Mendelian randomization studies are becoming increasingly common in cardiovascular research. It is, therefore, important for members of the cardiovascular medicine community to become familiar with this study design and comfortable interpreting the results of these studies.

The basic goal of a Mendelian randomization study is to introduce a randomization scheme into an observational study in an attempt to avoid the confounding, bias and reverse causation that can sometimes limit the validity of an observed epidemiologic association. Mendelian randomization studies are designed to determine if a non-genetic environmental exposure, such as a putative risk factor, is causally associated with the condition under study.¹ Ironically, these studies have very little to do with genetics. Instead, genetic polymorphisms that are associated with an environmental exposure are used merely as a convenient instrument to randomly allocate study subjects to higher or lower levels of the exposure of interest, thus allowing inferences to be made about whether that exposure is causally associated with the risk of disease.

Perhaps the easiest way to understand a Mendelian randomization study is by way of analogy with a randomized trial. Indeed, Mendelian randomization studies have been called "nature's randomized trials."² For example, there are many polymorphisms that are associated with plasma levels of low-density lipoprotein cholesterol (LDL-C).³ Each of these polymorphisms is allocated approximately randomly at the time of conception in a process sometimes referred to as Mendelian randomization.⁴ Therefore, inheriting an allele associated with lower LDL-C is analogous to being randomly allocated to an LDL-C lowering therapy at birth, while inheriting the other allele is analogous to being randomly allocated to usual care. If the polymorphism under study is associated only with LDL-C, but not with any other lipid or non-lipid pleiotropic effect; and if allocation is indeed random, then the only difference between the groups being compared should be their LDL-C level. As a result, comparing the risk of coronary heart disease among persons with and without an LDL-C lowering allele should provide a naturally randomized and unconfounded estimate of the causal effect of LDL-C on the risk of coronary heart disease in a manner analogous to a long-term randomized trial (Figure 1).

Figure 1

Indeed, the rationale for a Mendelian randomization study is the same as the rationale for a randomized trial. The goal is to randomly allocate study subjects into groups that are essentially identical in all characteristics except the exposure (or intervention) under study. The success of this natural randomization scheme can be tested objectively, just as in a randomized trial, by comparing the baseline characteristics among persons with and without the polymorphism under study. If there are no differences in any baseline characteristics between the groups being compared except the level of the exposure mediated by the polymorphism being used as an instrument to randomly allocate subjects, then allocation must have been random. As a result, the study should provide a naturally randomized and unconfounded estimate of the causal effect of that exposure on the risk of disease.

An important limitation of many Mendelian randomization studies is that most common polymorphisms have a very weak effect on environmental exposures such as LDL-C.³ It can be quite difficult to establish a causal association between an exposure and the risk of disease when evaluating polymorphisms that have a very weak effect on the exposure under study. This problem is analogous to the difficulty interpreting the results of a randomized trial evaluating the effect of a lipid lowering therapy on coronary heart disease when that therapy has a very weak effect on LDL-C, such as niacin or the fibrates.^5,6 Because of the weak effect of most polymorphisms, it is sometimes useful to combine the effects of multiple different polymorphisms on the exposure of interest into a genetic score to create an instrument that has a much greater effect on that exposure.

For example, polymorphisms in multiple different genes have been reported to be associated with both lower LDL-C and a corresponding lower risk of coronary heart disease.^7,8 Measuring the effect of lower LDL-C mediated by the combined effect of all of these polymorphisms in a genetic LDL-C score can provide greater statistical power to estimate the true magnitude of the causal effect of lower LDL-C on the risk of coronary heart disease. This type of genetic risk score can be calculated in two steps. First, the effect of each polymorphism on the risk of coronary heart disease is adjusted for a standard increment in LDL-C, e.g., 10 mg/dL lower LDL-C. Then, the adjusted effect estimates are combined in a meta-analysis to produce a summary estimate of the causal effect of each 10 mg/dL lower LDL-C on the risk of coronary heart disease. Interestingly, this method of calculating a genetic LDL-C score is exactly the same method used to perform a meta-analysis of randomized trials. For example, the Cholesterol Treatment Trialists Collaborators first adjusted the result of each statin trial for a standard increment of 1 mmol/L (38.67 mg/dL) lower LDL-C, and then combined the adjusted results in a meta-analysis to produce a summary estimate of the effect of statins on the risk of coronary heart disease per mmol/L lower LDL-C.⁹

It is clear from the above discussion that a Mendelian randomization study, if designed well, can be thought of as analogous to a "naturally randomized trial." This analogy can be extended further by recognizing that a group of Mendelian randomization studies evaluating the association between a specific exposure and the risk of disease can be thought of as analogous to a portfolio of "naturally randomized trials" that can be meta-analyzed to produce a summary estimate of effect called a genetic score that is calculated in exactly the same way that one would meta-analyze a portfolio of randomized trials. Framed in this way, it becomes clear that Mendelian randomization studies can be designed to answer clinically relevant questions.

For example, this powerful study design has the potential to determine if a risk factor is causally associated with the risk of disease and, therefore, can help to determine whether treating that risk factor is likely to reduce the risk of disease and by how much. These studies can also help determine who is likely to benefit from a treatment and when treatment should be initiated. In cardiovascular medicine, Mendelian randomization studies have already demonstrated that long-term exposure to lower LDL-C beginning earlier in life is associated with a 3-fold greater reduction in the risk of CHD than treatment with a statin started later in life thus leading to the emerging notion that with regard to LDL-C "lower is better and earlier is better";⁷ that the clinical benefit of exposure to lower LDL-C appears to be independent of the mechanism by which LDL-C is lowered, thus suggesting that the method used to lower LDL-C is less important the absolute magnitude of the achieved LDL-C reduction;⁸ and that the usual age-related rise in systolic blood pressure can be substantially slowed by interrupting the cycle of accumulating vascular injury soon after blood pressure begins to rise with age, thus introducing the notion that hypertension can be potentially prevented.¹⁰

Because large randomized trials are expensive and time-consuming, they can only be used to address a very small number of the many important unresolved questions in cardiovascular medicine. By contrast, Mendelian randomization studies can be performed much more rapidly and inexpensively than randomized trials, and they can address a much broader range of clinically relevant questions. As a result, Mendelian randomization studies have the potential to help reshape cardiovascular medicine by generating naturally randomized evidence that can help to fill evidence gaps when an actual randomized trial would be either impossible or impractical to conduct.

However, it is important to recognize that that bias can be introduced into a Mendelian randomization study, just as it can be introduced into a randomized trial, by a poor study design. It is, therefore, critically important that members of the cardiovascular medicine community understand this powerful methodology and actively participate in the critical appraisal of these studies to ensure that they are conducted well and interpreted properly.

References

1. Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med 2008;27:1133-63.
2. Hingorani A, Humphries S. Nature's randomised trials. Lancet 2005;366:1906-8.
3. Global Lipids Genetics Consortium, Willer CJ, Schmidt EM, Sengupta S, et al. Discovery and refinement of loci associated with lipid levels. Nat Genet 2013;45:1274-83.
4. Davey Smith G, Ebrahim S. 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 2003;32:1-22.
5. HPS2-THRIVE Collaborative Group. Effects of extended-release niacin with laropiprant in high-risk patients. N Engl J Med 2014;371:203-12.
6. ACCORD Study Group. Effects of combination lipid therapy in type 2 diabetes mellitus. N Engl J Med 2010;362:1563-74.
7. Ference BA, Yoo W, Alesh I, et al. Effect of long-term exposure to lower low-density lipoprotein cholesterol beginning early in life on the risk of coronary heart disease: a Mendelian randomization analysis. J Am Coll Cardiol 2012;60:2631-9.
8. Ference BA, Majeed F, Penumetcha R, Flack JM, Brook RD. Effect of naturally random allocation to lower low-density lipoprotein cholesterol on the risk of coronary heart disease mediated by polymorphisms in NPC1L1, HMGCR, or both: a 2 x 2 factorial Mendelian randomization study. J Am Coll Cardiol 2015;65:1552-61.
9. Cholesterol Treatment Trialists' (CTT) Collaboration. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. Lancet 2010;376:1670-81.
10. Ference BA, Julius S, Mahajan N, et al. Clinical effect of naturally random allocation to lower systolic blood pressure beginning before the development of hypertension. Hypertension 2014;63:1182-8.

Keywords: Alleles, Blood Pressure, Blood Pressure Determination, Cholesterol, Cholesterol, LDL, Coronary Artery Disease, Environmental Exposure, Fibric Acids, Hydroxymethylglutaryl-CoA Reductase Inhibitors, Hypertension, Lipoproteins, LDL, Niacin, Polymorphism, Single Nucleotide, Random Allocation, Risk Factors, Vascular System Injuries, Primary Prevention

< Back to Listings