Posted by Jost on July 18, 2006, at 19:41:23
In reply to Re: couldn't have said it better myself, posted by SLS on July 18, 2006, at 6:34:58
I'm late to this discussion, so maybe this has already been covered, but I have a few questions about the numbers in the Star-d study.
(My math is extremely rusty, so I could be wrong on my numbers-- so I guess I'm offering this with question marks.)
The numbers seem to be as follow:
original number considered for study: 4041
accepted into study: 2876
The first treatment was citalopram (celexa)
remission 30% (eg approximately 950)
responders 10-15 (eg 287-whatever 15% is) (these people were allowed to to into Level II because they didn't have "remissions" or become "symptom-free"
Ineffective, or too many Side effects to complete 70% (I infer that this number includes the responders, so there were 50-55% with no response, 10-15% with some response)
There were therefore 1914 people eligible to enter Level II.
Note: those who had remissions were to be followed for another 12 months, to evaluate the continuations of their remission, or other outcomes. I didn't notice any mention of them in the later discussion, however, they may be presumed to have some among them who were having a placebo effect, who might need another therapy later, if the study is continuing.
Of the 1914 eligible people, 1439 chose to continue into Level II
This means that almost 500 people dropped out at this point, leaving only 75% of the non-remitters in the study.
These people were given the choice to switch or augment citalopram with another AD. (People opted not to be randomized at this level of the study.)
51% (727) switched to another AD
39% (565) augmented
10% (149) chose to switch or augment with CBT, and were excluded from the final results.
So about 68% of the eligible people, or 1290 people participated in Level II.
Of these 68%, or 1290 people, there were remissions in:
25% of the switchers, or 188 people, and
30% of the augmenters, or 181 people
This means that of the second 1914 eligible, or 1439 participants, 369 had remissions. What 25% and 30% of 68% is, I haven't calculated, but I guess it's between 17-23% (I could be wrong on that...)
Altogether, of the 2876 people who entered level I, 1319 people (or about 46%) had remissions.
On one hand, remission is a higher standard than response, so this could be taken as a positive sign. On the other hand, it means that 64% had some sort of not-entirely-satisfying outcome.
The study, as reported online on the NIMH page, doesn't further divide these into those who had or didn't have a response less than remission, or further detail the quality of the less-than-remission responses. It also doesn't report on the progress of those who had remissions, and what, if any, further and different treatment they might have needed.
Maybe my math is wrong, because it seems to differ from the statistics that SLS quotes, which are more optimistic. If so, I'm interested in how one should analyze the results.
I should also note, however, that there must be some sense among scientists that treatments do work in a way that they take seriously. I say this because there are ethical concerns beginning to be reported about using placebos in studies. This ethical concern is about the use of placebos in patients for whom there is a proven treatment--for whom it is not considered ethically acceptable to give no treatment in a study of potentially helpful drugs.
I personally have had good results with certain ADs, so I'm not in any way debunking their value. I also believe that there are other ADs, which may work better, or other combinations that may work better, than those in the study.
I just wonder about exact outcome of the study, just on the simplest level, in the numbers. (Again, I apologize in advance, if I got this wrong.)