In utero or early in life exposure to antibiotics and the risk of childhood atopic dermatitis, a population-based cohort study

Abstract

Background:

Atopic dermatitis (AD) is a common inflammatory disease of the skin that begins early in life and can be lifelong. The purpose of our study was to evaluate whether fetal exposure and/or early life exposure of a child to antibiotics increases the risk of early onset AD.

Objective:

We hypothesize that antibiotic exposure in utero or early in life (e.g., first 90 days) increases the likelihood that children develop AD.

Methods:

Utilizing a large prospectively collected electronic medical records database, we studied the association of antibiotic exposure received in utero or very early in life and the relative risk of onset of AD in a population-based cohort study. Associations were estimated using proportional hazards models as hazard ratios (HR) with 95% confidence intervals (CI).

Results:

The risk of AD in childhood was increased after in utero or early life antibiotic exposure. For any in utero AB exposure the HR was 1.38 (1.36,1.39). However, penicillin demonstrated the strongest association with AD for both in utero exposure, 1.43 (1.41,1.44), and for childhood exposure, 1.81(1.79,1.82). HRs were higher in children born to mothers without AD than those with AD pointing to effect modification by maternal AD status.

Conclusion:

Children born to mothers exposed to antibiotics while in utero had, depending on the mother’s history of AD, approximately a 20 to 40% increased risk of developing AD. Depending on the antibiotic, children who received antibiotics early-in-life had a 40 to 80% increased risk of developing AD. Our study, supports and refines the association between incident AD and antibiotic administration. It also adds population-based support to therapeutic attempts to treat AD by modifying skin microbiome.

Introduction

Atopic dermatitis (AD) or atopic eczema is one of the most common inflammatory diseases of the skin and one of the most common chronic allergic/atopic diseases.¹ For many, it occurs early in life and it can be a lifelong disease.² AD may co-occur with other allergic illnesses (OAI) like asthma, seasonal allergies, and food allergies and often occurs prior to their onset.^2–4 While full elucidation of the pathogenesis of AD remains incomplete, it is thought to involve a complex interplay between genetics, skin barrier dysfunction, immune dysregulation, alterations in the skin microbiome, and exposure to environmental factors.^5–7 Many investigators have shown that AD is strongly associated with T-cell dysregulation and has therefore been classified as an autoimmune illness.^8–11 Patients with AD have altered skin microbiota as compared to those without AD and changes in the skin microbiota are associated with changes in AD disease severity.^12–17

Exposure to antimicrobials can induce changes to the gut and skin microbiota.^15,18 Antibiotics in particular are among the most commonly used drugs in infants and children and exposure to antibiotics has been associated with an increased risk of developing AD in this population.¹⁹ However, very few studies have evaluated antimicrobial exposures during pregnancy (i.e., in utero) and few within the first year of a child’s life and the potential association with AD.^20–28 While these studies have shown an association between antibiotic exposure and subsequent development of AD, the internal and external validity of many of these studies has often been hampered by recall bias, as most have relied on the mother’s recollection of the timing of the child’s antibiotic exposure, antibiotic exposure during pregnancy, as well as the onset of her child’s AD.^{20,24,26–31}

New therapeutics for AD have primarily focused on immune mechanisms, however, social and environmental factors have been associated with AD.^32,33 Exposure to “environmental factors” could begin in utero and/or early in the child’s life. The neonatal in utero environment may be directly affected by the administration of antibiotics, it could have metabolic effects to the mother that are transmitted across the placenta to the neonate, and/or change in the mother’s microbiome.^34–36 With respect to direct changes of the skin microbiome that might induce AD, recent early phase human studies have been based on changing the microbiome of an individual with active AD in order to treat AD in the short term and manage in the long term AD.³⁷

The purpose of our study was to evaluate whether fetal exposure and early life exposure of antibiotics increases a child’s risk of early onset AD. We hypothesize that a child who is born to a pregnant female who received antibiotics during pregnancy is at greater risk of developing AD compared to children born to mothers who did not receive antibiotics and that children exposed to antibiotics are more likely to develop AD compared to children without antibiotic exposure. For this study, we utilized a United Kingdom (UK) patient records database merging primary care health records of mothers and their babies focusing on in utero and early in life antibiotic exposures and a child’s subsequent diagnosis of AD.

Methods

Population:

In the United Kingdom (UK) there are several databases that hold anonymized patient electronic health records. For this study, we used the IQVIA Medical Research Data (IMRD) dataset, which incorporates data from The Health Information Network (THIN), a Cegedim Database. Any reference to THIN is intended to be descriptive of the licensed IMRD dataset. This database includes anonymized longitudinal general practice (GP) data from a patients’ clinical and prescribing records. Medications prescribed utilizing the electronic health record are generally provided via a program from the NHS at minimal cost to the patient and during pregnancies these medications are provided for free. IMRD includes data from 832 practices, across the UK, which cover approximately10% of the UK population, and is believed to be representative of the full UK population.^38,39

Definition of study population:

Diagnoses and symptoms are recorded by practice staff using Read codes, a numerical classification system developed to record health-care related diagnosis and symptoms.^38–40 Data collection for our study was from 2004 to 2021. Our study focused on children that were registered to their GP within 60 days of their reported delivery date and with at least two visits to the GP. To match mother and child, all mother-child pairs had to use the same GP and had to reside in the same household. This method is consistent with prior published work.^39,41 Approximately, one million babies were matched to their mothers using this method. This was a cohort study, the child was the primary unit of analysis, and the mother’s history of antibiotic exposure during pregnancy or the child’s first history of exposure to antibiotics as the primary exposure.

Exposure and outcome definition:

Antibiotic exposure was based on prescriptions entered into the electronic health record. Antibiotics were placed in common categories with the four most common groups being penicillin, macrolide, sulfa (e.g., sulfonamide and combinations), and cephalosporins. The primary outcome of interest was incident AD in the child. The method for ascertaining the diagnosis of AD has been previously validated in this database.⁴²

Analysis:

Descriptive statistics were used to summarize categorical and continuous variables using proportions (percentages, %) and means (standard deviation, SD), or medians (interquartile range. Statistical models were used to evaluate the association of maternal in utero antibiotic use or the association of childhood antibiotic use on the development of AD in the child. As appropriate, additional variables included in the analyses were the mother’s history of illnesses such as AD (well known to be associated with risk of childhood AD), and other allergic illnesses OAI (i.e. seasonal allergies, asthma and food allergy); children’s history of OAI (current or later document), Townsend’s score (index of socioeconomic deprivation reported as quintiles and dichotomized by the two quintiles of greatest social deprivation); ethnicity; and the child’s assigned sex at birth. All babies were followed from the date of delivery until they transferred out of the GP’s practice, died, developed the outcome of interest (AD) or until the end of database reporting period (administrative censoring). All evaluations were based on the presence or absence of the evaluated exposure/covariate. Cox proportional hazards models were used to evaluate the association between mother’s risk factors with the time of onset of AD in the child. Effect estimates are reported as hazard ratios (HR) with 95% confidence intervals (CI). Because a priori the mother’s history of AD was assumed to be strongly associated with a child’s risk of AD, this exposure was also viewed as an effect modifier, therefore HR are reported for children based on their maternal history of AD. With respect to antibiotic exposure in the child, the first exposure time varied. In addition, antibiotic exposure had to occur prior to the child’s diagnosis of AD. Our analyses allowed for clustering within mother, since a mother may have had more than one child. In secondary analyses, we created exposure matched sibling cohorts in which one sibling was antibiotic exposed and one was not exposed. If more than one sibling was available for matching, then one sibling was randomly chosen. The sibling pairs had the same mother, the same GP, and grew up in the same location. Separate cohorts were created for each type of exposure such as in utero penicillin, penicillin in childhood, etc. These secondary studies allowed us to control for environmental and genetic factors that could be associated with AD. These results are reported in a supplement.

