Almost every week, someone sitting across from me in Maysville tells me they have a chemical imbalance. They say it the way you would report a lab value, as settled fact. Usually a clinician told them so years ago, the phrase lodged, and nobody has revisited it since.
I understand why it lodged. It is a clean story. Serotonin runs low, the medication tops it back up, mood recovers. It borrows the shape of insulin and diabetes, which everyone already understands, and it hands a frightened person an explanation that is not a failure of character. As a piece of communication it was enormously effective.
There is one problem. The evidence for that specific story was never very good, and psychiatry has known it for longer than it has said out loud.
The serotonin story, and what happened to it
In July 2022, Joanna Moncrieff and colleagues published an umbrella review in Molecular Psychiatry pulling together the major strands of evidence for the serotonin hypothesis: serotonin metabolite levels, receptor and transporter imaging, tryptophan depletion studies, and genetic association work. Their conclusion was that none of it consistently supported the idea that depression is caused by low serotonin.
The paper landed hard, partly because the press coverage went considerably further than the paper did. Rebuttals followed quickly, and the methodological argument about how much weight an umbrella review of heterogeneous evidence can carry is still live. What has not been seriously contested is the narrower point. The simple version of the chemical imbalance story, the one my patients repeat back to me, is not something the evidence supports.
This is usually where people expect me to say the medications do not work.
They do work. That is the part the coverage mangled, and it is worth being precise about why the two claims sit together comfortably.
The delayed onset problem
Consider a fact that has sat awkwardly in the middle of this field for fifty years.
Give someone an SSRI and synaptic serotonin availability rises within hours. Give the same person two weeks and they usually feel no better. The clinical response, when it arrives, tends to land somewhere between weeks four and eight.
Were raising serotonin the mechanism, that gap should not exist. Aspirin does not take a month to work on a headache. The delay tells us fairly clearly that the neurotransmitter change sits upstream of whatever actually produces the benefit, and that the real work is something slower.
The leading account is that the slower process is structural. Chronic stress and depression are associated with dendritic atrophy in the hippocampus and prefrontal cortex, and antidepressants appear to promote synaptic remodeling and increase BDNF signaling over roughly the timescale on which patients start to improve. That evidence is strong in animal models and more indirect in humans, so I would call it the best current explanation rather than a settled one.
The rapid antidepressant effect of ketamine, which can lift mood within hours through a glutamatergic route entirely separate from the monoamines, points in the same general direction. Something about synaptic plasticity, rather than the resting level of any one transmitter, is doing the work.
Why the wrong story mattered
An inaccurate explanation might seem harmless as long as the treatment helps. It was not harmless.
Patients told they have a permanent chemical deficiency tend to conclude they will need medication forever, and research on causal beliefs in depression has generally found that purely biological explanations are associated with more pessimism about recovery rather than less stigma. The story meant to relieve blame quietly removed agency along with it.
It set up a particular kind of disappointment too. Someone who believes a medication is correcting a deficiency has no framework for what to do when the first one fails, and a fair number of people conclude the problem must be them.
What the numbers actually look like
STAR*D remains the largest real world trial of sequential antidepressant treatment, and it is worth knowing what it found. Of the roughly 2,876 patients who started on citalopram, about a third reached remission on that first medication. Not response, which is a lower bar, but remission.
The cumulative figures across all four treatment steps are genuinely disputed. The original reports put eventual remission near 67 percent, and a 2023 reanalysis by Pigott and colleagues argued the real figure was closer to 35 percent once protocol deviations and dropouts were handled differently. Reasonable people still disagree about this.
What survives the dispute is the shape of the thing. A first antidepressant helps a substantial minority outright, many people need a second or a third, and needing to switch is an ordinary feature of the process rather than evidence of anything about you.
This is also why pharmacogenomic panels have disappointed. They tell us something real about how quickly you metabolize certain drugs through CYP2D6 and CYP2C19, which occasionally matters for dosing, and the large trials have not shown that they reliably predict which antidepressant will work. I do not order them routinely and would want a specific reason before I did.
Coming off them
Antidepressants do not produce cravings or dose escalation, so they are not addictive in the ordinary sense of that word. They do produce physical dependence, meaning the brain adapts to their presence and registers their removal.
How common that is has been argued over with some heat. A 2019 review by Davies and Read reported that a majority of people experience withdrawal and that nearly half describe it as severe, figures critics argued were inflated by drawing on samples of people already seeking help for withdrawal. A 2024 meta-analysis in Lancet Psychiatry by Henssler and colleagues, working from randomized data and subtracting the placebo arm, put the attributable incidence closer to 15 percent, with severe symptoms around 3 percent.
Both are probably telling the truth about different populations. My working position is that discontinuation symptoms are real, are commoner than the field admitted for a long time, and are largely avoidable.
The mechanism points straight at the fix. Receptor occupancy does not fall linearly with dose. It follows a hyperbolic curve, so the last few milligrams strip away far more occupancy than the first several did, which is exactly why people sail through the early reductions and then hit a wall at the end. Horowitz and Taylor made this argument in Lancet Psychiatry in 2019 and it has changed how careful prescribers taper. The final steps should be the smallest and the slowest, which is the reverse of how tapers were written for decades.
What I tell people instead
The honest version is less tidy than the chemical imbalance, and in my experience patients handle it perfectly well.
We have no biomarker for depression. We have a set of medications that outperform placebo in a way that is modest at the group level and sometimes transformative for an individual, that appear to work by encouraging the brain to remodel rather than by refilling a tank, and that we still match to a person by trial rather than by test. Alongside them we have psychotherapy, which works through a different route and combines well, and the ordinary inputs of sleep and movement and daylight, which are not decoration.
Being told the truth about uncertainty tends to make people more willing to stay in treatment, not less. It also turns the eventual conversation about stopping into a normal one.
We handle medication management in person at our office on West Second Street in Maysville and by telehealth across Kentucky and Ohio, for patients ages six and up. Anyone who wants to talk it through first is welcome to book a free fifteen minute consultation, or call us at (606) 714-0056.