ESTRO 2020 Abstract book

S231 ESTRO 2020

nodal region. This warrants modification of the proposed guidelines for PALN region.

Poster Highlights: Poster highlights 13 CL : Gynaecology

PH-0400 Number of involved nodes and overall survival in node-positive endometrial cancer R. LI 1 , A. Shinde 1 , A. Amini 1 , A. Liu 1 , S. Glaser 1 1 City of Hope Medical Center, Department of Radiation Oncology, Duarte- CA, USA Purpose or Objective Number of involved lymph nodes (LNs) is a crucial stratification factor in staging of numerous disease sites, including breast, gastrointestinal, and head and neck cancers. However, this risk factor has not been incorporated for endometrial cancer and current staging systems stratify node-positive endometrial cancer solely based on para-aortic (PA) nodal involvement. We evaluated whether number of involved LNs better predict survival outcomes and therapeutic response compared to Patients diagnosed with node-positive, non-metastatic, endometrioid endometrial cancer from 2004 to 2015 were identified in the National Cancer Database. All patients underwent hysterectomy and lymph node dissection with at least 6 nodes dissected. We trained a Cox proportional hazards regression model as well as a machine-learning based, gradient-boosted tree (GBT) model of overall survival (OS) using age, insurance type, comorbidity score, number of involved LNs, PA node involvement, and adjuvant therapy. Shapley additive explanation (SHAP) values based on GBT model were used to identify inflection points to select cutoffs of number of LNs involved. Results We identified 11,514 patients with node-positive endometrial cancer meeting the inclusion criteria at median follow-up of 49.2 months. We selected these thresholds for number of involved nodes using visualization of non-linear inflection points in the GBT model: 1-3 involved LNs, 4-5 LNs, and 6+ LNs. Using these thresholds, the 3-year OS was 80.7% for 1-3 LNs, 73.8% for 4-5 LNs (hazard ratio [HR] 1.28; p<0.001), and 58.3% for 6+ LNs (HR 2.52; p<0.001). On univariate Cox regression analysis, PA nodal involvement was a significant predictor of OS (HR 1.44; 95% CI 1.29-1.62; p<0.001) but was not significant on multivariate analysis when number of LNs was included (HR 1.02; 95% CI 0.90-1.17; p=0.724). Similar results were seen using the machine-learning based approach. The concordance index of a GBT model using only PA node involvement and excluding number of LNs was 67.4, compared to 69.6 using only number of involved LNs and 69.8 with both (p<0.001). Additionally, we identified an interaction between adjuvant therapy received and number of involved LNs. Patients with 1-5 involved LNs had 3-year OS of 82.4% with combined chemoradiation (CRT) and 74.2% with single modality therapy, while patients with 6+ involved LNs had a larger difference of 66.8% with CRT and 48.9% with single modality therapy (p-interaction=0.013). Conclusion We used both conventional regression and machine- learning based methods to demonstrate that number of involved LNs is a stronger prognostic factor compared to PA node involvement. Additionally, number of positive LNs was predictive of the survival benefit of combined modality adjuvant therapy. Our findings strongly support more robust incorporation of lymph node burden into current staging systems. Material and Methods

PH-0399 Validation and Applicability of Para-aortic Lymph Nodal Contouring Atlas in Cervical Cancer S. Srinivasan 1 , L. Gurram 1 , S. Chopra 1 , A. Baheti 2 , P. Popat 2 , N. Sable 2 , U. Mahantshetty 1 1 Tata Memorial Centre, Department of Radiation Oncology, Mumbai, India ; 2 Tata Memorial Centre, Department of Radiodiagnosis, Mumbai, India Purpose or Objective Cervical cancer with para-aortic lymph node positivity are not uncommon. These patients are treated with extended field RT (EFRT) to include para aortic (PA) lymph nodal (LN) regions. Contouring of PALN CTV is a challenge and associated with major uncertainties. Recently, CTV delineation guidelines for PALN for patients with cervical cancer have been proposed (Keenan et al 2018). The purpose of this study was to validate these guidelines with the use of CT datasets of cervical cancer patients with macroscopic PALN treated with definitive (chemo)radiation (CTRT) at our centre. Material and Methods Planning CT datasets of 72 cervical cancer patients (between December 2016 to December 2018) with gross PA nodal disease treated with EFRT were used for the study. Two hundred and four positive PALN were identified based either on size and morphology on diagnostic CECT, PET CT or histologically proven PALN. LN regions were divided into upper (T12 to L1-L2 interspace), middle (L2 to L3-L4 interspace) and lower(L4-S1). Macroscopic PALN {left para-aortic(LPA), aortocaval(AC), right paracaval(RPC)} were contoured and the CTV for PALN irradiation was generated based on the proposed guidelines This included contouring of abdominal aorta and IVC and anisotropic margins of 1cm all directions except lateral (1.5cm) around aorta while 8 mm antero-medially and 6mm postero-lateral to IVC. These volumes were then combined to create a single CTV PALN with minor edits to exclude vertebrae, muscle and bowel on ECLIPSE (Version 13.5). The centre of mass (COM N ) was calculated for each gross PALN contoured. Evaluation was done to review the presence of COM N in relation to the CTV PALN. The data was analysed using SPSS version 21. Results Out of 72 patients, 13(18.1%) had a baseline PET-CT, 65(90.3%) had a diagnostic CECT and 12(16.7%) had histologically proven PALN involvement. The most common location of PALN was LPA (111LN- 54.4%) AC (75 LN-36.8%) and RPC (18 LN-8.8%). Middle PALN were the commonest (150/204 LN-73.5%). 33 were lower PALN (16.2%) and 21 were upper PALN (10.3%). Of the 204 PALN, COM N for 34 LN (16.7%) was not encompassed within the CTV PALN. Majority of the misses were in LPA region (25 LN ,73.5%) followed by RPC region (5 LN) and 4 in AC region. Majority of the COM N especially in LPA region (20LN) were marginal misses and could be adequately encompassed within the CTV PALN by additional 5mm expansion laterally. Four LN in LPA region required atleast a 1cm further expansion laterally for them to be covered. Two LN (1 in LPA region and 1 in RPC region) were missed cranially and required additional 1cm margin superiorly. Conclusion Our study for validation of the proposed PALN contouring guidelines suggest a possible miss of microscopic nodal region in nearly 20% patients especially in left para-aortic