P-140 THE IMPACT OF TOTAL NEO-ADJUVANT TREATMENT ON NONOPERATIVE MANAGEMENT IN PATIENTS WITH LOCALLY ADVANCED RECTAL CANCER: THE EVALUATION OF 66 CASES IN A SINGLE CENTER Handan Tokmak 1 , Oktar Asoglu 2 , Burak Koza 2 . 1 Acibadem University Maslak Hospital, Nuclear Medicine, Istanbul, Turkey; 2 Bosphorus Academy of Clinical Sciences, Surgery, Istanbul, Turkey Background: The study aimed to assess if adherence to a total-neo- adjuvant-treatment (TNT) protocol followed by observation(watch-and- wait) led to the successful nonoperative-management of low-rectal-cancer. Materials and Methods: In this study, patients with primary, resectable-T3- T4, N0eN1 distal-rectal-adenocarcinoma underwent-chemoradiotherapy+ consolidation-chemotherapy (TNT). During the-TNT-period, endoscopy, MRI, and FDG-PET/CT were performed. We allocated patients with complete- clinical-tumor-regression, who underwent endoscopy every two months, MRI every-four-months, and PET/CT every-six-months-after-treatment, to the observation-group( OG). All other patients were referred for surgery. The OG was followed-up. The primary endpoint was local tumor- recurrence after allocation to the OG. Exclusion criteria were the presence of: early-T-stage (cT1-2, N-any) tumors, proximal-tumors (>10 cm from the anal verge), synchronous colorectal or other primary tumors, and polyposis syndrome. Results: 2015-2018, we enrolled in 66-patients. Of 60-patients who were eligible to participate, 39 had complete-clinical-response(cCR) and were allocated to the OG, six underwent local excision (LE), and 15 underwent total-mesorectal-excision (TME). The median follow-up duration was 22 (9- 42) months. The local-recurrence-rate in the OG was 15.3%, and the LE and TME rates were 16.6% and 0%, respectively. All recurrence cases were salvaged through either LE or TME. The-distant-metastasis rate was 5.1%, 16.6%, and 12.5% in the OG, LE, and TME groups, respectively. The endoscopic negative-predictive-value(NPV) was 50%, and the positive-predictive-val- ue(PPV) was 76.9% in the surgerygroup (LE+TME). MRI; NPV-50%, PPV-76.9%. PET/CT; NPV-100%, PPV-93.3%. Six patients(28.57%) from surgery group achieved complete pathological response (cPR). APR indication rates were 70.90% in the NOM group, 9.67% in the LE group and 19.35% in the TME group. Conclusions: Our results indicated a high proportion of selected-rectal- cancers with-cCR after neo-adjuvant-therapy could potentially be managed non-operatively, and major surgery may be avoided. Table Tumor Stage and Treatment Parameters Univariate Analyses Ă P-141 A PREDICTIVE MODEL FOR FIVE-YEAR SURVIVAL FOLLOWING CYTOREDUCTIVE SURGERY AND HIPEC FOR PERITONEAL METASTASIS OF COLORECTAL CANCER USING A NOVEL ECONOMICAL APPROACH Dan Assaf 1, 2 , Haggai Benvenisti 1, 2 , Eyal Mor 1, 2 , Almog Ben-Yaacov 1, 2 , Gal Schtrechman 1,2 , Dov Zippel 1, 2 , David Hazzan 1,2 , Einat Shacham- Shmueli 3, 4 , Ofer Margalit 3, 4 , Daria Perelson 5, 6 , Dan Aderka 3, 4 , Aviram Nissan 1, 2 . 1 Sheba Medical Center, Tel Hashomer, Ramat Gan, Israel; 2 Affiliated with the Sackler School of Medicine, Tel Aviv University, Departments of General and Oncological Surgery, Ramat Gan, Israel; 3 Sheba Medical Center, Tel Hashomer, Ramat Gan, Israel; 4 Affiliated with the Sackler School of Medicine, Tel Aviv University, Oncology, Ramat Gan, Israel; 5 Sheba Medical Center, Tel Hashomer, Ramat Gan, Israel; 6 Affiliated with the Sackler School of Medicine, Tel Aviv University, Anasthesia, Ramat Gan, Israel Background: Given the complexity of cytoreductive surgery (CRS) and heated intraperitoneal chemotherapy (HIPEC) for treatment of peritoneal metastasis from colorectal cancer (CRC), a prediction model for overall survival (OS) may modify treatment and reduce morbidity and mortality. Objective. This study aims to develop a prognostic model in order to predict five year survival (5YS) following CRS/HIPEC among CRC patients. Materials and Methods: A retrospective analysis of a prospectively maintained database for all CRC patients who underwent CRS/HIPEC sur- gery between 2007 to 2018 (n¼417) was conducted. Only first completed HIPEC surgery were included (n¼291). Patients with colorectal cancer and at least one year of follow up were analyzed (n¼188) Results: Overall, 188 patients were included in the final analysis, the five years death rate was 27.1% (51/188), with a median OS of 68.5 months (CI ¼ 37.5 e 99.5). Kaplan-Mayer survival analysis revealed a significant different in OS be- tween the different primary tumor (p¼0.018, colon-mean 66.3, appendix- mean 103.8, rectum-mean 48.44). Univariate survival cox analysis identified eight significant factors associ- ated with OS, among them: operation length (p¼0.03), packed cell trans- fusion (p¼0.000), operative PCI (p¼0.003), and operation extent as presented by resected organs (pelvic peritonectomy p¼0.04, cystectomy p¼0.004, anterior resection p¼0.005, retroperitoneal dissection p¼0.005) and the need for ileostomy (p¼0.001).A conditional multivariate cox analysis identified only retroperitoneal dissection (p¼0.003), packed cell transfusion (p¼0.039), and the need of ileostomy (p¼0.006) as indepen- dently predictive of OS. Binary logistic regression model including multiple variables predicted 5YS with sensitivity of 42% and specificity of 92.5%, with AUC of 0.673 (p¼0.000). Using machine learning economical classification tree model based on 10- fold cross validation, including single variant to divide the sample into three risk groups by PCI level we reached sensitivity of 59.8% and speci- ficity of 80% for predicting five years survival. Conclusions: This machine learning predictive model demonstrated an economical approach using single variant and may enable clinical risk group classification. External validation of the models is needed, never- theless this model might be enhanced using a limited number of variables. P-142 AN ENSEMBLE ALGORITHM MODEL FOR THE DIAGNOSIS OF COLORECTAL CANCER BASED ON MACHINE LEARNING Jin-Hyeok Park 1 , Young-Ho Lee 2 , Hyun-Jin Kwon 2 , Won Jun Park 2 , Seung Uk Park 2 , Sun-Jin Sym 3 , Jeong-Heum Baek 4 . 1 Gachon University, IT Convergence Engineering, Incheon, Republic of Korea; 2 Gachon University, Computer Engineering, Incheon, Republic of Korea; 3 Gachon Univ. Gil Medical Center, Medical Oncology, Incheon, Republic of Korea; 4 Gachon Univ. Gil Medical Center, Surgery, Incheon, Republic of Korea Background: Machine learning technology has recently been used in various fields as a technology to extract the results through the informa- tion contained in given data without prior knowledge about prediction problems. As a result of searching using ‘Machine Learning’ and ‘Colorectal’ as main keywords in the life science and medicine database ‘PubMed’, the Abstracts / European Journal of Surgical Oncology 46 (2020) e30ee171 e74