Friday, December 02, 2005

What A Scientific Response (Rather than a Personal Attack) to My Helena/Pueblo Analysis Might Look Like

Dr. Siegel:
I read with interest your analysis which failed to show any immediate, substantial decline in hospital admissions for heart attacks in states that implemented smoke-free restaurant and/or bar smoking bans. The scientific evaluation of this issue is of great interest to me since as a public health practitioner, I feel it is important to be very careful about the claims that we make to the public and to be sure that the data fully supports our claims before we disseminate them widely to the media.

Unfortunately, I do not find your argument to be compelling for the following reason. Generally, when states implement smoking bans, they are not going from a state of widespread secondhand smoke exposure in bars and restaurants to a state of no exposure. The usual pattern is that a large number of cities and towns within those states first enact smoke-free laws; this occurs gradually over a period of time and only after years of local enactment of policies is a statewide ban put in place.

This is certainly the pattern in California, where localities began passing smoke-free restaurant and bar laws in large numbers during the late 1980s and early 1990s - a pattern which continued until the ultimate statewide implementation of smoking bans in restaurants in 1995 and in bars in 1998.

For this reason, one might not expect to see the dramatic decline in acute cardiovascular morbidity that was observed in Helena and Pueblo, cities where no smoking ban existed prior to the implemented law, and therefore, cities in which the reduction in secondhand smoke exposure was truly from an extremely high level to an extremely low one. In California, since many cities and towns had already enacted smoke-free bar and restaurant laws prior to the statewide bans, the change in secondhand smoke exposure in these venues was probably not dramatic enough to expect to see the kinds of effects that would be observed at the city-level, especially in a city that had no prior regulation of smoking in its establishments.

While your thought of looking at statewide trends in myocardial infarction admissions to see if an impact of statewide smoking bans can be observed is an interesting one, ultimately it fails because the reduction in secondhand smoke associated with the regulation of smoking in bars and restaurants is a trend that happens gradually over time, even in states where a statewide law is implemented. This type of analysis is therefore not sensitive enough to pick up the smaller and more incremental changes in cardiovascular morbidity that might be attributable to this gradual trend in reduction in secondhand smoke exposure.

Even if it is the case that the primary effect of smoking bans on acute cardiovascular morbidity is attributable to reductions in active smoking (as opposed to decreases in secondhand smoke exposure), the changes in active smoking due to the implementation of smoking bans is something that would be expected to occur gradually over time.

The Rest of the Story

Because of the concerns I raise about my own analysis, I specifically do not think that this analysis demonstrates that smoking bans are not associated with immediate and drastic declines in acute cardiovascular morbidity. As I was careful to state in my original commentary: "I am not concluding here that smoking bans do not reduce heart attacks. I am not even concluding that smoking bans did not reduce heart attacks in Helena or Pueblo."

All that I am suggesting is that this is a complex area of inquiry, and that more research is necessary before we can draw definitive and generalized conclusions about the effects of smoking bans on acute cardiovascular morbidity.

Looking at trends over time in a phenomenon as complex as hospital admissions for heart attacks, and having just an 18-month period (and essentially one data point) before the law and an 18-month period (and essentially one data point) after the law is a very tough research design to draw causal conclusions about the effects of a smoking ban. With two data points, it is simply impossible (or perhaps slightly less than impossible) to determine the baseline underlying secular trend and extent of variability in that trend to evaluate, with confidence, the degree to which the observed change from one data point to the second represents a truly significant difference from the underlying variability in this statistic. Moreover, even if an effect were to be real, it is unclear how generalizable the finding would be to all other communities.

What the examination of state and national trends in hospital heart attack admissions adds is simply that it makes it clear that there is no obvious, dramatic, and immediate decline in heart attacks from smoking bans, sufficient to conclude definitively that public policy makers can expect a 27% to 40% decline in heart attacks in their cities if they ban smoking in bars and restaurants.

No comments: