November 14, 2006

"Because I Said So"

I have written endlessly about how language controls psychiatric thought, and that it will be impossible for psychiatry to progress while semiotics trumps science.  Here is a recent example:

In the Oct 2006 JCP, there is an article about the efficacy of Depakote ER for acute mania. 

As I read the introduction to this useless paper, I get kicked in the throat by this:

"Currently approved treatments of the acute manic phase of bipolar disorder can be categorized primarily as mood stabilizers (e.g. divalproex sodium, lithium, and carbamazepine) or as atypical antipsychotics (i.e. aripiprazole, olanzapine, quetiapine, risperidone, and ziprasidone.(5)"

Note carefully that the authors have taken a set of medications and artificially divided them into "mood stabilizers" OR "antipsychotics." Ok, well, wouldn't it be great if reference 5 actually justified this?  Using data or logic?  Well, it doesn't. 

But the damage has been done.  Unless you have a computer with FIOS and three monitors and are reading every reference, a quick skim registers that there is a reference, which you assume has been checked, and move on.  In fact, the authors here don't even feel that a reference is necessary-- everyone knows what a mood stabilizer is.  It's too basic to even reference.

So, is there any reason that seizure drugs are "mood stabilizers" (read: prophylactic) while antipsychotics are not?  For antipsychotics, is there anything about their pharmacology, half-life, color, or pill size that a priori exclude them from the "mood stabilizer" category while including the seizure meds?

The artificiality of the terminology is confirmed when you actually look at the data: the only drugs listed here which actually are "mood stabilizers" are lithium, olanzapine and aripirazole (over 6 months).

A study may eventually show Depakote is a mood stabilizer after all, but that's not my point. My issue is that in the absence of data or logical necessity, how can we take an arbitrary set of names and make unjustified deductions?

This is the semiotic trap of psychiatry.  It doesn't actually matter what the data says (e.g. Depakote is not a mood stabilizer, Zypexa is), what matters is the language, the categories.  This isn't science.  Just because there are graphs and chi-squareds, doesn't make it science.   There's no science here at all.  At best it is linguisitics.  At worst, propaganda.

I'm not saying they are lying.  It's worse than that.  It's the structure of psychiatry.  It's a subtle manipulation of reality to make people believe what you "already know" to be true.  They are trying to convey a perspective, not report a finding. For example, later on the authors try to make the point that higher levels correlate with efficacy, but go too high and you get toxicity:

One analysis noted that serum valproate concentrations between 45-125 ug/ml were associated with efficacy, while serum valproate concentrations > 125 ug/ml were associated with an increased frequency of adverse effects. 19

This isn't what reference 19 says, exactly.  What it says is that 45 is a pivot point; below it is not as good as above it.  But it doesn't say that higher and higher levels give you better and better  efficacy.  What makes the omission of this rather important clarification all the more perplexing is that reference 19 was written by the same authors as this article.

But the damage has been done, again.  Now you think you have read a statement in support of what you already assumed to be true.  So you push the level.

You may argue that I am misinterpreting the author's words, that he never implied that efficacy had a linear relationship with level.  Ok: prior to reading this blog, did you think that there was?  Where did you learn that?  Did you pull it out of the ether?  No-- you skimmed articles like these that left you with half-truths, and never questioned it because everyone knows this already.

Let me show you what I mean.  Here's the relationship of the Depakote level to maintenance treatment:

Higher serum levels were modestly but significantly correlated with less effective control of manic symptoms in a maintenance study (26). The study therefore supports a somewhat lower serum level range for maintenance treatment than for treatment of mania.

Did you know that?  That the efficacy decreases as the level increases?  I'm not asking if you believe it, I'm asking if you had ever heard it.  Because if the answer is no, then there is something very, very wrong with the way we convey our knowledge. *



*Contrary to the opinions of former girlfriends, I am not an idiot.  I can plausibly explain this odd finding: the most manic patients got higher and higher doses, so the least responsive ended up getting the highest doses and levels.  So it looks like higher levels were associated with decreased efficacy, when really the highest doses went to the sickest people.  Ok, good explanation.  But this supports my earlier point: you can't take something which requires a post hoc justification and use it to make a leap in logic to conclude something else.