X-Message-Number: 18471
Date: Sat, 02 Feb 2002 22:19:33 -0500
From: <>
Subject: Re: CryoNet #18464 (flu vaccines and stroke)

> CryoNet - Sat 2 Feb 2002
Message #18464
From: "John de Rivaz" 
Subject: There is a theory linking infectious disease with atherosclerosis
Date: Fri, 1 Feb 2002 23:12:16 -0000

From Infobeat

Study Finds Flu Shots Prevent Strokes

A flu shot can protect you against a stroke, a French study finds.

The risk was 40 percent lower for people who had the influenza vaccine
compared to a carefully matched group of people who didn't, researchers at
the Denis Diderot University in Paris report in the February issue of
Stroke, a journal of the American Heart Association.

"There is a theory linking infectious disease with atherosclerosis
[hardening of the arteries]," explains Dr. Pierre Amarenco, professor of
neurology at the university and a member of the study group. "Especially, it
says that infections can promote or trigger occlusions of blood vessels,
either coronary arteries or brain arteries."

"We hypothesized that patients vaccinated against influenza may be at less
risk of brain infarction [stroke] because they are at less risk of
infection. So, we asked patients admitted for brain infarction whether they
had been vaccinated against influenza and compared them with control persons
who were matched for age, sex and place of residence. We found that the
stroke patients were much less likely to be vaccinated than the controls."

COMMENT:  If this is literally as reported, it's just another 
stupidly-controlled study which has found another association which is probably 
due to a dummy variable.  Why (for instance) would they possibly be so foolish 
as not to control for major known things which impact stroke-risk, like 
socioeconomic class and smoking behavior?  Poor people who don't take their 
blood pressure pills are (after all) the very same people who don't get 
vaccinated. For this study to have a chance to mean *anything,* they need to 
control for the fact that getting the flu vaccine may simply be a marker for 
whether or not you take care of yourself in other ways. 

Come on, folks! We talked about this before. The nature of regression analysis 
of post hoc data is that associations by themselves usually mean very little, 
because most of them are due to confounding effects (one variable is a dummy for
some other a truly causal one). Unless you believe that ice cream consumption 
causes juvanile delinquency because they both go up lock-step in the Summer....

In any positive association report you need to control for all other effects 
PLUS show a dose-response effect, to even make the study interesting enough to 
follow-up with other kinds of studies.  On the other hand, *null* associations 
between variables in such studies are important by themselves, because they come
closer to ruling out causal associations between the things measured.  This is 
because it's hard (though not quite impossible) to have a causal association 
between two variables which do not correlate at all. It's easy to have 
associations between variables which aren't causally connected. 

Steve Harris

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=18471