Childhood body mass index and subsequent physician-diagnosed asthma: a systematic review and meta-analysis of prospective cohort studies

Background Childhood asthma and obesity prevalence have increased in recent years suggesting a potential association. However, the direction of any association is poorly understood and the potential causal-relationship is unknown. Methods We examined the association between overweight/obesity, defined by body mass index (BMI) <18 years of age, and subsequent physician-diagnosed incident asthma at least one year after BMI assessment. We sought to explore potential effect modification by sex. PubMed and Embase were searched using keywords and restricted to subjects aged 0–18 years. There were no date or language restrictions. From each study we extracted: authors, publication date, location, overweight/obesity definitions, asthma definitions, number of participants, recruitment duration, description of cohort, follow-up time, adjusted effect estimates (with 95% CI) and estimates of subgroup analysis. Results Six prospective cohort studies which focused on children <18 years of age met criteria for inclusion. The combined risk ratio (RR) of overweight was associated with asthma (RR = 1.35; 95% CI = 1.15, 1.58). In boys, the combined RR of overweight on asthma was significant (RR = 1.41; 95% CI = 1.05, 1.88). For girls, when BMI was defined by Z-score, the combined RR of overweight on asthma was also significant (RR = 1.19; 95% CI = 1.06, 1.34). The combined risk ratio (RR) of obesity was associated with asthma in both boys and girls (RR = 1.50; 95% CI = 1.22, 1.83), in boys only (RR = 1.40; 95% CI = 1.01, 1.93) and in girls only (RR = 1.53; 95% CI = 1.09, 2.14). Conclusions Overweight and, especially, obese children are at increased risk of subsequent physician diagnosed asthma in comparison to normal weight children. Except for sex, no studies reported any other potential effect modifiers. The observed sex effects were inconsistent.


Background
Both asthma and obesity have increased in prevalence in recent years which suggests a possible association between the two conditions [1]. Childhood asthma/wheeze and obesity, measured by body mass index (BMI), have been linked in cross-sectional, case-control, and prospective epidemiologic studies. However, it is unknown whether the obesity and childhood physician-diagnosed asthma epidemics are causally related [2,3] and, if so, what is the direction of any potential association. Does obesity increase the risk for physician-diagnosed asthma, physician-diagnosed asthma increase the risk for obesity, or both? Obesity was implicated as a risk factor for asthma and wheeze in a meta-analysis of adult studies [4]. However, the sex-specific association between obesity, measured by age-and sex-specific BMI, and physician-diagnosed asthma in children less than 18 years of age is unclear.
Since 1980, rates of overweight (typically classified using International Obesity Task Force sex-and agespecific BMI standards [5]) have tripled among U.S. adolescents [6]. Epidemiological evidence suggests that the prevalence of asthma, a common childhood chronic illness [7,8], has also tripled since 1980 [9]. Asthma is characterized by airway inflammation, enhanced airway responsiveness to a variety of environmental stimuli, and reversible airway obstruction and is associated with significant medical and social morbidity [10,11]. Obesity in children, also a condition associated with significant medical and social morbidity, can lead to respiratory problems and has been linked to bronchial hyperactivity and reduced chest wall compliance [12]. Energy intake and energy expenditure, including appetite and metabolism are regulated by pro-inflammatory mediators that are derived from adipose tissue. These adipose-derived hormones include adiponectin [13], leptin [14,15], and, the cytokine-rich protein, resistin [16], among others. Adiponectin is exclusively secreted from adipose tissue and modulates both glucose regulation and fatty acid oxidation [13]. In adults, adiponectin is inversely correlated with body fat percentage but this relationship is not clear in children [17]. Leptin has a pro-inflammatory effect as it is responds specifically to adipose-derived inflammatory cytokines. These cytokines include resistin which affects insulin resistance, inflammation and energy homeostasis [16]. The pro-inflammatory role of leptin and resistin support the hypothesis that obesity may lead to new-onset (incident) asthma, which is also an inflammatory condition [14,15]. Some researchers hypothesize that asthma and obesity may have additive synergistic proinflammatory effects [18].
The prospective relationship between obesity, measured by body mass index (BMI), and asthma among children has not been evaluated in a systematic review. A metaanalysis of the prospective epidemiological data on the relationship between BMI and incident asthma among adults has been published [4]. The results suggest that the risk of adult incident-asthma increases with increasing BMI among both men and women although it is unclear how each study included in the meta-analysis defined incident asthma. Another meta-analysis of the effect of high birthweight (defined as ≥3.8 kg) or overweight during childhood (defined as BMI ≥85th percentile for age and sex, or Ponderal index ≥2.5 g/cm 3 or ≥27 kg/m 3 ) on future risk of asthma or wheeze reported that there is an increased risk of asthma and wheeze among those children with high birthweight (RR 1.2; 95% CI 1.1, 1.3) and among those with a high childhood body weight (RR 1.5; 95% CI 1.2, 1.8) [19]. A literature review focusing on the association between childhood nutritional status, defined by various measures of BMI, and risk of asthma or wheeze in adolescence concluded that childhood obesity may precede asthma and/or wheeze in adolescents [20]. This study reported no cumulative effect estimates of overweight/ obesity on asthma. Since body weight alone, in contrast to BMI, does not take into account height and the presence of wheeze is not necessarily indicative of incident physiciandiagnosed asthma, the focus of this updated meta-analysis is to assess the effect of overweight/obesity (defined by International Obesity Task Force sex-and age-specific BMI standards) during childhood on physician-diagnosed incident asthma at least one year after BMI measurement.
This systematic review focuses on the evidence for an association between childhood overweight/obesity and subsequent physician-diagnosed asthma. In addition, the review assesses the potential interaction between sex and BMI on asthma risk [21] since conflicting sex-specific associations have previously been reported [22,23]. We hypothesize that overweight/obese children are at increased risk of physician-diagnosed asthma and that this relationship differs by sex.
The objectives of this review were to: 1. Determine whether or not there is an association between overweight/obesity, as defined by BMI, before 18 years of age and subsequent physiciandiagnosed incident asthma. 2. Assess whether or not there is effect modification on the risk of developing asthma among overweight/ obese children by sex.

Study selection
The following inclusion and exclusion criteria were used to screen potentially eligible studies: 1) The primary objective was to investigate the relationship between overweight/obesity, as measured by BMI, and new-onset asthma in children and/or adolescents, at least 1 year after BMI measurement. 2) To establish temporality, the study used a prospective cohort design. 3) There were no language or date restrictions.

Types of participants
Identified study subjects under the age of 18 years, of both sexes and all ethnic groups, were included for review. Studies examining overweight/obesity as defined by child or adolescent BMI or BMI applied to pediatric growth charts were included for review. For BMI assessment, height and weight must have been measured by research staff, not collected from existing medical records, and then BMI calculated based on those measurements.

Types of exposure measures
BMI must have been categorized into age-and sexspecific Z-scores (continuous variable) and/or categorical overweight/obese variables. As the distribution of BMI at each age is skewed and variance increases with age, continuous age and sex-specific BMI Z-scores were used to represent each child's sex-specific, age-adjusted level of adiposity [24]. Likewise, categorical overweight/ obese variables based on age-and sex-specific percentiles on the CDC or International Obesity Task Force BMI growth charts can also be used [5,25].
Accordingly, overweight was defined as BMI at or above the 85th percentile and lower than the 95th percentile and obesity as BMI at or above the 95th percentile, for children of the same age and sex [26].

Types of outcome measures
Included studies must have used an asthma outcome definition that adhered to accepted diagnostic guidelines for childhood asthma [27]. Asthmatic subjects must have had new physician-diagnosed asthma at the time of study outcome assessment. Prescribed asthma medication usage was an acceptable outcome measure as it was assumed that physician-diagnosed asthma must precede use of prescription asthma medications. Over-the-counter asthma medication use was not assessed as information was unavailable in the included studies.

Primary outcomes
The primary outcome was the development of new physician-diagnosed asthma at least 1 year after height and weight were measured.

Data sources
PubMed and Embase were searched in November 2012. There were no restrictions on publication date or language. Both were searched using the terms "overweight" and "asthma" or "obesity" and "asthma" or "body mass index" and "asthma" or "body weight" and "asthma." The search was restricted to subjects aged 0-18 years. Reference lists of relevant studies identified in the electronic search were also checked to identify other potentially relevant studies.

