Predicting not just if, but also when, cognitively unimpaired individuals are likely to develop onset of Alzheimerʼs disease (AD) symptoms would be useful to clinical trials and, eventually, clinical practice. Although clock models based on amyloid and tau positron emission tomography have shown promise in predicting the onset of AD symptoms, a model based on plasma biomarkers would be more accessible. Using longitudinal plasma %p-tau217 (the ratio of phosphorylated to non-phosphorylated tau at position 217) from two independent cohorts ( n = 258 and n = 345), clock models were used to estimate the age at plasma %p-tau217 positivity. The estimated age at plasma %p-tau217 positivity was associated with the age at onset of AD symptoms (adjusted R 2 of 0.337−0.612) with a median absolute error of 3.0−3.7 years. Notably, the time from %p-tau217 positivity to onset of AD symptoms was markedly shorter in older individuals. Similar models were constructed with data from one p-tau217/Aβ42 immunoassay and four plasma p-tau217 immunoassays. These findings suggest that the time until onset of AD symptoms can be estimated using a single blood test within a margin of error that is acceptable for use in clinical trials.

BACKGROUND:Alzheimer's disease (AD) progression varies widely among individuals. Identifying factors influencing timing of pathology and clinical progression is crucial for optimizing early intervention trials. OBJECTIVES:To investigate how the estimated age at amyloid and tau PET positivity, and the time interval between these two key events ("amyloid-tau time interval"), relate to symptom onset and clinical progression, and to assess the effects of APOE-ε4 status and sex on these associations. DESIGN:This analysis used data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Harvard Aging Brain Study (HABS). SETTING:The ADNI is a multicenter observational cohort conducted at 55 sites across the United States; The HABS is a longitudinal, single-center observational cohort. PARTICIPANTS:This study included participants with at least one positive amyloid PET scan (ADNI n = 792; HABS n = 104) or at least one positive tau PET scan (ADNI n = 212; HABS n = 48). All participants had information on sex, APOE-ε4 status, and longitudinal cognitive assessments. MEASUREMENTS:We examined the influence of APOE-ε4 status, sex, and their interaction on the estimated age at biomarker positivity and the amyloid-tau time interval. Accelerated Failure Time (AFT) models were used to predict time to symptom onset (CDR > 0) based on estimated biomarker positivity age and the amyloid-tau time interval. Linear mixed-effects (LME) models evaluated differences in the rate of cognitive decline, as measured by CDR-SB, over five years following symptom onset according to estimated biomarker positivity age and amyloid-tau time interval. Additional models included interaction terms with sex or APOE-ε4 status. RESULTS:The amyloid-tau time interval varied markedly between individuals and was shorter in APOE-ε4 carriers, women, and those with older age at amyloid PET positivity. APOE-ε4 carriers and women became amyloid and tau PET positive at younger ages. Following amyloid PET positivity, a shorter time to tau PET positivity predicted earlier symptom onset. After symptom onset, faster cognitive decline was observed in individuals with younger ages at amyloid or tau PET positivity. The time to symptom onset following tau PET positivity, or the rate of cognitive decline after symptom onset, were not influenced by the amyloid-tau time interval. CONCLUSIONS:After becoming amyloid PET positive, APOE-ε4 carriers, women and older individuals may have a shorter window for detection and treatment before they become tau PET positive and develop symptoms. These findings should guide the identification of individuals at highest risk of rapid AD progression, enabling more efficient participant selection for clinical trials.

INTRODUCTION:While most studies of Alzheimer's disease (AD) examine cross-sectional relationships among biomarkers, longitudinal relationships are also highly relevant. METHODS:This study in the Alzheimer's Disease Neuroimaging Initiative cohort (n = 373) used non-parametric Spearman correlations to explore the relationships of baseline values and rates of change in plasma biomarkers and rates of change in key AD outcomes. RESULTS:Compared to rates of change of plasma biomarkers, baseline values of plasma biomarkers were more strongly associated with rates of change in key AD outcomes. Change in amyloid positron emission tomography (PET) was most strongly associated with baseline values of amyloid beta (Aβ)42/Aβ40 and phosphorylated tau (p-tau)217, especially in amyloid PET--negative individuals. Changes in cortical thickness and measures of cognition were most strongly associated with baseline p-tau217, especially in amyloid PET-positive individuals. DISCUSSION:Baseline p-tau217 is associated with rates of change of amyloid pathology and cognition. Visualization tools were developed to enable researchers to explore AD biomarker relationships.

