In epidemiology, Mendelian randomization is a method of using measured variation in genes of known function to examine the causal effect of a modifiable exposure on disease in non-experimental studies. The design was first described by Gray and Wheatley (1991) as a method for obtaining unbiased estimates of the effects of a putative causal variable without conducting a traditional randomised trial. These authors also coined the term Mendelian randomization. The design has a powerful control for reverse causation and confounding which otherwise bedevil epidemiological studies. 
Background: the problem of spurious findings in observational epidemiology
An important focus of observational epidemiology is the identification of modifiable causes of common diseases that are of public health interest. In order to have firm evidence that a recommended public health intervention will have the desired beneficial effect, the observed association between the particular risk factor and disease must imply that the risk factor actually causes the disease.
Well-known successes include the identified causal links between smoking and lung cancer, and between blood pressure and stroke. However, there have also been notable failures when identified exposures were later shown by randomised controlled trials (RCTs) to be non-causal. For instance, it has now been shown that hormone replacement therapy will not prevent cardiovascular disease, as was previously thought, and may have other adverse health effects (Rossouw et al. 2002). The reason for such spurious findings in observational epidemiology is most likely to be confounding by social, behavioural or physiological factors which are difficult to control for and particularly difficult to measure accurately. Moreover, many findings cannot be replicated by RCTs for ethical reasons.
The Mendelian randomization approach
"Genetics is indeed in a peculiarly favoured condition in that Providence has shielded the geneticist from many of the difficulties of a reliably controlled comparison. The different genotypes possible from the same mating have been beautifully randomised by the meiotic process. A more perfect control of conditions is scarcely possible, than that of different genotypes appearing in the same litter." --R.A. Fisher
Mendelian randomization is a method that allows one to test for, or in certain cases to estimate, a causal effect from observational data in the presence of confounding factors. It uses common genetic polymorphisms with well-understood effects on exposure patterns (e.g., propensity to drink alcohol) or effects that mimic those produced by modifiable exposures (e.g., raised blood cholesterol (Katan 1986)). Importantly, the genotype must only affect the disease status indirectly via its effect on the exposure of interest. Because genotypes are assigned randomly when passed from parents to offspring during meiosis, if we assume that choice of mate is not associated with genotype (panmixia), then the population genotype distribution should be unrelated to the confounders that typically plague observational epidemiology studies. In this regard, Mendelian randomization can be thought of as a “natural” randomized controlled trial. Because the polymorphism is the instrument, Mendelian randomization is dependent on on genetic association studies having provided good candidate genes for response to risk exposure.
From a statistical perspective, MR is an application of the technique of instrumental variables (Thomas & Conti 2004, Didelez & Sheehan 2007), with genotype acting as an instrument for the exposure of interest.
In addition to direct effects of the instrument on the disease, misleading the analyst, misleading conclusions may also arise in the presence of linkage disequilibrium with unmeasured directly-causal variants, genetic heterogeneity, pleiotropy, or population stratification (Davey Smith & Ebrahim 2003).
- Davey Smith, G. (2010). "Mendelian Randomization for Strengthening Causal Inference in Observational Studies: Application to Gene × Environment Interactions". Perspectives on Psychological Science: 527-545. doi:10.1177/1745691610383505.
- Fisher, R.A. (1952). "Statistical methods in genetics". Heredity 6 (1): 1–12.
- G. Davey Smith and S. Ebrahim (2003) Mendelian randomization: can genetic epidemiology contribute to understanding environmental determinants of disease? International Journal of Epidemiology 32: 1-22.doi:10.1093/ije/dyg070
- G. Davey Smith, S. Ebrahim, S. Lewis, A.L.Hansell, L.J. Palmer and P.R. Burton (2005) Genetic epidemiology and public health: hope, hype, and future prospects. Lancet 366: 1484-1498.doi:10.1016/S0140-6736(05)67601-5
- G. Davey Smith and S. Ebrahim (2005) What can Mendelian randomisation tell us about modifiable behavioural and environmental exposures? BMJ 330: 1076-1079.doi:10.1136/bmj.330.7499.1076
- V. Didelez and N. Sheehan (2007) Mendelian randomization as an instrumental variable approach to causal inference. Statistical Methods in Medical Research 16:309-330 doi:10.1177/0962280206077743
- M.B. Katan (1986) Apolipoprotein E isoforms, serum cholesterol and cancer. Lancet, 327:507-508.
- R. Gray and K. Wheatley. (1991). How to avoid bias when comparing bone marrow transplantation with chemotherapy. Bone Marrow Transplant, 7 Suppl 3, 9-12
- J.E. Rossouw et al. (2002) Risks and benefits of estrogen plus progestin in healthy post-menopausal women: principal results from the Women’s Health Initiative randomized controlled trial. JAMA 288: 321-333.
- D.C. Thomas and D.V. Conti (2004) Commentary: The concept of Mendelian randomization. International Journal of Epidemiology 32: 21-25 doi:10.1093/ije/dyh048
- Mendelian Randomization: A Perfect Causal Epidemiologic Approach to Simulate a Randomized Trial? Epidemiologic Inquiry 2006, 1: 16
- G. Davey Smith (2006). Capitalising on Mendelian randomization to assess the effects of treatments. James Lind Library.
- PHOEBE Biostatistics Group (2007) Mendelian randomisation: Inferring causality in observational epidemiology