Saturday, November 19, 2005

Premature Conclusions from Pueblo: More Information and More Research Needed Before Taking this to the Public

A press release issued on November 14 by the Pueblo City-County Health Department announced the results of a study which purported to show that the smoking ban in Pueblo, Colorado resulted in a 27% drop in hospital admissions for heart attacks. The study reported a 27% decline in heart attack admissions during the 18 months preceding the smoking ban (implemented in July 2003) compared to the 18 months after the smoking ban.

A press release issued the same day by the Campaign for Tobacco-Free Kids hailed the study as confirming that smoke-free laws reduce heart attacks, and concluded that this reduction in heart attacks occurred, at least in part, because of reductions in secondhand smoke exposure.

According to the Health Department press release: "The study validates previous scientific evidence that indoor smoke-free laws can dramatically reduce heart attacks and means that 108 fewer people had heart attacks in Pueblo in an 18-month period."

The headline of the Tobacco-Free Kids press release read: "New Study Confirms Smoke-Free Laws Reduce Heart Attacks, Shows Need to Make All Workplaces Smoke-Free."

I have already commented on what I think is the inappropriate and possibly inaccurate scientific conclusion of the Campaign for Tobacco-Free Kids that this study demonstrated a reduction in heart attacks due to reduced secondhand smoke exposure (rather than to reduced smoking by active smokers). Here, I comment on the overall conclusion that the study confirms that smoke-free laws reduce heart attacks.

The Rest of the Story

A critical piece of information is omitted from both the Pueblo Health Department press release and from the Campaign for Tobacco-Free Kids press release: that the expected number of heart attacks during the six-month period preceding the Pueblo smoking ban is substantially higher than the expected number of heart attacks during the six-month period after the Pueblo smoking ban.

The reason for this is that the six-month period preceding the Pueblo smoking ban includes the winter months, while the six-month period following the Pueblo ban includes the summer months, and heart attack admissions during the winter have been shown to be substantially higher than during the summer.

Since there were two winters and only one summer in the 18-month baseline period (before the smoking ban in Pueblo) and only one winter but two summers in the follow-up period (after the smoking ban), one would expect to see a decrease in the number of reported heart attacks, even in the absence of a smoking ban.

In fact, there are 53% more cases of acute myocardial infarction (heart attacks) during the winter compared to the summer (see: Spencer FA, Goldberg RJ, Becker RC. Seasonal distribution of acute myocardial infarction in the Second National Registry of Myocardial Infarction. Journal of the American College of Cardiology 1998; 31:1226-1233). In the Mountain region of the country (which includes Colorado), there are 50.3% more heart attacks during the winter than the summer.

A more important concern, however, is that an 18-month baseline period is probably inadequate to establish a stable baseline for comparison. It is not adequate, in my opinion, to:
  • establish the baseline seasonal variations in heart attack admissions;
  • establish the secular trend in heart attack admissions (changes in heart attacks over time); or
  • understand the variation in heart attacks from year to year, so that the differences in heart attacks from one period to the next can be meaningfully interpreted.
The problem is that we don't have an adequate idea of the number of, or range of the number of heart attack admissions that would have been expected in Pueblo in the absence of the smoking ban during the specific 18-month that included two summer seasons and only one winter season. This, in my mind, makes it extremely difficult to conclude that the observed changes in heart attacks in the study were attributable to the smoking ban, as opposed to simply chance variation or to some other factor.

Because of the inadequate baseline period of the study as well as the fact that there was a seasonal mismatch between the baseline and follow-up study periods, I do not think that the evidence presented supports a conclusion that 108 fewer people had heart attacks in Pueblo in an 18-month period due to the smoking ban (which I think the Pueblo Health Department implies) or that the study confirms smoke-free laws reduce heart attacks (as the Campaign for Tobacco-Free Kids states).

