The MAS Group recruited 499 newborn infants with risk factors for atopy (elevated cord blood IgE (
0.9 kU/l) or at least two atopic family members) and 815 newborn infants without these risk factors.10
The cohort children were followed up at the age of 1, 3, 6, 12, and 18 months, and from then on at yearly intervals within 3 months of the child's birthday up to the age of 7 years. The study was approved by the local ethics committees.
At each follow up, parents gave structured interviews to a study doctor on their child's development. Of greatest interest was asthmatic and atopic symptoms and diseases. Among other questions, parents were asked whether their child had had “a wheezy or whistling noise while breathing” since the previous follow up. When the children were 7 years old parents were asked whether their child had ever had a diagnosis of asthma.
Early childhood infectious diseases
We also assessed other illnesses at each follow up. If the parents responded affirmatively to the question “Was your child ill since your last visit?” the interviewing doctor assessed the reported symptoms and diagnosis of the illness and encoded them according to the Weidtman code, a German language version of the ICD-9 (international classification of diseases, ninth revision) for paediatric use.11
In addition, we assessed any drugs the children were given. To keep reporting bias low, we asked the parents to keep a non-structured diary of their child's diseases, which served as memory aid for the interview. By the time the children were 3 years old, we had recorded 598 different Weidtman codes during the follow up visits, comprising 106 codes for infections.
Because of a potential bias attributable to reverse causation, we analysed separately all lower respiratory tract infections irrespective of the infectious agent (pneumonia, bronchitis, pertussis, tracheobronchitis, flu, croup, bronchiolitis). We combined all other codes to assess the effect of the overall burden of infectious diseases and, in a second step, separated them into several distinct categories: viral infections (measles, rubella, mumps, hepatitis A and B, mononucleosis, runny nose (rhinitis), herpes, varicella, exanthema subitum, stomatitis, choriomeningitis, coxsackievirus); bacterial infections (meningitis, tonsillitis, lymphadenitis, otitis media, scarlet fever, septicaemia, abscess, impetigo, pyoderma, tuberculosis, urinary tract infections); fungal infections (nappy rash, candida infection); gastrointestinal infections (gastroenteritis, diarrhoea); and fever of unknown origin.
In addition, we assessed the number of antibiotic courses the children had received. Only antibiotic courses not related to the treatment of lower respiratory tract infections were included in statistical analysis.
Antibodies to specific antigens
We asked the parents to consent to blood sampling of their child at the ages of 1, 2, 3, 5, 6, and 7 years. We determined IgE concentrations to nine allergens (cow's milk, egg white, soya bean, wheat, house dust mite Dermatophagoides pteronyssinus
, cat, dog, mixed grass pollen, and birch pollen) by CAP-RAST FEIA (Pharmacia and Upjohn, Freiburg, Germany). We defined sensitisation to a specific allergen as a concentration of
0.35 kU/l of the specific IgE.
When the children were 7 years old we performed bronchial challenge, starting with 0.5 mg histamine/ml and increasing up to 8.0 mg/ml according to a standard procedure.12
We defined bronchial hyperreactivity as a PC20
(provocative concentration causing a 20% fall in forced expiratory volume in one second) greater than the 90th centile of the distribution of PC20
in a healthy subsample.13
We used χ2 tests to compare prevalences between groups and Mantel-Haenszel tests for analysing trends across categories. We calculated multivariate logistic regression models to analyse the effect of early childhood infectious diseases on asthma at the age of 7 years. For the longitudinal analyses, we used generalised estimation equation models to adjust for repeated measures. In order to take the stratified sampling scheme into account and to assess participation bias and potential effect modification, we initially stratified all multivariate analyses for elevated cord blood IgE concentration, family history of atopy, parental smoking at birth, and social status (defined as low, middle, or high parental education). If the results were similar over strata, however, we calculated the multivariate models in the total sample with indicator variables included for high risk of atopy at birth (elevated cord blood IgE concentration or two or more atopic family members), parental smoking at birth, and parental education. We also adjusted the longitudinal models for age. We used SAS software (version 6.12) for all statistical analyses.