Article Text
PDF
Abstract
Alzheimer’s disease (AD) is a common cause of dementia, characterised by cerebral amyloid-β deposition, pathological tau and neurodegeneration. The prodromal stage of AD (pAD) refers to patients with mild cognitive impairment (MCI) and evidence of AD’s pathology. At this stage, disease-modifying interventions should be used to prevent the progression to dementia. Given the inherent heterogeneity of MCI, more specific biomarkers are needed to elucidate the underlying AD’s pathology. Although the uses of cerebrospinal fluid and positron emission tomography are widely accepted methods for detecting AD’s pathology, their clinical applications are limited by their high costs and invasiveness, particularly in low-income areas in China. Therefore, to improve the early detection of Alzheimer's disease (AD) pathology through cost-effective screening methods, a panel of 45 neurologists, psychiatrists and gerontologists was invited to establish a formal consensus on the screening of pAD in China. The supportive evidence and grades of recommendations are based on a systematic literature review and focus group discussion. National meetings were held to allow participants to review, vote and provide their expert opinions to reach a consensus. A majority (two-thirds) decision was used for questions for which consensus could not be reached. Recommended screening methods are presented in this publication, including neuropsychological assessment, peripheral biomarkers and brain imaging. In addition, a general workflow for screening pAD in China is established, which will help clinicians identify individuals at high risk and determine therapeutic targets.
- Consensus
- Cognitive Dysfunction
- Neurocognitive Disorders
- Neuropsychological Tests
- Risk Assessment
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Introduction
Alzheimer’s disease (AD) is a common neurodegenerative disease that causes cognitive impairment in older adults, characterised by a series of pathological processes including the formation of amyloid-β plaques, hyperphosphorylation of tau proteins aggregated in neurofibrillary tangles, neuroinflammation and cell death. Mild cognitive impairment (MCI) is usually considered the (pAD), which refers to individuals with cognitive impairment that is not severe enough to result in a significant functional impact on daily activities.
China has the largest patient population of AD in the world, imposing a heavy social and economic burden on public health.1 2 In China, there were 9.83 million people aged ≥60 years with AD and 38.77 million with MCI.3 With the ageing population, the incidence is increasing rapidly, and the ranking of deaths due to AD in China rose from 10th in 1990 to 5th in 2019.4 The overall economic cost of AD in China was US$167.7 billion in 2015, which is expected to reach US$1.8 trillion by 2050.5 Despite the high prevalence and cost, diagnostic techniques and management strategies for AD in China are still inadequate, especially in rural areas. Efforts should be made to establish a screening flowchart for the early stages of AD using more efficient and inexpensive methods.
The neuropathological changes in AD can last for decades before the development of measurable cognitive symptoms, and timely intervention may delay the cognitive decline. To discover interventions that can prevent or delay the initial onset of AD, much attention should be paid to the predementia stage.
According to the 2018 National Institute on Aging-Alzheimer’s Association (NIA-AA) research framework, a biological definition of AD is established based on the AT(N) biomarkers.6 Biomarkers are grouped into those of amyloid-β deposition (A), pathological tau (T) and neurodegeneration (N), measured by positron emission tomography (PET), cerebrospinal fluid (CSF) and MRI. The cerebral deposition of amyloid-β protein is believed to be the core of AD’s pathogenesis, and an individual with biomarker evidence of amyloid-β deposition alone would be classified as having ‘AD’s pathologic change’.6
The cognitive symptoms of AD’s continuum are divided into six stages. Stage 1 is defined by biomarker evidence of AD in asymptomatic individuals. Stage 2 describes the earliest detectable clinical symptoms, including subjective cognitive decline (SCD), objectively defined subtle cognitive decline (obj-SCD) and neurobehavioural changes alone. Stages 1 and 2 both describe the preclinical phase of AD. Stage 3 refers to MCI. Stages 4–6 refer to mild, moderate and severe dementia, respectively.
The 2021 International Working Group (IWG) recommendations on the clinical diagnosis of AD introduce the concept of pAD, which refers to the early symptomatic and predementia phase and mainly includes the stage of MCI.7 The 2018 NIA-AA and 2021 IWG criteria both require AD’s pathologies for accurate diagnosis.
Since antiamyloid monoclonal antibodies have statistically improved cognitive and biomarker outcomes in recent AD phase III clinical trials, early and accurate identification of patients with positive amyloid-β deposition (Aβ+) becomes increasingly important. However, although PET and CSF are valid proxies for detecting AD’s pathology, they are either expensive or invasive. Meta-analysis showed the prevalence of Aβ+ in the AD population varied across different studies, and the number of patients at predementia stages remains uncertain in low- and middle-income regions where biomarker studies are missing.8 9 This highlights the need for less expensive and more widely accessible screening methods to identify individuals at risk for pAD in China.
Based on this background, the current consensus intended to recommend effective and cost-saving screening approaches to identify individuals at high risk of AD, particularly those with pAD. It should also be noted that the current consensus is not proposing methods that can fully replace amyloid-PET or CSF, but suggesting several options to flag individuals at-risk that warrant further diagnostic testing and avoid unnecessary examinations.
Materials and methods
A panel of 45 neurologists, psychiatrists and gerontologists (see the Acknowledgements section) was invited to review literature, vote and provide their opinions to reach this consensus. A majority (two-thirds) decision was used for topics for which consensus could not be reached. All members were required to disclose any conflicts of interest that may have a direct regulatory or commercial impact resulting from the publication of this consensus. No commercial funding was provided to support the literature review or the preparation of this paper.
Systematic literature search
We conducted a systematic literature review for promising pAD screening methods in the last 20 years. The PubMed database was used to search the literature for current screening methods related to pAD. The final search encompassed papers published from 2003 to 2023 (conducted on November 29, 2023) and was limited to clinical trials, meta-analyses, practice guidelines and research in humans, supplemented by major presentations at international meetings where abstracts were peer reviewed. The search terms included ‘prodromal Alzheimer’s disease’, ‘mild cognitive impairment’, ‘neuropsychological assessment’, ‘digital tests’, ‘blood tests’, ‘brain imaging’, ‘biomarkers’, ‘screening methods’ and ‘early diagnosis’. The search resulted in the retrieval of nearly 3000 manuscripts, which were screened by experts to include only articles with clinically accurate and relevant information and to remove duplicate papers, resulting in a final bibliography of 124 manuscripts.
Consensus statements and evidence-level categorisation
Following comprehensive discussion at national meetings, we achieved an expert consensus on the rapid screening of pAD, including neuropsychological assessment, peripheral biomarkers and brain imaging.
Levels of evidence are based on the Oxford Centre for Evidence-Based Criteria: Ia, systematic review of randomised controlled trials (RCTs) with homogeneity; Ib, individual RCT; Ic, all or none; IIa, systematic review of cohort studies with homogeneity; IIb, individual cohort study; IIc, outcomes research; IIIa, systematic review of case-control studies with homogeneity; IIIb, individual case-control study; IV, case series; V, expert opinion.
Grades of commendation include: A, consistent level 1 studies; B, consistent level 2 or 3 studies or extrapolations from level 1 studies; C, level 4 studies or extrapolations from level 2 or 3 studies; D, level 5 evidence or troublingly inconsistent or inconclusive studies of any level.
Data analysis
Neuropsychological evaluations
Recent developments in the field of cognitive testing have led to a growth of methodologies showing potential for pAD assessment, including conventional paper-and-pencil tests, assessment of metacognition, electronic cognitive assessment tools and digital behavioural markers. Non-cognitive assessment scales are also important for the clinical diagnosis of AD, such as the Activities of Daily Living Scale, Functional Activities Questionnaire, Hamilton Depression Scale, Hamilton Anxiety Scale and Neuropsychiatric Inventory, but their value in detecting pAD remains unclear.
Conventional paper-and-pencil tests
Traditional cognitive tests are the basis of clinical diagnosis. Several paper-and-pencil tests are widely used for the screening of MCI and dementia, such as the Mini Mental State Examination (MMSE), the Montreal Cognitive Assessment (MoCA) or MoCA Basic version, and the Third version of Addenbrooke’s Cognitive Examination.10–13 The effectiveness of these traditional tests to diagnose MCI is summarised in table 1.
Effectiveness of different cognitive assessments for detecting mild cognitive impairment (clinical diagnosis) and prodromal Alzheimer’s disease
Some cognitive tests are used for the evaluation of various cognitive domains, including the auditory verbal learning test (AVLT), Boston naming test, verbal fluency test, symbol digit modalities test and trail making test to assess the abilities of memory, language, attention and executive function, respectively. Meta-analysis showed that cognitive impairment related to amyloid-β is usually observed in semantic memory, visuospatial function and episodic memory, but such impact cannot identify the presence of amyloid-β deposits.14
Promising developments in the neuropsychological paradigm stress their association with biomarkers. For example, the visual short-term memory binding test,15 the Loewenstein-Acevedo Scales for Semantic Interference and Learning16 and the category switching test (CaST)17 have been reported to correlate strongly with cerebral amyloid burden (table 1). These tests rely on contextual information to support memory encoding and retrieval, semantic binding and controlled learning, which have recently demonstrated their use for the assessment of amyloid deposition18 and can be classified as ‘Aβ+ sensitive’ tests.
Metacognition
Metacognition reflects an individual’s reflection, regulation or evaluation of their knowledge or cognitive activity.19 As a core component of metacognition, metamemory represents an individual’s self-awareness and self-monitoring of memory activities. Approximately 80%–93% of MCI and mild AD cases have impaired metacognitive function.20 21 Previous research reported that decreased metacognition is associated with the accumulation of amyloid-β and tau proteins, as well as reduced brain metabolism and disturbed network connectivity.20 22–25
There are two key methods to evaluate metacognition. The first method indicated as ‘performance discrepancy’ is based on the discrepancy between the patient’s actual performance and their estimation scores on a certain neuropsychological test, which is usually combined in the semantic or episodic memory tasks,19 such as feeling of knowing judgements,26 judgement of learning27 28 and degree of confidence (DOC).29 The second method is ‘patient-informant discrepancy’, which is based on the calculation of discrepancy scores between questionnaires for the patient and their caregivers, such as the Everyday Cognition Scale,30 Measurement of Anosognosia Instrument31 and Memory Awareness Rating Scale.32
The presence of impaired metacognition in patients with MCI may be a risk factor for the transition to dementia, with individuals exhibiting this impairment being nearly three times more likely to progress to dementia within 2 years.33 In the preclinical phase of AD, some individuals may reflect a high level of awareness of subtle cognitive decline, evidenced by increased cognitive complaints, and this awareness declines as the disease continues to progress. Both longitudinal and cross-sectional studies have found that metamemory impairment precedes objective cognitive decline in Aβ+ patients.29 34
Previous studies reported that metamemory impairment occurs approximately 1.6 years before the diagnosis of MCI in Aβ+ patients,34 whereas the decline of anosognosia or metamemory can be detected approximately 3–4 years before reaching the clinical diagnosis of AD.34 35 Patients with MCI with metacognition deficit often exhibit overconfidence in their actual performance, such as making overestimated judgement in episodic memory tasks.29 The DOC,29 a subset of AVLT Huashan version, is sensitive to detecting SCD individuals with Aβ+ and warrants larger trials for further confirmation in the Chinese population. We summarised the current literature on metacognition for detecting pAD in table 1.
Electronic assessment tools
Electronic assessment tools are delivered by automated and intelligent cognitive measures, including the translation of existing standardised paper-and-pencil tests into computerised administration, and the development of novel electronic batteries based on promising techniques in neuropsychological approaches for the detection of cognitive impairment. We listed several commonly used electronic neuropsychological assessment tools that have been reported to be effective in screening for early cognitive impairment (table 1).
The advantages of electronic assessments include comprehensive documentation of both response speed and accuracy, independence from assessors, convenient data storage and remote administration. However, some electronic assessments still require manual assistance, and unfamiliarity with electronic devices may impact test results and lead to lower completion rates in populations with a lack of interest.
The digital clock-drawing test (DCTclock) has been found to be associated with abnormal amyloid and tau protein, with better discrimination ability than standard neuropsychological assessments such as the Preclinical Alzheimer Cognitive Composite (PACC). The DCTclock showed good discrimination performance between Aβ± cognitively normal groups with an area under the receiver operating characteristic curve (AUC) of 0.72, better than PACC (AUC=0.63) and hand-scored clock (AUC=0.58).36
The Brain Health Assessment (BHA) takes about 10 min to complete and consists of three subtests. The BHA subtests of Favorites (measuring associative memory), Match (measuring executive functions and speed) and Everyday Cognition Scale were observed to be significantly associated with Aβ+ (AUC=0.75).37
The Cogstate Brief Battery (CBB) takes about 10 min to complete, contains four individual card tasks and measures psychomotor function, attention, working memory and visual recognition memory. The CBB Learning/Working Memory Composite Score could discriminate between cognitively normal individuals (A−T−) and MCI (A+T+) with an AUC of 0.93, and differentiate MCI participants without biomarkers from pAD with an AUC of 0.86.38
However, these ‘Aβ+ sensitive’ tests have not been applied and verified in the Chinese population. Several electronic cognitive assessment tools have already been developed to screen individuals with predementia in China.39 For example, the Automated Memory and Executive Screening Instrument (AMES) is a self-rated screening scale that assesses individuals’ abilities of memory, language and executive function. It has good convergent validity with conventional tests and is good to discriminate patients with MCI (AUC: 0.88, sensitivity: 86%, specificity: 80%) and obj-SCD (AUC: 0.78, sensitivity: 89%, specificity: 63%) from normal controls.40
In addition, a voice recognition-based mobile cognitive assessment tool (Shanghai Cognitive Screening, SCS) takes about 6 min to guide users to self-administrate the cognitive assessment via voice interaction, and output instant reports about their test scores and voice features using machine learning techniques.41 Receiver operating characteristic (ROC) curve analysis showed that the SCS total score had an AUC of 0.92 to detect AD (sensitivity=90%; specificity=95%), and an AUC of 0.84 to detect MCI (sensitivity=79%; specificity=67%). The SCS subtests demonstrated moderate to high correlations with gold standard tests and correlated positively with hippocampal volumes.
The two electronic assessment tools discussed above are either programmed on a tablet or mobile app, which are easy to administer and effective to screen for early cognitive impairment in community-based settings. However, the relationship between such instruments and amyloid-β deposition needs to be explored.
With the growing popularity of smartphones and social apps among older adults in China, the population that can be reached and the scope of screening have greatly broadened, resulting in substantial savings in human resources. As an example, a 3 min game-based cognitive assessment tool (G3, a mini-program on the WeChat platform, https://www.bestcovered.com/products) has screened more than 17 million adults online. Social media on mobile phones helps establish pilot networks of people with cognitive impairment and related risks. However, validation studies are needed to clarify the effectiveness of the above tools in detecting AD’s pathology.
Notably, although electronic assessment tools have the potential to enable rapid, low-cost and large-scale screening in China, they are applicable to different scenarios. For example, G3/SCS could be used as a rapid dementia screening test on mobile devices for a community-based population, and AMES could be used as a tablet-based screening test to further classify individuals with MCI or subtle cognitive decline in primary care settings.
Some challenges also need to be tackled to ensure these digital instruments are ready for real-world application. First, the sample size of current electronic assessment tools is limited; further studies should include more participants to test their effectiveness, particularly in lower-educated and culturally diverse populations in China. Second, a quiet space and concentration on tests are required for most self-administered assessments, which may be an obstacle for at-home practice. Third, the digital divide caused by advanced age, low education and no access to or unfamiliar with electronic devices may also influence the application of these assessment tools.
Digital behavioural markers
Digital behavioural markers include physiological and behavioural information that is collected by digital techniques and quantifiable with clinical significance.42 With the development of digital technology and medical artificial intelligence (AI), behavioural information such as eye movement,43 44 olfactory identification,45–47 natural speech,41 48 49 driving50 and gait features51 are able to be detected via infrared, camera, recorder and wearable devices, contributing to the diagnosis and risk assessment of pAD (figure 1).
The digital behavioural markers for detecting prodromal Alzheimer’s disease. Parts of the figure were adapted from Servier Medical Art(https://smart.servier.com/), which by Servier is licensed under a Creative Commons Attribution 3.0 Unported License (https://creativecommons.org/licenses/by/3.0/). Aβ+, positive amyloid-β deposition; Aβ−, negative amyloid-β deposition; AR, augmented reality; ASR, automatic speech recognition; AUC, area under the receiver operating characteristic curve; EM, eye movement; GPS, Global Positioning System; MCI, mild cognitive impairment; mDEM, mild dementia; ML, machine learning; NC, cognitively normal adults; NLP, natural language processing; pAD, prodromal Alzheimer’s disease; SE, sensitivity; SP, specificity.
Gait
Reduced speed, extended step duration and varied trajectory are the gait characteristics associated with early AD’s pathology.52 Dual-task rather than single-task gait assessment is more recommended, during which motor and temporal parameters can be detected by wearable devices, infrared networks, cameras or gyroscopes inside mobiles and analysed via AI technologies.53 An augmented reality-based digital game uses gait features (ie, step distance, speed, time length, etc) to identify cognitive impairment and predict dementia, which correlates with amyloid-PET possibility.51
Driving
Digital driving features derived from vehicle Global Positioning System data, such as sudden braking or acceleration, night driving, single destination, frequent overspeed and number of drives, all correlate with amyloid burden measured by PET or CSF. The model based on machine learning also had good diagnostic accuracy to predict pAD (AUC=0.82).50
Natural speech
Language impairment frequently occurs in pAD, including word-finding difficulty, noun/verb reduction, repetitions, pauses and silences, grammatical variants and acoustic variations. Lexical semantic (AUC=0.80) and acoustic (AUC=0.77) features distinguished MCI from cognitively normal adults. Lexical semantic provides significantly better diagnostic accuracy than acoustic, with an AUC of 0.77 to detect amyloid status and an AUC of 0.61 to screen Aβ+ cognitively normal individuals.48
An automatic speech recognition software for cognitive impairment takes about 2–3 min to record and analyse the percentage of silence duration (PSD) of patients’ voice through a Cookie Theft picture description task to explore novel pAD screening methods. Results from the Chinese multicentre cohort found that PSD increased significantly in amnestic MCI (aMCI), with an AUC of 0.74 in the classification of aMCI and normal controls. In the Pitt centre cohort, PSD was verified as a reliable marker to differentiate mild AD from normal controls.49
Recommendations
Traditional paper-and-pencil neuropsychological tests are still the fundamental screening tools for evaluating and staging cognitive impairment (class V, level D).
Cognitive markers of memory binding, controlled learning and metacognition may facilitate early detection of pAD (class IIa, level B).
Electronic cognitive instruments show promise for the detection of underlying AD’s pathology (class IIb, level B).
Digital behavioural markers can contribute to the massive screening of cognitive decline (class IIIa, level B).
Peripheral biomarkers
Since the current PET or CSF tests indicating AD’s pathology are either expensive or invasive, we sought simple and accessible peripheral biomarkers as alternatives to identify pAD. At present, peripheral biomarkers with potential diagnostic and prognostic values are mainly derived from blood, urine and faeces. Although markers from saliva and tears have been studied, their diagnostic accuracy is far from conclusive.
Blood tests
Plasma amyloid-β (Aβ42/Aβ40) and phosphorylated tau (p-tau) are the most common blood-based biomarkers with potential clinical values. Given the relatively low levels of blood concentration, ultrasensitive methods such as single-molecule array (SIMOA), electrochemiluminescence immunoassay (ECLIA) and immunoprecipitation-mass spectrometry (IP-MS) are commonly applied for measuring plasma biomarkers.
Compared with other immunoassays, IP-MS assay showed much better accuracy and diagnostic value in measuring plasma Aβ, though numerous preanalytical steps were required.54 55 On the other hand, the measurement of plasma p-tau was more inclined towards using automated and high-throughput immunoassays, which also displayed an excellent value in clinical applications.56 The categorisation of various blood-based biomarkers for detecting AD is summarised in table 2.
Blood-based biomarker categorisation for Alzheimer’s disease
Across different clinical cohorts, the predictive accuracy for plasma Aβ42 or Aβ42/Aβ40 ratio measured via SIMOA to detect cerebral amyloid burden determined by CSF or PET was about 59%–82%, whereas a relatively higher accuracy of 72%–97% was observed using IP-MS.56–58 An automated diagnostic kit based on ECLIA for the quantitative determination of plasma Aβ42 and Aβ40 was first approved in Japan. Levels of plasma Aβ40 and Aβ42 measured by this assay were highly correlated with the results measured via IP-MS (r=0.91 and r=0.82), and the corresponding ratio of Aβ42/Aβ40 may effectively predict amyloid-PET with a sensitivity of 88.0% (95% CI 80.0% to 93.6%) and a specificity of 72.0% (95% CI 62.1% to 80.5%).59
Although the plasma Aβ42/Aβ40 ratio measured via IP-MS had high accuracy for predicting cerebral amyloid-β, the magnitude of the differences between Aβ± was only around 10%, less than the magnitude of 42% in CSF.60 61 This low magnitude undoubtedly limits the determination of cut-offs and their clinical application. In addition, an almost complete change of plasma Aβ42/Aβ40 was observed in the asymptomatic stage, which remained relatively stable during disease progression,62 making it difficult to detect disease progression. However, a lower level of plasma Aβ42/Aβ40 was still significantly associated with faster cognitive decline in the future.63
Plasma p-tau181, p-tau217 and p-tau231 are the most common p-tau proteins. In the populations with different cognitive status, plasma p-tau181 measured via SIMOA and IP-MS had an accuracy of 70%–88% and 67%–95% in discriminating brain amyloid-β positivity.56 64–66 In addition, plasma p-tau181 is tightly associated with the progression of brain tau pathology only in the population with Aβ+.67 68 Even in individuals without positive tau-PET, increased levels of plasma p-tau181 and p-tau217 were observed with positive amyloid-PET,69 70 and p-tau231 was further associated with cerebral amyloid burden in cognitively unimpaired individuals.71 These findings indicated that the levels of plasma p-tau may predate the amyloid-PET examination. In contrast to plasma Aβ42/Aβ40, plasma p-tau181 showed an increasing trend and peaked at the dementia stage,68 which assisted in monitoring disease progression. Its effectiveness in predicting conversion to AD significantly surpasses that of plasma Aβ42/Aβ40.72
In conclusion, plasma Aβ42/Aβ40 and plasma p-tau have potential value in screening pAD and predicting disease progression. For individuals with abnormal levels of plasma biomarkers, a clinical visit for further diagnostic tests is recommended as soon as possible. Meanwhile, using blood-based biomarkers as the first screening step may improve the efficiency of clinical trials. Though plasma biomarkers cannot currently be used as primary endpoints in clinical trials, they can be regarded as exploratory outcomes and have potential value to inform decisions. We summarised the accuracy of blood-based biomarkers for detecting brain amyloid pathology in online supplemental table S1.
Supplemental material
Besides Aβ42/Aβ40 and p-tau, blood-based biomarkers such as neurofilament light and glial fibrillar acidic protein are also associated with AD’s pathology and disease progression.73 Lipid metabolism indicators such as oxidised low-density lipoprotein, serum inflammatory-based indicators such as advanced oxidation protein products and transforming growth factor-β, and platelet-related markers such as β-secretase all showed significant differences between cognitively normal individuals and MCI.74 However, the relationship between these markers and AD still needs further investigation.
Urine tests
Urine samples have also been reported to contain potential biomarkers for AD, including urinary metabolites, proteins and DNA. However, available evidence is limited, and additional research is needed.
The urine formaldehyde and formic acid have been found to be correlated with global cognitive function, apolipoprotein E (APOE), plasma Aβ42 and p-tau181/t-tau and brain amyloidosis. The AUCs of urinary formic acid and formaldehyde in distinguishing normal controls from AD were 0.80 (sensitivity: 66.7%, specificity: 78.9%) and 0.57, respectively.75 76 Using urinary formic acid and formaldehyde levels could improve the prediction accuracy for disease status.
The urinary arginine levels and global arginine bioavailability ratio (GABR) in patients with aMCI are significantly reduced and positively correlated with MMSE. ROC analysis showed that to differentiate between aMCI and normal controls, the AUC of arginine is 0.68 (sensitivity: 80.8%, specificity: 42.3%), and the AUC of GABR is 0.80 (sensitivity: 84.6%, specificity: 80.8%).77
The CSF levels of Alu sequence-containing cDNA of neuronal thread protein (AD7c-NTP) overexpressed in AD have been reported to be associated with the severity of dementia. A significant difference in urine AD7c-NTP has also been found between Aβ± subjects. Using 1.46 ng/mL as a cut-off, 68.8% of Aβ+ individuals showed elevated urine AD7c-NTP level, and 92.9% of Aβ− subjects showed normal urine AD7c-NTP level.78
Faecal tests
Clinical, in vivo and in vitro studies have shown that the brain-gut-microbiome axis plays an important role in the onset and development of AD. A large number of gut microbes and their metabolites have shown promise as novel diagnostic and therapeutic targets for AD (figure 2).
Gut microbiota associated with Alzheimer’s disease.
Gut microbiota diversity and alteration are associated with cognitive decline and related pathological deterioration.79 80 It was reported that compared with normal controls, the diversity of faecal microbiota in AD was reduced, along with an increase in proinflammatory and a decrease in anti-inflammatory bacteria.81
The proinflammatory phyla such as Proteobacteria and Bacteroides and the genera of Dorea spp, Lactobacillus spp, Streptococcus spp, Bifidobacterium spp, Blautia spp and Escherichia spp increased, whereas the anti-inflammatory phylum Firmicutes and genera Bifidobacterium spp, Alistipes spp, Bacteroides spp, Parabacterium spp, Sutterella spp and Paraprevotella spp decreased in AD.82
The phyla γ-Proteobacteria and Enterobacteriaceae are progressively enriched in individuals with normal cognition, aMCI and AD.81 The abundance of some microbiota is positively correlated with cognitive scores.83 Patients with cognitive impairment and brain amyloidosis exhibited a much lower abundance of Eubacterium rectale and a higher abundance of genera Escherichia/Shigella.80
Notably, there are some inconsistent findings. For example, faecal microbes exhibited different abundances between normal control and AD, while no differential microbe was observed between MCI and AD.84 85 There was also no significant heterogeneity between MCI and AD regarding the CSF levels of the gut microbiome-dependent metabolite,Trimethylamine-N-oxide.85
Additionally, animal studies provide strong evidence supporting the association between AD and gut microbiota. The divergence of gut microbiota composition between the APP/PS1 and wild-type mice was proved to start at a young age (1–3 months), before the detection of amyloid deposition, which suggested that gut bacteria alteration could aid the early detection of AD prior to pathological evidence.86 87
Microbial metabolites associated with AD mainly include bile acids,88 short-chain fatty acids,89 branched-chain amino acids, indole and pyrimidine,90 and steroid hormones.91 Their catabolism and synthesis are regulated by the structure and function of gut microbiota.92 In addition, metabolites and microbes associated with chronic inflammation and immunity have also been linked to cognitive status and early brain neuropathic changes.93 Online supplemental table S2 lists some microbial metabolites associated with AD.
Machine learning models based on microbes, metabolites and the combination of microbes and metabolites are helpful for the early diagnosis of AD.81 Studies have shown that a series of random forest models based on 11 faecal microbe genera and their combination can distinguish cognitively normal patients and patients with MCI (considering results of MRI and PET), with the AUCs, sensitivities and specificities of 0.70–0.91, 67%–93% and 57%–93%.85
Faecal biomarkers for AD hold thriving prospects, yet they are still in their early phases and therefore a great deal of work needs to be done for their clinical application, including sample collection methods, measurement accuracy, interpretation of underlining mechanisms and the construction of independent or joint models.
Recommendations
Based on ultrasensitive methods such as SIMOA, ECLIA and IP-MS, plasma Aβ42/Aβ40 and plasma p-tau can be used to predict cerebral amyloid deposition and conversion to AD, with high sensitivity and specificity to detect pAD (class IIb, level B).
The popularisation of blood-based biomarkers depends on the standardisation of data from different laboratories (class V, level D).
Urinary formic acid, GABR and AD7c-NTP can help predict cerebral amyloid deposition (class IIIb, level B).
Models based on multiple microorganisms, metabolites and combinations of microorganisms and metabolites contribute to the early diagnosis of pAD (class IIIb, level B).
Brain imaging
Neuroimaging and electrophysiological examinations can effectively inform the diagnosis of pAD. Although PET imaging is not yet sufficient for rapid screening, a variety of other techniques have also shown promise in this field.
MRI
According to the 2018 NIA-AA biological framework,94 biomarkers in the (N) group indicate neurodegeneration and are usually tested by MRI. Although ‘N’ biomarkers are not specific to AD’s pathology, the use of MRI still plays an important role in several aspects: (1) excluding cognitive impairment caused by other diseases; (2) combined with other AD biomarkers to predict disease progression95; and (3) clinical staging and differential diagnosis.94
Structural MRI is one of the most important methods to detect pAD, which can detect structural brain changes 10 years before clinical cognitive decline in AD. Hippocampal volume can be regarded as an indicator to evaluate the pathological change caused by AD.96 The visual assessment scale of medial temporal lobe atrophy (MTA scale) is concise and commonly used, but it is not sensitive enough for younger patients with pAD and may lead to false negatives.
Both hippocampal volume and hippocampal texture provide valuable information for predicting the conversion from MCI to AD.97 98 The asymmetry of bilateral hippocampal atrophy, in which the right hippocampus is more atrophic than the left hippocampus, may also help in the early detection of AD. A systematic review showed that the volume of overall hippocampi could identify MCI with a sensitivity of 73% and a specificity of 71%; MTA with a sensitivity of 64% and a specificity of 65%; and lateral ventricles with a sensitivity of 57% and a specificity of 64%.99 Analysis of hippocampus evolution patterns increases the accuracy to 91.76% for conversion prediction.100
However, the screening value of other brain regions such as the entorhinal cortex, whole brain volume, lateral temporal lobe, amygdala, medial temporal gyrus or grey matter volume varied across studies. In Aβ+ patients with MCI, MTA and posterior atrophy were associated with an increased risk for progression to dementia, including the posterior cingulate sulcus, precuneus, parieto-occipital sulcus and parietal lobes.101–103 Furthermore, smaller medial temporal lobes were found in SCD subjects with abnormal CSF Aβ42.104
The atrophy of hippocampal subfields has a potential value in the differential diagnosis of AD.105 Atrophy in the insula, amygdala, precuneus, hippocampus and other temporal regions occurred before the clinical threshold for CSF amyloid positivity,106 and an automated classifier based on clinical, imaging and APOE can identify the presence of amyloidosis with a moderate level of accuracy.107
Longitudinal brain volumetric changes can also predict the presence of amyloid abnormalities and can avoid 55% unnecessary CSF or PET scans.108 Using radiomics models from MRI can help predict amyloid positivity in patients with MCI.109 We summarised the structural MRI regions associated with pAD in table 3.
Brain regions associated with prodromal Alzheimer’s disease
Some other MRI technologies, such as resting-state functional MRI and diffusion tensor imaging, as well as novel AI-based approaches, such as machine learning and convolutional networks, have improved the accuracy of MRI for the diagnosis of pAD.106 110–113 Multiscale graph-based grading of anatomical structures can accurately predict the conversion of MCI to AD.114 Although these methods lack practicality for simple screening purposes, they deserve continuous attention and further research.
Retinal imaging techniques
The retina shares the same embryological origin and physiological characteristics as the central nervous system and is structurally and functionally associated with the brain. As the only part of the central nervous system that can be directly visualised, biomarkers of retina imaging allow potential non-invasive assessments of pAD.
Recent advancements in retinal imaging techniques include the following: (1) optical coherence tomography (OCT), which provides measurement of the thickness of the retinal nerve fibre layer (RNFL) and ganglion cell-inner plexiform layer (GCIPL). The deposition of amyloid-β in the eye of patients with AD causes loss of ganglion cells and their axons, which ultimately leads to optic nerve degeneration and the thinning of RNFL and GCIPL115; (2) OCT angiography, which provides high-resolution images of the choroidal microvasculature to visualise the gradual changes in retinal blood vessels116; (3) electrophysiological examination of the retina, such as the electroretinogram; and (4) other new retinal imaging technologies for detecting amyloid deposition, such as retinal hyperspectral imaging and adaptive optics scanning laser ophthalmoscopy.117
Electrophysiological examination
Scalp electroencephalogram (electroencephalograph, EEG) records the sum of the postsynaptic potential generated by the pyramidal cells of the cerebral cortex, which can reflect the synaptic function of the brain. EEG is an economical, convenient and non-invasive screening method that can be used as a marker for pAD. Resting-state EEG, event-related potentials (ERP) and sleep EEG are the main EEG screening modules.
Resting-state EEG
Spectral analysis studies the EEG in terms of its dominant frequency, power (or amplitude), phase and synchrony of the EEG rhythm. Resting-state eye-closed EEG rhythms often change with physiological ageing. In the resting state, the EEG power density spectrum (power spectrum density) of healthy older individuals, MCI and AD showed differences in distribution and frequency.
Increased delta or theta power density, decreased alpha and beta power density and slowed mean EEG frequency have been shown to predict the progression from MCI to dementia. The alpha rhythm of high-power density in the back of the head also predicts more stable cognitive function in MCI subjects.118 In patients with MCI, increased power density of theta and delta rhythms and decreased power density of beta rhythms in temporal and occipital regions may indicate disease progression.119 As neurodegeneration develops, abnormalities in brain network connections that affect cognitive function develop.
The ideal approach would be to extract some indicators of functional brain connectivity abnormalities in neural networks to reveal such changes in cognitive function. Measurement of the functional coupling of rhythms between pairs of resting-state EEG electrodes with eyes closed is a promising marker of functional neural connectivity. Therefore, spectral analysis and ‘interrelatedness’ resting-state EEG measures (eg, directed transfer function, phase lag index, linear lagged connectivity, etc) at delta, theta and alpha frequency bands may be useful for stratification of AD and monitoring of disease progression and intervention.120
Event-related potential
ERP is brain electricity extracted from spontaneous potentials, known as evoked potentials. ERP can directly express the electrical response of the cerebral cortex to sensory, emotional or cognitive events.121 In the process of ERP research, researchers have made a finer division of ERP components, including ERP component polarity, cortical source location, amplitude and latency. Online supplemental figure S1 shows the waveforms and latencies of common ERP components. The ERP indicators related to early recognition of cognitive decline are listed in online supplemental table S3.
Sleeping EEG
Approximately two-thirds of patients with MCI subjectively report sleep-wake disturbances.122 Sleep monitoring using nocturnal polysomnography revealed abnormalities in the macrostructure of sleep in MCI relative to age-matched controls: prolonged sleep onset, delayed rapid eye movement (REM) sleep onset and decreased duration of REM and slow-wave sleep.123 Studies have shown lower overall δ in non-REM (NREM) sleep and lower overall power in both NREM and REM sleep in aMCI compared with controls.124 Pathological changes in NREM and REM sleep may predict the trajectory of cognitive decline in older adults.
The advantages of neuroimaging techniques, electrophysiological examinations and retinal imaging technologies lie in their relatively low cost, rapidity and non-invasiveness to allow massive screening in the context of large-scale screening applications, but their current applications are still limited due to a lack of validation studies with biomarker evidence.
Recommendations
T1-weighted MRI is a feasible and reliable imaging method for screening pAD. The atrophy in some brain areas (eg, hippocampus, amygdala, precuneus, temporal lobe) is sensitive to amyloid pathology (class IIb, level B).
Non-invasive retinal examinations (eg, OCT) have potential value for screening pAD (class IIIb; level B).
Brain electrophysiology examination is relatively cheap and easy to conduct, and can be employed as ancillary diagnostic tests for detecting early cognitive decline, yet their predictive value still needs more research (class IIIb, level B).
Screening process for pAD
In conclusion, we recommend the following three steps to screen for pAD (figure 3):
Workflow for screening prodromal Alzheimer’s disease. Aβ+, positive amyloid-β deposition; ACEIII, Third version of Addenbrooke’s Cognitive Examination; AD-8, Ascertain Dementia 8-item Questionnaire; CSF, cerebrospinal fluid; G3, 3 min version of game-based cognitive assessment; MoCA, Montreal Cognitive Assessment; pAD, prodromal Alzheimer’s disease; PDD, Parkinson’s disease dementia; PET, positron emission tomography; VaD, vascular dementia.
Step 1: A preliminary screening should be carried out in the community or primary care settings using brief screening scales. Individuals who are suspected to have cognitive impairment are then transferred to specialised outpatient clinics for further examination, including systemic medical history collection, physical examination, laboratory testing and brain imaging. After the initial procedures described above, individuals can be categorised into one of three groups: cognitively normal, dementia and transitional stage. Cognitively normal adults should be followed up. Patients with dementia should be given standardised treatment.
Step 2: Individuals in the transitional stage should proceed to further screening, including ‘Aβ+ sensitive’ cognitive tests, blood biomarkers and brain atrophy evaluation. Individuals with impaired scores on ‘Aβ+ sensitive’ tests (such as CaST and metamemory), abnormal blood biomarkers (decreased plasma Aβ42/Aβ40 and increased p-tau181) and brain atrophy in the hippocampus, amygdala, precuneus or temporal lobe should be assigned the label of ‘high-risk pAD’ and proceed to step 3 for further examination. Patients who are suspected of having other diseases (eg, cognitive impairment caused by vascular dementia or Parkinson’s disease) should be given appropriate interventions. Those with no identifiable cause (uncertain aetiology) should proceed to step 3.
Step 3: If the individual is still undiagnosed after all the above examinations, a PET scan or CSF testing should be performed for a definite diagnosis.
Conclusion and future directions
This consensus provides a series of non-invasive, low-cost and easy-to-use approaches for the rapid screening of pAD. The current literature review shows promising evidence that advances in ‘Aβ+ sensitive’ cognitive tests, novel blood biomarkers and MRI techniques are potential measurements of AD’s pathology. The application value of more advanced technologies, such as digital markers, urine and faecal tests and non-invasive retinal imaging, needs to be further explored. Since blood tests are cheaper and less invasive than PET/CSF, they might be the first step in the screening process for pAD in the near future. However, current peripheral biomarkers are tested in different laboratories using varied methods and measurement standards in China, which can impact the comparability and accuracy of screening. Therefore, establishing standardised testing methods and procedures to facilitate future advancements in precision medicine is of great importance.
Future research should include further validation of new screening methods (including strengths and limitations), improvements to existing methods and further large-scale studies to validate the efficacy of pAD screening tools. Moreover, if a vaccine for AD becomes available in the future, interventions should focus on the cognitively impaired individuals and on those with normal cognition underlying AD’s pathology. We believe that before novel markers become reliable screening methods, it is necessary to confirm that they are related to AD’s pathology. Having such screening methods available will greatly increase diagnostic accuracy in clinical trials and can be promoted in memory clinics or primary care settings. This expert consensus is proposing pAD screening methods for China’s current state of affairs and offering insights for the screening of cognitive disorders in low-income regions worldwide.
Ethics statements
Patient consent for publication
Acknowledgments
This consensus obtained advice from 32 experts apart from the 13 authors. We appreciate the contributions of Xinyi Cao, Yunpeng Cao, Qin Chen, Yan Zeng, Yifeng Du, Ying Han, Chuzhong Huang, Binyin Li, Fang Li, Xia Li, Yunxia Li, Wenlin Ma, Dantao Peng, Guoping Peng, Yanghua Tian, Jianjun Jia, Jiong Shi, Yi Tang, Huali Wang, Gang Wang, Yanjiang Wang, Fang Xie, Jun Xiao, Qun Xu, Jiewen Zhang, Min Zhang, Nan Zhang, Wei Zhang, Yu Yang, Jintai Yu, Junjian Zhang and Qianhua Zhao. Parts of this consensus have been published in the Chinese Journal of Neuromedicine previously.
References
Lin Huang has served as an attending physician at the Department of Gerontology in Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, China since 2019. She obtained her master's degree from the Medical school of Fudan University in China in 2019. Her research focuses on clinical neuropsychology and the identification of neurobehavioural mechanisms of cognitive dysfunction and dementia, including the early diagnosis and neural underpinnings of Alzheimer’s disease.
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors XC, JJ and QG designed and initiated the consensus. LH, QL, YLu, FP, LC, YW, YM, TC, YLi and JW carried out the literature review. LH drafted the manuscript. XC, JJ and QG critically reviewed the manuscript. QG served as the guarantor of this study. All authors read and approved the final manuscript.
Funding This work was supported by the National Natural Science Foundation of China (82171198, U20A20354) and the Sci-Tech Innovation 2030 Agenda of China (2022ZD0211603).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.