Posted by Klavot on December 30, 2006, at 11:32:37
In reply to Re: After the hype, posted by linkadge on December 30, 2006, at 10:02:57
No, the calculation is wrong; it assumes that 58% of trials find a 0% success rate for antidepressants.
Suppose, as you do, that 42% of trials find antidepressants to work for 55% of patients, and the remaining 58% of trials find antidepressants to work at placebo level - say for argument sake 40% success rate.
Then the overall proportion of patients who will benefit from antidepressants is
0.42 x 0.55 + 0.58 x 0.40 = 0.463
or 46.3% (assuming the given data).
Klavot
> More than half of all clinical trials fail to show that antidperessants perform better than placebo.
>
> So, if you really want accurate numbers for clinical trials, you do it this way.
>
> More than half fail to show, that means that less than half actually show, say 42%. Then we multiply that by the average number of people helped *within* a successfull clinical trial, say 55%.
>
> Ie clinical trials, overall, say that 0.42 x 55% = 23% of people are helped by antidepressants.
>
> But they don't do that. They simply state the percentage of people who are helped within sucessfull clinical trials.
>
> Linkadge
>
>
>
poster:Klavot
thread:717262
URL: http://www.dr-bob.org/babble/20061224/msgs/717567.html