目的：本研究拟基于ATX(N)生物标志物的框架：复现并优化基于液相色谱串联质谱（liquid chromatography–tandem mass spectrometry, LC–MS/MS）技术测定脑脊液（cerebrospinal fluid, CSF）β-淀粉样蛋白（β-amyloid, Aβ）1-42和Aβ1-40的经典方法，并基于磁珠辅助提取（magnetic-bead-assisted extraction, MBAE）前处理技术推动检测的自动化；比较自建LC–MS/MS方法和最常用的3种免疫学检测方法测定CSF核心AT(N)标志物的性能和结果；基于3种免疫学检测方法分别建立阿尔兹海默型痴呆（Alzheimer's disease dementia, ADD）患者的诊断切点值，并基于临床实验室真实世界数据评价其适用性，以期推动CSF核心AT(N)标志物在国内临床的应用；研制CSF Aβ1-42的定值质控品，基于最常用的免疫学检测方法定值，以期推动检测的一致化。此外，建立并优化基于LC–MS/MS技术测定CSF胶质纤维酸性蛋白（glial fibrillary acidic protein, GFAP）的方法，为该标志物的进一步研究提供可靠的检测手段；评估常见方法测定CSF和血浆GFAP的一致性，并初步探索CSF和血浆GFAP的临床应用价值。

方法：本研究通过文献综述的方式荟萃分析了既往发表的针对AD连续体患者建立CSF Aβ1-42和/或Aβ1-42/Aβ1-40诊断切点值的研究，识别了限制AD核心CSF标志物在临床广泛应用的关键点。基于传统的固相萃取（solid-phase extraction, SPE）和创新的MBAE前处理技术分别建立和优化了同时测定CSF Aβ1-42和Aβ1-40的LC–MS/MS方法，参考美国临床和实验室标准协会（Clinical & Laboratory Standards Institute, CLSI）指南进行方法学评价，并收集了68例CSF样本进行两种前处理方法的一致性评估。采集254例CSF样本用于自建LC–MS/MS方法与实验室最常使用的3种免疫学检测方法，即INNOTEST酶联免疫吸附（enzyme linked immunosorbent assay, ELISA）检测系统、Lumipulse和Roche Elecsys全自动化学发光免疫分析系统，测定核心CSF标志物Aβ1-42、Aβ1-40、总tau蛋白（total tau protein, t-tau）和181磷酸化tau蛋白（phosphorylated tau protein, p-tau181）的比对研究。自神经科招募ADD患者176人和年龄、性别、教育年限匹配的非AD型痴呆（non-Alzheimer degenerative dementia, NADD）患者114人，采用上述3种免疫学检测方法，基于受试者工作特征曲线（receiver operating characteristic curve, ROC）分析的原理，分别建立核心CSF标志物的诊断切点值并评估其诊断效能。选择实验室最多使用的Lumipulse全自动化学发光免疫分析系统开展临床样本检测，依据临床实验室标准进行质量管理，参与阿尔兹海默病协会质量控制（Alzheimer's Association Quality Control, AAQC）计划评估实验室检测能力，并基于临床样本的检测结果评估自建切点值的适用性。参考CLSI C37-A等指南文件，研制3个水平的CSF Aβ1-42的质控材料，评价其均匀性和稳定性，并基于Lumipulse全自动化学发光免疫分析系统进行赋值。

此外，为了评估2023年修正版指南草案中新推荐的标志物GFAP在AD诊断中的应用价值，本研究参考Cochrane手册及PRISMA 2020指南，开展了CSF和外周血GFAP对于AD连续体患者诊断价值研究的meta分析。基于SPE–LC–MS/MS技术建立和优化了CSF GFAP的检测方法，参照CLSI指南进行方法学评价，并采集75例CSF样本初步评估其与两个常见免疫学检测方法，即Quanterix SIMOA检测系统和迈克全自动化学发光免疫分析系统的一致性；并基于129例CSF样本初步探索了CSF GFAP的临床应用价值。此外，自神经科采集血浆样本302例，评估常用的3个免疫学检测方法，即Quanterix SIMOA检测系统、美联泰科磁微粒化学发光检测系统和迈克全自动化学发光免疫分析系统测定血浆GFAP的一致性，并基于139例ADD患者和116人年龄、性别、教育年限匹配的NADD患者探索其临床应用价值。

结果：文献综述分析发现限制AD核心CSF标志物在临床广泛应用的关键点包括可用的准确可靠且自动化程度较高的检测方法缺失、不同检测系统测定结果的一致性较差、统一的诊断切点值缺失、检测的标准化推进不足等。因此，本研究首先基于传统的SPE前处理技术和创新的MBAE前处理技术分别建立了准确、简便地同时测定CSF Aβ1-42和Aβ1-40的检测方法，方法总不精密度分别为3.7%~7.4%、3.0%~4.6%，定量限分别为0.1 ng/mL、0.05 ng/mL。与传统的SPE–LC–MS/MS方法相比，本研究首次基于创新的MBAE前处理技术建立的半自动化的LC–MS/MS方法更为简便、手工步骤更少、耗时更短（96 vs. 140 min）、成本更低（96个检测：34.98 vs. 493.96美元），与SPE–LC–MS/MS方法检测结果的一致性较好，该方法有望开发成为全自动化的检测方法，推动LC–MS/MS技术在临床检测中的应用。此外，我们首次系统地比较了自建的LC–MS/MS方法与目前实验室最常用的3种免疫学检测方法对于4个核心CSF标志物的检测情况，发现尽管采用不同检测方法测定同一标志物的结果的相关性较强（r: 0.878~0.971），但测定值存在显著差异，其中，基于LC–MS/MS方法测定的CSF Aβ1-42、Aβ1-40和t-tau的浓度均显著高于免疫学方法的检测结果。通过测定Aβ1-42的有证标准物质（certified reference material, CRMs），发现仅实验室自建的LC–MS/MS方法可溯源至参考测量程序，而基于3种免疫学检测方法的测定结果均低于CRMs的目标值并超出了允许不确定度的范围。进一步的校正实验显示，不同方法间检测结果的差异很难通过实验室内部建立的线性关系实现完全的校正。在标准化尚未实现的当下，仍需建立检测方法特异性的切点值。因目前的LC–MS/MS方法所需样本量较大且需两种方法才能实现Aβ1-42、Aβ1-40和t-tau共3个标志物的检测，因此，本研究进一步采用3种最常用的免疫学方法，首次基于PUMCH痴呆队列中的ADD患者（n = 176）和NADD患者（n = 114），根据Youden指数最大的原理分别建立了基于不同检测方法的各指标的诊断切点值。CSF Aβ1-42/Aβ1-40和p-tau181/Aβ1-42的比值呈现优于单指标的诊断准确性；基于INNOTEST ELISA检测系统和Lumipulse全自动化学发光免疫分析系统建立的各标志物的切点值较为接近，且与制造商来源的切点值较为接近；而基于Roche全自动化学发光免疫分析系统建立的CSF Aβ1-42的切点值显著高于其他两个检测平台的结果、CSF t-tau和p-tau181的切点值则显著低于其他两个检测平台的结果，且基于本研究建立的3个核心CSF标志物的切点值均显著低于制造商基于Aβ正电子发射断层扫描（positron emission tomography, PET）分组建立的切点值。此外，基于临床诊断需求，进一步建立了固定特异度为80%、85%、90%、95%和97.5%时CSF Aβ1-42、t-tau、p-tau181及其常见比值的切点值。基于3种免疫学检测方法可测定的所有指标联合诊断ADD患者和NADD患者的准确性无显著差异，因此，选择了最常用的Lumipulse全自动化学发光免疫分析系统，首次面向PUMCH神经科患者，以临床科研报告的形式开展了4个核心CSF标志物的检测。通过分析66例临床诊断明确的患者的核心CSF标志物的临床检测结果，进一步评价了制造商来源的切点值和实验室自建的切点值的适用性，当区分ADD和NADD患者时，对于CSF Aβ1-42、t-tau、p-tau181、t-tau/Aβ1-42和p-tau181/Aβ1-42，可采用制造商推荐的鉴别诊断的切点值，分别为599 pg/mL、404 pg/mL、56.5 pg/mL、0.784和0.123；对于CSF Aβ1-42/Aβ1-40，可采用实验室基于ROC分析自建的Youden指数最大时的切点值，为0.057。为了保证临床检测结果的准确可靠，我们通过室内质控图监测不同批次间检测结果的稳定性，并通过参与国际AAQC室间质量评价计划来评估实验室的检测能力。此外，为进一步推动AD核心CSF标志物在国内检测的一致化，本研究基于临床剩余CSF样本研制了3个水平的Aβ1-42的定值质控品，该质控品均匀性、稳定性良好，并基于Lumipulse全自动化学发光免疫分析系统为质控品定值，其浓度范围分别为89.73~110.02 pg/mL/125.16~151.34 pg/mL、394.36~461.48 pg/mL和772.22~852.20 pg/mL，并同时测定了Aβ1-40的浓度供使用者参考。

此外，本研究关注了美国国家衰老研究所—阿尔兹海默协会发布的2023年修订版指南草案中新推荐的标志物GFAP，首次基于诊断性meta分析的手段评估了CSF和外周血GFAP对于AD连续体患者诊断的准确性，发现GFAP（特别是外周血GFAP）对于AD的早期诊断、鉴别诊断和预后评估具有较优的性能，联合性别、年龄、ApoE基因型等基线信息可进一步提升其诊断准确性。本研究首次基于LC–MS/MS技术建立了CSF GFAP的检测方法，基于酶切后的特征性肽段SVSEGHLK进行表征定量，方法总不精密度为6.4%~7.9%，定量限为0.5 ng/mL，性能良好，并有望进一步开发外周血GFAP的检测方法。基于自建LC–MS/MS方法测定的CSF GFAP的浓度与Quanterix SIMOA免疫检测系统和迈克全自动化学发光免疫分析系统测定结果的相关性较好（r: 0.824~0.826），且与Quanterix SIMOA免疫检测系统的测定值较为接近。初步临床应用评估发现，基于自建LC–MS/MS方法测定的CSF GFAP的浓度分布在Aβ-PET阳性组和简易精神状态检查表（minimum mental state examination, MMSE）评分阳性组显著高于阴性组，在核心CSF标志物的基础上联合CSF GFAP可进一步提高Aβ-PET阳性组和阴性组、MMSE评分阳性组和阴性组的诊断效能，但无显著统计学差异，提示仍需基于更大样本量人群的评估与验证。此外，评估了基于3种常见免疫学检测方法测定血浆GFAP和神经丝蛋白轻链（neurofilament light chain, NfL）的一致性，发现不同检测方法测定结果的相关性较强（r: 0.882~0.947）但测定值存在显著差异。联合血浆GFAP和NfL可实现ADD组和NADD组间较好的鉴别诊断准确性（AUC = 0.829），与血浆核心标志物Aβ1-42、p-tau181、p-tau217进一步联合亦可进一步显著提升其诊断准确性（AUC: 0.924 vs. 0.898, p = 0.037）。

结论：本研究通过文献综述的方式荟萃调研了阻碍AD核心CSF标志物临床应用的关键点，在此基础上基于传统的SPE技术和创新的MBAE技术建立了准确、简便地同时测定CSF Aβ1-42和Aβ1-40的LC–MS/MS方法，并有望开发成为全自动化的检测方法；方法学比较研究发现基于4种检测方法测定4个核心CSF标志物的相关性较好，但绝对值存在差异，且该差异很难通过实验室内部的系数调整实现完全校正。因此，进一步基于不同的检测方法分别建立了用于ADD和NADD患者鉴别诊断的切点值，并通过临床真实世界数据的检测结果评价了制造商推荐的切点值和实验室自建切点值的适用性，给出特定使用场景下推荐的切点值。为推动不同实验室间检测结果的一致化，研制了国内首个CSF基质的Aβ1-42的定值质控品，并同步为Aβ1-40进行赋值。此外，本研究也关注到了ATX(N)框架中新提出的标志物GFAP，首次通过诊断性meta分析的手段评价其对于AD早期诊断、鉴别诊断、前瞻性诊断等过程的价值，并首次基于LC–MS/MS技术建立了CSF GFAP准确定量的方法，为外周血GFAP LC–MS/MS方法的建立奠定了基础。并初步评估了基于常见方法测定CSF和外周血GFAP的一致性及临床应用价值。综上，本研究基于ATX(N)框架开展了针对AD重要生物标志物的一系列研究，以期推动其准确、简便的测定，推广其临床应用，助力AD患者早期、精准的诊疗。

Aim: Based on the ATX(N) biomarker framework, the aims of this study include: to establish and optimize the classic liquid chromatography–tandem mass spectrometry (LC–MS/MS) method for determination of cerebrospinal fluid (CSF) β-amyloid (Aβ)1-42 and Aβ1-40, and further prompt the automation of the method based on the magnetic-bead-assisted extraction (MBAE) pre-treatment technology; compare the performance and measurements between the self-established LC–MS/MS method and the three most commonly used immunological methods for measuring CSF core AT(N) biomarkers, establish diagnostic cut-offs for Alzheimer’s disease dementia (ADD) based on three immunological methods, and evaluate the applicability of self-established cut-offs using real-world data from the clinical laboratory, to prompt the wide application of CSF core AT(N) biomarkers in domestic clinical laboratories; develop quality control materials for CSF Aβ1-42 based on the most commonly used immunological method, to promote the testing consistency. Besides, establish and optimize an LC–MS/MS method for the determination of CSF glial fibrillary acidic protein (GFAP), to prove a robust method for further study; evaluate the consistency of common methods for measuring CSF and plasma GFAP; and preliminarily explore the values of GFAP in CSF and plasma for clinical applications.

Methods: This study first conducted a review analysis of previous studies that established diagnostic cut-offs of CSF Aβ1-42 and/or Aβ1-42/Aβ1-40 for AD continuum patients, and summarized the key limitations hindering the widespread clinical application of core CSF biomarkers for AD. Thus, using both traditional solid-phase extraction (SPE) and innovative MBAE pre-treatment technology, we separately established and optimized an LC–MS/MS method for the simultaneous determination of CSF Aβ1-42 and Aβ1-40. Methodological evaluation was conducted referring to the Clinical & Laboratory Standards Institute (CLSI) guidelines, and a consistency evaluation of the two pre-treatment methods was performed by collecting 68 CSF samples. By collecting a total of 254 CSF samples, we compared the self-established LC–MS/MS method and the three most used immunological methods in laboratories, named INNOTEST enzyme linked immunosorbent assay (ELISA) detection system, Lumipulse and Roche Elecsys fully automated chemiluminescence immune detection system, for the determination of core CSF markers, including Aβ1-42, Aβ1-40, total tau protein (t-tau), and 181 phosphorylated tau protein (p-tau181). A total of 176 ADD patients and 114 non-Alzheimer degenerative dementia (NADD) patients, matched for age, sex, and educational years, were recruited from the Department of Neurology. Three immunological methods were utilized to establish separate cut-offs for CSF core biomarkers and evaluate their diagnostic values based on the receiver operating characteristic curve (ROC) analysis. The Lumipulse fully automated chemiluminescence immune detection system, the most used method in laboratories, was chosen for clinical testing. We implemented quality management procedures according to clinical laboratory standards and participated in the Alzheimer Association’s quality control (AAQC) program to verify testing reliability, and further evaluated the applicability of the self-established cut-offs based on the results of clinical samples. Referring to CLSI C37-A guideline, we developed three levels of CSF Aβ1-42 quality control materials for consistency evaluation and quality control purposes. The uniformity and stability of controls were evaluated, and concentrations were assigned based on the Lumipulse fully automatic chemiluminescence immune detection system.

Additionally, to assess the clinical values of the newly recommended GFAP by the 2023 revised guideline draft of National Institute on Aging-Alzheimer's Association, we conducted a meta-analysis on the diagnostic value of CSF and blood GFAP in AD continuum patients, referencing to the Cochrane Handbook and PRISMA 2020 Guideline. We further established and optimized an SPE–LC–MS/MS method for the determination of CSF GFAP and conducted methodological evaluations referring to the CLSI guidelines. The consistency between the self-established SPE–LC–MS/MS method and two common immunological methods, namely the Quanterix SIMOA detection system and Maccura fully automated chemiluminescence immune detection system, was preliminarily evaluated by collecting 75 CSF samples. The clinical application value was also preliminarily explored based on 129 CSF samples. Additionally, 302 plasma samples were collected from the Department of Neurology to evaluate the measurement consistency of plasma GFAP across three commonly used immunological methods, namely the Quanterix SIMOA detection system, Sophonix magnetic particle chemiluminescence detection system, and Maccura fully automated chemiluminescence immune detection system. The clinical application value of plasma GFAP was further explored based on 139 ADD patients and 116 NADD patients, matched for age, sex, and education years.

Results: The review analysis summarized the key limitations hindering the widespread clinical application of core CSF biomarkers for AD, including the lack of available accurate, reliable, and highly automated detection methods, poor consistency among results from different detection methods, absence of unified diagnostic cutoffs, and insufficient progress in standardization of detection. Therefore, this study initially established an accurate and simple LC–MS/MS method for the simultaneous measurement of CSF Aβ1-42 and Aβ1-40 based on both traditional SPE and innovative MBAE pre-treatment technology. The total imprecision of the method ranged from 3.7% to 7.4% for Aβ1-42 and 3.0% to 4.6% for Aβ1-40, and the limit of quantitation was 0.1 ng/mL and 0.05 ng/mL, respectively. This study, for the first time, established a semi-automated LC–MS/MS method based on the innovative MBAE pre-treatment technology. Compared to the traditional SPE–LC–MS/MS method, this method is simpler, requires fewer manual steps, takes less time (96 vs. 140 min), and is more cost-effective (96 tests: $34.98 vs. $493.96). Furthermore, the results of MBAE–LC–MS/MS showed good consistency with traditional SPE–LC–MS/MS method but shows greater potential to be developed into a fully automated detection method, thereby promoting the clinical application of LC–MS/MS technology. Additionally, we systematically compared the results based on our laboratory-developed LC–MS/MS method and the three most used immunological methods in laboratories for determining four core CSF biomarkers. We found a strong correlation between the results of same biomarkers measured by different detection methods (r: 0.878~0.971), but significant differences of measured values. Specifically, the concentrations of CSF Aβ1-42, Aβ1-40, and t-tau determined using the LC–MS/MS method were significantly higher than those obtained using immunological methods. By measuring the certified reference materials (CRMs) of CSF Aβ1-42, we found that only our laboratory-developed LC–MS/MS method could be traceable to the reference measurement procedure; however, the detection results from the three immunological methods were all below the target values of the CRMs and exceeded the allowable uncertainty ranges. Further calibration experiments showed that it was difficult to achieve complete calibration of the measurement results among different methods through linear relationships established within the laboratory. In the absence of standardization, it is still necessary to establish specific cutoffs based on different detection methods. Because the current LC–MS/MS method required larger sample volumes, and two separate methods could only detect three biomarkers including Aβ1-42, Aβ1-40 and t-tau, this study further utilized the three most used immunological methods, for the first time, selected patients with ADD (n=176) and NADD (n=114) from the PUMCH dementia cohort. Based on the principle of maximizing the Youden index, this study separately established cutoffs for various biomarkers based on different detection methods. The ratios of CSF Aβ1-42/Aβ1-40 and p-tau181/Aβ1-42 shown superior diagnostic accuracy compared to single biomarkers. Cutoffs of several biomarkers established using the INNOTEST ELISA detection system and the Lumipulse fully automated chemiluminescence immune detection system were relatively close and aligned closely with the cut-offs from the manufacturers; however, the cutoff established using the Roche fully automated chemiluminescence detection system for CSF Aβ1-42 was significantly higher, while cutoffs for CSF t-tau and p-tau181 were significantly lower than those from the other two platforms. Moreover, the cutoffs for the three core biomarkers established using Roche fully automated chemiluminescence immune detection system in this study were significantly lower than those established by the manufacturer based on Aβ positron emission tomography (PET) defined groups. Additionally, to meet clinical diagnostic needs, we further established cutoffs for CSF Aβ1-42, t-tau, p-tau181, and their common ratios at fixed specificities of 80%, 85%, 90%, 95%, and 97.5%. The study found no significant difference in the diagnostic accuracy for distinguishing ADD patients and NADD patients using all biomarkers that can be measured by the three immunological methods. Therefore, the most used Lumipulse fully automated chemiluminescence immune detection system was selected, for the first time, performed the detection of four core CSF biomarkers for patients from the Department of Neurology of PUMCH in the form of clinical research reports. By analyzing the clinical measurement results of core CSF biomarkers from 66 clinically diagnosed patients, we further evaluated the applicability of the cut-offs from the manufacturers and cutoffs established by our laboratory. When distinguishing patients with ADD and NADD, we recommended using the manufacturer-recommended cut-offs for CSF Aβ1-42, t-tau, p-tau181, t-tau/Aβ1-42, and p-tau181/Aβ1-42, which were 599 pg/mL, 410 pg/mL, 59 pg/mL, 1.051, and 0.160, respectively; for CSF Aβ1-42/Aβ1-40, we recommended using the cutoff established by maximizing the Youden index from our laboratory data, which was 0.057. To ensure the accuracy and reliability of clinical measurement results, we monitored the stability of measurements among different batches through internal quality control charts and assessed the laboratory's testing capabilities by participating in the international AAQC external quality assessment program. Additionally, to further promote the standardization of AD core CSF biomarker testing domestically, this study developed three levels of quality control materials for Aβ1-42 based on residual clinical CSF samples. These QC materials exhibit good uniformity and stability and were assigned values using the Lumipulse fully automated chemiluminescent immune detection system, and the concentration ranges were 89.73~110.02 pg/mL or 125.16~151.34 pg/mL, 394.36~461.48 pg/mL, and 772.22~852.20 pg/mL, respectively, moreover, Aβ1-40 concentrations were also concurrently measured as reference (AUC: 0.924 vs. 0.898, p=0.037).

Furthermore, this study also focused on GFAP, the novel biomarker recommended in the revised 2023 version of the guideline draft by National Institute on Aging-Alzheimer's Association. For the first time, we evaluated the accuracy of CSF and blood GFAP in diagnosing AD continuum patients using diagnostic meta-analysis method, and found that GFAP, especially blood GFAP, show good performance in the early diagnosis, differential diagnosis, and prognosis assessment of AD. Combining baseline characteristics such as sex, age, and ApoE genotype can further enhance its diagnostic accuracy. Furthermore, this study established, for the first time, a detection method for CSF GFAP using LC–MS/MS technology, quantifying it based on the enzymatically cleaved characteristic peptide SVSEGHLK. The method exhibited a total imprecision of 6.4% to 7.9% and a limit of quantification of 0.5 ng/mL, demonstrating good performance and the potential for further development for the measurements of blood GFAP. The concentration of CSF GFAP measured using the laboratory-developed LC–MS/MS method showed good correlation (r: 0.824~0.826) with results from the Quanterix SIMOA immunoassay system and the Maccura fully automated chemiluminescent immune detection system, and the measurements based on the LC–MS/MS method were close to that determined by the Quanterix SIMOA immunoassay system. Preliminary clinical application assessments revealed that the concentrations of CSF GFAP determined using the LC–MS/MS method were significantly higher in the Aβ-PET positive group and minimum mental state examination (MMSE) positive group compared to the negative groups. When combined with core CSF biomarkers, CSF GFAP could further improve the diagnostic performance of both the Aβ-PET positive and negative groups, as well as the MMSE positive and negative groups, although without significant statistical differences. Noticeably, further evaluation and validation based on a larger sample size are still required. Additionally, we evaluated the consistency of plasma GFAP and neurofilament light chain (NfL) measurements using three common immunoassay methods, finding good correlations between the results from different platforms (r: 0.882~0.947), but significant differences in measured values. Combining plasma GFAP and NfL shown good differential diagnostic capability for ADD and NADD groups (AUC=0.829), and the diagnostic accuracy would be further significantly improved when combined with plasma core biomarkers Aβ1-42, p-tau181, and p-tau217 (AUC：0.924 vs. 0.898, p=0.037).

Conclusion: This study first conducted a comprehensive review to analyze the key obstacles hindering the clinical application of AD core CSF biomarkers, and thus established accurate and simple LC–MS/MS detection methods for simultaneous measurement of CSF Aβ1-42 and Aβ1-40 using both traditional SPE and innovative MBAE techniques, which hold promise for further development into full-automated detection method. The methods comparison revealed good correlations among four commonly used methods for measuring four AT(N) core CSF biomarkers; however, there were significant discrepancies in measurement values, and completely correcting these differences by laboratory self-established proportional relationship proved challenging. Therefore, further efforts were made to establish cutoffs for differential diagnosis of ADD and NADD patients based on different detection methods. The applicability of cut-offs from manufacturer and laboratory-established cutoffs was evaluated through real-world data from the clinical laboratory, eventually providing recommended cutoffs for specific situation. To promote harmonization of results across different laboratories, the first domestically produced CSF matrix QC materials for Aβ1-42 was developed, with concurrent assigned values for Aβ1-40. Furthermore, GFAP, the newly proposed biomarker within the ATX(N) framework, attracted our attention. We, for the first time, conducted a diagnostic meta-analysis to evaluate its values in AD for early diagnosis, differential diagnosis, prognostic diagnosis et al. Thus, for the first time, an accurate method for CSF GFAP was established based on LC–MS/MS technology, laying the groundwork for the establishment of blood GFAP method. Preliminary assessments of methods consistency and clinical application value of common methods for measuring CSF and blood GFAP were also conducted. In summary, we conducted a series of studies and analysis on important AD biomarkers based on the ATX(N) framework, aiming to promote their accurate and widely measurement in clinical application, and contribute to early and precise diagnosis and treatment of AD patients.