Epidemiological association between male sexual dysfunction and depression: a systematic review and meta-analysis of observational studies published between 2020 and 2025

2.1 Eligibility criteria

This systematic review and meta-analysis was designed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 reporting guidelines [12, 13]. The completed PRISMA 2020 checklist is provided as Supplementary material. The protocol for this systematic review and meta-analysis was registered in the International Prospective Register of Systematic Reviews (PROSPERO, registration number: CRD420261280261; registration date: 07 January 2026). To precisely evaluate the epidemiological association between sexual dysfunction and mental health in the adult male population, we selected literature based on specific eligibility criteria. We restricted the study population to adult males aged 18 years and older. Following observations by Rajih et al. [1] and Saito et al. [3] regarding the rising prevalence trends among younger demographics, we adopted a strategy with no upper age limit to ensure the inclusion of data spanning the entire adult life spectrum.

Furthermore, to mitigate the confounding impact of antidepressant medication on sexual function, we prioritized studies involving drug-naive participants whenever feasible. The primary exposure for the main quantitative synthesis was ED, defined using validated ED-specific instruments (e.g., International Index of Erectile Function 5 (IIEF-5) questionnaire) or clinical diagnosis. Broader categories of male sexual dysfunction, including composite ASEX-defined sexual dysfunction or mixed PE/ED samples, were not treated as equivalent to ED and were not pooled with ED-focused studies in the primary meta-analysis. Such studies were retained only for limited narrative context when they provided clinically relevant complementary evidence.

Regarding antidepressant exposure, medication status was recorded during data extraction whenever reported. Studies were not excluded solely because some participants had used antidepressants; however, studies focused primarily on treatment-induced sexual dysfunction were excluded. Because the number of eligible studies with medication-stratified data was very small, subgroup analysis according to antidepressant use was not performed.

For this review, ED-focused studies were eligible for quantitative synthesis, whereas studies of broader male sexual dysfunction constructs (e.g., ASEX-defined general sexual dysfunction or mixed PE/ED samples) were retained only for narrative context when they contributed clinically relevant complementary evidence. Regarding the outcome variable, we accepted either formal clinical diagnoses of depression or symptom scores derived from standardized psychometric instruments, such as the PHQ-9 [14], SDS [15], and Beck Depression InventoryII (BDI-II) [16]. In terms of study design, we strictly included observational studies (cross-sectional, cohort, and case-control) that statistically validated independent correlations between the variables. Conversely, to maintain data reliability, we excluded randomized controlled trials (RCTs) assessing pharmaceutical or surgical interventions, reviews based solely on expert opinion, and abstracts where the full text was inaccessible.

2.2 Search strategy and information sources

We queried PubMed, EMBASE, the Cochrane Library, PsycINFO, and SCOPUS for academic literature published between January 2020 and December 2025. To optimize search sensitivity, we utilized Boolean logic combining Medical Subject Headings (MeSH) with free-text terms. Crucially, we expanded our keywords to include various subcategories of sexual dysfunction; this approach aimed to capture the cultural specificities and diverse clinical settings identified by Ghosh et al. [11] and Quang et al. [6]. Because the review was initially designed to capture the broader recent literature on male sexual dysfunction and depression, the search strategy included ED-related terms as well as broader sexual dysfunction terms. However, the primary quantitative synthesis in the revised manuscript was restricted to ED-focused studies. The PubMed query was, for example, as follows:

(“Sexual Dysfunction, Physiological”[MeSH] OR “Erectile Dysfunction”[MeSH] OR “Premature Ejaculation”[MeSH] OR “sexual dysfunction”[TIAB] OR “erectile function”[TIAB] OR “premature ejaculation”[TIAB] OR “low libido”[TIAB] OR “IIEF”[TIAB] OR “ASEX”[TIAB]) AND (“Depression”[MeSH] OR “Depressive Disorder”[MeSH] OR “depression”[TIAB] OR “depressive symptoms”[TIAB] OR “mood disorder”[TIAB] OR “PHQ-9”[TIAB] OR “BDI”[TIAB] OR “SDS”[TIAB]) AND (“Observational Study”[PT] OR “Cross-Sectional Studies”[MeSH] OR “Cohort Studies”[MeSH] OR “Case-Control Studies”[MeSH] OR “observational”[TIAB] OR “cross-sectional”[TIAB] OR “longitudinal”[TIAB] OR “prevalence”[TIAB]) AND (“2020/01/01”[PDAT] : “2025/12/01”[PDAT]).

