This study, recently published in Frontiers in Human Neuroscience, moves neurosurgery closer to an objective "scorecard" for deciding which patients should stay awake during brain tumor removal.
TL;DR
Deciding whether a patient should undergo awake brain surgery is currently a subjective process that varies between hospitals. Researchers have identified three objective predictors—patient age, specific language errors (paraphasias), and tumor location (eloquence)—that can form the basis of a standardized grading system to improve patient selection and surgical outcomes.
What Problem Does This Paper Address?
For patients with gliomas (brain tumors) located near language centers, "awake craniotomy" is the gold standard. By keeping the patient conscious and performing language tasks, surgeons can use direct electrical stimulation to map functional areas and avoid damaging them.
However, awake surgery isn't for everyone. It requires significant cognitive effort, emotional resilience, and a baseline level of language function. Currently, the criteria for selecting these patients are "highly heterogeneous"—essentially, it often comes down to a surgeon's "gut feeling" or local hospital tradition. There is a lack of a systematic, data-driven way to determine who is a good candidate, leading to potential inconsistencies in care.
Methods
The researchers performed a post-hoc analysis of 96 glioma patients treated at a single neurosurgical center between 2018 and 2021. They looked at:
- Preoperative Language Status: Patients performed an 80-item object-naming task. A speech therapist categorized errors into types like paraphasias (substituting the wrong word or sound), neologisms (making up words), and word-finding difficulties.
- Language Eloquence: They used a standardized 0–9 grading system based on MRI data (how close the tumor is to known language tracts and hubs) and clinical history.
- Statistical Modeling: They used binomial logistic regression and stepwise variable selection (specifically backward selection based on the Akaike Information Criterion) to identify which variables actually predicted whether a patient ended up in an awake vs. asleep surgery.
Datasets
The study utilized a clinical dataset from the Technical University of Munich, consisting of 96 patients with left-hemispheric, language-eloquent gliomas.
- Awake group: 25 patients.
- Asleep group: 71 patients.
- Tumor Type: Mostly high-grade gliomas (WHO grades 3 and 4).
Key Insights and Ideas
- The "25% Rule" Might Be Too Conservative: Traditional guidelines often suggest that if a patient makes more than 25% errors during baseline naming, they shouldn't have awake surgery. This study found that many patients who "failed" this rule still successfully completed awake surgery because their high tumor eloquence made the risk of asleep surgery even higher.
- Paraphasias are Key: Not all language errors are equal. While word-finding difficulties are common, the number of paraphasias (phonological or semantic errors) was a much stronger predictor of surgical choice. This suggests that specific types of language disruption are better indicators of how "involved" the language network is with the tumor.
- The Age Factor: Age was a significant predictor, with a clear preference for awake surgery in younger patients. The researchers identified a statistical "tipping point" around age 54, though they noted that healthy older patients can still be successful candidates.
Key Results
- The Predictive Trio: The final model showed that lower age, higher number of paraphasias, and high language eloquence were the most significant predictors for selecting awake surgery.
- Odds Ratios: Patients with "high" language eloquence (score 6–9) were 6.5 times more likely to be selected for awake surgery than those with "moderate" eloquence.
- Cut-off Values: ROC (Receiver Operating Characteristic) analysis suggested an optimal age cut-off of $\le$ 54 years and a language eloquence score of at least 6 for indicating awake surgery.
- Feasibility: In the awake cohort, the procedure was feasible in 96% of cases, suggesting that the current (albeit subjective) selection process is safe, but could be more efficient if standardized.
Limitations and Future Work
- Single-Center Bias: Because the data comes from one hospital, the "selection" criteria identified are essentially a reflection of that center's current expertise. Multicenter trials are needed to see if these predictors hold up across different surgical teams.
- Small Sample Size: With only 25 awake cases, the study lacked the statistical power to use more advanced machine learning or to split the data into training and validation sets.
- Missing Intraoperative Data: The study focused on preoperative predictors. It did not analyze how these patients actually performed during the surgery or their long-term post-operative recovery, which is the ultimate measure of a "good" selection.
- Future Direction: The authors propose integrating these variables into a "standardized grading tool" that can be easily used in a clinical workflow to make the decision-making process more objective and reproducible.
Summary updated