Thursday, 4 February 2016

Garbage in, garbage out

Heart miracles are once more in the news. This story, for instance, is on the front page of the Telegraph...

Smoking ban sees 40 per cent cut in heart attacks in UK since 2007 law was introduced

Heart attack rates in the UK have fallen by up to 42 per cent since the 2007 smoking ban, major research suggests.

A review of 77 studies found that reduced exposure to passive smoking has caused a “significant reduction” in heart problems across the population.

The basis of the report is a literature review from the Cochrane Collaboration (not to be confused with a Cochrane Review). Smoking ban/heart attack miracles were invented by Stanton Glantz with the notorious Helena study. Glantz has since published at least three meta-analyses/literature reviews, all of which cobble together terrible studies in the hope of masking their obvious flaws and providing the illusion of strong evidence. You can do this as many times as you like but it won't stop the underlying evidence base being outrageously, hilariously and often fraudulently awful. A meta-analysis of smoking ban miracles is the ultimate garbage in, garbage out study.

Michael Siegel and myself have written about dozens of these studies over the years. For a while, it was mildly entertaining because those responsible would try audacious new tricks, but they have generally settled for relying on the simple post hoc ergo propter hoc fallacy (as in the Telegraph headline above) or by using one of two techniques I described in 2014:

Firstly, dredging the data for any town that saw a large decline (in percentage terms) in heart attacks at around the time of a smoking ban. Nobody decided to do a study of Helena, Montana or Bowling Green, Ohio before the bans took place. The decision to focus on such obscure places came about only once it was clear that they were anomalous (not unlike Derren Brown's horse-racing trick). They were then presented to the media with the implication that they had been randomly selected.

Secondly, although less frequent, studies of larger populations have portrayed rather small declines in the heart attack rate as being the result of a smoking ban, without acknowledging that that there had been a secular decline of the same magnitude long before the ban was enacted. As the authors of the above study note, the secular decline is simply ignored in such cases.

As that blog post highlighted, there is also a great deal of publication bias in this corner of quackademia.

I've pretty much exhausted the subject on this blog. There's a large archive of posts if you want to catch up, but I will leave you with a couple of graphs showing hospital admissions rates and mortality rates from heart attacks in the UK between 2002 and 2010 (click to enlarge). I trust you will notice the lack of change in the trends when the smoking ban was enacted in 2007.

These graphs come from a 2012 BMJ study which looks at trends in heart attack mortality and the reasons for them. It doesn't mention the smoking ban at all, nor does this study from Heart which looks at the epidemiology of cardiovascular disease in the UK. That's because (a) the smoking ban didn't have any effect on cardiovascular disease or rates of acute myocardial infarction, and (b) the authors are not tobacco control activists and therefore do not have the same exceptionally large incentives to make stuff up.

No comments: