Spillover effects of childhood obesity prevention interventions: A systematic review

Childhood obesity prevention initiatives are complex interventions that aim to improve children's obesity‐related behaviors and provide health promoting environments. These interventions often impact individuals, communities, and outcomes not primarily targeted by the intervention or policy. To accurately capture the effectiveness and cost‐effectiveness of childhood obesity prevention interventions, an understanding of the broader impacts (or spillover effects) is required. This systematic review aims to assess the spillover effects of childhood obesity prevention interventions.


| INTRODUCTION
Childhood overweight and obesity is a significant public health concern worldwide, with serious health, economic, and social consequences. 1A large international evidence base for strategies to prevent childhood obesity has accumulated over recent years, 2 with obesity prevention interventions commonly aiming to improve children's dietary, physical activity, and other obesity-related behaviors and environments, within specific settings such as schools, home, or communities.A recent Cochrane review suggests that obesity prevention interventions that include strategies for changing diet and/or activity levels may result in modest reductions in body mass index (BMI) in children aged up to 12 years, with less conclusive evidence for adolescents and young adults. 2Many childhood obesity prevention interventions seek to address multiple determinants of childhood obesity across individual, interpersonal, institutional, community, and/or policy levels (reflecting the five levels of the Socio-Ecological Model [SEM]). 3 While these determinants are targeted to influence childhood overweight and obesity, they also influence other health and nonhealth outcomes. 4Obesity prevention interventions are complex interventions, and it is recognized that complex interventions are likely to have a range of health and nonhealth sector impacts. 5[8] Four types of spillover effects have been defined in the literature: (i) effects in unexposed populations (i.e., externalities); (ii) effects in nontarget populations via social, physical, or economic proximity; (iii) context equilibrium effects, where behavioral or social norms are affected as a result of intervention; and (iv) general equilibrium effects, where intervention affects supply and demand. 7,9For example, an effect in an unexposed population may be those benefitting from subsidised healthy food initially intended for a narrow target group; nontargeted family members may also change behaviors in response to efforts to improve a child's dietary patterns; a school intervention targeting reducing sugar sweetened beverages may lead to a broader community impact as behavioral or social norms are influenced via role modeling of changes in food preferences; behavioral changes as a result of intervention may lead to increased demand for physical activity facilities and impacts on local community.Economic spillovers may also occur as a result of the availability of additional resources due to an intervention. 6For instance, when one household member receives additional resources to engage in healthy behaviors as a result of intervention, this may spillover to other household members due to the reallocation of existing resources. 6illover effects of childhood obesity prevention interventions may occur at the individual level (such as among parents, caregivers, family members of intervention participants, school teachers, and community leaders) 10 or at the wider community or settings level (e.g., in communities or settings not directly exposed to an intervention). 8Institutional or community level spillovers may occur due to a range of factors, including as a "prevention viral effect", whereby prevention knowledge, skills, and ideas spread across nonintervention settings or communities to stimulate action. 8Beneficial effects for nonparticipants have previously been considered signs of possible contamination; 7 however with increasing emphasis on the move to real-world implementation and scale-up, these effects are an important component of overall intervention effects, although they are not yet well understood. 11 achieve effective and cost-effective prevention interventions at a scale required to reduce the population prevalence of obesity, positive spillovers from effective interventions may be critical. 8Failure to consider these spillover effects in economic evaluations may result in undervaluing the cost-effectiveness of the interventions. 12,13Given that childhood obesity prevention interventions are often funded from limited public health budgets, effects that go beyond a strict health perspective on the individual -to consider potential impacts on a broader range of stakeholders such as families, peers, and communities -should be considered. 12 date, and to the best of our knowledge, no systematic review has been published that synthesizes the evidence on spillover effects arising from childhood obesity prevention interventions.This paper aims to systematically review the evidence for spillover effects of childhood obesity prevention interventions on individuals and communities not specifically targeted by the intervention but who may be connected to recipients through physical or social proximity. 6,7Findings from this systematic review will be useful to decision-makers and should be considered by both researchers designing and evaluating interventions and health economists exploring the cost-effectiveness of intervention.

| METHODS
The systematic review followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines 14 and was registered with PROSPERO (42022202545).The PRISMA checklist is in Table S1. 14S2.The results of the academic database searches were imported into Endnote, and duplicates were removed.Covidence software 15 was used for independent title and abstract screening by three reviewers (VB, HT, JJ).Full text review was undertaken by two independent reviewers (HT, JJ), with disagreements resolved through discussion with a third independent reviewer (VB).Reference lists of included studies were hand-searched by one reviewer (VB) for additional relevant study inclusions.

| Search strategy and screening procedures
Trial registry searches have recently been recognized as an important component in systematic reviews that aim to incorporate as much relevant evidence as possible. 16Systematic searches of two publicly available clinical trial registries (clinicaltrials.govand the World Health Organization International Clinical Trials Registry Platform [WHO ICTRP]) were also undertaken by two reviewers (VB, MN) using a predefined search strategy (Table S3).Searches were conducted on June 30, 2022, following best-practice guidance. 16vanced searches were conducted in each database using keywords, and searching for studies in children at any stage of completion (Table S3).The two clinical trial registries were selected as they are both recognized by the International Committee of Medical Journal Editors and are comprehensive in their inclusions. 16The WHO ICTRP includes 17 primary registries, including the Australian New Zealand Clinical Trials Registry and the EU Clinical Trials Register. 17The US National Library of Medicine's clinicaltrials.govdatabase contains records for over 420,000 research studies in 221 countries worldwide. 18Data from the search of trial registries were automatically exported from the trial registries and saved in Microsoft Excel.Trial registration records were screened independently by two reviewers (VB, MN), and conflicts were solved through discussion.The database and trial registry search strategies were updated in January 2023 just prior to paper submission, to ensure all recent relevant studies were included.

| Inclusion and exclusion criteria
Studies meeting the following criteria were included: • Reported outcomes of interventions to prevent obesity or to improve obesity-related health behaviors (including physical activity, healthy eating, sedentary behavior, sleep) with an intention of influencing weight status in children aged between 0 to 19 years; • Reported quantifiable spillover effects in individuals or communities not exposed or primarily targeted by an intervention but who may be connected to participants through physical or social proximity (e.g., family member spillover from an obesity prevention intervention targeted at children, community level spillover); or

• Intervention effects reported as follows:
Quantifiable obesity-related outcomes (including obesityrelated behaviors) measured in those not exposed or not primarily targeted by the intervention (e.g., increase in fruit and vegetable intake in parents/caregivers of intervention participants); Other quantifiable outcomes (not specifically obesity-related) in those not exposed or not primarily targeted (e.g., educational outcomes in siblings of participants and community capacity scores); • Any study design; • Published in English, from January 2007 until January 2023.
The exclusion criteria for studies included the following: • Studies published in a language other than English; • Reviews, meta-analyses, expert opinions, dissertations, conference abstracts; • Studies that only report outcomes in third parties that are related to the child's behaviors or exposure to the intervention (e.g., parenting practices related to obesity-related behaviors, such as feeding practices, parental self-efficacy, parental attitudes, and beliefs specifically about the child's obesity-related behaviors) or the child's environment.These studies have been excluded as childhood obesity prevention interventions often target parenting practices related to the child's obesity-related behaviors and/or changes to obesogenic environments and so these effects cannot be considered as "spillovers."

| Data extraction
A data extraction template for the systematic database search was developed in Microsoft Excel, based on the Consolidated Standards of Reporting Trials (CONSORT) checklist. 19Data were extracted on publication information, country, study aim and/or hypothesis, study design, setting, intervention methods summary, comparator summary, participant inclusion criteria, sample size, primary and secondary outcomes reported in child participants, primary and secondary outcomes reported in others (i.e., spillover effects), statistical methods, summary of findings, limitations, source of funding, registration number (if applicable), and conflicts of interest.Data were extracted by one reviewer (MS) and checked by another reviewer (VB).
Data extracted from the clinicaltrials.govdatabase included trial registration number, website address of the trial registration record, title, acronym, status, condition, intervention, outcome/s, sponsor/ collaborator, gender, age, phases, enrolment, study type, study design, start date, primary completion date, and completion date.Data extracted from the WHO ITRCP database included trial identification, title (public and scientific), acronym, primary sponsor, date of registration, website address of the trial registration record, inclusion ages, target enrolment size, study type, study design, phase, inclusion criteria, intervention summary, primary outcome/s, and secondary outcome/s.

| Data analysis and quality assessment
A narrative synthesis of the characteristics, methods, findings, and limitations of included studies was undertaken and reported using the Synthesis without meta-analysis (SWiM) guidelines (Table S4). 20rrative synthesis was undertaken as the outcome data were too heterogeneous for meta-analysis.
Outcomes from studies published in the academic literature were grouped by outcome domain (anthropometry, diet and/or nutrition, physical activity and/or sedentary behaviors, sleep, or other), and findings were summarized for each domain.Where studies reported multiple outcomes, the study was listed separately for each outcome domain.For trial registration records, vote counting was undertaken and recorded for each outcome and for the population group that the outcomes were measured in (e.g., parent/caregiver, educator, family, and community).Two researchers were involved in data synthesis (VB, MS), and all study co-authors provided critical review of the narrative synthesis.
Critical appraisal was undertaken for all studies reporting statistically significant spillover effects using JBI Critical Appraisal Tools relevant to the study design (Table S5). 21Critical appraisal only for studies reporting statistically significant results was a pragmatic decision, based on the resources available to undertake the review.
The critical appraisal was undertaken by one reviewer (MA) and checked by a second reviewer (VB).
Three studies explored spillover effects in children not exposed to the intervention, 8,11,30 with two reporting statistically significant effects. 8,30Both studies utilized measured BMI in analyses, with the study by Quinto Romani et al 30 using an RCT design (Table S6).
Savage et al 33 found that second-born children whose parents received the INSIGHT intervention with their first child had lower BMI than controls, but the difference was not statistically significant.
The significance of findings did not appear to vary by study characteristics, including study design or outcome measurement method (i.e., self-report or measured).

| Diet and/or nutrition outcomes
,27,35,38 Studies were heterogenous both in terms of the outcomes measured and in whom they were measured, making synthesis of findings challenging. Three studiesexplored dietary intake or healthy eating indices among siblings, 34,35,38 three studies explored dietary intake in parents, 26,32,36 one study explored the frequency of food items served in families, 27 one study explored dietary intake in children not exposed to the intervention 8 and one study explored dietary intentions and knowledge in children not exposed to the intervention.11 Statistically significant positive effects were found for frequency of eating vegetables in second-born siblings, 38 "high energy snack and processed foods" and "high-fat food" scores in mothers, 26 healthier eating habits in families with children from intervention schools, 27 consumption of sugar-sweetened beverages and fruit drinks/cordial 8 and healthy eating index scores 35 in children not exposed to the intervention.Critical appraisal identified that statistically significant effects all arose from parent-or self-report measures, 8,26,27,35,38 with the study by Hohman et al reporting no loss to follow up (Table S6).35

| Physical activity and/or sedentary behavior outcomes
Eight studies examined physical activity or sedentary behavior-related spillover effects. 8,11,23,26,28,30,32,36Outcomes were heterogenous between studies, and included physical activity of parents 23,26,32,36 or children not exposed to the intervention 8 ; aerobic capacity or fitness of children not exposed to the intervention 11,30 ; and, screen behaviors of parents 26,32,36 and children not exposed to the intervention. 8Only one study 8 estimated a statistically significant improvement from preintervention in favoring active transport to school in control participants, versus intervention participants.

| Sleep or other outcomes
Two studies examined spillover effects on sleep-related outcomes, 37 with the study by Hohman et al 37 reporting statistically significant effects on sleep-related outcomes on second-born siblings of children exposed to the INSIGHT intervention.As previously noted, there was no loss to follow-up from the SIBSIGHT study (Table S6). 37One study reported statistically significant improvements in body image-related outcomes in children not exposed to the intervention. 8Other spillover effects that were not statistically significant, included parent depression and fatigue, 24 overall child health 31 and community capacity and coalition effects. 25,29tailed characteristics of studies identified from the academic literature search are given in Supporting Information Table S6.
Comprehensive summaries of the 41 included trial registration records are given in Supporting Information Table S7.Counts of the spillover-related outcomes included in the 41 trial registration records by impacted population group (parent/caregiver, educator, family or community) are given in Figure 2. Of the 41 included trial registration T A B L E 1 Summary of selected study characteristics and spillover findings from the academic literature.(Continues) records, 16 (39%) measured diet-related spillover outcomes (including diet quality, fruit and vegetable intake and sugar-sweetened beverage intake), followed by 14 (34%) measuring BMI spillovers and physicalactivity spillovers respectively.

| DISCUSSION
We found multiple examples of studies, using a range of designs, which measured spillover effects from childhood obesity prevention interventions.For each of the outcome domains considered (anthropometry, diet and/or nutrition, physical activity and/or sedentary behaviors, sleep and other), studies were identified that reported significant spillover effectsin other words, positive health impacts for populations who were not specifically targeted by the intervention, but who were connected to the target population through physical or social proximity.The most commonly reported positive spillover was for diet or nutrition-related outcomes among siblings and parents.
However, these positive spillover effects were not uniform, with several studies reporting null spillover effects across outcomes and domains.In addition, there were several areas where there was limited or no evidence for spillovers (for instance, community outcomes such as community capital or capacity building).
Although we found mixed evidence for spillover effects from child obesity prevention interventions, it should be noted that the search identified only 20 eligible studies from the academic literature.
Considering the size of the obesity prevention literature and the decades of interventions that have now been implemented and evaluated, our review shows that very few studies actively collected data, analyzed, and/or reported on spillover effects in the academic literature.This could be due to various reasons, including a lack of understanding or consideration regarding the importance of spillover effects and limited research budgets to collect relevant data.Without strong empirical and cost-benefit studies, which demonstrate the spillover effects of obesity prevention interventions, we risk underestimating the benefits of intervention, and therefore the case for investment in obesity prevention.
The field of measuring spillover effects appears to be more advanced for vaccines and treatment interventions, than for broader public health prevention interventions.For instance, the herd effects (commonly referred to as positive externalities) of vaccines have been well-documented in epidemiological studies. 7,39The majority of studies published in a 2019 systematic review exploring the inclusion of family spillover effects in pediatric cost-utility analyses focused on infectious diseases. 40Vaccines, followed by pharmaceuticals, were the most common intervention types included. 40A systematic review on spillover effects on health outcomes in low and middle income countries found the strongest evidence for spillovers from vaccines and mass drug administration, with little high quality evidence of spillovers for other interventions. 6r review of spillover effects reported in childhood obesity pre- groups these were measured for.The majority of spillover outcomes identified were directly related to the intervention (BMI, diet or physical activity), with fewer studies reporting on more distal outcomes.
Further, the spillover outcomes were predominantly measured and reported for those closest in terms of physical or social proximity to the children for whom the intervention was targeted (i.e., parents/ caregivers).The more distal the spillover outcome to an intervention and the larger or more distal the population group to the targeted population, the harder it is to measure and attribute any change to the intervention.Few, if any, of the included studies seemed to take a comprehensive approach to investigating the effects of the intervention beyond the primary target group.
The broad range of outcomes reported in the studies included in our review poses challenges in synthesizing the results.The development of Core Outcome Sets (COS), recommending the minimum set of outcomes that should be collected in studies in specific areas in order to better facilitate data synthesis, is a growing field. 41 on children could ensure that resources and strategies are also appropriate for promoting health among teachers, which in turn may have a reinforcing effect on the intervention's impact on children, given teachers' potential as important role models. 43However, spillovers through social and/or physical proximity rely on complex social dynamics, and these factors may vary contextually. 6And although an intervention may provide resources, the ways in which those resources are shared, and the impacts of this sharing on both primary and spillover populations, are not well understood and need further exploration. 6recent paper exploring the inclusion of spillover effects that an individual's illness can have on their informal or unpaid caregivers and other family members posited that consensus guidelines should be established to promote best practices in adequately accounting for these effects in both policy and practice. 13Our results suggest that the scope of use for such guidelines could be broadened into the field of prevention, and that these guidelines could not only promote best practice in accounting for negative spillover effects (e.g., on caregivers arising from caring for a person in ill health) as the authors propose, but extend to positive spillover effects (e.g., on caregivers or others arising from health promotion intervention spillovers on health behaviors or environments).The Second Panel on Cost-Effectiveness in Health and Medicine 44 recommend including health and nonhealth effects borne by patients (participants) and "third-party payers" in an "impact inventory" to inform cost-effectiveness.The findings of our review, however, suggest that much more evidence-building is Strengths of our study included the rigorous, systematic method to identify studies, and the critical appraisal of studies reporting statistically significant results.Limitations included the depth and quality of study information available to inform selection into the review.While we did our best to systematically determine whether interventions were targeted specifically at children, or if intervention content was targeted more broadly (for instance, child-and parent-targeted content), the variations in intervention descriptions or limited information available describing the design and objectives of some interventions made this challenging.We acknowledge that some studies may have been misclassified due to this limited information.In addition, some trial registry records were incomplete, not updated, or lacked sufficient detail and so some trial registrations may have been misclassified.A strength of our search strategy using multiple databases (i.e., academic literature databases, trial registration records) addressed this limitation to the best of our ability.Finally, the potential for publication bias in the academic literature is evident, given the number of trials identified in the search of trial registry records reporting planned or completed collection of data on spillover outcomes (n = 41) but the relatively few papers identified in our search of the academic literature that reported these data (n = 20).While some trials may still be ongoing and therefore findings have not been published in the academic literature, 65% of included trial records were registered >5 years ago.

| Implications for practice and future research
This review suggests we have limited knowledge of the spillover effects for child obesity prevention interventions because few studies consider these outcomes, and of those that do, the quality of methods to measure spillover effects vary substantially and is rarely comprehensive.Nonetheless, our review shows where spillover effects are measured, several, but not all, show positive impacts.
The corollary is that future studies should actively engage with the study of spillover effects to close this evidence gap.This call needs to be tempered with the caveat that incorporating any sort of comprehensive spillover analysis will require collection and analysis of a far broader set of data than a typical trial.At its simplest, this may include additional surveys or measurement among family members, but at its more complex, it will involve trying to understand the impacts on outcomes as broad as those related to the wider community, the health system and nonhealth impacts such as environmental impacts. 45This requires additional funding and increases the burden on data collection.The utility of measuring spillover effects should therefore be a central concern in future research.The types of spillover effects and the methods and metrics for measuring them should be driven by the needs of the key stakeholders, and evidenceinformed judgments on the spillover effects that are most likely to occur, and most likely to be consequential, impacting on our estimation of overall intervention impacts and cost-effectiveness.Such data should be collected in a way that can be readily accessible to policy makers, community planners and others with an interest in population health.A further avenue for research is the science of actively creating spillover effects, within or between communities, and the methods to create ripple effects, 46,47 to maximize and multiply the effects achieved by interventions.

| CONCLUSION
The nature and magnitude of spillover effects from childhood obesity prevention interventions are generally unmeasured, unknown, and therefore unaccounted for when evaluating the overall benefits and costs of these types of interventions.We have shown, from the limited data available, that spillover effects can represent positive benefits for parents or caregivers, families, and communities.Greater attention to spillover effects in childhood obesity prevention trials and interventions will build a stronger case for obesity prevention in the future.

A
systematic search of six databases (EBSCOHost, Academic Search Complete, CINAHL Complete, Global Health, MEDLINE Complete, and Web of Science) was undertaken for studies published between January 2007 and June 2022, to identify those that fit the inclusion criteria.The search was limited to studies published from January 2007 to include contemporary studies published in the last 15 years.Search strategies, keywords, and Medical Subject Headings were developed in conjunction with a subject specific librarian.Academic database search strategies are given in Table Treatment interventions (i.e., interventions specifically targeting children with overweight and/or obesity) or interventions for obesity prevention in special sub groups (e.g., interventions targeting children with diabetes); However, it is currently unclear how comprehensively spillover effects are incorporated in published COS in different health areas and this is an area for future investigation and should be considered in future COS development.For instance, the recently published Core Outcome Set for Early intervention trials to Prevent Obesity in Childhood (COS-EPOCH) recommends the collection of three outcomes in parents/caregivers of young children and five environmental outcomes (including the family meal environment and the sedentary behavior/ physical activity home environment). 42Any future development of a COS for systems-based obesity prevention interventions should consider potential spillovers, given the multi-sector, multi-actor, community-wide focus of these interventions.While several studies included in our review were able to measure and report spillover effects, less is known about how these effects were achieved.Understanding these spillover effect mechanisms are important so that interventions can be adapted to maximize effectiveness.Previous research 6 has suggested that the spillover benefits from a range of health interventions occurred through reduced disease transmission, social proximity and substitution or sharing of resources within households.While reduced disease transmission is not relevant in the context of noncommunicable disease prevention, it may be possible to promote mechanisms such as social proximity and resource sharing within childhood obesity prevention interventions, to actively augment spillover effects.For example, as part of an intervention where children are provided with resources, the sharing of these resources with family members may be actively promoted.Similarly, school-based interventions focused

F I G U R E 2
Count of spillover outcomes in included trial registration records, by stakeholder.required before spillover effects can be rigorously accounted for in evaluating the effectiveness and cost-effectiveness of child obesity prevention interventions.
No statistically significant effects on BMI percentile in either children in the control group or the spillover group.
RCT• 12 week after-school program that ran twiceweekly, for a total of 24 sessions per school• No intervention • Spillover (intervention classroom students who did not participate in the after school intervention) BMI of control and spillover group participants; baseline, post intervention; measured À0.75 to 0.03) lower than controls, respectively.
O • INSIGHT-SIBSIGHT, the INSIGHT responsive parenting intervention was delivered to first-born BMI of second-born siblings; 3, 16, 28, and 52 weeks; measured No statistically significant effects in second-born children.Second-born children whose parents received the INSIGHT intervention with their first T A B L E 1 (Continued) No significant intervention vs. control effects were observed regarding total physical activity or TV viewing time.
• Usual care PA of mother; baseline, postintervention; self-report TV viewing behavior of mother; baseline, postintervention; self-report 95% CI19-64)and in total (53 min, 95% CI 17-90), and more likely to self-soothe to sleep (OR 2.0, 95% CI 1.1-3.7)than second-born children of intervention group participants compared with second-born children of control participants.
Be Active Eat Well; C-RCT, cluster randomized controlled trial; MVPA, moderate-vigorous physical activity; NA, not applicable; O, observational; PA, physical activity; POI, Prevention of Overweight in Infancy; Q, quasi-experimental; RCT, randomized controlled trial; SNAP, Scouting Nutrition and Activity Program; SSB, sugar sweetened beverage; SUS, Shape Up Somerville; TV, television; WIC, United States Department of Agriculture's Special Supplemental Nutrition Program for Women, Infants and Children program.