Wednesday, September 21, 2016

Helena Miracle? Not So Much; New Study Casts Doubt on Conclusions of Anti-Smoking Groups

For many years, anti-smoking groups have been claiming that smoking bans result in an immediate and dramatic decline in hospitalizations for heart attacks. This claim is based on studies like that conducted in Helena, in which the authors concluded that there was a 40% reduction in heart attack admissions within 6-18 months after a bar and restaurant smoking ban was implemented.

When these studies were first published, I warned anti-smoking groups not to use these conclusions to promote smoking bans because I believed that the conclusions were not adequately supported by the data. In particular, I criticized these studies and questioned their conclusions because they did not adequately account for secular trends in heart attack rates that were occurring even in the absence of smoking bans.

I also argued that it was not plausible to see such large effects in so short a time span because it takes many years for heart disease to develop. In contrast, I noted that respiratory effects might be observed immediately.

Finally, I argued that these findings were largely a result of publication bias. Cities for study were not chosen randomly. Instead, in places where there appeared to be a dramatic decline in heart attacks, that community was chosen for study. The few studies that did systematically examine this research question failed to find an immediate effect of smoking bans on heart attack hospitalization rates.

This week, a new study was published in the journal Medical Care Research and Review which re-examines the relationship between smoking bans and heart attack hospitalization rates.

(See: Ho V, et al. A nationwide assessment of the association of smoking bans and cigarette taxes with hospitalizations for acute myocardial infarction, heart failure, and pneumonia. Medical Care Research and Review 2016. Published online ahead of print on September 12, 2016. DOI: 10.1177/10775587/16668646.)

The authors summarize the study as follows:

"We examine the association between county-level smoking-related hospitalization rates and comprehensive smoking bans in 28 states from 2001 to 2008. Differences-in-differences analysis measures changes in hospitalization rates before versus after introducing bans in bars, restaurants, and workplaces, controlling for cigarette taxes, adjusting for local health and provider characteristics. Smoking bans were not associated with acute myocardial infarction or heart failure hospitalizations, but lowered pneumonia hospitalization rates for persons ages 60 to 74 years."

The Rest of the Story

The results of this study support the arguments that I made back in 2006, and which I have continued to emphasize on this blog over the past 10 years. The key finding of the study is that once you account for time-specific local trends in heart attack hospitalization rates, there is no longer a significant observed decline in heart attack rates associated with the implementation of these smoking bans.

The key data are in Table 2. Helpfully, this table provides regression estimates both with and without the inclusion of region-specific (in this case, county-specific) time trends in the rate of hospitalization for heart attacks. Note that without the county-specific time trends, the authors do find a significant decrease in hospitalizations for heart attacks associated with the implementation of smoking bans. But once they control for these local time trends, there is no longer a significant effect.

What this suggests is that communities that implemented smoking bans happened to be those which had a higher pre-existing rate of decline in heart attacks and that once you adjust for that, there is no difference associated with the implementation of the smoking bans.

The authors further demonstrate that the previous conclusions were spurious by showing that when you examine the impact of smoking bans on hospitalizations for hip fractures (which are not plausibly related to smoking bans), you find a significant effect of smoking bans when you don't include regional time trends. That spurious effect disappears once you control for the regional time trends.

The regional time trends are important to consider in these models because without them, you are making the assumption that the secular rates of decline in heart attacks at baseline were the same in communities with and without smoking bans. Apparently, this is not a fair assumption because the nature of cities that enacted smoking bans differed from that of cities which did not enact such laws. And those differences were reflected in higher rates of decline in heart attacks in the intervention communities.

It is interesting to note that it was my expression of the above opinions about these studies back in the mid-2000's that led to my "expulsion" from the tobacco control movement, including being thrown off several list-serves, ostracized by many of my colleagues, accused of being a "tobacco mole," being characterized by my hero and mentor - Stan Glantz - as being "a tragic figure," having copyright to one of my articles violated by an anti-smoking organization, no longer being invited to speak at tobacco conferences, not being able to present at tobacco control conferences anymore, not being able to obtain further research grants, and having colleagues refuse to appear with me at conferences to discuss these or any other scientific issues. In fact, it was this censorship that led to the creation of the Rest of the Story in the first place.

Nearly three million page views later, perhaps these groups knew what they were doing because it appears that I may have been right all along. By silencing me, these groups were able to disseminate their pre-determined conclusions widely to the public through the media long enough for the conclusions to be generally accepted. Now, it is too late to undo the damage. The media and the public have already made up their minds, and one article noting the results of this new study is not going to correct or undo 10 years of dissemination of unsupported and errant scientific conclusions.

To be clear, this issue has nothing to do with my support for smoking bans. My readers understand that I have and continue to support 100% smoke-free policies for all workplaces (including bars, restaurants, and casinos). However, I do believe that we can advocate for such policies without misrepresenting scientific evidence or exaggerating findings beyond what is scientifically defensible.

No comments: