JAMA Psych Meta-Metanalysis

[Armadillos]

http://archpsyc.jamanetwork.com/article.aspx?articleid=1865002&JamaNetworkReader=True

Saw this paper in JAMA psych, I have to admit I'm pretty bad at trying to figure out what information is actually useful in meta-analyses because everything has been reduced down to simple numbers after tons of stats. So was wondering what you all find interesting, helpful, etc from this study? Most of the results seem not so surprising.

Did find it interesting in the discussion when it said "notably the effect sizes were overall similar to many common general medicine drugs [citation], providing important perspective." I dont really know enough statistics to know what that actually means

---

Members don't see this ad.

 

[Frazier]

http://archpsyc.jamanetwork.com/article.aspx?articleid=1865002&JamaNetworkReader=True

Saw this paper in JAMA psych, I have to admit I'm pretty bad at trying to figure out what information is actually useful in meta-analyses because everything has been reduced down to simple numbers after tons of stats. So was wondering what you all find interesting, helpful, etc from this study? Most of the results seem not so surprising.

Did find it interesting in the discussion when it said "notably the effect sizes were overall similar to many common general medicine drugs [citation], providing important perspective." I dont really know enough statistics to know what that actually means

In regards to your third paragraph, here is a different study summary that indicates similar findings...but explains a little more about what effect size means in boiled down form.

http://www.rcpsych.ac.uk/mediacentre/pressreleasearchives/2012/efficacyofpsychiatricdrugs.aspx

...and here is more of an indepth explanation of effect size:

http://www.leeds.ac.uk/educol/documents/00002182.htm

 

[shan564]

The only thing I see there is

http://archpsyc.jamanetwork.com/article.aspx?articleid=1865002&JamaNetworkReader=True

Saw this paper in JAMA psych, I have to admit I'm pretty bad at trying to figure out what information is actually useful in meta-analyses because everything has been reduced down to simple numbers after tons of stats. So was wondering what you all find interesting, helpful, etc from this study? Most of the results seem not so surprising.

I don't think it's you being bad at understanding anything... the problem with meta-analyses in psychiatry is that study designs are so variable and so often fraught with bias from drug companies. Most of the studies I see are problematic for a variety of different specific subtle reasons, and those reasons are too complex to control for in a large meta-analysis.

Did find it interesting in the discussion when it said "notably the effect sizes were overall similar to many common general medicine drugs [citation], providing important perspective." I dont really know enough statistics to know what that actually means

It means very little. I guess they're addressing the common critique that psych drugs don't work that well, and countering that by saying that we have pretty good effect sizes. Still, it's not necessarily a meaningful claim, since the meaning of the term "effect size" is so highly dependent on the type of effect that you're measuring.

 

[slappy]

Not quite, because the effect measured is usually symptom resolution (effect sizes are also adjusted for scale and the accuracy of measurements). Of course one could argue that for syndromes like Schizophrenia, the only meaningful effect is a complete cure. But until we do have a cure, symptom resolution is all we aim for in clinical practice (not dissimilar to asthma or diabetes for example) and is the most meaningful effect to measure in clinical studies.

Effect sizes are superior to the usual statistical tests in showing the significance because not only does it show that the intervention has a significant (arbitrarily assigned) effect, but it also shows how big that effect is (in the particular situations it's been tried). Also, another important property of the effect sizes is that it inherently allows comparison between studies (something that statistical significance alone cannot do because of the varying sample sizes between studies). So that quoted line implies that, because of comparable effect sizes, it can be concluded that the psychiatric drugs work at least as well as the general medicine drugs.

Hope that helps.

 

[shan564]

It's useful to measure effect sizes, but the point is that it's hard to use that meaningfully in a meta-analysis because different studies measure it differently.

 

[slappy]

What do you mean? You use effect sizes IN meta-analyses to combine measures from different studies.

 

[shan564]

If the effect size is measured the same way in each study.

 

[slappy]

As long as they measure the same thing, they can be safely combined with the right statistical methods.

Here's an excellent article on doing exactly that: http://psych.colorado.edu/~willcutt/pdfs/Morris_2002.pdf

 

[AnnoyedByFreud]

Anyone notice that the effect size for medication treatment of BPD is listed as 0.63 but is only 0.29 for DBT treatment? Drug company bias?

 

[shan564]

Anyone notice that the effect size for medication treatment of BPD is listed as 0.63 but is only 0.29 for DBT treatment? Drug company bias?

Placebo effect is powerful, especially for BPD, if you only measure the effect for a short-term period.