To investigate whether delivery mode (vaginal versus by caesarean section), maternal pre-pregnancy body mass index (BMI) and early exposure to antibiotics (<6 months of age) influence child's risk of overweight at age 7 years, hence supporting the hypotheses that environmental factors influencing the establishment and diversity of the gut microbiota are associated with later risk of overweight.
Longitudinal, prospective study with measure of exposures in infancy and follow-up at age 7 years.
A total of 28 354 mother–child dyads from the Danish National Birth Cohort, with information on maternal pre-pregnancy BMI, delivery mode and antibiotic administration in infancy, were assessed. Logistic regression analyses were performed with childhood height and weight at the 7-year follow-up as outcome measures.
Delivery mode was not significantly associated with childhood overweight (odds ratio (OR):1.18, 95% confidence interval (CI): 0.95–1.47). Antibiotics during the first 6 months of life led to increased risk of overweight among children of normal weight mothers (OR: 1.54, 95% CI: 1.09–2.17) and a decreased risk of overweight among children of overweight mothers (OR: 0.54, 95% CI: 0.30–0.98). The same tendency was observed among children of obese mothers (OR: 0.85, 95% CI: 0.41–1.76).
The present cohort study revealed that a combination of early exposures, including delivery mode, maternal pre-pregnancy BMI and antibiotics in infancy, influences the risk of overweight in later childhood. This effect may potentially be explained by an impact on establishment and diversity of the microbiota.
The genetic transmission of risk of overweight and obesity from parents to children is well established.1, 2, 3, 4 In addition to the genetic risk, other factors may promote the transmission of risk within families, among which the transfer of the maternal gut microbiota to the child during delivery has recently received particular attention on the basis of several lines of evidence.
The intestinal tract is sterile at delivery but harbours more than 1017 microorganisms in adulthood. It is not clear how the microbiota is established and whether a potential obesogenic effect is determined during infancy.5 Early environmental exposures as delivery mode (vaginal delivery versus delivery by caesarean section (CS)) and administration of antibiotics in infancy have previously been found to affect the establishment and diversity of the infants' intestinal microbiota.6, 7, 8, 9, 10, 11, 12, 13, 14 Vaginally delivered children are colonised with bacterial strains from the mothers' vaginal- and gastrointestinal tract during delivery in contrast to children delivered by CS, and these differences seem to persist throughout infancy.10 Furthermore, exposure to (especially broad spectrum) antibiotics in infancy may have long-term implications on the gut microbial composition.6, 11
From twin studies, it has been shown that the gut microbiota differs between lean obese subjects, with increased bacterial diversity in lean individuals.15 Furthermore, studies of rodents have found that the gut microbiota may have a role in energy harvesting, hence suggesting a link to development of obesity.16, 17, 18 If so, the mother's pre-pregnancy body mass index (BMI) may also be expected to influence whether the child (if vaginally delivered) obtains an obesogenic microbiota. In addition to the described obesogenic features, the gastrointestinal microbiota has also been suggested of importance in relation to overall metabolism, hormonal regulation and immunisation.18, 19
The aim of the present study therefore was to investigate whether factors presumably influencing the gut microbiotal transmission from mother to the newborn (that is, mode of delivery, early use of antibiotics and pre-pregnancy BMI) were associated with development of overweight later in childhood, while carefully taking into account several other factors possibly confounding the association, such as socioeconomic status,20 smoking during pregnancy21 and breastfeeding.20, 22
Materials and methods
We studied 28 354 of 40 640 mother–child dyads who participated in the ongoing 7-year follow-up of the Danish National Birth Cohort (DNBC) according to criteria shown in Figure 1 and described below. The birth cohort DNBC was established in 1997–2002 and included 30% of the Danish pregnant population in this period. The mothers were interviewed by telephone, two times during pregnancy (interviews 1 and 2) and two times after pregnancy when the children were 6 and 18 months old, respectively (interviews 3 and 4). The 7-year follow-up was carried out by a questionnaire, filled in by the parents by internet or in paper format. A thorough description of the cohort can be found elsewhere.23, 24
For this study, child inclusion criteria were: live born, term delivered singleton, whose mothers participated in interview 1 and in the 7-year follow-up. Siblings were excluded. Therefore, the mothers only participated with their first pregnancy during the cohort period, which was not necessarily the mothers' first pregnancy in general. Mothers were excluded if they had diabetes (insulin dependent diabetes mellitus or non-insulin dependent diabetes mellitus) or if they developed gestational diabetes or pre-eclampsia during pregnancy. Children were excluded if they had missing, unrealistic or different-date time measurements of height and weight at 7-year follow-up. Children with an age <5 years or age >9 years at the follow-up were also excluded. Left for inclusion in the analyses were 28 354 mother–child dyads. The Data Protection Agency and The Danish National Committee on Biomedical Research Ethics have approved the use of the cohort for this study.
