Friday, July 13, 2007

Study of Trends in State Heart Attack Admissions Refutes Conclusions of Helena et al. Studies

A study of trends in annual heart attack admissions through the emergency room by state refutes the conclusions of the Helena, Pueblo, Piedmont, and Bowling Green studies which concluded that smoking bans cause immediate, drastic declines in heart attacks. The study was conducted by Michael J. McFadden and David W. Kuneman. A preliminary version of the study appeared online in November 2005, while the final version, discussed on the American Council on Science and Health (ACSH) FactsAndFears blog, confirmed and strengthened the original findings.

McFadden and Kuneman overcame the two chief limitations of the existing research: (1) the small number of heart attacks in the cities studied, which resulted in tremendous baseline year-to-year variability in heart attack rates; and (2) the bias inherent in relying only upon the experiences of only a few small cities in making generalizations intended to apply to larger populations. They accomplished this by considering the following premise: if smoking bans cause an immediate decrease in heart attack admissions on the order of a 25% to 50% decline, then if one examines trends in heart attack admissions on an entire state level, then one should certainly observe a notable decline immediately following the implementation of statewide smoking bans.

Using data on emergency room admissions for acute myocardial infarction (heart attacks) from the Agency for Healthcare Research and Quality's HCUP database, McFadden and Kuneman examined trends in four states that implemented statewide bar and/or restaurant smoking bans - California, Oregon, Florida, and New York - and five states with neither a statewide smoking ban nor widespread local smoking bans - Arizona, New Jersey, South Carolina, and Iowa. They also examined trends in heart attack admissions for the United States as a whole.

California's restaurant smoking ban took effect in January 1995. In that year, heart attack admissions in California increased by 0.6%. In 1996, admissions increased by 2.9%. The corresponding changes for the United States were an increase of 3.2% in 1995 and 3.9% in 1996. California's bar smoking ban took effect in January 1998. In that year, heart attack admissions in California increased by 6.0%, compared to an increase of 6.2% in the nation as a whole. In 1999, admissions increased by 3.7% in California and decreased by 1.3% in the U.S.

Florida banned smoking in all restaurants in July 2003. Its heart attack admissions decreased by only 0.7% that year, and by only 2.0% the following year. In comparison, admissions in the U.S. decreased by 2.8% in 2003 and 8.2% in 2004.

Trends in heart attack admissions following the Oregon and New York smoking bans were also not found to be substantially different from national trends, or from trends in the comparison states without smoking bans.

The Rest of the Story

While this study certainly does not prove that smoking bans have no effect on heart attack admissions, what it does is demonstrate that when one examines population-based data for an entire state, one does not find any evidence of a dramatic decline in heart attacks immediately following the implementation of smoking bans. This casts serious doubt on the conclusion of the Helena, Pueblo, Piedmont, and Bowling Green studies. If smoking bans truly cause an immediate and dramatic decline in heart attacks, on the order of a 25% to 50% reduction, then why do we not observe any evident decline in heart attacks when entire states implement smoking bans.

The chief limitation of the study is that some localities within these states had already enacted smoking bans, so one would not expect to see as dramatic an effect on a statewide level. However, the proportion of residents in these states covered by local smoking bans was not particularly high. For example, the authors cite data that only 14% of Californians were covered by a smoke-free restaurant law in 1993. This increased to 100% in 1995. Yet the number of heart attack admissions increased from 1993 to 1995 (and increased further in 1996). The corresponding changes in the U.S. data were similar. So it does not seem that the existence of some local smoking bans is enough to invalidate the study conclusion.

I have already discussed in detail the reasons why I believe the conclusions of the Helena et al. studies are unjustified. Briefly, the chief flaw of these studies is that they are unable to rule out the very likely possibility that the observed changes in heart disease admissions in these cities during the study period are due primarily to random variation, rather than to the smoking bans. There is tremendous natural (random) variation in the heart disease admission rates in these cities because of the small population sizes. Because we are dealing with such small numbers of admissions, the percentage change in admissions from one year to the next is very high, even without any smoking ban.

While these papers make some attempt to account for these baseline trends, I do not believe that they go back far enough in time to do so adequately.

Because McFadden and Kuneman have examined heart attack trends among large populations, the number of heart attacks is much higher, and therefore the degree of underlying variability in the annual number of heart attacks is much lower. It is therefore easier to identify any major changes in the underlying trends.

I think it is high time that my fellow tobacco control researchers and practitioners recognize that the Helena et al. studies are examples of shoddy science that apparently now passes as acceptable in tobacco control research. While I support workplace smoking bans, I do not believe that we should be using shoddy science to promote them.

No comments: