|
Abstract:
|
Recently, many birth cohort studies have been started. Their target is to detect the effect of risk factors at fetal life to onset of diseases after grown up (ex. Neuropsychiatric developmental disorders, allergies, cancers, obesity etc.). In traditional cohort study, we consider exposures and confounders only at baseline (fetal or newborn period) as risk factors but now we have to consider their life course exposure (changing exposure status at long-term life stage). So it is not sufficient to regard baseline risk factors and covariates. Our study aim is to draw DAG (Directed Acyclic Graph) and apply joint and marginal models to evaluate direct and indirect effects. Our birth cohort study has been conducted in Japan since 2003. Study subject is 1703 children. We focus on the relationship between start time of baby food and onset of allergy at 5 yrs. Covariates and intermediate factors are history of parent's allergy, frequency of cleaning and washing, holding pets.etc. in 1-4yrs. Comparing with crude analysis (include all factors into regression model simultaneously), estimated direct effect and indirect effects is small, and total effect is larger than crude results.
|
