Monday 31 January 2022

A desperate attempt to deny the benefits of moderate drinking

There is mounting desperation in 'public health' to debunk the benefits of moderate drinking - or, at least, to make the public think the benefits have been debunked. Various academics are offering their services. It's as if there is a reward on offer to whoever can make the best case against the J-Curve. In a sense there is, because whoever manages it will be set for life in the neo-prohibitionist 'public health' industry.

In recent years, we have seen the J-Curve challenged by ridiculous modelling, lying by omission and the inappropriate use of Mendelian Randomisation. None of this casts serious doubt on the existence of health benefits of moderate alcohol consumption, although the drip-drip effect no doubt leaves the public confused about what the science shows (which is presumably the intention). 

Two weeks ago, the gloriously named Beatriz Champagne from the World Heart Federation falsely claimed that any amount of alcohol harms the heart. She said that any claims to the contrary are "at best misinformed and at worst an attempt by the alcohol industry to mislead the public about the danger of their product." I didn't write about this at the time because it was just an assertion with no evidence to back it up.

But then on Friday, a new study was published which claimed to disprove the J-Curve. The study was grossly misrepresented by its own author and was (unsurprisingly) misunderstood by the media.

The study should have set off warning bells to any decent health journalist. For a start, it was accompanied by some hyperbolic, media-baiting quotes from the lead author, who is a physiologist at Anglia Ruskin University...

“The so-called J-shaped curve of the cardiovascular disease-alcohol consumption relationship suggesting health benefit from low to moderate alcohol consumption is the biggest myth since we were told smoking was good for us."

The title of the study reads more like a headline than a piece of academic research...

Alcohol - The myth of cardiovascular protection

And the study ends with a dog whistle call-to-action for the government that is not justified by the findings...

Our results do not support the current alcohol consumption guidelines for the United Kingdom of up to 14 units per week in relation to CV risk in the general population.

As informed readers know, there is 50 years of evidence showing that moderate drinkers live longer than teetotallers, an effect that is mostly due to their lower heart disease risk. The authors of the new study used UK Biobank data to monitor the health outcomes of drinkers and people who have never drunk, but things start badly for them when they find that teetotallers have higher rates of heart disease and have more heart attacks than drinkers. 

Never drinkers were older (P < 0.0001), had a higher body mass index (P < 0.0001), higher systolic and diastolic blood pressure (both P < 0.0001), were less physically active (P < 0.0001), had a higher prevalence of diabetes (P < 0.0001) and higher incidence of overall CV events (P < 0.0001), ischaemic heart disease (P < 0.0001) and cerebrovascular disease events (P < 0.0001). The incidence of overall CV events, ischaemic heart disease and cerebrovascular disease were comparable between never drinkers and drinkers consuming more than 14 units per week.

This suggests, a priori, that teetotalism is not particularly healthy, although it could be the result of confounding variables. However, when the results are adjusted for confounders, the picture remains the same. (Note that former drinkers are already excluded to rule out a “sick quitter” effect).

Compared to current drinkers in adjusted models, never drinkers were at higher risk for overall CV events (hazard ratio, 1.31; 95% confidence interval, 1.20-1.42; P < 0.0001), ischaemic heart disease (hazard ratio, 1.51; 95% confidence interval, 1.29-1.76; P < 0.0001) and cerebrovascular disease (hazard ratio, 1.46; 95% confidence interval, 1.14-1.87; P < 0.0001). Consequently, using never drinkers as reference resulted in alcohol from all drink types combined exhibiting protection in relation to all outcome measures...

This is rather awkward for the authors. To deal with this inconvenient finding, they have a simple solution. They exclude ALL the non-drinkers from their analysis and compare light, moderate and heavy drinkers instead! They split the drinkers into five quintiles based on their alcohol consumption and show the results in the figures below. The pink data points are what you get when you compare them to non-drinkers. The blue data points show how the drinkers compare to each other. (Click to enlarge.)

As you can see, the drinkers have a lower risk of ischaemic heart disease at every level of consumption. For wine, the risk drops as people drink more, but for beer, cider and spirits (which are grouped together for some reason) the benefits decline somewhat as people drink more (although none of the changes are statistically significant). 

It's a similar story when you look at the relationship with all types of cardiovascular disease, although the authors exclude ischaemic heart disease from one of these plots (red) for reasons that are never properly explained. Here we see a reduction in risk for all wine drinkers and for moderate drinkers of other alcoholic drinks.

In short, the authors are making some pretty extreme claims about moderate drinking without looking at the only reference group that matters (non-drinkers - or never-drinkers). They effectively use light drinkers as the control. And even then, they barely achieve their desired result. 

We’ve seen this trick performed before in a study from 2018 which made similar claims by comparing light drinkers to moderate and heavy drinkers - and generated similar headlines. The graph used in the main body of that study appeared to show no evidence of a J-Curve...


This was sleight of hand designed to catch readers out at a glance. The authors simply ignored the non-drinkers and used light drinkers as their reference point. But if you digged deep into the appendix, there were graphs that showed that the J-Curve was alive and well. Never-drinkers had a 30 per cent increased risk of heart disease and a 20 per cent increased risk of premature death.

Everything about Friday's study seemed designed to mislead. The press release for it said... 
New study: 14 units a week still harmful to health

Compared to what? Not compared to zero units, that's for sure. It's difficult to find any interpretation of that headline that isn't trivial or wrong.

I see no academic merit in the methodology used in this study. It certainly can't be used to say anything about the relative risks of drinking versus not drinking because it deliberately excludes the non-drinkers. 

It should never have been published.

No comments: