1 INTRODUCTION

Childhood obesity has emerged as one of the most serious worldwide problems. This is recognized by the World Health Organization (WHO) which has included childhood obesity as a priority in the 2013–2020 Global Plan of Action for the prevention and control of chronic noncommunicable diseases.1 In fact, this institution has already defined a set of global targets for detaining the increase of obesity among children and adolescents by 2025. Recently, data from a systematic review and meta-analysis2 indicated that despite a recent stabilization of the trend of increasing prevalence of excessive weight during childhood in most European countries, the prevalence continues to be high. According to the latest Childhood Obesity Surveillance Initiative (COSI) report, the prevalence of overweight and obesity for boys and girls aged 6 to 9 years was 29% and 27%, respectively, worldwide.3 However, when considering Mediterranean countries, including Spain, more than 40% of children are estimated to have overweight and obesity.

Childhood overweight and obesity are the result of a complex interaction of biological, behavioral and environmental factors that have an impact on the long-term energy balance.4 Therefore, identification of modifiable risk factors, including dietary components, is a crucial step in the prevention of childhood obesity and its comorbidities. For example, in the last few decades, sugar-sweetened beverage consumption has increased globally,5 paralleling the rise in childhood obesity.6 It has been suggested that the increase in sugar-sweetened beverage consumption has been associated with a decline in dairy consumption.7-9 Displacement of dairy from the diet with sugar-sweetened beverages could result in insufficient daily intake of essential nutrients because, although the nutrient content varies by type of dairy product, dairy, in general, provides a dietary source of proteins, fat, calcium, phosphorus, vitamins, and other milk-derived bioactive compounds important for good health.

A few reviews10-14 and meta-analyses10, 15, 16 of epidemiological studies, exploring the potential relationship between the consumption of dairy products and risk of obesity in children and adolescents, have been published to date. Findings suggest an inverse or null relationship between dairy product consumption and the prevalence or incidence of overweight and obesity. However, these analyses present with some methodological issues, for instance, the combination of studies with different exposures (e.g., full fat milk and total milk) in the total dairy product analysis16 or the inclusion of different outcomes (obesity and abdominal obesity) together in the main analysis.15 Moreover, since the publication of the last meta-analysis in 2016, new epidemiological studies evaluating these associations have been published. Therefore, in order to update the evidence and clarify the aforementioned limitations, the aim of this work was to systematically review and meta-analyze the associations between total dairy consumption and its different subtypes with the prevalence and incidence of overweight, obesity, and overweight/obesity in children and adolescents.

2 METHODS 2.1 Search strategy and study selection

For the present systematic review and meta-analysis, we followed the methodological guidelines of the Cochrane Handbook for Systematic Reviews of interventions17 and results were reported in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.18 The protocol was registered at https://www.clinicaltrials.gov/ with the identifier NCT04297722.

We conducted a comprehensive electronic systematic search of two databases (Medline through PUBMED and the Cochrane Library) until October 18, 2021, combining relevant MeSH terms and key words. Table S1 depicts the detailed search strategy. Additionally, the reference lists of the retrieved articles were manually reviewed to ensure all relevant studies were identified.

First, duplicate studies from the identified articles through the search strategy were discarded. Next, two independent reviewers (LL and IPG or JFGG) performed an initial screening of the titles and abstracts of the retrieved papers against eligibility. To aid in performing this task, the online screening program, Abstrackr (http://abstrackr.cebm.brown.edu/) was used. Cross-sectional or prospective cohort studies, with at least 1 year of follow-up, were included if they were conducted in children or adolescents (aged between 2 and 21 years old). Included articles had to reported risk estimates as odd ratios (OR), relative risks (RR), or hazard ratios (HR) and their corresponding 95% confidence intervals (CI) for the association between dairy product consumption (including total dairy products, full-fat or low-fat total dairy, total milk, full-fat or low-fat milk, total yogurt, full-fat or low-fat yogurt, and cheese) and the risk of overweight and/or obesity. We did not consider published abstracts or proceedings for inclusion.

2.2 Data extraction

To verify that the articles which passed the initial title and abstract screening process met the eligibility criteria, two independent researchers (LL and IPG or JFGG) reviewed the full text. Additionally, relevant information was extracted from articles that met inclusion criteria using a standardized proforma. Collected data included authors, article title, journal and year of publication, study name, participant characteristics, sample size, follow-up (only for cohort studies), type of exposure, dietary assessment method, type of outcome and assessment method, number of cases, statistical analyses, and multivariable-adjusted risk estimates (OR, RR or HR, and 95%CI) for the association of interest. A third researcher (NB-T or SKN) was consulted to provide consensus and solve any disagreements.

2.3 Quality assessment of the included studies

Two different quality assessment tools were utilized to assess the quality of the included studies. For cross-sectional studies, we used The Study Quality Assessment Tool for Observational Cohort and Cross-sectional Studies from the National Heart, Lung, and Blood Institute.19

For prospective cohort studies the Newcastle–Ottawa Scale20 was used. This is a rating scale ranging from 0 to 9 points that are given to the studies based on three domains, population selection, outcome assessment, and comparability. Studies with a total punctuation of at least 7 points were considered to be high quality (i.e., low risk of bias). Any disagreement between researchers (MM and LL or IPG and JFGG) was solved by consensus or consulting a third researcher (NB-T or SKN). When additional information from included articles was required for the analyses, the corresponding author was contacted by e-mail.21-31 We received responses from seven authors,22, 24-26 where two provided additional information relevant for inclusion in the analyses.25, 31

2.4 Outcomes

Obesity, using body mass index (BMI) as a proxy where applicable, was the primary outcome. Secondary outcomes were overweight and overweight/obesity combined. There is no standard worldwide BMI cut-off point to define overweight and obesity in children. Hence, most of the studies followed the international criteria suggested by Cole et al. (2000)32 and (2007),33 which define overweight and obesity based on age- and sex-specific BMI cut-off points that correspond to the worldwide accepted cut-off points for defining overweight (BMI ≥ 25 kg/m2) and obesity (BMI ≥ 30 kg/m2) in adults. Other studies used BMI percentiles as a cut-off point to define overweight or obesity for the children and adolescents. For instance, the criteria proposed by the Centers for Disease Control and Prevention (CDC) growth charts for United States based on BMI-for-age cut-off point, which defined overweight as BMI between the 85th and 95th percentiles, and obesity as BMI greater than or equal to the 95th percentile. The China Obesity Task Force used the same criteria as CDC to define overweight and obesity.34 One study also classified severe obesity in children as BMI ≥ 99th percentile.35 Another possible definition is that proposed by WHO based on BMI-z scores for each specific age and sex.36 Moreover, there are other specific national definitions for each country.37-40 Therefore, definitions were not homogeneous across eligible studies since they used different standards.

2.5 Statistical analyses

All analyses were conducted using STATA software, versions 14.2 and 15 (StataCorp LP). The natural log-transformed ORs, RRs, and HRs and 95% CI comparing highest versus lowest categories of dairy product consumption were pooled using the generic inverse-variance method with fixed-effects or random-effects models when there were five or fewer or at least five comparisons, respectively. Studies using continuous risk per dose were excluded from the analyses, but their findings were described within the text. Separate meta-analyses were conducted for cross-sectional and prospective cohort studies.

For all meta-analyses, the Cochran Q statistic was used to estimate interstudy heterogeneity and it was quantified by the I2 statistic. We considered substantial heterogeneity as I2 ≥ 50% and Pheterogeneity < 0.10.

Subgroup analyses were not conduced as there were less than 10 studies per outcome available. Likewise, dose–response meta-regression analyses could not be conducted due to the few studies with applicable data available. Less than three studies reported sufficient data to perform dose–response meta-regression, and we were unable to obtain the missing information after several attempts to contact the corresponding author of these papers.

Sensitivity analyses were conducted when more than three study comparisons were available in the analyses by the removal of one study at a time from the meta-analyses and recalculating the summary risk estimates. We considered an influential study as one that changed the evidence of heterogeneity or the magnitude by more than 20%, the significance, and/or direction of the association.

Potential publication bias was not possible to be tested since less than 10 study comparisons were included in each analysis.41

3 RESULTS 3.1 Study characteristics

A total of 4367 articles were identified after a primary search of MEDLINE-PubMed and Cochrane databases as shown in Figure S1. Of these, 4290 articles were excluded according to title and abstract based on eligibility criteria. Therefore, 77 articles were collected as full texts and were further assessed for inclusion. A total of 24 articles were included in the qualitative synthesis,9, 21-31, 42-51 and 159, 22-24, 26, 28, 30, 31, 39, 40, 42, 43, 47, 48, 51 (which reported risk estimates comparing extreme categories) were included in quantitative synthesis meta-analyses (Figure S1). Two of the studies (DeBoer et al.9 and Huus et al.26) conducted both cross-sectional and prospective cohort analyses.

The characteristics of each study are presented in Table 1. This meta-analysis included a total of 101,330 subjects from different continents: 10 studies were performed in America, 6 in Europe, 7 in Asia, and 1 in Oceania. The duration of follow-up in the prospective cohort studies ranged from 1 to 3.2 years. Most of them assessed dairy exposure through FFQs or using a 24-h dietary recall and evaluated the outcome according to Cole et al. (2000)32 criteria or CDC growth charts.52 Some studies (i.e., Berkey et al.,22 Abreu et al.,42 Nasreddine et al.,28 Guo et al.,39 and Xu et al.51) stratified the analysis by age or sex. Consequently, we considered those results separately in the meta-analyses.

TABLE 1. Characteristics of studies included in the systematic review and meta-analysis Study Country Study name Population Agea (years) Dairy consumption assessment method Type of dairy products Comparisons Outcome and assessment method Follow-upa Quality b Cross-sectional studies Abreu et al. (2013)1 Portugal Azorean Physical Activity and Health Study II 1209 (503 boys and 706 girls) 15–18 FFQ Total dairy (milk, yoghurt, and cheese) T3 versus T1

Overweight/obesity

Cole et al. (2000)2 and (2007)3

NA Total milk (whole, semiskimmed, and skimmed) T3 versus T1 Total yoghurt T3 versus T1 Cheese T3 versus T1 Al-Hazzaa et al. (2012)4 Saudi Arabia Arab Teens Lifestyle (ATLS) 2845 (boys and girls) 14–19 FFQ Total milk ≥5 days/week versus <3 days/week

Overweight/obesity Cole et al. (2000)2 and

BMI > =25 kg/m2 for participants aged 18–19 years

NA Amin et al. (2008)5 Saudi Arabia - 1148 (only boys) 10–14 FFQ Total dairy (milk, cheese and other dairy) 5–6 times/week versus ≤once/week

Overweight/obesity

BMI ≥ 85th percentile

NA POOR Barba et al. (2005)6 Italy The BRAVO Project and the Gabbiano Study 884 (456 boys and 428 girls) 3–11 FFQ Whole milk ≤1/week versus ≥2/day

Overweight

Cole et al. (2000)2

NA Beck et al. (2017)7 USA Birth cohort study of Latino children and mothers 145 (71 boys and 74 girls) 3 FFQ and 24-h dietary recalls Total milk Ounce per day

Severe obesity

BMI ≥ 99th percentile8

NA Beck el al. (2014)9 USA Cohort of Mexican American children 319 (150 boys and 169 girls) 8–10 FFQ Flavored milk 240-ml servings per week Obesity CDC (2000)10 NA Whole milk 240-ml servings per week 2% milk 240-ml servings per week DeBoer et al. (2015)11 USA Early childhood longitudinal survey birth cohort 8950 (4550 boys and 4400 girls) 4 Not specified Total milk (whole, 2%, 1%, skim, and soy) ≥3 servings/day versus ≤2 servings/day

Overweight CDC (2000)10

Obesity CDC 200010

NA Govindan et al. (201)312 USA Project healthy schools 848 (girls) 10–12 24-h dietary recalls Total milk (any type) ≥2 servings/day Obesity CDC 200010 NA Guo et al. (2020)13 China The Chinese Environmental Exposure-Related Human Activity Patterns Survey-Children (CEERHAPS-C) 40,607 (19,618 boys and 20,989 girls) 6–17 Standard questionnaire administered in one-on-one interviews with the aid of standard measuring containers. Total milk (plain whole, low-fat, and skim cow's milk but not other types of milk such as goat milk or soymilk) >100 ml per month versus 0 ml per month

Overweight/obesity

China NHaFPCotPsRo15

NA POOR Huus et al. (2009)16 Sweden All babies in Southeast Sweden (ABIS) 7356 5 FFQ Total milk ≥4 times/day versus 2 times/day

Overweight/obesity

Cole et al. (2000)2

NA Cheese ≥4 times/day versus 1 to 3 times/week Cheese Daily versus 3 to 5 times/week Li et al. (2007)10 China China National Nutrition and Health Survey 6828 (3988 boys and 2840 girls) 7–17 24-h dietary recall for three consecutive days Total dairy ≥100 g/day versus <100 g/day

Overweight/obesity

Group of China Obesity Task Force (2004)17

NA Marcos-Pasero et al. (2018)18 Spain GENYAL study 201 (108 boys and 93 girls) 6–9 48-h dietary recall Total dairy (milk, yogurt, cheese, commercial milkshakes, and other preparations containing milk such as flans or ice creams) 1 serving/day

Overweight/obesity

Cole et al. (2000)2

NA Matthews et al. (2011)19 USA Child-Adolescent Blood Pressure Study 1764 (879 boys and 885 girls) 6–19 FFQ Total dairy (whole milk, chocolate milk, cottage cheese, cheeses, yogurt, pudding, ice cream, frozen yogurt, and milkshake) Q4 versus Q1

Overweight

Expert Committee Recommendations (2007)20

NA Nasreddine et al. (2014)21 Lebanon National Survey of Household living conditions 868 (439 boys and 429 girls) 6–19 24-h dietary recalls Total dairy (milk and dairies) T3 versus T1

Overweight WHO (2007)22

Obesity WHO (2007)22

NA Nilsen et al. (2017)23 Sweden Swedish WHO COSI Study 2620 (1365 boys and 1255 girls) 7–9 FFQ

Full-fat milk

(whole-fat milk, 3% fat)

≥4 days/week versus <4 days/week

Overweight Swedish national growth reference from Werner24

Obesity Swedish national growth reference from Werner24

NA Low-fat milk (skimmed/semiskimmed milk, (0.1%–0.5% fat)) ≥4 days/week versus <4 days/week O'Sullivan et al. (2015)25 Australia Western Australia Pregnancy Cohort (Raine) Study 1418 13.0–14.9 FFQ Total dairy (core dairy products + noncore dairy products ➔ butter, cream and ice cream) 1 serving/day

Overweight/obesity

Cole et al. (2000)2 and (2007)3

NA Core dairy (milk—including milk-based beverages such as smoothies, milkshakes and flavored milk, yoghurt, cheese, and custard). 1 serving/day Olivares et al. (2004)26 Chile - 1723 (927 boys and 774 girls) 8–13 FFQ Total dairy (milk and yogurt) g/day Obesity CDC (2000)10 NA