|
Abstract:
|
We need life course approach to estimate effect of early life exposure or early life status on their disease developed when they grew up. In the birth cohort data, many exposures, covariates and outcome variables are repeatedly measured over time. When we draw DAG(directed acyclic graph) to organize the relationship between variables and to verify the specific hypothesis, DAG would be very complicated. We apply joint and marginal model to estimate effect of birth weight on childhood allergy with whether study subjects have pets(dogs or cats) in the house as a repeatedly measured confounder. In our birth cohort study, we have followed up about 1700 mothers and children as a study subject in 10 years with annual questionnaire survey and face to face survey for every five years. As a result, the risk of develop allergy (asthma and atopic dermatitis) was 1.02(95%CI:0.93-1.24) in low-birth-childrenlow birth weight children(< 2500g) without adjusting correlation between repeatedly measured confounders using conditional model, 1.13(1.01-1.45) with joint modeling, 1.24(1.10-1.48) with marginal modeling. When there are time dependent confounders, we recommend to use marginal modeling.
|
