Trial link.

Cipriani updates her last major review of all head to head trials between antidepressants to give us a new line up of best ones to use for either efficacy or tolerability. There is not clear winner, but there are some losers. The new line up: Best choices: agomelatine, amitriptyline, escitalopram, mirtazapine, paroxetine, venlafaxine, and vortioxetine. Losers: Prozac, Luvox.

Update:  N. Ghaemi, a prominent mood disorder specialist in the US (at my alma Mater), has written strongly against the conclusions of Cipriani.  It turns out, according to careful reading of the study, that two conclusions can be drawn:

1. He admits there is good evidence that antidepressants are significantly better than placebo, up to 1.5x better.
2. However, he asks what 1.5x better really means, in terms of baseline effect.  In other words, when compared to placebo, what is the actual effect size of the difference? (ie. okay they’re better, but by how much?); he shows that its actually meaninglessly better.  In fact, all antidepressants save amitriptyline are only mildly better (effect size <0.5), where “mildly better” basically means only a few points difference on a 27 or 30 point scale.  Big deal, he says.  Maybe amitriptyline is much better, (meaning, lots of points better than placebo on a scale).

So, in sum, as I explained to a patient today, if you don’t really expect the antidepressant to work, it probably won’t, because there is no placebo effect…except if I give you amitriptyline.  That’s a tough pill to swallow…literally loaded with side effects.  It works, though!   Tough choices..

1. Side note: MAOi were not studied, and may actually be seriously better. Nortriptyline, Imipramine, clomipramine not mentioned and could hold up to amitriptyline.  It’s well known TCA > SSRI or SNRI.

I need to get my head wrapped around this bad news.

N.B.  Dr. Ghaemi, actually goes so far as to question Major Depression as a valid construct.  Either antidepressants don’t work or major depression is not a real entity.  The latter suggests it can be parsed into bipolar disorder, PTSD, personality traits, and probably up to 4-5 different constructs with different responds to different drugs.  (ie. an antidepressant wouldn’t work in all these cases, and would look badly in any study of Major Depression and come out poorly vs. placebo.  More to get my head around…