A new study, published online ahead of print in the American Journal of Public Health, concludes that the Massachusetts statewide smoking ban, implemented in July 2004, caused a 7.4% reduction in the heart attack mortality rate in the state (see: Dove et al. The impact of Massachusetts' smoke-free workplace laws on acute myocardial infarction deaths. American Journal of Public Health 2010).
The study examined trends in the mortality rates from myocardial infarction (heart attacks) from 1999 through 2006. Using a regression model, the authors estimated the change in heart attack rate following the statewide smoking ban, controlling for the baseline trend. The major finding of the study was that the overall heart attack mortality rate decreased by an estimated 7.4% in the two years following the statewide smoking ban. The study concludes that this decline was attributable to the smoking ban.
The Rest of the Story
The rest of the story is that without a comparison group, it is not possible to conclude that the observed decline in heart attack mortality in Massachusetts from 2004 to 2006 was attributable to the statewide smoking ban, as opposed to some other factor, such as changes in the treatment of severe heart disease and increased use of statin drugs to control cholesterol.
While this study demonstrates quite convincingly that the rate of heart attack death in Massachusetts declined significantly from 2004 to 2006, might it be possible that heart attack mortality was declining significantly during that time period anyway, even without the smoking ban?
We can't answer that question unless we examine heart attack mortality data from some comparison group. Since the study lacks any comparison group, I do not believe it can draw the conclusion that the observed reduction in heart attack mortality was due to the smoking ban, rather than to other factors, such as the tremendous advances in treatment of severe heart disease, better control of hypertension and hypercholesterolemia, and reductions in other risk factors such as smoking.
To evaluate this issue, we can examine national heart attack mortality data during the same period (2004 to 2006) to find out what was occurring in the nation as a whole.
From 2004 to 2006, the age-adjusted acute myocardial infarction mortality rate in the United States dropped from 52.2 to 45.0, a decline of 13.8%. Thus, even without the smoking ban, one would have expected the heart attack mortality rate in Massachusetts to drop by 13.8% during the same time period examined in the study. In light of this rather dramatic nationwide decline in heart attack mortality, the Massachusetts data do not appear as impressive as when they are viewed in isolation.
The study actually found no significant change in the heart attack mortality rate during the first year in which the ban was in effect and then a large decline in the second year. If you look at similar rates for the nation as a whole, the decline in heart attack mortality rate from 2004 to 2005 was 5.9%, but from 2005 to 2006, the rate accelerated to 8.4%. Controlling for the baseline trend in heart attack deaths from 2004 to 2005, the decline in 2005 to 2006 nationally represented a drop of 5.1% in the mortality rate. Thus, the observed decline in Massachusetts of 7.4% no longer looks so impressive.
My point here is not that the Massachusetts statewide smoking ban had no effect on heart attack deaths. My point is simply that we cannot determine whether or not the smoking ban was the cause of the observed decline because we have no comparison group. One cannot rule out as an alternative explanation for the study findings the possibility that secular changes were occurring anyway by which there was an accelerated decline in heart attacks during the study period.
To demonstrate the dangers of drawing causal conclusions from a study such as this one, suppose we just look at the data from the first year after the smoking ban. In the United States as a whole, heart attack deaths dropped by 5.9% from 2004 to 2005. But in Massachusetts, heart attack deaths dropped by only 1.6%. Does this mean that the statewide smoking ban led to an increase in heart attack deaths in Massachusetts during its first year of implementation?
Might it not instead be the case that what we are seeing here is simply the quirks of a short follow-up period, where the true decline in heart attacks was relatively stable, but the data happen to show very little decline during the first year and then a huge decline in the second year? If you put too much stock in any one year of data, then you would be forced to conclude that the smoking ban actually led to a deceleration in the decline in heart attack deaths during its first year.
Interpreting this type of data is tricky and I believe that this paper goes too far by drawing a causal conclusion. The paper goes so far as to estimate the exact number of heart attack deaths averted because of the smoking ban. I believe this is over-stating what the data actually show.
There is no question that the heart attack death rate in Massachusetts declined significantly following the statewide smoking ban in 2004. But the same can be said of the heart attack death rate in the United States as a whole, in the absence of any nationwide smoking ban.
Finally, I would point out that the entire basis for the conclusion in this paper is the last four months of data. To me, it would make more sense to at least wait until the 2007 data are available before drawing any definitive conclusions. The 2007 data will tell us whether or not the marked decline in heart attack deaths in the later half of 2006 was a real change in the trend, or just an anomaly.
While I would love to be able to conclude that the statewide smoking ban in Massachusetts caused the observed decline in heart attack mortality, I do not feel that the data presented in this study allow such a conclusion to be drawn. Without any comparison group, it is impossible to draw conclusions about what would have been expected in the absence of the statewide smoking ban. I don't see how one can draw such a conclusion without at very least examining the secular trends in heart attack mortality occurring elsewhere.
No comments:
Post a Comment