Data extraction and analysis
All identified studies were loaded into EndNote Web (2011). A single reviewer examined the titles for relevant articles and duplicates were identified and removed. Abstracts for remaining relevant studies were reviewed. Obviously ineligible studies (studies in adults or not involving asthma or obesity) were excluded. If it was not clear whether the study met the inclusion criteria, then the full text of the article was assessed. All studies excluded and reasons for exclusion were documented. Uncertainties were jointly discussed and resolved by consensus.
The following information, if available, was extracted from each included study: authors' names, publication date, location of study, type of asthma, asthma medication use, and a brief description of the cohort, including: duration of recruitment, participant demographics, follow-up time, subgroup analyses (if any), crude and adjusted effect estimates reported and corresponding 95% confidence intervals, covariates included in analyses, and how overweight/obesity was assessed. The information was entered into Review Manager 5.1 [28]. Reviewers were not blinded to the names of authors, journals, or study institutions.
Studies were assessed for evidence of differential or non-differential selection bias using Cochrane Collaboration criteria [29]. Ascertainment bias was assessed by examining whether overweight or obese children were more likely to be screened and, subsequently, diagnosed with asthma. Attrition bias was determined by assessing completeness of follow-up, whether or not there were systematic differences in withdrawals, and/or differences in follow-up by exposure status.
Heterogeneity of studies in combined estimates was addressed by computing the I 2 statistic which is designed to estimate the proportion of variation across studies due to differences rather than chance alone [30,31]. Low heterogeneity is defined as an I 2 value 25-49%; moderate heterogeneity 50-74% and high heterogeneity ≥75%.

Data collection and synthesis
The crude relative risk (RR) or odds ratio (OR) for individual studies and 95% confidence intervals for all studies was computed and compared to reported crude RRs and ORs, when sufficient information was provided. If the computed effect estimates differed from the reported effect estimates, it was assumed that the reported effect estimate was an adjusted estimate. Comparisons were drawn between the crude and adjusted estimates. Estimates for the fixed effects model were calculated using the generic inverse variance method in which each study estimate is given a weight equal to the inverse of the variance of its effect estimate. Covariate-adjusted estimates were derived from all included studies.
Subgroup estimates were computed when at least two studies reported sufficient information to compute RRs for each subgroup. Data was available to examine the effects of age (young vs. older child) and sex (male vs. female). Publication bias could not be assessed using funnel plots because so few eligible prospective studies were eligible for inclusion [32]. Figure 1 summarizes the number of articles identified and the search process. The initial literature search identified 1431 articles. Examination of the titles eliminated 1322 articles that did not meet the search criteria. Of 109 candidate abstracts reviewed, 91 studies were eliminated: 48 did not use a prospective cohort design, two were reviews, 10 investigated diseases other than asthma, and 31 did not fulfill other inclusion criteria. Eighteen studies were identified for full text review. Of those, twelve studies, listed in Table 1 by year, were excluded: six did not report physician-diagnosed asthma [33][34][35][36][37][38], two did not categorize overweight/obesity by BMI [39,40], one did not report incident asthma as the primary outcome [41], two reported only adult populations [42,43], and one used a cross-sectional analysis [44].

Included studies
Six prospective cohort studies examining the association between BMI and subsequent physician-diagnosed asthma in children less than 18 years of age were 1431 records identified through database searching 109 abstracts included for review after duplicates and not-relevant articles based on title review were removed 109 abstracts screened Excluded: 40 cross-sectional studies, 8 case-control studies, 2 literature reviews, 10 articles of diseases other than asthma, and 31 articles that did not fulfill other inclusion criteria Excluded: 6 full-text articles that did not report physician diagnosed asthma, 2 articles that did not report weight by BMI, 1 article that did not report incident asthma as the primary outcome, 2 studies using non-pediatric populations, and 1 cross-sectional analysis  Outcome measure defined as at least 4 episodes of wheeze or 1 episode of wheeze with use of prescribed inhaled steroids, rather than physician-diagnosed asthma.
included and are described in Tables 2 and 3. The results of these six studies are summarized in Table 4 [50].