The birth registry was linked to the DNBC, and information on delivery mode, obtained as either vaginal delivery or delivery by CS, was collected. Information on antibiotic prescription in infancy (<6 months of age) was obtained from interview 3 in the DNBC. Antibiotics were given on indication of either ear or lung infection and was used as a dichotomous variable.
The outcome measurement was childhood overweight at 7 years of age. Childhood overweight was assessed by calculating BMI=weight in kilograms per height in m2 and using International Obesity Task Force's (IOTF) age- and sex-specific BMI cutoff scores for overweight.25
We also generated a variable categorising childhood BMI at the 7-year follow-up into four groups: thinness, normal weight, overweight and obesity using IOTF and Cole et al.'s25, 26 extended age- and sex-specific cutoff values for thinness and obesity. Thinness is defined by a BMI <18.5 kg m−2 as an adult if the child remains on the same growth curve.
From interview 1, we used information on maternal age, socioeconomic status, maternal pre-pregnancy height and weight, parity and smoking status. Maternal age was primarily used as a continuous variable, but was also categorised into three categories (age 18–24, 25–35 and >35 years) for group comparisons. Socioeconomic status was categorised into high, medium and low socioeconomic status. The women classified with high socioeconomic status had a job that required a higher education, generally 4 years beyond high school. Parity was coded in three categories: 0 sibling, 1 sibling and 2 or more siblings. The smoking variable was categorised into non-smokers, smoking between 1 and 10 cigarettes per day or above 10 cigarettes per day. Maternal pre-pregnancy BMI was coded into three categories: BMI <25, 25BMI30, BMI >30 kg m−2. From interview 3, we used information on total gestational weight gain and exclusively breastfeeding. Maternal gestational weight gain was grouped into three categories: <12, 12–19.9 and 20 kg. Exclusively breastfeeding was categorised into four categories: never or never fully exclusively breastfeeding, 0<weeks12, 12<weeks20, 20<weeks41. From interview 4, information on paternal height and weight was collected. Paternal BMI was categorised into three categories: BMI <18.5, 18.5–24.9 and 25 kg m−2, but was also used as a continuous variable. Birth weight and delivery season were obtained from the Danish National Birth Registry. Birth weight was coded into three categories: <2500, 2500–4000 and >4000 g, and the date of delivery was used to categorise delivery season into spring (March–May), summer (June–August), fall (September–November) or winter (December–February) seasons. Also, information on child sex and gestational age at birth were obtained from the Danish National Birth Registry.
First, we summarised the covariates in relation to delivery mode and antibiotic exposure by testing group differences using χ2-tests. Furthermore, we examined distributions of overweight and obesity according to covariates. We then tested interactions between exposures and covariates, including smoking status, pre-pregnancy BMI, gestational weight gain, parity, paternal BMI, delivery season, birth weight, breastfeeding and sex, in relation to childhood overweight. Interaction with pre-pregnancy BMI and sex was found. Gender-separated analyses showed less interaction with pre-pregnancy BMI.
We then performed multiple logistic regressions analysing delivery mode (vaginal versus CS) and antibiotic exposure in infancy (No/Yes), and risk of childhood overweight, including analyses either stratified for pre-pregnancy BMI or with an interaction term consistent with the pre-pregnancy BMI group and antibiotic exposure included in the analyses. Our adjusted regression model included maternal age, socioeconomic status, pre-pregnancy BMI, gestational weight gain, parity, smoking status, paternal BMI, birth weight, breastfeeding and age at 7-year follow-up. In the adjusted regression model, we fixated the covariates that were categorised into groups. The reference groups were mothers with high socioeconomic status, no smoking, pre-pregnancy BMI 18.5–25 kg m−2, gestational weight gain 9–16 kg, birth weight 2500–4000 g and exclusively breastfeeding 12–20 weeks. All analyses were first performed for sex together, then for boys and girls separately.
Analyses were performed in STATA (StataCorp version 9.2, College Station, TX, USA).
The prevalence of overweight and obesity in the study population was, respectively, 10.9 and 1.6% among the 7-year old girls, and 8.3 and 1.1% among boys. Overweight and obesity were more often seen in children with siblings (crude odds ratio (OR): 1.12, 95% confidence interval (CI): 1.06–1.18), children with higher birth weight, children of fathers with higher BMI (crude OR: 1.15, 95% CI: 1.14–1.17), children of smoking mothers (crude OR: 1.48, 95% CI: 1.39–1.58) and mothers with higher pre-pregnancy BMI (crude OR: 1.13, 95% CI: 1.12–1.14; Table 1).
