Abstract Book

ESTRO 37

S372

the competing event. Cox regression was performed for univariate analyses. Results From 1,465 potential subjects, 710 small- to medium- sized BM (≤ 2 cm) were evaluated; of those, 555 lesions (78.1%) were treated with DP ≥ 20 Gy and 155 lesions (21.9%) with 15 Gy. Median radiographic follow-up after SRS was 9.2 months (IQR 3-80 months). Overall, there were 31 local failures (4.37%) resulting in a CI at 2 years of 7.2% (95%IC 0-16.5%) for DP of 15 Gy and 4.8% (95%IC 0-9.7%) for ≥ 20Gy (P=0.45) . Overall, CI of LF for lesions ≤ 1cm, and > 1 cm, were 2.1% (95%CI 0-5.7%), and 10.0% (95CI 0.1-20%), respectively (p=0.0004). Overall, 26 lesions developed RN (3.66%). CI of RN at 2 years was 2.1% (95%IC 0-7.2%) for 15 Gy and 4.8% (95%IC 0-9.9%) for ≥ 20Gy (P=0.02) . Overall, CI of RN at 2 years for lesions ≤ 1cm, and > 1 cm, were 1.4% (95%CI 0-4.4%), and 8.9% (95CI 0-19%), respectively (p=0.0014). For lesions ≤ 1cm, there were no LFs or RN events for DP of 15 Gy. For lesion > 1 cm, there was no significant difference (p=0.344) when comparing DP of 15 Gy (10.2%) versus ≥ 20 Gy (7.7%) for LF, but there was a significantly higher CI of RN for lesions treated with ≥ 20 Gy (9.3% versus 3.8%, P=0.049). Conclusion Our results suggest that, with respect to LF, 15 Gy is equivalent to ≥ 20 Gy for small- to medium-sized (≤ 2 cm) BM treated with Gamma Knife SRS. However, DP ≥ 20 Gy correlated with a higher CI of RN, especially for lesion greater than 1 cm. PO-0728 Pre-Radiation Tumor Progression Improves Prognostic Stratification In Glioblastoma Patients J. Palmer 1 , D. Bhamidipati 2 , G. Shukla 2 , E. James 3 , C. Farrell 3 , K. Judy 3 , L. Kim 4 , J. Glass 4 , D. Andrews 3 , M. Werner-Wasik 2 , W. Shi 2 1 Ohio State University, Radiation Oncology, Columbus, USA 2 Sidney Kimmel Medical College at Thomas Jefferson University, Radiation Oncology, Philadelphia, USA 3 Sidney Kimmel Medical College at Thomas Jefferson University, Neurosurgery, Philadelphia, USA 4 Sidney Kimmel Medical College at Thomas Jefferson University, Neuro-Oncology, Philadelphia, USA Purpose or Objective The goal of our study was to assess the prevalence of early radiographic progression (EP) in GBM patients, and its associated outcomes. We hypothesize that early progression represents a prognostic factor for poorer survival in patients with GBM. Material and Methods After IRB approval; consecutive patients with newly diagnosed Glioblastoma from 2006-2014 were identified. All imaging, from the preoperative scan up until recurrence, was reviewed independently by an experienced neuroradiologist. Patients received SOC therapy. Demographic variables were compared using Fisher’s exact test for categorical variables and two- sample t-test for quantitative variables. Kaplan-Meier curves were generated to compare progression free survival and overall survival between the study groups and log-rank test was performed to evaluate the differences between the groups. Univariate analysis was performed using the log-rank test for categorical variables and Cox proportional hazards for continuous variables. Multivariable logistic regression was performed to assess factors related to EP and Cox proportional hazards model was used for multivariate analysis of OS. P-values <=0.05 were considered statistically significant. Results 107 patients were reviewed and 87 patients met entry criteria. The OS of 11.5 months in the EP group was significantly shorter than the OS of 20.1 month in

patients without EP (p=0.013). The median PFS of 6.3 months with EP was shorter than the median PFS of 10.9 months without EP (p=0.064). Survival based on presence of EP and MGMT was 10.2 months in patients with evidence of EP and unmeth; 16.5 months in patients with evidence of EP and meth; 19.6 months in patients without evidence of EP and umeth, and 34.7 months in patients without evidence of EP and meth (p=0.033). MGMT status, age, tumor location in deep brain structures, and time from surgery to the pre-RT MRI were not associated with the development of EP (p=0.330, p=0.227, p=0.316, and p=0.766 respectively). Of the EP patients, 27 patients had radiographic evidence of progression available for review. Of this cohort, 20 patients (74%) were noted to have a recurrence at the site of EP, with 12 patients (44%) progressing within the surgical bed near the site of EP, and 3 patients (11%) progressing distantly.

Conclusion Approximately half of the GBM patients evaluated by pre- radiation MRI were noted to have evidence of tumor progression, which was found to be associated with worse overall survival. Moreover, assessment of EP improved prognostic stratification by MGMT methylation status. We found that EP before the initiation of radiation could independently predict outcomes in patients with newly diagnosed glioblastoma, further stratifying patients based on known prognostic variables such as MGMT. Furthermore, we found that many of the posttreatment recurrences were preceded by patterns of EP. These findings support the use of pre-radiation MRI to evaluate for interval changes before treatment. PO-0729 Preoperative Prediction of Local Failure and Overall Survival for Meningioma E. Gennatas 1 , O. Morin 1 , S. Braunstein 1 , T. Solberg 1 , G. Valdes 1 , D. Raleigh 1 1 University of California- San Francisco, Radiation Oncology, San Francisco, USA Purpose or Objective Meningioma is the most common intracranial tumor in the United States, and the majority of patients are effectively treated with maximal safe resection and radiotherapy. Approximately 25-30% of meningiomas are atypical or anaplastic (WHO grade II and III, respectively) and recur in 40 – 70% of patients despite optimal therapy. Currently, the best predictor of local failure for atypical or anaplastic meningioma is tumor grade, which is established postoperatively. The aim of this study was to estimate the accuracy of preoperative clinical features in predicting local failure and overall survival for meningioma in comparison to the accuracy of using meningioma grade alone. Material and Methods 257 patients (161 females) who underwent surgery for intracranial meningioma at the University of California, San Francisco were selected (Mean Age = 58.0 years, SD = 13.4; Grade I, N = 128; Grade II, N = 104; Grade III, N = 25). We trained a series of algorithms to predict two outcomes, local failure (LF) and overall survival (OS),