Tuesday, August 05, 2008

No Reduction in Heart Attacks in Wales During First Nine Months Following Smoking Ban; Anti-Smoking Groups Cherry-picked Data to Try to Show an Effect

(Note: This post draws upon an analysis of heart attack data by Christopher Snowdon over at Velvet Glove Iron Fist. The original piece upon which this post is based can be found here).


Data from Wales released by Christopher Snowdon yesterday reveal that there was no reduction in hospital admissions for myocardial infarction (heart attacks) during the first 9 months after implementation of the smoking ban throughout Wales. Snowdon obtained monthly data on heart attack admissions from all Welsh hospitals for the years 2006 and 2007. The smoking ban went into effect on April 2, 2007.

There were 4,199 heart attack admissions in 2006 and 4,155 in 2007. Thus, there was essentially no change in heart attacks between these two years. In contrast, there was a 6.3% decline in heart attack admissions from 2005 to 2006 and a 10.3% decline in admissions from 2004 to 2005, according to Snowdon.

When analyzing the data by month, Snowdon found that comparing 2007 to 2006, there was an increase in heart attack admissions during the first five months after the smoking ban (April through August) and a decline in heart attack admissions during the next four months (September through December).

The Rest of the Story

Although it seems quite clear that the data show that the smoking ban in Wales was not associated with a decline in heart attack admissions during the first 9 months it was in effect (and if anything, the decline in heart attacks came to a halt), this did not stop anti-smoking groups from claiming that the smoking ban in Wales led to a reduction in heart attacks.

How did the anti-smoking groups pull off this miraculous feat, given that the data so clearly indicate no reduction in heart attacks?

They used a technique known as cherry-picking. By citing data for the few specific months in which there was a decline in heart attacks from 2006 to 2007, they purported to show that the smoking ban had resulted in a reduction in heart attacks.

In an article published this past June in The Daily Post, Action on Smoking and Health (UK) was quoted as saying: "It seems likely that the drop in hospital admissions for heart attacks is linked to the implementation of the smoking ban. It shows just how quickly the benefits can be felt."

What data were ASH referring to in making this pronouncement? Clearly, it was not the data for the first nine months after implementation of the smoking ban (April through December).

It turns out that ASH was citing data cherry-picked for the months October through December. Compared to the corresponding period in 2006, this three-month period in 2007 saw a 13% decline in heart attacks.

Had ASH examined data for the full period of April through December (which was obviously available at the time of its pronouncement), it could have come to no conclusion other than that there was no decline in heart attacks associated with the first 9 months of the smoking ban.

In an article published in Wales Online, the British Heart Foundation is quoted as stating: "These new statistics are very significant, and indicate the smoking ban has had a beneficial effect on the number of heart attacks quicker than many people predicted."

What "new statistics" is the British Heart Foundation referring to which purportedly demonstrate that the smoking ban in Wales resulted in a reduction in heart attacks?

Once again, it turns out that the British Heart Foundation is referring to the cherry-picked data from October through December of 2007 which reveal a 13% decline in heart attacks during these months compared to the corresponding period in 2006. But a broader look, which includes all 9 months following the smoking ban, confirms that there actually was no change in heart attack admissions. Apparently, these three months were cherry-picked in order to show an effect.

Implications of the Rest of the Story

There are three important implications to this story.

First, these data cast doubt on the conclusion of studies - such as the report from Scotland published last week - that smoking bans result in a dramatic, immediate drop in heart attacks. There seems little doubt that such a decline was not observed in Wales, a result which seems inconsistent with the conclusion of these other studies.

Second, it will be interesting to see whether anti-smoking groups acknowledge these data or ignore them. Are anti-smoking groups interested in the truth, or are they only interested in results which support their pre-determined conclusions?

Third, this story demonstrates that anti-smoking groups are not relying on solid science to draw their conclusions about the effects of smoking bans on heart attacks. They are willing to rely on cherry-picking of a fraction of the relevant data to draw their conclusions and make their pronouncements to the public.

It's quite clear that anti-smoking groups are prepared to rally around and tout the results of any study that draws favorable conclusions, regardless of the quality of the science. But what's most interesting to me is that we in tobacco control have denounced smoking ban opponents for relying on studies with exactly the same tactics - cherry-picking of data - in order to "demonstrate" that smoking bans result in a decline in restaurant sales. How can we on the one hand denounce studies that do not consider all relevant data and come to unfavorable conclusions and then on the other hand tout the results of studies using the same methodology which come to favorable conclusions?

Will the British Heart Foundation and ASH now retract their earlier statements and apologize to the public for drawing and disseminating premature conclusions? I doubt it. This is why it is essential that we get it right the first time around.

But when your sole criterion for scientific quality is whether the study results support your pre-determined conclusion, it is going to be difficult to get it right the first time, or any time.

No comments: