Development of clinical models for predicting erectile function after localized prostate cancer treatment.

Document Type


Publication Date



CORE, Urology

Journal Title

International journal of urology : official journal of the Japanese Urological Association

MeSH Headings

Aged, Erectile Dysfunction, Forecasting, Humans, Male, Middle Aged, Models, Theoretical, Penile Erection, Prognosis, Prostatectomy, Prostatic Neoplasms, Quality of Life


OBJECTIVES: To develop clinical prediction models estimating the probability of maintaining erections adequate for intercourse 2 years after prostate cancer treatment, based on pretreatment characteristics.

METHODS: Study participants consisted of prostate cancer patients with localized disease and functional erections before undergoing surgery (n = 536) or radiation therapy (n = 240) at a single USA institution. Baseline patient- and treatment-related data were collected from a clinical database and through chart review. Erectile function at 2 years post-treatment was prospectively assessed through a self-administered single-item measure. Multivariate logistic regression using backward selection was used to derive clinical prediction models to predict erectile function at 2 years for surgery and radiation therapy patients; the models were internally validated using bootstrapping methods.

RESULTS: The final prediction model for surgery patients included the predictor variables of age, body mass index, smoking, diabetes, hypertension and nerve-sparing procedures, whereas the model for radiation therapy patients included hypertension, risk category and radiation technique. The new models showed acceptable calibration and discrimination: c-statistic = 0.71 (95% confidence interval 0.68-0.76) for surgery and 0.66 (95% confidence interval 0.61-0.74) for radiation therapy models.

CONCLUSIONS: New clinical prediction models based on patient and treatment characteristics show promising accuracy in predicting erectile function at 2 years in patients treated with surgery and radiation for localized prostate cancer. More work is required to confirm and validate these models in different patient populations.



First Page


Last Page