Layperson-Led vs Professional-Led Behavioral Interventions for Weight Loss in Pediatric Obesity

This systematic review and meta-analysis investigates the association of professional-led and layperson-led weight loss interventions with short- and long-term weight reduction among children and adolescents with overweight and obesity.


Introduction
Child and adolescent obesity are a global public health concern. 1 Intensive behavioral lifestyle therapy is considered the cornerstone for treatment of obesity in this age group. 1,2 Systematic reviews of the literature show that intensive behavioral lifestyle interventions elicit modest shortterm weight loss 3,4 and improved cardiometabolic health 4,5 among children and adolescents with overweight and obesity. While efficacious, these approaches are costly and often impractical in realworld settings. Less intensive interventions delivered in community settings are less costly but often yield less significant weight loss. [6][7][8] Few studies have directly compared the association between short-term and sustained interventions on weight management in children and adolescents with overweight and obesity. [9][10][11] Engaging laypersons or community-based health workers to deliver health interventions is an attractive public health model for disease management, as it is cost-effective and can be tailored to local needs. 12,13 In some settings, community health workers or peer leaders yield meaningful improvements in lifestyle behaviors and health outcomes among persons living with chronic disease. 14,15 A series of recent trials suggested that behavioral interventions led by nonprofessionals yield similar results to those led by trained professionals. 7,8 In the context of pediatric obesity, a limited number of trials suggest that peer-or layperson-led approaches may be associated with the achievement of successful weight loss. [6][7][8] To the best of our knowledge, this has yet to be investigated by a systematic literature review with meta-analysis.
Network meta-analysis allows for the comparison of multiple treatments in 1 statistical model. 16 Network meta-analyses can assess treatment outcomes or safety when few direct head-to-head trials exist. 17,18 With the abundance of therapeutic trials for weight loss among children and adolescents with overweight or obesity, 2 a network meta-analysis is an attractive model for comparing the associations of layperson-and professional-led approaches with weight loss.
Accordingly, we conducted a systematic review and network meta-analysis to assess the association of behavioral interventions led by lay individuals vs those led by professionals, compared with the standard of care, with short-and long-term weight loss among children and adolescents younger than 18 years with overweight or obesity.

Primary and Secondary Outcomes and Subgroup Analyses
The primary outcomes were the change from baseline in weight and any measure of BMI (BMI z score and BMI percentile) at the end of the intervention period. As outcomes for BMI were not consistently reported, we used the J correction factor for an unbiased estimate of the standardized mean difference (SMD). 21 We also assessed change in percent body fat, waist circumference, and overall study withdrawals as secondary outcomes. For RCTs that reported long-term follow-up data after the end of the intervention, we also examined changes in these outcomes to assess the sustainability of the interventions. The change in the outcome variables was calculated as the difference between baseline and immediate postintervention measurements to calculate the short-term weight loss. The difference from baseline to the last follow-up after the intervention was completed was used to calculate the sustainability of the intervention.

Risk of Bias Assessment
We evaluated the internal validity of included RCTs using the Cochrane risk of bias tool. 22 This tool consists of 6 domains (sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and other sources of bias) and a categorization of the overall risk of bias. Each separate domain was judged as low, unsure, or high risk of bias. The overall assessment was based on the responses to individual domains. If 1 or more individual domains were assessed as having a high risk of bias, the trial was judged as having a high risk of bias. The overall risk of bias was considered low only if no domain was rated as having an either high or unclear risk of bias. The source of funding was also extracted.

Statistical Analysis
To rank the intervention types for relative effectiveness and to compare every intervention to each other using all available evidence, even when no studies contributed data directly, we used network meta-analyses (also termed as multiple, or mixed, treatment comparisons). We used a bayesian framework and Markov chain Monte Carlo simulation methods to combine direct and indirect evidence implemented in WinBUGS software, version 1.4.3 (University of Cambridge). 23 We fit both random-effects and fixed-effect network meta-analysis models 24 (code provided in eTable 2 of the Supplement). The preferred model was chosen by comparing the deviance information criteria 25 (eTable 2 in the Supplement). For all analyses, we assessed model convergence using the Brooks-Gelman-Rubin diagnostic tool, 26 history plots, autocorrelation, the form of the posterior density for the between-study heterogeneity, and the basic parameters (eFigures 1-4 in the Supplement). We used vague prior distributions for all parameters, a burn-in period of 50 000 iterations, a sampling period of 100 000 iterations, and 3 chains with varied initial values in all analyses (eTable 2 in the Supplement). The goodness of fit model was measured by the posterior mean of the residual deviance; in a well-fitting model, the residual deviance should be close to the number of data points included in the analysis. 27 Where possible, we evaluated the consistency between the direct and indirect evidence by calculating a bayesian 2-sided P value for the difference between the direct and indirect estimates using the Bucher method, 28 where the direct estimates were obtained from the inconsistency model (eTable 2 in the Supplement). 29-31 P < .05 was considered significant.

Intervention Details
Summary information for professional-and layperson-led interventions is provided in Table

Primary Outcomes
Data from each randomized trial on primary outcomes are presented in eFigure 2 in the Supplement.
Random-effects models yielded better deviance information criteria than fixed-effects models for

Secondary Outcomes
Results for associations between layperson-and professional-led behavioral interventions and secondary outcomes are presented in Table 2. Professional-led interventions were associated with significant reductions in percent body fat (MD, -1.70%; 95% CI, -2.60% to -0.81%; P < .001) and   interventions were seen compared with standard care (Table 2). There were insufficient data to analyze long-term associations of secondary outcomes. No differences were observed in either professional-or layperson-led interventions for study withdrawals.

Treatment Rankings
Treatment rankings for the 3 intervention types for both primary outcomes, immediately following the intervention and during long-term follow-up, are presented in Figure 3. interventions were ranked nearly equal for their association with achieving absolute and relative weight loss (Figure 3).

Discussion
The main finding from this systematic review and network meta-analysis was that professional-led behavioral weight loss interventions were associated with significant short-term, but not long-term, obesity. 4 This review tested for differences in treatment outcomes across studies that had different contact time with participants, but did not directly compare different intervention models. Among the 42 behavioral therapeutic trials included in the analysis, several were not randomized, cluster trials were included, and no long-term follow-up data were provided. Despite these differences between the network meta-analysis presented here and the analysis from the US Preventive Services Task Force, the effect size for professional-led interventions was similar and comparable to previous systematic reviews of weight loss interventions among children and adolescents with overweight or obesity. 3,116,117 The data presented here extend previous systematic reviews by demonstrating that  the short-term benefits of professional-led weight loss interventions are not sustained following the end of the intervention and that similar effect sizes may be achieved with layperson-led interventions. Adherence to lifestyle change is a critical determinant of intervention effectiveness.

JAMA Network Open | Pediatrics
Unfortunately, very few trials reported adherence to prescribed lifestyle change, and therefore it is unclear whether the lack of maintenance and the relatively modest weight loss following these interventions are associated with low adherence to intervention attributes. Collectively, these data provide robust evidence that professional-led behavioral interventions are associated with achieving modest weight loss among children and adolescents with overweight or obesity; however, this association was not sustained in the long term.
Systematic reviews of home-, 119 school-, 118 and community-based 120 behavioral interventions for obesity prevention in children and adolescents suggest that these nonprofessional-led interventions yield minimal or no weight change. In contrast to previous systematic reviews, we excluded cluster randomized trials, quasi-experimental trials, trials lasting less than 12 weeks, and trials that included children and adolescents of a healthy weight. For the current review, layperson-led interventions were delivered either by parents or older peers without formal training in a health profession. In contrast to results from quasiexperimental 6,7,121   layperson-led interventions, however, lacked precision. Larger trials of layperson-led interventions, particularly trials directly comparing layperson-and professional-led interventions, are needed to understand the association of this approach for weight management among children and adolescents with overweight or obesity.
There is some evidence that layperson-or peer-led approaches support positive behavioral change and improved health outcomes among adults living with obesity [122][123][124][125] or obesity-related comorbidities. 126 Layperson-or community-led interventions have proved to be associated with low-resource areas or settings in which culturally tailored approaches are preferred by community members. 12,13 The meta-analysis conducted here found that layperson-led behavioral trials were not associated with statistically significant reductions in body weight among children and adolescents living with obesity. Weight status is only 1 of multiple measures of health that can be influenced by behavioral change in children and adolescents with overweight or obesity, particularly cardiometabolic risk factors. Children and adolescents with overweight or obesity also are more likely to live in families with low income, 127,128 to have been exposed to adverse childhood experiences, and to suffer from mental health comorbidities. We did not include these outcomes in our analysis; however, it is possible that these outcomes could be responsive to layperson-led interventions. The promising association of layperson-led approaches in other settings and populations 129,130 reinforces the need for large-scale, well-designed, multiarm RCTs to determine the effectiveness of interventions led by lay individuals for supporting weight change among children and adolescents with overweight or obesity.
The advantage of conducting a network meta-analysis, relative to a conventional meta-analysis, lies in the capacity to estimate the relative efficacy of 2 given interventions when few or no direct head-to-head trials exist. We were only able to identify 3 to 5 trials that directly compared laypersonto professional-led interventions for weight loss in children and adolescents with overweight or obesity. Performing a meta-analysis on the results of these RCTs did not reveal an association with either intervention. The few RCTs directly comparing layperson-and professional-led interventions were relatively low-powered and were considered to have a high risk of bias. Based on the limited available evidence, professional-led approaches were ranked as being associated with short-term weight loss; however, over the long-term, layperson-and professional-led interventions appeared to perform equally. Adequately powered, head-to-head trials of layperson-and professional-led approaches with long-term follow-up are needed to confirm these observations.

Strengths and Limitations
The study is strengthened by limiting the analyses to trials focused exclusively on children and adolescents with overweight or obesity, an a priori published protocol, and the relatively large number of RCTs available for the network meta-analysis. Despite these strengths, there are limitations to consider. The strict criteria we imposed on the search limited the inclusion of designs, including cluster RCTs and quasi experiments, which limits the generalizability of our findings.
Additionally, differences in intervention designs could have influenced the point estimates between professional-and layperson-led trials; however, with only 5 trials led by laypersons, we were underpowered to adjust for these differences. We also recognize that age-and sex-standardized measures of adiposity are the best practice for reporting weight-related outcomes in children. As we relied on published outcome data and not individual-level data, we were largely unable to use BMI or waist circumference z scores. Additionally, we only searched for trials appearing in the last 20 years in an effort to limit the number of low-quality RCTs. We did not include non-English publications or RCTs that were unpublished in order to increase feasibility and the homogeneity between weight loss interventions; this may have introduced selective reporting bias (eg, publication bias). Furthermore, only 25% of the included trials were judged as having a low risk of bias. As these were behavioral trials, blinding was not possible; however, as the outcomes are semiobjective, blinding may not be as effective as in pharmaceutical trials with subjective outcomes. 131 As mentioned previously, this review was restricted to weight-related outcomes and did not include