Mendelian Randomisation: A Tool for Assessing Causality in Observational Epidemiology
Detection and assessment of the effect of a modifiable risk factor on a disease with view to informing public health intervention policies are of fundamental concern in aetiological epidemiology. In order to have solid evidence that such a public health intervention has the desired effect, it is necessary to ascertain that an observed association or correlation between a risk factor and a disease means that the risk factor is causal for the disease. Inferring causality from observational data is difficult, typically due to confounding by social, behavioural, or physiological factors which are difficult to control for and particularly difficult to measure accurately. A possible approach to inferring causality when confounding is believed to be present but unobservable, as it may not even be fully understood, is based on the method of instrumental variables and is known under the name of Mendelian randomisation if the instrument is a genetic variant. While testing for the presence of a causal effect using this method is generally straightforward, point estimates of such an effect are only obtainable under additional parametric assumptions. This chapter introduces the concept and illustrates the method and its assumptions with simple real-life examples. It concludes with a brief discussion on pitfalls and limitations.
Key wordsCausal inference Instrumental variable Confounding
We acknowledge research support for all authors from the Medical Research Council through a collaborative project grant (G0601625) addressing causal inference in observational epidemiology using Mendelian randomisation.
- 4.T.O. Scholl and W.G. Johnson. Folic acid: influence on the outcome of pregnancy. American Journal of Clinical Nutrition, 71 (Suppl.):12955–13035, 2000.Google Scholar
- 5.W.C. Willett. Vitamin A and lung cancer. Nutritional Review, 48:201–211, 1990.Google Scholar
- 13.R.J. Bowden and D.A. Turkington. Instrumental Variables. Cambridge University Press, Cambridge, 1984.Google Scholar
- 18.S.L. Lauritzen. Causal inference from graphical models. In O.E. Barndorff-Nielsen, D.R. Cox, and C. Kluppelberg, editors, Complex Stochastic Systems, Chapter 2, 63–107. Chapman & Hall, Boca Raton, 2000.Google Scholar
- 20.A.P. Dawid. Causal inference using influence diagrams: the problem of partial compliance. In P.J. Green, N.L. Hjort, and S. Richardson, editors, Highly Structured Stochastic Systems, 45–81. Oxford University Press, Oxford, 2003.Google Scholar
- 22.J. Pearl. Causality. Cambridge University Press, Cambridge, 2000.Google Scholar
- 24.V. Didelez and N.A. Sheehan. Mendelian randomisation: why epidemiology needs a formal language for causality. In F. Russo and J. Williamson, editors, Causality and Probability in the Sciences, volume 5, Texts in Philosophy, 263–292. London College Publications, London, 2007.Google Scholar
- 25.J.M. Robins, T.J. VanderWeele, and T.S. Richardson. Comment on: Causal effects in the presence of non compliance: a latent variable interpretation. Metron, 64:288–298, 2006.Google Scholar
- 26.S. Geneletti and A.P. Dawid. The effect of treatment on the treated: a decision theoretic perspective. In P. McKay Illari, F. Russo and J. Williamson, editors, Causality in the Sciences,Oxford University Press, 2010.Google Scholar
- 27.V. Didelez, S. Meng, and N.A. Sheehan. Assumptions of IV methods for observational epidemiology. Statistical Science, 25: 22-40, 2010.Google Scholar
- 32.A.A. Balke and J. Pearl. Counterfactual probabilities: computational methods, bounds and applications. In R.L. Mantaras and D. Poole, editors, Proceedings of the 10th Conference on Uncertainty in Artificial Inteligence, 46–54, 1994.Google Scholar
- 36.P. Clarke and F. Windmeijer. Instrumental variable estimators for binary outcomes. Working Paper 10/239, Centre for Market and Public Organisation, University of Bristol, 2010.Google Scholar