Monday, July 21, 2008

News Article Invokes Smoking Ban as Major Cause of 49% Decline in Heart Attack Admissions in North Cumbria; Junk Science on this Issue Running Rampant

Junk science on the issue of smoking bans and dramatic, immediate reductions in heart attacks continues to run rampant. The latest example is an article in the News & Star (Cumbria, England) which reports that the national smoking ban in England is a major cause of an observed 49% reduction in heart attacks in north Cumbria during a three-month period from October through December, 2007, compared to the same three-month period one year earlier (prior to the smoking ban).

According to the article: "The number of patients suffering heart attacks in north Cumbria has almost halved since the introduction of the smoking ban a year ago. ... According to the Department of Health, there was a 49 per cent reduction in admissions across north Cumbria over a three-month period. This compares to a national drop of two per cent, with the total admissions falling from 4,979 to 4,889. Experts claim that one of the reasons for this drop has been the ban on smoking in public places, which has encouraged more people to give up. The latest figures look at the three-month period from October to December 2006, compared to the same period a year later – after the ban. When broken down they show that, in 2006, 33 people were admitted to the Cumberland Infirmary in Carlisle with heart attacks. This compares with 21 in the same period of 2007. Meanwhile 26 patients were admitted to Whitehaven’s West Cumberland Hospital in 2006, compared with ten over same period last year."

The Rest of the Story

There's just one problem with this "research." The smoking ban is being invoked to explain the 49% decline in heart attack admissions in North Cumbria compared to just a 2% decline in England as a whole. The problem? The control group in this case - England - is not a control group but is, in fact, the intervention group. In other words, the smoking ban was implemented in all of England, so a 49% decline in heart attacks in North Cumbria in light of a much smaller 2% decline in the rest of the country cannot be attributed to the smoking ban.

If anything, these data suggest that the conclusions from Helena, Pueblo, and Bowling Green are wrong and that smoking bans do not produce the kind of dramatic declines in heart attacks that anti-smoking researchers have claimed. After all, a 2% decline in heart attacks is small enough that one cannot be sure any substantial reduction in heart attacks occurred as a result of the smoking ban. But certainly, there was no 40% decline, as claimed in Helena, or a 27% decline, as claimed in Pueblo, or anything close to the 19% decline claimed in a meta-analysis of all the published studies.

Perhaps more alarming than the fact that the conclusions from these data are completely unsupported is the apparent bias in the way these data are being analyzed. One cannot simply compare the number of heart attacks during a chosen period before and after the smoking ban and attribute any change to the smoking ban. Using that same reasoning, one would be forced to conclude that the smoking ban in England had no effect on heart attacks, because in 42% of England's hospital trusts, there was no change (or an increase in heart attacks) during the first year after implementation of the smoking ban.

In fact, the finding that heart attacks declined in 58% of the hospital trusts is not significantly different from the expectation that heart attacks would decline in 50% of the trusts under the assumption that the smoking ban had no effect on heart attacks. If one is really going to employ the reasoning being used in the above analysis, then one would have to conclude that there is no evidence to support the conclusion that the smoking ban had any effect on heart attack admissions.

What I don't really understand are two things:

1. Who is it who is putting out all of this junk science, and why are they doing it? It seems hard to believe that newspaper reporters themselves have some sort of agenda to show that the smoking ban is causing massive declines in heart attacks. It seems more likely that the health department or some other entity is publicizing these data and misleading reporters into believing that the data support such a conclusion.

2. Why is it that these data are not being reported as refuting the conclusions from Helena et al. that there is a dramatic reduction in heart attacks caused immediately by implementation of smoking bans? Whether the decline in heart attacks was only 3% as previously reported or 2% as reported here, these data do not support the conclusion that smoking bans lead to substantial, immediate declines in heart attacks.

What does seem clear is that the studies of the Helena ilk have somehow opened the door for junk science as an acceptable tactic in tobacco control research.

And here is the most amazing thing: very few people or organizations within tobacco control are speaking out against the deterioration of the science.

I think these data from England actually explain why the Helena et al. studies are so flawed. If you just pick several specific locations to study, you are bound to find some locations that experience major declines in heart attacks. Within England, sure enough - Cumbria experienced dramatic reductions in heart attacks. Most of the 49% decline was probably just due to random variation given the low numbers involved.

What you need to do is a systematic study where you look at many locations affected by a smoking ban - preferably examining many countries or many states with such bans and comparable control locations without such bans. When one does this in England, it becomes immediately apparent that the finding in Cumbria was just a fluke. There is no consistent finding of a 49% decline in heart attacks across England. This is called random variation or chance.

Most likely, locations chosen for study are those in which investigators notice that there seems to be a decline in heart attacks. If heart attacks went up after a smoking ban, most likely investigators in those locations would not opt to conduct and publish such a study.

Interestingly, this serious problem of publication bias is not even mentioned in the meta-analysis of these studies which was recently published online in Preventive Medicine.

To really answer this research question, what is needed is a systematic study of multiple locations with smoking bans. I believe that McFadden and Kuneman have conducted such a study, in which they examined heart attack trends in a number of U.S. states with and without smoking bans for which data were available. This study has not yet been published, as not surprisingly, it is probably getting unfair reviews from anti-smoking researchers.

I conducted my own analysis of the data, using McFadden and Kuneman's methods, for all U.S. states with and without smoking bans for which data were available. I found no evidence that there is any substantial, immediate decline in heart attack admissions in states with smoking bans.

It doesn't make sense to me that anti-smoking researchers and groups aren't willing to criticize the conclusions of these studies and to decry the junk science that is being used. After all, deterioration of the science being used in tobacco control is not an asset; it is a severe liability.

However, as I am learning, science in tobacco control is judged not by its quality, but by the direction of the results.

(Thanks to Michael J. McFadden for the tip).

No comments: