The search strategy was updated according to best practice recommendations (5,6). Literature searches were performed using MEDLINE (between 1996 and April 2015; https://www.nlm.nih.gov/bsd/pmresources.html) and PubMed (between 1990 and November 2015; www.ncbi.nlm.nih.gov/pubmed) databases. The search queries included: i) Blood, ii) cerebrospinal fluid, iii) urine, iv) Alzheimer's disease and v) mild cognitive impairment.

A reviewer scrutinized abstracts found by electronic search in order to identify articles meriting a full review. Entire articles were reviewed before data were extracted from pertinent papers.

The inclusion criteria used for the review protocol were as follows: i) Age ≥60 years, ii) diagnosis of AD according to the criteria of the National Institute on Aging-Alzheimer's Association (NIAAA) (7) or diagnosis of MCI according to NIAAA criteria (8) with a confirmed progression to AD assessed by clinical follow-up, and iii) suitable clinical assessments of cognitive and functional impairment, quality of life and clinical evaluations. The exclusion criteria were: i) No English editing (as we lacked resources for translation), ii) diagnosis of non-AD dementia, and iii) MCI that did not progress to AD. No limits were defined on the grounds of disease duration or drug treatment.

A total of 2,243 papers, and 92 additional articles from recent reviews were found. After reviewing the abstracts, 648 papers were obtained on the basis of the aforementioned inclusion/exclusion criteria and duplicates. After screening, 259 papers were reserved, of which 36 reviews and 158 full-text articles were excluded after more in-depth examination, on the basis of the same inclusion/exclusion criteria. Therefore, 65 published studies were considered eligible for the current review (Fig. 1). Data extraction was divided into four categories among the authors by expertise: Potential blood-derived markers of AD and MCI, potential CSF-derived markers of AD and MCI, potential urinary-derived markers of AD and MCI, and cell-based techniques for analyzing tau pathogenicity. Co-authors provided a detailed abstract of all studies describing their strengths and weaknesses, as well as a general evaluation of the category (9,10). The quality of studies was assessed using the Standards for the Reporting of Diagnostic Accuracy Studies in dementia (11).

The potential blood-derived biomarkers for AD and MCI are presented in Table I. The ratio of amyloid precursor protein (APP) forms contained in the platelets was significantly decreased in individuals with mild AD, very mild AD and MCI (12), individuals with MCI that progressed to AD dementia (13), and patients with AD that exhibited cognitive decline (14). Compared with patients with non-amnestic MCI, patients with amnestic MCI exhibited increased levels of coated-platelets, which are produced by collagen and thrombin activation. In addition, the increases in coated-platelet levels were associated with an increased risk for the progression to AD (15). There were significantly lower levels of Aβ1-42 and Aβ1-42/1-40 in the plasma of patients with AD compared with patients with non-AD dementia, or in patients with AD-type MCI compared with patients with other types of MCI (16). A cross-sectional study demonstrated that plasma Aβ1-40 levels were independently associated with microvascular brain injury (17). However, plasma Aβ peptide levels only weakly predicted the conversion to AD in normal individuals (18). In addition, plasma Aβ monomer biomarkers were not sensitive enough to discriminate patients with AD-type MCI from patients with early AD, or to distinguish patients with other types of MCI from patients with other types of early dementia.

Tau proteins in the serum are potential biomarkers for neuronal degeneration or damage. A novel ultrasensitive digital immunoassay technology, which is referred to as a single-molecule ELISA, is able to detect clinically relevant proteins in the serum at concentrations <10⁻¹⁵ M (19). Compared with the classic ELISA, this ultrasensitive technique is 1,000-fold more sensitive. The lower limit of quantification is 0.02 pg/ml, which is 100 times higher than the plasma levels of tau proteins (~5 pg/ml). This ultrasensitive technique has detected dynamic changes in tau protein in the serum after hypoxic brain injury, and a mean 2-fold increase in AD individuals, however with some overlaps with the levels in patients with MCI and the controls (20). However, rare protein isoforms that result due to cleavage of the tau protein and disease-specific tau phosphorylation are not recognized by currently available assays (21).

Aberrant lipid metabolism is associated with the pathophysiological processes in AD. Some plasma phospholipids may therefore be considered biomarkers for AD. Reductions in the levels of these phospholipids may accurately predict the progression of an individual from normal cognition to MCI or AD within 2 years (22,23). A biomarker panel that detected decreases in the levels of 10 phospholipids in blood plasma that reflect cell membrane integrity may be able to predict which cognitively normal elderly adults will develop MCI and AD with >90% accuracy (22). Early studies have reported that phosphatidylcholine molecules in AD brains are decreased and that their main metabolite, glycerophosphocholine, is increased (23,24). However, some phosphatidylcholines produced by the sequential methylation of another phosphatide, such as phosphatidylethanolamine, were increased in AD brains. It has previously been demonstrated that three phosphatidylcholine molecules were significantly diminished in patients with AD compared with patients with MCI or age-matched individuals (23).

A previous study measured the plasma levels of 27 vascular-related proteins using multiplex assays or ELISA in patients with MCI or AD, and in healthy controls. The results indicated that N-terminal pro-brain natriuretic peptide (NT-proBNT) was significantly increased in patients with MCI or AD, and could be considered a potential biomarker for AD diagnosis and prognosis (25). Multiplex assays were used to measure 26 previously discovered AD-associated plasma biomarkers in plasma samples from patients with AD and MCI, and elderly non-demented individuals from three multicenter cohorts. These assays demonstrated that 10 proteins, namely transthyretin (TTR), clusterin, cystatin C, alpha-1-acid glycoprotein, intercellular adhesion molecule 1, complement C4, pigment epithelium-derived factor (PEDF), alpha-1 antitrypsin, normal T cells, and expressed and secreted Apolipoprotein C3, were strongly associated with the severity and progression of AD. The 10 proteins, plus the Apolipoproteina E (ApoE) genotype, had the greatest predictive power for identifying dementia from prodromal disease (26). In addition, butyrylcholinesterase K-variant alleles had synergistic effects with the APOE ε4 genotype on the conversion of MCI to AD (27). Another study measured the levels of 146 plasma proteins at baseline and after 1 year using a multiplex immunoassay panel; the results verified that the levels of eotaxin 3, pancreatic polypeptide and NT-proBNT were increased in subjects with AD and MCI, which was similar to changes detected in the CSF, and also demonstrated that the Apo ε3/ε4 or ε4/ε4 genotype depended on the biochemical profile (28). Another study quantified 1,001 proteins by SOMAscan, and demonstrated that protein expression levels of prostate-specific antigen complexed to α1-antichymotrypsin, pancreatic prohormone, clusterin and fetuin B were the most strongly associated with AD. Pancreatic prohormone was significantly associated with left entorhinal cortex and hippocampus atrophy, whereas fetuin B was only related to left entorhinal atrophy. Clusterin was significantly related to the rate of cognitive decline (29).

Other potential blood-derived markers include mitochondrial function indicator N-acetylaspartate (NAA), neuroinflammatory indicator chitinase-3-like 1 (YKL-40), stress protein heme oxygenase-1 (HO-1), nutritional biomarkers, histone-binding protein RbAp48, plasma ketone body (KB), and a cumulative risk index based on several pathways associated with cognitive decline for blood-derived markers. NAA levels are coupled to neuronal mitochondrial function and are correlated with Aβ42 in patients with AD (30). The mean plasma and CSF levels of YKL-40 were enhanced in individuals with a clinical dementia rating between 0.5 and 1, and YKL-40 is considered a potential fluid biomarker for preclinical AD (31). HO-1 protein levels in plasma and CSF, and mRNA levels of lymphocyte HO-1 in individuals with sporadic AD are decreased compared with in normal controls, and other subjects with chronic neurological and medical disorders (32). A previous study suggested that plasma levels of α-/γ-tocopherols, or plasma levels of γ-tocotrienols in combination with structural magnetic resonance imaging (sMRI) measures, may be used to differentiate patients with AD or MCI from cognitively healthy individuals, and predict the conversion of MCI to AD after 1 year of follow-up (33). High levels of homocysteine, a risk factor for microvascular impairment, have been associated with alterations in electroencephalographic rhythms in mild AD, but not in MCI subjects, including unselective increases in cortical delta, theta and alpha rhythms (34). Compared with control individuals, plasma levels of non-enzymatic and enzymatic antioxidants were similarly decreased in patients with MCI and AD (35). Baseline higher levels of free copper may be used to predict severe cognitive decline in AD, and a faster and more obvious progression of disability at 1 year (36). Patients with higher levels of free copper combined with hyperlipidemia were also prone to severe cognitive impairment. RbAp48 modifies histone acylation and is associated with age-related memory impairment. Post-mortem measures demonstrated that levels of RbAp48, which are associated with hippocampus-dependent memory deficits in elderly individuals, were decreased in the dentate gyrus. Similar memory deficits were also observed in RbAp48-deficient transgenic mice (37). Unlike the amyloid cascade hypothesis, which suggests that amyloid pathways are the main therapeutic target, the interventions based on the mitochondrial cascade hypothesis, which suggests that mitochondrial function decline triggers AD pathophysiological cascade, are associated with the restoration and maintenance of mitochondrial function. The increase in plasma KB levels induced by ketone ester in several animal models of AD may improve mitochondrial metabolism and postpone the appearance of AD-like pathological alterations (38). Therefore, low plasma KB levels may be a potential biomarker for AD pathology in individuals with preclinical sporadic AD or reversible CF. A cummulative risk index, including blood-derived markers ApoE ε4 and Aβ42/40, telomere length, blood glucose, cystatin, C-reactive protein, interleukin (IL)-6 and albumin (39), may provide valuable predictive information regarding future cognitive trajectories independent of age and baseline cognitive status. It is a question worthy of further research whether this same risk index could also predict CF. In addition, the ApoE genotype, IL-6 receptor and clusterin plasma levels, together with Auditory-Verbal Learning Test and Trails B have been reported to be useful for predicting brain amyloidosis and MCI progression to AD with modest accuracy (40). In genetic analyses of non-familial AD, the ApoE ε4 genotype contributes to the heterogeneity of disease processes associated with sporadic AD (41–43). In addition, alongside the ApoE receptor, the low density lipoprotein receptor 5 repeated allele has been reported to be associated with the risk of dementia, and the correlation was more evident in individuals with mixed or vascular dementia (VaD) compared with AD (44).

The CSF-derived biomarkers for AD and MCI are presented in Table II. The monomer form Aβ42, tau and phosphorylated (p)-tau are widely used to screen AD and MCI caused by AD, and to predict the conversion of preclinical AD and MCI to AD (45,46), independent of the ApoE genotype (43). Among the CSF biomarkers, CSF Aβ42 has been reported to have the best diagnostic accuracy for the discrimination of AD from frontotemporal dementia (47). CSF Aβ42 and florbetapir were inversely correlated across cognitively normal, MCI and AD groups with 86% consistency (48). Abnormal CSF Aβ levels together with abnormal imaging and serum markers exhibited a better predictive value for the earliest stage of AD (49). Individuals with abnormal CSF proteins, particularly abnormalities of both tau and Aβ42 appeared to be simultaneously at greatest risk of cognitive decline (50) and progression from MCI into incipient AD (51). In addition, the p-tau/Aβ42 ratio in CSF could longitudinally predict cognitive performance loss in subjects with MCI (52). Injury markers in CSF, including total (t)-tau and p-tau, could also predict conversion time, from MCI with amyloid pathology, to dementia (53). However, MRI was slightly better at predicting future clinically defined disease stages and functional decline compared with Aβ42, tau or p-tau in the CSF (54). A combination of MRI or fluorodeoxyglucose-positron emission tomography and CSF promote the predictive value of the conversion of MCI to AD (55,56). However, a combination of CSF t-tau/Aβ42 together with sMRI or neuropsychological tests was not superior, as compared with single trail making test-B or sMRI measures of the right entorhinal cortex (57). Unlike the evaluated CSF values of t-tau, p-tau and Aβ42, the order of the annual increase in ventricular volume detected by serial sMRI was AD group>amnestic MCI group>no AD group. Furthermore, the annual increase in ventricular volume was consistent with the annual decrease in general cognitive and functional indices. The alterations in structure and cognitive function were influenced by the ApoE ε4 genotype (58).

The potentially associated CSF-derived biomarkers for AD and MCI are presented in Table III. A previous study reported that the albumin ratio between CSF and serum is normal in patients with AD independent of age, and is increased in patients with cerebrovascular diseases (59). However, an elevation in the ratio was associated with AD and VaD in an 85-year-old population, and was associated with the conversion from non-dementia to dementia across 3 years of follow-up (60). A previous study indicated that age-dependent elevation in the ratio could be observed in ApoE ε4 allele carriers without cognitive impairment (61). An elevation in the albumin ratio was also closely associated with severity in patients with AD with medial temporal atrophy (62). Another study demonstrated that an age-dependent elevation in the ratio could only be detected in the hippocampal CA1 and dentate gyrus regions of cognitively normal subjects (63). Compared with cognitively normal subjects, patients with MCI exhibited an increase in the ratio and the levels of soluble platelet-derived growth factor receptor-β. These biomarkers of neurovascular unit damage appeared earlier than other biomarkers, including CSF Aβ42 and tau. Increased CSF β-secretase (BACE 1) activity in individuals with sporadic AD may be involved in the amyloidogenic process and axonal degeneration (64). Elevated BACE 1 activity in subjects with AD and MCI has been shown to be associated with the ApoE ε4 genotype, and decreased levels of Aβ42 were only observed in ApoE ε4 carriers with MCI (65). CSF levels of the soluble amyloid precursor proteins (sAPP) α and β were similar between AD and healthy controls; however, sAPP β levels were significantly higher in patients with MCI compared with healthy controls (66). Significantly higher levels of sAPP in the CSF were superior to significantly lower levels of CSF Aβ42 with regards to early prediction of the progression from MCI to AD, and the differential diagnosis of AD from MCI and frontotemporal dementia (67). Compared with the stable MCI group and the control group, the levels of sAPP α and β were significantly higher in the AD group and the MCI progression to AD group; the ApoE ε4 allele had no effects on the levels of sAPP α and β (68). CSF Aβ42/Aβ40 values were more sensitive than CSF levels of Aβ42 for the identification of incipient AD in MCI individuals (69). Soluble Aβ oligomers are potential biomarkers for AD; however, compared with the changes in CSF Aβ42 or tau, changes in the levels of soluble oligomers were not superior for the discrimination of AD from controls (70,71). In a comparative study of brain expression of soluble Aβ oligomers, Aβx56, and Aβ trimers and dimers, it was demonstrated that levels of Aβ dimers were highest in subjects with probable AD, and the levels of Aβx56 and Aβ trimers were lowest when compared with age-matched unimpaired and young unimpaired subjects (72). Furthermore, only Aβx56 was correlated with pathological tau proteins and postsynaptic proteins. These results suggested that Aβx56 may contribute to the very early stages of AD pathogenesis. Savage et al reported that higher Aβ oligomer levels predicted more severe dementia (73). However, analogous to the detection of various monomer forms, including Aβ40 and Aβ42, developing even more specific assays to measure the precise nature of oligomers via more sensitive amplification platforms will improve the early diagnostic potential of these biomarkers (74). The amide I band reflects the structural destruction of all types of Aβ peptides. Its downshift in CSF and blood has been detected in patients with MCI that progressed to AD, and its downshift frequency was superior to a single Aβ misfold or the level of specific oligomers (75).

