Agomelatine: Independent Evidence on Depression, Sleep, Liver Risk & Withdrawal

Table of contents

• Evidence summary
• What agomelatine is
• How it works
• What it is prescribed for
• What works and what does not
• Benefits by claim
• Risks and all side effects
• All interactions
• Who should avoid agomelatine
• Dosage and how to take it
• Follow the money: who makes it and who funded the evidence
• Related research
• Frequently asked questions
• Sources and funding notes

Evidence summary

Independent evidence and credibility scorecard

Independence tier: 1 = no money from anyone selling the product; 4 = seller-funded. Credibility: A (highest) to D.

What agomelatine is

Agomelatine is an antidepressant with an unusual mechanism. Its ATC code is N06AX22, which places it among the "other antidepressants" (EU SmPC). It is chemically related to melatonin, but it is not a melatonin supplement and it is not a sleeping pill. It was developed by Les Laboratoires Servier of Suresnes, France, which holds the EU marketing authorisation under the brand name Valdoxan. The same product was also licensed as Thymanax (EU SmPC, MHRA 2012).

Its path to market was not smooth. Servier first applied to the EMA in March 2005. The EMA's scientific committee (CHMP) raised "major objections regarding the doubtful clinical relevance of the effect size" and gave a negative opinion in July 2006. The European Commission formally refused authorisation in January 2007. Servier submitted a completely new application in September 2007, adding a new relapse-prevention study. The European Commission authorised Valdoxan across the EU on 19 February 2009 (EMA assessment report, EMA EPAR).

UK status: prescription-only. Valdoxan 25 mg has a Great Britain licence held by Les Laboratoires Servier, marketed by Servier Laboratories Limited; the UK SmPC was revised on 15 April 2026 (UK SmPC). A generic version is listed from Zentiva (Zentiva SmPC), and another generic (Neuraxpharm) is listed as discontinued. The NHS website has no patient page for agomelatine, and NICE has no positive recommendation for it (see below).

US and Canada: not approved. Novartis had licensed US rights but "discontinued the development of agomelatine for the US market in 2011" (Guaiana et al., Cochrane 2013). The Canadian CANMAT guideline notes that agomelatine is "not available in Canada or the USA" (CANMAT 2023). This is why it does not appear in US references such as the AGS Beers Criteria (AGS Beers 2023).

How it works

According to its product information, agomelatine has two actions. It is a "melatonergic agonist (MT1 and MT2 receptors) and 5-HT2C antagonist". Binding studies show it "has no effect on monoamine uptake and no affinity for α, β adrenergic, histaminergic, cholinergic, dopaminergic and benzodiazepine receptors". Blocking 5-HT2C receptors increases noradrenaline and dopamine release specifically in the frontal cortex, and it "has no influence on the extracellular levels of serotonin" (EU SmPC). This explains two features of its side-effect profile. Because it does not boost serotonin like SSRIs, it is less linked to sexual side effects. Because it has no anticholinergic or antihistamine action, it lacks the dry mouth, constipation and heavy sedation of older tricyclics.

The melatonin-receptor action is meant to shift the body clock. In humans it "induces a phase advance of sleep, body temperature decline and melatonin onset" (EU SmPC). This is why it is taken at bedtime. How much of its antidepressant effect comes from each action is not established. For how melatonin itself works as a supplement, see our melatonin review.

How the body handles it: agomelatine is well absorbed, but less than 5% reaches the bloodstream because the liver breaks most of it down on first pass. This is done mainly by the enzyme CYP1A2 (about 90%), with CYP2C9 and CYP2C19 contributing the rest. Its half-life is only 1 to 2 hours. Exposure varies a great deal between people. It is higher in women and people taking oral contraceptives, and lower in smokers. In people with mild or moderate cirrhosis, exposure was 70 and 140 times higher. In people aged 75 and over, average exposure (AUC) was about 4 times higher and peak levels about 13 times higher than in people under 75 (EU SmPC). These numbers explain most of its interaction and liver warnings.

