Alcohol and dementia - more dishonest reporting from the BBC

A study was published in the British Medical Journal today showing that moderate to moderately heavy drinkers have a lower risk of dementia than teetotallers. As always happens when epidemiology finds benefits from drinking, there has been a rush to cast doubt on the findings by talking about 'sick quitters', confounding variables, recall bias etc. These are valid caveats to make, but they are mostly generic criticisms that apply to all observational epidemiology. And yet these points are only raised when a study shows benefits from alcohol consumption. They are almost never raised when risks are reported.

An example of this is Tom Chivers' article for CNN but if you want an absolute classic of the genre, read the BBC's report. The headline sets out the anonymous author's stall straight away...

Alcohol and dementia - is moderate drinking safe?

There is no question that moderate drinking is 'safe' for dementia. The question is whether being teetotal is 'safe'. This study provides yet more evidence that it is not, but by reframing the question, the BBC moves the Overton Window and implies that there are only two possibilities for moderate drinkers: greater risk or no greater risk.

There is no doubt that alcohol abuse is bad for the brain - but could there be health benefits for moderate drinkers?

How do we know that alcohol 'abuse' is bad for the brain? Because that's what the epidemiology shows. When it comes to harm, the BBC is satisfied that epidemiological studies can provide sufficient proof. But when it comes to benefits, it requires an impossibly heavy burden of proof.

The research is contradictory and so the answer isn't straightforward.

The research is actually pretty consistent. A meta-analysis of 15 studies published in 2009 concluded that light and moderate drinkers were 28 per cent less likely to develop dementia. A further meta-analysis of 20 studies published last year concluded that...

Modest alcohol consumption (≤12.5 g/day) is associated with a reduced risk of dementia with 6 g/day of alcohol conferring a lower risk than other levels while excessive drinking (≥38 g/day) may instead elevate the risk.

Does the BBC discuss these evidence reviews or any of the studies in them? Does it heck. So why does the BBC reckon the science is unsettled?

Some research suggests that drinking one or two units of alcohol a day - particularly red wine - could be of benefit to brain health, but other scientists are more sceptical.

The link is to a BBC story about a study which found no evidence that resveratrol - an ingredient in red wine - prevents heart disease. It doesn't mention dementia at all because the study wasn't about dementia.

Another study found that, even in moderation, drinking alcohol could increase the risk of dementia.

The link is to another BBC story, this time reporting the findings of an unpublished study that was presented at a conference in San Francisco in 2012. It is telling that the BBC chose to report this at all, but it did so at length and with the headline 'Drinking alcohol, even in moderation, "a dementia risk"'.

The only peer-reviewed study by the named researcher on this subject was published in 2014. It found that old women who reduced the amount they drank were at increased risk of dementia. You won't be surprised to hear that the BBC did not cover the study when it was published.

The BBC's claim that the research is contradictory starts and ends with those two stories. That's it.

Another study has come out to add to the confusing picture of public health advice around drinking.

It's not confusing. There is a J-Curve of risk that has been shown repeatedly for dementia and several other diseases. The advice should be to drink moderately.

The British Medical Journal study found that a group of people who did not drink alcohol in middle age were more likely to develop dementia later on than people who drank moderately. 

You can read it here. It used the famous Whitehall II cohort of civil servants, following them up over a period of 23 years. It found familiar J-Curves across the board, with those who were teetotal throughout the period being 47 per cent more likely to develop dementia that light/moderate drinkers.

The authors try to argue that this vindicates the new drinking guidelines (14 units) that were set in 2016 after a rigged process. This is a strange interpretation. Although risk is at its lowest at around 14 units per week, it does not follow that drinking more than this is unsafe. Common sense and convention dictate that drinking becomes 'unsafe' when the risk exceeds that of a teetotaller's, not when it exceeds that of the lowest risk drinker. As the graphs above show, you have to drink much more than 14 units before your risk reaches that of a teetotaller's.

There is a big difference between a safe amount and an optimal amount. If we are going to tell drinkers to consume 14 units a week because that is where the optimal health benefits lie, we should also be telling non-drinkers to consume 14 units a week. Neither the BMJ nor the BBC are going to do that, however.

So should non-drinkers take up the habit for the sake of their health? The answer is almost certainly no.

Why? The BBC won't explain its reasoning. Instead it makes the usual generic criticisms about epidemiology which are conspicuous by their absence when it is reporting that, for example, half a glass of wine increases breast cancer risk.

First of all, it can only really say that more of the people observed who didn't drink in midlife went on to develop dementia - it cannot say that abstaining from drinking itself is causing dementia. 

Yes, correlation doesn't prove causation. This applies to literally every epidemiological study ever published.

And people in this group may have drunk heavily in the past or had to give up drinking for health reasons. 

Hello sick quitter hypothesis, my old friend. I'm here to obfuscate with you again. The non-drinkers in this study were not all necessarily lifelong teetotallers, but they were certainly long term teetotallers of at least 23 years standing. It is possible that some of them may have been heavy drinkers when they were young. Is there evidence that heavy drinking in youth followed by decades of abstinence increases the risk of dementia? I don't know of any and the BBC doesn't cite any. It just throws out the possibility to cast doubt on the findings.

The study only looked at whether people drank during a particular snapshot in time, so some of that group might already be in poor health.

That sentence doesn't make any sense whatsoever. It takes two separate issues - 'sick quitters' and misreporting of consumption - and garbles them together in a way that suggests that the author doesn't have the first idea what he or she is talking about. In any case, the study didn't look at what people drank in one 'snapshot in time'. The civil servants were asked how much they drank on eight occasions over the course of the study.

Misreporting is an issue for all epidemiological studies that are based on questionnaires but the real issue in alcohol epidemiology is under-reporting of consumption. Since we know that people greatly under-report the amount they drink, it is highly likely that the people who have the lowest dementia risk are consuming around 20 units, not 14 units. The BBC doesn't mention this.

There have been a number of studies in this area with conflicting results...

As discussed above, this is misleading at best. A study was published earlier this year which was reported to show that light drinking increases dementia risk but, as David Spiegelhalter explained, it actually confirmed the J-Curve.

...and this one doesn't provide enough evidence to suggest that anyone should go ahead and change their drinking habits.

Somehow I suspect that the message would be different if moderate drinkers were 47 per cent more likely to get dementia than non-drinkers, but never mind. They don't call the BBC 'Auntie' for nothing.