Predicting disease progression in high-grade glioma with neuropsychological parameters: the value of personalized longitudinal assessment.

Progressive disease in patients with high-grade glioma may be reflected in cognitive decline. However, the cognitive functions most sensitive to progression may differ between patients. We investigated whether decline on a personalized selection of tests predicted progressive disease according to RANO criteria in high-grade glioma patients. Starting one day before surgery, patients underwent neuropsychological assessment every three months during standard treatment and clinical follow-up. We first made a personalized selection of three tests that showed the highest Reliable Change Index (RCI) values, i.e., most positive change, at the first post-surgical assessment for each patient. In subsequent follow up, a decline of RCI ≤ - 1 on at least two of the three tests in the selection was considered cognitive decline. We performed a discrete Cox proportional hazards model including a time-dependent coefficient cognitive decline (vs. stability) and covariate age to predict progressive disease. Twenty five patients were included. Cognitive decline on the personalized test selection preceded or had occurred by the time progression was established in 9/15 patients with RANO confirmed progressive disease (60%). Decline was absent in 8/10 patients (80%) with stable disease during participation. The independent hazard ratio for progression in case of cognitive decline was 5.05 (p < 0.01) compared to stable performance. Using only three patient-specific neuropsychological tests, we found a fivefold increased chance of disease progression in case of cognitive decline as compared to stable performance. Brief, patient-tailored cognitive assessment may be a noninvasive addition to disease monitoring without overburdening patients and clinical care.