Drinking
There’s been a spate of stories recently about “middle class” drinking, which stem from a report produced by Liverpool John Moores University. The BBC’s coverage is reasonably typical.
I’ve not been able to find the report itself. It appears from the way it's been reported that it considers the geographical distribution of two classes of drinkers: (i) “hazardous drinkers”, being men who drink 22-50 units a week and women who drink between 15-35 units a week and (ii) “harmful drinkers”, being men who drink more than 50 units a week and women who drink more than 35 units a week.
The conclusion appears to be that whereas harmful drinking is associated with poorer areas, hazardous drinking is associated with more prosperous areas, although I’ve not seen anything like an analysis of the correlation between income and drinking habits. The list of top ten hazardous drinking hotspots falls a long way short of being Where’s Where of middle-class prosperity. It seems likely that the class based aspects of this story have been spun out of all proportion to the original report, but in any case, that isn’t what I want to focus on.
What I want to focus on is the tendentious nature of those labels. What does “hazardous” mean, any way? How does hazardous differ from harmful? The obvious way of reading the labels is that people who drink at a “hazardous” level are at risk, whereas those who drink at a “harmful” level are definitely harming themselves.
Dr Gray is quoted as saying, “There are increased risks of illness in people who are drinking in the hazardous category… The biggest risk to your health is chronic liver disease, followed by heart disease.” I have added the ellipsis where the BBC have put a paragraph break, because I am moderately confident that Dr Gray’s comments have been edited and that there were some other words between those sentences. The reason I am confident of this, apart from the fact that that’s what journalists do, is that it is simply not true that drinking in the “hazardous” category gives rise to a measurable risk of chronic liver disease (more on this below).
One good aspect of the story is that, for once, the alcohol content of wine has been correctly reported. As it says at the bottom of the linked-page, a large glass of wine contains about three units of alcohol. In my (fairly wide) experience, a typical wine glass is 175ml, and typical wine is 12% abv, so a typical glass will contain a little over two units, whereas one still often sees the advice that a glass of wine is about one unit.
They then rather spoil the effect by indicating that drinking a large glass of wine a day falls within the hazardous range, when in fact it does not. 3 units a day is in line with government health guidelines (for a man). I saw the same thing reported on the television news as well: that 7 large glasses a week for a man or 5 large glasses for a woman would constitute hazardous drinking. This is inconsistent (it’s also wildly misleading for reasons I will set out below). It may be that they’re thinking that at least some of those glass will be of wine over 12% abv, which will push the 7 glasses over 21 units. If so, this does rather indicate the folly of the approach, because it blatantly is not the case that there is a significant difference between 21 and 22 units a week.
Marcel Berlins wrote about this in yesterday’s Guardian. It appears that on the Today Programme, Prof Bellis said that two bottles of wine a week constituted hazardous drinking for men. Two bottles would be 18 units, maybe 20 units if the wine is high in alcohol. There are a very few bottles of wine which are 15% abv, which would reach 22 units, but by and large Prof Bellis’ comment (assuming he did say that) is not consistent with his own categories. Berlins then equates this to two “ordinary sized” glasses a day, which means his glasses are a lot smaller than mine. (That may well be true; glasses have got bigger and I’m younger than Berlins, even though I am now very old.)
These details, while giving an impression of the general confusion on this subject, are not of substantial importance. Whether the measure is two ordinary sized glass or one large glass, Berlins’ point stands that it is just not credible that drinking these sorts of amounts poses a risk to human health. Like him, I have seen not a single example (either from my own experience or in the media) of a person who has suffered ill effects from drinking these sorts of amounts and all the medical discussions I have seen of alcohol related illness relate to much larger amounts.
My suspicion, and it does seem to be born out by Bellis’ comment that drinking at moderate levels could progress into drinking at heavy levels, is that the public health concern is not really about people drinking 22-50 units a week at all. The point of the government’s alcohol strategy is to reduce drinking generally in order to create a culture less likely to give rise to problem drinkers.
