How a Network Meta-Analysis Ranks 21 Antidepressants

The short answer

A network meta-analysis does not crown a single best antidepressant. It pools hundreds of randomized trials into one connected web of comparisons, then estimates how each drug performs on two separate yardsticks: efficacy (how often symptoms respond) and acceptability (how often people stay on the drug rather than dropping out). The landmark 2018 Lancet analysis by Cipriani and colleagues did exactly this across 21 antidepressants, 522 trials, and 116,477 adults, and its central lesson is often lost in the headlines. The differences between active drugs are modest, the rankings carry wide uncertainty, and even the medicines that score well on both measures do not settle the choice for an individual.

What a network meta-analysis is doing

A standard meta-analysis pools trials that compared the same two things. The problem in psychiatry is that most antidepressants have never been tested directly against each other. Drug A was compared with placebo, Drug B was compared with placebo, and a trial of Drug A versus Drug C may not exist at all.

A network meta-analysis connects those trials. If A and B were each tested against a shared comparator, you can estimate A versus B indirectly through that common anchor, then combine the indirect estimate with any direct A-versus-B trials that do exist. Stack enough of these links together and you get a network in which every drug can be compared with every other, including pairs that were never studied side by side. The Cipriani team released the full dataset openly, which lets other researchers rerun the analysis and test its assumptions.

That power rests on a condition called transitivity: the trials being linked have to be similar enough, in patients, severity, and design, that borrowing strength across them is fair. When the populations differ a lot, the indirect comparisons can drift, which is one reason the method rewards careful reading over a glance at the top of a chart.

How the two rankings are built

Cipriani and colleagues measured two things that people often blur together.

Efficacy was response: roughly a halving of depression-scale scores over about eight weeks of acute treatment. On this measure every one of the 21 drugs beat placebo, with odds ratios reported from about 2.13 for amitriptyline down to 1.37 for reboxetine. In head-to-head data, a cluster including agomelatine, amitriptyline, escitalopram, mirtazapine, paroxetine, venlafaxine, and vortioxetine tended to rank higher.

Acceptability was the opposite lens: how many participants dropped out for any reason, a rough proxy for tolerability and real-world staying power. Here a different cluster, including agomelatine, citalopram, escitalopram, fluoxetine, sertraline, and vortioxetine, came out more favorable, while several effective drugs such as amitriptyline and venlafaxine carried higher dropout.

Notice what that split means. The most-efficacious list and the most-acceptable list overlap only partly. A drug can push symptoms harder and still be harder to stay on.

Why the ranking is not a leaderboard

The tempting move is to read the numbers as a ranked table and pick number one. The authors warned against exactly that, and the reasons are built into the method.

Rankings are point estimates wrapped in wide, overlapping credible intervals. When the uncertainty bands of the third-place and the tenth-place drug cross, the apparent order is mostly noise. Ranking statistics such as SUCRA can make a shaky order look crisp, flattering small gaps into decisive ones.

Evidence quality limits how hard the numbers can be pushed. The authors graded certainty as moderate to very low, and most trials carried moderate or high risk of bias. Many were short, industry-funded, and enrolled tidier patients than clinics usually see, so the estimates describe an average acute response in trial populations, not a forecast for one person with a particular history, comorbidities, and set of prior medication trials.

The analysis also covers acute treatment only. It does not resolve maintenance over months, individual side-effect profiles, drug interactions, or personal preference, all of which shape a real decision.

What the evidence actually supports

Read carefully, the study is genuinely useful. It confirms that antidepressants as a class separate from placebo for acute major depression, it shows that the differences among reasonable first choices are small, and it hands clinicians a defensible short list rather than a lone winner. A few drugs sit in a favorable region of both distributions in this dataset, escitalopram and vortioxetine among them. That placement simply marks where the two curves overlap in these particular trials. It is a pattern worth discussing with a clinician, never a verdict that either drug is objectively best or a reason to prefer it over an equally reasonable alternative.

That is the honest use of a ranking: a map of where the sensible options cluster, with the size of the gaps taken seriously. A claim that one antidepressant is definitively the best treats a probability distribution as a verdict, and the underlying data will not carry that weight.

This article is educational and not medical advice; decisions about antidepressants belong in a conversation with a qualified clinician who knows the individual.