Last week, I reported on a study, by Dr. Robin Mathews of the Duke Clinical Research Institute, that examined rates of heart attacks among persons ages 65 and older in 74 cities across the U.S. which adopted strong smoking bans during the period 2000-2008. The researchers compared the heart attack rate in each city during the year before and after the smoking ban was implemented.
When Mathews included all 74 cities that enacted smoking bans during the study period (regardless of strength of the ordinance), he found an overall decline in heart attack rates of just 3%. However, when the analysis was restricted to the 43 cities whose newly enacted ordinances represented a significant increase in protection from secondhand smoke, Mathews reports that there was absolutely no change in the heart attack rates across the sample of cities. A figure shows that heart attacks decreased in some cities and increased in others. All told, heart attack rates decreased by an insignificant 1% among these 43 cities.
Based on these findings, I pointed out that the study fails to support the conclusion - being widely disseminated by anti-smoking groups - that smoking bans result in dramatic, immediate reductions in hospital admissions for acute myocardial infarction. The Institute of Medicine report, for example, cites a reduction of 17% in heart attacks across the studies that it evaluated.
Now, a number of anti-smoking researchers, presumably in response to my blog post, have argued that the Mathews study actually supports the prior research, demonstrating that while small (3%), there is a "significant effect" of smoking bans on heart attacks within one year of implementation of these laws.
The Rest of the Story
There are two reasons why the argument disseminated by these anti-smoking advocates is scientifically flawed.
1. The argument ignores the relevant analysis, relying on the result of an essentially meaningless analysis.
First, it ignores a major result of the study, choosing instead the one result which found a significant effect but ignoring what is actually the most relevant analysis, which failed to find an effect.
The analysis of all 74 cities, which did find a significant decline in heart attacks of 3%, included cities that had enacted ordinances that failed to "meaningfully" increase protection from secondhand smoke. Thus, these cities were misclassified. Rather than being in the intervention group, they should have been considered to be communities that did not increase their secondhand smoke protection.
The relevant analysis is the analysis of the 43 cities in which the level of protection from secondhand smoke meaningfully increased after implementation of the smoking ordinance. It was that analysis which found only a 1% (non-significant) decline in heart attacks.
If an ordinance provides no meaningful increase in protection from secondhand smoke, then why would one classify it as an intervention community?
2. The study lacks a control or comparison group.
Second, because the study lacks a control group, all one can conclude from it is the change in heart attacks from before to after the ordinances. One cannot infer that this change was due to the ordinance because there is no evidence that these declines exceed those that were occurring simultaneously in communities without smoking bans.
Even if we accept the results of the full analysis of 74 cities (a 3% decline in heart attacks overall in these cities), the only way one could credibly conclude that this 3% decline was due to the smoking bans is to compare this decline to that in communities without smoking bans and show that the heart attack decline in the intervention (i.e., smoking ban) communities was significantly greater.
The rest of the story is that heart attack rates during the study period declined substantially throughout the United States, even in communities without smoking bans, and the average annual rate of decline in heart attack admissions among 65+ year olds during the study period appears to be in the approximate range of 3% per year.
Based on data from the Health Care Utilization Project, I calculated the year-to-year declines in heart attack admissions among persons ages 65 and up nationally for the years 2000 to 2006 (more recent data are not yet available on admissions). The average annual decline in admissions was 3.4%.
Thus, the observed decline in heart attacks in the 74 communities that enacted smoking bans during the study period appears to be a little lower than the decline in heart attacks that occurred nationally, with or without smoking bans in place.
To demonstrate how "meaningless" the observed 3% decline in heart attacks is in the context of a control group (the United States as a whole), the annual decline in heart attack admissions among 65+ year-olds during the more recent years was as follows:
Without even seeing these data, the anti-smoking researchers have nevertheless concluded that a mere 3% reduction in heart attacks in a one-year period is a significantly greater decline than what was occurring nationally anyway based on secular trends, in the absence of a smoking ordinance.
Of course, what we really need as a comparison group are cities that are similar to those which enacted smoking bans in other ways, but did not themselves enact smoking bans. The national data include cities with and without smoking bans.
My point is simply that without the comparison data from communities that did not enact smoking bans, one cannot possibly conclude that the observed 3% decline in heart attacks in the cities that enacted smoking bans was greater than what would have occurred in the absence of these bans.
The Mathews study certainly cannot be used to support the contention that smoking bans result in a significant short-term decline in heart attacks. At best, all it can support is the conclusion that in the cities which enacted smoking bans, there was an overall average decline of 3% in heart attacks during the first year of implementation. This 3% decline could be more than, less than, or the same as what would have occurred in the absence of the smoking bans (i.e., the observed change in heart attacks during the same period of time in comparison cities without smoking bans).
While I am a passionate supporter and promoter of smoking bans, I do my best to try not to let that interfere with my interpretation of the scientific evidence. As is hopefully apparent to regular readers, I will report findings to the public, even if they are "unfavorable" to us in tobacco control.
I want to close by emphasizing that in the long-term, smoking bans will almost certainly reduce heart disease (and therefore heart attacks). The question is whether this effect can and will happen immediately (within one year). Even if we fail to see an effect on heart attacks within one year, this does not argue against the enactment of smoking bans. In fact, even if smoking bans had no long-term effect on heart disease, the respiratory effects of secondhand smoke alone would be enough to support protecting the public from secondhand smoke exposure.
I do hope, however, that we in tobacco control will base our support of smoke-free policies on credible scientific evidence, rather than on a study like the Mathews one which lacks a control group and makes the conclusion being disseminated by these researchers an unsupportable one.