Dairy Consumption and Overweight and Obesity in Children and Adolescents: a Meta-analysis.

Searches: We will search the following electronic bibliographic databases: MEDLINE (PubMed) and The Cochrane Library (Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials.

The search strategy will include only terms relating to or describing the exposure. There will be no language restrictions.

Data extraction (selection and coding):

In the first step duplicate studies will be removed. Secondly, studies will be screened based on the title and abstract in order to exclude those that do not meet the inclusion criteria. In the last step, articles will be reviewed in full to select those that meet all the inclusion criteria.

Two reviewers (LL and MM) will independently extract relevant data and assess the risk of bias following a standardize proforma. Relevant data will include:

• Authors, journal and year of publication
• Title
• Study design
• Study name
• Country/location of study
• Sample size
• Characteristics of the subjects
• Follow-up duration (for prospective cohort studies)
• Health status
• Sources of funding
• Type of exposure and assessment method
• Outcome and assessment method
• Odd ratios, risk ratios or hazard ratios and 95%CI
• Statistical analyses

Risk of bias (quality) assessment:

Two review authors will independently assess the risk of bias in included studies by considering the following tools: The New-Castle Ottawa scale will be used to assessed the quality of the included prospective cohort studies, whereas the NIH National Heart, Lung and Blood Institute. Study quality assessment tools will be used for cross-sectional studies. Any disagreement will be solved by consensus or consulting with a third author (NB-T).

Strategy for data synthesis The meta-analysis will be conducted according to the Cochrane handbook for systematic reviews of interventions and results will be reported following the meta-analysis of observational studies in epidemiology (MOOSE) guidelines. The generic inverse-variance method with random-effects models or fixed-effects models, when at least 5 or less than 5 comparisons are available, respectively, will be used to obtain summary estimates. Inter-study heterogeneity will be assessed by the Cochrane Q statistics (p<0.10 will be considered significant) and quantified by the I2 statistic (≥ 50% will indicate substantial heterogeneity). Possible sources of heterogeneity will be explored by conducting sensitivity analyses, in which one study will be systematically removed at a time. If 10 or more comparisons are available, researchers will also explore sources of heterogeneity by a priori subgroup analyses (including, age, gender, geographical area and study quality). Meta-regression will assess the significance of categorical and continuous subgroup analyses. Linear and non-linear dose response relationship will be assessed by using generalized least squares trend (GLST) estimation models and spline curve modelling (MKSPLINE procedure), respectively. Publication bias will be assessed by visual inspection of funnel plots and the Egger and Begg tests when at least 10 trial comparisons are available.

Analysis of subgroups or subsets:

Prior subgroup analysis are planned to be conducted if at least 10 study comparisons are available. This will include: age (< or > of the median), gender (boys, girls and mixed), geographical area (America, Europe, Africa, Asia and Oceania), study quality (high vs low quality).