PsychoTropicalResearch, serotonin and serotonin syndrome research.
Antidepressants - SSRIs vs TCAs Latest on efficacy
Antidepressants - SSRIs vs TCAs Latest on efficacy
Date Created: 20/09/1999 Last Modified: 03/06/2000 Last Checked: 29/06/2003
The most recent meta-analysis comparing individual SSRIs is now available. The author (Ian Anderson) has published several previous meta-analyses in reputable journals on similar themes; this is the first that has sufficient data to look at the question of the efficacy of individual SSRIs vs various reference TCAs.
Summary
The paper analyses a total of ~11,000 patients from 102 randomised control trials that met his predefined criteria for quality and methodology.
*** SSRIs are better tolerated, overall discontinuation rate 3% less, and discontinuation rate from side effects 3% less (NNT=33, (NNT=number needed to treat) ie you have to treat 33 patients to get one who continues with drug 'A' who would otherwise have ceased with drug 'B')
*** Individual SSRIs; only fluvoxamine different, side effects were just as bad as the TCAs (nausea was common)
*** For inpatients (25 studies with 1377 patients in total) the magnitude of treatment benefit ('effect size' was 0.2 better) was significantly (P=0.012) in favor of TCAs (as a group) over SSRIs.
*** amitriptyline (30 studies with 3053 patients) was significantly superior to SSRIs (P=0.01)
*** However, figures for clomipramine (18 studies, total 2264 patients) do not support its possible superior effect (this contrasts with a previous result from the same author, 1994)
*** Counter-intuitively, the SSRIs seem to show no side effect advantages over TCAs in a sub-analysis of the elderly group. Also in all patients dothiepin had less side effect discontinuations than SSRIs.
*** There was a strong trend for citalopram and paroxetine to be less effective than TCAs which did not quite reach statistical significance at the P 0.02 level. Nevertheless a betting man would have to go with the odds. They are 10:1 in favor of tricyclic antidepressants being significantly better than citalopram and paroxetine.
*** Note that for comparison with above figures:-- the effect size for antidepressants generally, vs placebo, is 0.4
Also note that all these studies were sponsored by the makers of the new drug ie. the SSRI.
Dr Anderson has also advanced evidence in another analysis that the efficacy demonstrated in trials is heavily biased towards superiority of the sponsored drug. Sponsorship is a bigger influence on treatment effect size than any other factor: including dual action (like venlafaxine), SSRI vs TCA, other drug properties, age, methodolgy of trial, dose. One need hardly go on !
Conclusions
The interpretation of meta-analyses is so fraught with difficulties that it cannot be taken at face value, so much so that the more experience one has of it the more one wonders, what is it useful for?
It would seem that the side effect advantages of SSRIs are very modest indeed; one needs to treat 33 patients (NNT=number needed to treat) to gain one less dropout (see above). One wonders what the result would be if the blood levels of the TCA were optimised and if presently unknown or underestimated side effects of SSRIs were factored in. It seems likely to me from my experience that the less side effect prone TCAs (eg nortriptyline or desipramine) would probably turn out to have less side effects than serotonin reuptake inhibitors.
Fluvoxamine may be best avoided altogether, or reserved as a last resort drug (see other notes). Indeed in view of its extensive and potentially serious interactions (via blockade of cytochrome P450 enzymes) it is best avoided in some clinical settings. To some extent similar reservations apply to both paroxetine and fluoxetine, again because of the potentially serious cytochrome P450 enzyme interactions.
Amitriptyline probably remains the most effective drug.
The finding of less drop outs from TCAs in older vs younger patients is rather strange. It may be a reminder that the side effects reported in trials are as unreliable as the clinical assessment of improvement. Remember the absurdly low estimates of sexual dysfunction produced by almost all SSRI trials?
All these patients were subject to the usual trial exclusion criteria (re general health etc), so they are not very representative of those we all treat. This really is an under-recognised problem (see especially the paper by Licht).
The trials were short term (ie 4-6 weeks). That factor alone substantially restricts the usefulness of the results since patients with major depression require treatment for much longer.
I suspect Dr Anderson sometimes wonders if the effort was worth it.
The uncovering of the large unpublished trial of the serotonin and noradrenalin reuptake inhibitor (SNRI) sibutramine makes a nonsense of meta-analysis by demonstrating the magnitude of the hidden distorting biases.
All text ©Dr Ken Gillman MRC Psych - PsychoTropical Research®
|Return to the menu at the top of this page|