Reacción a "Spanish study claims that drinking up to 35 glasses of wine a month is associated with lower cardiovascular risk in some people"

The question of whether drinking small or moderate amounts of wine is associated with lower risk of cardiovascular disease (CVD, that is, heart attacks, strokes, etc.) has been widely discussed amongst experts and elsewhere. Epidemiologists still differ on this question (and that’s discussed both in the new research paper and in the accompanying editorial). This new research adds to the evidence. It’s a pretty competent study – I need to say that, because I’m going to be quite critical about some aspects of it. But it doesn’t get near a definitive answer to whether drinking a little wine is good for one’s heart and circulation. 

To see why not, I’ll mention what the difficulties are. Generally the best way to investigate whether exposure to something (like drinking wine) affects health outcomes (such as heart disease risk) is to do a randomised trial, but there are many ethical and practical reasons why that’s very difficult or impossible if the exposure is alcohol consumption. 

So instead most of the existing evidence comes from observational studies. In these, people drink what they would have drunk anyway. Generally, the researchers measure the participants’ wine consumption, and then follow them up for years and record who gets CVD. 

This kind of research suffers from the same kind of issues that affect all observational studies. People who drink different amounts of wine will differ in many other ways too, and one or more of these other differences – so-called potential confounders – could be the real cause of any correlations between wine consumption and CVD that are found. Then it isn’t the wine consumption that’s the cause of differences in CVD risk – it’s something else, in part or entirely. 

But with consumption of wine, or indeed any alcoholic drink, there are other issues too. How do the researchers measure how much wine the participants drink? Typically, they ask the participants in various ways, perhaps with a questionnaire that asks how much of what kinds of food and drink they consume. There’s evidence from other studies, and indeed from this new study as well, that people don’t always report accurately how much alcohol they drink. Maybe they don’t remember, or maybe they just don’t want to have to say that they drink more than they feel is acceptable. 

The principal new feature in this new study is that they used a different method of estimating wine consumption, as well as basing some of their results on the usual kind of questionnaires. The new method was to measure the level of tartaric acid in the participants’ urine. Grapes, and hence wine, contain quite large amounts of tartaric acid, and it seems there’s little of it in other food and drink. Measuring the tartaric acid content of someone’s urine gives a biomarker for alcohol consumption, and the tartaric acid levels are used in this new study to produce a measure of the association between wine consumption and CVD risk. 

I should point out that the new study was indeed observational, even though the participants were all in a randomised trial (called PREDIMED) that aimed to find out whether advice to follow one of two kinds of Mediterranean diet had benefits for health. Participants from this trial were used because the researchers already had good data on them and had followed them up already – but the data they provided for this new study is observational. The data came from food (and drink) frequency questionnaires, from measurements of the tartaric acid content in urine, and from records of who became ill during follow-up. But people weren’t randomised to consume different amounts of wine – they drank what they would drink anyway – to the study is observational. 

“The researchers made statistical adjustments for several potential confounders in their analysis. That’s standard in observational studies of this kind. In their research paper, they do acknowledge that there remains possible confounding – that is, their findings on the association between tartaric acid levels and CVD diagnoses might be caused, in part or completely, by something else than the wine-drinking that was measured by the tartaric acid levels. I'm not a physiologist so can't comment on whether they adjusted for everything relevant. But you can never be totally confident that you adjusted for every potential confounder, so you still can't be sure that the differences in CVD risk are caused by the different amounts of wine consumption as measured by the tartaric acid biomarker. They might be, but it remains possible that they aren’t. 

So, if indeed tartaric acid levels in the urine are a good biomarker for people’s wine consumption, what the researchers found is that people who drank between 3 and 35 glasses of wine per month had a lower risk of CVD, during the follow-up period of about five years on average, than people who drank no wine (or less than one glass a month). People who drank less than 3 or more than 35 glasses a month didn’t have a reduced CVD risk. 

But the pattern isn’t entirely clear-cut – even if the association is one of cause and effect, the results seem to indicate that CVD risk first decreases as one drinks small amounts rather than none at all, and after a point it starts to increase again, in a kind of U-shaped pattern. 

I think it’s worth noting that the amounts of wine involved, for monthly consumptions where the CVD risk is below the risk for people who drink nothing or next to nothing, are pretty small, in terms of what some people consider to be moderate drinking. 

Above 35 glasses a month, there’s no evidence of a lowered CVD risk. Now 35 glasses a month is less than one and a quarter glasses a day. The research paper doesn’t say how big a ‘glass’ is, but I looked at the questionnaire for this study and it indicates that a glass is 100ml. So 35 glasses a month is below 125ml a day (so less than one sixth of a standard wine bottle). If the association really is one of cause and effect, drinking more than one fairly small glass of wine a day takes you out of the area of reduced CVD risk. To be in the part of the U-shaped curve that appears to have the largest risk reduction, compared to not drinking, maybe one would have to drink, very roughly, about 3 such small glasses a week. 

There is a further issue about how good a biomarker tartaric acid in urine is for wine consumption. A previous study by a research team that overlaps with the researchers for the new study, found a very high measure of correlation (0.922) between how much wine twenty-one men had consumed and their urinary tartaric acid levels the next morning. But the amount of wine consumption in that study was randomised and it was generally much more tightly controlled than the new one. 

The new paper does report that the correlation between the wine consumption that the participants reported, and their tartaric acid levels, was statistically significant. But that simply tells us that there's good evidence that the correlation wasn't zero, not whether the correlation was actually high. They report a correlation measure of 0.46. That's way smaller than in the randomised study I mentioned above. One interpretation is that the square of 0.46, that is, about 0.21, or 21%, is the proportion of the variability in the tartaric acid level explained by the participants' reported alcohol consumption. That's certainly not tiny but nor is it huge. 

So a lot of the variability in tartaric acid could arise from other things, such as the respondents not all being honest about how much they drank, or by there being a certain length of time between the self-reported alcohol consumption and the tartaric acid measurement, or because the tartaric acid levels are also affected by consumption of other foods or some processes in the body (which you'd have to ask a physiologist about). So this is yet another reason we can't be at all sure that drinking certain amounts of alcohol causes people to have CVD risks that look like those observed in the new study. 

One further point. The participants in this study all lived in Spain. They had agreed to be in a randomised trial of advice about Mediterranean diets. That already makes them untypical of people in the UK. Also, they were not young when the trial began (55-75 for men and 60-80 for women), and all were at relatively high risk of CVD. (They either had type 2 diabetes or three or more known CVD risk factors from a list.) This makes it even harder to know how the results might apply to people in different countries and cultures, who are not all necessarily at high CVD risk. 

“It would perhaps be all be clearer if we knew clearly what was going on inside the body to cause small amounts of wine to reduce CVD risk, if indeed it does. This study does give some suggestions (based on other research) as to how that might happen, but it provides no new information on the matter itself. Together with all the other limitations I’ve mentioned, it’s clear that this new study isn’t going to do away with the controversy about whether consumption of small amounts of wine could reduce your CVD risk.