Maternal antibiotic exposure during pregnancy and hospitalization with infection in offspring: a population-based cohort study

Abstract

Background

The early life microbiome contributes to immune development. Antibiotics during pregnancy alter the microbiome and may influence disease risks in the offspring. We investigated the relationship between maternal antibiotic exposure before and during pregnancy, and risk of childhood hospitalization with infection.

Methods

We used population-based Danish national databases for pregnancies between 1995 and 2009. Infants were followed from birth until their first infection-related hospitalization, death, 14th birthday, emigration or end-2009. Exposure was maternal antibiotics prescribed before and during pregnancy. Outcome was infection-related hospitalization.

Results

141 359 (18%) mothers had at least one antibiotic prescription during pregnancy, 230 886 (29.4% of those with complete data) in the 18 months before pregnancy. Of 776 657 live-born singletons, 443 546 infection-related hospitalizations occurred in 222 524 (28.6%) children. Pregnancy antibiotic exposure was associated with increased risk of childhood infection-related hospitalization [hazard ratio (HR) 1.18, 95% confidence interval (CI) 1.17–1.19]. In mothers prescribed antibiotics only during pregnancy whose child did not receive pre-hospitalization antibiotics, this association was present only in those born vaginally. Higher risks of infection-related hospitalization occurred when pregnancy antibiotic prescriptions were closer to birth and in mothers receiving more pregnancy antibiotics. Children born to mothers exposed to antibiotics before (but not during) pregnancy also had increased risk of infection-related hospitalization (HR 1.10, 95% CI 1.07–1.12).

Conclusions

Antibiotic exposure before or during pregnancy was associated with increased risk of childhood hospitalized infections. Alteration of the maternally derived microbiome and shared heritable and environmental determinants are possible contributory mechanisms.

antibiotic, pregnancy, infection, microbiome, hospitalization, child

Key Messages

• Perinatal antibiotics may alter the maternal microbiome during pregnancy and consequently may affect the neonatal microbiome, immune development and infection risk.

• Pre- and perinatal antibiotic exposure is associated with increased risk of infection in offspring. Modulation of the maternal microbiome together with other inherited and environmental factors shared between mother and child are possible mechanisms.

• Prudent antibiotic use prior to and during pregnancy may reduce the severe infection burden in offspring, as well as minimizing broader adverse effects of profligate antimicrobial exposure.

Introduction

Infection is a leading cause of childhood morbidity and mortality worldwide.¹ Approximately 20–30% of children in industrialized countries are hospitalized at least once with an infection,^2,3 largely due to pathogens that infect or colonize the majority, but only cause severe disease in a proportion of children. The mechanisms underlying this differential susceptibility are largely unknown, but may partly reflect shared heritable factors and environmental exposures around pregnancy and in early life.⁴

In addition to the contribution from host genetics, there is increasing evidence that the initial postnatal microbiome is critical to immune development. Fetal microbial colonization may begin in utero and perinatal exposures are major determinants of the infant microbiome.⁵ Mode of delivery has a substantial impact on the infant microbiome;⁶ vaginally born infants largely acquire a maternally derived gastrointestinal and vaginal (endogenous) microbiome, whereas caesarean-section-born infants are initially colonized by skin and nosocomial commensals.⁷ Perturbation of the maternal microbiome during pregnancy, e.g. by antibiotics, may affect the infant microbiome and immune development, and may be associated with differential susceptibility to childhood infection. For example, children exposed to antibiotics during pregnancy have an increased risk of otitis media and insertion of ventilation tubes.⁸

We conducted a population-based cohort study using Danish registry data to investigate the hypothesis that maternal pregnancy antibiotic exposure is associated with increased risk of severe infection in childhood, stratified by mode of delivery. We hypothesized that the risk would be greater in vaginally born children who acquire the maternally derived endogenous microbiome. Since prenatal antibiotic exposure may be indicative of heritable and environmental determinants of infection shared between mother and child, we also hypothesized that maternal pre-pregnancy antibiotics would be independently associated with increased infection-related hospitalization in the child.

Methods

Ethics statement

According to Danish law, when personal identifiers are encrypted and stored by a trusted third party (Statistics Denmark), individual consent is not required for register-based studies. The study was approved by the Danish Data Protection Agency (J.nr. 2008–41–2680) and the Royal Children’s Hospital (Parkville, Australia) Human Ethics Research Committee (34042).

Study population

Denmark has a comprehensive registration system with extensive data on health and social conditions.⁹ Danish residents are assigned a unique personal identification number at birth or immigration that enables linkage of individual information among all national registries. The Danish Medical Birth Register contains data on all live births and stillbirths nationally, including pregnancy and delivery characteristics of mother and child.¹⁰ We identified all live-born, non-adopted singletons born in Denmark where pregnancy began 1 January 1995 through 31 December 2008 (n = 825 106). Children with congenital malformations, identified in the Danish National Hospital Register by ICD-10 codes Q00–Q99 and diagnosed before aged 2 years, were excluded (n = 48 449, 5.9% of total). Of the remaining children, 776 657 were alive at their birth-related hospital discharge and were included in the analysis. 141 359 (18.2%) children were born to mothers with at least one recorded antibiotic prescription during pregnancy (‘exposed’) and 635 298 children were born to mothers without any recorded antibiotic prescription during pregnancy (‘unexposed’).

Prescribed antibiotics

Antibiotic prescription data were obtained from the Danish National Prescription Registry, which contains individual-level information on all prescription medication dispensed at Danish community pharmacies since 1994.¹¹ Data on medication not requiring a prescription, those dispensed at hospitals for outpatient treatment (such as anti-neoplastic drugs or anti-retroviral therapy) and those administered intravenously or during hospitalization are not available.¹¹

Each medication is coded using the World Health Organisation Anatomical Therapeutic Chemical (ATC) classification system with ATC code and date of sale stored in the database when the prescription was redeemed. Antibiotics were defined with ATC codes ‘J01’ (any systemic antibacterial), ‘J01C’ (beta-lactam antibacterials, penicillins), ‘J01A’ (tetracyclines), ‘J01D’ (other beta-lactam antibacterials), ‘J01E’ (sulfonamides and trimethoprim), ‘J01F’ (macrolides, lincosamides and streptogramins), ‘J01M’ (quinolone antibacterials) and ‘J01X’ (other antibacterials).

Mothers were classified as exposed during pregnancy if they filled an antibiotic prescription with the date of sale between 2 weeks prior to the start of pregnancy and the child’s date of birth. Mothers were classified as unexposed during pregnancy if they did not have a filled prescription for any systemic antibiotic in the analogous period. Initiation of pregnancy was calculated by subtracting gestational age in days from date of birth. Pre-pregnancy antibiotic exposure was defined by the 18 months preceding the start of pregnancy, where data were available (n = 711 580, 91.6%).

Infection-related hospitalizations

Data on infection-related hospitalizations were obtained from the Danish National Hospital Register,¹² which holds diagnostic data for all patients discharged from Danish hospitals since 1977. Diagnostic codes are based on the Danish version of the International Classification of Diseases, 10th Revision (ICD-10) from 1994 onward and reported to the register after each hospitalization. Hospital care is tax-paid for Danish residents.

Children were classified as having an infection-related hospitalization if they had an inpatient hospital admission that incurred at least one primary or secondary infectious disease discharge code, at least 1 day after the birth-related discharge date and if they were ≤14 years of age at discharge. Date of onset was defined as the first recorded day of contact with the hospital when patients were hospitalized. Re-hospitalizations for infection within 7 days were considered as single admissions for the same cause. ICD infection codes were classified a priori into seven clinical groups: invasive bacterial, gastrointestinal, lower respiratory tract, skin and soft tissue, upper respiratory tract, genitourinary and viral infections, as previously described.¹³

Covariates

Information on smoking during pregnancy, maternal age, number of siblings, gestational age (weeks), birthweight (grams), 5-minute Apgar score, sex, mode of delivery (vaginal, any caesarean section, elective caesarean section) and birth year were obtained from the Danish Medical Birth Registry.¹⁰ Gestational age was estimated from the date of last menstrual period and adjusted, if appropriate, by ultrasound dating during early pregnancy.¹⁴

Data on parents’ highest level of completed education at the year of birth of the child and population density were obtained from the Danish Civil Registration System database.⁹ Missing values for education were replaced by available information from the preceding or following 5 years, whichever was closest to the date of birth. Education level was grouped as ‘low’ (high-school education or less), ‘middle’ (college or vocational training) or ‘high’ (graduate-level education).

Statistical analysis

We modeled risk of infection-related hospitalizations over time in different maternal antibiotic exposure groups. All children were followed from their birth-related hospital discharge date until the date of their first infection-related hospitalization, death, 14th birthday, emigration or 31 December 2009, whichever occurred first. Directed acyclic graphs (DAGs) were used to identify possible confounders and mediators. DAGs can be used to illustrate and encode the variables in the statistical model and in particular avoid bias arising model selection, e.g. conditioning on a mediator.¹⁵ Multivariable analyses included maternal age at birth (<20, 20–<25, 25–<30, 30–<35, ≥35 years), parents’ education at birth (low, middle, high), smoking during pregnancy (no/yes), sex, number of siblings at birth (0, 1, 2, ≥3), season of birth (spring, summer, autumn, winter), population density [Copenhagen, >100 000 (ex Aarhus, Odense, Aalborg), <100 000 inhabitants (other urban and rural areas)] and birth year (1995–99, 2000–04, 2005–08). Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated by Cox proportional hazard regression models with age as the time-to-event variable and hospitalization as the event. Robust standard errors accounted for interdependency between multiple pregnancies of women during the study period. Plots of log [-log (survival)] against log (survival time), where survival refers to no infection-related hospitalization, were used to evaluate the proportionality assumptions for the use of Cox models. Covariates, other than education, had very few missing data and no imputations were warranted.

We conducted analyses in all children then separately in vaginally and caesarean-section-born children to investigate possible effect measure modification by mode of delivery. We considered type of caesarean section (elective or emergency) in the overall analyses. We estimated risk of infection in different age groups (<1, 1–4, 5–9, 10–14 years of age). We performed analyses stratified by gestational age and birthweight.¹³ These variables were not included in the final model, as they may be intermediates in the causal pathway linking maternal antibiotics and infection-related hospitalization and inclusion may introduce bias.¹⁶

We analysed the number of infection-related hospitalizations (0, 1, 2, ≥3) during the follow-up period until age 5 years using a continuation-ratio model for the analysis of an ordered response categorical outcome.¹⁷ We categorized number of filled maternal prenatal antibiotics (1, ≥2) to examine for a dose–response relationship with infection-related hospitalizations. For this analysis, vaginally and caesarean-section-born children were compared with unexposed vaginally born children. We also investigated timing of prenatal antibiotic prescription, based on the week before birth when the last prescription was filled.

Finally, we investigated antibiotic exposure pre-pregnancy in the mother and postnatally in the child. We defined “pre-pregnancy” as a maternal antibiotic prescription within 18 months preceding pregnancy and “postnatally” as a primary-care-prescribed antibiotic for the child that preceded an infection-related hospitalization, within the first year of life. We categorized the antibiotic exposure by timing relative to birth: pre-pregnancy, during pregnancy or postnatally. Since inpatient prescription data were not available, we restricted these analyses to births ≥36 gestational weeks, as empiric antibiotics are prescribed infrequently for late preterm/term infants. Children were followed to 5 years of age given the considerable cumulative antibiotic exposure by mid-childhood. Statistical analyses were performed in SAS version 9.2 (SAS Institute, Cary, NC, USA).

Results

Children were followed from birth-related hospital discharge to a maximum of 14 years of age (median 4.9/mean 5.7 years of age). During follow-up, a total of 443 546 first-time infection-related hospitalizations occurred in 222 524 children (28.6% of study population). 141 359 (18%) children had mothers with at least one antibiotic prescription during pregnancy. Compared with unexposed, mothers with a filled pregnancy antibiotic prescription were slightly younger, had more caesarean sections and later birth years, and were more likely to report smoking during pregnancy and be primigravida (Table 1).

Table 1

Characteristics of the study population

		Maternal prenatal antibiotic exposure				Chi-squared P-valuea
		Unexposed		Exposed
		N	%	N	%
Characteristic		635 298	82	141 359	18	0.19
Female		311 356	49	69 550	49
Male		323 942	51	71 809	51
Maternal age at birth
	<20	8759	1.4	3537	2.5	<0.0001
	20–<25	75 027	12	21 287	15
	25–<30	223 667	35	49 323	35
	30–<35	225 481	35	46 738	33
	≥35	102 363	16	20 474	14
	Missing	1	0	0	0
Caesarean section		102 229	16	26 402	19	<0.0001
Planned caesarean section
	No	67 306	11	18 229	13	<0.0001
	Yes	34 923	5	8173	6
Parents’ education level
	Low	99 708	16	25 155	18	<0.0001
	Medium	234 491	37	54 817	39
	High	292 633	46	58 627	41
	Missing	8466	1	2760	2
Smoked during pregnancy
	No	495 119	78	104 034	74	<0.0001
	Yes	119 226	19	32 566	23
	Missing	20 953	3	4759	3
Birth weight (grams)
	>2500	611 434	96	135 449	96	<0.0001
	<2500	19 631	3	4938	3
	Missing	4233	1	972	1
Gestational age (weeks)
	>37	606 791	96	134 182	95	<0.0001
	<37	28 463	4	7158	5
	Missing	44	0	19	0
Number of siblings at birth
	0	276 411	43	73 896	52
	1	237 505	37	50 585	36	<0.0001
	2	92 257	15	13 519	10
	≥3	29 125	5	3359	2
	Missing	0	0	0	0
Season of birth
	Winter	150 674	24	33 011	24
	Spring	158 493	25	35 611	25	0.0021
	Summer	165 063	26	37 151	26
	Autumn	161 068	25	35 586	25
	Missing	0	0	0	0
Population density of maternal residence
	Copenhagen	175 391	28	41 098	29
	>100 000 inhabitants	84 864	13	17 434	12	<0.0001
	<100 000 inhabitants	369 150	58	81 996	58
	Missing	5893	1	831	<1
Birth year
	1995–99	211 153	33	42 297	30	<0.0001
	2000–04	246 146	39	46 742	33
	2005–08	177 999	28	52 320	37
Apgar score, 5 minutes
	Mean/SD	9.9	0.6	9.9	0.7
Person-years
	Mean/SD	5.8	4.2	5.1	4.2
Number of infection-related hospitalization diagnoses
	0	457 489	72	96 644	68
	1	102 757	16	24 767	18
	2	36 608	6	9324	7
	≥3	38 444	6	10 624	8

a

Chi-square test of independence for exposure status and dichotomous and categorized measures.

Antibiotic exposure during pregnancy was associated with increased risk for infection-related hospitalization in the child, compared with children born to unexposed mothers (HR 1.18, 95% CI 1.17–1.19). The increased risk persisted throughout childhood in different age groups (Supplementary Table 9, available as Supplementary Data at IJE online). Males had an overall higher risk for infection-related hospitalization (HR 1.28, 95% CI 1.26–1.29) but other analyses were comparable in analyses stratified by sex (data not shown). Increased risks were observed in all antibiotic classes in vaginally born children and in most classes in caesarean-section-born children (Table 2). Risks were similar for elective and non-elective caesarean-section-born children (elective caesarean HR 1.15, 95% CI 1.10–1.20; non-elective caesarean HR 1.14, 95% CI 1.11–1.180). The increment in infection-related hospitalization risk, compared with unexposed, was modestly increased in vaginally than in caesarean-section-born children (vaginal HR 1.18, 95% CI 1.17–1.20; caesarean HR 1.14, 95% CI 1.11–1.17). Unexposed vaginally born children had the lowest rate of infection-related hospitalizations (46.1/1000 children, per year). Exposed caesarean-section-born children had the highest rate (74.8/1000 children, per year). Increased risks were comparable across gestational age and birthweight strata (data not shown).

Table 2

HRs and 95% CIs for any infection-related hospitalization (first occurrence) among antibiotic exposure groups, by all, vaginally born and caesarean-section-born children

All children	Total N	Cases	Incidence rate/year (*103)	Crude HR	Adjusted HRa	95% CI
Unexposed to any antibacterial for systemic use	635 298	177 809	48.1	1.00	1.00	–
Any antibacterial for systemic use	141 359	44 715	61.7	1.21	1.18	(1.17–1.19)
Beta-lactam antibacterials, penicillins	112 787	33 566	59.1	1.22	1.19	(1.17–1.20)
Sulfonamides and trimethoprim	29 956	9715	62.5	1.23	1.20	(1.18–1.23)
Macrolides, lincosamides and streptogramins	14 843	5264	63.5	1.33	1.26	(1.23–1.30)
Other antibacterials	6306	2069	72.2	1.29	1.24	(1.19–1.30)
Tetracyclines	1153	404	66.2	1.35	1.28	(1.15–1.41)
Quinolone antibacterials	613	198	66.8	1.29	1.30	(1.13–1.49)
Other beta-lactam antibacterials	278	110	91.4	1.67	1.59	(1.32–1.92)
Vaginally born
Unexposed to any antibacterial for systemic use	533 069	146 230	46.1	1.00	1.00	–
Any antibacterial for systemic use	114 957	35 834	59.1	1.21	1.18	(1.17–1.20)
Beta-lactam antibacterials, penicillins	91 752	28 502	60.0	1.22	1.19	(1.18–1.21)
Sulfonamides and trimethoprim	24 181	7710	60.0	1.23	1.20	(1.17–1.23)
Macrolides, lincosamides and streptogramins	12 097	4223	60.7	1.33	1.26	(1.22–1.30)
Other antibacterials	4909	1569	68.6	1.29	1.23	(1.17–1.30)
Tetracyclines	896	318	66.3	1.40	1.33	(1.19–1.49)
Quinolone antibacterials	467	138	57.8	1.18	1.22	(1.03–1.44)
Other beta-lactam antibacterials	210	89	93.5	1.80	1.74	(1.41–2.15)
Caesarean-section-born
Unexposed to any antibacterial for systemic use	102 229	31 579	60.3	1.00	1.00	–
Any antibacterial for systemic use	26 402	8881	74.8	1.15	1.14	(1.11–1.17)
Beta-lactam antibacterials and penicillins	21 035	7064	76.6	1.17	1.15	(1.12–1.18)
Sulfonamides and trimethoprim	5775	2005	74.3	1.17	1.17	(1.11–1.22)
Macrolides, lincosamides and streptogramins	2746	1041	78.3	1.28	1.26	(1.18–1.34)
Other antibacterials	1397	500	86.8	1.26	1.21	(1.11–1.33)
Tetracyclines	257	86	65.9	1.12	1.06	(0.85–1.31)
Quinolone antibacterials	146	60	104.0	1.55	1.46	(1.13–1.89)
Other beta-lactam antibacterials	68	21	83.6	1.16	1.07	(0.68–1.67)

a

Adjusted for parents’ education level (low, middle, high), maternal age at birth (<20, 20–<25, 25–<30, 30–<35, ≥35), smoked during pregnancy (no, yes), sex, number of siblings at birth (0, 1, 2, 3+), season of birth, population density (Copenhagen, >100 000, <100 000 inhabitants), birth year (5-year period).

Higher risks for infection-related hospitalization in the child were observed when the last pregnancy antibiotic prescription was filled closer to birth and lowest from 29 to 40 weeks before birth (Figure 1). Increased risks for hospitalization were observed in all clinical infection groups (Figure 2). The lowest risks were observed for invasive bacterial and skin and soft-tissue infections. There was some evidence of a mode of delivery effect for gastrointestinal infections (vaginal HR 1.34, 95% CI 1.30–1.38; caesarean HR 1.25, 95% CI 1.18–1.32).

Figure 1

Hazard ratios and 95% CIs for any infection-related hospitalization (1st occurrence) among those exposed to any antibiotic, by week before birth of last prescription, in all children. Adjusted for parents' education level (low, middle, high), maternal age at birth (<20, 20-<25, 25-<30, 30-<35, ≥35), smoked during pregnancy (no, yes), sex, number of siblings at birth (0, 1, 2, 3+), season of birth, population density (Copenhagen, >100 000, <100 000 inhabitants), birth year (5 year period).

Figure 2

Hazard ratios and 95% CIs for ever having infection-related hospitalization infection subgroup among those exposed to any antibiotic, by mode of delivery. (Solid circle) vaginal–born children, (empty circle) cesarean section-born children. Adjusted for parents' education level (low, middle, high), maternal age at birth (<20, 20-<25, 25-<30, 30-<35, ≥35), smoked during pregnancy (no, yes), sex, number of siblings at birth (0, 1, 2, 3+), season of birth, population density (Copenhagen, >100 000, <100 000 inhabitants), birth year (5 year period).

Sixty-four percent (n = 91 084) of mothers with pregnancy antibiotic exposure had one filled prescription and 36% (n = 50 275) had at least two filled prescriptions. Compared with unexposed vaginally born children, a dose–response was observed with one and at least two pregnancy antibiotic prescriptions, regardless of mode of delivery (Table 3), with the highest risk for one, two or at least three infection-related hospitalizations observed in caesarean-section-born children whose mothers received at least two prescriptions of antibiotics in pregnancy [odds ratio (OR) 1.56, 95% CI 1.50–1.61]. Caesarean-section-born children had overall higher risks for one, two or at least three infection-related hospitalizations, compared with unexposed vaginally born children, even if unexposed (unexposed caesarean OR 1.24, 95% CI 1.23–1.26).

Table 3

ORs and 95% CIs for number of infection-related hospitalizations until age 5 years among those exposed to 0, 1 or ≥2 antibiotics in vaginally and caesarean-section-born children

Mode of delivery	Number of antibiotic prescriptions in pregnancy	1, 2 or ≥3 vs 0 infection- related hospitalizations OR (95% CI)	2 or ≥3 vs 1 infection- related hospitalizations OR (95% CI)	≥3 vs 2 infection- related hospitalizations OR (95% CI)
Vaginally born	0	1.00 (ref)a	1.00 (ref)a	1.00 (ref)a
	1	1.11 (1.09–1.13)	1.04 (1.02–1.07)	1.08 (1.03–1.13)
	≥2	1.26 (1.24–1.29)	1.15 (1.11–1.18)	1.09 (1.03–1.16)
Caesarean-section-born	0	1.24 (1.23–1.26)	1.16 (1.14–1.19)	1.13 (1.08–1.18)
	1	1.34 (1.30–1.37)	1.28 (1.22–1.34)	1.23 (1.12–1.34)
	≥2	1.56 (1.50–1.61)	1.37 (1.30–1.46)	1.22 (1.10–1.35)

a

Reference for all comparisons is unexposed, vaginally born children.

Adjusted for parents’ education level (low, middle, high), maternal age at birth (<20, 20–<25, 25–<30, 30–<35, ≥35), smoked during pregnancy (no, yes), sex, number of siblings at birth (0, 1, 2, 3+), season of birth, population density (Copenhagen, >100 000, <100 000 inhabitants), birth year (5-year period).

When adjusting for pre-pregnancy antibiotic use in the model, the increased risks for antibiotics during pregnancy on infection-related hospitalization were vaginally born HR 1.14, 95% CI 1.12–1.15 and caesarean-section-born HR 1.09, 95% CI 1.05–1.12. Risk for infection-related hospitalization was present during all exposure time frames. Compared with children of mothers unexposed pre-pregnancy, during pregnancy and in the child without postnatal antibiotic exposure, children of mothers exposed to antibiotics only pre-pregnancy had an increased risk for infection-related hospitalization (pre-pregnancy only HR 1.10, 95% CI 1.07–1.12) (Table 4). The risk was similar to that of children of mothers exposed only during pregnancy (HR 1.08, 95% CI 1.05–1.12). In children of mothers who were exposed to antibiotics solely during pregnancy, only those born vaginally had increased infection risk (vaginal HR 1.09, 95% CI 1.05–1.13; caesarean section HR 1.03; 95% CI 0.95–1.11). If exposed both before and during pregnancy, an additive effect was observed (Table 4). Children exposed only to antibiotics postnatally, compared with children of mothers unexposed pre-pregnancy, during pregnancy and postnatally in the child, had greater risk for infection-related hospitalization (HR 1.64, 95% CI 1.62–1.67). Being exposed during all three time frames was associated with the greatest risk for an infection-related hospitalization (HR 2.06; 95% CI 2.02–2.11) (Table 4).

Table 4

HRs and 95% CIs for any infection-related hospitalization (first occurrence) until age 5 years in children born ≥36 weeks gestation, by exposure to antibiotics, either pre-pregnancy*, during pregnancy or postnatally** (in the child) in all, vaginally and caesarean-section-born children

	All children	Vaginally born	Caesarean-section-born
	HR (95% CI)	HR (95% CI)	HR (95% CI)
No antibiotic pre-pregnancy, during pregnancy or postnatally (in the child)	1 (ref)	1 (ref)	1 (ref)
Antibiotic pre-pregnancy only	1.10 (1.07–1.12)	1.09 (1.07–1.12)	1.09 (1.04–1.15)
Antibiotic during pregnancy only	1.08 (1.05–1.12)	1.09 (1.05–1.13)	1.03 (0.95–1.11)
Antibiotic pre AND during pregnancy only	1.21 (1.18–1.24)	1.21 (1.18–1.25)	1.16 (1.09–1.23)
Antibiotic postnatally (in the child) only	1.64 (1.62–1.67)	1.65 (1.63–1.68)	1.59 (1.53–1.64)
Antibiotic pre-pregnancy AND during pregnancy AND postnatally (in the child)	2.06 (2.02–2.11)	2.06 (2.01–2.10)	2.02 (1.93–2.11)

Adjusted for parents’ education level (low, middle, high), maternal age at birth (<20, 20–<25, 25–<30, 30–<35, ≥35), smoked during pregnancy (no, yes), sex, number of siblings at birth (0, 1, 2, 3+), season of birth, population density (Copenhagen, >100 000, <100 000 inhabitants), birth year (5-year period).

