Abstract
The number of patients with Alzheimer’s disease (AD) and non-Alzheimer’s disease (non-AD) has drastically increased over recent decades. The amyloid cascade hypothesis attributes a vital role to amyloid-β protein (Aβ) in the pathogenesis of AD. As the main pathological hallmark of AD, amyloid plaques consist of merely the 42 and 40 amino acid variants of Aβ (Aβ42 and Aβ40). The cerebrospinal fluid (CSF) biomarker Aβ42/40 has been extensively investigated and eventually integrated into important diagnostic tools to support the clinical diagnosis of AD. With the development of highly sensitive assays and technologies, blood-based Aβ42/40, which was obtained using a minimally invasive and cost-effective method, has been proven to be abnormal in synchrony with CSF biomarker values. This paper presents the recent progress of the CSF Aβ42/40 ratio and plasma Aβ42/40 for AD as well as their potential clinical application as diagnostic markers or screening tools for dementia.
INTRODUCTION
The number of patients with dementia is increasing rapidly because of societal aging. Alzheimer’s disease (AD), a severe chronic progressive neurodegenerative disease, represents 75% of all dementia cases [1] and affects approximately 50 million people worldwide [2]. It is estimated that the number of people suffering from AD in the world will climb to 152 million by 2050 [2]. AD is clinically characterized by a comprehensive neurocognitive disorder, such as progressive decline in cognitive function and basic activities of daily living and changes in personality and behavior, which eventually leads to clinical adverse events or death [3]. The disease greatly increases the burden on social psychology and public health and influenced the World Health Organization to identify key areas for the improvement of AD diagnosis. Due to the incurability of AD, early, accurate, and biomarker-based AD diagnosis is particularly important today, as disease-modifying therapies to slow the progression of disease have become available.
The definitive diagnosis of AD relies on neuropathological evidence of extracellular amyloid-β (Aβ) plaques and intracellular tau tangles [4]. Through more than a century of research, we have learned that the pathological process of AD can be recorded by means of autopsy, neuroimaging, and cerebrospinal fluid (CSF) biomarkers [5]. In 2018, the ATN diagnostic framework presented by the National Institute on Aging and Alzheimer’s Association (NIA-AA) [6] enabled accurate in vivo diagnosis of AD patients through PET examination and/or lumbar puncture. Positive amyloid-PET and abnormal levels of CSFAβ42 or Aβ42/40 can verify A+, that is, cerebral amyloidosis, and are necessary conditions to classify an individual on the AD continuum. Many studies have compared the overall availability of two CSF indicators reflecting the same pathological features. They found that the CSFAβ42/40 ratio in PET+ was 0.54–0.58 times of that in PET- cases [7–10], showed higher consistency with the PET value of amyloid than CSFAβ42 alone, with a concordance of 86–96%, sensitivity of 86–100%, and specificity of 73–94% [7, 11], and was of clinical diagnostic value [12–15]. Here, we summarized existing corollaries about these findings. 1) As the reference peptide, the Aβ40 level (the most abundant amyloid peptide in CSF) compensated for interindividual variation in the overall Aβ production and CSF circulation to “normalize” the CSF Aβ42 level. For healthy controls who were wrongly classified as positive due to low Aβ42 and AD patients wrongly classified as negative due to high Aβ42, the ratio has been shown to improve classification performance [16, 17]. 2) Non-AD lesions (e.g., subcortical injury) may lower the overall levels of all Aβ subtypes (including Aβ38, Aβ40, and Aβ42) because of the decrease in neuroid activity, whereas only the low level of CSF Aβ42 is related to AD-specific neurodegeneration (i.e., hippocampal atrophy). The Aβ42/40 ratio thereby diminishes the confounding effect caused by non-AD-specific pathological changes [13] and has more advantages in distinguishing AD from non-AD dementia. 3) Many studies have demonstrated that Aβ42/40 can mitigate the confounding effects induced by preanalysis factors, such as adsorption-derived Aβ42 concentration loss, additional freeze/thaw cycles of PP storage tubes, tube type, and CSF volume [12, 19]. In a clinically based multicenter study, Aβ42/40 was found to be the only consistent biomarker in AD and non-AD patient cohorts from three different centers, which further supported the insensitivity of the Aβ ratio to changes in preanalysis and analysis conditions [20]. 4) Both Aβ42 and Aβ42/40 changed before amyloid-PET was positive, but the longitudinal time point at which Aβ42/40 changed lagged behind that of Aβ42 [21]. Such a time difference may be one of the reasons why Aβ42/40 was more consistent with amyloid-PET than Aβ42. 5) Independent of the pathophysiological effect of the disease, the theoretical derivation at the pure probability level showed that the variance of the ratio of two random variables with a strong positive correlations (where the expected value of the denominator was at least 1) was less than the variance of the numerator. Compared with Aβ142, the improvement of the diagnostic performance of Aβ42/40 in distinguishing the control and AD groups seemed to be assigned by the basic law of probability [22].
As mentioned above, although the existing routine biomarkers of CSF significantly facilitated the diagnosis of AD, the perceived invasiveness in the sampling process of CSF inevitably limited its widespread use. In recent years, with the rapid development of targeted proteomic techniques, Aβ pathology has also been detected in plasma Aβ42/40 [10, 23–26]. According to the results of a clearly defined cohort, a low concentration of Aβ42/40 in plasma was an ideal marker of cortical amyloid deposition at all stages of the AD continuum [27, 28] to provide a new choice in clinical practice and trials. To date, there is a lack of a comprehensive review of the analytical and clinical availability of different fluid Aβ42/40 ratios in the diagnosis of AD. In this paper, we reviewed studies that used the Aβ42/40 ratio as an information tool for the whole AD spectrum. That is, 1) preclinical exploratory studies, which provided the clue of the potential biomarker of fluid Aβ42/Aβ40, 2) differential diagnostic studies, which determined the ability of fluid Aβ42/40 biomarkers to distinguish AD from cognitively unimpaired (CU) patients and, more importantly, from non-AD dementia patients, 3) prognostic studies, which assessed the ability of biomarkers to predict the progression from preclinical to dementia and determined their positive threshold. Moreover, we also took into account the technical analysis of fluid biomarkers, summed up existing biochemical detection techniques of Aβ, and stressed matters of priority for correct use of fluid Aβ42/40 biomarkers clinically (including standardizing preanalysis parameters and analytical procedures to diminish measurement variability, and evaluating the effectiveness of common covariates on discrimination ability, and minimizing interference from nonpathological factors). In summary, the purpose of this review was to evaluate Aβ42/40 biomarkers in CSF and plasma in three aspects: early detection of diseases, accurate identification of prodromal diseases, and detection of prodromal diseases.
PRECLINICAL EXPLORATORY STUDIES
CSF Aβ42/40
Aβ is a group of endogenous peptides containing 36–43 amino acids, produced by a large transmembrane protein, amyloid-β protein precursor (AβPP), which cuts sequentially at different sites via β-secretase (BACE1) and γ-secretase (a kind of transmembrane complex) under physiological conditions [29]. The deposition of Aβ plaque in the brain is an invariable feature of AD, but the delay between clinical onset and autopsy diagnosis prevents us from determining the correlation between brain pathology and clinical symptoms. Additionally, accurately diagnosing patients within their life cycle has become a challenge. In 1985, Master et al. [30] identified the complete sequence of Aβ from purified amyloid plaques and discovered that some Aβ had a truncated N-terminal and a strong tendency to form oligomers. In 1992, Seubert [31] found that Aβ can be secreted into CSF, which underlies the subsequent development of in vivo diagnostic tests of AD-CSF biomarkers. The earliest CSF researchers mainly measured the level of total Aβ peptide and, thus, were more inclined to measure Aβ40 [32] (approximately 50–70% of the total CSF Aβ content [33]). However, these studies failed to define the differences between healthy controls and AD patients. In 1993, Jarrett and colleagues [34] posited that Aβ42 alone but not Aβ40 was probably a key pathogenic protein involved in the formation of amyloid plaques. At the beginning of the 21st century, people learned about the value of low-level CSF Aβ42 in the diagnosis of AD [35]. In 2004, the first diagnostic application of the Aβ42/40 ratio in addition to Aβ142 was proposed by Lewczuk et al. [36]. Subsequently, two extra amino acids (isoleucine and alanine) at the C-terminus of Aβ42 were proven to be crucial to different aggregation mechanisms [37, 38]. Moreover, studies have also shown that only 10% of Aβ42 tends to accumulate and mediate strong neurotoxicity. It is deposited in the earliest stages of AD and is always a key component of most amyloid plaques throughout the disease [39]. Although Aβ42 and Aβ40 may have different associations with and effects on AD pathology [40, 41], interestingly, recent studies have documented that the combination of Aβ42 and Aβ40 prorata (expressed as Aβ42/40) seems to be more associated with the mechanism of AD than the absolute value of the amounts of Aβ42 and Aβ40 [42–44]. Taken together, the identification of CSF Aβ42/40 is a clue about the potential useful liquid biomarker for AD, and the prior identification of the identified clue has been fully achieved.
Plasma Aβ42/40
Based on the advantages of plasma acquisition approaches (i.e., they are safe, minimally invasive, cheap, and repeatable), the concept of dementia plasma biomarkers and relevant exploratory studies have arisen [45–48]. However, according to early studies, Aβ42/40 based on CSF has no similar ability to identify potential AD pathology in the plasma matrix. Two meta-analyses [49, 50] consistently showed substantial overlap between the AD disease group and the normal control group in biomarkers associated with the amyloid peptide in plasma, including Aβ42, Aβ40, and Aβ42/40, and there was no significant difference. This result was interpreted as follows. 1) The facts that only a small amount of cerebral protein entered the blood and the total protein concentration of the blood matrix was several orders of magnitude higher than that of CSF further resulted in dilution of target trace cerebral protein [51]. 2) Peripheral Aβ peptides from platelets and other nonbrain tissues weakened the correlation between plasma and CSF levels [52]. 3) The cerebral proteins entering the blood were degraded by protease via the metabolism of liver and kidney [53], which would undoubtedly introduce mutations unrelated to amyloid into the central nervous system. Factors related to Aβ physiology inevitably restrict the potential for the accurate determination of target proteins in plasma. However, with the development of technology, many reliable methods are now available to bring new hope for the sensitive and robust measurement of plasma Aβ. The SimoaAβ test in the first item lowered the matrix effect by diluting the samples highly linearly in advance and achieved a highly sensitive and accurate quantification of plasma Aβ, with a recovery rate of nearly 100% (the limit of quantification was 0.032 pg/mL) [54]. Then, Pannee’s team [55] developed a tandem mass spectrometry for immunoprecipitation to be applied in AD research for the first time. The target protein was separated by antibodies coupled with beads, and then plasma Aβ42 and Aβ40 were quantified with internal standards labeled by mass spectrometry and isotope. High-precision analysis tools were used for exploratory evaluation, and the plasma Aβ42/40 level significantly dropped in the mild cognitive impairment (MCI) and AD groups compared with the control group [25, 57]. In similar studies, the correlation between peripheral Aβ and Aβ in the central nervous system has also begun to take shape. Specifically, there was a weak but significant correlation between plasma and CSF Aβ42/40 [55], and the plasma Aβ42/40 level of individuals with positive amyloid-PET significantly fell [24, 57]. Moreover, the finding of moderate to high consistency between plasma Aβ42/40 and amyloid-PET also added value to the potential of this plasma biomarker [10, 58]. Aβ42/40 seemed to be the most promising Aβ-related biomarker in blood. All the above preliminary findings have spurred larger studies on the potential of plasma Aβ42/40 for use in clinical practice.
DIFFERENTIAL DIAGNOSIS STUDIES
Aβ42/40 distinguishes normal controls from AD patients
CSF Aβ42/40
The severity of neuropathy in AD patients varies and partially overlaps with the pathological changes found in the brains of elderly individuals with normal cognition [59–62]. According to Beach [63], even in expert academic centers, the accuracy of pure clinical diagnostic criteria for AD seems to be unsatisfactory, and the sensitivity and specificity are only approximately 70%. The discovery of AD-CSF biomarkers has deepened our insight into the mechanism of disease. After more than 20 years of research, changes in CSF Aβ42 and Aβ40/42 are now widely used as “the characteristic of Alzheimer’s disease” in the routine diagnostic evaluation of suspected AD [6]. In addition, some literature has reported that the addition of CSF Aβ42/Aβ40 can dramatically optimize the diagnostic efficacy of AD-related scales in clinical settings [64, 65].
To begin, in studies that took clinical diagnosis as the reference (using a case—controlled design), an accumulated body of evidence clearly reveals that Aβ42/40 can promote sensitivity and specificity, and the area under the ROC of Aβ42/40 was significantly larger than that of Aβ42 [66, 67]. Previous studies also compared the Aβ42/40 value as an information tool of the whole AD spectrum with amyloid-PET, which is widely accepted as the pathological agent of AD and found that the negative correlation between CST Aβ42/40 and standardized uptake value ratios (SUVRs) of amyloid of different radioactive tracers was consistently proven [16, 69]. Sacchi et al. [70] also offered reliable evidence showing a strong correlation between CSF Aβ42/40 and amyloid-PET SUVRs, which was particularly strong in the posterior cingulate/isthmus cingulate cortex; this encephalic region was one of the earliest and most distinctively affected regions in AD [71, 72]. CSFAβ42/40 was almost 100% consistent with the Aβ state labeled by PET [6]. Interestingly, even if there were individuals with conflicting results (usually isolated positive CSF Aβ42/40), they usually turned PET-positive within a few years [6, 73]. Subsequent studies documented that CSF biomarkers not only identified early Aβ pathology with the same accuracy as PET but also displayed abnormalities earlier than PET [22, 75]. In summary, although the CSF Aβ42/40 ratio had no explainable significance, it had high performance in identifying individuals who were clinically diagnosed with AD with an unknown amyloid state in the brain and could identify abnormal cortical deposition of amyloid, as shown by PET. In this way, fewer patients were diagnosed with a false positive (low CSF Aβ42) or false negative (high CSF Aβ42) by conventional CSF biomarkers. The fact that CSF Aβ42/40 can identify AD-specific changes facilitated the universal acknowledgement of the CSF Aβ ratio as a surrogate marker of amyloid state in clinical trials and the ultimate clinical nursing environment of AD. Many working groups have suggested always calculating the CSF Aβ42/Aβ40 ratio when analyzing biomarkers but not CSF Aβ42 [8, 70]. However, Dumurgier [20] did not find any added value in the systematic review of Aβ42/40. Instead, given that the results of p-tau and Aβ42 were inconsistent, the application of the secondary classification of the Aβ ratio may draw biological conclusions in more than 50% of uncertain results and improve patient classification (NRI = 10.5%, p = 0.003). This finding seemed to call into question the idea that CSF Aβ42/40 can routinely replace CSF Aβ42. CSF Aβ42/40 was directly applied in first-line diagnosis when the initial CSF characteristic was uncertain, and more tests were required to use the ratio as a determinant in secondary diagnosis.
In addition, we also reviewed the progress of the standardization of CSF biomarkers and emphasized the importance of biomarker interpretation in a full clinical context. As far as the technical level was concerned, a unified protocol for CSF sample processing [76] in combination with reference materials and methods certified by CSF Aβ42 [77] as well as the introduction of fully automated analysis [78] together with standards for proper use of lumbar puncture [79] helped consolidate CSF data from different clinical research centers and allowed the development of a universal CSF Aβ42/40 cutoff level. Additionally, as the AD-CSF biomarkers are more often routinely measured in clinical neurochemistry, some methods and protocols were established, aiming at improving AD diagnosis and enabling comparison of interpretations of AD biomarkers measurements across laboratories applying different preanalytical handling procedures, analytical methods, cutoffs, or even sets of biomarkers. In 2009, the Erlangen Score (ES) interpretation algorithm [80] was first proposed to standardize diagnosis-oriented interpretation of CSF biomarker profiles and has been widely confirmed in different cohorts thus far [81–83]. The ES model was different from the dichotomy applied earlier (CSF normal/pathologic) [84], incorporating the concept of border zones for the first time. This led us to think prudently about how to interpret results close to the cutoff value of CSF Aβ42/40. For example, with the reference value of Aβ42/40 ratio in the Erlangen laboratory being 0.05, all results above this cutoff are considered entirely normal, whereas results ranging from 0.045 to 0.05 are interpreted as “borderline pathologic”, considering the reduction of overdiagnosis (more false-positive errors) was ethically more correct when there is a lack of effective therapy [85]. In 2014, Lehmann et al. [86] defined an intuitive classification scale based on the numbers (from 0 to 3) of pathologic CSF biomarkers (Aβ42, tau, and p-tau) named the Paris–Lille–Montpellier (PLM) scale, with the objective of identifying groups of patients with different predictive values for AD. Then, the optimized scale, the PLMR scale [87], which integrated the Aβ42/40 ratio and was compatible to multiple cutoffs resulting from different preanalytical protocols and kit providers, was subsequently presented in 2018. In the same year, the nominal-scale ATN system [6] for AD research, where “A” is the biomarker of Aβ, including positive cortical amyloid-PET and pathologic CSF Aβ42 or the Aβ42/Aβ40 ratio, was recommended by NIA-AA as an open standard framework that could be adapted by individual research groups to apply to their own research goals and environment. Separate from the subjective interpretation of the clinical status of the patients, it was intended purely to describe their biomarker status and, thus, was not a diagnostic criterion or guideline. Compared with the ES model, the CSF results in ATN were a fully descriptive category, rather than a rank variable, restricting the correlation analysis (at least the semiquantitative correlation) between CSF results and other indicators (odds ratio, progression hazards, or time from MCI to dementia) [85]. However, in different regions and AD research centers, there is no consensus regarding the interpretation of multivariate CSF data and the classification of patients in clinical practice and trials at present, which seems to depend on the personal wishes of clinicians or researchers.
Plasma Aβ42/40
As one of the few proteins related to AD neuropathology that can penetrate the blood—brain barrier [88], plasma Aβ is considered to be a metabolic byproduct of brain-derived amyloid plaques, and the establishment of new technology has made it possible to take plasma Aβ as a surrogate marker. Recently, the development of Aβ assays in blood mainly focuses on ultrasensitive immunoaffinity-based Simoa [23], Elecsys [58], and IMR platforms [89] and IPMS detection [10, 91]. Surprisingly, a considerable amount of evidence generated by other methods supported the notion that the plasma Aβ42/40 level of MCI/AD was obviously lower than that of CU. However, the application of the IMR detection technique resulted in the opposite conclusion. A series of studies [92–94] performed by Chiu’s team through IMR steadily showed that the plasma Aβ42 and Aβ42/40 ratios increased in MCI and AD patients in Taiwan. The findings of Lue [95] and Teunissen [96] confirmed this finding. The differences in plasma Aβ results may be attributed to distinct principles and designs between platforms. Based on the design of IMR analysis, the C-terminus of Aβ42 exposed in various conformations (e.g., isolated, complex, or oligomeric forms) can be specifically captured [96]. This probably involved the possible detection of oligomerized Aβ142 in plasma. A recent study reported a higher oligomer level of MDS plasma in AD patients than in normal controls [97].
On the other hand, a recent cross-sectional study reported that since the early days of AD, plasma Aβ42/40 has shown a significant correlation with Aβ deposition on CSF or amyloid-PET scans [23, 98–101]. It was particularly noted that despite a strong correlation between plasma Aβ42/40 and the latter two, they did not approximate 1.0, implying that the information provided by the three indicators was not exactly the same, and the differences and relationships between them need to be assessed in future studies. Moreover, although plasma Aβ42/40 showed consistent utility in determining abnormal CSF or PET states on the AD continuum, both showing very high AUCs [10, 58], there are differences between the analysis methods of different platforms in terms of accuracy. For mass spectrometry, the accuracy of plasma Aβ42/40 was 82–97% [10, 102–104], and the accuracy of fully automatic Elecys analysis was > 80% [58], whereas the accuracy of commercially available Simoa detection reportedly fell between 60–64% and 62–68% [23, 101]. To date, there are no reports on the application of IMR and MSD to verify the in vivo prediction of pathological features of the brain by plasma Aβ42/40. These improved methods have yielded desirable signs in the search for plasma biomarkers of AD. The results of plasma Aβ42/40 obtained by different research centers and diverse detection methods were collected, and the comparison can help introduce a uniform critical value. In this regard, we have only achieved a small piece of the work.
Research on plasma Aβ42/40 has focused more on its ability to detect amyloid positivity; however, the nuance of plasma Aβ42/Aβ40 between populations defined by the amyloid-PET state, a minor decrease of 14–20% compared with CSF at 50%, accordingly led to greater overlap of plasma Aβ between amyloid-PET(+) and amyloid-PET(-) individuals [10, 56]. This introduced problems when ad hoc testing in clinical routine or multicenter laboratory testing in trials are done to classify patients into amyloid +/-, as there will be variation and bias. To explore and resolve the issues in biomarker measurements, some studies examining the robustness of the plasma Aβ42/40 ratio have emerged. Benedet et al. [28] focused on random variation at the single biomarker level, demonstrating that the accuracy of plasma IP-MS Aβ42/40, which was the best predictor of elevated Aβ pathology, deteriorated with only a modest increase in assay variability. They found that IP-MS Aβ in combination with GFAP and p-tau181, two plasma biomarkers with lower overall accuracy but higher robustness, was the simplest model with the highest accuracy at the symptomatic phase of AD. In contrast, Cullen [105] showed that plasma p-tau217, rather than Aβ42/Aβ40, was the highest-performing individual biomarker in terms of separating Aβ–from Aβ+ participants. However, he drew a similar conclusion as Benedet et al.; that is, the clinical predictive performance of the multibiomarker panel was largely unaffected by test-retest variability. The existence of biologically doubtful cases (nonconcordant between two visits in a short time span) interestingly indicated that the relationship between random error and overlap between normal and abnormal groups are general problems in AD clinical chemistry; therefore, individuals with unstable predicted outcomes (“gray zone”) should be recommended for additional tests, such as CSF- and PET-based methods.
Furthermore, when the assay is characterized by a low dynamic range or when there is high overlap between positive and negative groups, in addition to the above random errors, systematic error caused by uncontrolled preanalytical factors (e.g., sample collection/handling) can significantly affect biomarker values [105]. Therefore, the key prerequisite for achieving consistency is to establish preanalysis standardization procedures. In 2015, the standard operating procedures (SOPs) for blood sample processing were presented for the first time. Although they were not based on experiments, the emergence of these guidelines provided a starting point for the coordination of procedures needed in the scientific validation stage of potential plasma biomarkers for AD [51]. With the emergence of new studies, these guidelines have been updated as needed. Recently, Rózga et al. [106] published an easy-to-use preanalysis sample processing protocol that is compatible with all plasma biomarkers of AD to supplement the Elecsys method. Walter et al. [107] reported the impact of temperature on Aβ stability in venous blood and clearly demonstrated that storing blood samples at a lower storage temperature (4°C) led to stable Aβ peptide concentrations for up to 72 h. In the latest report, the Standardization of Alzheimer’s Blood Biomarkers (SABB) workgroup sent blood samples to six laboratories, which corresponded to six different detection methods (two MS assays, two Simoa assays, and two ELISAs) to analyze Aβ42 and Aβ40 quantitatively and put forth an evidence-based, technology-, and biomarker-independent SOP. 1) The whole blood obtained by venipuncture was collected by a K2EDTA blood collection tube. 2) The whole blood was centrifuged at room temperature (RT) at 1800x for 10 min, and the resulting plasma was placed in a 250- to 1000-μl polyethylene storage tube. 3) The delay in centrifugation had a negative impact on the concentration of Aβ at room temperature in all technical analyses, so it was recommended that blood samples < 3 h from blood collection to centrifugation should be stored at room temperature, 2–8° for 3–24 h, –20° for 24 h-2 w, and –80° after a 2 w-time window for a long time [19].
In summary, some results of plasma Aβ determination have been obtained by referring to the CSF Aβ42/40 study. However, the field of blood biomarkers undoubtedly lacks a comparison of analytical platforms, clear preanalysis guidance, and the effects of common covariables on biomarker levels, which will inform us of the accurate cutoff value of biomarkers [108].
Aβ42/40 distinguishes AD from non-AD neurodegenerative diseases
CSF Aβ42/40
Frontotemporal dementia (FTD) and dementia with Lewy bodies (DLB) are two common types of neurodegenerative dementia following AD. As FTD and DLB overlap with AD in terms of clinical symptoms, cognitive test performance, and neuropathological features [109–113], it is relatively easy to detect dementia, but it remains challenging and critical to diagnose disease subtypes later [114]. A recent meta-analysis showed that 20% of DLB patients were inaccurately diagnosed [115], and misdiagnosing what should have been AD was the most frequent. Similarly, the prominent manifestation of AD neuropathology in patients clinically diagnosed with FTD suggests the possibility of AD being misdiagnosed as FTD [116–118]. As mentioned above, not only can AD be misdiagnosed as FTD and DLB, but FTD and DLB can also be misdiagnosed as AD. Beach et al. [63] discovered that approximately 20% of the pathological phenotypes of the brain of clinical AD patients did not support their prenatal diagnosis, and inconsistent cases were usually identified as Lewy body dementia, FTD, corticobasal degeneration, and progressive supranuclear palsy neuropathology. In terms of drug use, misclassifying FTD as AD can lead to serious adverse outcomes. Mendez et al. [119] followed up FTD patients who took donepezil (commonly used for AD) orally for six months and found that 1/3 of them showed more serious disinhibiting and compulsive behavior. Likewise, taking inappropriate drugs (e.g., anticholinergic and antipsychotic drugs) in DLB misdiagnosed as AD can raise its morbidity and mortality accordingly [115]. This dilemma pushed us to seek a new breakthrough in the diagnostic accuracy of biomarkers to ensure the application of a disease-specific therapeutic schedule in the course of the disease. The overlap of Aβ42 in AD and non-AD dementia patients made it useless in the differential diagnosis as an independent measuring method [120]. Therefore, whether taking the Aβ42/Aβ40 ratio as a combined measurement can promote discrimination ability has become a research hotspot. Below, we will describe the latest progress on the Aβ ratio of different fluids in relevant differential diagnosis studies.
Gabelle pointed out that the CSF Aβ40/42 of AD (the Aβ peptide ratio was reported as Aβ40/42) was higher than that of FTD, and this ratio can differentiate them with higher sensitivity and specificity (79% and 76%, respectively). Interestingly, however, the AUC values of Aβ40/42 and isolated Aβ42 for subtype differentiation were very close, without a significant difference [121]. Nevertheless, more studies resulted in the opposite conclusion. The Aβ42/40 ratio was noted to significantly improve the discrimination ability for AD and non-AD controls [38, 122–126]. Likewise, Struyfs [127] reported that the calculation of CSF Aβ42/40 can increase the diagnostic performance of CSF Aβ42 when distinguishing AD from FTD and MCI from non-AD. Hansson’s team [16] estimated the percentage of dementia patients who were only misdiagnosed by Aβ42 and correctly classified by Aβ42/40 in their review (16 studies using the Aβ142/Aβ140 ratio were included) and conservatively held that not using the Aβ42/40 indicator might lead to the misdiagnosis of 5–10% of cases. Meanwhile, Paterson et al. [128] evaluated the specificity of CSF Aβ42/40 for disease differentiation and found that at a fixed 85% sensitivity, the optimal specificity of the CSF Aβ ratio for the differentiation of AD and FTD subtypes was strikingly similar to the ratio free of AD pathology at autopsy. The specificity of each group in descending order was semantic dementia 100%, behavioral variant FTD 85%, progressive nonfluent aphasia (PNFA) 50%, and DLB 50%. They unanimously agreed that if the differential diagnosis problem was correctly set, the ratio of Aβ can help ameliorate the distinction between dementia. However, in the same study, the authors pointed out that the current CSF biomarkers cannot distinguish the syndromes generally induced by AD (PNFA) or mixed AD pathology (DLB) from AD, which emphasized the need for biomarkers that were pathologically specific to non-AD dementia. Moreover, Chaudhry et al. [129] showed through a meta-analysis that the CSF Aβ42/40 level in DLB patients was dramatically higher than that in AD patients, but the results were highly heterogeneous. The genesis of heterogeneity may be attributed to the fact that most of the current research subjects are clinical patients who lack autopsy confirmation, and potential misclassification undoubtedly restricts the research results. In the future, it will be necessary to evaluate the discrimination ability of CSF Aβ42/Aβ40 in patient groups defined by neurology.
Plasma Aβ42/40
At present, studies in search of biomarkers to distinguish AD from non-AD dementia have focused to a lesser extent on the blood matrix. There are few differential diagnostic tests involving plasma Aβ42/40. Janelidze’s team [69] found that compared with MCI Aβ- individuals (inferred as non-AD pathology), the plasma Aβ42/40 dramatically decreased in MCI Aβ+ patients. Subsequently, in the first study that verified multiple mesoscale detection methods and detected plasma Aβ42 and Aβ40 at the same time, Vogelgsang et al. [130] confirmed that there was a significant difference between AD dementia and “dementia control” without any AD pathological ground in the plasma Aβ42/40 level, but the AUC value of 0.76 was not sufficient to make the blood-derived Aβ42/40 measured by this method replace CSF or PET as a diagnostic tool. Palmqvist [131] also proved that the plasma Aβ42/40 level of AD patients diagnosed by the latest NIA-AA standard was lower than that of the non-AD neurodegenerative disease group, but its concentration was significantly lower than those of other biomarkers included. There is limited literature on the diagnostic performance of plasma Aβ in non-AD dementia patients, and larger-scale studies may be needed to evaluate whether these results are highly reproducible in patient groups and clinical settings. Moreover, a point of concern is that Lin [132] compared the plasma Aβ42 level of non-AD dementia patients and individuals with normal cognition and discovered that the plasma Aβ42 concentration of DLB was the lowest, but without a significant difference. Regrettably, the authors did not have corresponding Aβ40 data, and it was unclear whether the calculation of the Aβ42/40 ratio could make the difference between AD and non-AD dementia significant by normalizing Aβ42. Taken together, due to the limitation of research and the low sensitivity of early measurement techniques, amyloid became less specific, and research on the diagnostic value of plasma Aβ42/40 in distinguishing AD from non-AD dementia did not yield satisfactory results.
Factors associated with heterogeneity
It is highly important to identify covariates that may significantly affect the level of liquid biomarkers in CU or AD patients. It is conducive to better clinical interpretation to incorporate the positive threshold of each related subgroup when establishing it. At present, there is no study that systematically explored the effect of age, gender, lifestyle and other factors, and genetic variation factors on fluid Aβ42/40, but many studies have incorporated some information that interested us in the description of participants’ demographic factors. When the CSF biomarker is used as the reference, it is not recommended to adjust for age, gender, or the apolipoprotein E (APOE) ɛ4 allele [133], but the information related to blood biomarkers has not yet been confirmed. Most of the findings showed that plasma Aβ42/40 decreased with age [134, 135]. However, Palmqvist et al. [58] reported the age effect of plasma Aβ levels across all diagnostic categories and proved that there was no difference among age groups in the prediction of positive Aβ (the age subgroups were < 72 and > 73). Of course, there are also inconsistent perceptions in relation to other covariates. Schlinder [10] described plasma Aβ42/40 in males and carriers of the APOE ɛ4 allele, whereas other studies indicated that plasma Aβ42/40 was not affected by gender or the APOE ɛ4 allele [136]. Toledo [137] stated that only the blood glucose level, rather than all the indicators mentioned above, could affect the determination results of plasma Aβ42/40. At present, there is only preliminary evidence for factors with possible additional effects on biomarker performance. Subsequently, additional work is needed to promote more complete insight into individual clinical information.
Recently, some studies have begun to fully evaluate the correlation between blood markers and other factors. O’Bryant and his colleagues [138] were the first to explore whether race affected the discussion on plasma biomarker concentrations in AD in a multiethnic population without dementia from the community. They found that marked changes had taken place in both unadjusted and adjusted models in the plasma Aβ42/Aβ40 of Mexican Americans. On the other hand, many factors, including chronic kidney disease and the Charlson Comorbidity Index, have also been shown to be significantly correlated with higher plasma Aβ42/40 in CU participants, whereas smoking and chemotherapy were related to lower plasma Aβ42/40 in MCI/AD patients [139].
A particular point to note is that some scholars have stressed the importance of studying MCI and dementia in the LGBT population (including lesbian, gay, bisexual, and transgender people) [140–143]. In fact, compared with their heterosexual peers, LGBT people may experience more dementia risk factors, which can in turn lead to greater cognitive aging [144]. However, to date, we know very little about cognitive impairment in elderly LGBT individuals, let alone whether there is any difference among LGBT subtypes at the cognitive level. At present, no one has paid attention to whether the fluid level of AD biomarkers in LGBT individuals changes, which remains to be discussed in the future.
It is seemingly crucial to accurately interpret biomarker levels and develop reference ranges in different populations to synthesize the above information, which undoubtedly spurs us to obtain the whole picture of AD biomarkers. It should be noted that it is necessary to consider the concentration of blood biomarkers in the context of other factors, but this does not mean overinterpretation. Given the interaction between covariates, simply pointing out whether there is any difference between specific dependent variables in biomarker proper may confound the presentation of effective information; it seems more logical to consider the frequency of a given covariate in complications or other factors and then make a comprehensive interpretation [145].
PROGNOSTIC TEST
Correlation between plasma Aβ42/Aβ40 and disease phenotype
The pathological process of AD that anti-Aβ monoclonal antibodies aim to prevent begins as early as 10 years or more before the occurrence of cognitive symptoms [146–149], and the failure of clinical trials of secondary prevention involving mild to moderate AD at an advanced stage of the disease seems predictable. We urgently need biomarkers that can identify high-risk AD patients before symptoms present to promote the recruitment efficiency of disease-modifying trials for AD in the future. It is undeniable that both CSF biomarkers and amyloid-PET have high diagnostic and prognostic value [150–152]. Nevertheless, considering the high cost and low availability of this technique in the setting of primary health care institutions, it is mostly used as a one-time clinical diagnostic tool in groups with chief complaints concerning cognition and is not fit for extensive screening of asymptomatic community populations. Moreover, the fact that the diagnostic model based on plasma is not worse or even better than the CSF model [108] supports the notion that plasma Aβ detection with higher accessibility has greater specificity and scalability, which is needed for effective screening of the AD population. There has been important research progress in clinical practice thus far. In this paper, we only reported on the latest progress related to the potential of plasma Aβ42/40 as a predictor of AD and the future direction of development.
First, based on prior knowledge related to AD, the correlation between plasma Aβ42/40 as a potential screening biomarker and common neuroimaging phenotypes of AD is brought into focus. MRI and FDG-PET measure the secondary effects of disease on brain structure and function by showing different topographies of atrophy or hypometabolism, whereas amyloid-PET directly measures molecular pathology [153]. Notably, Simrén [25] offered clear evidence to prove that the baseline level of plasma Aβ42/40 was slightly correlated with gray matter loss induced by AD. Moreover, compared with other biomarkers, its longitudinal measurement can capture AD-related brain changes in cognitively unimpaired disease with the strongest ability. This information suggests that the plasma Aβ ratio can be adopted to identify AD throughout the clinical process. Few studies have reported the correlation between plasma Aβ levels and brain metabolism, and conflicting conclusions have impeded the achievement of consistency. Grijalba [100] subclassified an amnestic MCI (aMCI) group according to the visual interpretation of FDG-PET; the plasma Aβ42/40 level of aMCI FDG (+) was low. Throughout the follow-up, up to 72.4% of the group progressed to dementia, and their conversion rate was twice that of the aMCI FDG (-) group, which further supported the applicability of plasma Aβ42/40 in detecting short-term disease progression in individuals with prodromal AD. However, the potential connection between plasma Aβ and FDG-PET was not observed in all studies. For example, Lemercier [154] found that there was no statistically significant correlation between the concentration of plasma Aβ42/40 and the quantitative value of FDG-PET in the broad encephalic region at the cross-sectional and longitudinal levels in SMC individuals. The author explained that the absence of a correlation between them probably suggested that the plasma Aβ ratio was not an appropriate metabolic marker that can indicate neuron loss and disease progression. At present, there are limited data on the correlation between plasma Aβ42/40 and FDG-PET values. In the future, more studies may be needed to supplement the existing research. Most importantly, increasing evidence has indicated that plasma Aβ42/40 demonstrates a very high AUC in predicting the amyloid state of the brain characterized by amyloid-PET in the future [99, 156]. Moreover, Schindler [10] pointed out that those who were positive for plasma Aβ42/40 had a 15-fold risk of progressing to positive amyloid-PET findings after 18 months of follow-up, which was clearly of great potential value for recruitment for therapeutic trials.
Correlation between plasma Aβ42/Aβ40 and disease endpoint
Second, while evaluating the correlation between plasma Aβ42/40 levels and AD endpoints, we reviewed the existing literature and observed that the studies achieved controversial results. Differences in the surveyed population, detection method, and follow-up time may explain the heterogeneity in the effect of plasma Aβ related to MCI/AD [157]. Despite the initial setbacks, a large body of recent studies involving participants with good characteristics has tended to show that low-level plasma Aβ42/40 is associated with an increased risk of AD [100, 158–160] and a more distinct decline in composite cognition score [160–162]. Researchers found that patients with low baseline plasma Aβ42/40 had a 70% increased risk of progressing from aMCI to AD within two years [100]. These data were similar to the total risk ratio of progressing to AD under low plasma Aβ42/40 levels obtained from a previous meta-analysis of 13 studies involving 10,303 participants (RR = 1.60; 95% CI, 1.04–2.46; p = 0.03) [157]. Additionally, it is worth noting that in a 25-year large-scale longitudinal study by Sulivan and his colleagues [163] in 2021, the nonlinear relationship between plasma Aβ42/40 levels and the risk of cognitive impairment in middle-aged/elderly people without dementia was reported for the first time, thereby creating an impressive new threshold; that is, when it was < 0.20 pg/ml, the decline in plasma Aβ42/40 led to an increase in the risk of cognitive impairment, which probably suggests that individuals with a very low plasma Aβ ratio were particularly vulnerable. Low plasma Aβ42/40 may lead to a decline in longitudinal cognition, but little attention has been given to the effectiveness of different plasma ratio ranges in predicting the risk of cognitive impairment. To the best of our knowledge, a similar nonlinear relationship has not been reported in patients with memory chief complaints, such as subjective cognitive decline (SCD) and MCI. Later, Pan [164] pointed out that the predictive value of plasma Aβ42/40 in brain Aβ pathology also showed interesting nonlinear characteristics on the whole AD continuous spectrum. The increase in plasma Aβ42/40 in the MCI group indicated that the risk of positive amyloid-PET was low. In SCD with the same follow-up time, plasma Aβ42/Aβ40 did not show a similar protective effect. The baseline increases in Aβ42 and Aβ40 alone have conversely become powerful predictors of positive amyloid-PET. These findings encouraged us to substantiate the effectiveness of the predictive power of Aβ in different cognitive function subgroups, and the plasma Aβ42/40 range that was truly of predictive value in different subgroups deserves careful deliberation. The nonlinear evolution of plasma Aβ with the disease was reported at the same time [165, 166]. This was consistent with the following facts: Aβ42 and Aβ40 changed in parallel at the early stage, and then Aβ42 dropped continuously, with Aβ40 being stable. The synchronous growth of plasma Aβ42 and Aβ40 during SCD led to the inevitable weakening of Aβ42/40 changes at this cognitive stage, which supported the failure of the predictive power of plasma Aβ42/40 in SCD patients mentioned by Pan above. The determination of the degree of cognitive impairment was obviously vital for the evaluation of predictive value because at different stages on the AD continuous spectrum, the biological follow-up of disease progression corresponded to different clinical reactions. Here, we stress the need to better understand the biology and kinetics of plasma Aβ. Thus far, the “peripheral sink” theory provides a reasonable hypothesis for the nonlinear changes in plasma Aβ; that is, at the early stage of AD, compensatory brain-to-blood transport increases, and at the late stage, the dysfunction of the system in clearing Aβ from the brain to blood is removed [164]. However, the marked decline in Aβ42/40 in AD patients can also be explained by the selective deposition of Aβ42 as insoluble plaques in the brain, rather than the inherent defect of the clearance system [157, 167]. Obviously, the mechanism of plasma Aβ42/40 changes in the neurodegeneration of brain amyloid deposition requires more in-depth longitudinal studies. This information can help us gain deep insight into the driving mechanism of the predictive power of plasma Aβ.
In addition, in view of the desirable predictive power of plasma Aβ42/40, Pereira [168] performed a more detailed exploration. In addition to the fact that plasma Aβ42/40 had similar predictive power to CSF Aβ42/40, they also discovered that the analysis results of the continuous values of the dichotomous plasma Aβ42/40 and Aβ ratio as predictors were similar, and both were found to be the only independent predictors of amyloid-PET accumulation in nondementia individuals. The former, of course, was obviously easier to achieve in clinical practice. As described by Leuzy [133], many statistical methods have been put forth to classify continuous AD biomarkers as normal or abnormal. However, based on the difference in biomarker determination methodology and the heterogeneity of the surveyed population, it is still challenging to horizontally standardize a dichotomous positive Aβ42/40 threshold that can identify future longitudinal pathological changes and cognitive deterioration of AD globally. At present, there is no independent laboratory that has comprehensively tested the various cutoff values of plasma Aβ42/40 reported by scholars.
Potential utility of plasma Aβ42/40 in predicting AD
As mentioned above, despite many problems to be discussed in depth, existing research results support the potential use of plasma Aβ42/40 in the selection of high-risk individuals in clinical trials. The existing detection of Aβ42/40 in the blood tends to be highly sensitive, rather than specific. Therefore, in the clinical setting, where the amyloid state of patients is unknown, this plasma biomarker is probably the most useful. In other words, after most asymptomatic individuals are excluded by taking plasma Aβ measurement with a high negative predictive value as a first-line screening tool in primary health care institutions, those with abnormal results are referred to specialized memory clinics for invasive CSF Aβ tests or expensive neuroimaging for confirmation in Step 2; this seems to be a more acceptable and obviously cost-effective sequential diagnostic procedure. Li Yan [169] estimated the number of participants and PET scans required to identify 1,000 positive amyloid-PET individuals with and without the predetection of plasma Aβ42/40 to obtain the time- and cost-effectiveness induced by blood Aβ42/40 screening. Ultimately, they were surprised to find that prescreening with plasma Aβ42/40 can reduce the number of PET scans required by cognitively unimpaired individuals by 59%, decreasing the number of PET scans required by cognitively impaired individuals by approximately 19%, and halve the time for enrollment and screening costs (amyloid-PET scans + blood test). Moreover, the correlation between plasma biomarkers and cognitive deficits has been reported to be consistently statistically significant and even stronger in middle age, regardless of whether plasma Aβ42/40 was measured in middle age or old age [166]. This finding seems to lay a fairly strong theoretical foundation for active plasma screening 20 years before the onset of AD. In fact, Zetterberg et al. [170] recently suggested applying plasma biomarkers in clinical trials of AD to screen individuals without CSF or PET examinations, thus identifying cases at higher risk of AD.
CONCLUSION
With the development of AD, the physiological process of Aβ in vivo changes. Compared with CSF Aβ42 alone, CSF Aβ42/40 has greater potential to become a surrogate marker for the characteristic brain Aβ pathology of AD. Aβ42/40 seems to be the most promising Aβ-related metabolic marker in the CSF matrix thus far. With the emergence of ultrasensitive detection techniques, the experience of using CSF biomarkers for in vivo evaluation has further facilitated the recognition of blood Aβ42/40 as a new diagnostic and prognostic biomarker. Fluid Aβ42/40 can be adopted in clinical diagnosis together with other biomarkers or appear earlier in clinical trials. Quite a few studies provide promising information on the evaluation of the ability of Aβ42/40 in different media to distinguish normal controls from MCI/AD patients and highlight the prognostic value of future dementia risk. However, in the comparison between different detection techniques and research centers, more exploration and improvement are needed to establish the optimal threshold. It should be mentioned that the evaluation of the ability of the plasma Aβ42/40 ratio to distinguish AD from non-AD dementia is still a gap in the current research field. This is especially useful for the identification of different neuropathological mechanisms in individuals with dementia with overlapping clinical symptoms. The development of precision medicine has undoubtedly perfected the diagnosis and treatment of AD, and personalized treatment and clinical nursing may become a reality in the near future.
Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/22-0673r1).