Is the Relationship Between Breastfeeding and Childhood Risk of Asthma and Obesity Mediated by Infant Antibiotic Treatment?

Abstract

Background:

Studies of the protective effect of breastfeeding on asthma have not brought unequivocal results, and thus this issue remains controversial. Antibiotic use, known to increase asthma risk, may be involved in this relationship. The objective of this study was to assess the influence of breastfeeding duration on obesity and asthma risk in childhood and to test a mediating role of antibiotic use in infancy.

Subjects and Methods:

A cross-sectional anthropometric and questionnaire study was conducted on 1,277 schoolchildren 8 years of age. Data on weight status, asthma, breastfeeding duration, antibiotic administration in infancy, socioeconomic status, and lifestyle were analyzed. Multivariate standard and logistic regression and mediation analyses, controlling for confounders, were applied.

Results:

Total duration of breastfeeding was negatively related to the child's body mass index (p=0.038), fat percentage (p=0.030), and obesity risk (p=0.032). Dropping the variable of antibiotic use from the model made the breastfeeding duration a significant predictor of low asthma risk (p=0.027). Antibiotic treatment mediated the relationship between breastfeeding duration and asthma risk (Sobel's z=−2.61, p=0.009).

Conclusions:

Our findings support protective effects of longer duration of breastfeeding against obesity and asthma. We propose a new mechanism for a relationship between breastfeeding and asthma: shorter breastfeeding compromises infant health and thereby leads to antibiotic treatment, which in turn increases the risk of asthma.

Introduction

Evidence is accumulating that breastfeeding provides short- and long-term benefits for infant health in both low-income societies and developed countries, leading to the reduction of the incidence of, among others, respiratory tract, gastrointestinal tract, and urinary tract infections and otitis media.^1–5 Previous research suggested that exclusive and long-term breastfeeding may also lower the risk of overweight,⁶ obesity,⁷ and asthma^8–10—disorders with still increasing incidence worldwide,^11,12 bearing serious physical and psychological consequences for adults and children.^2,13,14 However, due to methodological and ethical limitations and also the complex etiology of those conditions with both genetic and environmental factors involved,^12,14,15 an unambiguous conclusion has not been reached.^5,7,16

The protective effects of breastfeeding may result from the composition of human milk because many of its ingredients, such as hormones, immune, anti-inflammatory, and growth factors, enzymes, and viable cells, act against obesity and asthma and are absent in infant formulas.^17,18 Furthermore, behavioral factors may be involved in the relationship because the amount of food ingested by a child depends to a higher degree on the child when breastfed or the caregiver when bottle-fed.⁷ Finally, the association between breastfeeding and the risk of obesity and asthma may arise from factors with a causal impact on both breastfeeding duration and obesity/asthma risk. These confounding factors may include the mother's body weight, education level, socioeconomic status, and age.⁷

Here we propose another mechanism linking breastfeeding duration to later obesity and asthma, which combines the effects of breastmilk content and parental behavior. Specifically, we propose that these relationships are mediated, at least partly, by antibiotic use in infancy. The proposition builds on the following premises: (1) shorter breastmilk feeding is associated with worse infant health,^2–4,10 and this may result in more frequent administration of antibiotics, and (2) the use of antibiotics in the first year of life increases the risk of developing asthma^19,20 and obesity²¹ in later life.

Using data on an ethnically and geographically homogeneous sample of Polish urban children 8 years of age, we aimed at realizing two goals: first, to verify whether breastfeeding duration affects the risk of asthma, increased body weight, and fatness; and second, to determine whether the application of antibiotics in infants mediates the presumed relationship between breastfeeding duration and the risk of asthma and obesity in later life.

Subjects and Methods

Participants

The participants were pupils of 31 randomly selected primary schools in Poznań, a relatively large city (approximately 560,000 inhabitants) in the western part of Poland. Of 1,745 children invited to take part in our project, 1,353 (77.5%) provided written informed consents signed by their parents. The study finally enrolled a sample with all anthropometric data available consisting of 1,290 children (672 girls and 618 boys) 8 years of age (born in 1999, examined in 2007). Some children were later excluded from the analysis for the following reasons: posture defect (n=2), gross measurement error (n=5), and one twin from each of same-sex pairs (n=6). This reduced the sample to 1,277 individuals (668 girls and 609 boys). Mean age was 7.81 (standard deviation=0.31) and 7.82 (standard deviation=0.31) years for girls and boys, respectively. The sample was ethnically homogeneous, and all children were white.

Measurements

Anthropometric examination was carried out at schools. Three trained researchers measured stature (with a stadiometer, to the nearest 1 mm), body weight (bathroom balance, to the nearest 0.1 kg), and thickness of triceps, subscapular, and abdominal skinfolds (Harpenden [Ann Arbor, MI] skinfold caliper, to the nearest 0.1 mm). Body mass index (BMI) was calculated as body weight (in kg) divided by the squared height (in m). Body fat percentage was estimated from skinfold measurements with formulas established for prepubescent children.²² Children with BMI values above the 85th or 95th percentile were classified as overweight or obese, respectively.²³ The percentile values for Polish urban 8-year-old children were taken from Nawarycz and Ostrowska-Nawarycz.²⁴

Questionnaires

The participants' parents filled in two anonymous questionnaires. The first one gathered data on biological and social variables. The crucial question for this study was about breastfeeding duration: “How long was the child breastfed?” We thus asked about the duration of total rather than exclusive breastfeeding because of more accurate recall of the former²⁵ and the evidence of the protective effect of partial and prolonged breastfeeding,²⁶ and total duration of breastfeeding may even be a more important determinant of asthma in childhood than its exclusivity.¹⁶ Other questions inquired about the characteristics likely associated with asthma and body mass (Tables 1 and 2); they were used as covariates in further analyses (see below).

Table 1.

Basic Statistics of Quantitative Variables

	Girls		Boys
Variable	n	Mean (SD)	n	Mean (SD)
Breastfeeding duration (months)	462	8.35 (8.53)	421	8.04 (8.26)
Body height (cm)	668	128.8 (5.75)	609	129.68 (5.96)
Body weight (kg)	668	27.96 (5.81)	609	28.96 (6.09)
Fat percentage (%)	556	19.29 (6.59)	526	18.16 (7.53)
BMI (kg/m²)	668	16.73 (2.52)	609	17.10 (2.62)
Birth mass (kg)	458	3.282 (0.578)	428	3.417 (0.600)
Gestational age (weeks)	381	39.51 (1.94)	368	39.14 (2.09)
Maternal BMI (kg/m²)	445	22.34 (3.24)	414	22.27 (3.42)
Paternal BMI (kg/m²)	405	26.31 (3.14)	389	26.22 (3.46)

BMI, body mass index; SD, standard deviation.

Table 2.

Basic Statistics and Coding of Qualitative Variables

Variable	n	%
Sex
0=girl	668	52.3
1=boy	609	47.7
Sibship
0=no sibling	269	31.2
1=at least one sibling	592	68.8
Antibiotic use in infancy
1=during first 12 months	433	53.1
0=otherwise	383	46.9
Passive smoking
1=in infancy or currently	403	49.6
0=otherwise	410	50.4
Milk consumption
0=no more than twice a week	195	24.3
1=otherwise	609	75.7
Fast-food consumption
0=less than once a week	725	92.1
1=otherwise	62	7.9
Physical activity with increased breathing
0=less than once a week	194	24.3
1=at least once a week	605	75.7
TV watching
0=up to 3 hours/day	652	80.7
1=exceeding 3 hours/day	156	19.3

The second questionnaire was the Polish version of the one developed within the International Study of Asthma and Allergy in Childhood (ISAAC) (see the original version at http://isaac.auckland.ac.nz/phases/phasetwo/phasetwomodules.pdf). It screens for three allergic diseases in childhood—asthma, rhinitis, and eczema—and inquires about diet and physical activity. An allergist (A.Z.), based on the data from the ISAAC questionnaires, classified the children examined into three categories: asthma absent, asthma present, and probable asthma. Asthma absent included children with negative answers in the ISAAC module 1.1, and asthma present embraced children with doctor-diagnosed cases (established asthma) and the ones with positive answers to all questions in module 1.1 of the ISAAC protocol,²⁷ whereas all remaining subjects were classified into the probable asthma category. The parents returned 902 biosocial questionnaires (for 472 girls and 430 boys) and 908 allergological ones (for 464 girls and 444 boys).

