Finasteride and Depression: Recalibrating the Warning considering New Evidence

Abstract

Background Regulatory agencies including the MHRA and FDA have issued prominent warnings about finasteride-associated depression and suicidality. These warnings are based largely on observational studies and spontaneous adverse event reports. In clinical practice, the current framing could cause patients to decline or discontinue effective treatment for benign prostatic hyperplasia and androgenetic alopecia. We synthesize four independent lines of evidence emerging post 2024 to evaluate whether the magnitude of risk implied by current warnings is supported by the best available science.

Evidence synthesis We draw on a published meta-analysis of five longitudinal studies (N > 2.5 million), a two-sample Mendelian randomisation study with UK Biobank validation (N = 388,894; median follow-up 13.7 years), and two independent pharmacovigilance analyses of the FDA Adverse Event Reporting System (FAERS) and WHO VigiBase.

Findings When finasteride users are compared against patients receiving treatment for the same condition — the appropriate active comparator design — there is no meaningful increase in depression risk (HR 0.89, 95% CI 0.86–0.92). The elevated risk reported in earlier studies (HR up to 2.00) is a product of confounding by indication arising from comparisons with untreated controls. Mendelian randomization finds no causal association with depression or suicide across ten outcome definitions after false discovery rate correction, with incidence rates of 2.89 versus 2.90 per 1,000 person-years between users and non-users. Pharmacovigilance analyses demonstrate that the adverse event signal emerged only after 2012, in temporal alignment with media coverage rather than changes in prescribing volume or formulation. The signal concentrates in younger men taking the 1 mg dose for hair loss rather than older men on the pharmacologically stronger 5 mg dose — a distribution that is biologically paradoxical for a neurosteroid mechanism. Dutasteride, a dual 5-alpha reductase inhibitor with greater DHT suppression than finasteride, produces no signal across 17 years of FAERS data.

Conclusions Four methodologically independent approaches converge on the conclusion that the depression signal associated with finasteride is driven by confounding, stimulated reporting, and nocebo amplification rather than pharmacological causation. The current framing overstates the risk, may itself worsen psychiatric outcomes through nocebo mechanisms, and could create clinically meaningful friction that leads patients to decline effective therapy. The evidence base does not support the magnitude of risk reported previously with benign prostatic hyperplasia.

What you need to know

Regulatory warnings about finasteride and depression — including the MHRA’s 2024 patient alert card — were driven by a surge in spontaneous reports that tracked media coverage of observational studies that did not use appropriate control groups.
A meta-analysis of over 2.5 million patients shows that when finasteride users are compared to patients receiving treatment for the same condition, there is no meaningful increase in depression risk (HR 0.89, 95% CI 0.86–0.92). The elevated risk seen in studies using untreated controls (HR 2.00) is a product of study design.
Pharmacovigilance data from both the FDA and WHO databases show that the signal appeared only after 2012, tracks media coverage rather than prescribing volume, concentrates in the 1 mg dose rather than the pharmacologically stronger 5 mg dose, and is absent for dutasteride — a more potent 5-alpha reductase inhibitor. These patterns are incompatible with the proposed neurosteroid mechanism.
A mendelian randomisation finds no causal association across ten outcome definitions, though the available genetic instruments have limitations. A secondary RCT analysis suggests any pharmacological effect is small — approximately 10%, not the twofold increase implied by earlier studies.
The current warning framing may worsen psychiatric outcomes through the nocebo effect and is creating friction that leads patients to decline effective treatment. The latest evidence does not support the warning at the magnitude implied in patients with benign prostatic hyperplasia.

A warning that has become its own clinical problem

In April 2024, the MHRA introduced a patient alert card for finasteride, warning of psychiatric side effects including depression and suicidal ideation [1]. The FDA had added depression to finasteride’s label in 2011 and suicidal behaviour in 2022; the EMA followed a similar path. These actions were precautionary responses to accumulating spontaneous reports and a body of observational literature reporting up to a twofold increase in depression risk among 5-alpha reductase inhibitor users [2, 3].

The MHRA’s 2023 safety review was thorough. It assessed 25 studies identified through a systematic literature search, 281 Yellow Card reports, stakeholder engagement sessions with patients and professional bodies, and regulatory actions taken in other jurisdictions. The review identified the right questions: the evidence was mixed, the inverse dose-response was unexplained, depression was acknowledged to be multifactorial, and the possible contribution of media attention and the nocebo effect was noted. The patient card was a reasonable decision given the evidence available.

