The Serotonin Theory of Depression: How to Read an Umbrella Review and Its Rebuttals

The short version

An umbrella review is a review of reviews: it gathers the existing systematic reviews on a question, grades their quality, and reports what the assembled evidence supports. That design answers one narrow thing well, whether a specific claim holds up across the literature, and the 2022 umbrella review by Joanna Moncrieff and colleagues in Molecular Psychiatry used it to conclude there is "no consistent evidence" that low serotonin causes depression. What it did not do, and did not set out to do, was test whether antidepressants work. Mechanism and efficacy are separate questions, and keeping them apart is most of the skill of reading this field.

What the design can and cannot settle

An umbrella review inherits the strengths and weaknesses of the reviews it summarizes. It rarely reaches back into raw patient data; instead it appraises whole bodies of evidence, in this case studies of serotonin metabolites, receptor and transporter imaging, tryptophan depletion experiments, and genetic association work. When those underlying literatures are small, inconsistent, or confounded, the review will say so, and the Moncrieff group did: weak and conflicting findings across the serotonin domains, plus a note that long-term antidepressant use might itself lower serotonin markers.

Here is the limit. "No consistent evidence for a causal claim" is not the same as "the claim is false," and it is different again from "the treatments built around the claim do not help." A review can only weigh the evidence that exists. Absence of a clean signal can mean the signal is not there, or that the measurements are too crude, or that depression is too varied for a single-molecule story. This is why the grading step matters so much: a formal framework like GRADE forces reviewers to state how confident the pooled evidence lets them be, and honest umbrella reviews report which alternatives they can rule out and which they cannot. Careful readers hold the conclusion at the altitude the design supports.

Two questions people keep merging

The first is mechanistic: is depression caused by a serotonin deficiency? The second is therapeutic: do drugs acting on serotonin reduce depressive symptoms? A drug can work while the popular story about why it works is wrong. Aspirin relieved pain for decades before anyone understood prostaglandins.

The therapeutic question has its own evidence base. The largest network meta-analysis of acute antidepressant treatment, by Andrea Cipriani and colleagues in The Lancet in 2018, pooled 522 trials and more than 116,000 participants across 21 drugs and found all of them more effective than placebo, with odds ratios running from about 1.37 to 2.13 and certainty rated moderate to very low. That is a real average effect of modest size, and it rests on its own trials, not on the serotonin hypothesis. The Moncrieff review did not analyze that literature, which is why treating it as a verdict on whether medication helps asks the design to carry weight it was never built for.

How the 2022 review was contested

The review drew unusual public attention because it landed on a phrase millions of people had heard from advertising and clinicians alike, the idea of a serotonin "chemical imbalance." The scientific pushback was fast and largely methodological. In Molecular Psychiatry, Sameer Jauhar and a large group of co-authors published a reply titled "A leaky umbrella has little value," arguing that the review applied its own rules unevenly. Their charges were specific: a modified quality-grading scheme introduced after the fact, selective inclusion of primary studies, omission of some large effect sizes in the tryptophan-depletion work, and a description of 5-HT1A receptors as presynaptic autoreceptors when many are postsynaptic. A separate reply from Lucie Bartova, Siegfried Kasper and colleagues in the same journal argued that molecular imaging findings show the serotonin system is involved in depression, and that the review understated them.

Notice what the critics did and did not concede. Few defended the crude "chemical imbalance" slogan as a full account of depression; several called it a simplification that outran the science. The dispute was about whether the grading was fair, whether the primary studies were read correctly, and whether the framing implied more about treatment than the data could support. Moncrieff and colleagues defended their appraisal and argued that the replies reasserted the hypothesis rather than supplying the consistent evidence the review had gone looking for.

How to read the exchange

A few habits help. Separate the claim being tested from the claim being defended: the review tested a causal deficiency model, while some replies defended the looser idea that serotonin is "involved," which is easier to support and harder to falsify. Watch the grading rules, because umbrella reviews live or die on how they rate evidence quality, so a charge that the rules were applied inconsistently is serious and deserves a direct answer rather than a restatement of belief. Check scope: a review that excludes efficacy trials cannot speak to efficacy in either direction.

None of this tells an individual what to do. Whether a particular person benefits from a particular treatment is a clinical question settled with a qualified prescriber, and this article is educational, not medical advice. The scientific point is narrower and more durable: the mechanism behind a symptom and a drug's effect on that symptom are separate empirical questions, tested with separate methods. The 2022 review and its rebuttals are a useful case study because the loudest reactions, on every side, collapsed the two.