MIAMI, FL – MARCH 23: A bottle of anti-depressant pills named Paxil are shown March 23, 2004 photographed in Miami, Florida. The Food and Drug Administration asked makers of popular anti-depressants to add or strengthen suicide-related warnings on their labels as well as the possibility of worsening depression especially at the beginning of treatment or when the doses are increased or decreased.
Illustration by Joe Raedle/Getty Images
The largest study ever conducted on the long-term effectiveness of antidepressants has found that three common medications — escitalopram (Lexapro), duloxetine (Cymbalta), and agomelatine (Valdoxan) — are significantly more likely to keep patients well over six months or more than fluoxetine (Prozac), the drug that has remained America’s most familiar antidepressant for decades.
The findings, published November 10 in Lancet Psychiatry, come from a network meta-analysis of 115 randomized controlled trials encompassing more than 30,000 participants — the largest such analysis ever focused specifically on antidepressant maintenance treatment rather than the short-term, eight-week efficacy that most prior drug comparisons have measured.
For the roughly 21 million Americans diagnosed with major depressive disorder (MDD) each year, and for the many more who continue on antidepressants for months or years after their acute episode resolves, the question the study asks is unusually practical: does it matter which antidepressant you stay on long-term, not just which one lifts your mood in the first few weeks?
The answer, according to the data, is yes — and by a margin that clinicians may find difficult to ignore.
How the Study Measured Long-Term Success
Traditional antidepressant trials measure response after six to eight weeks. The new analysis took a different approach, defining success as sustained response — continuing to meet the threshold for treatment response at six months or longer, with all-cause discontinuation (stopping the drug for any reason, including side effects) as a secondary measure.
The research team used a statistical technique called network meta-analysis, which allows researchers to compare treatments against one another even when no head-to-head trial between them exists, by threading together a web of direct and indirect comparisons simultaneously. The primary analysis drew on 35 head-to-head continuation trials — studies that enrolled patients who had already responded in the acute phase and tracked whether they maintained that response over time.
Fluoxetine (Prozac) served as the reference comparator, chosen for its long history and widespread use. Against that baseline, the results were clear:
Escitalopram (Lexapro): odds ratio 1.62 (95% confidence interval 1.21–2.18) — patients on escitalopram had 62% higher odds of maintaining response than those on fluoxetineAgomelatine (Valdoxan): odds ratio 1.57 (95% CI 1.08–2.26) — 57% higher oddsDuloxetine (Cymbalta): odds ratio 1.40 (95% CI 1.06–1.85) — 40% higher odds
An odds ratio of 1.62 means this: for every 100 fluoxetine patients who maintained their response at six months, an estimated 162 escitalopram patients would have done the same under otherwise comparable conditions. All three confidence intervals exclude 1.0, meaning the results are statistically significant and unlikely to be explained by chance.
Escitalopram and duloxetine also showed lower rates of all-cause discontinuation than fluoxetine — an important secondary finding, since stopping an antidepressant is itself a clinical event associated with relapse risk.
At the other end of the spectrum, two antidepressants showed lower odds of sustained response than fluoxetine: bupropion (Wellbutrin) and reboxetine — a finding that may prompt clinicians currently prescribing bupropion for maintenance to revisit that choice.
What Makes Agomelatine Different — and Why US Patients Cannot Access It
The second-ranked drug in the analysis, agomelatine, operates through a mechanism entirely unlike the SSRIs and SNRIs that dominate antidepressant prescribing. Rather than blocking serotonin reuptake, agomelatine acts as an agonist at melatonin receptors (MT1 and MT2) while simultaneously blocking serotonin 5-HT2C receptors. The dual action is thought to resynchronize disrupted circadian rhythms — the internal biological clocks that govern sleep, mood, and cortisol release — while also modulating serotonergic signaling.
