ESTRO 38 Abstract book

S327 ESTRO 38

Conclusion This preliminary study demonstrates the possibility of introducing patient-specific MB information into NTCP model, through use of unsupervised clustering to exploit the whole MB information (176 classes) without dramatically increasing the number of features to be included in the model. Results obtained in a small sample of PCa pts seem promising in indicating that pts with/without radio- induced acute tox have different constitutional gut MB profiles. Introduction of MB clustering into NTCP highly improves model performance.If confirmed in the whole accrued population, this could represent an important finding, not only for prediction of tox but also for design of possible interventional trials to reduce tox by modification of gut microbiota before RT start. OC-0617 Work interruptions in radiotherapy and their impact on patient safety. S. Cucchiaro 1 , M. Delgaudine 2 , N. Gourmet 1 , P. Coucke 1 1 C.H.U. - Sart Tilman, Radiotherapy Department, Liège, Belgium ; 2 C.H.U. - Sart Tilman, STA Quality Department, Liège, Belgium Purpose or Objective Task Interruptions (TI) are common in hospitals and especially in Radiation Therapy (RT) Department. The main objective is to evaluate the number and characteristics of those interruptions in our RT department in order to propose solutions to reduce them to a minimum. There are many TI sources (phone calls, talk, noise,...), often of short duration and most often induced by team members. They affect attention, can generate stress, as well as treatment errors. The idea is to understand the TI, when they occur, where and who produces them and how the work is resumed. The goal is to implement prevention barriers to tasks identified as critical in the process and to allow a safe recovery of the task. Material and Methods Following an adverse event where an identified root cause was TI, we decided to carry out an exploratory study whose main objective is to be able to identify the sources of TI in the treatment unit as well as the timing of these interruptions. We used an observation grid (figure 1) with several items such as time, duration, nature, reason and consequence of the interruption. The TI were observed at the 5 service treatment units. The collected data has been analyzed in order to confirm that there are TI at the Radiation Therapists (RTTs) during a radiotherapy treatment, and that these TI can be a source of errors. Proffered Papers: RTT 6: Education and quality management for optimising patient care

Purpose or Objective A mono-institutional trial was set up in 2017 to investigate the role of gut/saliva microbiota (MB) in driving radio- induced toxicity (tox) after RT for prostate (PCa) and head&neck cancers. We here focus on introduction of information on gut MB into a normal tissue complication probability model (NTCP) for acute gastro-intestinal (GI) tox in the PCa cohort. Material and Methods 130 consecutive PCa patients (pts) receiving conventional (78Gy @2Gy/fr) or moderately hypofractionated (65Gy @2.6Gy/fr) VMAT+IGRT in 5 fr/week were enrolled. A detailed evaluation was done pre-RT, during RT and at RT end, including gut MB measurement. Stool samples were collected using gut-OMNIgene devices (Oragene). DNA extraction was carried out using the QIAamp-DNA- Stool-Mini-Kit (Qiagen). The bacterial 16S ribosomal-RNA reads were analyzed with the QIIME software and pooled in Operational Taxonomic Units (OTUs) with Uclust software. Grade 2 (G2) CTCAE acute G Itox was the primary endpoint. For this preliminary evaluation 20 pts were selected: 10 with G0 and 10 with G2 tox.Unsupervised clustering (fuzzy c-means algorithm) was used to separate the pts into 2 MB clusters, based on relative abundance of OTUs at bacterial class level in MB before RT start. Information on MB clustering was introduced as a dose- modifying factor (dmf) into a logit NTCP model (characterized by D50=dose associated to 50% tox probability and steepness parameter k). Mean dose to the rectum was chosen as dosimetric predictor (as already found in the literature). Results Unsupervised clustering identified 13 pts included in a first MB cluster (A) and 7 in a second cluster (B), average OTU composition for pts in clusters A and B are presented in figure 1. 4/13 (31%) and 6/7 (86%) pts with tox were found in clusters A and B, respectively (p=0.019). MB clustering resulted in AUC=0.75 (95%CI=0.51-0.91) for tox discrimination.NTCP model including only mean rectal dose had D50=49Gy, k=16 (AUC=0.85, 95%CI=0.62-0.97). When clustering was introduced, k=20.5, D50=42Gy for cluster A vs D50=32Gy for cluster B were found: MB clustering dmf (B vs A)=0.76 (AUC=0.87, 95%CI=0.65-0.98), with significant improvement in goodness of fit and calibration. Model curves are reported in figure 2.

Results A 5-hour observation at each treatment unit was performed for a total of 25 hours. 145 TI have been recorded. We were able to highlight an interruption every