Tuesday, August 05, 2008

Siegel & Gorski on Scottis heart attack study

Dr. Michael Siegel, whose many secondhand smoke studies helped form the
basis of Surgeon General Carmona' report, is disputing the significance of
the Scottish heart attack study with Dr. David H. Gorski, MD, PhD a cancer
surgeon/researcher. Both agree that the number 17 percent is greater that
the true reduction, if there was any.

Here are some relevant statements by Siegel:
"..one cannot rule out the very plausible alternative hypothesis that the
observed decline in heart attacks is explained by random variation in the
data and the already existing secular trend of declining heart attacks in

"Just to make it clear, prior to the use of troponin, a number of cases of
unstable angina were likely missed. If these patients were sent home
without definitive treatment, they were likely to return again in time
with a full-blown heart attack. As the diagnosis of unstable angina
becomes more sensitive, these patients are getting appropriate treatment
and this will reduce future admissions for heart attacks. The study fails
to consider this very plausible explanation for at least a portion of the
observed decline in heart attack admissions." (from the comments for
previous Siegel blog post)

"I think it is very misleading to take the 17% drop in the acute coronary
syndrome (which clearly includes heart attacks as its major component as
the diagnostic criterion was simply a detectable level of cardiac troponin
which occurs in heart attacks as well as unstable angina) as diagnosed in
these 9 hospitals under this study protocol and to argue that this drop is
much greater than the 10% drop observed in national data. Why? Because the
national data is based on a completely different data source and
definition and is subject to far less random variation. In fact, if you
use the national data to examine the change in heart attacks associated
with the smoking ban, you find that it is nowhere near 17%, but it is
actually between 4 and 8%, depending on which year you look at. In either
case, the decline is less than that observed from 1999 to 2000 and is
clearly within the range of changes in heart attacks observed during the
baseline period.
Again, I'm not arguing that these data suggest that there was NO effect of
the smoking bans - in fact, there almost certainly was SOME effect. I just
think that there is no way one can attribute the 17% observed decline to
the smoking ban.
I have conducted studies very similar to this myself (for example,
analyzing the effects of smoking bans on trends in smoking prevalence). In
doing such studies, you need to have a consistent data source over a long
period of time (I used a period of over 10 years) to be able to draw
credible conclusions."

Dr. Gorski:
"In retrospect, I actually agree that the effect is almost certainly not
17% and that that number shouldn't have been emphasized in the paper or in
the press releases. As I said before, I would have characterized the
results as being a drop that is 13% greater than the long-term trend, and
if I were writing an accompanying editorial I'd say that that's probably
the upper bound for how large the effect was likely, in fact, to have
been. The "true" effect was likely smaller.

Even so, what if the effect of the smoking ban was only 10% Or only 5%? Or
only 2-3%, as you seemed to be implying with your use of the 2005-2007
heart attack data (17 - 4.6 - 10.2 = 2.2)? That would still be a benefit.
Of course, trrll nailed it when he pointed out that you are actually the
one comparing apples and oranges; so my guess is that the "true" effect
was significantly higher than 2.2% and probably lower than 13%."

Here is link to Siegel's discussion of a yet unpublished heart attack
study that Siegel really likes which found no effect:

Dr. Gorski also asked Dr. Siegel:
"What specific data would it take for you to believe that smoking bans can
significantly reduce admissions for coronary artery disease?"

Siegel's answer:
"Very simple - a study that uses a consistent data set over a long enough
period of time so that a clear baseline trend can be established and which
also uses a comparison population so that secular trends can be ruled out
in addition to random variation in the data. Ideally, a systematic study
of multiple locations that enacted smoking bans would be conducted."

I believe the Kuneman-McFadden heart attack study meets Siegel's criteria: