Metamodeling of late rectal bleeding in patients undergoing radiotherapy for prostate cancer.

Abstract

61 Background: Predictive models for late rectal bleeding (LRB) after radiotherapy (RT) for prostate cancer were established by several studies, with good performance on development population (POP). Nonetheless, they were found to be unsatisfactory in their generalization to independent validation cohorts. Aim of the study is to build a metamodel for grade 2-3 (G23) and grade 3 (G3) LRB (LENT/SOMA) starting from literature evidence and to validate it on a new POP. Methods: The metamodel was constituted by clinical/dosimetric features and by their coefficients. Available predictive models for LRB were selected by literature search. Models including rectal Equivalent Uniform Dose (EUD) with/without patient-related dose-modifying factors were retrieved. The process allowed identification of “evidence-based” clinical/dosimetric features associated to LRB. Coeff. for identified features were resolved by weighted mean of published values, using their standard deviation as weight and then inserted in a sigmoid-shaped logit-EUD model. The resulting metamodel was validated on a pooled POP of radically treated pts (3DCRT & IMRT). Performance was assessed through calibration. Results: The search identified rectal EUD, previous abdominal surgery (SURG), hormone therapy (HT) and use of cardiovascular drugs (CARDIO) as relevant features (coeff. presented in table). Validation POP included 1591 pts with 240 (15%) LRBG23 and 98 (6.2%) LRBG3 pts. Calibration was very good: LRBG23 slope = 1.4 with R2 = 0.86; LRBG3 slope = 1.1 with R2 = 0.96. Conclusions: A metamodel for prediction of LRB has been developed from literature, including all currently available information on association between LRB and clinical/dosimetric factors. The model was successfully validated on a large POP, proving to be a valuable tool for predicting LRB before RT. [Table: see text]