Introduction

Childhood obesity has emerged as a challenging global public health problem,1 with adverse health consequences in adulthood.2, 3 It has been hypothesized that prenatal exposure to microbiome-altering agents, such as systemic antibacterials, may lead to postnatal obesity.4, 5

In humans, prenatal infection is associated with a 30% increased risk of overweight in later life, and antibiotic exposure in the neonatal period is associated with a 22% increased risk.6, 7 These findings are consistent with the fetal origins hypothesis, which ascribes a critical role of the perinatal period in the development of childhood obesity.4 Antibacterials act by altering fetal exposure to microbe-associated molecular patterns, which have a role in the programming of development of gut epithelium and of the immune system, through Toll-like receptor stimulation.8 Prenatal antibacterials may also act through endocrine disruption9 and metabolic imprinting,10 both of which program postnatal energy homeostasis, food intake and weight set-point,11 leading to childhood overweight and obesity.12 Additionally, evidence in humans13, 14, 15 and in animals15, 16, 17, 18 indicates that antibacterial exposure during pregnancy changes the composition and function of maternal and placental microbiota, potentially configuring the fetal gut microbiomes in an obesogenic direction. It is therefore plausible that treatment with antibacterials during pregnancy can affect intergenerational transmission of maternal microbiota to the offspring.

Because of the rising prevalence of childhood obesity1 and high antibacterial use among women during pregnancy,19, 20 it is important to understand the interplay between maternal antibacterial use during pregnancy and childhood obesity. Previous studies6, 7 were either restricted to adult men or examined the effect of early-life exposure after birth with different hypothesis. To our knowledge, no study to date has examined the association between prenatal exposure to antibacterials and body mass index (BMI) among adolescent schoolchildren. We conducted a prevalence study to examine the association between prenatal exposure to systemic antibacterials and prevalence of overweight and obesity in schoolchildren.

Materials and methods

Study population and setting

We conducted the study among schoolchildren in the Aalborg municipality of Denmark, linking records from several medical and administrative databases. Our research database included all children (N=9886) from 7 to 16 years of age who were born between 1994 and 1998, attended public or private schools in the Aalborg municipality and who underwent routine school anthropometric evaluation during 2002–2013. The median age at the time of the anthropometric evaluation was 14.5 years (interquartile range 14.1, 14.9 years) among schoolchildren with and without prenatal exposure to systemic antibacterials. The Danish National Health Service provides tax-funded universal access to health care, including access to primary care and hospitals, and a partial reimbursement of costs of prescription medications, including antibacterials.21 Individual-level data from all Danish registries can be linked via the unique central personal registry number, assigned at birth or immigration, and registered in the Danish Civil Registration System.22

Prenatal exposure to systemic antibacterials

We linked the children’s school records to their records in the Danish Medical Birth Registry,23 which has tracked all births in Denmark since 1973 and contains a variable for maternal central personal registry numbers. Prenatal exposure to systemic antibacterials was defined as one of the following: (1) a maternal prescription for a systemic antibacterial agent from 30 days before the estimated conception date until childbirth (this period was chosen to capture exposure between fertilization and diagnosis of pregnancy, which could occur via a prescription that was filled before and lasted at least until conception), and/or (2) a maternal inpatient, emergency or outpatient hospital contact with infection between the estimated conception date and childbirth.

Records of maternal prescriptions were obtained, via the maternal central personal registry number, from the Aarhus University Prescription Database, which electronically tracks reimbursed prescriptions sold at all outpatient pharmacies in northern Denmark, including Aalborg municipality, since 1989.24 The prescription record contains date of sale, the amount purchased and the active substance, coded according to the Anatomical Therapeutic Chemical classification. Similarly, the Danish National Registry of Patients was used to obtain data on maternal hospital contacts.25 The Danish National Registry of Patients records complete hospitalization history of all Danish residents since 1977, including dates of admission and discharge and up to 20 discharge diagnoses, coded according to the eighth revision of the International Classification of Diseases until the end of 1993 and the tenth revision thereafter.25

Women without a prescription for a systemic antibacterial agent or a hospital encounter with an infection in the periods defined above were used as the reference group.

Outcome variables

BMI was calculated from height and weight, as recorded by school physicians and nurses during routine anthropometric examinations.26 Weight was measured to the nearest 0.1 kg on an electronic scale (at most schools) or on a mechanical sliding scale. Height was measured to the nearest 0.5 cm using a stadiometer. BMI (weight (kg)/height (m)2) was classified into three non-overlapping categories of normal weight, overweight and obesity, using the International Obesity Task Force age- and sex-specific cutoffs.27

Covariates

Potential confounders measured in the linked data sources were identified a priori and included multiple gestation (yes/no), maternal marital status at the time of childbirth (married/cohabiting or single/divorced/widowed), maternal age at delivery (⩽20/21–35/⩾35 years), maternal smoking during pregnancy (yes/no), maternal diabetes (yes/no), maternal gestational diabetes (yes/no) and child’s gestational age (full-term (⩾37 weeks)/preterm (<37 weeks)).23

All diagnostic and medication codes are listed in the Supplementary Appendix.

Statistical analyses

We described the distributions of the study variables among the schoolchildren according to prenatal exposure to systemic antibacterials. We calculated sex-specific means with s.d. and medians with range of BMI and sex-specific prevalences of overweight and obesity. We computed sex-specific crude and adjusted prevalence ratios (aPRs) of overweight and obesity with associated 95% confidence intervals (CIs) using log-binomial regression28 with generalized estimating equations to model correlated outcomes between children of the same mothers.29 We adjusted the analysis for maternal age at delivery, marital status, smoking during pregnancy and multiple gestation. We did not include gestational age in the regression model, because we worked from the hypothesis that gestational age is a causal intermediary on the pathway between prenatal antibiotic exposure (maternal infection) and obesity; furthermore, we did not include the diabetes variables in the final model, as their prevalence did not differ by the prenatal antibiotic exposure status. We examined the association for exposure in each trimester of pregnancy, for the number of filled prescriptions (0 (reference), 1, 2 and ⩾3) and type of antibacterials. Chi-square test for trend was used to evaluate the dose–response relation between the number of antibiotic prescriptions and the study outcomes. We lacked information on maternal obesity, which could be a potential confounder. Maternal obesity confers a relative risk of 3.1 for childhood obesity and 2.7 for childhood overweight.30, 31 Reported prevalence of obesity is 33% among women taking antibacterials and 32% among women not taking antibacterials. Based on this information, we computed estimates externally adjusted for maternal obesity and compared it with our crude estimates to assess the impact of confounding by maternal obesity.32 Additionally, we examined effect modification by birth weight and mode of delivery.

Data were managed using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA) and analyzed using STATA version 12 (StataCorp. 2011. Stata Statistical Software: Release 12; College Station, TX, USA: StataCorp LP). This study was approved by the Danish Data Protection Agency (number 2009-41-4051). As this registry-based study did not include human biological material, approval by the Danish Scientific Ethical Committee was not needed according to Danish legislation.

Results

Characteristics of the study population

Anthropometric records were available for 9886 schoolchildren, of whom 33% (3280) had prenatal exposure to systemic antibacterials. Median age at the time of anthropometric evaluations was 14 years (range: 7–16 years), and 48% (4718) of the children were girls. Compared with unexposed schoolchildren, exposed children were more likely to be boys, were less likely to be first-born and were more likely to be born to a married or cohabiting mother. In addition, prenatal exposure to systemic antibacterials was associated with a higher prevalence of maternal smoking, caesarean delivery and preterm birth (Table 1). Distributions of exposure to specific antibacterial types and specific hospital infection diagnoses are provided in Supplementary Table S1.

Table 1 Maternal and perinatal characteristics of schoolchildren who underwent anthropometric evaluation during 2002–2013 and were born in the Aalborg municipality of Denmark, according to prenatal exposure to systemic antibacterials
Full size table

Prevalence of overweight

Among all schoolchildren, 7.8% (768/9886) were overweight, including 7.6% (390/5168) of boys and 8.0% (378/4718) of girls (Table 2). The crude PR for overweight was 1.31 (95% CI: 1.14–1.51) and aPR was 1.26, 95% CI: 1.10–1.45). Among boys, the aPR was 1.37 (95% CI: 1.13–1.66), whereas among girls the aPR was 1.16 (95% CI: 0.95–1.42) (Table 3). Stratification by birth weight suggested higher prevalence of overweight in schoolchildren born with weight <3500 g and exposed to prenatal antibacterials compared with non-exposed (aPR: 1.30, 95% CI: 1.05–1.61) than in schoolchildren with birth weight ⩾3500 g (aPR: 1.18, 95% CI: 0.95–1.46). The aPRs were similar in those delivered by vaginal and Caesarean delivery (Table 3). Second-trimester exposure (aPR: 1.39, 95% CI: 1.11–1.73) and third-trimester exposure (aPR: 1.43, 95% CI: 1.17–1.76) were associated with increased prevalence of overweight among schoolchildren but not first-trimester exposure (aPR: 1.09, 95% CI: 0.91–1.31). Prevalence of overweight increased with the increasing number of maternal prescriptions for antibacterials (P=0.001 for trend) (Table 4). PRs did not vary according to antibacterial type (Supplementary Table S2). External adjustment for unmeasured maternal obesity changed the crude PR from 1.31 to 1.33 for the ‘overweight’ outcome.

