Psychological distress, decision-making, and health behaviors among prostate cancer survivors with urinary incontinence: A scale- and symptom-level network analysis Authors: Yan He, Zhiguang Huang, Jiayu Gong, Zunguang Bai, Wenfeng He, Liling Zhang Published: 2026 ABSTRACT AbstractObjectiveThis study aimed to investigate the interrelations among depressive symptoms, anxiety symptoms, decision-related factors, and health behaviors in prostate cancer survivors with urinary incontinence using a network analysis approach.MethodsA cross-sectional study was conducted among 251 prostate cancer survivors experiencing urinary incontinence at a tertiary hospital in China between October 2025 and February 2026. Participants completed self-administered electronic questionnaires assessing depression, anxiety, shared decision making, decisional conflict, loneliness, urinary incontinence, and exercise adherence. Scale-level and symptom-level networks were estimated using Gaussian graphical models with graphical LASSO regularization and EBIC model selection. Centrality, predictability, and network stability were examined using bootstrap procedures.ResultsAt the scale level, anxiety and depression emerged as the most central nodes, forming a tightly connected psychological cluster with loneliness. Shared decision making was strongly associated with decisional conflict (partial r = 0.68), while urinary incontinence was moderately associated with exercise adherence (partial r = 0.43) and weakly associated with anxiety and depression. At the symptom level, suicidal ideation and low energy were the most central depressive symptoms, while difficulty controlling worry occupied a central position in the anxiety network.ConclusionsDepression and anxiety occupy central positions within integrated psychosocial and clinical networks among prostate cancer survivors with urinary incontinence. Identifying these core nodes underscores their importance for clinical monitoring and comprehensive symptom assessment. Longitudinal studies are needed to determine whether these central symptoms predict or influence broader psychological distress, treatment-related symptoms, and health behaviors.Keywords: Psychological distress, Shared decision making, Prostate cancer, Network analysis Introduction Prostate cancer has become a chronic disease among a growing population of survivors, with an estimated 1.4 to 1.5 million new cases diagnosed annually.1, 2, 3 Along with continuous improvements in early detection and treatment outcomes, the focus of survivor care has shifted from solely reducing mortality rates to managing long-term, interconnected health-related outcomes,3 including mental health, functional impairment, and health-related behaviors. Approximately 17% of prostate cancer survivors experience significant depression and anxiety symptoms, with 10% reporting recent suicidal ideation, which is higher than that of the general male population.4 Psychological issues frequently coexist with functional impairments, and the majority of survivors experience varying degrees of postoperative urinary incontinence.5,6 Health-related behaviors are increasingly recognized as modifiable factors, and regular physical activity has been demonstrated to improve functional outcomes and alleviate emotional distress in cancer survivors.7 Decision-making represents another critical dimension in prostate cancer treatment. Patients frequently confront complex therapeutic choices, requiring trade-offs between tumor control and quality of life.8 Shared decision-making, promoted as a patient-centered model, aims to enhance engagement and satisfaction; however, its relationship with decision conflict and psychological distress remains controversial.9 Network analysis conceptualizes symptoms and psychosocial factors as interacting components within a dynamic system, allowing identification of central and influential elements. Sharpley et al.10, 11, 12 identified uncontrollable worry and loss of pleasure as central symptoms within anxiety and depression networks. Shin et al.13 examined networks of depression, anxiety, perceived stress, and social support among South Korean adults, highlighting key central and bridging symptoms. However, research integrating anxiety and depression with disease-related symptoms and health behaviors remains limited among prostate cancer survivors. This study applied network analysis to examine the interrelationships among depression, anxiety, shared decision-making, decisional conflict, urinary incontinence symptoms, and exercise adherence in prostate cancer survivors following radical prostatectomy. To capture both broad psychosocial dynamics and fine-grained symptom-level patterns, we employed a two-tiered network analysis framework. Here, multi-level refers to two levels of analytic granularity, namely a scale-level network with total scores as nodes and symptom-level networks with individual items as nodes, rather than to hierarchical or multilevel statistical modeling. Specifically, one scale-level network was constructed to map interrelationships among seven clinical and behavioral constructs, and two separate symptom-level networks were estimated for depression and anxiety to identify central symptoms within each domain. By identifying core nodes, cross-domain associations, and network stability across both levels, this study aims to clarify pathways linking psychological distress, clinical symptoms, and health behaviors, thereby informing more comprehensive risk assessment strategies and helping prioritize specific symptom clusters for investigation in future