'No safe level' booze guidelines? Nonsense, thunder stats profs

When the Royal Statistical Society slaps you down, you've got it badly wrong

The heads of Britain’s statistics society have written to the Health Secretary to point out that the government’s alcohol guidelines don’t accurately reflect the numbers.

Two weeks ago the British civil servant at the Department of Health responsible for issuing medical advice (who rejoices in the corporate-inspired title of “Chief Medical Officer”), Dame Sally Davies, declared that there was “no safe level” of alcohol drinking.

This followed a report produced for her by an “expert group” which contradicts evidence that teetotallers are at higher risk than moderate drinkers, and drinking after middle age correlates with a substantially lower level of risk of heart disease and strokes. Drinkers live longer.

Professors Peter Diggle and Sir David Spiegelhalter – the current and next presidents of the Royal Statistical Society – say the report was unbalanced (a nice way of not saying “biased”) and that the bureaucrat’s claims don’t reflect the evidence available to the government’s working group on safe alcohol levels.

The report recommend an upper limit of 14 units per week for both adult men and women, and then included the much-derided “no safe limits” observation. Members of the expert group include prohibitionists and anti-alcohol campaigners.

In their letter to Health Minister Jeremy Hunt, the two stats profs argue that the derision which greeted Hawkins report will make more pressing public health campaigns less credible. Once the public has ignored Hawkins crying wolf over alcohol they may be inclined to ignore the others, too. They add that Hawkins’ group contradicts the principle that the public deserves an “informed choice”.

Spiegelhalter, the Winton Professor for the Public Understanding of Risk in the Statistical Laboratory at Cambridge University*, had already commented on the relative “dangers” of alcohol.

“An hour of TV watching a day, or a bacon sandwich a couple of times a week, is more dangerous to your long-term health,” he noted.

Medical statistician Adam Jacobs makes a thorough analysis of the modelling used campaigners “expert group” on his StatsGuy blog, here.

He finds much to ponder, as the assumptions behind Sheffield University’s model are not documented. One expected outcome of the model is the conclusion that women can drink more, safely, than men can – “which contradicts most of the empirical research in this area,” he notes.

It appears an undocumented model has produced numbers directly contradicted by the empirical real world evidence, and then the real world evidence has been discarded. It must be some model.

Reminder: today is Friday.

For the sake of your health: You know what you must do. ®


Spiegelhalter has also written lucidly about the dubious use of statistics in alcohol campaigns on his blog, Understanding Uncertainty. Last February he wrote that a more accurate analysis of a report claiming “no safe levels”, on which the Expert Group subsequently based its claim, would more accurately be headlined “Study supports a moderate protective effect of alcohol”.

Biting the hand that feeds IT © 1998–2019