Objective To examine the association between birth weight for gestational age and asthma in childhood and adolescence while controlling for potential confounders and considering smoking as an effect modifier.
Methods A retrospective cohort of all singleton term births in Nova Scotia, Canada, between 1989 and 1993 was identified in the provincial perinatal database and followed through 2014 by linking with administrative health data. The outcome, asthma, was defined as having one hospitalisation or two physician visits with an International Classification of Diseases code for asthma in a 2-year period. Birth weight was categorised as small (SGA), large (LGA) or appropriate (AGA) for gestational age. Multivariable-adjusted Cox proportional hazards models were used to examine the association between the birth weight for gestational age and asthma and to test for effect modification by maternal smoking in pregnancy.
Results Of the 40 724 cohort children, 10.5% and 11.7% were born SGA and LGA, respectively, and the risk of developing asthma to age 18 years was 30.2%. The adjusted HRs for SGA and LGA (relative to AGA) and asthma were 1.07 (95% CI 1.02 to 1.14) and 0.96 (95% CI 0.91 to 1.02), respectively. Relative to AGA children born to non-smoking mothers, SGA children were not at increased risk of asthma (HR 1.02), whereas both AGA and SGA children born to smoking mothers were at significantly increased risk (HR 1.14 and 1.29, respectively).
Conclusions Our findings suggest that SGA in term infants is not associated with asthma in childhood in the absence of smoking in pregnancy.
Statistics from Altmetric.com
What is already known on this topic?
Low birth weight infants are at a higher risk for asthma in childhood but this association may be modified by smoking in pregnancy.
The role of high birth weight in the aetiology of childhood asthma is unclear.
What this study adds?
In term infants, being born small for gestational age is not associated with asthma in childhood in the absence of smoking in pregnancy.
Large for gestational age infants do not have increased risk of asthma.
Asthma is one of the most common chronic diseases in children1 and is associated with considerable burden to the affected children, their families and the healthcare system.2 Besides environmental, family, sex, socioeconomic and lifestyle factors,3 several perinatal factors have been identified as risk factors for asthma.4–6 Low birth weight (LBW,<2500 g) may be linked to asthma6 through adaptive processes in utero that lead to impaired lung growth with smaller airways.7 Many studies, however, have not considered the gestational age of the infant; LBW may rather be a proxy for prematurity, which has also been linked to lung disease in childhood.8
The risk of asthma in childhood is also influenced by other perinatal factors, such as caesarean delivery, maternal pre-pregnancy weight or smoking in pregnancy.4 9 10 Few studies were able to examine the contribution of all of these factors to the risk of asthma in children and youth.11 12 Moreover, a Swedish study in 2011 found a strong interaction between LBW and maternal smoking during pregnancy.13 It has been speculated that prenatal smoking modifies the association between impaired fetal growth and later asthma such that LBW infants become more susceptible to environmental asthma risk factors in childhood.8 This interaction has not been assessed in other studies so far. Therefore, the objective of this study was to examine the association between birth weight for gestational age and asthma in childhood and adolescence while considering a range of perinatal confounders and smoking as an effect modifier.
Study design and setting
This study was a retrospective cohort study of all children of singleton gestation born in the Canadian province Nova Scotia between 1 January 1989 and 31 December 1993 who were followed until 31 December 2014. The cohort was assembled through linkage of the Nova Scotia Atlee Perinatal Database (NSAPD) and with provincial administrative health data. The linkage was conducted by Health Data Nova Scotia based on the health card number. Prior to linkage, health card numbers were encrypted to ensure anonymity. For the present analysis, only term infants (≥37 weeks) were included.
