The meta-analysis by Cipriani et al. (2018) comparing the efficacy of different antidepressants has led to a lot of press and discussion among clinicians.
But, I’m not sure how many people have personally read the paper.
Summary of the study
The authors did a thorough search of the literature to identify all two types of randomized, controlled clinical trials of 21 different antidepressants for the treatment of major depressive disorder in adults—those comparing an antidepressant to a placebo and those comparing one antidepressant to another in the same study (called “head-to-head” trials).
Trials that included patients who had psychotic depression, treatment-resistant depression, or patients with a serious concomitant medical illness were excluded.
A total of 522 trials comprising 116 477 participants were included in the analysis.
Here are two key findings from the paper, presented verbatim from the Abstract:
1. “In head-to-head studies, agomelatine, amitriptyline, escitalopram, mirtazapine, paroxetine, venlafaxine, and vortioxetine were more effective than other antidepressants (range of ORs 1·19-1·96), whereas fluoxetine, fluvoxamine, reboxetine, and trazodone were the least efficacious drugs (0·51-0·84).”
2. “For acceptability, agomelatine, citalopram, escitalopram, fluoxetine, sertraline, and vortioxetine were more tolerable than other antidepressants (range of ORs 0·43-0·77), whereas amitriptyline, clomipramine, duloxetine, fluvoxamine, reboxetine, trazodone, and venlafaxine had the highest dropout rates (1·30-2·32).”
The authors concluded that, “These results should serve evidence-based practice and inform patients, physicians, guideline developers, and policy makers on the relative merits of the different antidepressants.”
The full article is available free to everyone at THIS LINK. A commentary on the article was also published in The Lancet and that paper available free to everyone HERE.
Calling all clinician readers!
Do you accept the finding of this study that some antidepressants are more effective than others? If so, which ones? Most importantly, will this paper change your choice of antidepressants or not?
We would like to get your comments on these questions. Please leave them under “Leave a Reply” at the bottom of this page.
Rajnish Mago, MD, has the following comments on this paper:
I have serious concerns about the conclusions being drawn from this paper.
Looking at the head-to-head studies, i.e., direct comparisons rather than a network meta-analysis (see their figure 4 at the bottom of this page), only a few of the comparisons are statistically significant. But, in their Conclusions, the authors state that some antidepressants were more efficacious than others even when the differences were not statistically significant. For example, some clinician readers were surprised to read that agomelatine was more efficacious than other antidepressants. But, the only antidepressant that agomelatine was statistically significantly more efficacious than was reboxetine. For all other antidepressants included in the analysis, it was not statistically significantly different.
In figure 4 (head-to-head comparisons; see figure at the bottom of this page), if we exclude comparisons to trazodone and reboxetine, which are special types of antidepressants and did poorly compared to many antidepressants) the only statistically significant comparisons were the following:
1. Several antidepressants were statistically significantly more efficacious than fluoxetine (alphabetically: amitriptyline, escitalopram, mirtazapine, paroxetine, venlafaxine).
2. In addition to its comparison with fluoxetine, escitalopram was also statistically significantly more efficacious than citalopram, clomipramine, and fluvoxamine.
3. Mirtazapine was statistically significantly more efficacious than fluvoxamine.
That’s it. So, in my opinion, the statement in the Abstract that has received so much attention and has caused consternation among clinician readers is, in fact, misleading:
“In head-to-head studies, agomelatine, amitriptyline, escitalopram, mirtazapine, paroxetine, venlafaxine, and vortioxetine were more effective than other antidepressants (range of ORs 1·19–1·96), whereas fluoxetine, fluvoxamine, reboxetine, and trazodone were the least efficacious drugs (0·51–0·84).”
Heinz Grunze, MD, who is on the Editorial Board of Simple and Practical Mental Health had the following response to my comments above:
I fully agree with your conclusion, as the mentioned antidepressants were only more effective than one or more of the following: reboxetine, trazodone, fluvoxamine or fluoxetine but not more effective than “other antidepressants” which, in my understanding, would include also standard treatments such as sertraline, paroxetine, duloxetine etc.
As a clinician, I would also agree that they should not base such conclusion on superiority over reboxetine which, in previous meta-analyses, has been shown to be no better than placebo. I wonder whether they dug out any unpublished data that changed the signal or whether this result is due to the omission of studies from the analysis.
There are also other issues, e.g., why did they not include a single MAO-I? I don’t think that this class is obsolete, at least not in more severe depression.
And, I also feel that the analysis is not very helpful when it comes to treating inpatients as it was done almost exclusively on outpatient trials- which clearly is a distinct group of patients with distinct needs.
Mark Zimmerman, MD, who is on the Editorial Board of Simple and Practical Mental Health had the following comments on this paper:
This article has no impact on my selection of an antidepressant.
I must confess to having my suspicions about this type of network analysis. I know it is an oversimplification, but I tend to think of it this way: Brown beat Harvard, Harvard beat U Mass, which beat Rutgers, which beat U Maryland, which beat Michigan, which beat Duke. Thus, Brown is better than Duke. I know it is more complicated than that, but given how studies have changed in how they recruit patients over time and given how the make-up of patients has likely changed, I am skeptical of these type of analyses.
They omit one crucial factor—exclusion from studies. You cannot extrapolate to clinical practice because of the restrictive nature of these studies. They do not address this issue at all. Thus I disagree with their conclusion that “These results should serve evidence-based practice and inform patients, physicians, guideline developers, and policy makers on the relative merits of the different antidepressants.”
