ESTRO 2023 - Abstract Book

S1227

Digital Posters

ESTRO 2023

Conclusion Urinary toxicity after prostate salvage stereotactic reirradiation is the main toxicity, with a 2 year grade ≥ 2 rate of 55%. It is mostly related to anticoagulant treatment and the V37 in the bladder wall.

PO-1509 Predictors of Patient-Reported decline in Social activity 2-years after WPRT for prostate cancer

M. Pavarini 1 , G. Girelli 2 , E. Villa 3 , B. Avuzzi 4 , A. Bresolin 1 , E. Garibaldi 5 , A. Faiella 6 , A. Magli 7 , D. Cante 8 , J.M. Waskiewicz 9 , T. Statuto 10 , L. Ferella 11 , B. Noris Chiorda 4 , L. Rago 12 , P. Ferrari 13 , M. Gatti 14 , C. Piva 2 , T. Rancati 15 , R. Valdagni 15 , V.L. Vavassori 3 , F. Munoz 11 , G. Sanguineti 6 , N.G. Di Muzio 16 , C. Fiorino 1 , C. Cozzarini 16 1 IRCCS San Raffaele Scientific Institute, Medical Physics, Milano, Italy; 2 Ospedale degli Infermi, Radiotherapy, Biella, Italy; 3 Cliniche Gavazzeni-Humanitas, Radiotherapy, Bergamo, Italy; 4 Fondazione IRCCS Istituto Nazionale dei Tumori, Radiotherapy, Milano, Italy; 5 A.O. SS. Antonio e Biagio, Radiotherapy, Alessandria, Italy; 6 Istituto Nazionale dei Tumori "Regina Elena", Radiotherapy, Roma, Italy; 7 Azienda Ospedaliero Universitaria S. Maria della Misericordia, Radiotherapy, Udine, Italy; 8 Ospedale di Ivrea, Radiotherapy, Ivrea, Italy; 9 Comprensorio Sanitario di Bolzano, Radiotherapy, Bolzano, Italy; 10 Ospedale San Carlo, Biology, Potenza, Italy; 11 Ospedale Regionale Parini-AUSL Valle d’Aosta, Radiotherapy, Aosta, Italy; 12 Ospedale San Carlo, Radiotherapy, Potenza, Italy; 13 Comprensorio Sanitario di Bolzano, Medical Physics, Bolzano, Italy; 14 Istituto di Candiolo - Fondazione del Piemonte per l'Oncologia IRCCS, Radiotherapy, Candiolo, Italy; 15 Fondazione IRCCS Istituto Nazionale dei Tumori, Prostate Cancer Program, Milano, Italy; 16 IRCCS San Raffaele Scientific Institute, Radiotherapy, Milano, Italy Purpose or Objective Intestinal toxicity (IT) after radiotherapy (RT) with pelvic nodal irradiation (PNI) for prostate cancer (PCa) may have a non negligible impact on patients’ (pts) health-related quality of life (HRQoL). This work aims at identifying predictors of self perceived decline of Social Domain 2 years after RT. Materials and Methods A multi-Institute study recruited 886 pts between 2011 and 2021 and prospectively collected the Inflammatory Bowel Disease Questionnaire (IBDQ) at baseline, at RT mid-point and end, at 3 and 6 months and every 6 months up to 5 years. The 24-item Eysenck Personality Questionnaire Revised-Abbreviated (EPQR-A) and the 14-item Hospital Anxiety and Depression Scale (HADS) were also filled in by pts at baseline. The 32 IBDQ items (on a 1 to 7 scale, 1 corresponding to worse symptoms) pertain to four Domains: PNI-induced Bowel symptoms (BOWEL) and their impact on Emotional (EMOT), Social (SOCIAL) and Systemic (SYST) domains of HRQoL. This analysis refers to 554 pts with complete baseline, 2-years (or 30 or 18 months, if missing) and at least one between RT mid-point and end questionnaires. Thirty-one clinical, dosimetric and patient-reported covariates were considered for the prediction of the worst decline of Social Domain 2 years after RT as compared to baseline. The first decile of scores distribution was the event for Logistic Regression (LR). Pts were first split into training (n=415) and validation (n=139) cohorts. Feature selection was performed by checking for any zero-variance covariate, standardizing the training set using Robust Scaling and balancing it via data augmentation and limiting variable redundancy. Bootstrap resampling (1000 iterations) was carried out to identify the best combination of three most significant features in terms of p-values and AUC.

Goodness of fit was assessed by means of the Hosmer and Lemeshow (HL) test.

Results The decile threshold value identifying pts (43/415 pts, 10.4%) with larger 2-year Social impairment was -1.4. The best three-variable model (p=0.0004) included Bowel Domain acute worsening compared to baseline, (p=0.049), Social Domain’s Score at baseline (p=0.0073) and adjuvant RT intent (p= 0.0059).

Odds Ratios (and 95% C.I.) of the variables were 0.66 (0.44-1.00), 2.07 (1.22-3.53) and 2.59 and (1.32-5.09), respectively.

The goodness of fit of the trained model was satisfactory (HL p-value=0.976).

In Fig. 1. the area under curve (AUC) of the ROC curves of the training (0.809) and the validation (0.721) sets showed the ability of the model's predicted values to distinguish between positive and negative cases.