There is a critical practical limitation for American readers: agomelatine is not approved by the US Food and Drug Administration and is not available in the United States. It has been marketed in the European Union under the brand name Valdoxan since 2009, but the FDA never approved it. The drug’s novel mechanism, combined with a hepatotoxicity risk that required mandatory liver monitoring protocols in Europe, contributed to a regulatory pathway that has not resulted in US approval.
The liver risk is not trivial. Regulatory bodies including the UK’s Medicines and Healthcare products Regulatory Agency (MHRA) and the European Medicines Agency (EMA) have documented that agomelatine causes clinically significant elevations in liver enzymes (more than three times the upper limit of normal) in roughly 1 to 2.5% of patients, a rate that is dose-dependent — higher at the 50mg dose than at 25mg. As of 2012, six cases of hepatic failure worldwide — some fatal, some requiring liver transplantation — had been linked to agomelatine use, most involving patients with pre-existing liver risk factors such as obesity, alcohol use, or diabetes. European prescribers are required to perform liver function tests before starting agomelatine, then at weeks 3, 6, 12, and 24, and again when the dose is increased. The drug is contraindicated in anyone with existing hepatic impairment.
For patients outside the US: agomelatine’s ranking in this study is genuine and notable, but that ranking must be weighed alongside this monitoring burden. For US patients and clinicians: the first-ranked drug, escitalopram, is both FDA-approved and widely available.
A Major Limitation: More Than Half the Evidence Is Flagged for Bias
The study’s authors are forthright about the quality of their evidence base, and their disclosure warrants careful attention before drawing clinical conclusions.
Of the 35 head-to-head continuation trials that formed the primary analysis, 19 — more than half — were assessed as having high risk of bias. A substantial portion of this bias stems from what researchers call enrichment design: continuation trials commonly enroll only patients who already responded to the drug being tested in the acute phase, then randomize them to continue the same drug or switch to a comparator. This creates an inherent advantage for the drug being continued, because the study population is pre-selected for responding to it.
External experts have flagged this limitation prominently. Dr. Sameer Jauhar, a psychiatrist at Imperial College London who was not involved in the study, noted that the high proportion of biased trials means the findings should be interpreted with caution. While the direction of the results — that escitalopram and duloxetine outperform fluoxetine over the long term — is consistent with prior real-world cohort data, the magnitude of the differences may be overstated by enrichment.
This does not render the study uninformative. A network meta-analysis of this scale, covering 115 trials and three decades of research, represents the most comprehensive synthesis of long-term antidepressant evidence to date. But it does mean that the odds ratios above should be understood as estimates with meaningful uncertainty bands, not precise clinical prescriptions.
Why This Fills a Gap That the Famous 2018 Study Left Open
The 2026 findings extend the work of a landmark 2018 paper by Cipriani and colleagues, also published in The Lancet, which compared 21 antidepressants across 522 randomized trials involving more than 116,000 participants. That study established the most comprehensive efficacy hierarchy for antidepressants then available — but it focused almost entirely on acute treatment: response within eight weeks of starting medication.
The 2018 analysis found escitalopram and sertraline to be the most efficacious and best-tolerated antidepressants for the acute phase. The 2026 study asks a different question: does that ranking hold when treatment extends beyond the acute phase into the months that most patients with recurrent MDD actually spend on these medications?
The answer, based on the new data, is that escitalopram’s superiority holds — but that the acute rankings for duloxetine and agomelatine appear stronger in the maintenance setting than in acute comparisons, while some drugs that perform adequately at eight weeks (including bupropion) may underperform at six months.
What Current Guidelines Already Say
For clinicians wondering whether the new study changes their practice, it is worth noting that the top-ranked drug, escitalopram, is already a preferred first-line antidepressant in major treatment guidelines. The UK’s NICE guideline NG222 — reviewed as recently as January 2026 — lists escitalopram and sertraline as default first-line options. The Canadian CANMAT guidelines, the RANZCP guidelines, and the American College of Physicians guidance converge on SSRIs, with escitalopram 10mg and sertraline 50mg identified as standard starting points.
The new study provides what those guidelines largely lacked: specific evidence that escitalopram’s advantages over fluoxetine persist and may be larger in the maintenance setting than in the acute phase. For clinicians who have defaulted to fluoxetine for maintenance — perhaps from inertia or historical familiarity — the study offers a research-grounded reason to reconsider.
Duloxetine (Cymbalta), the SNRI that placed third, may be particularly relevant for patients with comorbid pain conditions, where its dual serotonin-norepinephrine mechanism offers documented advantages beyond mood.
How Network Meta-Analysis Works — and Why It Matters Here
A standard meta-analysis can only pool trials that compare the same two treatments directly. Network meta-analysis extends that approach by simultaneously incorporating all available comparisons — direct and indirect — across a web of studies. If Drug A has been compared with Drug B, and Drug B has been compared with Drug C, network meta-analysis can estimate how A compares with C, even if no trial has ever put them head-to-head. This allows researchers to build a full ranking across 21 or more antidepressants from a dataset where most drugs have been compared against only a handful of others.
For long-term antidepressant comparisons specifically, this methodology is particularly valuable because head-to-head maintenance trials are rare and expensive — running a randomized trial for 12 months or longer in a population already stabilized on medication is logistically demanding. The network approach allows existing data to answer comparative questions that individual trials were not designed to address.
The methodology’s main vulnerability is what statisticians call the transitivity assumption: for indirect comparisons to be valid, the patient populations and trial designs across the network must be sufficiently similar that comparisons can be legitimately combined. When they are not — when trial designs differ in systematic ways that correlate with treatment assignment — indirect comparisons can be misleading. The enrichment design problem described above is one specific form of this vulnerability, and it is why the 19/35 high-bias figure is not merely a boilerplate limitation caveat but a substantive caveat about the reliability of the specific estimates.
Key Takeaways for Patients and Prescribers
If you are in the US and currently on fluoxetine (Prozac) for long-term depression treatment: The new study provides the strongest evidence to date that escitalopram (Lexapro) may sustain remission more effectively over six months or longer. This is a conversation worth having with your prescriber — not a reason to stop fluoxetine without medical guidance.
If you are on bupropion (Wellbutrin) for maintenance: The study found bupropion underperformed fluoxetine for sustained long-term response — a finding that, combined with prior real-world cohort data showing similar results, is worth discussing with your clinician.
If you are outside the US and your prescriber suggests agomelatine: The study’s data support its efficacy for long-term treatment. Be aware that it requires regular liver function monitoring, particularly if you have risk factors including obesity, diabetes, non-alcoholic fatty liver disease, or significant alcohol use. It is contraindicated in hepatic impairment.
What the study cannot tell you: Which drug is right for you individually. Network meta-analysis produces population-level estimates; individual patients differ in side effect profiles, comorbidities, prior treatment history, and response. The odds ratios in this study are averages across diverse clinical populations, not predictions for any specific patient.
Study Details
Publication: Comparative efficacy and acceptability of antidepressants for the long-term treatment of adults with major depressive disorder: a systematic review and network meta-analysis. Lancet Psychiatry. Published November 10, 2026. DOI: 10.1016/S2215-0366(26)00326-3
Research team: An international team of researchers including collaborators from Radboud University Medical Center (Netherlands) and partner institutions; see the full study in Lancet Psychiatry.
Scale: 115 randomized controlled trials; 30,000+ participants; 21 antidepressants; 35 head-to-head continuation trials in primary analysis; 19/35 rated high risk of bias.
Primary outcome: Sustained response at 6 months or longer (all-cause discontinuation as secondary outcome).
This article is for informational purposes. Never start, stop, or change antidepressant medications without guidance from a licensed healthcare provider. Stopping antidepressants abruptly can cause discontinuation syndrome and relapse risk.