What it is prescribed for

Off-label means the prescriber takes responsibility for using the medicine outside its licence. It does not automatically mean the use is unsafe, but it does mean the regulator has not judged the evidence for that use.

What works and what does not

Benefits by claim

Depression: a real but small effect, and a publication-bias problem

The regulator's own files are the clearest place to start. The EMA assessment report describes the five main short-term placebo-controlled studies (1,893 adults). Two studies without an active comparator showed agomelatine beating placebo. Three studies that also included fluoxetine or paroxetine found no difference between agomelatine and placebo, but in two of them the comparator drug also failed. The EMA's pooled estimate across the six pivotal short-term studies was a difference of 1.5 points on the Hamilton scale (95% CI 0.80 to 2.22). The assessors wrote that agomelatine 25 mg "is probably less efficacious than other antidepressants" (EMA assessment report, EMA summary). The current product information puts it this way: "significant efficacy of agomelatine 25–50 mg was demonstrated in 6 out of the ten short-term double-blind placebo-controlled trials" (EU SmPC).

Two independent teams then collected published and unpublished trials. Koesters and colleagues (Ulm, Verona and Oxford) included 10 acute and 3 relapse-prevention trials, seven of which were unpublished. Agomelatine beat placebo by 1.51 HRSD points (99% CI −2.29 to −0.73). They found that "None of the negative trials were published" and concluded that "a clinically important difference between agomelatine and placebo… is unlikely" (Koesters et al., BJPsych 2013). Taylor and colleagues (King's College London) found 20 trials with 7,460 participants. The effect size was 0.24 (95% CI 0.12 to 0.35), and the relative risk of response was 1.25 (1.11 to 1.4). They also found that "published studies were more likely than unpublished studies to have results that suggested advantages for agomelatine". Their overall conclusion was more positive: agomelatine is "an effective antidepressant with similar efficacy to standard antidepressants" (Taylor et al., BMJ 2014). Taylor's lead author declared fees and grants from Servier, and the Koesters team declared none in the abstract. Yet both teams report essentially the same numbers. They differ mainly in whether they call the effect clinically meaningful.

Honest reading: a gap of about 1.5 Hamilton points is similar in size to the average drug–placebo gap for antidepressants in general, which an analysis of 232 trials held by the US FDA put at 1.75 points (95% CI 1.63 to 1.86) (Stone et al., BMJ 2022). It is small at the level of the average patient. Agomelatine is not a clear outlier in either direction. The more specific concern is transparency: the drug looked better in the journals than in the full dataset. For the wider debate on how much antidepressants help, see our stress, anxiety and depression guide.

Against other antidepressants: about the same

The Cochrane review compared agomelatine with paroxetine, fluoxetine, sertraline, escitalopram and venlafaxine in 13 trials (4,495 participants). Response was no different from SSRIs (RR 1.01, 95% CI 0.95 to 1.08) or venlafaxine (RR 1.06, 0.98 to 1.16). Remission was also no different (RR 0.83, 0.68 to 1.01 vs SSRIs). The reviewers noted that "almost all of the studies were sponsored by the pharmaceutical company that manufactures agomelatine (Servier)" and that "attempts to contact the pharmaceutical company Servier for additional information on all unpublished studies were unsuccessful". Their conclusion was that "no firm conclusions can be drawn" (Guaiana et al., Cochrane 2013). The Cipriani network meta-analysis is broader, with 522 trials and 21 drugs. It placed agomelatine among the drugs that were more effective in head-to-head trials (ORs 1.19–1.96) and among the more tolerable ones (ORs 0.43–0.77). The certainty of agomelatine comparisons was "often rated as moderate" (Cipriani et al., Lancet 2018). The EU product information cites company trials showing superiority over sertraline in 2 trials and non-inferiority to SSRIs and SNRIs in 4 (EU SmPC). Given the funding pattern and the Cochrane findings, we treat claims of superiority as unproven.

Tolerability: the strongest part of the case

This is where agomelatine stands out in independent analyses. In Cipriani 2018, only agomelatine (OR 0.84, 95% CrI 0.72–0.97) and fluoxetine (0.88) had fewer all-cause dropouts than placebo. For dropouts caused by side effects, every drug except agomelatine had significantly higher rates than placebo; agomelatine's OR was 1.21 (95% CrI 0.94–1.56) (Cipriani 2018). Cochrane found fewer dropouts than on venlafaxine (RR 0.40, 95% CI 0.24 to 0.67) and less dizziness (RR 0.19). Tolerability was similar to SSRIs (RR 0.95) (Cochrane 2013). CANMAT lists agomelatine among antidepressants with lower rates of sexual side effects. It rates its discontinuation-symptom risk as "low or minimal" and notes that agomelatine and mianserin are the only antidepressants not associated with low sodium (hyponatraemia) (CANMAT 2023).

Relapse prevention

The EMA describes two relapse-prevention studies in 706 people whose depression had already responded to agomelatine. In the first, agomelatine was no better than placebo over 26 weeks. In the second, symptoms returned in 21% on agomelatine (34 of 165) against 41% on placebo (72 of 174) (EMA). The current product information reports relapse rates of 22% vs 47% in its cited relapse trial (EU SmPC). Pooling three relapse trials, the Koesters review found no significant effect (RR 0.78, 99% CI 0.41–1.48) (Koesters 2013). The evidence that it keeps depression away long-term is therefore weaker than for many older antidepressants.

Sleep

Company studies reported in the product information found that agomelatine 25 mg increased slow-wave (deep) sleep without changing REM sleep. They also found that "from the first week of treatment, onset of sleep and the quality of sleep were significantly improved" (EU SmPC). The EMA assessors described its advantage over venlafaxine on one sleep scale as "modest" (5 mm on a 100 mm scale) (EMA assessment report). A 2026 independent meta-analysis of 30 studies compared agomelatine, mirtazapine and trazodone. All three improved how people rated their own sleep, but "agomelatine may lack a definitive effect on improving objective sleep parameters" (Zhang et al., J Affect Disord 2026). In one Cochrane comparison, daytime sleepiness was more common than with sertraline (RR 4.65, one study) (Cochrane 2013). Agomelatine is not a treatment for insomnia on its own. Our sleep supplements review covers melatonin and other options.

Anxiety and other uses

A pooled analysis of three placebo-controlled trials in generalised anxiety disorder (669 patients) found a 6.30-point advantage on the HAM-A anxiety scale at 12 weeks. Response was 67.1% vs 32.5% (Stein et al., Adv Ther 2021). A 2025 network meta-analysis concluded that agomelatine "can be considered as a drug of choice" for anxiety (Hood et al., 2025). Both papers have co-authors employed by Servier, and the 2021 analysis lists Servier as a funder. We found no independent confirmation, and the EU and UK licences do not cover anxiety. For seasonal affective disorder, a Cochrane review found one trial with an "indeterminate result" (RR 0.83, 95% CI 0.51 to 1.34; very low certainty) (Nussbaumer-Streit et al., Cochrane 2019).

Risks and all side effects

According to the product information, most side effects "were usually mild or moderate and occurred within the first two weeks of treatment". The most common were headache, nausea and dizziness (EU SmPC). Agomelatine has no US label, so it has no FDA boxed warning. The EU product information carries the class warning on suicidal thoughts in people under 25.