Synaptic biomarkers in the CSF are potential early biomarkers for AD. Neurogranin, which is mainly located in dendritic spines, is associated with long-term potentiation and memory consolidation. CSF neurogranin is composed of a series of C-terminal peptides. All neurogranin peptides detected by the ELISA method suggest that CSF neurogranin may be used to monitor synaptic degeneration and reflect the rate of cognitive decline in individuals with prodromal AD (76). Even if it is very rare in CSF, the levels of CSF neurogranin peptide 48–76 exhibit the most obvious increase in patients with AD. A progressive decline in CSF levels of neurogranin could be observed from individuals with AD, MCI that progressed to AD, stable MCI to controls (77). Patients with AD or MCI that progressed to AD demonstrated significantly higher baseline CSF levels of neurogranin compared with stable MCI or controls (77,78). CSF levels of the cytoskeleton light neurofilament protein were significantly higher in patients with late-onset AD or frontotemporal dementia compared with controls (79). In addition, alterations in the CSF levels of axonal growth-associated protein-43 were associated with changes in CSF tau and sAPP levels, and patients with AD exhibited a significant increase compared with controls (80). However, the detection of specific peptides and other biomarkers for synaptic dysfunction will benefit from the development of novel ultrasensitive assays.

The onset of cognitive decline is estimated to lag AD pathology by 10–15 years. By the time of the appearance of clinically detectable cognitive impairment, substantial neuronal loss has occurred. As a marker of neuronal damage, the AD-associated neuronal thread protein AD7c-NTP is a ~41 kD membrane-spanning phosphoprotein. The increase in immunoreactive AD7c-NTP in the brain is associated with the increase in p-tau-immunoreactive cytoskeletal lesions, but not the increase in Aβ accumulation. The protein may be involved in neuronal apoptosis and neurite sprouting, and in the pathological alterations of AD. Dying cells in the brain are able to secrete or release the protein into the CSF. Post-mortem CSF levels of AD7c-NTP in AD were significantly higher compared with in age-matched controls. The CSF levels of patients with probable AD were also significantly higher compared with controls or other neurological disease controls, and were associated with the severity of cognitive impairment (81). The CSF levels of visinin-like protein-1 (VILIP-1) and the ratios of VILIP-1/Aβ42, similar to tau and tau/Aβ, may predict future cognitive decline in cognitively normal subjects, differentiate patients with AD from non-AD dementia and healthy controls (82), and predict rates of global cognitive decline in individuals with early AD (83). CSF VILIP-1 levels may also predict rates of whole-brain and regional atrophy with a similar power to CSF levels of tau and p-tau181 in individuals with very mild AD or preclinical AD (84). Longitudinal CSF biomarker patterns, including low Aβ42 in early middle age, markedly increased t-tau, p-tau and VILIP-1 in mid- and late middle age, and increased levels of the neuroinflammatory marker YKL-40 throughout middle age, were useful for screening middle-aged, asymptomatic individuals with AD (85).

Systematic analysis of metabolite profiling of the CSF by magnetic resonance spectroscopy demonstrated that increased cortisol and cysteine levels, and decreased uridine levels, may be involved in the progression of AD. Individuals with severe AD exhibited increased cortisol levels, and individuals with mild AD [Mini Mental State Examination (MMSE)>22] exhibited increased cysteine and decreased uridine levels. Specificity and sensitivity >75% could be obtained for the paired combination of cysteine and uridine to identify mild AD (86). Lactate is a product of glycolytic metabolism; CSF levels of lactate in patients with AD were significantly higher compared with in controls and VaD controls (87). Patients with mild AD exhibited higher lactate levels compared with in moderate and severe AD, and higher lactate levels corresponded with lower levels of t-tau and p-tau. A biomarker panel that analyzed the levels of seven proteins in the CSF by proteomic analysis and mass spectroscopy was able to classify AD individuals from controls with an accuracy of 84.5% (sensitivity 93.3%, specificity 75.7%) (88). Target proteomic analysis detected another four proteins that could reflect obvious longitudinally dynamic alterations during AD progression (89). Other potential CSF-derived markers include α-synuclein, heart fatty acid binding protein (HFABP), PEDF, complexed prostaglandin-d-synthase (PDS) and TTR, isoprostane and ferritin. CSF α-synuclein levels were not only significantly increased in patients with MCI and AD compared with controls, the increase in CSF α-synuclein levels was also significantly associated with the decrease in MMSE scores (90). CSF α-synuclein levels only offered modest sensitivity and specificity as a diagnostic marker of AD. Significantly increased CSF levels of HFABP were detected in patients with AD compared with controls, and in patients with MCI that progressed to AD or VaD when compared with individuals with stable MCI (91). However, HFABP had a lower predictive value than Aβ42, t-tau and p-tau in identifying the progression of MCI to AD and VaD. Cortical neurons and astrocytes in AD brains demonstrated strong immunostaining of PEDF, and CSF PEDF as a biomarker may improve the diagnosis of AD (92). Complexed PDS/TTR exhibited a significant increase in post-mortem ventricular CSF in MCI and late-stage AD compared with diseased control subjects, and lumbar CSF levels of the complex showed a six-fold increase in living subjects with probable AD compared with normal control subjects (93). CSF levels of isoprostane were shown to possess robust longitudinal effects on the progression of MCI to AD. The annual rate of isoprostane was significantly different in the following order: MCI that progressed to AD group>stable MCI group>controls (94). The elevated CSF levels of ferritin were not only associated with cognitive performance but could also predict AD progression (95).

The potential urinary-derived biomarkers for AD and MCI are presented in Table IV. Patients with AD demonstrated elevated levels of AD7c-NTP in CSF and urine, and the elevated levels of AD7c-NTP were associated with the severity of dementia, i.e., very mild, mild or moderately severe AD. Several studies have suggested that AD7c-NTP is a useful biomarker, and may be widely used to screen the risk of elderly individuals with MCI and AD (96–99).

Oxidative stress and oxidative DNA damage have important roles in the process of AD. A major product of oxidative DNA damage is 8-hydroxy-2′-deoxyguanosine (8-OHdG). The antioxidant enzyme paraoxonase 1 (PON1) is able to prevent the oxidation of low-density lipoproteins. Patients with AD exhibited significantly elevated levels of 8-OHdG in urine and significantly decreased PON1 activity in serum, as compared with in healthy elderly volunteers (100). Isoprostanes, which are products of arachidonic acid peroxidation by free radicals, are also biomarkers for oxidative injury. Urinary levels of F2-isoprostanes are significantly elevated in AD subjects (101). In addition, 3-hydroxypropyl mercapturic acid/creatinine reduction in urine is not only associated with stroke but is also an ideal biomarker for differentiating patients with AD from patients with MCI (102). Furthermore, high levels of urinary polyphenols may exert protective effects of polyphenol intake against cognitive impairment, and are associated with a lower risk of obvious cognitive decline in elderly individuals over a 3-year follow-up period (103).

Recently, two novel methods to detect tau pathogenicity have been developed. One method detects tau seeding activity in various AD brain lysates using a cell-based biosensor assay. In vitro studies have reported that pathogenic tau oligomers can move between cells similar to mechanisms of prion pathogenesis, this is referred to as transcellular spread (104). Briefly, biosensor cells are generated to express human tau proteins fused with cyan fluorescent protein (CFP) or yellow fluorescent protein (YFP); when pathogenic tau from AD samples is added to a culture of biosensor cells, which contain tau proteins fused with CFP or YFP, pathogenic tau promotes the aggregation of two fused proteins and results in a positive fluorescence resonance energy transfer signal (105). The signal intensity reflects the pathogenic tau seeding activity of the AD sample. Detection of pathogenic tau seeding activity from human brain samples could reliably differentiate AD from Huntington's disease and aged controls (106). In addition, P301S tau transgenic mice demonstrated that tau seeding activity is an ideal biomarker of tauopathy. Increases in tau seeding activity occurred 1.5 months earlier than tau deposition. Future verification of pathogenic tau seeding activity in human CSF or blood will improve the diagnostic accuracy of tauopathies in a clinical setting, and provide a noninvasive biomarker to evaluate the therapeutic efficacy of tau-modifying agents in future clinical trials.

The second method quantitatively detects tau pathogenicity in biological samples using monoclonal antibodies against seeding tau. The antibodies can also be used as an immunotherapy that traps pathogenic tau and prevents transcellular spread. Anti-pathogenic tau antibodies that inhibit the ability for tau pathogenicity in vitro could markedly reduce the levels of p-tau, aggregated and insoluble tau, and inhibit microglial activation and improve cognition in P301S tau transgenic mice (107). If these antibodies can be used to quantify tau pathogenicity in CSF and blood from different individuals, they may be considered potentially noninvasive biomarkers for detecting tau pathology.

Aging-related major neurocognitive disorders, particularly AD-induced neurodegeneration, can progress over decades before clinical symptoms become apparent. Biomarkers for brain pathological processes allow for their early diagnosis in preclinical stages and for the development of objective prognostic assessments in clinical intervention trials. Some fluid biomarkers, including Aβ42, t-tau and p-tau, have been widely used in clinical practice and clinical trials. The development of novel measurement techniques greatly promotes the production of novel biomarkers and improves the accuracy of old biomarkers (19,20,73,74). Several novel biomarkers associated with different aspects of AD neuropathology are being developed, including those related to synaptic dysfunction, neuronal damage and apoptosis, neuronal activity alteration, neuroinflammation, oxidative stress, metabolites, mitochondrial function and aberrant lipid metabolism. A bioinformatics approach for identifying specific single nucleotide polymorphisms (108–110) and epigenetic markers, including microRNAs (111) in the CSF or blood of patients with MCI, AD or non-AD dementia may also be useful for AD diagnosis and differentiation.

Based on the ordering of AD biomarkers, particularly CSF biomarkers (Aβ42, t-tau and p-tau), and AD-associated biomarkers, such as neural injury and neuroinflammation biomarkers, we classified which one was the earliest event resulting in the heterogeneity of cognitive impairment. These biomarkers may be helpful for the diagnosis of CF and the differential diagnosis of cognitive impairment from CF, including AD or VaD. CF consists of reversible and potentially reversible cognitive impairment subtypes (112). The former is based on SCD, and the latter is comparable to MCI. The most common cause of cognitive impairment of CF is AD. The severity of SCD has been associated with biomarkers for AD in subjects with MCI, including low CSF Aβ42 and high CSF tau or p-tau levels (46). Longitudinal CSF biomarkers, such as Aβ42 reductions in early middle age; markedly increased tau, p-tau and VILIP-1 in mid- and late middle age, and increases in YKL-40 throughout middle age were associated with the severity of cognitive impairment in prelinical AD (85). The combination of fluid biomarkers and imaging biomarkers could further improve diagnostic accuracy of cognitive impairment. In addition, physical frailty and AD may share similar pathophysiological mechanisms. Certain AD-associated fluid biomarkers, such as oxidative stress and inflammatory markers, may also contribute to the screening of physical frailty in CF subjects (112,113). Furthermore, the dynamic changes of AD-specific and AD-associated fluid biomarkers may be helpful for the screening of candidate drugs that affect cognitive impairment and physical frailty.