Late Fecal Incontinence After High-Dose Radiotherapy for Prostate Cancer: Better Prediction Using Longitudinal Definitions

Purpose To model late fecal incontinence after high-dose prostate cancer radiotherapy (RT) in patients accrued in the AIROPROS (prostate working group of the Italian Association of Radiation Oncology) 0102 trial using different endpoint definitions. Methods and Materials The self-reported questionnaires (before RT, 1 month after RT, and every 6 months for ≤3 years after RT) of 586 patients were available. The peak incontinence (P_INC) and two longitudinal definitions (chronic incontinence [C_INC], defined as the persistence of Grade 1 or greater incontinence after any Grade 2-3 event; and mean incontinence score [M_INC], defined as the average score during the 3-year period after RT) were considered. The correlation between the clinical/dosimetric parameters (including rectal dose–volume histograms) and P_INC (Grade 2 or greater), C_INC, and M_INC of ≥1 were investigated using multivariate logistic analyses. Receiver operating characteristic curves and the area under the curve were used to assess the predictive value of the different multivariate models. Results Of the 586 patients, 36 with a Grade 1 or greater incontinence score before RT were not included in the present analysis. Of the 550 included patients, 197 (35.8%) had at least one control with a Grade 1 or greater incontinence score (M_INC >0). Of these 197 patients, 37 (6.7%), 22 (4.0%), and 17 (3.1%) were scored as having P_INC, M_INC ≥1, and C_INC, respectively. On multivariate analysis, Grade 2 or greater acute incontinence was the only predictor of P_INC (odds ratio [OR], 5.9; p = .0009). Grade 3 acute incontinence was predictive of C_INC (OR, 9.4; p = .02), and percentage of the rectal volume receiving >40 Gy of ≥80% was predictive of a M_INC of ≥1 (OR, 3.8; p = .008) and of C_INC (OR, 3.6; p = .03). Previous bowel disease, previous abdominal/pelvic surgery, and the use of antihypertensive (protective factor) correlated highly with both C_INC and M_INC ≥1. The predictive values of the models for C_INC (area under the curve, 0.83) and M_INC ≥1 (area under the curve, 0.73) were greater than the ones for P_INC (area under the curve, 0.62) and more reliable ( p = .0001–.0003 against p = .02). Nomograms for the two longitudinal definitions were derived. Conclusions The longitudinal definitions of fecal incontinence (C_INC and M_INC ≥1) were helpful in accounting for both the persistence and the severity of the incontinence. A significant fraction of peak events was consequential to acute incontinence, and a longer duration of symptoms mainly depended on the rectal dose bath (percentage of rectal volume receiving >40 Gy), and pretreatment clinical factors.