The liver: what the regulators found

The product information reports "cases of liver injury, including hepatic failure (few cases were exceptionally reported with fatal outcome or liver transplantation in patients with hepatic risk factors)". It also reports liver enzymes above 10 times normal, hepatitis and jaundice. Most cases "occurred during the first months of treatment", and enzyme levels usually returned to normal after stopping (EU SmPC). In 2012 the UK regulator reported that raised transaminases were dose-dependent: 2.5% on 50 mg vs 1.4% on 25 mg. It put the rate of liver failure at "less than 1 in every 1000 patients treated" (MHRA 2012). In 2014 it added that "a recent European review revealed poor clinical compliance with recommended liver function monitoring", and it introduced a patient booklet (MHRA 2014). An academic systematic review from Rostock found liver-injury rates "as high as 4.6% for agomelatine compared to 2.1% for placebo" and lower rates for the comparators: escitalopram 1.4%, paroxetine 0.6%, fluoxetine 0.4% and sertraline 0% (Freiesleben & Furczyk, 2015). CANMAT groups agomelatine with bupropion, duloxetine and nefazodone as having "a higher risk of adverse liver effects" (CANMAT 2023).

Stopping: discontinuation symptoms

This is one of the few antidepressants whose licence says "no dosage tapering is needed on treatment discontinuation". A company trial using a standard discontinuation checklist found that agomelatine "did not induce discontinuation syndrome after abrupt treatment cessation" (EU SmPC). CANMAT independently places it in the "low or minimal" risk group (CANMAT 2023). There is one important exception. When people switch to agomelatine from an SSRI or SNRI, the old drug can still cause withdrawal. In a switching trial, 56.1% to 79.8% of patients had at least one discontinuation symptom a week after stopping their SSRI or SNRI; the figure was lowest with a two-week taper and highest with abrupt switching (EU SmPC). Stopping any antidepressant should still be planned with a prescriber, because the depression can return.

Overdose

Experience is limited, but the product information records "one person having ingested 2,450 mg agomelatine, recovered spontaneously without cardiovascular and biological abnormalities". It found no effect on heart ion channels (hERG) in safety studies (EU SmPC). This contrasts with tricyclic antidepressants, which NICE describes as "dangerous in overdose" (NICE NG222). Any suspected overdose still needs urgent medical assessment.

All interactions

Almost all of agomelatine's interactions come from one enzyme, CYP1A2. Drugs that block it can raise agomelatine levels many times over, and drugs or habits that speed it up (smoking, rifampicin) can lower them. Agomelatine itself "will not modify exposure to medicinal products metabolised by CYP 450" (EU SmPC).

Interaction source: EU SmPC sections 4.3–4.5 and 5.2. Herbal products such as St John's wort have not been formally tested with agomelatine, so tell the prescriber and pharmacist about all supplements.

Who should avoid agomelatine

Must not take it (contraindications): people with a known allergy to agomelatine or its ingredients; people with liver impairment such as cirrhosis or active liver disease; people whose transaminases exceed 3 times the upper limit of normal; and anyone taking potent CYP1A2 inhibitors such as fluvoxamine or ciprofloxacin (EU SmPC).

Needs careful benefit–risk review: people with liver-injury risk factors, meaning obesity, overweight or fatty liver disease, diabetes, alcohol use disorder or heavy drinking, and those taking other medicines that can harm the liver. Caution is also advised with a history of bipolar disorder, mania or hypomania, and with moderate or severe kidney impairment, because clinical data are limited. The tablets contain lactose (EU SmPC).

Older adults: no effect has been documented in people aged 75 or over, so "agomelatine should not be used by patients in this age group". It should also not be used for depression in older people with dementia (EU SmPC, MHRA 2014). Between 65 and 74, one placebo-controlled trial showed benefit, and tolerability was similar to that in younger adults. The US Beers Criteria do not cover agomelatine because it is not sold in the US (AGS Beers 2023).

Children and teenagers: not recommended under 18. In pooled data, adolescents had more serious adverse events than adults (10.4% vs 3.5%), more liver events (6.3% vs 1.7%) and more suicidal events (3.1% vs 1.2%) (EU SmPC).

Pregnancy and breastfeeding: human data are limited (fewer than 300 pregnancy outcomes). Animal studies show no harm, but "as a precautionary measure, it is preferable to avoid the use of Valdoxan during pregnancy". It is not known whether agomelatine passes into human milk; it does in animals. The decision whether to stop breastfeeding or stop the medicine must weigh both (EU SmPC). Anyone pregnant or planning pregnancy should discuss options with their prescriber and not stop treatment abruptly without advice.

Dosage and how to take it

How to take it: swallow the tablet at bedtime, with or without food (EU SmPC). Keep every blood-test appointment. The MHRA has warned that monitoring is often missed in practice (MHRA 2014).

How long before it works: the licence allows a dose increase if there is no improvement after two weeks, and the trials measured their main results at 6 to 8 weeks (EU SmPC). As with other antidepressants, the NHS advises that it can take weeks for mood to improve, and some people feel worse at first (NHS).

How to stop: the licence says no taper is needed. Still, decide the timing with your prescriber. NICE advises monitoring for both withdrawal symptoms and the depression returning when any antidepressant is stopped (NICE NG222). After stopping because of raised liver enzymes, blood tests continue until the levels return to normal (EU SmPC).

Follow the money: who makes it and who funded the evidence

Who makes it and who sells it

• Originator and licence holder: Les Laboratoires Servier, 50 rue Carnot, Suresnes, France, which holds the EU marketing authorisation (EU SmPC). In the UK the licence is also held by Les Laboratoires Servier, with Servier Laboratories Limited as the local company (UK SmPC).
• Ownership: Servier says it is "governed by a non-profit foundation, the Fondation Internationale de Recherche Servier (FIRS)", a "non-profit governance foundation without capital". It adds that "no persons own any stake in the Group's capital" (Servier governance page). There are no outside shareholders to pay, but the company still depends on product sales.
• Revenue: Servier reported group revenue of €6.9 billion for 2024/25, up 16.2%, driven mainly by oncology and cardiometabolic products. The results release does not mention Valdoxan or agomelatine (Servier, 27 January 2026). We found no published sales figure for agomelatine.
• US partner: Novartis held US development rights and stopped in 2011 (Cochrane 2013).
• Generics: in the UK a Zentiva generic is listed (Zentiva SmPC), and a Neuraxpharm generic is listed as discontinued.

Who funded the evidence

• The efficacy trials were almost all Servier's. The Cochrane reviewers found that "almost all of the studies were sponsored by… Servier", and that the company did not respond to requests for unpublished data (Guaiana et al., Cochrane 2013). The later independent review found that "none of the negative trials were published" (Koesters 2013). The unpublished trials were available only because regulators had them. Four trials in the Taylor review came from the EMA file (Taylor 2014).
• The independent syntheses were publicly funded or unfunded. Cipriani 2018 was funded by UK NIHR and the Japan Society for the Promotion of Science. One co-author (Leucht) declared lecture fees from companies including Servier (Cipriani 2018). The Cochrane group's largest funder is NIHR, and the review authors declared no conflicts (Cochrane 2013). Taylor 2014 had "no specific grant", but its lead author declared Servier fees and grants (Taylor 2014).
• Evidence for newer uses is still company-led. The anxiety analyses list Servier funding or Servier-employed authors (Stein 2021, Hood 2025). The adolescent trial was company-run and did not convince the regulator (EMA 2024).
• The regulator's funding. The EMA is funded mostly by fees from industry. That did not stop it refusing agomelatine in 2006–2007 (EMA assessment report).
• NICE appraisal. NICE was "unable to recommend" agomelatine because Servier "did not provide an evidence submission" (NICE TA231, 2011). This means no cost-effectiveness judgement was ever made, positive or negative.
• A documented company integrity event, unrelated to agomelatine. On 21 December 2023 the Paris Court of Appeal found the Servier group guilty over its weight-loss and diabetes drug Mediator (benfluorex), including fraud and improperly obtaining marketing authorisations. Penalties totalled about €431 million, including roughly €416 million in reimbursements and €9 million in criminal fines. Servier said it would appeal to France's Court of Cassation (pharmaphorum, December 2023; Servier Mediator information page). This case does not involve agomelatine. We include it because readers weighing company-generated evidence may reasonably want to know about it. It is not evidence that agomelatine data are unreliable.

Related research

• Stress, anxiety and depression: prevention and treatment guide
• Supplements for stress, anxiety and depression: the evidence
• Melatonin: how the hormone agomelatine mimics works as a supplement
• Sleep: evidence-based prevention guide and sleep supplements evidence
• St John's wort, saffron for depression, magnesium, L-theanine and ashwagandha
• Sibling medicine reviews: fluoxetine, escitalopram, sertraline, vortioxetine, mirtazapine, trazodone, venlafaxine, fluvoxamine (do not combine) and nortriptyline

Frequently asked questions

Sources and funding notes

• EMA: Valdoxan EPAR overview — EU regulator (Amsterdam); mostly industry-fee funded.
• Valdoxan EU product information (SmPC), last updated 15/04/2025 — written by Les Laboratoires Servier (France) and approved by the EMA.
• EMA: Valdoxan European public assessment report — EU regulator; covers the 2006 negative opinion and the 2007 refusal.
• EMA: Outcome of assessment to extend the use of Valdoxan (31 May 2024) — EU regulator.
• Valdoxan UK SmPC (revised 15/04/2026) — licence holder text, MHRA-approved.
• Agomelatine Zentiva UK SmPC — generic licence holder text, MHRA-approved.
• MHRA Drug Safety Update, October 2012 and November 2014 — UK regulator.
• NICE TA231: agomelatine (terminated appraisal, 2011) — UK government-funded.
• NICE NG222: Depression in adults (2022) — UK government-funded.
• NHS: Antidepressants overview — UK government health service.
• Cipriani et al., Lancet 2018 — UK/Japan/international; NIHR and JSPS funding; 409 of 522 trials (78%) industry-funded; one co-author declares lecture fees from Servier among others.
• Guaiana et al., Agomelatine versus other antidepressive agents, Cochrane 2013 (full text) — Canada/UK/Germany; authors declare no conflicts; NIHR is the largest funder of the Cochrane group.
• Koesters et al., Br J Psychiatry 2013 — Germany/Italy/UK academics; funding not shown in the abstract.
• Taylor et al., BMJ 2014 — UK; no specific grant; lead author declared Servier fees and grants.
• Freiesleben & Furczyk, J Mol Psychiatry 2015 — Germany (University of Rostock); funding not verified.
• CANMAT 2023 depression guideline (Lam et al., 2024) — Canada; internal funds, no industry funding for the guideline; many authors declare pharmaceutical fees.
• Zhang et al., J Affect Disord 2026 — China; academic.
• Stein et al., Adv Ther 2021 — international; Servier funding and Servier-employed co-authors.
• Hood et al., Int Clin Psychopharmacol 2025 — Australia/France; co-authors from Servier International and contracted consultancies; lead author declares Servier fees among others.
• Nussbaumer-Streit et al., Melatonin and agomelatine for preventing SAD, Cochrane 2019 — Austria/international; Cochrane review.
• Stone et al., BMJ 2022 — USA; analysis of 232 trials submitted to the FDA; funding statement not verified in the abstract.
• American Geriatrics Society 2023 Beers Criteria — USA; "no sponsor"; agomelatine not listed.
• Servier: governance and 2024/25 results (27 January 2026) — company statements (France).
• pharmaphorum: Servier Mediator appeal judgment (December 2023) — UK trade press, advertising-funded; Servier Mediator information page — company.

Last reviewed: 26 September 2026. Educational information from Pure City Research, not medical advice. Medicine decisions belong with a qualified prescriber.