Michael Thase, MD, who is on the Editorial Board of Simple and Practical Mental Health had the following comments on this paper:
In meta-analyses of head-to-head trials, escitalopram, venlafaxine, and mirtazapine do have stronger effects than fluoxetine (and possibly the remainder of the SSRIs as a class).
Vortioxetine, on the other hand, has never shown superiority to any antidepressant and is marginally less effective than venlafaxine and duloxetine.
Related Pages
Which antidepressants are better than the SSRIs?
Combining antidepressants: Are two better than one?
SSRI/ SNRI plus mirtazapine: What three clinical trials found
Ketamine for major depression: An overview
Pramipexole for major depressive disorder or bipolar depression
Tips on using pramipexole for mood disorders
How to choose between phenelzine (Nardil®) and tranylcypromine (Parnate®)
References
Cipriani A, Furukawa TA, Salanti G, Chaimani A, Atkinson LZ, Ogawa Y, Leucht S, Ruhe HG, Turner EH, Higgins JPT, Egger M, Takeshima N, Hayasaka Y, Imai H, Shinohara K, Tajika A, Ioannidis JPA, Geddes JR. Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. Lancet. 2018 Feb 20. pii: S0140-6736(17)32802-7. PubMed PMID: 29477251.
Parikh SV, Kennedy SH. More data, more answers: picking the optimal antidepressant. Lancet. 2018 Feb 20. pii: S0140-6736(18)30421-5. doi: 10.1016/S0140-6736(18)30421-5. [Epub ahead of print] PubMed PMID: 29477249.
Copyright © 2018, Simple and Practical Mental Health. All rights reserved. May not be reproduced in any form without express written permission.
Disclaimer: The content on this website is provided as general education for medical professionals. It is not intended or recommended for patients or other laypersons or as a substitute for medical advice, diagnosis, or treatment. Patients must always consult a qualified health care professional regarding their diagnosis and treatment. Healthcare professionals should always check this website for the most recently updated information.
Thank you for your analysis and comments. To me the one useful finding, as mentioned by Dr. O’Reardon above, is that antidepressants do work. However it is mentioned in the paper’s discussion that the effect sizes were moderate. In addition, “46 (9%) of 522 trials were rated as high risk of bias, 380 (73%) trials as moderate, and 96 (18%) as low; and the certainty of evidence was moderate to very low.”. This isn’t very convincing, so I find their conclusion lacking in nuance.
I will continue to recommend psychotherapy, exercise and behavioral activation as first line to most patients with mild to moderate depression who are interested candidates with access to such treatment, as I remain convinced that often the work done in therapy might protect from recurrence by changing behavioral patterns and life choices.
For those who will need or prefer a medication, my choice will continue to be guided, as mentioned by JPA above, by side effect profiles, drug interactions, comorbidity…and obviously patient choice.
The results of network meta-analyses which use robust statistical method and so called mixed evidence represent relative “hierarchy” of medication. They are taking into account direct and indirect comparisons of effect sizes, and are conducted only if there is low to moderate heterogeneity and moderate to high consistency of included studies. It is clear that, in this way, increase in sample size increases power and precision of estimate. Since network metaanalyses produce relative “order” of efficacy they seem to be very useful and informative for treatment guidelines and clinical practice. Clinicians need to know, at least with some certainty which drug is probably the best, second best, third best and so on. For example Canadian and British Guidelines for the treatment of depressive and bipolar disorders respectively, incorporated into their treatment recommendations results of certain network meta-analyses. Hence, results of this network-meta-analysis should be carefully reviewed and its implications fully appraised. The “hierarchy” of antidepressants, based on this network metaanalysis goes as follows : Amitriptyline, Mirtazapine Duloxetine, Venlafaxine, Paroxetine, Milnacipran, Fluvoxamine, Escitalopram, Nefazodone, Sertraline, Vortioxetine, Agomelatine, Vilazodone Levomilnacipran Bupropion, Fluoxetine, Citalopram, Trazodone, Clomipramine, Desvenlafaxine, Reboxetine (ORs ranging between 2,13 for amitriptyline and 1,37 for reboxetine). However, it is questionable whether thereare any meaningful differences in efficacy ( ORs) between Paroxetine, Milnacipran, Fluvoxamine, Escitalopram, Nefazodone, Sertraline, Vortioxetine or Agomelatine. After all, it is interesting that summary SMD for all antidepressants was only 0, 30.
Thanks for highlighting and critiquing it as it has been having a big impact on public discourse.
I think the most impressive element is the sample size of 116,000 which allows psychiatrists to put to rest finally in the public domain the old shibboleth that antidepressants don’t work and that it is fraud by psychiatrists and pharmaceutical companies (David Healy, Thomas Szasz et al.). That in itself is a great thing. When you have shown they work in 522 RCTs in 116,000 patient what more debate does there need to be on this question?
Re agomelatine, I happen to know a psychiatrist who sat on the European panel for Drug Approval and he told me that the opinion of the panel on agomelatine was that it was “was as safe and effective as an M&M.”
If MAOIs came out in 1955 why was one of them or several not included? Prozac did not come out until more than 30 years later in 1987.
I don’t think there is anything very persuasive to tell us to use one individual antidepressant over an other.
The main virtue is my point in the first paragraph.
John O’Reardon, MD (johnoreardonmd.com)
Agree with Dr. Mago and the members of the editorial board. Side effect profile, drug interactions, and co-morbidity are the most is the most important factors in my selection of an antidepressant. Furthermore, if the statistical analysis is correct, I am not sure that what averages out to best in 100,000 patients is going to be appropriate for the unique patient in my office.