6

The UK government issues public health advice about how much booze people ought to consume. It has recently revised those guidelines. The previous guidelines were the subject of this SK.SE question: Recommended alcohol drinking quantity: are they just making it up?

The answers hint that there was not enough evidence to come up with clear guidance but that guidance was issued anyway. This time the 2016 advice claims to have been based on more robust evidence.

The new guidelines are lower and, according to England's chief medical officer:

Drinking any level of alcohol regularly carries a health risk for anyone

But the science behind the new guidelines has been challenged. For example, this Register article reports:

Civil servant Dame Sally Davies has drawn up the lowest recommendations in the West: there is no “safe drinking level”, her team declared.

The question is what justification was used to get there. The answer isn’t pretty for “evidence based” policy.

Repeated studies have shown that alcohol in moderation prolongs life: it reduces the risk of heart disease and strokes. In fact the benefits of alcohol in preventing strokes and heart disease are far clearer than the negatives of drinking.

So are the new guidelines robustly evidence based?

Community
  • 1
matt_black
  • 56,186
  • 16
  • 175
  • 373
  • Possible duplicate of [Recommended alcohol drinking quantity: are they just making it up?](http://skeptics.stackexchange.com/questions/8197/recommended-alcohol-drinking-quantity-are-they-just-making-it-up) – DJClayworth Jan 09 '16 at 20:19
  • 4
    @DJClayworth I specifically referenced the other question to show that this question is different. It is about the recently launched guidance which professes a much stronger scientific basis. So the question is less about the process and more about the content of the science. – matt_black Jan 09 '16 at 20:22
  • 1
    @matt_black I've made two tiny edits: "now" will be meaningless in the future, so "2016" seems more descriptive. "based on sound science" leaves the door open to speculation about what constitutes "sound" science, "evidence based" seems to be more accurate, especially since you use it yourself in the body. – Sklivvz Jan 09 '16 at 21:43
  • I've heard of *wine* giving health benefits, but never alcohol. –  Jan 09 '16 at 23:17
  • @fredsbend Alcohol consumption in general has been argued by some to give benefits, especially for reducing strokes and heart attacks. The question is whether other things outweigh that. IF there was no evidence of benefit the question would not be interesting, but there is longstanding evidence. The question is whether recent work has overturned that. – matt_black Jan 09 '16 at 23:53
  • @Matt Can you get me a link? –  Jan 10 '16 at 00:21
  • I think this question would be easier to answer, and more interesting, if it focussed on one of their claims: that there is "no safe drinking level" or "any level of alcohol carries a health risk", and circumvented the question of how a bureaucracy reached its conclusions. Even that question will be difficult to answer if alcohol is, say, good for the heart and bad for the liver. – Oddthinking Jan 10 '16 at 00:57
  • The problem with studies that compare teetotalers against people who drink small amounts is that teetotalers include a disproportionate number of (reformed/recovered/ex-) alcoholics. Hence the average health of teetotalers tends to be worse, but not because small amounts of alcohol are good for you. – Paul Johnson Jan 23 '16 at 12:16
  • @PaulJohnson This has often been proposed and is an interesting hypothesis. But attempts to estimate the size of the effect have not significantly altered the picture. – matt_black Jan 23 '16 at 12:35
  • @matt_black, do you have any references for that? – Paul Johnson Jan 23 '16 at 12:49
  • @PaulJohnson It is discussed [here](http://archinte.jamanetwork.com/) and in some of the other studies. – matt_black Jan 24 '16 at 12:14

1 Answers1

7

The recommendations come from this report to the Chief Medical Officer:

The headline recommendation was chosen to be

the level of drinking where alcohol would be expected to cause an overall 1% lifetime risk of death

That is, a 1% lifetime risk of dying of an alcohol-related condition, including accident and injury due to drinking, as well as alcohol-related medical conditions like liver failure and oesophageal cancer.

The reason for giving a single headline recommendation ("you are safest not to drink regularly more than 14 units of alcohol per week") that does not take into account sex, body mass, age, and other variables, is in order to keep it

simple, specific, measurable, timebound, and realistic

If you're happier to accept a higher lifetime risk than 1%, or if you think the differences between men and women might be significant, then you can look at the CMO's report to get the extra detail. Or you could just ignore the guidelines and carry on drinking as you were—it's just a guideline, not the law.

The criticism made by the Register is based on a confusion between two uses of the word "safe". When the Chief Medical Officer says that there is no "safe" level of alcohol consumption, she means that there is no level of drinking at which the risk of dying from an alcohol-related condition goes to zero. Whereas the Register is using "safe" to mean the level at which the harm due to drinking is balanced by the benefits. Note that the protective effect of moderate consumption of alcohol is discussed in detail in the CMO's report (see pages 14–19).

So where did the lifetime risk levels in the CMO's report come from? They are from this report commissioned from the Sheffield Alcohol Research Group:

The numbers are the output of a computer model, the Sheffield Alcohol Policy Model (SAPM):

The key data inputs into SAPM are baseline levels of current UK alcohol consumption, current levels of alcohol-related mortality and morbidity (defined as person-specific hospital admissions) and, most importantly, international and UK-specific evidence relating different levels and patterns of alcohol consumption to risk of mortality or morbidity from 43 health conditions causally related to alcohol consumption. This evidence is taken from a combination of published meta-analyses of risk relationships, analyses of the proportion of cases of alcohol-related conditions attributable to alcohol and UK mortality and morbidity rates for the 43 conditions.

The model and its inputs are described in detail in section 3 of the Sheffield report. Take a look and see if the approach makes sense to you.

Update

A comment points to this article by Adam Jacobs ("The Stats Guy") as a criticism. Unfortunately, Jacobs misunderstood the figures in the Sheffield report, and so much of his criticism is nonsensical.

Consider figure 11 from the Sheffield report.

This graph plots lifetime risk of hospitalization due to an alcohol-related cause, relative to someone who does not drink, against peak day alcohol consumption. Hence the relative risk is 1.0 when peak consumption is 0.

But Jacobs describes this as "how many times more likely you are to be admitted to hospital for an alcohol-related cause than you are on a day when you drink no alcohol at all".

I think the confusion here must be over "alcohol-related cause". The Sheffield group uses this to mean "a condition that can be caused by alcohol"—the point being that you can still be hospitalized for liver cirrhosis even if you have never drunk alcohol (liver cirrhosis has multiple causes). Whereas Jacobs seems to understand this as "a condition definitely known to have been caused by alcohol" and then struggles to find an interpretation of the graph that would make this work (because non-drinkers can't be hospitalized for such a condition, and so there would be no base rate to derive a relative risk from). The Sheffield report could have done a better job of explaining this.

Community
  • 1
Gareth Rees
  • 1,767
  • 1
  • 15
  • 14
  • The CMO report ends up with a different balance of risk and benefit than previous studies (at much lower consumption) by *modelling* rather than by observing harms (that what the Sheffield model is about). This seems odd and subject to uncheckable bias (the model assumptions are fairly opaque). See [this discussion](http://www.statsguy.co.uk/?p=172). – matt_black Jan 11 '16 at 21:58
  • @matt_black: See updated article. – Gareth Rees Jan 22 '16 at 18:57