longitudinal studies. Methods Study design and participants This study employed a cross-sectional design. Participants were recruited from the Department of Urology at the Guangdong Provincial Hospital of Traditional Chinese Medicine between October 2025 and February 2026. We utilized a purposive sampling strategy, and eligible patients were identified through the hospital's electronic health record system during their routine outpatient or inpatient visits. The inclusion criteria focused on patients with a confirmed diagnosis of prostate cancer to capture a broad spectrum of the disease's psychosocial impact. Eligible participants were those who: (1) were aged 18 years or older; (2) were able to communicate in Mandarin; (3) had a histologically confirmed diagnosis of prostate cancer; (4) had undergone radical prostatectomy; and (5) experienced urinary incontinence, defined as involuntary urine leakage occurring at least once per week during the preceding three months. The exclusion criteria were: diagnosis of cognitive impairment or other severe diseases precluding reliable participation or informed consent. A standardized recruitment and survey-administration protocol was implemented. Trained research assistants provided a comprehensive oral briefing of the study goals and explained the content and individual items of the questionnaire to each participant. Participants then completed the questionnaire themselves via an electronic self-report platform; research assistants remained available to clarify the meaning of items upon request but did not read items aloud sequentially or enter responses on the participants' behalf. All participants provided informed consent prior to data collection. Patients were informed of their right to withdraw at any stage without consequence to their clinical care. To ensure data integrity, all completed questionnaires were screened for response inconsistency prior to analysis using pre-specified criteria: (1) completed the entire survey in less than 4 minutes; (2) provided invariant responses across all items within any single scale, such as selecting the identical response option for every item; or (3) gave internally contradictory responses on items measuring the same construct in opposing directions. Participants meeting any one of these criteria were excluded. The 4-min threshold was derived as follows. Because participants read and completed the questionnaire themselves, total completion time directly reflected each individual's own reading and responding pace; applying a silent-reading-speed benchmark was therefore appropriate. The full questionnaire comprised approximately 1300 words across all instruments, instructions, and demographic items. Assuming an average silent reading speed of 250 words per minute for adult Chinese readers,14 the minimum plausible reading time alone was approximately 5.2 minutes. The 240-s cutoff was therefore set as a conservative lower bound, flagging only completion times at which participants could not plausibly have read the questionnaire content, regardless of response speed. This threshold was intentionally set below the estimated reading time to minimize false-positive exclusions of genuinely fast but attentive responders. Measures Psychological symptoms Depressive and anxiety symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9)15 and the Generalized Anxiety Disorder-7 (GAD-7),16 respectively. Both instruments have demonstrated strong psychometric properties and are widely used in clinical and research settings.16,17 Participants rated the frequency of each symptom over the preceding two weeks using a four-point Likert scale ranging from 0 (“not at all”) to 3 (“nearly every day”). Total scores ranged from 0 to 27 for the PHQ-9 and from 0 to 21 for the GAD-7, with higher scores indicating greater symptom severity. Loneliness was measured using the eight-item short-form of the UCLA Loneliness Scale (ULS-8).18 Participants rated eight items, including “I lack companionship” and “I feel left out”, on a four-point Likert scale ranging from one for never to four for always. Total scores were calculated by summing the responses. The total score ranges from 8 to 32, where higher scores indicate greater perceived social isolation. Decision-related measures Shared decision making was assessed using the Shared Decision Making Questionnaire (SDM),19 and decisional conflict was measured using the Decisional Conflict Scale (DCS).20 Higher scores on the SDM reflect greater perceived involvement in decision making, whereas higher DCS scores indicate greater uncertainty and difficulty in treatment-related decision making. Functional and behavioral measures Urinary incontinence was assessed using the International Consultation on Incontinence Questionnaire (ICIQ),21 with higher scores indicating more severe urinary symptoms. Exercise adherence was assessed using a self-reported, guideline-based checklist developed according to international physical activity guidelines proposed by the World Health Organization22 and the American College of Sports Medicine.23 Participants were asked whether they met four predefined criteria: (1) engaging in physical exercise at least three times per week; (2) exercising for at least 30 minutes per session; (3) performing at least moderate-intensity physical activity; and (4) maintaining regular exercise for at least six months. Each item was rated dichotomously (yes/no), with higher total scores indicating greater adherence to recommended exercise behaviors. Statistical analysis Descriptive statistics and data preprocessing Continuous variables were