Exposures and covariates
As shown in Table 2, delivery mode and administration of antibiotics in infancy were related to the covariates in various ways. Delivery by CS was more frequent in mothers with higher pre-pregnancy BMI, higher gestational weight gain and with children of lower or higher birth weight.
Exposure to antibiotics during the first 6 months of life was the same between children delivered vaginally and children delivered by CS (crude OR: 1.02, 95% CI: 0.88–1.19). Children delivered in summer or fall and children with a higher birth weight (crude OR: 1.16, 95% CI: 1.03–1.30) were at increased risk of receiving antibiotics during the first 6 months of life. Also, children with higher birth weight or children delivered in the winter season were fully breastfed shorter than children with normal birth weight or children delivered in the winter season. Administration of antibiotics in infancy was also more likely in children with siblings (crude OR: 1.63, 95% CI: 1.53–1.75) and children of mothers with higher pre-pregnancy BMI (crude OR: 1.21, 95% CI: 1.06–1.37) and lower socioeconomic status (crude OR: 1.26, 95% CI: 1.16–1.38), as shown in Table 2.
Exposures and childhood overweight
Crude analyses revealed that delivery by CS was associated with the risk of childhood overweight (OR: 1.15, 95% CI: 1.02–1.29), although not after adjustment for covariates (Table 3). The association between delivery mode and childhood overweight was similar for boys (unadjusted OR: 1.18, 95% CI: 1.00–1.47; adjusted OR: 1.18, 95% CI: 0.95–1.47) and girls (unadjusted OR: 1.15, 95% CI: 0.98–1.35; adjusted OR: 1.03, 95% CI: 0.76–1.40). Maternal pre-pregnancy BMI did not modify these associations.
Exposure to antibiotics in infancy was in a crude model found to increase the OR for childhood overweight (OR: 1.20, 95% CI: 1.02–1.40), but the association did not persist when adjusting for covariates (OR: 1.04, 95% CI: 0.79–1.37). However, when assessing the potential modifying role of maternal pre-pregnancy BMI (categorised in three subgroups), an increased risk of childhood overweight in antibiotic-treated infants of normal-weight mothers (adjusted OR: 1.54, 95% CI: 1.09–2.17) was observed (Figure 2). In contrast to this, a small protective effect (inverse association) of exposure to antibiotics in infancy, in relation to childhood overweight, was found in children of overweight mothers (OR: 0.54, 95% CI: 0.30–0.98) compared with children of overweight mothers not receiving antibiotics during the first 6 month of life (Figure 2). A similar tendency was observed among children of obese mothers. Risk associations between antibiotic exposure in infancy and childhood overweight using children of normal-weight mothers not receiving antibiotics in infancy as the reference group revealed that, even though administration of antibiotics in infancy seems to have a protective effect of overweight in children of overweight mothers (without antibiotics: adjusted OR: 1.97, 95% CI: 1.67–2.33 and with antibiotics: adjusted OR: 1.05, 95% CI: 0.58–1.89), these children still carry an increased risk of overweight compared with children of normal-weight mothers (Figure 3). The same observation was made in children of obese mothers (without antibiotics: adjusted OR: 2.85, 95% CI: 2.25–3.62 and with antibiotics: adjusted OR: 2.59, 95% CI: 1.28–5.23).
In unadjusted gender-specific subanalyses, we found that any antibiotic administered during infancy revealed an increased risk of overweight for both boys (OR: 1.33, 95% CI: 0.99–1.79) and girls (OR: 1.37, 95% CI: 1.01–1.87) of normal-weight mothers. However, with adjustments for maternal age, smoking, socioeconomic status, birth weight and breastfeeding, the increased risk only persisted in boys (OR: 1.75, 95% CI: 1.18–2.60) and not in girls (OR: 1.09, 95% CI: 0.66–1.79). The contemporary decreased risk of overweight in antibiotic-treated children of overweight mothers was, on the other hand, only significant for girls (OR: 0.56, 95% CI: 0.32–0.96) in gender-separate unadjusted analyses, which disappeared with adjustment (OR: 0.77, 95% CI: 0.31–1.93). Boys, on the other hand, were observed to have an increased risk of overweight after adjustment (OR: 2.01, 95% CI: 1.05–3.85), which was the same risk as in boys of overweight mothers without having antibiotics early in life (OR 2.03, 95% CI: 1.62–2.54). Type of antibiotics administered to children was known in ∼10% of cases, which limits the type of analyses we can perform.
Sex-separate analyses also revealed a high risk of overweight in boys of normal-weight mothers delivered by CS, with contemporary exposure to antibiotics in infancy (crude OR: 2.18, 95% CI: 1.03–4.61; adjusted OR: 2.33, 95% CI: 0.70–7.77).
When investigating modification from other covariates than pre-pregnancy BMI, we found that another subgroup at increased risk of childhood overweight was children with siblings, in particular boys delivered by CS (unadjusted OR: 1.71, 95% CI: 1.19–2.3). This increased risk persisted with adjustment for covariates (OR: 1.56, 95% CI: 1.03–2.37), but not after additional adjustment for antibiotics early in life (OR: 1.50, 95% CI: 0.98–2.29).
In the present population-based cohort study of 28 354 mother–child dyads followed up for 7 years from birth of the child, we found no overall significant association between delivery mode and risk of childhood overweight. However, a tendency towards increased risk of overweight in boys delivered by CS was found. Exposure to antibiotics in infancy was found to increase the risk of childhood overweight in offspring of normal-weight mothers, while to some extent reducing the risk of overweight in children of overweight or obese mothers.
In the literature, differences have been described in the microbial composition between vaginally delivered children and children delivered by CS, with lower counts of Bifidobacteria and higher counts of Clostridium difficile (that is, a potential obesogenic microbiota) in children delivered by CS.10, 13, 18 In the present study, we only observed a tendency towards an increased risk of overweight in boys delivered by CS. However, the adjustments for many covariates may have decreased the power to observe significant differences, as the ORs did not change with adjustments, but the CIs increased.
Therefore, our results may actually suggest that inoculation with maternal microbiota through vaginal delivery compared with delivery by CS may be particularly protective for childhood overweight in boys. This is further supported by the observed increased susceptibility of overweight in boys delivered by CS and receiving antibiotics during early life.
Further, we found that the risk of childhood overweight depended strongly on maternal pre-pregnancy BMI. This can be explained by the known genetic risk, by epigenetic transmission of risk or by simple transmission of the mother's obesogenic gut microbiota through vaginal delivery. The risk was, however, found to be modifiable by early administration of antibiotics. Interaction analyses revealed that children of mothers with high maternal pre-pregnancy BMI had a decreased risk of overweight with exposure to antibiotics in infancy, whereas children of normal-weight mothers exposed to antibiotics early in life had an increased risk of later overweight. We can only speculate on the reason for these findings. It is possible that differences in the microbial composition of lean and overweight mothers are passed on to the infants during delivery, and that subsequent antibiotic administration in infancy could modify the microbial diversity and composition in, respectively, a positive or negative direction depending on the obtained composition and hence modify the subsequent risk of overweight. This hypotheses is supported by the findings from Collado et al.27 who reported that overweight pregnant women as well as women with high gestational weight gain (>16 kg) had higher microbial concentrations of Bacteroides, Clostridium and Staphylococcus, which have also been found to be higher in 7-year-old overweight children.28
In accordance with our findings, a 7-year follow-up study by Kalliomäki et al.28, investigating the microbial diversity between overweight and lean children, reported that 40% of overweight children had received antibiotics in infancy against 13% (3/24) in the normal-weight group by the age of 6 months. Still, these results were based on only 49 children, with no indication of maternal pre-pregnancy BMI or sex distribution.28 Penders et al.10, 29 reported that oral use of antibiotics during the first month of life resulted in decreased numbers of Bifidobacteria and B. fragilis group species, whereas the first mentioned have been found in higher numbers in lean subjects. Also, perinatal exposure to antibiotics, especially antibiotics of broad spectrum, have been associated to risk of asthma and atopic diseases in childhood, which was explained by long-term changes in the microbiotal composition early in life with lower counts of Bifidobacterium and higher counts of Bacteriodes.6
Before regulatory restrictions, antibiotics were widely used in veterinary foods because of the growth-promoting effect and it has been suggested that this known growth-promoting effect of antibiotics should be considered as a contributor to the obesity epidemic observed in humans.30
As an additional finding, we observed that children with siblings were at increased risk of overweight, which stands in contrast to findings for asthma and atopic dermatitis, and hence does not directly support the hygiene hypothesis suggested for obesity.31 Children with siblings in our study were at increased risk of early infectious diseases and also more frequently received antibiotics early in life than children without siblings. In addition, boys delivered by CS with siblings showed an increased risk of overweight, but this was not the case with vaginally delivered boys having sibling (data not shown). Our data on siblings were obtained when the mothers were pregnant, and therefore reflect older siblings, namely, the number of siblings at the time of delivery. It is likely that the studied children did not remain only-children. The previous literature findings of increased risk of overweight in children without siblings are perhaps due to the children who remain only children.32
The primary strength of our study was the assessment of the DNBC cohort, which originally included over a 100 000 mother–child dyads, with prospectively collected information on a broad range of early environmental exposures. We used this information in the multiple regression analyses to control for a priori confounding factors and covariates, which is another major strength of our study. Our information on delivery mode is complete and we believe that the reliability of this information that was drawn from the birth registry is very high. Also, we used the international classifications of childhood overweight from IOTF. Thereby, it is possible to compare our results with other studies also classifying overweight in children with this international standard.