Combined effect estimates Overweight
As shown in Figure 2, the combined RR for the three studies examining overweight (BMI ≥85th percentile for age and sex) and physician-diagnosed asthma in boys and girls combined was significant (RR = 1.35; 95% CI = 1.15, 1.58; I 2 = 2%). When the analyses were stratified by sex, the combined RR for overweight and physiciandiagnosed asthma in boys was significant (RR = 1.41; 95% CI = 1.05, 1.88; I 2 = 62%). No significant association was observed for girls.

BMI Z-scores
As shown in Figure 4, the association between BMI, measured by continuous Z-scores, and physiciandiagnosed asthma was significant in girls (RR = 1.19; 95% CI = 1.06, 1.34; I 2 = 54%). BMI Z-score was not significantly associated with asthma in the combined cohort or among boys alone. For the fixed effects models, all combined RRs are listed in Table 5. Data were not available to measure the effects of any other variables as potential only available effect modifiers of the obesity/overweight and asthma associations. Table 6 lists the risk of bias in each study.

Ascertainment bias
Given similar presenting symptoms, it is unclear whether or not overweight/obese children were more likely than normal weight children to be diagnosed with asthma in any of the included studies, so this bias could not be assessed.

Classification of exposure bias
Mannino et al. has a risk of differential misclassification of BMI as only 69% of height and 61% of weight measurements were by study personnel [48]. The remaining values were self-report height and weight which are known to be over-and under-estimated, respectively [51]. The other five studies do not appear to be at risk.

Attrition bias
Gilliland et al., Gold et al., and Mannino et al. did not report the number of participants lost to follow-up [45,46,48]. Therefore, it is unclear whether attrition bias is present in these studies. Mamun et al. had a high risk of attrition bias as 46.5% of the original cohort was lost to follow-up by age 5 years and 59.6% of original cohort had incomplete data at either age 5 or 14 years and was excluded from the analysis [47]. Zhang et al. has low risk of attrition bias as the authors sampled 89.6% of the original cohort at the 6 year visit and 82.4% of the original cohort at the 8 year visit [49]. Ho et al. also has low risk of attrition bias as 82% of the selected original cohort had complete data at 12 month follow-up [50].

Discussion
Overweight and/or obese children were at increased risk of physician-diagnosed asthma (RR = 1.35; 95% CI = 1.15, 1.58). When stratified by sex, overweight boys remained at significantly increased risk of asthma. The overall increased risk appears to be more pronounced in the children who were obese (RR = 1.50; 95% CI = 1.22, 1.83). When stratified by sex, this association remained      Figure 2 Forest plot of overweight/obesity (categorical variable) and physician-diagnosed asthma (total and by sex).   Figure 3 Forest plot of obesity (categorical variable) and physician-diagnosed asthma (total and by sex).  Figure 4 Forest plot of BMI z-score (continuous variable) vs. physician-diagnosed asthma (total and by sex). significant in both boys, although with high heterogeneity that could not be explained, and girls. When BMI was defined using Z-scores, only girls were at significantly increased risk of physician-diagnosed asthma (RR = 1.19; 95% CI = 1.06, 1.34). The effect of obesity (a fifty percent increased risk of asthma) was observed in both boys and girls suggesting that sex does not appear to be a modifier of the effect of childhood obesity and asthma. However, overweight boys were at increased risk of asthma in comparison to girls and when BMI was measured by z-score girls were at increased risk of asthma in comparison to boys. This suggests that the type of overweight/obesity BMI measurement could influence study results. BMI alone cannot distinguish between fat and muscle mass and may, therefore, inadequately reflect fat distribution which can differ by sex. In children, this may under-or over-estimate obesity as wide variations in body fat distribution can occur within the same BMI percentile group [52]. BMI assessment could also potentially be modified by timing of puberty. Sex hormonal fluctuations, a biological phenomenon occurring during puberty known to influence asthma, especially among girls during the menstrual cycle, may explain the relationship between asthma and puberty [53]. Additionally, leptin has been proposed to be peripherally involved in respiratory function regulation and sexual maturation [54]. Included studies were either of pre-pubescent children or did not control for puberty or body fat distribution and, therefore, further research is needed to tease apart this complex association and determine whether these factors could confound the sex-specific overweight/obesity and asthma association.
The evidence for these associations, assessed in this systematic review using the most valid methodology available, supports the hypothesis that overweight, and especially obese, children are at significantly increased risk of incident physician-diagnosed asthma. However, additional high quality research is needed to confirm this finding as there are a limited number of prospective studies focusing on this issue. In two studies [45,49], where BMI was analyzed as a categorical variable, the combining of normal weight and underweight children into one reference category may have led to an underestimation of the true effect. In two other studies [46,47], where BMI was assessed as a continuous Zscore, the association between BMI and asthma was only statistically significant in girls and had a moderate level of heterogeneity (54%). Other possible subgroup analyses of potential effect modifiers, such as: race/ethnicity, geographic location, family history of asthma and/or allergy, in utero exposure to maternal smoking, and birthweight, were not possible due to insufficient reporting of information, a limited number of eligible studies, and variations in duration of follow-up and child ages. Whether or not the observed association between obesity and asthma is causal, or it results from a risk factor common to both conditions cannot be ascertained from this analysis.

