We provided a condensed form of the protocol for systematic reviews, adhering to the International Prospective Register of Systematic Reviews (PROSPERO) [CRD42023402162]. The results are reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines (22).

The PRESS Checklist was used to build the search strategy, which was later adapted for all databases (23). No limitations were imposed in terms of date or language of publication. On February 13, 2023, a simultaneous systematic search was performed across PubMed, Scopus, Embase, LILACS, Ovid-Medline, and Web of Science databases. Furthermore, the reference lists of the included studies and preprint databases were manually screened. The full search strategy is provided in Supplementary Table S1. The research question was based on the Population, Exposure, Comparison, and Outcome (PECO) strategy: Do children and adolescents (P) with obesity (E) have lower levels of vitamin B12, folate, and homocysteine (O) compared to children and adolescents without obesity (C)?

The research question was based on Population (Children and adolescents), Exposure (Obesity), Comparison (no obesity), and Outcome (Vitamin B12, folate, and Hcy) strategy.

Cross-sectional, case–control, or cohort studies that assessed the association between vitamin B12, folate, or Hcy levels with obesity in children/adolescents were eligible for inclusion. The exclusion criteria were as follows: duplicate publications, conference abstracts, scoping reviews, systematic reviews, randomized controlled trial (RCT), and narrative reviews. Among the mentioned, RCT were excluded because they are intervention studies that aim to assess a diseased population; whereas observational studies allow us to study the levels of vitamin B12, folate and Hcy in both healthy and diseased populations.

The articles retrieved from the databases were transferred to a software for data management called Rayyan © (24). After the elimination of duplicates, four authors individually screened the titles and abstracts of all articles against the selection criteria. Subsequently, full texts of the short-listed articles were independently reviewed by two authors. Any article that did not meet the entire set of selection criteria was excluded from the review. For articles with missing information, the authors were contacted. Any conflict of opinion or discrepancy in any of the phases was resolved by consensus.

The selection process was depicted using the PRISMA flowchart. A standardized data collection sheet developed in Google Sheets© was used for data extraction. Data pertaining to the following variables were independently extracted by two authors: first author, study location, publication date, study design, sample size, age and sex distribution of the study population, definition of obesity, number of patients with obesity, vitamin B12 levels in patients with and without obesity, folate levels in patients with and without obesity, Hcy levels in patients with and without obesity, and assay technique. Obesity was defined based on the parameters established by each author of the included articles. The definitions of obesity are presented in Supplementary Table S2.

The risk of bias was evaluated independently by two authors. For cohort and case–control studies, the Newcastle-Ottawa Scale (NOS) was employed, while an adaptation of the NOS for cross-sectional studies (NOS-CS) was used for cross-sectional studies. The NOS assesses the methodological quality (risk of bias) of a study based on three main aspects: the selection of study groups, the comparability of the groups, and the assessment of outcomes or exposures. The NOS consists of eight items (seven items if it is the NOS-CS), with each item scored up to one star, except for comparability, which can be scored up to two stars. A score of ≥7 stars was regarded as a low risk of bias (high methodological quality), while a score of <7 stars was deemed a high risk of bias (low methodological quality). The effect of potential publication bias on the results of the meta-analysis was evaluated using funnel plots, the Egger test, and the trim-and-fill method (25). For funnel plots, the minimum number of studies required was 10.

Statistical analysis was performed using STATA 17.0©. Standardized mean difference (SMD) with the corresponding 95% confidence interval (CI) was used as the only effect size. Median values and their interquartile ranges were converted into mean and the corresponding standard deviation (SD) using Hozo’s method (26). For variables with the standard errors (SE) reported, SD was determined using the following equation: SE × √ (sample size). In addition, the natural logarithm of the odds ratio (lnOR) and its standard error were transformed into SMD and its 95% CI using Chinn’s method (27). A random-effects meta-analysis was performed using the Sidik-Jonkman method and corrected 95%CI using the truncated Knapp-Hartung standard error (28, 29). The Cochran’s Q test and the I² statistic were employed to evaluate between-study heterogeneity; high heterogeneity was defined as I² ≥ 60% and a p-value <0.05. Subgroup analyses were performed by study design and assay method. A sensitivity analysis was conducted according to risk of bias and obesity definition.

The systematic literature search yielded 1,524 records; after the elimination of duplicates, 682 records remained. After the screening of titles/abstracts, 132 articles were selected for full-text review. Finally, only 20 records complied with all the selection criteria and were included in this systematic review (30–49). The selection process is summarized in the PRISMA flow diagram (Figure 1).

The characteristics of the included studies are summarized in Table 1. A total of 20 studies (5 case–control, 14 cross-sectional, and one cohort study) that assessed the association of vitamin B12, folate, and Hcy levels with obesity in children and adolescents were included. These studies were conducted between 2004 and 2022 in various countries: Turkey (6 studies), India (3 studies), Spain (2 studies), Israel (2 studies), United States (1 study), South Korea (1 study), Egypt (1 study), China (1 study), Brazil (1 study), Romania (1 study), and Tunisia (1 study). The combined total study population in the 20 studies was 7,791. Only 12 studies (n = 4,066) reported the number of male (n = 1834) and female (n = 2,232) participants included.

The methods used to measure Hcy, vitamin B12, and folate levels were Chemiluminescence Immunoassay (CLIA), Electrochemiluminescence Immunoassay (ECLIA), Enzyme-linked Immunosorbent Assay (ELISA), Fluorescence Polarization Immunoassay (FPI), High-Performance Liquid Chromatography (HPLC), Radioimmunoassay (RIA), and Enzyme immunoassays (EIA).

The risk of bias assessment of the included studies using the NOS and the NOS-CS revealed that 3 studies had a high risk of bias (low methodological quality) while the remaining 17 articles had a low risk of bias (high methodological quality; Supplementary Table S3).

These differences were evaluated in 10 studies (n = 6,295). No significant difference was observed in the vitamin B12 levels between children/adolescents with and without obesity (SMD: −0.26; 95% CI: −0.52 to 0.00; p > 0.05, I² = 75.15%; Figure 2). Subgroup analyses performed according to study design (Supplementary Figure S1), assay method (Supplementary Figure S2), and continents (Supplementary Figure S3) revealed no significant differences in these subgroups. No high heterogeneity (I² ≥ 60%) was found in the subgroups that evaluated B12 using RIA or in case–control studies. In the sensitivity analysis according to risk of bias (Supplementary Figure S4), the lack of significant difference and high heterogeneity was maintained (SMD: −0.26; 95% CI: −0.59 to 0.07; p > 0.05, I² = 73.97%).

The funnel plot (Supplementary Figure S5) and the Egger test (p < 0.1) indicated potential publication bias, which was corrected by the trim-and-fill method (SMD = −0.42, 95% CI −0.69 to −0.14; Supplementary Figure S6).

These differences were evaluated in 07 studies (n = 3,430). No significant difference was observed in the folate levels between children/adolescents with and without obesity (SMD: −0.12; 95% CI: −0.29 to 0.06; p > 0.05, I² = 19.6%; Figure 3). Subgroup analyses performed according to study design (Supplementary Figure S7), assay method (Supplementary Figure S8), and continents (Supplementary Figure S9) revealed no significant differences in these subgroups. No high heterogeneity (I² ≥ 60%) was found in the subgroups that evaluated folate using RIA, as well as in case–control studies, cross-sectional studies, and studies conducted in Asia. In the sensitivity analysis according to risk of bias (Supplementary Figure S10), the lack of significant difference was maintained (SMD: −0.13; 95% CI: −0.42 to 0.15; p > 0.05, I² = 21.97%). The Egger test indicated no significant publication bias (p = 0.48).

These differences were evaluated in 14 studies (n = 1704). Children and adolescents with obesity had significantly higher Hcy levels than those without obesity (SMD: 0.77; 95% CI: 0.39 to 1.14; p < 0.001, I² = 86.4%; Figure 4). Subgroup analyses were performed according to study design (Supplementary Figure S11), assay method (Supplementary Figure S12), and continents (Supplementary Figure S13). Significant differences were only observed in the subgroups that evaluated Hcy with CLIA, cross-sectional studies, and studies conducted in Asia. The only subgroup that did not exhibit high heterogeneity was the one that evaluated Hcy using CLIA. The sensitivity analysis according to risk of bias (Supplementary Figure S14) maintained a significant difference and high heterogeneity (SMD: 0.86; 95% CI: 0.44 to 1.27; p < 0.001, I² = 85.5%). Similarly, in the sensitivity analysis according to the definition of obesity (Supplementary Figure S15), the significant difference and high heterogeneity were maintained (SMD: 0.65; 95% CI: 0.19 to 1.12; p < 0.05, I² = 87.58%), despite the exclusion of studies that did not report the definition for obesity (42, 46, 47), as well as the study (49) that defined obesity as a BMI > 90th percentile, due to the risk of including overweight participants. The funnel plot (Supplementary Figure S16) and the Egger test revealed no significant publication bias (p = 0.13).

This study has some limitations. First, differences with respect to the study population (e.g., age, sex, ethnicity, and underlying health conditions) and study characteristics might have led to substantial heterogeneity across the included studies. However, subgroup and sensitivity analyses were performed to explore potential sources of heterogeneity. Second, the pooled estimates were based on crude measures of effect size, which may have potentially been affected by non-controlled confounders. Third, although there is at least one included study per continent, the generalizability of our findings is limited. Fourth, the sensitivity and specificity of optimal cut-off levels of the biomarkers assessed have not been reported. Further studies should establish these cut-off levels to evaluate the prognostic value of these biomarkers in children and adolescents with obesity. Nonetheless, our study has several strengths. Multiple databases were searched, allowing a comprehensive synthesis of the available literature. A substantial number of participants were included, which ensured adequate statistical power.