Association of Early Life Exposure to Antibiotics With Risk of Atopic Dermatitis in Sweden

This cohort study examines the association between exposure to antibiotics in prenatal and early postnatal life and the development of atopic dermatitis during childhood.


Introduction
Atopic dermatitis (eczema) is a common inflammatory skin disorder, 1,2 with a global prevalence in children ranging from 7% to 25%, and is associated with significant morbidity and health care costs. 2,3 Atopic dermatitis typically begins in early childhood and is characterized by fluctuating intensely pruritic skin lesions 4 that may or may not persist into adulthood. 2 Children who have atopic dermatitis with specific IgE antibodies are at an increased risk for development of allergic rhinitis, food allergies, and asthma. 4 Atopic dermatitis has a multifactorial origin and is associated with genetic, inflammatory, and environmental factors. 2,5 Although a familial (genetic) history of atopy is the main risk factor, 2 several environmental factors are also associated with disease onset, but these associations are less consistent. 5 One such factor is maternal use of antibiotics during pregnancy. [5][6][7] Exposure to antibiotics in early life has been associated with delayed maturation of the gut microbiome, and the resulting disturbances may negatively affect bacterial diversity. These factors may be associated with altered intestinal compositional states in children and lead to atopy. 8 Despite the potential disruption to the microbiome, prenatal exposure to antibiotics has not been consistently shown to be associated with atopic dermatitis and atopy. 2,6,[9][10][11] In a Belgian study of 773 children by Dom et al, 6 prenatal exposure to antibiotics was found to be associated with atopic dermatitis (odds ratio [OR], 1.82; 95% CI, 1.14-2.92). However, conflicting results were found in other Postnatal exposure to antibiotics may also be associated with a risk of developing atopic dermatitis in early childhood. Some studies have shown that children who are treated with antibiotics in early childhood are at high risk of atopic dermatitis. [13][14][15][16] In a study of 3306 participants by Mai et al 13  To address the observed discrepancies, we conducted a large, population-wide, register-based cohort study of Swedish mother and child pairs to examine the association of antibiotic exposure in early life with the risk of atopic dermatitis in childhood. We further investigated this association within families using sibling control participants to adjust for familial factors and address potential associations. 19 (established in 1973). 20 Standardized maternal and child characteristics for the register are prospectively collected starting at the first prenatal care visit at the antenatal care clinic. 20 Information about family relationships was retrieved from the Multi-Generation Register, 21 which contains links between parent and child or children; vital statistics and migration information were retrieved from the Register of the Total Population. 22 Institutional review board approval was obtained for this project from the Regional Ethics Review Board in Stockholm, Sweden. Informed consent was not required because the study did not involve human participants, and all data were deidentified before analyses were performed. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
The study period was selected to ensure full coverage of dispensed medications from the Swedish Prescribed Drug Register (established in July 2005) 23 during the mothers' pregnancy period and during early childhood. In addition, disease diagnoses from the antenatal period and early childhood were available from the National Patient Register, which has covered all inpatient diagnoses since 1987 and specialist outpatient care since 2001. 24 This register does not include general practice outpatient care.
Children born abroad or whose parents immigrated during the pregnancy period were not included in the study to ensure availability of intrapartum events (n = 8006). All children were then followed up to emigration, death, or the study end on December 31, 2015 (n = 722 767) (eFigure 1 in the Supplement). We also identified 530 590 sibling combinations within the cohort, of which 74 663 were discordant. We linked the different national registers using the unique 10-digit personal identity number that is given to each Swedish resident at birth or immigration.

Exposure
Maternal exposure to systemic antibiotics during pregnancy was defined as a dispensed prescription for antibacterial drugs. Antibiotics were identified using the Anatomic Therapeutic Chemical Classification System code J01. Using a classification developed by Marra et al 10 and modified by Örtqvist et al, 25 we classified the antibacterial drugs according to indication for use into 4 categories: airway, urinary tract, skin, and subcutaneous and other infections. In addition, we divided the aforementioned antibiotics into broad-spectrum and narrow-spectrum categories using a classification previously described by Örtqvist et al 25

Covariates
We selected covariates based on direct acyclic graphs to identify potential confounders recorded in the population registers (eFigure 2 in the Supplement). Independent variables extracted from the

Statistical Analysis
We investigated the association of maternal use of antibiotics during pregnancy and antibiotic use during the first year of life with atopic dermatitis in the offspring separately. Cox proportional hazards regression with attained age as time scale was used to estimate HRs and 95% CIs for atopic dermatitis. The proportional hazards assumption was verified using log-log plots and Schoenfeld residuals plots. Participants were censored at emigration, death, or the end of the study (whichever came first).
For maternal antibiotic use during pregnancy, models were adjusted for sex and for multiple variables, including sex, birth weight, mother's age, family situation, parity, level of education, area of residence, smoking history, maternal history of asthma, and mode of delivery. We included antibiotic use at any time during pregnancy and trimester-specific exposure as exposure variables. We then analyzed the association between maternal antibiotic use during pregnancy and atopic dermatitis by using the likely indication for antibiotic prescription as the exposure. For antibiotic use by the child during the first year of life, adjustments were made for similar covariates: sex, mother's age, family situation, parity, level of education, area of residence, smoking history, maternal history of asthma, and mode of delivery. The models were additionally adjusted for maternal antibiotic use during pregnancy.
A sibling-control analysis was then conducted to account for a shared familial environment for both prenatal life and the first year of life as exposures. Sibling control participants were identified by shared maternity in the Multi-Generation Register. All sibling pairs were included in the analysis, but only those discordant for exposure and outcome were informative for the estimation of the HRs for exposure. The prenatal analysis models were adjusted for sex, mother's age, family situation, smoking, and mode of delivery. In the models in which the first year of life was used as an exposure, maternal use of antibiotics was also included as a covariate. Analysis was done using a stratified Cox proportional hazards regression, and results were reported as HRs and 95% CIs.
We also included a sensitivity analysis that investigated the association of risk of atopic dermatitis with the number of courses of antibiotics dispensed during pregnancy. The association of use of broad-spectrum or narrow-spectrum antibiotics vs no antibiotic use with risk of atopic dermatitis was also assessed. Analyses were performed using Stata, version 16 (StataCorp LLC). 28 Data were analyzed from June 1, 2020, to October 31, 2020.
regard to sex distribution, maternal age, maternal weight, and birth weight. The baseline characteristics of the study population are shown in Table 1.
When the trimester of antibiotic use was used as an exposure, the risk of atopic dermatitis was increased in the multivariable-adjusted models regardless of trimester of use (first trimester: aHR, Sensitivity analysis showed that when children exposed to antibiotics were compared with children who had not been exposed to antibiotics in utero, no difference in risk between children exposed to broad-spectrum antibiotics (HR, 1.08; 95% CI, 1.04-1.14) and those exposed to narrowspectrum antibiotics (HR, 1.08; 95% CI, 1.06-1.10) was observed; however, the risk of atopic dermatitis was higher among children exposed to broad-spectrum or narrow-spectrum antibiotics during the first year of life compared with children not exposed to any antibiotics during the first year of life even after adjustment for maternal use of antibiotics during pregnancy (broad-spectrum vs no observed with antibiotic use during pregnancy; however, among children exposed to antibiotics during the first year of life compared with children not exposed to antibiotics, the risk of atopic dermatitis among children exposed to broad-spectrum antibiotics was similar to that among children exposed to narrow-spectrum antibiotics (eTable 5 in the Supplement).

Discussion
In this large, population-based cohort study, we found that maternal use of antibiotics during pregnancy was associated with a higher risk of atopic dermatitis during early childhood. There was a dose-dependent association of maternal antibiotic use with risk of atopic dermatitis in childhood.
The positive association observed in the whole-cohort analyses with multivariable adjustment did not persist in the sibling-control analyses, in which the risks were significantly attenuated and, in some cases, reduced to null. We also found that use of antibiotics during the first year of life was associated with an increased risk of atopic dermatitis in children, and this association persisted even after controlling for familial and environmental factors in the sibling-control analysis, although the magnitude was attenuated.

Comparison With Other Studies on Antibiotic Use In Utero
Studies conducted on prenatal antibiotic exposure and the risk of atopic dermatitis have had inconsistent results. 6,12,29 In a study of approximately 31 000 Danish children, Stensballe et al 12 found no association between maternal use of antibiotics during the third trimester and risk of atopic dermatitis during childhood. In addition to the factors that we adjusted for, they adjusted for diet, daycare attendance, and seasonality; however, their study population was smaller and had a shorter follow-up period. In contrast, in a study by Timm et al 30  also that maternal allergy may be associated with increased effects of environmental exposures in the offspring. 30 Although we adjusted for maternal history of asthma and did not adjust for maternal atopy, we still found an increased risk of eczema in children whose mothers had more prescriptions.
In addition, similar to the present study, children in the study by Dom et al 6 were followed up from birth to at least 4 years, and these investigators also corrected for postnatal exposure to antibiotics. Our study replicated these results, and adjustment for the trimester of exposure, the likely indication for antibiotic use, and the use of broad-or narrow-spectrum antibiotics did not alter our findings of greater risk. Prenatal exposure to antibiotics may modify gut microbial diversity and delay its maturation. 8 This alteration may be associated with immune dysregulation that leads to atopic dermatitis and other atopic conditions in offspring. 31 Although some previous studies adjusted for birth order or bacterial infections in siblings, to our knowledge, no sibling analysis or co-twin studies investigating prenatal antibiotic use and the risk of atopic dermatitis have been published. The lack of association observed in our study shows that familial and environmental confounding may have affected our observations. antibiotic use in the first years of life was associated with an increased risk of atopic dermatitis even after controlling for genetic and environmental confounders. In addition, they found a greater risk of atopic dermatitis in their co-twin analyses, 14 which was also observed in our sibling-control analyses. Twin pairs are known to have more concurrently shared familial and genetic environments than non-twin siblings. The possibility of unmeasured confounders that change across pregnancies and across circumstances may be associated with upbringing cannot be ruled out in non-twin siblings. 33

Strengths and Limitations
This study has strengths. The large, register-based population allowed for a long follow-up period while minimizing both selection and recall bias. The registers provided access to a large number of confounding factors and allowed us to explore the role of familial factors. The sibling-control design allowed us to adjust for unmeasured genetic and environmental confounders shared by siblings. 19 We were able to adjust for some important covariates, including maternal history of asthma, mode of delivery, maternal history of tobacco use, and area of residence. In an additional analysis, we adjusted for level of education, but this did not alter the main results.
This study also has limitations. We were not able to adjust for all variables; important variables, such as cigarette smoking, which is known to be associated with atopic dermatitis, were only available at antenatal booking, and no information was available about exposure during the first year of life. 34 However, by performing the sibling analysis, we tried to account for smoking during the first year of life for both siblings. In addition, we could not rule out the possibility of residual confounding owing to factors not shared by siblings. We are aware that use of the sibling-control design may have resulted in some potential crossover effects and that the analysis design is more sensitive to measurement error. 35 However, we are confident that the use of sibling comparisons accounted for some important genetic variants and shared environments. Although well-established, validated definitions for diagnosis of established atopic dermatitis exist, finding validated measures of eczema at the population level for children remains a challenge. In this study, we used the algorithm created by Henriksen et al. 26 They included more than 1.5 million children in their cross-national, registerbased cohort study, with approximately 530 000 of those children born in Sweden and overlapping with the population in the present study. We used longer follow-up period, a larger Swedish-born population, and an added sibling analysis.
Another important limitation was potential misclassification of the outcome. Atopic dermatitis medication is usually supplied over the counter, and in general, parents do not take their children to a health care facility for minor cases of the condition. Thus, it is possible that only pediatric patients with the most severe forms of atopic dermatitis or those seen for other medical conditions are reported in the Prescribed Drug Register and National Patient Register. It is also possible that the exposure of 1 sibling influenced the outcome of another. 19

Conclusions
In this cohort study, antibiotic use in prenatal and early postnatal life was associated with risk of atopic dermatitis in early childhood. Although the association between maternal antibiotic use at any time during pregnancy and atopic dermatitis during childhood was confounded by familial (genetic and environmental) factors, the use of antibiotics during the first year of life was associated with risk of atopic dermatitis, although it was partly confounded by familial factors. Further research should be conducted to determine the underlying mechanisms and the atopic dermatitis phenotypes that are more likely to be associated with other conditions, such as asthma and allergic rhinitis.