In situations where multiple environmental factors seem to influence a human trait but a genetic variant exists, which may be quite rare in the population, that is associated with one of the environmental factors, then “Mendelian randomization” gives us as close to a controlled experiment as is possible.
Mendelian here refers to the existence of variants at some locus on the gene that are associated with increased levels of the risk factor [and r]andomization refers to other risk factors having no association with the presence or absence of that variant (Davey Smith and Ebrahim 2007). It turns out that for four CRP variants, each associated with up to 30% higher CRP levels, no association with coronary heart disease has been found (C Reactive Protein Coronary Heart Disease Genetics Collaboration 2011).
CRP may be thought of as an environmental factor possibly associated with the medical trait in the sense that CRP levels can be altered by medication. Mendelian randomization can also be used to investigate hypotheses about environmental factors in a more conventional sense. Consider the apparent association of moderate alcohol consumption with reduced incidence of coronary heart disease (Davey Smith and Ebrahim 2007, 344ff). Alcohol consumption is also associated with other factors associated with coronary heart disease, such as age, smoking, and an increased level of protective high-density lipoprotein. Moreover, nondrinkers may include people who have stopped drinking because of deterioration in their health. To disentangle these factors, we can look at Japanese populations that have a high frequency of a variant of an enzyme aldehyde dehydrogenase (ALDH2). This variant makes the after-effects of drinking very unpleasant. The average alcohol consumption of these people is low for people who are homozygous for the variant and intermediate for those who are heterozygous. The variant is not associated with—is randomized with respect to—age and smoking, so these factors are not involved in any association that remains between coronary heart disease and the ALDH2 variant—as a proxy for alcohol consumption. It turns out that there is an association between the variant and lower levels of high-density lipoprotein, which supports the causal status of moderate alcohol consumption as a protective factor for coronary heart disease (i.e., associated with a reduced incidence) through the effect of alcohol consumption on levels of high-density lipoprotein.
Davey Smith and Ebrahim 2007 provide a review of other examples and address potential limitations of Mendelian randomization.
[In order for] application of the results—assuming the results show that the modifiable factor is causally associated with the trait—… the factor must be reduced in the population by means that do not exacerbate the other factors associated with the trait. The means of securing that reduction may involve a clinical focus on treating the subset of individuals who have high levels of the factor in question or a public health focus on reducing the average levels of the factor in the population. (The tension between a clinical and public health focus is longstanding in epidemiology and applies not only to the results of Mendelian randomization; Rose 2008 ).
The quotes above are from pages 144ff in Taylor, Peter J. (2014) Nature-Nurture? No: Moving the Sciences of Variation and Heredity Beyond the Gaps.
C Reactive Protein Coronary Heart Disease Genetics Collaboration (2011). “Association between C reactive protein and coronary heart disease: Mendelian randomisation analysis based on individual participant data.” British Medical Journal 342(15 February): d548.
Davey Smith, G. and S. Ebrahim (2007). “Mendelian randomization: Genetic variants as instruments for strengthening causal influences in observational studies. Pp. 336-366 in Biosocial Surveys. M. Weinstein, J. W. Vaupel and K. W. Wachter (Eds.) Washington, DC: National Academies Press.
Rose, G. (2008 ). Rose’s Strategy of Preventive Medicine. Oxford: Oxford University Press.
(Introduction to this series of posts)