Monday, 06 January 2020 15:19

Critical review of Mendelian randomisation used in research of alcoholic beverages

This critical review proposes to regard Mendelian Randomization (MR) as another type of observational study, not as an alternative to observational studies to determine causality. The researchers highlight the limitations of MR when applied to the associations of alcohol consumption with cardiovascular disease (1).

MR is a method that is increasingly being used to conduct studies on alcoholic beverages and health and as result challenges the findings of previous research, such as the J-shaped curve.

What does it mean:

This paper (1) is a critical analysis of the MR method used in recent research on alcoholic beverages that has generated debate in the scientific community and the public. Some researchers have claimed that the results of certain MR studies (2,3) disprove the protective effect of moderate drinking for cardiovascular disease (CVD). The authors of the current publication argue that - because of a number of limitations - MR should be viewed as another type of observational study and therefore, should not be seen to negate the existing body of evidence which supports the J-curve relationship between drinking and CVD.

References:

  1. Mukamal, K. J., Stampfer, M. J., & Rimm, E. B. (2019). Genetic instrumental variable analysis: Time to call Mendelian randomization what it is. The example of alcohol and cardiovascular disease. European Journal of Epidemiology. doi:10.1007/s10654-019-00578-3.
  2. Holmes, M. V., Dale, C. E., Zuccolo, L., Silverwood, R. J., Guo, Y., & Engmann, J. (2014). Association between alcohol and cardiovascular disease: Mendelian randomisation analysis based on individual participant data. British Medical Journal, 349
  3. Millwood, I. Y., Walters, R. G., Mei, X. W., Guo, Y., Yang, L., Bian, Z., et al. (2019). Conventional and genetic evidence on alcohol and vascular disease aetiology: A prospective study of 500,000 men and women in China. The Lancet, 393, 1831-1842.

 

Mendelian Randomization (MR) - Background information:

In recent years, epidemiologists have increasingly sought to use genetic data to find “causal” relationships between the exposures of interest and various endpoints, for example exposure to alcoholic beverages and cardiovascular disease as endpoint. This approach is called Mendelian randomization (MR).

The problem in observational studies is establishing causality. A best way to test cause and effect associations are long-term clinical trials, however, long-term trials on diet or drinking patterns, are not readily tested in randomized trials.

 

What is Mendelian randomization?

Assumption:

Fig. 1

MR is a research method that uses an individual’s genes to estimate their behavior, rather than asking them to report their behavior. Genes vary across a population and these variations can influence behavior (known as a ‘genetic proxy’ for the behavior). In MR, researchers use these genetic proxies to see, if certain behaviors are linked to health outcomes, such as specific diseases.     

        

Why is MR used?

MR can be used as an alternative to a traditional method of analyzing the consumption of alcoholic beverages (epidemiology) [1, 2], which has previously identified a J-shaped relationship between alcohol intake and cardiovascular disease (CVD). The J-curve shows that moderate drinking is associated with a lower risk of CVD compared to non-drinking or heavy drinking. But because of the nature of these epidemiological studies, this evidence on its own cannot show that the relationship is causal, and it may be that moderate drinkers are healthier than non-drinkers. Therefore, MR has been promoted as an alternative method to identify, if the link is causal or not. When used correctly, MR can remove the influence of other factors and identify direct causation between a behavior and a health outcome.

How is it used to address alcohol and health questions?

MR studies

Scientists have identified several gene variants that affect the way individuals metabolize alcohol. These gene variants lead to unpleasant symptoms, such as facial flushing, and are generally associated with a reduced consumption of alcoholic beverages. Researchers have exploited this effect as a “genetic proxy” for alcohol consumption, since individuals who experience unpleasant symptoms are less likely to drink, and compare health outcomes between individuals with and without these genes.

Epidemiological studies

On the other hand, epidemiological studies compare health outcomes between different types of drinkers based on their reported alcohol consumption, however, self-reports may not be accurate, potentially biasing these studies. Also, it is important to determine how patterns of consumption affect health outcomes [3], but it can be difficult to accurately identify a person’s lifetime drinking habits. MR may be a way to address this, if the genetic basis for limited alcohol intake was consistent and did not change over a person’s lifetime.

What have MR studies on alcohol indicated so far?

Where epidemiological studies find a J-curve with CVD risk and all-cause mortality risk, MR has shown an increase in risk as the level of consumption goes up with no “protective” effect being shown [4, 5]. However, there are questions [6-8] about whether the current approach to MR research for alcohol consumption and health satisfies the required assumptions for this type of analysis [9, 10].

What are the limitations of the current MR research to study alcohol and health?

  • MR may not work correctly, if the genetic proxy affects the outcome independently of the behavior, an effect known as pleiotropy (*) [9] (Fig 1). For example, if individuals with a gene variant that affects alcohol consumption are more or less prone to CVD (regardless of whether they drink alcohol), then an MR study using these genes would not be able to accurately tell, if the alcohol consumption, or the genes themselves, are influencing the risk of CVD.
  • Currently, MR research focuses on the variants of two genes: the alcohol dehydrogenase (ADH) and the aldehyde dehydrogenase (ALDH). Studies have found that both these variants affect risk factors for CVD independently of alcohol consumption: the ADH1B gene affects Body Mass Index (BMI) and blood pressure [4], and the ALDH2 gene has an effect on HDL cholesterol [11, 12]. This suggests that the genes may be pleiotropic (*).
  • There is another issue with using ADH and ALDH genes as a proxy: they are both involved in the breakdown of alcohol in the body. These variants of the proxy genes, which are often used in MR studies for “lifetime low level consumption”, increase the buildup of a toxic substance called acetaldehyde. This buildup can make drinkers feel unwell and limit their drinking habits, and it can affect the health of the individual [13], which is not comparable to people who don’t carry these genetic variants.
  • MR works best when the genetic proxy strongly and consistently predicts behavior. If people with the gene don’t always behave as predicted, or if the influence of the gene on behavior is very small, then MR will be less able to determine, if the behavior is driving the results: this is known as “weak instrument bias” [9]. ADH1B and ALDH2 gene variants may not be ideal genetic proxies for the behavior of lifetime low level consumption of alcohol. This is because carriers of the gene variant have still been documented binge drinking [14] and regularly heavy drinking [15], even though carriers of either of these gene variants show lower average alcohol consumption.
  • Another study found that even among those with the ALDH2 gene variant, alcohol intake was influenced by individual’s behavior [16]. Ideally, an MR study which shows the relationship between alcohol and CVD would use a gene variant which had a strong, lifetime effect on alcohol consumption. For example, if it limited an individual’s consumption to very low but consistent levels, which wouldn’t be increased but also wouldn’t stop them drinking entirely. This gene would be a proxy for “lifetime low level consumption” and researchers would analyze this carrier group’s CVD risk, compared to those who drank more and drank nothing, to see if the J-curve exists or not. But this ideal MR method for alcohol consumption has not yet been identified or utilized.
  • Currently, the strongest “genetic proxy” is ALDH2, but this gene variant is restricted within ethnic populations, for instance the ALDH2 gene variant rs671 only exists within south East Asian populations, and therefore the results may not be applicable to other ethnic groups. The South East Asian population has a different risk profile associated with their diet and lifestyles as well as their genetics, which may not be applicable to non-South East Asian populations or other cultures/geographical areas.

What are the implications of MR for alcohol and health research?

MR is a relatively novel and potentially useful tool in health research. Questions remain about its application to alcohol research [7, 12, 17-20], and the relationship between different genes and alcohol consumption, particularly across different populations. The evidence base on alcohol and health is large and broad and it is too early to suggest, if results from MR studies invalidate previous findings, particularly as MR results have not been consistent. Further research is needed to refine MR approaches to alcohol and to reconcile its results with results from epidemiological, biomedical, and experimental research.

 

References:

  1. Di Castelnuovo, A., Costanzo, S., Bagnardi, V., Donati, M. B., Iacoviello, L., & de Gaetano, G. (2006). Alcohol dosing and total mortality in men and women: An updated meta-analysis of 34 prospective studies. Archives of Internal Medicine, 166(22), 2437-2445.
  2. Ronksley, P. E., Brien, S. E., Turner, B. J., Mukamal, K. J., & Ghali, W. A. (2011). Association of alcohol consumption with selected cardiovascular disease outcomes: A systematic review and meta-analysis. British Medical Journal, 342(7795), 479.
  3. Britton, A., Ben-Shlomo, Y., Benzeval, M., Kuh, D., & Bell, S. (2015). Life course trajectories of alcohol consumption in the United Kingdom using longitudinal data from nine cohort studies. BMC Med, 13, 47.
  4. Holmes, M. V., Dale, C. E., Zuccolo, L., Silverwood, R. J., Guo, Y., & Engmann, J. (2014). Association between alcohol and cardiovascular disease: Mendelian randomisation analysis based on individual participant data. British Medical Journal, 349, g4164.
  5. Millwood, I. Y., Walters, R. G., Mei, X. W., Guo, Y., Yang, L., Bian, Z., et al. (2019). Conventional and genetic evidence on alcohol and vascular disease aetiology: A prospective study of 500,000 men and women in China. The Lancet. 10.1016/S0140-6736(18)31772-0.
  6. Mukamal, K. J., & Ding, E. L. (2016). Pinpointing the health effects of alcohol. Bmj, 353, i3043.
  7. Gmel, G. (2017). Beneficial effects of moderate alcohol use—a case for Occam's razor? Addiction, 112(2), 215-217.
  8. Costanzo, S., de Gaetano, G., Di Castelnuovo, A., Djoussé, L., Poli, A., & Van Velden, D. P. (2019). Moderate alcohol consumption and lower total mortality risk: Justified doubts or established facts? Nutrition, Metabolism and Cardiovascular Diseases. https://doi.org/10.1016/j.numecd.2019.05.062.
  9. Davies, N. M., Holmes, M. V., & Davey Smith, G. (2018). Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. Bmj, 362, k601.
  10. Pierce, B. L., Ahsan, H., & Vanderweele, T. J. (2011). Power and instrument strength requirements for Mendelian randomization studies using multiple genetic variants. International Journal of Epidemiology, 40(3), 740-752.
  11. Kato, N., Takeuchi, F., Tabara, Y., Kelly, T. N., Go, M. J., Sim, X., et al. (2011). Meta-analysis of genome-wide association studies identifies common variants associated with blood pressure variation in east Asians. Nat Genet, 43(6), 531-538.
  12. Wada, M., Daimon, M., Emi, M., Iijima, H., Sato, H., Koyano, S., et al. (2008). Genetic association between aldehyde dehydrogenase 2 (ALDH2) variation and high-density lipoprotein cholesterol (HDL-C) among non-drinkers in two large population samples in Japan. J Atheroscler Thromb, 15(4), 179-184.
  13. Xiangwei, L., & Aijun, S. (2017). Aldehyde dehydrogenase-2 roles in ischemic cardiovascular disease. Current Drug Targets, 18(15), 1817-1823.
  14. Doran, N., Myers, M. G., Luczak, S. E., Carr, L. G., & Wall, T. L. (2007). Stability of heavy episodic drinking in Chinese- and Korean-American college students: effects of ALDH2 gene status and behavioral undercontrol. J Stud Alcohol Drugs, 68(6), 789-797.
  15. Wall, T. L. (2005). Genetic associations of alcohol and aldehyde dehydrogenase with alcohol dependence and their mechanisms of action. Therapeutic Drug Monitoring, 27(6), 700-703.
  16. Luczak, S. E., Yarnell, L. M., Prescott, C. A., Myers, M. G., Liang, T., & Wall, T. L. (2014). Effects of ALDH22 on alcohol problem trajectories of Asian American college students. J Abnorm Psychol, 123(1), 130-140.
  17. Glymour, M. M. (2014). Alcohol and cardiovascular disease. Bmj, 349, g4334.
  18. Katan, M. B. (2014). Association between alcohol and cardiovascular disease: Mendelian randomisation analysis based on individual participant data [Response to Author]. British Medical Journal, 349, g4164.
  19. Israel, Y. (2014). Association between alcohol and cardiovascular disease: Mendelian randomisation analysis based on individual participant data. [Response to Author]. British Medical Journal, 349, g4164.
  20. Rehm, J. (2014). Association between alcohol and cardiovascular disease: Mendelian randomisation analysis based on individual participant data [Response to Author]. British Medical Journal, 349, g4164.
  21. IARD

 

(*) Pleiotropy – the gene variants used in the study affect the health of a person regardless of their drinking behavior. For example, some genes result in the build-up of a toxic substance (acetaldehyde) unrelated to drinking alcohol, which does not happen in individuals who do not carry the gene variant (and can metabolise alcohol normally), so these subgroups are not directly comparable.

 

 

Disclaimer

The authors have taken reasonable care in ensuring the accuracy of the information herein at the time of publication and are not responsible for any errors or omissions. Read more on our disclaimer and Privacy Policy.