Abstract Book

ESTRO 37

S416

Purpose or Objective This study was designed to investigate the impact of Intensity Modulated Radiation Therapy (IMRT) on overall survival (OS) in patients treated with chemoradiation (CRT) for anal cancer (AC) Material and Methods We performed a case-control, propensity score (PS) matched analysis of the National Cancer Data Base (NCDB) of patients diagnosed with non-metastatic AC from 2003 to 2013. Only patients receiving concurrent CRT were included. Patients were stratified into two groups based on the RT technique: IMRT vs. non-IMRT. Multivariate analysis (MVA) and Kaplan-Meier (KM) plots for OS were obtained for the matched and unmatched A total of 8,108 patients diagnosed between 2004 and 2013 were eligible for the study, of which 3,307 (40.8%) and 4,801 (59.2%) were in the IMRT and non-IMRT groups, respectively. Median follow-up for all patients was 54.4 months. After PS matching, MVA for OS showed that IMRT was associated with improved OS compared to non-IMRT (HR 0.83, CI 0.74-0.94; p=0.002). Adjusted KM analysis showed that the 5-yr OS for patients treated with IMRT was 74.6% vs 70.5% (p=0.002). Conclusion To our knowledge, this is the largest study to date that evaluates the impact of IMRT on OS for patients with AC. Our investigation shows that IMRT based concurrent CRT for non-metastatic AC is associated with improved survival when compared to similar patients treated with non-IMRT based therapy. In the absence of randomized evidence, our analysis might provide additional support for increasing the use of IMRT for patients with AC receiving concurrent CRT. groups. Results PO-0802 Lymph flow guided radiotherapy in patients with cervical cancer S. Novikov 1 , P. Krzhivitskiy 1 , S. Kanaev 1 , I. Berlev 1 , M. Kargopolova 1 , Z. Ibragimov 1 , M. Bisyarin 1 1 Prof. N.N. Petrov Research Institute of Oncology, Radiation Oncology and Nuclear Medicine, St. Petersburg, Russian Federation Purpose or Objective Idea of “lymph flow guided radiotherapy” represents logical continuation and development of sentinel lymph nodes (LN) hypothesis that proposed sequential fashion of regional dissemination. The corn stone of this concept is visualization of individual patterns of lymph flow from the primary tumour and subsequent irradiation of those lymph nodes that received lymph fluid from the primary cancer. Purpose: to evaluate patterns of lymph flow from primary lesions in patients with cervical cancer and to determine how useful can be this information for SPECT-CT visualization of sentinel lymph nodes (LNs) was performed in 36 primary patients with IB-IIB cervical cancer. Acquisition started 120-240 min after peritumoral injections of 99mTc-radiocolloids (150-300MBq in 0.4- 1ml). All lymph nodes that accumulated radiocolloids were regarded as sentinel LN. They were assigned to 6 nodal groups against their anatomic location in relation to blood vessels: obturator, internal, external and common iliac, pre-sacral, para-aortic. Additionally we determined type of lymph flow (mono-, bilateral) and lymph flow patterns (supra-ureteral paracervical, infra- radiotherapy planning. Material and Methods Poster: Clinical track: Gynaecological (endometrium, cervix, vagina, vulva)

to assess treatment response with certainty, an histological examination of the TME is required. Predicting complete response from pre-treatment features could represent a major step towards conservative treatment. We developed a Deep Learning model to predict the outcome after chemoradiation of Patients with a T2-4 N0-1 rectal adenocarcinoma treated between June 2010 and October 2016 with neo-adjuvant chemoradiation (45 to 50.4 Gy) with Capecitabine (800 mg/m 2 twice a day) from three academic institutions were included. Chemoradiation was performed in our department and surgery was then performed 6 to 8 weeks by the surgery department in each initial institution. All clinical and treatment data was integrated in our clinical data warehouse, from which we extracted the following features: age, sex, smoking status, tumour differentiation and size, T and N stages, baseline haemoglobin, neutrophils and lymphocytes counts, total dose to the GTVp and CTV, dose per fraction and treatment length. A Deep Neural Network was created with TensorFlow, an open source software library for machine learning using the DNNClassifier estimator. The network included 14 inputs to the model and three layers with respectively 10, 20 and 10 neurons per layers. The output predicted complete response, as assessed on the pathological examination of the TME. A two-fold cross validation was performed. Results Ninety-five patients were included in the final analysis. Median follow-up was 16 months (range: 3-65). There were 51 males (54%) and 44 females (46%). Median age was 66 years old (32-84). Median tumour size was 48 mm (15-130). There were 9 T2 (9%), 75 T3 (79%) and 11 T4 (12%) tumours. Median baseline haemoglobin, neutrophils and lymphocytes counts were 13.6 g/dl (9.7-17.5), 1734/mm 3 (336-3760) and 4050/mm 3 (1100-11160) respectively. Median doses were 50.4 Gy (45-50.4) to the GTVp and 45 Gy to the CTV. Median dose per fraction was 2 Gy (1.8-2.25 Gy). Median treatment length was 39 days (32-69). 22 patients (23%) had pathologic complete response after chemoradiation. Two patients had a local relapse (2.1%) and 7 a distant relapse (7.3%) during follow-up. Two thirds of the study population, i.e. 67 patients’ files, were used as training set. The classifier was tested on one third of the study population. The DNNClassifier predicted complete response on 14 clinical and treatment features with an 80.12% accuracy. Conclusion Our model correctly predicted complete response after neo-adjuvant rectal chemoradiotherapy in 80.12% of the patients of this multicentre cohort. This approach could help guide treatments to avoid any unnecessary surgical treatment and perform a conservative treatment with confidence. Validation of this model on another cohort is warranted. PO-0801 Intensity Modulated Radiation Therapy is Associated with Improved Survival in Anal Cancer J. Jhaveri 1 , L. Rayfield 2 , Y. Liu 3 , M. Chowdhary 4 , R. Cassidy 5 , S. Tian 5 , T. Gillespie 6 , P. Patel 5 , J. Landry 5 , K. Patel 7 1 Emory University, Radiation Oncology, Decatur, USA 2 Emory University, Rollins School of Public Health, Atlanta, USA 3 Emory University, Biostatistics and Bioinformatics, Atlanta, USA 4 Rush University, Radiation Oncology, Chicago, USA 5 Emory University, Radiation Oncology, Atlanta, USA 6 Emory University, Department of Surgery, Atlanta, USA 7 Yale University, Radiation Oncology, New Haven, USA rectal adenocarcinoma. Material and Methods