Investigating Intergenerational Associations: Revisiting the Children-of-Twins Design I

Datasets comprising twins and their children can be a useful tool for those interested in understanding the nature of intergenerational associations between parent and offspring phenotypes. For example, by using such data it... [ view full abstract ]

Datasets comprising twins and their children can be a useful tool for those interested in understanding the nature of intergenerational associations between parent and offspring phenotypes. For example, by using such data it is possible to ask whether parent-child associations can be explained via genetic transmission, and/or whether associations remain significant after accounting for genetic relatedness between parent and child. By applying structural equation models (SEMs) to children-of-twins (CoT) data, it is possible to quantify the relative importance of genetic vs. non-genetic (social) pathways.

In this talk I will review the SEMs that have been used to analyse CoT data to date. I will highlight some of the limitations that these models have, and discuss some limitations of the CoT datasets that these models have previously been applied to. For example, previous CoT models have had low power to parameterise the aetiology of the offspring phenotype, leading to questions about the models ability to identify the nature of intergenerational transmission. It is also not possible to detect offspring shared environmental effects using cousin data. I will present new variants of the CoT model, showing that including multiple offspring per parent solves many of the shortcomings of previous CoT models, dramatically increases power, and makes it possible to ask questions about the direction of effects between parent and child using a single CoT dataset (i.e. parent-to-child vs. child-to-parent effects). I will demonstrate the power and utility of these models using simulated data and real data from the Intergenerational Transmission of Risk Project (a subsample of the Norwegian Mother and Child Cohort Study). I will finish by highlighting further planned model developments including multivariate and longitudinal CoT models.