Karin Gehring

Background: Patients with a glioma often suffer from cognitive impairments both before and after anti-tumor treatment. Ideally, clinicians can rely on predictions of post-operative cognitive functioning for individual patients based on information obtainable before surgery. Such predictions would facilitate selecting the optimal treatment considering patients’ onco-functional balance. Method: Cognitive functioning 3 months after surgery was predicted for 317 patients with a glioma across 8 cognitive tests. Nine multivariate Bayesian regression models were used following a machine-learning approach while employing pre-operative neuropsychological test scores and a comprehensive set of clinical predictors obtainable before surgery. Model performances were compared using the expected log pointwise predictive density (ELPD), and pointwise predictions were assessed using the coefficient of determination (R2) and mean absolute error. Models were compared against models employing only pre-operative cognitive functioning, and the best-performing model was interpreted. Moreover, an example prediction including uncertainty for clinical use was provided. Results: The best-performing model obtained a median R2 of 34.20%. Individual predictions, however, were uncertain. Pre-operative cognitive functioning was the most influential predictor. Models including clinical predictors performed similarly to those using only pre-operative functioning (ΔELPD = 14.4 ± 10.0, ΔR2 = −0.53%). Conclusion: Post-operative cognitive functioning could not reliably be predicted from pre-operative cognitive functioning and the included clinical predictors. Moreover, predictions relied strongly on pre-operative cognitive functioning. Consequently, clinicians should not rely on the included predictors to infer patients’ cognitive functioning after treatment. Furthermore, our results stress the need to collect larger cross-center multimodal datasets to obtain more certain predictions for individual patients.

Computerized neuropsychological test batteries (CNTs), such as Central Nervous System Vital Signs (CNS VS), are increasingly used for measuring cognitive functioning, but empirical evidence of how they measure cognition is scarce. We investigated the factor structure of CNS VS using exploratory factor analyses four samples: healthy adults ( n = 169), patients with meningioma (392), low-grade glioma (99), and high-grade glioma (247). We tested model fit and investigated measurement invariance. Differences in factor interpretation existed between healthy participants and patients. Factor structures among patient groups were approximately the same but differed in non-zero loadings. Overall, factor structures largely did not support the “clinical domains” provided by CNS VS for clinical interpretation. Confirmatory models did not have a good fit, and measurement invariance could not be established. Our results indicate that (weighted) sum scores of CNS VS results may lack validity. We recommend researchers and clinicians to use scores on individual test measures.

Introduction Patients with a glioma often suffer from cognitive impairments both before and after anti-tumor treatment. Ideally, clinicians can rely on predictions of post-operative cognitive functioning for individual patients based on information obtainable before surgery. Such predictions would facilitate selecting the optimal treatment considering patients’ onco-functional balance. Method Cognitive functioning three months after surgery was predicted for 317 patients with a glioma across eight cognitive tests. Nine multivariate Bayesian regression models were used following a machine-learning approach while employing pre-operative neuropsychological test scores and a comprehensive set of clinical predictors obtainable before surgery. Model performances were compared using the Expected Log Pointwise Predictive Density (ELPD), and pointwise predictions were assessed using the Coefficient of Determination (R²) and Mean Absolute Error. Models were compared against models employing only pre-operative cognitive functioning and the best-performing model was interpreted. Moreover, an example prediction including uncertainty for clinical use was provided. Results The best-performing model obtained a median R² of 34.20%. Individual predictions, however, were uncertain. Pre-operative cognitive functioning was the most influential predictor. Models including clinical predictors performed similarly to those using only pre-operative functioning (ΔELPD 14.4±10.0, ΔR² −0.53%.). Conclusion Post-operative cognitive functioning cannot yet reliably be predicted from pre-operative cognitive functioning and the included clinical predictors. Moreover, predictions relied strongly on pre-operative cognitive functioning. Consequently, clinicians should not rely on the included predictors to infer patients’ cognitive functioning after treatment. Moreover, it stresses the need to collect larger cross-center multimodal datasets to obtain more certain predictions for individual patients. Importance of the study Patients with a glioma often suffer from cognitive impairments both before and after anti-tumor treatment. Ideally, clinicians would be able to rely on predictions of cognitive functioning after treatment for individual patients based on information that is obtainable before surgery. Such predictions would facilitate selecting the optimal treatment considering patients’ onco-functional balance and could improve patient counseling. First, our study shows that cognitive functioning three months after surgery cannot be reliably predicted from pre-operative cognitive functioning and the included clinical predictors, with pre-operative cognitive functioning being the most important predictor. Consequently, clinicians should not rely on the included predictors to infer individual patients’ cognitive functioning after surgery. Second, results demonstrate how individual predictions resulting from Bayesian models, including their uncertainty estimates, may ultimately be used in clinical practice. Third, our results show the importance of collecting additional predictors and stress the need to collect larger cross-center multimodal datasets. Key points - Cognitive functioning after treatment cannot yet reliably be predicted - Pre-operative cognitive functioning was the most important predictor - Additional predictors and larger cross-center datasets are needed

Introduction: Cognitive impairments of patients with a glioma are increasingly considered when making treatment decisions considering a personalized onco-functional balance. Predicting cognitive functioning before surgery can serve as a steppingstone for the clinical goal of predicting cognitive functioning after surgery. However, in a previous study, machine-learning models could not reliably predict cognitive functioning before surgery using a comprehensive set of clinical variables. The current study aims to improve predictions while making the uncertainty in individual predictions explicit. Method: Pre-operative cognitive functioning was predicted for 340 patients with a glioma across eight cognitive tests. This was done using six multivariate Bayesian regression models following a machine-learning approach while using a comprehensive set of clinical variables. Four models included interactions with- or a multilevel structure over histopathological diagnosis. Point-wise predictions were compared using the coefficient of determination (R ² ) and the best-performing model was interpreted. Results: Bayesian models outperformed machine-learning models and benefitted from using shrinkage priors. The R ² ranged between 0.3% and 21.5% with a median across tests of 7.2%. Estimated errors of individual prediction were high. The best-performing model allowed parameters to differ across histopathological diagnoses while pulling them toward the population mean. Conclusion: Bayesian models can improve predictions while providing uncertainty estimates for individual predictions. Despite this, the uncertainty in predictions of pre-operative cognitive functioning using the included clinical variables remained high. Consequently, clinicians should not infer cognitive functioning from these variables. Different histopathological diagnoses are best treated as distinct yet related. Highlights: Bayesian regression outperformed machine-learning models. Predictions were uncertain despite improvements. Different histopathological diagnoses are best treated as distinct yet related. Importance of the study: Cognitive impairments of patients with a glioma are increasingly considered when making treatment decisions considering a personalized onco-functional balance. Predicting cognitive functioning before surgery serves as a steppingstone for the clinical goal of predicting cognitive functioning after surgery. The current study is important for two reasons. First, it demonstrates that Bayesian models can improve predictions of pre-operative cognitive functioning over popular machine-learning models. Second, it explicitly shows that individual predictions of pre-operative cognitive functioning based on a comprehensive set of readily available clinical variables included in the current study are uncertain. Consequently, clinicians should not infer cognitive functioning from these variables. Last, it shows that prediction models may benefit a multifaceted view of patients and from treating patients with different histopathological diagnoses as distinct yet related.

Cognitive functioning is increasingly considered when making treatment decisions for patients with a brain tumor in view of a personalized onco-functional balance. Ideally, one can predict cognitive functioning of individual patients to make treatment decisions considering this balance. To make accurate predictions, an informative representation of tumor location is pivotal, yet comparisons of representations are lacking. Therefore, this study compares brain atlases and principal component analysis (PCA) to represent voxel-wise tumor location. Pre-operative cognitive functioning was predicted for 246 patients with a high-grade glioma across eight cognitive tests while using different representations of voxel-wise tumor location as predictors. Voxel-wise tumor location was represented using 13 different frequently-used population average atlases, 13 randomly generated atlases, and 13 representations based on PCA. ElasticNet predictions were compared between representations and against a model solely using tumor volume. Preoperative cognitive functioning could only partly be predicted from tumor location. Performances of different representations were largely similar. Population average atlases did not result in better predictions compared to random atlases. PCA-based representation did not clearly outperform other representations, although summary metrics indicated that PCA-based representations performed somewhat better in our sample. Representations with more regions or components resulted in less accurate predictions. Population average atlases possibly cannot distinguish between functionally distinct areas when applied to patients with a glioma. This stresses the need to develop and validate methods for individual parcellations in the presence of lesions. Future studies may test if the observed small advantage of PCA-based representations generalizes to other data.

Patients with meningiomas frequently exhibit impairments in executive functioning. There are few studies specifically examining the role of frontal meningioma localization in executive functioning impairments. This study examines whether frontally located meningiomas are specifically associated with executive functioning impairments in a large sample of meningioma patients before treatment, using an axis-wise and lobe-based approach to meningioma localization. We retrospectively examined cognitive performances in 353 patients with frontal, frontally-involved and non-frontal meningiomas on a battery of tests including tests of executive functioning. We applied an axis-based approach to meningioma location, in addition to qualitative lobe-based localization. We examined the association between meningioma coordinates on an anterior-posterior axis and continuous cognitive performance scores in univariate correlations and linear regression analyses. We also examined the association between meningioma coordinates on an anterior-posterior axis with cognitive impairments in multivariable logistic regression analyses. Meningioma position on the anterior-posterior axis was only univariately associated with mean performance on the Stroop test Interference ratio and Symbol Digit Coding task. There was no (multivariable) association with impairments on tests of executive or non-executive domains. Increased odds of impairment on executive functioning tasks were associated with left-localization (Verbal Fluency) and larger meningioma volumes (Shifting Attention). We did not find a specific relation between a frontal meningioma location and executive functioning impairments, which may be explained by widespread organization of executive functioning throughout the brain, diffuse cognitive effects of the mass of meningiomas, functional reorganization due to neuroplasticity, or functional involvement of less-anteriorly located frontal areas.

Introduction : Cognitive functioning is increasingly considered when making treatment decisions for patients with a brain tumor in view of a personalized onco-functional balance. Ideally, one can predict cognitive functioning of individual patients to make treatment decisions considering this balance. To make accurate predictions, an informative representation of tumor location is pivotal, yet comparisons of representations are lacking. Therefore, this study compares brain atlases and principal component analysis to represent voxel-wise tumor location. Methods Pre-operative cognitive functioning was predicted for 246 patients with a high-grade glioma across eight cognitive tests while using different representations of voxel-wise tumor location as predictors. Voxel-wise tumor location was represented using 13 different frequently-used population average atlases, 13 randomly generated atlases, and 13 representations based on principal component analysis. ElasticNet predictions were compared between representations and against a model solely using tumor volume as a predictor. Results Preoperative cognitive functioning could only partly be predicted from the representations of tumor location. Representations based on principal component analysis resulted in better predictions compared to those based on population average- and random atlases. Population average atlases did not result in better predictions when compared to random atlases. Discussion For the prediction of cognitive performances, representations based on principal component analysis should be favored over population average atlases for prediction if the decrease in interpretability is acceptable. Population average atlases possibly cannot distinguish between functionally distinct areas when applied to patients with a glioma. Researchers should develop and validate methods for individual parcellations in the presence of lesions.

Background: Previous research identified many clinical variables that are significantly related to cognitive functioning before surgery. It is not clear whether such variables enable accurate prediction for individual patients’ cognitive functioning because statistical significance does not guarantee predictive value. Previous studies did not test how well cognitive functioning can be predicted for (yet) untested patients. Furthermore, previous research is limited in that only linear or rank-based methods with small numbers of variables were used. Methods: We used various machine learning models to predict preoperative cognitive functioning for 340 patients with glioma across 18 outcome measures. Predictions were made using a comprehensive set of clinical variables as identified from the literature. Model performances and optimized hyperparameters were interpreted. Moreover, Shapley additive explanations were calculated to determine variable importance and explore interaction effects. Results: Best-performing models generally demonstrated above-random performance. Performance, however, was unreliable for 14 out of 18 outcome measures with predictions worse than baseline models for a substantial number of train-test splits. Best-performing models were relatively simple and used most variables for prediction while not relying strongly on any variable. Conclusions: Preoperative cognitive functioning could not be reliably predicted across cognitive tests using the comprehensive set of clinical variables included in the current study. Our results show that a holistic view of an individual patient likely is necessary to explain differences in cognitive functioning. Moreover, they emphasize the need to collect larger cross-center and multimodal data sets.