OS01.5.A THE IMPORTANCE OF METHOD IN EVALUATING COGNITIVE CHANGE IN IDH-MUTATED GRADE 2 GLIOMAS

Abstract BACKGROUND Accurately assessing cognitive change over time in patients with low-grade glioma is complicated by normal fluctuations related to test-retest variability, practice effects, and day-to-day performance differences. Traditional methods based on z-score changes (typically z 1) often fail to fully account for these sources of variance. In contrast, the Reliable Change Index (RCI) offers a more robust approach for determining whether observed cognitive changes exceed what would be expected by chance. However, it remains unclear to what extent these different methods yield divergent results. Therefore, we applied both z-score and RCI to estimate cognitive change and compare their outcomes. MATERIAL AND METHODS We followed 88 patients with IDH-mutated WHO grade 2 glioma, assessing cognitive function preoperatively and 12 months postoperatively using multiple neuropsychological tests. The proportion of patients with cognitive declines (∆z≥ -1) were calculated for each test. These results were then compared to those derived from the RCI (90% CI, corresponding to z ≥ 1.645), calculated using a matched control group of 88 healthy individuals assessed at corresponding intervals. In addition, changes using z ≥ 1.645 were calculated for comparison. RESULTS The proportion of patients classified as having a significant cognitive decline varied depending on the method used. In 9 out of 16 tests, the traditional z-score method (∆z≥ -1) overestimated the frequency of decline compared to RCI. This discrepancy may be attributed to the more lenient threshold used in the z-score approach (1 SD vs. 1.645 SD for RCI). However, when applying the same threshold to the z-score method (i.e. z ≤ -1.645), the method underestimated cognitive decline in 12 out of 16 tests relative to RCI. This may reflect the impact of practice effects, as we observed variable but substantial improvements in the control group (e.g. an increase of >1 word on Boston Naming Test (BNT) and a reduction of >3 seconds on Color-World Interference Test 4). Furthermore, in some tests, such as BNT, the proportion of patients classified as declined increased from 10.45% using z ≤ -1.645 to 29.85% with RCI. This illustrates a significant discrepancy between the two methods, even when identical thresholds are used. These findings underscore the importance of using RCI, as reliance on z-scores alone may mask cognitive decline due to unaccounted practice effects, potentially leading to false assumptions about post-surgical outcomes. CONCLUSION The use of RCI enhances the accuracy of longitudinal cognitive assessments in neuro-oncology. By accounting for expected variability in test performance, RCI minimizes false positives and better captures true cognitive change in patients with IDH-mutated glioma.