Construction and internal validation of a prediction nomogram for acquired premature ejaculation (APE) in PE patients

Abstract Background A predictive model for acquired premature ejaculation (APE) in PE patients has not yet been established. Objectives This study was aimed at determining which factors were independently associated with the possibility of predicting APE in PE patients, and whether an effective pre‐treatment nomogram for predicting their individual chances of being APE in PE patients can be developed. Materials and methods We analyzed the medical histories of 915 PE patients diagnosed at Xijing Hospital (Xi'an, China) and Northwest Women's and Children's Hospital (Xi'an, China) between May 2019 and May 2020. The diagnostic nomogram was developed using a multivariate logistic regression model by integrating selected significant variables determined through univariate analysis. Receiver operating characteristic curves were used to measure the predictive accuracy of the nomogram and its constituted variables, and calibrations were performed by making a comparison of nomogram‐predicted probability with actual rate of APE. Results The independent predictors for APE that were identified include Age, Intra‐vaginal Ejaculation Latency Time (IELT), Frequency of sexual desire (FSD), and Eysenck Personality Questionnaire‐Revised Short Scale for Chinese (psychoticism) [EPQ‐RSC(P)] scores. The predictive accuracy of the nomogram was 0.782 (95% CI: 0.723–0.841). Also, excellent agreement was demonstrated between the nomogram‐predicted probability and the actual rate of APE. Discussion and conclusion We identified 4 independent predictors for APE and demonstrated the potential significant differences in psychoticism between LPE and APE patients. This was the first internally validated predictive APE nomogram where good discrimination and calibration were applied, and it offers a promising role in clinical practice. More studies are necessary for verification of its universal applicability.


| INTRODUC TI ON
Premature ejaculation (PE) is characterized as the most frequent sexual complaint, demonstrating a prevalence rate of 20-30%. [1][2][3] In 2014, the International Society for Sexual Medicine (ISSM) defines PE and divides it into acquired PE (APE) and lifelong PE (LPE). 4 Variable and subjective PEs have also been proposed for the classification of PE syndromes, but they have always been considered provisional. 5 A Turkish study found that 20% of subjects complained of PE, and the rate of LPE was 2.3%, APE was 3.9%, variable PE was 8.5%, and subjective PE was 5.1%. 6 Approximately 5% of prevalence of APE and LPE among the general population is consistent with epidemiological data, thus indicating that in nearly 5% of the population ejaculation latency is less than 2 minutes. 7 It is believed that the pathogenesis of LPE is mainly caused by neurobiological and genetic issues, 8 whereas APE is usually associated with sexual performance anxiety, 9 psychological or relationship problems, 9 erectile dysfunction (ED), 2 prostatitis, 10 and thyroid diseases, 11 among others. It is imperative to determine the subtypes of PE and discuss patient expectations in detail prior to treatment.
Although treatment of the underlying causes, such as ED, is the primary goal for APE, drug therapy is considered the first-line treatment for LPE. 7 Diagnosis and classification of PE are mainly based on the medical and sexual histories of patients, 7 and intra-vaginal ejaculation latency time (IELT) plays an important role in the classification of PE. Nomograms, graphical representation of multivariable prediction models, has been applied in the field of andrology and shows a good accuracy. 12 However, nomogram prediction for APE in PE patients has not yet been established to date and it may be useful to establish a nomogram to predict APE in PE patients.
The predictors of PE subtypes are currently controversial, and the nomogram that predicts APE in PE patients has not yet been established. As such, this study aimed at determining which factors were independently associated with the possibility of predicting APE in PE patients, and whether an effective pre-treatment nomogram can be developed to predict their individual chance of being APE in PE patients.

| Sample size assessment
Referring to previous studies, 13 the P (Proportion) of APE in PE (including APE and APE) is at 60%; if the allowable error is 10% of P, it is at 0.06% and the Confidence Interval Width (Two-Sided) is 0.12 and Alpha is 0.05. After the calculation of PASS software, version 11.0.7 (https://www.ncss.com/softw are/pass/), the minimum sample size was 271 which was a sufficient sample size for this study.

| Patient population
We conducted an observational cross-sectional study between May 2019 and May 2020 at Xijing Hospital (Xi'an, China) and Northwest Women's and Children's Hospital (Xi'an, China). A total of 915 PE patients (327 from Xijing Hospital and 588 from Northwest Women's and Children's Hospital) were included based on the diagnostic criteria for ISSM. The inclusion criteria involved a stable sexual relationship of at least 6 months and at least one sexual attempt within the last month. All participants were heterosexual and practiced vaginal sexual intercourse. A total of 252 LPE and 663 APE patients were identified through evaluation of the sexual and medical histories of patients by 2 experienced andrologists. All procedures performed in studies were in accordance with the 1964 Helsinki Declaration and its later amendments, or comparable ethical standards, and with the ethical standards of the institutional and national research committee. The Xijing Hospital Ethics Committee approved the study (KY20192108-F-1) and it was registered in Clinical Trials (https://clini caltr ials.gov/, NCT04235192). All participants provided written informed consent.

| Study covariates
The 5-item International Index of Erectile Function (IIEF-5) 14,15 was used to evaluate erectile function, which has been widely used around the world including China. 16 A patient was considered to have ED when the total IIEF-5 score was below 22. Ejaculation function was assessed with the Premature Ejaculation Diagnostic Tool (PEDT), 17,18 which was linguistically validated in China in a previous study. 19 Total of PEDT >10 indicates PE. IELT is the duration between vaginal penetration to ejaculation, and this was self-reported in our study. History of masturbation was also assessed and frequency of sexual desire (FSD) was reported as "> 50% of the time" or "≤ 50% of the time". We used the Patient Health Questionnaire-9 (PHQ-9) 20 to assess the symptoms of depression, while the Generalized Anxiety Disorder (GAD-7) 21 was used to evaluate symptoms of anxiety, and both of them were validated in the Chinese population. 22,23 A total PHQ-9 or GAD-7 score ≥5 indicates the existence of depression or anxiety symptoms, respectively.
In our study, the Eysenck Personality Questionnaire-Revised Short Scale for Chinese (EPQ-RSC) version was used to evaluate patients' personality status, which was translated and revised by Qian et al. in

| Statistical analysis
R for Windows, version 3.6.1 (http://www.r-proje ct.org/) with the "rms" package was used for all statistical analyses. Categorical data were presented as frequencies (proportions) and were compared between subgroups by the Chi-squared test. Continuous variables were presented as medians (ranges) and were compared by the Mann-Whitney U test. We performed a univariate logistic regression model to identify statistically significant variables, which were then incorporated in the multivariate model. The nomogram was developed using this final multivariate logistic regression model by integrating selected significant variables.
We measured the discriminatory capability of the nomogram according to the area under the receiver operating characteristic curve (AUC). The values ranged from a maximum of 1.0, which indicated a perfect discrimination, to a minimum of 0.5, which indicated a random chance. We generated calibration plots to test the consistency of prediction between the nomogram-predicted probability and the actual outcome. In these calculations, bootstrapping with 1,000 resamples was performed.
Statistical significance was considered as P < 0.05.

| Independent predictors for APE
Age, PEDT, IELT, FSD, and EPQ-RSC(P) were significantly associated with APE (P < 0.05) according to the univariate analysis. PEDT was excluded in the multivariate analysis due to significant negative correlation shown between PEDT and IELT (r = −0.362, P < 0.001) in the Spearman correlation analysis. Results of the multivariate analysis showed that age, IELT, FSD, and EPQ-RSC(P) were all independent predictors for APE in PE patients ( Table 2). Characteristics of independent predictors between the LPE and APE group were shown in Table 3.    Figure 3A show that the accuracy of our nomogram was 0.782 (95% CI: 0.723-0.841), which was higher than models established by integrating any 1, 2, or 3 of the variables among age, IELT, FSD, and EPQ-RSC(P). Moreover, we achieved an excellent agreement between nomogram prediction and the actual situation in the calibration plot ( Figure 3B).

| DISCUSS ION
It is necessary to distinguish these PE subtypes in clinical practice,    Although more studies are required to verify its universal applicability in clinical practice, our nomogram is promising and could be clinically valuable.

ACK N OWLED G EM ENTS
The

CO N FLI C T S O F I NTE R E S T
None of the authors have any conflicts of interest to disclose.