Do antidepressants actually work? Two respected researchers disagree.
The honest answer sits between "antidepressants are miracle drugs" and "antidepressants are basically placebo." Here's what the largest meta-analysis found, why one influential critique disagrees, and why the most-cited real-world trial's headline number is more complicated than it looks.
The evidence landscape
| Study | N / Scope | Position | Finding | Source |
|---|---|---|---|---|
| Cipriani et al. 2018 | 522 trials, 116,477 participants | Supportive | All 21 antidepressants studied outperformed placebo for acute treatment of major depressive disorder; effect sizes varied notably by specific drug. | The Lancet 2018 |
| Kirsch et al. (ongoing critique) | Meta-analyses of FDA trial data, incl. unpublished | Skeptical | Argues the average drug-placebo difference is small and clinically marginal for mild-to-moderate depression once unpublished (often less favorable) trial data is included; effect is larger for severe depression. | Kirsch et al., PLOS Medicine 2008 |
| STAR*D (original, 2006) | 4,041 | Real-world effectiveness | Reported cumulative 67% remission rate across up to four sequential treatment steps. | Rush et al., Am J Psychiatry 2006 |
| STAR*D reanalysis, 2023 | Same dataset, reassessed | Methodological correction | Identified protocol deviations in how remission was counted across steps; a like-for-like remission estimate under the trial's own pre-registered criteria was roughly 35%, well below the widely cited 67%. | Pigott et al. reanalysis, 2023 |
What this means in plain language
"67% remission" is the number in marketing. It's not the whole story
STAR*D's headline 67% cumulative remission rate gets cited constantly because it's genuinely encouraging, but it counted remission cumulatively across four treatment attempts using criteria that a 2023 reanalysis found were applied inconsistently with the trial's own pre-registered protocol. The more conservative, protocol-consistent estimate is closer to 35%. Neither number is "the lie": 67% overstates first-line response, and 35% may understate the real-world value of persisting through multiple treatment attempts. Both are worth knowing.
The Cipriani-vs-Kirsch debate is a real scientific disagreement, not marketing spin
Cipriani's 522-trial meta-analysis is the most comprehensive comparative-efficacy dataset available and found a consistent placebo-beating effect across all 21 drugs studied. Kirsch's critique (using regulatory data that includes unpublished, less-favorable trials) argues the average effect is smaller than it looks in the published literature alone, particularly for milder depression. Both researchers are credentialed and published in top journals; treat this as an open, evolving debate rather than a settled question either direction.
Severity matters more than any single average number
Across most analyses, including Kirsch's, the antidepressant-vs-placebo gap widens as baseline depression severity increases. The debate is largely about mild-to-moderate depression; for severe depression, medication benefit is more consistently supported across both supportive and skeptical analyses.
Decision framework
| Your priority | What the evidence supports |
|---|---|
| Severe depression | Medication benefit over placebo is more consistently supported across both supportive and skeptical analyses. |
| Mild-to-moderate depression | Benefit is real on average but smaller and more debated: talk therapy alone or combined with medication is a reasonable evidence-based option to discuss. |
| First medication doesn't work | STAR*D (even under the more conservative reanalysis) supports that persisting through additional treatment steps meaningfully improves cumulative odds of remission. |
| Choosing therapy vs. medication vs. both | Evidence generally favors combined treatment for moderate-to-severe depression over either alone; mild depression has reasonable evidence for therapy-first approaches. |
Sources
- Cipriani A, et al. Comparative Efficacy and Acceptability of 21 Antidepressant Drugs for Adults With Major Depressive Disorder. The Lancet. 2018.
- Kirsch I, et al. Initial Severity and Antidepressant Benefits. PLOS Medicine. 2008.
- Rush AJ, et al. Acute and Longer-Term Outcomes in Depressed Outpatients (STAR*D). American Journal of Psychiatry. 2006.
- Pigott HE, et al. STAR*D: A Reanalysis. 2023.
- 988 Suicide & Crisis Lifeline
See how 7 providers score
Every provider scored on evidence-based prescribing, rating, insurance support, and prescribing discipline.
View the scores