Posted by Larry Hoover on January 3, 2009, at 16:20:45
In reply to Re: Actually, that's not what the study found., posted by SLS on January 3, 2009, at 9:24:39
> In my opinion, the study you cited at the beginning of this thread is worthy of consideration. I think it was designed pretty well. I'll leave it to Larry to change my mind.
>
> :-)Its design is a post-hoc analysis of data that were collected between 1989 and 1999, primarily for the purpose of assessing cardiovascular risk factors and their influence on cognitive decline. In all, 70 variables were assessed. White matter decline was observed with smoking, use of diuretics and statins, as examples. (Reference 11 in the full-text.)
That same dataset was subsequently used for the purposes of the present study, a post-hoc analysis of data collected for another purpose altogether. I happen to hold to a particular skeptical view of data subjected to post-hoc analysis, what some call "data-mining". Essentially, you just don't know how complete the data really are, or if bias or uncontrolled variables influence the data that were collected.
I've previously questioned the meaning attributable to the correlation between antidepressant use and white matter decline, particularly because data with respect to diagnosis and course of major depression were not collected. It's impossible to know if depression caused both antidepressant use and white matter decline, or if antidepressant use alone was the dominant factor.
If you plotted height vs use of make-up for people under 20, you'd find a positive correlation. But, each factor is really an outcome of maturation. Despite the correlation, height doesn't cause make-up use, nor the other way around. Moreover, there would be other factors influencing the correlation, e.g. gender, ethnicity, and religion. A discussion of the correlation wouldn't be complete without taking those into account.
Getting back to the present study, depression is already in a well-established correlation with cardiovascular disease, for which causation need not even be considered in the present context. Diuretics and statins are often used in CV treatment, yet these variables, previously shown by these same authors to significantly promote white matter decline, were neither mentioned nor adjusted for in the present study. Nor was smoking. Why not?
The link between depression per se and white matter loss may even be inferred from the very small sample (15 subjects) who used both serotinergic agents and TCAs during the study period. Fully 60% of that small group experienced white matter declines, but with their exclusion from the stats, no significant effect of antidepressant use remains. I'm tempted to consider that unremitted depression, in some cases calling for multiple treatments, is the underlying issue, rather than medication use. The data allow such an interpretation, whether I may be right or wrong.
So long as there are multiple reasonable interpretations of the data, I can only conclude that questions/concerns have been raised that have not been answered. The authors' own hypothesis was that serotonergic meds would be protective, and upon a quite reasonable basis. Their data-mining excursion did not support their hypothesis, but nor did they manage to exclude explanations that did not support a possible causative relationship between antidepressant use and white matter decline.
I'm not trying to change anybody's mind. I am saying that I haven't made up my mind, due to insufficient and incomplete data for consideration.
Lar
poster:Larry Hoover
thread:871107
URL: http://www.dr-bob.org/babble/20081223/msgs/872121.html