Finasteride Side Effects: The Trial Rates, the Label and the Reports

How common are finasteride side effects?

In the trials that got finasteride 1 mg approved for hair loss, 36 of 945 men (3.8%) reported a sexual adverse experience in year one, against 20 of 934 (2.1%) on placebo. The gap is 1.7 percentage points, or roughly one man in 59. That is the only randomised figure that exists, and a 2015 meta-analysis found the trials behind it were not designed well enough to detect much else.

Almost every page about finasteride side effects picks one number and defends it. Clinics quote 3.8% and call it reassuring. Forums quote thousands of case reports and call it a cover-up. Both are describing real data, and the reason they never converge is that a randomised incidence rate, a line on a regulator's label and a count of voluntary reports answer three different questions. This page separates them, gives the number attached to each, and says which questions the evidence still cannot answer.

Everything below refers to finasteride 1 mg daily, the dose used for male pattern hair loss and sold as Propecia. Where a figure comes from the 5 mg prostate dose, which is five times higher and studied in much older men, it is labelled as such, because blending the two is the most common error in this category.

Table of contents

• The three tiers of evidence, and why they disagree
• Finasteride side effects in the randomised trials
• What those trials could not have detected
• Two meta-analyses, opposite conclusions
• What the label lists that no trial measured
• Depression and suicidal ideation: what regulators decided in 2025
• Post-finasteride syndrome
• The side effects that are not sexual
• Reversibility, and the part the label does not settle
• The numbers in proportion
• What to ask at the appointment

The three tiers of evidence, and why they disagree

Three kinds of evidence exist for finasteride safety, and each one has a different denominator. That single fact explains most of the argument.

Tier	What it is	What it can tell you	What it cannot
Randomised trial incidence	Adverse events counted in men assigned to drug or placebo	A rate, with a control group to subtract	Anything rare, delayed, or that the trial did not ask about
Regulatory label	What the FDA or EMA requires the manufacturer to print	That an association was considered credible enough to list	How often it happens; most label entries carry no frequency
Spontaneous reports	Voluntary reports to FAERS, EudraVigilance, VigiBase	That cases exist, and roughly how the count changes over time	A rate of any kind, because nobody counts the people who took the drug and were fine

The last row is where most of the public argument lives, and it is the row least able to settle it. A spontaneous report database has a numerator and no denominator. Ten thousand reports mean something very different against ten thousand users than against ten million, and the databases do not know which.

Finasteride side effects in the randomised trials

The registration trials produced one table, and it is short. In three 12-month placebo-controlled studies, the FDA prescribing information for Propecia reports the following drug-related adverse experiences at 1% or above.

Adverse experience, year 1	Finasteride 1 mg (n=945)	Placebo (n=934)
Decreased libido	1.8%	1.3%
Erectile dysfunction	1.3%	0.7%
Ejaculation disorder	1.2%	0.7%
Discontinued for a sexual adverse event	1.2%	0.9%
Any sexual adverse experience (integrated)	3.8%	2.1%

Two things in that table get quoted constantly and one gets ignored. The quoted parts are 3.8% and the fact that placebo produced 2.1%, meaning more than half the men reporting a problem on the drug would have reported one on a sugar pill.

The ignored part is the placebo column itself. Erectile dysfunction at 0.7% in a group of men aged 18 to 41 who were not taking anything is a reminder that this population reports sexual problems at a baseline rate, and that the honest measure of harm is the difference between columns, not the drug column alone.

The label also records something the summaries almost never mention. A sexual function questionnaire given to men in the two vertex baldness trials found statistically significant differences favouring placebo in three of four domains at month 12, covering sexual interest, erections and perception of sexual problems. There was no significant difference on overall satisfaction with sex life. A more sensitive instrument found more signal than the open-ended adverse event reporting did, which is exactly what you would expect, and it is in the label.

For context on what the drug is doing while this is happening, the same document reports that at five years 100% of the placebo group had lost further hair against 35% of men on finasteride. That is the trade being weighed, and where you sit on the Norwood scale changes how much the upside is worth.

What those trials could not have detected

The trials behind the label were short, small and mostly funded by the manufacturer. That is not an accusation, it is a published finding.

A meta-analysis in JAMA Dermatology assessed the quality of safety reporting across 34 clinical trials of finasteride for androgenetic alopecia (Belknap et al, JAMA Dermatol 2015, PMID 25830296). Its results are worth reading slowly:

• None of the 34 trials had adequate safety reporting. Nineteen were partially adequate, 12 inadequate, and three reported no adverse events at all.
• Funnel plots were asymmetric, biased toward lower odds ratios for sexual adverse effects, which the authors read as systematic underdetection.
• No report assessed whether blinding held. In a trial where the drug produces a visible cosmetic effect and the outcome is self-reported, that matters.
• Twenty-six of 34 trials (76%) evaluated safety for one year or less. Finasteride is taken indefinitely.
• Nineteen (56%) were manufacturer-funded and 18 (53%) disclosed conflicts of interest.
• In a real-world cohort of 5,704 men prescribed finasteride at 1.25 mg or less for hair loss, only 31% would have met the entry criteria for the pivotal trials, and only 33% took it for more than a year.