That may or may not be fair enough (as a libertarian, I rather think not), but the important point is it won’t work. The reason is that most people who drink moderately, like myself and Berlins, will react to the guidance with incredulity. The effect of telling middle class people that they should not drink a large glass of wine a day will not be that middle class drinkers will cut back. The effect will be that they conclude (correctly) that the Government is scaremongering. The Government’s health advice is discredited as a result, which is potentially very dangerous.
I was sufficiently interested by this that I looked on the Department of Health web-site in an effort to find out what the Government’s policy actually is. I discovered that a report on the “next steps” in the National Alcohol Strategy was published recently (5th June 2007). It’s available for download here.
The report sets out the Government’s advice on drinking levels. No more than 3-4 units a day for men, 2-3 women. As ever when I see that advice, I puzzle as to how a safe maximum can be expressed as a range. Pregnant women should avoid drinking alcohol, but if they do drink they should drink no more than 1-2 units of alcohol once or twice a week. That advice seems to be a very mixed message: if I’m a pregnant woman, can I drink lightly or not? The absence of medical basis for the “don’t drink” message has been widely discussed elsewhere, so I won’t go into that here.
“Harmful” drinking is defined as “drinking at levels that lead to significant harm to physical and mental health and at levels that may be causing substantial harm to others.” The section goes on to say regularly drinking over 6 units a day (or over 35 units a week) for women and over 8 units a day (or 50 units a week) for men puts place people “at highest risk of such alcohol related harm”. The weekly numbers correspond to those used in the John Moores report for harmful drinking, but if I’m reading the guidance correctly, “harmful drinking” is defined (as one would expect) by reference to consequences and those numbers are merely indicative of those likely to be at risk.
The notion of risk in public health is one which requires some care. I recently saw something suggesting that people who didn’t get sufficient exercise were “at risk” of obesity. That’s obviously nonsense from the individual’s point of view. If I have a BMI of 20 and don’t exercise, then I’m not at risk of becoming obese. One needs to distinguish between a statistical treatment of populations, where one can talk about such risk factors and a causal treatment at the level of the individual, where one cannot. It’s clearly true that if more people followed the Government’s recommendations on exercise, there would be less obesity. In the case of a given individual, they are either obese or they are not. There’s no random element involved.
Similarly, I’ve no doubt that people who are harmful drinkers by the above definition tend to fall in the over 50 units a week category (if men, and as the report lately reveals, they are mostly men). However, at the level of the individual, it’s all much more complicated. It depends on patterns of drinking as well as overall volume, and it depends very much on the individual’s physical and psychological make-up.
I can see that, from the point of view of collecting and analysing statistics, it is helpful to replace a qualitative definition with a quantitative one, but at the individual level we have to revert to the qualitative definition, or we end up with the absurdity of saying that a man who drinks 49 units a week is not a harmful drinker, whereas one who drinks 50 units a week is.
Under “binge drinking” the report mentions that it is usually analysed by reference to exceeding 8 units a day for men or 6 units a day for women. It remarks, “In practice, many binge drinkers are drinking substantially more than this level, or drink this amount rapidly, which leads to the harm linked to drunkenness.” This amounts to admitting that a woman who drinks a large glass of wine at lunch and another at dinner is not binge drinking, either in the colloquial sense of the phrase or as defined in the report. Again what we have is a statistical convenience (because to conduct a survey one needs to have a quantitative way of classifying something a binge) which has some how been converted (largely I think by the media) into a classification.
There is no category of “hazardous” drinking at all. This seems to have been introduced ad hoc in the John Moores report to include everyone who drinks more than the recommended maximums but less than the “harmful” levels (bearing in mind that those levels themselves are essentially an ad hoc statistical convenience). But this introduces an element of confusion: the Government has never claimed that drinking above the recommended levels is necessarily hazardous. The levels are no more than a public policy guideline.
One fascinating snippet on p14: the HMRC figure data on “duty paid” clearances (which will certainly be an underestimate of alcohol consumed) show that the average adult consumes 11.3 litres of alcohol (or 1,130 units) a year, amounting to an average of just over three units a day. Bearing in mind the number of adults who don’t drink, it is clear that the average drinker is exceeding the Government guideline. ONS data, on the other hand, based on self-reporting, suggests the average adult drinks 10.8 units a week, around half of the HMRC figure. From this we can conclude what I have always suspected: that many people seriously underestimate their alcohol consumption. It follows that figures suggesting around 35% of men and 20% of women exceed the Government limits are likely to be a significant underestimate.
