Social disparities in obesity treatment for children age 3–10 years: A systematic review

Socio‐economic status and ethnic background are recognized as predictors of risk for the development of obesity in childhood. The present review assesses the effectiveness of treatment for children according to their socio‐economic and ethnic background. Sixty‐four systematic reviews were included, from which there was difficulty reaching general conclusions on the approaches to treatment suitable for different social subgroups. Eighty‐one primary studies cited in the systematic reviews met the inclusion criteria, of which five directly addressed differential effectiveness of treatment in relation to social disparities, with inconsistent conclusions. From a weak evidence base, it appears that treatment effectiveness may be affected by family‐level factors including attitudes to overweight, understanding of the causes of weight gain and motivation to make and maintain family‐level changes in health behaviours. Interventions should be culturally and socially sensitive, avoid stigma, encourage motivation, recognize barriers and reinforce opportunities and be achievable within the family's time and financial resources. However, the evidence base is remarkably limited, given the significance of social and economic disparities as risk factors. Research funding agencies need to ensure that a focus on social disparities in paediatric obesity treatment is a high priority for future research.


Funding information
European Union's Horizon 2020 research and innovation programme, Grant/Award Number: 774548 Summary Socio-economic status and ethnic background are recognized as predictors of risk for the development of obesity in childhood. The present review assesses the effectiveness of treatment for children according to their socio-economic and ethnic background. Sixty-four systematic reviews were included, from which there was difficulty reaching general conclusions on the approaches to treatment suitable for different social subgroups. Eighty-one primary studies cited in the systematic reviews met the inclusion criteria, of which five directly addressed differential effectiveness of treatment in relation to social disparities, with inconsistent conclusions. From a weak evidence base, it appears that treatment effectiveness may be affected by family-level factors including attitudes to overweight, understanding of the causes of weight gain and motivation to make and maintain family-level changes in health behaviours. Interventions should be culturally and socially sensitive, avoid stigma, encourage motivation, recognize barriers and reinforce opportunities and be achievable within the family's time and financial resources. However, the evidence base is remarkably limited, given the significance of social and economic disparities as risk factors. Research funding agencies need to ensure that a focus on social disparities in paediatric obesity treatment is a high priority for future research. In 2016, 60 million children aged between 5 and 10 years was living with obesity worldwide, equivalent to one child in every eight in this age group. 1 Childhood obesity has long-term detrimental effects on individual health and has wider social and economic consequences: it is directly linked with endocrine and orthopaedic complications and early onset of cardiovascular disease and type-2 diabetes and affects children's psychosocial well-being by reducing self-esteem, quality of life and increasing social stigmatization. 2,3 The prevalence of obesity is not spread uniformly across child populations. Variability is associated with parental weight status, maternal smoking, infant feeding patterns and, of particular interest in the present study, ethnicity and socio-economic status. 2,[4][5][6][7] In high-income countries, evidence from epidemiological studies have shown that obesity levels are higher in children of the lowest socio-economic status, whereas in lower income countries, overweight tends to be more prevalent in urban and higher income households. 2,6 Evidence also suggests that ethnicity is an independent risk factor, with children in southern Asian, Afro-Caribbean and Hispanic families tending to show higher overweight prevalence levels than those in far-Eastern and White Caucasian families. 2 In order to reduce the prevalence of childhood overweight and obesity, two approaches are needed: (i) reducing the incidence of new cases through prevention and (ii) reducing the number of existing cases through treatment and weight management services. In this review, we will examine the latter approach, with a focus on paediatric services for younger children. This is an area in which a substantial amount of research has been undertaken, and the results examined in many systematic reviews in the last decade.
Although surgical and pharmaceutical interventions are rarely considered in preadolescent children, interventions using diet and physical activity are commonly undertaken, but the results show only small average intervention effects on sustained improvements in adiposity. 8,9 Despite the limited effects, these trials have helped to identify features that are associated with a better likelihood of success, including a focus on younger children, a multidisciplinary approach, intensive delivery, parental or family involvement and a focus on school or group settings. 8,10 Rarely mentioned, however, are the barriers to successful treatment that may be associated with social disparities. 4 The purpose of the present review is to focus on treatment interventions in health care settings for younger children experiencing overweight or obesity, with a specific focus on the evidence for differential effectiveness of interventions to treat paediatric obesity in relation to socio-economic and ethnic disparities, and to examine evidence on the challenging phases of the interventions such as recruitment, adherence and follow-up in relation to these disparities. The review was registered with the PROSPERO International Prospective Register of Systematic Reviews (CRD42019128687) with additional searches undertaken, as described here.

| METHODS
This paper focuses on social disparities (defined here as disparities linked to ethnicity, migrant status, educational status, household income, health insurance status or other related socio-economic measure such as area deprivation index) in relation to paediatric obesity treatment and outcome, as provided through health care services to younger children (defined here as children aged between 3 and 10 years).
The search for evidence was undertaken in two stages: an examination of systematic reviews and an examination of primary studies of paediatric obesity treatment. The two stages were found to be necessary when it became clear in pilot searches that the systematic reviews did not provide sufficient evidence on social disparities in paediatric obesity treatment.

| Stage 1
In the first stage, we undertook a systematic search for evidence on social disparities contained within systematic reviews of paediatric obesity treatment published in the last decade (2009 onwards). Papers were included if they provided evidence on younger children (age 3.0-9.9 years) being treated for overweight or obesity. For each systematic review, we examined the Methods, Results, Discussion and Conclusion sections in order to identify evidence relating to social disparities in outcomes or in the recruitment and retention of participants. Relevant information was extracted to provide a narrative review.

| Stage 2
In the second stage, we examined all the primary studies of paediatric treatment that had been accepted for inclusion in the systematic reviews identified in the first stage. The primary studies were included according to the population, intervention, comparison and outcome (PICO) criteria shown in Table 1, which specifies age (children aged between 3.0 and 9.9 years), treatment for excess bodyweight provided through health care services to children, assessed in a controlled trial with at least 6 months of follow-up. Outcome variables included weight-related measures and treatment process indicators.
Social status variables followed a qualified PROGRESS-Plus recommendations, 11 (for exclusions, see Table 1). Data were extracted from these studies according to a template designed to capture salient information on social disparities, intervention procedures and treatment outcomes (see supporting information, section 3).
Following concern that additional papers may have been missed under the search strategy outlined in Stage 2, we undertook a rapid review for recent primary studies using Medline, restricted to studies published 1/1/2018 through 1/7/2019. The search terms and results are shown in the supporting information (section 2.2).

| Search methods
In Stage 1, searches were undertaken in Medline, Cochrane Database and Embase (Ovid) for systematic reviews focusing on socio-economic aspects of paediatric obesity treatment. Search terms are shown in the supporting information (section 2), and in brief form were (Child+ OR Pediatric) AND (Overweight OR Obes+) AND (Treatment or Management), limited to systematic reviews and meta-analyses, and published between 1/1/2009 and the date of the search, 24/6/2019. From the identified publications, further potential reviews were sought by examining the references cited. In addition, a Google Scholar search (first 100 returns) was undertaken to identify additional reviews. Text in each of the systematic reviews was examined and relevant sections extracted by one researcher and subsequently verified independently by a second researcher. Differences were resolved by discussion. The quality of the reviews was assessed using the AMSTAR2 rating scheme 12 and reported in Table 2 below.
In Stage 2, all primary studies of paediatric obesity treatment, which had been cited in the systematic reviews examined in Stage 1, were considered as eligible for further analysis. These primary studies were assessed according to the PICO eligibility criteria described in Table 1 and the included studies processed for data extraction. Data from primary studies were extracted independently by two researchers using a standard data template (see supporting information, section 3). The completed templates for each study were then compared and differences resolved by discussion with a third researcher. Where the individual studies provided stratified results based on social disparities, a GRADE rating system * was used as an evaluation tool and reported in Table 4 below.

| FINDINGS
The numbers of papers identified in each of the stages of the present review are shown in the PRISMA chart below. This shows the identification of 64 systematic reviews included in the present study and the identification of 82 primary studies of paediatric obesity treatment, which conform to the PICO inclusion criteria.  Table 2 shows the narrative text extracted from these three systematic reviews. It can be seen that the quantity of information is remarkably limited and the level of detail is poor. The interpretation provided by the authors in their narrative text needs to be taken in the context of the critical appraisal shown in the third column, based on AMSTAR2 criteria, where it can be seen that the applicability of the authors' comments to the population of interest (children under age 10 years, treated for obesity through paediatric services) is limited. As the review by Ligthart et al 7 noted, most studies had small sample sizes, and therefore, the opportunity to examine the effects of interventions on subgroups defined by social disparities was very limited.

| Results from systematic reviews
The paucity of results from these three reviews led the authors to examine the remaining 61 systematic reviews addressing paediatric treatment identified in the literature search. For each review, the T A B L E 2 Summary statements from three systematic reviews identified in Stage 1

Review Key statements in the review's text Comments and AMSTAR2 quality concerns
Brown et al 5 Abstract: "There was no evidence that interventions were more or less effective according to whether the intervention was set in South Asia or not, or by socio-economic status." Conclusions: "One high quality RCT in South Asian children found that a school-based physical activity intervention that was delivered within the normal school day which was culturally sensitive, was effective. There is also evidence of culturally appropriate approaches to, and characteristics of, effective interventions in adults which we believe could be transferred and used to develop effective interventions in children." No PICO shown. Duplicate data extraction was not stated. Risk of bias and publication bias was not mentioned in the Discussion. Included only three RCT studies of children. Results for south Asians were not compared with non-south Asians. Review included adults, and included preventive interventions. Of seven studies, none complied with the present reviews' PICO criteria. AMSTAR2: LOW Hillier-Brown et al 6 Abstract: "At the individual level (n = 4), there was indicative evidence that screen time reduction and mentoring health promotion interventions could be effective in reducing inequalities in obesity. … The review has found only limited evidence although some individual and community based interventions may be effective in reducing socio-economic inequalities in obesity-related outcomes amongst children but further research is required, particularly of more complex, societal level interventions and amongst adolescents." Discussion: "Treatment interventions are more likely to show positive effects than prevention ones.
[A] targeted approach … has limitations as even when interventions are effective amongst low income groups they are only able to reduce the health inequalities gap, they have little effect on the wider social gradient." No PICO shown. The quality of studies was assessed but not reported. Risk of bias and publication bias were not mentioned in the discussion. The review included preventive and treatment interventions. Age range 6-12 years old. Race/ethnicity was not examined. Of 23 studies, 2 complied with present reviews' PICO criteria. AMSTAR2: LOW Ligthart et al 7 Discussion: "We found that Black ethnicity seems to be associated with higher intervention dropout and that low family income appears to be associated with lower compliance with the intervention. … The associations between other ethnicities (such as White and Hispanic and White and other ethnic minorities) and SES categories and intervention or study dropout and non-compliance were mainly non-significant. … In the literature, ethnicity and SES are considered to be related: ethnic minorities often have a lower SES than Whites … This relationship was reflected in our study results; outcomes for ethnicity and SES pointed in the same direction. Studies that reported on both ethnicity and SES found corresponding associations with study and intervention dropout and non-compliance. …" No PICO shown. The review included adolescents up to age 20 years. Some interventions included non-obese children. Publication bias was not mentioned in the Discussion. Of 30 studies, 6 complied with the present reviews' PICO criteria.  Table 3. Several reviews noted that many primary studies involve families with higher income and higher levels of general functioning, with resources to make changes to their health behaviour, and with parenting skills and capacity to ensure good family involvement in the treatment programme. Studies of subgroups, such as Latino or Mexican populations are inconclusive, and do not demonstrate whether any specific treatment requirements were advantageous. Overall, there is considerable difficulty reaching general conclusions on the forms and approaches to paediatric obesity treatment suitable for different social subgroups within a general population.

| Results from primary studies
The systematic reviews were not able to answer the research questions with a high level of confidence. We therefore examined the 1,699 primary studies cited in the systematic reviews, and from these identified 81 which fulfilled the PICO criteria in Table 1 for data extraction (see Figure 1B). These 81 studies are listed in the supporting information, with the relevant information from each of them summarized from their data extraction templates.

