Wednesday, January 11, 2006

Science Behind Helena Claim Continues to Deteriorate

In the response, written by one of the authors of the original Helena study, to my questioning of the plausibility of the claim that the Helena smoking ban reduced heart attack admissions by 40% within six months, it was suggested that:

"The fact that there was a 13% drop in heart attacks in New York City also provides more evidence for a large immediate effect of eliminating exposure to SHS. (One would expect a smaller change in NYC than in isolated places like Helena or Pueblo because not all people covered by the ordinance would be hospitalized in NYC and vice versa. In addition, there were some earlier ordinances in some surrounding jurisdictions, which would 'smear out' the effect in time.)"

A similar response, posted as a rapid response in the British Medical Journal, made the identical point.

The Rest of the Story

First of all, I don't see how a 13% drop in heart attacks in New York City (assuming that it were due to the smoking ban) supports the plausibility of a 40% drop in heart attacks due to the smoking ban in Helena. The reasoning eludes me, because it seems to me that to support the plausibility of a smoking ban in Helena causing a 40% decline in heart attacks, you'd have to present data from another city showing that there was a 40% decline in heart attacks, or at least something approaching that magnitude of an effect.

But second of all, there was clearly not a 13% drop in heart attacks in New York City attributable to the smoking ban. The only data upon which this claim is made is apparently the observation that the number of heart attack deaths in New York City in 2004 (3,680) was 13.9% lower than in 2003 (4,275).

Comparing these two numbers to estimate the effect of the New York City smoking ban is not, in my view, science. It is just playing with numbers.

One could almost just as easily conclude that the smoking ban in New York City caused a 9.3% reduction in drug and alcohol-related deaths in New York, since the number of deaths from drugs and alcohol dropped by this amount from 2003 to 2004. And it's "plausible" that such an effect could have occurred because if people were less likely to smoke in bars, perhaps they were also less likely to be drinking as much, or to be using other drugs (since smoking is known to be related, to some extent, to use of other drugs). Maybe the smoking ban caused people to quit smoking, and with that, they dropped their other addictions as well.

And just as easily, one could conclude that the smoking ban caused almost a 10% increase in deaths from hypertensive heart disease, since deaths from this cause increased from 1,337 to 1,459 (an increase of 9.1%) from 2003 to 2004. Perhaps what happened was that smokers, unable to smoke in bars and restaurants, became more anxious and their hypertension worsened.

The point is, you can't simply observe that there was a reduction in deaths from a certain disease from one year to the next and conclude that a particular policy was the cause of that drop, without any other data or evidence.

Interestingly, this is the kind of science that the tobacco companies at one time used to try to convince policy makers that smoking bans caused major economic devastation, and that was criticized by at least one of the same people behind the New York City claim as being shoddy science. It's interesting that when applied to smoking bans, that's shoddy science, but when applied to something that is favored by anti-smoking advocates, it becomes "more evidence for a large immediate effect of eliminating exposure to SHS."

The Rest of the Rest of the Story

Well the truth of the matter is that even if one accepted the shoddy methodology of simply concluding that whatever drop in heart attack deaths was observed between 2003 and 2004 was attributable to the smoking ban, the 13.9% figure is just plain wrong, and deceiving as well.

The truth is that heart attack deaths in New York City have been declining for at least the past 5 years!

In fact, between 2002 and 2003 (the smoking ban took effect late in 2003), there was a 6.3% decline in heart attack deaths in New York City. So if heart attack deaths had simply followed the pre-existing annual trend, it would have dropped by 6.3% between 2003 and 2004. Thus, the decline "attributable to the smoking ban" would have been not 13.9%, as claimed, but only 7.6%.

And, in fact, not only had heart attack deaths been declining in New York City, but the rate of decline in heart attack deaths had accelerated during the previous three year period. The rate of decline in heart attack deaths was only 0.8% from 2000-2001, but rose to 4.5% from 2001-2002, and to 6.3% between 2002 and 2003.

Based on the trend in the pre-existing observed decline in heart attack deaths in New York City, the expected decline in heart disease deaths from 2003 to 2004 was in fact 9.3%. Thus, the true decline "attributable to the smoking ban" was in fact not 13.9% as claimed, but only 4.6%.

So even if one accepts that any observed difference in the heart attack death trend in New York City from 2003 to 2004 had to be attributable to the smoking ban and nothing else, then the "actual" effect was only a 4.6% decline (which I actually find to be quite plausible).

Obviously, that doesn't provide any evidence for the plausibility of a 40% decline in heart attack admissions in Helena attributable to the smoking ban.

To make matters worse, the claim in Helena was a 40% decline in the incidence of heart attacks due to the smoking ban, but the New York City data relates to the mortality (deaths) from heart attacks. There is no evidence of which I am aware that the number of heart attacks in New York City declined after the smoking ban beyond what would have been expected from the existing secular trend.

There are many factors that can and do affect heart disease death rates, not the least of which is the treatment for heart disease, which has drastically improved in recent years. It is much more difficult, and most certainly inaccurate, to attribute changes in heart attack deaths solely to changes in risk factors for heart disease, since treatment is definitely a factor that is going to affect the death rate.

In fact, heart disease mortality rates declined by a whopping 46% from 1970 to 1995, but it has been estimated that only about one-third of this decline was due to a reduction in heart disease incidence. The majority of the decline was due to advances in the medical treatment of heart disease (see Sytkowski PA, Kannel WB, & D'Agostino RB. Changes in risk factors and the decline in mortality from cardiovascular disease: The Framingham Heart Study. New England Journal of Medicine 1990;322:1635-1641 and Siegel M, Doner L. Marketing Public Health: Strategies to Promote Social Change. Gaithersburg, MD: Aspen Publishers, 1998).

The rest of the story suggests that the science behind the claim of a 40% reduction in heart attacks due to smoking bans is shoddy. It is not plausible, and the attempts so far to defend its plausibility are based on poor science.

My point is simply that the credibility of tobacco control scientists and practitioners is threatened when scientific claims that are not adequately justified are made. People are just not going to be able to discriminate between adequately supported claims and those which are poorly supported, such as this one. And the result will be that all claims, even the legitimate ones, are going to eventually be disregarded or seriously challenged. And that's going to undermine our ability to advocate effectively for the policies that are needed to protect people from secondhand smoke and other public health hazards.

No comments: