Friday 8 January 2016

The Chief Medical Officer is misleading the public

The change to the alcohol guidelines today was an opportunity to bring official advice closer in line with the evidence. Instead, the Chief Medical Officer has taken it further away.

I have predicted for years that the guidelines will one day be reduced to zero. This is another step in that direction. In the 1960s, people were told not to drink more than a bottle of wine a day (or the equivalent). Then, in 1979, Britons were advised to drink no more than 56 units a week. This fell in 1984 to 14-21 units for women and 21-36 units for men. By the end of the 1980s it had become 14 units for women and 21 units for men. They have now fallen to 14 for both sexes - one pint of beer a day.

Most men will find their new 14 unit 'limit' laughable and rightly ignore it but guidelines are not really designed for the public. They are designed for 'public health' campaigners. The effect of today's change will be to drag hundreds of thousands of people into the at-risk category and revive the flagging narrative of Booze Britain.

It is also being done because the 'public health' lobby hates a complex message. The reality is that there is no safe drinking level. Being teetotal offers protection against a few cancers but drinking offers protection from other diseases, including heart disease and stroke. So what do you want to be 'safe' from - cancer or heart disease? You have to choose.

Sally Davies has got around this thorny question by ignoring or downplaying the benefits of moderate drinking to health. She relied on one piece of quack research to do so.

Last year, a study claimed that drinking only had a protective effect for certain groups. The authors used a simple statistical trick. They gathered data which clearly showed health benefits from moderate drinking and then divided it into so many subgroups that it was almost impossible for them to produce statistically significant results. By the time the authors had sliced and diced the data, the only people who appeared to benefit from drinking were post-menopausal women. I wrote about it at the time, as did David Spiegelhalter. It was absolute junk.

Everybody knew the guidelines were under review. I said at the time that the study was a concerted effort to overturn decades of evidence. Countless studies have found a J-shaped relationship between drinking and disease/mortality. This can be illustrated with this graph from Doll and Peto (1994)...

Richard Doll returned to the subject in 2005 and found much the same effect...

The J-curved relationship between alcohol consumption and coronary heart disease is particularly strong. It has been summarised in this meta-analysis of 84 studies and can be illustrated with this graph from Corrao et al. (2000)...

What about overall mortality? It's the same picture. As many researchers have shown, for example Gronbaek et al. (1994), the relationship is also J-shaped.

The only real question is how much do drinkers have to consume before they assume as much risk as non-drinkers? Di Castelnuovo et al. (2006) was quite typical in finding a lower rate of mortality up until drinkers consume 40 units of alcohol per week. Beyond this, risk increases above that of the teetotaller. A unit of alcohol is 8 grams so this works out at five standard drinks or more (nb. a unit is 10 grams in the US hence it is four drinks in the graph). 

From time to time, 'public health' people who don't like to admit that drinking can be good for you wheel out the 'sick quitter' hypothesis, a zombie argument that was debunked years ago. (Even the slice-and-dice study showed that the risks of not drinking were not due to unhealthy former drinkers - only lifelong non-drinkers were studied. Doll's 2005 study found the same thing).

The facts are clear. Moderate drinking - or by today's puritanical standards, even relatively heavy drinking - reduces not only heart disease risk but overall mortality risk. And the relationship is causal.

So how does Davies - or rather Davies' advisory group, which was made up of familiar anti-drinking zealots such as Ian Gilmore and Gerard Hastings, with the research shipped out to Sheffield University's minimum pricing modellers - get around this dilemma? She cites that one terribly flawed study from last year which split the drinkers into lots of non-significant subgroups. That's it. Actually she doesn't cite it explicitly - the report contains remarkably few citations to academic studies - but it is clear what she is referring to.

Meta-analyses have identified that for some conditions, notably ischaemic heart disease (IHD), drinking alcohol at low levels may have a protective effect (compared to not drinking), particularly for all-cause mortality. However, the group noted that any potential protective effect seems mainly relevant to older age groups

This claim is repeated again in the main text of the report...

Evidence for a net protective effect of alcohol from risk of death (which has been linked to possible reduced risks of heart disease late in life) is considered less strong than it was. A reduced risk still exists, but, in the UK, it now appears to matter overall in a significant way only for women aged 55 or older.

And again...

Previous analyses suggested the protective effect was only likely to be relevant to men from age 40 onwards and for post-menopausal women. 

She follows this up with some classic merchants of doubt stuff...

The evidence for a direct, protective, effect of alcohol on mortality is a subject of continuing scientific discussion

Whereas there is no scientific discussion about one contrarian study based on nothing more than statistical chicanery?! That study is sufficiently robust to overturn everything else in the literature, is it?

The reality is that the J-curve has been repeatedly shown in a body of evidence that is much larger, stronger and more consistent than the evidence for a relationship between moderate drinking and most of the cancers Davies focuses on and yet she portrays the former as controversial and the latter as rock solid. As Eric Crampton said some years ago, in the world of alcohol research 'every risk is conclusively proven; every benefit needs further study.'

She then seeks to say, in effect, that reducing heart disease risk isn't very important...

Deaths from this type of disease have been falling in the UK population for some years, which means there is less risk for which low alcohol consumption might give protection

Maybe heart disease declined partly as a result of the increase in drinking? Just a thought. In any case, coronary heart disease still kills more people than all the 'alcohol-related' cancers Davies mentions combined.

And then there is this remarkable statement...

Lifestyle changes, such as stopping smoking, increasing levels of physical activity, and eating a healthy diet, can help protect against heart disease, so any potential protective effects from alcohol could be achieved in other ways, which avoid the other health risks which come with any drinking of alcohol.

Imagine them saying this about anything else! Imagine them saying that people don't need to bother about eating too much salt because they can always reduce the risk of having a heart attack by losing weight. In any case, it's not true. A non-smoking teetotaller is as greater risk than a non-smoking moderate drinker.

This whataboutery is a blatant attempt to downplay the significance of alcohol's protective effect on heart disease to such an extent that they are even happy to downplay the significance of heart disease as a cause of death. The lengths these people will go to is extraordinary.

Anyone who cares about scientific probity should be outraged by the Chief Medical Officer's behaviour in this report. Casting out decades of research in favour of a single claim from a single study is a shocking abuse of her authority. People have a right to receive honest and accurate information from the government. I would call for her to be sacked but she'd only be replaced by somebody cut from the same corrupt cloth.

No comments: