Plasma β-amyloid in Alzheimer’s disease and vascular disease

Implementation of amyloid biomarkers in clinical practice would be accelerated if such biomarkers could be measured in blood. We analyzed plasma levels of Aβ42 and Aβ40 in a cohort of 719 individuals (the Swedish BioFINDER study), including patients with subjective cognitive decline (SCD), mild cognitive impairment (MCI), Alzheimer’s disease (AD) dementia and cognitively healthy elderly, using a ultrasensitive immunoassay (Simoa platform). There were weak positive correlations between plasma and cerebrospinal fluid (CSF) levels for both Aβ42 and Aβ40, and negative correlations between plasma Aβ42 and neocortical amyloid deposition (measured with PET). Plasma levels of Aβ42 and Aβ40 were reduced in AD dementia compared with all other diagnostic groups. However, during the preclinical or prodromal AD stages (i.e. in amyloid positive controls, SCD and MCI) plasma concentration of Aβ42 was just moderately decreased whereas Aβ40 levels were unchanged. Higher plasma (but not CSF) levels of Aβ were associated with white matter lesions, cerebral microbleeds, hypertension, diabetes and ischemic heart disease. In summary, plasma Aβ is overtly decreased during the dementia stage of AD indicating that prominent changes in Aβ metabolism occur later in the periphery compared to the brain. Further, increased levels of Aβ in plasma are associated with vascular disease.

Scientific RepoRts | 6:26801 | DOI: 10.1038/srep26801 40/Aβ 42 ratio and Aβ ligand retention on PET was observed in APOE ε 4-negative subjects only 14 . The Australian Imaging Biomarkers and Lifestyle (AIBL) research team have reported that plasma levels of either Aβ 40 or Aβ 42 do not associate with AD or neocortical Aβ burden 15 . However, the Aβ 42/Aβ 40 ratio (note inverse ratio as compared to the ADNI results above) was slightly reduced in patients with AD and correlated inversely with amyloid burden as determined by amyloid PET.
The difficulties in getting consistent data could be, at least in part, related to poor performance and insufficient sensitivity of available analytical methods (mostly ELISA-based techniques) for adequate quantification of the minute amounts of Aβ present in peripheral blood 9 . The recently developed ultrasensitive Simoa technology offers improved analytical sensitivity 16 that makes it suitable for measurements of AD-related biomarkers in serum and plasma 17 . In the present study, we measured plasma levels of Aβ 42 and Aβ 40 using Simoa assays in a cohort of 719 individuals including patients with subjective cognitive decline (SCD, n = 174), mild cognitive impairment (MCI, n = 214), AD (n = 57) and cognitively healthy elderly (n = 274). We combined plasma measurements with the analysis of CSF samples, amyloid PET, magnetic resonance imaging (MRI) and cognitive assessments in order to establish whether plasma Aβ 42 and Aβ 40 may be useful biomarkers of AD.

Materials and Methods
Study populations. All participants gave written informed consent to participate in the study. Ethical approval was given by the Ethical Committee of Lund University, Lund, Sweden and all the methods were carried out in accordance with the approved guidelines. [ 18 F]flutemetamol PET imaging approval was obtained from the Swedish Medicines and Products Agency and the local Radiation Safety Committee at Skåne University Hospital, Sweden.
The study population stemmed from three cohorts from the prospective and longitudinal Swedish BioFINDER study (www.biofinder.se). The first cohort consisted of 274 cognitively normal elderly participants who were recruited from the population-based Malmö Diet Cancer study. Subjects were eligible for inclusion if they 1) were aged ≥ 60 years old, 2) scored 28-30 points on the Mini-Mental State Examination (MMSE) at the screening visit, 3) did not have cognitive symptoms as evaluated by a physician, 4) were fluent in Swedish, 5) did not fulfill the criteria of MCI or any dementia. The exclusion criteria were 1) presence of significant neurologic or psychiatric disease (e.g. stroke, Parkinson's disease, multiple sclerosis, major depression), 2) significant systemic illness making it difficult to participate, 3) refusing lumbar puncture and 4) significant alcohol abuse. Data was collected between 2009 and 2014 in accordance with a standardised protocol.
In the second cohort, 388 non-demented patients were enrolled consecutively at three memory outpatient clinics in Sweden. They were referred for assessment of cognitive complaints and evaluated by physicians with special interest in dementia disorders. Patients were included between 2010 and 2014. The inclusion criteria were: 1) referred to the memory clinics because of cognitive impairment; 2) not fulfilling the criteria for dementia; 3) an MMSE score of 24-30 points; 4) age 60-80 years and 5) fluent in Swedish. The exclusion criteria were: 1) cognitive impairment without doubt explained by another condition (other than prodromal dementia); 2) significant systemic illness making it difficult to participate, 3) refusing lumbar puncture and 4) significant alcohol abuse. Classification into SCD and MCI was based on a neuropsychological battery assessing the cognitive domains of verbal ability, visuospatial construction, episodic memory, executive functions and the clinical assessment by a senior neuropsychologist. These criteria resulted in a clinically relevant population where 45% were classified as SCD and 55% as MCI.
In the third cohort, we included 57 patients with AD at baseline, who were recruited consecutively at the Memory Clinic, Skåne University Hospital, Sweden between 2010 and 2014. The patients were assessed by a medical doctor specialized in dementia disorders. All cases met the DSM-IIIR criteria for dementia 18 as well as the NINCDS-ADRDA criteria for AD 19 . The exclusion criteria were: 1) significant systemic illness making it difficult to participate and 2) significant alcohol abuse.
In all three cohorts, a medical doctor made the diagnosis of hypotension, diabetes and ischemic heart disease. Ischemic heart disease was defined as stable angina, unstable angina, and myocardial infarction. Thiazide diuretics, calcium channel blockers, ACE inhibitors, angiotensin II receptor antagonists, and beta blockers were categorized as anti-hypertensive/cardio-protective medications.
Plasma and CSF sampling and analysis. Blood and CSF samples were collected on the same day and at the same time of day (plasma was obtained within 15 min of CSF sampling). For plasma collection, blood was drawn into tubes containing EDTA as anticoagulant. After centrifugation (2000 g, + 4 °C, 10 min), plasma samples were aliquoted into polypropylene tubes and stored at − 80 °C pending biochemical analyses. Lumbar CSF samples were collected according to a standardized protocol 2,20 . CSF samples were centrifuged (2000 g, + 4 °C, 10 min) after collection and aliquoted into polypropylene tubes followed by storage at − 80 °C. The analysis of CSF followed the Alzheimer's Association Flow Chart for CSF biomarkers 2 .
Plasma Aβ 42 and Aβ 40 were analyzed using ultrasensitive Simoa immunoassay (Quanterix, Lexington, MA, USA). In traditional sandwich ELISA of complex matrices such as plasma and serum, issues with spike recovery and lack of dilutional linearity (suggesting matrix interferences) have been reported 21,22 . The ultrasensitivity of the Simoa assays, allowing dilution of the plasma/serum samples at 1:4 minimizes these matrix effects. The overall benefits of the Simoa assays are not only its high sensitivity and precision, but also the elimination of matrix interferences. The Simoa Aβ 40 and Aβ 42 assays both utilize the same capture antibody targeting the N-terminus of β -amyloid and different C-terminus detection antibodies specific to Aβ 40 and Aβ 42. The Aβ 40 assay uses β -amyloid (1-40) peptide from AnaSpec (AnaSpec, Fremont, CA, USA) as standard and the Aβ 42 assay uses the β -amyloid (1-42) peptide from Covance (Covance Inc., Princeton, NJ, USA) as standard. For each assay, capture antibody was first covalently conjugated to magnetic particles utilizing a standard EDC coupling procedure and detection antibody was biotinylated. In the first step of the assay, antibody coated paramagnetic capture beads, biotinylated detection antibodies, and samples were combined, during which target molecules present in the sample were captured by the capture beads and labeled with the biotinylated detection antibodies. After washing, a conjugate of streptavidin-β -galactosidase (Sβ G) was mixed with the capture beads where Sβ G bound to the biotin, resulting in enzyme labeling of captured target molecules. Following a second wash, the capture beads were resuspended in a resorufin β -D-galactopyranoside (RGP) substrate solution and transferred to the Simoa array disc for detection 16 . All samples were diluted 4-fold for Aβ 42 and 8-fold for Aβ 40 using a proper sample diluent (PBS containing carrier protein and detergent) for measurement. The lower limit of detection (LLoD), defined as a concentration corresponding to a signal level of 2.5 SD above assay background, was 0.019 and 0.16 pg/mL for Simoa Aβ 42 and Aβ 40 assays, respectively. The lower of limit of quantification was 0.167 pg/ml for Aβ 42 (11% dose CV and 90% recovery) and 1.939 pg/mL for Aβ 40 (5% dose CV and 99% recovery). Mean spike recoveries of the Simoa Aβ 42 and Aβ 40 assays were 78.4% and 95.6%, respectively. Both Intra-assay (n = 3) and Inter-assay (n = 13) CVs were less than 10% for both assays. The average CV of measurement of Aβ 42 and Aβ 40 in all tested plasma samples during this study was 7% and 3%, respectively.
CSF levels of Aβ 42 and Aβ 40 were analyzed with Euroimmun immunoassay (EUROIMMUN AG, Lübeck, Germany) 23,24 in all study participants. This was done before the analysis of plasma samples using Simoa platform. Because CSF levels of Aβ obtained with Simoa and Euroimmun immunoassays strongly correlated in a subset of 69 of patients, (Aβ 42, Pearson's r = 0.912; Aβ 40, Pearson's r = 0.913; all p < 0.001, Supplementary Fig. S1) we used Euroimmun-derived CSF measurements of Aβ 42 and Aβ 40 in the present study.
Brain imaging. [ 25 . PET/CT scanning of the brain was conducted at two sites using the same type of scanner (Gemini, Philips Healthcare, Best, the Netherlands). Summed PET images from 90-110 min post injection representing the average uptake of [ 18 F]flutemetamol over this time were analyzed using NeuroMarQ software (provided by GE Healthcare, Cleveland, OH). A volume of interest (VOI) template was applied for the following 9 bilateral regions: prefrontal, parietal, lateral temporal, medial temporal, sensorimotor, occipital, anterior cingulate, posterior cingulate/precuneus and a global neocortical composite region 26 . The standardized uptake value ratio (SUVR) was defined as the uptake in a VOI normalized for the mean cerebellar cortex uptake.
Magnetic Resonance Imaging. A total of 620 individuals underwent MRI imaging including 266 control subjects, 161 SCD and 193 MCI patients. MR imaging was performed at a 3 T Siemens ® Trio system equipped with a standard 12 channel head coil. Axial T2 FLAIR (27 slices, voxel size 0.7 × 0.7 × 5.2 mm3), coronal GRE (25 slices, voxel size 0.9 × 0.9 × 6.5 mm3) and coronal MPRAGE (180 slices, voxel size 1 × 1 × 1.2 mm3) images were acquired. Visual rating of WML was performed according to the ARWMC scale (0-30 points) 27 . For statistical analysis, scores from the left and right hemispheres were summarized. The presence of cerebral microbleeds (CBM) was rated according to the MARS scale. This variable was dichotomized as CMB in any hemisphere, present or non-present 28 .
Statistical analyses. SPSS (IBM, Armonk, NY, US) was used for statistical analysis. Associations between plasma and CSF Aβ as well as between plasma Aβ and composite [ 18 F]flutemetamol SUVR in each diagnostic group and in the total sample were first evaluated with Pearson's correlation analysis. When significant correlations were found the associations between plasma Aβ , CSF Aβ and composite [ 18 F]flutemetamol SUVR were further investigated using reduced major axis regression (RMA). The 95% confidence intervals (CI) for the slope estimates were calculated using bootstraping. RMA and all subsequent statistical analysis were adjusted for age and gender. For comparisons of plasma Aβ levels between the diagnostic groups, we used univariate general linear models. The effects of APOE genotype, CBM, hypertension, ischemic heart disease and anti-hypertensive/cardio-protective medications on plasma Aβ levels were assessed with univariate general linear models additionally adjusting for diagnosis (with controls, SCD, MCI and AD as diagnostic categories). To test associations between plasma Aβ and cognitive function (cognitive measures of global function (MMSE) or delayed memory recall (ADAS-cog item 3)) and WML we performed linear regression analysis also adjusting for diagnosis. We categorized the study participants into groups with normal and pathological CSF signature using the CSF Aβ 42/Aβ 40 ratio cutoff ≤ 0.1 20,23 . ROC curves were used to determine how well plasma Aβ could distinguish individuals with a normal versus pathological CSF signature. Similar analysis was conducted for amyloid PET status using the SUVR cutoff > 1.42 20 . When comparing all markers between the diagnostic groups the Bonferroni correction was used to adjust for multiple comparisons.

Results
Demographic and clinical data for the study participants are shown in Table 1.  Table 2). Correlations within individual diagnostic groups are given in Table 2 and shown in Supplementary Fig. S2. The levels for Aβ 42 and Aβ 42/Aβ 40 in CSF and plasma correlated significantly in the control, SCD and MCI groups. In AD patients, there were significant correlations between plasma and CSF levels but only for Aβ 42 and Aβ 40.   Table 3). The correlations with plasma Aβ 42 and Aβ 40 were significant in the MCI group while the Aβ 42/Aβ 40 ratio correlated with [ 18 F]flutemetamol SUVR in the SCD group (Table 3).

Plasma biomarkers, CSF biomarkers and [
In both plasma and CSF, there were strong correlations between Aβ 42 and Aβ 40 levels (all r ≥ 0.511, p < 0.001 in the total samples and individual diagnostic groups; Supplementary Table S1).
The results were similar when the associations between plasma Aβ biomarkers, CSF Aβ biomarkers and [ 18 F] flutemetamol SUVR were examined using RMA adjusting for age and gender (95% CIs not containing 0; Tables 1  and 2).
Plasma and CSF Aβ levels and diagnostic groups. We next compared the levels of plasma Aβ between the diagnostic groups. Plasma Aβ 42 was reduced in AD compared with control, SCD and MCI groups (all p < 0.0001; Fig. 2A; Table 1). However, there were no differences in Aβ 42 levels between SCD or MCI patients and controls (Table 1). Plasma levels of Aβ 40 were decreased in the AD group compared with controls (p < 0.001), SCD (p < 0.0001) and MCI (p < 0.0001) ( Fig. 2B; Table 1). The Aβ 42/Aβ 40 ratio was lower in the MCI group than in control subjects (p = 0.002) and in AD patients compared with controls (p < 0.0001), SCD (p < 0.0001) and MCI (p = 0.003) ( Fig. 2C; Table 1). The levels of Aβ 42 and Aβ 40 in CSF were found to be in agreement with existing data 6 ( Fig. 2D-F; Table 1).
We also compared diagnostic subgroups with pathological CSF signature (control-P, SCD-P, MCI-P, AD-P with CSF Aβ 42/Aβ 40 ratio ≤ 0.1) with control subjects showing normal CSF status (control-N with CSF Aβ 42/Aβ 40 ratio > 0.1). Plasma Aβ 42 levels were slightly, but significantly, reduced in control-P (p < 0.001), SCD-P (p < 0.001), MCI-P (p < 0.0001) groups compared with the control-N group (Fig. 3A and Table 4). In the AD-P dementia group, levels were more clearly decreased compared to control-N subject (p < 0.001), and levels were also significantly lower in this group compared to control-P (p < 0.0001), SCD-P (p < 0.0001), MCI-P (p < 0.0001) (Fig. 3A). Plasma Aβ 40 was decreased in the AD-P group compared with the control-N (p < 0.0001), control-P (p < 0.001), SCD-P (p < 0.001) and MCI-P (p < 0.0001) groups ( Fig. 3B and Table 4). The plasma Aβ 42/Aβ 40 ratio was reduced in all the diagnostic groups with pathological CSF compared to control individuals with normal CSF (control-P, p < 0.0001; SCD-P, p < 0.001; MCI-P, p < 0.0001; AD-P, p < 0.0001) ( Fig. 3C and Table 4). Notably, although the differences in Aβ 42 and the Aβ 42/Aβ 40 ratio between the diagnostic subgroups in plasma were similar to those observed in CSF, they were more pronounced for CSF than for plasma ( Fig. 3 and Table 4).  Table 2. Associations between plasma and CSF Aβ biomarkers. AD, Alzheimer's disease; CSF, cerebrospinal fluid; MCI, mild cognitive impairment; SCD, subjective cognitive decline. Plasma and CSF Aβ were measured using Simoa and Euroimmun immunoassays, respectively. Associations between plasma and CSF Aβ were first evaluated with Pearson's correlation analysis and if significant correlations were found further investigated using RMA adjusting for age and gender. Data are presented as r, p from Pearson's correlation analysis and slope estimates (95% CI) from RMA; significant results are shown in bold. Plasma Aβ and APOE4. Given that individuals with one or two APOE ε4 alleles have a several fold higher risk for AD and that CSF levels of Aβ 42 are affected by APOE genotype [29][30][31] we assessed the effects of APOE4 on plasma levels of the Aβ isoforms.

All cases
In the total sample, APOE ε4 carriers showed lower levels of Aβ 42 (p < 0.001), Aβ 40 (p = 0.009) and lower Aβ 42/Aβ 40 ratio (p = 0.015) in plasma than non-carriers. When analyzed within individual diagnostic groups, plasma levels of Aβ 42 were decreased in APOE ε4 carriers in controls (p < 0.001) and SCD (p < 0.001), but not in MCI and AD dementia patients. There were no differences in Aβ 40 and the Aβ 42/Aβ 40 ratio between APOE ε4 carriers and non-carriers in any of the groups. Plasma Aβ and cognitive function. We did not find any significant associations between cognitive meas-  (Fig. 4J-L). Notably, we did not observe any changes in CSF Aβ 42 or Aβ 40 in relation to cardiovascular factors (Supplementary Fig. S3).
There were no associations between plasma Aβ and smoking or hyperlipedmia.

Discussion
In this study, we report that plasma Aβ levels correlate with CSF levels and with Aβ plaque burden in the brain assessed using amyloid PET imaging. We show that plasma Aβ 42 and Aβ 40 are reduced in AD patients, especially during dementia stages, compared with cognitively healthy control individuals. We also demonstrate increased plasma levels of Aβ were associated with WML, CMB, hypertension, diabetes and ischemic heart disease.  Numerous previous studies investigated plasma levels of Aβ 42 and Aβ 40 in patients with prodromal AD and AD dementia using conventional ELISA 2,9,12,[32][33][34] . However, because of the inconsistency of the available data, it has been difficult to draw definite conclusions with respect to changes in plasma Aβ concentration in AD. Heterogeneity of sample population, small sample size, confounding factors (in particular age) and insufficient analytical sensitivity for the ELISA methods are all thought to contribute to low reproducibility of the reported results 9 . Here we employed an ultrasensitive digital ELISA to measure plasma Aβ 42 and Aβ 40. Compared to conventional analog immunoassays, the digital Simoa platform offers improved sensitivity and lower imprecision , ischemic heart disease (n = 73 and n = 637, respectively), diabetes (n = 69 and n = 641, respectively) or anti-hypertensive/cardio-protective medications (n = 325 and n = 385, respectively). Data are presented as mean ± 95% confidence interval (CI); p values are from univariate general linear models controlling for age, gender and diagnosis; ** p < 0.05; ** p < 0.01; *** p < 0.001. IHD, ischemic heart disease.
for the detection of blood proteins 16 . Using the Simoa platform and large cohorts of well-characterized patients and cognitively healthy controls, we found that plasma levels of Aβ 42 and Aβ 40 are decreased in AD whereas the plasma Aβ 42/Aβ 40 ratio is decreased in MCI and even more in AD. Although most of the previous investigations showed no differences in plasma Aβ 42 between AD patients and cognitively healthy controls 2,12,35 our data are in agreement with some studies demonstrating low plasma Aβ 42 and/or low plasma Aβ 42/Aβ 40 ratio in AD 15,36 . Further, our results of decreased concentration of plasma Aβ 42 in APOE ε4 carriers compared with non-carriers are also consistent with earlier reports 12,37 .
We found reduced levels of Aβ 42 and the Aβ 42/Aβ 40 ratio in plasma of patients with preclinical and prodromal AD (e.g. cognitively healthy individuals, SCD and MCI patients with pathological CSF signature). However, the differences were small in comparison with a marked decline in CSF levels of Aβ 42 (Fig. 3) that is observed decades before the onset of clinical symptoms 4,5 . These results indicate that AD pathology can be identified in CSF years before overt changes in peripheral blood. Other studies demonstrated that individuals with low plasma Aβ 42/Aβ 40 ratio (but not plasma Aβ 42) at baseline have a somewhat greater risk of future AD and that decrease in plasma Aβ 42 levels over time is linked to cognitive decline 10,13 . Thus, the low plasma levels of Aβ 42 in AD patients could be due to the slow decline along the disease course.