Table 2 Sex-stratified prevalence of overweighta and obesitya among schoolchildren in Denmark, who underwent anthropometric evaluation during 2002–2013 in the Aalborg municipality of Denmark, by prenatal exposure to systemic antibacterials
Full size table
Table 3 Overall and stratified prevalence ratios (PRs) of overweighta and obesitya among 9886 schoolchildren examined during 2002–2013 who were born in the Aalborg municipality of Denmark
Full size table
Table 4 Prevalence ratios (PR) of overweighta and obesitya, according to trimester of exposure, and the dose–response relation of the number of prenatal antibacterial prescriptions, among 9886 schoolchildren examined during 2002–2013 who were born in the Aalborg municipality of Denmark
Full size table

Prevalence of obesity

The overall prevalence of obesity was 3.1% (309/9886): 3.5% (181/5168) among the boys and 2.7% (128/4718) among the girls (Table 2). The crude PR for obesity was 1.39 (95% CI: 1.11–1.73) and the aPR was 1.29 (95% CI: 1.03–1.62). We observed no difference in aPRs among boys and girls (aPR: 1.29, 95% CI: 0.96–1.73 for boys; and aPR: 1.28, 95% CI: 0.90–1.82 for girls). Among schoolchildren who weighed ⩾3500 g at birth, the association between prenatal antibacterials and obesity was higher (aPR: 1.35, 95% CI: 1.00–1.81) than in those who weighed <3500 g at birth (aPR: 1.16, 95% CI: 0.82–1.65). The aPRs for obesity did not differ across mode of delivery and trimesters of exposure (Tables 3 and 4). We observed an increasing prevalence of obesity with higher numbers of maternal antibacterial prescriptions (P=0.016 for trend; Table 4). There was no difference associated with exposure to specific antibacterials (Supplementary Table S2). The external adjustment for unmeasured maternal obesity did not change the crude PR for obesity among their children.

We further analyzed the effect of prepregnancy antibacterial use on the overweight and obesity among schoolchildren. For overweight, the crude PR was 1.14 (95% CI: 0.99–1.30) and the aPR was 1.08 (95% CI: 0.94–1.24), and for obesity, the crude PR was 1.31 (95% CI: 1.05–1.63) and the aPR was 1.24 (95% CI: 1.00–1.55).

Discussion

Prenatal exposure to antibacterials was associated with a 26–29% increased prevalence of overweight and obesity at school age after adjustment for measured confounding. This association varied by birth weight.

The primary strength of our study was its population-based design and well-defined population receiving care within a uniform health care system. We used independent and highly valid population-based registries.23, 24 The registries prospectively collect data on dispensed prescriptions, and all other variables used in this analysis, years before collection of anthropometric data among the schoolchildren, thus ruling out investigator and recall biases.33 Selection bias based on socioeconomic status can be ruled out, as school anthropometric evaluations were performed at all public and private schools. The prevalence of overweight and obesity among schoolchildren observed in our study was similar to that reported in another recent Danish report,34 indicating that our study population was representative of the source population.

Study limitations include potential unmeasured confounding by maternal gestational weight gain and genetic predisposition to obesity and infection. Second, we did not have information on maternal obesity. The risk of obesity in early childhood is twofold to fourfold higher in children born to obese women compared with children born to non-obese women.31, 35 Obese women have an increased risk of genital tract infections (odds ratio (OR): 1.30, 95% CI: 1.07–1.56) and urinary tract infections (OR: 1.39, 95% CI: 1.18–1.63)36 and are therefore more likely to receive systemic antibacterials. However, external adjustment for maternal obesity had little effect on our estimates. Third, the fact and the timing of maternal medication intake may be misclassified by use of dispensations data, with the extent of misclassification depending on individual adherence to treatment. A recent Danish study reported 37% prevalence of antibacterial use during pregnancy based on self-report, which is close to the 33% prevalence identified in our study and other studies,37 indirectly indicating strong adherence. Fourth, because we used a prevalence design and examined schoolchildren who survived to the age at school anthropometric evaluation, we were able to measure only prevalence instead of incidence of overweight and obesity. However, only 1% of children die between birth and reaching school age in Denmark, suggesting that such mortality is a negligible source of selection bias.38 Finally, confounding by the underlying infection might have attenuated the effect of exposure in our study. We tried to address this issue by exploring the effect of specific antibacterial agents, as they may differ in their effect on normal microbiota (Supplementary Table S2). The number of children with maternal-dispensed prescriptions for antifungal and antiviral agents was too low to address the class effect. Nor did we have data on untreated prenatal infections.

Our results accord with the finding by Cocoros et al.6 of an increase in the prevalence of obesity at age 17–24 years among Danish conscripts born to mothers after hospitalization with an infection during pregnancy (adjusted OR: 1.34, 95% CI: 0.82–2.19). In a recent study, Mueller et al.39 observed that the risk of obesity among children at 7 years of age was increased by 84% (95% CI: 33–154) among those who were exposed to antibacterials during second or third trimester compared with those who were not exposed and by 46% (8–98) among those who were delivered by caesarean section compared with vaginally delivered children. The higher risk of obesity among schoolchildren born by Caesarean delivery observed in our study corroborates findings by Blustein et al.40 who reported 1.83 times higher odds of obesity (95% CI: 1.24–2.70) at age 11 years among children delivered by caesarean section. Additionally, early-life exposure to antibacterials was associated with consistent increase in BMI in later life according to a longitudinal birth cohort study of 11 532 children.7

Plausible mechanisms underlying the observed association may involve altered maternal microbiota through use of antimicrobials during pregnancy. In humans, maternal gut microbiota provides newborns with their first natural microbial exposure,41, 42 while human placenta harbors commensal organisms responsible for antenatal infections.14 Exposure to antimicrobials during gestation15 and in early life43 may affect the postnatal metabolism by altering the composition of the ‘pioneer’ microbiota, ultimately leading, in some persons, to overweight or obesity.40, 44 In addition, use of antibiotics in childhood was associated with an increased risk of obesity in a recent large cohort study.43, 44 Prenatal antibacterial exposure impacts intergenerational transfer of microbiota to infants, which leads to persistent alteration in host immunity45 and can make newborns more susceptible to infections and results in early exposure to antibiotics.46

Another possible mechanism for the findings of this study involves epigenetic pathways. The relevant evidence comes primarily from animal studies.15, 16, 47, 48, 49 In adult rats, prenatal infection-associated exposure to interleukin-6 and tissue necrosis factor-alpha can induce neuroendocrine upregulation of adipose tissue.50 In mice, Cho et al.16 and Cox et al.15 recently demonstrated that antibiotics disrupt early-life metabolic homeostasis and can alter gene expression in ileal tissue, causing microbe-induced obesity in later life. Prenatal exposure to antibiotics may differentially alter methylation at regulatory regions of imprinted genes and may be associated with low birth weight in humans,30 lending support to our hypothesis that somatic epigenetic changes may occur in response to prenatal exposure to antibacterials. This could lead to overweight and obesity at school age mediated by altered development and long-term dysfunction in various organs, including the liver.51 The role of prenatal antibacterial exposure as a possible source of abnormal imprinting raises the issues of its potential effects on postnatal BMI and on other key imprinted genes controlling growth and development.52, 53 However, similar estimates were observed for prepregnancy antibacterial use and most different antibacterial classes increased the risk of obesity (Supplementary Table S2), which favors the hypothesis of intergenerational transmission of maternal microbiota. Alternatively, antibacterial use might be a marker of other risk factors, for instance, lifestyle; otherwise the effect would not have been similar with antibacterial exposure outside pregnancy.

Conclusions

Prenatal exposure to systemic antibacterials was associated with an increased risk of both overweight and obesity at school age. The role of the underlying infection in this association cannot be ruled out. This association varied by birth weight. In addition to lifestyle intervention necessary to reduce obesity epidemic among adolescent, we should also try to find early-life links that may help in finding the key to reduce the epidemic further. There is urgent need to explore early-life modifiable risk factors that can help to reduce later burden on health care as well as on community.