*

Pre-pregnancy = antibiotic prescription for the mother within the 18 months preceding pregnancy.

**

Postnatally (in the child) = antibiotic prescription for the child, preceding an infection-related hospitalization, within the first year of life.

Discussion

Maternal antibiotic exposure was associated with increased risk of infection-related hospitalization in the child. For exposure during pregnancy, a dose–response effect occurred and risk increased when antibiotic exposure was closer to birth. Associations were observed with all modes of delivery and were present for all clinical groups of infection, with the highest risk for hospitalization with gastroenteritis in vaginally born children. In subgroup analysis of late preterm/term children whose mothers received antibiotics only during pregnancy, the association was only observed in vaginally born children. Maternal antibiotic exposure pre-pregnancy and postnatally in the child were also associated with increased risk. The greatest risk was for children of mothers exposed both pre-pregnancy and during pregnancy, and who received postnatal antibiotics themselves.

Population data linkage studies seldom provide data to evaluate mechanistic hypotheses. Notwithstanding, there are a number of possible mechanisms that underlie these associations, including shared genetic, social and environmental factors. As antibiotics profoundly affect the microbiome, it is plausible that pregnancy antibiotic exposure contributes to a maternal dysbiosis shared with the offspring. This could increase susceptibility to infection, especially in early childhood, possibly by sub-optimal immune development.¹⁸ The increased risk with greater antibiotic exposure, exposure later in pregnancy and in gastrointestinal infection in vaginally born children is in keeping with a potential role for the microbiome. Similarly, in mothers who only had antibiotics during pregnancy, the increased risk was only observed in vaginally born children, whose initial microbiome is derived from the maternal endogenous microbiome.⁷

Maternal antibiotic exposure pre-pregnancy is less likely to have pervasive effects on the pregnancy microbiome but was also associated with increased risk of infection-related hospitalization. This pre-pregnancy exposure may be indicative of heritable and environmental factors such as general immunogenetic and/or anatomical variation that increase overall susceptibility to infection and are partially shared with the child. Therefore, the increased risk for antibiotics during pregnancy after adjusting for this pre-pregnancy exposure provides additional support for a role of the maternally derived microbiome. Postnatal antibiotics in the child compound the risk of infection-related hospitalization, which may partly reflect an adverse effect on establishment of the postnatal microbiome in childhood. To account for this effect, the risk for first occurrence of an infection-related hospitalization, rather than for multiple occurrences, was modeled. The associations, including those with postnatal antibiotics, were independent of factors that increase childhood infection risk (such as low birthweight), suggesting additional causal pathways are involved.

Associations were observed with all classes of antibiotics and clinical groups of infection. This suggests that putative effects of antibiotics on the microbiome may be via non-specific innate immune responses that predominate early in life, when infection burden is greatest and when adaptive responses are maturing.¹⁹

Previous reports have largely investigated prenatal antibiotics and childhood asthma.²⁰ An analysis of maternal pre-, peri- and postnatal antibiotic exposure and childhood asthma concluded that increased asthma risk likely reflected genetic factors rather than the microbiome, as the effect size was similar for peri- and postnatal antibiotics.²¹ In our study, the modest association between pre-pregnancy antibiotics only and infection-related hospitalization suggests a contribution of shared heritable and environmental factors, but subgroup analysis of those only exposed during pregnancy suggests the maternally derived microbiome is also an important determinant. The associations in children only exposed to antibiotics during the first year of life are partly reflective of a strong correlation (79%) between childhood antibiotics and subsequent infection-related hospitalization.

Our study has a number of strengths, including large, complete, well-maintained population-based registries, with minimal loss to follow-up. The use of data collected for non-research purposes may reduce the risk of differential misclassification. The hospitalization data have been validated²² and the inclusion of both primary and secondary diagnostic codes, as done by others,²³ captures infection burden more accurately than the primary code alone. Hospitalization as an outcome measure largely reflects severe infection, particularly as hospital care is free in Denmark. This outcome is less influenced by health-seeking behaviour, socio-economic status and physician management than primary care or emergency department presentations;^24,25 these data were not available. We were able to adjust for key covariates such as maternal education, birth parameters, smoking and mode of delivery. Changes in infectious disease epidemiology or immunization schedule are unlikely to have influenced the findings, as analysis spanned 14 years and were adjusted for year.

We acknowledge limitations, particularly the possibility that maternal infection itself rather than antibiotic exposure may contribute independently to the association. Details of maternal infections managed in primary care are not available. Our exposure data do not indicate whether the antibiotics were taken as prescribed; compliance in pregnancy is not widely reported.²⁶ Using prescription data assumes that women categorized as exposed took their antibiotics, and that women categorized as unexposed did not take antibiotics during pregnancy that were prescribed prior to pregnancy. If this assumption does not hold, then exposure misclassification (non-differential or differential) is possible and may bias effect estimates in either direction, most likely towards the null. Poor drug compliance may lead to an under-estimation of effect size if correlated with factors that increase risk of childhood infection, such as smoking and lower maternal education. Maternal education, age, smoking, asthma and previous childbirth may be associated with prenatal antibiotic administration,²⁷ which, apart from maternal asthma, were adjusted for in the analyses.

It is possible that the significant association with vaginal but not caesarean-section mode of delivery observed in the subgroup analysis of mothers only receiving pregnancy antibiotics may partly reflect a relative lack of power in the smaller proportion of deliveries by caesarean section (128 000 vs 648 000, respectively). Confirmation of these findings in other large population-based studies is warranted.

Residual confounding by socio-economic status not represented by maternal education may be present; however, other measures were not available for this study. We did not observe variation in risk within strata of gestational age or birthweight, although small numbers of children born very preterm resulted in wide CIs. Data were not available on potential unmeasured confounders, such as prenatal diet, growth and cigarette exposure.²⁸ It is likely that unmeasured exposures later in life contributed to the increased risks at older ages.

Peripartum antibiotics at the time of caesarean section may impact the infant microbiome,²⁹ but these data were not available. Prior to the revised 2012 Danish national guidelines, antibiotics were only routinely administered to emergency caesarean sections, although the extent of variation in use and timing of antibiotics between centers is unknown. Peripartum antibiotic exposure of the mother or neonate might affect the results, in particular the suggested difference between children born vaginally or by caesarean section. To partially address this issue, we showed that the increased risk for hospitalization with infection was similar for all infants born by caesarean section and for those by elective cesarean section when peripartum antibiotics were unlikely to be given. In most scenarios, peripartum antibiotic exposure is an intermediate in the causal pathway and further adjustment might cause bias.³⁰

Population-level data preclude information on potential determinants of the maternal and infant microbiome, such as maternal diet, body composition, stress, animal exposure, infant feeding and introduction of solids. Data on antibiotics administered in hospital were not available. We addressed this in part by subgroup analyses restricted to late preterm/term infants who were unlikely to have received empiric antibiotics postnatally. The strength of associations between maternal pregnancy antibiotics and hospitalization with infection in this subgroup was equivalent to those observed overall, indicating that a high risk group of newborn infants who may have received antibiotics during the birth-related admission were unlikely to have skewed the findings. Group B streptococcus screening is not routinely performed in pregnant women in Denmark. Data on primary care visits for childhood infections were not available. This population-based study was not designed to evaluate specific hypotheses arising from our epidemiological observations; rather, our findings should inform future mechanistic investigations of the maternally derived microbiome.

In conclusion, pre-pregnancy, pregnancy and postnatal antibiotic exposures are associated with increased risk of childhood infection-related hospitalization. Shared heritable and environmental determinants together with alterations of the maternally derived and postnatal microbiome may contribute to differential risk. Further studies with biological sampling and detailed individual-level data would allow investigation of underlying mechanisms, including the relative importance of shared heritable and environmental determinants.

Supplementary Data

Supplementary data are available at IJE online.

Funding

J.E.M. is funded by the Australian National Health and Medical Research Council (NHMRC) Program Grant (572742). D.P.B. is supported by an NHMRC Senior Research Fellowship (1064629) and an Honorary Future Leader Fellowship from the National Heart Foundation Australia. Research at Murdoch Childrens Research Institute is supported by the Victorian Government’s Operational Infrastructure Support Program. L.H.P. is funded by the Central Denmark Region (project 490–79–5601). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Acknowledgements

All authors contributed to the planning and design of the study. J.E.M. and D.P.B. were involved in review of the raw data and directly involved in the analysis. C.S.W., L.H.P., J.O. and N.d.K. provided analytical feedback based on aggregated results. D.P.B. and J.E.M. developed clinical infection categories. J.E.M., D.P.B. and L.H.P. drafted the manuscript, with input from all authors. All authors provided substantive review and commentary on multiple drafts and approved the final version. J.E.M. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Conflict of interest: The authors have no conflicts of interest to declare.