Increased Risk of Childhood Asthma From Antibiotic Use in Early Life^* FREE TO VIEW

Asthma is one of the most common chronic diseases worldwide, significantly impacts quality of life, and represents a significant cost to the health-care system.¹²³ The increasing prevalence of asthma in the industrialized world over the last quarter century has produced several theories on its origins.⁴ In particular, the “hygiene hypothesis” postulates that growing up in a more hygienic environment with less microbial exposure may promote the fetal immune response, which is skewed in the atopic T-helper (Th) type 2 direction, whereas microbial pressure would drive the immune system toward a balanced Th-1 and Th-2 immunity.⁵⁶ Studies⁷ of farmer children have suggested that exposure early in life to endotoxin from Gram-negative bacteria may be the key element of less hygienic environments, which results in a lower prevalence of allergy and asthma. However, many researchers have argued⁸ that the regulation of the immune response is not likely to be dependent on external microbial exposure and have proposed the “microflora hypothesis” of allergic disease. This theory posits that the maturation of the mucosal immune system during infancy, namely, the development of immunologic tolerance via regulatory T cells, requires the presence of commensal microbial flora in the GI tract.⁹ Evidence for this thesis comes from epidemiologic studies¹⁰¹¹¹² that link variations in GI microflora and probiotic administration with less allergy and asthma, and from murine models¹³¹⁴¹⁵ that document that antibiotic administration causes altered intestinal flora, impaired barrier function, diminished Th-1 immune responses, and allergic airway disease.

To date, the findings from epidemiologic studies have supported^{16171819202122} and refuted²³²⁴²⁵ an association between antibiotic use in early life and the development of asthma. Since oral antibiotics are frequently prescribed for upper and lower respiratory tract infections in children,²⁶²⁷ an understanding of the relation between antibiotic use and asthma is critical to clinicians and health-care policymakers worldwide. Previous attempts to assess whether the relationship between antibiotic use in early life and asthma is causal have been hampered by cross-sectional or retrospective study design, in which it is difficult to discern whether the association is subsequent to antibiotic use for wheeze-related respiratory illnesses that precede asthma. A recent metaanalysis²⁸ of antibiotic use in the first year of life has reported a twofold increased risk of childhood asthma following antibiotic use, but no association among studies conducted prospectively. Further, some studies have been limited to high-risk cohorts²³²⁴ or to urban populations.²⁵ We examined the association between oral antibiotic use in the first year of life and asthma at age 7 years in a large cohort of children who were followed up from birth and were living in urban and rural environments with universal access to health-care insurance. As a secondary objective, we tested this association in subgroups of antibiotics and subgroups of children. Preliminary findings from this research have been published in an abstract.²⁹

This was a longitudinal study (known as the Study of Asthma, Genes and the Environment) of a cohort of 13,980 children born in Manitoba in 1995 and continuously registered with the Manitoba Health Services Insurance Plan (MHSIP) until 2003. The likelihood of asthma at age 7 years according to antibiotic prescription use during the first year of life was determined. Data sources were the complete health-care administrative records for the cohort, including all physician visits, hospitalizations, and prescription drugs collected by MHSIP in the provision of universal health insurance to Manitoba residents. MHSIP databases are reliable and valid data sources.³⁰³¹ Database record linkages were achieved through anonymized personal identifiers. A family registration number permitted linkage of maternal and child records. This study was approved by the Health Research Ethics Board at the University of Manitoba and the Health Information Privacy Committee.

Current asthma at age 7 years was defined as at least two physician visits for asthma, one asthma hospitalization, or two prescriptions for any asthma drug (eg, β-agonists, inhaled corticosteroids, cromones, or leukotriene receptor antagonists) in the year following the seventh birthday. This definition was chosen from a validation study³² in a subset of 539 cohort children who were recruited for clinical assessment by an allergist, on the basis of a high positive predictive value (94%; 95% confidence interval [CI], 82 to 99%) and high specificity (92%; 95% CI, 78 to 98%).

Antibiotic use during the first year of life was categorized by the number of oral antibiotic prescriptions (zero, one to two, three to four, and five or more courses, as classified in other publications²⁵ testing the association between antibiotic use and asthma development). Penicillin, cloxacillin, cephalexin, cefadroxil, and erythromycin were defined as narrow-spectrum (NS) antibiotics. The remaining antibiotics fell into the category of broad-spectrum (BS) antibiotics. Physician visits for childhood infections were classified by the number of lower respiratory tract infections (eg, bronchitis, bronchiolitis, and pneumonia), upper respiratory tract infections (eg, otitis media, pharyngitis, and sinusitis), and non-respiratory tract infections (eg, genitourinary infections, cellulitis, and impetigo). Risk and protective factors for asthma, which were derived from health-care administrative records, included gender, urban location (municipality population, > 40,000) or rural location, neighborhood income, total number of siblings at age 7 years, the number of health-care visits made during the first year of life, and maternal history of asthma (at least one physician visit or hospitalization for asthma or one prescription for an asthma drug).

In order to control for reverse causation, 864 children who had received asthma diagnoses during the first year of life were excluded from the study. In a subset of 2,859 children, health-care database records were linked to parental reports of the presence of pets in the home during the birth year of their child. This information was obtained from a mail survey of the 1995 cohort on health and home environmental exposures.

Multivariate logistic regression analysis was conducted with a statistical software package (SAS; SAS Institute; Cary, NC). Variables were retained in models at the 95% level of confidence. Separate multivariate models were tested by antibiotic type (ie, NS and BS antibiotics), and for children living in urban and rural areas, for children with and without a maternal history of asthma, and for children exposed and not exposed to dogs at birth. Although the stratified analyses were based on a priori questions, the interaction term for antibiotic use and the stratum was tested for statistical significance. The type of antibiotic and the timing of its administration were compared among these subpopulations of children.

In the study cohort of 13,116 children at age 7 years, 50% were male, 57% lived in urban areas, 24% were from low-income families, 90% had siblings, 6% had current asthma at age 7 years, and 5% had a maternal history of asthma; 65% of children had received at least one antibiotic prescription during the first year of life, 3% of children had received NS antibiotics and no BS antibiotics, 52% of children had received BS antibiotics and no NS antibiotics, and 10% of children had received both types of antibiotics. The majority of children (55%) had received at least one prescription for a BS penicillin, 9% had received a BS cephalosporin, and 1% had received a BS macrolide (Fig 1 ). The linkage of a physician diagnosis with an antibiotic prescription dispensed within 7 days following the physician visit showed that 40% of children received antibiotics for otitis media, 28% for other upper respiratory tract infections, 19% for lower respiratory tract infections, and 7% for non-respiratory tract infections.

Following adjustment for gender, maternal history of asthma, number of siblings, urban/rural location, and the number of health-care visits (model 1 in Table 1 ), antibiotic use in the first year of life (vs no use) was significantly associated with greater odds of the development of asthma at age 7. This likelihood increased with the number of antibiotic courses, as confirmed by analyses that treated antibiotic use as a continuous variable (odds ratio [OR], 1.09; 95% CI, 1.06 to 1.12). Children who had received more than four courses of antibiotics were almost twice as likely to have asthma develop. Separate adjustments for the number of lower respiratory tract infections and non-respiratory tract infections reduced but did not eliminate the association with asthma.

In a model that adjusted for all risk factors for asthma (Table 2 ), asthma was significantly more likely to develop in children receiving antibiotics in a dose-dependent manner. The multivariate model included all of the factors tested, with the exceptions of neighborhood income and the number of upper respiratory tract infections, which were not significantly associated with asthma at age 7 years. The highest risk of asthma occurred among children receiving more than four courses of antibiotics (OR, 1.46; 95% CI, 1.14 to 1.88). The association between asthma and the use of BS antibiotics was statistically significant (OR, 1.50; 95% CI, 1.16 to 1.93), but this was not the case for NS antibiotics (OR, 1.35; 95% CI, 0.29 to 6.23), although the ORs were not appreciably different from each other.

Asthma at age 7 years was almost twice as likely (OR, 1.86; 95% CI, 1.02 to 3.37) in children receiving one or more antibiotic prescriptions for non-respiratory tract infections in comparison to children who had not received antibiotics. These analyses were adjusted for maternal history of asthma, number of health-care visits, number of siblings, household income, urban/rural location, and gender. A total of 148 children had received antibiotics for non-respiratory tract infections only (33 children had urinary tract infections and the remainder had skin infections), as determined from the linkage of the antibiotic prescription to the preceding physician visit. This risk for asthma following antibiotic use in treating non-respiratory tract infections was greater than that use in treating non-respiratory tract infections alone (OR, 1.15) [Table 2].

The interaction term for antibiotic use and presence/absence of maternal asthma was statistically significant (p = 0.03), as was the interaction term for antibiotic use and urban/rural location (p = 0.04). Limiting the analysis to 7,517 children living in urban areas reduced the association with antibiotic use to nonsignificance (Table 3 ). However, in 5,599 rural children the association remained among children receiving more than four courses of antibiotics (OR, 1.88). Antibiotic use was associated with asthma in children with no history of maternal asthma, especially in children who had received multiple courses of antibiotics (OR, 1.57). The interaction term for antibiotic use and the presence/absence of a dog during the birth year was also statistically significant (p = 0.03). The absence of a dog in the home at birth increased the strength of the association between more than four courses of antibiotics and the development of asthma (OR, 2.02).

The time to the receipt of the first antibiotic was shorter in rural children than in urban children (p = 0.06) [Table 4 ]. While 100% of children in each subpopulation received BS antibiotics, there were differences in the use of BS cephalosporins (eg, cefixime, cefprozil, and cefuroxime). In comparison to urban children, a significantly higher proportion of rural children received BS cephalosporins. The percentage of use of BS cephalosporins was higher, but not statistically significant, in children with no history of maternal asthma or no dog exposure vs their counterparts. No other group differences in antibiotic type were observed.

In a cohort of 13,116 children born in Manitoba in 1995, we found an association between antibiotic use in the first year of life and asthma at age 7 years. Children receiving more than four courses of antibiotics were at 1.5 times the risk of having asthma develop than were children not receiving antibiotics. Our analysis was adjusted for reverse causation, health-care utilization bias, and many well-known risk factors for asthma, and used a health-care database definition of asthma with a high positive predictive value for allergist-diagnosed asthma.³³³⁴ Moreover, the association was observed for antibiotic use in the treatment of children for non-respiratory tract infections, for which the risk of asthma was doubled.

While our findings are consistent with many retrospective surveys and some prospective health-care database studies, they do not agree with a similar database study conducted by Celedon et al.²⁵ These authors reported no association for asthma until age 5 years in children receiving multiple course of antibiotics. We speculate that the discrepant findings are related to differences in the populations studied. Almost half of our cohort of children lived in rural areas, while the study by Celedon et al²⁵ drew its data from an urban population. We also did not find an association with antibiotic use in urban Manitoba children, but the risk was twofold greater in rural children receiving multiple courses of antibiotics.

Similar to its distribution worldwide, asthma prevalence was lowest in our rural children,³⁵ which may be attributed to the protective effect of endotoxin exposure in farming communities in Manitoba.⁷³⁶³⁷ Urban-rural differences in allergy and asthma are also supported by the microflora hypothesis. The gut microbiota of infants in more rural countries has higher levels of anaerobic bacteria, such as lactobacilli or bifidobacteria, and lower levels of Clostridium difficile.¹¹³⁸ These same patterns of gut colonization are more common in nonallergic infants than in allergic infants.³⁹⁴⁰ Antibiotics may alter the protective effect of gut flora in rural children but have little effect on the gut flora of urban children, who are already predisposed to the development of atopic disease. Of interest, Voor et al⁴¹ also reported an association between antibiotic use and atopy in Estonian children and not in more urbanized Swedish infants.

An alternate explanation for the increased risk of asthma from antibiotic use in rural children may be related to the type of antibiotic used, which has not been investigated in intestinal microflora studies of infants.³⁸³⁹⁴⁰ All antibiotics decrease anaerobic microflora in infants, but use of the BS cephalosporins leads to the significant suppression of lactobacilli and bifidobacteria, and to the overgrowth of C difficile.⁴²⁴³⁴⁴ The microflora existing in rural Manitoba children may be subject to the more potent effect of BS cephalosporins. Voor and colleagues⁴¹ have offered a similar account for their findings of increased atopy following antibiotic use in Estonian children, but not in Swedish children; the former children were more likely to have received BS antibiotics. Still another explanation for the increased risk of the development of asthma among the rural children in our study may be related to the earlier administration of antibiotics following birth.⁴⁵

Children who have received therapy with multiple antibiotics and were born to women without a history of asthma were at greater risk of the development of asthma than those who did not receive antibiotics. No antibiotic effect was observed in the presence of maternal asthma, which is consistent with findings from high-risk cohort studies.²⁴ An increased sensitivity to antibiotics in the absence of maternal asthma may be analogous to the protective effects of dog ownership in children with no parental history of asthma.⁴⁶ Postnatal maturation of Th-1 immunity is faster in genetically low-risk children vs high-risk children, so low-risk children are potentially more susceptible to the effects of antibiotic administration on intestinal microflora early in life than are high-risk children,⁴⁷ in whom changes in microflora have already occurred.⁴⁸ Alternatively, as with our rural children, the increased use of BS cephalosporins may explain the increased occurrence of asthma among children with no maternal history of asthma.

Lack of dog exposure during the birth year also increased the association of antibiotic use with the development of asthma among children receiving multiple courses of antibiotics. This concentration of risk for atopy with antibiotic use in children who were exposed to fewer pets in the first year of life has been reported by others.⁴⁵ We hypothesize that lesser contact with dogs during infancy results in a lower microbial load and makes infants more vulnerable to the effects of antibiotics, especially if they are BS cephalosporins.

The strongest evidence against reverse causation in our study is the finding of an association between asthma and antibiotic use for the treatment of non-respiratory tract infections. The majority of non-respiratory infections were skin infections, which may represent misdiagnosed atopic dermatitis. Atopic dermatitis in early life is a major risk factor for asthma, as between 50% and 80% of children with atopic dermatitis will have asthma in childhood.⁴⁹ It is the earliest disease manifestation of future allergic asthma. However, recent findings⁵⁰ on the immunogenetics of asthma and atopic dermatitis have identified the epithelium as a common pathway for the development of both of these diseases. It is plausible that skin infections early in life are manifestations of an impaired barrier function of the epithelium, including the GI epithelium, which leads to allergen penetration and subsequent inflammation. Antibiotic disruption of intestinal microflora may further increase this inflammation by preventing immunologic tolerance via regulatory T cells.⁸ This is suggested in our findings of a greater risk of asthma following antibiotic use in children with non-respiratory tract infections than in those with non-respiratory tract infections alone.

Our population-based study of a 1995 cohort of children identified antibiotic use as a risk factor for the development of asthma at age 7 years. While we have constructed our study to diminish the likelihood of reverse causation and confounding bias, and have implemented a validated definition of childhood asthma, we can neither confirm nor refute the causative role of antibiotics in the development of asthma. However, our study has yielded some interesting findings in subpopulations of children, which we postulate are due to the use of BS cephalosporins or to increased sensitivity to the antibiotic effect among children with a genetic predisposition to impaired barrier function of the epithelium. Further large-scale studies are required to determine the longitudinal associations between the composition of intestinal microflora, antibiotic use, and atopic dermatitis during infancy, and the development of asthma in low-risk and high-risk children. In the interim, it would be prudent to avoid the unnecessary use BS antibiotics in the first year of life when other antibiotics are available.

We would like to acknowledge the computer analysis support of Matthew Dahl and Shamima Huq; as well as the research support provided by Marilyn Lilley, Brenda Gerwing, Michelle Tillett, Ingrid Loewen, Rishma Chooniedass, Tanya Lilley-Chan, Donna Everette, and Joel Liem in the collection of data for the nested case-control study used to validate the asthma definition.