However, as a potential limitation to the study, many families have been lost to follow-up and this may have affected the number of overweight children represented in the cohort as well as the representation of women in the highest pre-pregnancy BMI groups. Thereby, selection bias may exist. The maternal part of the cohort has been found to be representative, although with overall higher socioeconomic status.33 If selection bias exists with lack of representation of the women with the highest pre-pregnancy BMI, this may have influenced the different relations found with pre-pregnancy BMI strata and exposure to antibiotic in infancy. Further, it is possible that mothers underestimated their weight, especially in the highest BMI groups.34 This could also have biased our findings and the interaction we found with pre-pregnancy BMI.
As another potential limitation, the measurements on the children's height and weight were self-measured and some families gave a height and weight that were from different dates. We decided to exclude children with different date measurements of height and weight from the analyses, which has reduced our number of subjects in the cohort but has increased the validity of our outcome measurement. Several mothers actually gave a height and weight that were measured by their general practitioner or primary nurse. This increased the validity of the measurements, but it also increased the follow-up time span because several of these measurements were from the 5-year child examination. Overall, we believe that height and weight information given on the children are of good validity and no systematic bias should exist.
Exposure to antibiotics during the first 6 months of infancy was obtained from interview 3 and only contained information on antibiotic exposure because of either ear or lung infection. However, the infants may have had antibiotics for other reasons, which may have biased the estimates we found. Antibiotics were given in 8.5% of boys and 6% of girls during the first 6 months of life. Several mothers did, however, not answer the questions about antibiotic use, which could have caused selection bias. We examined the distribution of covariates for these non-respondents (n=5986), and we found no differences (results not shown). Information on the type of antibiotics administered existed only for ∼10% of the children. Of these, the majority who reported the type of antibiotic received either Amoxicillin or Phenoxymethylpenicillin, which are both antibiotics with a broad-spectrum range.
Gender-specific analyses indicated that antibiotics administered in early childhood increased the risk of overweight among both boys and girls. After adjustments, the increased risk only persisted for boys, whereas the protective effect of administration of antibiotics among children of overweight mothers was only apparent among girls. This may either represent a random finding or reflect gender-specific differences in how drugs are metabolised or differences in (intestinal) susceptibility and adaptation to environmental changes. In line with this, it has been suggested that both age, gender and the gut microbiota influence drug metabolism and clearance.35 It therefore appears relevant in future studies to assess modification by gender on the impact of specific types of antibiotics on risk of overweight, while also taking maternal pre-pregnancy BMI into account.
In conclusion, we found no overall significant associations between delivery mode and risk of childhood overweight, but a tendency towards increased risk of obesity in boys delivered by CS. We found that the relation between antibiotic exposure in infancy and the risk of childhood overweight differed across maternal pre-pregnancy BMI groups, with increased risk of overweight in children of normal-weight mothers along with an overall decreased risk of childhood overweight in children of overweight mothers when exposed to antibiotics in infancy. In addition, the transmission of risk of childhood overweight, having an overweight or obese mother, was still stronger than the protective effect found in antibiotic administration early in life, especially in boys.
The study was financed through a Female Research Leader grant (no. 09-066323) from the Danish Council of Independent Research to Dr Tine Jess. Additional support for the DNBC (http://www.dnbc.dk) was obtained from the Pharmacy Foundation, the Egmont Foundation, the March of Dimes Birth Defects Foundation, the Augustinus Foundation and the Health Foundation. The 7-year follow-up study within the DNBC has received financial support from the Lundbeck Foundation (195/04) and the Danish Medical Research Council (SSVF 0646). We acknowledge all the families who are represented in the cohort for their thorough contribution in completing questionnaires. This work was also carried out as a part of the research programme of the Danish Obesity Research Centre (DanORC, see http://www.danorc.dk), granted by the Danish Council for Strategic Research (grant no. 2101-06-0005), and as part of the TORNADO collaboration (http://www.fp7tornado.eu).
About this article
BMC Public Health (2018)