Between 2012 and 2024, the FDA approved 22 new psychiatric drugs and supplemental indications. The market grew. Press releases announced breakthroughs. Patients were told they had more options than ever.
Then a team from Stanford and Yale ran them through an independent rating system. The result: zero were rated clinically helpful.
The Study
Havlik et al. published in the Journal of Clinical Psychiatry (January 2026) a quantitative analysis of every FDA-approved psychiatric drug from 2012 to 2024. They didn't just count approvals — they measured clinical utility using Prescrire, an independent French drug bulletin that takes no industry funding.
Prescrire's rating scale runs from "Bravo" (major therapeutic advance) down through "Offers an advantage," "Possibly helpful," "Nothing new," "Not acceptable," and "Judgment reserved." Historically, about 10% of all drugs earn ratings in the top three categories.
For the 22 psychiatric approvals in this window: zero reached the top three categories.
Havlik et al. — 22 FDA-Approved Psychiatric Drugs, 2012–2024
That "judgment reserved" category deserves emphasis. It doesn't mean "we haven't looked yet." It means the data were too poor in quality for Prescrire to determine whether the drug works. Nineteen of 22 psychiatric drugs approved by the FDA were backed by evidence so weak that an independent evaluator couldn't form a judgment.
The Gap No One Talks About
The FDA asks one question: Is this drug statistically significantly better than placebo on a rating scale?
Prescrire asks a different question: Does this drug help patients more than what already exists?
These are not the same question. A drug can clear the FDA bar — statistically separating from sugar pills — while offering patients nothing they couldn't already get. The absolute difference can be too small for patients to notice. The comparison group can be placebo instead of the current standard of care. The trial duration can be too short to reveal what happens in practice.
This gap between "approved" and "useful" is structural. It isn't an accident. It's the design of the system.
Fifty Years of the Same Drugs
David Nutt, writing in The Lancet Psychiatry (March 2025), put the Havlik findings in historical context: "The past 50 years have seen remarkable advances in pharmacological treatments for hypertension, immune disorders, and cancer... The same cannot be said for psychiatric medicines."
Most psychiatric drugs prescribed today are functionally descendants of compounds discovered by accident in the 1950s. SSRIs refine the monoamine hypothesis. Atypical antipsychotics refine dopamine blockade. The mechanisms are the same. The molecular structures are tweaked. The clinical utility improvement is, according to the independent data, zero.
Nutt's explanation: we never understood the causal mechanisms of psychiatric disorders well enough to design drugs rationally. The NIMH's Research Domain Criteria (RDoC) project — meant to replace symptom-based diagnosis with neuroscience-based categories — hasn't produced new treatments. Hundreds of billions in R&D. Same drug classes.
The Third Position
Peter Gøtzsche, writing in response to Nutt (October 2025), argued that Nutt doesn't go far enough. Nutt's proposed solutions — more investment, fewer regulatory barriers — assume the enterprise is worth saving. Gøtzsche's position: psychiatric drugs as a class do more harm than good, and calling for better versions of fundamentally flawed tools is the wrong frame.
Three positions, then. Industry says innovation is happening. Nutt says it's stalled and needs resuscitation. Gøtzsche says the patient is better off without the treatment. The Havlik data doesn't resolve the debate, but it makes the first position — that approvals mean progress — very difficult to maintain.
The Honest Counterpoints
This picture is not entirely bleak, and intellectual honesty demands the exceptions.
| Drug | Mechanism | What Happened | But... |
|---|---|---|---|
| Cobenfy | Muscarinic M1/M4 agonist — first non-dopamine schizophrenia drug in decades | FDA approved Sept 2024. Genuinely novel. Avoids weight gain and EPS. | Prescrire hasn't rated it (US-only). Early real-world reports show high discontinuation. Long-term data limited. |
| Esketamine | NMDA antagonist — different from monoamines | FDA approved 2019 for TRD. Real-world data confirms some effectiveness. | NICE rejected it (TA854, Dec 2022). Modest effects. Needs repeated in-clinic dosing. |
| Zuranolone | Neurosteroid GABA modulator | FDA approved Aug 2023 for postpartum depression. Oral, 14-day course. | FDA rejected it for MDD. Predecessor brexanolone withdrawn Jan 2025 (low sales). |
| COMP360 Psilocybin | Classic psychedelic — 5-HT2A agonist | Both Phase 3 trials positive (Feb 2026). Rolling NDA planned Q4 2026. | Falls outside the 2012–2024 window. Not yet approved. Came from outside the standard pipeline. |
Cobenfy is the strongest counterpoint. It's a genuinely novel mechanism, within Havlik's window, with real clinical differentiation — PANSS reduction of -21.2 vs. -11.6 for placebo, with a moderate effect size (Cohen's d = 0.61). Early real-world experience shows roughly 40% of patients demonstrating meaningful improvement in add-on use, with 52-week extension data showing sustained benefit. But its Prescrire "judgment reserved" rating may reflect a geographic limitation — US-only approval, unavailable for EU evaluation — rather than clinical failure. A genuine wrinkle in the methodology that Havlik's study can't resolve.
COMP360 psilocybin is the most promising development in the entire field — both Phase 3 trials positive, rolling NDA planned. But it's not yet approved, falls outside the study window, and came from academic research on Schedule I compounds, not the pharmaceutical pipeline. The fact that the most credible psychiatric innovation in a generation came from outside the system is itself data about the system.
The Market Doesn't Care
While clinical utility flatlined, the market grew. The global psychiatric medications market is projected to reach $78.9 billion by 2033 (CAGR 5.2%). Antidepressants alone represent 44% of the psychotropic drug market.
This is the signature of a regulatory surrogate at work. Approval is the event that generates prescribing, marketing, and revenue. Whether the approved drug represents a clinical advance over existing options is a separate question — one the approval system was not designed to answer and the market has no incentive to ask.
The Regulatory Surrogate — Mechanism #10
When a proxy metric replaces the thing it was meant to measure, the system optimizes for the proxy. FDA approval was designed as a safety and efficacy gate — a minimum standard. Over time, it became interpreted as a signal of clinical value. Approval leads to "new option" leads to prescribing leads to revenue. The question "Is this better than what we already have?" drops out of the loop.
This is Goodhart's Law at the scale of a $50+ billion industry: when a measure becomes a target, it ceases to be a good measure.
What This Means
The Havlik study doesn't claim that psychiatric drugs are useless. Older drugs — lithium for bipolar disorder, clozapine for treatment-resistant schizophrenia — have strong evidence bases. What the study shows is that 13 years of new approvals added nothing to the clinical toolkit, as measured by the most rigorous independent evaluator available.
This is not an argument against pharmaceutical research. It's an argument against mistaking approvals for progress. When the system produces 22 approved drugs and zero clinical advances, the system is optimizing for something other than patient outcomes. The approval-utility gap isn't a failure of individual drugs — it's a structural feature of how psychiatric drug development currently works.
A note on sourcing: the Havlik study has been amplified by anti-psychiatry outlets and organizations with ideological agendas. The data doesn't need ideological framing. It was published in the Journal of Clinical Psychiatry by researchers at Stanford and Yale. It stands on its primary source. Cite the JCP paper directly.