Obesity in young children and its relationship with diagnosis of asthma, vitamin D deficiency, iron deficiency, specific allergies and flat-footedness: A systematic review and meta-analysis

Summary There is evidence that a number of medical conditions and co-morbidities are associated with obesity in young children. This review explored whether there is evidence of associations with other conditions or co-morbidities. Observational studies of young children (mean age < 10 years) were identified using electronic searches of five databases (MEDLINE, Embase, CINAHL, AMED and SPORTDiscus). Of 27 028 studies screened, 41 (comprising 44 comparisons) met the inclusion criteria. These studies provided data on five distinct diseases/conditions: asthma (n = 16), vitamin D deficiency (n = 10), iron deficiency (n = 10), allergies (n = 4) and flat-footedness (n = 4). Thirty-two studies were appropriate for meta-analysis using random-effects models, and revealed obesity was significantly associated with having asthma (OR 1.5, 95% CI 1.3–1.7), vitamin D deficiency (OR 1.9, 95% CI 1.4–2.5) and iron deficiency (OR 2.1, 95% CI 1.4–3.2). Heterogeneity (I 2) ranged from 57% to 61%. Narrative synthesis was conducted for all studies. There was no evidence of a consistent association between obesity in young children and eczema, dermatitis or rhinitis due to the low number of studies. However, there was an association with flat-footedness. These results have implications for health policy and practice and families. Further research leading to a greater understanding of the associations identified in this review is suggested.


Funding information
Thirty-two studies were appropriate for meta-analysis using random-effects models, and revealed obesity was significantly associated with having asthma (OR 1.5, 95% CI 1.3-1.7), vitamin D deficiency (OR 1.9, 95% CI 1.4-2.5) and iron deficiency (OR 2.1, 95% CI 1.4-3.2). Heterogeneity (I 2 ) ranged from 57% to 61%. Narrative synthesis was conducted for all studies. There was no evidence of a consistent association between obesity in young children and eczema, dermatitis or rhinitis due to the low number of studies. However, there was an association with flat-footedness.
These results have implications for health policy and practice and families. Further research leading to a greater understanding of the associations identified in this review is suggested.

| INTRODUCTION
While having obesity in childhood is a known predictor of numerous health conditions in adulthood, 1,2 a growing evidence base has demonstrated the adverse health effects obesity has during childhood and adolescence. Specifically, an abundance of research has demonstrated the link between having childhood obesity and cardio-metabolic disease markers such as high cholesterol, hypertension and abnormal glucose tolerance, with children with obesity at a threefold increased risk of hypertension than children without obesity for example. [3][4][5] A recent systematic review of observational studies and randomized trials demonstrated that 5-to 15-year-old children with obesity had 7.49 mmHg higher systolic blood pressure, 0.15 mmol L −1 higher total cholesterol, 0.26 mmol L −1 higher triglycerides and significantly higher fasting insulin and insulin resistance than children without obesity. 6 The latter finding further supports recent research demonstrating the association between having childhood obesity and the development of type 2 diabetes in youth. 7 While the relationship between having childhood obesity and cardio-metabolic risk factors has been well established by published research, including several systematic reviews, the potential relationship between having obesity in childhood and other co-morbid conditions is not as clearly understood. Furthermore, a focus on these comorbidities in younger children with obesity is limited to date. Epidemiological research has demonstrated associations between childhood obesity and increased risk of asthma, sleep apnoea, vitamin D deficiency, non-alcoholic fatty liver disease, dental caries, eye disorders, atopic disease and musculoskeletal complaints among other conditions. [8][9][10][11][12][13][14][15][16] However, recent systematic reviews investigating the health impacts of having childhood obesity have not been definitive regarding these conditions. Specifically, asthma has increasingly been linked to having childhood obesity; however, a review by Pulgaron identified a number of studies in which no association between childhood obesity and asthma was reported. 17 Another systematic review and meta-analysis of 48 epidemiological studies demonstrated a weak but significant link between asthma and overweight/obesity in children. 18 However, a systematic review of 10 longitudinal studies demonstrated that children who had obesity as a child were more likely to suffer from asthma either in childhood or in adolescence, 19 a finding supported by an umbrella review of risk factors for childhood obesity. 20 There are a number of plausible explanations for the equivocal findings observed for the relationship between having some co morbidities and childhood obesity. However, one issue that is rarely discussed sufficiently in the literature is the methodological issues that may arise from the practice of combining children with overweight and obesity during participant recruitment and subsequent analyses.
Many observational studies grouped children with obesity and overweight as a combined exposure variable, instead of considering both conditions as two distinct groups. Participants with overweight and obesity would rarely be combined in adult studies of co-morbidities but are routinely combined in paediatric studies. Possible reasons for this may be that there is relatively low prevalence of children with obesity in some populations historically, in comparison with overweight; therefore, recruiting an adequately powered sample to measure the outcomes of interest may be more difficult if the inclusion criteria are limited to individuals with obesity. While overweight without obesity in childhood is associated with numerous health conditions, 21 the grouping of participants with overweight and obesity together without appropriate stratification or subgroup analysis may dilute the real relationship between true obesity and the outcome of interest, which in some cases is considerably more pronounced than the effects of having overweight alone. 6,22 Such issues are evident from studies that have stratified by body mass index (BMI) percentile, where the risk of co-morbidity rises as BMI increases, or prevalence is higher in groups with obesity compared with groups with overweight as defined by standardized cut-offs. 11,22 Another potential source of inconsistent findings in research on the co-morbidities of child and adolescent obesity may be the numerous different age ranges used to define 'childhood' among the published literature. While some studies will distinguish childhood from adolescence using either internationally recognized categories, or researcher-defined cut-offs, others will take an all-encompassing approach and group all participants together from early childhood up to late adolescence. An issue with this approach is that the physiological, behavioural and metabolic differences between a young child and an older adolescent may have a significant influence on the outcome of interest. 23 The aim of this current systematic review was therefore to update and synthesize the evidence base on the physical comorbidities of childhood obesity, focusing specifically on the impact of obesity (rather than overweight) in children under 10 years old and excluding co-morbidities of childhood obesity that have been wellestablished by previous systematic reviews and meta-analyses.

| METHODS
This systematic review was prospectively registered with PROSPEROregistration number: CRD42018079387. The Preferred Reporting Items for Systematic reviews and Meta-analyses (PRISMA) checklist was used to inform the conduct and reporting of this review. The methods used were guided by two expert Cochrane reviewers (AM and CS).

| Inclusion and exclusion criteria
Observational studies of a cross-sectional, longitudinal or casecontrol design were included if they reported one measure of adiposity (e.g., BMI) in childhood (WHO definition; ≤9.9 years) and measured at least one physical health outcome in either childhood or adolescence (age 0-19 years). Studies that included children older than 10 years of age were included if the mean age of the overall sample was ≤9.9 years. Therefore, the overall age range of children within the included studies was 2-19 years (a number of studies stratified by age groups). Included studies were required to contain both groups with obesity and groups without obesity as a comparison within the sample and be published in English. Studies were also required to include both children with and without the respective co-morbid conditions. Studies that only recruited children with co-morbidities were excluded.
A number of co-morbid conditions including cardiovascular disease markers, diabetes, dental carries, sleep apnoea and metabolic syndrome were not incorporated into the search strategy for this review as the relationship is either well established, or a recent goodquality systematic review has been published relating to these conditions. [24][25][26] Therefore, any studies reporting exclusively on these outcomes concerning childhood obesity were excluded. However, all other potential co-morbidities of childhood obesity were deemed eligible for inclusion in the review providing they met the inclusion criteria. Studies were excluded if • The study population had a prior health condition that would limit generalizability of the study findings (such as children born preterm or who had a disability).
• Studies that exclusively recruited children who had the outcome of interest (co-morbidity) without an unaffected comparison group (as this would not allow for the assessment of weight status on the incidence of the condition).
• All participants in the study population had the exposure (obesity) without a comparison group that did not have obesity.
• Recognized definitions of obesity were not used or reported in the study (e.g., ≥95th percentile for age and gender using national or international growth charts).
• Children with obesity and overweight were combined without stratification during analysis.
• The mean age of the sample was over 9.9 years at the time of obesity exposure.
• Anthropometry or presence of morbidities was obtained through parental or self-report (parental reporting of doctor diagnosis of condition was included). where broad headings were initially used (e.g., 'health,' 'comorbid* or co-morbid*') before becoming more focused on specific health conditions that were identified as potential co-morbidities of obesity from the published literature (e.g., 'pes planus,' 'asthma,' 'musculoskeletal diseases'). Where possible, MESH-headings were used in addition to free-text words to account for databases without the MESH function and specific conditions not covered under MESH-headings. The search was restricted to children, human subjects and primary research studies.