Analyses were conducted using Stata MP version 18. This study was reported in accordance with STROBE guidelines for reporting of observational studies using routinely collected data. The study was approved by IMRD, the Scientific Review Committee (SRC) for UK Ethics as protocol number 22SRC042, and the University of Pennsylvania Institutional Review Board (IRB).

Results:

Between 2004 and 2021, 1,023,140 mother-child pairs were identified. Children were followed on average for 10.2 (sd:7.9) years resulting in more than 10 million person-years of follow-up. The average age of diagnosis of AD was 3.2 (sd:4.6) years. Characteristics of the children are displayed in Table 1 and Supplement Table 1. Supplement Table 1 describes characteristics based on antibiotic exposure status. As expected, a history of AD in the mother was highly associated with AD in childhood (HR: 1.71 (95%CI:1.96,1.72)) as was a diagnosis of asthma, seasonal allergies, and food allergies (Table 2). However, risks associated with ethnicity, gender assigned at birth, and Townsend index were much less. In general, during follow up, children with AD were more likely to receive an antibiotic (88.21% (88.46, 88.33) than those without AD 67.34% (67.45,67. 56), respectively).

Table 1:

Characteristics and demographics of the children in the study. Characteristics are also displayed based on the final AD^a outcome. Characteristics based on antibiotic exposure status are available in supplement Table 1. Frequencies are area based on the presence or absence of the covariate and presented as percentage and 95% CI except for age, which is presented as mean and sd.

	Full cohort	Child without AD	Child with AD
Average follow up* (mean, sd)	10.2(sd:7.9)	9.7(7.9)	11.6(7.6)
Ethnicity white (%)	71.22(71.13,71.31)	71.87(71.77,71.97)	69.30(69.12,69.48)
Townsend index (≥ 4)	29.02(28.93,29.11)	29.67(29.57,29.77)	27.10(26.93,27.27)
Child with AD(%)	25.32(25.24,25.40)	0.00(0.00,0.00)	100.00(100.00,100.00)
Child with food allergy(%)	0.90(0.89,0.92)	0.45(0.44,0.47)	2.23(2.18,2.29)
Child with asthma(%)	12.98(12.92,13.05)	10.31(10.24,10.37)	20.88(20.73,21.04)
Child with seasonal allergies(%)	7.96(7.91,8.01)	5.81(5.76,5.87)	14.29(14.16,14.43)
Mother with AD(%)	14.96 (14.90,15.03)	12.54(12.47,12.62)	22.10(21.94,22.26)
Mothers receiving any antibiotic during pregnancy(%)	18.11(18.04,18.19)	17.08(17.00,17.17)	21.15(20.99,21.30)
Children receiving any antibiotic during observation(%)	72.74(72.65,72.82)	67.45(67.34,67.56)	88.33(88.21,88.46)
Penicillin(%)	68.61(68.52,68.70)	63.17(63.07,63.28)	84.64(84.50,84.78)
First 90 days(%)	3.03(3.00,3.07)	2.56(2.53,2.60)	4.42(4.34,4.50)
Cephalosporin(%)	9.73(9.67,9.79)	8.20(8.14,8.26)	14.24(14.10,14.37)
First 90 days(%)	0.14(0.13,0.14)	0.11(0.10,0.12)	0.21(0.20,0.23)
Sulfa (%)	13.22(13.16,13.29)	11.61(11.54,11.68)	17.99(17.84,18.13)
First 90 days(%)	0.29(0.28,0.30)	0.26(0.25,0.27)	0.37(0.34,0.39)
Macrolide(%)	25.33(25.25,25.41)	21.51(21.41,21.60)	36.61(36.42,36.79)
First 90 days(%)	0.42(0.41,0.44)	0.35(0.34,0.36)	0.65(0.61,0.68)
Penicillin mother(%)
First trimester (%)	2.08(2.05,2.11)	2.03(2.00,2.07)	2.22(2.17,2.28)
Second trimester (%)	3.03(2.99,3.06)	2.89(2.86,2.93)	3.42(3.35,3.49)
Second or third trimester (%)	10.76(10.70,10.82)	10.01(9.94,10.08)	12.96(12.83,13.09)
Last trimester(%)	7.73(7.68,7.78)	7.12(7.06,7.17)	9.54(9.43,9.65)
Sulfa mother(%)
First trimester (%)	0.22(0.21,0.23)	0.23(0.22,0.24)	0.20(0.19,0.22)
Second trimester (%)	0.20(0.19,0.21)	0.20(0.19,0.21)	0.21(0.19,0.23)
Second or third trimester (%)	0.74(0.73,0.76)	0.71(0.70,0.73)	0.82(0.79,0.86)
Third trimester (%)	0.54(0.53,0.55)	0.52(0.50,0.53)	0.61(0.58,0.64)
Cephalosporin mother(%)
First trimester (%)	1.31(1.29,1.33)	1.29(1.26,1.31)	1.38(1.34,1.43)
Second trimester (%)	1.24(1.22,1.26)	1.18(1.16,1.20)	1.42(1.38,1.47)
Second or third trimester (%)	3.61(3.57,3.64)	3.37(3.33,3.41)	4.30(4.22,4.38)
Third trimester (%)	2.36(2.33,2.39)	2.19(2.16,2.22)	2.88(2.82,2.94)
Macrolide mother(%)
First trimester (%)	0.33(0.32,0.34)	0.33(0.32,0.34)	0.35(0.33,0.37)
Second trimester (%)	0.41(0.39,0.42)	0.39(0.38,0.40)	0.46(0.43,0.48)
Second or third trimester (%)	1.40(1.38,1.42)	1.29(1.27,1.32)	1.73(1.68,1.78)
Third trimester (%)	0.98(1.01,0.99)	0.88(0.92,0.90)	1.23(1.32,1.27)

^aAD=atopic dermatitis;

^bsd=standard deviation;

^c(%)=percentage;

^dtop 40% most socially deprived

^*Birth to end of study period (e.g., age of child at the end of the study period)

Table 2:

The association of the presence or absence of exposures and factors in Mothers on their children with respect to the onset of childhood atopic dermatitis. Effect estimates are expressed as hazard ratios with 95% confidence intervals. AD- atopic dermatitis

	Full cohort	Moms without AD	Moms with AD
Mother with AD	1.71(1.69,1.72)		100(100,100)
White ethnicity	0.93(0.92,0.94)	0.93(0.93,0.94)	0.98(0.97,1.00)
Townsend index	0.94(0.93,0.95)	0.93(0.92,0.94)	0.94(0.93,0.96)
Seasonal allergies	1.45(1.44,1.47)	1.40(1.38,1.42)	1.25(1.23,1.27)
Asthma	1.18(1.17,1.19)	1.11(1.10,1.13)	1.09(1.07,1.11)
Food allergy	1.59(1.50,1.69)	1.38(1.27,1.49)	1.47(1.35,1.60)
Penicillin	1.43(1.41,1.44)	1.42(1.40,1.44)	1.20(1.18,1.22)
First trimester	1.31(1.28,1.35)	1.35(1.30,1.39)	1.05(1.00,1.11)
Second trimester	1.43(1.40,1.46)	1.45(1.41,1.48)	1.18(1.13,1.23)
Second or third trimester	1.43(1.41,1.45)	1.42(1.40,1.44)	1.22(1.19,1.24)
Third trimester	1.40(1.38,1.41)	1.38(1.36,1.40)	1.21(1.18,1.24)
Cephalosporin	1.35(1.32,1.37)	1.35(1.33,1.38)	1.13(1.09,1.16)
First trimester	1.34(1.30,1.38)	1.35(1.29,1.40)	1.11(1.05,1.18)
Second trimester	1.43(1.39,1.48)	1.44(1.39,1.50)	1.22(1.15,1.30)
Second or third trimester	1.34(1.31,1.36)	1.35(1.32,1.38)	1.13(1.09,1.17)
Third trimester	1.29(1.26,1.32)	1.29(1.26,1.33)	1.08(1.04,1.13)
Sulfa	1.24(1.20,1.29)	1.23(1.17,1.28)	1.10(1.02,1.18)
First trimester	1.12(1.03,1.22)	1.10(0.99,1.22)	1.04(0.89,1.22)
Second trimester	1.32(1.21,1.44)	1.35(1.22,1.49)	1.08(0.92,1.27)
Second or third trimester	1.27(1.22,1.33)	1.26(1.20,1.33)	1.11(1.03,1.20)
Third trimester	1.26(1.20,1.32)	1.23(1.16,1.31)	1.12(1.02,1.22)
Macrolides	1.36(1.32,1.40)	1.32(1.28,1.37)	1.21(1.15,1.27)
First trimester	1.29(1.21,1.37)	1.22(1.13,1.32)	1.19(1.07,1.33)
Second trimester	1.36(1.29,1.44)	1.39(1.30,1.49)	1.09(0.98,1.21)
Second or third trimester	1.37(1.33,1.41)	1.34(1.29,1.39)	1.21(1.15,1.27)
Third trimester	1.37(1.33,1.42)	1.32(1.27,1.38)	1.25(1.18,1.33)

