Wednesday, December 16, 2009

Error Disclosed in Meta-Analysis of Smoking Bans' Effect on Heart Attacks: Will Anti-Smoking Groups Report the Error?

In October, Dr. David Meyers and colleagues published a paper in the Journal of the American College of Cardiology in which they reported the results of a meta-analysis of published studies on the effect of smoking bans on heart attack admissions. The paper concluded that smoking bans were associated with a 17% decline in heart attack admissions in the 11 studies that were reviewed.

These results were disseminated widely in the media and heavily touted by anti-smoking groups as supporting the conclusion that smoking bans immediately and dramatically reduce heart attacks and that brief exposure to secondhand smoke in bars and restaurants causes a large number of heart attacks.

According to an article in the Lawrence Journal-World & News, Dr. David Meyers, professor of cardiology and preventive medicine at Kansas University Medical Center and lead investigator of the study claimed that: "Within minutes of the ban, it is going to start having an effect on heart attacks."

Meyers was quoted as supporting his claim with the following argument: "Heart attacks are caused in large part by blood clots. With 20 minutes or so of tobacco smoke exposure, people’s blood becomes hypercoagulable and sticky and clots easily, and bam, you have a heart attack."

I criticized the study's conclusions on a number of grounds, including the fact that no control or comparison groups were examined, so that there was no way to determine whether the 17% decline in heart attacks might have occurred anyway, even in the absence of the smoking bans. I also noted the severe weaknesses in the underlying studies and the fact that the review only included published data, while ignoring a large body of unpublished data which refutes the study conclusion, while employing much larger sample sizes and thus having greater validity.

The Rest of the Story

As it turns out, the study findings were due to a careless error. In the original study, the authors had inadvertently reported the Pueblo study has having reported a 70% reduction in heart attacks (a result that is completely implausible and clearly should have been noticed as having been in error). Instead, that study actually reported a 34% reduction in heart attacks. The meta-analysis authors published a correction in which they re-analyzed the correct data.

It turns out that the 11 studies did not find a 17% reduction in heart attacks, but only found an 8% reduction in heart attacks.

This level of decline in admissions for heart attacks is obviously not significantly different from the levels of decline in heart attacks that are being observed in the absence of smoking bans, which have varied between 5% and 10% per year in many communities.

For example, in the United States as a whole, heart attack admissions declined by 8.2% in 2004. The decline of 8% in communities/nations with a smoking ban is comparable to this. Therefore, the meta-analysis result fails to provide any evidence that the smoking bans resulted in a decline in heart attacks.

The failure of the meta-analysis to include any comparison groups is a fatal flaw that is indicative of very poor analytic methods. For example, we know that hospital admissions for heart attacks decreased by 28.5% in Nebraska in 2004 and by 12.5% in South Carolina during the same year. These are states without statewide smoking bans or even local smoking bans in place during 2004. Thus, it is obvious why it is essential to include a control or comparison group in conducting a meta-analytic review of these studies. One needs to be able to separate out the effect of the smoking ban from the secular decline in heart attacks that is being observed anyway.

But even without the comparison group analysis, the meta-analysis - as it stands - fails to provide any evidence that smoking bans resulted in a decline in heart attacks. The observed 8% decline in heart attack admissions could easily be indicative of the underlying decline in heart attacks, which was at exactly the same level (8%) in 2004, which is the approximate time of many of the studies included in the meta-analysis.

Predictably, this correction of the error in the meta-analysis has received no media attention. No effort was apparently made to disseminate the revised results. Of course not, because they are less favorable to anti-smoking groups. Remember, it's not the truth that 's important, it's whether the results are favorable or not.

So once again, I offer a prize to any anti-smoking group which originally disseminated the results of the meta-analysis - reporting a 17% decline in heart attacks associated with smoking bans - and which now corrects that error and publicly disseminates the correct finding: that the actual decline was only 8%, which is not significantly different from what is occurring anyway in the absence of smoking bans.

Since I am so sure that no groups will respond to this challenge, I will increase the size of the award to $200.

The money is not at risk because these anti-smoking groups are really not interested in scientific rigor and accuracy. They are interested in presenting favorable data which supports their agenda. While I largely share and support that agenda, I do not share and support the tactic of misleading the public in order to achieve these goals.

(Thanks to Klaus K. for the tip)

No comments: