The Imbalance That Was Never Found

In 2022 Joanna Moncrieff and colleagues published an umbrella review (a review of reviews) of the evidence that depression is associated with lowered serotonin.

They found none that held up. Not in serotonin metabolite levels, not in receptor binding studies, not in studies depleting tryptophan in healthy volunteers, not in the genetics.

The coverage was enormous and mostly framed as a scandal about antidepressants, which is the wrong conclusion drawn from a real finding.

Start with where the serotonin story came from, because it was never a discovery. In the 1950s two drugs were noticed to affect mood: iproniazid, a tuberculosis treatment, lifted patients' spirits, and reserpine, used for blood pressure, was associated with depression. Both affect monoamines. The monoamine hypothesis was proposed as an inference from the drugs backwards to a cause, if a drug that raises serotonin helps, perhaps low serotonin is the problem.

That inference is not valid, and its invalidity is easy to see with a different example. Aspirin relieves headaches, and headaches are not caused by an aspirin deficiency.

The hypothesis was then adopted with enthusiasm by pharmaceutical marketing in the 1990s, because "correcting a chemical imbalance" is a comprehensible story that positions a pill as the obvious remedy. It appeared in advertisements, in patient leaflets and in a great deal of journalism. It was always a simplification that the research literature had substantially abandoned.

So what about the drugs?

Trials measure outcomes, not mechanisms. A drug can work while the story about why it works is wrong, which has been true of most drugs for most of history. Cipriani's 2018 network meta-analysis covered 522 trials and found all twenty-one examined antidepressants more effective than placebo, with effect sizes that were modest. Irving Kirsch's analyses, using unpublished as well as published FDA trial data, found the drug-placebo difference falling below conventional thresholds for clinical significance except in the most severely depressed patients.

These two results are compatible, and together they say something reasonably clear: a real average effect, small, distributed unevenly across severity, and substantially oversold for mild depression. That is a more useful summary than either "antidepressants don't work" or "antidepressants correct an imbalance".

What might actually be going on? The most productive current account is about plasticity rather than about levels of anything.

Chronic stress demonstrably reduces dendritic branching and neurogenesis in the hippocampus, and effective treatments (SSRIs, exercise, electroconvulsive therapy, ketamine) increase them. This account has a specific advantage: it explains the delay. Serotonin levels in the synapse change within hours of the first SSRI dose, and symptoms take two to six weeks to shift. A levels account has never had a good explanation for that gap. A plasticity account predicts it, because growing connections takes time.

A second strand concerns inflammation. Somewhere between a quarter and a third of depressed patients show elevated inflammatory markers. Giving healthy volunteers an inflammatory challenge reliably produces low mood, fatigue, social withdrawal and disturbed sleep: the syndrome, arriving from the immune system. Patients treated with interferon develop depression at high rates. Trials of anti-inflammatory agents show benefit concentrated in patients with raised markers and little in those without.

That pattern is what you would expect if "depression" names several different conditions that share a description. Which is also what the comorbidity and the uneven treatment response have been suggesting all along.

Now the part that the biological framing tends to obscure. Poverty, unemployment, isolation, discrimination and childhood adversity are among the strongest predictors of depression that anyone has measured, and it is a mistake to file this under "social causes" as though it were an alternative to a biological account.

Chronic stress acts through cortisol, through inflammation, through sleep architecture and through the plasticity changes described above. The social causes are biological causes; they are simply upstream. Treating "biological" and "social" as competing explanations makes the causal chain invisible rather than choosing between two options, and the chain is where every intervention actually acts.

Finally, an uncomfortable finding about how to talk about this.

Telling patients their depression is a chemical imbalance was recommended for years partly on compassionate grounds. It reduces self-blame, and depressed people blame themselves relentlessly. It does reduce blame.

It also, reliably across studies, increases pessimism about recovery. People given biogenetic explanations rate their condition as more permanent, less responsive to their own efforts and longer-lasting. Clinicians shown biologically framed case descriptions report less empathy for the patient than when the same case is framed in terms of life events.

So the kinder-sounding explanation carries a cost that nobody intended and which shows up in exactly the belief, that nothing I do will change this, that the treatment has to work against. Which is an argument for being accurate rather than for choosing whichever story sounds most consoling: that depression is common, that it responds to several different kinds of intervention, and that it usually lifts.