Breastfeeding and the prevalence of allergic diseases in schoolchildren: Does reverse causation matter?

Authors


Takashi Kusunoki, Department of Pediatrics, Shiga Medical Center for Children, 5-7-30 Moriyama, Moriyama, Shiga 524 0022, Japan
Tel.: +81 77 582 6200
Fax: +81 77 582 6304
E-mail: kusutaka@gamma.ocn.ne.jp

Abstract

Kusunoki T, Morimoto T, Nishikomori R, Yasumi T, Heike T, Mukaida K, Fujii T, Nakahata T. Breastfeeding and the prevalence of allergic diseases in schoolchildren: Does reverse causation matter?
Pediatr Allergy Immunol 2010: 21: 60–66.
© 2010 John Wiley & Sons A/S

Infants at higher risk of allergic diseases might be breastfed for longer periods compared with infants at lower risk in the hope that breastfeeding might reduce the risk of atopic disorders. Therefore, this intention could manifest as an apparent allergy-promoting effect of breastfeeding or reverse causation. To analyze the effect of breast feeding on the prevalence of allergic diseases at school age, a large questionnaire survey was administered to the parents of schoolchildren aged 7–15 yrs. 13,215 parents responded (response rate, 90.1%). Prevalence rates of allergic diseases were compared according to the type of feeding in infancy (either complete breastfeeding, mixed feeding or complete artificial feeding). In both univariate and multivariate analysis, compared with those with complete artificial feeding, those with mixed and complete breastfeeding showed a significantly lower prevalence of bronchial asthma (BA) (p = 0.01 and 0.003, respectively). On the other hand, in univariate analysis, the prevalence of atopic dermatitis (AD) and food allergy (FA) were significantly higher in those with complete breastfeeding (p = 0.04 and 0.01, respectively). There was a significantly higher proportion of complete breastfeeding among those with greater risk of allergic diseases (presence of family history, either eczema or wheeze within 6 months after birth, or FA in infancy). Therefore, our multivariate analysis included these risks as confounding factors, and we found that the promoting effects of breastfeeding on AD and FA disappeared. In conclusion, our data clearly showed the inhibitory effect of breastfeeding on the prevalence of BA at school age. The apparent promoting effect of breastfeeding on the prevalence of AD and FA is most likely because of reverse causation.

There have been conflicting data with respect to the effect of breastfeeding on the development of allergic diseases in childhood (1, 2). Whereas breastfeeding used to be regarded as protective against bronchial asthma (BA) and other allergic diseases, a number of recent prospective studies have suggested that breastfeeding might in fact be a risk factor for their development (3, 4). This unexpected finding prompted us to reconsider the preventive effect of breastfeeding on allergy. Only a few studies have been published regarding this issue from Japan, in which Miyake et al.(5, 6) showed in cross-sectional studies, that breastfeeding may be associated with an increased prevalence of atopic dermatitis (AD) in Japanese schoolchildren.

A criticism of the ‘allergy promoting’ effect of breastfeeding has been that it might be attributable to reverse causation. Infants at the highest risk of allergic diseases (because of family history of allergic diseases or presence of early signs of allergy, such as infantile eczema and wheeze) might be breastfed for longer periods in the hope that breastfeeding might reduce the risk of long-term allergic diseases (3). Actually, it was shown in prospective birth cohort studies, that early signs of allergic diseases or sensitization were independently associated with a reduction in risk of ceasing exclusive breastfeeding (3). Thus, to analyze the true relationship between breastfeeding and allergic diseases, one needs to take into account reverse-causation-associated factors, such as family history of allergy and early signs of allergic diseases (e.g. eczema, asthmatic symptoms or food reactions), as potential confounders.

In 2006, we did a large questionnaire-based survey on the prevalence of allergic diseases in schoolchildren in Kyoto, Japan (7), in which we asked about feeding patterns in infancy, as well as family history of allergies and time of onset of allergic symptoms in infants. Thus, we utilized the data to evaluate the effect of breastfeeding on the prevalence of allergic diseases at school age, paying special attention to how reverse-causation-associated confounding factors affected the results.

Subjects and methods

Epidemiological studies on the prevalence of allergic diseases in schoolchildren

A questionnaire survey dealing with the prevalence of allergic diseases was administered to the parents of schoolchildren aged 7–15 yrs. The questionnaire was based on and comparable with the one used by the International Study of Asthma and Allergies in Childhood (ISAAC) (8) and was prepared and validated by the Study Group of Epidemiology of Allergic Diseases founded by the Japanese Ministry of Health and Welfare in 1993 (9). This study was designated as the Allergic Schoolchildren in Kyoto study, and was approved by the Ethics Committee of Kyoto University Graduate School of Medicine.

Details of the questionnaire, as well as the definitions of BA, AD, allergic rhinitis (AR) and allergic conjunctivitis (AC) have been published previously (7). Briefly, BA was defined as repeated episodes of wheezing with dyspnea during the preceding 2 yrs. AD was defined as chronic eczema with more than three of the typical symptoms of AD. AR was defined as the presence of at least three of four chronic nasal symptoms: sneezing, rhinorrhea, scratching around the nose and nasal obstruction. AC was defined as the presence of bilateral itchy, watery eyes.

With respect to food allergy (FA), we asked the following questions as described previously (10): (i) Does your child ever have allergic symptoms, such as skin symptoms like hives, or respiratory symptoms like cough/wheeze, within 1–2 h after ingesting a particular food? (2) Does your child ever avoid particular foods because of these symptoms? (3) If so, what are the kinds of foods and the duration of avoidance? Those who answered ‘yes’ to both questions 1 and 2 were considered to have either a past history or present illness of immediate-type FA. Among them, those who avoided any foods at the time of the survey were regarded as having FA.

The questionnaire was distributed by teachers to the parents of schoolchildren aged 7–15 yrs in 30 randomly selected schools in Kyoto City, one of the representative cities in Japan, with residents who have a typical urban lifestyle. The questionnaires were answered by the parents and collected at school. The number of questionnaires distributed was 14,669. Of those, 13,215 were returned (response rate, 90.1%). Of the respondents, 13,100 (99.1%) answered questions about feeding pattern in infancy and were enrolled in the study.

Feeding pattern in infancy

In the questionnaire, we asked the parents to choose one of five infancy feeding pattern options: (1) complete breastfeeding, (2) mixed with mostly breastfeeding, (3) mixed with equal artificial feeding and breastfeeding, (4) mixed with mostly artificial feeding and (5) complete artificial feeding. The feeding patterns of children whose parents chose (1) were designated as complete breastfeeding. Those who chose (2), (3) or (4) were classified as mixed feeding. And those who selected (5) were designated as complete artificial feeding.

Confounding factors

Sex, age, birth order (1st, 2nd and 3rd or more), family history (defined as the presence of either BA, AD, AR, AC or FA in first-degree relatives), eczema or wheeze within 6 months after birth, and FA in infancy (defined as the presence of FA because of egg, milk, or wheat within the first year of life) were chosen as confounding factors in multivariate logistic regression analysis, unless otherwise indicated.

Statistical analysis

Univariate and multivariate logistic regression analyzes were performed using spss software (Version 17.0; SPSS, Chicago, IL, USA). The level for statistical significance was determined to be p < 0.05.

Results

Feeding pattern in infancy and the prevalence of allergic diseases in schoolchildren – complete breastfeeding vs. complete artificial feeding

The distribution of selected characteristics in the 13,100 children enrolled in the study is shown in Table 1. The prevalence of allergic diseases in school age children who received complete artificial feeding was compared with children who were completely breastfed. Compared with those with complete artificial feeding, those with complete breastfeeding showed a significantly lower prevalence of BA (5.0% vs. 7.3%) in both univariate and multivariate analysis [adjusted odds ratio (OR) = 0.58] (Table 2a). With respect to FA, those with complete breastfeeding had a higher prevalence than those with complete artificial feeding (4.1% vs. 2.2%). However, while the difference was statistically significant in univariate analysis (p = 0.02), it was no longer significant in multivariate analysis (p = 0.22). No significant difference was observed for the prevalence of AD, AR or AC (Table 2a).

Table 1.   Distribution of selected characteristics in 13,110 schoolchildren in Kyoto, Japan
Variablen (%)
Male sex6655 (50.9)
Birth order
 1st6515 (49.7)
 2nd4831 (36.8)
 ≥3rd1734 (13.2)
Family history of allergy9850 (75.1)
Eczema or wheeze within 6 months after birth1423 (10.9)
Food allergy in infancy556 (4.2)
Feeding pattern in infancy
 Complete breastfeeding3982 (30.4)
 Mixed feeding8365 (63.8)
 Complete artificial feeding763 (5.8)
Table 2.   Feeding pattern in infancy and the prevalence of allergic diseases in schoolchildren
 PrevalenceCrude OR (95% CI)p-Value (univariate)Adjusted OR (95% CI)*p-Value (multivariate)
  1. BA, bronchial asthma; AD, atopic dermatitis; FA food allergy; AR, allergic rhinitis; AC, allergic conjunctivitis; OR, odds ratio; CI, confidence interval.

  2. *Adjusted for sex, age, birth order , family history, eczema or wheeze within 6 months after birth, and FA in infancy.

(a) Complete artificial feeding vs. complete breastfeeding
 BA
  Complete artificial feeding56/763 (7.3%)0.67 0.58 
  Complete breastfeeding200/3982 (5.0%)(0.49–0.61)0.01(0.42–0.80)0.001
 AD
  Complete artificial feeding37/763 (4.8%)1.30 1.1 
  Complete breastfeeding248/3982 (6.2%)(0.91–1.86)0.14(0.77–1.60)0.60
 AR
  Complete artificial feeding207/763 (27.1%)1.03 0.92 
  Complete breastfeeding1106/3982 (27.7%)(0.87–1.23)0.71(0.77–1.11)0.37
 AC
  Complete artificial feeding183/763 (24.0%)1.09 0.97 
  Complete breastfeeding1020/3982 (25.6%)(0.91–1.31)0.34(0.81–1.17)0.75
 FA
  Complete artificial feeding17/763 (2.2%)1.87 1.39 
  Complete breastfeeding163/3982 (4.1%)(1.13–3.11)0.02(0.82–2.35)0.22
(b) Complete artificial feeding vs. mixed and complete breastfeeding
 BA
  Complete artificial feeding56/763 (7.3%)0.65 0.60 
  Mixed and complete breastfeeding604/12347 (4.9%)(0.49–0.86)0.003(0.45–0.81)0.001
 AD
  Complete artificial feeding37/763 (4.8%)1.17 1.05 
  Mixed and complete breastfeeding695/12347 (5.6%)(0.83–1.64)0.36(0.74–1.49)0.78
 AR
  Complete artificial feeding207/763 (27.1%)1.02 0.94 
  Mixed and complete breastfeeding3390/12347 (27.5%)(0.86–1.20)0.84(0.79–1.11)0.48
 AC
  Complete artificial feeding183/763 (24.0%)1.08 0.99 
  Mixed and complete breastfeeding3127/12347 (25.3%)(0.91–1.28)0.41(0.83–1.18)0.89
 FA
  Complete artificial feeding17/763 (2.2%)1.67 1.33 
  Mixed and complete breastfeeding452/12347 (3.7%)(1.02–2.72)0.04(0.80–2.20)0.27
(c) Mixed and complete artificial feeding vs. complete breastfeeding
 BA
  Mixed and complete artificial feeding460 (5.0%)1.00 0.92 
  Complete breastfeeding200 (5.0%)(0.84–1.18)0.97(0.77–1.10)0.36
 AD
  Mixed and complete artificial feeding484 (5.3%)1.19 1.09 
  Complete breastfeeding248 (6.2%)(1.01–1.39)0.03(0.92–1.28)0.33
 AR
  Mixed and complete artificial feeding2491 (27.3%)1.03 0.98 
  Complete breastfeeding1106 (27.7%)(0.94–1.11)0.56(0.90–1.07)0.67
 AC
  Mixed and complete artificial feeding2290 (25.1%)1.03 0.97 
  Complete breastfeeding1020 (25.6%)(0.94–1.12)0.52(0.89–1.06)0.53
 FA
  Mixed and complete artificial feeding306 (3.4%)1.23 1.06 
  Complete breastfeeding163 (4.1%)(1.01–1.49)0.04(0.86–1.31)0.58

Feeding pattern in infancy and the prevalence of allergic diseases in schoolchildren – complete artificial feeding vs. mixed and complete breastfeeding

Next, those with complete artificial feeding were compared with those with mixed and complete breastfeeding for the prevalence of allergic diseases at school age. Again, compared with those with complete artificial feeding, those with mixed and complete breastfeeding had a significantly lower prevalence of BA (4.9% vs. 7.3%) in both univariate (p = 0.003) and multivariate analysis (p = 0.001, adjusted OR = 0.60) (Table 2b). However, while they showed a higher prevalence of FA (3.7% vs. 2.2%) that was statistically significant in univariate (p = 0.04) analysis, the difference did not remain statistically significant in multivariate analysis (p = 0.27).

Inhibitory effect of breastfeeding on the prevalence of bronchial asthma is age dependent

The difference in BA prevalence between those with complete artificial feeding and those with mixed and complete breastfeeding was further analyzed by stratifying them into different age groups (7–9, 10–12, and 13–15 yrs) (Table 3). The most significant difference was observed in those who were 10–12 yrs old. The adjusted OR for BA prevalence in the mixed and complete breastfeeding groups compared with that in the complete artificial feeding group was 0.48 (p = 0.002). The difference was also observed in those who were 7–9 yrs old (adjusted OR = 0.65, p = 0.04). However, the difference was no longer observed in 13- to 15-yr olds.

Table 3.   Feeding pattern in infancy and the prevalence of BA in schoolchildren: comparison between different age groups – complete artificial feeding vs. mixed and complete breastfeeding
Age group (yrs)BA prevalenceCrude OR (95% CI)p-Value (univariate)Adjusted OR (95% CI)*p-Value (multivariate)
  1. BA, bronchial asthma; OR, odds ratio; CI, confidence interval.

  2. *Adjusted for Sex, age, birth order, family history, eczema or wheeze within 6 months after birth, and FA in infancy

7–9
 Complete artificial feeding78/373 (7.51%)0.72 0.65 
 Mixed and complete breastfeeding318/5749 (5.53%)(0.48–1.08)0.11(0.43–0.99)0.04
10–12
 Complete artificial feeding24/274 (8.76%)0.50 0.48 
 Mixed and complete breastfeeding200/4330 (4.62%)(0.32–0.79)0.002(0.30–0.75)0.002
13–15
 Complete artificial feeding4/114 (3.51%)1.06 0.98 
 Mixed and complete breastfeeding83/2244 (3.70%)(0.38–2.93)0.92(0.35–2.74)0.96

Feeding pattern in infancy and the prevalence of allergic diseases in schoolchildren – mixed and complete artificial feeding vs. complete breastfeeding

The prevalence of allergic diseases in school age of children who received mixed and complete artificial feeding was compared with those who were completely breastfed (Table 2c). In univariate analysis, those who were completely breastfed showed a significantly higher prevalence of AD and FA compared with those who received mixed and complete artificial feeding (6.2% vs. 5.3%, p = 0.03; 4.1% vs. 3.4%, p = 0.04, respectively). However, these differences were no longer statistically significant in multivariate analysis with the confounding factors described previously.

Reverse-causation-associated factors and the frequency of complete breastfeeding

Among the confounding factors, family history of allergy, eczema or wheeze within 6 months after birth, and FA in infancy are risk factors for allergic diseases in children, and thus are believed to be the basis for reverse causation. In other words, children with these factors might have been more likely to have been breastfed in infancy, with the expectation that breastfeeding might reduce the risk of allergy. To check for this possibility, the frequencies of subjects with those confounding factors among those with complete breastfeeding were examined. The frequency of subjects with any of these reverse-causation-related factors among those with complete breastfeeding was significantly higher than among those with mixed and complete artificial feeding (Fig. 1).

Figure 1.

 Frequencies of subjects with reverse-causation-related factors among those with different feeding patterns in infancy. Frequencies of subjects with either (I) family history of allergy, (II) eczema or wheeze within 6 months after birth or (III) food allergy in infancy, among those with (A) complete breastfeeding and (B) mixed or complete artificial feeding.

Effect of reverse-causation-associated factors on analysis of the prevalence of atopic dermatitis and food allergy in schoolchildren

To examine whether the reverse-causation-associated factors affected the higher prevalence of AD and FA in those with complete breastfeeding, multivariate analysis with and without those factors as confounders was performed. Statistical significance for the higher prevalence of both AD and FA in those with complete breastfeeding, which was observed in univariate analysis, was still observed when sex, age and birth order were included as confounding factors. However, for both AD and FA, when the analysis was performed with any one of the three reverse-causation-associated factors as confounding factors, the significance disappeared (Table 4).

Table 4.   Odds ratio (OR) and 95% confidence intervals (CI) (complete breastfeeding/mixed and complete artificial feeding) for the prevalence of atopic dermatitis (AD) and food allergy (FA) according to confounding factors
 OR (95% CI)p-ValueConfounding factors
AD1.19 (1.01–1.39)0.03None
1.18 (1.01–1.38)0.04Sex, age, birth order
1.15 (0.98–1.34)0.09Family history
1.14 (0.97–1 .34)0.12Eczema or wheeze within 6 months after birth
1.14 (0.97–1.33)0.12FA in infancy
1.09 (0.92–1.28)0.33All of these
FA1.23 (1.01–1.49)0.04None
1.25 (1.03–1.52)0.02Sex, age, birth order
1.19 (0.98–1.45)0.08Family history
1.18 (0.97–1.44)0.10Eczema or wheeze within 6 months after birth
1.07 (0.87–1.32)0.52FA in infancy
1.06 (0.86–1.31)0.58All of these

Complete breastfeeding and the prevalence of atopic dermatitis and food allergy – comparison among those with reverse-causation-associated factors

The prevalence of AD and FA in those with complete breastfeeding and with mixed and complete artificial feeding was compared among those with reverse-causation-associated factors. No statistical difference was observed among those with any of the reverse-causation-related factors (Table 5).

Table 5.   Feeding pattern in infancy and the prevalence of food allergy (FA) and atopic dermatitis (AD) in schoolchildren: comparison among those with reverse causation-related factors – mixed and complete artificial feeding vs. complete breastfeeding
 AD prevalencep-ValueFA prevalencep-Value
With family history (n = 9850)
Mixed and complete artificial feeding427/6740 (6.3%) 271/6740 (4.0%) 
Complete breastfeeding227/3110 (7.3%)0.08149/3110 (4.8%)0.08
With eczema or wheeze within 6 month after birth (n = 1423)
Mixed and complete artificial feeding179/944 (19.0%) 111/943 (11.8%) 
Complete breastfeeding95/479 (19.8%)0.6961/479 (12.7%)0.6
With FA in infancy (n = 556)
Mixed and complete artificial feeding65/339 (19.2%) 113/339 (33.3%) 
Complete breastfeeding41/217 (18.9%)0.9467/217 (30.9%)0.89

Discussion

In this study, we found that, compared with those with complete artificial feeding, those with mixed or complete breastfeeding had a significantly lower prevalence of BA at school age, whereas those with complete breastfeeding showed a higher prevalence of AD and FA. The statistical significance of the differences in the prevalence of AD and FA, however, disappeared in multiple logistic regression analysis that included reverse-causation-associated factors as confounders. Overall, these data suggest that, in schoolchildren, breastfeeding has a preventive effect on BA, while its promoting effect on AD and FA is most likely because of reverse causation.

The strength of our study is because of the large sample size with a high response rate, strict diagnostic criteria for allergic diseases using a firmly evaluated questionnaire and the assessment of children at high risk of allergic outcomes (reverse causation). On the other hand, there are some limitations, one of which is reliance on maternal recall of feeding patterns. However, we do not think the recall bias is a serious problem because we asked parents to choose one of five options for the feeding pattern in infancy: complete breastfeeding; mixed with mostly breastfeeding; mixed with equal artificial and breastfeeding; mixed with mostly artificial feeding; and complete artificial feeding. We then combined the middle three options as ‘mixed feeding.’ As a result of these simplified questions, it should not have been difficult for parents to recall the feeding pattern of their children in infancy, although more detailed studies on breastfeeding duration could not be performed. Another limitation involves the questions about family history of allergic diseases. Although we asked about the presence of several allergic diseases in the child’s first-degree relatives, we did not ask who had the allergic diseases. Hence, we could not find the influence of family history in more detail, such as the effects of maternal asthma, which has been shown to influence the effect of breastfeeding on allergy in several previous reports (11–13).

The preventive effect of breastfeeding on BA at school age, shown in our study, is in accord with several previous observations. Some data suggest that the effect of breastfeeding on BA prevalence is age dependent, and while the preventive effect can be observed during early childhood, the effect disappeared or even reversed at older ages (11, 13, 14). Our data also confirmed these age dependencies. The inhibitory effect could be observed in the 7–12 yrs age group, while it disappeared in the 13–5 yrs age group, although reversal of the effect was not found. One of the reasons might be that as children become older, BA prevalence diminishes and thus the effect can no longer be observed clearly. Possible mechanisms by which breastfeeding prevent BA might include prevention of early respiratory viral infections (1), although it is also possible that prevention may lead to Th2 bias and increase BA prevalence at a later age. It would be interesting to follow these children into middle age to analyze if there is a reverse effect of breastfeeding, as observed by Matheson et al. (13). Another mechanism might be, as in the mouse model, tolerance induction of airborne allergen transferred with transforming growth factor-β from mother to child through breast milk (15).

In accord with the data by Miyake et al., who observed the AD-promoting effect of breastfeeding in Japanese schoolchildren (5, 6), we found a slight but statistically significant AD- and FA-promoting effect of breastfeeding in univariate analysis. However, this significance was lost after adjusting for only one of the three reverse-causation-associated factors (family history, eczema or wheeze within the first 6 months after birth, and FA in infancy), suggesting that the effect was most likely because of reverse causation. The discrepancy of our data with those by Miyake et al. could be that they did not take into account early atopic symptoms as confounding factors. In fact, we observed in our study population that those with complete breastfeeding were more likely to have such reverse-causation-related factors compared with those with mixed or complete artificial feeding. This phenomenon was observed in several studies (3, 16, 14), but not by others (13, 17), suggesting that influence of reverse causation might be different depending on the study population.

In conclusion, large-scaled cross-sectional data suggest the existence of a preventive effect of breastfeeding on BA in schoolchildren. Thus, we can recommend breastfeeding to mothers for the purpose of BA prevention at school age. Apparent AD- and FA-promoting effects of breastfeeding are most likely because of reverse causation, and there are neither promotive nor preventive effects of breastfeeding on either AD, AR, AC or FA. These data reconfirm the need to take reverse causation into account in studying the relationships between breastfeeding and allergy.

Ancillary