| Differential outcomes
Of the 81 studies identified, 37 did not mention social disparities in the published reports. The remaining 44 studies stated that some social disparity measure had been taken at baseline but 39 of these 44 studies did not describe body-weight-related outcomes in relation to the socio-economic disparity measures taken. The remaining five studies had undertaken some quantitative analysis of treatment outcomes in relation to one or another measure of social disparity, and a summary is given in Table 4.
Of these five studies, one 17 found no significant differential outcome between social groups. Two studies 1315, 16  Bond et al 20,21 Of the three studies included in this pair of reviews, one, the Hip-Hop Jr study, "… took great care to be culturally sensitive to the minority groups it was working with. The Hip-Hop Jr authors identified several components from their pilot work that were important in engaging these families: easy and safe access to the programme; being situated in the preschool that the children were already attending; having the parental element take place in the home; encouraging identification between those delivering the intervention and participants; addressing cognitive and environmental barriers to exercise and dietary change; emphasis on modelling lifestyle change; and consideration of all levels of literacy" Colquitt et al 9 "Five of the seven trials reported ethnicity. … Five trials reported socioeconomic status using different indicators …. No trials investigated all-cause mortality, morbidity, or socioeconomic effects." Eisenberg et al 22 (Review focused on interventions targeting Latino population groups, suitable for application in Mexico.) "… it is recognized that parents and the home environment can influence children's dietary and physical activity behaviors. As such, parental components should be highly considered in designing obesity interventions." Ells et al 10 Concern about self-selection for treatment "… whether the study population … may have attracted a subset of the community amenable to the availability of free treatment." Foster et al 23 One study 18 found no change in BMI at 1 year compared with controls but "a post hoc analysis showed significant effects on BMI in female subjects … and those in households with incomes less than $50,000". The Taveras study is reported in Table 4, below.
Kitzmann and The practicalities of delivering effective advice on lifestyle changes to obese children and adolescents will vary with the wide span of social, ethnic and economic circumstances, as well as with the many variations in available resources for local health service delivery. … the majority of research in the field has been conducted in motivated, middle class, Caucasian populations" Park et al 30 "The results of this review must be interpreted with caution: the studies were short-term and based on small samples; participants were mainly from the U.S., and large portions were from ethnic backgrounds known to be at increased risk of metabolic disorders, limiting the generalizability of findings; and the studies presented unadjusted measures without any intention-to-treat analyses, which may have overestimated treatment effects." Staniford et al 31 "A large number of studies did not identify the ethnicity (49.2%) or the socio-economic status (67.2%) of the participants and in studies that identified these demographics, samples with a majority of white participants (36.1%), from middle to upper class backgrounds (21.3%), were the most common." "Limited research has addressed recommendations to actively recruit and tailor treatment interventions to ethnically diverse and immigrant populations … When reported, studies generally involved white, middle/upper class samples. Future research targeting diverse populations, specifically groups with the highest prevalence of obesity are still required to avoid taking a 'one size fits all' approach." Viner et al 32 Results section notes that "subjects were predominantly white or Hispanic" but this is not referred to in the Discussion.
the intervention countered a significant rise in BMI experienced by lower socio-economic status children over the period. In the Epstein et al study, 14

| Recruitment, adherence and follow-up
From both the systematic reviews and the primary studies, we extracted statements referring to recruitment of participants, adherence to treatment, drop-out from treatment and availability for follow-up, in relation to the social disparities of interest in this study.
A total of 15 documents contained relevant material. Table 5 provides a brief summary of the text and quantitative data found in the 15 documents. Loss to recruitment or to treatment due to the reasons stated by participants such as "no time", "no transport" or similar were disregarded unless these were linked to the subjects' social disparity status.
Few general conclusions can be made from these extracted texts.
Participation in paediatric treatment, and especially in controlled trials of paediatric interventions, requires a degree of commitment, family resources and capacity, and motivation from the family and the child.  Davis et al 34 " The clinical implications of this study are many. First, for rural families facing the issue of pediatric obesity, telemedicine or other methods of interactive televideo seem to be feasible for the delivery of empirically supported interventions. Families from rural areas who commit to this type of intervention are likely to show up for treatment and to encounter few technical difficulties." Jang et al 35 "Although none of the studies we reviewed discussed the reason for high attrition, prior research has found that high attrition was associated with low socio-economic status, the single-parent family, and ethnic minorities … Further research is indicated to develop methods to ameliorate these discrepancies, particularly since studies included in this review did not reach families of diverse race/ethnicity or low socioeconomic status. … Understanding family dynamics within a family system and how this relates to intervention program participation is also important to address in order to eliminate obstacles. In addition, family and social support as well as culturally relevant intervention programs should be considered in future research as a means to enhance program participation and effectiveness." Kelishadi et al 36 "Participants were selected … to avoid socioeconomic bias." Kirk et al 37 "Children were recruited from referrals to a pediatric weight management programme at Cincinnati Children's Hospital Medical Center (CCHMC) who lacked health insurance coverage for the CCHMC program."

Kitzmann and Beech 24
"It is important to note that families who have participated in research on family-based interventions for pediatric obesity are likely to be relatively high functioning. These families must show a certain level of organization and cohesion to successfully initiate participation in an intervention program and to complete the program over the course of many weeks. In this sense, current research on family-based interventions for pediatric obesity could be considered a form of efficacy research in that the treatments are being implemented with families who are relatively well positioned to take advantage of the program. Tests of these interventions in a wider range of families would thus constitute a form of research on effectiveness rather than efficacy. We believe that a more general family focus may be a helpful framework for modifying these programs so that they also may be implemented with a wider range of families. Some familiessuch as those characterized by destructive conflict or poor parenting skills, or those experiencing multiple stressors associated with socioeconomic disadvantagemay need more basic support and preparation in order for treatment to be effective. For these families, intervention programs may need to include a greater emphasis on conflict resolution, basic parenting skills, and stress reduction so that parents are in a better position to influence their children's eating and exercise. As such, we are arguing for a more ecological approach to treatment, one that focuses not just on the immediate context of parent-child interactions but also on the larger social context of the family and community. This ecological perspective has been shown to be useful in targeting behavior problems in high-risk youth … and is becoming increasingly common as a perspective for understanding and treating children's behaviors related to physical health." Lochrie et al 38 "Compared with those who completed the study, those who did not complete the study had significantly lower SES, were less likely to be living with both biological parents, and caregivers were less likely to be married." Nagle et al 29 (Review focused on interventions targeting Latino population groups.) "The healthcare setting facilitates interaction with health professionals who are knowledgeable about the health effects of obesity. … this setting would not be ideal for populations and communities that do not have regular access to clinics and/or do not seek out healthcare on a regular basis." Resnicow et al 39 "We lost 30% of the baseline sample. Although this was the anticipated range of attrition and consistent with previous studies, the fact that those lost to follow-up differed on several demographic variables (e.g. race, income and education) limits generalizability. … those lost to follow-up were significantly more likely to be black or Hispanic patients and to come from households with <$40,000 income and lower parental education. There were also more likely to have Medicaid." Taveras 18 "Although we attempted to match pediatric sites to obtain similar participant characteristics in intervention and usual care, unbalanced participant characteristics at baseline occurred. This imbalance may have also affected differences in parent obesity and household income." Taylor et al 40 "Multivariate regression predicting intervention uptake showed pacific ethnicity and university degree influenced uptake-see table II. Socioeconomic status differed in intervention participants (n = 197) 4.9(2.8) vs non-participants (n = 74), 5.4 (2.9). Information on the socioeconomic status of their place of residence using the New Zealand Index of Deprivation (ranges from 1-least deprived to 10-most deprived). Few differences in demographic variables were observed between intervention participants and non-participants with age, sex, ethnicity, maternal BMI, or household structure differing little by intervention uptake (Table III). However, non-participants were more likely to be from homes in more deprived areas (P = 0.039) and participant mothers also tended to be more highly educated (P = 0.051, Table III Walker et al 43 "Children with private insurance appeared to have a benefit in that they were less likely to drop out compared to children with public insurance." West et al 44 "Although the sociodemographic characteristics of the sample were typical for the Australia general population, participants were mainly white, well-educated for parents with moderate levels of employment and income. The sample included some sole-parent and low-income families, and some children of mixed ethnicity; however, further research is needed to clarify whether similar findings would be obtained with higher-risk families (e.g. families experiencing poverty, minority families or parents from non-English speaking background." need more basic support and preparation in order for treatment to be effective. For these families, intervention programs may need to include a greater emphasis on conflict resolution, basic parenting skills, and stress reduction" (p58). 24

| LIMITATIONS
In the present review, we limited our search for primary studies to those which had been cited in the initial 64 identified systematic reviews. This identified 81 primary studies of which only five provided data on differential outcomes according to social disparities. American, 36% Hispanic). The authors did not describe BMI-relevant outcomes in relation to the social disparity measures taken, but they noted that the intervention was designed to be applicable to a "low income and diverse population", by being flexible and relatively unstructured, with adaptable enrolment and attendance schedules: "This flexibility is a strength in terms of inclusivity, but the lack of structure and accountability is also a limitation" (p8).
A second limitation is the narrow range of countries from which evidence is available: the large majority of primary studies were conducted in North America and Europe and only one study in a non-OECD economy (Brazil).

| DISCUSSION
The objective of this review was to assess the evidence of differential effectiveness of interventions undertaken through health services to treat paediatric obesity with a particular focus on social disparities, and the potential impact of social disparity during the challenging phases of the interventions such as recruitment, adherence and follow-up. This review was conceived on the premise that it would be a "review of reviews" looking specifically at the influence of social and economic variables on treatment effectiveness, as defined in current systematic reviews of the issue. However, an initial scoping exercise raised concerns that insufficient evidence might be available, and a two-stage process was designed. on treatment processes, such as differential recruitment and adherence issues. A follow-up database search found one additional paper 45 that met the inclusion criteria and contained some evidence on optimal intervention design.
From the material examined in the present review, we make a number of observations.

| Treatment outcomes
• There is a remarkable lack of high-quality evidence concerning the influence of social disparities on the effectiveness of paediatric obesity treatment, and on recruitment, drop-out and follow-up phases of interventions.
• Where base-line data on social disparities are collected in treatment trials, they are heterogeneous in nature and may include ethnicity or racial descriptors, household income, parents' education, a composite index of deprivation used in one country only or an indirect indicator such as health insurance status. We found no evidence of data collected for migrant status for the younger children included in this review.
• Where baseline data are collected and reported, there is often no further analysis, with neither the processes nor the outcomes differentiated by social subgroup.
• When reported, the most common ethnic subgroup is Caucasian/White, followed by African-American or Black, and Hispanic or Latino. These categories reflect the dominance of treatment studies undertaken in the United States.

| Research implications
There is a clear and continuing high level of policy concern over health inequities and universal health coverage at global, national and community levels. Action to mitigate disparities needs evidence, yet this need for evidence is not being addressed.
• Many intervention studies, paid for with public funds or philanthropic grants, appear not to be collecting the relevant information on social disparities, or collecting it in inconsistent forms, and then not analysing or reporting on the processes and outcomes in relation to these disparities. We urge academics, clinicians and funding bodies to make socio-economic disparities a priority for research trials.
• In studies where the relevant social status information has been collected at baseline, but not subsequently used to analyse differential responses, reanalyses could be considered to exploit the data already available.
• Steps may be taken to increase the collection of data from uncontrolled observational studies as additional sources of valid evidence. In addition, steps can be taken to encourage academics and service providers to work with the populations known to suffer disadvantages, including higher obesity prevalence levels, to develop new studies and participant-led interventions.

| CONCLUSION
There is an extraordinary lack of information on social and economic influences on trials of paediatric obesity treatment administered through health services. This is despite the well-recognized evidence of disparities in obesity prevalence, which shows that among most middle-and high-income countries, there is a greater prevalence of obesity among families with lower incomes or parental education and in specific ethnic groups. The causes of these disparities are likely to have major relevance for the success or failure of paediatric treatment, yet such disparities are rarely examined in treatment studies and, as a consequence, not featuring in systematic reviews.
The lack of high-quality information on differential treatment impact among socially disparate groups is likely to be hampering the development of good practices and coherent national guidance on paediatric obesity treatment for those most in need. Use of weight management and obesity treatment services is likely to be affected by familial attitudes to overweight in children, their understanding of the underlying causes of weight gain, their motivation to make familylevel changes and, above all, the resources they may have available to make and maintain these changes.