Three developments since the MHRA’s review cut-off of February 2023 change the picture. First, a meta-analysis has shown that the elevated depression risk in the literature is entirely attributable to comparator selection bias [4]. Second, two pharmacovigilance analyses have demonstrated that the pharmacovigilance signal appeared only after 2012 and is absent for dutasteride — a more potent inhibitor of the same pathway [5, 6]. Third, a Mendelian randomisation study found no causal association across ten outcome definitions [7]. The ambiguities the MHRA identified can now be resolved. This Analysis examines these evidence lines and argues that the current regulatory framing should be recalibrated.

The evidence behind the warning — and the flaw within it

A meta-analysis of five longitudinal studies involving over 2.5 million patients found an I² of 95.5% — statistical heterogeneity so extreme that it points to a design-driven effect rather than a pharmacological one [4]. The source of the disagreement was a single methodological variable: who the finasteride users were compared against. Studies using non-drug-user controls — men with benign prostatic hyperplasia (BPH) who did not receive pharmacological treatment — reported a hazard ratio of 1.61 (95% CI 1.20–2.16). Studies using active comparators — patients receiving alpha-blockers such as tamsulosin for the same condition — reported a hazard ratio of 0.89 (95% CI 0.86–0.92). The same drug, the same indication, entirely opposite risk estimates — determined by comparator choice alone.

The explanation is confounding by indication: a bias where treated patients systematically differ from untreated controls. Men who seek and receive pharmacological treatment for BPH have more severe symptoms, attend clinic more often, and are more likely to receive a diagnosis of depression because they are more visible to the healthcare system. Comparing finasteride users to untreated men does not estimate the drug’s effect on depression; it estimates the combined effect of the drug and all the ways in which treated patients differ from untreated ones.

The active comparator design removes this bias by restricting the comparison to patients receiving treatment for the same condition at a similar level of clinical engagement. The elevated risk reported in earlier studies is a product of study design, and no synthesis had yet brought this evidence together at the time of the MHRA review.

The pharmacovigilance signal contradicts the proposed mechanism

The proposed mechanism linking 5-alpha reductase inhibitors to depression involves the neurosteroid pathway: 5-alpha reductase converts progesterone to dihydroprogesterone, which is subsequently metabolised to allopregnanolone — a potent positive allosteric modulator of GABA-A receptors with anxiolytic and mood-stabilising properties [8]. If this mechanism were responsible for the reported depression signal, the pharmacovigilance data should show four specific patterns: a signal that predates media awareness, a stronger signal at higher doses and with longer exposure, a signal for dutasteride (which suppresses DHT by approximately 98% compared with finasteride’s 70% and more completely inhibits allopregnanolone production), and a signal independent of the comparator drug used for the same indication. The data show none of these.

Nguyen et al. analysed VigiBase — the WHO’s global database of individual case safety reports — and found significant disproportionality for suicidality (reporting odds ratio [ROR] 1.63, 95% CI 1.47–1.81) and psychological adverse events (ROR 4.33, 95% CI 4.17–4.49) [5]. Suicidal ideation drove the entire suicidality signal (ROR 4.39, 95% CI 3.90–4.95); attempted and completed suicide did not meet the threshold for disproportionate reporting. The signal concentrated in men aged 45 or under taking finasteride for hair loss (ROR 3.47, 95% CI 2.90–4.15) and was absent in older men taking the 5 mg dose for BPH. The 5 mg dose itself did not produce a suicidality signal. Reports doubled after 2012 compared with before (ROR 2.13, 95% CI 1.91–2.39). The authors tested three comparator drugs — minoxidil (same alopecia indication, different mechanism), tamsulosin (same BPH indication, different mechanism), and dutasteride (same 5-alpha reductase mechanism) — and none produced a signal in any subgroup. The minoxidil null is particularly informative: if the signal reflected confounding by indication — younger men distressed by hair loss being inherently more vulnerable — minoxidil users should show a similar pattern.

Gupta et al.’s analysis of the US FDA Adverse Event Reporting System (FAERS) across three defined epochs confirms and extends these findings [6]. Before 2012 — before “post-finasteride syndrome” entered lay and clinical discourse — there was no finasteride signal for any of the five psychiatric outcomes studied. The signal then appeared and escalated: suicidal ideation ROR rose from 2.83 (95% CI 2.54–3.15) in 2013–2018 to 4.96 (95% CI 4.47–5.50) in 2019–2023, with similar escalation for completed suicide and depression-related suicidality. The FAERS data replicated the inverse dose-response: 1 mg signals consistently and significantly exceeded 5 mg signals (Breslow-Day p<0.05). Across all three FAERS epochs — covering 17 years of data — dutasteride produced no signal for any of the five outcomes.

Both groups of investigators concluded that stimulated reporting could explain the pattern. Together, the two analyses document when the stimulation occurred (after 2012), in whom (young men with alopecia, not older men with BPH), at what dose (1 mg, not 5 mg), and demonstrate through the dutasteride and minoxidil nulls that the proposed neurosteroid mechanism cannot account for the observed signal.

The genetic evidence

Mendelian randomisation (MR) uses naturally occurring genetic variants as proxies for a drug exposure, effectively constructing a randomised trial from population data. Wang et al. tested genetic proxies for finasteride exposure against ten depression and suicide outcome datasets comprising over 2.7 million cases and controls [7]. After correcting for multiple hypothesis testing, none of the ten outcomes showed a significant causal association. The same study included a prospective analysis of 388,894 UK Biobank participants (2,010 finasteride users, mean age 62.9 years) followed for a median of 13.7 years, which found virtually identical depression incidence: 2.89 per 1,000 finasteride users per year against 2.90 per 1,000 in controls. The two approaches complement each other: the MR analysis tests whether the causal mechanism exists across the population using genetic proxies, while the prospective cohort tracks depression incidence in actual finasteride users over more than a decade. Both converge on a null finding.

However, it’s important to interpret the results with some caveat. The number of genetic instrumental variables was insufficient to conduct formal heterogeneity or pleiotropy tests, meaning one of the three core MR assumptions — the exclusion restriction — could not be verified. The UK Biobank cohort predominantly comprises older men prescribed finasteride for BPH and may underrepresent the younger alopecia population where pharmacovigilance signals concentrate. These limitations do not invalidate the null finding, but they constrain how confidently it can be generalised. How to integrate this uncertainty into regulatory policy, and how to design trials that can resolve the question in a media-saturated information environment, is discussed below.

Convergent evidence across four methodological layers

Evidence layer	Source	Study design	Key result	How confounding was controlled
Observational meta-analysis	Nguyen et al. 2026, Postgrad Med [4]	5 longitudinal studies; N > 2.5 million	Non-user comparator: HR 1.61 (95% CI 1.20–2.16). Active comparator: HR 0.89 (95% CI 0.86–0.92). Difference entirely attributable to comparator choice.	Active comparator design (alpha-blockers for same indication)
Pharmacovigilance (VigiBase)	Nguyen DD et al. 2021, JAMA Dermatol [5]	WHO global ICSR database; age, indication, and comparator drug analyses	Suicidality ROR 1.63 (95% CI 1.47–1.81) — driven by ideation, not attempts or completions. Signal in men ≤45 yr with alopecia only. 5 mg: no signal. Post-2012: ROR 2.13. No signal for minoxidil, tamsulosin, or dutasteride.	Pre/post-2012 stratification; three comparator drugs
Pharmacovigilance (FAERS)	Gupta et al. 2025, J Cosmet Dermatol [6]	FAERS disproportionality across three epochs: 2006–2011, 2013–2018, 2019–2023	No signal 2006–2011. Suicidal ideation ROR 2.83 (2013–2018), rising to 4.96 (2019–2023). 1 mg > 5 mg (Breslow-Day p<0.05). Dutasteride: zero signal across all epochs.	Dutasteride as internal pharmacological control; inverse dose-response
Mendelian randomisation	Wang et al. 2025, J Cosmet Dermatol [7]	Two-sample MR + UK Biobank cohort; N = 388,894; median 13.7 yr follow-up	No causal association across 10 outcomes after FDR correction. Incidence 2.89 vs 2.90 per 1,000/yr.	Genetic instrumental variables bypass observational confounding

BPH = benign prostatic hyperplasia; CI = confidence interval; DHT = dihydrotestosterone; FAERS = FDA Adverse Event Reporting System; FDR = false discovery rate; HR = hazard ratio; MR = Mendelian randomisation; ROR = reporting odds ratio; 5-ARI = 5-alpha reductase inhibitor.

Recalibrating the warning

The clinical situations stemmed from the current warnings merit consideration alongside the new evidence. Urologists report that the warnings have shifted the risk-benefit conversation, with some calling for mental health screening before initiating finasteride [9]. My experience working with patients across the UK, Vietnam, South Korea, and the United States paints a consistent picture: the warnings have created tremendous friction in initiating effective treatment, with patients increasingly focused not on efficacy but on whether finasteride will cause depression.

The warning is not a neutral safety measure — it is a clinical intervention with consequences. The nocebo literature documents this consistently: informing patients about sexual side effects multiplied reporting tenfold for atenolol (3% to 31%); including gastrointestinal warnings for aspirin increased discontinuation sixfold without changing objective bleeding rates [10]. In antidepressant trials, placebo groups report side-effect profiles matching the expected drug class, confirming that patients experience what they are primed to expect [11]. The Mondaini RCT demonstrated this for finasteride directly: counselling about sexual side effects tripled symptom reporting under identical drug exposure [12]. Patients primed by an alert card to expect psychiatric harm become hypervigilant to normal mood fluctuations, attribute them to the drug, and report them — feeding a cycle that generates further regulatory concern. Meanwhile, patients who decline finasteride face different harm: androgenetic alopecia is itself associated with depression and impaired quality of life [13].

For older patients taking finasteride for BPH, the evidence converges on a clear answer. The meta-analysis shows no increase in depression risk when active comparators are used (HR 0.89) [4]. Neither VigiBase nor FAERS detects a signal at the 5 mg dose [5, 6]. The MR study — whose UK Biobank cohort predominantly comprises older men prescribed finasteride for BPH (mean age 62.9 years) — finds virtually identical depression incidence between finasteride users and controls (2.89 vs 2.90 per 1,000 per year) [7]. Dutasteride, which suppresses the same neurosteroid pathway more completely, produces no signal across 17 years of FAERS data [5]. For this population, the available evidence does not support the current warning language, which should be substantially softened or dropped.

For younger patients taking finasteride for alopecia, the question remains open. A small pharmacological effect cannot be excluded — the PCPT secondary analysis suggests approximately 10% [14], not the 200% increase implied by the studies that informed current warnings. Neither VigiBase nor FAERS can distinguish a genuine vulnerability in this population from stimulated reporting, confounding by the psychological burden of alopecia itself, and nocebo effects amplified by widespread media coverage of “post-finasteride syndrome.” Patients with a personal history of depression warrant a clinical conversation before starting treatment, and this guidance should remain.

The road ahead

Resolving this question requires randomised evidence, but conducting an interpretable trial of finasteride’s psychiatric effects in the current information environment presents a distinct methodological challenge. Finasteride now carries such cultural and media baggage that any trial is compromised before enrolment begins: patients who fear the drug will not participate, and those who do enrol carry expectations that contaminate psychiatric outcomes even under double-blinding. Trial designs that manipulate the information given to participants — such as authorised deception protocols in which patients consent to receive incomplete framing as part of the study design [15] — are conceptually appealing but face uncertain ethical review.

A cleaner path exists. Dutasteride inhibits the same 5-alpha reductase pathway more potently than finasteride but carries none of the expectation burden — there is no “post-dutasteride syndrome” in public discourse, and 17 years of pharmacovigilance data show no psychiatric signal. A double-blind, placebo-controlled trial of dutasteride in young men with androgenetic alopecia, using validated psychiatric outcome measures such as the PHQ-9 as a secondary endpoint, would directly test the neurosteroid mechanism in the population where the question is most pressing — without the nocebo contamination that makes any finasteride trial in this population uninterpretable. If dutasteride produces psychiatric effects, the pharmacological concern is genuine and the drug class warrants a warning. If it does not, the finasteride signal is an artefact of expectation and media-driven reporting, and the warning should be revised accordingly.

Conclusion

Three independent lines of evidence converge: the depression signal associated with finasteride is driven by comparator selection bias and stimulated reporting, not by the drug’s pharmacology. For older patients with BPH, the evidence supports no excess risk and the warning should be softened. For younger patients with alopecia, the question requires a trial designed for the information environment we now inhabit — one that tests the mechanism without the noise.

Until that evidence arrives, the patient alert card should contextualise the risk, acknowledge the uncertainty, and avoid language implying established causation. The framing of a warning is itself a clinical intervention — and the current framing may be causing more harm than the risk it describes.

References

[1] Medicines and Healthcare products Regulatory Agency. Safety review of Finasteride - Public Assessment Report. 2024.

[2] Welk B, McArthur E, Ordon M, et al. Association of Suicidality and Depression With 5α-Reductase Inhibitors. JAMA Intern Med 2017;177:683. doi:10.1001/jamainternmed.2017.0089

[3] Garcia-Argibay M, Hiyoshi A, Fall K, et al. Association of 5α-Reductase Inhibitors With Dementia, Depression, and Suicide. JAMA Netw Open 2022;5:e2248135. doi:10.1001/jamanetworkopen.2022.48135

[4] Nguyen M-H, Tran TH, Donovan J. Depression risk of 5-alpha reductase inhibitors: impact of active-comparator versus non-drug user control groups on risk measurement. Postgrad Med 2026;0:1–8. doi:10.1080/00325481.2026.2633464

[5] Nguyen D-D, Marchese M, Cone EB, et al. Investigation of Suicidality and Psychological Adverse Events in Patients Treated With Finasteride. JAMA Dermatol 2021;157:35–42. doi:10.1001/jamadermatol.2020.3385

[6] Gupta AK, Bamimore MA, Williams G, et al. Finasteride Use: Evaluation of Depression and Suicide Risk. J Cosmet Dermatol 2025;24:e70102. doi:10.1111/jocd.70102

[7] Wang J, Lin D, Zhao J, et al. Finasteride Use Does Not Lead to Depression or Suicide: Insights From a Large-Scale Cohort Study and Mendelian Randomization Analysis. J Cosmet Dermatol 2025;24:e70579. doi:10.1111/jocd.70579

[8] Melcangi RC, Santi D, Spezzano R, et al. Neuroactive steroid levels and psychiatric and andrological features in post-finasteride patients. J Steroid Biochem Mol Biol 2017;171:229–35. doi:10.1016/j.jsbmb.2017.04.003

[9] Al Saffar H, Xu J, O’Brien JS, et al. US Food and Drug Administration Warning Regarding Finasteride and Suicidal Ideation: What Should Urologists Know? Eur Urol Open Sci 2023;52:4–6. doi:10.1016/j.euros.2023.04.009

[10] Häuser W, Hansen E, Enck P. Nocebo Phenomena in Medicine. Dtsch Arztebl Int 2012;109:459–65. doi:10.3238/arztebl.2012.0459

[11] Mora MS, Nestoriuc Y, Rief W. Lessons learned from placebo groups in antidepressant trials. Philos Trans R Soc Lond B Biol Sci 2011;366:1879–88. doi:10.1098/rstb.2010.0394

[12] Mondaini N, Gontero P, Giubilei G, et al. Finasteride 5 mg and sexual side effects: how many of these are related to a nocebo phenomenon? J Sex Med 2007;4:1708–12. doi:10.1111/j.1743-6109.2007.00563.x

[13] Aukerman EL, Jafferany M. The psychological consequences of androgenetic alopecia: A systematic review. J Cosmet Dermatol 2023;22:89–95. doi:10.1111/jocd.14983

[14] Unger JM, Till C, Thompson IM, et al. Long-term Consequences of Finasteride vs Placebo in the Prostate Cancer Prevention Trial. JNCI J Natl Cancer Inst 2016;108:djw168. doi:10.1093/jnci/djw168

[15] Colloca L. The Nocebo Effect. Annu Rev Pharmacol Toxicol 2024;64:171–90. doi:10.1146/annurev-pharmtox-022723-112425

Citation

Cited as:

Nguyen, Minh-Ha. “Finasteride and Depression: Recalibrating the Warning considering New Evidence”. mh-nguyen.cv (Apr 2026). https://mh-nguyen.cv/posts/finasteride-depression-warning/

@article{nguyen2026finasteride,
  title   = "Finasteride and Depression: Recalibrating the Warning considering New Evidence",
  author  = "Nguyen, Minh-Ha",
  journal = "mh-nguyen.cv",
  year    = "2026",
  month   = "Apr",
  url     = "https://mh-nguyen.cv/posts/finasteride-depression-warning/"
}