Friday 15 January 2016

The research behind those alcohol guidelines

This is a guest post by Chris Oakley...

“The benefits for heart health of drinking alcohol are less, and apply to a smaller group of the population, than previously thought. The Sheffield report commissioned for the expert group included a UK analysis which has found that the net protective effect from mortality that may be attributable to drinking regularly at low levels appears now to be significant only for women aged 55+ (with men aged over 55+ showing such a net protective effect only of negligible size).”

This statement is taken from the executive summary of the recently issued Alcohol Guidelines Review. The review is the work of “expert” groups and the implication of this carefully worded text is that experts have reviewed a new body of evidence that shows the well-established and well researched health benefits of alcohol to be very much less important than previously thought and to be dependent on age and sex.

This seems to be how the press has interpreted the report and if they are to be believed so has the DH. According to the Daily Mail

“The Department of Health said researchers at Sheffield University analysed a number of studies showing alcohol only protected the hearts of women over 55. Even so, this was only for small amounts – less than one unit a day …”

This devastating news for moderate drinkers was reported very widely indeed, usually with some reference to protective effects only applying to post-menopausal women and always with the implication that the findings reflected the latest published research.

The problem is that this interpretation is not backed by published research and what the Daily Mail is reporting the DH as saying is not true.  The claim that a protective effect is only seen in women over 55 is in fact based entirely on an unpublished speculative report produced for Public Health England by a team from Sheffield University led by Dr John Holmes. 

Holmes and his team are the same people who came up with a magic bullet in the form of a minimum unit price (MUP) model that they claimed would target problem drinkers without impacting everyone else. Many people were intrigued by the model including me so I went to the trouble of wading through the extensive analysis that was supposed to be the basis for MUP and was shocked to find that it was not only not supported by the literature but was at odds with many of the most relevant publications. The Sheffield team ploughed ahead regardless and were paid by two governments and the BBC for a model that is economically counterintuitive and contradicts much of the published evidence. Fortunately, I was not alone and many people came to understand that MUP would not target the people Holmes said it would and would be a regressive unpopular tax. Some persist in an irrational belief in MUP including Nicola Sturgeon and some of the members of the “expert” alcohol guideline groups who are still lobbying for it when they are not being impartial experts. 

It is hard to imagine, based on the academic quality of the MUP modelling, how anyone might consider the Sheffield team suitable candidates for remodelling UK alcohol guidelines and utterly incredible that its findings should be given precedence over knowledge gleaned from decades of oft tested published research, but that appears to be what has happened.  

The Sheffield report repeatedly challenges the widely held belief that moderate alcohol consumption has significant benefits. Both it and the DH guidelines claim or imply that there is widespread scientific controversy on this issue but an in depth analysis using the most relevant references included and omitted from the Sheffield report demonstrates that the controversy is largely confined to John Holmes et al and Tim Stockwell, a man who is still trying to claim that MUP reduced deaths in parts of Canada based on evidence that nobody else seems able to see and which is at odds with official figures. 

The Sheffield report claims that protective effects are disputed, may be overestimated and are probably limited to particular groups within society. We might assume that those groups would be the post-menopausal women and perhaps older men mentioned in the guidelines and that there is a body of evidence proving it but in fact the report provides only one reference, a Canadian paper written by Michael Roerecke and Jürgen Rehm, who conclude that their findings...

“ current low-risk drinking guidelines, if these recognize lower drinking limits for women”. 

This is at odds with the Sheffield report, the proposed UK guidelines (which now give men and women the same 'limits') and the attendant publicity.

Elsewhere in the Sheffield report we are told that the same Roerecke and Rehm paper is used as a basis for risk functions that suggest that “drinking up to 8 units a day for males and 4 units a day for females is associated with a reduced risk of IHD (Ischaemic heart disease) relative to abstainers.  This is true but in a separate analysis Roerecke and Rehm fail to find a statistically significant dose response relationship between people consuming at the lower and upper ends of the moderate spectrum, prompting speculation that protective effects peak at just one drink per day - speculation that the Sheffield report and the expert group seize upon as fact.*

Confused? That’s hardly surprising. The evidence is less than clear and to their credit Roerecke and Rehm explain this in terms of the limitations of the epidemiological data that they are working with.
Interestingly they go on to say: 

“Regarding causality of effects, a potential cardioprotective association is supported by short-term experimental evidence on surrogate biomarkers, such as increasing HDL cholesterol, reducing fibrinogen levels, and inhibition of platelet activation. Indeed, this might be the strongest argument for causality given that observational findings are always prone to residual confounding and bias due to study design.”

This translates as “there is biological evidence to support a protective effect and it is more robust than the epidemiological evidence.”

But the Sheffield report in a statement thought so important that it appears in a special box claims that:

“many researchers…point to a lack of well evidenced biological processes that could explain the effect”. 

Clearly those many researchers don’t include Roerecke and Rehm who are just about the only people other than the Sheffield team and Tim Stockwell who might be claimed to be contributors to scepticism over the protective effects of alcohol.** 

In a considered response to Tim Stockwell in the journal Addiction Roerecke and Rehm observe that...
“...some researchers in the field may be using different standards in assessing the cardioprotective effect of alcohol vs. its detrimental effect.” 

They go on to point out that they...

“...sense a desire by some in the field to apply tough standards on protective effects and more lenient standards on other effects”.

It is hard to disagree that these statements also apply to key elements of the evidence that underpins the proposed new UK guidelines. The Alcohol Guidelines Review is careful to talk about net benefits that take into account detrimental effects but some serious questions need to be asked about the manner in which the net effects were arrived at. 

There is one recent UK study that claims alcohol benefits are only relevant in older women. It isn’t referred to in either the DH Guidelines Review or the Sheffield report.*** The reason for its omission is perhaps that it was discredited as statistically illiterate by the UK’s most famous statistician. That same statistician, David Spiegelhalter was apparently consulted on the proposed alcohol guidelines. I would love to reads his analysis of the Sheffield report with which his name is likely to be inextricably linked.****

In my view, the guidelines and supporting data need to be assessed by competent statisticians and presented by a much more objective group of “experts” before anyone can take them seriously. Someone in Westminster should perhaps mull on the value of exercises such as this review and consider what they tell us about standards at the Department of Health.  

I can't resist adding a few footnotes - CJS
* The peak in protection is almost certainly below 14 units per week but that has no bearing on what the 'safe' level should be. If it is considered safe to not drink, any mortality risk below that of a non-drinker must also be safe. In the Neverland of Sheffield's computer model that level is reached at 14 units. In the real world, as observational data show, it is at least twice that.

** The claim that there are no biological mechanisms to explain the protective effect is such a tired zombie argument - see p. 7 of the previous evidence review from 21 years ago.

*** There is an obvious point that I haven't heard anyone make regarding the protective effects of alcohol 'only' applying to people over the age of 55, namely that heart disease overwhelmingly affects people aged over 55. Sally Davies has strongly implied, if not explicitly stated, that there are no benefits to drinking under this age, but the beneficial impact on heart health is likely the result of long term moderate consumption. The benefits may only pay off in late middle and old age, but they accrue earlier.

**** It should be remembered that even that study - which went out of its way to downplay the benefits of drinking - found that men aged 50-64 years who consumed 15.1 to 20 units of alcohol per week halved their mortality risk compared with nondrinkers - and this finding was statistically significant. This makes the CMO’s decision to reduce the guidelines from 21 units to 14 units per week for men all the more baffling. Moreover, the study did not merely show a reduction in mortality for women over 55 consuming up to 5 units per week, as Sally Davies suggested on the Today programme. It showed a statistically significant reduction in female mortality at every level of drinking up to and over 20 units per week. The strongest effect was shown at the level of 15.1 to 20 units per week with mortality rates falling by around 40 per cent for women over 65 and by around 46 per cent for women aged 50-64 years. In other words, the optimal effect is seen only when people drink above the existing guidelines.

No comments: