Posted by Larry Hoover on January 6, 2008, at 14:05:01
In reply to Re: STAR*D confirmed what patients already knew, posted by linkadge on January 5, 2008, at 15:24:47
> >The methodology could not include placebo.
> Well, then change the methodology. Its not so much that doctors can offer placebos, but if they knew that medications were essentially no better than placebos, they might decide against riskier treatments in favor of other treatments.
But medications are better than placebo. Some studies fail to show a significant difference, but there are methodological issues to consider. I don't want to turn this into a methodology lecture. The failure to find a difference between two groups is *not* evidence that they are similar.
If placebos and antidepressants had similar efficacy, then placebos would have come out ahead as often as behind antidepressants. We do not see that.
> Well, strong response to a placebo might have helped support the finding that pretty much all choices are essentially equivilant.
I can hypothesize too. Placebo and antidepressants have not been shown to be essentially equivalent.
> It might even help to uncover exactly what factors predict placebo response.
We are learning those things. The most commonly used methodology in efficacy trials actually promotes placebo response. Simply introducing variable dosing, and limiting subjects to the more severely depressed, drastically reduces the placebo artefact. I would argue that any study that failed to include these two considerations should be discarded as innately flawed.
> I personally think that the study purposly did not include a placebo.... The last thing they wanted in this study was to have a pesky little placebo response disclaimer tagged to the end.
You're right, but for the wrong reason.
> Consider the fist round, %30 of people responded to citalopram. That is slightly less than what previous studies for citalopram have stated.
This is more than a trivial error. The standard applied in this study is remission, not response. This study "raised the bar".
> However, in many of such studies, the placebo response comes in about the same rate.
Many methodologically flawed studies, yes.
> On to round two...Eventually, you're going to get a final result saying yeah sure %70 of patients can improve,
No, the standard here was remission.
> ...with one drug or another, yet %60-80 can improve with placebo. Wow, all of a sudden that %70 is meaningless.
Where did you come up with that placebo figure? I have never seen even so much as a single study that assessed serial placebo treatments. To presume that a person who failed one placebo treatment might subsequently respond to another, and with the same likelihood as in the first instance, is a huge assumption.
> Its one of those things where sometimes the most basic assumptions are wrong IMHO.