Statistical analyses

The purpose of statistical analysis was to ascertain whether the BMI value, fat percentage, and the risk of asthma, overweight, and obesity at the age of 8 years were predicted by breastfeeding duration while statistically adjusting for potential confounders. These covariates were sex, age, socioeconomic status (obtained with factorial analysis of declared income, number of books at home, mother's education, and father's education, with factor loadings varying from 0.54 to 0.84), birth weight, gestational age, having a sibling, exposure to cigarette smoke in infancy (a parent or babysitter smoked in the presence of the child as an infant or smokes currently), antibiotic treatment in the first 12 months of life, paternal and maternal BMI (calculated from declared father's and mother's height and weight), and life style, including milk consumption, fast-food consumption, physical activity, and duration of daily TV watching (Tables 1 and 2). In the analysis of asthma risk, the set of covariates included the child's BMI instead of parental BMI values. Furthermore, the analysis of the risk of asthma was conducted only on children assigned to the category asthma present (coded as 1) or asthma absent (coded as 0), whereas those with symptoms suggestive of asthma were not taken into consideration.

We applied a generalized linear model with the logit link function (effectively, the logistic regression) for dichotomous dependent variables (risk of obesity, overweight, and asthma) or the identity link function (equivalent to standard analysis of regression) for continuous variables (BMI and fat percentage). Because of the many independent variables, the backward elimination method was used so to identify the most parsimonious set of important predictors. We started with a model including all potential predictors and at each step dropped from it the nonsignificant predictor (p>0.05) with the impact on the coefficient for breastfeeding duration being the lowest and below 10%. The breastfeeding duration, as the predictor of primary interest, was forced in the model. Therefore, the final model contained breastfeeding duration and variables that were significant at the 0.05 level or changed the coefficient for breastfeeding duration by 10% or more.²⁸ Following recommendations, breastfeeding duration was treated as a continuous quantitative variable rather than being transformed in a dichotomous one.²⁹

We applied the Sobel test (in Aroian version)³⁰ to establish whether a variable M mediates an effect of variable X on variable Y. To this end we performed three regressions as suggested by Baron and Kenny³¹: (1) of Y on X, (2) of M on X, and (3) of Y on X and M. In addition, each of these regressions included variables having been previously identified as meaningful covariates in the above-mentioned backward stepwise regression analyses of the variable Y. The Sobel test relies on coefficients obtained in regression (2) and (3). The proportion of the effect of X on Y that is mediated by M was calculated as the relative change in the coefficient for X between regression (1) and (3).³² Although the mediation analysis was originally developed for continuous variables, it is also valid for dichotomous variables (such as the presence of asthma or obesity), provided that the regression coefficients were appropriately standardized³²; thus we applied such standardization to our data.

All calculations were performed using StatSoft (Tulsa, OK) Statistica version 8.0 software, and reported p values are two-tailed. Values of p<0.05 were considered statistically significant.

Ethics

The research project was approved by the Institutional Review Board at Poznań University of Medical Sciences, the Education Office in Poznań that supervises the schools in the region, and the headmasters of each school taking part in the project. An informed written consent was obtained from each participant's parent.

Results

Duration of breastfeeding averaged 8.20 months with a median of 6 months and standard deviation of 8.40 months (Table 1), and its distribution was as follows: less than 3 months, 256 children (28.99%); 3–6 months, 249 (28.20%); 7–12 months, 201 (22.76%); and over 12 months, 177 (20.05%). Overweight was found in 96 (14.4%) girls and 101 (16.6%) boys, and obesity was found in 36 (5.39%) girls and 33 (5.42%) boys. Asthma was diagnosed in 53 (11.6%) girls and 77 (17.5%) boys; the diagnosis was uncertain in 17 (3.7%) girls and 18 (4.1%) boys. Basic descriptive statistics of other variables are provided in Tables 1 and 2.

Simple regression analyses indicated that the longer the duration of breastfeeding, the lower the BMI (n=883; B=−0.020; 95% confidence interval [CI], −0.040, −0.001; p=0.037) and the lower the risk of obesity (n=883; odds ratio=0.934; 95% CI, 0.884, 0.987; p=0.015). Similar, although nonsignificant, associations were observed for fat percentage (n=785; B=−0.050; 95% CI, −0.104, 0.005; p=0.076), risk of overweight (n=883; odds ratio=0.980; 95% CI, 0.956, 1.004; p=0.105), and risk of asthma (n=818; odds ratio=0.975; 95% CI, 0.951, 1.000; p=0.054). The prevalence of asthma in categories of breastfeeding duration was as follows: less than 3 months, 20.7%; 3–6 months, 18.1%; 7–12 months, 8.9%; and over 12 months, 15.2%. The prevalence of obesity by breastfeeding was as follows: less than 3 months, 5.9%; 3–6 months, 6.8%; 6–12 months, 4.0%; and over 12 months, 1.1%.

Multiple regression analyses showed that the longer the duration of breastfeeding, the lower the BMI and the fat percentage (Table 3) at the age of 8 years, and also the lower the risk of obesity but not of overweight (Table 4). It appears from the regression coefficients that for each succeeding month of breastfeeding, the odds of obesity decrease by 10%, which means that the odds of obesity decrease by 73% in children breastfed for 12 months and 48% for 6 months compared with nonbreastfed children. The risk of asthma was significantly predicted only by antibiotic use in infancy and current BMI, although breastfeeding duration had a marginally significant effect (Table 4). If antibiotic use was dropped from the model, the regression coefficient for breastfeeding duration increased by 30.8% and became significant (Table 4), suggesting a strong confounding effect of antibiotic administration to infants.

Table 3.

Body Mass Index and Fat Percentage Predictors in Backward Stepwise Multiple Regression

	B (SE)	p value	95% CI
BMI (n=535)
Intercept	7.467 (1.071)	0.000	5.367, 9.567
Breastfeeding duration	−0.025 (0.012)	0.038	−0.049, −0.001
Has sibling	−0.278 (0.105)	0.008	−0.485, −0.072
Antibiotics	0,181 (0.098)	0.064	−0.011, 0.372
Birth mass	0.721 (0.164)	0.000	0.399, 1.043
Paternal BMI	0.164 (0.029)	0.000	0.108, 0.221
Maternal BMI	0.132 (0.029)	0.000	0.075, 0.189
Socioeconomic status	−0.151 (0.099)	0.129	−0.345, 0.044
Fat percentage (n=474)
Intercept	−7.481 (3.278)	0.023	−13.906, −1.055
Sex	−0.988 (0.296)	0.001	−1.569, −0.408
Breastfeeding duration	−0.076 (0.035)	0.030	−0.145, −0,007
Has sibling	−0.988 (0.320)	0.002	−1.616, −0.360
Birth mass	1.509 (0.503)	0.003	0.524, 2.495
TV watching	0.907 (0.385)	0.018	0.153, 1.660
Paternal BMI	0.550 (0.088)	0.001	0.377, 0.723
Maternal BMI	0.372 (0.088)	0.001	0.200, 0.544

BMI, body mass index; CI, confidence interval; SE, standard error.

Table 4.

Overweight/Obesity and Asthma Predictors in Backward Stepwise Multiple Logistic Regression

	B (SE)	p value	Odds ratio (95% CI)
Overweight (n=535)
Intercept	10.689 (1.561)	0.000
Breastfeeding duration	−0.017 (0.018)	0.343	0.983 (0.948, 1.019)
Has sibling	−0.181 (0.141)	0.199	0.834 (0.633, 1.100)
Birth mass	0.586 (0.246)	0.017	1.798 (1.110, 2.910)
Paternal BMI	0.170 (0.037)	0.000	1.186 (1.103, 1.276)
Maternal BMI	0.109 (0.035)	0.002	1.115 (1.041, 1.195)
Socioeconomic status	0.307 (0.141)	0.029	0.736 (0.558, 0.970)
Obesity (n=535)
Intercept	10.462 (2.030)	0.000
Breastfeeding duration	−0.109 (0.050)	0.032	0.887 (0.813, 0.990)
Paternal BMI	0.192 (0.055)	0.001	1.217 (1.093, 1.356)
Maternal BMI	0.113 (0.053)	0.032	1.120 (1.010, 1.242)
Socioeconomic status	−0.263 (0.235)	0.264	0.769 (0.485, 1.219)
Asthma (n=563)
Intercept	3.751 (0.886)	0.000
Breastfeeding duration	−0.031 (0.018)	0.087	0.970 (0.936, 1.005)
Has sibling	0.268 (0.146)	0.067	1.308 (0.982, 1.742)
Antibiotics	0.534 (0.144)	0.000	1.706 (1.287, 2.261)
Passive smoking	0.186 (0.127)	0.142	1.204 (0.940, 1.543)
BMI	0.111 (0.049)	0.023	1.118 (1.016, 1.230)
Asthma, antibiotics excluded (n=563)
Intercept	3.723 (0.867)	0.000
Breastfeeding duration	−0.040 (0.018)	0.027	0.960 (0.926, 0.995)
Has sibling	0.290 (0.144)	0.044	1.336 (1.008, 1.771)
Passive smoking	0.187 (0.125)	0.134	1.206 (0.944, 1.541)
BMI	0.123 (0.048)	0.010	1.131 (1.029, 1.242)

BMI, body mass index; CI, confidence interval; SE, standard error.

To test the supposition of antibiotic use mediating the relationship between breastfeeding duration and asthma risk, we conducted a series of regression analyses as proposed by Baron and Kenny³¹ on a sample of 747 children with available data on breastfeeding duration and antibiotic use, with three covariates (siblings, passive smoking, and BMI [Table 4]) and with definite asthma status (uncertain diagnoses omitted). These were logistic regressions, each conducted on the same sample and controlled for these three covariates. Because the variable of antibiotic administration embraced the first year of a child's life and hence breastfeeding beyond the age of 12 months could not influence the prior antibiotic use, we used winsorized data for breastfeeding duration (i.e., values exceeding 12 months were replaced by 12 months [this pertained to 178 or 20% of cases]). Breastfeeding duration predicted the asthma risk both when it was the only predictor in the model and when the variable of antibiotic treatment was added into the model; in the latter case, the antibiotic use was also a significant predictor. Furthermore, breastfeeding duration predicted antibiotic use (Fig. 1 gives numerical data).

FIG. 1.

Estimation of (A) total and (B) direct and indirect effects of breastfeeding duration on asthma risk (n=747). Estimator values (a, b, c, c') and their standard errors (SEs) were obtained with logistic regression analyses. Standardized values of the estimators and their SEs were obtained with formulas provided by Herr.³² The Sobel test conducted on these values revealed a significant mediation of the breastfeeding duration–asthma risk relationship by antibiotic use (z=−2.61, p=0.009). The mediation by antibiotic use accounts for 24% of the total effect of breastfeeding duration on asthma risk ([0.142 – 0.108]/0.142=0.24).

Most important is that the Sobel test strongly confirmed the mediation of the relationship between breastfeeding duration and asthma risk by antibiotic use (Sobel's z=−2.61, p=0.009). The raw value of the coefficient for the total effect (−0.058) informs that for each succeeding month of breastfeeding, the odds of asthma decrease by 6%, which means that the odds of asthma decrease by 50% in children breastfed for 12 months and 29% for 6 months compared with nonbreastfed children. Comparison of standardized coefficients for breastfeeding duration for total and direct effects (Fig. 1) indicates that mediation by antibiotic use accounts for 24% of the total effect of breastfeeding duration on asthma risk ([0.142 – 0.108]/0.142=0.24).

To facilitate interpretation of our data and results, we also present and analyze frequencies of children with regard to breastfeeding duration (dichotomized), antibiotic administration, and asthma occurrence in a group of 789 participants with these data available (Table 5). Effects of breastfeeding duration and antibiotics were tested with the z test for two proportions. Frequency of antibiotic administration was higher in children breastfed for a short (233/389=0.599) rather than for a long time (188/400=0.470) (z=3.63, p=0.00028). Also, asthma frequency was higher in children breastfed for a short (75/389=0.193) rather than a long time (47/400=0.118) (z=2.91, p=0.0036). In addition, the frequency of children who both received antibiotics in infancy and developed asthma was higher in the case of a short (58/389=0.149) rather than a long duration of breastfeeding (31/400=0.078) (z=3.15, p=0.0016). Asthma frequency was higher among children who had been treated with antibiotics in infancy than among those who had not, and this was true both among children breastfed for a short time (58/233=0.249 versus 17/156=0.109) (z=3.43, p=0.0006) and for a long time (31/188=0.165 versus 16/212=0.075) (z=2.79, p=0.0053). Finally, in accordance with our hypothesis of mediation by antibiotics, frequency of antibiotic use was decisively higher among asthmatic children who were breastfed for a short time (58/75=0.773) than among nonasthmatic children who were breastfed for a long time (157/353=0.445) (z=5.15, p<0.000001).

Table 5.

Frequencies of Children with Regard to Breastfeeding Duration (Dichotomized), Antibiotic Administration, and Asthma Occurrence in a Group of 789 Participants with Available Data

	Asthma
Breastfeeding duration, antibiotics in infancy	No	Yes	Sum
Short (<6 months)
No	139	17	156
Yes	175	58	233
Sum	314	75	389
Long (≥6 months)
No	196	16	212
Yes	157	31	188
Sum	353	47	400
Sum	667	122	789

Antibiotic treatment did not predict any measure of body mass or fatness in multiple regression analyses (Table 3). Bivariate relationships between antibiotic use and each of these measures were nonsignificant too (all p values >0.3, according to Pearson's chi-squared test for independence or Mann–Whitney U test applied appropriately). However, antibiotic use was a marginally significant predictor in the regression of BMI (Table 3), and dropping it from the model changed the coefficient for breastfeeding duration by 11.1%, suggesting some confounding/mediating potential for this variable. We therefore conducted another test of mediation, specifically, the mediation by antibiotic use in relationship between breastfeeding duration and BMI. The Sobel test indicated no significant mediation (Sobel's z=−0.84, p=0.40). Summing up, our data suggest that antibiotic use is not a mediator in the influence of breastfeeding duration on BMI, fat percentage, or risk of obesity.

Discussion

The previous studies on impact of breastfeeding on health outcomes frequently used combined data on children of different age, varying in ethnicity, and deriving from the whole country area, both urban and rural sites, whereas this study has been carried out on a sample homogeneous in terms of age, ethnicity, and geographic location. Because asthma and obesity are complex disorders affected by numerous interacting determinants, in the present analysis we controlled for many variables. Thus we excluded their possible confounding effects on the impacts of breastfeeding duration that we found.

We found a significant effect of breastfeeding duration on child's BMI, body fat percentage, risk of asthma, and obesity (but not overweight) at the age of 8 years. Those findings agree with most other studies.^5,6,9,33

The impact of breastfeeding duration on obesity and asthma is being explained in several ways: impact of the mother's milk composition on the child's growth and development, feeding mode–dependent effect of child's and caregiver's behaviors on the amount of food ingested by the child, or impact of demographic, psychological, and social factors on both duration of breastfeeding and the child's development. Although we are not able to differentiate between the first and second mechanism, the probability of the last hypothesis is strongly limited as we statistically controlled for many possible confounders. Additionally, we proposed a new factor responsible for the relationship between breastfeeding duration and the risk of later developing asthma and excessive fatness, namely, antibiotic administration to the child in infancy. Although no evidence was obtained for fatness measures, we found that antibiotic use statistically mediated the influence of breastfeeding duration on asthma risk at the age of 8 years.

This mediation can potentially operate through at least two mechanisms. First, a short period of breastfeeding results in more frequent and/or severe diseases in infancy,^2–4,10 and this increases the probability of treatment with an antibiotic. Antibiotic administration during infancy, in turn, heightens the risk of developing asthma in later life.^19,20 This mechanism would therefore be a causal chain: shorter duration of breastfeeding→worse health→more frequent antibiotic administration→higher risk of asthma.

The second possibility arises from the fact that some parents have better nurturing knowledge than others. This knowledge results in a longer period of breastfeeding³⁴ and more prudent (i.e., less frequent) administration of antibiotics to the child.³⁵ Although we had no data on the health of the examined children during infancy or on parental knowledge on child rearing, we checked whether breastfeeding duration and antibiotic use in infancy correlated with variables that appear to be related to parental knowledge,³⁶ namely, mother's education, father's education, number of books at home, and number of sport disciplines in which the child participates (excluding physical education classes). A series of simple regression analyses were conducted with breastfeeding duration (standard regression) or antibiotic use (logistic regression) as the dependent variable, with the putative indicators of parental knowledge as independent variables (one per analysis). Although each of these variables significantly predicted breastfeeding duration (all p values <0.001), none predicted the antibiotic use (all p values >0.3). In a similar way we checked a possible effect of parental wealth: their declared income proved to be associated with longer breastfeeding (p=0.012) but not with antibiotic administration (p=0.95). This challenges mechanisms that refer to parental knowledge or wealth, supporting thereby the causal chain mechanism (breastfeeding duration→antibiotic use→asthma risk).

There is also another mechanism, relying on reverse causality, that could explain our findings: early symptoms of asthma (including cough and wheeze) result in asthma occurrence at the age of 8 years and antibiotic use in infancy and lead to shortened breastfeeding. This could be possible because infants with preexisting symptoms of asthma (mistakenly attributed to respiratory infections) receive more antibiotics than nonallergic children.^5,37 However, it has been reported that mothers of frequently ill children tend to prolong, rather than shorten, breastfeeding.³⁸ Additionally, sick infants appear to prefer breastmilk to other foods.²⁶

Thus it seems that early signs of asthma would lead to prolonged breastfeeding duration, which would cause positive correlations between breastfeeding duration and asthma in 8-year-olds and antibiotic administration to the infant, whereas in our sample these relationships are negative. This questions the mechanism based on reverse causation.

Conclusions

The present study showed that the total duration of breastfeeding inversely affected BMI, body fat percentage, and risk of asthma and obesity (but not overweight) of the children at the age of 8 years. Our results indicate that the child benefits from each succeeding month of breastfeeding, that is, the longer the breastfeeding, the better the health outcomes. In infants breastfed for 6 months the odds of asthma decreased by 21%, and it decreased by 38% in children breastfed for 12 months compared with nonbreastfed children. Thus our findings strongly support World Health Organization guidelines for infant feeding emphasizing that breastmilk should always be the preferred choice of feeding for healthy infants^3,26 and recommending exclusive breastfeeding for 6 months with the introduction of complementary foods and continued breastfeeding thereafter.^8,26 Because pediatric research institutions recommend that a mother's decision to breastfeed her baby should be perceived as a critical and fundamental health decision for an infant's short and long-term well-being,³⁹ we think that our findings can contribute to support the promotion of breastmilk feeding.

We also found that in the relationship between breastfeeding duration and asthma risk, a mediating role was played by antibiotic use during the first year of the child's life. The analyses suggest that the mechanism might be the following: shorter breastfeeding impairs health, which increases the probability of antibiotic administration in infancy, which in turn increases the risk of asthma in later life. The mediation through antibiotics accounted for 24% of the breastfeeding effect on asthma risk, leaving the remaining 76% for possible action of other mechanisms.

A limitation of this study is that we had no data on the health of the examined children during infancy. Future research with such data included could perform a more stringent test of the proposed mechanism.

Footnotes

Acknowledgments

The study was supported by grant 2P04C01429 from the Polish Ministry of Science and Higher Education. We thank Professor Grzegorz Lis for making available the Polish version of the ISAAC questionnaire. We extend special thanks to Professor Jan Strzałko for critical review of the manuscript.

Disclosure Statement

No competing financial interests exist.

References

Stepans

MBF

, Wilhelm

, Hertzog

, et al. Early consumption of human milk oligosaccharides is inversely related to subsequent risk of respiratory and enteric disease in infants. Breastfeed Med, 2006; 1:207–215.

Abrahams

, Labbok

. Breastfeeding and otitis media: A review of recent evidence. Curr Allergy Asthma Rep, 2011; 11:508–512.

, Chung

, Raman

, et al. Breastfeeding and maternal and infant health outcomes in developed countries. Evid Rep Technol Assess (Full Rep), 2007;(153):1–186.

Jackson

, Mathews

, Pulanic

, et al. Risk factors for severe acute lower respiratory infections in children: A systematic review and meta-analysis. Croat Med J, 2013; 54:110–121.

Matheson

, Allen

, Tang

. Understanding the evidence for and against the role of breastfeeding in allergy prevention. Clin Exp Allergy, 2012; 42:827–851.

Harder

, Bergmann

, Kallischnigg

, et al. Duration of breastfeeding and risk of overweight: A meta-analysis. Am J Epidemiol, 2005; 162:397–403.

Bartok

, Ventura

. Mechanisms underlying the association between breastfeeding and obesity. Int J Pediatr Obes, 2009; 4:196–204.

Kneepkens

CMF

, Brand

. Clinical practice: Breastfeeding and the prevention of allergy. Eur J Pediatr, 2010; 169:911–917.

Silvers

, Frampton

, Wickens

, et al. Breastfeeding protects against current asthma up to 6 years of age. J Pediatr, 2012; 160:991–996.

10.

Al-Makoshi

, Al-Frayh

, Turner

, et al. Breastfeeding practice and its association with respiratory symptoms and atopic disease in 1–3-year-old children in the city of Riyadh, central Saudi Arabia. Breastfeed Med, 2013; 8:127–133.

11.

Hansen

, Evjenth

, Holt

. Increasing prevalence of asthma, allergic rhinoconjunctivitis and eczema among schoolchildren: Three surveys during the period 1985–2008. Acta Paediatr, 2013; 102:47–52.

12.

Kopelman

. Obesity as a medical problem. Nature, 2000; 404:635–643.

13.

Fletcher

, Green

, Neidell

. Long term effects of childhood asthma on adult health. J Health Econ, 2010; 29:377–387.

14.

Han

, Lawlor

, Kimm

. Childhood obesity. Lancet, 2010; 375:1737–1748.

15.

Ober

, Yao

. The genetics of asthma and allergic disease: A 21st century perspective. Immunol Rev, 2011; 242:10–30.

16.

Nwaru

, Takkinen

, Niemelä

, et al. Timing of infant feeding in relation to childhood asthma and allergic diseases. J Allergy Clin Immunol, 2013; 131:78–86.

17.

Carver

. Advances in nutritional modifications of infant formulas. Am J Clin Nutr, 2003; 77(Suppl):1550S–1554S.

18.

Hanson

. Breast-feeding and immune function. Proc Nutr Soc, 2007; 66:384–396.

19.

Foliaki

, Pearce

, Björkstén

, et al. Antibiotic use in infancy and symptoms of asthma, rhinoconjunctivitis, and eczema in children 6 and 7 years old: International Study of Asthma and Allergies in Childhood Phase III. J Allergy Clin Immunol, 2009; 124:982–989.

20.

Murk

, Risnes

, Bracken

. Prenatal or early-life exposure to antibiotics and risk of childhood asthma: A systematic review. Pediatrics, 2011; 127:1125–1138.

21.

Ajslev

, Andersen

, Gamborg

, et al. Childhood overweight after establishment of the gut microbiota: The role of delivery mode, pre-pregnancy weight and early administration of antibiotics. Int J Obes, 2011; 35:522–529.

22.

Slaughter

, Lohman

, Boileau

, et al. Skinfold equations for estimation of body fatness in children and youth. Hum Biol, 1988; 60:709–723.

23.

Ogden

, Flegal

. Changes in terminology for childhood overweight and obesity. Natl Health Stat Rep, 2010;(25):1–5.

24.

Nawarycz

, Ostrowska-Nawarycz

. Wskaźnik masy ciała u dzieci i młodzieży łódzkiej w wieku szkolnym. Pol Merk Lek, 2007; 23:264–270.

25.

, Scanlon

, Serdula

. The validity and reliability of maternal recall of breastfeeding practice. Nutr Rev, 2005; 63:103–110.

26.

World Health Organization. Infant and Young Child Feeding. World Health Organization, Geneva, 2009.

27.

Lukrafka

, Fuchs

, Moreira

, et al. Performance of the ISAAC questionnaire to establish the prevalence of asthma in adolescents: A population-based study. J Asthma, 2010; 47:166–169.

28.

Greenland

. Modeling and variable selection in epidemiologic analysis. Am J Public Health, 1989; 79:340–349.

29.

MacCallum

, Zhang

, Preacher

, et al. On the practice of dichotomization of quantitative variables. Psychol Methods, 2002; 7:19–40.

30.

Preacher

, Leonardelli

. Calculation for the Sobel Test: An Interactive Calculation Tool for Mediation Tests. 2010. Available at www.quantpsy.org/sobel/sobel.htm (accessed July 1, 2013).

31.

Baron

, Kenny

. The moderator-mediator variable distinction in social psychological research—Conceptual, strategic, and statistical considerations. J Pers Soc Psychol, 1986; 51:1173–1182.

32.

Herr

. Mediation with Dichotomous Outcomes. 2006. Available at www.nrhpsych.com/mediation/logmed.html (accessed July 1, 2013).

33.

Parsons

, Power

, Logan

, et al. Childhood predictors of adult obesity: A systematic review. Int J Obes, 1999; 23(Suppl 8):S1–S107.

34.

Arora

, McJunkin

, Wehrer

, et al. Major factors influencing breastfeeding rates: Mother's perception of father's attitude and milk supply. Pediatrics, 2000; 106:E67.

35.

Mangrio

, Wremp

, Moghaddassi

, et al. Antibiotic use among 8-month-old children in Malmö, Sweden—In relation to child characteristics and parental sociodemographic, psychosocial and lifestyle factors. BMC Pediatr, 2009; 9:31.

36.

Bornstein

, Cote

, Haynes

, et al. Parenting knowledge: Experiential and sociodemographic factors in European American mothers of young children. Dev Psychol, 2010; 46:1667–1693.

37.

Kummeling

, Thijs

. Reverse causation and confounding-by-indication: Do they or do they not explain the association between childhood antibiotic treatment and subsequent development of respiratory illness?. Clin Exp Allergy, 2008; 38:1249–1251.

38.

Simondon

, Costes

, Delaunay

, et al. Children's height, health and appetite influence mothers' weaning decision in rural Senegal. Int J Epidemiol, 2001; 30:476–481.

39.

Eidelman

. The AAP's 2012 Breastfeeding Policy Statement: Is there anything new?. Breastfeed Med, 2012; 7:203–204.