The Early-Life Risk Factors of Childhood Wheezing: Results from the NHANES During 2005-2012

Objectives: Wheezing is the most common cause of hospitalization during childhood. The aim of the present study was to identify early-life risk factors predisposing to childhood wheezing using directed acyclic graph (DAG) analysis. Methods: In this study, we used the National Health and Nutrition Examination Survey (NHANES) data to identify the risk factors associated with wheezing in children aged 1-15 years old in the United States. The data was related to four 2-year cycles from 2005 to 2012. All the adjusted analyses were based on the directed acyclic graph(DAG). The applied logistic regression was performed by survey analysis to adjust the weights of samples. Results: From 13203 children, 87.2% (n = 11 519) had not experienced wheezing during the past year, and 12.8% (n=1684) experienced wheezing during the same time. The mean ages (SD) were 6.6 (4.4) and 7.6 (4.4) years old in wheezing and healthy children respectively. Boys comprised a signicantly higher ratio in wheezing compared with healthy children (P = 0.001). Mother smoking during pregnancy (MSDP) was a signicant risk factor for childhood wheezing (OR= 1.42, 95% CI: 1.15-1.74, P = 0.001). All ethnicities other than the Mexican-American were established as signicant risk factors for childhood wheezing. On the other hand, childhood wheezing was insignicantly associated with Socio-economic status, birth weight, maternal age at pregnancy, and exposure to smoke at home. Conclusion: After adjusting for established confounders, history of MSDP, as well as ethnicity delivered important risk factors for childhood wheezing in a large population of the US.


Introduction
Wheezing is most often triggered by the blockage of the lower respiratory tract (1). Wheezing can in ict anywhere in the intrathoracic trachea, induce in ammation and mucus production in the airways, and promote reversible tightening of the smooth muscles in the airways walls (2). The sound of wheezing is induced by the turbulent air ow in the large airways which collapse upon expiration. Due to smaller airways of children, wheezing is an important clinical condition in childhood (1).
Wheezing presents a global problem and a multifactorial disease (3). It is one of the most common causes of hospitalization during childhood (4). With the prevalence of about 26.4%, wheezing comprises a common problem in children aging 2-3 years old in the United States (US). (5). The prevalence of childhood wheezing has been noted as 22.1% in United Kingdom (6). In order to prevent childhood wheezing, it is important to identify early-life risk factors for this phenomenon (3). In this study, we aimed to ascertain the early-life risk factors of childhood wheezing using a dataset of a large population from 2005 to 2012. Discrepancies have been observed in the estimated risk factors of childhood wheezing in previous studies. These discrepancies may be due to recruiting inappropriate inclusion criteria for variables in the analysis model. In this study, we tried to obviate this problem and avoid selection bias by using a multivariable analysis based on the directed acyclic graph (DAG) and modern epidemiology recommendations (7,8).
The aim of the present study was to investigate the potential effects of some early-life factors, Socioeconomic status (SES), and ethnicity on childhood wheezing. For this, we used DAG analysis to identify possible confounding factors contributing to childhood wheezing. This was a large population-based survey performed on the data related to a large population in the US from 2005 to 2012.

Methods
The study population and data acquisition The National Health and Nutrition Examination Survey (NHANES) represents a cross-sectional survey conducted on the US pediatric population. Details on the methodology have been described elsewhere (9,10). In particular, NHANES is a population-based survey designed to assess the health status of the US population. The data of the NHANES study has been being collected at 2-year cycles since 1999. The NHANES interviews addressed the demographic, socioeconomic and other health related variables. The survey was designed as a strati ed multistage probability study so that oversampled of certain age and minority groups.
In this study, we used NHANES data related to children aged 1-15 years living in the US. The analyzed data corresponded to four 2-year cycles from 2005 to 2012.
Regarding the four cycles of the survey, 13212 children aged 1 to 15 years old were interviewed (either the children themselves or their parents). Children were excluded if they had incomplete data on the wheezing status.

