Monday 17 June 2019

Critiques of the Lancet's 'no safe level' study

A study appeared in the Lancet last August which claimed to have virtually erased the J-Curve from alcohol epidemiology. The authors used an unconventional methodology, modelling aggregate data from whole countries rather than looking at risk to individuals, and came up with this:

According to their model, there is no reduction in mortality risk from moderate levels of alcohol consumption. It did not quite show that there is 'no safe level', despite the authors claiming that '[o]ur results show that the safest level of drinking is none', but it came close. It has since been cited by the Lancet and others as conclusive evidence that no amount of drinking is safe and that alcohol should be treated like cigarettes.

The methodology was too opaque to allow a full immediate critique, but David Spiegelhalter discussed it here and I wrote about it at the time. Among its flaws was the inclusion of tuberculosis as an alcohol-related disease (debatable in itself) which lifted the risks of drinking even in countries where TB is virtually nonexistent.

The Lancet has now published criticisms from three groups of scientists. You can read them here, here and here, along with the authors' reply here. The authors defend themselves by saying that their conclusion is broadly supported by two reviews by Tim Stockwell and another study published in The Lancet last year. In fact, the latter study found clear evidence of a J-Curve and Stockwell's one man crusade against the J-Curve is based on extreme cherry-picking and 'questionable statistical methods'. In the end, the authors say:

Debates concerning whether the safest level of consumption is zero or near zero are missing the point.

You see what they did there? That's what you call moving the Overton window. It's not actually a question of zero or near zero. It's a question of moderate consumption versus not so moderate consumption; 20-odd units a week versus 30-odd units a week. Either way, more than the UK government's evidence-free 'safe drinking' guideline of 14 units.

If you're interested in this issue, hit the links above.

No comments: