follow-up. Dose differences were tested for signiﬁcance using permutation testing. The most statistically signiﬁcant area of dose difference was out- lined and individual patient doses collected. A Cox regression survival analysis was performed with the patients grouped into those above and below the median dose, with a log rank test performed, and adjusted for tumor size (continuous variable). Results: For all time points (except 3 months) a statistically signiﬁcant dose difference (P < 0.001) was seen across the upper portion of the mediastinum. The most signiﬁcant dose difference was found in the base of the heart, centered over the aorta and the origin of the left coronary artery. The median dose to this area was 20.2 Gy across all patients with a 6 Gy difference between survivors and non-survivors at 12 months. The Kaplan- Meier curves for patients receiving above or below 20.2 Gy were well separated on their 95% conﬁdence intervals; difference in median survival of 6 months (log rank P < 0.001). Controlling for tumor size resulted in a signiﬁcantly improved survival model (change in cumulative hazard 0.4%) but a high dose region remained an independent predictor for death. The hazard ratio for death at 12 months (median survival across the cohort) was 1.16 and 1.28 with and without controlling for tumor size. Conclusion: This work utilized a novel data mining approach to identify dose-sensitive anatomical regions for overall survival. To our knowledge this is the ﬁrst application of such a method to lung cancer. The analysis showed a high dose to the base of the heart independently predicts for shorter survival, even when controlled for tumor size. Further work will aim to identify other covariates and validate the ﬁndings using other co- horts. Studies correlating dose to substructures of the heart and survival in lung cancer patients are also ongoing. Author Disclosure: A. McWilliam: None. C. Faivre-Finn: None. J. Kennedy: None. L. Kershaw: None. M.B. van Herk: None. 108 National Cancer Data Base Analysis of Uterine Carcinosarcoma (UC): Improvement in Survival With the Use of Radiation Therapy E.M. Nichols, 1 S.M. Bentzen, 1 D. Scartoni, 2 I. Giacomelli, 2 D.M. Roque, 1 G. Rao, 1 N. Hanna, 1 and P. Mohindra 3 ; 1 University of Maryland School of Medicine, Baltimore, MD, 2 University of Florence, Florence, Italy, 3 University of Maryland Medical Center, Baltimore, MD Purpose/Objective(s): The value of radiotherapy in the management of UC is controversial. Several multi-institutional studies have suggested a beneﬁt from radiation therapy (RT), although sample sizes are small due to rarity of disease. The purpose of this analysis was to estimate the asso- ciation between RT and overall survival of UC using the NCDB which provides the largest report of this patient population to date. Materials/Methods: The NCDB was queried to identify women diagnosed with UC between 1998 and 2012. A total of 9317 cases were identiﬁed. After excluding patients with stage IV disease or missing stage, survival, or RT data, there were 4255 evaluable cases. The primary endpoint was overall survival (OS). Chi-square tests and Kaplan-Meier estimates and Cox proportional hazards model were used in the analysis. Results: Median age at diagnosis was 68 years; 52% had stage I (n Z 2169), 13% stage II (n Z 541), and 36% stage III (n Z 1545) disease. Nearly 25% were African American. Among all patients, 35% did not receive either chemotherapy (CT) or RT, 24% received RT alone, 18% received CT alone, and 20% received CT and RT. Data was missing data for 3%. Among the 1912 patients who received RT, 1387 received external beam with or without brachytherapy, 440 brachytherapy only and treat- ment was not otherwise speciﬁed in 85 patients. RT use did not change over time when divided into quartiles. With a median follow up of 23 months (maximum 191 months), estimated median and 5-year OS with RT was 48 months and 45% and without RT 29 months and 35% (P < 0.001). When stratiﬁed by stage, receipt of RT was associated with improved OS in all stages, with the largest magnitude of beneﬁt seen for stage I patients with a median OS of 97 versus 56 months (P < 0.001) followed by 38 versus 26 months for stage II (P < 0.014) and 23 versus 17 months for stage III (P < 0.001). Improved OS with the use of RT persisted, inde- pendent of receipt of CT. In patients who did not receive CT, median OS with and without RT was 38 months versus 27 months (P < 0.001); in patients who received CT, median OS with and without RTwas 70 months versus 33 months without (P < 0.001). Five-year OS was 51.3% for pa- tients receiving CT and RT; 39.1% for patients receiving CT alone; 40.1% for patients receiving RT alone and 33% with no CT or RT. On multivariate analyses, receipt of RT (HR Z 0.75: 95% CI Z 0.68-0.83, P < 0.001) and receipt of CT (HR Z 0.59; 95% CI Z 0.53-0.66, P < 0.001) were favorable while African American race (HR Z 1.34; 95% CI Z 1.2-1.5), higher stage (HR Z 1.7; 95% CI Z 1.6-1.8), older age (HR Z 1.03; 95% CI Z 1.02-1.03), and higher Charlson/Deyo comorbidity score (HR Z 1.12; 95% CI Z 1.03-1.2) were associated with worse survival. Conclusion: This large population based analysis provides strong evidence for an association between survival and the use of RT in the treatment of stage I-II UC. When combined with existing data, RT should be considered in patients with stage I, II, or III disease until randomized data can be obtained. Author Disclosure: E.M. Nichols: None. S.M. Bentzen: Executive Di- rector; International Commission on Radiation Units and Measurements. D. Scartoni: Visiting Resident; University of Maryland. I. Giacomelli: Visiting Resident; University of Maryland. D.M. Roque: None. G. Rao: None. N. Hanna: None. P. Mohindra: None. 109 Updated 4-Year Results for NRG Oncology/RTOG 0921: A Phase 2 Study of Postoperative Intensity Modulated Radiation Therapy With Concurrent Cisplatin and Bevacizumab Followed by Carboplatin and Paclitaxel for Patients With Endometrial Cancer A.N. Viswanathan, 1 K. Winter, 2 B.E. Miller, 3 Y. Xiao, 4 A. Jhingran, 5 L. Portelance, 6 W.R. Bosch, 7 U. Matulonis, 8 N. Horowitz, 9 R. Mannel, 10 L. Souhami, 11 B.A. Erickson, 12 W. Small, Jr, 13 and D.K. Gaffney 14 ; 1 Brigham & Women’s Hospital/Dana-Farber Cancer Institute, Boston, MA, 2 NRG Oncology Statistics and Data Management Center, Philadelphia, PA, 3 Carolinas Medical Center, Concord, NC, 4 University of Pennsylvania, Philadelphia, PA, 5 MD Anderson Cancer Center, Houston, TX, 6 Radiation Oncology Department, University of Miami/Sylvester Comprehensive Cancer Center, Miami, FL, 7 Washington University School of Medicine, St. Louis, MO, 8 Dana-Farber Cancer Institute, Boston, MA, 9 Brigham and Women’s Hospital, Boston, MA, 10 Oklahoma University Health Services Center, Oklahoma City, OK, 11 McGill University Health Centre, Montreal, QC, Canada, 12 Department of Radiation Oncology, Medical College of Wisconsin, Milwaukee, WI, 13 Stritch School of Medicine, Loyola University Chicago, Maywood, IL, 14 University of Utah Huntsman Cancer Hospital, Salt Lake City, UT Purpose/Objective(s): To report 4-year outcome results including late adverse events (AEs) >1 year from treatment start, overall survival (OS), para-aortic failure (PaF), distant failure (DF), and disease-free survival (DFS) in a prospective Phase II clinical trial of bevacizumab (Bev) and pelvic intensity modulated radiation therapy (IMRT) with chemotherapy in high-risk endometrial-cancer patients. Materials/Methods: Patients had a hysterectomy, lymph node removal, and 1 of the following high-risk factors: grade 3 carcinoma with >50% myometrial invasion (n Z 15); grade 2 or 3 disease with any cervical stromal invasion (n Z 10); or known extrauterine extension conﬁned to the pelvis (n Z 13). Treatment included pelvic IMRT and concurrent cisplatin on days 1 and 29 of radiation and Bev (5 mg/kg on days 1, 15, and 29 of radiation) followed by adjuvant carboplatin and paclitaxel for 4 cycles. Univariate OS and DFS estimates were obtained using the Kaplan-Meier method and PF and DF rates were determined by the cumulative incidence method. Results: Thirty-four patients were accrued from November 2009 through December 2011, 30 of whom were eligible and received study treatment. The median follow up is 3.92 years (min-max: 1.13-4.96). There were 5 patients with reported grade 3 or 4 AEs > 1 year from treatment start as follows: grade 3 diarrhea/ vomiting/ low lymphocyte count/ dyspnea/ acute kidney injury; grade 3 hematuria; grade 3 vaginismus; grade 3 skin/pelvic infection; and grade 4 low lymphocyte/neutrophil count. Four-year OS was Volume 96  Number 2S  Supplement 2016 Oral Scientiﬁc Sessions S49