leaving 7409 abstracts for the first screening, thereafter another 7205 were excluded as they failed to meet the inclusion criteria, i.e., not an intervention, not the primary outcome, not the study design. Two hundred and four papers were screened for full text; 159 papers were excluded because they did not meet inclusion criteria. After the second screening, 30 studies from 45 papers [22][23][24][25][26][27][28][29] (including protocol and long term follow-up) were eligible to be included in this review. Of these studies, only three were conducted in ethnic/racial minorities [33,36,51] and four included analyses by ethnic and racial groups [28,29,61,68]. Figure 1 shows the review process in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses(PRISMA) flowchart. Please see Table S2 for the list of full text articles excluded and supported reasoning.

Setting and Study Design
Of the 7 RCTs conducted in racial and ethnic minority groups, two were cluster RCTs [28,68]. Papers were published from 2006 to 2014. Five studies were performed in the United States of America (USA) [29,33,36,51,68], and two in the Netherlands [28,61]. Three interventions were conducted in schools [28,61,68], four were carried out in the community [29,33,36,51]

Study Characteristics
The duration of the seven interventions in minority groups ranged from 8 weeks [36] to 8 months [61]. Follow-up duration ranged from 6 [36] to 24 months [28] from baseline. Three did not conduct follow-up measures [29,51,68]. Two studies were single component dietary interventions [29,51], six were multicomponent with four deploying a combination of dietary and physical strategies [28,33,36,68] only to prevent overweight and obesity in adolescents from ethnic and racial minority backgrounds. All were educational and six were behaviour-change-based interventions [28,33,36,51,61,68]. Five were reported to be underpinned by an explicit theoretical framework(s) such as the following: Social Cognitive Theory [33,36], Social Learning Theory [68], Theory of Interactive Technology [68], Theory of Planned Behaviour [28], Trans Theoretical Model [36], Intervention Mapping [61] and the Precaution Adoption Process Model [28]. One study applied Motivational Interviewing [33], and two studies incorporated parental involvement [29,36] to aid adherence to the intervention and ultimately a healthier lifestyle for the long term. One intervention was grounded in weight control principles [51]. Lastly, all studies had one intervention and one control arm, with the control group not being exposed to the intervention.

Participants' Characteristics and Recruitment Strategies
Participant characteristics for studies that tested on efficacy across ethnic and racial backgrounds were as follows; mean age of participants was from 11.3 to 15.3 years old, with an overall range from 11 to 17 years. All studies except one [51] recruited both sexes, with the majority being female and the remaining study recruiting females only. Of the seven studies, six reported on ethnic/racial group of its participants. The study by Singh et al. [61] did not report, as they found no interaction of ethnic/racial group with intervention effect and hence did not provide stratified analysis. Of the seven, two were conducted exclusively in a singular racial minority; one recruited from African Americans [33] and the other from American Chinese backgrounds [36]. Four [28,29,61,68] were universal interventions that included both White and minority populations [28,29,68], and one study recruited participants from various ethnic groups only [51]. The studies by Black et al. [33] and Nollen et al. [51] targeted adolescents from low-income neighbourhoods to access those from racial and ethnic backgrounds. Three recruited from schools [28,61,68], three from community and one intervention recruited from an existing longitudinal study and schools [33]. One study did not report recruitment methods [33]. Please see Table 1a for further information.  Participants' characteristics for the remaining studies were as follows; mean age of participants within the studies was 11.3 to 15.8 years old, with overall range 11 to 17 years. Seven recruited only females [24,26,38,44,47,53,54], three recruited only males [45,53,64] and one did not report sex [27]. The remaining 12 studies recruited indiscriminately with the majority recruiting more females than males. Ten of the 23 studies reported ethnic/racial background of its participants [24,25,27,32,41,42,52,55,57,64] with two only reporting the proportion of participants from Caucasian background [27,57]. One study recruited only white [32]. Three studies had a diverse ethnic demographic [24,25,52,57,58,68,69] and six studies had mostly white participants [26][27][28]48,57,66]. One study targeted suburban areas for recruitment for their diverse student bodies [24]. Nineteen studies recruited participants from schools, four other studies recruited from within the community [27,52,55,57], of which one study recruited from both within the community and within schools [55]. Four did not report recruitment strategy [29,34,35,40,52]. Please see Table 1b for further information.

Primary and Secondary Outcomes; General and Targeted
Only seven studies reported on primary outcomes in racial/ethnic minority groups. All reported on BMI except one which used BMI z-score [33] as the primary indicator. One study [68] did not report mean difference or effect size. All studies failed to detect a significant difference in the primary outcome of BMI or BMI z-score between the intervention and control group [28,29,33,36,51,61,68]. Mean differences in BMI ranged from −0.0 [61] to −0.14 [29] kg/m 2 . The mean difference in BMI z-score was −0.03 [33] and another study reported β = 0.14 [28] for BMI [28]. One study reported a significant BMI difference of −0.75 kg/m 2 (p = 0.03) for participants in the upper BMI tertile [29]. It is important to note that statistical significance was achieved for other obesity measures such as a decline (−0.25, p = 0.006) in the prevalence of overweight and obesity [33], waist to hip ratio (0.01, p = 0.02) [36] and bicep skin fold thickness for boys (−0.1) and girls (−0.3) [61].
Twenty studies reported secondary outcomes [22][23][24][25][26][27]29,32,42,44,45,47,49,52,53,55,57,62,64,67], with 14 of them having significant changes in either diet, physical activity or sedentary time. None of these studies carried out a subgroup analysis to determine the effect of ethnic/racial background on primary or secondary outcomes. Lastly, of the 23 studies, only two studies reported as to whether ethnic/racial background moderated attrition and both reported that there was no significant effect on the completion of the intervention [52,57]. Please see Table 2b for further information.

Risk of Bias
Using the Cochrane Risk of Bias tool, four of the seven RCTs conducted in ethnic minorities were rated as unclear risk of selection bias for randomization, as they did not specify method for random sequence generation [29,33,51,68]. Two of the seven studies reported allocation concealment for selection bias [28,29]. All seven studies were rated as low risk for attrition bias [28,29,33,36,61,68]. More than half (n = 4) of the studies were scored as unclear risk for detection bias due to insufficient information [29,36,51,68], and two were rated high as the researchers and research assistants were not blinded to group allocation [28,61]. One study was classified as low risk, as research assistants were blinded to group allocation and baseline findings [33]. Half of the studies (n = 3) were classified as having a low risk of reporting bias [28,61,68]. Please see Table S3 for further information.

Grading of Recommendations Assessment, Development and Evaluation Quality Rating
Of the 30 studies identified, seven studies reported findings for ethnic minorities for intervention effects on primary and secondary outcomes [28,29,33,36,51,61,68]. Thus, the GRADE tool was only applied to these studies to address the research question. These included a total number of 2763 adolescents. Please see Table 3 for further information. Table 3. Overall assessment of quality in seven studies (2763 participants in total) in the systematic review of effectiveness of prevention interventions for adolescents from ethnic/racial minorities using the Grading of Recommendations Assessment, Development and Evaluation system.

Rating with Reasoning
Limitations −2 quality due to limitations Consistency No subtraction Directness −1 quality level due to population Precision −1 due to lack of precision Publication −1 quality levels, as publication bias cannot be ruled out Overall Quality Low: effect confidence is limited

Study Limitations
Four of the seven studies did not state method of randomization [29,33,51,68] and only two studies reported allocation concealment method [28,29]. All used intention to treat analysis. Four studies [29,36,51,68] did not state method of blinding of personnel or participants, and two were deemed high risk of bias [28,61]. Blinding was low risk in only one study [33] and only three had a low risk of reporting bias [28,61,68].

Directness
Only seven of the thirty studies formed the evidence base [28,29,33,36,51,61,68] with an additional ten studies [24,25,27,32,41,42,52,55,57,64] capturing participant ethnicity without subgroup analyses of the primary outcome of change in BMI or BMI Z-Score. All interventions directly reported on the outcome of interest as they were all interventions aimed at preventing weight gain.

Publication Bias
Efforts were made in ensuring all papers were captured, including an extensive search through seven major databases, hand searches of references list and contacting authors for additional information.

Discussion
To our knowledge, this is the first systematic review to investigate the efficacy of lifestyle interventions for the prevention of harmful weight gain in adolescents from ethnic and racial minority backgrounds. Analysis of the literature has revealed that despite their heightened vulnerability to overweight and obesity, there remains a dearth of lifestyle interventions that recruit and report on effectiveness in this priority population. Of the 30 studies captured in the search, only six were successful in preventing increases in BMI/BMI z-score [25,34,40,41,49,67]. Of the 30 studies reviewed, seven targeted ethnic minorities exclusively or reported subgroup analyses to determine if ethnic/racial minority status moderated intervention effect on BMI Z-score or BMI [28,29,33,36,51,61,68]. It is acknowledged that some studies may not have included subgroup analyses as sample power calculations did not allow for these comparisons. Furthermore, the small body and quality of evidence limits the interpretation and generalizability of results reported.

Effectiveness of Interventions in Preventing Harmful Weight Gain in Adolescents from Ethnic Minorities: Primary Outcomes
It is now well established that certain groups, such as those from ethnic and racial minority backgrounds, are disproportionately affected by obesity, and as such there has been a call for interventions to be targeted towards specific groups and be differentiated on factors of sex, age and SES [70]. Of the seven studies, no studies were successful in demonstrating a significant difference in BMI between the intervention and control group. Rather, most fell short of the BMI difference needed (0.57 [29]-0.8 [36] kg/m 2 ) to reach the required effect size (e.g., 0.5) [29]. The greatest BMI change across the studies was 0.14 kg/m 2 , less than the 0.15 BMI effect considered to be clinically meaningful by Waters et al. [71]. Two web-based interventions offered the short follow-up periods (6 months from baseline) as a possible explanation for the lack of change in BMI [36,68]; however the small body of evidence limits any conclusions on intervention length to be drawn and the effectiveness of web-based interventions for ethnic and racial minority groups to remain unclear. Another possible explanation for the lack of efficacy is that, despite all four studies [28,29,61,68] including adolescents from ethnic/racial backgrounds, they did not incorporate co-design or consider ethnic minorities in the design of the universal interventions limiting influence and championship of the programs. They also did not employ specific recruitment strategies that targeted ethnic and racial minorities. It should also be considered that of the 23 interventions included in the review that did not consider ethnic/racial minority background, only seven were successful in preventing increases in BMI/BMI z-score [25,34,40,41,45,49,67].
An additional two of the seven studies conducted in ethnic/racial minority groups demonstrated some positive impact. While both studies were conducted in the community, the limited number prevents inferences from being made. The study of Black et al. [33], which targeted African-Americans exclusively, showed a reduced prevalence of overweight and obesity among the target population from 54% to 36% and from 32% to 34% in the intervention and control group respectively (−0.25 (0.09), p = 0.006). The absence of a significant change in BMI z-score was attributed to the inclusion of adolescents across a wide BMI range. Nevertheless, the intervention was still considered effective as it halted an increase in BMI category, with evidence also indicating that a 1% reduction in the prevalence of overweight and obesity in 16-17-year-old adolescents today has been projected to reduce the number of obese adults by 52,821 in the future, decrease lifetime medical costs by $586.3 million, and increase of quality-adjusted life years by 47,138 [41,72]. The intervention of Ebbeling et al. [29] was successful in decreasing BMI in those in the upper baseline-BMI tertile, which is the most important group to target. A subgroup analysis showed that this was consistent for all ethnic/racial subgroups. Similarly, there have been suggestions that while population-based primary prevention interventions should persist to target all children, the study aim and primary outcomes should be evaluated in the highest risk subgroup as opposed to the cohort at large [73].

Effectiveness of Interventions in Preventing Harmful Weight Gain in Adolescents from Ethnic Minorities: Secondary Outcomes
Five of the seven studies [28,29,33,36,39,61,74] demonstrated improvements in diet, physical activity or sedentary time, and this was similar to the ethnic groups. These included decreases in consumption of SSB [28,29,36,61,74] and snack foods, and increases in healthy foods such as fruit and vegetable intake [28,36], more physical activity [36] and reduction of screen time [36].

Setting: School or Community?
Traditionally, schools have been the predominant choice of setting for the delivery of interventions. This is expected, as schools offer ready and continuous access to adolescents as well as resources such as school policies, necessary personnel, curriculum, staff and facilities to promote physical activity and healthy eating [84,85].
Three [28,61,68] of the studies examining ethnic minorities [28,29,33,36,51,61,68] were conducted in the school setting, but none were effective in changing primary outcome of BMI, but changes in SSB consumption were reported [28,61]. Two of the three studies that were conducted in the community demonstrated positive impacts with one demonstrating significant changes in BMI for the most overweight/obese of the group and another reduced the prevalence of overweight and obesity. Community interventions support a family based approach by enabling parents to positively impact diet and physical activity habits [86].
A novel study by Chen et al. [36,74] was the first to explore the feasibility of a culturally specific, family based program delivered online for at-risk adolescents from an ethnic background in primary care clinics. This intervention was able to demonstrate significant changes in BMI and secondary outcomes and have them maintained at follow-up. The combination of a community setting and an online delivery allowed this intervention to leverage the strengths of each mode, to reach and engage this at-risk audience. The community setting allowed the early involvement of key stakeholders, with adolescents influencing the design and implementation of the intervention. The locally engineered nature of this intervention allowed the online and mobile technologies to be capitalised upon, as adolescents could choose from a variety of online learning methods, which is not possible in traditional face to face interventions. The use of mobile and online delivery might have increased equitable accessibility of the program with 95% of adolescents reporting they have access or own a smartphone [87]. A previous systematic review has highlighted the efficacy of this technology in increasing engagement in weight interventions and in decreasing adolescents' dropout rate, which could have mitigated the often-reported difficulty in retaining and recruiting overweight youth in community-based programs [88,89].

Indigenous and First Nations
Of the six interventions carried out in Australia [22,41,45,47,53,64], only two reported ethnicity of participants [41,64] and only one study reported and recruited participants from Aboriginal and Torres Strait Islander background [41]. This study by Hollis et al. [41] was also the only intervention of the six that was able to demonstrate significant changes in BMI. However, no subgroup analyses were conducted to determine if the result was equally efficacious for all subgroups. Only five other interventions recruited and reported on First Nation Populations such as American Native Indians and African Americans [24][25][26]52,55], and similarly, only one study by Melnyk et al. was effective in demonstrating changes in BMI but did not test for differences [25]. The studies by Melnyk et al. [25] and Hollis et al. [41] were both multicomponent interventions underpinned by a theoretical framework, involved parents and employed behaviour change strategies including goal setting, social support, feedback and monitoring, identification and demonstration of the behaviour and used antecedents. These are only two studies of the six studies that impacted on BMI [25,34,40,41,49,67], of the thirty studies included in the review. This small body of evidence and lack of subgroup analysis prevents any meaningful conclusions to be drawn.
The mismatch between the evidence available to inform policy makers on how to intervene in this priority population and the existing disparities prevalent between First Nations and White populations is severe. In 2012-2013, more than a quarter of Indigenous people (37.4%) aged 10-14 and 34.9% of those aged 15-17 were overweight and obese, more likely than their non-Indigenous counterparts [90]. This is alarming considering that obesity was identified to be the second main contributor to the health gap between Indigenous and non-Indigenous people in Australia [91]. Furthermore, the intersection of socio-economic status and ethnicity/race should also be highlighted given that ethnic minorities tend to dominate the lower socio-economic bracket. For example, American Indians and Alaskan Natives have identified to be a racial group with the highest poverty rate in the U.S. [92] and Hispanics to be identified as an ethnic group with a poverty rate of 16% compared to the 8% experienced by the White population [93]. This was evident in selected studies choosing to recruit from low-income and suburban neighbourhoods due to their higher proportion of residents from ethnic and racial backgrounds [24,33,51]. Obesity prevention efforts for Aboriginal communities needs to be nested within the context of their history of colonisation, the major factor contributing towards their poor nutrition and health with the removal of traditional lands (and food sources) and prolonged financial stress associated with food insecurity which among other social inequalities has pressed Aboriginal communities to shift from traditional foods and consume the energy-dense Western diet [94]. Obesity intervention efforts focussed on adolescents from first nations should be at the forefront of prevention intervention priorities and should promote co-design, championship and governance with the First Nation population [95]. Of the thirty interventions, only one study conducted co-design with its participants from ethnic and racial backgrounds [50,51].

Study Strengths and Limitations
The lack of a standardized definition of ethnic/racial background and its indicators makes analysis difficult. Of the 30 studies, only five reported indigenous participants [24][25][26]41,51] and three studies divided race and ethnicity [29,51,69]. Concepts such as 'race' and 'ethnicity' are persisting issues in research [96] due to inconsistent definitions, lack of transparency in methodology employed to measure these concepts, and inappropriate classification of ethnicity and race. Similarly, variation in the use and combination of obesity indicators meant some studies were deemed unsuccessful despite eliciting some significant changes in other outcomes such as skinfold thickness [60,61], abdominal adiposity [62], waist circumference [26,42,43,62], waist hip ratio [62], and an increase in the maintenance and reduction of BMI for age respectively [55]. These limitations have been highlighted previously in a similar review among young adults from ethnic minorities [97].

Review Strengths and Limitations
A strength of the current review is the use of RCTs, which also ensures a higher level of evidence and contacting authors via email, therefore more information was obtained than was originally published. Among the limitations are that only interventions that were published in the English language were included. It is possible some successful interventions for ethnic minorities were not identified. Another strength was that efforts were made in ensuring all papers were captured, including an extensive search through seven major databases, hand searches of references list and contacting authors for additional information to minimise chances of publication bias.

Conclusions
Childhood programs for the prevention of harmful weight gain are important and interventions among adolescents are central to tackling the current obesity epidemic and halting its progression and its co-morbidities into adulthood. This review emphasizes that despite the need to focus efforts on priority populations such as those from ethnic/racial minorities and First Nations populations, to produce a meaningful decrease in the overall prevalence of overweight and obesity, there is an absence of studies. In effect, the review establishes the need for researchers to; actively engage and recruit from priority populations such as ethnic/racial minorities and to consider minority groups in the design and analysis of universal interventions.  Table S1: Search Strategy Medline (Search 1), Table S2: Full-Text Articles excluded and supported reasoning, Table S3: A Risk of bias as assessed by the Cochrane Collaboration Tool 32 for the randomized controlled trials included in the systematic review of effectiveness of preventions interventions for adolescents from ethnic/racial minorities (n = 7), Table S4: Risk of bias as assessed by the Cochrane Collaboration Tool 32 for the randomized controlled trials included in the systematic review of effectiveness of preventions interventions for adolescents from ethnic/racial minorities (n = 23).