Deaths from alcohol foretold

Six executives from leading drink companies have written to The Times today protesting about a story on alcohol-related deaths. The common reaction will probably be “They would, wouldn’t they?” but actually their argument deserves serious attention.
I have been puzzling for the past week about the same story, which was based on a Comment article in The Lancet (February 23, p1) . So puzzled was I that I e-mailed two of the authors last week seeking clarification. 
There is a general presumption that drinking in the UK is out of control. Fed by media stories and health warnings from doctors such as the article in The Lancet, this idea is now firmly rooted.
An opinion poll commissioned by the Home Office as part of its alcohol pricing initiative found that 65 per cent of people agreed with the statement “The amount people drink in this country is out of control”. Among non-drinkers, 78 per cent agreed.
When asked whose drinking was out of control, 71 per cent of respondents agreed that it applied to only a minority of problem drinkers. Just under half (49 per cent) agreed that the media exaggerated the problem, while 42 per cent disagreed. 
The sense of a juggernaut careering down the road with nobody at the wheel is reinforced by warnings such as the one delivered by The Lancet. Headed “Projections of alcohol deaths – a wake-up call”, it was written by Nick Sheron, Chris Hawkey and Ian Gilmore, all well-known in the field. Two of them, Dr Nick Sheron, a liver specialist from the University of Southampton and Professor Sir Ian Gilmore, are members of the Department of Health’s Responsibility Deal Alcohol Network.
They said: “Few can doubt that there is a particular problem in the UK” and pointed to rapidly rising levels of deaths from liver disease. Since 1986 they say, they have more than doubled, from 4.9 to 11.4 deaths per 100,000. Over much the same period the death rate in France from liver disease has halved, from 21.76 to 10.36 (a 2006 figure) which was less than the UK’s equivalent figure for that year.
More recent trends are, however, more equivocal (see chart below). Enthusiasts for European integration might well see the trends since 2001 as a good example of convergence. UK death rates from liver disease have changed relatively little since 2001, and are close to the average for the pre-2004 EU countries. Ireland is lower but gaining: Germany higher but declining. The source for this data is the WHO European health for all database, the same source used by the Lancet authors.
Among those not charted, Norway and Sweden have much lower rates, Denmark and Finland significantly higher.
In The Lancet the three experts argue that if the UK could improve its rates as quickly as France did in the past, it could save 22,000 lives by 2019. If it could improve as rapidly as the average of the EU as a whole, 8,900 lives would be saved over the same period. (They must mean the “old” EU, though they don’t say so, as liver deaths have actually increased if you include the more recent entrants.)
While true – indeed a truism – this isn’t terribly realistic. France’s big decline was from a very high level, and as my chart shows, it has now plateaued at much the same levels as the UK. It is beyond argument that lower rates of liver disease would result in fewer deaths, and that alcohol is an important contributor to liver disease.
Where I take issue with the Lancet article is in the extrapolations the authors use to reach the conclusion that “between 160,000 and 250,000” lives could be lost or saved over a 20-year period.
This depends on two extrapolations: one suggesting that it is possible for the UK to reduce deaths at the same rate as the historic French decline, which took it from around 35 deaths per 100,000 people to the present level of just over 10, and another suggesting that UK deaths are actually on a rising trend which will lead to an additional 8,900 deaths by 2019.
The Lancet's chart (below) shows these extrapolations, with the green projection being the reduction at the same historic rate as France and the black projection the way they say the UK is heading.
As my chart shows, recent death rates in the UK do not justify the second of these projections. The UK figure for 2009 (10.82) is lower than that for any year since 2002 (10.44). The Lancet team says this projection is based on “a similar rate to the UK as a whole over the past ten years” but it is not.
When the Lancet originally written (Nick Heron tells me) the 2009 data, which showed a fall in liver deaths, were not available. He insists the extrapolation, based on 1998-2008 data, is correct. I continue to believe that the trend since 2001 does not justify it.
So the headline figures of lives potentially lost or saved are based on two projections, one unrealistic and the other (I submit) wrong. The conclusion that the UK could save 77,000 liver deaths over 20 years is, therefore, also wrong, and so is the 160,000 – 250,000 figure which comes from multiplying this up by “considering the wider harms of alcohol” and which the authors call a conservative estimate.
They do not count any of the benefits of alcohol, coincidentally the subject of a meta-analysis in the latest BMJ which pooled data from 84 previous studies. This showed that the risk of dying from all causes was 13 per cent lower in drinkers than in non-drinkers. The effect is largely due to a 25 per cent reduction in heart deaths among moderate drinkers. 
Shouldn’t any reasonable projection of alcohol deaths include this? It can be argued that causation has not been clearly established and there are possible confounding factors, but an abundance of studies have now shown that moderate drinking is protective against heart disease, a much bigger killer than liver disease.
In 2008, the North West Public Health Observatory estimated that 8,838 deaths a year might have been saved in 2005 as a result of drinking – more than were lost as a result of liver disease. This estimate appears to have alarmed them, and they backtracked, lowering the estimate to 3,813 by saying that alcohol has not been shown to be protective in the over-75s, where most deaths occur. The report also suggests that moderate drinking prevents about 1,400 ischaemic strokes a year in men, while heavy drinking contributes to around 250 – a net benefit of over 1,100.  Whatever the figures are, The Lancet authors might have mentioned them.
Instead, they repeat the claim that three quarters of the alcohol sold in the UK is consumed by hazardous and harmful drinkers, attributing it to Safe, Sensible, Social – Consultation on further action, a 2008 consultation report from DH. This, in turn, attributes the claim to the NWPHO report, where I cannot find it. That report finds, to the contrary, that alcohol-attributable deaths are, for the most part, evenly distributed among light, moderate and heavy drinkers.
In women, the paradoxical result is that the highest number of alcohol-related deaths occur in the lowest consumption category. This is because there are far more women who drink lightly than there are who drink heavily. There are more deaths from cancer of the oesophagus attributed to alcohol (156) in women in the lowest alcohol category than there are in the highest (23), even though this cancer has a well-established link to alcohol.
In men, this is less marked – 278 deaths from oesophageal cancer attributed to alcohol in the lowest category of drinking (1-19 g/day) compared to 273 in the highest (75+ g/day). Cirrhosis of the liver is one of relatively few conditions that does not follow this pattern in men, with 112 deaths in the lowest category and 277 in the highest. 
This report says alcohol was responsible for 459,842 hospital admissions in 2005, which is a lot. But it is far fewer than the 1.05 million admissions attributed to alcohol by Dr Foster last week (Health Service Journal, 24 February 2011, p9). In this field, figures vary widely, and not all of them can be right.
Finally, to return to the question of whether drinking is out of control. Recent statistics do not suggest that it is, as I have discussed before. The letter in today’s Times says that average alcohol consumption has fallen by 11 per cent since 2004. The writers do not give a source, but ONS provides these figures for the past decade. Figures for consumption outside the home do not exist before 2001-2.
These figures imply a greater fall in consumption than the letter claims, but whatever the precise figure, do not suggest that alcohol is out of control.
The Lancet comment gained widespread coverage: The Times, The Guardian, The Independent, The Daily Mail, the Daily Express, Metro, even the International Herald Tribune. The 250,000 figure featured in almost all the stories, yet as I have shown, there must be serious doubts about its derivation.
Given the reluctance of so many journalists to dig beneath the surface, it’s hardly surprising that so many people think drinking is out of control, and the only escape is to force us all to pay more for it. There’s a good case for sensible drinking, but it isn’t improved by exaggeration.