Serum 25(OH)D concentrations and atopic diseases at age 10: results from the GINIplus and LISAplus birth cohort studies

The analysis was based on samples from two German birth cohort studies, namely the German Infant Study on the Influence of Nutrition Intervention plus Air pollution and Genetics on Allergy Development (GINIplus) study and the Influence of Life-style factors on Development of the Immune System and Allergies in East and West Germany plus Air Pollution and Genetics on Allergy Development (LISAplus) study, that were recruited from 1995 to 1998 in Munich and Wesel, and from 1997 to 1999 in Munich, Wesel, Bad Honnef and Leipzig, respectively.

Briefly, participants were recruited in maternity wards, and parents were invited to fill in self-administered questionnaires. Since then, families were regularly contacted, and information on disease outcomes in the children was collected by questionnaires and medical examinations. Within the GINIplus framework, children were allocated to either an interventional study arm or an observational study arm, based on parental history of allergic diseases and consent for intervention. GINI aimed to investigate the allergy-preventive effect of 3 differently hydrolyzed infant formulas compared with a conventional cow’s milk formula in infants at risk for atopy in a randomized and double-blind design. During the strict intervention period of four month, mothers were encouraged to exclusively breast-feed and asked to do this preferably for 6 months. It was left to the mother to introduce the study formula, which was provided until the participating child was 6 months old. Mothers were asked to feed no solid foods during the first four months and weekly diaries were filled out documenting the kind of milk the child was fed. In the LISAplus study children were not preselected on parental history of allergic diseases. The LISA study aimed to investigate immunoglobulin E (IgE)-dependent allergic symptoms or diseases in the first 2 years of life as well as to determine parameters of the immune system’s maturation and activation. Both studies were approved by the local ethical committees (Bavarian Board of Physicians, University of Leipzig, Board of Physicians of North-Rhine-Westphalia), and families gave their informed consent prior to their inclusion in the study.

In the first wave about 6000 newborns were recruited within the GINI study and about 3100 newborns within the LISA study. We analyzed data of 2815 children who participated in the 10 year follow-up and for which at least data on serum 25(OH)D concentration was available. The following data from the 10-year follow-up questionnaire and medical examination was used: measured weight and height, single-parent status, and net equivalent income. The 10-year follow-up medical examinations took place from April 2005 to December 2009. Detailed descriptions of both studies were published elsewhere [20],[21].

When examining the distributions of the covariates from those participating in the questionnaire at age 1 year and in the assessment at age 10 years (questionnaire and the corresponding medical examination, serum vitamin D), we found only slight changes (see Additional file 1). Solely for the distribution of `study’ we found a higher loss of follow-up in the observational arm of GINIplus.

We investigated different atopic diseases, either diagnosed at the age of 10 years, or ever diagnosed since birth. All these outcomes are assessed on the basis of self-administered questionnaires data. For asthma, eczema and hay fever or rhinitis, participants were categorized as `case’s at the age of 10 years if a physician first diagnosed or confirmed the diagnosis of the disease during the last 12 month before the 10 year questionnaire was filled in, thereby not taking into account any information on allergic sensitization. For the definition of `ever’ diagnosis of these diseases information from all questionnaires from birth up to the 10 year follow-up was used (self-reports of physician diagnosed diseases). If at least at one time point a disease was reported the participating child was treated as a case for that outcome.

For definition of allergic sensitization, the serum concentrations of specific IgE antibodies against common inhalant (SX1) or food allergens (FX5) at the age of 6 and 10 years were measured. Children with specific IgE values higher than 0.35 kU/l were regarded as sensitized (Pharmacia CAP System (Pharmacia Diagnostics, Freiburg, Germany). For the definition of `ever’ sensitization, at least one of the two tests for specific IgE, performed at age 6 and 10, must have been positive.

For all outcomes regarding the `ever diagnosis’, to minimize the number of missing observations, the status `missing’ was only assigned when all relevant answers to determine the status were missing. In return, this means that the status `non-diseased control’ was assigned even when one or more answers were missing, as long as no diagnosis by a physician was affirmed.

Blood samples were collected at age of 10 years during the clinical examination, centrifuged after collection, and stored frozen at –80° until assayed for vitamin D. Only one measurement per sample was conducted.

Total serum 25(OH)D concentration was determined by Roche Vitamin D total on the fully automated Modular system (E170, Roche Diagnostics, Mannheim, Germany). The specificity is reported by the manufacturer as 25(OH)D₂ = 81%; 25(OH)D₃ = 98%; 1,25(OH)₂D₂ = 6%; 1,25(OH)₂D₃ = 5%; 24,25(OH)₂ = 121%, and the lower limit of detection as 3 ng/mL. The intra-assay CV is 2.2-6.8% for sera with levels between 8.35 - 69.6 ng/mL, the inter-assay CV as provided by the manufacturer is 3.4-13.1% for levels between 8.35-69.6 ng/mL.

All analyses were conducted using SAS for Windows, version 9.2 (SAS Institute, Cary, N.C., USA).

It is well known, that the vitamin D status which is assessed best by serum 25(OH)D concentration, strongly depends on sunlight (UVB) exposure which induces the endogenous production of vitamin D in the skin. Therefore, we adjusted the serum 25(OH)D values for date of blood collection by fitting a non-parametric LOESS regression. Our LOESS regression explained the observed serum 25(OH)D value by the date of blood collection. Season-adjusted serum 25(OH)D values were computed by adding the overall mean to the residuals derived from the LOESS regression. The rationale behind this is that the residuals represent the remaining variation of the serum values, whereas the seasonal effect was accounted for by the date of blood collection. Adding the residuals to the overall mean of the serum values yields a well interpretable adjusted variable. These season-adjusted serum 25(OH)D values were introduced as continuous variable in the logistic regression models, or classified in quartiles based on the distribution of the entire cohort, where the first quartile served as reference category.

To assess the association between serum 25(OH)D levels and the atopic disease status we fitted logistic regression models. Minimally adjusted odds ratio estimates were obtained by adjustment for age (continuous), sex, study (GINI observational arm/GINIinterventional arm/LISA), and study location (Munich/Wesel/Bad Honnef/Leipzig). The fully adjusted logistic regression models further included the following variables: breastfeeding (exclusively breastfed/breast and formula fed/exclusively formula fed) during the first four months, child’s BMI (continuous), parental education (categorized as low/medium/high based on the highest number of school years being lower, equal or higher than 10 years), single parent status (yes/no), and parental history of atopic diseases (none/one/both parents).

Whenever the variable sex was identified as an effect modifier, e.g. the interaction term between sex and continuous 25(OH)D serum level was significantly associated with the outcome (p < 0.05), a stratified analysis was performed.

Additionally, we fitted three models comparing those that never suffered from eczema versus those who suffered transiently and those who suffered persistently.

We ran sensitivity analysis examining higher cut-points for the specific IgEs, considering net equivalent income and by performing a complete case analysis to examine a possible influence of missing observations on the definition of the control status of the clinical outcomes.

A sub-analysis was carried out within the framework of the LISAplus study, using the monthly status on vitamin D supplementation available during the first year of life. We extended the models described above by including the main effect of length of supplementation and an interaction term thereof and 25(OH)D serum level. We investigated categories of at least 4, 6 or 12 months of supplementation, respectively.

Measurement of vitamin D status is challenging as results vary strongly with the methods applied [39]. The 25(OH)D levels analyzed in serum samples of children participating in the GINIplus and LISAplus study are considerably higher than reported elsewhere for this age group. This might be due to the assay applied with a comparably low specificity for 25(OH)D and a high cross-reactivity (see Methods section). We used the serum concentration for ranking (in quartiles) and as a continuous variable (p trend). However, we avoided interpretation of serum concentrations in terms of sufficiency of vitamin D supply that requires the application of cut-off values.

Our evaluation is based on serum 25(OH)D concentrations measured in serum samples obtained at one time point, about 10 years after birth. With the applied LOESS regression we controlled adequately on a continuous scale for the seasonal variation in vitamin D status; in other studies often only the date or season of blood collection was included as a confounder in the analysis. All relevant potential confounders were included in the adjusted models.

The cross-sectional analysis at age 10 is suitable to investigate the association of vitamin D status and atopic diseases. Although the occurrence of atopic diseases from birth up to age 10 years (`ever diagnosis’) was included in our evaluation, serum 25(OH)D concentrations were analyzed in serum samples obtained at age 10 years. Thus, reverse causation effects cannot be ruled out (see e.g. [7]). There is little known about the stability of vitamin D levels in children with a fast changing lifestyle. Our data suggest that the lifetime prevalence of eczema as estimated at the age of 10 years represents predominantly the persistent cases of eczema.

Our analysis of clinical endpoints is based on self-reported physician-based diagnoses and therefore at risk of reporting bias. Furthermore eczema, allergic rhinitis, and asthma are each inexact, clinical diagnoses which might amplify this imprecision. The definition of the `non-diseased control’s as described in the Methods section is debatable. Among the `non-diseased control’s may be cases who missed to report a physician diagnosis. The sensitivity analysis performed included only complete cases in the definition of the control status. This led to a remarkably reduced number of controls. However, no substantial changes regarding the direction of the OR estimates were observed, supporting the validity of the approach taken. We included a non-responder analysis showing that a selective lost for follow up between the study population and the recruited newborns is limited.