CSF Aβ 42 measurements and amyloid PET imaging are increasingly integrated in the clinical work-up of AD as biomarkers of amyloid pathology. However, development of less expensive and less invasive blood biomarkers that could predict CSF Aβ 42 and/or amyloid PET status will greatly facilitate widespread implementation of amyloid biomarkers in routine clinical practice. In the present study, we observed significant positive correlations between plasma and CSF concentrations for Aβ 42 and the Aβ 42/Aβ 40 ratio. Moreover, low plasma Aβ 42 and Aβ 42/Aβ 40 ratio were significantly associated with high total brain binding of [ 18 F]flutemetamol. Yet similar to previous reports, we found that both associations were relatively weak 12 . Furthermore, neither plasma Aβ 42 nor the plasma Aβ 42/Aβ 40 ratio showed sufficient accuracy to identify individuals with pathologic CSF signature or abnormal amyloid PET. Collectively, our findings suggest that blood Aβ levels reflect only to some extent the dysregulated Aβ metabolism and aggregation in the brain. Other factors that are unrelated to brain amyloid pathology might be modulating the peripheral levels of these peptides. First, Aβ entering peripheral blood may be degraded by circulating enzymes or metabolized in the liver, which would reduce the potential to monitor brain Aβ metabolism. Second, production outside the central nervous system by platelets, skeletal muscle cells and other cells types 38 probably contributes to the circulating pool of Aβ . Consequently, while cerebral amyloid deposition is accompanied by a considerable decline in CSF Aβ 42 levels, the peripheral effects of plaque accumulation in the brain might be more diluted. Notably, we found decreased plasma levels of Aβ 40 in AD patients compared with cognitively healthy elderly, whereas in line with available evidence, no change was observed in CSF samples 2 . Altered levels of Aβ have been shown in the skeletal muscle and liver of AD patients 38 indicating that AD-related changes in the periphery might affect plasma levels of Aβ 40 while the CSF levels remain unaltered.
In our study, elevated Aβ 42 and Aβ 40 in plasma, but not CSF, were associated with WMLs and the plasma, but not CSF, Aβ 42/Aβ 40 ratio was increased in individuals with CBM. Plasma Aβ 42 and Aβ 40 have been previously linked to WMLs in non-demented elderly as well as in AD and MCI patients 37,39 . Increased plasma levels of Aβ 40 have also been described in individuals with infarctions in the ADNI study 12 . There are several potential mechanisms that could explain the association between plasma Aβ and cerebrovascular pathology. Plasma Aβ may affect endothelial cell function and vascular tone thereby leading to cerebral hypoperfusion that eventually results in WMLs 40 . Alternatively, reduced cerebral blood flow, which is an early clinical feature of AD 41 could promote overproduction of Aβ in endothelial cells and its secretion into the circulation 42 . Increased production of Aβ with marked increases in plasma levels are in fact found after severe ischemia due to cardiac arrest in patients that are resuscitated 43 . In this context it would be of interest to establish if vascular amyloid deposition in cerebral amyloid angiopathy is accompanied by altered blood levels of Aβ 42 and Aβ 40. Studies in small patient groups have so far produced inconclusive results and warranty future investigations [44][45][46] . Our finding also indicate that increased Aβ 42 and Aβ 40 in plasma, but not CSF, are associated with hypertension, diabetes and ischemic heart disease, conditions that adversely impact the function of the vascular system. This is in agreement with previous studies reporting association between plasma Aβ and hypertension 12,47 . Further, high plasma Aβ 40 has been recently linked to increased cardiovascular mortality in patients with coronary heart disease 48 .
In conclusion, we demonstrate that elevated plasma Aβ is associated with vascular disease both in the brain and in the periphery. In AD, plasma Aβ 42 and Aβ 40 are markedly reduced during the dementia stages, which is in contrast to the CSF where there is a clear drop in Aβ 42, but not Aβ 40, already during preclinical stages. However, although low plasma Aβ 42 and Aβ 42/Aβ 40 ratio were associated with amyloid deposition in the brain, these markers did not show diagnostic value in AD. Several panels of plasma AD biomarkers have been recently reported 49,50 . Future studies need to determine whether inclusion of plasma Aβ measures might potentially improve the diagnostic performance of the plasma biomarker panels, especially during the dementia stage where we show a clear decrease in plasma Aβ levels.