Normal Tissue Complication Probability (NTCP) model for erectile dysfunction (ED) following external beam radiotherapy (RT) for prostate cancer.

Abstract

135 Background: ED remains a common toxicity of prostate RT despite technological advances. Penile bulb (PB) dose has been proposed as a predictor of ED post RT. The main objective of this study was to develop NTCP models for ED. Methods: 162 men treated within the CHHiP IGRT substudy (CRUK/06/16) had baseline clinical data, PB dosimetric data & evaluation of ED using EPIC-26 at least 3 years post RT. Planning CT and reference dose distributions were imported into analysis software (VODCA, MSS GmbH) and PB retrospectively contoured by one clinician. The defined endpoint (severe ED) was a standardised average value of 0-33 for EPIC-26 sexual domain. Predictive models of ED were generated using PB dose in EQD2 (α/β ratio = 3Gy) & clinical data (age, diabetes, hypertension, NCCN risk group, baseline PSA, hormone therapy, IGRT, margin size, PB volume). Multivariate logistic regression method using resampling methods was applied to select model order and parameters. Models were fitted using logistic regression of the form Probability = eA(x)/1+eA(x), where A(x) = constant + sum of (variables * associated regression coefficients). Model performance was evaluated through area under the receiver operating characteristic curve (AUC) and Hosmer-Lemeshow (HL) goodness-of-fit test. Results: 101/162 (62%) men had severe ED with statistically significant difference in PB max and mean dose between those patients with or without severe ED (max: 61.8Gy vs 43Gy & mean: 27.4Gy vs 14Gy respectively; p = 0.001). In the univariate analyses, age, diabetes, risk group, PB mean and max doses were significantly associated with EPIC calculated severe ED. The optimal NTCP model (AUC 0.78; CI 0.71-0.86: p for HL = 0.75) for EPIC calculated severe ED included age, PB mean dose and diabetes where A(x) = -10.13+(0.14*age)+(0.03*PB mean dose)+(2.88 if diabetic). A comparable model using clinician completed outcomes will be reported. Conclusions: This study provides the first known clinical prediction model for ED including PB dose, with good model performance. The determined predictors for the NTCP model of severe ED in this cohort were PB mean dose, age & diabetes. External validation of this model is desirable. Clinical trial information: 97182923.