summarized as means ± standard deviations (SD), whereas categorical variables were presented as frequencies and percentages. Prior to analysis, the data were examined for potential outliers. Outliers in age were identified using the interquartile range (IQR) method (values below Q1 − 1.5 × IQR or above Q3 + 1.5 × IQR). Identified outliers were treated as missing values. A small number of age values (< 5%) were identified as outliers. To minimize data loss, missing or outlying age values were imputed using the sample median. No significant outliers were detected in other study variables. All node variables were examined for distributional properties, including skewness and kurtosis. Mild positive skewness was observed across several variables, which is commonly observed for symptom-level data. No extreme skewness or kurtosis was detected, and therefore no data transformation was applied. Network estimation All analyses were conducted using R (version 4.5.0). To avoid confusion with nested or hierarchical statistical modeling, our network architecture is operationally split into separate, independent analytical tiers. Two distinct types of network structures were estimated: (1) a scale-level network including shared decision making (SDM), decisional conflict (DCS), anxiety (GAD-7), depression (PHQ-9), loneliness (ULS-8), urinary incontinence (ICIQ), and exercise adherence; and (2) symptom-level networks for depression (PHQ-9) and anxiety (GAD-7). Given the approximately continuous nature of the scale and symptom severity scores, networks were estimated using Gaussian graphical models (GGMs).24 Spearman rank-order correlation matrices were used as input for the GGMs estimation to account for the ordinal nature and potential non-normality of the data. This approach is commonly applied in psychological network studies and has been shown to be robust for Likert-type symptom data, particularly under regularization. Partial correlation networks were constructed using the graphical least absolute shrinkage and selection operator (glasso) with the extended Bayesian Information Criterion (EBIC) for model selection. The hyperparameter γ was set to 0.5, consistent with recommendations for psychological network analysis,25 to balance sensitivity and specificity. Regularization shrinks small partial correlations toward zero, yielding sparse and interpretable network structures. Networks were estimated and visualized using the qgraph and bootnet packages.26,27 In the symptom-level analyses, nodes represented individual questionnaire items (e.g., PHQ9 [“suicidal ideation”], GAD2 [“difficulty controlling worry”]), whereas in the scale-level network, nodes represented total scores of the corresponding measures. All edge weights reported in this study are standardized partial correlation coefficients. Sensitivity analysis for mixed variable types Given that the scale-level network included variables derived from different measurement types, sensitivity analyses were conducted using mixed graphical models (mgm). All variables were specified as Gaussian (type = "g"), as they represent total scores derived from validated scales and are commonly treated as approximately continuous in psychological research. Model selection was performed using cross-validation (CV), as implemented in the mgm package. This approach prioritizes predictive accuracy and is commonly used in mgm-based analyses. To evaluate the consistency between modeling approaches, the edge-weight matrices obtained from the mgm and EBICglasso models were compared using Pearson correlation coefficients. Item redundancy and information value For symptom-level networks, the informational value of each item was evaluated using average standard deviations, and potential item redundancy was assessed using the goldbricker function from the networktools package.28 A redundancy threshold of 0.25 was applied, with statistical significance set at P < 0.01. No items met criteria for redundancy and were therefore retained in the final networks. Item redundancy assessment was not applied to the scale-level network, as total scores rather than individual items were used. Centrality and predictability Node centrality was assessed using strength, betweenness, and closeness indices calculated via the qgraph package.26 Strength centrality, defined as the sum of absolute edge weights connected to a node, was the primary centrality metric interpreted. Betweenness and closeness centrality were computed but interpreted only when stability criteria were met. Node predictability, defined as the proportion of variance explained by neighboring nodes (R2), was estimated using the mgm package.29 Predictability provides complementary information to centrality by quantifying how well each node is accounted for by the rest of the network. Accuracy and stability analyses Network accuracy and stability were evaluated using non-parametric bootstrap procedures implemented in the bootnet package.27 Edge weight accuracy was assessed by estimating 95% confidence intervals based on 1000 bootstrap samples. Narrow confidence intervals indicate greater precision of edge weight estimates. Centrality stability was examined using case-dropping bootstrap procedures (1000 permutations) to compute the correlation stability coefficient (CS–C). The CS-C reflects the maximum proportion of cases that can be removed while retaining a correlation of at least 0.70 between the original and subset centrality estimates.30 Following an established guideline,27 CS values below 0.25 indicate poor stability and should not be interpreted, whereas values above 0.50 