Authors:
Victor E. Staartjes (Zurich | CH)
Morgan Broggi (Milan | IT)
Costanza Maria Zattra (Milan | IT)
Flavio Vasella (Zurich | CH)
Julia Velz (Zurich | CH)
Silvia Schiavolin (Milan | IT)
Carlo Serra (Zurich | CH)
Jiri Bartek Jr (Stockholm | SE)
Alexander Fletcher-Sandersjöö (Stockholm | SE)
Petter Förander (Stockholm | SE)
Darius Kalasauskas (Mainz | DE)
Mirjam Renovanz (Mainz | DE)
Florian Ringel (Mainz | DE)
Konstantin R. Brawanski (Innsbruck | AT)
Johannes Kerschbaumer (Innsbruck | AT)
Christian F. Freyschlag (Innsbruck | AT)
Asgeir S. Jakola (Gothenburg | SE)
Kristin Sjåvik (Tromsö | NO)
Ole Solheim (Trondheim | NO)
Bawarjan Schatlo (Göttingen | DE)
Alexandra Sachkova (Göttingen | DE)
Hans-Christoph Bock (Göttingen | DE)
Abdelhalim Hussein (Göttingen | DE)
Veit Rohde (Göttingen | DE)
Marike L.D. Broekman (Leiden | NL)
Claudine O. Nogarede (Leiden | NL)
Cynthia Lemmens (Leiden | NL)
Julius Kernbach (Aachen | DE)
Georg Neuloh (Aachen | DE)
Oliver Bozinov (Zurich | CH)
Niklaus Krayenbühl (Zurich | CH)
Johannes Sarnthein (Zurich | CH)
Paolo Ferroli (Milan | IT)
Luca Regli (Zurich | CH)
Martin N. Stienen (Zurich | CH)
Background: Decision-making for intracranial tumor surgery requires balancing the oncological benefit against the risk for resection-related impairment. There is no reliable and objective way to predict an individual patient’s risk of experiencing functional impairment. We aimed to develop and externally validate a prediction model for functional impairment after intracranial tumor surgery.
Methods: We developed a machine learning model using data from two prospective registries (2013 - 2017), and subsequently carried out multicenter external validation (2007 - 2018) in 7 European centers. Consecutive series of adult patients with primary or secondary intracranial tumors who underwent resection based on a “maximum safe resection” philosophy were included. Diagnostic biopsies were excluded. We recorded Karnofsky Performance Status (KPS) at admission and at 3 to 6 months postoperatively, as well as age, gender, prior surgery, tumor histology and maximum diameter, expected major brain vessel or cranial nerve manipulation, resection in eloquent areas and/or the posterior fossa, and surgical approach. The primary endpoint was functional impairment at 3 to 6 months postoperatively, defined as a decrease in KPS of ≥10 from baseline.
Results: In the development (2437 pts., 48.2% male; mean [SD] age, 55 [15] years) and external validation (2427 pts. 42.4% male; mean [SD] age, 58 [13] years) cohorts, rates of functional impairment were 21.5% and 28.5%, respectively. The final model included all 11 baseline variables. In the development cohort, area-under-the-curve (AUC) values of 0.72 (95% CI: 0.69 – 0.74) were observed. In the pooled external validation cohort, AUC was 0.72 (95% CI: 0.69 – 0.74), confirming the generalizability of the prediction tool. Calibration plots indicate fair calibration of the predicted probabilities in the development and external validation cohorts. The prediction tool has been incorporated into a surgeon- and patient-friendly web-app, and is available at https://neurosurgery.shinyapps.io/impairment/.
Conclusions: Functional impairment after intracranial tumor surgery is extraordinarily difficult to predict. The proposed, externally validated prediction tool can serve as basis for case-by-case discussions on the risk-to-benefit estimation of surgical treatment in the individual patient.