Study population
The study data was collected within a large study in East and West Germany investigating the health effects in school beginners (6-year-old) of the changing environmental and socio-economic conditions after re-unification between 1991 and 2000. Details of the study design were presented elsewhere [21–23]. In brief, all boys and girls living in geographically pre-defined areas of East and West Germany and entering the elementary school between 1991 and 2000 in East and in 1991, 1994, 1997 and 2000 in West Germany were eligible to participate. Rural areas without heavy industrial impact were located in Salzwedel, Gardelegen, Osterburg and Kloetze in East Germany and in Borken and Borken District in West Germany. Urban areas with industrial impact and strong traffic burden were selected in Halle, Leipzig and Magdeburg in East Germany and in Duisburg, Essen and Cologne in West Germany. The financial support of the study and the participation potential of the local health authorities were not constant over the whole study duration. Therefore, different numbers of pre-defined areas determined the entire investigation area in each year and, in consequence, led to different sample sizes. A letter was mailed to the parents inviting the child to participate in the study and to complete a standardised questionnaire. On the day of investigation, the questionnaire was checked by a physician and any missing answers subsequently completed by the parents. In the investigation years 1991, 1994, 1997 and 2000 and in pre-selected parts of the study areas, the children were asked to undergo a dermatological examination and to provide a blood sample. Additionally, with the exception of the areas in East Germany in the year 2000, every second child was invited to have its lung function tested. The overall response rate to the questionnaire was 83% and 71% in East and West Germany, respectively. In those children eligible, the response rates for participation in the dermatological examination, donation of a blood sample and testing the lung function were 84%, 73% and 73%, respectively, in East Germany and 88%, 69% and 69% in West Germany. Written informed consent of the parents of the participating children was obtained. The ethical committee of the Medical Association of Saxony-Anhalt approved the study.
Parental education
Information about school education of the parents was taken from the questionnaire., A suitable measure of the parental educational level was chosen by classifying the school education into the two categories of no more than 10 years school and more than 10 years school. The parent or partner with the highest school grade defined the parental educational level of the child. To avoid ambiguity in classification of the parental education, we only included children whose parents were of German nationality. This exclusion resulted in an average loss of 8% in the industrialized areas of East and of 26% of West Germany, and of 1% in the rural areas of East and 3% of West Germany. Obviously, the use of a dichotomous variable for the educational level does not sufficiently account for the very low educated parents whose children were the most disadvantaged in many aspects, e.g. the living conditions [19, 20]. However, a further subdivision of the lower educational level into the two categories of "less than 10 years school" and "10 years school" result in a classification which differs in interpretation between the two parts of Germany considering their different schooling systems before re-unification. In the former German Democratic Republic, ten years of schooling was compulsory. Consequently in East Germany, only six per cent of the participating children had parents with less than 10 years school.
Individual living conditions
In the questionnaire, the parents were asked to provide information on the child's living conditions. Altogether, six individual living conditions, which were considered relevant for the child's health, were investigated: The question about the number of sisters or brothers of the child resulted in the family condition "single child". The per capita living space of the child's home could be defined by using the two questions "How many people are living in the child's home?" and "How many square meters is the child's home?". The living space was considered small if it was below 20 m2 per person ("small living space"). If an oven heated with fossil fuel or cooking with gas existed then "unfavourable indoor air" in the child's home was assumed. Affirmation of the question "Would you characterise your home as damp?" indicated a "damp housing condition". "Maternal smoking during pregnancy" was recorded. It is often highly correlated with current maternal smoking and as such can be considered as an indication of the child's exposure to environmental tobacco smoke throughout life. Traffic exposure of the children was determined by the distance of the children's home to the nearest busy road, and "living near a busy road" was stated if the distance was less than 50 m.
Health indicators
Altogether, 14 variables indicating the child's health were investigated in this study. Of these, six health indicators representing a range of infectious and atopic diseases and symptoms were selected from the questionnaire data, namely "bronchitis ever diagnosed by a physician", "more than 4 colds during the last 12 months", "frequent cough without cold", "sneeze attacks during the last 12 months", "allergy ever diagnosed by a physician" and "eczema ever diagnosed by physician". Information regarding "birth weight" of the child was also provided by questionnaire. Seven further indicators were determined independent from the questionnaire: On the day of investigation, a physician examined the child's skin and decided on the diagnosis of an atopic eczema ("atopic eczema on the day of investigation"). Across all surveys, a standardized diagnostic procedure (SCORAD) [24] was used, and one of the authors (JR) has supervised the training of the physicians. Atopic sensitisation was tested by determining specific immuno-globulin-E (IgE) antibodies against common allergens in the blood serum of the child by an enzyme immuno assay (Radio Allergo Sorbent Test (RAST), Pharmacia & Upjohn, Uppsala, Sweden). Based on the results of the RAST, participants were classified as having "specific IgE grass pollen positive" or "specific IgE house dust mite positive" if the specific IgE concentrations against grass pollen and house dust mite, respectively, were greater than 0.35 kU/l. Height and weight were measured on the day of examination and "body height" used as a general measure of physical development. A body mass index (BMI = body weight/(body height)2 [kg/m2]) greater than 19 kg/m2 provided an indication of "overweight" Using a body plethysmograph for a sub-group of the participants, "airway resistance" and "total lung capacity" were available as indicators of lung function. Details of the measurement procedure were presented elsewhere [23]. An increased airway resistance indicates an obstruction of the airways. For children, higher total lung capacity is related to better lung development. To correct for height, gender and age, total lung capacity was provided as percent predicted. A group of healthy children aged 5 to 10 years and living in the West German study areas in 1991 was used as standard for prediction.
Statistical analysis
As the changes of living and environmental conditions took very different courses in East and West Germany after re-unification and the educational systems in East and West Germany featured many differences before re-unification a separate statistical analysis of the study data was done for the two parts of Germany.
The statistical significance of the difference between two groups was tested using t-test and chi-square test for continuous and dichotomous variables, respectively. A linear time trend across the surveys was checked using the Cochran-Armitage test for dichotomous variables, and for continuous variables, a significance of linear regression on the year of survey was determined using the t-test.
A two-step process of multiple regression analysis was applied to reveal, in the first step, associations between parental education and health indicators and to examine, in the second step, the additional explanatory potency of the living conditions. As health indicators were the dependent variables parental educational level was the independent variable in the first step regression model which also included the three co-variables year of investigation, residency and gender as potential confounders. For the second step, the regression model from the first step was expanded by adding the six living conditions as further independent variables. For dichotomous outcome variables, we used logistic regression, and the linear multiple regression approach was used for continuous outcome variables. Interaction terms between year of investigation and parental educational level were incorporated, and the results of a stratified analysis are presented, if the interaction was statistically significant (p < 0.05). Only children with complete information regarding all dependent and independent variables were included in the respective regression model. The regression results with respect to the associations of the health indicators with educational level and the six living conditions were presented in the case of logistic regression as adjusted odds ratios and in the case of linear regression as adjusted standardized means differences, both together with 95% confidence intervals. The adjusted standardized means difference (MD) was calculated as the ratio of the expected adjusted change of the dependent variable Δy by a unit-increase of the respective independent variable and the sample mean plus unity: MD = 1 + Δy/.
All calculations of the statistical analysis were performed using the SAS statistical software package, Version 9.01 (SAS Institute, Cary, NC).