ESTRO 2020 Abstract Book

S197 ESTRO 2020

conventionally- (n=125, 66.6-79.8 Gy) or hypo- fractionated (n=94, 65.8-76.8 Gy, 2.1-2.7 Gy/fr) RT during years 2011-2017. Their clinical and dosimetric data were prospectively collected within a registered multi- institutional cohort study. The pt-reported ICIQ-SF questionnaire was used as a measure of UI evaluation at baseline (BL) and every six months thereafter. The present analysis focused on three ICIQ-based clinically relevant endpoints: (a) very frequent leakage (FREQ), defined as an ICIQ3 score >3, (b) moderate to severe AMOUNT of urine loss (ICIQ4 >2) and (c) objective severe symptoms (OBJ, ICIQ3+4 scores >5). Predictors of the risk of 2-years endpoints were investigated focusing only on pts exhibiting none to mild incontinence (ICIQ3≤2, ICIQ4≤2, ICIQ3+4≤4) at BL. Firstly, univariate logistic regression analysis (UVA) was performed in order to assess the best fit between the three selected endpoints and 2-Gy equivalent dose (EQD2) calculated for different α/β ratios (0.8, 3, 5 Gy). Then, UVA was used to identify for each endpoint the most significant clinical variables (p-value <0.2), that were entered into a backward stepwise multi-variable (MVA) logistic regression model. Goodness of fit was assessed by means of the Hosmer and Lemeshow test (H&L) and the calibration plot (slope and R 2 ). The Brier scores after 1000 bootstrap resampling were considered as internal validation. Analyses were performed with MedCalc and R software. Results Pts reporting none to mild symptoms, with 68/219 (31%) being “completely dry” (ICIQ3+4=0), at BL experienced (a), (b) and (c) 2-yr endpoints in 41/130 (32%), 40/192 (21%) and 41/129 (32%) of the cases, respectively. The EQD2 for α/β=0.8 Gy showed the best logistic fit (with the highest likelihood). MVA defined the two-variable models reported in Table with ICIQ score at BL as the most significant predictor for all the endpoints. Both FREQ and OBJ were modulated also by EQD2 (0.8), whereas AMOUNT by age. The goodness of the fit and the calibration were both good; the validation confirmed the robustness of the results (Brier score corrected for optimism changed of only 0.005-0.009).

Conclusion Baseline urinary incontinence strongly modulated the 2- year risk of all the endpoints. In addition, FREQ was characterized by a marked dose-effect relationship that strongly influenced also the trend of OBJ (Figure), with FREQ being more reliable than AMOUNT as an objective index. These findings are suggestive of an independent detrimental effect of both fractionation and dose on severe urinary incontinence after PORT, consistently with previously reported results of a large retrospective study.