indicate acceptable stability. All statistical analyses were conducted with methodological guidance and were independently verified to ensure accuracy and appropriateness. Results Descriptive statistics A total of 264 individuals were invited to participate, of whom 9 declined (response rate: 96.6%). Four of the remaining 255 were excluded due to inconsistent responses (2 straight-liners, 2 contradictory responses). The final analytical sample included 251 participants (Fig. 1). Flow diagram of participant recruitment and inclusion. The demographic and clinical characteristics of the participants are presented in Table 1, with key characteristics summarized below. The mean age of the participants was 73.09 years (SD = 7.24). The majority were partnered (89.2%), and educational levels were distributed across all categories. Nearly half of the participants reported an annual household income below 100,000 CNY (47.8%). In addition, 84.1% of participants reported having chronic diseases, with an average of 1.60 (SD = 0.75) comorbidities. Table 1Characteristics of participants (N = 251).Table 1CharacteristicValueAge, years (mean ± SD)73.09 ± 7.24Educational level, n (%) Primary school or below49 (19.5) Junior high school75 (29.9) High school/secondary67 (26.7) College or above60 (23.9)Marital status, n (%) Partnered224 (89.2) Non-partnered27 (10.8)Annual household income, n (%) < 100,000 CNY120 (47.8) 100,000–200,000 CNY83 (33.1) 200,001–400,000 CNY36 (14.3) > 400,000 CNY12 (4.8)Chronic diseases, n (%) No40 (15.9) Yes211 (84.1)Number of comorbidities (mean ± SD)1.60 ± 0.75SD, standard deviation. Characteristics of participants (N = 251). SD, standard deviation. Descriptive statistics for the main study variables are presented in Table 2. The mean score for shared decision making (SDM) was 56.42 (SD = 12.22), suggesting a moderate level of perceived involvement in treatment-related decision making. The mean decisional conflict (DCS) score was 52.76 (SD = 6.05). Regarding psychological symptoms, the mean anxiety score (GAD-7) was 5.48 (SD = 6.41), and the mean depression score (PHQ-9) was 5.97 (SD = 8.43). The average loneliness score (ULS-8) was 15.91 (SD = 5.31). For functional and behavioral measures, the mean urinary incontinence score (ICIQ) was 6.91 (SD = 4.18), and the mean exercise adherence score was 1.31 (SD = 1.53). Distributional properties of all node variables, including skewness and kurtosis, are reported in Supplementary Table S1. Most variables exhibited mild positive skewness (skewness range: 0.89–1.73), which is expected for symptom-level measures. Regarding clinical severity, the majority of participants exhibited minimal to mild levels of psychological distress. Specifically, 21.1% (n = 53) of survivors met the criteria for clinically relevant anxiety (GAD-7 ≥ 10), and 21.5% (n = 54) met the threshold for clinically relevant depression (PHQ-9 ≥ 10). Table 2Descriptive statistics of study variables (N = 251).Table 2VariablesMean (SD)n (%)Shared decision making (SDM)56.42 (12.22)–Decisionalconflict (DCS)52.76 (6.05)–Anxiety (GAD-7)5.48 (6.41)– Minimal to mild (0–9)–198 (78.9) Moderate to severe symptoms (≥ 10)a–53 (21.1)Depression (PHQ-9)5.97 (8.43) Minimal to mild (0–9)–197 (78.5) Moderate to severe symptoms (≥ 10)b–54 (21.5)Suicidal ideation (PHQ-9 item 9 ≥ 1)c–52 (20.7)Loneliness (ULS-8)15.91 (5.31)–Urinary incontinence (ICIQ)6.91 (4.18)–Exercise adherence1.31 (1.53)–SD, standard deviation.aDenotes clinically relevant anxiety thresholds;bDenotes clinically relevant depression thresholds.cSuicidal ideation was defined as a positive endorsement (score ≥ 1) on Item nine of the PHQ-9 (“Thoughts that you would be better off dead, or of hurting yourself in some way”) over the past two weeks. Descriptive statistics of study variables (N = 251). SD, standard deviation. Denotes clinically relevant anxiety thresholds; Denotes clinically relevant depression thresholds. Suicidal ideation was defined as a positive endorsement (score ≥ 1) on Item nine of the PHQ-9 (“Thoughts that you would be better off dead, or of hurting yourself in some way”) over the past two weeks. Scale-level network Network structure and predictability A scale-level Gaussian graphical model was estimated including shared decision making (SDM), decisional conflict (DCS), anxiety (GAD-7), depression (PHQ-9), loneliness (ULS-8), urinary incontinence (ICIQ), and exercise adherence. The resulting network was sparse and well-connected, indicating that most associations represented direct conditional relationships after controlling for all other variables (Fig. 2). Network structure and predictability of patient-reported outcomes among prostate cancer survivors with urinary incontinence (Shared Decision Making, Decisional Conflict, Anxiety, Depression, Loneliness, Urinary Incontinence, and Exercise Adherence). A tightly connected psychological cluster was observed, with anxiety and depression showing a strong positive association and both being positively connected to loneliness. Shared decision making was strongly associated with decisional conflict and displayed weaker associations with psychological and functional variables. Urinary incontinence showed positive associations with anxiety, depression, and exercise adherence. Node predictability varied across nodes, with psychological symptoms demonstrating higher predictability (R2 = 0.85–0.86) than functional and behavioral variables (Supplementary Table S2), suggesting that a substantial proportion of variance in psychological symptoms could be explained by their neighboring nodes within the estimated network. Centrality and stability Centrality analyses indicated that anxiety (GAD-7) and depression (PHQ-9) exhibited the highest strength and expected influence, identifying them as the most central nodes in the scale-level network (Fig. 3; Supplementary Table S2). In contrast, loneliness, decisional conflict, and exercise adherence showed comparatively lower centrality values. Centrality indices of the psychosocial and clinical network among prostate cancer survivors with urinary incontinence. Case-dropping bootstrap analyses indicated high robustness of the estimated network structure (CS = 0.75). Strength (CS = 0.59) and expected influence (CS = 0.52) showed acceptable stability and were therefore interpreted. In contrast, closeness (CS = 0.21) and betweenness (CS = 0.00) demonstrated insufficient stability and were not interpreted further (Supplementary Table S3). Bootstrap confidence intervals indicated relatively high precision for the strongest associations, whereas weaker edges were estimated with greater uncertainty (Supplementary Fig. S1). Detailed centrality and stability results are provided in Supplementary Fig. S1 and S2 and Table S2 and S3. Edge weight analysis The strongest association in the scale-level network was observed between shared decision making and decisional conflict (partial r = 0.68), followed by the association between anxiety and depression (partial r = 0.56). Urinary incontinence showed a moderate positive association with exercise adherence (partial r = 0.43) and weaker positive associations with anxiety (partial r = 0.15) and depression (partial r = 0.17). Several negative associations were identified. Shared decision making was negatively associated with urinary incontinence (partial r = −0.20) and anxiety (partial r = −0.08). A weak negative association was also observed between decisional conflict and loneliness (partial r = −0.10). Complete edge weight estimates are reported in Supplementary Table S4. Sensitivity analysis The mixed graphical model demonstrated a moderate-to-high correlation with the EBICglasso network (r = 0.76), indicating that the overall network structure was largely consistent across modeling approaches. Minor differences in edge weights were observed but did not alter the main pattern of associations. Symptom-level depression network (PHQ-9) Network structure and predictability The PHQ-9 symptom network exhibited a well-defined and sparse structure (Fig. 4), indicating that most symptom associations represented direct conditional relationships after controlling for all other symptoms. Node predictability was consistently high across depressive symptoms (R2 range = 0.81–0.95), indicating strong interdependence among depressive symptoms (Supplementary Table S5). Network structure and predictability of depressive symptoms (PHQ-9) among prostate cancer survivors with urinary incontinence. Centrality and stability Case-dropping bootstrap analyses indicated that the network metrics demonstrated varying levels of stability. Strength (CS = 0.52) and expected influence (CS = 0.59) showed adequate stability in the PHQ-9 network, supporting their interpretation. Based on these indices, suicidal ideation (PHQ9) and low energy (PHQ4) emerged as the most central symptoms, exhibiting the highest strength and expected influence (Supplementary Table S5). Poor concentration (PHQ7) showed moderate centrality, whereas sleep problems (PHQ3) and sad mood (PHQ2) demonstrated comparatively lower centrality values. The stability of edge weights was also acceptable (CS = 0.67). In addition, edge-weight accuracy was supported by bootstrap confidence intervals, with the strongest associations (e.g., PHQ5–PHQ9) estimated with relatively high precision (Supplementary Fig. S5). In contrast, closeness (CS = 0.13) and betweenness (CS = 0.00) exhibited insufficient stability (CS < 0.25) and were therefore not interpreted (Supplementary Table S5). Edge weight analysis Several strong symptom-to-symptom associations were identified within the PHQ-9 network. The strongest edge was observed between appetite changes (PHQ5) and suicidal ideation (PHQ9; partial r = 0.66), followed by the association between low energy (PHQ4) and sleep problems (PHQ3; partial r = 0.57). Additional moderate associations included links between low energy (PHQ4) and poor concentration (PHQ7), between poor concentration (PHQ7) and anhedonia (PHQ1), and between psychomotor changes (PHQ8) and suicidal ideation. The overall sparsity of the network indicated that many symptom pairs were conditionally independent after adjustment for other symptoms. Complete edge weight estimates are reported in Supplementary Table S7. Symptom-level anxiety network (GAD-7) Network structure and predictability The GAD-7 symptom network exhibited a sparse and interpretable structure (Fig. 5), with conditional associations distributed across multiple symptom pairs. Node predictability was moderate to high across anxiety symptoms (R2 range = 0.78–0.92), though more heterogeneous compared with the depression network, suggesting greater variability in anxiety symptom interconnections (Supplementary Table S8). Network structure and predictability of anxiety symptoms (GAD-7) among prostate cancer survivors with urinary incontinence. Centrality and stability Strength and expected influence centrality suggested that “Difficulty controlling worry” (GAD2) tended to show higher centrality within the anxiety network (Supplementary Table S8). “Nervousness” (GAD1) and “Trouble relaxing” (GAD4) also demonstrated relatively high centrality, followed by “Restlessness” (GAD5). In contrast, “Irritable” (GAD6) and “Afraid” (GAD7) showed low and negative expected influence, indicating more peripheral positions. Case-dropping bootstrap analyses indicated that strength (CS = 0.52) and expected influence (CS = 0.59) demonstrated acceptable stability in the GAD-7 network, supporting cautious interpretation of these centrality indices. The stability of edge weights was lower but remained acceptable (CS = 0.44). In addition, edge-weight accuracy was supported by bootstrap confidence intervals, with the strongest associations (e.g., GAD2–GAD4) estimated with relatively high precision (Supplementary Fig. S7). In contrast, closeness (CS = 0.21) and betweenness (CS = 0.00) exhibited insufficient stability (CS < 0.25) and were therefore not interpreted (Supplementary Table S9). Edge weight analysis The strongest association in the GAD-7 network was observed between “Difficulty controlling worry” (GAD2) and “Trouble relaxing” (GAD4; partial r = 0.55). Additional notable associations included links between “Nervousness” (GAD1) and “Restlessness” (GAD5; partial r = 0.35), between “Too much worry” (GAD3) and “Nervousness” (GAD1; partial r = 0.28), and between “Too much worry” (GAD3) and “Irritable” (GAD6; partial r = 0.28). Overall, the anxiety network displayed a distributed pattern of associations without a single dominant symptom pair. Complete edge weight estimates are reported in Supplementary Table S10. Discussion Main findings Using a network analytic approach, this study examined the interrelationships among psychological symptoms, decision-related factors, and health behaviors in prostate cancer survivors. Network modeling allowed the identification of statistically central constructs and associations that may be obscured by traditional analytic methods. The resulting network revealed anxiety and depression as the central psychological symptoms, with shared decision making serving as a key connecting node linking psychological distress to other symptoms and behavioral domains. By examining the structure of individual symptoms and psychosocial factors, this study provides an understanding of the mental health burden in prostate cancer and identifies central symptom clusters that may serve as focal points for routine psychosocial screening and monitoring. At the scale level, anxiety and depression emerged as the most central nodes in the network, indicating that these symptoms were highly interconnected with multiple psychosocial, behavioral, and functional variables. Previous network analyses in prostate cancer survivors using the GAD-7 or GAD-10 have shown that the most central anxiety-related symptoms include being unable to control worry and difficulty relaxing,11,12 while depression networks based on the PHQ-9 have identified loss of interest or pleasure in activities as a central depressive symptom.10 Although these studies did not simultaneously model anxiety and depression within a single network, their results provide indirect support for the central role of these symptom domains. In addition, evidence from large-scale studies31,32 and systematic reviews33 employing the PHQ-9 and GAD-7 consistently identifies anxiety and depression as core components of psychological distress. From a network perspective, highly central nodes are theoretically positioned to influence broader symptom network dynamics.33, 34, 35 In the current cross-sectional context, these findings highlight that anxiety and depression require close monitoring in clinical practice. In addition, the sensitivity analysis using mixed graphical models showed a moderate-to-high level of agreement with the EBICglasso network, supporting the robustness of the findings despite the inclusion of variables derived from different measurement scales. The slightly lower correlation between models may reflect differences in underlying assumptions and the use of aggregated scale scores, which can introduce stronger overall associations among variables. A notable finding was the positive association between SDM and decisional conflict. This relationship appears counterintuitive, as greater involvement in decision making is commonly expected to reduce uncertainty and conflict.36, 37, 38 However, in the context of prostate cancer, due to significant variations in quality of life across treatment options, while SDM or decision support tools may reduce decision conflict, the magnitude and duration of these effects vary across studies.39, 40, 41 Increased engagement in decision making may heighten patients’ awareness of potential risks, thereby temporarily amplifying perceived uncertainty.39 Alternatively, the observed association may partly reflect conceptual and measurement proximity between SDM and decisional conflict.42 Notably, the edge between SDM and DCS was the strongest association in the scale-level network. Given the substantial conceptual proximity between these constructs, this association is best interpreted cautiously and may be partially confounded by measurement overlap rather than reflecting a purely substantive clinical relationship. Both SDM and DCS assess closely related dimensions of the treatment decision-making process, including perceived involvement, uncertainty, and confidence regarding treatment choices. Because the present analysis relied on scale-level total scores, it was not possible to disentangle substantive clinical associations from shared construct variance. Therefore, interpretation of this edge should remain provisional until future studies replicate these findings using item-level network models or alternative measurement approaches. Urinary incontinence was