Reviews
Comparative Long-Term Efficacy and Acceptability of Antidepressants in Major Depressive Disorder
van Haaren et alThe lancet. Psychiatry

A systematic review and network meta-analysis evaluated the comparative long-term (>12 weeks) efficacy and acceptability of 21 antidepressants in adult unipolar major depressive disorder. Researchers searched multiple databases, including MEDLINE, Embase, Cochrane Register, PsycINFO, and regulatory agency registries from inception to March 27, 2026. The review encompassed 115 double-blind randomised controlled trials comprising 30,428 participants (mean age 45.3 years [SD 8.5]; 20,066 [66%] female). Eligible study designs included enrichment, non-enrichment, continuation, and continuation-extension trials. Primary endpoints were efficacy (sustained response rate) and acceptability (all-cause discontinuation), synthesized via random-effects pairwise and network meta-analyses using fluoxetine as the reference comparator.
In the primary analysis of head-to-head continuation and continuation-extension trials (35 studies; n=9,888), sustained response differed across agents. Escitalopram (OR 1.62, 95% CI 1.21–2.18), duloxetine (OR 1.57, 95% CI 1.08–2.26), and agomelatine (OR 1.40, 95% CI 1.06–1.85) showed higher sustained response rates than fluoxetine. In league table pairwise comparisons, bupropion and reboxetine showed lower sustained response rates against four other antidepressants. Regarding acceptability, escitalopram (OR 0.60, 95% CI 0.43–0.83) and duloxetine (OR 0.64, 95% CI 0.44–0.95) demonstrated lower all-cause discontinuation relative to fluoxetine, whereas reboxetine and milnacipran showed higher discontinuation rates compared with six and four other antidepressants, respectively.
Sensitivity analyses incorporating enrichment designs corroborated these outcomes. However, the authors noted key limitations: the certainty of evidence and high risk of bias in the included trials—with 19 (54%) of 35 primary-analysis studies assessed as having high risk of bias using the Cochrane Risk of Bias 2 tool—might limit the reliability of the findings. In addition, sex or gender distribution other than female could not be derived, race and ethnicity data were not extracted, and no individuals with lived experience were involved in study design, conduct, or interpretation. The authors concluded that while antidepressants can differ in sustained response efficacy and acceptability to inform shared clinical decision-making, additional head-to-head comparative studies are needed to determine long-term effects.
Key Takeaways
- In long-term treatment (>12 weeks), escitalopram (OR 1.62, 95% CI 1.21–2.18), duloxetine (OR 1.57, 95% CI 1.08–2.26), and agomelatine (OR 1.40, 95% CI 1.06–1.85) showed higher sustained response rates than fluoxetine.
- Escitalopram (OR 0.60, 95% CI 0.43–0.83) and duloxetine (OR 0.64, 95% CI 0.44–0.95) demonstrated lower all-cause discontinuation compared with fluoxetine.
- Among the 35 head-to-head continuation and continuation-extension trials analyzed, 19 (54%) had a high risk of bias, which might limit the reliability of the findings.
Read the original sourceThe lancet. Psychiatry
