Discriminant analysis and logistic regression on genetic history and environmental factors in children with asthma

Abstract

Background: Asthma is known to be related to genetic and environmental factors, we aimed to identify the predictors discriminating between children with asthma and a control group in order to build typical profiles of these children.

Methods: A multidimensional analysis covered children (58 with asthma and 217 as control group), under 17 years of age, involving environmental variables and medical history of these children and their families.

Results: Chi-square tests highlighted significant links between variables as rhinitis and conjunctivitis (P<0.001). The results showed, in group of asthmatic children, significant high frequencies of allergies, mainly seasonal (P<0.001), rhinitis, family history more present in mothers (P=0.002) and in maternal aunts and uncles (P<0.02). Allergies were mostly present in mothers of asthmatic children (P=0.03). Children whose father, mother or both had asthma were significantly more numerous in asthmatic group (P=0.0007). A multiple correspondence analysis (MCA) identified two typical profiles of children, a first group of asthmatic children with positive modalities of family history, medical and environmental factors, a second, the control group (nA, non-asthmatic children), with essentially negative modalities of the variables. Logistic regression (LR) resulted in a final model which retained four significant predictors, rhinitis (P=0.01), atopic dermatitis (P=0.04), mother antecedents (P=0.03) and paternal uncle antecedents (P=0.008) with a globally appreciable predictive value (82%) of the Hosmer-Lemeshow Test.

Conclusions: These results allowed the drafting of a typical profile quantifying through a function of a few predictors, the variation of the probability for a child to develop an asthma.