The study variables
The wheezing as the outcome variable was de ned as self-reported wheezing or whistling in chest during the past year. During the interviews, the parents were asked: "In the past 12 months, have children had wheezing or whistling in their chest?". The response options included "yeas" and "no".
Other variables have been described in the following. The MSDP (did children biological mother smoke at any time during the pregnancy) was de ned as a binary variable (yes / no). Race/ethnicity was categorized as Mexican-American, non-Hispanic white, non-Hispanic black, other Hispanic, and other (including multiracial status). Socio-economic status (SES) was estimated by the number of rooms in the house, annual household income, and the total number of family members. The SES was nally determined and categorized into ve quartiles by merging these items through principle component analysis (11,12). Birth weight was recorded as a continuous variable in pounds (How much did the child weigh at birth?). In addition, maternal age at pregnancy was categorized as 14-19, 20-29, 30-34, and ≥ Page 4/13 female. In the present analysis, the question of "Does anyone smoke at home? (Yes / No)" was incorporated to determine the exposure of children to smoke at home.

Statistical analysis
Directed acyclic graphs (DAGs) are valid causal inferences that are increasingly used in modern epidemiology (10,13,14). In order to address confounding effects, the DAG was used to determine the variables eligible to be included in this study (Fig. 1). All the backdoor paths that were related to the studied variables were identi ed to disclose any confounding effect (7,13). The impact of each of MSDP, SES, smoker family members, birth weight, maternal age at pregnancy, and ethnicity on childhood wheezing was calculated through conditioning by adjusting confounding paths in logistic regression analysis. For this purpose, the SES, smoker family members, and ethnicity, as the variable located on the confounding paths, were included in the logistic regression model to determine the impact of MSDP on childhood wheezing. Furthermore, as the NHANES survey depends on the weighting approach (10), the applied logistic regression was run by survey analysis to calculate unadjusted and adjusted ORs and respective 95% con dence intervals (95% CIs) for potential confounders. By default, the sampling weighting in NHANES public dataset is speci ed at 2-year time intervals. Because the analysis was performed on four 2-year cycles, and in order to achieve valid weights, we transformed the 2-year weights into 8-year weights in compliance with the NHANES instructions. The analyses were performed in STATA 12.0 software.

Results
Of the 13212 children who were present at all the four cycles (i.e. eight years from 2005-2012), 13203 (99.93%) had full data on wheezing status. Among these, 87.2% (n = 11 519) declared no wheezing or whistling in chest during the past year. On the other hand, 12.8% (n = 1684) children experienced wheezing in the same period. Table 1 shows the characteristics of the study population strati ed by wheezing status. The mean ages (SD) were 6.6 (4.4) and 7.6 (4.4) years old in wheezing and healthy children respectively. The ratio of boys was signi cantly higher among the wheezing than healthy children (P = 0.001). The proportion of households with the annual income ≥ $20000 was higher in wheezing than healthy children (P = 0.914). The total number of family members was signi cantly lower in wheezing than healthy children (P = 001). Finally, the number of rooms per house was slightly higher in wheezing than healthy children (P = 0.45, Table 1).
All the adjusted analyses were based on the confounding paths identi ed in the DAG analysis (Fig. 1). The adjusted and unadjusted associations between each studied variable and wheezing status have been shown in

