Predicting Late Fecal Incontinence Risk After Radiation Therapy for Prostate Cancer: New Insights From External Independent Validation

Purpose: This study aimed to validate a previously published predictive model for late fecal incontinence (FI) in a contemporary population of prostate cancer patients treated with radical radiation therapy. Methods and Materials: The validation included patients treated with intensity-modulated radiation therapy (IMRT) (2010-2014). Prescribed dose range was 6580 Gy, including conventional and moderate hypo-fractionated treatments. Rectal toxicity was scored using LENT/SOMA, a minimum 2-year follow up was considered. We chose to validate the model published by Rancati et al for predicting chronic FI, developed on a 3-dimensional conformal radiation therapy (3DCRT) population. It considered a longitudinal endpoint defined as the average toxicity grade during the follow up. This continuous endpoint was dichotomized using a cut-off value of mean FI grade > 1. The model included mean rectal dose (Dmean), previous diseases of the colon (COLO) and previous abdominal surgery (SURG). Doses were corrected to 2 Gy/fraction using the linear-quadratic model and applying alpha/beta ratio = 4.8 Gy. Results: 228 patients constituted the validation population. A mean FI grade > 1 was scored in 25 patients (11%). Logistic regression confirmed risk factors reported in the literature, with similar odds ratios (ORs) for Dmean (1.04 +/- 0.03 vs 1.06 +/- 0.04) and SURG (1.9 +/- 1.7 vs 1.6 +/- 1.45); COLO was not confirmed. Consequently, the predictive models including Dmean/Dmean + SURG were evaluated using calibration plots. Both showed a clear discriminative trend, but the absolute observed toxicity rates were underestimated (ie, absolute predicted rates were always lower than corresponding absolute observed rates). This result was consistent with an unexpected effect of hypofractionation (OR=2.20, conventional=8.1% vs hypofractionated = 17.4%) beyond the standard correction using linear-quadratic model. Nevertheless, the FI rate in the conventionally treated group was almost double the rate observed in the previously studied cohort (4.3% vs 8.1%). Conclusions: The study confirms previously published results indicating that abdominal surgery and rectal mean dose are risk factors for late FI. Calibration plots highlight a possible role of hypofractionation beyond linear-quadratic correctio