There are now 2 major papers that show antidepressants to be less
effective than expected to the point where the clinical value of
antidepressants over a sugar pill is questionable.
This is a critique of the Newsweek coverage. What's extraordinary is
major magazines cover psychiatric research papers. This really doesn't
happen much in the UK, not unless it's a crap paper about caffeine
causes psychosis.
The Kirsch meta-analysis is the significant one because it uses
unpublished data and that's where the big story is. The author
criticises the meta-analytic technique for only being as good as the data.
His criticism of the latest Fournier study is valid. I'm surprised the
authors of the paper selected subclinical doses for inclusion in their
review - I suspect they would have a response to this criticism. They'd
also have a reason why they chose to exclude placebo washout trials but
I can't think why.
He hasn't pushed aside the problem of publication bias and it's effect
on inflating the effect size from other meta-analyses of
antidepressants. Publication bias is significant and affects all areas
of mental health research. It's estimated in a paper published earlier
this year that publication bias accounts for about a third of the effect
size of psychological therapies research.
He makes a very important point though.
"
As research psychiatrist Dr. Sheldon Preskorn recently wrote me, "…
there is much more treatment [provided] by being on a placebo in a study
than most depressed patients get in routine clinical practice,
particularly in the primary care setting" (personal communication,
2/03/10). Indeed, Preskorn estimates that in a typical 8-week trial, a
subject in the placebo group may receive 10 to 12 hours of contact time
with knowledgeable and empathic healthcare practitioners. In effect, the
placebo control is a kind of substantive, supportive intervention."
His criticism that the placebo group do better because they're under
trial conditions makes sense initially however it's notable that this
effect is an important part of the placebo-controlled trial. In any
trial there should only be 1 variable the effect of the social contact
is shared between both trial groups. The research is to show the
effectiveness of antidepressants compared to the placebo effect. It's
not to compare to no treatment at all. (He's forgotten the studies that
show 85% of people recover from depression in one year).
There's another interesting thing on the following line. This is where I
wish people would cite their sources.
"
Furthermore, placebo group response rates in depression studies have
been mysteriously and substantially rising in recent decades—perhaps in
part because less severely depressed subjects are being recruited.
"
There's an easier to read critique of the Fournier meta-analysis
published in Newsweek.
http://www.newsweek.com/2010/01/29/a-doctor-disagrees.html
This is the Fournier meta analysis
http://jama.ama-assn.org/cgi/content/short/303/1/47
Antidepressant Drug Effects and Depression Severity - A Patient-Level
Meta-analysis
JAMA. 2010;303(1):47-53.
I rambled on another paper showing similar results that was published
later this year in the journal Psychotherapy and psychosomatics. It's
the third paper that's calling into question the value of SSRI-type
antidepressants in the treatment of depression.
The massive problem for psychiatrists is the effect sizes regardless of
these criticisms. In physical medicine effect sizes of 2 and 3 are
considered significant. In psychiatry an effect size of 1 is enough to
be considered significant. It's because psychiatry's armentarium isn't
very good.
No comments:
Post a Comment