Skip to content

The antidepressant efficacy debate

TL;DR — The central dispute is not whether antidepressants separate from placebo on average; they do. It is how large, clinically meaningful, generalizable, and unbiased that separation is. Cipriani's 522-trial network meta-analysis found all 21 drugs superior to placebo (OR 1.37–2.13) (Cipriani 2018, PMID 29477251). FDA-data analyses found selective publication inflated apparent effect sizes by 32% and suggested clinically important separation was concentrated at high baseline severity (Turner 2008, PMID 18199864; Kirsch 2008, PMID 18303940). Patient-level analysis also found severity-dependent benefit (Fournier 2010, PMID 20051569). These findings can coexist: a statistically robust average effect may still be modest, heterogeneous, and selectively reported.

Claims that should be separated

Claim Evidence verdict
Antidepressants have zero specific effect Inconsistent with large meta-analytic evidence
Every antidepressant works well for most patients False
Trial averages are clinically modest Generally supported
Benefit may rise with baseline severity Supported in influential analyses; exact form disputed
Publication bias distorted the historical literature Directly demonstrated in FDA comparisons
Placebo response means illness is unreal False; placebo arms include natural history, attention, expectation, and measurement

Network meta-analysis

Cipriani et al. synthesized 522 randomized trials and 116,477 participants. All drugs beat placebo for acute response, with ORs spanning 1.37–2.13; head-to-head differences and acceptability profiles also varied (Cipriani 2018, PMID 29477251). Strengths include scale and simultaneous comparison. Limitations include short duration, average effects, heterogeneous sponsorship, outcome definitions, and trial populations.

Odds ratios exaggerate intuition when event rates are common. Absolute benefit depends on placebo response and baseline risk. A high placebo response can yield a small risk difference despite an apparently notable OR.

Severity analyses

Kirsch et al. analyzed FDA-submitted data and concluded mean drug–placebo differences crossed a conventional clinical-significance threshold mainly in very severe depression, driven largely by lower placebo improvement rather than greater drug improvement (Kirsch 2008, PMID 18303940). Fournier et al. used patient-level data from six trials and similarly found minimal average benefit at mild-to-moderate severity and larger benefit in very severe depression (Fournier 2010, PMID 20051569).

Methodological issue Why it matters
Restricted severity range Trials may exclude mild or extremely severe cases
Regression to mean Baseline extreme scores tend to improve
Scale nonlinearity Same point difference may not mean same clinical change
Individual vs trial-level analysis Ecological relationships can differ from patient-level effects
Missing data Imputation can alter group separation

Publication bias

Turner et al. compared FDA judgments with publications for 74 trials. Of FDA-positive studies, 37/38 were published as positive; among FDA-negative or questionable studies, most were unpublished or published in a way that conveyed a positive outcome. Ninety-four percent of published reports appeared positive versus 51% by FDA assessment; published effect size was 32% larger (Turner 2008, PMID 18199864).

This is direct evidence of reporting distortion, not proof that the true effect is zero. Registration, results reporting, protocol access, and regulatory-data synthesis are the preventive controls.

STAR*D: effectiveness or optimistic arithmetic?

STAR*D enrolled 3,671 people at step 1. Reported QIDS-SR remission fell from 36.8% at step 1 to 13.0% at step 4, with a theoretical cumulative remission of 67% (Rush 2006, PMID 17074942). Attrition increased and relapse was more frequent among those needing more steps.

Pigott et al. reanalyzed patient-level data using prespecified protocol criteria and reported materially lower remission/response estimates, challenging denominator choices and post hoc outcome substitutions (Pigott 2023, PMID 37491091). The dispute concerns which population and rule set answer the practical question; both estimates should be labeled, not blended.

Placebo response and trial design

Placebo-arm change includes spontaneous improvement, regression to mean, clinician contact, expectancy, concomitant care, and measurement artifacts. Rising placebo response reduces power without proving that active treatment has stopped working. Functional unblinding from adverse effects can inflate apparent drug benefit; active placebos are rarely used.

Better design element What it addresses
Public protocol and analysis plan outcome switching
Regulatory-data inclusion publication bias
Patient-level estimands severity and subgroup effects
Functional outcomes scale-only significance
Blinding assessment expectancy and unmasking
Long follow-up durability and harm

Acute response versus durable value

Acute trials do not answer recurrence prevention. Continuing medication after response reduced pooled relapse from 41% to 18% in an older systematic review (Geddes 2003, PMID 12606176), although enriched randomized-withdrawal designs may overestimate general-population benefit. Psychotherapy plus medication can outperform either alone (Cuijpers 2020, PMID 31922679), so the debate should not be framed as drugs versus no care.

Estimands behind the efficacy dispute

The literature often appears contradictory because it asks different questions. Drug–placebo separation in randomized participants, probability of patient-important improvement, effectiveness after attrition, and durability after withdrawal are distinct estimands.

Dispute Evidence for larger benefit Evidence for smaller benefit
Published acute efficacy Network meta-analysis finds all 21 drugs superior to placebo FDA-based analyses recover unpublished negative or equivocal trials
Baseline severity More severe groups can show larger drug–placebo separation The interaction may be driven by declining placebo response and regression artifacts
Individual variability Clinical experience shows striking responder/nonresponder differences Variance meta-analyses question whether drug-specific response variability exceeds random outcome variation
Sequenced effectiveness STAR*D's original cumulative calculation reaches 67% Protocol-faithful denominators and attrition yield much lower estimates
Maintenance Randomized discontinuation favors staying on medication Enrichment and withdrawal confounding constrain causal interpretation

No single effect size resolves clinical significance. Absolute response, remission, functioning, deterioration, discontinuation, and time horizon should be shown together, with confidence intervals and missing-data assumptions.

Additional live-search evidence ledger

The records below were added after full PubMed E-utilities retrieval on 2026-08-30. The ledger states the evidentiary role of each record and preserves the design limitation that should travel with its citation.

  • Oliva V 2024 — Pharmacological treatments for psychotic depression: a systematic review and network meta-analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Oliva V 2024, PMID 38360024)

  • Yang X 2024 — Vortioxetine for depression in adults: A systematic review and dose-response meta-analysis of randomized controlled trials. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Yang X 2024, PMID 38957929)

  • Brown JVE 2021 — Antidepressant treatment for postnatal depression. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Brown JVE 2021, PMID 33580709)

  • Vita G 2023 — Antidepressants for the treatment of depression in people with cancer. Primary or secondary empirical evidence; interpretation should follow its design and comparator rather than the headline alone. (Vita G 2023, PMID 36999619)

  • Seshadri A 2024 — Efficacy of intravenous ketamine and intranasal esketamine with dose escalation for Major depression: A systematic review and meta-analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Seshadri A 2024, PMID 38537759)

  • Kalfas M 2025 — Incidence and Nature of Antidepressant Discontinuation Symptoms: A Systematic Review and Meta-Analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Kalfas M 2025, PMID 40632531)

  • Bahji A 2022 — Efficacy and safety of racemic ketamine and esketamine for depression: a systematic review and meta-analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Bahji A 2022, PMID 35231204)

  • de Vries YA 2019 — Hiding negative trials by pooling them: a secondary analysis of pooled-trials publication bias in FDA-registered antidepressant trials. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (de Vries YA 2019, PMID 30261934)

  • Dean RL 2021 — Ketamine and other glutamate receptor modulators for depression in adults with unipolar major depressive disorder. Systematic review; useful for mapping consistency and gaps, not automatically a pooled causal estimate. (Dean RL 2021, PMID 34510411)

  • Moreno SG 2009 — Novel methods to deal with publication biases: secondary analysis of antidepressant trials in the FDA trial registry database and related journal publications. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Moreno SG 2009, PMID 19666685)

  • Mavridis D 2014 — Exploring and accounting for publication bias in mental health: a brief overview of methods. Primary or secondary empirical evidence; interpretation should follow its design and comparator rather than the headline alone. (Mavridis D 2014, PMID 24477532)

  • Pigott HE 2010 — Efficacy and effectiveness of antidepressants: current status of research. Primary or secondary empirical evidence; interpretation should follow its design and comparator rather than the headline alone. (Pigott HE 2010, PMID 20616621)

  • Turner EH 2022 — Selective publication of antidepressant trials and its influence on apparent efficacy: Updated comparisons and meta-analyses of newer versus older trials. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Turner EH 2022, PMID 35045113)

  • Naudet F 2025 — Efficacy and safety of esketamine for "treatment resistant depression": registered report for a systematic review with an individual patient data meta-analysis of randomized, double-blind, placebo-controlled trials. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Naudet F 2025, PMID 41310599)

  • Pillinger T 2025 — The effects of antidepressants on cardiometabolic and other physiological parameters: a systematic review and network meta-analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Pillinger T 2025, PMID 41135546)

  • Nelson JC 2013 — Moderators of outcome in late-life depression: a patient-level meta-analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Nelson JC 2013, PMID 23598969)

  • Locher C 2015 — Moderation of antidepressant and placebo outcomes by baseline severity in late-life depression: A systematic review and meta-analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Locher C 2015, PMID 25917293)

  • Weitz ES 2015 — Baseline Depression Severity as Moderator of Depression Outcomes Between Cognitive Behavioral Therapy vs Pharmacotherapy: An Individual Patient Data Meta-analysis. Meta-analysis; pooled estimates depend on eligibility, heterogeneity, and reporting bias. (Weitz ES 2015, PMID 26397232)

  • Zimmerman M 2019 — Severity and the Treatment of Depression: A Review of Two Controversies. Review-level synthesis; conclusions inherit limitations of the underlying designs. (Zimmerman M 2019, PMID 30920477)

Open questions

  • What absolute benefit do modern antidepressants provide within prespecified severity strata using patient-level data?
  • How much historical effect inflation remains after including unpublished regulatory trials (Turner 2008, PMID 18199864)?
  • Which STAR*D estimand best answers patient-centered sequential-care questions (Rush 2006, PMID 17074942; Pigott 2023, PMID 37491091)?
  • Can active-placebo or blinding-assessment designs quantify expectancy inflation?

References

  1. Cipriani A, et al. Comparative efficacy and acceptability of 21 antidepressant drugs. Lancet. 2018. PMID 29477251
  2. Turner EH, et al. Selective publication of antidepressant trials and its influence on apparent efficacy. New England Journal of Medicine. 2008. PMID 18199864
  3. Kirsch I, et al. Initial severity and antidepressant benefits. PLoS Medicine. 2008. PMID 18303940
  4. Fournier JC, et al. Antidepressant drug effects and depression severity. JAMA. 2010. PMID 20051569
  5. Rush AJ, et al. Acute and longer-term outcomes in STAR*D. American Journal of Psychiatry. 2006. PMID 17074942
  6. Pigott HE, et al. Protocol-faithful reanalysis of STAR*D. BMJ Open. 2023. PMID 37491091
  7. Geddes JR, et al. Relapse prevention with antidepressant drug treatment. Lancet. 2003. PMID 12606176
  8. Cuijpers P, et al. Psychotherapies, pharmacotherapies and their combination. World Psychiatry. 2020. PMID 31922679
  9. Oliva V, et al. Pharmacological treatments for psychotic depression: a systematic review and network meta-analysis. The lancet. Psychiatry. 2024;11:210-220. PMID 38360024
  10. Yang X, et al. Vortioxetine for depression in adults: A systematic review and dose-response meta-analysis of randomized controlled trials. Psychiatry and clinical neurosciences. 2024;78:536-545. PMID 38957929
  11. Brown JVE, et al. Antidepressant treatment for postnatal depression. The Cochrane database of systematic reviews. 2021;2:CD013560. PMID 33580709
  12. Vita G, et al. Antidepressants for the treatment of depression in people with cancer. The Cochrane database of systematic reviews. 2023;3:CD011006. PMID 36999619
  13. Seshadri A, et al. Efficacy of intravenous ketamine and intranasal esketamine with dose escalation for Major depression: A systematic review and meta-analysis. Journal of affective disorders. 2024;356:379-384. PMID 38537759
  14. Kalfas M, et al. Incidence and Nature of Antidepressant Discontinuation Symptoms: A Systematic Review and Meta-Analysis. JAMA psychiatry. 2025;82:896-904. PMID 40632531
  15. Bahji A, et al. Efficacy and safety of racemic ketamine and esketamine for depression: a systematic review and meta-analysis. Expert opinion on drug safety. 2022;21:853-866. PMID 35231204
  16. de Vries YA, et al. Hiding negative trials by pooling them: a secondary analysis of pooled-trials publication bias in FDA-registered antidepressant trials. Psychological medicine. 2019;49:2020-2026. PMID 30261934
  17. Dean RL, et al. Ketamine and other glutamate receptor modulators for depression in adults with unipolar major depressive disorder. The Cochrane database of systematic reviews. 2021;9:CD011612. PMID 34510411
  18. Moreno SG, et al. Novel methods to deal with publication biases: secondary analysis of antidepressant trials in the FDA trial registry database and related journal publications. BMJ (Clinical research ed.). 2009;339:b2981. PMID 19666685
  19. Mavridis D, et al. Exploring and accounting for publication bias in mental health: a brief overview of methods. Evidence-based mental health. 2014;17:11-5. PMID 24477532
  20. Pigott HE, et al. Efficacy and effectiveness of antidepressants: current status of research. Psychotherapy and psychosomatics. 2010;79:267-79. PMID 20616621
  21. Turner EH, et al. Selective publication of antidepressant trials and its influence on apparent efficacy: Updated comparisons and meta-analyses of newer versus older trials. PLoS medicine. 2022;19:e1003886. PMID 35045113
  22. Naudet F, et al. Efficacy and safety of esketamine for "treatment resistant depression": registered report for a systematic review with an individual patient data meta-analysis of randomized, double-blind, placebo-controlled trials. BMC medicine. 2025;23:677. PMID 41310599
  23. Pillinger T, et al. The effects of antidepressants on cardiometabolic and other physiological parameters: a systematic review and network meta-analysis. Lancet (London, England). 2025;406:2063-2077. PMID 41135546
  24. Nelson JC, et al. Moderators of outcome in late-life depression: a patient-level meta-analysis. The American journal of psychiatry. 2013;170:651-9. PMID 23598969
  25. Locher C, et al. Moderation of antidepressant and placebo outcomes by baseline severity in late-life depression: A systematic review and meta-analysis. Journal of affective disorders. 2015;181:50-60. PMID 25917293
  26. Weitz ES, et al. Baseline Depression Severity as Moderator of Depression Outcomes Between Cognitive Behavioral Therapy vs Pharmacotherapy: An Individual Patient Data Meta-analysis. JAMA psychiatry. 2015;72:1102-9. PMID 26397232
  27. Zimmerman M. Severity and the Treatment of Depression: A Review of Two Controversies. The Journal of nervous and mental disease. 2019;207:219-223. PMID 30920477