Here's a fact both sides of the antidepressant debate agree on: the drugs beat placebo by a standardized mean difference of about 0.30. One side calls this proof they work. The other calls it clinically meaningless. A 2026 Cochrane review just added a third voice: exercise performs about as well as either. The same numbers. Three incompatible conclusions. Welcome to the fault line.
The Largest Meta-Analysis Ever Conducted
In 2018, Andrea Cipriani and colleagues published the definitive meta-analysis in The Lancet. They examined 522 randomized controlled trials covering 116,477 patients across 21 antidepressant drugs. The headline was triumphant: all 21 antidepressants were more effective than placebo.
This was reported worldwide as settling the debate. Antidepressants work. Case closed.
Except for a number buried in the results: the overall effect size was an SMD of 0.30. The strongest drug, amitriptyline, managed 0.48. These are not large numbers. In clinical terms, 0.30 translates to roughly two points on the Hamilton Depression Rating Scale — a 17-item questionnaire scored from 0 to 52.
The Two-Point Problem
Irving Kirsch had been pointing at this number for a decade. His 2008 meta-analysis of FDA-submitted trial data — including the unpublished trials the drug companies preferred to forget — found an almost identical effect size (SMD 0.32). He'd been saying since then that this gap was clinically meaningless.
How meaningless? The UK's National Institute for Health and Care Excellence (NICE) set three points on the Hamilton scale as the minimum for clinical significance. Kirsch's and Cipriani's numbers both fall below it. Research comparing Hamilton scores to clinical impression ratings suggests a physician wouldn't even notice a two-point difference — you need at least seven or eight points for a patient to register as "minimally improved."
Here's where it gets interesting: Cipriani's team agreed with the number. They just disagreed about what it meant. Their response was that many accepted medical treatments have similar effect sizes. In other words: the drugs work about as well as lots of things we consider effective.
Kirsch's response: that's not the reassurance you think it is.
The Severity Threshold
Kirsch's most provocative finding wasn't just the small effect size — it was where the effect appeared. When he broke the data down by initial severity, there was virtually no drug-placebo difference for mild and moderate depression. The drugs only beat placebo meaningfully for the most severely depressed patients (Hamilton scores above 28).
And even then, the apparent effectiveness wasn't because severely depressed patients responded better to the drugs. It was because they responded worse to placebo. The drug response stayed roughly flat across severity levels. What changed was the placebo response — it dropped as depression got worse.
The implication: for the majority of people prescribed antidepressants — those with mild to moderate depression — 75% of the improvement they experience would have happened with a sugar pill.
The Serotonin Gap
If antidepressants barely outperform placebo for most patients, why do so many people feel they help?
For decades, the answer was the serotonin hypothesis: depression is caused by low serotonin, and SSRIs fix it. In 2022, Joanna Moncrieff and colleagues published a comprehensive umbrella review in Molecular Psychiatry concluding there was "no convincing evidence that depression is caused by serotonin abnormalities."
This made international headlines as a bombshell. But here's the contradiction within the contradiction: most researchers already knew this. The scientific response to Moncrieff wasn't shock — it was irritation. The simple serotonin hypothesis had been considered outdated for decades in the research community. The paper debunked something scientists had quietly moved past years ago, but that patients, prescribers, and the public still believed.
The gap between what researchers knew and what patients were told is itself a contradiction. For years, doctors explained to patients that depression was "a chemical imbalance" and that SSRIs would correct it. This was, at best, a useful simplification. At worst, it was a marketing narrative that outlived its evidence by decades.
Enter the Treadmill
In January 2026, the Cochrane Collaboration published its latest updated review on exercise for depression — 73 randomized controlled trials, 4,985 participants. The conclusion: "there is little to no difference in the reduction of symptoms of depression provided by exercise compared to those provided by psychological therapies or antidepressants."
Read that again. Not "exercise helps a bit." Not "exercise is a useful complement." Exercise — walking, running, swimming, lifting weights — produces outcomes statistically indistinguishable from the drugs.
This wasn't a lone finding. Multiple meta-analyses from 2025 and 2026 converge on the same result, though with an important caveat: the evidence quality is rated "low" because exercise trials are hard to blind (you know if you're running), sample sizes tend to be small, and the people who volunteer for exercise studies may not represent the broader depressed population.
But here's what makes the comparison devastating: exercise comes with cardiovascular benefits, improved sleep, reduced inflammation, better metabolic health, and no withdrawal effects. Antidepressants come with weight gain, sexual dysfunction, emotional blunting, and withdrawal syndromes that can persist for months. When two interventions produce similar depression outcomes but radically different side-effect profiles, the tie doesn't go to the pill.
Follow the Prescriptions
If the evidence is this muddled, you'd expect prescription rates to plateau or decline while the field sorts it out. The opposite has happened.
The global antidepressant market was valued at $13.8 billion in 2023 and is projected to reach $30.5 billion by 2033. In the United States, 11.4% of adults now take antidepressant medication — up from 9.8% in 2019. Among young people aged 12 to 25, antidepressant prescriptions increased by 66% from 2016 to 2022, with an even sharper spike for teenage girls (up 130% since the pandemic).
Meanwhile, a study published in the Journal of Clinical Psychiatry found that all 22 psychiatric drugs approved by the FDA between 2012 and 2024 failed to demonstrate meaningful clinical advantages over older options. The market is growing, the patient population is expanding, and the drugs aren't getting better.
Who's doing the prescribing? Increasingly, not psychiatrists. By 2020, nurse practitioners prescribed the most antidepressant prescriptions (33%), surpassing psychiatric specialists (26%) and family medicine physicians (14%). The medications are being prescribed more widely, by less specialized providers, for less severe cases, while the evidence for their effectiveness in those cases remains weakest.
What the Contradiction Looks Like
This isn't a simple story of one side being right and the other being wrong. Both sides are using the same data. The collision is in the interpretation:
- Cipriani says: An SMD of 0.30 is real and comparable to other accepted treatments. Antidepressants work.
- Kirsch says: An SMD of 0.30 falls below clinical significance thresholds. The improvement is real — but it's mostly placebo.
- Cochrane 2026 says: Exercise produces equivalent outcomes. If antidepressants "work," so does a jog.
- The market says: Prescriptions are up 66% among young people and the industry is heading toward $30 billion.
- The mechanism says: The theory that justified these drugs was wrong. We're still not sure why they do what they do.
This is what a knowledge fault line looks like. Not a clean break between truth and falsehood, but a place where the same evidence supports contradictory narratives — and the narrative with the most money behind it wins the prescription pad.
The Uncomfortable Middle
None of this means antidepressants don't help anyone. For severe depression, the evidence is clearer: the drugs do separate from placebo in ways that matter. People in crisis need options, and medication is one. The newer drugs targeting novel pathways — kappa opioid antagonists, glutamate modulators, brain stimulation devices — may eventually change the calculus.
But for the millions of people with mild to moderate depression who are prescribed SSRIs as a first-line treatment — which is most antidepressant users — the evidence says something the $30 billion industry has no incentive to broadcast: a structured exercise program would likely produce the same outcome, with a better side-effect profile, at zero pharmaceutical cost.
The gap between what the evidence says and what the prescriptions reflect isn't a conspiracy. It's a system doing what systems do: following the path of least resistance. Pills are faster to prescribe than exercise programs. Insurance covers medication but rarely covers a gym membership. Pharmaceutical companies market drugs; no one markets running. And a 15-minute appointment ends more naturally with a prescription than with a training plan.
The contradiction persists not because anyone is lying about the data, but because the data alone was never what drove the prescriptions.