The risks of AD in children exposed to antibiotic in utero are summarized in Table 2. The HR (95% CI) of AD in childhood was increased after in utero antibiotic exposure (HR:1.38 95% CI:(1.36,1.39)). The effect of in utero antibiotic exposure did not depend on the trimester administered (Table 2). Overall, the association between any antibiotic exposure and a child’s risk of AD was greater in children born to mothers who did not have a history of AD (Tables 1 and 2). In utero exposure to penicillin was associated with the largest hazard ratio,1.43 (1.41,1.44). This effect was not significantly altered after adjusting for maternal AD and OAI, maternal exposure to antibiotics, sex of the child assigned at birth, Townsend’s index, and ethnicity. Maternal history of AD is an effect modifier with respect to the risk of all antibiotics (p-value < 0.00001 in all cases) (Table 2).

Table 3 summarizes the risk of AD after exposure to antibiotics in childhood. As noted above, children exposed to penicillin were more likely to develop AD than children not exposed to antibiotics (1.81(1.79,1.82)). This association changed minimally in the adjusted models (Table 3). As noted above these effects were also modified by the mother’s AD status. For example, for penicillin, the effect was greater in children from mothers who did not have AD (1.81(1.79,1.82)) versus mother with AD (1.52(1.50,1.55)). Table 4 summarizes the association between early (within the first 90 days of life) antibiotic and development of AD. For example, children with early penicillin exposure had a 70% increased risk of developing AD compared to children not exposed early to penicillin (1.70(1.67,1.73)). This effect was greater in children born to mothers with no history of AD (1.71(1.68,1.75)) than children born to mothers with AD (1.45(1.40,1.50)). Sensitivity analysis matching siblings exposed to antibiotics and not exposed to antibiotics born to the same mothers and living in the same household resulted in similar findings. These results are available in the supplement (Supplement Table 2).

Table 3:

Association (hazard ratio [95% confidence interval]) between child’s first exposure to antibiotics and development of atopic dermatitis. Fully adjusted model covariates include maternal atopic dermatitis, maternal exposure to antibiotics, gender assigned at birth of the child, child OAI, Townsend’s index, and ethnicity.

	Penicillin N=701,966(68.6%)	Macrolides N=259,159(25.3%)	Cephalosporin N=99,539(9.7%)	Sulfa N=135,291(13.2%)
Unadjusted	1.81(1.79,1.82)	1.54(1.52,1.56)	1.44(1.41,1.46)	1.43(1.40,1.46)
Maternal atopic dermatitis	1.74(1.72,1.76)	1.48(1.46,1.50)	1.38(1.36,1.41)	1.38(1.36,1.41)
Maternal antibiotic exposure during pregnancy	1.74(1.73,1.76)	1.47(1.45,1.49)	1.36(1.34,1.39)	1.37(1.365,1.40)
Fully adjusted model	1.76(1.75,1.78)	1.52(1.50,1.54)	1.42(1.39,1.44)	1.43(1.40,1.46)

Table 4:

Associations (hazard ratio [95% confidence interval]) between antibiotic use within the first 90 days of life and the development of atopic dermatitis.

