In May 2019, the European Medicines Agency’s safety committee reviewed a signal and recommended a change to the product information for a group of antidepressants. The sentence they settled on is worth reading exactly as written, because every word in it is load-bearing: “There have been reports of long-lasting sexual dysfunction where the symptoms have continued despite discontinuation of SSRIs/SNRI.”
Reports. Have continued. No frequency, no mechanism, no claim about cause. Ten substances were named: citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline, desvenlafaxine, duloxetine, milnacipran and venlafaxine. The wording is public, and so are the minutes behind it.
Note: This is not medical advice. If any of this is relevant to your own medication, that conversation belongs with the person who prescribed it, and stopping an antidepressant on the strength of an article is a genuinely bad idea.
A label warning is not a finding
The most common misreading of the 2019 decision runs in both directions. One camp treats it as regulatory confirmation that the condition is real, common and drug-caused. The other treats the absence of a frequency estimate as evidence that regulators found nothing. Neither is what happened.
Regulators operate on a precautionary standard that is deliberately lower than a scientific one. A signal reaches the threshold for a label change when the reports are numerous enough and consistent enough that patients should be told, whether or not causation has been established. That is the whole design. The label exists to inform consent, not to settle a question.
Other agencies followed. Australia’s TGA updated its warnings in 2024. The UK’s MHRA announced a review of 28 antidepressants in December 2025, which is still running, and reporting an outcome for it would be inventing one. New Zealand’s Medsafe looked at the same evidence in September 2019 and wrote the sentence that has aged best: the prevalence, it said, is unknown.
Seven years later it still is.
Why the obvious study cannot be run
The design that would answer the question is easy to describe and impossible to execute. Take people with no sexual dysfunction, randomise half of them to an antidepressant and half to placebo, measure sexual function with a validated instrument at baseline, during treatment and for several years after discontinuation, and count.
No ethics committee will approve randomising healthy people to a psychiatric medication for years. Randomising depressed people fixes the ethics and destroys the measurement, because depression itself impairs sexual function, so the baseline is already confounded before anyone takes a tablet.
David Healy and Dee Mangin set out the full set of obstacles in a 2024 paper, and the striking thing about their list is how little of it is anyone’s fault. Patients rarely raise sexual symptoms unprompted. Clinicians rarely ask. Withdrawal effects in the first weeks after stopping mask exactly the window that matters. And because no MedDRA code for the condition has been adopted, pharmacovigilance databases cannot be searched for it in the ordinary way, which means the reporting system that exists to detect signals like this one cannot count this one.
The one incidence estimate, and what it actually counted
There is a single formal incidence figure in the literature, and it is a proxy.
Ben-Sheetrit and colleagues went through nineteen years of records from Israel’s largest health maintenance organisation, identifying 12,302 men aged 21 to 49 who met a strict set of inclusion and exclusion criteria. Of those, 866 had been prescribed an SSRI or SNRI; the remaining 11,436, who had not, served as controls. Among the 866 who had taken the drugs, the rate of later being prescribed a drug for erectile dysfunction that persisted was 0.46 percent, or one in 216.
What that number counts is a prescription for a PDE-5 inhibitor. It therefore excludes women entirely. It misses genital numbness, loss of orgasm and loss of desire, none of which produce that prescription. And it captures only the men who raised the problem with a doctor at all. The authors say in their own paper that it is likely an underestimate.
Both of the readings circulating online get this wrong in mirror image. It is not a finding that one in 216 people who take an antidepressant will develop persistent sexual dysfunction, because the denominator is male users and the numerator is one specific prescription. It is also not a finding that the condition is vanishingly rare, for the same reason. The paper is a floor built out of the only countable trace the health system happened to leave.
The court case that gets reported as a verdict
In May 2024 a petition was filed asking the FDA to add a similar warning to US labelling, and when it was not acted on, litigation followed. Csoka v. FDA was dismissed on 31 March 2025.
It was dismissed on standing. A court concluded the plaintiffs had not established the legal basis to bring the claim, which is a procedural ruling about who may sue and says nothing whatsoever about whether the condition exists. It gets cited as though a court weighed the science and found against it. No court has weighed the science.
What was never measured on the way in
The most useful methodological point in this whole area is not about the persistent form at all. It is about how sexual side effects were counted during treatment, because that failure shaped everything downstream.
The registration trials mostly relied on spontaneous reporting: side effects were recorded when patients raised them. When Angel Montejo and colleagues asked 1,022 outpatients directly, using a structured questionnaire, 59.1 percent reported sexual dysfunction, with wide variation by drug, from 72.7 percent on citalopram down to 3.9 percent on moclobemide. That study was open-label with no placebo arm, and depression impairs sexual function on its own, so the absolute figures carry real uncertainty. The gap between asking and waiting to be told does not.
A 2026 meta-analysis of thirteen randomised trials covering 5,941 participants puts orgasmic dysfunction at a relative risk of 3.28 against placebo, with high certainty, while the effect on desire was not statistically significant. That is on-treatment evidence and belongs strictly in the on-treatment conversation, not the post-discontinuation one.
Which brings the argument to its sharpest point, made by the psychiatrist Ronald Pies and quoted secondhand in discussions of this literature: clinicians almost never document baseline sexual function before prescribing. If nobody wrote down what was normal for a person beforehand, then years later, when that person says something has not come back, there is no record to compare against. Not for the individual, and not, multiplied across a population, for the field.
What the absence of a number does to people
Three mechanisms are under discussion, involving serotonin receptor desensitisation, neurosteroid depletion and epigenetic effects on dopaminergic circuitry. All three are hypotheses. There is no biomarker, no imaging finding and no human mechanistic evidence for any of them. A group led by Healy published consensus diagnostic criteria in 2022, which no manual and no regulator has adopted and which have not been validated against an independent sample. No professional psychiatric body has issued a position statement, and that silence is reportable as silence, not as rejection.
There are also clinicians who think the evidence base is far too thin to support the attention it gets. The psychiatrist George Dawson has argued at length that the reporting is unreliable and the implied frequency unsupported, and it is worth being precise about his position: he disputes the quality of the evidence, not the existence of anyone’s symptoms.
What sits underneath all of it is a group of people carrying a symptom that appears on a regulatory label, has consensus criteria nobody has adopted, has one incidence estimate built from a proxy nobody thinks is adequate, and cannot be quantified by any study that could ethically be run. The label acknowledges them. The literature cannot yet count them. Those two facts have coexisted for seven years, and there is currently no design on the table that would end the standoff.
If this is your situation and it is heavy to read about, the person to raise it with is your prescribing clinician, and support from a professional is worth more than anything an article can offer.
