Abstract Book

ESTRO 37

S364

by FL-V75 (p<0.001), V70 (p<0.001) and V65 (p=0.02) as well as RT-V75 (p=0.003), V50 (p=0.03), V45 (p=0.02), V40 (p=0.03), V35 (p=0.04), V10 (p=0.03) and V5 (p=0.02). The remaining domains of executive function, naming, memory, attention and language did not correlate with brain dosimetry. Patient-reported cognitive function correlated with FL-V65 (R=0.242, p=0.02) and RT-V75 (R=0.19, p=0.058). Although post-treatment FrSBe apathy, disinhibition and executive dysfunction subscale scores did not correlate with brain dosimetry, associations were found in patient- reported changes from pre-illness to post-treatment. These significant correlations with brain dosimetry were seen in each domain: (1) total score (RT-V75 R=0.25, p=0.01); (2) apathy (RT-V75, R=0.20, p=0.02; LT-V75 R=0.23, p=0.02; (3) disinhibiton – (RT-V75, R=0.20, p= 0.05); and (4) executive dysfunction (RT-V75, R=0.27, p<0.01). The total volume of temporal lobe (RT+LT) receiving V70, V65, V60 and V55 correlated with the presence of radiological TLN. Conclusion Although overall cognitive function showed only borderline association with frontal and temporal lobe dosimetry, the individual domains of visuospatial function, abstraction and orientation did show weak but significant correlation. Self-reported changes in neurobehavioural function were most affected by high doses (V75) to the temporal lobes. PO-0711 Re-irradiation in recurrent high grade gliomas: An analysis of survival and radionecrosis outcomes M. Shanker 1 , B. Chua 2 , C. Bettington 2 , M. Pinkham 3 1 The University of Queensland, Faculty of Medicine, Herston, Australia 2 Royal Brisbane and Women's Hospital, Department of Radiation Oncology, Brisbane, Australia 3 Princess Alexandra Hospital, Department of Radiation Oncology, Woolloongabba, Australia Purpose or Objective We aim to review the current literature and assess the efficacy and toxicity outcomes of conventional fractionated radiotherapy, fractionated stereotactic radiotherapy (FSRT) and stereotactic radiosurgery (SRS) in the management of recurrent high grade gliomas. Material and Methods We performed a systematic analysis of relevant articles in the MEDLINE and Pubmed databases published from 1990 to 2016. The studies were analysed using the linear- quadratic model with an α/β=2 to determine to cumulative equivalent total doses in 2-Gy fractions (EQD2). A population weighted multiple linear regression analysis was performed to evaluate the relationships between medial overall survival (MOS) and clinical radionecrosis (RN) outcomes and radiotherapy modality, total EQD2, re-irradiation EQD2 and re-irradiation dose per fraction. Results 73 articles were included based on the selection criteria with a total of 3403 patients analysed. In all studies, patients received an initial median dose of 54-60Gy conventionally fractionated external beam radiotherapy with or without concurrent chemotherapy. In the multivariate analysis, there was a significant difference in (MOS) and radiotherapy type when adjusted for total dose, age, interval between initial and salvage radiotherapy and PTV (p<0.0001) with SRS having the greatest MOS post completion of re-irradiation (12.6 Poster: Clinical track: CNS

months, 95%CI 12.1-13), followed by FSRT (10.1 months, 95%CI 9.6-10.5 months) and conventional (8.8 months, 95%CI 8.3-9.29). Interval had a significant impact on survival with MOS increasing by 0.25 months (p<0.0001, 95%CI 0.02-0.29) for every month increase in interval. There was no statistically significant association between re-irradiation dose per fraction and survival adjusted for other covariates (p=0.3398). There is an overall association between treatment type and clinical RN after adjusting for age, total and re-irradiation EQD2, interval and PTV ([<0.0001) with the highest incidence of RN in FSRT (7.1%, 95%CI 6.6 to 7.7%), followed by SRS (6.1%, 95%CI 5.6-6.6%) and conventional (1.1%, 95%CI 0.5 to 1.7%). The incidence of RN decreased by 0.5% for every month increase in interval (p<0.0001, 95%CI 0.42 to 0.54%). Conclusion Re-irradiation is an acceptable option in the management of recurrent high grade gliomas. Our pooled analysis of the literature demonstrates survival outcomes are significantly greater with SRS, followed by FSRT and conventionally fractionated radiotherapy however the incidence of RN is significantly higher in SRS and FSRT compared to conventionally fractionated radiotherapy. There is a statistically and clinically significant association with increased interval between initial and salvage radiotherapy treatments showing improved survival and toxicity outcomes. Re-irradiation dose per fraction when adjusted for other confounders is not statistically significantly associated with survival or toxicity. PO-0712 Radiomics to predict medulloblastomas molecular subgroups: prospective blinded data of 111 patients R. Jalali 1 , A. Dasgupta 1 , N. Shirsat 2 , E. Sridhar 3 , S. Pungavkar 4 , G. Chinnaswamy 5 , R. Krishnatry 1 , R. Tonse 1 , T. Gupta 6 1 Tata Memorial Hospital, Radiation Oncology, Mumbai, India 2 Tata Memorial Centre, Neuro Oncology Lab, Mumbai, India 3 Tata Memorial Centre, Neuro pathology, Mumbai, India 4 Global Hospital, Radiology, MUMBAI, India 5 Tata Memorial Centre, Pediatric Oncology, MUMBAI, India 6 Tata Memorial Centre, Radiation Oncology, Mumbai, India Purpose or Objective To identify the medulloblastoma molecular subgroups pre-operatively using conventional MRI features. Material and Methods Based on our previous experience and relevant literature we had predefined a set of 22 radiological features, each classified as categorical data. The pre-operative scans of 111 patients with histological diagnosis of medulloblastoma (MB) were discussed in the joint neuro- oncology meeting comprising of radiologists and oncologists who were blinded from the molecular subgrouping on tissue samples. Molecular profiling was done using differential expression of 12 protein coding genes and 9 microRNAs. Of 111 patients 2/3 rd were randomly selected from each of four subgroups to create the training cohort (TC-76 patients) and the others were labelled as validation cohort (VC-35 patients). Multinomial logistic regression in TC led to identification of parameters based on which binary nomograms were constructed for each subgroup and findings validated in VC. Results Of the four nomograms developed SHH nomogram (features used-tumour on horizontal and vertical aspect, brainstem involvement, peritumoural oedema and