The report refers to a 1995 report Sensible Drinking, from which the guidelines were taken.
So I went to look at the report. It’s available for download here.
The first thing I noticed was that the UK (in 1995) ranked 14th out of 20 OECD countries for alcohol consumption and 25th out of 40 WHO countries. UK alcohol consumption is therefore not particularly high by world standards (although I believe we have moved up the rankings slightly since then). The next thing I noticed was that consumption had remained fairly steady over the period 1980-1995, although admittedly the HMRC figures (but not the ONS figures) in the more recent report show a slow and steady increase from 1995 to 2004 (totalling 24%), with a slight fall in 2005 and 2006. The point I am making is that, contrary to the impression sometimes received, current UK alcohol consumption is high neither by international nor by historic standards (alcohol consumption was “considerably higher” in the late nineteenth century).
The second thing I notice is that the authors are keen to stress the distinction between association and causality, a point which seems to have been lost in subsequent discussion. For instance, one often sees it implied that the association between depression and alcohol consumption shows that alcohol makes people depressed, whereas it might show that depression leads people to drink.
The third thing I notice is that “One UK authority considers that around 20 units of alcohol daily for 5 years is probably the minimum associated with significant liver damage.” Admittedly, that’s only one authority, but there is nothing in the report to suggest that their conclusion is wrong. 20 units of alcohol is a huge amount: half a litre of vodka or two and half bottles of wine. So the implication in the BBC report I referred to above and elsewhere that people slightly exceeding Government guidelines are putting their livers at risk is seriously misguided.
Other than for liver cancer, the report finds no evidence of any causal connection between cancer and alcohol consumption. There does appear to be a clearly established link between alcohol consumption and hypertension (although I note that hypertension itself is a risk factor rather than a substantive health condition), although I note that the report makes an elementary mathematical error in paragraph 6.16 which appears to lead them to conclude that the safe consumption for men is 4 units/day when their logic should have led them to a figure of 6 units/day. In terms of mental illness, the report only identifies an association in the case of extreme heavy drinking (over 10 units/day) for long period of time.
The report then goes on to consider “all cause mortality” as a function of alcohol consumption, which they explain is a J-curve, with non-drinkers being more at risk than light to moderate drinkers, with the risk then increasing so that very heavy drinkers are most at risk. They identify a band of minimal mortality between 7 and 28 units for men. Note, in the first place, that the upper boundary is 28 units. There does not seem to be any reason why the John Moores’ report uses 22 units as the lower boundary for the “hazardous” band. Note in the second place, that a person drinking just over 28 units a week is “at risk” in the same sense that a person drinking just less than 7 units a week: both of them are marginally missing the band of minimum mortality. So to describe the 22-50 unit band as “hazardous” is grossly misleading. At least for the lower half of that band, it would make just as much (or more) sense to describe the 0-7 band as hazardous. What we are talking about here is a minor variation in the risks of certain conditions.
The fourth thing I notice is that when it comes to advice for women, the report basically says, “We have no sufficient data because women don't drink enough to suffer interesting drink related illnesses, but we know that women are smaller than men and we think they might metabolise alcohol more slowly, so we’ll make something up.”
The fifth thing I noticed was this paragraph (10.15), “We see value in setting benchmarks to enable people to monitor their own drinking levels. However, advice to the general population is just that. By definition it is not applicable to all individuals. There is considerable variation in many individual characteristics - for example, body weight - and general advice needs to be considered in the light of these differences.”
It goes on, "The current message has sometimes been interpreted as a rigid limit, suggesting there is a critical difference for men between drinking 20 and 22 units a week. We do not believe that this approach reflects the scientific evidence."
This obviously sensible point seems to have been ignored in much recent discussion, of the "if you drink a large glass a day, you're putting your health at risk" type.
In conclusion (congratulations if you've got this far!), I feel vindicated in my suspicion that the government and media are feeding us a considerable quantity of disinformation on this topic. The intention presumably is to achieve an overall benefit to public health, but I must seriously question the wisdom of attempting to persuade people to do what you think is in their own interest by providing them with partial and tendentious information. It does not seem like the way a mature government should treat its populus.
I’ve not been able to find the report itself. It appears from the way it's been reported that it considers the geographical distribution of two classes of drinkers: (i) “hazardous drinkers”, being men who drink 22-50 units a week and women who drink between 15-35 units a week and (ii) “harmful drinkers”, being men who drink more than 50 units a week and women who drink more than 35 units a week.
The conclusion appears to be that whereas harmful drinking is associated with poorer areas, hazardous drinking is associated with more prosperous areas, although I’ve not seen anything like an analysis of the correlation between income and drinking habits. The list of top ten hazardous drinking hotspots falls a long way short of being Where’s Where of middle-class prosperity. It seems likely that the class based aspects of this story have been spun out of all proportion to the original report, but in any case, that isn’t what I want to focus on.
What I want to focus on is the tendentious nature of those labels. What does “hazardous” mean, any way? How does hazardous differ from harmful? The obvious way of reading the labels is that people who drink at a “hazardous” level are at risk, whereas those who drink at a “harmful” level are definitely harming themselves.
Dr Gray is quoted as saying, “There are increased risks of illness in people who are drinking in the hazardous category… The biggest risk to your health is chronic liver disease, followed by heart disease.” I have added the ellipsis where the BBC have put a paragraph break, because I am moderately confident that Dr Gray’s comments have been edited and that there were some other words between those sentences. The reason I am confident of this, apart from the fact that that’s what journalists do, is that it is simply not true that drinking in the “hazardous” category gives rise to a measurable risk of chronic liver disease (more on this below).
One good aspect of the story is that, for once, the alcohol content of wine has been correctly reported. As it says at the bottom of the linked-page, a large glass of wine contains about three units of alcohol. In my (fairly wide) experience, a typical wine glass is 175ml, and typical wine is 12% abv, so a typical glass will contain a little over two units, whereas one still often sees the advice that a glass of wine is about one unit.
They then rather spoil the effect by indicating that drinking a large glass of wine a day falls within the hazardous range, when in fact it does not. 3 units a day is in line with government health guidelines (for a man). I saw the same thing reported on the television news as well: that 7 large glasses a week for a man or 5 large glasses for a woman would constitute hazardous drinking. This is inconsistent (it’s also wildly misleading for reasons I will set out below). It may be that they’re thinking that at least some of those glass will be of wine over 12% abv, which will push the 7 glasses over 21 units. If so, this does rather indicate the folly of the approach, because it blatantly is not the case that there is a significant difference between 21 and 22 units a week.
Marcel Berlins wrote about this in yesterday’s Guardian. It appears that on the Today Programme, Prof Bellis said that two bottles of wine a week constituted hazardous drinking for men. Two bottles would be 18 units, maybe 20 units if the wine is high in alcohol. There are a very few bottles of wine which are 15% abv, which would reach 22 units, but by and large Prof Bellis’ comment (assuming he did say that) is not consistent with his own categories. Berlins then equates this to two “ordinary sized” glasses a day, which means his glasses are a lot smaller than mine. (That may well be true; glasses have got bigger and I’m younger than Berlins, even though I am now very old.)
These details, while giving an impression of the general confusion on this subject, are not of substantial importance. Whether the measure is two ordinary sized glass or one large glass, Berlins’ point stands that it is just not credible that drinking these sorts of amounts poses a risk to human health. Like him, I have seen not a single example (either from my own experience or in the media) of a person who has suffered ill effects from drinking these sorts of amounts and all the medical discussions I have seen of alcohol related illness relate to much larger amounts.
My suspicion, and it does seem to be born out by Bellis’ comment that drinking at moderate levels could progress into drinking at heavy levels, is that the public health concern is not really about people drinking 22-50 units a week at all. The point of the government’s alcohol strategy is to reduce drinking generally in order to create a culture less likely to give rise to problem drinkers.
That may or may not be fair enough (as a libertarian, I rather think not), but the important point is it won’t work. The reason is that most people who drink moderately, like myself and Berlins, will react to the guidance with incredulity. The effect of telling middle class people that they should not drink a large glass of wine a day will not be that middle class drinkers will cut back. The effect will be that they conclude (correctly) that the Government is scaremongering. The Government’s health advice is discredited as a result, which is potentially very dangerous.
I was sufficiently interested by this that I looked on the Department of Health web-site in an effort to find out what the Government’s policy actually is. I discovered that a report on the “next steps” in the National Alcohol Strategy was published recently (5th June 2007). It’s available for download here.
The report sets out the Government’s advice on drinking levels. No more than 3-4 units a day for men, 2-3 women. As ever when I see that advice, I puzzle as to how a safe maximum can be expressed as a range. Pregnant women should avoid drinking alcohol, but if they do drink they should drink no more than 1-2 units of alcohol once or twice a week. That advice seems to be a very mixed message: if I’m a pregnant woman, can I drink lightly or not? The absence of medical basis for the “don’t drink” message has been widely discussed elsewhere, so I won’t go into that here.
“Harmful” drinking is defined as “drinking at levels that lead to significant harm to physical and mental health and at levels that may be causing substantial harm to others.” The section goes on to say regularly drinking over 6 units a day (or over 35 units a week) for women and over 8 units a day (or 50 units a week) for men puts place people “at highest risk of such alcohol related harm”. The weekly numbers correspond to those used in the John Moores report for harmful drinking, but if I’m reading the guidance correctly, “harmful drinking” is defined (as one would expect) by reference to consequences and those numbers are merely indicative of those likely to be at risk.
The notion of risk in public health is one which requires some care. I recently saw something suggesting that people who didn’t get sufficient exercise were “at risk” of obesity. That’s obviously nonsense from the individual’s point of view. If I have a BMI of 20 and don’t exercise, then I’m not at risk of becoming obese. One needs to distinguish between a statistical treatment of populations, where one can talk about such risk factors and a causal treatment at the level of the individual, where one cannot. It’s clearly true that if more people followed the Government’s recommendations on exercise, there would be less obesity. In the case of a given individual, they are either obese or they are not. There’s no random element involved.
Similarly, I’ve no doubt that people who are harmful drinkers by the above definition tend to fall in the over 50 units a week category (if men, and as the report lately reveals, they are mostly men). However, at the level of the individual, it’s all much more complicated. It depends on patterns of drinking as well as overall volume, and it depends very much on the individual’s physical and psychological make-up.
I can see that, from the point of view of collecting and analysing statistics, it is helpful to replace a qualitative definition with a quantitative one, but at the individual level we have to revert to the qualitative definition, or we end up with the absurdity of saying that a man who drinks 49 units a week is not a harmful drinker, whereas one who drinks 50 units a week is.
Under “binge drinking” the report mentions that it is usually analysed by reference to exceeding 8 units a day for men or 6 units a day for women. It remarks, “In practice, many binge drinkers are drinking substantially more than this level, or drink this amount rapidly, which leads to the harm linked to drunkenness.” This amounts to admitting that a woman who drinks a large glass of wine at lunch and another at dinner is not binge drinking, either in the colloquial sense of the phrase or as defined in the report. Again what we have is a statistical convenience (because to conduct a survey one needs to have a quantitative way of classifying something a binge) which has some how been converted (largely I think by the media) into a classification.
There is no category of “hazardous” drinking at all. This seems to have been introduced ad hoc in the John Moores report to include everyone who drinks more than the recommended maximums but less than the “harmful” levels (bearing in mind that those levels themselves are essentially an ad hoc statistical convenience). But this introduces an element of confusion: the Government has never claimed that drinking above the recommended levels is necessarily hazardous. The levels are no more than a public policy guideline.
One fascinating snippet on p14: the HMRC figure data on “duty paid” clearances (which will certainly be an underestimate of alcohol consumed) show that the average adult consumes 11.3 litres of alcohol (or 1,130 units) a year, amounting to an average of just over three units a day. Bearing in mind the number of adults who don’t drink, it is clear that the average drinker is exceeding the Government guideline. ONS data, on the other hand, based on self-reporting, suggests the average adult drinks 10.8 units a week, around half of the HMRC figure. From this we can conclude what I have always suspected: that many people seriously underestimate their alcohol consumption. It follows that figures suggesting around 35% of men and 20% of women exceed the Government limits are likely to be a significant underestimate.
