gms | German Medical Science

69. Jahrestagung der Deutschen Gesellschaft für Neurochirurgie (DGNC)
Joint Meeting mit der Mexikanischen und Kolumbianischen Gesellschaft für Neurochirurgie

Deutsche Gesellschaft für Neurochirurgie (DGNC) e. V.

03.06. - 06.06.2018, Münster

Article

Send article

The role of resection in primary glioblastoma – revisited

Meeting Abstract

Search Medline for

  • Marco Skardelly - Universitätsklinikum Tübingen, Klinik für Neurochirurgie, Tübingen, Deutschland
  • Marlene Kaltenstadler - Universitätsklinikum Tübingen, Klinik für Neurochirurgie, Tübingen, Deutschland
  • Felix Behling - Universitätsklinikum Tübingen, Klinik für Neurochirurgie, Tübingen, Deutschland
  • Susan Noell - Universitätsklinikum Tübingen, Klinik für Neurochirurgie, Tübingen, Deutschland
  • Irina Gepfner-Tuma - Universitätsklinikum Tübingen, Neurologie, Tübingen, Deutschland
  • Jens Schittenhelm - Universitätsklinikum Tübingen, Institut für Pathologie und Neuropathologie, Tübingen, Deutschland
  • Benjamin Bender - Universitätsklinikum Tübingen, Neuroradiologie, Tübingen, Deutschland
  • Frank Paulsen - Universitätsklinikum Tübingen, Radioonkologie, Tübingen, Deutschland
  • Ulrike von Hehn - Medistat GmbH, Kronshagen, Deutschland
  • Mirjam Renovanz - Universitätsmedizin der Johannes Gutenberg-Universität Mainz, Klinik und Poliklinik für Neurochirurgie, Mainz, Deutschland
  • Jens Gempt - Technische Universität München, Klinikum rechts der Isar, Klinik für Neurochirurgie, München, Deutschland
  • Melanie Barz - Technische Universität München, Klinikum rechts der Isar, Klinik für Neurochirurgie, München, Deutschland
  • Bernhard Meyer - Technische Universität München, Klinikum rechts der Isar, Klinik für Neurochirurgie, München, Deutschland
  • Ghazaleh Tabatabai - Universitätsklinikum Tübingen, Neurologie, Tübingen, Deutschland
  • Marcos Tatagiba - Universitätsklinikum Tübingen, Klinik für Neurochirurgie, Tübingen, Deutschland

Deutsche Gesellschaft für Neurochirurgie. 69. Jahrestagung der Deutschen Gesellschaft für Neurochirurgie (DGNC), Joint Meeting mit der Mexikanischen und Kolumbianischen Gesellschaft für Neurochirurgie. Münster, 03.-06.06.2018. Düsseldorf: German Medical Science GMS Publishing House; 2018. DocV041

doi: 10.3205/18dgnc042, urn:nbn:de:0183-18dgnc0423

Published: June 18, 2018

© 2018 Skardelly et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License. See license information at http://creativecommons.org/licenses/by/4.0/.

Outline

Text

Objective: In primary glioblastoma (GBM) the role of resection is well established. However, different thresholds for a supposed clinical benefit have been published ranging from about 70% to total resection. We compared parametric and non-parametric models to investigate the nature of relationship of extent of resection (EOR) and overall survival (OS).

Methods: We included patients >18years of age with primary isocitrate dehydrogenated wild type (IDH) GBM who were operated between 2006 and 2014 in our department. Univariate Cox regressions identified significant predictors of OS, which were included in multivariate Cox regression to determine the impact of EOR on OS. The parametric accelerated failure time model (AFT) was also used to examine the relationship between EOR and OS. Models were validated by internal cross-validation and by an external patient cohort.

Results: Of 425 patients 303 complete datasets were available for multivariate Cox regression and AFT model. Residual tumor volume, age, methylation of O6-Methylguanin-DNA-Methyltransferase (MGMT), therapy modality and extent of white matter infiltration related to ventricles were significant predictors of OS. AFT model demonstrated a continuous almost linear relationship between residual tumor volume and OS. Both models showedsimilar coefficients of determination (R2) of 0.38 and 0.4, which demonstrate a good prediction of OS. The median prediction error of the model was about 1.3 months. External validation with a cohort of 253 patients demonstrated a slight overfitting of the model but still a median prediction error of less than 2.5 months for.

Conclusion: The residual tumor volume predicts OS in a continuous but not "all or nothing" relation. Therefore maximal safe resection is recommended in primary glioblastoma treatment. The predictability of parametric and non-parametric models is closely correlated but parametric models are more complex to calculate However, they need less assumptions and are far better in visualizing the data.