The NSAPD captures information on all births to mothers who were resident in Nova Scotia. For each delivery record, the NSAPD includes extensive information on routine demographic variables, medical conditions, reproductive history, delivery events and neonatal outcomes.14 Trained coders enter this information from standardised clinical forms into the NSAPD. The Reproductive Care Program of Nova Scotia maintains the database; periodic abstraction and validation studies form an ongoing data quality assurance programme and have shown that the data are accurate and reliable.15
Administrative health data
All Canadian provinces have a single-payer universal healthcare system that pays for essential health services for all residents. The Medical Services Insurance physician billing database contains administrative records for each insured service rendered by a physician and paid for by the provincial healthcare system since 1989. The database contains diagnostic codes in International Classification of Diseases (ICD)-9 format (one from 1989 to 1996, up to three after 1996). Hospital databases comprised the Admissions/Separations/Day Surgery hospital database (1989–1994) and the Canadian Institute for Health Information Discharge Abstract Database (1995 and later). Information in the hospital databases includes patient demographic information, diagnostic and procedure codes and specialty services received. Diagnostic codes are in ICD-9 format from 1989 to 2000 and ICD-10CA format from 2001 forward. The Insured Patient Registry keeps an ongoing record of every beneficiary of the provincial healthcare plan and was used to determine loss to follow-up due to death or relocation.
Main exposure of interest
The main exposure of interest was birth weight for gestational age and sex relative to a Canadian reference population.16 Infants were categorised as small (SGA, <10th percentile), large (LGA, >90th percentile) or appropriate (AGA, 10th–90th percentiles) for gestational age. Birth weight z-scores were calculated using sex and gestational age-specific means and SDs from the same reference population.
The outcome was asthma, defined as having at least one hospitalisation or two physician visits with a diagnosis code for asthma (ICD-9: 493; ICD-10: J45) in a 2-year period. The date of the earliest of visit that met this definition was considered as the date of asthma diagnosis. This definition has been validated using chart abstraction studies and has a sensitivity of 89% and a specificity of 72% to detect asthma cases in children 0–17 years of age.17
A directed acyclic graph was used to determine confounding variables (figure 1). These included maternal age, area of residence (urban vs rural, based on the Canadian postal code), area-level income quintile at birth (from Census of Canada information18), parity (1, 2, 3 or ≥4, including the current pregnancy), pre-pregnancy weight status (normal weight, overweight or obese), asthma during pregnancy, mode of delivery (vaginal delivery vs caesarean section) and smoking in pregnancy (yes/no). Information on all confounding variables was obtained from the NSAPD.
Mothers’ pre-pregnancy weight was recorded on standard provincial prenatal forms and would be either self-reported or measured at the first prenatal visit. Self-report of pre-pregnancy weight has been shown to agree closely with measured weight.19 Maternal height has been routinely collected for the NSAPD since 2003 and has been retroactively entered where available for women with pregnancies before that year. Where maternal height was not available, we predicted maternal body mass index (BMI) category from maternal weight based on data from 77 297 mothers for whom height and weight were recorded in the NSAPD from 2003 to 2015 (normal weight:<68.0 kg; overweight: 68.0–76.6 kg; obese: ≥76.7 kg). The sensitivity of these weight-based categories was between 89% and 96%, and specificity was between 90% and 95%. When regressing common perinatal outcomes (gestational diabetes, mode of delivery, 5 min Apgar score <7, birth weight for gestational age, pre-eclampsia) based on imputed and recorded weight status, the effect estimates from the corresponding models did not differ by more than 10%, and the imputation-based estimates were conservative (ie, closer to the null than the complete case estimates). Smoking during pregnancy was assessed by care providers based on maternal self-report at two timepoints (at a mother’s first prenatal visit and on admission to the labour ward); any smoking at either of these times was considered smoking during pregnancy.
Descriptive statistics were calculated by birth weight for gestational age category. Unadjusted and adjusted Cox proportional hazards regression with robust standard errors was used to estimate the association between birth weight for gestational age category and diagnosis of asthma. Observations were censored when a child’s healthcare coverage ended due to relocation outside the province or death. The Efron approximation was used to handle ties. Hot deck multiple imputation with five iterations was used to impute missing values of the confounding variables.
We also examined potential interactions between birth weight for gestational age category and smoking in pregnancy. Multiplicative interaction was investigated by assessing the statistical significance of the cross product terms of the two variables. Additive interaction was estimated by the relative excess risk due to interaction (RERI) and the attributable proportion (AP) due to interaction.20
Lastly, to describe the shape of the relationship between birth weight for gestational age and asthma, we developed a generalised additive model regressing asthma on birth weight for gestational age z-score, stratified by maternal smoking status during pregnancy.
Approval for this study was given by the IWK Health Centre Research Ethics Board (File No 1015756), the Joint Data Access Committee of the Reproductive Care Program and the Health Data Nova Scotia Data Access Committee.
A total of 48 790 births were recorded in the NSAPD between 1 January 1989 and 31 December 1993. Plausible birth weight and gestational age information was available for 47 450 children, and 42 999 of these could be linked with administrative health data. After exclusion of children born before 37 weeks of gestation, the final sample consisted of 40 724 children (83.5%). The administrative data algorithm17 identified 10 155 (23.6%) and 12 997 (30.2%) of children as having had asthma at some point during the first 6 and 18 years of life, respectively. The sociodemographic and clinical characteristics of the women and children are summarised in table 1 by birth weight for gestational age category. Compared with mothers of AGA infants, mothers of SGA infants were more likely to be younger, single, primiparous, and of normal weight, live in areas with a lower average household income and smoke during pregnancy, whereas mothers of LGA infants tended to be older and obese, have a partner, live in higher income neighbourhoods, deliver by caesarean section and not smoke in pregnancy.
HRs for the association between birth weight for gestational age and the development of asthma are shown in table 2. In both the unadjusted and adjusted models, SGA was associated with an elevated risk for asthma compared with AGA. LGA was not associated with the risk of developing asthma. Further adjustment for the hypothesised mediator, mode of delivery, did not change the estimates (data not shown). Analysis of the Schoenfeld residuals showed that the proportionality assumption was met. There was no multiplicative but an additive interaction between SGA and smoking (RERI 0.13, 95% CI 0.01 to 0.26) as shown in table 3. Relative to AGA infants of non-smoking mothers, the asthma risk for SGA infants of non-smoking mothers was not different (HR 1.02), whereas the HRs were elevated for AGA/smoking (HR 1.14) and SGA/smoking (HR 1.29), respectively. The AP due to interaction was 0.10, indicating that approximately 10% of asthma risk in the SGA/smoking group is due to the interaction.
The generalised additive model (figure 2) showed that among children whose mothers smoked in their pregnancy, the lowest risk of asthma was in children with birth weight for gestational age z-scores >1; risk increased steadily as birth weight for gestational age decreased below 0.5 units. By contrast, virtually no change in asthma risk was observed across the birth weight z-score spectrum among children whose mothers did not smoke during pregnancy.
We demonstrated in a large population-based sample that SGA in term infants was not associated with asthma risk in the absence of smoking. We found no association between LGA and asthma.
An association between SGA or LBW and asthma has been described in several studies; two meta-analyses have reported pooled ORs of 1.16 and 1.34, respectively, for later asthma in LBW compared with normal birthweight (2500–4000 g) infants.6 21 Some studies did not consider smoking as a confounder or effect modifier and, therefore, may have overestimated the effect of LBW. One study from Sweden recently reported a strong multiplicative interaction between LBW and smoking. The authors found no association of LBW with asthma and a moderate association of smoking with asthma but a strong combined association of both factors with asthma.13 Similarly, an analysis from the Prevention and Incidence of Asthma and Mite Allergy (PIAMA) study found only a weak association of LBW with asthma in the absence of maternal smoking in pregnancy but a strong association in the presence of smoking during pregnancy.22 In keeping with these two studies, our study found no association between SGA and asthma in the children of non-smoking mothers, and a strong combined effect in SGA infants whose mothers smoked during pregnancy. A possible explanation is that among children born to mothers who smoked in pregnancy, being born SGA may indicate heavier smoking compared with children born LGA or AGA, and, hence, the association appears more pronounced in the SGA group. However, limited information on the number of cigarettes smoked in the present study does not support this hypothesis; the proportion of smokers who reported smoking more than half a pack per day during pregnancy did not differ between the SGA and AGA group (67% and 62%, respectively). Since we did not have information on postnatal smoking for our cohort, we are unable to attribute the observed association to adverse intrauterine influences of smoking as opposed to highly correlated postnatal smoke exposure. Two previous studies that compared the prenatal and postnatal influences of smoking found that prenatal smoking had a stronger association with asthma or wheezing than early life exposure to smoking.23 24 A review of studies on parental smoking and childhood asthma has also suggested that (postnatal) environmental smoke exposure triggers wheezing attacks rather than being the cause of an asthmatic predisposition.25
The literature on the association between LGA (or high birth weight) and asthma is conflicting; some studies did not find an association,26 27 while others reported a positive association.5 28 29 Two of the studies that reported a positive association were Canadian and also used administrative health data. To et al using Ontario data found a slightly lower risk of asthma in infants with high birth weight but a trend towards a higher risk among infants with extremely large birth weight (>6.5 kg).29 The authors suggested that confounding by parity or ethnicity may explain the findings. Sin et al using Alberta data found that high birthweight infants (>4.5 kg) had 16% more emergency visits for asthma than infants with a normal birth weight.5 The results from our survival analysis and the generalised additive model do not suggest an increase in asthma risk with increasing birth weight. The finding is somewhat surprising considering that LGA infants are at a higher risk for later obesity,30 which in turn is a risk factor for asthma.31
Our study has several notable strengths, including the use of a large population-based data set with a long follow-up. By contrast to many previous studies, we used birth weight for gestational age and sex to define our exposure as opposed to simple birth weight-based cut-offs and restricted our sample to term infants to remove the effect of prematurity from the association. We were also able to control for a broad range of confounders in our analysis. The outcome of our study was derived from diagnosis codes assigned by physicians and not parent report. The algorithm used to determine the presence of asthma from ICD codes in the administrative data had a sensitivity of 89% and a specificity of 72% for identifying asthma as per standardised diagnostic criteria from a chart review in a sample of Ontario children.17 The relatively low specificity probably contributed to the high cumulative incidence of asthma over the first 18 years of life in our sample. In 2017, the Canadian Chronic Disease Surveillance System reported an asthma prevalence of 14.7% for children and youth aged 1–19 years.32 The asthma prevalence in Nova Scotia is above the Canadian average.33 It should be noted that the 30.2% in the current study represents the cumulative incidence of ever having had asthma over the first 18 years of life, which is not directly comparable to the prevalence in a given year. The children identified as having asthma in the present study probably represent a mix of children with true asthma and children with asthma-like respiratory symptoms who were investigated for the presence of asthma. The misclassification would most likely have been non-differential and would, therefore, have resulted in an underestimation of the true association. Antenatal and neonatal care at the time our cohort was born (1989–1993) was different from today’s care in some aspects, and therefore, our findings may not be applicable to children born during this decade. Lastly, there is the potential for residual confounding due to the use of maternal pre-pregnancy weight instead of BMI, which may have also lead to an attenuation of the association.
In summary, our study found that SGA in term infants may not be associated with asthma in childhood in the absence of smoking in pregnancy. We also found no association between LGA and asthma. Future studies should examine the interaction between SGA and smoking dose to clarify their roles in the development of asthma in childhood.
Portions of the data used in this report were made available by Health Data Nova Scotia of Dalhousie University.
Contributors JHC contributed to the design of the study, wrote the initial manuscript draft, reviewed and revised the manuscript and approved the final manuscript as submitted. CGW contributed to the design of the study, reviewed and revised the manuscript and approved the final manuscript as submitted. LL carried out the analysis, reviewed and revised the manuscript and approved the final manuscript as submitted. SK conceptualised and designed the study, carried out the analysis, reviewed and revised the manuscript and approved the final manuscript as submitted.
Funding This work was funded by an operating grant from the Canadian Institutes of Health Research (CIHR) awarded to SK (FRN 134531). JHC held a Dalhousie Faculty of Medicine Summer Studentship.
Disclaimer Although this research is based on data obtained from the Nova Scotia Department of Health and Wellness, the observations and opinions expressed are those of the authors and do not represent those of either Health Data Nova Scotia or the Department of Health and Wellness.
Competing interests None declared.
Patient consent Not required.
Ethics approval IWK Health Centre Research Ethics Board (File No 1015756), Halifax, NS, Canada.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The data used in this study are available from the Reproductive Care Program of Nova Scotia and Health Data Nova Scotia, Canada. As these data constitute personal health information, they can only be accessed from within Nova Scotia as per provincial privacy laws. Researchers must submit a data access application to the database custodians and upon approval sign a data sharing agreement.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.