Sexual Dysfunction in Male Patients After Initiating Treatment with Antidepressants
Venlafaxine raised sexual dysfunction risk 27% vs citalopram overall — and 93% higher among hospitalized depression patients
Journal: Drug Safety | Published: 2025-11-08 | Type: Cohort Study | PMID:41203969Authors: Westphal L, Gräf DD, Le H, Hallgreen CE, Andersen M (Copenhagen Centre for Regulatory Science, University of Copenhagen)
Funding/COI: No direct industry funding for this study. Lead author Westphal's PhD fellowship is funded by a Novo Nordisk grant to CORS. Senior author Andersen's professorship was previously Novo Nordisk Foundation-funded (2016–2022); he has received past project grants from AstraZeneca, Lundbeck, Janssen, Novartis, MSD, and Pfizer, paid to his institution.
Summary
Using 15 years of Danish national health register data, this study compared healthcare-recorded sexual dysfunction rates across 310,105 new antidepressant treatment episodes in adult men. Venlafaxine (an SNRI) showed the highest risk, 27% above citalopram overall and nearly double among men with prior hospital-based depression care. Mirtazapine showed a modest protective signal, though smaller than earlier studies suggested.
Claims
Incidence rates of healthcare-recorded sexual dysfunction ranged from 18.80 (95% CI 17.60–20.00) to 28.5 (95% CI 26.00–31.30) per 1,000 person-years across drugs
Venlafaxine vs citalopram: HR 1.27 (95% CI 1.02–1.46)
Venlafaxine vs citalopram in prior-depression-hospital-contact subgroup: HR 1.93 (95% CI reported as 1.14–1.27 — see Red Flags)
Sertraline vs citalopram: HR 0.99 (95% CI 0.90–1.09) — no significant difference
Mirtazapine vs citalopram: HR 0.87 (95% CI 0.81–0.93) — statistically significant but modest
Mirtazapine's protective effect was substantially smaller than the 50%+ reduction reported in prior literature
Study Quality
This is a large, well-designed new-user active comparator cohort study with 310,105 treatment episodes across a 15-year national dataset — a more rigorous design than most prior work in this area, which relied on cross-sectional surveys or short-term RCT data. The use of propensity score matching and multiple sensitivity analyses (varying washout periods from 1 to 5 years, restricting to hospital-diagnosed depression) adds credibility. STROBE reporting standards were followed.
The outcome measure — healthcare-recorded sexual dysfunction via PDE5 inhibitor prescriptions or ICD-10 F52 diagnoses — captures only cases that surfaced in formal healthcare encounters. The authors acknowledge this produces incidence rates far below self-reported estimates in the literature (typically 40–70% for SSRIs). This is a known limitation of administrative data, not a fatal flaw, but it means the absolute numbers substantially undercount the true burden.
Red Flags
Obvious data error in the abstract: The subgroup hazard ratio for venlafaxine vs citalopram among prior-depression patients is reported as HR 1.93 (95% CI 1.14–1.27). A confidence interval of 1.14–1.27 cannot contain a point estimate of 1.93. Either the point estimate or the CI is wrong. This should have been caught in peer review.
Massive outcome undercount: Incidence rates are a fraction of self-reported rates, because men have to formally seek care and receive a diagnosis or PDE5 prescription. This undercount is likely non-random — younger men and those with less severe dysfunction are probably less likely to seek formal care — which could distort comparisons.
Channeling bias for mirtazapine: Patients already worried about sexual side effects, or those with known risk factors, may have been preferentially prescribed mirtazapine. This inflates the baseline sexual dysfunction risk in the mirtazapine group and attenuates the apparent protective effect — meaning the true protective effect may be larger than the observed HR 0.87 suggests.
No treatment indication data: The Danish prescription register doesn't record why a drug was prescribed. The primary analysis can't rule out confounding by indication. The depression-subgroup sensitivity analysis partially addresses this but doesn't fully resolve it.
Residual confounding from disease severity: Sicker patients may preferentially receive certain drugs and may also have more baseline sexual dysfunction — biasing estimates away from the null.
Largest dataset on this question to date: 310,105 treatment episodes from comprehensive national registers
New-user design with washout period eliminates prevalent-user bias
Active comparator design (vs placebo) reduces confounding by indication relative to uncontrolled studies
Multiple sensitivity analyses with 1-, 3-, and 5-year washout periods; results consistent across all
15-year study period (2001–2015) provides long-term follow-up unavailable in most prior work
Propensity score matching applied
No direct pharmaceutical company involvement in study design or execution
Verdict
This is a methodologically serious attempt to fill a gap in pharmacovigilance data on antidepressant-induced sexual dysfunction. Its sheer scale and national-register design make it more reliable than the cross-sectional and clinical-trial evidence that dominates this literature. The venlafaxine signal is clinically plausible and the consistency across sensitivity analyses adds confidence. That said, the abstract contains an arithmetic impossibility (HR 1.93 with CI 1.14–1.27) that no reviewer caught, which demands caution before citing the subgroup figure. The outcome undercount is also severe enough that absolute rates from this paper should not be used to counsel patients on how likely they are to experience problems — the relative comparisons are more defensible than the raw incidence numbers.