Introduction:Studies of the risk and timing of symptomatic Alzheimer's disease (AD) in cognitively unimpaired individuals are challenging due to the relatively small number of clinical progressors and limited clinical follow-up, which can lead to design-related associations. Clock models can be used to anchor the timing of events to biological events such as biomarker positivity. We hypothesized that estimated age at plasma %p-tau217 positivity based on clock models is less affected by design-related associations as compared to baseline age. Methods:Data from the Knight Alzheimer Disease Research Center (Knight ADRC) and Alzheimer's Disease Neuroimaging Initiative (ADNI) were analyzed. Age at %p-tau217 positivity was estimated using two clock model approaches, TIRA and SILA. The C-index of estimated age at plasma %p-tau217 positivity and age at the baseline plasma sample (baseline age) for ranking age of AD symptom onset was evaluated in initially cognitively unimpaired individuals, including progressors and non-progressors. In progressor sub-cohorts, baseline age and time from %p-tau217 positivity to baseline were associated with time from baseline until symptom onset; baseline age and estimated age at %p-tau217 positivity were associated with age at symptom onset. Commonality analyses partitioned the variance unique to each predictor and shared between predictors. Randomization analyses evaluated whether observed associations exceeded those expected by chance. Results:Estimated age at %p-tau217 positivity enabled analyses of a greater number of progressors in the research cohorts, which did not have plasma %p-tau217 data from every clinical assessment. The estimated age at %p-tau217 positivity had a higher C-index than baseline age for ordering the likelihood of AD symptom onset when all follow-up was considered; when follow-up was truncated, the C-index for estimated age at %p-tau217 positivity remained stable while the C-index for baseline age became inflated. In progressors, estimated age at %p-tau217 positivity contributed unique variance beyond baseline age in associations with age at symptom onset. Randomization analyses in the larger Knight ADRC found that associations between clock-derived measures and time from baseline until symptom onset and age at symptom onset exceeded the permuted null distribution, with some mixed results in the smaller ADNI cohort. Conclusions:Compared to baseline age, the biologically-anchored estimated age at %p-tau217 positivity is less susceptible to design-related associations and incrementally improves prediction of age at symptom onset in analyses conditional on progression.

BackgroundNeuroinflammation actively contributes to the pathophysiology of Alzheimer's disease (AD); however, the value of neuroinflammatory biomarkers for disease-staging or predicting disease progression remains unclear.ObjectiveTo investigate diagnostic and prognostic utility of inflammatory biomarkers in combination with conventional AD biomarkers.MethodsData from 258 participants in the Alzheimer's Disease Neuroimaging Initiative (ADNI) with cerebrospinal fluid (CSF) biomarkers of amyloid-β (Aβ), tau, and inflammation were analyzed. Clinically meaningful cognitive decline (CMCD) was defined as a ≥ 4-point increase on the Alzheimer's Disease Assessment Scale Cognitive Subscore 11. Predictor variables included demographics (D: age, sex, education), APOE4 status (A), inflammatory biomarkers (I), and classic AD biomarkers of Aβ and p-tau181 (C). Models incorporating inflammatory biomarkers assessed their contribution to improving baseline diagnostic classification and 1-year CMCD prediction.ResultsAt 1-year follow-up, 27.1% of participants experienced CMCD. Adding inflammatory biomarkers to models with D and A variables (DA model) improved classification of cognitively normal (CN) versus mild cognitive impairment (MCI) and CN versus Dementia (p < 0.001). Similarly, inflammatory markers enhanced classification in models including C (DAC model), for CN versus MCI (p < 0.01) and CN versus Dementia (p < 0.001). Predictive performance for CMCD was improved in individuals with MCI and dementia in both models (all p < 0.05). In addition, the DAI model outperformed the DAC model in predicting CMCD for MCI and Dementia groups (both p < 0.05).ConclusionsAddition of CSF inflammatory biomarkers to biomarkers of AD improves diagnostic accuracy of clinical disease stage at baseline and add incremental value to AD biomarkers for prediction of cognitive decline.

Blood tests that accurately determine the presence of amyloid pathology are critically needed. Compared to amyloid PET and CSF tests, blood tests are less expensive, less invasive, more accessible, and highly scalable. The Foundation for the National Institutes of Health (FNIH) Biomarkers Consortium evaluated the accuracies of leading AD blood tests in classifying amyloid PET status. Assays from C 2 N Diagnostics (Aβ42, Aβ40, p-tau217, p-tau217 ratio [phosphorylated to non-phosphorylated tau at position 217]), Fujirebio Diagnostics (Aβ42, Aβ40, p-tau217), ALZPath (p-tau217), Janssen (p-tau217), Roche Diagnostics (Aβ42, Aβ40, p-tau181, GFAP, NfL), and Quanterix (Aβ42, Aβ40, p-tau181, GFAP, NfL) were run on plasma samples from ADNI participants with corresponding florbetapir-PET scans. Unadjusted logistic regression models were constructed for each measure or combination of measures from each assay platform. The classification accuracies of the plasma biomarker measures for amyloid PET status (Centiloids>20) were assessed with receiver operating characteristic area under the curve (AUC) analyses and compared using DeLong’s tests. The cohort included data from 392 individuals with a median age of 78.1 years; 193 (49.2%) were female; 132 (33.7%) were APOE ε 4 carriers; 191 (48.7%) were amyloid PET positive; and 192 (49.0%) were cognitively impaired (Clinical Dementia Rating>0), ( Table 1 ). Models with high classification accuracies for amyloid PET status included the C 2 N p-tau217 ratio + Aβ42/Aβ40 (AUC 0.929 [95% confidence intervals 0.902-0.956]) and Fujirebio p-tau217 + Aβ42/Aβ40 (0.911 [0.882-0.940]), ( Table 2 ). The AlzPath p-tau217 (0.885 [0.851-0.920]), Janssen p-tau217 (0.882 [0.848-0.916]), Roche p-tau181 + Aβ42/Aβ40 + GFAP + NfL (0.873 [0.838-0.909]), and Quanterix p-tau181 + Aβ42/Aβ40 + GFAP + NfL (0.808 [0.763-0.854]) models had lower accuracies than the C 2 N p-tau217 ratio + Aβ42/Aβ40 model (p = 0.0006, 0.0006, 0.0006, and <0.0001, respectively), ( Figure 1 ). In 122 individuals with CSF Roche Elecsys p-tau181/Aβ42 data, the classification accuracy for amyloid PET status was 0.915 (0.864-0.967). In this head-to-head study, logistic regression models of C 2 N p-tau217 ratio + Aβ42/Aβ40 and Fujirebio p-tau217 + Aβ42/Aβ40 were among the most accurate in classifying amyloid PET status. The classification accuracies of the blood tests for tau PET status, brain atrophy status, and clinical dementia symptoms will also be assessed.

BACKGROUND:Time proxies based on blood biomarkers may facilitate understanding of the timing of biomarker change and symptom onset in Alzheimer's disease (AD) and could potentially enable biological staging of AD. Compared to CSF biomarkers or brain imaging, blood biomarkers are more accessible and less burdensome, making a "plasma clock" a practical and useful tool. METHOD:Longitudinal plasma %p-tau217 (p-tau217 concentration divided by np-tau217 concentration x 100) measurements from individuals enrolled in the Knight ADRC (n = 420) or ADNI (n = 392) cohorts were analyzed to develop plasma clocks for each cohort. The %p-tau217 values were transformed into a time scale by integrating the inverse of the modeled rate of change. A %p-tau217 value of 4.06%, which corresponds to amyloid PET Centiloid=20, was considered the threshold for positivity and set as time zero. The estimated age at %p-tau217 positivity was then used to predict symptom onset, defined as a change in Clinical Dementia Rating® (CDR) from CDR=0 to CDR>0. RESULT:The rate of change for plasma %p-tau217 was relatively consistent between 0.65-8.16% (Figure 1) and was used to construct independent plasma %p-tau217 clocks for the Knight ADRC and ADNI cohorts (Figure 2). For individuals with plasma %p-tau217 measured before and after positivity (4.06%), the estimated age at positivity based on the clock was strongly correlated with the actual age at conversion (Knight ADRC: R2=0.646, r=0.807; ADNI: R2=0.790, r=0.891). The Knight ADRC and ADNI plasma clocks provided highly similar estimates of the age at %p-tau217 positivity for individuals in either study (adjusted R2=0.993, Pearson r=0.997, Figure 2). The estimated age at plasma %p-tau217 positivity predicted the age at symptom onset (Figure 3) in the Knight ADRC (Spearman r=0.421, adjusted R2=0.151) and ADNI cohorts (Spearman r=0.648, adjusted R2=0.397). CONCLUSION:Two independently constructed %p-tau217 plasma clocks had strong correlations with actual time measures and were consistent across the two cohorts. Additionally, the age at %p-tau217 positivity predicted symptom onset. Overall, these findings suggest that plasma %p-tau217 may be useful in predicting symptom onset and in biological staging of AD.