positively associated with both anxiety and depression, consistent with findings from previous research.43,44 Persistent physical discomfort, skin irritation, and activity limitations associated with urinary incontinence are associated with impaired quality of life, feelings of embarrassment, and reduced self-worth, which may co-occur with higher psychological distress.45,46 In addition, urinary incontinence demonstrated a moderate positive association with exercise adherence.47,48 However, this relationship must be interpreted with strict caution and treated purely as an exploratory finding. The exercise adherence measure was based on a self-developed, four-item dichotomous checklist that has not undergone formal psychometric validation. This limited variability and reduced measurement granularity may have artificially inflated the strength of its edge with urinary incontinence. While this exploratory pattern could tentatively suggest that patients with more severe functional impairment report higher levels of physical activity as a symptom-management strategy, definitive clinical conclusions cannot be drawn. Future studies utilizing fully validated psychometric instruments are required to explore whether structured exercise programs adapted for urinary incontinence might improve urinary control and alleviate psychological distress. At the symptom level, the depression network exhibited a highly cohesive structure with strong symptom interconnections and high predictability, indicating that individual depressive symptoms were largely explained by their relationships with other symptoms. Notably, the predictability estimates within the PHQ-9 network were exceptionally high. However, these metrics should be interpreted with caution. In symptom networks of this scale, such pronounced predictability values must be evaluated alongside the sample-to-node ratio and the potential for shared conceptual variance among adjacent depressive symptoms. While the regularization procedure minimized overfitting, these high proportions of explained variance may partially reflect tight local dependencies within the PHQ-9 items rather than an unmitigated systemic cascade. Considering this structural context, our findings can be compared with Sharpley et al.'s network analysis of depressive symptoms in prostate cancer patients, which identified loss of interest or pleasure as a central symptom, with suicidal ideation forming a stable symptom cluster.10 While our study identified suicidal ideation and fatigue as the most central depressive symptoms, this difference may reflect variations in sample characteristics. In cancer populations, depressive symptoms vary across phases of diagnosis, treatment, and recurrence, potentially reshaping symptoms over time.11 Compared with the depression network, the anxiety symptom network demonstrated a more dispersed structure. Difficulty controlling worry tended to show higher centrality within the anxiety network, although these findings should be interpreted cautiously given that the edge-weight stability demonstrated only moderate stability (CS = 0.44). This pattern is broadly consistent with previous network studies.11,12 Importantly, suicidal ideation emerged as the most central node in the depression network. While this statistical centrality reflects its tight interconnection with other symptoms, it must not be interpreted as a guarantee that intervening on suicidal ideation will alter the broader network structure. Instead, this finding carries profound clinical urgency, strongly supporting the critical need for integrating suicide risk assessment into routine psychological screening for prostate cancer survivors. Given the cross-sectional nature of this study, longitudinal research is required to evaluate whether clinical management of this symptom can yield downstream improvements across the wider network. Implications for nursing practice and research Anxiety and depression occupied central positions within the psychosocial and functional network, suggesting that these domains may warrant particular attention in survivorship screening and psychosocial assessment. Although the cross-sectional nature of the present study precludes causal inference, the findings indicate that symptoms related to psychological distress may represent clinically relevant components within the broader symptom network. At the symptom level, suicidal ideation emerged as a highly central depressive symptom. This finding suggests that suicide risk assessment may deserve prioritization within routine psychological screening for prostate cancer survivors, particularly among patients with elevated depressive symptom burden. However, whether intervening on central symptoms leads to broader downstream improvements remains to be determined through longitudinal and intervention-based research. While the association between urinary incontinence and exercise adherence tentatively highlights the potential value of exercise-based rehabilitation, this exploratory finding must be viewed as strictly preliminary due to the use of an unvalidated checklist. Future longitudinal studies utilizing formally validated psychometric instruments are needed to clarify temporal and causal relationships, verify this exploratory hypothesis, and replicate these findings across diverse populations. Limitations This study has several limitations. First, participants were recruited from a single tertiary hospital in Guangdong Province, China, using a purposive sampling strategy. This introduces potential selection bias, as patients who seek care at tertiary centers or agree to participate in research may differ systematically from the broader population of prostate cancer survivors. In addition, the purposive sampling approach may further limit representativeness by selectively including individuals who were accessible and willing to participate. Therefore, the generalizability of the findings to other regions, health care settings, and cultural contexts should be interpreted with caution. Second, the cross-sectional design captures associations at a single time point and does not allow for conclusions regarding causality or the temporal ordering of variables. The observed network structure reflects conditional associations rather than directional relationships. Third, the findings cannot fully account for disease-related and treatment-related heterogeneity due to the absence of key clinical variables in our dataset. Specifically, we lacked data on tumor stage, Gleason score, and time since surgery, which are major determinants of urinary incontinence severity, psychological symptoms, and health behaviors. Consequently, the network configurations presented here may not reflect the diverse clinical trajectories of prostate cancer survivors. Furthermore, the relatively dense network structure observed in the EBICglasso model may reflect high intercorrelations among scale-level variables, potentially increasing the likelihood of weaker edges being retained. Therefore, small edge weights should be interpreted with caution. Additionally, several clinically relevant variables were not systematically collected as part of the study protocol and could therefore not be included in the analyses, such as tumor stage, Gleason score, and time since surgery. As these factors are known to influence both urinary incontinence severity and psychological outcomes, their absence limits the ability to account for disease-related heterogeneity within the sample and should be considered when interpreting the findings. Finally, exercise adherence was assessed using a self-developed checklist without formal psychometric validation. Although the items were constructed based on established international guidelines, the lack of validated reliability and validity metrics may limit the precision of adherence classification. Conclusions This study explored the network characteristics of prostate cancer survivors across psychological, behavioral, and symptomatic dimensions, identifying anxiety and depression as core nodes within the overall network structure. The complex association between shared decision making and decisional conflict suggests that individuals experiencing emotional distress may benefit from tailored decisional support. By adopting a network analysis perspective, this study contributes to identifying core symptoms and stable interconnections. While network theory suggests that addressing these central nodes could foster broader improvements across psychosocial, decision-related, and functional domains, future studies employing multicenter and longitudinal research designs are necessary to determine whether targeting these nodes actually improves patient outcomes. CRediT authorship contribution statement Yan He: Resources, Conceptualization, Methodology, Writing – Original Draft. Zhiguang Huang: Data Curation, Formal Analysis, Visualization. Jiayu Gong: Resources, Writing – Review & Editing. Zunguang Bai: Supervision, Funding Acquisition. Wenfeng He: Resources, Supervision. Liling Zhang: Project Administration, Software. All authors have read and approved the final manuscript. Ethical statement The study protocol was approved by the Ethics Committee of Guangdong Provincial Hospital of Traditional Chinese Medicine (Approval No. ZE2025-360-01), and conducted in accordance with the 1964 Helsinki Declaration and its later amendments. All participants provided written informed consent. Data availability The data supporting the findings of this study are available on request from the corresponding author, L.Z. The data are not publicly available due to ethical and privacy considerations, as well as the ongoing nature of this research series. Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work, the authors used Gemini (Google) in order to improve the language, clarity, and readability of the manuscript. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. Funding This project is funded by the Guangzhou Municipal Science and Technology Program (Grant No. 2025A03J3144). The funders had no role in the study design; data collection, analysis, or interpretation; preparation of the manuscript; or the decision to submit the article for publication Declaration of competing interest The authors declare no conflict of interest. Acknowledgments The authors would like to express their gratitude to the Department of Infectious Diseases and Public Health at the City University of Hong Kong for providing expert statistical consultation and methodological support throughout the data analysis and verification process of this study. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.apjon.2026.100995. Contributor Information Wenfeng He, Email: hwfzll@gzucm.edu.cn. Liling Zhang, Email: zllhwf@126.com. Supplementary materials The following is the Supplementary data to this article. Multimedia component 1 References Associated Data Supplementary Materials Multimedia component 1 Data Availability Statement The data supporting the findings of this study are available on request from the corresponding author, L.Z. The data are not publicly available due to ethical and privacy considerations, as well as the ongoing nature of this research series.