Is it likely that seasonal variation could explain a 27% drop in heart attacks in Pueblo during the 18-month period following the ban? Probably not. The seasonal variation is large, but probably not large enough to in and of itself explain a 27% reduction in heart attacks. Is it likely that the observed 27% decline reflects secular trends of decreasing heart attack incidence? Very possibly, since in at least one other city (Kent, Ohio) there was apparently a 34% drop in heart attack incidence during roughly the same time period. Is there a normal variation in heart attack incidence in Pueblo great enough so that a 27% decline would fall outside of the range of normal variation? That is impossible to know from the evidence that has been presented. I'd have to examine the actual data. And even having the actual data may not be sufficient, because as I stated above, an 18-month baseline period is probably not adequate to establish the underlying variation in heart attack rates over time.

The bottom line is that although the study presents data that is strongly suggestive of an effect of the smoking ban on reduced heart attacks, leaping to a definitive conclusion is not scientifically solid at this time. The new data may suggest an effect of the smoking ban, but they hardly validate or confirm that the smoking ban caused a reduction in heart attacks.

Thus, I think the conclusions that have been drawn from the evidence and disseminated to the public are premature. I think that the Pueblo Health Department and especially the Campaign for Tobacco-Free Kids have jumped the gun. The data were merely presented at a scientific conference - they have not yet been published or opened up to scrutiny. They have not even been released so that we can make our own judgment about the validity of the conclusions. In some circumstances, prematurely taking an unpublished study to the media is appropriate, but in this case, I tend to think not.

In my career so far I have almost never gone to the media with a scientific conclusion before my paper has been published. Why? Because I think there is a responsibility that comes with being in public service, and part of that responsibility is making data and methods publicly available so that independent judgment is possible. In this case, it is not. All we have are the biased conclusions of the Campaign for Tobacco-Free Kids, which, frankly, I do not trust.

I'll be very honest here. If this were a study of the economic impact of a smoke-free law in Pueblo, and the researchers compared restaurant sales data for an 18-month period prior to the ordinance to sales during the 18-month period following the ordinance, found a 27% decline in revenues and attributed that decline to the smoking ban, anti-smoking groups would blast the study as being seriously flawed.

In fact, anti-smoking groups would probably point out the 34% drop in "revenues" in Kent, Ohio during roughly the same period and suggest that the sales pattern in Pueblo could simply be reflecting secular trends.

They would, I am quite certain, attack the study for not having a long enough baseline period, for not having a sufficiently large comparison group, for not having accounted for secular trends in restaurant sales, for not having accounted for seasonal trends in restaurant sales, and for not being able to rule out the possibility that the 27% decline in sales was not simply within the bounds of normal variation in sales from year to year.

However, with the same study design, anti-smoking groups seem to be perfectly content to draw definitive causal conclusions when the observed effect is a positive one.

I am not concluding here that the observed reduction in heart attacks in Pueblo was not real or that it was not caused by the smoking ban, or even that there is not reason to believe that the reduction may have been attributable to the smoking ban. What I am saying is that I do not believe the evidence from this study, based on the information available to us, is sufficient to conclude that the smoking ban caused a significant decline in heart attacks and that the conclusion that the Campaign for Tobacco-Free Kids made and communicated to the public is scientifically shaky, in my opinion.

Most importantly, this is another example that is demonstrating to me that the science is not driving the anti-smoking agenda. Rather, the anti-smoking agenda appears to be driving the interpretation of the science.

The shame of this is that I think it is going to hurt the credibility of legitimate tobacco control research conclusions. If the public and policy makers realize that we are drawing scientifically shaky conclusions with respect to this research, what is to prevent them from dismissing the results of studies we publicize where the conclusions truly are scientifically solid?

The rest of the story suggests to me that anti-smoking groups are more concerned with the direction of the results of scientific studies than with the validity of the study conclusions, and that they are letting the agenda drive their interpretation of the science, rather than demanding that the science be used to drive the agenda. Anyone familiar with my work knows that I highly value the particular agenda that is being promoted here - but I would never promote it myself on the kind of weak science that is being used to do so. And I think that hurts, rather than helps, the pursuit of some important public health goals.

No comments: