Abstract
Purpose
Erection dysfunction (ED) represents a globally prevalent men's health problem and affected by a variety of factors. This study aimed to develop a dynamic nomogram model to assess the probability of ED in a population through a multitude of factors.
Materials and Methods
A total of 2,668 subjects from the National Health and Nutrition Examination Survey were included in this study. The entire dataset was randomly divided into training and validation sets, with the training set comprising 70% of the data and the validation set comprising 30%. The Least Absolute Shrinkage and Selection Operator and multivariate logistic regression analysis determined the predictors for constructing the nomogram, and the model was evaluated by Concordance Index (C-index), calibration curve, Hosmer-Lemeshow test, and decision curve analysis.
Results
The nomogram model consisted of 9 predictors, which were age, education, stroke, lymphocyte, diabetes, poverty income ratio, prostate disease, activity, and hypertension. The C-index for the training set was 0.828 and for the validation set was 0.825, indicating that the model shows good clinical applicability and calibration of the model on both the training and validation sets. Additionally, we created an online dynamic nomogram (https://wvknly-liu-guodong.shinyapps.io/dynnomapp/) that anyone can evaluate on a web page.
Conclusions
Our dynamic nomogram can integrate multiple risk factors to provide a personalized risk assessment that is highly clinically predictive and provides a valuable tool for early intervention and prospective management of ED. This could help physicians to identify and manage high-risk populations early and provide personalized treatment plans.
Keywords: Erectile dysfunction, Nomograms, Nutrition surveys, Predictive learning models
INTRODUCTION
Erectile dysfunction (ED) is a prevalent condition characterized by the inability to achieve or maintain an erection sufficient for a satisfying sexual encounter [1]. It affects millions of men worldwide, severely impacting their quality of life and psychosocial well-being [2]. A longitudinal study of a randomly sampled population in Massachusetts showed that the annual risk of ED is approximately 26 per 1,000 men [3], and the global incidence of ED is projected to rise dramatically due to increasing ED-related comorbidities, potentially affecting nearly 320 million men by 2025 [4]. The etiology of ED is multifactorial, with extensive research identifying risk factors broadly categorized into physiological, psychological, and lifestyle-related domains [5].
First, age remains the most significant non-modifiable risk factor, predominantly affecting men over 40 years [6]. In addition, cardiovascular disease (CVD) is one of the major physiological risk factors for ED [7]. Conditions such as hypertension, atherosclerosis, and hyperlipidemia impair vascular function critical for erectile hemodynamics [8]. Several cross-sectional and longitudinal studies have further found that the occurrence of ED was associated with diabetes, metabolic syndrome, depression, and lower urinary tract symptoms [9]. A meta-analysis of 12 prospective cohort studies provided strong evidence demonstrating that ED indeed is statistically significantly and independently associated with an increased risk of CVD, coronary heart disease, stroke, and all-cause mortality [10]. Furthermore, environmental and lifestyle factors such as obesity, smoking, and physical inactivity have been identified as modifiable predictors of ED [10].
Given the multifactorial nature of ED, comprehensive risk factor assessment enables healthcare providers to better predict high-risk populations and implement targeted interventions. This study aimed to quantitatively identify and assess risk factors associated with ED by extracting large-scale data from the National Health and Nutrition Examination Survey (NHANES), using statistical techniques and machine learning algorithms to construct predictive models, to comprehensively analyze and construct risk prediction models for ED, and to explore their potential impact on clinical practice and future research directions.
MATERIALS AND METHODS
1. Study design and participants
NHANES is a nationally representative research program administered by the National Center for Health Statistics (NCHS), a division of the Centers for Disease Control and Prevention (CDC), designed to assess the health status of adults and children in the US. Utilizing a multi-stage, stratified, clustered sampling methodology, NHANES annually selects approximately 5,000 participants to evaluate the health and nutritional status of the US civilian population. Data collection includes standardized interviews, physical examinations, and laboratory tests. Additional survey design and protocol details are available on the official NHANES website (https://www.cdc.gov/nchs/nhanes/). All participants provided written informed consent prior to inclusion in the NHANES database, with protocols reviewed and approved by the NCHS Ethics Review Committee.
In our study, we selected datasets from two NHANES study cycles (2001–2002 and 2003–2004) for analysis based on the information on ED in NHANES, since information on ED data was only available for these two cycles. From 2001 to 2004, 21,161 individuals participated in NHANES, to ensure the accuracy of the predictive model, we included as many indicators as possible to analyze the factors influencing ED comprehensively. Missing data were excluded as follows: women (n=10,860); missing information on ED (n=6,143); age >70 years (n=776); missing information on body mass index (BMI; n=43); missing information on hypertension (n=98); missing information on education (n=1); missing information on marriage (n=2); and missing data on laboratories (n=156); Information on poverty income ratio (PIR) was missing (n=168); information on alcohol consumption was missing (n=200); and information on questionnaires was missing (n=46). Finally, a total of 2,668 cases were included in our study, including 371 ED patients and 2,297 controls. Fig. 1 illustrates the data screening process.
Fig. 1. Flow chart of the development and validation groups. NHANES: National Health and Nutrition Examination Survey, ED: erection dysfunction.
2. Data collection and definition
In the two NHANES cycles selected, there was an evaluation item specific to ED that requested participants to assess their ability to achieve and maintain an erection sufficient for sexual intercourse, including “never,” “sometimes,” “usually,” and “always or almost always,” and we categorized subjects who answered ‘never’ or ‘sometimes’ as patients with ED. Based on this, we began to conduct relevant statistical analyses to identify factors associated with ED and construct predictive models.
3. Potential predictors
To construct a more realistic and comprehensive prediction model of ED, we comprehensively analyzed the data collected by NHANES. We incorporated information on as many variables as possible by the current relevant research and clinical experience. The following variables were included: age, BMI, PIR (a socioeconomic status index published annually by the US. Department of Health and Human Services that reflects the ratio of household income to poverty, categorized as PIR (≤1.3, 1.3–3.5, and >3.5), cholesterol (mmol/L), triglyceride (TG, mmol/L), white blood cell count (WBC), lymphocyte (%), hemoglobin (Hb, g/dL), hematocrit (%), mean cell volume (fL), mean cell Hb (pg), mean Hb concentration (g/gL), platelet count, mean platelet volume (fL), C-reactive protein (CRP, mg/dL), albumin (g/L), alanine transaminase (µ/L), aspartate aminotransferase (µ/L), blood urea nitrogen (BUN; mmol/L), Uric acid (µmol/L), creatinine (µmol/L), globulin (g/L), vitamin D (nmol/L), race (Mexican American, non-Hispanic black, non-Hispanic white, or other), marital status (married/living with a partner or living alone), educational level (less than high school, high school, more than high school), diabetes (no, yes), hypertension (no, yes), coronary heart disease (CHD, no, yes), stroke (no, yes), low back pain in the last three months (no, yes), number of bad mood days in the last month (divided into four groups; 0 days is no, 1–7 days is mild, 7–14 days is moderate, and 14–30 days is severe), alcohol intake (severe is more than five drinks per day and or greater than four days drunk per month, moderate is 3–5 drinks per day and or 1–4 days drunk per month, and mild is drinking but does not meet the above conditions, the never), smoking (no, yes), daily exercise (less than peers, same as peers, more than peers), and prostate disease (no, yes). These variables were recognized as potential factors influencing ED and included as initial predictive model screening variables.
4. Model development and validation
To develop and validate the nomogram, we randomly divided the total dataset into two parts, the training set and the validation set, with the training set accounting for 70% of the dataset and the validation set accounting for 30%. The optimal predictors in the training and validation sets were identified through the Least Absolute Shrinkage and Selection Operator regression model. Subsequently, multivariate logistic regression analysis was performed on these predictors to identify independent predictors with p-values less than 0.05 and the nomogram was constructed using these predictors. To validate the efficacy of the constructed model, we utilized the area under the curve (AUC) of the receiver operating characteristic (ROC) curves and the Concordance Index (C-index) to quantify the power of the nomogram to assess the ED. Calibration curves and the Hosmer-Lemeshow test were also utilized to assess the consistency of the real results with the predicted probabilities. Finally, we evaluated the clinical efficacy of the nomogram with a clinical decision curve analysis (DCA). Meanwhile, we constructed a web-based dynamic nomogram to facilitate the application in clinical practice.
5. Statistical analysis
Since NHANES adopts a complex multi-stage sampling design, we applied sample weights, grouping, and stratification information in our statistical analyses in order to obtain more realistic statistical results that would allow us to assess the included population as accurately as possible. When describing continuous variables, we used the mean and standard error or median (interquartile spacing), while when describing categorical variables, we used percentages. In order to analyze the differences between the two groups, we performed the weighted t-test or the weighted Mann–Whinney U test (continuous variables) and the weighted chi-square test (categorical variables). The study was analyzed and processed by using R statistical software (4.2.1.) and SPSS ver. 26.0 statistical package (IBM Corp.). All tests were two-sided and a p-value of less than 0.05 was considered statistically significant.
RESULTS
1. Baseline characteristics of participants
According to the process in Table 1, we finally recruited a total of 2,668 participants, of which 1,868 were in the training set and 800 in the validation set. Table 1 details the results of the weighted analyses of the total participant demographics, disease information, and biochemical metrics. The results showed that 13.9% of the participants had ED. Age, BMI, TG, CRP, BUN, and globulin were significantly higher in people with ED than in those without ED, whereas vitamin D, lymphocyte, Hb, Hematocrit, and albumin levels were significantly lower (p<0.05). In the ED population, lower PIR and education, being married or living with a partner, suffering from diabetes, hypertension, or CHD, ever had a stroke, experienced back pain in the last three months, being moody in the last month, exercising less, and suffering from prostate disease were significantly higher than in the control group. In addition, we conducted a comparative analysis of the baseline characteristics between the training and validation sets to confirm the randomness and reasonableness of the division of the total dataset. The results showed no statistically significant difference in all data between the two datasets (all p>0.05, Supplement Table 1).
Table 1. Baseline characteristics of participants with or without ED, weighted.
| Characteristic | Total participants (2,668) | No ED (2,297) | ED (371) | p-value | |
|---|---|---|---|---|---|
| Age (y) | 42.3±0.25 | 40.39±0.25 | 54.11±0.64 | <0.0001 | |
| BMI (kg/m2) | 28.10±0.10 | 27.39±0.11 | 29.43±0.30 | <0.0001 | |
| Cholesterol (mmol/L) | 5.23±0.02 | 5.23±0.02 | 5.24±0.06 | 0.8641 | |
| TG (mmol/L) | 1.75±0.03 | 1.71±0.03 | 1.99±0.07 | <0.0001 | |
| WBC count | 7.18±0.04 | 7.17±0.04 | 7.23±0.10 | 0.6364 | |
| Lymphocyte (%) | 29.79±0.15 | 30.00±0.16 | 28.49±0.44 | <0.0001 | |
| Hb (g/dL) | 15.51±0.02 | 15.54±0.02 | 15.32±0.06 | <0.0001 | |
| Hematocrit (%) | 45.72±0.06 | 45.80±0.06 | 45.23±0.18 | <0.0001 | |
| MCV (fL) | 90.64±0.09 | 90.53±0.09 | 91.28±0.25 | 0.0030 | |
| MCH (pg) | 30.75±0.03 | 30.73±0.04 | 30.92±0.10 | 0.0518 | |
| MHC (g/dL) | 33.93±0.01 | 33.94±0.01 | 33.87±0.04 | 0.0773 | |
| Platelet count | 256.42±1.16 | 257.92±1.25 | 247.17±3.14 | 0.0013 | |
| MPV (fL) | 8.15±0.02 | 8.14±0.02 | 8.16±0.04 | 0.6932 | |
| CRP (mg/dL) | 0.15 (0.26) | 0.14 (0.25) | 0.24 (0.39) | <0.0001* | |
| Albumin (g/L) | 44.03±0.06 | 44.24±0.06 | 42.78±0.15 | <0.0001 | |
| ALT (µ/L) | 26.00 (13) | 26.00 (13) | 26.00 (13) | 0.4861* | |
| AST (µ/L) | 24.00 (8) | 24.00 (8) | 24.00 (9) | 0.3219* | |
| BUN (mmol/L) | 4.83±0.03 | 4.76±0.03 | 5.24±0.09 | <0.0001* | |
| UA (umol/L) | 362.07±1.38 | 362.16±1.48 | 361.55±3.89 | 0.8806 | |
| Creatinine (µmol/L) | 87.87±0.47 | 87.94±0.52 | 87.36±0.99 | 0.6621 | |
| Globulin (g/L) | 29.06±0.08 | 28.96±0.08 | 29.71±0.25 | <0.0001 | |
| Vitamin D (nmol/L) | 64.19±0.40 | 64.66±0.44 | 61.33±1.02 | 0.0043 | |
| Race (%) | 0.6112 | ||||
| NH-black | 9.1 | 9.1 | 9.0 | ||
| NH-white | 74.3 | 74.6 | 72.4 | ||
| Mexican | 8.7 | 8.6 | 9.0 | ||
| Others | 8.0 | 7.7 | 9.7 | ||
| PIR (%) | 0.0454 | ||||
| ≤1.3 | 15.6 | 15.2 | 18.1 | ||
| 1.3–3.5 | 34.2 | 33.7 | 37.6 | ||
| >3.5 | 50.1 | 51.1 | 44.2 | ||
| Marital (%) | <0.0001 | ||||
| Cohabitation | 69.5 | 68.0 | 78.6 | ||
| Solitude | 30.5 | 32.0 | 21.4 | ||
| Education (%) | <0.0001 | ||||
| Less than high school | 15.6 | 13.5 | 28.4 | ||
| High school | 27.7 | 28.4 | 23.5 | ||
| More than high school | 56.7 | 58.1 | 48.2 | ||
| Diabetes (%) | <0.0001 | ||||
| No | 92.2 | 95.4 | 72.5 | ||
| Yes | 7.8 | 4.6 | 27.5 | ||
| Hypertension (%) | <0.0001 | ||||
| No | 68.4 | 71.8 | 47.2 | ||
| Yes | 31.6 | 28.2 | 52.8 | ||
| CHD (%) | <0.0001 | ||||
| No | 96.5 | 97.8 | 88.2 | ||
| Yes | 3.5 | 2.2 | 11.8 | ||
| Stroke | <0.0001 | ||||
| No | 98.9 | 99.3 | 95.8 | ||
| Yes | 1.1 | 0.7 | 4.2 | ||
| Back pain past 3 months | 0.0002 | ||||
| No | 61.2 | 62.6 | 52.3 | ||
| Yes | 38.8 | 37.4 | 47.7 | ||
| Days of bad mood | 0.0024 | ||||
| No | 65.3 | 65.4 | 64.6 | ||
| Mild | 22.1 | 22.8 | 17.4 | ||
| Moderate | 3.8 | 3.7 | 4.7 | ||
| Severe | 8.8 | 8.1 | 13.2 | ||
| Alcohol user (%) | <0.0001 | ||||
| No | 15.8 | 13.7 | 29.0 | ||
| Mild | 33.1 | 32.8 | 34.8 | ||
| Moderate | 41.5 | 43.9 | 26.7 | ||
| Severe | 41.5 | 9.6 | 9.6 | ||
| Smoke (%) | 0.2116 | ||||
| No | 57.9 | 57.5 | 60.9 | ||
| Yes | 42.1 | 42.5 | 39.1 | ||
| Activity | <0.0001 | ||||
| Less than peers | 17.4 | 16.1 | 25.9 | ||
| Same as peers | 43.9 | 43.6 | 45.9 | ||
| More than peers | 38.7 | 40.4 | 28.2 | ||
| Prostate disease | <0.0001 | ||||
| No | 92.6 | 94.3 | 81.9 | ||
| Yes | 7.4 | 5.7 | 18.1 | ||
ED: erection dysfunction, BMI: body mass index, TG: TRiglyceride, HDL-c: high density lipoprotein cholesterol, WBC count: white blood cell count, Hb: hemoglobin, MCV: mean cell volume, MCH: mean cell hemoglobin, MHC: mean hemoglobin concentration, MPV: mean platelet volume, CRP: C-reactive protein, ALT: alanine transaminase, AST: aspartate aminotransferase, BUN: blood urea nitrogen, UA: uric acid, PIR: poverty income ratio, CHD: coronary heart disease.
For continuous variables: survey-weighted mean, p-value was by survey-weighted student t-test and Mann–Whitney U test. For categorical variables: survey-weighted percentage, p-value was by survey-weighted chi-square test.
*Indicates that the results were obtained using weighted Mann–Whitney U test statistics.
2. Screening predictors and constructing nomogram model
We included a total of 36 relevant feature variables in the training set in the LASSO regression analysis, and we filtered out 9 potential predictors with non-zero coefficients, including age, education, stroke, lymphocyte, diabetes, PIR, prostate disease, activity, hypertension (Fig. 2). The above factors were further confirmed as independent predictors of ED by multivariate logistic regression analysis with the following results: age (OR: 1.08, 95% CI: 1.06–1.09, p<0.001), Lymphocyte (OR: 0.98, 95% CI: 0.96–1.00, p=0.015), PIR 1.3–3.5 (OR: 0.77, 95% CI: 0.53–1.11, p=0.163) and PIR >3.5 (OR: 0.61, 95% CI: 0.41–0.92, p=0.017) compared to PIR ≤1.3, high school (OR: 0.57, 95% CI: 0.38–0.83, p=0.004), and more than high school (OR: 0.50, 95% CI: 0.36–0.72, p<0.001) compared to less than high school, diabetes (OR: 2.18, 95% CI: 1.50–3.15, p=0.037), hypertension (OR: 1.37, 95% CI: 1.02–1.84, p=0.037), and stroke (OR: 4.04, 95% CI: 1.67–10.69, p=0.003), exercise normally (OR: 0.85, 95% CI: 0.58–1.25, p=0.411) and more often (OR: 0.44, 95% CI: 0.29–0.67, p<0.001) compared to physical inactivity, suffer from prostate disease (OR: 1.92, 95% CI: 1.28–2.89, p=0.002) (Supplement Table 2). We included these variables in the construction of the nomogram, shown in Fig. 3, which demonstrates the risk scores corresponding to each factor and the coefficients corresponding to the presence of ED. Additionally, we created an online dynamic nomogram (https://wvknly-liu-guodong.shinyapps.io/dynnomapp/) that anyone can evaluate on a web page, as shown in Supplement Fig. 1, where users can obtain the corresponding probability of disease and 95% confidence interval based on the influencing factors.
Fig. 2. Variable selection using the Least Absolute Shrinkage and Selection Operator (LASSO) regression model. (A) Generated coefficient curves plotted against log (lambda) series. (B) Plot of partial likelihood deviation (binomial deviation) versus log (lambda).
Fig. 3. Nomogram for assessing the risk of ED. The points for each characteristic were summed to give a total number of points, resulting in a corresponding ED risk point score. PIR: poverty income ratio, ED: erection dysfunction.
3. Evaluation and validation of the nomogram model
First, we assessed the discrimination of the nomogram by the ROC curves of the training and validation sets (Fig. 4). The AUC of the ROC curve for the training set is 0.828 (95% CI: 0.803–0.853), while the AUC of the validation set is 0.825 (95% CI: 0.786–0.864), which indicates that the nomogram model has a better discriminative ability and predictive value. Supplement Table 3 shows the performance of the clinical model. In the training set, the specificity was 82.17%, the sensitivity was 70.33%, and the accuracy was 80.3%. In the validation set, specificity was 82.67%, sensitivity was 74.32%, and accuracy was 81.13%. Next, calibration curves were plotted to assess the agreement between predicted and observed outcomes (Fig. 5). The horizontal axis represents the predicted ED risk, while the vertical axis indicates the actual diagnosed ED risk. Both training and validation set curves closely followed the diagonal reference line, indicating strong alignment between predicted probabilities and observed frequencies. Meanwhile, the Hosmer-Lemeshow test returns a p-value of 0.798 for the training set and a p-value of 0.108 for the validation set, indicating that the nomogram exhibits good agreement between prediction and actuality. Finally, DCA was performed to evaluate clinical utility (Fig. 6). The blue line indicating the model, the gray line indicating the net benefit for all ED patients, and the black line indicating the net benefit for the ED-free population. The area of the model curve between the black line and the gray line represents the model with a clinical net benefit, showing that the model has a very significant clinical net benefit rate.
Fig. 4. The receiver operating characteristic (ROC) curves for the prediction model of erection dysfunction. (A) ROC curve for the Training group. (B) ROC curve of the validation group. AUC: area under the curve.
Fig. 5. Calibration curves for the nomogram of the prediction model. (A) Calibration curve for the training group. (B) Calibration curve for the validation group.
Fig. 6. Decision curve for the nomogram. (A) Decision curve for the training group. (B) Decision curve for the validation group.
DISCUSSION
To date, there are no studies incorporating large amounts of data to develop predictive models regarding ED. The development and validation of predictive models for the risk of ED hold significant potential for advancing medical research and clinical practice. ED is a common condition that affects a significant proportion of men, especially as they age, and is associated with a variety of physical, psychological, and lifestyle factors. Our analysis of baseline characteristics revealed statistically significant differences between participants with and without ED, underscoring the multifactorial nature of this condition. Key contributors included demographic variables, biochemical indicators, and modifiable lifestyle elements. Based on these findings, we developed a clinically interpretable nomogram incorporating nine critical predictors: age, education, stroke, lymphocyte count, diabetes, PIR, prostate disease, physical activity, and hypertension. The model demonstrated robust discriminative performance, with the AUC values of 0.828 (training set) and 0.825 (validation set). Calibration curves exhibited strong alignment between predicted probabilities and observed outcomes, further validated by non-significant Hosmer-Lemeshow test results (p>0.05 for both cohorts). DCA confirmed substantial clinical utility across multiple probability thresholds, indicating favorable benefit-to-harm ratios for clinical decision-making.
Regarding demographic data, we found a significant relationship between age and ED in our study, and the nomogram shows an increasing score of age on ED risk with advancing age. This aligns with previous studies demonstrating that ED is strongly associated with age-related factors [11]. Aging is associated with endothelial dysfunction, decreased testosterone levels, and other physiological changes contributing to ED [12]. In addition to age, BMI is another more common risk factor for ED. Consistent with prior research, our study confirms that obesity plays a significant role in ED development. The regulation of obesity and ED involves multiple mechanisms, including altered sex hormone metabolism and insulin resistance, both of which are associated with endothelial dysfunction—a key pathway in ED pathogenesis [13]. Furthermore, our study highlights PIR (household income/poverty threshold income) as a risk factor for ED. PIR reflects the economic status of households, and our findings are supported by studies suggesting that individuals with lower incomes are more likely to experience ED symptoms. Lower PIR may lead to chronic stress accumulation, which can result in physical and emotional strain. Additionally, low-income groups are more likely to engage in unhealthy behaviors, such as smoking and alcohol abuse, which are established risk factors for ED [14]. Moreover, individuals with higher socioeconomic status tend to have better access to healthcare and greater awareness of physical health, whereas lower-income populations often face poorer healthcare access and community environments, further explaining the link between socioeconomic status and ED [15]. Education level also emerged as a strong predictor of ED in our study, with higher education levels associated with a lower risk of ED. A large meta-analysis including Europe, the Americas, and Asia concluded that lower levels of education are associated with CVD [16]. A Study in the Netherlands explained that the level of education may be related to the development of CVD through biologically mediated and behavioral variables such as smoking, obesity, physical inactivity, and hypertension [17]. Given the strong connection between CVD and ED, these findings provide a potential explanation and future research direction for the association between education level and ED.
Patients with ED exhibited distinct biochemical profiles compared to controls, including elevated levels of TG, CRP, BUN, and globulin, alongside reduced levels of vitamin D, lymphocytes, Hb, hematocrit, and albumin. Notably, lymphocyte percentage emerged as a protective factor in the final nomogram. These biochemical indicators suggest several potential pathological mechanisms: First, there is the inflammatory mechanism, as CRP levels are strongly correlated with systemic levels of inflammation, which is a known contributor to endothelial dysfunction and atherosclerosis, both of which are key mechanisms leading to ED [18]. A meta-analysis has also demonstrated an association between CRP and ED [19]. Additionally, lymphocytes serve as a protective marker for inflammation [20], with higher levels of lymphocytes being protective against ED. Their role in immune regulation may also support vascular health [21]. Subsequently, some biochemical markers are closely related to metabolic dysregulation, which is another important factor that may contribute to ED. Lipid abnormalities, particularly hyperlipidemia, are closely linked to atherosclerosis and reduced penile blood flow, as evidenced by their strong association with CVD mechanisms. In recent years, it has been found that there is also a link between vitamin D deficiency and ED, and the mechanism behind this may be related to endothelial dysfunction and CVD [22]. Lastly, there are indicators related to nutritional status, such as low levels of albumin and Hb indicating poor nutritional status and hypoxia, respectively, which can impair overall health and vascular function, leading to ED [23].
Our research has found a close connection between ED and several other conditions as well, starting with the more common ones of diabetes and hypertension. Diabetes is a significant predictor of ED, which may be inextricably linked to the well-known effects of diabetes on vascular and nerve function [24]. Hyperglycemia leads to microvascular damage, oxidative stress, and neuropathy, all of which can impair erectile function. Hypertension, in turn, is a CVD that causes endothelial damage to blood vessels, reduces nitric oxide supply, and leads to atherosclerosis, all of which can affect penile blood flow [25]. In addition, stroke is often associated with vascular disease, which can directly affect penile blood flow [26]. Prostatic disorders, including benign prostatic hyperplasia and prostatitis, elevate ED risk via vascular compromise, inflammatory cascades, and neurological dysfunction [27].
One important finding that cannot be overlooked is the protective effect of physical activity on erectile function. Lack of exercise is a significant contributing factor to the development of ED [28]. Exercise can improve cardiovascular health, enhance endothelial function, and reduce obesity and stress, all of which have beneficial effects on erectile function. Consequently, lifestyle modification emphasizing physical activity remains a cornerstone of ED management [29].
The development and validation of our predictive model for assessing ED risk represent a significant advancement in the field of men's health. By integrating multiple risk factors, the model offers a personalized risk assessment tool that holds great potential for early intervention and proactive management of ED. For example, entering the following predictor variables into the dynamic nomogram: age (48 years), education level (less than high school), PIR (1.0), lymphocyte percentage (28), hypertension (yes), diabetes mellitus (yes), stroke (no), prostate disease (no), and physical activity (less than peers) yields a risk of 0.515 (0.396–0.633) for developing ED, which is a medium risk group. Based on the results, the patient can then be advised to undergo lifestyle interventions (weight loss, exercise) and to monitor blood pressure and blood glucose.
Our study underscores the complex interplay of demographic, biochemical, and lifestyle factors in ED development, while the dynamic nomogram provides a user-friendly, individualized approach to ED risk evaluation. However, our study has several limitations. First, the cross-sectional design limited our ability to establish causal relationships between variables and ED. Although we controlled for confounders through multivariate adjustment and rigorous statistical methods, it was not possible to determine the temporal order and causal relationship between risk factors and ED. The lack of causality may affect the effectiveness of intervention strategies and thus the utility of the model in guiding clinical interventions. Future studies are needed to track the relationship between dynamic changes in risk factors and ED onset using cohorts with time-series data in subsequent studies and to design randomized controlled trials to assess the effects of interventions. Exploring potential mediating mechanisms such as endothelial dysfunction, chronic inflammation, and oxidative stress would also be helpful in understanding the pathophysiologic processes of ED. Additionally, although our model incorporates multidimensional factors such as demographic, biochemical, and lifestyle variables, some important factors were not included. Psychological factors (e.g., stress, depression, and anxiety) are well-documented contributors to ED, potentially affecting erectile function through neuroendocrine pathways. Similarly, medication history (e.g., use of antihypertensive or antidepressant drugs) may directly or indirectly influence sexual function. However, due to limitations in the NHANES dataset, these variables were not included in our analysis. Future studies should aim to incorporate these factors to refine ED risk prediction models. Furthermore, our study sample was drawn from the US population, which may limit the generalizability of our findings to men of other ethnicities and regions. Future research should focus on validating the global applicability of the model. For example, validating models using external datasets from different regions (e.g., cohort data from Europe, Asia, and Africa) and exploring differences in risk factors across cultures.
CONCLUSIONS
We developed a nomogram using NHANES data to assess the risk of ED in the US male population. Moreover, we have developed a web page dynamic nomogram prediction model that can be used by all individuals to make predictions about the risk of developing ED. The model has strong clinical predictive power, enabling clinicians to use this tool to identify patients at high risk for ED. Potentially, this could reduce the incidence of ED at the preventive level and significantly improve the prognosis of patients with ED.
Acknowledgements
The authors would like to thank all the participants and staff of NHANES.
Footnotes
Conflict of Interest: The authors have nothing to disclose.
Funding: The National Natural Science Foundation of China (Grant No. 82371635).
- Conceptualization: GL, YZ.
- Data curation: GL, YZ.
- Formal analysis: GL, YZ, XW, HG.
- Funding acquisition: XZ.
- Investigation: GL.
- Methodology: GL, YZ.
- Project administration: XZ.
- Resources: GL, XZ.
- Software: GL.
- Supervision: HJ, XZ.
- Validation: GL, HJ, XZ.
- Visualization: XZ.
- Writing – original draft: GL.
- Writing – review & editing: GL, XZ.
Data Sharing Statement
Data can be downloaded from the cdc.gov website.
Supplementary Materials
Supplementary materials can be found via https://doi.org/10.5534/wjmh.250018.
Baseline characteristics of participants in the training and validation sets, weighted
Logistic regression analyses of the included factors
Diagnostic efficacy of the nomogram model for assessing erection dysfunction risk
The web-based online dynamic nomogram tool. PIR: poverty income ratio.
References
- 1.Yafi FA, Jenkins L, Albersen M, Corona G, Isidori AM, Goldfarb S, et al. Erectile dysfunction. Nat Rev Dis Primers. 2016;2:16003. doi: 10.1038/nrdp.2016.3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Korfage IJ, Pluijm S, Roobol M, Dohle GR, Schröder FH, Essink-Bot ML. Erectile dysfunction and mental health in a general population of older men. J Sex Med. 2009;6:505–512. doi: 10.1111/j.1743-6109.2008.01111.x. [DOI] [PubMed] [Google Scholar]
- 3.Johannes CB, Araujo AB, Feldman HA, Derby CA, Kleinman KP, McKinlay JB. Incidence of erectile dysfunction in men 40 to 69 years old: longitudinal results from the Massachusetts male aging study. J Urol. 2000;163:460–463. [PubMed] [Google Scholar]
- 4.Pallangyo P, Nicholaus P, Kisenge P, Mayala H, Swai N, Janabi M. A community-based study on prevalence and correlates of erectile dysfunction among Kinondoni District Residents, Dar Es Salaam, Tanzania. Reprod Health. 2016;13:140. doi: 10.1186/s12978-016-0249-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Nguyen HMT, Gabrielson AT, Hellstrom WJG. Erectile dysfunction in young men-a review of the prevalence and risk factors. Sex Med Rev. 2017;5:508–520. doi: 10.1016/j.sxmr.2017.05.004. [DOI] [PubMed] [Google Scholar]
- 6.Nicolosi A, Glasser DB, Kim SC, Marumo K, Laumann EO GSSAB Investigators' Group. Sexual behaviour and dysfunction and help-seeking patterns in adults aged 40-80 years in the urban population of Asian countries. BJU Int. 2005;95:609–614. doi: 10.1111/j.1464-410X.2005.05348.x. [DOI] [PubMed] [Google Scholar]
- 7.Terentes-Printzios D, Ioakeimidis N, Rokkas K, Vlachopoulos C. Interactions between erectile dysfunction, cardiovascular disease and cardiovascular drugs. Nat Rev Cardiol. 2022;19:59–74. doi: 10.1038/s41569-021-00593-6. [DOI] [PubMed] [Google Scholar]
- 8.Gandaglia G, Briganti A, Jackson G, Kloner RA, Montorsi F, Montorsi P, et al. A systematic review of the association between erectile dysfunction and cardiovascular disease. Eur Urol. 2014;65:968–978. doi: 10.1016/j.eururo.2013.08.023. [DOI] [PubMed] [Google Scholar]
- 9.Rosen RC, Link CL, O'Leary MP, Giuliano F, Aiyer LP, Mollon P. Lower urinary tract symptoms and sexual health: the role of gender, lifestyle and medical comorbidities. BJU Int. 2009;103 Suppl 3:42–47. doi: 10.1111/j.1464-410X.2009.08370.x. [DOI] [PubMed] [Google Scholar]
- 10.Dong JY, Zhang YH, Qin LQ. Erectile dysfunction and risk of cardiovascular disease: meta-analysis of prospective cohort studies. J Am Coll Cardiol. 2011;58:1378–1385. doi: 10.1016/j.jacc.2011.06.024. [DOI] [PubMed] [Google Scholar]
- 11.Gebremedhin HT, Mezgebo HM, Geberhiwot GT, Gebru TT, Tesfamichael YA, Ygzaw HB, et al. Erectile dysfunction and its associated factors among the male population in Adigrat Town, Tigrai Region, Ethiopia: a cross-sectional study. PLoS One. 2021;16:e0242335. doi: 10.1371/journal.pone.0242335. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Leitão AE, Vieira MCS, Pelegrini A, da Silva EL, Guimarães ACA. A 6-month, double-blind, placebo-controlled, randomized trial to evaluate the effect of Eurycoma longifolia (Tongkat Ali) and concurrent training on erectile function and testosterone levels in androgen deficiency of aging males (ADAM) Maturitas. 2021;145:78–85. doi: 10.1016/j.maturitas.2020.12.002. [DOI] [PubMed] [Google Scholar]
- 13.El Assar M, Ruiz de Adana JC, Angulo J, Pindado Martínez ML, Hernández Matías A, Rodríguez-Mañas L. Preserved endothelial function in human obesity in the absence of insulin resistance. J Transl Med. 2013;11:263. doi: 10.1186/1479-5876-11-263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Pearson AL, Zhou Y, Buxton RT, Horton TH, Pfeiffer KA, Beyer KMM. The effects of contemporary redlining on the mental health of Black residents. SSM Popul Health. 2023;23:101462. doi: 10.1016/j.ssmph.2023.101462. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.McMaughan DJ, Oloruntoba O, Smith ML. Socioeconomic status and access to healthcare: interrelated drivers for healthy aging. Front Public Health. 2020;8:231. doi: 10.3389/fpubh.2020.00231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Khaing W, Vallibhakara SA, Attia J, McEvoy M, Thakkinstian A. Effects of education and income on cardiovascular outcomes: a systematic review and meta-analysis. Eur J Prev Cardiol. 2017;24:1032–1042. doi: 10.1177/2047487317705916. [DOI] [PubMed] [Google Scholar]
- 17.Kershaw KN, Droomers M, Robinson WR, Carnethon MR, Daviglus ML, Monique Verschuren WM. Quantifying the contributions of behavioral and biological risk factors to socioeconomic disparities in coronary heart disease incidence: the MORGEN study. Eur J Epidemiol. 2013;28:807–814. doi: 10.1007/s10654-013-9847-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Gimbrone MA, Jr, García-Cardeña G. Endothelial cell dysfunction and the pathobiology of atherosclerosis. Circ Res. 2016;118:620–636. doi: 10.1161/CIRCRESAHA.115.306301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Liu G, Zhang Y, Zhang W, Wu X, Jiang H, Huang H, et al. Novel predictive risk factor for erectile dysfunction: Serum high-sensitivity C-reactive protein. Andrology. 2022;10:1096–1106. doi: 10.1111/andr.13206. [DOI] [PubMed] [Google Scholar]
- 20.Bergmann CB, Beckmann N, Salyer CE, Crisologo PA, Nomellini V, Caldwell CC. Lymphocyte immunosuppression and dysfunction contributing to Persistent Inflammation, Immunosuppression, and Catabolism Syndrome (PICS) Shock. 2021;55:723–741. doi: 10.1097/SHK.0000000000001675. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Lužnik Z, Anchouche S, Dana R, Yin J. Regulatory T cells in angiogenesis. J Immunol. 2020;205:2557–2565. doi: 10.4049/jimmunol.2000574. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.de la Guía-Galipienso F, Martínez-Ferran M, Vallecillo N, Lavie CJ, Sanchis-Gomar F, Pareja-Galeano H. Vitamin D and cardiovascular health. Clin Nutr. 2021;40:2946–2957. doi: 10.1016/j.clnu.2020.12.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kurt HA, Demirci E, Alan C. Endothelial dysfunction and ischemia-modified albumin levels in males with diabetic and nondiabetic erectile dysfunction. Dis Markers. 2022;2022:3661822. doi: 10.1155/2022/3661822. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Defeudis G, Mazzilli R, Tenuta M, Rossini G, Zamponi V, Olana S, et al. Erectile dysfunction and diabetes: a melting pot of circumstances and treatments. Diabetes Metab Res Rev. 2022;38:e3494. doi: 10.1002/dmrr.3494. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Kloner R. Erectile dysfunction and hypertension. Int J Impot Res. 2007;19:296–302. doi: 10.1038/sj.ijir.3901527. [DOI] [PubMed] [Google Scholar]
- 26.Massi DG, Mountap GN, Moby HE, Ngalle FGE, Mouliom S, Doumbe J, et al. Poststroke erectile dysfunction in cameroon: prevalence, associated factors, and quality of life. Stroke Res Treat. 2021;2021:9988841. doi: 10.1155/2021/9988841. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Fan D, Mao W, Wang G, Shi H, Wu Z, Xie J, et al. Study on the relationship between sex hormone changes and erectile dysfunction in patients with chronic prostatitis/chronic pelvic pain syndrome. Ann Palliat Med. 2021;10:1739–1747. doi: 10.21037/apm-20-985. [DOI] [PubMed] [Google Scholar]
- 28.Silva AB, Sousa N, Azevedo LF, Martins C. Physical activity and exercise for erectile dysfunction: systematic review and meta-analysis. Br J Sports Med. 2017;51:1419–1424. doi: 10.1136/bjsports-2016-096418. [DOI] [PubMed] [Google Scholar]
- 29.Franklin BA, Rusia A, Haskin-Popp C, Tawney A. Chronic stress, exercise and cardiovascular disease: placing the benefits and risks of physical activity into perspective. Int J Environ Res Public Health. 2021;18:9922. doi: 10.3390/ijerph18189922. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Baseline characteristics of participants in the training and validation sets, weighted
Logistic regression analyses of the included factors
Diagnostic efficacy of the nomogram model for assessing erection dysfunction risk
The web-based online dynamic nomogram tool. PIR: poverty income ratio.
Data Availability Statement
Data can be downloaded from the cdc.gov website.