Tuesday, 30 April 2013

The Irish smoking ban miracle

Another smoking ban miracle, this time from Ireland and this time not limited only to heart attacks but to, well, nearly everything...

Following the implementation of the ban, an immediate 13% decrease in all-cause mortality was observed (RR: 0.87; 95% CI: 0.76–0.99).

Boom. Take that, sceptics!

Mortality decreases were primarily due to reductions in passive smoking.

Blimey. No quibbling about correlation and causation for these folks.

Who are these folks, anyway? Step forward Luke Clancy, currently of the Tobacco Free Research Institute, previously the chairman of ASH Ireland and the lead figure in the smoking ban campaign back in 2003. What better man to mark his own homework?

Let's return to that extraordinary claim...

Following the implementation of the ban, an immediate 13% decrease in all-cause mortality was observed (RR: 0.87; 95% CI: 0.76–0.99).

What does "immediate" mean here? Overnight? Within a month? A year? The authors do not tell us, but they also—quite rightly—look at diseases which they don't think are caused by passive smoking (a small and dwindling category) to use as a control.

In contrast, an immediate 15% decrease was observed for non-smoking related mortality (RR: 0.85; 95% CI: 0.75–0.97)

Doesn't that undermine the central claim?
...followed by a 5% increase each post-ban year (RR: 1.05; 95% CI: 1.02– 1.08).

Phew! That's alright then.

For ages 35–64 years, an immediate post-ban decrease in all-cause mortality was observed (RR: 0.79; 95% CI: 0.67–0.93)

Huzzah! So the smoking ban saved lives in the 35-64 age group then?

...followed by annual trend increases in all-cause mortality (RR: 1.06; 95%: 1.02–1.10), resulting in a net post-ban increase of 2%.

Maybe not. It seems that if you weren't a pensioner, the smoking ban was bad for your health (if, like Clancy et al., we assume that correlation = causation).

Of course, it is mainly pensioners who die of cancer, heart disease, stroke etc., even if it is mainly younger people who go to pubs and workplaces, and we know that all-cause mortality has been declining for years (see top blue line).

The graph below, also published by Ireland's Department of Health, tells much the same story but brings us up to 2011...

The Irish smoking ban was introduced in March 2004. Can you spot the dramatic decline in mortality around 2004-05? To me, it looks very much the existing downward trend continued and, if anything, slowed down in the post-ban period.

In the end, Clancy et al. resort to apply the statistical trick method of simply guessing how much higher rates of disease might have been if Clancy hadn't succeeded in getting the Irish government to ban smoking in 'public' places.

In the absence of a national smoking ban, an estimated 3,726 (95% CI: 2,305–4,629) additional smoking-related deaths would have occurred.

With that bit of 'science' completed all that remains is to type out the study, issue the press release and wait for the media to announce...

Smoking ban saves 3,700 lives thanks to cleaner air

All in a day's work for tobacco control.


BrianB said...

As ever, you beat me too it, Chris.

I have posted my take on this at Simon's place. It isn't currently past moderation, yet, but hopefully soon.

It might add a few thoughts to your own analysis, but I can't compete with the sheer, sarcastic beauty of your writing-style!

Good piece.

Xopher said...

Your/their 1992-2011 graph says it all.
In each case a simple straight line from the 1992 or more markedly from 1993 rate through the 2004 rate shows the 2011 rate above the line.
Only Tobacco Control can ignore that the rates of decrease have slowed and the smoking ban is to blame.
Common-sense tells us that the smoking ban has cost lives not saved them.

JohnB said...

The Tobacco Control morons are at it again. And the Drivel Machine™ is only in a “lazy” 3rd-gear.

Snowdon said...

I recommend BrianB's post (linked to in his comment above) for further reading.

Ivan D said...

BrianB has pretty much nailed it in that the immediate effect observed is entirely a consequence of the authors wishful thinking combined with (wilful?) statistical incompetence.

At least the BBC appear to have ignored this one. I am sure that Michelle Roberts would have loved to talk it up but she was busy covering a non-study about bowel cancer and smoking from Norway presumably in her professional capacity as a PR officer for CRUK.


Or perhaps in her private capacity as a public health obsessive?

paulp127x said...

Could someone tell me why the figures for mortality in the past have confidence intervals? Do we not know the actual figures by now ?

BrianB said...


Can you clarify which numbers and confidence intervals you are referring to?

As far as I can see the only numbers in the report with confidence intervals are estimates of one kind or another - ie they aren't real counts of deaths, which should, as you rightly suggest, not be subject to CIs.

But ther problem with this kind of 'study' - as has been shown time and time again - is that these people ignore the real numbers, as they are an incovenience to their message, so they invent silly 'models' that provide the usual "compelling" evidence that black is really - erm - white!

It's a dismally sad reflection of the state of our educational system - in mathematics in particular - that they get away with such blatant untruths.

They are either knowingly lying, or they are just thick!

I suspect both.

paulp127x said...

Brian, I was thinking of quotes such as the first in the post, about the immediate 13% decrease in mortality. Thanks for confirming my layman's reading