Our analyses of first-borns demonstrated that elective caesarean sections were associated with modestly increased risk of dispensed ICS, a marker of asthma. The sibling-pair analysis, which inherently adjusts for shared environmental and genetic factors, confirmed this association in elective caesarean in the two to five year olds, but not in the older children. The associations between emergency caesarean section and vacuum extraction and risk of dispensed ICS observed in the analyses of first-borns where not confirmed in the sibling-pair analyses, suggesting these effects are due to residual confounding.
Several large cohort studies from Scandinavian countries have suggested that emergency as well as elective section could affect the risk of asthma[5, 14] or asthma hospitalisation in young children. The conventional, unconditional logistic regression, analysis in our study indicated that emergency caesarean sections and vacuum extraction were associated with an increased risk of asthma medication but these associations disappeared completely in the sibling pair analyses suggesting that these effects in the unconditional logistics regression were due to residual confounding by familial risk factors. It remains unclear what these unmeasured confounding factors in the conventional analysis could be. We have adjusted for a range of potential confounders, including socio-economic status, parental asthma, and pregnancy and neonatal complications. Adjustment for these factors did not greatly alter the observed associations. Maternal anxiety is associated with an increased risk of pregnancy complications and emergency caesarean sections and maternal stress during pregnancy could contribute to an increased risk of childhood asthma. Although we did adjust our analysis for proxy indicators for parental asthma (dispensed ICS), there may well be residual genetic confounding that connects perinatal complications with asthma later in life.
The current study underlines the difficulties to control for confounding in conventional analyses of mode of delivery and asthma. It is based on register data covering the whole Swedish population. There is no recall or selection bias. However, some of the indicators in the registers are crude proxies for the true exposures. It is particulary difficult to cover life style factors, and the specific indications for elective caesareans. These factors may affect mode of delivery[19, 20] as well as the risk of asthma, use of medication and access to health care.
Studies with a sibling design are a powerful tool to control for familial confounding due to genetic or environmental factors. With the exception of the recent study by Almqvist et al, this design has not been used in epidemiological studies of mode of delivery and asthma. A prerequisite is a sufficient number of discordant sibling pairs. Our study is based on all ethnic Swedish children born term between 1992 and 2008 with data available on over 110 000 sibling pairs. Despite these quite impressive sibling populations, only 1965 sibling pairs at age 2-5 and 569 at age 6-9 were discordant for both mode of delivery and asthma medication use, and thus were informative for this analysis. These sample sizes were insufficient to obtain precise estimates of these comparatively small effects, thus requiring cautious interpretation. In comparison, Almqvist et al. had 1005 informative sibling pairs available in their study. Although the majority of our findings are consistent with those of Almqvist et al. they observed an association between emergency caesarean section and increased risk of asthma, while we did not. This may be due to multiple differences in the design of these two studies. Specifically, we excluded all children born pre-term, and those who had malformations or cerebral palsy. These conditions may result in delivery by emergency caesearan section and increased risk of developing various respiratory illnesses including asthma.
Childhood asthma is a heterogenous disease and a number of phenotypes with partly different risk factors have been identified. Almqvist et al did not observe any association between elective caesarean section and an increased risk of asthma in sibling-pairs aged 9 and 12 years. Our study suggests that elective caesarean section could be associated with a slightly increased risk of transient wheeze up to five years of age. Maternal complications during pregnancy have previously been linked to an increased risk of childhood wheeze. Hypertension and diabetes were more likely in women who were delivered by elective caesarean section but the increased risk for asthma medication persisted also after adjustment for these factors. However, we cannot exclude that other undetected complications contributed to the increased risk of asthma medication in preschoolers.
This study has a number of limitations. Defining asthma within epidemiological studies is a challenge. We have chosen to use prescription of inhaled corticosteroids (ICS) to define current symptoms of asthma. We have previously used a similar definition as has a similar study from Finland A number of studies have assessed the validity of prescription registry based information on asthma medicine use against parent report of asthma, standardised questionnaire based definitions of asthma, and doctor diagnosis in similar settings to the current study. These studies generally find that prescription registry data provides a reasonably valid definition of asthma. We have elected to exclude use of beta-agonists from our primary definition of the outcome, as these are a less specific marker for asthma.
However, dispensed medication as a proxy for asthma is affected by a number of factors at different levels such as awareness of the symptoms by the child and the parents, severity of asthma, health seeking behavior, diagnostic criteria by the doctor, attitude to medication and its related costs. If one sibling has asthma, other siblings are more likely to be diagnosed and get a prescription for an asthma medication than children in families where the symptoms of asthma are less familiar. Sharing of medication may occur in families where several members suffer from asthma. When asthma is diagnosed compliance with medication probably differ in different families, and this will have an effect on subsequent prescriptions. Many of these family effects could contribute to a negative outcome and some of these such as sharing of medication could weaken the associations also in the sibling analyses. Using purchase of ICS may imply that we have missed individuals (parents or children) with undiagnosed or mild asthma. Therefore, residual confounding may explain some of the reduced effect sizes in the sibling analysis.
In young children, dispensed ICS is a proxy for respiratory illness, but not always for asthma. as diagnosis of asthma is very difficult in preschool children. By the age of 6 years, the prevalence of transient viral wheeze has largely passed, and a diagnosis of asthma is much easier to confirm, making this a reasonable time point to divide the cohort of children. Variability in prescription pattern could dilute the association between asthma medication and mode of delivery in young children. For this reason, we have also excluded children aged less than two years, and used ICS rather than all asthma medication as our main outcome. Children with nonspecific clinical symptoms sometimes get one prescription of ICS as a test for asthma. One may therefore ague that at least two prescriptions of ICS would have been a better proxy for asthma in our study but we have tested analyses based on at least two prescriptions of ICS and the findings were fairly similar.
An association between caesarean section and asthma could also be weakened or even missed, if potential effects are restricted to specific genetic variants or subgroups of asthma. We do not have direct measures for atopy in this study. Therefore, we cannot exclude that mode of delivery could affect allergic induced symptoms. Moreover, we have no information in the registers concerning postnatal exposures. However, a recent birth cohort study failed to identify modification of the association of between mode of delivery and childhood asthma by a range of key postnatal exposures, including breast-feeding behavior and age at day care entry.
We cannot exclude the risk of some misclassification of the caesarean sections. “Elective” caesarean sections were not necessarily optional. We considered a caesarean section to be emergent if the operation was made after the onset of labour. Therefore, caesarean sections due to emergencies such as fetal distress or a severe maternal complication were defined as elective if the caesarean sections were performed without preceding labour. Although this would have resulted in some misclassification between emergency and elective caesarean sections, as this reason for cesarean section is relatively uncommon, we consider it to be an unlikely source of a major bias in this study. The major indications in Sweden for elective caesarean sections after 36 weeks are breech presentation, previous caesarean section, cephalopelvic disproportion and psychosocial factors. They accounted for more than 80% of all elective caesarean sections between 1996 and 2007.
To conclude, elective caesarean section contributes to a modestly increased risk of asthma medication, but only up to five years of age. The associations between emergency caesarean section or vacuum extraction and asthma medication seen in the firstborns appear to be due to residual confounding by familial factors.