INTRODUCTION:Alzheimer's disease (AD) has a long preclinical phase in which individuals may accumulate amyloid beta (Aβ) and tau pathology without noticeable cognitive impairment. Subjective cognitive impairment reports can provide early insights into cognitive decline. METHODS:In the A4 Study, 339 cognitively unimpaired, Aβ-positive individuals underwent tau positron emission tomography imaging. Tau status was classified based on medial temporal lobe tau standardized uptake value ratios (tauMTL). Participants and study partners assessed cognitive changes using the 15-item Cognitive Function Index (CFI) questionnaire. We explored the relationship among tauMTL, hippocampal volume (HVa), and CFI reports. RESULTS:Higher tauMTL was associated with participant-reported concerns about memory and navigation, and with study partner-reported difficulty remembering appointments. Lower HVa showed a marginal association with participant-reported driving difficulty. DISCUSSION:These findings support the utility of participant- and study partner-reported concerns as early indicators of preclinical AD pathology, with potential value for early detection and trial enrichment strategies. Highlights:Higher tau in the medial temporal lobe (tauMTL) was linked to participant-reported memory and orientation decline such as needing reminders or getting lost.Higher tauMTL was associated with increased memory-related concerns, such as needing help with appointments and asking repetitive questions.Lower hippocampal volume was associated with spatial memory and navigation such as driving difficulties and greater memory decline as reported by study partners.

Self-awareness of declining cognition in the absence of objective cognitive impairment is called subjective cognitive decline (SCD). Herein, we examine the association of SCD measured by the Cognitive Function Index (CFI; Amariglio, et al. 2015) with the Stages of Objective Memory Impairment (SOMI) system (Grober, et al. 2018) that classifies cognitively normal individuals into one of five stages based on the Free and Cued Selective Reminding Test (FCSRT; Grober & Buschke, 1987) and that maps onto the presence of AD pathology (Grober, et al. 2022 & Petersen, et al. 2023). In early SOMI stages (SOMI-1, 2) there is increasing retrieval difficulty; in later stages (SOMI-3/4) storage deficits emerge. We used baseline data from 4109 neuropsychologically normal, older adults from the Anti-Amyloid Anti-Amyloid Treatment in Asymptomatic Alzheimer’s (A4) Study. All participants had amyloid PET measures and completed the FCSRT used for determing SOMI classification. Additionally, all participants and study partners completed the CFI. An adjusted CFI score was used that excluded three items that assess functional rather than cognitive ability. CFI positivity was defined by cutoffs, 7.8 and 6.5 for participant and partner versions respectfully, representing one standard deviation above the mean. Analysis of Variance (ANOVA) with post hoc analysis was used to assess differences of CFI scores between SOMI stages. We found fewer A+ participants were SOMI-0 compared to A- participants (34.1% vs. 42.7%, p<0.001, Figure 1). Additionally, more A+ participants were SOMI-3/4 compared to A- participants (15.4% vs. 10.3%, p<0.001). Among A+ participants, memory impairment (SOMI-2/3/4) was more prevalent among those with self-perceived SCD (34.9%) than those without SCD (29.5%, p = 0.016). Similarly, among A+ participants, prevalence of memory impairment was higher among those with partner-perceived SCD (40%) than those without (28%, p<0.001). Post hoc results found higher SOMI stages were associated with higher partner and study partner CFI scores except between SOMI-2 and SOMI-3/4 (p<0.05 for all, Figure 2). Memory impairment was found to be more prevalent among A+ participants with self- and partner-rated SCD. SCD increases with increasing memory impairment.

Objectives The aim of this work is to use a machine learning framework to develop simple risk scores for predicting β-amyloid (Aβ) and tau positivity among individuals with mild cognitive impairment (MCI). Methods Data for 657 individuals with MCI from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) data set were used. A modified version of AutoScore, a machine learning-based software tool, was used to develop risk scores based on hierarchical combinations of predictor categories, including demographics, neuropsychological assessments, APOE4 status, and imaging biomarkers. Results The highest area under the receiver operating characteristic curve (AUC) for predicting Aβ positivity was 0.79, which was achieved by 2 separate models with predictors of age, Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-cog), APOE4 status, and either Trail Making Test Part B (TMT-B) or white matter hyperintensity. The best-performing model for tau positivity had an AUC of 0.91 using age, ADAS-13, and TMT-B scores, APOE4 information, abnormal hippocampal volume, and amyloid status as predictors. Discussion Simple integer-based risk scores using available data could be used for predicting Aβ and tau positivity in individuals with MCI. Models have the potential to improve clinical trials through improved screening of individuals.

Alzheimer’s disease (AD) is a heterogeneous disease with different clinical phenotypes and pathophysiological subtypes. Identifying cognitive/functional subtypes in AD could elucidate the diverse clinical progression patterns. The Cognitive Function Index (CFI), a 15-item questionnaire completed by participants and study partners, captures aspects of cognitive and functional decline. We investigated participant heterogeneity using the CFI and compared differences between subtypes based on participant or study partner reports. Participants were 4486 cognitively unimpaired older adults from the Anti-Amyloid Treatment in Asymptomatic Alzheimer’s (A4) Study. All participants and study partners completed the CFI. Participants had baseline demographics, neuropsychological test scores, and amyloid PET measures available. The Staging and Subtype Inference (SuStaIn) algorithm was independently applied to participant (PT) and study partner (SP) CFI reports to identify subtypes. Differences between subtypes were assessed using Analysis for Variance (ANOVA) with post hoc analysis. Table 1 summarizes participant characteristics. Analysis of PT and SP CFI responses independently identified three subtypes with distinct progression patterns. Notably, 35.2% and 57.5% of participants showed no progression based on PT and SP reports, respectively. All subtypes, except SP-2, had CFI items related to difficulties with recall, needing written reminders, and misplacing items as the first items of suspected decline (Figure 1). Clusters PT-1 (66.5%) and SP-1 (45.8%) were the subtypes with most frequent progression pattern (Figure 1). Subtypes PT-2 and SP-2 had the highest scores on Digit Symbol Substitution test, walked the most, and for SP-2 lowest Geriatric Depression Scores (p<0.05 for all; Figure 3). Subtypes PT-3 and SP-3 scored lowest on the participant Activities of Daily Living (ADL; p<0.05). Additionally, PT-3 scored lowest on the Study Partner ADL. We identified distinct subtypes from both participant and study partner CFI reports. Post hoc analysis indicated consistency within subtypes identified from the two sources of information. Subjective reports on cognitive decline could serve as a valuable tool for classifying participants into different risk groups.

OBJECTIVE:The objective of this study was to evaluate the timing of change of Alzheimer's disease (AD) plasma biomarkers (Aβ42/Aβ40, p-tau217, p-tau181, GFAP, and NfL) from six different assay platforms, alongside established AD biomarkers, using amyloid and tau positron emission tomography (PET)-based AD progression timelines. METHODS:Data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), including 784 individuals with longitudinal amyloid PET and 359 with longitudinal tau PET, were analyzed to estimate the age at amyloid and tau PET positivity, respectively. Longitudinal plasma biomarker measurements were available from 190 individuals with an estimated amyloid PET positivity age and from 70 individuals with an estimated tau PET positivity age. In a subset of 17 clinical progressors, age at tau PET positivity strongly predicted symptom onset, allowing for estimation of symptom onset age. Biomarker trajectories based on time from amyloid or tau PET positivity or symptom onset were modelled using Generalized Additive Mixed models. Time intervals of significant biomarker change and the earliest timepoints at which biomarkers exceeded predefined abnormality thresholds were identified. RESULTS:All plasma biomarkers except NfL became abnormal prior to established thresholds for amyloid and tau PET positivity. Plasma Aβ42/Aβ40 became abnormal very early in both amyloid PET and tau PET timelines, while plasma GFAP became abnormal early in the tau PET timeline. Plasma Aβ42/Aβ40 levels plateaued, whereas plasma p-tau217, p-tau181, GFAP, and NfL levels increased throughout the modeled disease progression. Some variations in the timing of these changes were observed across different biomarker assays. INTERPRETATION:These findings suggest that the plasma Aβ42/Aβ40 may be useful in identifying individuals with very low levels of amyloid pathology, whereas p-tau, GFAP, and NfL may be useful in staging disease progression. ANN NEUROL 2025;98:508-523.