Quality of the evidence
Assessing the quality of evidence was difficult due to a number of limiting factors. First, there was inadequate reporting of unadjusted effect estimates. None of the included articles listed the unadjusted effect estimates and few included adequate data to calculate them. Second, recruitment methods and characteristics of the study populations were not always complete. Detailed inclusion and exclusion criteria for each cohort were rarely provided. Third, study participation rates and losses to follow-up were only included in three articles [47,49,50]. Due to the inadequacy of reporting, it was difficult to assess: how and when participants in each Table 6 Risk of bias table   Bias Studies cohort were recruited, exact ages at which exposure and outcome were assessed, and the overall internal validity and generalizability of each study. None of the studies reported on diet or physical activity levels.

Agreements and disagreements with other studies or reviews
Results of this meta-analysis: that overweight and, particularly, obese children are at a 40-50% increased risk of physician-diagnosed asthma, support other published reviews examining the association between overweight/ obesity and asthma and/or wheeze. Previous metaanalyses of adults [4] and children [19] have been published. Flaherman and Rutherford [19] analyzed wheeze as a primary outcome in addition to asthma; however, that assessment differs from this meta-analysis as we only included studies that reported physician-diagnosed asthma as the primary outcome measure. Wheeze is an unreliable indicator of physician-diagnosed asthma [55]. Further, Flaherman and Rutherford [19] defined overweight/obesity as high birthweight and/or body weight instead of using age-and sex-specific BMI. Obese girls and boys were at significantly increased risk of subsequent physician-diagnosed asthma suggesting that sex is not a modifier of the effect of obesity, defined as age-and sex-specific BMI ≥95th%, on subsequent asthma diagnosis. However, when overweight/obesity was assessed via a continuous BMI Z-score, girls, but not boys, were at significantly increased risk of physician-diagnosed asthma. The opposite was true when overweight was categorically defined as age-and sex-specific BMI ≥85th%. In this case, boys were at significantly increased risk of asthma. This suggests that for the sex-specific overweight-to-asthma association the classification used to define the exposure is important and may influence conclusions drawn. It is also unclear whether the sex-specific associations observed could be influenced by timing of puberty, which we were unable to assess. Interestingly, Beuther and Sutherland reported a significant association between adult overweight/obesity, defined as BMI ≥ 25 kg/m 2 , and asthma incidence for both men (OR = 1.46, 95% CI 1.05, 2.02) and women (OR = 1.68; 95% CI = 1.45, 1.94) [4]. This further supports our findings which suggest that the overweight/ obese variable classification may be influential in assessing the sex-specific overweight/obese and asthma association as different classifications may yield varying results.

Conclusion
This systematic review provides an up-to-date assessment of the published prospective studies among children on the association between overweight/obesity and subsequent incident physician-diagnosed asthma in childhood. In children, the association between overweight and physician-diagnosed asthma appears to vary by sex based on how BMI is classified but these effects are inconsistent and limited to a few studies. Future research needs to better account for different durations of follow-up, ages at baseline and assessment, and assessment of obesity (categorical or continuous variables).
In addition, further research should investigate the potential impact of other risk factors, such as antibiotic exposure that may be independently associated with risk of asthma and obesity. It is important to understand whether asthma and obesity are causally associated and, if so, the directionality of the causal pathway or whether both are outcomes from a common exposure.