Tuesday, April 13, 2010

Study Attributes Fall in Cardiovascular Admissions in Toronto to Smoking Ban and Decrease in Secondhand Smoke Exposure, But Methods are Faulty

Yet another study with severely faulty methods has concluded that an observed decline in cardiovascular disease admissions is attributable to a smoking ban. This time, the place is Toronto and the study was published yesterday online ahead of print in the Canadian Medical Association Journal.

The study examined trends in hospital admissions for cardiovascular and respiratory disease in Toronto from 1996 to 2006. A smoking ban in restaurants (excluding designated rooms) was implemented in 2001 and a smoking ban in bars (excluding designated rooms) was implemented in 2004. Two comparison cities - Durham Region and Thunder Bay - were used to control for secular trends in cardiovascular and respiratory disease.

The study findings were reported as follows: "Crude rates of admission to hospital because of cardiovascular conditions decreased by 39% (95% CI 38%–40%) and admissions because of respiratory conditions decreased by 33% (95% CI 32%–34%) during the ban period affecting restaurant settings. No consistent reductions in these rates were evident after smoking bans affecting other settings."

The study concludes that the observed declines in cardiovascular and respiratory hospital admissions were attributable to the smoking ban.

The paper notes that "the ecological design of this study prevents the delineation of individual smoking status and other individual-level information, such as actual exposure to second-hand smoke and other risk factors (e.g., prior acute myocardial infarction and other comorbidities).
This methodological limitation makes it possible that our observed decreases in both cardiovascular- and respiratory related admissions to hospital were the result of active
smokers smoking less or quitting."

Nevertheless, this limitation does not stop the authors from concluding: "that the reduction in hospital admissions observed in our study is unlikely to have resulted from decreased rates of smoking." In other words, they are attributing the declines in cardiovascular admissions to decreased secondhand smoke exposure due to the smoking ban in restaurants and bars.

The Rest of the Story

There are huge problems with this study. While it is plausible that a smoking ban would result in an immediate decline in respiratory disease admissions, the observed declines in cardiovascular disease admissions are not plausibly related to decreased secondhand smoke exposure associated with a bar and/or restaurant smoking ban. This commentary will focus on the study's conclusions regarding the reduction in cardiovascular disease admissions.

The first problem is that an examination of the trend in cardiovascular disease admissions in the control communities demonstrates that there was a dramatic decline during the study period, completely negating the study's primary conclusion.

While the data for the comparison cities is not presented in the paper, it is provided in an online appendix. If you look at the data on angina admissions in Durham Region, you'll notice that between 2001 and 2004, there was a whopping 53% decline in angina admission rates. The paper discounts any declines in cardiovascular disease in the control cities by stating that there were "no consistent reductions." However, this is a highly biased presentation of the data. There is a very striking 53% decline in angina admissions in Durham Region, and such a finding completely negates the study conclusion, because it demonstrates that declines in cardiovascular disease admissions of an even greater magnitude than observed in Toronto occurred in this comparison city during the study period.

To see why I would argue that the paper's presentation of the findings and its conclusions are highly biased, note that the paper does not even mention this whopping 53% decline in angina admission rates. It is literally buried. Ignored. The same reasoning used by this paper to conclude that the smoking ban led to the observed decline in angina in Toronto would also argue that the 53% decline in angina admissions in Durham Region was attributable to the absence of a smoking ban. Clearly, the paper's conclusion is simply not supported by the actual data.

Also ignored is the fact that heart attack rates in Durham Region were on the rise during the baseline period, thus making it inappropriate to compare changes in rates of heart attacks in this city with those in Toronto.

Also ignored: stroke admissions in Durham Region declined by a whopping 47% from 2000-2006.

And if this were not enough to invalidate the study conclusions, look at the trends in cardiovascular disease admissions in Thunder Bay. From 2000 to 2006, there was a whopping 43% decline in angina admissions. Furthermore, there was a 28% decline in heart attack admissions in Thunder Bay from 2003 to 2006.

Frankly, in the midst of these huge declines in cardiovascular disease admissions in both of the comparison cities, I simply cannot see how the paper can possibly mislead readers into thinking that there were no declines in cardiovascular disease in these control areas. An examination of the actual data reveals that the paper is hiding the key fact that huge secular declines in cardiovascular disease admissions were occurring anyway in the control cities, even in the absence of a smoking ban. This completely negates the study's conclusion that the observed decline in Toronto's admissions for cardiovascular disease are attributable to its smoking ban.

While the finding that there were also striking declines in cardiovascular disease admissions in these control communities does not prove that the changes in Toronto were independent of the smoking ban, it does prevent any reasonable conclusion that the decrease in cardiovascular disease admissions is attributable to the smoking ban. Again, I can't understand how one could obscure the clearly evident and quite massive declines in both of these comparison cities by stating that there are no "consistent declines." That is obviously a very subjective term as we don't know exactly what consistent is. But what it appears to mean is that the study authors didn't like what they saw when they examined the control city trends because it didn't lend itself to a clean conclusion, so they appeared to have obfuscated the findings by simply casting them aside under the presumption that the pattern was not consistent.

Perhaps the pattern is not entirely consistent, but it certainly is evident.

We hardly need to go further because this first problem completely negates the study findings and conclusion, but I will point out a second major problem with the study. The comparison cities are very different from Toronto and are not the least bit comparable. Most striking is the difference in smoking prevalence near the end of the study period (2005), which was only 18.1% in Toronto but 26.6% in Thunder Bay and 24.8% in Durham Region. This suggests that declines in smoking observed in Toronto did not occur in these other areas. It makes a decline in smoking in Toronto a likely explanation for any true differences in the cardiovascular disease trends. In the face of this difference in smoking prevalence, I don't see how one can conclude that reduction in secondhand smoke exposure is the most likely cause for the purported difference in cardiovascular disease trends.

The rest of the story is that we have yet another study purporting to show a dramatic and immediate effect of smoking bans on acute cardiovascular event admissions which turns out to be severely flawed. In this case, the data presented in the paper (or at least in the appendix) not only fail to support the study conclusion, but they actually refute that conclusion. They also refute the findings themselves as presented in the paper.

If anyone out there can look at the Appendix and tell me that they see no evident reduction in cardiovascular disease admissions in either of the comparison cities, I'd like to know how you came to such an impression. I believe the actual data is quite clear in showing that this study failed to accurately present the findings and therefore resulted in an unsupported and unjustified conclusion.

While I strongly support smoking bans (and would actually go further than Toronto did by arguing that it makes no sense to allow designated smoking rooms), I do not support the sloppy or biased interpretation of scientific data to try to support a particular conclusion. I would love nothing more than to see that smoking bans are reducing acute coronary events within a couple of years of their implementation. However, I will not disseminate such conclusions unless they are fully supported by the underlying scientific evidence.

No comments: