Cognitive–emotional associations between non-restorative sleep and erectile dysfunction: the role of rumination profiles and resilience

Abstract

Introduction

Erectile dysfunction (ED) is a prevalent male health condition that substantially correlates with impaired sexual function and reduced overall quality of life. Non-restorative sleep (NRS), a prevalent subjective sleep disturbance, has recently been identified as an under-recognized factor linked to ED. Nonetheless, the psychological factors involved in the association between NRS and ED remain poorly clarified. Rumination and psychological resilience are two promising psychological variables that may participate in this correlational pathway yet have not been empirically examined in the NRS-ED context. The present study aimed to explore the potential indirect correlational link between NRS and ED through rumination, and further evaluate whether resilience moderates this associative pathway.

Methods

Between February 2024 October 2025, we conducted a cross-sectional study of 880 men with ED recruited from the Andrology Clinic of the First Affiliated Hospital of Anhui Medical University. Participants completed the NRS Scale (lower scores indicate more severe NRS), Ruminative Responses Scale, 10-item Connor–Davidson Resilience Scale, and the 5-item International Index of Erectile Function (IIEF-5). Pearson correlations, bootstrapped mediation, and moderated mediation models with covariate adjustment, and latent profile analysis (LPA) in Mplus were performed to identify rumination subgroups and compare erectile function across profiles.

Results

Non-restorative sleep severity was significantly associated with worse ED and elevated rumination (all P < .001). Rumination partially mediated the association between NRS and ED. Exploratory LPA identified 3 statistical patterns representing sample heterogeneity in rumination severity (low 28.86%, moderate 43.18%, high 27.96%), rather than clinically distinct phenotypes. Higher rumination severity corresponded to significantly poorer erectile function (high vs. low severity IIEF-5 scores: 8.92 ± 3.84 vs. 14.95 ± 5.03). Resilience significantly moderated this pathway, acting as a buffer that attenuated both the effect of NRS on rumination and the subsequent impact of rumination on ED.

Conclusion

Our findings demonstrate that NRS is linked to ED via a ruminative associational pattern, and this association varies across levels of resilience. These findings suggest that integrated therapeutic strategies targeting these cognitive-emotional factors could facilitate more personalized clinical management for patients.

Introduction

Erectile dysfunction (ED) is a highly prevalent men’s health condition that extends beyond reproductive impairment to significantly compromise psychological well-being and quality of life. With epidemiological estimates suggesting a lifetime prevalence of up to 71%, ED represents a substantial public health burden.^1,2 While the physiological etiology—encompassing endocrine, vascular, and neurogenic dysregulation—is well characterized, emerging evidence necessitates a parallel focus on psychological factors and sleep disturbance.^3,4 Historically, the literature has prioritized quantitative sleep metrics (eg, duration) or distinct clinical syndromes like insomnia; yet, the subjective quality of sleep remains under-examined.^5,6 Of particular relevance is non-restorative sleep (NRS), a qualitative phenotype defined by the subjective sense of unrefreshing sleep despite adequate duration. Non-restorative sleep is distinctively linked to heightened psychological strain and incomplete homeostatic recovery, marking it as a clinically salient dimension of sleep health.^7,8 However, the mechanistic pathways bridging this subjective sleep deficit and erectile impairment remain underexplored. What is clear, however, is that subjective sleep deficits profoundly disrupt emotion regulation. Since psychological distress is a potent inhibitor of sexual arousal, the cross-sectional association between NRS and erectile function is likely reflected in cognitive–emotional tendencies. Prominent among these tendencies is rumination, which frequently emerges as a cognitive sequela associated with unrefreshing sleep.⁹ By co-occurring with prolonged psychological distress, this specific cognitive process may serve as a pivotal mechanistic link bridging the gap between subjective sleep deficits and erectile impairment.

Rumination is defined as a maladaptive cognitive style characterized by recursive, passive fixation on distress and its antecedents, typically devoid of active problem-solving. As a form of cognitive perseveration, this entrenched thought process undermines emotion regulation and impedes the attentional disengagement from negative stimuli.^10–12 Emerging evidence suggests that NRS is closely associated with such emotional and cognitive dysregulation. When the restorative function of sleep is compromised—even in the presence of adequate duration—the resulting deficits in cognitive control are strongly linked to a greater propensity for ruminative thinking. These maladaptive cognitive–emotional patterns, in turn, are concurrently associated with sustained physiological arousal and intensified stress reactivity, which track with altered hypothalamic–pituitary–adrenal (HPA) axis function and poorer vascular endothelial status—factors closely associated with ED.^13–15 Integrating these bio-behavioral correlates, we observe a correlated pattern: compromised emotional recovery coexists with persistent rumination, which further correlates with elevated physiological dysregulation and poorer erectile function. Translating this framework into a testable model, we hypothesize that rumination mediates the relationship between NRS and ED (Hypothesis 1).

Notwithstanding the robust association between disturbed sleep and adverse physiological sequelae, the clinical impact of sleep loss is far from uniform. The observation that some individuals remain clinically unaffected implies the operation of protective mechanisms that mitigate the downstream consequences of sleep loss. Psychological resilience—conceptualized not as a fixed trait but as a dynamic adaptive process—provides a mechanistic rationale for these divergent clinical outcomes.¹⁶ By facilitating the restoration of psychological equilibrium and physiological homeostasis under stress, resilience enhances emotion regulation and accelerates recovery. Individuals with robust resilience typically exhibit superior cognitive control, reduced engagement in ruminative thinking, and attenuated HPA axis activation—a constellation of features that may shield against the stress-related dysregulation implicated in ED.^17–20 From this vantage point, resilience may serve to mitigate the disruptive interference of NRS on erectile function. Specifically, lower rumination levels among resilient individuals correspond to a weaker link between sleep disturbances and poorer erectile function. Accordingly, we hypothesized that resilience functions as a critical boundary condition, moderating the rumination-mediated association between NRS and ED, such that the indirect effect is attenuated among individuals with higher adaptive capacity (Hypothesis 2).

Despite the growing recognition of the psychological links between sleep and sexual health, current research often relies on analytic strategies that implicitly assume homogeneous responses to sleep disturbance. This assumption contradicts clinical reality, where individuals exhibit substantial variability in stress appraisal and emotion regulation when confronted with NRS. Ignoring this heterogeneity risks masking meaningful biological and behavioral nuances behind population means. Addressing this gap, we adopted a comprehensive methodological framework combining mediation analysis, latent profile analysis (LPA), and moderated mediation. Our specific aims were to: (1) test whether rumination mediates the link between NRS and ED; (2) identify exploratory statistical subgroups of rumination based on LPA; and (3) assess resilience as a protective buffer that attenuates the indirect impact of sleep problems. To visually consolidate this analytic framework and the corresponding hypotheses, we present the proposed conceptual model in Figure 1. By synthesizing variable-centered and person-centered inquiries, this study explicitly accounts for the non-uniform experience of sleep loss, offering a nuanced perspective on the interplay between sleep, cognitive patterns, and sexual function. Accordingly, does NRS relate to ED through heterogeneous rumination subtypes, and does psychological resilience weaken such indirect association?

Figure 1.

Hypothesized moderated mediation model.

Materials and methods

Subjects and study procedure

This cross-sectional study was performed at the Andrology Clinic of the First Affiliated Hospital of Anhui Medical University Hospital from February 2024 to October 2025. To ensure diagnostic precision and methodological rigor, all evaluations were supervised by a multidisciplinary team of board-certified urologists and clinical psychologists.

A total of 1256 men presenting with self-reported ED were screened. Initial assessment included the 5-item International Index of Erectile Function (IIEF-5). Final eligibility was adjudicated based on a comprehensive clinical workup—including medical and sexual history, physical examination, and laboratory panels—against prespecified criteria. Inclusion criteria were: (1) IIEF-5 score ≤21; (2) age ≥18 years within a stable heterosexual partnership; and (3) regular sexual activity. Exclusion criteria encompassed: (1) recent use of pharmacotherapies affecting erectile function (eg, antidepressants, anxiolytics); (2) substance or alcohol dependence; and (3) significant organic comorbidities, including diabetes, hypertension, cardiovascular disease, spinal cord injury, or other clinically severe conditions.

Ultimately, 880 men satisfied all stringent criteria and constituted the final analytic cohort. Participants completed self-administered instruments covering socio-demographics and key psychological variables in a private clinical setting to ensure confidentiality and mitigate social desirability bias. The study workflow is depicted in Figure 2. Written informed consent was obtained from all participants. The protocol adhered to the Declaration of Helsinki and received approval from the Ethics Committee of the First Affiliated Hospital of Anhui Medical University (approval No. Quick-PJ 2023-09-45), and written informed consent was obtained from all participants before enrollment.

Figure 2.

Study flowchart and participant screening. Abbreviations: ED = erectile dysfunction; IIEF-5 = the five-item International Index of Erectile Function; NPTR = nocturnal penile tumescence and rigidity.

Measures

Socio-demographic information

To rigorously control for confounding, we adjusted for a comprehensive set of covariates selected a priori. These were categorized into anthropometric and clinical metrics (age, body mass index [BMI]), relational and sexual history (relationship duration, coital frequency over the preceding 4 weeks), and sociodemographic/lifestyle profiles (educational attainment, occupational status, residence, smoking, and alcohol consumption). These variables were integrated into all multivariable models.

Erectile function

Erectile function was quantified using the 5-item International Index of Erectile Function (IIEF-5), a widely validated instrument with robust psychometric properties.²¹ This scale comprises 5 items rated on a 5-point Likert basis, yielding a composite score ranging from 5 to 25, where lower values reflect greater impairment. In accordance with established diagnostic thresholds, ED was operationalized as a total score ≤ 21.

Non-restorative sleep

Subjective sleep quality was quantified using the validated Chinese version of the non-restorative sleep scale (NRSS).²² This 12-item inventory encompasses domains of perceived physical and mental restoration as well as daytime functioning. Items are rated on a 5-point Likert scale, generating a total score range of 12-60, where lower values signify more severe NRS symptoms. The NRSS is recognized for its robust psychometric properties across clinical and community settings and exhibited high internal consistency in the current cohort (Cronbach’s α = 0.85).

Rumination

Rumination was quantified using the Chinese version of the Ruminative Responses Scale.²³ Designed to index the frequency of repetitive, passive self-focused thinking regarding distress and its antecedents, this 22-item inventory utilizes a 4-point Likert scale (1 = “almost never” to 4 = “almost always”). Total scores range from 22 to 88, with higher values indicating a more pronounced tendency toward rumination. The scale is noted for its sound psychometric properties and exhibited good internal consistency in the current study (Cronbach’s α = 0.82).

Resilience

Psychological resilience was quantified using the validated Chinese version of the 10-item Connor–Davidson Resilience Scale (CD-RISC-10).²⁴ This instrument gauges an individual’s capacity for stress adaptation and recovery from adversity. Items are rated on a 5-point Likert scale (0-4), generating a cumulative score ranging from 0 to 40, where higher values denote superior adaptive coping. The CD-RISC-10 is distinguished by its robust psychometric validity across diverse clinical and cultural settings and demonstrated excellent reliability in the current cohort (Cronbach’s α = 0.88).

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics 23.0 (IBM Corporation, Armonk, New York, USA) and Mplus 8.3 (Muthén & Muthén, Los Angeles, California, USA). Following the computation of descriptive statistics and Pearson correlations to establish bivariate associations, we employed the PROCESS macro (Model 4) to test for mediation. Indirect effects were assessed using 5000 bias-corrected bootstrap resamples, with significance inferred from 95% confidence intervals (CIs) excluding zero. All regression models were adjusted for a prespecified set of covariates: age, BMI, health behaviors, and comorbidities. To delineate unobserved heterogeneity in ruminative processing, we conducted LPA in Mplus. The optimal class solution was adjudicated based on a synthesis of model fit indices—Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), sample-size adjusted BIC (aBIC), entropy, the Lo–Mendell–Rubin likelihood ratio test, and the bootstrap likelihood ratio test (BLRT)—alongside classification precision and theoretical interpretability. Bayes factor analyses for pairwise comparisons among latent profiles were conducted in JASP 0.19.3 (University of Amsterdam, Amsterdam, The Netherlands). Finally, we interrogated the conditional nature of these pathways using PROCESS Model 49, testing whether resilience moderates the indirect trajectory from NRS to ED. Continuous predictors were mean-centered to mitigate multicollinearity, with significance defined as a two-sided P < .05.

Results

Characteristics of participants

Supplementary Table S1 outlines the baseline socio-demographic and clinical characteristics of the analytic cohort (N = 880). The study population was characterized by a mean age of 36.89 ± 7.72 years and a mean BMI of 23.35 ± 2.33 kg/m². Comprehensive distributions of lifestyle factors and socio-demographic strata—including educational attainment, occupational status, and place of residence—are detailed in the Supplementary Table S1.

Bivariate associations among key study variables

Bivariate correlations among the study variables are summarized in Table 1. As anticipated, erectile function (IIEF-5; where lower scores indicate more severe ED) was significantly associated with each psychometric construct. In particular, NRSS scores—where lower scores indicate more severe NRS—showed strong positive correlations with both erectile function (r = 0.63, P < .001) and resilience (r = 0.57, P < .001), while displaying a significant negative association with rumination (r = -0.54, P < .001). Collectively, these interrelations substantiate the hypothesized bio-behavioral links, providing an empirical foundation for the subsequent multivariable mediation analyses.

Table 1.

Means, SDs and correlations among variables.

	Mean	SD	1. NRSS	2. Rumination	3. Resilience	4. IIEF-5
1. NRSS	37.07	14.54	1
2. Rumination	51.84	12.94	−0.54***	1
3. Resilience	17.98	10.52	0.57***	−0.49***	1
4. IIEF-5	11.88	4.58	0.63***	−0.50***	0.52***	1

Abbreviations: IIEF-5 = the five-item International Index of Erectile Function; NRSS = total score of non-restorative sleep scale; SD = standard deviation.

^*** P<0.001.

Mediation effect of rumination between NRS and ED

To explore the associational pattern connecting NRS to ED, we tested a covariate-adjusted mediation model (Table 2). Initially, we found that NRS (lower NRSS scores) was strongly correlated with increased rumination (Model 1: β = −.54, t = −19.07, P < .001). While NRS exhibited a strong total association with erectile function (Model 3: β = .63, t = 24.06, P < .001), the inclusion of rumination in the full model (Model 2) revealed a significant inverse link between rumination and IIEF-5 scores (β = −.22, t = −7.37, P < .001). Crucially, the association between NRS and erectile function remained significant but was attenuated in magnitude (β = .51, t = 16.85, P < .001). These findings point to partial mediation, supporting Hypothesis 1 that rumination represents a prominent psychological pathway linking NRS and ED within this cross-sectional framework.

Table 2.

Mediation analysis of rumination between NRS and ED.

Predictors	Model 1 (Rumination)		Model 2 (IIEF-5)		Model 3 (IIEF-5)
	β	t	β	t	β	t
NRSS	−.54	−19.07***	.51	16.85***	.63	24.06***
Rumination			−0.22	−7.37***
R  2	0.29		0.43		0.40
F	363.53***		334.24***		579.07***

Abbreviations: ED = erectile dysfunction; NRSS = total score of non-restorative sleep scale; R² = coefficient of determination. ^***P < .001.

Latent rumination profiles identified through LPA

To further characterize the heterogeneity of rumination based on its mediating role in the NRS–ED association, we adopted LPA as an exploratory statistical classification method. We evaluated models ranging from 1 to 6 profiles (Table 3). Although AIC, BIC, and aBIC decreased progressively with added complexity, the BLRT favored a three-profile structure. Of note, extending the model to 4 profiles yielded no statistically significant gain in fit (P = .47), a plateau that persisted for 5- (P = .56) and 6-profile (P = .62) solutions. Given its high classification precision (entropy = 0.992) and good interpretability, the three-profile model was selected. This analysis divided the total sample into three statistical subgroups according to rumination severity: low (n = 254, 28.86%), moderate (n = 380, 43.18%), and high (n = 246, 27.96%) rumination groups. These subgroups only differ quantitatively in rumination severity, without distinct qualitative differences or clinically independent subtypes. Item-level patterns are depicted in Supplementary Figure S1.

Table 3.

LPA of rumination.

Indices	Unconditional model
	1 profile	2 profile	3 profile	4 profile	5 profile	6 profile
Fit statistics
LL	−24612.84	−20267.97	−19754.78	−19577.25	−19450.60	−19354.15
AIC	49 313.67	40 669.94	39 689.56	39 380.50	39 173.20	39 026.31
BIC	49 523.99	40 990.19	40 119.75	39 920.63	39 823.27	39 786.31
aBIC	49 384.25	40 777.41	39 833.93	39 561.76	39 391.36	39 281.36
Entropy		0.962	0.992	0.94	0.94	0.95
BLRT		<0.001	<0.001	0.47	0.56	0.62
LMR		<0.001	<0.001	<0.001	<0.001	<0.001
Group sizes, n (%)
C1	880 (100%)	322 (36.59%)	254 (28.86%)	203 (23.06%)	160 (18.18%)	106 (12.05%)
C2	–	558 (63.41%)	380 (43.18%)	120 (13.64%)	399 (45.34%)	74 (8.41%)
C3	–	–	246 (27.96%)	392 (44.55%)	79 (8.98%)	399 (45.34%)
C4	–	–		165 (18.75%)	118 (13.41%)	87 (9.89%)
C5	–	–			124 (14.09%)	122 (13.86%)
C6	–	–				92 (10.45%)

Bold figures highlight the selected class solution. Abbreviations: AIC = Akaike Information Criterion; BIC=Bayesian Information Criterion; aBIC = adjusted BIC; BLRT = bootstrapped likelihood ratio test; LL = Log-likelihood; LMR = Lo–Mendell–Rubin likelihood ratio test; LPA = latent profile analysis.

LPA-based differences in ED

Following profile identification, we assessed differences in erectile function (IIEF-5; where lower scores indicate more severe ED) across the subgroups. A one-way analysis of variance (ANOVA) revealed a pronounced main effect of profile membership (F = 143.40, P < .001), which accounted for 25% of the total variance in IIEF-5 scores (R² = 0.25). As detailed in Table 4, we observed a gradual changing trend in erectile function across subgroups: IIEF-5 scores decreased gradually as rumination severity increased. Participants in the low-rumination subgroup exhibited the highest scores (M = 14.95 ± 5.03), which were significantly higher than those of the moderate (M = 11.74 ± 3.21) and high (M = 8.92 ± 3.84) subgroups. Bayesian post hoc analysis substantiated these distinctions, yielding decisive evidence for all pairwise contrasts (BF₁₀ ranging from 1.08 × 10¹⁸ to 6.74 × 10³⁸) with negligible numerical error. The stability of these findings was confirmed via robustness checks using alternative Cauchy prior widths and sequential evidence plots (Supplementary Figure S2), underscoring a robust inverse graded association between ruminative tendency and erectile function. The stability of these findings was confirmed via robustness checks using alternative Cauchy prior widths and sequential evidence plots (Supplementary Figure S2), underscoring a robust inverse graded association between ruminative tendency and erectile function.

Table 4.

LPA-based differences in ED.

LPA-based differences in IIEF-5 score
Subgroups	N	M ± SD	R  2	F	P
Class 1	254	14.95 ± 5.03	0.25	143.40	<.001
Class 2	380	11.74 ± 3.21
Class 3	246	8.92 ± 3.84
Bayesian post-hoc comparisons
Groups			BF10	Error%
Low rumination vs medium rumination			1.08 × 10+18	1.29 × 10−24
Low rumination vs high rumination			6.74 × 10+38	1.80 × 10−41
Medium rumination vs high rumination			3.38 × 10+18	3.94 × 10−25

Abbreviations: BF10 = Bayes factor; ED = erectile dysfunction; LPA = Latent profile analysis; SD = standard deviation.

The moderated mediation analysis combining rumination and resilience

We applied a moderated mediation framework to examine whether resilience correlates with changes in the associational patterns linking NRS, rumination and erectile function. As detailed in Table 5, resilience significantly modulated the link between NRSS (where lower scores reflect greater impairment) and rumination (β = −.13, t = −3.91, P < .001, 95% CI [−0.19 to −0.06]). Simple slope analyses (Supplementary Figure S3A) revealed that this inverse association was more pronounced at higher levels of resilience (b = −0.57, t = −9.94, P < .001 compared to lower levels (b = −0.32, t = −8.56, P < .001). Crucially, resilience correlates with a weaker associational trend between rumination and erectile function. A significant rumination × resilience interaction (β = .14, t = 5.46, P < .001, 95% CI [0.09-0.19]) demonstrated that the association between rumination and poorer IIEF-5 scores was weaker among highly resilient individuals. Consistent with this interaction, the simple slope plots (Supplementary Figure S3B) showed that rumination was more strongly associated with lower IIEF-5 scores among men with low resilience (b = −0.33, t = −8.41, P < .001, 95% CI [−0.40 to −0.25]) than among those with high resilience (b = −0.05, t = −1.36, P < .001, 95% CI [−0.13 to −0.01]). Conversely, the direct NRSS × resilience interaction was non-significant (β = .03, t = 1.43, 95% CI [−0.15 to 0.08]). These findings confirm that the indirect associational pattern among NRS, rumination and ED varies across resilience levels, validating Hypothesis 2.

Table 5.

The moderated mediation analysis combining rumination and resilience.

Variables	Model 1 (Rumination)				Model 2 (IIEF-5 Score)
	β	t	LLCI	ULCI	β	t	LLCI	ULCI
NRSS	−0.45	−12.31***	−0.52	−0.37	0.38	10.78***	0.31	0.45
Resilience	−0.20	−5.25***	−0.27	−0.12	0.27	7.96***	0.21	0.34
NRSS × Resilience	−0.13	−3.91***	−0.19	−0.06	0.03	1.43	−0.15	0.08
Rumination					−0.19	−6.34***	−0.25	−0.13
Rumination × Resilience					0.14	5.46***	0.09	0.19
R2	0.35				0.49
F	160.00***				166.31***

All variables were standardized before analyzed. Abbreviation: NRSS = Total score of Non-restorative Sleep Scale. ^***P<0.001.

Discussion

By synthesizing NRS, rumination and psychological resilience into a unified analytical framework, this study describes the cognitive–emotional associational patterns between disturbed sleep and ED. Leveraging a multi-methodological approach—combining mediation, LPA and moderated mediation—we provide convergent evidence that sleep-related affective processes are jointly associated with erectile function. Our results confirm that NRS shows direct and indirect associations with ED via rumination. Crucially, this indirect associational pattern differs across resilience levels: individuals with higher adaptive capacity tend to show a weaker association between rumination and erectile impairment. Furthermore, LPA identified three statistical subgroups based on rumination severity, which corresponded to a gradual changing trend of erectile function. These subgroups reflect sample heterogeneity in ruminative thinking, rather than clinically distinct groups or inherent cognitive vulnerability phenotypes. These findings challenge the traditional focus on quantitative indices (eg, total sleep duration) by highlighting the psychological salience of perceived restoration. Specifically, the subjective sense of unrefreshing sleep—captured here by the NRSS—is intrinsically tied to emotion regulation, which in turn correlates with sexual health. By situating these factors within a coherent biopsychosocial model, our work offers a robust empirical basis for understanding ED beyond purely physiological confines. The biopsychosocial model advocates that health and illness are determined by the interplay of biological, psychological, and social factors. This study closely aligns with the core philosophy of the biopsychosocial model: From the biological perspective, we focused on sleep status, neuroendocrine activity, and penile physiological function related to ED. From the psychological perspective, we explored the roles of rumination and psychological resilience. From the social and behavioral perspective, we took demographic characteristics, lifestyle and partnership-related factors into account as confounding variables in all statistical analyses. This work further enriches the application of the biopsychosocial model (BPS) model in the field of male sexual health research, while simultaneously reaffirming the importance of sleep restoration itself.

Dissecting this model reveals that the robust direct association between NRS and ED, which merits further attention. Even after adjustment for demographic and lifestyle confounders, NRS remained significantly associated with poorer erectile function. This aligns with accumulating evidence that sleep quality—a qualitative dimension distinct from duration—is a clinically consequential determinant of men’s sexual health.^25,26 While prior literature has predominantly focused on objective metrics such as total sleep time, apnea, or insomnia symptoms, our findings position NRS as a unique and meaningful construct with dual relevance to the physiological and psychological substrates of sexual functioning. Specifically, at the neuroendocrine level, perceived non-restoration likely indexes disrupted slow-wave sleep and incomplete homeostatic recovery. Such disruptions are known to potentiate HPA axis activity, blunt nocturnal testosterone pulsatility, and elevate sympathetic tone—a triad of neuroendocrine and vascular dysregulations that compromise erectile capacity.^27–29 Concurrently, the recurring burden of unrefreshed awakening may exacerbate daytime fatigue and erode self-efficacy, thereby dampening sexual desire and depleting the cognitive–affective reserves necessary for arousal. From a translational standpoint, the convergence of these dysregulations positions elevated NRS burden as a key clinical correlate of ED. Consequently, therapeutic avenues targeting perceived restoration—whether via cognitive–behavioral or physiological interventions—could confer downstream benefits for sexual health. This suggests that while physiological perturbations are critical, the NRS–ED nexus is unlikely to be driven solely by these direct biological mechanisms.

Beyond these physiological correlates, our mediation analysis identifies rumination as a central intermediate associational pathway between NRS and ED. Individuals reporting higher NRS also tend to show lower executive control and higher levels of perseverative cognition. Higher rumination correlates with increased self-monitoring, negative mood, and psychogenic inhibition, which are linked to poorer erectile function. By linking these processes—traditionally studied in affective disorders such as depression and anxiety—to sexual health, we show that rumination characterizes a key intermediate association between NRS and erectile function.³⁰ Moreover, rumination likely operates through tangible somatic channels. Chronic ruminative thinking is associated with systemic dysregulation, including HPA-axis disturbance, cortisol elevation, and sympathetic overactivation, these physiological changes co-occur with impaired endothelial and neurovascular integrity.^31–33 Taken together, these observations indicate that rumination, psychological states, and physiological indicators show parallel correlational patterns alongside the NRS–erectile function association. From a clinical standpoint, these findings suggest that maladaptive rumination may serve as a valuable therapeutic target. Therapeutic strategies designed to disrupt these perseverative loops may offer downstream benefits for erectile function by mitigating both performance anxiety and the physiological arousal that constrains vascular response.

However, treating rumination as a uniform construct obscures critical inter-individual variability. We adopted LPA to explore heterogeneous statistical patterns of rumination within the sample. This approach identified three discrete subgroups with high classification precision (entropy = 0.992). Comparisons across subgroups showed that erectile function gradually declined as rumination severity increased. This graded trend supports a dimensional view of emotion-related cognition, indicating that rumination levels correlate with ED severity, rather than representing fixed cognitive vulnerabilities. Clinically, these statistical groupings only reflect differences in rumination severity among participants, and cannot be defined as high-risk clinical subtypes. Men in the high-rumination subgroup tended to report poorer emotion regulation, greater sexual distress, and more severe ED symptoms. By moving beyond aggregate scores to capture this meaningful heterogeneity, our analysis clarifies which patients carry the heaviest cognitive–emotional burden. Consequently, these findings validate the utility of routine screening, providing a rationale for precision medicine strategies where psychological interventions are tailored to the specific vulnerability profile of the patient.

To inform such targeted interventions, we sought to identify modifiable factors capable of buffering this risk. Specifically, we investigated whether resilience correlates with a weaker associational trend across this set of interrelated variables. Our moderated mediation modeling confirmed that resilience is associated with altered correlational strength at two key points: the association between NRS and rumination, and the association between rumination and erectile function. Functionally, individuals with higher resilience tend to use more adaptive coping strategies, and this pattern coincides with lower negative self-focus and a weaker link between sleep distress and altered physiological arousal. Aligning with robust evidence across a spectrum of stress-related conditions and affective disturbances,^34,35 resilience shows important correlations with sexual health in the present sample. Accordingly, integrating resilience-building protocols (eg, mindfulness-based counseling) into ED management could yield synergistic benefits, underscoring the necessity of a holistic approach that addresses not only the physiological contributors to ED but also the cognitive–emotional assets that foster positive adaptation.

Synthesizing these multi-layered findings, our study describes the cognitive–emotional associational network between NRS and ED. Specifically, rumination represents a central intermediate variable, while resilience correlates with variations in the strength of these associations. These findings advocate for a comprehensive management paradigm that jointly evaluates subjective sleep restoration, ruminative tendencies, and adaptive capacity to personalize ED care. We acknowledge, however, several limitations inherent to the study design. First, the cross-sectional nature of the data restricts our ability to infer causality or temporal dynamics. Second, as the sample was comprised exclusively of men with ED, findings may not extrapolate to non-clinical populations or those with subclinical symptoms. Third, the exclusive use of self-report instruments may be subject to shared method variance. Fourth, although we adjusted for age, BMI, lifestyle factors, and several comorbidities in the statistical models, this study did not collect and control for multiple well-documented confounding variables that may influence the observed associations. These important unmeasured confounders include depressive and anxiety symptoms, relationship quality, and dyadic functioning, the duration and clinical history of ED, as well as prior use of PDE5 inhibitors and other ED-related treatments. All these factors have been proven to affect subjective sleep perception, ruminative thinking, and erectile function, which may influence the interpretation of our results. To advance this field, future research mandates longitudinal designs and the integration of objective physiological indices—specifically electroencephalography metrics and heart rate variability—to further explore the neurobiological correlates of the sleep–ED association. Conclusively, by identifying rumination as a modifiable vector and resilience as a protective shield, this study redefines the therapeutic horizon, demonstrating that the remission of ED requires silencing the cognitive noise of sleep loss as much as restoring vascular flow.

Conclusions

In conclusion, this study identifies rumination as a key intermediate variable in the associational network between NRS and ED. We demonstrated notable heterogeneity in rumination across the sample. Latent profile analysis revealed heterogeneous rumination levels across the sample. The 3 statistical subgroups derived from LPA differ only in the severity of rumination, and correspond to varying degrees of erectile impairment. There are no essential qualitative differences among these subgroups. Crucially, resilience correlates with weaker associations between sleep distress, rumination and sexual function. These findings refine the biopsychosocial model of ED, indicating that perceived sleep restoration, ruminative thinking, and resilience are important factors to consider in intervention design. While longitudinal validation is required to establish causality and temporal order, our results provide a robust rationale for integrating these psychological dimensions into more individualized ED management strategies.

Author contributions

P.G.: Conceptualization, Writing—original draft, Project administration, Funding acquisition, Writing—review & editing, Validation. Y.J.: Conceptualization, Writing—review & editing, Validation. X.Z.: Conceptualization, Project administration, Funding acquisition, Writing—review & editing, Validation. Z.C.: Data curation, Writing—review & editing, Validation. F.W.: Data curation, Writing—review & editing, Validation. Y.G.: Data curation, Writing—review & editing, Validation. J.D.: Data curation, Writing—review & editing, Validation. Y.M.: Formal analysis, Writing—review & editing, Validation. P.Y.: Formal analysis, Writing—review & editing, Validation. Y.L.: Formal analysis, Writing—review & editing, Validation. H.S.: Writing—original draft, Writing—review & editing, Validation.

Pan Gao, Yao Jiang, Zhi Cao, and Fan Wu contributed equally.

Funding

This study was founded by the University Natural Science Research Project of Anhui Province (2023AH053308) and the National Natural Science Foundation of China (82401890).

Conflicts of interest

The authors declare no potential conflicts of interest.

Data availability

The data used during the current study are available from the corresponding author on reasonable request.

Consent for publication

Not applicable.

Ethics approval and consent to participate

This study was reviewed and approved by the Ethics Committee of the First Affiliated Hospital of Anhui Medical University (approval No. Quick-PJ 2023-09-45).