No language restriction was prespecified at the search stage; however, only studies with sufficient full-text information for data extraction and quality assessment were included in the final synthesis.

2.3 Study selection

After removing duplicates, the researchers independently reviewed the collected records. We performed an initial screening based on titles and abstracts to assess thematic relevance. Following this, we conducted a full-text examination adhering to the inclusion criteria outlined by Xiao et al. [7], which led to the final selection of eight studies for analysis. A PRISMA flow diagram is shown in Fig. 1.

Fig. 1.PRISMA flow diagram. RCT: randomized controlled trial.

2.4 Data extraction

We systematically harvested data from the selected literature using a standardized extraction form. We categorized the extracted items into study-level characteristics (author, publication year, country, study design, recruitment period, sample size, and traits) and measurement specifications (instruments and cut-off points for sexual dysfunction and depression, recall window). To facilitate comparison between multidimensional assessments—such as the ASEX tool [17] employed by Sayed et al. [5]—and domain-specific instruments like the IIEF-5 [10], we meticulously documented the operational definitions of each measurement tool.

2.5 Quality assessment

We assessed the risk of bias in the included observational studies using the Newcastle-Ottawa Scale (NOS) [13]. This instrument evaluates three core domains: Selection (representativeness), Comparability, and Outcome, allowing each study to earn a maximum of nine stars (★). Specifically, we weighted the “Comparability” domain based on whether the study adequately controlled for key confounders—such as age, education level, and smoking status—through multivariable regression analysis. This rigorous approach ensures our evidence-based conclusions reflect both the quantitative findings and the methodological quality of the individual studies.

2.6 Statistical analysis

To synthesize prevalence data, we applied a Logit transformation to individual study proportions and employed a DerSimonian-Laird random-effects model, conservatively accounting for potential between-study heterogeneity. We evaluated statistical heterogeneity using the I² statistic, considering I² > 50% or Cochran’s Qp < 0.10 as indicative of meaningful heterogeneity. To investigate sources of variation, we conducted pre-planned subgroup analyses based on measurement instruments (e.g., PHQ-9, SDS) and participant characteristics, calculating tool-specific prevalence rates. For comparing associations and severity, we utilized adjusted odds ratios (aOR), regression coefficients (β), correlation coefficients (r), and standardized mean differences (SMD; Hedges’ g) as effect size indicators, presenting all results with 95% confidence intervals (CIs). All statistical tests were two-sided. Because the number of studies available within each ED-focused quantitative analysis was small, meta-regression and extensive subgroup exploration by region, age, or study quality were not considered statistically reliable and were, therefore, not performed. Single-study estimates, such as adjusted odds ratios, regression coefficients, correlation coefficients, and standardized mean differences, were not interpreted as pooled meta-analytic findings. Rather, they were presented as narrative or descriptive individual-study results. Exploratory publication-bias assessment used funnel-plot inspection and Egger’s regression test, but these results were interpreted cautiously because of the limited number of included studies and the marked imbalance in sample sizes. All meta-analyses were conducted using R software with standard meta-analytic procedures.

3.1 Study characteristics

The eight observational studies finally selected for this systematic review were published between 2021 and 2025 (Table 1, Ref. [1, 3, 5, 6, 7, 8, 9, 11]). Regarding study design, six investigations employed a cross-sectional approach, while two studies (Pitta et al. [9] and Jiang et al. [8]) utilized a retrospective design based on medical records. The cumulative sample size was 24,563 participants. Notably, Pitta et al. [9] significantly bolstered the statistical power of the analysis by contributing a large-scale general population dataset of 21,139 individuals. Conversely, Sayed et al. [5] and Ghosh et al. [11] focused on smaller clinical cohorts of 68 and 100 patients, respectively, allowing for in-depth analysis within specific clinical settings.

In terms of participant characteristics, the studies diverged in their target populations. Rajih et al. [1] and Saito et al. [3] aimed to assess sexual dysfunction within the general population, recruiting community-dwelling young men in Saudi Arabia and web-based samples in Japan, respectively. In contrast, Sayed et al. [5] and Jiang et al. [8] examined the prevalence of sexual dysfunction within cohorts already diagnosed with Major Depressive Disorder (MDD). Jiang et al. [8] specifically selected drug-naive, first-episode patients to rigorously control for the confounding effects of antidepressant medication. Finally, Quang et al. [6] and Xiao et al. [7] investigated the comorbidity of depressive symptoms among patients visiting hospitals specifically for ED treatment. Because most of the quantitatively informative and conceptually consistent studies focused on ED, the revised main synthesis centered on ED-focused studies, whereas the broader sexual dysfunction studies were retained only as contextual evidence. The general characteristics and corresponding study references are summarized in Table 1.

3.2 Measurement instruments

The most frequently used instrument for assessing ED across the selected literature was the IIEF-5 (also known as Sexual Health Inventory for Men (SHIM)), the abbreviated version of the International Index of Erectile Function [17]. Rajih et al. [1], Saito et al. [3], Quang et al. [6], Xiao et al. [7], and Pitta et al. [9] uniformly established a cut-off score of 21 or lower to identify the presence of ED.

By contrast, Sayed et al. [5] and Jiang et al. [8] used the Arizona Sexual Experience Scale (ASEX), which captures broader sexual dysfunction, rather than ED alone. Ghosh et al. [11] used the International Classification of Diseases (ICD-10) based assessment in a mixed PE/ED sample. These studies were, therefore, not treated as ED-equivalent in the main quantitative synthesis.

Regarding depressive symptom measurement, Saito et al. [3], Quang et al. [6], and Ghosh et al. [11] commonly employed the Patient Health Questionnaire-9 (PHQ-9), generally setting the threshold for suspected depression at a score of 10 or higher. In studies conducted in China, Xiao et al. [7] and Jiang et al. [8] adopted the Self-rating Depression Scale (SDS), considering an index score of 53 or above as indicative of clinically significant depression. Furthermore, Pitta et al. [9] applied a cut-off of 14 or higher using the Beck Depression Inventory-II (BDI-II), while Rajih et al. [1] utilized the PHQ-2 to enhance screening efficiency in a community setting, demonstrating the selection of tools optimized for specific research environments. Most studies designed their measurement time windows to reflect the “current” state or the “past 1–2 weeks”, focusing on elucidating the concurrent correlation between sexual dysfunction and depression. In this paper, “ED” is used for erectile dysfunction, “sexual dysfunction” is reserved for broader constructs when appropriate, and instrument abbreviations, such as PHQ-9, IIEF-5, SHIM, ASEX, SDS, and BDI-II, are used consistently. The detailed measurement characteristics and corresponding study references are summarized in Table 2 (Ref. [1, 3, 5, 6, 7, 8, 9, 11]).

3.3 Study quality

Using the Newcastle-Ottawa Scale (NOS) to assess the risk of bias across the eight included observational studies, we observed a quality spectrum ranging from “Low” to “Very Good”. This evaluation focused on three critical domains: Selection, Comparability, and Outcome/Exposure.

In the Selection domain, Pitta et al. [9] demonstrated the highest validity, securing strong scores for representativeness by leveraging a massive general population dataset of 21,139 individuals. Conversely, Sayed et al. [5] and Ghosh et al. [11] faced limitations regarding representativeness and potential selection bias due to their reliance on convenience sampling from small, hospital-based cohorts. While Rajih et al. [1] and Saito et al. [3] utilized community and web-based samples from Saudi Arabia and Japan, respectively, we evaluated them as retaining some risk of selection bias inherent to voluntary response protocols.

Within the Comparability domain, our assessment centered on whether studies employed multivariable regression to control for key confounders, such as age, socioeconomic status, and comorbidities. Saito et al. [3], Quang et al. [6], Xiao et al. [7], Jiang et al. [8], and Pitta et al. [9] achieved high comparability—earning two stars (★★)—by adequately adjusting for these confounding variables within their statistical models. However, Ghosh et al. [11] relied primarily on simple correlation or descriptive statistics; consequently, we determined that they failed to sufficiently rule out the influence of confounding variables when establishing an independent association between sexual dysfunction and depression.

Finally, in the Outcome/Exposure domain, most studies received favorable evaluations for utilizing validated self-report instruments, such as the IIEF-5, PHQ-9, and SDS. Jiang et al. [8] particularly secured objectivity by precisely measuring sexual function in patients prior to medication. In contrast, Ghosh et al. [11] recorded lower scores due to insufficient descriptions of their clinical diagnostic procedures and a lack of measurement standardization.

Synthesizing these assessments, Pitta et al. [9] emerged as the highest quality study (“Very Good”) with a score of 7. The studies by Saito et al. [3], Quang et al. [6], Xiao et al. [7], and Jiang et al. [8] were rated as “Good” with 6 points. Rajih et al. [1] and Sayed et al. [5] maintained a “Moderate” level with scores of 5 and 4, respectively, whereas Ghosh et al. [11] was classified as “Low” quality with a score of 3. We incorporated these quality gradings to appropriately weigh the evidence when deriving our final conclusions. The NOS domain scores and corresponding study references are summarized in Table 3 (Ref. [1, 3, 5, 6, 7, 8, 9, 11]).

3.4 Prevalence of depression in patients with ED

Three studies (Saito et al. [3], Nguyen Quang et al. [6], Xiao et al. [7]) reported the prevalence of depression within samples suffering from ED. Given the disparity in assessment tools, we prioritized a subgroup analysis based on the specific instrument used, adhering to the study protocol. Fig. 2 (Ref. [3, 6, 7]) illustrates the depression prevalence in the ED cohort and the results stratified by instrument.

Fig. 2.Depression prevalence among ED samples (subgrouped by instrument: PHQ-9 ≥10 vs. SDS ≥53). PHQ: Patient Health Questionnaire; SDS: Self-rating Depression Scale.

Pooling two studies [3, 6] that defined depression using a PHQ-9 cutoff of ≥10 via a logit-transformed random-effects model yielded a combined prevalence of 22.1% (95% CI: 7.6%–49.5%). Heterogeneity between these studies was substantial. Conversely, the single study employing the SDS with a cutoff of ≥53 [7] reported a significantly higher prevalence of 65.0% (95% CI: 60.7%–69.0%). While the aggregate prevalence across all three studies—regardless of the tool—was 34.7% (95% CI: 13.2%–64.9%), and between-study heterogeneity was extreme. Therefore, the overall pooled estimate should be interpreted with considerable clinical caution.

3.5 Prevalence of sexual dysfunction in male MDD patients

Two studies (Sayed et al. [5], Jiang et al. [8]) reported the prevalence of sexual dysfunction, defined by the Arizona Sexual Experience Scale (ASEX), among male patients with Major Depressive Disorder (MDD). Integrating these findings using a logit-based random-effects model resulted in a pooled prevalence of 54.0% (95% CI: 23.2%–82.0%), with substantial heterogeneity observed (I² = 95.9%). Fig. 3 (Ref. [5, 8]) presents the results of this meta-analysis.

Fig. 3.Sexual dysfunction prevalence among males with MDD (ASEX-defined; random-effects model). MDD: major depressive disorder; ASEX: Arizona Sexual Experience Scale; SD: Sexual Dysfunction.

3.6 Association between ED and depressive symptoms (adjusted OR)

One study (Pitta et al. [9]) presented the association between ED and depressive symptoms in a general male population using an adjusted odds ratio (aOR) to control for covariates. The analysis revealed a significant association between ED and depressive symptoms, with an aOR of 2.021 (95% CI: 1.797–2.272). This indicates that men with ED had approximately double the odds of depressive symptoms compared with those without ED in that study. Fig. 4 (Ref. [9]) displays these effect estimates. Because this estimate was derived from a single large study, rather than a pooled meta-analysis, it is interpreted in this paper as an important individual-study finding, rather than definitive meta-analytic evidence.

Fig. 4.Association between ED and depressive symptoms (adjusted odds ratio) in a population-based male sample.

3.7 Differences in erectile function scores by depression screening status

Rajih et al. [1] utilized regression coefficients to report the linear relationship between depression screening results and erectile function scores in a cohort of young married men. Multivariable regression analysis indicated that the group screening positive for depression had significantly lower IIEF scores (β = −12.65; 95% CI: −17.37 to −7.92) and lower SHIM scores (β = −4.63; 95% CI: −6.64 to −2.62) compared to the negative screening group. These findings demonstrate a significant negative association between depressive state and erectile function. Fig. 5 (Ref. [1]) illustrates these results.

Fig. 5.Adjusted regression coefficients for erectile-function scores (IIEF and SHIM) by depression screening status. IIEF: International Index of Erectile Function; SHIM: Sexual Health Inventory for Men.

3.8 Correlation between depression severity and sexual dysfunction

According to Sayed et al. [5], who analyzed the correlation between depression severity and the degree of sexual dysfunction in male MDD patients, the correlation coefficient between DASS-21 scores and ASEX total scores was r = 0.640 (95% CI: 0.474–0.762). This indicates a strong positive correlation, where the intensification of depressive symptoms parallels the worsening of sexual dysfunction. Fig. 6 (Ref. [5]) presents these correlation estimates.

Fig. 6.Correlation between depressive symptom severity tool (DASS-21) and sexual dysfunction severity (ASEX) in males with MDD.

3.9 Standardized mean difference (SMD) in depression severity by sexual dysfunction status

Jiang et al. [8] compared depression severity based on the presence of sexual dysfunction using Propensity Score Matching (PSM). Post-matching analysis revealed that the sexual dysfunction group had statistically significantly higher SDS scores than the control (non-dysfunction) group. The calculated Standardized Mean Difference (Hedges’ g) was 0.624 (95% CI: 0.403–0.845), representing a medium effect size according to Cohen’s criteria. Fig. 7 (Ref. [8]) details this result.

Fig. 7.Standardized mean difference (Hedges’ g) in depressive symptom severity (SDS) between sexual dysfunction vs. non-dysfunction groups (after PSM). PSM: Propensity Score Matching; SDS: Self-rating Depression Scale.

3.10 Other single-study prevalence estimates

Additionally, an observational study [11] focusing on samples related to premature ejaculation and erectile dysfunction reported a prevalence of “depressive symptoms (or depression accompanied by anxiety)” at 57.0% (95% CI: 47.2%–66.3%); this is shown in Fig. 8 (Ref. [11]). Separately, Rajih et al. [1] reported a positive response rate for depressive symptoms of 63.8% (95% CI: 56.8%–70.2%) using PHQ-2 screening in a specific cohort of young married men.

Fig. 8.Depression prevalence in a PE/ED-related observational sample (definition: depression or anxiety + depression). PE: premature ejaculation; ED: erectile dysfunction.

3.11 Evaluation of publication bias

To explore potential publication bias among the eight included studies, we used both visual inspection of the funnel plot (Fig. 9) and Egger’s regression test [18]. The funnel plot did not show a strongly directional visual pattern; however, Egger’s regression test was statistically significant (p < 0.001), raising the possibility of publication bias or small-study effects. Because only a small number of studies were available and the quantitative evidence base was heterogeneous in both study design and effect-size metric, these results were treated as exploratory, rather than confirmatory. Publication bias or small-study effects, therefore, cannot be excluded. Although methods such as trim-and-fill correction [19] or leave-one-out influence analysis can be informative, we did not emphasize them here because the evidence base was limited and the reported effect measures were not sufficiently homogeneous for robust comparative sensitivity testing. Future reviews based on larger and more methodologically consistent datasets should incorporate these additional approaches.

Fig. 9.Funnel plot for evaluating publication bias.