Monday, January 05, 2009

CDC: Pueblo Smoking Ban Reduced Heart Attacks by 41%, Due Mostly to Decreased Secondhand Smoke Exposure; But Conclusions are Biased and Invalid

In a new study published in the current issue of MMWR (Morbidity and Mortality Weekly Reports), researchers from Colorado and from the Centers for Disease Control and Prevention (CDC) have concluded that the smoking ban in Pueblo, Colorado caused a 41% reduction in heart attacks during the three years following its implementation, primarily due to a reduction in secondhand smoke exposure associated with the ban (see: Alsever RN, et al. Reduced Hospitalizations for Acute Myocardial Infarction After Implementation of a Smoke-Free Ordinance --- City of Pueblo, Colorado, 2002--2006. MMWR 2009; 57(51);1373-1377).

The study compared the rate of hospitalizations for acute myocardial infarction (heart attacks) in the city of Pueblo with similar rates in Pueblo county (outside of Pueblo) and El Paso county (which includes Colorado Springs) for the 18-month period prior to the implementation of Pueblo's smoking ban and for the two 18-month periods following the smoking ban, which was implemented in July 2003. While there was no significant reduction in heart attack admissions for Pueblo or El Paso counties, there was a reduction in the heart attack rate of 19% and 41% from pre-implementation to the first and second post-implementation periods, respectively, in the city of Pueblo.

The study concludes: "These findings suggest that smoke-free policies can result in reductions in AMI [acute myocardial infarction] hospitalizations that are sustained over a 3-year period and that these policies are important in preventing morbidity and mortality associated with heart disease. This effect likely is mediated through reduced SHS [secondhand smoke] exposure among nonsmokers and reduced smoking, with the former making the larger contribution."

The Rest of the Story

Before you jump to any conclusions here (something the study did prematurely), consider this: let's accept the study's conclusion as correct - that smoking bans do lead to a dramatic, immediate reduction in heart attacks, in part because of a large reduction in smoking prevalence. Let's suppose that you want to demonstrate this "fact" by showing that compared to a similar city, heart attack rates in the city with the smoking ban fell substantially more after the ban was implemented.

Now you have to choose a comparison city. You have two choices, with the following information available about the smoking prevalence changes in those cities from pre-implementation to post-implementation:

City A - The smoking prevalence increased from 19% to 24%.
City B - The smoking prevalence remained relatively unchanged, dropping only from 24% to 23%.

Which city would you choose as the comparison city?

If you choose city B, you would be justified. There was little change in smoking prevalence, which mirrored the changes nationally during that time period, so one could argue that this is a reasonable comparison group.

If you choose city A, where there was a large increase in smoking prevalence, you are going to expect to see an increase in heart attacks due to the rise in smoking alone. This is going to artificially reduce any secular decline in heart attacks occurring in the comparison city and bias your results towards finding a larger decline in heart attacks in the city with the smoking ban.

A researcher who chose city A as the comparison city would certainly be suspected of having intentionally biased the results towards finding an effect of the smoking ban on heart attacks.

The last thing in the world that you want for a comparison city is one in which there was actually an increase in smoking prevalence, defying all odds about what the national trends in smoking are throughout the nation.

Unfortunately, this is exactly what this study does: it knowingly uses a comparison county in which it has been documented that the smoking prevalence over the study period has increased from 17.4% to 22.3%.

The study doesn't try to hide this fact. It openly acknowledges that the reported smoking prevalence in El Paso County (the comparison group) increased from 17.4% in 2002-2003 to 22.3% in 2004-2005.

Given this finding, El Paso County simply cannot be used as a comparison population. You can't take a population in which you know that smoking prevalence increased substantially and "pretend" that it represents a reasonable area in which to evaluate the baseline secular trends in heart attack admission rates that would have occurred in the smoking ban city in the absence of the smoking ban.

Of course you are going to find that the rate of heart attacks in El Paso County did not decline all that much, given the increase in smoking. El Paso County is clearly not going to give you a good, representative picture of what the actual secular trend in heart attack admissions is.

Now if smoking rates throughout the country had increased substantially during the same time period, one could argue that El Paso county is representative of the nation as a whole, or of Colorado as a whole. But clearly, the trends in smoking reported in El Paso are an anomaly - they are very different from the rest of the nation and from Colorado, where we know that smoking has continued to decline during the study period.

While I am not arguing here that the study intentionally used El Paso county in order to try to create the finding of a smoking ban effect on heart attacks, the fact that the study failed to even consider this problem suggests to me that there is a great deal of bias inherent in the paper. Yes, I do think that the study wanted to find an effect of the smoking ban and that it lost its neutrality somewhere in the process. It's natural to want to see the positive effects of a public health policy. But you have to separate your desires from the science itself. More about that later.

Another important problem is the other comparison group that was used: the rest of Pueblo county. Since this area is directly adjacent to Pueblo, which is the one city in this area, it would be expected that many residents of Pueblo county work in, and/or spend time in Pueblo, including eating in restaurants in the city. Thus, one would expect that if the smoking ban reduced heart attack rates, it would reduce rates among Pueblo county residents as well. It's not like those residents were somehow shielded from the intervention.

For this reason, the study should have combined the heart attack admissions from Pueblo and Pueblo county. Doing this, the reduction in the heart attack rate from pre-implementation to the second post-implementation period is 33%, rather than 41%.

Two logical comparison groups that one would want to consider are the state of Colorado as a whole and the nation as a whole. Heart attack admission rates for Colorado during the approximate period of the study (2002-2005) dropped by 18.4%. For the United States as a whole, the heart attack admission rate dropped by 17.2% during this period.

It is quite a different situation to claim that the smoking ban in Pueblo reduced heart attacks by 41% (because there was no significant decline in the inappropriate comparison county of El Paso) than it is to view the whole picture, and see that a 33% decline in heart attacks in Pueblo must be compared with about an 18% drop throughout the state of Colorado and a 17% decline nationally during the same time period.

The fact that these comparisons were not made is problematic, since the data are readily available (it took me about half hour to access and run the numbers). Why wouldn't the study want to look at the statewide trends in Colorado, rather than simply rely on the biased control group of El Paso county? In 30 minutes, the study could have determined that there was an impressive 18% decline in heart attacks in the whole state during the study period, thus making it clear that the present conclusion of the study is inaccurate.

The bottom line is that the study fails to appropriately determine the baseline secular trends in heart attacks in order to be able to judge the differences observed in Pueblo from the trends that would have been expected in the absence of the smoking ban. For this reason, the study cannot conclude that the observed changes in heart attacks are due to the smoking ban, rather than to other changes that took place over time, including changes in medications being used to treat heart disease, better diagnosis and more aggressive treatment of heart disease, and a substantial decline in smoking prevalence in Pueblo county during the study period, which may or may not be due to the smoking ban itself.

More troubling to me than the fact that the study draws a conclusion that is premature and inadequately supported by the data is the appearance of bias in the study. Not only in the choice of a comparison community where smoking prevalence dramatically increased during the study period, but also in the conclusion itself.

Even if we stipulate that the smoking ban did cause the decline in heart attacks, how can the study possibly conclude that the effect was due primarily to reduced secondhand smoke exposure? The study made no attempt to determine the smoking status of the heart attack victims, so there is no evidence that the reduction in heart attacks occurred primarily among nonsmokers. Neither did the study measure changes in population-based exposure to secondhand smoke.

Moreover, the study itself documents that there was a substantial decline in smoking prevalence in Pueblo county during the study period, from 25.9% to 20.6%. Wouldn't this documented decline in active smoking prevalence be the presumed major reason for the observed decline in heart attacks, as opposed to reductions in secondhand smoke exposure? At very least, wouldn't a study simply remark that both mechanisms may be operating, but that it can't be determined to what extent each is contributing?

The fact that the study concludes that it must primarily be the secondhand smoke reduction is curious. The fact that the editorial note of the study begins by claiming that evidence shows that brief secondhand smoke exposure can trigger a heart attack is revealing. If you look at the report to which that claim refers (the 2006 Surgeon General's report), you will not find any conclusion that brief secondhand smoke exposure triggers heart attacks. And you certainly won't find any evidence in that report that if we reduce secondhand smoke exposure, we can reduce heart attacks triggered by secondhand smoke exposure.

You may remember that I have previously called attention to the poor science by CDC and the Department of Health and Human Services in their communications regarding the acute cardiovascular effects of secondhand smoke, when they went out on a limb, against the advice of respected and expert scientists in the tobacco control field, and told the public that brief secondhand smoke exposure is enough to trigger heart attacks, cause heart disease, and cause lung cancer.

It seems odd that even if we stipulate that the overall conclusion of the study is valid (that the smoking ban caused a dramatic reduction in heart attacks in Pueblo), the study would emphasize that the effect must be primarily due to the reduction in secondhand smoke and thus a reduction in heart attacks among nonsmokers that would have otherwise been triggered by brief secondhand smoke exposures in restaurants or other public places.

Even if I were writing this editorial as a highly biased advocate, I would have simply concluded that the effect is likely due to the combination of a reduction in smoking prevalence and a reduction in secondhand smoke, but that the study provides no way of teasing out the degree to which these two phemomena are operating.

In fact, given the large decline in smoking prevalence reported in Pueblo county, even the above conclusion seems biased, since it is clear that if the effect were real, the smoking prevalence reduction would likely have been a major reason.

The study goes overboard not only in its overall conclusion, but in its attempt to paint these data as somehow proving that eating in a smoky restaurant for a half hour is causing lots of people to keel over from heart attacks. The study does nothing of the sort.

Let me finish by emphasizing that I would like nothing more than to have strong evidence presented that smoking bans are resulting in immediate and dramatic reductions in heart attacks. As I have devoted much of my life's work to promoting smoking bans, especially in bars and restaurants, it would bring a great sense of fulfillment to now that these policies are immediately saving lives and that we can document these acute effects.

However, I am first a scientist and I believe that in public health, our conclusions must be based on solid science, not just on conjecture or our deeply felt desire to see the success of our policies.

No comments: