From the Univ. Bordeaux (V.B., G.C., C.D.), Inserm, Bordeaux Population Health, UMR1219, Bordeaux; CIC 1401 EC (V.B., G.C., C.D.), Pôle Santé Publique, CHU de Bordeaux; Laboratory of Immunology and Immunogenetics (I.P.), Resources Biological Center (CRB), CHU Bordeaux; Univ. Bordeaux (I.P.), CNRS, ImmunoConcEpT, UMR 5164, Bordeaux; Alzheimer Research Center IM2A (B.D.), Salpêtrière Hospital, AP-HP, Sorbonne University, Paris; Univ. Bordeaux (V.P.), CNRS, Institut des Maladies Neuroégénératives, UMR 5293, Bordeaux; Pôle de Neurosciences Cliniques (V.P.), Centre Mémoire de Ressources et de Recherche, CHU Bordeaux, France.
Neurology. 2024 Apr 23;102(8):e209219. doi: 10.1212/WNL.0000000000209219. Epub 2024 Mar 25.
Patients' comorbidities can affect Alzheimer disease (AD) blood biomarker concentrations. Because a limited number of factors have been explored to date, our aim was to assess the proportion of the variance in fluid biomarker levels explained by the clinical features of AD and by a large number of non-AD-related factors.
MEMENTO enrolled 2,323 individuals with cognitive complaints or mild cognitive impairment in 26 French memory clinics. Baseline evaluation included clinical and neuropsychological assessments, brain MRI, amyloid-PET, CSF (optional), and blood sampling. Blood biomarker levels were determined using the Simoa-HDX analyzer. We performed linear regression analysis of the clinical features of AD (cognition, AD genetic risk score, and brain atrophy) to model biomarker concentrations. Next, we added covariates among routine biological tests, inflammatory markers, demographic and behavioral determinants, treatments, comorbidities, and preanalytical sample handling in final models using both stepwise selection processes and least absolute shrinkage and selection operator (LASSO).
In total, 2,257 participants were included in the analysis (median age 71.7, 61.8% women, 55.2% with high educational levels). For blood biomarkers, the proportion of variance explained by clinical features of AD was 13.7% for neurofilaments (NfL), 11.4% for p181-tau, 3.0% for Aβ-42/40, and 1.4% for total-tau. In final models accounting for non-AD-related factors, the variance was mainly explained by age, routine biological tests, inflammatory markers, and preanalytical sample handling. In CSF, the proportion of variance explained by clinical features of AD was 24.8% for NfL, 22.3% for Aβ-42/40, 19.8% for total-tau, and 17.2% for p181-tau. In contrast to blood biomarkers, the largest proportion of variance was explained by cognition after adjustment for covariates. The covariates that explained the largest proportion of variance were also the most frequently selected with LASSO. The performance of blood biomarkers for predicting A+ and T+ status (PET or CSF) remained unchanged after controlling for drivers of variance.
This comprehensive analysis demonstrated that the variance in AD blood biomarker concentrations was mainly explained by age, with minor contributions from cognition, brain atrophy, and genetics, conversely to CSF measures. These results challenge the use of blood biomarkers as isolated stand-alone biomarkers for AD.
患者的合并症可能会影响阿尔茨海默病（AD）的血液生物标志物浓度。由于迄今为止仅探索了有限数量的因素，因此我们的目的是评估 AD 的临床特征以及大量非 AD 相关因素对液体生物标志物水平的方差的解释比例。
MEMENTO 纳入了 26 家法国记忆诊所的 2323 名有认知主诉或轻度认知障碍的个体。基线评估包括临床和神经心理学评估、脑 MRI、淀粉样蛋白-PET、CSF（可选）和血液采样。使用 Simoa-HDX 分析仪测定血液生物标志物水平。我们对 AD 的临床特征（认知、AD 遗传风险评分和脑萎缩）进行线性回归分析，以建立生物标志物浓度模型。接下来，我们使用逐步选择过程和最小绝对收缩和选择算子（LASSO）在最终模型中添加常规生物学检测、炎症标志物、人口统计学和行为决定因素、治疗、合并症以及分析前样本处理中的协变量。
共有 2257 名参与者纳入分析（中位年龄 71.7 岁，61.8%为女性，55.2%具有高教育水平）。对于血液生物标志物，AD 临床特征对神经丝（NfL）、p181-tau 的解释比例分别为 13.7%、11.4%，Aβ-42/40 为 3.0%，总 tau 为 1.4%。在纳入非 AD 相关因素的最终模型中，主要由年龄、常规生物学检测、炎症标志物和分析前样本处理解释了方差。在 CSF 中，AD 临床特征对 NfL、Aβ-42/40、总 tau 和 p181-tau 的解释比例分别为 24.8%、22.3%、19.8%和 17.2%。与血液生物标志物不同，在调整协变量后，认知解释了最大比例的方差。在 LASSO 中，也选择了解释方差比例最大的协变量。在控制了变异驱动因素后，血液生物标志物对 A+和 T+状态（PET 或 CSF）的预测性能保持不变。
这项全面分析表明，AD 血液生物标志物浓度的方差主要由年龄解释，认知、脑萎缩和遗传学的贡献较小，与 CSF 测量结果相反。这些结果对血液生物标志物作为 AD 的孤立的独立生物标志物的使用提出了挑战。
