Do Antidepressants Work Better Than Placebo?

Do antidepressants work? There’s a lot of stigma around psychiatric medications, and plenty of people would argue that no, antidepressants don’t work. There are also people, that argue that medication is life-saving. Those are personal experiences, but what does the research have to say?

A recent paper published in the Lancet, and reported on in the media, looked at what a large number of studies had to say about the effectiveness of antidepressants. Since media outlets don’t necessarily have strong research literacy, let’s look at what the paper itself has to say. The post on research literacy explains some of the terms I’ll be using here.

A systematic review involves collecting a body of relevant literature on a topic using academic databases and search terms that are clearly specified, then narrowing that down by applying certain criteria to find studies that are academically rigorous and fit with the research parameters being considered. Typically, different investigators go through this process independently, then come to a consensus on which studies to include in the review. The results are then evaluated to get a picture of the current state of the evidence.

A meta-analysis goes a step further by pooling the data from the various studies and then performing statistical analysis. This information is then used to answer the question – do antidepressants work?

Inclusion criteria for the meta-analysis

• randomized, double-blinded, controlled trials (either placebo-controlled or head-to-head trials of different antidepressants)
• study participants were adults 18+
• diagnosis of major depressive disorder
• no more than 20% of participants in a study had bipolar disorder, treatment-resistant depression, psychotic depression, or serious concurrent medical condition (while this might sound like a bad thing, when pooling numbers for a meta-analysis, you want to make sure you’re comparing apples to apples)
• evaluation of quality of evidence and risk of bias met specified academic standards

The outcome measures were response rate and acceptability (as measured by the number of discontinuations due to side effects). While ideally patients should be treated to full remission of symptoms, response rate is often used in research studies. Response rate is defined as a 50% reduction in score on a standardized depression rating scale like the HAM-D.

For this meta-analysis, they chose to evaluate outcomes at the 8-week point. For included studies that didn’t take ratings at 8 weeks, this was imputed using statistical methods. This time frame matters because it takes at least 4 weeks to clearly see the therapeutic effect of antidepressants; earlier on is mostly side effects.

Findings

• All 21 antidepressants considered were more effective than placebo in adults with major depressive disorder.
• In head-to-head studies, agomelatine, amitriptyline, escitalopram, mirtazapine, paroxetine, venlafaxine, and vortioxetine were more effective than other antidepressants.
• Fluoxetine, fluvoxamine, reboxetine, and trazodone were the least efficacious drugs.
• For acceptability, agomelatine, citalopram, escitalopram, fluoxetine, sertraline, and vortioxetine were the most tolerable.
• The most discontinuations occurred with amitriptyline, clomipramine, duloxetine, fluvoxamine, reboxetine, trazodone, and venlafaxine.
• Agomelatine was the only antidepressant with a lower dropout rate than placebo.

Information the analysis does NOT give us

It’s important to recognize that the meta-analysis only tells us what it tells us. We can’t assume answers to questions that weren’t even addressed. These are some of the issues that the study didn’t address, and therefore no conclusions about those elements can be extrapolated from the results.

• How effective these antidepressants are in terms of achieving full remission
• Efficacy measures are only based on a 50% reduction in rating scale scores; there is no information about which particular clusters of symptoms are more or less likely to respond
• What outcomes are beyond the 8-week point
• Whether a specific antidepressant is or is not likely to be a) effective in a specific individual or b) more or less effective than any other antidepressant in that same individual
• Whether a specific antidepressant is or is not likely to be tolerable for a specific individual
• How often people experience side effects due to any of the antidepressants (it only looks at discontinuations due to intolerable side effects)
• Whether any specific individual would or would not benefit from antidepressants at all
• How well antidepressants work outside of the population considered in the review (i.e. under 18 years old or with treatment-resistant depression, psychotic depression, or bipolar depression)

Do antidepressants work?

What does all of this mean—do antidepressants work? It’s important to keep in mind that a systematic review/meta-analysis such as this is only answering certain questions. The authors are deliberately comparing apples to apples so they can pool large groups of numbers and draw conclusions from that. There’s a lot of real-world information that it doesn’t give us, but it’s worth keeping in mind that it makes no claims that it is.

The authors do not suggest that their findings can be extrapolated to answer any of the questions I’ve mentioned that the paper doesn’t give us information about. What it does tell us is that antidepressants belong in our arsenal of available treatment strategies. Anything more specific than that always needs to be a collaborative decision between the individual and their treatment team.

Overview

Reference

Cipriani, A., et. al. (2018). 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. The Lancet, 391(10128), 1357-1366.