| Search strategy and study selection
Identified studies were independently screened for eligibility by two reviewers (SM and JG), initially by title and abstract (a random 20% of identified papers were double-screened), before double fulltext screening was conducted for all papers that were deemed eligible after title and abstract screening. In addition to the two reviewers discussing any inconsistencies in identification, where consensus could not be reached, a third reviewer (JRR) was included in the discussion to support resolution of the decision. Following full-text screening, eligible studies had their reference lists searched for potentially relevant studies, as did relevant published systematic reviews. Relevant titles were exported to an excel spreadsheet and screened using the same methods as initially applied to the articles identified through database searches.

| Quality assessment
The quality of eligible studies was assessed using an adapted version of the Newcastle-Ottawa scale (NOS) previously used in a review by Herzog et al. 27 The tool has been adapted to assess the quality of cross-sectional studies in addition to cohort and case-control studies.
The scale assesses studies against the following criteria: (i) selection of the sample; (ii) comparability of the sample/participants; (iii) assessment of exposure and outcomes. Stars are awarded for high-quality aspects of each study against the three aforementioned criteria, with a total of nine stars available for case-control and cohort studies and eight stars available for cross-sectional designs. Studies awarded less than five stars were classed as having a high risk of bias, while an award of ≥5 stars indicated low risk of bias. The scale has been specifically designed for non-randomized studies and does not report summary scores, which have been shown to be unreliable. 28 Two researchers independently appraised each study before discussing any disagreements. No formal overall assessment of the quality of evidence was undertaken, as this review only included observational studies, which are deemed by GRADE 29 to be of either low to very low quality of evidence.

| Data extraction
The following data were extracted from each study: authors, publication year, study design, study population characteristics (sample size, geographic location, age ranges, % female), method of recruitment, exposure (exposure assessment method, the definition of obesity), outcomes (outcome assessment method, the definition of outcome, number of outcomes), method of analysis, reported effect estimates (odds ratios, relative risk, proportions, prevalence and relevant confidence intervals), level of significance reported and confounders controlled for in the analysis. We also planned to collect any data on the socio-economic status (SES) of children under study, given the known association between some medical conditions and SES in adults, and this field was included in our data extraction form. Data were extracted by two researchers independently using the pre-designed form, which was piloted on a random sample of studies prior to full data extraction commencing.

| Data synthesis and meta-analysis
Both meta-analysis and narrative synthesis were performed for this review. Due to the inconsistent nature of information reported in a number of included studies, coupled with considerable study heterogeneity, it was not possible to include all the studies in the meta-analyses. Where narrative synthesis was adopted, recommendations outlined in the Cochrane handbook of systematic reviews were followed, 30 whereby the characteristics of each study were summarized in terms of the study design, risk of bias and study context for each outcome. This was then followed by an exploration of the similarities and differences between each studies' findings.
Random-effects meta-analysis was conducted on three outcomes in this review: asthma, vitamin D deficiency and iron deficiency. Odds ratios and corresponding 95% confidence intervals were collated from studies reporting these results. For studies that did not report odds ratios, information was collected on the number of children with obesity versus children without obesity, in addition to the number who presented with the outcome of interest within these two exposure groups. Pooled odds ratios and 95% confidence intervals were generated using random-effects method on MetaXL meta-analysis software (Version 5.3; EpiGear International Pty Ltd). Forest plots were generated for each outcome, and funnel plots were used to visually assess publication bias. Heterogeneity and inconsistency were assessed using Cochran's Q statistical test, with the inconsistency test (I 2 > 50%) used to indicate moderate heterogeneity.
Where a significant proportion of the studies were of high risk of bias, a sensitivity analysis was conducted to assess whether their removal from the model significantly affected the overall result. For all analyses, the level of significance was set at ≤.05.

| Description of included studies
Of the 27 028 studies identified following database searches and deduplication, 41 met the inclusion criteria ( Figure 1). These studies presented results investigating relationships between childhood obesity and five distinct health outcomes: asthma (n = 16), 31-45 vitamin D F I G U R E 1 PRISMA flow diagram for study identification and inclusion deficiency (n = 10), 46-55 iron deficiency (n = 10), 34,56-64 flat-footedness/pes planus (n = 4) 65-68 and allergies (n = 4). 34,39,69,70 Two of the studies identified reported results for more than one of the outcomes.
Although one of the original aims of the review was to investigate obesity in children under <10 years and co-morbidity in later childhood or adolescence, we did not identify any longitudinal studies that followed young children into adolescence. Therefore, all analyses investigated the associations between having obesity and co-morbid conditions during childhood defined as under 10 years of age.

| Study characteristics
Study characteristics are summarized in Tables 1-5. The majority of included studies were cross-sectional studies (n = 29), followed by case-control studies (n = 8). Four studies were longitudinal, three being prospective cohort studies, and one a Mendelian randomization study. Studies varied considerably by sample size, from a case-control study with 100 participants to a repeat cross-sectional study totalling 36 152 participants. All studies involved the objective measurement of anthropometry by trained practitioners. All outcomes were objectively measured using established protocols, with the exception of asthma and allergies, for which the majority of studies employed valid diagnostic survey methods to obtain confirmation via parental report of diagnosis by a health professional.

| Study quality
Overall, cohort studies were rated as having a lower risk of bias than the case-control studies, with mean NOS scores of 7.4/9 and 6.3/9, respectively (a higher score equates to a lower risk of bias). Crosssectional studies had a mean score of 6.3 out of a possible 8 stars. In general, studies of all designs did not adequately describe or justify sample sizes or demonstrate the representativeness of the sample to the general population. Because of the stringent inclusion criteria adopted for this review, the included studies all scored highly on the NOS items pertaining to 'assessment of exposures and outcomes.' Six studies received a rating of below 5 stars (four cross-sectional and two case-control) and were deemed to have a high risk of bias. The NOS score and corresponding risk of bias for each study are summarized in Tables 1-5.

| Association between childhood obesity and asthma
Fifteen studies were grouped comparing the odds of asthma diagnosis between children with and without obesity. [31][32][33][34][35][36][37][38][39][40][41][42][43][44][45]71 Six studies presented results separately for different subgroups within the study sample (i.e., by age group and ethnicity), 32,34,38,39,42,44 and these results are presented separately in the forest plot output (Figure 2). Additionally, subgroup analysis of four studies that presented results separately for boys and girls is also included in the forest plot. The metaanalysis demonstrated that having childhood obesity significantly increased the odds of asthma diagnosis by over 50% in comparison with children without obesity (OR 1.5; 95% CI 1.3-1.7). Inconsistency was moderate (I 2 = 57%). In subgroup analysis by sex, boys showed higher odds than girls for having asthma and obesity (OR 2.0; 95% CI 1.4-2.9 and OR 1.6; 95% CI 1.2-2.2, respectively). However, this finding was not statistically significant (p > .05). Two of the studies were assessed as having a high risk of bias. 31,33 However, removal of these studies from the model in a sensitivity analysis did not lead to a statistically significant change in the pooled result (OR 1.5; 95% CI 1.3-1.7). No publication bias was indicated by the funnel plot ( Figure 3).
One additional study met the inclusion criteria for the review but was not appropriate for inclusion in the meta-analysis, as it was a large prospective cohort study, 71 which presented time-toevent analysis that would be poorly interpreted by conversion to odds. The study was evaluated as having a low risk of bias and reported that having moderate obesity (≥95th centile) and extreme obesity (≥99th centile) both significantly increased the risk of asthma diagnosis. 71

| Association between childhood obesity and vitamin D deficiency
Nine separate studies were included in the meta-analysis of vitamin D deficiency and obesity, [46][47][48][50][51][52][53][54][55] resulting in a pooled odds ratio of 1.9 (95% CI 1.4-2.5) (Figure 4). The studies showed moderate heterogeneity (I 2 = 58%). The majority of studies defined deficiency as <20 ng ml −1 , and one study reported separate odds ratios for deficiency and severe deficiency, which have been added to the forest plot separately. 50 Studies that reported serum levels in nmol L −1 were less uniform, with deficiency cut-offs ranging from <17.5 to <30 nmol L −151,55 (see Table 2). Two of the studies included in the meta-analysis had a high risk of bias. 46,48 When these studies were removed during sensitivity analysis, the pooled effect size was reduced to OR 1.7 (95% CI 1.3-2.3).
One additional study 49 reported insufficient data to allow for pooling within the meta-analysis and was judged as having a high risk of bias. 49 This study reported significantly lower vitamin D levels in children with obesity, due to its small sample size, this study would not have significantly influenced the overall effect size in pooled analysis had it been included.

| Association between childhood obesity and iron deficiency
Ten studies investigated the relationship between childhood obesity and iron deficiency 34 Low risk of bias (7/8) for meta-analysis ( Figure 5). Two of these studies conducted separate analyses by sex, 34 One case-control study 60 assessed as having a low risk of bias was not appropriate for meta-analysis due to insufficient reporting of data necessary for calculation of odds and found children with obesity to have significantly different markers of iron deficiency than the control group. Specifically, children with obesity had significantly lower iron, transferrin saturation and total-iron binding capacity along with higher ferritin, soluble transferrin receptors and hepcidin-25 than children of normal weight.

| Association between childhood obesity and pes planus (flat-footedness)
Four studies investigated the relationship between childhood obesity and having flat-footedness. [65][66][67][68] Of these, one was a longitudinal study, 65 two were of a cross-sectional design 66,67 and one was a case control study. 68 All four studies were of low risk of bias and reported a statistically significant association between having flat-footedness and obesity. One study 66 investigated both bilateral and unilateral flat-footedness but only found having obesity to significantly increase the odds of the bilateral condition (OR 1.9; 95% CI 1.2-2.9), while the remaining three studies investigated bilateral flat-footedness only. All four studies were assessed as having a low risk of bias following quality assessment (Table 4).

| Association between childhood obesity and allergies
Four studies assessed the relationship between childhood obesity and allergic conditions. 34,39,69,70 Two distinct conditions were investigated within the studies; all four studies assessed eczema/dermatitis, and one study also included rhinitis as an outcome. 39 Additionally, one study reported results from the skin prick test. 70 Three of these studies had a cross-sectional design, and one was a case-control study. 69 All four studies had a low risk of bias (Table 5). Three of the four studies found having obesity to increase the odds of eczema/dermatitis diagnosis; however, these effects were small to moderate. 39,69,70 One study reported no association between eczema and obesity/severe obesity. 34 Having obesity was found to slightly increase the odds of rhinitis diagnosis in one study (OR 1.3; 95% CI 1.0-1.7) when the sample was analysed collectively. However, a differential effect by sex was reported, as the association was only evident in girls and not boys. 39

| DISCUSSION
This systematic review and meta-analysis investigated associations between obesity in young children and multiple co-morbid conditions.
Though this topic has been studied through both primary research and recent systematic reviews, 17,18,[72][73][74][75][76][77][78] we investigated obesity as a distinct condition from overweight. This is in contrast to similar systematic reviews in the subject area, which have included studies that combine individuals with overweight or obesity, or do not stratify by weight status in the analysis. [17][18][19]73,78 We also used more stringent inclusion criteria to define childhood as children under 10 years, potentially offsetting the physiological, cultural and behavioural effects that later childhood/adolescence can have on obesity comorbidities. 79,80 F I G U R E 2 Forest plot for random effects meta-analysis of studies investigating relationships between childhood obesity and asthma (OR = odds ratio; CI = confidence interval) F I G U R E 3 Funnel plot for studies reporting on the relationship between obesity and asthma in Meta-analysis. (ES = effect size) The results of our review offer a number of important findings.
Firstly, the meta-analysis of childhood obesity and asthma diagnosis appears to support previous results from systematic reviews in this area, 18,77,78 while also further distinguishing the effects of having obesity considered explicitly from overweight. Chen et al. 77 reported a significantly higher risk of incident asthma among children and adolescents with obesity compared with children without obesity (relative risk 2.02; 95% CI 1.16-3.50), while a narrative synthesis by Papoutsakis et al. 18 concluded that there was a clear relationship between childhood obesity and asthma incidence. Our finding that having obesity at a young age increased the odds of asthma by over 50% indicates that there is a possible relationship between the two conditions. However, we did not identify any longitudinal studies that were appropriate for meta-analysis based on our inclusion criteria, meaning the cross-sectional data that our findings are based on cannot offer any indication of a causal link between obesity and asthma.
One cohort study included in our narrative synthesis found that not only did having a higher BMI predispose children to subsequent asthma development, but children with both asthma and obesity or overweight were also more likely to develop a severe asthma phenotype than healthy weight children with asthma. 71 A meta-analysis of six longitudinal studies also found that asthma risk increased by 35% among children with obesity/overweight. 78 87 It is also important to consider that vitamin D deficiency has been shown to increase asthma severity, which may indicate each condition may mediate any relationship with obesity. 88 Despite this, few studies included in this review controlled for this potential confounding (Tables 1 and 2). Furthermore, it is theorized that having obesity may impair the bioavailability of vitamin D for bloodstream absorption, as it is fat-soluble, and instead stored in adipose tissue reservoirs, 89 further highlighting the complexity of the relationship between the two conditions. This review also found that childhood obesity increased the odds of having iron deficiency, which to our knowledge is only the second such meta-analysis to demonstrate this relationship and the first to do so exclusively in children aged <10 years. 76 An important observation that applies to both vitamin D and iron deficiency is that they are both nutritional deficiencies. It could therefore be that causation may be related to diet quality, as children with obesity have been shown to have poorer nutritional intake (lower nutrient density, consuming less iron-rich foods) than children with normal weight in epidemiological studies. 90 A small number of studies included in our review controlled for diet in their analysis, 47,49,58 still finding the conditions to be associated with obesity. However, the majority of studies concerning vitamin D and iron status did not include diet/nutrient intake as a covariate, which may have confounded the results obtained for these studies. In the case of vitamin D, this can be extended to include time spent outdoors (or physical activity as a proxy measure of ultraviolet light exposure), as it has been demonstrated that children with obesity spend less time in outdoor play and longer periods sedentary indoors. 91 With vitamin D levels mediated by sunlight exposure, 47 this could potentially explain the differences observed in children with obesity from a behavioural perspective. Therefore, the results of a number of included studies that did not control for these covariates should be interpreted with caution. 46,48,[50][51][52][53]55 In the case of iron deficiency, while mechanisms explaining effects of obesity on the condition are not fully understood, individuals with obesity have both increased iron requirements (secondary to increased blood volume) and reduced iron absorption (secondary to increased inflammation). 92 Interventions to reduce weight status in children with vitamin D and iron deficiency would therefore enhance understanding of the causal effects of having obesity in these conditions.
While the inclusion criteria adopted for this review are a strength of this study, there are a number of limitations that should be considered when interpreting the findings. We decided against conducting an overall assessment of the quality of evidence for each outcome by using GRADE assessment criteria. 29  reported, in addition to the novel finding that iron deficiency is a potential co-morbidity of childhood obesity. Additionally, it appears that a better understanding of any important inequalities (by SES) in the relationship between obesity and health conditions in young children is needed to help support policy and practice with regard to obesity prevention in children. 93 Healthcare professionals may find our results helpful when treating pediatric patients with obesity, in terms of additional assessment and consideration for the co-morbid conditions identified and investigated in this review. The potential for obesity to cause harm as early as childhood is apparent, and efforts to prevent obesity in the early years could, in turn, alleviate the health burden of conditions associated with having excess weight in childhood.