What SSRIs Actually Do — And What the Evidence Says About Stopping Them

If you are in crisis, stop reading and get help now. Samaritans, free from any phone in Ireland and the UK: 116 123, 24 hours, every day. At immediate risk: A&E, or 999 or 112. Thoughts of ending your life are a medical emergency and they are treatable — tell someone tonight, not eventually.

If you take an SSRI, or you've been offered one, you have probably been told some version of "it corrects a chemical imbalance." That sentence is doing a lot of work it can't support. What an SSRI does mechanically is well established and fairly simple. What that mechanism means — about depression, about why the drugs help some people, about what happens when you stop — is genuinely contested in the published literature right now, by named researchers, in print, with numbers.

This article is the mechanism, the efficacy evidence, and the stopping evidence, with the disagreements left in rather than smoothed over. We sell supplements, not medicines, and some of what we sell interacts with these drugs — so the last section is about that, honestly. We are not doctors and nothing here is medical advice or a reason to change a prescription.

What an SSRI does mechanically

"SSRI" stands for selective serotonin reuptake inhibitor. The mechanism is in the name. When a neuron releases serotonin into a synapse, a protein called the serotonin transporter — SERT — pulls it back into the cell that released it, clearing it from the gap. An SSRI binds SERT and blocks that reuptake. Serotonin stays in the synapse longer.

This part is not disputed. It is measurable, it is dose-dependent, and it happens within hours of the first dose. Sertraline, escitalopram, citalopram, fluoxetine, paroxetine and fluvoxamine are the SSRIs in common use; they differ in how selectively they hit SERT versus other targets, how long they persist in the body, and which other proteins they touch, but they share that core action.

Here is the thing worth sitting with: the mechanism takes effect in hours, and the clinical benefit, when it arrives, takes weeks. Whatever makes an SSRI help somebody, it is not simply "more serotonin in the synapse," because that happens on day one. Something downstream and slower — changes in receptor sensitivity, in gene expression, in the way circuits adapt — has to be involved. That gap between the immediate pharmacology and the delayed effect is one of the open problems in the field, not a settled story.

The "chemical imbalance" claim is contested — and that matters less than you'd think

In 2022 a team led by Joanna Moncrieff published an umbrella review in Molecular Psychiatry — a review of existing systematic reviews and meta-analyses — asking whether depression is actually associated with lowered serotonin.1 They pulled 17 studies across six research areas. The headline findings:

  • Two meta-analyses of the serotonin metabolite 5-HIAA in body fluids found no association with depression (largest n = 1,002).
  • One meta-analysis of plasma serotonin found no relationship with depression — and found that lowered serotonin was associated with antidepressant use (n = 1,869).
  • Meta-analyses of 5-HT1A receptor binding (largest n = 561) and SERT binding (largest n = 1,845) showed weak and inconsistent evidence of reduced binding in some brain regions — a pattern that, if anything, would be consistent with more available serotonin, not less. The authors note that prior antidepressant use was not reliably excluded.
  • Tryptophan depletion — experimentally lowering the raw material the body makes serotonin from — did not produce depression in most healthy volunteers (n = 566), with weak evidence of an effect in those with a family history (n = 75). No systematic review of these studies had been done since 2007.
  • The two largest and highest-quality SERT gene studies — a genetic association study of 115,257 people and a collaborative meta-analysis of 43,165 — found no association with depression and no gene-by-stress interaction.

Their conclusion was that the main areas of serotonin research provide no consistent evidence of an association between serotonin and depression.

That review has been disputed in print, and you should know that before you treat it as the final word. Jauhar, Cowen and Browning published a narrative review in the Journal of Psychopharmacology summarising the evidence for a serotonergic contribution, drawing on psychopharmacology, molecular imaging and systems neuroscience.2 A separate correspondence piece in Molecular Psychiatry, titled "A leaky umbrella has little value," argued the review's methodology could not support its conclusion. This is an active argument between researchers, not a case of one side having won.

But notice what the argument is actually about. It is about whether depression is caused by low serotonin. It is not about whether SSRIs do anything. Those are separate questions, and conflating them is the single most common error in writing about this. Aspirin works on headaches; headaches are not caused by an aspirin deficiency. A drug that acts on a system can help without that system being the origin of the problem. The efficacy question has its own evidence, and that evidence is better than the mechanism story.

Do they work? The largest comparison we have

The reference point is a 2018 network meta-analysis in The Lancet led by Andrea Cipriani, covering 21 antidepressants.3 It screened 28,552 citations and included 522 double-blind randomised trials with 116,477 participants, including unpublished trials and regulatory-agency data — which matters, because publication bias in this field runs in one direction.

What it found:

  • All 21 drugs beat placebo on response rate. Odds ratios ranged from 2.13 (95% credible interval 1.89–2.41) for amitriptyline down to 1.37 (1.16–1.63) for reboxetine.
  • The differences between the active drugs were much smaller than the gap to placebo — odds ratios between drugs spanned 1.15 to 1.55 for efficacy, with wide credible intervals on most comparisons.
  • On acceptability — people dropping out for any reason — only agomelatine (0.84) and fluoxetine (0.88) did better than placebo. Clomipramine did worse (1.30).
  • In head-to-head studies, escitalopram, paroxetine, venlafaxine, mirtazapine, amitriptyline, agomelatine and vortioxetine came out more effective than other options; fluoxetine, fluvoxamine, reboxetine and trazodone came out least effective.
  • Quality was a real limitation. 46 of 522 trials (9%) were rated high risk of bias, 380 (73%) moderate, and only 96 (18%) low. The authors rated the certainty of evidence as moderate to very low.

The honest summary: these drugs outperform placebo in randomised trials, the effect is real but modest, no single drug is clearly best for everyone, and the evidence base has meaningful quality problems that the authors themselves flag. Anyone telling you either "antidepressants don't work" or "antidepressants are highly effective" is reading past the confidence intervals.

Two further points of context. First, these trials measured acute treatment of major depressive disorder in adults — not long-term maintenance, not adolescents, not treatment-resistant depression, all of which have their own separate literatures. Second, medication is one option among several with trial evidence behind them; we went through the wider picture, including the psychological therapies and the exercise data, in depression: what actually works. SSRIs are also prescribed for anxiety disorders, OCD and panic, which the efficacy figures above do not cover — if what you are weighing up is anxiety specifically, what actually works for anxiety is the relevant one.

Side effects, and one that's relevant to what we sell

The common ones are well known from trial data: nausea and other gastrointestinal effects early on, headache, sleep disruption in either direction, and sexual side effects — reduced desire, delayed or absent orgasm, erectile difficulty — which are common, frequently under-discussed, and for some people the reason they stop.

One worth naming specifically, because it touches a product category we sell into: SSRIs are associated with increased bleeding risk. Serotonin is involved in platelet aggregation, and platelets cannot make their own — they take it up via the same transporter the drug blocks. An analysis of the FDA's adverse event reporting system covering 2004 to late 2025 identified 5,604 bleeding-related cases across 380,241 total reports naming an SSRI as a suspect drug.4 Unspecified haemorrhage was the most reported event, followed by contusion and haematoma; gastrointestinal and central-nervous-system bleeds appeared consistently. Patients over 65 were the largest age group, and a medication known to increase bleeding risk was also in use in 541 cases (9.7%) — most often an antiplatelet agent, then NSAIDs, then direct oral anticoagulants.

We want to be precise about what that kind of data can and cannot tell you, and so were its authors: they state plainly that their findings are descriptive and should not be read as comparative risk differences between the individual drugs. Reporting systems like FAERS collect reports, not rates. There is no denominator of people taking the drug, reporting is voluntary and uneven, and nothing in the dataset establishes that the drug caused the event. Bleeding shows up as a class-wide safety concern worth knowing about. It is not a risk figure, and the per-drug percentages in that paper are not a league table.

It is relevant here because fish oil also has a reputation for affecting bleeding, and the two together is a question people ask us. We looked at the actual trial evidence on the fish oil half of that question in does fish oil interact with blood thinners, and the short version is that the controlled data is far less alarming than the consumer databases imply. That does not settle the combination, and we are not going to pretend it does — if you are on an SSRI and considering omega-3, that is a conversation for your prescriber, who knows the rest of your medication list.

Stopping them: the part where the literature openly disagrees

This is the section most worth your time, because the published estimates differ by a factor of nearly four and you will be quoted whichever one suits the source.

The lower estimate. Henssler and colleagues published a systematic review and meta-analysis in The Lancet Psychiatry in 2024 covering 79 studies and 21,002 patients.5 Incidence of at least one discontinuation symptom was 31% (95% CI 27–35%) after stopping an antidepressant — and 17% (14–21%) after stopping placebo. Subtracting that non-specific component, they concluded the attributable incidence is approximately 15%, or one in six to seven people. Severe symptoms came out at 2.8% on drug versus 0.6% on placebo. Desvenlafaxine, venlafaxine, imipramine and escitalopram showed higher frequencies; imipramine, paroxetine and venlafaxine/desvenlafaxine higher severity. The authors flagged substantial heterogeneity and noted that factors not explained by diagnosis or drug — possibly including how investigators and patients report — were in play.

The middle estimate. A 2025 meta-analysis in Molecular Psychiatry by Zhang and colleagues pooled 35 studies and found withdrawal-symptom incidence of 42.9% overall and 44.4% across 11 RCTs.6 By class: SNRIs 29.7%, SSRIs 45.6%, tricyclics 59.7% — differences that were not statistically significant (p = 0.221). Two findings from this one are practically useful. First, longer treatment meant more withdrawal, in a dose-response pattern: 35.1% at 6–12 weeks, 42.7% at 12–24 weeks, 51.4% beyond 24 weeks. Second, and counter to standard advice, tapering versus stopping abruptly gave 34.5% versus 42.5% — a difference that did not reach significance (p = 0.484). Risk factors identified: being female, being younger, having had side effects early in treatment, higher doses, longer duration, abrupt cessation, and slower drug clearance. Symptoms usually appeared within two weeks.

The higher estimate. In 2025 Moncrieff, Horowitz and colleagues published an appraisal and reanalysis in Psychological Medicine arguing the Henssler review rests on unreliable data.7 Their objections: most of its data came from industry-sponsored efficacy trials where withdrawal was an afterthought; it used spontaneously reported adverse events as a proxy for withdrawal; withdrawal may have been misclassified as relapse; prior treatment duration and observation periods were short. Restricting the analysis to the five studies that used a systematic, purpose-built withdrawal measure, they got a pooled figure of 55% — with a 95% confidence interval of 31% to 81%, n = 601, high heterogeneity, and no subtraction of nocebo effects.

One thing you should know to read these properly. These are not three independent teams. Mark Horowitz is an author on both the 42.9% paper and the 55% reanalysis, and Joanna Moncrieff is an author on both the 55% reanalysis and the serotonin umbrella review discussed earlier. That does not make their numbers wrong — their methodological criticisms of using spontaneous adverse-event reports as a withdrawal measure are substantive, and the same objection we raised about pharmacovigilance data above applies with full force. But a reader comparing "31%" with "55%" should know the second figure comes from researchers reanalysing the first, with a prior position in the debate, and that the wide confidence interval and n = 601 reflect how little purpose-built data exists.

What all three agree on: withdrawal symptoms are real, they are not rare, longer use raises the odds, and they can be mistaken for the original condition returning — which is a clinically consequential confusion in both directions. What nobody has yet produced is good-quality prospective data on long-term users tapering slowly with long follow-up. Every one of these papers says so.

Where supplements come into this, and where we draw our own line

This is the part that is actually our business, so we will be direct about it.

Several supplements act on the same serotonin system an SSRI acts on. Combining them is not a neutral act. The specific concern is serotonin syndrome — too much serotonergic activity at once, which ranges from uncomfortable to life-threatening and has been documented in published case reports.

  • 5-HTP is a direct serotonin precursor. We don't sell it as a product, and the full case — including why the contamination history matters — is in 5-HTP: real evidence, real risk.
  • L-tryptophan is the precursor one step further back, and it is in Sleep Formula at a disclosed 100mg. The published serotonin-syndrome case reports involving it are covered in L-tryptophan, sleep and serotonin.
  • St John's Wort has both a serotonergic action and a serious enzyme-induction problem — it accelerates the clearance of a long list of drugs. Our write-up is here, and it includes the published Lancet heart-transplant rejection case.

The disclosure that matters most on this page: St John's Wort and 5-HTP are both ingredients in our Sleep Support blend. Their individual milligram amounts are not disclosed — they sit inside a 905mg, 14-ingredient blend, and our manufacturer has confirmed those per-ingredient amounts cannot be published. We are not going to print a dose we can't stand behind, and we are not going to imply a timeline for getting one. The serotonin-interaction warning naming SSRIs, SNRIs and MAOIs is on that product's page and label. The plain consequence: if you take an SSRI or any other medication affecting serotonin, do not take Sleep Support without clearing it with your doctor first. If you want a sleep product where every ingredient's dose is on the label, that is Sleep Formula — and here is how the two compare. It contains no 5-HTP and no St John's Wort.

The broader interaction picture across our whole range, drug by drug, is in Proco products and mental health medication interactions. And if the question underneath all of this is whether a supplement has any business in a low-mood plan at all, we answered that as straight as we could in can supplements help with anxiety and low mood — the answer involves the word "no" more often than a supplement company would like.

The short version

  • SSRIs block the serotonin transporter. That is immediate, measurable and not in dispute.
  • Whether depression involves lowered serotonin is in dispute, publicly, between named researchers. The drugs' effects don't depend on that question being settled.
  • Across 522 trials and 116,477 people, all 21 antidepressants studied beat placebo. The effect is real and modest; differences between drugs are small; evidence certainty is moderate to very low.
  • Benefit takes weeks while the pharmacology takes hours. Nobody has fully explained that gap.
  • Published withdrawal estimates range from ~15% attributable to 55%, depending on who measured and how. Longer use means higher odds. The best-quality prospective data does not exist yet.
  • Serotonergic supplements and SSRIs are a combination to clear with a prescriber, not to experiment with. Two of ours contain ingredients that qualify, and one of them doesn't disclose their doses.

The rest of this series

This is one of four explainers on the medication classes that come up most often alongside what we sell. Each one is the mechanism, the best current evidence, and the places where that evidence is weaker or stranger than the standard account — including where it counts against us.

  • What benzodiazepines actually do — GABAA amplification, why Z-drugs are not a separate category, and why stopping badly is the genuinely dangerous part.
  • What ADHD stimulants actually do — dopamine and noradrenaline transporters, where efficacy stops climbing with dose, and the placebo study that should unsettle every focus supplement on the market.
  • What antipsychotics actually do — four dopamine pathways and the side effects that follow, tardive dyskinesia rates, and the one prescribed off-label as a sleeping tablet.

This article is for informational purposes only and has not been evaluated by the FDA. It is not intended to diagnose, treat, cure, or prevent any disease. We are not doctors, pharmacists or prescribers. Nothing here is a reason to start, stop, change or skip a prescribed medication — antidepressant discontinuation in particular should be planned with the clinician who prescribed it, and stopping abruptly carries risks this article does not cover. If you take any medication affecting serotonin, speak to your doctor before taking any supplement.

References

  1. Moncrieff J, Cooper RE, Stockmann T, Amendola S, Hengartner MP, Horowitz MA. "The serotonin theory of depression: a systematic umbrella review of the evidence." Molecular Psychiatry, 2023, 28(8):3243–3256. doi:10.1038/s41380-022-01661-0
  2. Jauhar S, Cowen PJ, Browning M. "Fifty years on: Serotonin and depression." Journal of Psychopharmacology, 2023, 37(3):237–241. doi:10.1177/02698811231161813
  3. Cipriani A, Furukawa TA, Salanti G, et al. "Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis." The Lancet, 2018, 391(10128):1357–1366. doi:10.1016/S0140-6736(17)32802-7
  4. Chan ACY. "Pharmacovigilance Analysis of Bleeding Events Associated With Selective Serotonin Reuptake Inhibitors (SSRIs) Using the US Food and Drug Administration Adverse Event Reporting System (FAERS)." Cureus, 2026, 18:e110163. doi:10.7759/cureus.110163
  5. Henssler J, Schmidt Y, Schmidt U, Schwarzer G, Bschor T, Baethge C. "Incidence of antidepressant discontinuation symptoms: a systematic review and meta-analysis." The Lancet Psychiatry, 2024, 11(7):526–535. doi:10.1016/S2215-0366(24)00133-0
  6. Zhang MM, Tan X, Zheng YB, et al. "Incidence and risk factors of antidepressant withdrawal symptoms: a meta-analysis and systematic review." Molecular Psychiatry, 2025, 30(5):1758–1769. doi:10.1038/s41380-024-02782-4
  7. Moncrieff J, Hobday H, Sørensen A, et al. "Evidence on antidepressant withdrawal: an appraisal and reanalysis of a recent systematic review." Psychological Medicine, 2025, 55:e191. doi:10.1017/S0033291725100652