Aim: The aim of this study was to analyse the risk factors for preschool wheeze with special reference to the early introduction of fish and early antibiotic treatment. To avoid reverse causation regarding antibiotics, we focused on the influence of broad-spectrum antibiotics given during the first week of life.
Methods: Data were obtained from a prospective, longitudinal study of a cohort of children born in western Sweden where 50% of the birth cohort was randomly selected. The parents answered questionnaires at 6 and 12 months and at 4.5 years of age. The response rate at 4.5 years was 83% (4496 of 5398 questionnaires distributed).
Results: In the multivariate analysis, broad-spectrum antibiotics in the first week increased the risk of recurrent wheeze (≥3 episodes) during the last 12 months at age 4.5 years (adjusted OR 2.2; 95% CI 1.3–3.8) and multiple-trigger wheeze (aOR, 2.8; 1.3–6.1). The introduction of fish before the age of 9 months reduced the risk of recurrent wheeze (aOR, 0.6; 0.4–0.8).
Conclusion: Treatment with broad-spectrum antibiotics during the first week of life increased the risk of recurrent wheeze and multiple-trigger wheeze at preschool age. The early introduction of fish reduced the risk of recurrent wheeze.
Antibiotic treatment in early life has been reported to increase the risk of future allergic disease (1,2). Alteration of the microflora of the gut following antibiotic treatment has been suggested to affect the development of oral tolerance, inducing changes in the developing mucosal immune system towards a Th2-polarization (1,3). We have previously shown an increased risk of wheezing in infancy following antibiotic exposure during the first week of life (4). Antibiotic exposure during infancy has also been reported to increase the risk of childhood asthma (2). However, the possibility of reverse causation or confounding by indication has been hard to exclude (5).
The composition of fat in the diet, increasing the ratio of n-6 to n-3 polyunsaturated fatty acids (PUFA), has been reported to influence the pathogenesis of allergic disease, acting on inflammatory and immunological pathways (6). Fish, being rich in n-3 PUFA, has been suggested to oppose the action of n-6 PUFA, thus decreasing the risk of allergy (6). In line with this, we have reported a beneficial effect from the early introduction of fish on the risk of eczema in infancy and on allergic rhinitis at preschool age (7,8). It has been suggested that the intake of fish also has a protective effect on the development of asthma (6,9).
Wheezing in early childhood is heterogeneous, with differences in prognosis, pathophysiology and response to treatment (10–12). The temporal pattern and trigger factors of preschool wheeze have been used to delineate the phenotypes of ‘multiple-trigger’ wheeze and ‘episodic viral’ wheeze (11). Children with multiple-trigger wheeze are more likely to develop allergic asthma as they grow up (11). Moreover, differences in pulmonary function and inflammatory response have been reported (13). However, the classification has been questioned, and it has been suggested that it is of transient value in a clinical setting (14).
The aim of this study was to analyse the risk factors for preschool wheeze, with special reference to the effects of early antibiotic treatment and feeding strategies in infancy. To avoid reverse causation regarding antibiotics, we focused on the influence of broad-spectrum antibiotics given during the first week of life. A secondary aim was to analyse possible differences between multiple-trigger and episodic viral wheeze.
Data were obtained from a prospective, longitudinal cohort study of children born in the region of western Sweden in 2003. The region has 1.5 million inhabitants, one-sixth of the Swedish population. It comprises urban, rural and coastal areas, and the largest city is Gothenburg, with 500 000 inhabitants. The random sample comprised 8176 families (50% of the birth cohort).
After written informed consent was obtained, the parents answered questionnaires at 6 and 12 months and at 4.5 years of age. The questionnaires were based on the Swedish version of the ISAAC questionnaire and the questionnaire from the Swedish BAMSE study. Details regarding the questionnaires at 6 and 12 months have been published previously (4,7). For the families that were initially contacted, the response rate at 6 months of age was 69%, while it was 60% at 12 months of age. At 4.5 years of age, questionnaires were distributed to the responders at 6 and/or 12 months, except for those who had declared that they did not want to participate any more. The response rate at 4.5 years was 83%, i.e. 4496 of the 5398 questionnaires distributed (dropout rate 17%). This equals 55% of the families that were initially contacted. After supplementation with data from the Swedish Medical Birth Register (MBR), the database consists of 4171 infants with data from all questionnaires and the MBR.
Information regarding pregnancy and post-natal factors was collected at 6 months of age. In addition, supplementary information regarding pregnancy and delivery was obtained from the MBR, which was the basis of information of the following parameters: caesarean section, gestational age, small for gestational age, large for gestational age, gender and Apgar score.
Admission to a neonatal ward during the first week of life and treatment with broad-spectrum antibiotics during this period was recorded from the 6 months questionnaire. We attempted to verify the questionnaire-based record of neonatal antibiotic treatment with an infectious diagnosis obtained from the MBR. Unfortunately, because of incomplete data on diagnoses in the MBR, this was not possible.
Information regarding the duration of breastfeeding and introduction of different foods (including fish) was collected at 12 months. At 4.5 years of age, questions were asked regarding current health and disease, family, environment and feeding habits.
Questions relating to the outcome variables and important covariates are summarized in Table S1.
In the statistical analysis, contingency tables with χ2-test and binary logistic regression were used. Odds ratios (OR) were estimated with 95% confidence intervals (CI). Crude ORs are indicated as ‘OR’ and adjusted as ‘aOR’.
Factors considered in the univariate analyses are summarized in Table S2. Factors that were significant with a p-value of <0.01 in the univariate analysis (Table 1) were analysed in the multivariate model (Table 2). The cut-off of 0.01 was chosen to focus on the factors with substantially large marginal effect. The multivariate model also controlled for maternal smoking during pregnancy and any breastfeeding for 4 months or more because these factors have previously been shown to affect the risk of childhood asthma (15,16). Parental level of education was included as a marker of socio-economic status. Adjustments were made for all factors simultaneously in one multivariate model.
Table 1. Risk factors with a significance level of p < 0.01 in the univariate analysis for recurrent wheeze at 4.5 years of age
|Atopic heredity (mother or father with asthma, eczema or rhinoconjunctivitis)||181 (73.9)||2505 (59.7)||1.9||1.4–2.6|
|Male gender||153 (62.7)||2190 (51.9)||1.6||1.2–2.0|
|Maternal medication during pregnancy||92 (38.0)||1154 (27.6)||1.6||1.2–2.1|
|Gestational age <37 weeks||22 (9.1)||204 (4.9)||2.0||1.2–3.1|
|Caesarean section||51 (21.2)||582 (14.0)||1.7||1.2–2.3|
|Admission to a neonatal ward||50 (20.5)||426 (10.2)||2.3||1.6–3.2|
|Treatment with antibiotics during the first week||26 (10.7)||176 (4.2)||2.7||1.8–4.2|
|Doctor-diagnosed food allergy during the first year||27 (12.2)||182 (4.7)||2.8||1.8–4.4|
|Eczema during the first year||77 (34.2)||781 (19.9)||2.1||1.6–2.8|
|Introduction of fish before 9 months of age||147 (73.5)||3033 (84.0)||0.5||0.4–0.7|
|Fish once a month or more at 1 year of age||195 (86.3)||3619 (92.5)||0.5||0.3–0.8|
|Fermented food once a month or more at 1 year of age||187 (83.5)||3539 (90.8)||0.5||0.4–0.7|
|Choice of spread at 1 year of age|
| Butter||134 (60.6)||2772 (71.4)||1||Ref.|
| Margarine||68 (30.8)||927 (23.9)||1.5||1.1–2.1|
| No spread||19 (8.6)||182 (4.7)||2.2||1.3–3.6|
Table 2. Multivariate analyses for recurrent wheeze, multiple-trigger wheeze and episodic viral wheeze at 4.5 years of age. Adjustments were made for all factors simultaneously
|Atopic heredity (mother or father with asthma, eczema or rhinoconjunctivitis)||1.7||1.2–2.5||3.6||1.8–7.4||1.2||0.7–1.8|
|Parental education level||0.8||0.6–1.1||0.6||0.4–0.97||1.1||0.7–1.6|
|Smoking during pregnancy||1.1||0.6–1.9||1.1||0.5–2.5||1.1||0.6–2.4|
|Maternal medication during pregnancy||1.3||0.95–1.8||1.9||1.2–3.1||1.0||0.6–1.5|
|Gestational age <37 weeks||1.6||0.9–2.8||2.0||0.9–4.3||1.4||0.7–3.0|
|Treatment with antibiotics during the first week||2.2||1.3–3.8||2.8||1.3–6.1||1.5||0.7–3.4|
|Breastfeeding for 4 months or more||0.7||0.5–1.04||0.9||0.5–1.6||0.6||0.4–1.001|
|Doctor-diagnosed food allergy during the first year||1.9||1.1–3.4||2.8||1.4–5.8||0.98||0.3–2.8|
|Eczema during the first year||1.5||1.05–2.2||1.7||0.97–2.8||1.4||0.8–2.3|
|Introduction of fish before 9 months of age||0.6||0.4–0.8||0.6||0.3–0.99||0.6||0.4–0.99|
|Fish once a month or more at 1 year of age||0.9||0.5–1.8||0.6||0.3–1.2||2.7||0.7–11.5|
The protective effect of the frequent consumption of fermented foods and the increased risk with margarine or ‘no spread’ at 12 months of age were closely associated with own food allergy in infancy (i.e. infants with cow’s milk allergy avoiding yogurts and butter) and were therefore excluded. Admission to a neonatal ward was closely related to receiving neonatal antibiotics and was also excluded.
The results of the multivariate analyses were confirmed using an extended model including factors with p < 0.1 in the univariate analysis (Table S3). The SPSS statistical package version 17.0 (SPSS Inc., Chicago, IL, USA) was used for calculations.
As primary outcome variable, recurrent wheeze (≥3 episodes of wheezing during the last 12 months) at 4.5 years of age was used. However, information was available on no wheeze, 1–2 episodes of wheeze and ≥3 episodes of wheeze during the last 12 months, making it possible to perform a trend analysis and a multinomial regression analysis using the number of wheezing episodes as the outcome variable. Furthermore, among the children with recurrent wheeze, the phenotypes of multiple-trigger wheeze and episodic viral wheeze were analysed. These terms have been suggested by the European Respiratory Society Task Force on wheezing disorders in preschool children (11). ‘Multiple-trigger wheeze’ denotes children who wheeze both during and between viral episodes, while ‘episodic viral wheeze’ denotes children who wheeze intermittently and are well between viral episodes.
The study was approved by the ethics committee at the University of Gothenburg.
Representativeness of the study sample
As reported earlier, the material appears to be largely representative of the population (4). We analysed the differences between children who did not participate in this follow-up and children for whom we had data from the follow-up at 12 months. Among the non-responders, the frequency of parents with a low educational level and mothers smoking during pregnancy and preterm birth was higher. There also was a slightly lower prevalence of atopic heredity and breastfeeding for 4 months or more. No other significant differences were found (Table S4).
At 4.5 years of age, 20% reported at least one episode of wheezing, while 5.5% had recurrent wheeze (i.e. ≥3 episodes of wheezing) during the last year. Of these, 75% reported treatment with asthma medications and 55% reported doctor-diagnosed asthma. Of the children with recurrent wheeze at preschool age, 43% had multiple-trigger wheeze, while 57% had episodic viral wheeze.
Univariate and multivariate analyses
Univariate risk factors (with a p-value of <0.01) for recurrent wheeze are shown in Table 1. The results of the multivariate analyses of risk factors for recurrent wheeze, multiple-trigger wheeze and episodic viral wheeze are shown in Table 2. The results for neonatal antibiotics and early introduction of fish were similar using the extended multivariate model as shown in Table S3.
No significant interaction between atopic heredity and neonatal antibiotic treatment was seen (p = 0.29). When controlling for parental asthma separately in the multivariate analysis, the effect of neonatal antibiotic treatment was still independently significant. Furthermore, no significant interaction was seen between parental asthma and neonatal antibiotic treatment (p = 0.28).
There was some indication of a combined effect between neonatal antibiotics and early fish introduction. The risk of recurrent wheeze in children exposed to neonatal antibiotics was reduced in early fish users (OR 0.3; 0.1–0.7) compared with late fish users (OR 1.00 ref). However, the interaction did not reach significance (p = 0.084).
The impact of neonatal antibiotics, i.e. broad-spectrum antibiotics during the first week of life, could be seen more clearly in children with more frequent wheeze, with 3.6% having received neonatal antibiotics in the ‘no wheeze’ group, 6.5% in the ‘1–2 episodes of wheeze’ group and 10.7% in the ‘≥3 episodes of wheeze’ group (chi-square for linear trend p < 0.0001). The association was confirmed in a multinomial logistic regression using ‘no wheeze’ as the reference category. For ‘1–2 episodes of wheeze’ vs. ‘no wheeze’, aOR for the predictor ‘broad-spectrum antibiotics during the first week of life’ was 1.4 (0.89–2.2); for ‘≥3 episodes of wheeze’ vs. ‘no wheeze’, aOR was 2.4 (1.3–4.2), considering the same confounders as stated in Table 2.
Early fish introduction
The most commonly consumed type of fish was white fish (79%), followed by salmon/game fish (17%), flat fish (3%) and herring/mackerel (1%).
Fish was introduced 1–2 weeks later in children with atopic heredity, eczema or doctor-diagnosed food allergy during infancy. These variables were all controlled for in the multivariate analysis, and we found no significant interaction between them and early fish introduction (parental atopic heredity, p = 0.77; eczema during infancy, p = 0.79; and food allergy in infancy, p = 0.48).
In this study, we report an increased risk of preschool wheeze in children treated with antibiotics during the first week of life. Furthermore, the early introduction of fish reduced the risk of wheeze at preschool age.
Antibiotic treatment in infancy has previously been reported to increase the risk of preschool wheeze, with a dose–response relationship (2). However, in those studies, the possibility of reverse causation cannot be fully excluded. In our study, we have only considered broad-spectrum antibiotic treatment during the first week of life to minimize the risk of reverse causation.
However, neonatal antibiotic treatment could be a marker for asthma rather than a causative factor. A delayed immune maturation at birth is seen in infants born to parents with allergy/asthma, as indicated by cord blood cytokine levels or responses to stimulation in vitro (17,18). Furthermore, there are data linking delayed maturation of the immune system at birth and in early post-natal life with an increased risk of subsequent asthma in children and with an increased frequency of lower respiratory illnesses (19,20). Low production of interferon-gamma at birth or shortly after predicts recurrent wheezing later on (21), and excessive production of IL-5 by T cells at birth is associated with an increased risk of subsequent severe respiratory infections (19). It could be speculated that a delayed immune maturation at birth could increase the susceptibility not only to viral infection but also to other infections, leading to antibiotic treatment.
To minimize the influence of susceptibility to infections at birth, because of a delayed immune maturation, we adjusted for caesarean section and preterm birth, as markers for post-natal vulnerability. Furthermore, we performed the multivariate analysis controlling also for Apgar <7 at 5 min, as a proxy for asphyxia, with stable results for all included variables.
On the other hand, a causal relationship is supported by the findings of an altered mucosal immune response following antibiotic treatment (1). The alteration of gastrointestinal (GI) microflora following antibiotic treatment with long-term changes in GI colonization has been suggested to impair the development of immunological tolerance (1). Alteration of the mechanisms of antigen presentation with disturbance of the T-regulatory-cell response has been discussed (1). Furthermore, impaired gut barrier function and a Th2-polarization of the immune response have been seen in murine models (3,22). Thus, disturbance of the intestinal environment in early life, altering the microflora of the gut and the exposure to microbes and allergens, seems to increase the risk of future allergic asthma. Our results are compatible with this by showing a more pronounced effect of neonatal antibiotic treatment in children with multiple-trigger wheeze, a phenotype more prone to develop allergic (‘true’) asthma (11).
A high maternal intake of fish during pregnancy and the intake of fish by the child itself during infancy have been reported to reduce the risk of later allergic manifestations in the child (6,7,9,23–25). Increasing the ratio of n-6 to n-3 PUFAs, because of an increased intake of linoleic acid, has been reported to influence the development of allergic disease by acting on inflammatory and immunological pathways (6). Fish, being rich in n-3 PUFA, has been suggested to oppose the action of n-6 PUFA, thereby reducing the risk of allergy (6). Other studies have reported that the protective effect of fish on wheeze might be independent of the type of fish ingested, indicating that the effect cannot be ascribed to n-3 PUFA alone (24,26). However, our data do not allow conclusions on the influence of different types of fish on asthma development.
The protective effect we found following the early introduction of fish was independent of atopic heredity, educational level of the parents and allergic disease during infancy. These findings are in line with the reports from the BAMSE study and the LISA study group finding no evidence supporting a delayed introduction of solids to prevent allergic disease and asthma (23,27).
Our results show some differences in risk factors for multiple-trigger wheeze and episodic viral wheeze. For example, the association with antibiotic treatment during the first week of life was only significant for multiple-trigger wheeze. Droste et al. (28) reported an effect of antibiotics early in childhood only in children with atopic heredity. As children with multiple-trigger wheeze are more prone to develop allergic asthma, it makes sense that the impact of neonatal antibiotic treatment was greater in this group. In line with the Asthma Predictive Index, the effect of atopic heredity, as well as food allergy in infancy, was more pronounced in children with multiple-trigger wheeze (29).
The differences in risk factors between the phenotypes indicate differences in pathophysiological mechanisms. Also, differences in terms of pulmonary function and inflammatory markers have recently been reported, supporting the more persistent and allergic nature of multiple-trigger wheeze (13). However, patients have been reported to move from one phenotype to the other over time (14) and the differences seen in clinical features may be obvious when comparing groups but not when classifying individual patients (13). This should be borne in mind when using the terminology in a clinical setting. Furthermore, the phenotypes have been suggested to reflect severity and frequency of wheeze rather than different pathophysiological mechanisms, with multiple-trigger wheeze representing a more severe disease, more susceptible to external exposures and thus triggered more often than the milder episodic viral wheeze (30).
The weaknesses of this study are those inherent in questionnaire-based studies, i.e. there can always be some uncertainty regarding the validity of answers. However, our questionnaire is based on the Swedish part of ISAAC and the Swedish BAMSE study. For example, we have used the Swedish term for wheeze (‘pipande eller väsande’ breathing) used in these two studies. There is good agreement between the Swedish and English wordings. The Swedish word ‘pipande’ means whistling or wheezing, ‘eller’ means or, and ‘väsande’ means hissing or wheezing. Like the English word wheezing, ‘väsande’ is onomatopoetic. Furthermore, 75% of the children in our study reporting recurrent wheeze at preschool age also report current asthma medication. In addition, recurrent symptoms of wheezing in this age group (preschoolers) can be considered to represent asthma. As expected, the responders at 4.5 years were somewhat more diligent and health conscious than the non-responders. This could result in some bias of the sample, but is difficult to avoid in cohort studies.
The strengths of the study are the large size of the birth cohort and the good response rate. In all, we have data from all the questionnaires for more than 4000 children. We consider our multivariate model to be stable and reliable, adjusting for the plausible confounders available. This is supported by the similar results using an extended multivariate model that included factors with p < 0.1 in the univariate analysis. In addition, the reliability of the model was supported by the almost identical results in point estimates when only factors with p < 0.01 in the univariate analysis were included. As reported earlier, the material appears to be largely representative of the population (4).
In conclusion, we report an increased risk of recurrent wheeze and multiple-trigger wheeze at preschool age in children treated with antibiotics during the first week of life. The early introduction of fish significantly reduced the risk of recurrent wheeze.
We thank our colleagues Laslo Erdes, Per Möllborg and Gunnar Norvenius for valuable discussions. The study was supported by the Sahlgrenska Academy at the University of Gothenburg, the Research Foundation of the Swedish Asthma and Allergy Association, the Swedish Foundation for Health Care Sciences and Allergy Research and the Health & Medical Care Committee of the Regional Executive Board, Västra Götaland Region.