Posted by Larry Hoover on January 28, 2009, at 10:20:17
In reply to Re: Stats...ooops » Larry Hoover, posted by SLS on January 28, 2009, at 8:58:22
> Hi Larry.
>
> http://medicine.plosjournals.org/perlserv/?SESSID=117d5ff3d1cfb7ab00f072415cdba934&request=slideshow&type=figure&doi=10.1371/journal.pmed.0050045&id=96831
>
> How do you interpret the dashed line representing "Very Severe Only"? What are its implications regarding study design?
>
> Thanks.
>
>
> - ScottThe different slopes of the regression lines are likely due to the inclusion or exclusion of the one study way to the left, where initial Hamilton Depression scores were less severe. Critics of antidepressant drugs argue that the implication is that the drugs are unnecessary or ineffective for mild to moderate depression. In fact, the likelihood of placebo response is inversely correlated with initial severity of depressive symptoms. In effect, it is more likely to mask drug response in those with lesser severity. If you plot drug response alone against HRSD, the line is virtually horizontal, indicating stable response across initial severity.
Over time, placebo response in antidepressant clinical trials is increasing. What that indicates, IMHO, is that we are becoming better at treating depression overall. The uncontrolled independent variables that promote placebo response, the caring expert physician (run of the mill family doctors don't participate....it's the cream of pdocs, typically), regular and detailed follow-up (weekly or biweekly), the symbolism of being in a clinical trial (expectation of better treatment than is otherwise available), co-morbidities excluded, brief duration, and so on.
I don't know that study design needs to change. I think interpretation needs a shake-up. Getting hung up on statistical significance or the magnitude of difference scores is an overly restrictive viewpoint. Clinical trials do not represent typical clinical practise. If a drug surpasses placebo despite the artificial circumstances, it is a true difference. Failure to find a significant difference does not mean there isn't one. The homoskedacity (see my post immediately prior to this one) in clinical trial outcomes has only one reasonable interpretation.
You're welcome.
Lar
poster:Larry Hoover
thread:876214
URL: http://www.dr-bob.org/babble/20090104/msgs/876745.html