The prevalence rates of children with obesity have increased rapidly over the past decade in low- and middle-income countries and in Latin America in particular(Reference Abarca-Gómez, Abdeen and Hamid1,Reference Corvalán, Garmendia and Jones-Smith2) . According to estimates from 2008–2013, approximately 20 % of children in Latin America had overweight or obesity(Reference Rivera, De Cossío and Pedraza3), and projected trends suggest that these numbers may be higher(Reference Abarca-Gómez, Abdeen and Hamid1,4) . Moreover, children in Latin America and the Caribbean have experienced one of the most rapid increases in age-standardised mean BMI over the past decade, with children in this region now ranking among the highest globally in terms of mean BMI(Reference Abarca-Gómez, Abdeen and Hamid1). At the same time, Latinx children in high-income countries such as the United States (US) are disproportionately affected by obesity(Reference Ogden, Carroll and Kit5). From 2017–2018, Latinx children had the highest rate of obesity (25·8 %) among all racial and ethnic groups(6). Given these health disparities, there is a need to identify culturally appropriate, community-engaged approaches to prevent obesity in Latinx children in the US and Latin America.
The early years (<6 years of age), in particular, represent an important period for obesity prevention(Reference Dietz7,Reference Lakshman, Elks and Ong8) . This period can be defined as the years after birth and before entry into kindergarten, which usually occurs before age 6. There is increasing recognition that the first few years of development lay a foundation for most health-related behaviours, including self-regulatory capacities. Having obesity during early childhood may have important short- and long-term health consequences, including greater likelihood of suffering from psychological comorbidities(Reference Quek, Tam and Zhang9), asthma(Reference Lang, Bunnell and Hossain10) and a greater risk of musculoskeletal problems and metabolic disorders later in life(Reference Smith, Sumar and Dixon11,Reference Friedemann, Heneghan and Mahtani12) . In the US, there is some evidence of an overall levelling off of obesity during early childhood in recent years(Reference Ogden, Carroll and Kit13), but this trend has not extended to Latinx populations. The prevalence of having obesity among Latinx children 2–5 years old was four times that of their non-Hispanic white peers from 2011 to 2012(Reference Ogden, Carroll and Kit5). More recent estimates show that among Special Supplemental Nutrition Program for Women, Infants and Children (WIC) programme participants, 16·4 % of Latinx children 2–4 years had obesity, representing the highest rate among all racial and ethnic groups with the exception of American Indians and Alaska Natives(6).
Several authors have reviewed the body of literature on obesity prevention interventions in Latinx children. However, most are roughly a decade old and in need of an update(Reference Branscum and Sharma14–Reference Holub, Elder and Arredondo16). Given the parallel rise in young children with obesity in the US and Latin America, there is a need to update the literature to identify and highlight successful interventions targeting Latinx children during early childhood(Reference Vorkoper, Arteaga and Berrigan17). Therefore, the objective of this study was to systematically review the efficacy and effect of obesity prevention interventions in Latinx children during the early years, ages birth to 6 years.
Methods
This review is part of a series of reviews that aim to examine obesity prevention interventions in Latinx children from birth to 18 years of age. Given the wide variation in invention types between early childhood and later childhood and adolescence, we conducted a separate review for young children. The protocol for the larger systematic review and meta-analysis is registered with the PROSPERO database for systematic reviews under registration number CRD42020161339 and has been reported elsewhere(Reference Bleiweiss-Sande, Jiménez-Cruz and Bacardí-Gascón18). We conducted this review according to the guidelines specified by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement(Reference Moher, Liberati and Tetzlaff19).
Search strategy and selection criteria
We included studies published between 2010 and 2020 to provide an update to the existing reviews that have been published on this topic(Reference Branscum and Sharma14–Reference Holub, Elder and Arredondo16). We included articles in peer-reviewed journals published in English, Spanish or Portuguese. We identified studies reporting the results of interventions aimed at preventing obesity in Latinx children that included children ages birth to 6 years at baseline, thereby excluding interventions targeting mothers during pregnancy. We included studies with a minimum of 50 % of children identified as Latinx, a term used to describe the ethnicity of Mexican, Central American, South American and Caribbean origin individuals and those of Latin American descent living in the US and in other countries(20). For interventions that took place in Latin America and Spanish-speaking countries in the Caribbean, we assumed that all children were Latinx, unless stated otherwise.
We included randomised controlled trials, quasi-experimental studies and non-randomised interventions such as natural experiments. We excluded studies without a control or comparison group. We included interventions that targeted obesogenic behaviours or risk factors for obesity, including diet, physical activity, sedentary behaviour, screen time exposure, stress and sleep – or any combination of these behaviours, as well as interventions targeting obesogenic environmental influences, such as community food access, nutrition programmes and policies, and physical activity environments. We did not require that studies report outcome measures related to obesogenic behaviours or environmental influences, but we excluded studies without interventions targeting a behaviour or environmental influence. We included studies that evaluated and reported outcomes such as change in adiposity, measured by age- and sex-standardised BMI (BMI z-score), BMI (BMI), prevalence of overweight and obesity, percentage body fat, waist or hip circumference, skinfold thickness or other anthropometric measures. If a study reported multiple weight-related outcomes, we attempted to extract all relevant quantitative measures. We excluded studies that reported only weight or height(Reference Mei, Grummer-Strawn and Pietrobelli21,Reference Horan, Gibney and Molloy22) . We included studies with direct measurement of adiposity by researchers, study staff, or clinicians or health care providers only (v. parent report). We excluded studies evaluating interventions to treat, rather than prevent, having obesity during childhood and studies targeting children with a specific known medical condition such as diabetes or CVD.
Search methods
We searched PubMed, ERIC, PsycINFO, Scopus, Scientific Electronic Library Online (SciELO) and Google Scholar databases from January 1 2010 to January 1 2020, using a search strategy developed a priori (Reference Bleiweiss-Sande, Jiménez-Cruz and Bacardí-Gascón18). We selected databases based on expertise of the authors. Due to considerable overlap between two databases – SciELO and LILACS – we searched SciELO only. We included studies published between 2010 and 2020 to provide an update to the two existing reviews that have been published on this topic. We included articles published in peer-reviewed journals only to ensure that the review comprised high-quality research. Searches were conducted on May 1 2020 and repeated on January 1 2021 and November 1 2022. This search strategy used a combination of medical subject headings and keyword terms informed by search strategies used in related systematic reviews(Reference Branscum and Sharma14–Reference Holub, Elder and Arredondo16). We translated the final search strategy for each database into Spanish and Portuguese to capture publications written in languages other than English. However, we did not find any non-English studies that met our inclusion criteria. We performed forward and backward citation searches of included studies and reviewed the reference lists of relevant review articles to identify additional publications. We have provided a sample search strategy in English for PubMed and Scopus in Appendix 1.
Data extraction and management
We uploaded all search results into Covidence Software (Covidence Systematic Review Software, Veritas Health Innovation, Melbourne, Australia), an online tool developed for systematic review management. Two independent reviewers conducted title, abstract and full-text screening to assess article eligibility. The same two reviewers used a pre-piloted and standardised form for data abstraction. Information abstracted during this phase included publication details; country; study; intervention details (setting, content, format, delivery, control or comparison group); baseline child demographics and characteristics (e.g. geographic location, gender, family income, parental education); recruitment and intervention complement rates; weight-related outcomes and method of ascertainment; statistical methods; results; and limitations. For interventions taking place in the US, we also noted any culturally tailored intervention elements, such as the use of promotoras (community health educators) to deliver the intervention. The reviewers resolved differences at the screening and abstraction phase through discussion.
We contacted study authors to request missing data regarding child demographics (n 3) and adiposity outcome measures for use in calculating effect sizes (n 18). Authors of eleven studies did not respond, five responded that data were unavailable, and two responded with the requested data.
Quality assessment
Two reviewers independently assessed bias for individual studies using the Cochrane Collaboration’s risk of bias tool (ROB) for randomised(Reference Higgins, Altman and Gotzsche23) or non-randomised studies (ROBINS-I)(Reference Sterne, Hernán and Reeves24), as appropriate. The ROB tool is used to rate studies according to their level of bias (high, low, or unclear) across seven domains: random sequence generation; treatment allocation concealment; blinding of participants and personnel; blinding of outcome assessment; completeness of outcome data; selective outcome reporting; and other sources of bias(Reference Higgins, Altman and Gotzsche23). The ROBINS-I tool also includes seven domains, including confounding; selection of participants into the study; classification of interventions; deviations from intended interventions; missing data; measurement of outcomes; and selection of the reported result(Reference Sterne, Hernán and Reeves24). We resolved discrepancies in judgements through discussion. If we deemed a study to have a high or unclear risk of bias for two or more criteria, we assigned it an overall ROB of high or unclear. In the case of multiple criteria deemed to be high or unclear for the same study, we assigned an overall ROB based on whichever rating was more frequently assigned for that study. Otherwise, we assigned the study a low ROB.
We summarised the quality of included studies using the Grading Quality of Evidence and Strength of Recommendations approach(Reference Guyatt, Oxman and Akl25). Grading Quality of Evidence and Strength of Recommendation rates the quality of studies as high, moderate, low, or very low across four areas, including methodological flaws, consistency of results across studies, generalisability to the target population and effect size(Reference Atkins and Best26). In order to include all studies in our quality assessment, we took into account the precision of estimates for studies that did not have enough information to calculate an effect size.
Evidence synthesis
We conducted a narrative synthesis of included studies by intervention characteristics including design (randomised v. quasi-experimental), primary setting where the intervention took place (early care and education centres such as Head Start or another preschool facility; community sites such as churches or community centres; WIC clinics; primary care or hospital settings and the home) and behavioural target, including diet-only interventions, diet and physical activity interventions (targeting physical activity, sedentary time or a combination) and multiple targets including diet, activity, screen time and sleep. We further described studies by population characteristics, including age group (mean age < 2 years or mean age > 2 years), number of children, percentage of females and the percentage of Latinx children in the study sample. We also aggregated studies by region (US or Latin America) and country. For studies taking place in the US, we synthesised studies by cultural elements included in the interventions. Finally, we aggregated the studies by the weight or adiposity-related outcomes reported.
Despite widespread adoption of culturally tailored interventions in the US, there are no published guidelines to develop culturally appropriate dietary, physical activity or other weight-related interventions among minority populations in the US(Reference Aycinena, Jennings and Gaffney27). However, several publications have reviewed strategies and approaches to developing interventions for specific subgroups, which include cultural adaption through modifications to evidence-based interventions (cultural tailoring); culturally grounded interventions involving active participation from subcultural group members to create intervention materials, and community-initiated indigenous interventions instigated by a community agent(Reference Barrera, Castro and Steiker28,Reference Kreuter, Lukwago and Bucholtz29) . Due to the heterogeneity of these approaches, we documented all cultural components reported by study authors, which included offering study materials in multiple languages, having bilingual study staff, incorporating culturally tailored intervention elements (such as using programmes developed specifically for Latinx families), developing interventions based on research with members of the study population, reporting parent or caregiver acculturation, reporting parent or caregiver place of birth and employing promotoras (health workers from the Latinx community) to lead intervention activities.
As outlined in our protocol paper, we planned to conduct meta-analyses if more than two studies with comparable exposure and outcome variables were available(Reference Bleiweiss-Sande, Jiménez-Cruz and Bacardí-Gascón18). We examined randomised and non-randomised studies separately due to the major methodological differences in these study designs. We also examined post-intervention and follow-up outcomes separately, if available. We used unadjusted outcome estimates to calculate effect sizes. For continuous outcomes (BMI, BMI percentile and BMI z-score), we calculated adjusted, unstandardised mean differences (Hedge’s g) and for dichotomous outcomes (risk of obesity), and we calculated risk ratios and transformed them using the natural log for use in meta-analyses. We combined effect sizes across outcomes using random effects meta-analyses. We computed the between-study variance component (τ2) using the restricted maximum likelihood method, which has been demonstrated to perform well in the case of large τ2 estimates(Reference Veroniki, Jackson and Viechtbauer30). We also specified a modified Knapp–Hartung adjustment be applied to the se of the overall effect size; this approach corrects for type-I error probabilities in the case of meta-analyses of a small number of studies(Reference Röver, Knapp and Friede31).
We assessed the homogeneity of effects among studies using forest plots and Higgins I 2 statistics(Reference Higgins32). We used funnel plots to assess the risk of publication bias. We conducted all analyses in Stata version 16 (StataCorp; Stata Statistical Software: Release 16; 2019). We produced ROB plots using R software (R Core Team; 2013).
Studies included in this review had considerable variation in results and some inconsistency in the direction of the effect for certain outcomes, including BMI, BMI percentile and BMI z-score. In addition, bias was present in some of the individual studies. In this situation, meta-analysis is likely to compound the errors and produce a misleading result(33). Given the high clinical, methodological and statistical heterogeneity of included studies, it would be inappropriate to perform a meta-analysis of included studies(34). Therefore, we do not include meta-analysis results in this study.
Results
Study selection
Through our literature search, we identified 11 861 records including 3260 duplicates. After deduplication and title and abstract screening, we identified 313 articles that potentially met our eligibility criteria. We have presented full-text exclusions by study in Appendix 2. Of those, forty were included in the systematic review(Reference Barkin, Gesell and Po’E35–Reference Taveras, Perkins and Boudreau74) and twenty-five were included in meta-analyses (Fig. 1)(Reference Barkin, Gesell and Po’E35,Reference Barkin, Heerman and Sommer36,Reference Berry, Colindres and Sanchez-Lugo38,Reference Céspedes, Briceño and Farkouh40–Reference Cloutier, Wiley and Huedo-Medina42,Reference Costa, Campagnolo and Lumey44–Reference Heerman, Teeters and Sommer53,Reference Linville, Mintz and Martinez55,Reference Louzada, Campagnolo and Rauber56,Reference Martínez-Andrade, Cespedes and Rifas-Shiman58,Reference Natale, Lopez-Mitnik and Uhlhorn59,Reference Phelan, Hagobian and Ventura62,Reference Romo and Abril-Ulloa63,Reference Sharma, Vandewater and Chuang67,Reference Slusser, Frankel and Robison68,Reference Yin, Parra-Medina and Cordova70) .
Fig. 1 Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) flow diagram for the inclusion of studies
Study design and sample
Study design
Table 1 provides an overview of the main characteristics of the included studies. Two manuscripts(Reference Hughes, Power and Beck54,Reference Hughes, Power and Beck72) reported findings from the same intervention and study population at different time points, so are treated as a single study in syntheses. Most studies (n 30) were randomised or cluster randomised controlled trials(Reference Barkin, Gesell and Po’E35–Reference Céspedes, Briceño and Farkouh40,Reference Cloutier, Wiley and Kuo43–Reference Louzada, Campagnolo and Rauber56,Reference Martínez-Andrade, Cespedes and Rifas-Shiman58–Reference Phelan, Hagobian and Ventura62,Reference Salazar, Vasquez and Concha65,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Slusser, Frankel and Robison68,Reference Washio, Humphreys and Colchado69) , and nine were quasi-experimental studies, including one natural experiment(Reference Chaparro, Crespi and Anderson41) and seven non-equivalent group designs(Reference Cloutier, Wiley and Huedo-Medina42,Reference Machuca, Arevalo and Hackley57,Reference Romo and Abril-Ulloa63,Reference Sadeghi, Kaiser and Hanbury64,Reference Sharma, Vandewater and Chuang67,Reference Yin, Parra-Medina and Cordova70,Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71,Reference Taveras, Perkins and Boudreau74) .
Table 1 Characteristics of the included studies (n 40*)
* The review sample included 40 manuscripts reporting findings from 39 unique studies. Hughes et al. 2020 and Hughes et al. 2021 report findings from the same intervention and study population at different time points.
† Not Reported.
‡ Not applicable.
§ Women, Infants and Children Supplemental Feeding Program.
|| Randomised controlled trial.
¶ Cluster randomised controlled trial.
** Control.
†† Intervention.
‡‡ Non-randomised study.
Setting
The most common intervention setting was early care and education centres (n 13)(Reference Bellows, Davies and Anderson37,Reference Céspedes, Briceño and Farkouh40,Reference Davis, Myers and Cruz46–Reference Fitzgibbon, Stolley and Schiffer48,Reference Grummon, Cabana and Hecht50,Reference Hughes, Power and Beck54,Reference Natale, Lopez-Mitnik and Uhlhorn59,Reference Natale, Messiah and Asfour60,Reference Sadeghi, Kaiser and Hanbury64,Reference Salazar, Vasquez and Concha65,Reference Sharma, Vandewater and Chuang67,Reference Yin, Parra-Medina and Cordova70) , followed by community sites (n 8)(Reference Barkin, Gesell and Po’E35,Reference Berry, Colindres and Sanchez-Lugo38,Reference Haines, Rifas-Shiman and Gross52,Reference Linville, Mintz and Martinez55,Reference Machuca, Arevalo and Hackley57,Reference Slusser, Frankel and Robison68,Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71,Reference Taveras, Perkins and Boudreau74) , WIC clinics (n 4)(Reference Bonuck, Avraham and Lo39,Reference Chaparro, Crespi and Anderson41,Reference Palacios, Campos and Gibby61,Reference Phelan, Hagobian and Ventura62) , primary care or hospital settings (n 4)(Reference Cloutier, Wiley and Huedo-Medina42,Reference Costa, Campagnolo and Lumey44,Reference Martínez-Andrade, Cespedes and Rifas-Shiman58,Reference Sangalli, Leffa and Valmórbida73) and the home (n 2)(Reference Haines, McDonald and O’Brien51,Reference Louzada, Campagnolo and Rauber56) . Eight studies took place in combined settings involving the home and a community, WIC or early care and education setting(Reference Barkin, Heerman and Sommer36,Reference Cloutier, Wiley and Kuo43,Reference Crespo, Elder and Ayala45,Reference French, Sherwood and Veblen-Mortenson49,Reference Heerman, Teeters and Sommer53,Reference Romo and Abril-Ulloa63,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Washio, Humphreys and Colchado69) .
Demographic characteristics of the study population
Studies included 70 458 children at baseline overall, and we included 59 147 intervention children in quantitative analyses. Baseline sample sizes ranged from nineteen(Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71) to 57 171(Reference Chaparro, Crespi and Anderson41). The mean age of children at baseline ranged from newborns to 70·8 months. Thirteen studies targeted children younger than 24 months of age at baseline(Reference Bonuck, Avraham and Lo39,Reference Chaparro, Crespi and Anderson41,Reference Cloutier, Wiley and Kuo43,Reference Louzada, Campagnolo and Rauber56,Reference Machuca, Arevalo and Hackley57,Reference Palacios, Campos and Gibby61,Reference Phelan, Hagobian and Ventura62,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Washio, Humphreys and Colchado69–Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71,Reference Sangalli, Leffa and Valmórbida73,Reference Taveras, Perkins and Boudreau74) . The percentage of females ranged from 47 to 60 %; however, six studies did not report the characteristics of children by gender or sex(Reference Cloutier, Wiley and Kuo43,Reference Crespo, Elder and Ayala45,Reference Washio, Humphreys and Colchado69,Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71,Reference Sangalli, Leffa and Valmórbida73,Reference Taveras, Perkins and Boudreau74) .
Intervention duration and follow-up
Online Supplementary 1 and 2 provide further details of the included studies. Twenty interventions lasted fewer than 12 months in duration(Reference Barkin, Gesell and Po’E35,Reference Bellows, Davies and Anderson37,Reference Berry, Colindres and Sanchez-Lugo38,Reference Céspedes, Briceño and Farkouh40,Reference Fernandez-Jimenez, Jaslow and Bansilal47,Reference Fitzgibbon, Stolley and Schiffer48,Reference Grummon, Cabana and Hecht50–Reference Hughes, Power and Beck54,Reference Martínez-Andrade, Cespedes and Rifas-Shiman58,Reference Natale, Lopez-Mitnik and Uhlhorn59,Reference Palacios, Campos and Gibby61,Reference Romo and Abril-Ulloa63,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Slusser, Frankel and Robison68–Reference Yin, Parra-Medina and Cordova70,Reference Sangalli, Leffa and Valmórbida73,Reference Taveras, Perkins and Boudreau74) , with the shortest intervention lasting 6 weeks(Reference Martínez-Andrade, Cespedes and Rifas-Shiman58). Six interventions lasted 12 months(Reference Bonuck, Avraham and Lo39,Reference Cloutier, Wiley and Huedo-Medina42–Reference Crespo, Elder and Ayala45,Reference Phelan, Hagobian and Ventura62) , and nine interventions lasted for longer than 12 months(Reference Barkin, Heerman and Sommer36,Reference Davis, Myers and Cruz46,Reference French, Sherwood and Veblen-Mortenson49,Reference Louzada, Campagnolo and Rauber56,Reference Machuca, Arevalo and Hackley57,Reference Natale, Messiah and Asfour60,Reference Sharma, Vandewater and Chuang67,Reference Taveras, Perkins and Boudreau74,Reference Sadeghi, Kaiser and Schaefer75) . One study reported the results of a natural experiment that followed children for 48 months(Reference Chaparro, Crespi and Anderson41), and three studies did not report the intervention length(Reference Linville, Mintz and Martinez55,Reference Salazar, Vasquez and Concha65,Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71) . Twenty-two studies reported follow-up outcomes ranging from one month to 7 years(Reference Barkin, Gesell and Po’E35,Reference Barkin, Heerman and Sommer36,Reference Berry, Colindres and Sanchez-Lugo38,Reference Céspedes, Briceño and Farkouh40,Reference Chaparro, Crespi and Anderson41,Reference Costa, Campagnolo and Lumey44,Reference Crespo, Elder and Ayala45,Reference Fernandez-Jimenez, Jaslow and Bansilal47,Reference Fitzgibbon, Stolley and Schiffer48,Reference Haines, Rifas-Shiman and Gross52–Reference Natale, Messiah and Asfour60,Reference Romo and Abril-Ulloa63,Reference Sadeghi, Kaiser and Hanbury64,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Slusser, Frankel and Robison68,Reference Washio, Humphreys and Colchado69) .
Behavioural targets
Most interventions targeted a combination of three or more obesogenic factors including diet, physical activity, sedentary time, screen time or media use, parenting skills or the home and community environment (n 21)(Reference Barkin, Gesell and Po’E35,Reference Barkin, Heerman and Sommer36,Reference Berry, Colindres and Sanchez-Lugo38,Reference Cloutier, Wiley and Huedo-Medina42,Reference Cloutier, Wiley and Kuo43,Reference Crespo, Elder and Ayala45–Reference French, Sherwood and Veblen-Mortenson49,Reference Haines, McDonald and O’Brien51–Reference Linville, Mintz and Martinez55,Reference Martínez-Andrade, Cespedes and Rifas-Shiman58,Reference Natale, Lopez-Mitnik and Uhlhorn59,Reference Phelan, Hagobian and Ventura62,Reference Romo and Abril-Ulloa63,Reference Sharma, Vandewater and Chuang67,Reference Yin, Parra-Medina and Cordova70,Reference Taveras, Perkins and Boudreau74) ; ten interventions targeted diet or infant feeding only(Reference Bonuck, Avraham and Lo39,Reference Chaparro, Crespi and Anderson41,Reference Costa, Campagnolo and Lumey44,Reference Grummon, Cabana and Hecht50,Reference Louzada, Campagnolo and Rauber56,Reference Palacios, Campos and Gibby61,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Washio, Humphreys and Colchado69,Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71,Reference Sangalli, Leffa and Valmórbida73) ; four targeted parenting skills and parent feeding behaviour(Reference Hughes, Power and Beck54,Reference Machuca, Arevalo and Hackley57,Reference Natale, Messiah and Asfour60,Reference Slusser, Frankel and Robison68) , and four targeted diet and physical activity or movement skills only(Reference Bellows, Davies and Anderson37,Reference Céspedes, Briceño and Farkouh40,Reference Salazar, Vasquez and Concha65,Reference Sadeghi, Kaiser and Schaefer75) . All studies involved parents in one or more intervention elements except one school-based study, which did not engage parents in intervention activities(Reference Romo and Abril-Ulloa63).
Intervention approach
Of the thirty-nine included studies, six assessed policy, systems or environmental approaches to obesity prevention. The first examined the impact of changes to the 2009 WIC food package through a natural experiment(Reference Chaparro, Crespi and Anderson41). These changes included the addition of fruits, vegetables and whole grains; reduction in the amount of juice, milk, cheese and eggs offered; reductions in the fat levels allowed in milk; inclusion of culturally diverse replacement options; and reduction in the amount of formula for breastfeeding mothers(Reference Chaparro, Crespi and Anderson41). The second study evaluated a beverage intervention targeting changes in children’s food environment, including the adoption and integration of the Healthy Beverages in Childcare Policy in California early care and education centres(Reference Grummon, Cabana and Hecht50). Two studies by Natale et al. also included changes to policies in early care and education centres. One intervention developed policies to increase physical activity and healthy eating(Reference Natale, Lopez-Mitnik and Uhlhorn59), and one integrated the American Academy of Pediatrics Caring for Our Children policies into early care and education centre practices(Reference Natale, Messiah and Asfour60). This policy promotes healthy drinks and snacks, adequate physical activity and minimal screen time in childcare centres(Reference Natale, Messiah and Asfour60). One study targeted systems-level changes to prevent early childhood obesity through clinical staff obesity prevention training, family-level behavioural knowledge and lifestyle changes and individual-level supports for women and infants considered to be high risk for obesity(Reference Taveras, Perkins and Boudreau74). Finally, a study by Salazar et al. evaluated the impact of a newly developed national preschool education curriculum in Chile(Reference Salazar, Vasquez and Concha65). Three of these studies found a positive impact on adiposity in favour of the intervention(Reference Natale, Messiah and Asfour60,Reference Salazar, Vasquez and Concha65,Reference Taveras, Perkins and Boudreau74) , but two only found an impact among obese children(Reference Natale, Messiah and Asfour60,Reference Salazar, Vasquez and Concha65) .
Study location
Most studies took place in the US (n 30)(Reference Barkin, Gesell and Po’E35–Reference Bonuck, Avraham and Lo39,Reference Chaparro, Crespi and Anderson41–Reference Cloutier, Wiley and Kuo43,Reference Crespo, Elder and Ayala45–Reference Linville, Mintz and Martinez55,Reference Machuca, Arevalo and Hackley57,Reference Natale, Lopez-Mitnik and Uhlhorn59–Reference Phelan, Hagobian and Ventura62,Reference Sadeghi, Kaiser and Hanbury64,Reference Sharma, Vandewater and Chuang67–Reference Yin, Parra-Medina and Cordova70,Reference Taveras, Perkins and Boudreau74) , and nine took place in Latin American countries, including Colombia (n 1)(Reference Céspedes, Briceño and Farkouh40), Brazil (n 4)(Reference Costa, Campagnolo and Lumey44,Reference Louzada, Campagnolo and Rauber56,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Sangalli, Leffa and Valmórbida73) , Mexico (n 2)(Reference Martínez-Andrade, Cespedes and Rifas-Shiman58,Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71) , Chile (n 1)(Reference Salazar, Vasquez and Concha65) and Ecuador (n 1)(Reference Romo and Abril-Ulloa63). One study took place in the US and Puerto Rico (included as a US-based study)(Reference Palacios, Campos and Gibby61).
Cultural components of studies in the US
Of the thirty studies that took place in the US, the percentage of Latinx children ranged from 51(Reference Haines, McDonald and O’Brien51) to 100 %(Reference Barkin, Gesell and Po’E35,Reference Berry, Colindres and Sanchez-Lugo38,Reference Crespo, Elder and Ayala45,Reference Heerman, Teeters and Sommer53,Reference Hughes, Power and Beck54,Reference Sadeghi, Kaiser and Hanbury64,Reference Slusser, Frankel and Robison68,Reference Washio, Humphreys and Colchado69) . Of these, eight recruited Latinx children exclusively(Reference Barkin, Gesell and Po’E35,Reference Berry, Colindres and Sanchez-Lugo38,Reference Crespo, Elder and Ayala45,Reference Heerman, Teeters and Sommer53,Reference Hughes, Power and Beck54,Reference Sadeghi, Kaiser and Hanbury64,Reference Slusser, Frankel and Robison68,Reference Washio, Humphreys and Colchado69) . Most studies that took place in the US offered study and intervention materials in English and Spanish (n 25)(4,Reference Barkin, Gesell and Po’E35–Reference Bonuck, Avraham and Lo39,Reference Cloutier, Wiley and Huedo-Medina42,Reference Cloutier, Wiley and Kuo43,Reference Crespo, Elder and Ayala45,Reference Davis, Myers and Cruz46,Reference Fitzgibbon, Stolley and Schiffer48,Reference French, Sherwood and Veblen-Mortenson49,Reference Haines, McDonald and O’Brien51–Reference Linville, Mintz and Martinez55,Reference Natale, Lopez-Mitnik and Uhlhorn59–Reference Phelan, Hagobian and Ventura62,Reference Sadeghi, Kaiser and Hanbury64,Reference Sharma, Vandewater and Chuang67–Reference Yin, Parra-Medina and Cordova70) and employed bilingual study staff (n 21) (Table 2)(4,Reference Barkin, Gesell and Po’E35,Reference Barkin, Heerman and Sommer36,Reference Berry, Colindres and Sanchez-Lugo38,Reference Bonuck, Avraham and Lo39,Reference Cloutier, Wiley and Huedo-Medina42,Reference Cloutier, Wiley and Kuo43,Reference Crespo, Elder and Ayala45,Reference Fitzgibbon, Stolley and Schiffer48,Reference Haines, McDonald and O’Brien51–Reference Linville, Mintz and Martinez55,Reference Natale, Lopez-Mitnik and Uhlhorn59–Reference Phelan, Hagobian and Ventura62,Reference Sadeghi, Kaiser and Hanbury64,Reference Sharma, Vandewater and Chuang67,Reference Slusser, Frankel and Robison68,Reference Yin, Parra-Medina and Cordova70) . About one-third of these studies (n 10) described culturally tailored intervention elements, including programmes developed by the National Latino Children’s Institute(Reference Barkin, Gesell and Po’E35), provision of Latinx culture-specific foods(Reference Chaparro, Crespi and Anderson41,Reference Linville, Mintz and Martinez55) and intervention materials developed for low-literacy populations(Reference Palacios, Campos and Gibby61). Eight studies included interventions that were developed or piloted within a similar study population or community(Reference Berry, Colindres and Sanchez-Lugo38,Reference Davis, Myers and Cruz46,Reference Fitzgibbon, Stolley and Schiffer48,Reference Haines, McDonald and O’Brien51,Reference Hughes, Power and Beck54,Reference Linville, Mintz and Martinez55,Reference Slusser, Frankel and Robison68,Reference Yin, Parra-Medina and Cordova70) . One study employed promotoras, or health workers from the Latinx community, to deliver the intervention(Reference Crespo, Elder and Ayala45).
Table 2 Cultural components included in studies conducted in the United States (n 30)
* Hughes et al.’s 2020 and 2021 report findings from the same intervention and study population at different time points.
Measures of adiposity
The most reported measure of adiposity was BMI z-score (n 18), followed by BMI (n 13) and BMI percentile (n 7). Other adiposity outcomes included risk of overweight and obesity; weight-for-length, weight-for-age and weight-for-height z-score; waist circumference; waist circumference to height ratio; skinfold thickness and body fat percentage. Overall, twelve studies reported a desirable outcome effect in favour of the intervention(Reference Barkin, Gesell and Po’E35,Reference Berry, Colindres and Sanchez-Lugo38,Reference Chaparro, Crespi and Anderson41,Reference Cloutier, Wiley and Kuo43,Reference Haines, McDonald and O’Brien51,Reference Heerman, Teeters and Sommer53,Reference Machuca, Arevalo and Hackley57,Reference Romo and Abril-Ulloa63,Reference Sadeghi, Kaiser and Hanbury64,Reference Slusser, Frankel and Robison68,Reference Taveras, Perkins and Boudreau74,Reference Shamah Levy, Morales Ruán and Amaya Castellanos76) ; two studies reported a positive effect among children having obesity only (in these two studies, all children attending a childcare centre were enrolled in the intervention, regardless of their baseline weight status)(Reference Natale, Messiah and Asfour60,Reference Salazar, Vasquez and Concha65) .
Risk of bias
Online Supplementary 3 and 4 present the overall ROB ratings for the included studies by domain and individual domain ratings by study, respectively. Appendix 3 and 4 provide ROB assessments by study and individual outcome, with reasons. Of randomised studies, eight received an overall high ROB rating(Reference Barkin, Gesell and Po’E35,Reference Bellows, Davies and Anderson37,Reference Davis, Myers and Cruz46,Reference French, Sherwood and Veblen-Mortenson49,Reference Haines, Rifas-Shiman and Gross52,Reference Louzada, Campagnolo and Rauber56,Reference Martínez-Andrade, Cespedes and Rifas-Shiman58,Reference Sangalli, Leffa and Valmórbida73) , 20 received an unclear rating(Reference Berry, Colindres and Sanchez-Lugo38–Reference Céspedes, Briceño and Farkouh40,Reference Cloutier, Wiley and Kuo43,Reference Crespo, Elder and Ayala45,Reference Fernandez-Jimenez, Jaslow and Bansilal47,Reference Fitzgibbon, Stolley and Schiffer48,Reference Grummon, Cabana and Hecht50,Reference Haines, McDonald and O’Brien51,Reference Heerman, Teeters and Sommer53–Reference Linville, Mintz and Martinez55,Reference Natale, Lopez-Mitnik and Uhlhorn59–Reference Phelan, Hagobian and Ventura62,Reference Salazar, Vasquez and Concha65,Reference Schwartz, Vigo and Dias De Oliveira66,Reference Slusser, Frankel and Robison68,Reference Washio, Humphreys and Colchado69) , and 2 received a low rating(Reference Barkin, Heerman and Sommer36,Reference Costa, Campagnolo and Lumey44) . Among non-randomised studies, one received an overall high ROB rating(Reference Machuca, Arevalo and Hackley57), and eight received an unclear rating(Reference Chaparro, Crespi and Anderson41,Reference Cloutier, Wiley and Huedo-Medina42,Reference Romo and Abril-Ulloa63,Reference Sadeghi, Kaiser and Hanbury64,Reference Sharma, Vandewater and Chuang67,Reference Yin, Parra-Medina and Cordova70,Reference Zaragoza Cortes, Trejo Osti and Ocampo Torres71,Reference Benjamin Neelon, Reyes-Morales and Haines77) .
Strength of evidence
Tables 3 and 4 present the strength of evidence by study design and setting for randomised and non-randomised studies, respectfully. Among randomised studies, the strength of evidence was low for all settings, due to risk of bias or indirectness of the evidence. Interventions taking place in community or home settings were the most likely to report a desirable outcome effect. Among the non-randomised studies, the strength of evidence was moderate for early care and education, and insufficient or low for all other settings. Two of the three studies taking place in early care and education settings reported a desirable effect, and one study taking place in a community setting reported a desirable effect. The single studies taking place in primary care, WIC and combined settings all reported a desirable intervention effect.
Table 3 Summary of findings for randomised studies, by setting (n 30)*
* Hughes et al.’s 2020 and 2021 report findings from the same intervention and study population at different time points so are not counted as independent study populations.
† Assessed using the Cochrane Risk of Bias assessment for randomised trials. Detailed explanation of risk of bias judgements for individual studies is presented in online Supplementary 4.
‡ Combined settings include community and home; community, home and early care and education; primary care and home, and WIC and home.
Table 4 Summary of findings for non-randomised studies, by setting (n 9)
* Assessed using the Risk Of Bias In Non-randomised Studies – of Interventions (ROBINS-I) assessment tool. Detailed explanation of risk of bias judgements for individual studies is presented in online Supplementary 5.
Discussion
In this comprehensive systematic review, we identified forty manuscripts reporting findings from thirty-nine relevant studies reporting adiposity measures from obesity prevention interventions in Latinx children during early childhood. Of the thirty randomised studies included in this review, studies taking place in community or home settings were more likely to report significant reductions in adiposity or weight-related outcomes as a result of the intervention compared to early care and education, WIC, primary care or combined settings. Almost all (n 7) of the non-randomised studies reported a significant adiposity or weight-related outcome in favour of the intervention. We found moderate evidence that early care and education settings may be effective in preventing obesity for non-randomised study designs. Overall, we found low or insufficient evidence by setting the effectiveness of obesity prevention interventions in Latinx children, and a lack of consistent exposure and outcome variables prevented further tabulation by study characteristics.
Our findings of low intervention quality and inconsistent results are aligned with previous reviews examining interventions in similar populations. A review of obesity prevention interventions in Hispanic children in the first 1000 d identified only five relevant interventions and found that most were of low or moderate quality(Reference Ismaeel, Weems and McClendon78). Although all but one intervention led to an improvement in the outcome measure assessed, none of the included studies reported change in adiposity or weight-related measures(Reference Ismaeel, Weems and McClendon78). The authors point out that the lack of assessment of any clinical outcome measures was a major limitation of the included studies(Reference Ismaeel, Weems and McClendon78). Branscum and Sharma reviewed obesity prevention interventions in Latinx children from 2000 to 2010(Reference Branscum and Sharma14). Of the nine studies included in their review, two targeted children 6 years and younger. Neither study found an impact on weight-related measures or intermediate outcomes including diet and physical activity(Reference Branscum and Sharma14). A review by Pérez-Morales et al. that focused on obesity prevention interventions in Latinx children in the US from 2001 to 2012 found that the quality of evidence of the included studies was low, with inconsistent improvements in weight-related outcomes(Reference Perez-Morales, Bacardi-Gascon and Jimenez-Cruz15). Only two studies targeted children 6 years of age and younger, both of which are included in this review(Reference Crespo, Elder and Ayala45,Reference Fitzgibbon, Stolley and Schiffer48) . Two other reviews examining childhood obesity prevention interventions in Latin America and the US focused on school-aged children only(Reference Holub, Elder and Arredondo16,Reference Chavez and Nam79) .
Although including interventions that took place in both the US and Latin America in this review represents a strength of our research, this also led to substantial heterogeneity in terms of the child, study and intervention characteristics. Indeed, the major drivers of obesity among Latinx children in the US and Latin America are diverse and may include contributors such as dietary factors, the local food environment and physical activity patterns(Reference Popkin and Reardon80). There is substantial evidence that points to an ongoing shift in dietary intake and energy expenditure in less developed regions such as Latin America, referred to as the nutrition transition(Reference Popkin, Adair and Ng81). Researchers have pointed to broad changes in the food system at the national and local level, which have led to increases in low-nutrient-dense, highly processed food and sugar-sweetened beverage consumption and an uptick in away-from-home eating(Reference Popkin and Reardon80). These dietary shifts are exacerbated by changes to the local food environment including increased access to supermarkets and fast-food restaurants(Reference Popkin and Reardon80,Reference Reardon and Berdegué82) and exposure to targeted food and beverage marketing that promotes unhealthful products(Reference Gunderson, Clements and Benjamin-Neelon83,Reference Adeigbe, Baldwin and Gallion84) . Research has also demonstrated that Latinx children in the US and Latin America may be at risk for physical inactivity due to limited access to greenspace, high neighbourhood crime rates and transportation barriers(Reference Trost, McCoy and Vander Veur85–Reference Hallal, Andersen and Bull87).
Factors associated with acculturation may impact weight status among Latinx children in the US, beginning as early as infancy. For example, Latinx mothers are more likely to initiate breastfeeding than the national average, but they are also more likely to supplement with formula feeding, often due to beliefs regarding cultural norms and the need to return to work(Reference Sloand, Budhathoki and Junn88,Reference Hohl, Thompson and Escareño89) . In addition, although recent Latinx immigrants experienced lower rates of chronic disease compared to their non-Latinx white peers, studies have demonstrated that more time spent in the US was associated with having obesity(Reference McLeod, Buscemi and Bohnert90). Clearly, the multifaceted nature of factors that may influence obesity in Latinx children necessitates a multidimensional response to obesity prevention.
We acknowledge several limitations to this review. Our decision to focus on studies reporting measures of adiposity as an outcome may lead us to exclude studies focused on strategies to improve behaviours associated with obesity, such as changes in diet, physical activity or sleep. However, other reviews have examined specific obesogenic behaviours, such as sugar-sweetened beverage intake(Reference Vercammen, Frelier and Lowery91) and physical activity(Reference Eisenberg, Sánchez-Romero and Rivera-Dommarco92,Reference Yuksel, Şahin and Maksimovic93) . Second, the limited number of interventions for obesity prevention with available data to compute effect sizes restricted our ability to conduct meta-analyses and the number of reviews with comparable study designs and outcome variables. However, by including all available studies in a narrative synthesis, we have reviewed the available literature as rigorously as possible. Finally, our decision to include both randomised and non-randomised study designs introduced analytic complexities, further precluding meta-analysis. However, by including both randomised and non-randomised study designs, this review sheds light on potential policy, systems and environmental approaches to obesity prevention in Latinx populations.
Implications for policy and practice
Our review found that interventions taking place in community or home settings were more likely to report significant reductions in adiposity or weight-related outcomes as a result of the intervention compared to early care and education, WIC, primary care or combined settings. Community-based interventions, in particular, involve multiple stakeholders and buy-in from diverse community groups. These interventions may have greater success due more rigorous formative research with the study population and a better understanding of important culturally relevant intervention components. There is a need for more culturally appropriate, community-engaged approaches in future research to address the broad inequities in health in Latinx children.
Studies employing quasi-experimental designs may hold promise for future obesity prevention interventions. Specifically, interventions that use policy, systems and environmental strategies for obesity prevention have emerged as effective strategies to address complex public health issues. These strategies target the broader social and environmental context to support diet and physical activity changes, thus addressing underlying determinants of health and social inequity. Policy, systems and environmental strategies are particularly important for populations at a greater risk of obesity, including Latinx children. Five studies in our review assessed policy, systems or environmental strategies for obesity prevention in Latinx children, with two demonstrating a positive intervention effect among obese children. It is important to note that the effects of obesity prevention interventions may take much longer to appear than treatment interventions. In addition, studies targeting changes in policy or environmental factors may not have noticeable effects in the short term. Longer study duration and long-term follow-up with participants may be necessary to ascertain the true impact of preventative intervention strategies and is critical to advancing successful approaches on a broader scale.
Recent research has highlighted the need to use novel approaches to adapt and scale up intervention strategies for obesity prevention and control in the US and Latin America. Using a case study approach to understand how successful obesity policies and programmes have been implemented in the US and Latin America, Perez-Escamilla et al. found that evidence-based advocacy and evidence of scalability and advocacy were key factors to the launch and implementation of successful interventions(Reference Pérez-Escamilla, Vilar-Compte and Rhodes94). The authors argue that the use of implementation science, which aims to promote integration of research findings into policy and practice, may be an important strategy to use during intervention implementation as well as during the maintenance phase to ensure ongoing success and sustainability. Implementation science can offer a forward-thinking approach to designing, implementing and adapting obesity prevention research in Latinx communities.
Conclusion
In this systematic review and meta-analysis, we found that randomised interventions taking place in community or home settings were more likely to report significant reductions in adiposity or weight-related outcomes compared to other settings. Studies with less-rigorous study designs, such as quasi-experimental studies, were also more likely to report a favourable intervention effect. This review provides an important update to the literature regarding interventions to prevent obesity in Latinx child populations globally over the past decade. Preventing obesity among Latinx children is an issue of critical global public health importance. Results are relevant to stakeholders across multiple sectors engaged in obesity prevention in Latinx children, including community health workers, researchers and policymakers.
Acknowledgements
We would like to acknowledge Silvia Costa, PhD for her assistance with Portuguese translations.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Conflicts of interest
There are no conflicts of interest.
Authorship
R.B.S. and S.E.B.N. conceived of the study. R.B.S., M.B.G. and K.S. developed the search strings. R.B.S., K.S. and D.Z. contributed to data extraction. R.B.S. designed and conducted the analyses. All authors read and approved the final manuscript.
Ethics of human subject participation
N/A
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980023001283
Open access
