The 2018 Lancet network meta-analysis by Cipriani and colleagues remains the largest evidence synthesis in psychiatry: 522 double-blind randomised trials, 116,477 patients, 21 antidepressants — every one of them more effective than placebo for acute major depression, with odds ratios from about 1.4 to 2.1. The average effect size is modest (a standardised mean difference around 0.3), placebo response in depression trials is unusually large, and roughly a third of the benefit measured in published trials evaporated when unpublished trials were added back in a landmark 2008 New England Journal analysis. 'Modestly better than a strong placebo, for the average trial patient' is the accurate summary — a sentence that satisfies neither the drugs' marketers nor their abolitionists.

The 2022 umbrella review led by Joanna Moncrieff in Molecular Psychiatry found no consistent evidence that depression is caused by low serotonin — and the honest response from academic psychiatry was that it had quietly stopped believing the simple version decades earlier. The problem is that patients had not: surveys find around 88% of Americans believe antidepressants correct a chemical imbalance, a narrative pharmaceutical advertising built deliberately. Crucially, the collapse of the mechanism story does not decide the efficacy question — paracetamol relieved pain for a century before its mechanism was understood. The two claims — 'the serotonin-deficiency story was oversold' and 'the drugs do nothing' — are routinely fused in public debate and must be kept apart.

Trial averages conceal enormous heterogeneity. Evidence suggests benefit concentrates in more severe depression, while in mild depression drug-placebo differences shrink toward clinical irrelevance — which is why NICE and most guidelines put psychological therapy first for mild cases. The famous STAR*D pragmatic trial reported cumulative remission near 67% after four treatment steps, a figure a contested 2023 BMJ reanalysis argued falls to roughly 35% under the original protocol — a dispute about outcome-switching that remains genuinely unresolved. Long-term outcomes, optimal treatment duration and effects in children and adolescents (where evidence is weakest and fluoxetine stands nearly alone) are open questions, not settled ones.

On 21 July 2025 the US FDA convened an expert panel on SSRIs in pregnancy in which nine of ten panellists were established public critics of the drugs, including Moncrieff and David Healy. Panellists cited studies without appropriate controls and, outside experts noted, rarely weighed the risks of untreated depression — despite suicide being a leading cause of maternal death in the first postpartum year. The American College of Obstetricians and Gynecologists, the Society for Maternal-Fetal Medicine and ultimately 114 organisations formally objected that the panel misrepresented evidence. A legitimate scientific debate about effect sizes and withdrawal now has a parallel political track, and readers should know which one they are watching.

The credible core of the critical position is real: effect sizes were inflated by publication bias; the chemical-imbalance narrative was misleading and aided overprescription; withdrawal was denied for two decades (see the antidepressant withdrawal hub); prescribing has drifted far beyond the severe depression where evidence is strongest, with around one in six adults in some Western countries taking the drugs; and trials are too short to inform the multi-year use that is now typical. Acknowledging all of this is compatible with the drugs being genuinely useful — and is precisely what distinguishes evidence-based criticism from the claim that psychiatry knowingly pushes placebos.

For moderate-to-severe depression, antidepressants are a reasonable, evidence-supported option, ideally alongside psychological therapy, with the largest benefits in the sickest patients. For mild depression, guidelines prefer non-drug approaches first. Nobody should start one expecting a corrected chemical imbalance, and nobody should stop one abruptly — discontinuation belongs in the withdrawal hub. Pregnancy decisions deserve individualised risk-benefit discussion with a clinician, not a verdict from either a marketing department or an advocacy panel: untreated perinatal depression carries its own well-documented risks to mother and child.