Discussion
In this large population-based 8-year study conducted on children in the US, we indicated that odds of wheezing were higher in children who had MSDP than children without this exposure. Although SES and birth weight were negatively associated with childhood wheezing, these associations were not statistically signi cant. Maternal age at pregnancy and having smoker family member were also recognized as potential risk factors for childhood wheezing; however, the results did not establish signi cant associations. Importantly, all the studied ethnicities (i.e. non-Hispanic White, non-Hispanic black, other Hispanic, and other ethnicity including multi-racial) except for the Mexican-American were identi ed as signi cant risk factors for childhood wheezing. All these associations were adjusted for the potential cofounding effects of the variables identi ed in the DAG analysis. The previous reports were consistent with our nding on the association between the history of MSDP and childhood wheezing. In a recent cohort study in which infants were followed during the rst year of life, maternal active smoking signi cantly increased the risk of wheezing in infants (OR = 2.09, 95% CI: 1.54-2.84, p < 0.0001) (15). In accordance with our observation, no associations were reported between wheezing and neither SES (OR = 0.81, 95% CI: 0.61-1.07, p = 0.151) nor birth weight (OR = 0·98, 95% CI: 0·89-1·07, P = 0·614) in the recent report (15). The impact of cigarette smoking on in-utero biological development of fetus has been ascertained (15). Overall, these statements are consistent with our ndings. In fact, smoking exposure may impair the lung development and growth secondary to the detrimental effects of nicotine which interferes with lung growth and increases collagen deposition in airways (15).
Vardavas et al (16,17) in a pooled analysis reported that prenatal exposure to smoke increased the risk of postnatal wheezing in children. This was consistent with our ndings as we also observed that MSDP had a signi cant association with childhood wheezing, while post-natal exposure to smoke by having smoker family members did not exert such effect. Furthermore, according to the report of Moradzadeh et al in 2018 (10), maternal smoking moderately increased the odds of childhood asthma after adjustment for the misclassi cation bias of self-reported exposure. As the rate of MSDP has been unacceptably high (as high as 15.9%) among US women (18,19), the detrimental effects of this event on fetus and infants are concerning. Whitney et al (18) underscored that the rate of MSDP was signi cantly different among various ethnicities. In fact, MSDP may role as an intermediate variable in uencing the association between ethnicity and childhood wheezing. As we observed here, there was a signi cant association between ethnicity and wheezing. This may be partly due to the high prevalence of smoking in mothers of the at risk ethnicities. Similar to our nding, the SES was not signi cantly associated with childhood wheezing in the study of Vanker et al as well (15). Therefore, these observations indicated that ethnicity may present a stronger interrelationship with cigarette smoking than SES to determine the risk of childhood wheezing.
Our results further con rmed the ndings of the Millennium Cohort Study (a national wide study on children born in UK) by . In the recent study, no signi cant association was identi ed between low birth weight and childhood wheezing (OR = 0.97) (20). However, the birth weight may be considered as a mediator in uencing the interaction between MSDP and wheezing. In fact, the non-signi cant association between birth weight and wheezing may be due to the superior impact of MSDP on the risk of childhood wheezing. In line with our study, no statistically signi cant association was reported between birth weight and childhood wheezing in a cohort study by Vanker et al (15) and a systematic review and meta-analysis by Mebrahtu et al (21).
Our study did not support any statistically signi cant link between maternal age at pregnancy and the odds of childhood wheezing. This was consistent with the results of two cohort studies by Caudri et al in Netherlands (22) and Kotecha et al in United Kingdom (23).
The strengths of this study included using the data related to a relatively long period of time (i.e. 8 years) and incorporating a large sample size obviating any sampling error. Our results also presented a low possibility of selection bias because of a high response rate and the complex sampling method.
Some limitations should be considered for this study. The variables studied in NHANES are based on selfreporting (10); therefore, there is a risk for misclassi cation bias (10,24,25) due to problems with the recalling and under-reporting (14). The adjusted ORs (considering the possible misclassi cation of selfreported MSDP) were bigger than the respective unadjusted values in this study. So, assuming misclassi cation bias in self-reported MSDP, the obtained ORs in this study were smaller than those in reality. Nevertheless, self-reporting has been the standard method for collecting data on wheezing in large epidemiological studies such as the International Study of Asthma and Allergies in Childhood. In order to assess the impact of misclassi cation bias on the ndings, it is recommended to obtain bias-adjusted ORs by performing bias analysis for exposures, confounders, and self-reported wheezing. In addition, mediation analysis is necessary to calculate the direct and indirect impacts of any other variable on the risk of childhood wheezing.
In conclusion, a history of MSDP, as well as ethnicity remained independent and important risk factors of childhood wheezing in a large US population after adjustment for well-known confounders. This study highlighted the need for implementing effective programs to prevent women from smoking before and during pregnancy. The study also highlighted the importance of ethnicity as a risk factor for childhood wheezing. Therefore, it may be logical to consider preventive measures among high-risk ethnicities. Our results also underscored that exposures during prenatal period may exert a larger impact on the risk of childhood wheezing. Further studies are needed to divulge other unknown confounders and their effects on respiratory tract health.

Declarations
Con ict of interest: