April 4, 2019

expert reaction to research on alcohol intake and stroke

Research published in The Lancet demonstrates that moderate alcohol consumption does not protect against stroke as previous studies have claimed.

Dr Stephen Burgess, Group Leader at the MRC Biostatistics Unit, University of Cambridge, said:

“While it is well known that heavy alcohol consumption is hazardous, previous studies have shown that non-drinkers and ex-drinkers are at higher risk of cardiovascular disease compared with light drinkers. This has led some to suggest that light drinking may protect individuals against cardiovascular disease. This phenomenon of higher risk in non-drinkers and ex-drinkers was observed in the Chinese cohort investigated here. However, non-drinkers and ex-drinkers in this cohort have lower levels on average of education, lower average income, lower physical activity, and higher rates of previous disease. The increased risk of cardiovascular disease may be explained by these factors and not by drinking behaviour.

“To address this question, the authors consider genetic variants that predispose people to drinking more or less alcohol. These genetic variants are not commonly encountered in people with European ancestry, but they are common in those with Chinese ancestry. Although groupings of people with genes that predispose them to drinking more or less alcohol have marked different average levels of alcohol consumption, in contrast average levels of education, income and physical activity did not differ across these genetic groupings. These genetic variants therefore act in an analogous way to treatment allocation in a randomized controlled trial – they influence people’s level of alcohol consumption without also influencing other risk factors. The genetic variants allow us to compare groupings within the population that have different average levels of alcohol consumption, but similar levels of other risk factors.

“When comparing these genetic groupings, we observe no protective effect of low alcohol consumption for any cardiovascular disease, but instead a steady increase in risk of stroke that is proportional to the average level of alcohol consumption. This suggests that drinking alcohol, even at moderate levels, increases the risk of stroke. The association between genetically predicted alcohol consumption and coronary heart disease was flat, with no evidence of increasing risk with increased alcohol consumption, but also no evidence for a protective effect of moderate drinking. However, this null finding reflects the absence of strong evidence, rather than strong evidence of the absence of an effect.

“A limitation of the study is that it was carried out in a Chinese population in which alcohol consumption mainly consisted of beer and spirits, and so results may not be applicable to other types of alcohol. However, China is a very heterogeneous country and similar results were observed across the different provinces in which the research was conducted. Additionally, risk was proportional to the amount of alcohol consumed, suggesting that findings are not sensitive to the type of alcohol consumed.

“This research reflects the culmination of many years of research into the impact of alcohol consumption. It strongly suggests that there is no cardiovascular benefit of light drinking, and that risk of stroke increases even with moderate light alcohol consumption. Risk of stroke increases proportionally with the amount of alcohol consumed, so if people do choose to drink, then they should limit their alcohol consumption.”

Prof David Spiegelhalter, Winton Professor for the Public Understanding of Risk, University of Cambridge, said:

“This is a very impressive study which shows that men who, by chance, have a combination of genes that put them off drinking alcohol, have a lower risk of stroke compared with those without these genes.  The fact that this is not true for Chinese women, who tend not to drink whatever their genes, suggests this effect is due to the alcohol rather than the genes themselves.

“The increase in total stroke risk was around 38% for every 40g of alcohol drunk per day; that’s 5 UK units, or over a half-bottle of wine.  To give this some perspective, it is very roughly the opposite effect of taking a statin.

“I have always been reasonably convinced that moderate alcohol consumption was protective for cardiovascular disease, but now I am having my doubts.”

Prof Deborah Lawlor, Professor of Epidemiology at the MRC Integrative Epidemiology Unit, University of Bristol, said:

“This paper provides evidence of a dose-response effect of alcohol on stroke. The main genetic analyses show that for each 280g more per week of alcohol someone drinks their risk of haemorrhagic (bleeding into the brain) stoke increases on average by 58% and of ischaemic (clot in the brain blood vessels) stroke by 27%. To make this easier to digest a can of 5% strength beer, standard glass of 12% wine or a single shot of spirit each contain about 14g of alcohol. So if one group of people each 10 shots of spirit per week and another group drank 30 shots, we would expect the rates of haemorrhagic strokes to be 58% higher and of ischaemic strokes to be 27% higher in the second group.

“This is a large study and for the main analyses uses genetic factors (instead of relying on what study participants reported) to measure how much alcohol participants drank. The genetic factors they used affect how much someone drinks because one version causes you to feel really unwell when you drink alcohol. The version of any gene that we get from our parents (e.g. the version that would decrease how much we drank or the version that would not) occurs by a random process. This means that the genetic factors that were used to measure alcohol are less likely to be related to any other factors, such as smoking, that might be what is really causing the increased risk of stroke. Also, we get our genes at conception and these cannot be changed by later disease. So, someone who started to feel ill, for example if they had started to get thickened arteries that could later cause a stroke, cannot change their genes, whereas they might start to drink less. Therefore, using genetic variants to measure alcohol rather than self-report can get rid of some of the problems of other factors causing an apparent (non-causal) association of alcohol with stroke that are common in conventional methods.  These results support alcohol having a causal effect on stroke risk. This conclusion is further strengthened by the fact that the alcohol related genetic variants did not affect stroke risk in women. Women from the regions of China where this study was undertaken do not tend to drink alcohol for cultural reasons. As the genetic factors used in this study influence how much someone drinks by making them feel ill when they drink they are only good measure of alcohol in populations that drink. In populations that do not (as in the women in this study) seeing no effect of the genes on stroke risk is evidence that the effects seen in men are due to alcohol and not anything else that the genes might affect.

“Three important questions remain. First, as this study was undertaken in a Chinese population can we assume that the results apply to everyone else (e.g. African or European populations)? It is possible. When the authors used more ‘conventional’ analyses their results were similar to those seen in European populations. However, most of the alcohol consumed in this study was spirits and in populations where beer and/or wine are more common the results may be different. Second, the absolute effect is not known. The authors focus primarily on relative risk and absolute effects are only hinted at in the discussion. If the assumptions of the genetic method are correct these results imply that about 16% of haemorrhagic strokes and 8% of ischaemic strokes are due to alcohol in Chinese men. Though this sound less dramatic than a 58% and 27% increased relative risk, it nonetheless demonstrates if alcohol consumption were reduced in this population there would be a reduction in strokes. On the other had in populations with low risk of stoke the impact in absolute terms would be less.

“Lastly, despite finding evidence of an effect of alcohol on blood pressure, which is an important risk factor for strokes and heart attacks, the results in this study did not find evidence of an effect of alcohol on heart disease, which is contrary to previous evidence using the genetic approach in European populations and the authors acknowledge that the results for heart disease are hard to interpret.”

Prof Kevin McConway, Emeritus Professor of Applied Statistics at The Open University, said:

“It’s not easy to do research on associations between drinking alcohol and health. You can’t carry out experiments where people are chosen at random to drink different amounts for many years, so generally you can just observe what people drink and record who gets ill from diseases like heart attacks and strokes. The trouble is that there will be differences between people that drink different amounts (or drink no alcohol), other than their drinking habits, for example in the amounts they smoke or their average prosperity. These differences could possibly be the cause of any differences in the risk of disease, rather than the differences in drinking. Some statistical adjustments can help in sorting things out, but they are never perfect, so it’s very hard to tell what causes what.

“This new study gets around this problem, in part at least, because of two characteristics of the Chinese population. First, there are genetic variants that affect the chance that a person will drink, by making the unpleasant after-effects even more unpleasant, and it happens that these variants are relatively common in China. Second, probably mainly for cultural reasons, drinking alcohol is very much rarer in Chinese women than in Chinese men. With the men in this study, the researchers found that those who had genes that made it likely that they drank less, did actually drink less, on average. But they also had lower average blood pressure and HDL cholesterol, and lower risk of strokes. That could have been because of some effect of these genes other than their effect on drinking amounts. But no such effects of the genes on blood pressure, cholesterol, or stroke risk were found in women. This provides evidence that the increased stroke risk in men who drink more alcohol is caused by the alcohol, and not something else.

“One thing I found particularly interesting in this study is that, using the same methods, it did not find evidence of increased risk of heart attacks due to alcohol, at least not for relatively moderate alcohol consumption. Generally, alcohol is considered to be a risk factor for heart attacks as well as strokes. But perhaps that relates to a general point about this study. It has certainly advanced what we know about the role of alcohol in some diseases, but it can’t be the last word. The new study doesn’t tie down exactly how alcohol works to increase stroke risk but (in this study) doesn’t appear to increase heart attack risk. The effect of alcohol on disease risk could, for example, be related to some lifestyle features, such as diet or exercise, that are (on average) the same in China for drinkers and non-drinkers but are different in other countries. I mean, perhaps something about Chinese lifestyles protects Chinese men from effects of alcohol on the heart attack risk, even if alcohol tends to raise their blood pressure, but perhaps that protection doesn’t work in places with a different lifestyle. This study can’t tell us about any of that because it looked only at Chinese people, although the people studied did come from a range of different parts of China which will differ in lifestyle choices and in other ways.”

‘Conventional and genetic evidence on alcohol and vascular disease aetiology: a prospective study of 500 000 men and women in China’ by Iona Y Millwoodet al. was published in The Lancet at 23.30 UK TIME on Thursday 4 April 2019.

Declared interests

Prof McConway: Trustee of the Science Media Centre.

Prof Spiegelhalter: nothing to declare.

Dr Burgess: nothing to declare

Prof Deborah Lawlor:Deborah Lawlor works in the same Unit with one of the middle authors of the paper (George Davey Smith). She wrote this quote independently of any of the authors (i.e. no discussion with any of them about what she has written)