Thursday, February 14, 2008

New Study Concludes that Italian Smoking Ban Reduced Heart Attack Rates Within First Year

An article published online ahead of print in the journal Circulation concludes that the ban on smoking in public places in Italy has resulted in an 8%-11% decline in acute coronary events (i.e., heart attacks) in the first year following implementation of the ban (see: Cesaroni G, Forastiere F, Agabiti N, et al. Effect of the Italian Smoking Ban on Population Rates of Acute Coronary Events. Circulation 2008).

The study compared the annual age-standardized rates of acute coronary events among adults in three different age groups during the five years preceding the smoking ban (2000-2004) to the rates during the first year the ban was in effect (2005). The post-smoking ban acute coronary event rate was 11% lower than the pre-ban rate for adults ages 35-64 and 8% lower for adults ages 65-74. There was no significant difference in the pre- and post-ban rates for adults ages 75-84.

The authors conclude that the observed reduction in the heart attack rate in 2005 among 35-74 year-olds was atributable to the smoking ban. Furthermore, the authors conclude that at least a portion of the effect is due to a reduction in secondhand smoke exposure among nonsmokers.

One study author stated: "Smoking bans in all public and workplaces result in an important reduction of acute coronary events."

These conclusions have been widely reported. For example, this article was headlined: "Public smoking ban decreases acute coronary events in Italy."

Results of this and related studies (with similar methodology) are being used by anti-smoking advocates in testimony before policy makers that smoking bans have an immediate effect on heart attack rates.

The Rest of the Story

There's just one problem with all of this: the conclusions of this study are not supported by the data. The data clearly show that the decline in heart attack rates among adults in these two age groups began prior to the implementation of the smoking ban. Thus, it is evident that the decline is not attributable to the smoking ban.

Take a look at the data for yourself:

A. Age 65-74

From 2003 to 2004 (prior to the smoking ban), the heart attack rate declined from 7.86 to 7.39, a drop of 6.0%.

From 2004 to 2005 (first year of the smoking ban), the heart attack rate declined from 7.39 to 6.95, a drop of 6.0%.

In other words, the decline in the heart attack rate from 2003 to 2004 was exactly the same as the decline from 2004 to 2005.

These data clearly do not support the conclusion that the smoking ban resulted in a sudden drop in the heart attack rate. Instead, these data document that the decline in the heart attack rate in this age group was exactly the same post-ban as it was pre-ban.

In light of these data, I find it impossible and highly invalid to conclude that the smoking ban resulted in the observed decline in heart attacks from 2004 to 2005. A more likely, and certainly plausible, explanation is that there was already a trend of declining heart attack rates and that this trend simply continued from 2004 to 2005.

The precise reasoning being used by this paper could equally be used to argue that the smoking ban had no effect on heart attacks. I'm not making such a claim, but I am suggesting that these data by no means support the conclusion of the article. In fact, they refute such a conclusion.

B. Age 35-64

From 2002 to 2003 (prior to the smoking ban), the heart attack rate declined from 2.13 to 1.95, a drop of 8.5%.

From 2004 to 2005 (the first year of the smoking ban), the heart attack rate declined from 1.92 to 1.80, a drop of 6.3%.

In other words, the observed decline in the heart attack rate one year prior to the smoking ban was actually greater than the decline in the heart attack rate after the smoking ban.

The average decline in the heart attack rate for the two year period preceding the smoking ban (2002 to 2004) was 4.9%.

The decline in the heart attack rate for the first year following the smoking ban was 6.4%.

Thus, one can see that the decline in the heart attack rate in this age group after the smoking ban was comparable to the decline in the heart attack rate in this age group before the smoking ban.

Once again, these data clearly do not support the conclusion that the smoking ban resulted in a sudden drop in the heart attack rate. If anything, these data document that the decline in the heart attack rate in this age group was about the same post-ban as it was pre-ban.

In light of these data, I find it impossible and highly invalid to conclude that the smoking ban resulted in the observed decline in heart attacks from 2004 to 2005. A more likely, and certainly plausible, explanation is that there was already a trend of declining heart attack rates and that this trend simply continued from 2004 to 2005.

The precise reasoning being used by this paper could equally be used to argue that the smoking ban had no effect on heart attacks. Again, I'm not making such a claim, but I am suggesting that these data by no means support the conclusion of the article. In fact, they refute such a conclusion.

What is so alarming about the conclusion of this Italian smoking ban study is not so much that the authors have drawn a conclusion that follows from the data but have failed to consider alternative explanations for the cause of the decline in heart attack rates. What is so alarming is that they have drawn a conclusion that is completely unsupported by the data itself.

In other words, what appears to be operating here is an investigator bias, by which the authors seem to have been determined to find an effect even if one did not exist. It appears that this bias probably skewed their judgment in evaluating these data.

If you want to see what I mean, plot the heart attack rates for yourself on a graph. I used Microsoft Excel to do this and it took just a few minutes.

Then, examine the trend in heart attack rates among the age groups visually. For the 35-64 year-old group, you will readily see that there is literally a straight line from 2002 to 2005. There is an apparent decline in the heart attack rate that has remained relatively stable during the past few years. But that decline started in 2002, not in 2004.

Similarly, if you look at the graph for the 65-74 year-old group, you will see that there is literally a straight line from 2003 to 2005. There is indeed a trend of a declining heart attack rate, but that decline starts in 2003.

The appropriate conclusion from these data is that there is indeed a significant decline in the heart attack rates among 35-74 year-olds, but that this decline preceded the implementation of the smoking ban. The smoking ban certainly does not seem to have done anything to alter the existing observed declines.

This does not prove that the smoking ban had no effect. But what it does is indicate that the conclusion of the study is completely invalid.

There is another serious problem which, independently of the improperly interpreted data, renders the study conclusion invalid. That problem is simple: there is no comparison group. Without a comparison group, it is impossible to know whether the observed changes in heart attack rates in Italy are unique to Italy or whether they mirror similar secular changes occurring elsewhere.

We happen to know, for example, that in the United States, there have been substantial declines in heart attack rates - even in states without smoking bans - during the same time period. There is no way that the study can conclude that the observed decline in Italy is not simply a reflection of the underlying secular trend of declining coronary event rates that is occurring anyway, even in the absence of the smoking ban.

What I find highly intriguing is the apparent bias that is plaguing the interpretation of these smoking ban/heart attack studies, from Helena to Pueblo to Saskatoon to Bowling Green to Ireland to Scotland to Piedmont to Rome.

What I think this shows is that if one wants strongly enough to be able to find an effect of something, that can skew ones interpretation and analysis of the data such that one is able to do so. That is somewhat reasonable, since there is always going to be an inherent bias in any scientific study.

But what is not so reasonable to me is that these inherent biases and the skewed and invalid interpretation of the data are not being picked up by journals. It suggests that perhaps the peer reviewers who are being chosen to review these studies are affected by the same bias. If these papers are not being reviewed by scientists who are neutral, then those reviewers are far less likely to pick up the glaring flaws in these study conclusions. It certainly appears that may have been what happened in the case of Rome.

No comments: