ESTRO 2020 Abstract Book

S47 ESTRO 2020

Purpose or Objective A Bayesian network (BN) is a probabilistic graphical model that represents a set of variables and their dependencies in a directed acyclic graph (DAG). Within the graph, each node signifies a variable, and the direction of the arrow link between nodes represents the direction of causality, from the cause (parent node) to the effect (child node). Typically, BN structures can be specified by an expert in the domain of interest or consulted from the data via a learning algorithm. However, these methods have some limitations in the medical field. Generating a Bayesian network from a dataset can include casual relationships that are not possible or have no clinical meaning (e.g., causal links like gender to age). On the other hand, a network constructed by an expert might be biased based on prior experts' knowledge and experience of the domain. In this study, we develop and validate a Bayesian network structure to predict local recurrence in patients diagnosed with locally advanced rectal cancer. Material and Methods A retrospective cohort of 8566 diagnosed locally advanced rectal cancer patients from 2004 to 2014 from 14 international trial cohorts are analyzed for this study. A stratified 80-20 percent split per trial cohorts is used to train and validate the developed BN structure, respectively. Continuous variables are categorized, and missing values are considered as a category (Unknown) for all variables but the response. Multiple expert's domain knowledge from different radiotherapy institutions is employed to develop and validate the Bayesian network structure. Experts from Rome defined the casual relationship links between the variables which were independently reviewed by experts from Maastricht and Korea. Our analysis was conducted in R version 3.6.1 using the bnlearn package and GeNIe (Graphical Network Interface) application. The SMOTE function was used to adjust the rare event on the response of interest. The model performance is assessed by generating calibration plots and calculating the area under the receiver operating characteristics curve (AUC) on both training and validation datasets. Results We excluded 1023 (15%) patients from the training cohort and 268 (16%) from the validation cohort due to information missing-at-random. The radiotherapy dose and adjuvant chemotherapy are excluded from the final structure because they did not influence the response of interest (no direct or indirect parental link). Figure 1 shows the developed Bayesian network structure and its calibration performance on the training data.

area under the curve (AUC), and model calibration was evaluated by the Hosmer-Lemeshow (HL) test. Results In the final multivariate models, both GR2A2B and GR2B (Fig. 1) LRB was associated with age, anti-coagulant use, rectum volume, rectum D5%, and V75 of the rectum wall. The Grade 3+ GU endpoint was associated with anti- coagulant use, prostate volume, bladder D5%, equivalent uniform dose (EUD) (n=0.18) of the bladder, and V75 of the bladder wall. The AUC and HL test values were 0.66 and p = 0.49 for the GR2A2B LRB model, 0.73 and p = 0.35 for the GR2B LRB model, and 0.64 and p = 0.09 for the Grade 3+ GU morbidity model (Fig. 2), indicating good model fits.

Conclusion Multivariate NTCP models were developed for both LRB and GU morbidity following passive scattered PT for prostate cancer. All endpoints were associated with many of the same predictors. In particular, anti-coagulant use, V75 of either the full structure or wall, as well as D5%, were important prognostic factors in this cohort. Consequently, these predictors may be used to optimise proton treatment planning. OC-0100 A Bayesian network structure for predicting local recurrence in rectal cancer patients A. Biche 1 , C. Masciocchi 2 , A. Damiani 3 , I. Bermejo 1 , E. Meldolesi 3 , G. Chiloiro 3 , V. Valentini 2,3 , A. Dekker 1 , J. Van Soest 1 1 Department of Radiation Oncology MAASTRO- GROW School for Oncology and Developmental Biology- Maastricht University Medical Centre+- Maastricht- the Netherlands, MAASTRO clinic, Maastricht, The Netherlands ; 2 Fondazione Policlinico Universitario A. Gemelli IRCCS- Roma- Italia, Radiotherapy, Rome, Italy ; 3 Università Cattolica del Sacro Cuore- Roma- Italia, Radiotherapy, Rome, Italy

Table 1 shows the mean Accuracies, AUCs, and confidence intervals when the structure was used to predict two, three, and five-years local recurrence on the training and validation data.