Friday, 7 January 2011

Antidepressants have no more than a minor effect in Minor Depression

Neuroskeptic reports on another paper that:
"...has added to the growing ranks of studies finding that antidepressant drugs don't work in people with milder forms of depression"
The study looked at 'Minor Depression', which is a DSM diagnosis that is similar to 'Major Depression' except that patients only have 2-4 rather than the 5 or more depressive symptoms necessary for the latter diagnosis.

Neuroskeptic reproduces the 'response rate' figure which shows no significant effect of antidepressants in this diagnosis but I think the findings from the continuous outcomes are more informative:

This shows the 3 studies that reported continuous outcomes (mean HRSD score) looking at paroxetine in nursing home residents, fluoxetine, and amitriptyline respectively. The effect size is a small* -0.93 HRSD points (95% CI: -2.27-0.41) and not statistically significant**. However, there are only about 200 patients included (which is about the minimum number to justify combining data in a meta-analysis) of which the vast majority came from one trial and I don't think there is enough evidence there to conclude that antidepressants do not work in minor depression***, although the effect size is likely to be small even if they do have a statistically significant effect when enough patients are studied.

* See my previous post for an antidepressant medication that is widely recommended and similarly shows a statistically non-significant HRSD score improvement around 1 point.
** The data presented are for random effects but fixed effects produce exactly the same results.
*** To compare with previous studies of major depression, where mean baseline severity is usually around 24 points on the HRSD in these studies they were something like 12 points on the HRSD. If you extrapolate from studies in major depression then baseline severity of this magnitude would fall well within the region of 'no effect' and so the detected effect size of around 1 point is actually larger than you would have expected.


Neuroskeptic said...

As you say, with 200 patients and only one decent study, it's hard to know.

However, I think the evidence at best limits the possible effect to being tiny.

There's also the question of whether the HAMD is appropriate to measure Minor Depression. I would say no. If you only score on 2 DSM-IV items, by definition, you should score low on the HAMD because you won't have multiple sleep, appetite, suicide, etc. scores. The only way you could score high on the HAMD with 2 DSM-IV items would be if you scored on the anxiety items of the HAMD. But then you're anxious, not depressed, I would say.

* although it kindof depends on which items you score on.

pj said...

Oh, I agree, it is likely to be tiny, or rather, 'small' as Cohen would put it.

I suppose it is good that HRSD scores are low in minor depression, you'd rather hope that they were, and so you'd expect it to be more difficult to detect a change (with an average of 12 at baseline you've only got 4 points to play with before you reach 'remission' on the HRSD). The mechanics of developing a more sensitive or suitable scale for minor depression I wouldn't like to imagine.

As for anxiety items, for mild depression there is, in my experience, often a great overlap between anxiety and depressive symptoms (there's even a diagnosis of 'mixed anxiety & depression') so it isn't completely outrageous for these to be prominent.