Unmeasured confounding is a general problem in epidemiological research and one example is the controversy regarding whether the often observed association between low educational and Coronary Heart Disease (CHD) is a causal effect or spurious association due to confounding by childhood environment. Since higher education is self selected, i.e. not randomized, it has been difficult to draw causal conclusions from observed associations.

Photo: congerdesign/Pixabay.com

Attempts has been made to estimate the causal effect by adjusting for potential confounders such as childhood socioeconomic status, e.g. parental education and income, smoking during pregnancy and other lifestyle factors, or by using family design by sibling comparison. Possible paths of how education could influence CHD risk has also been studied. In recent years Genom Wide Association studies have detected a large set of genetic variants associated with educational attainment. These genetic variants has been used as Instrumental Variables (IVs), i.e. Mendelian Randomization (MR), to estimate a potential causal effect of educational attainment on the risk of CHD. In order to estimate the causal effect in IV analysis a range of assumptions needs to be fulfilled. This presentation will cover an introduction to IV analysis, previous work on using MR to estimate the causal effect of educational attainment on CHD risk and limitations in these studies. I will also show my recent work on using data from the UK biobank to estimate the Attributable Fraction (AF) of educational qualification on CD risk.

We wish you all welcome!