Comparative efficacy and acceptability of antidepressants in the long-term treatment of major depressive disorder: a systematic review and network meta-analysis.
van Haaren Paul C F PCF, van Gaalen Jakob M JM, Ostinelli Edoardo G EG, Shinohara Kiyomi K et al.
Major depressive disorder carries high risks of relapse and chronicity, necessitating treatments that achieve and maintain response (sustained response). This study aimed to compare antidepressants on their ability to achieve sustained response, and on acceptability, by synthesising data from randomised controlled trials. For this systematic review and network meta-analysis, we searched the Cochrane Register, AMED, CINAHL, Embase, LILACS, MEDLINE, MEDLINE In-Progress, PsycINFO, the UK National Research Register, and regulatory agency websites from inception to March 27, 2026. We included double-blind randomised controlled trials that compared any of 21 antidepressants head-to-head or with placebo for long-term treatment (>12 weeks) of adult unipolar major depressive disorder. We included enrichment, non-enrichment, continuation, and continuation-extension designs, and extracted study-level data. We could not derive the sex or gender distribution of participants other than female and did not extract data on race or ethnicity. Primary outcomes were efficacy (sustained response rate) and acceptability (all-cause discontinuation). We synthesised data in random-effects pairwise and network meta-analyses to estimate pooled odds ratios (ORs), using fluoxetine as a reference in the ranked forest plots. The primary analysis included head-to-head continuation and continuation-extension design studies. Enrichment design studies were incorporated in the analysis using a chain-of-evidence approach. Risk-of-bias assessment was done using the Cochrane Risk of Bias 2 tool. No people with lived experience were involved in the design, conduct, or interpretation of this study. From 54 231 records, 115 studies (30 428 participants; mean age 45·3 years [SD 8·5], 20 066 [66%] female) were included. In head-to-head continuation and continuation-extension design studies (35 studies; n=9888), sustained response differed across antidepressants. Escitalopram (OR 1·62, 95% CI 1·21-2·18), duloxetine (OR 1·57, 1·08-2·26), and agomelatine (OR 1·40, 1·06-1·85) showed higher sustained response rates than fluoxetine. Bupropion and reboxetine showed lower sustained response rates against four other antidepressants in the league table of pairwise comparisons. Escitalopram (OR 0·60, 0·43-0·83) and duloxetine (OR 0·64, 0·44-0·95) demonstrated lower all-cause discontinuation compared with fluoxetine, while reboxetine and milnacipran showed higher all-cause discontinuation against six and four other antidepressants, respectively, in the league table of pairwise comparisons. Sensitivity analyses incorporating enrichment-design studies corroborated these results. However, the certainty of evidence of the network meta-analysis results and the risk of bias in the included studies (19 [54%] of 35 had high risk of bias) might limit the reliability of the current findings. Antidepressants can differ in their efficacy for achieving sustained response and in their acceptability. These findings can inform the shared decision making for patients, prescribers, and carers in real-world practice. However, more head-to-head comparative studies are needed to provide reliable information about long-term effects of all licensed antidepressants. None.