Predicting late fecal incontinence risk after RT for prostate cancer: External independent validation.

Abstract

116 Background: To validating a predictive model for late fecal incontinence (FI) on a recent population (pop) of prostate cancer patients (pts) treated with radical radiotherapy. NTCP model was derived from literature. Methods: Pop included 267 pts treated with IMRT in 2010-2014. Prescribed dose was between 68 and 80 Gy with conventional and hypo-fractionated (HF, from 2.2 to 2.8 Gy) treatment. Rectal toxicity was scored using the LENT/SOMA. Follow-up (FU) was considered up to 2 years. We chose to validate a model for prediction of chronic FI through multiple measures during FU. Mean FI was defined as the average score during the FU period after RT (Mean incontinence > 1). Literature based multivariate model included: mean rectal dose (Dmean), previous diseases of colon (COLO) and previous abdominal surgery (SURG). Dose distributions were corrected EQD in 2 Gy fractions. Results: 186 pts were available. Mean grade > 1 FI was scored in 18 patients (〜10%). Univariate logistic analysis confirmed the risk factors reported in literature. Similar Odds Ratios (OR) were found for Dmean (1.04vs1.05) and SURG (1.90vs1.50). COLO was not a risk factor for this pop. As consequence, NTCP models including Dmean and Dmean+SURG were evaluated through calibration plot. The models showed a clear trend (increasing observed toxicity rates with predicted risk), but the observed toxicity rates were underestimated (slope〜3, R2〜0.7). Including HF (OR = 2.20, 8.6% vs 17.6%) as a variable into the previous model the calibrations improved significantly (slope〜1, R2〜0.9). Conclusions: The study confirms formerly published results on effect of abdominal surgery and dose to large rectal volumes as potential risk factors for late FI. The overfitting in calibrations could be due to an effect of HF, not included in previous models developed on normofractionated treatments. This effect goes beyond the applied standard correction using LQ model for late effects and also beyond the time recovery correction (slope〜2, R^2〜0.8). Probably we should found a more suitable alpha/beta value for the longitudinal definition (toxicity starting in acute phase and persisting during follow-up) instead of using the assumption settled on incidence of late peak events.