Monday 19 November 2018

Teetotallers still dying too young

In April, the Lancet published one of those 'no safe level' studies about alcohol. It was garbage but it got the kind of media coverage its authors doubtless anticipated, with the Guardian declaring that ‘Drinking is as harmful as smoking’ and the BBC asserting that ‘One drink a day “can shorten life”‘. Interestingly, the BBC headline has since been changed to the more credible 'Regular excess drinking can take years off your life' but the damage has been done.

It's all part of the drip-drip campaign to erase the health benefits of alcohol consumption. Lies like this travel the world while the truth is getting its shoes on. Last week, I mentioned the article by Alexandra Freeman and David Spiegelhalter criticising a similar study that was published in August. It did not receive the same level of interest from the media as the original, to put it mildly.

Nor did the letters in Friday's Lancet criticising the April study so I will flag them up here. The trick employed in that study was fiendishly simple. The authors simply removed non-drinkers from the analysis, leaving graphs that imply no health benefits from moderate drinking...

If you put the never-drinkers and ex-drinkers in, the usual J-Curve emerges. Drinkers have to consume quite a lot of booze before their health risks exceed those of teetotallers.

I don't know why it has taken seven months for a rebuttal to this deeply misleading study be published in the Lancet, but two have now come along at once. As the first letter says:

Taking out the non-drinkers as a reference group is the only novelty in this study compared with existing scientific literature, and causes the complete elimination of the left rising arm of the J curve (as shown in the figure and in the Article's appendix, p 31). This is also not the first time a study has suggested that the mortality curve bends at a drink per day or less. By removing the non-drinkers, Wood and colleagues make it difficult to establish whether any amount of moderate alcohol consumption has a different effect to abstaining. Thus, we believe that the study has little to add to existing scientific literature and cannot contribute to public health advice.

The second letter makes a similar point...

Over the past decades, observational studies have taught us a lot about the relation between alcohol consumption and health. The effects of alcohol on health are now understood to differ depending on the amount of alcohol consumed, the age and gender of the consumer,1 and the various endpoints studied. Most studies find that individuals who abstain entirely from alcohol consumption might be worse off than moderate drinkers, and virtually none have reported a beneficial effect of abstention on mortality compared with moderate consumption.

.. By disregarding ex-drinkers and never-drinkers and the differing effects patterns of alcohol consumption had on mortality, the main message emerged that less alcohol consumption was better. A more informative message based on the findings hidden in the appendix could have been that the study added further evidence to support a beneficial effect of moderate alcohol consumption (especially wine) on mortality. The headlines might have then resembled those of just a year ago: “A glass of wine or pint of beer a day can help people to live longer, according to new research”.

The authors have replied to these criticisms but are unable to credibly defend the decision to exclude non-drinkers - who are, after all, the only relevant control group if you're going to encourage total abstinence. They don't deny that non-drinkers have higher rates of mortality than moderate drinkers, nor do they deny controlling for every conceivable confounding factors. Instead they make vague assertions about non-drinkers possibly being unhealthy for reasons nobody can think of.

...never-drinkers might differ systematically from drinkers in ways that are difficult to measure, but might be relevant to disease causation.

Well, they might. Just as heavy drinkers might be more likely to get liver cirrhosis for reasons that have nothing do with drinking, but since every alternative explanation has been studied and found wanting, we would reject this notion as desperate straw-clutching. Same rules apply.

NB. I wrote about this study at the time for Spectator Health.

No comments: