Zilioli A, Rosenberg A, Mohanty R, Matton A, Granberg T, Hagman G, Lötjönen J, Kivipelto M, Westman E
Department of Neurology, University-Hospital of Parma, Parma, Italy.
Population Health Unit, Finnish Institute for Health and Welfare, Helsinki, Finland.
J Neurol. 2024 Dec 21;272(1):84. doi: 10.1007/s00415-024-12853-9.
Predicting amyloid status is crucial in light of upcoming disease-modifying therapies and the need to identify treatment-eligible patients with Alzheimer's disease. In our study, we aimed to predict CSF-amyloid status and eligibility for anti-amyloid treatment in a memory clinic by (I) comparing the performance of visual/automated rating scales and MRI volumetric analysis and (II) combining MRI volumetric data with neuropsychological tests and APOE4 status. Two hundred ninety patients underwent a comprehensive assessment. The cNeuro cMRI software (Combinostics Oy) provided automated computed rating scales and volumetric analysis. Amyloid status was determined using data-driven CSF biomarker cutoffs (Aβ42/Aβ40 ratio), and eligibility for anti-Aβ treatment was assessed according to recent recommendations published after the FDA approval of the anti-Aβ drug aducanumab. The automated rating scales and volumetric analysis demonstrated higher performance compared to visual assessment in predicting Aβ status, especially for parietal-GCA (AUC = 0.70), MTA (AUC = 0.66) scores, hippocampal (AUC = 0.68), and angular gyrus (AUC = 0.69) volumes, despite low global accuracy. When we combined hippocampal and angular gyrus volumes with RAVLT immediate recall and APOE4 status, we achieved the highest accuracy (AUC = 0.82), which remained high even in predicting anti-Aβ treatment eligibility (AUC = 0.81). Our study suggests that automated analysis of atrophy rating scales and brain volumetry outperforms operator-dependent visual rating scales. When combined with neuropsychological and genetic information, this computerized approach may play a crucial role not only in a research context but also in a real-world memory clinic. This integration results in a high level of accuracy for predicting amyloid-CSF status and anti-Aβ treatment eligibility.
鉴于即将出现的疾病修饰疗法以及识别符合阿尔茨海默病治疗条件患者的必要性，预测淀粉样蛋白状态至关重要。在我们的研究中，我们旨在通过以下方式预测记忆门诊患者的脑脊液淀粉样蛋白状态及抗淀粉样蛋白治疗的适用性：（I）比较视觉/自动评分量表和MRI体积分析的性能；（II）将MRI体积数据与神经心理学测试及APOE4状态相结合。290名患者接受了全面评估。cNeuro cMRI软件（Combinostics Oy）提供自动计算评分量表和体积分析。使用数据驱动的脑脊液生物标志物临界值（Aβ42/Aβ40比值）确定淀粉样蛋白状态，并根据美国食品药品监督管理局批准抗淀粉样蛋白药物阿杜卡努单抗后发布的最新建议评估抗Aβ治疗的适用性。在预测Aβ状态方面，自动评分量表和体积分析显示出比视觉评估更高的性能，特别是对于顶叶-全球认知衰退量表（AUC = 0.70）、内侧颞叶萎缩（AUC = 0.66）评分、海马体（AUC = 0.68）和角回（AUC = 0.69）体积，尽管总体准确性较低。当我们将海马体和角回体积与雷氏听觉词语学习测验即时回忆及APOE4状态相结合时，我们获得了最高的准确性（AUC = 0.82），即使在预测抗Aβ治疗适用性时也保持较高水平（AUC = 0.81）。我们的研究表明，萎缩评分量表和脑容量的自动分析优于依赖操作者的视觉评分量表。当与神经心理学和遗传信息相结合时，这种计算机化方法不仅可能在研究背景中发挥关键作用，而且在现实世界的记忆门诊中也可能发挥关键作用。这种整合在预测淀粉样蛋白-脑脊液状态和抗Aβ治疗适用性方面具有很高的准确性。