The authors' conclusion is blunt: published trial reports "provide insufficient information to establish the safety profile for finasteride in the treatment of AGA."

That is the correct frame for 3.8%. It is not a ceiling on how often finasteride causes problems. It is the floor that a set of one-year trials, not built to detect much, managed to find anyway. Someone who reads it as proof of safety and someone who reads it as proof of a cover-up are both overreading a number that was never precise.

Two meta-analyses, opposite conclusions

Two systematic reviews pooled randomised trials of 5-alpha reductase inhibitors and reached opposite answers about sexual dysfunction in hair loss patients. Both are real, both are peer reviewed, and the disagreement is the most informative thing in this whole literature.

	Lee et al, Acta Derm Venereol 2019	Liu et al, J Sex Med 2016
Trials pooled	15 randomised, double-blind, placebo-controlled	17 randomised controlled
Subjects	4,495	17,494
Population	Male androgenetic alopecia only	BPH and alopecia, analysed separately
Risk of sexual dysfunction	RR 1.57 (95% CI 1.19 to 2.08) for 5ARIs overall	RR 1.21 (95% CI 0.85 to 1.72) in alopecia
Finasteride specifically	RR 1.66 (95% CI 1.20 to 2.30)	ED RR 0.66 (0.20 to 2.25) in alopecia
Conclusion	Significant increased risk	Not statistically significant in alopecia

The 2019 analysis restricted itself to hair loss patients on finasteride 1 mg or dutasteride 0.5 mg and found a 1.66-fold risk with finasteride. The 2016 analysis covered both indications and found a clear signal in men with benign prostatic hyperplasia (RR 2.56) and no significant signal in men with hair loss.

The gap is explained by dose, age and trial selection, not by one team being wrong. BPH trials use 5 mg in men whose mean age across that review was 60, a group with a high baseline rate of erectile dysfunction and more to lose. Alopecia trials use 1 mg in men in their twenties and thirties. Confidence intervals in the alopecia arm are wide enough to contain both a real effect and none, which is what "not statistically significant" means and is not the same as "no effect".

Anyone citing a single relative risk for finasteride without saying which population it came from is quoting half a result. The same problem shows up in the dutasteride and finasteride comparison, where the head-to-head trials used the 5 mg prostate dose rather than the 1 mg hair loss dose.

What the label lists that no trial measured

The postmarketing section of the Propecia label is where the events the trials never captured appear. It carries an explicit warning that these come from a population of uncertain size and that frequency cannot be estimated from them.

The listed events, verbatim from the FDA label:

• Reproductive system: sexual dysfunction that continued after discontinuation of treatment, including erectile dysfunction, libido disorders, ejaculation disorders and orgasm disorders; male infertility and poor seminal quality, with normalisation or improvement reported after stopping; testicular pain; haematospermia.
• Nervous system and psychiatric: depression, suicidal ideation and behaviour.
• Neoplasms: male breast cancer.
• Breast disorders: breast tenderness and enlargement.
• Hypersensitivity: rash, pruritus, urticaria and angioedema including swelling of the lips, tongue, throat and face.

"Sexual dysfunction that continued after discontinuation of treatment" is on the label of the drug, printed by the manufacturer at the FDA's direction. That single line ends the argument about whether persistent symptoms are acknowledged. It does not begin to answer how often they occur, because a postmarketing listing never carries a rate.

Note also what appears here and not in the trial table. Breast tenderness and enlargement showed no difference from placebo in the Propecia studies, according to the label. They are listed in postmarketing anyway, because a one-year trial in fewer than a thousand men cannot rule out an event that affects one in several thousand.

Depression and suicidal ideation: what regulators decided in 2025

European regulators completed a formal review in 2025 and added suicidal ideation to the product information for finasteride tablets. This is the most consequential change to finasteride labelling in a decade, and it came with numbers.

The EMA's safety committee assessed clinical trial data, literature and the EudraVigilance reporting database. It identified 325 relevant cases of suicidal ideation, 313 for finasteride and 13 for dutasteride, against an estimated exposure of around 270 million patient years for finasteride and around 82 million for dutasteride.

The committee's findings, in its own framing:

• Suicidal ideation is confirmed as a side effect of finasteride 1 mg and 5 mg tablets.
• Most cases were reported in people using the 1 mg tablets, the hair loss dose.
• The frequency is unknown, meaning it cannot be estimated from the available data.
• Packs of 1 mg finasteride now carry a patient card.
• No link was found for finasteride skin sprays, and their product information was not changed.
• Dutasteride showed no direct link, but gets the same wording as a precaution because it works the same way.

Hold the 325 against 270 million patient years before drawing a conclusion in either direction. That ratio is not an incidence rate, because spontaneous reporting captures a small and unknowable fraction of real events. It is still the best available sense of proportion in this niche, and it is the number most pages arguing either side leave out.

A separate pharmacovigilance study reached a more complicated conclusion. Analysing the WHO's VigiBase, researchers found a significant disproportionality signal for suicidality with finasteride (Nguyen et al, JAMA Dermatol 2021, PMID 33175100): reporting odds ratio 1.63 overall, rising to 3.47 in patients under 45 and 2.06 in those treated for alopecia, with no signal in older men treated for prostate symptoms. Reports also rose sharply after 2012 (ROR 2.13), when media coverage of the issue began.

The authors offer two readings of their own result and refuse to choose between them: stimulated reporting after publicity, or younger patients being genuinely more vulnerable. Both mechanisms are plausible, they are not mutually exclusive, and a disproportionality analysis cannot separate them. Any page that cites this study as settling the question in either direction has not read past the abstract.

Post-finasteride syndrome

Post-finasteride syndrome describes sexual, physical and psychological symptoms that persist after the drug is stopped. What is established and what is claimed can be stated separately, and should be.

What is established. Persistent sexual dysfunction after discontinuation is listed in the postmarketing section of the FDA label. The US National Institutes of Health added a post-finasteride syndrome entry to its Genetic and Rare Disease Information Center. Case series exist. Symptoms reported include low libido, erectile dysfunction, reduced arousal, difficulty reaching orgasm, depression, anxiety and cognitive complaints.

What is not established. There is no incidence rate. There is no diagnostic test, no agreed case definition and, as reviews state plainly, no evidence-based effective treatment. There is no controlled trial that followed men after discontinuation with a comparison group, which is the study design that would settle whether the syndrome represents a drug effect, a coincidence of timing, or a combination.

A 2016 systematic review of 5-alpha reductase inhibitor safety put the reported range of sexual effects at 3.4% to 15.8% of men across studies and concluded these drugs were well tolerated but not without risk. A 2026 systematic review of 5-alpha reductase isoenzymes and neuropsychiatric outcomes found impaired neurosteroid synthesis a plausible mechanism and pharmacovigilance data strengthening the association, while calling for translational studies with standardised psychiatric outcomes (PMID 41718149).

That mechanism is why researchers who have never seen an incidence rate still take the condition seriously. Finasteride blocks 5-alpha reductase, and the enzyme does not only make dihydrotestosterone. It also produces neuroactive steroids including allopregnanolone, which acts on GABA receptors in the brain. A drug suppressing that pathway has a route to psychiatric effects with nothing to do with hair.

The honest position is that persistent symptoms are recognised by regulators, undercounted by trials that mostly ran for a year, and unquantified by anyone. That is where the evidence currently stops.

The side effects that are not sexual

Three non-sexual effects matter enough to state, and two of them are about the 5 mg dose being read across to the 1 mg one.

PSA suppression is real and clinically important. In men aged 18 to 41 taking finasteride 1 mg, mean serum PSA fell from 0.7 ng/mL at baseline to 0.5 ng/mL at month 12. At the 5 mg prostate dose in older men, PSA drops by roughly 50%. The label's instruction is that any confirmed rise from the lowest PSA value while on the drug should be evaluated, even when it still sits inside the normal range. A man who starts finasteride at 30 and gets screened at 50 needs whoever orders that test to know he is on it.

High-grade prostate cancer. In the seven-year Prostate Cancer Prevention Trial, men aged 55 and over taking finasteride 5 mg had Gleason 8 to 10 prostate cancer at 1.8% against 1.1% on placebo. This is a 5 mg finding in a much older population, and the label states directly that the clinical significance for men taking Propecia is unknown. It belongs on the page, with that caveat attached.

Pregnancy exposure. Finasteride is contraindicated in women who are or may become pregnant, because blocking DHT interferes with development of the male external genitalia. Propecia tablets are coated, which prevents contact with the active ingredient during normal handling, but the label instructs that women who are or may be pregnant should not handle crushed or broken tablets. This is a household consideration, not just a patient one, and it is the reason the drug carries a contraindication rather than a warning.

Reversibility, and the part the label does not settle

The label's position on reversibility is specific and partial. It reports that events of erectile dysfunction and decreased libido "lasted for at least several weeks after drug discontinuation" where information was available, that resolution occurred both in men who stopped and in most of those who kept taking it, and that the incidence of each event fell to 0.3% or below by the fifth year of treatment.

That is the strongest argument that most sexual side effects settle, and it sits on the same document as a postmarketing entry for dysfunction that continued after stopping. There is no contradiction: the trials describe what happened to most men in them, and the postmarketing section describes a category of report the trials were not built to capture.

The effect on hair is unambiguously reversible, and the label puts a timeline on it. Withdrawal of treatment leads to reversal of effect within 12 months, and in the extension studies men switched from finasteride to placebo at month 12 had lost the hair count gain by month 24. Stopping means returning to the trajectory you were on, which is why a transplant does not remove the decision either.

The numbers in proportion

The arithmetic nobody prints is the absolute risk, so here it is.

The trial difference is 3.8% minus 2.1%, which is 1.7 percentage points. Inverted, that is one additional man reporting a sexual adverse event for every 59 treated for a year. Put the other way, 58 of every 59 men in those trials did not report one that placebo would not also have produced.

Now the counterweight. Of every 100 men in those same trials, 14 on finasteride had further hair loss at 12 months against 58 on placebo, and at five years the split was 35 against 100. Whether a one in 59 chance of a reported sexual event is worth that depends entirely on how much the hair matters to the person deciding, which is not a question a study can answer and not one this page will try to.

Three framings to avoid, because they show up constantly on both sides:

1. Quoting 3.8% without the 2.1%. More than half of that number is the placebo rate. The drug-attributable figure is 1.7 points.
2. Quoting spontaneous report counts as rates. Thirty-two FAERS reports, 325 EudraVigilance cases or several thousand forum accounts are numerators without denominators. The EMA's 270 million patient years is the closest thing to a denominator anyone has published.
3. Reading 5 mg data as 1 mg data. The prostate cancer finding, the 50% PSA drop and the higher sexual dysfunction rates in the BPH meta-analysis all come from five times the hair loss dose, in men decades older.

For anyone weighing this against the alternatives, topical finasteride reaches roughly a hundredfold lower plasma concentration than the oral drug while still suppressing serum DHT. Topical minoxidil works through a different mechanism with a side effect profile that has no hormonal component at all, and the cheapest version of it is Kirkland's generic, whose ingredient list is identical to the brand's. Oral minoxidil carries its own dose-dependent cardiovascular considerations. None of these is a free option, and none is finasteride with the risks removed.

What to ask at the appointment

Five questions that produce a useful answer from a dermatologist or hair restoration surgeon rather than a reassuring one:

• What is my baseline, and is any sexual or mood symptom already present before I start? Without that, neither of us can attribute anything later.
• What is my PSA now, and will it be recorded that I am taking this drug?
• What is the plan if a side effect appears, and how long do we wait before stopping?
• Would starting at a lower frequency than daily change the calculation, and what evidence is there either way?
• Given my stage and rate of loss, what happens if I do nothing for a year?

The last one matters most. Someone at Norwood 3 with rapid recent change and someone stable at the same stage are making different decisions, and the way stages are assessed is set out in the methodology.

Nothing here is a recommendation about whether to take finasteride. That decision depends on individual history and belongs with a dermatologist or hair restoration surgeon who can see the person in front of them.

Frequently Asked Questions

What percentage of men get side effects from finasteride?

In the trials supporting approval, 3.8% of men on finasteride 1 mg reported a sexual adverse experience in year one against 2.1% on placebo, a drug-attributable difference of about 1.7 percentage points. A 2015 JAMA Dermatology meta-analysis found none of the 34 trials in this literature had adequate safety reporting, so that figure is best read as a floor rather than a complete rate.

Do finasteride side effects go away?

The FDA label reports that resolution occurred in men who stopped the drug and in most of those who continued, with incidence falling to 0.3% or below by year five. The same label also lists, under postmarketing experience, sexual dysfunction that continued after discontinuation, with no frequency attached. Both statements are on the label and neither has been quantified against the other.

Does finasteride cause depression or suicidal thoughts?

European regulators confirmed suicidal ideation as a side effect of finasteride 1 mg and 5 mg tablets in 2025, based on 325 EudraVigilance cases against an estimated 270 million patient years of exposure, and stated that the frequency is unknown. The FDA label lists depression and suicidal ideation under postmarketing experience. Anyone with a history of depression has a reason to raise it before starting.

Does finasteride affect PSA and prostate cancer screening?

Yes. Finasteride 1 mg lowered mean PSA from 0.7 to 0.5 ng/mL over 12 months in men aged 18 to 41, and the 5 mg dose halves PSA in older men. The label instructs that any confirmed increase from the lowest value while on the drug should be evaluated even if it remains within the normal range, which means whoever orders a PSA test needs to know the drug is being taken.