When bad science appears in the media, who is to blame? Sometimes it is the fault of the journalists, sometimes it is the scientists and sometimes it is the person who wrote the press release.
There was a classic example last Friday when news outlets around the world covered a study in the Lancet with headlines such as ‘One drink a day “can shorten life”‘ (BBC) and ‘Just one alcoholic drink a day will shorten your life, study shows’ (Evening Standard). As if this were not scary enough, Yahoo warned that ‘Alcohol guidelines in many countries may not be safe’ and the Guardian declared that ‘Drinking is as harmful as smoking’.
At first glance, the study appeared to support some of these claims. Its authors concluded that their data ‘support adoption of lower limits of alcohol consumption than are recommended in most current guidelines’ and they provided graphs that appear to show few health benefits from moderate drinking.
Decades of epidemiology have shown that the relationship between alcohol consumption and mortality is J-shaped. At moderate rates of consumption, mortality risk falls below that of a teetotaller and then rises, exceeding that of a teetotaller at around 30-40 units per week. The exact shape of the curve varies depending on your criteria for inclusion but the basic conclusion is always the same: moderate drinkers live longer than teetotallers. Here are some examples from a meta-analysis of 34 studies.
In the new Lancet study, a J-curve is only apparent with cardiovascular disease and the protective effect is weaker than in previous studies. There seems to be no overall mortality benefit. This is a surprising result which, if true, could have justified some of the scary headlines.
But there is something unusual about the Lancet graphs. As the statistician Adam Jacobs pointed out on Friday morning, non-drinkers have been excluded from them. If you want to see how drinkers compare against teetotallers you have to turn to page 31 of the study’s appendix where you will find these graphs…
With the non-drinkers included, we can see that the new study does not contradict the existing evidence. The mortality risk of those who have never drunk alcohol is 20 per cent higher than that of those who drink 100g per week (12.5 units) and is the same as those who drink 300g per week (37.5 units). (NB. Forget about the ex-drinkers who have an even higher risk because some of those will be former heavy drinkers. It is the never-drinkers who are the relevant control group.)
The study’s authors got rid of the J-Curve by cutting off part of the J! Instead of using non-drinkers as their baseline, they used the most moderate of moderate drinkers and buried the findings for non-drinkers in the appendix. If you look at the data in the study and ignore the editorialising of its authors, the study doesn’t tell us anything we did not already know. Moderate drinking reduces mortality risk and is particularly good for the heart. Light drinkers have the best outcomes, but drinkers who consume double the 14 units recommended by the Chief Medical Officer do better than those who do not drink at all.
The Lancet study appeared to produce different findings from its predecessors because it was asking a different question. The conventional, common sense way of setting drinking guidelines is to identify the point on the J-Curve at which a drinker’s mortality risk rises to the same level as that of a teetotaller. If it is considered safe to abstain from alcohol, it makes sense to use non-drinkers as the baseline for acceptable risk. But the Lancet study is based on the unspoken assumption that the guidelines should be set at the point at which the health benefits of drinking are maximised, ie. at the lowest point of the J-Curve.
It’s an unusual assumption with little practical relevance. It may be true that 12.5 units of alcohol per week is the ideal quantity to consume if you want the best health outcomes, but an ideal quantity is different to a safe limit. If we are going to start setting guidelines based on what is ideal, 12.5 units should be not just the maximum recommendation but also the minimum. In other words, we should tell non-drinkers to start drinking and tell light drinkers to drink more. Needless to say, the ‘public health’ lobby has no intention of doing this.
Either way, reports in the media about people being killed by a single glass of wine were totally misleading and as the truth emerged on Friday morning, the BBC changed its headline. ‘One drink a day “can shorten life”‘ was changed to the blander but tautologically correct ‘Regular excess drinking can take years off your life, study finds’. It was too late for the newspapers which had, of course, already gone to press.
Who was to blame for this mess? There was talk of ‘ropey reporting’ and ‘misleading journalism’ on social media but I am inclined to give the Fourth Estate a break on this occasion. Nothing in the Lancet’s press release hinted at the truth. As has become routine under the editorship of Richard Horton, the journal focused on the supposed political relevance of the study. The press release was headlined ‘Alcohol limits in many countries should be lowered, evidence suggests’ and it included quotes like this:
“Doctors and other healthcare professionals must heed this message and transmit it to their patients. This study has shown that drinking alcohol at levels which were believed to be safe is actually linked with lower life expectancy and several adverse health outcomes,” Dr Dan G Blazer, co-author, Duke University, USA
Professor Edoardo Casiglia, co-author, University of Padua, Italy, adds: “This study provides clear evidence to support lowering the recommended limits of alcohol consumption in many countries around the world.”
The British Heart Foundation, which part-funded the study, was no better. Its press release was titled ‘Even moderate drinking linked to heart and circulatory diseases’ and said that ‘the authors say their findings challenge the widely held belief that moderate drinking is beneficial to cardiovascular health’. They quoted their ‘senior dietician [sic]’ who claimed that the study ‘does seem to broadly reinforce government guidelines for the UK’ and said:
“We should always remember that alcohol guidelines should act as a limit, not a target, and try to drink well below this threshold.”
But by focusing on the ideal level of consumption, rather than the ‘safe’ level, the Lancet study did effectively create a ‘target’ rather than a ‘limit’. It portrays five or six drinks a week as optimal. Zero drinks a week is not. The senior dietitian is trying to have it both ways.
The British Heart Foundation came close to claiming that there were no health benefits from moderate drinking at all, and Tim Chico of Sheffield University finished the job by saying:
“This study makes clear that on balance there are no health benefits from drinking alcohol, which is usually the case when things sound too good to be true.”
The way this story was reported was lamentable, but most of the blame lies with the authors of the study and the accompanying press releases. None of them explained that the control group was made up of moderate drinkers rather than total abstainers. Nor did the abstract of the study, although the authors did find room in it to make another appeal to governments around the world to lower drinking guidelines. The quotes given to the press were largely misleading and were more focused on defending the controversial lowering of the UK guidelines than explaining the study’s findings.
A close reading of the study and appendix would have led journalists to the truth, but the Lancet’s press officers must know that health correspondents rely on abstracts and press releases. Journalists have seen many studies of this kind over the years. They should have been informed that this one used a very different methodology.
By the time the truth had got its shoes on and the corrections were issued, the lie had spread around the world. I’m not convinced that this was an accident. It has all the hallmarks of fake news. No matter how easy it is to debunk a bald lie, it will always spread faster, and live longer, than a subtle truth. Last week’s news reports will be remembered by those who want to believe that the benefits of moderate drinking are imaginary. The authors’ claim that drinking guidelines are still too low will be cited by campaigners in Britain for years to come. I doubt that anyone associated with this study was upset when they saw the headlines on Friday morning.