Can early infection explain the sibling effect in adult atopy?

In a UK birth cohort, atopy and hay fever were more common among children from small families 1. The same patterns have been replicated elsewhere, for example among German 2 and Italian 3 children. These epidemiological observations are commonly ascribed to lower rates of infection among first-born children or those with few siblings; it is argued that early encounters with infection encourage the preferential development of T‐helper cell type 1 lymphocytes at the expense of T‐helper cell type 2‐subtypes associated with allergic responses to foreign proteins. Direct confirmation of this explanation is lacking although protective roles for measles in West Africa 4, mycobacteria in Japan 5 and gastrointestinally-acquired infections, particularly hepatitis A, in southern Italy 6, have been reported for atopy.

The present study sought a critical test for the hygiene theory among a population of UK adults. The current authors hypothesised that rates of atopy in this population would be inversely related to family size (an indirect marker of early infection), and also that direct markers of infection would be more common in larger families. The null hypothesis that the pattern of atopic risk in relation to family size would not be explained by direct evidence of early infection was subsequently tested.

Materials and methods

Results

Atopy and infection

Atopy was unrelated to a positive test for either hepatitis A or H. pylori (table 3⇓). This lack of association was maintained across all age groups and was not modified by family size.

View this table:

• View inline
• View popup

Table 3

Relationships between atopy and markers of early infection

When all recorded infective episodes were considered, there was also no relationship with atopic status (table 3⇑). Lower respiratory infections, however, were positively associated with atopy (OR 1.08 (1.01–1.14) per recorded episode) whereas gastrointestinal (and other) infections demonstrated a statistically significant inverse relationship (OR 0.73 (0.58–0.92) per recorded gastrointestinal episode). These relationships were maintained after adjustment for age, sex, socioeconomic status, parental allergic disease and number of brothers. Importantly the inverse association between numbers of brothers and atopy was unaffected by adjustment for any of the measures of early infection (table 3⇑).

The adjusted odds ratios for recorded infective episodes at each age are plotted in the figure 1⇓. Those for respiratory infections increased each year to age 5 yrs; the opposite pattern was found for gastrointestinal infections. Atopy was less common among those with a recorded measles infection (30% versus 38%) but the difference was not significant (p=0.11). When this was examined by age of infection, atopy was significantly (p=0.004) less common in the 28 adults who had a record of measles by age 3 yrs (11% versus 38%).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.—

Atopy odds ratios (adjusted for age, sex, social class, parental allergy and number of siblings) associated with infectious episodes before the age of 5 yrs and for 1–5 yrs separately.

The prescription of antibiotics before 5 yrs was unrelated to subsequent atopy (OR 1.01 (0.97–1.05) per prescription). This lack of an association was found for all classes of and indications for antibiotics (data not shown). Risk estimates for antibiotic prescriptions (all types) tended to rise with age of prescription but the increases were small and none was statistically significant: OR 0.85, 1.01, 1.07, 1.09 and 1.09 per prescription at ages <1–5 yrs respectively (all p>0.05).

Discussion

This analysis confirms a clear inverse association between adult atopy and family size in early childhood, a pattern largely driven by the number of brothers. However, this could not be accounted for by several different indices of childhood infection. Conversely, apparently protective effects associated with early gastrointestinal infections with dog ownership and with moving home in the first 5 yrs of life were found. Although the current study's population is relatively small, it has the distinction of being highly representative of adults from South East England; both the participation and follow-up rates were high, and higher than those in other, similar studies. In addition the current authors used not only an objective outcome but were careful also to ensure blinded assessments of serological results and contemporary general practice records. The relationships found were unchanged by adjustment for several potential confounders suggesting that other important factors had not been neglected.

General practice records are inevitably an incomplete and potentially biased source of information about early infection. Many extraneous factors, including birth order, but also socioeconomic status, sex and era determine whether a parent will take her child to the doctor. In the present study's population, the existence of early, contemporary records was in part determined by age and was less common for males. The records examined were more likely to contain a mention of three or more infections before the age of 5 yrs among those adults who were brought up in small families. However a systematic difference in seeking medical advice between parents with small and larger numbers of children 7 may have hidden a truly higher rate of infection among those from larger families. This is suggested by the higher rates of Helicobacter and hepatitis A infection in those with several siblings. As it happens, among contemporary UK families there is remarkably little evidence of a true relationship between infection and family size or birth order, although in Sweden 8 hospitalisation for respiratory infection and in the Netherlands 9 meningococcal carriage are both higher in large families. Even if not associated with a reduced risk of early infection, growing up in a small family may however increase the median age of first, relevant infection(s) 10, a pattern suggested by the records in this study.

Positive serology for hepatitis A, a postulated marker for an unhygienic early childhood, was inversely associated with atopy among southern Italian military recruits 11. The current study could not replicate this finding in a population where atopy was more common (37% versus 28% in Italy) and evidence of hepatitis infection less frequent (9% versus 27%). Similarly, in a small case-referent analysis from Aberdeen 12, where positive hepatitis A serology was much commoner (61%), atopy was more often associated with a positive test although the difference was not statistically significant. The current study's rates of positivity to H. pylori were lower than those in Italy and Aberdeen (each 50%) although the differences were less marked; in Italy this was also related, inversely, to atopic status a finding which again could not be replicated in the present study. These variations probably reflect different epidemiologies of early hepatitis A and Helicobacter infection in the UK and Italy.

Early lower-respiratory infections were more commonly recorded for atopic adults, a pattern which increased with the age of infection and which the current authors suggest reflects the emergence of accompanying asthma over the first 5 yrs of life. Indeed the difficulty in separating cause from effect probably renders the epidemiological study of respiratory infections and allergic disease impossible, even when specific markers of infection are available. Conversely, recorded gastrointestinal infections were less common among atopic adults, a pattern which also strengthened with age of infection, which was seen despite the probable behavioural bias above and which remained after adjustment for possible confounding factors. A protective effect of such infections in early life is broadly consistent with the findings of a large survey in the USA 13; and there is evidence that the gut colonisation in early life may affect the subsequent differentiation of T-cell populations 14. Conversely, among a very large cohort of contemporary British children 15 no protective effect of early gastrointestinal infections on clinical diagnoses of asthma, hayfever or eczema was found; IgE antibody production was not examined, however, and the average period of follow-up was short (3 yrs). The information collected was insufficiently detailed to allow analysis of particular gastrointestinal infections although, at this age, most were probably viral. The current study's finding of an apparently protective effect of measles infection at age 3 yrs should be interpreted with care since it is based on small numbers; and the infrequency with which this disease was recorded suggests a degree of underascertainment. It does however accord with findings of a study of asthma in Aberdeen 16.

In several respects the current findings are at variance with those of an earlier UK study of allergy and infections recorded in general practice records up to the age of 12 yrs 17. There, no relationship between sibship and recorded atopic disease (asthma, eczema or hayfever as described in the medical records) was found. Gastrointestinal infections were not reported as a separate category; but treatments with antibiotics for respiratory (particularly lower respiratory), urinary and skin infections in the first year of life were associated with markedly increased risk estimates for subsequent atopic disease. A survey of children aged 5–10 yrs attending Rudolph Steiner schools in New Zealand reported remarkably strong associations between antibiotic use and a history of asthma; particularly when symptoms of the latter started after the age of one and the drugs had first been prescribed during infancy 18. Similar findings have been reported among German schoolchildren 19. In none of these studies can recall or protopathic biases be excluded. Conversely, the present authors, using an objective definition of atopy derived separately from general practice records, were unable to confirm any significant increase in risk with any class of (or indication for) documented antibiotic use in the first 5 yrs. This is an important negative finding; and probably corroborates the recent findings for British children concerning early antibiotic use and clinical allergies 15.

In some other aspects the current findings are similar to those of a comparable, albeit larger, survey of adult atopy carried across Europe 20 although this did not include any direct markers of early infection. In that study individuals from large families, particularly those with many brothers, and, independently those who kept dogs, were less likely to be atopic. Indeed, the evidence indicating a protective effect of dog ownership is increasingly consistent 21 although not entirely so 22. In the current study's population, the effect was strong and could not be accounted for by statistical adjustment for confounding factors, in particular family size. The specificity of the effect, both here and in the pan-European study, is suggested by the absence of any such relationship with pet cats. A bias, whereby allergic families tend not to keep pets, is an improbable explanation for these findings. The relationship was not affected by adjustment for parental allergy, and an earlier report of this cohort showed that the keeping of a pet cat is unrelated to general or specific allergies within the family 23. Bacterial, viral and parasitic infections may each be transmitted from dogs to humans 24 although with uncertain frequency, and the current authors suggest that the present findings support a protective role for early (unidentified) infection(s) in the development of atopy; alternatively levels of endotoxin, itself postulated to be protective, may be higher in homes where pets are kept 25. The finding of reduced rates of atopy among those who had moved home in early life was unexpected; indeed an earlier study of childhood asthma suggested the opposite effect 26. It is possible that moving home is associated with an increased rate of childhood infection (viral mixing) although the current authors are not aware of any direct evidence for this.

The current study was planned as a critical test of the plausible explanation for the inverse association between family size and atopic status: that higher rates of early infection among larger families are protective. The present study could not account for the relationship with sibship by several markers of early infection including evidence of infection with hepatitis A and Helicobacter pylori as observed in other populations. Within the Ashford birth cohort the current authors have recently reported 27 that mothers of small families are more likely to be atopic, a pattern that could not be accounted for by maternal age. This observation is as yet unexplained but together with the results reported here and elsewhere 28 suggests that the well-established inverse association between atopy and family size may be a manifestation of maternal or intra-uterine events rather than solely a reflection of early-life infection.