ESTRO 36 Abstract Book

S325 ESTRO 36 _______________________________________________________________________________________________

determined with microscopic investigation of hematoxylin-eosin (HE) stained whole-mount slices. The purpose of this study is to investigate the accuracy of delineating microscopic tumor growth on HE stained slices by comparing it with the gold standard for squamous cell carcinoma (SCC): immunohistochemical pan-cytokeratin (CKAE1/3) staining. Material and Methods Twenty-seven patients with laryngeal or hypopharyngeal SCC underwent total laryngectomy (TLE). The laryngeal specimen was fixated and preserved with formalin and sliced up into 3 mm thick slices. For each patient, four subsequent slices with tumor were selected. From each slice, two 4µm-thick whole-mount sections were obtained and stained with HE and CKAE1/3. The tumor was delineated both on the HE and CKAE1/3 stained sections by a dedicated head-and-neck pathologist using a microscope. The stained sections were three- dimensionally reconstructed and rigidly registered to each other. Registration between CKAE1/3 and HE stainings was performed by manually selecting 3-5 corresponding landmarks. The HE-based tumor delineations were compared with the CKAE1/3-based tumor delineations. The maximum distance between HE and CKAE1/3 tumor delineations was measured for each patient. Furthermore, the conformity index (CI) was measured between the corresponding delineations. Results The 99 th percentile distance for microscopic disease outside the HE sections was 1.7 mm (0.9-3.4). This distance includes some registration error between HE and CKAE1/3 stained sections. The coverage of the CKAE1/3 delineated volume by the HE volume (sensitivity) was 0.93 (0.85-0.98) and the positive predictive value (PPV) was 0.89 (0.76-0.95). This shows that the tumor volume delineated on HE is larger than on CKAE1/3. As the distances between the contours on HE and CKAE and reversely between CKAE and HE is comparable, the effect on the CTV margin might be considered small. The CI was 0.83 (0.73-0.89), which is smaller than the CI for inter- observer variation in tumor delineation on HE stained sections. This might be partially attributed to the extra registration step which is required for the comparison between HE and CKAE1/3.

Conclusion Nearly all tumor is delineated on HE within 2 mm accuracy of the CKAE1/3 delineation. This is comparable with the interobserver delineation accuracy on HE stained sections. Consequently, the effect on the CTV margin for delineations made on MRI or CT is not clinically relevant.

Poster: Clinical track: CNS

PO-0622 Medulloblastoma in adults: a retrospective single institution analysis I. Hadi 1 , O. Roengvoraphoj 1 , F. Roeder 1,2 , C. Belka 1 , S. Nachbichler 1 1 Ludwig-Maximilians-Universität München, Radiation Oncology, München, Germany 2 German Cancer Research Center DKFZ, Molecular Radiation Oncology, Heidelberg, Germany Purpose or Objective Adult medulloblastoma is a rare disease. The therapy is based on pediatric treatment protocols. This retrospective analysis investigated the clinical outcomes and prognostic factors of adult patients receiving multimodal therapy in our institution. Material and Methods Treatment charts of patients with medulloblastoma older than 15 years old, who have been treated in our institution between 2001 and 2014, were analyzed retrospectively. Supratentorial tumors were excluded. Patients‘ demographic parameters, histology, initial