	Fully cohort	Full cohort*	Mother no AD*	Mother with AD*
Penicillin	1.70(1.67,1.73)	1.76(1.72,1.80)	1.71(1.68,1.75)	1.45(1.40,1.50)
Sulfa	1.46(1.37,1.56)	1.42(1.31,1.53)	1.49(1.39,1.61)	1.24(1.09,1.41)
Cephalosporin	1.70(1.56,1.85)	1.96(1.76,2.18)	1.68(1.52,1.86)	1.53(1.31,1.78)
Macrolide	1.77(1.69,1.86)	1.86(1.75,1.98)	1.81(1.71,1.92)	1.42(1.30,1.56)

Fully adjusted models* used the following covariates maternal atopic dermatitis (full cohort only), maternal exposure to antibiotics, gender assigned at birth of the child, child’s OAI, Townsend’s index, and ethnicity.

Discussion:

In our study, children born to mothers exposed to antibiotics while in utero had approximately a 20 to 40% increased risk of developing AD. Interestingly, we observed that the hazard ratio of a child developing AD following first exposure to antibiotics was greater in children born to mothers who did not have a personal history of AD and that the risk was higher among those exposed to penicillin. Using models that allowed the first exposure to antibiotics to vary with time, children who received antibiotics had a 40 to 80% increased risk of developing AD by age 3. This risk was not statistically different after adjustment and not confounded by the mother’s history of AD and OAIs nor mother’s history of having received antibiotics during her pregnancy. Furthermore, the risk was not significantly altered by the child having a future tendency towards having allergic illnesses as represented by OAI. We also observed that the effect of antibiotic exposure was larger during the first 90 days of life. In addition, similar results to our primary analyses were found when we controlled for unmeasured common family exposures like environmental exposures by matching children who did or did not have antibiotic exposure but were born to the same mother and lived in the same house.

Previous studies have investigated in utero and childhood exposure to antibiotics and the onset of AD. These studies include a retrospective cohort survey study that relied on subject recall using data from the Growing Up Today Study, a cohort of the Nurses’ Health study II.²⁸ This study relied on questionnaires answered retrospectively by mothers and evaluated antibiotic exposure during pregnancy as well as during the first 18 months of life.²⁸ Physician diagnosed AD in the child as reported by the mother was increased after antibiotic exposure (Odds Ratio (OR)=1.44 (1.21,1.72)) and early in life (OR:1.37 (1.19,1.57)).²⁸ The authors noted that their study could have been limited by recall bias and AD self-reporting.²⁸ A separate study of 492 mother-baby pairs, intrapartum antibiotics for more than 24 hours increased the risk of AD by age 2 (Relative Risk (RR)=1.99 (1.13,3.49)) but if administered for less than 24 hours during a vaginal delivery it did not.²¹ A prospective birth cohort of 976 mother baby pairs in China also found an increased risk of eczema (OR=3.59 (1.19,10.85)) in babies exposed to in utero antibiotics.²⁷ A study of 1,080 children from a European birth cohort that used questionnaires to collect information on antibiotic use reported an increased risk of AD after prenatal antibiotic exposure (OR=1.55(1.08,2.24)) and an increased risk of AD after exposure of antibiotics in the first year of life (OR=2.57(1.91,3.44)).²⁴ It is important to note that for most of these studies, recall and ascertainment bias could be problematic as they relied primary of questionnaires administered to the mother later in life. Our findings help support and expand our current understanding of this association and suggest that antibiotic exposure is a potential risk factor for childhood onset AD.

We also observed that use of certain antibiotic classes confers an increased risk of early onset AD. Similar findings have been observed in other studies although the risk varies by type of antibiotics. A prospective study of 370 children initially without AD demonstrated an increased risk of AD in those exposed to macrolide (RR=2.15(1.18,3.91)) and cephalosporin (RR=1.93(1.07,3.49)) antibiotics given in the first year of life.²⁹ Another study utilizing questionnaires also established an increased risk of eczema in a birth cohort of over 62,000 mother baby pairs after early prenatal antibiotics (OR=1.45(1.19,1.76)), however, exposure in the last trimester was not associated with eczema.³⁰ The largest previous study, with similar design to ours, also used health records and merged the Swedish Prescribed Drug Register, National Patient Register, and the Swedish Medical Birth Register.²² About 21.2% of children were exposed to antibiotics in utero and 23.8% were exposed in the first year of life.²² There was an increased risk of AD in children exposed in utero (HR=1.10 (1.09,1.12)) and in the first year of life (HR=1.52 (1.50,1.55)). Using a sibling matched analysis, the in utero effect became null but the early exposure effect was maintained.²² In contrast, a prospective UK based study attempted to better control potential confounders and noted that prenatal antibiotic administration was not associated with childhood AD but antibiotic exposure during a child’s first year of life was associated with childhood AD.²⁰ However, the methods utilized for this study including the selection of confounders appears to be problematic.⁴³ In addition, a study of 2,909 mother baby pairs in Eastern China that used questionnaires and hospital records noted that intra-partum use of antibiotics to treat GBS was associated with an increased risk (OR=2.54 (1.80,3.61)) of AD by age 2.²⁶ Other investigations have demonstrated that intrapartum antibiotic prophylaxis for GBS does have an effect on the infant’s microbiota.²⁵

Our study has several strengths including its large sample size and its rich data source collected prospectively by healthcare providers, from birth to up to on average 10 years of follow up. This allowed us to uniquely analyze the effect of history of maternal atopic illnesses on the association of antibiotic exposure. We also used time varying models to capture variation in antibiotic exposure over time. The limitations of our study include potential errors in linking mother and child. However, the method that we used to link pregnant women to their newborn has been validated in a similar database.^39,41,44 It is possible that protopathic bias (i.e., the antibiotic was prescribed for an early manifestation of AD) was associated with our findings. We were careful to assure that antibiotic exposures occurred prior to the child’s diagnosis of AD (e.g., in utero and first 90 days of life). In addition, the effect of antibiotic administration early in life was mitigated in children with the highest anticipated risk of early onset AD (i.e., those born to mothers with AD, seasonal allergies, food allergies, and/or asthma). Information bias could influence the recording of medical information or the willingness of a parent to bring their child to the GP for examination and treatment. Furthermore, mothers with AD could be more likely to seek healthcare. While this is might be possible, in our cohort, the mothers and children used the same GP (i.e., the medical practice already being used by the mother), all pregnant women and children received free healthcare and prescriptions in the UK. The effect of antibiotics on childhood AD was greater in children whose mothers did not have AD, thereby, decreasing the likelihood that maternal antibiotic use was confounded by other variables. In our secondary analysis, which focused on babies born to the same mother, the effects of antibiotic were like our primary analyses. Medical information evaluated in this study were from the GP record so it is possible for that antibiotics administered only in hospital may not have been recorded in the GP record and therefore not available for evaluation.

In conclusion, this longitudinal cohort study that utilized medical records, supports, and more importantly refines the association between incident AD and antibiotic administration during pregnancy as well as in the early life period. It is not known why exposure to antibiotics in utero and early in life might be associated with AD, but it has been hypothesized that antibiotics may play a role in immune dysregulation due to an impaired development of a robust gut microbiome.^15,18,45 While this is not a new practice guideline, healthcare providers should carefully consider the need for antibiotics before using them.^46,47 While our findings are consistent with this message, causation has not been established and other alternative methods for altering the human microbiome while treating AD are still being developed.^37,48,49