OBJECTIVE:The Stages of Objective Memory Impairment (SOMI) system, based on the Free and Cued Selective Reminding Test (FCSRT), is a potential marker of subtle cognitive impairment in cognitively normal persons defined by a Clinical Dementia Rating (CDR) = 0. We investigated SOMI's ability to predict incident cognitive impairment (CDR >0) in combination with demographic features and neuroimaging biomarkers. METHODS:Cognitively unimpaired participants (CDR = 0) from the Harvard Aging Brain Study had baseline FCSRT scores, MRI, FDG-PET, and PiB-PET as well as follow-up CDRs for 5 years. Cox proportional hazards models with correction for multiple testing assessed the predictive validity of SOMI and neuroimaging biomarkers for progression (CDR >0). Comprehensive sensitivity analyses examined alternative outcomes and stricter screening criteria. RESULTS:Participants (N = 231) were 73.7 years (SD = 6.0), 60.2% were female, 29.0% were APOE4 positive, and 54 (23.4%) progressed to CDR >0. At baseline, 67% were SOMI-0, 22% were SOMI-1, 4% were SOMI-2, and 7% were SOMI-3/4. After multiple testing correction, hazard ratios (HRs) using SOMI-0 as reference were: SOMI-1 = 2.06 (CI: 1.09 - 3.88), SOMI-2 = 2.85 (CI: 1.08 - 7.54), and SOMI-3/4 = 3.73 (CI: 1.58 - 8.79, p = 0.016). SOMI-3/4 remained significant across most biomarker models. Entorhinal thickness emerged as the most robust biomarker predictor (HR = 0.57 - 0.65, p ≤ 0.015). Sensitivity analyses confirmed robustness across alternative outcomes and stricter screening criteria. CONCLUSIONS:SOMI stages predict progression to incident cognitive impairment with SOMI-3/4 maintaining significance after rigorous multiple testing correction. Entorhinal thickness provides the strongest biomarker enhancement to prediction models. SOMI demonstrates substantial incremental predictive value beyond standard demographic and biomarker predictors.

The Cognitive Function Index (CFI) is a validated test used to assess changes in self-perceived cognitive and functional status as reported by an individual and their study partner. Previous studies have demonstrated an inverse correlation between higher amyloid-beta (Aβ) burden and CFI, with certain CFI items exhibiting stronger associations than others. However, there is limited understanding of the association between declines in cognition and function, as assessed by CFI, and Tau levels measured by PET. Participants were 339 cognitively unimpaired, Aβ positive, individuals enrolled in the Anti-Amyloid Asymptomatic Alzheimer’s (A4) Study who underwent tau-PET imaging. Participants were classified as tau-PET positive (T+) or negative (T-) based on tau levels in the medial temporal lobe (T MTL + /T MTL - ). Participants and their study partners assessed subjective changes in cognition and function over the past year using a 15-item CFI questionnaire. For each CFI item, the relationship between Tau and CFI reports (Yes/Maybe vs No) was investigated using logistic regression models. Participants were on average 72.38(SD = 4.87) years old, 58.1% were female, and 23.6% was T MTL + . Table 1 summarizes the sample characteristics. Higher T MTL was significantly associated with participant report of decline on six items, seeing a doctor about memory concerns (OR = 1.746, p = 0.002) and getting lost while travelling to another city (OR = 1.557, p = 0.003) showing the highest odds ratios. By contrast, Higher T MTL was associated with study partner report of decline on only one item: needing help from others to remember appointments/occasions (OR = 1.592, p = 0.002). Unique CFI items were associated differentially with higher T MTL for participants and their study partners. The value of the CFI questionnaire with participant as the source of information may be higher than responses provided by informant in terms of the association with Tau-PET positivity. This is consistent with the idea that self-awareness of cognitive decline might precede the observations made by others.

AD is characterized by the sequential accumulation of amyloid and tau, but the interplay between these pathologies and other factors remains unclear. Using amyloid and tau PET clocks, we examined the temporal relationships between amyloid and tau positivity and potential modifying effects of sex and APOE ε4. We studied 757 ADNI participants with longitudinal amyloid (18F-Florbetapir) or tau (18F-Flortaucipir) PET imaging. Of these, 635 had at least one positive amyloid PET scan (SUVR>0.78), and 212 had at least one positive tau PET scan (mesial-temporal SUVR>1.41), enabling the estimation of age at amyloid or tau positivity (Table 1). Linear regression models assessed the effects of sex and APOE ε4 on the ages at amyloid and tau positivity. Ninety individuals had estimates of the age at positivity for both amyloid and tau, allowing estimation of the amyloid-tau time interval. Regression analyses examined the association between amyloid and tau positivity ages. Additional models included interaction terms to explore potential modifying effects of sex and APOE ε4. Men and APOE ε4 non-carriers had an older age at amyloid and tau positivity. For amyloid positivity age, there was a significant interaction between sex and APOE ε4 status, with sex modifying amyloid positivity age in APOE ε4 carriers. (Figure 1A, B). APOE ε4 status, but not sex, modified the association between age at amyloid and at tau positivity (Figure 1C, D). Most individuals (84.5%) became amyloid positive before tau positive and those who became tau positive first were more likely to be women (Figure 2A, B). The amyloid-tau interval was shorter in individuals with an older age at amyloid positivity (Figure 2C, D), and this association was more pronounced in APOE ε4 non-carriers (Figure 2C). Men had a longer amyloid-tau interval than women, regardless of the age at amyloid positivity (Figure 2D). Our findings suggest that a later age at amyloid positivity is associated with earlier tau positivity and this association differs by APOE ε4 carriership status and sex. These results may imply differences in tau-related disease progression and highlight the need to consider sex and APOE ε4 status in clinical trials aimed at slowing the onset of tau pathology.

INTRODUCTION:Understanding the heterogeneity of brain structure in individuals with the Motoric Cognitive Risk Syndrome (MCR) may improve the current risk assessments of dementia. METHODS:We used data from six cohorts from the MCR consortium (N = 1987). A weakly-supervised clustering algorithm called HYDRA (Heterogeneity through Discriminative Analysis) was applied to volumetric magnetic resonance imaging (MRI) measures to identify distinct subgroups in the population with gait speeds lower than one standard deviation (1SD) above mean. RESULTS:Three subgroups (Groups A, B, and C) were identified through MRI-based clustering with significant differences in regional brain volumes, gait speeds, and performance on Trail Making (Part-B) and Free and Cued Selective Reminding Tests. DISCUSSION:Based on structural MRI, our results reflect heterogeneity in the population with moderate and slow gait, including those with MCR. Such a data-driven approach could help pave new pathways toward dementia at-risk stratification and have implications for precision health for patients. Highlights:Different patterns of brain atrophy were observed among the people with moderate and slow gait speedsSlower gait speeds were associated with substantial cortical atrophy, higher rates of Motoric Cognitive Risk Syndrome (MCR), and worse cognitive performanceThis approach can aid patient stratification at early asymptomatic stages and have implications for precision health.

The Motoric Cognitive Risk Syndrome (MCR) is a predementia stage characterized by slow gait speed and subjective cognitive complaints. Defining the heterogeneity of brain volumetrics in individuals with MCR will improve current dementia risk assessments. We used data from 6 cohorts from the MCR consortium (N=2,007). We used K-means clustering algorithm guided by volumetric MRI to identify distinct subgroups of participants. We compared the differences in cortical and subcortical volumes, comorbidities, and gait speeds across the identified subgroups using one-way ANOVA and post-hoc pairwise group comparisons. The sample had a mean age of 71.89 (±7.05) years, 48.7% were women, 32.9% were White and 63.2% were Asian (see Table 1). Four subgroups (A to D) were identified through MRI-based clustering with significant differences in brain region volumes (Figure 1A), gait speeds (Figure 1B) and proportion of individuals with MCR (Figure 2). Subgroups A and C had the least amount of atrophy in all brain regions and had the least proportion of MCR. Subgroup D had the highest proportion of MCR subjects and highest atrophy specifically in hippocampus and cortical regions. The average gait speed of Subgroup D was lower than other subgroups. Subgroup D also had the highest rate of hypertension and diabetes among the subgroups (Figure 2). Our results validate the previous findings linking MCR syndrome to MRI evidence of neurodegeneration. Heterogeneity in cortical and subcortical signatures are present in older adults and provides insights into brain substrates of MCR.

Alzheimer’s disease (AD) is a complex heterogeneous neurodegenerative disease. Unsupervised clustering techniques have been used to identify disease subtypes, but such approaches are limited since subtypes may not directly be related to disease progression. Herein, we implement a novel supervised clustering approach that aims to identify MRI-derived subtypes that are likely to experience incident cognitive impairment (ICI). We used data from 822 cognitively normal individuals (68.4 ± 9.5 years old) with Clinical Dementia Rating® (CDR®) of 0 at baseline, from the Knight ADRC. We performed supervised clustering of eight volumetric MRI regional measures at baseline. MRI measures were adjusted for age, sex, and education. Mean aggregated time-dependent Shapley Additive exPlanation (SHAP) values were derived from random survival forest models with an ICI outcome based on time-to-conversion from a CDR=0 to CDR>0 over 10 years of follow-up. K-means clustering was then applied to SHAP values. Analysis of Variance was used to assess cluster differences and linear regression was used to understand longitudinal trajectories of MRI volumetrics and CSF biomarkers. In the entire sample, hippocampus, entorhinal, and amygdala SHAP values were determined to be the most important MRI measures in terms of estimated contribution to the model prediction with the hippocampus differentiating from other regional measure about 24 months from baseline (Figure 1). The optimal number of clusters was determined to be four. Cluster characteristics and significant differences among them are presented in Table 1 and Figure 1. Cluster 4 is characterized by the largest mean hippocampus SHAP value with approximately 89% of participants experiencing ICI. Figure 2 shows longitudinal linear trajectories for MRI and cerebrospinal measures. Cluster 1 was the healthiest across all longitudinal biomarkers. Whereas, Cluster 4 had the smallest hippocampal volume and abnormal CSF measures both at baseline and over time. We implemented a novel application of SHAP values from survival models in a supervised clustering approach to identifying MRI-subtypes with different probabilities of incident cognitive impairment. This approach could be used to elucidate the heterogeneity inherent in the relationships among AD biomarkers and longitudinal cognitive changes.

Alzheimer’s disease exhibits heterogeneity through varied phenotypic and pathological manifestations. Here, we aimed to investigate the potential of semi-supervised pattern classification applied to volumetric MRI data in identifying relatively homogeneous subgroups of individuals exhibiting cognitive decline (CD) throughout the study period. We used data from the placebo arm of trial of Solanezumab for mild dementia due to AD (EXPEDITION-3 trial). Participants were classified as showing CD if there was any increase in Clinical Dementia Rating (CDR, Sum of Boxes) from baseline to the 80 weeks of follow-up, and as stable cognition (SC) otherwise. We applied a non-linear, semi-supervised clustering algorithm, HYDRA, that captures subtypes within the decliner group accounting for the heterogeneous variations that are also present in the stable group. Two models with different feature-sets were developed: model I used only baseline volumetric MRI measures, while the model II included both baseline volumetric MRI measures and longitudinal changes observed after 80 weeks. Participants were 710 individuals with longitudinal MRI and CDR measures (mean age: 72.57[±7.69] years, 60.1% Female). Models I and II each produced two distinct subgroups of CD with notable differences in their baseline MRI volumes, and different rates of change in cognitive outcomes over the course of trial ( Table 1 ). In both the models, one of the subgroups (Slow Decliner) had significantly lower change in scores than the other (Fast Decliner) on the 11-item version of the Alzheimer’s Disease Assessment Scale–Cognitive subscale (ADAS-Cog) at the end of Week 80 ( Figure 1 ). In Model-I, both the decliner subgroups had smaller temporal and parietal lobe volumes at baseline compared to the Stable group ( Figure 2 ). In Model-II, the Slow Decliner group had higher baseline MRI volumes than the stable group and significantly lower change in ADAS-Cog compared to that of the Fast Decliner group ( Figure 2 ). Heterogeneity in brain atrophy is linked to distinct patterns of cognitive decline. Using longitudinal changes in MRI measures revealed decliner subgroups with greater difference in ADAS-cog scores compared to those obtained by baseline measures only.

BACKGROUND AND OBJECTIVES:Among the participants of Alzheimer disease (AD) treatment trials, 40% do not show cognitive decline over 80 weeks of follow-up. Identifying and excluding these individuals can increase power to detect treatment effects. We aimed to develop machine learning-based predictive models to identify persons unlikely to show decline on placebo treatment over 80 weeks. METHODS:We used the data from the placebo arm of EXPEDITION3 AD clinical trial and a subpopulation from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Participants in the EXPEDITION3 trial were patients with mild dementia and biomarker evidence of amyloid burden. For this study, participants were identified as those who demonstrated clinically meaningful cognitive decline (CMCD) or cognitively stable (CS) at final visit of the trial (week 80). Machine learning-based classifiers were trained to classify participants into CMCD vs CS groups using combinations of demographics, APOE genotype, neuropsychological tests, and biomarkers (volumetric MRI). The results were developed in 70% of the EXPEDITION3 placebo sample using 5-fold cross-validation. Trained models were then used to classify the participants in an internal validation sample and an external matched sample ADNIAD. RESULTS:Eight hundred ninety-four of the 1,072 participants in the placebo arm of the EXPEDITION3 trial had necessary follow-up data, who were on average aged 72.7 (±7.7) years and 59% female. 55.8% of those participants showed CMCD (∼2 years younger than those without) at the final visit. In the independent validation sample within the EXPEDITION3 data, all the models showed high sensitivity and modest specificity. Positive predictive values (PPVs) of models were at least 11% higher than base prevalence of CMCD observed at the end of the trial. The subset of matched ADNI participants (ADNIAD, N = 105) were aged 74.5 (±6.4) years and 46% female. The models that were validated in ADNIAD also showed high sensitivity, modest specificity, and PPVs of at least 15% higher than the base prevalence in ADNIAD. DISCUSSION:Our results indicate that predictive models have the potential to improve the design of AD trials through selective inclusion and exclusion criteria based on expected cognitive decline. Such predictive models need further validation across data from different AD clinical trials.