The report refers to a 1995 report Sensible Drinking, from which the guidelines were taken.
So I went to look at the report. It’s available for download here.
The first thing I noticed was that the UK (in 1995) ranked 14th out of 20 OECD countries for alcohol consumption and 25th out of 40 WHO countries. UK alcohol consumption is therefore not particularly high by world standards (although I believe we have moved up the rankings slightly since then). The next thing I noticed was that consumption had remained fairly steady over the period 1980-1995, although admittedly the HMRC figures (but not the ONS figures) in the more recent report show a slow and steady increase from 1995 to 2004 (totalling 24%), with a slight fall in 2005 and 2006. The point I am making is that, contrary to the impression sometimes received, current UK alcohol consumption is high neither by international nor by historic standards (alcohol consumption was “considerably higher” in the late nineteenth century).
The second thing I notice is that the authors are keen to stress the distinction between association and causality, a point which seems to have been lost in subsequent discussion. For instance, one often sees it implied that the association between depression and alcohol consumption shows that alcohol makes people depressed, whereas it might show that depression leads people to drink.
The third thing I notice is that “One UK authority considers that around 20 units of alcohol daily for 5 years is probably the minimum associated with significant liver damage.” Admittedly, that’s only one authority, but there is nothing in the report to suggest that their conclusion is wrong. 20 units of alcohol is a huge amount: half a litre of vodka or two and half bottles of wine. So the implication in the BBC report I referred to above and elsewhere that people slightly exceeding Government guidelines are putting their livers at risk is seriously misguided.
Other than for liver cancer, the report finds no evidence of any causal connection between cancer and alcohol consumption. There does appear to be a clearly established link between alcohol consumption and hypertension (although I note that hypertension itself is a risk factor rather than a substantive health condition), although I note that the report makes an elementary mathematical error in paragraph 6.16 which appears to lead them to conclude that the safe consumption for men is 4 units/day when their logic should have led them to a figure of 6 units/day. In terms of mental illness, the report only identifies an association in the case of extreme heavy drinking (over 10 units/day) for long period of time.
The report then goes on to consider “all cause mortality” as a function of alcohol consumption, which they explain is a J-curve, with non-drinkers being more at risk than light to moderate drinkers, with the risk then increasing so that very heavy drinkers are most at risk. They identify a band of minimal mortality between 7 and 28 units for men. Note, in the first place, that the upper boundary is 28 units. There does not seem to be any reason why the John Moores’ report uses 22 units as the lower boundary for the “hazardous” band. Note in the second place, that a person drinking just over 28 units a week is “at risk” in the same sense that a person drinking just less than 7 units a week: both of them are marginally missing the band of minimum mortality. So to describe the 22-50 unit band as “hazardous” is grossly misleading. At least for the lower half of that band, it would make just as much (or more) sense to describe the 0-7 band as hazardous. What we are talking about here is a minor variation in the risks of certain conditions.
The fourth thing I notice is that when it comes to advice for women, the report basically says, “We have no sufficient data because women don't drink enough to suffer interesting drink related illnesses, but we know that women are smaller than men and we think they might metabolise alcohol more slowly, so we’ll make something up.”
The fifth thing I noticed was this paragraph (10.15), “We see value in setting benchmarks to enable people to monitor their own drinking levels. However, advice to the general population is just that. By definition it is not applicable to all individuals. There is considerable variation in many individual characteristics - for example, body weight - and general advice needs to be considered in the light of these differences.”
It goes on, "The current message has sometimes been interpreted as a rigid limit, suggesting there is a critical difference for men between drinking 20 and 22 units a week. We do not believe that this approach reflects the scientific evidence."
This obviously sensible point seems to have been ignored in much recent discussion, of the "if you drink a large glass a day, you're putting your health at risk" type.
In conclusion (congratulations if you've got this far!), I feel vindicated in my suspicion that the government and media are feeding us a considerable quantity of disinformation on this topic. The intention presumably is to achieve an overall benefit to public health, but I must seriously question the wisdom of attempting to persuade people to do what you think is in their own interest by providing them with partial and tendentious information. It does not seem like the way a mature government should treat its populus.