Citations

AbstractBackgroundDecreased image self‐similarity of repeated FDG‐PET scans has been suggested to indicate increased risk for disease progression. Here, we compared several self‐similarity measures based on conventionally used voxel‐ and ROI‐based approaches and those based on patterns of cortical metabolism.MethodFrom the MCSA/ADRC, we included 376 cognitively unimpaired (CU) individuals and those with a clinical diagnosis of MCI (n = 55) or AD (n = 20) who had two FDG‐PET images within 2‐3 years and clinical follow‐up available. We compared image self‐similarity for three approaches: voxel, ROI and “eigenbrain” (Fig.1). For the voxel approach, FDG‐PET images were normalized to template space, grey matter masked, and intensity normalized to the pons to create standard uptake value ratios (SUVR) for each voxel. From these images, mean SUVR were extracted for 122 ROIs (MCALT‐ADIR122) for the ROI approach. For the “eigenbrain” approach, the SUVR images were transformed to a latent space embedding composed of 253 components (“eigenbrains”) and reflecting metabolism patterns, that was generated in an independent sample of 7008 FDG‐PET images. For each approach, image self‐similarity between baseline and follow‐up values (voxel, ROI or eigenbrain) was determined using three metrics (Pearson correlation, cosine, Euclidean distance). We computed AUC to determine which approach and metric would best discriminate between groups. We used Cox regression analyses to assess whether the metrics could predict disease progression.ResultComparing each distance metric across approaches, AUC was highest in differentiating clinical diagnoses using self‐similarity measures based on eigenbrains, then voxels, then ROIs (e.g., Euclidean distance CU vs AD AUC eigenbrain: 0.86, voxel: 0.75, ROI: 0.71; Fig.1). Across all approaches, Euclidean distance always discriminated best between classes. Eigenbrain‐based Euclidean distance was associated with risk for progression to MCI in CU (HR [95%CI]: 1.25 [1.05, 1.49]), and eigenbrain‐based cosine and Euclidean distances were associated with risk for progression to AD in MCI (1.61 [1.1, 2.34] and 1.53 [1.01, 2.31], respectively), while we found no associations for voxel‐ or ROI‐based distance measures.ConclusionEigenbrain‐based measures outperformed other, more conventional approaches in discriminating clinical diagnoses, and were associated with risk for disease progression. Our results suggest that decreased self‐similarity in metabolism patterns informs on disease stage and prognosis.

AbstractBackgroundThe mapping between risk factors and biomarkers with cognitive performance and decline in the older adults is complex and varies significantly across different age and sex groups. Our goals were to (i) develop a machine learning (ML) model to predict baseline cognition and cognitive decline over a 5‐year follow‐up using only baseline assessments, and (ii) identify important modifiable and non‐modifiable predictors of cognitive decline across different age strata.MethodWe utilized longitudinal data from 2219 participants (mean age (SD): 70 (12) years, 49% male) from Mayo Clinic Study of Aging who had 43 features belonging to cardiovascular health, plasma, MRI (diffusion and structural), and lifestyle measures. Cognition was measured using a global cognition z‐score. We built ML models to predict baseline cognition and rate of cognitive decline over 5‐years (linear regression slopes over all individual followup measurements for each participant) using baseline assessments as inputs. We then built a model for each category separately and as a pooled model with all measures included, incorporating education and sex as demographic features in all models (Fig. 1). We applied a multi‐layer stack ensemble ML technique that utilized repeated k‐fold bagging to maximize accuracy. Importance of the top predictors were assessed in three age strata (55‐65 years (31%), 65‐80 years (46%), 80‐95 years (23%)).ResultThe pooled model explained the largest variance for prediction of baseline cognition and cognitive decline (Fig. 2). In individuals ≤80 years, lower NfL was most predictive of higher baseline cognition. A combination of imaging and plasma features predicted the 5‐year cognitive decline. Additionally, in individuals ≤65 years, diabetes and hypertension were important predictors of faster cognitive decline and in those >80, lower BMI and midlife physical/cognitive activity was an important predictor of faster cognitive decline (Fig. 3).ConclusionThe plasma model was best for baseline cognition prediction and imaging model was best for cognitive decline prediction. There was significant heterogeneity in important features across the age strata. In individuals >65 years, plasma GFAP was the most important predictor of cognitive decline. Our model allowed identification of key risk factors that predicted cognitive decline.

AbstractBackgroundThe implications of a positive tau (T) Positron Emission Tomography (PET) with a negative beta amyloid (A) PET are not well understood. We investigated cognitive performance in participants who were positive on tau PET but negative on beta amyloid PET.MethodWe evaluated 98 participants from the Mayo Clinic Study of Aging (MCSA) and the Mayo Clinic Alzheimer’s Disease Research Center (ADRC) who were positive for Flortaucipir (FTP) and negative for PiB PET. Participants were classified as A‐ and T+ based on Standardized Uptake Value Ratios (SUVr) of FTP and PiB PET scans, and were matched 2:1 to amyloid negative and tau negative cognitively unimpaired controls from the MCSA. Cognitive test scores were compared between the combined ADRC/MCSA group and MCSA only group. We also examined clinical diagnoses in our cohort. The most common were Alzheimer’s Disease, mild cognitive impairment (MCI), Dementia with Lewy Bodies (DLB) and Frontotemporal dementia (FTD) amongst others.ResultThere were no significant differences in demographic features or PiB SUVr between groups. There were significant differences in APOE genotype between groups. In the combined ADRC/MCSA group, 25% of A‐T‐ and 13% of A‐T+ were APOE e4 carriers (p = 0.016), and in the MCSA only group, 27% of A‐T‐and 10% of A‐T+ were APOE e4 carriers (p = 0.010). In the combined ADRC/MCSA group, the A‐T+ group demonstrated lower performance than the A‐T‐ group on the Mini Mental Status Exam (MMSE) (p<0.001), Wechsler Memory Scale‐Revised Logical Memory I (p<0.001) and Logical Memory II (p<0.001), Auditory verbal Learning Test Delayed Recall (AVLT) (p = 0.004), Category fluency animals (p = 0.005), Category fluency vegetables (p = 0.021), Trailmaking test A (p<0.001), Trail making test B (p<0.001), Wechsler Adult Intelligence Scale‐IV (WAIS‐IV) (p = 0.002), and Unified Parkinson’s Disease Rating Scale part III (UPDRS III) (p = 0.003). In the MCSA only group, the A‐T+ group had lower scores on Logical Memory II (p = 0.022), AVLT delated recall (p = 0.04), and WAIS‐R block design (p = 0.035) compared to the A‐T‐group.ConclusionPositive tau PET in the absence of beta‐amyloid positivity is associated with lower cognitive performance.

AbstractBackgroundQuantified accelerated aging, or “age gap,” defined as the difference between a person’s estimated age using a biological measurement, such as a brain image, and their true age, has strong associations with multiple neurodegenerative etiologies. As an alternative to imaging‐based measurements, in this study we developed an age prediction model based on gait patterns and used this model to quantify accelerated aging as an indicator of cognitive impairment and predictor of future cognitive decline.MethodA database of 16,917 pressure sensitive walkway exams from 5,307 participants in the Mayo Clinic Study of Aging was used to develop an age estimation model that included two main components. First, a novel dimensionality reduction and feature selection algorithm was used to extract gait features. Second, a regularized linear model was trained using these features to produce a sparse, clinically interpretable estimator. This model was used to calculate a gait age gap metric, which was tested for association with measures of cognitive impairment and future cognitive decline.ResultThe distributions of gait age gap in age‐matched participants defined as having no cognitive impairment (CU, n = 573), mild cognitive impairment (MCI, n = 520), or a clinical diagnosis of dementia due to Alzheimer’s (AD, n = 53) are shown in Fig 1a. A significant shift in the mean is observed between all three groups. Fig 1b. shows the distributions of baseline gait age gap values amongst individuals categorized by cognitive decline: stable CU (n = 644), MCI converters (n = 89), stable MCI (n = 78), and significant impairment converters (SI, n = 35). The baseline gait age gap in stable CU individuals was found to be significantly lower than in the already impaired or converter groups.ConclusionIn this study, we developed a model for estimating accelerated aging from gait variables in the form of a gait age gap metric. This metric was shown to be strongly associated with AD‐driven cognitive impairment as well as predictive of future cognitive decline. These results highlight the potential clinical utility of quantitative gait analysis in behavioral neurologic assessment of cognitive impairment in AD.

AbstractBackgroundPathogenic mutations in the MAPT (tau) gene cause familial frontotemporal lobar degeneration (f‐FTLD). No disease‐modifying treatments exist. Tau therapeutics being tested in other tauopathies may be suitable candidates, but the small number of identified mutation carriers prohibits traditional clinical trial designs from achieving adequate power. Platform trial designs permit multiple therapies to be tested in parallel and improve power by sharing placebo groups across treatment arms. We evaluate whether a platform design could improve the likelihood of identifying an effective treatment for MAPT mutation carriers.MethodALLFTD and GENFI observational data were used to simulate MAPT prevention and early symptomatic clinical trials (CDR®+NACC‐FTLD≤1). Only presymptomatic participants (CDR®+NACC‐FTLD = 0) within 2.5 years of expected symptom onset based on disease progression models (DPM) were included. First, we evaluated the time and sample size required to test five treatments using a platform design compared to sequential randomized control trials. Platform designs assumed a max of three treatment arms at any time, with new arms enrolling upon completion of prior arms. Each arm randomized 60 mutations carriers 3:1 treatment to placebo, with a placebo group shared across arms. Minimum follow up was 1.5 years with common close. Sequential trials assumed 90 participants allocated 1:1 treatment to placebo, with three‐month intervals between trials. Using a platform design, we compared a novel method for testing therapeutic effects using DPMs to account for differential rates of expected progression compared to traditional mixed models for repeated measures (MMRM) tests controlling for baseline CDR®+NACC‐FTLD.ResultFive treatments could be tested in a platform design using 300 participants (75% treated) over five years, compared to 450 participants (50% treated) over 13 years using traditional, sequential trials. Platform trial simulations indicated that with 80% power, DPM‐based analytic approach could detect a 25% treatment effect, whereas traditional MMRM could only detect a 40% effect.ConclusionPlatform designs combined with DPM‐based analyses provide a path to testing more treatments with fewer participants. Adequately powered MAPT platform trials are feasible but require a global clinical trial infrastructure.

AbstractBackgroundGiven the rarity of genetic frontotemporal dementia (FTD), researchers across the world have come together to form the FTD Prevention Initiative (FPI) in an effort to improve prevention trials design. As this initiative begins to bring together large‐scale data from worldwide cohort series, including ALLFTD in North America and GENFI in Europe and Canada, it is critical for FPI to quantify the level of relatedness between all participants. Here we provide the most recent update to these ongoing analyses.MethodsGenome‐wide SNP genotyping data from 1,684 ALLFTD and 568 GENFI participants was used to perform lineage analyses using PLINK. Briefly, QC was performed similarly in all datasets to remove individuals with low call rate and filter autosomal SNPs for missingness, frequency, and deviation from Hardy‐Weinberg equilibrium. Genetic ancestry was inferred by projecting genotyped samples into the principal components of the 1000 Genomes reference panel, using R package bigsnpr. Overlapping ALLFTD and GENFI genotyping data was then used, in a two‐stage approach, to calculate pairwise identity‐by‐descent (IBD) estimates and KING coefficients, followed by family‐network identification and pedigree reconstruction using PRIMUS.ResultsFirst, we calculated IBD estimates among all participants by restricting pairs to those with estimates>0.1875 (up to second‐degree relatives). Overall, we identified a total of 292 second‐degree family networks, including 168 ALLFTD and 120 GENFI families, mostly associated with pathogenic variants in the 3 major FTD‐causing genes. We also identified 4 family networks with participants enrolled in both the ALLFTD and GENFI series, as well as several multi‐site families within the ALLFTD consortium. This first, overall approach allowed us to predict close relationships even between individuals with different ancestral backgrounds, including at least 4 confirmed admixed families. More distant relationships were also detected within ALLFTD and GENFI by performing ancestry‐based analysis among participants with estimated European ancestry, using the KING‐robust algorithm.ConclusionsThese lineage analyses allowed us to identify, otherwise unknown, close (and distant) relatives from different study sites, as well as within the ALLFTD and GENFI series. This dataset will be a crucial resource to increase statistical accuracy and power in upcoming collaborative FPI studies.

AbstractBackgroundEasily accessible and self‐administered cognitive assessments that can aide early identification of individuals at risk for Alzheimer’s disease (AD) dementia are critical for timely intervention. Mayo Test Drive (MTD) is a remote, self‐administered, multi‐device compatible, web‐based cognitive assessment (Stricker et al., 2022). We investigated continuous associations between MTD and amyloid and tau PET biomarkers and compared the strength of association to PET biomarkers and in‐person cognitive measures.Method556 adults from the Mayo Clinic Study of Aging participated (mean age = 70.6; mean education = 16.0 years; 50.7% female; 97.5% white). Participants were predominantly cognitively unimpaired (95.3%; 4.7% had a consensus diagnosis of mild cognitive impairment). Participants completed (1) brain amyloid and tau PET scans, (2) in‐person neuropsychological assessment including Mayo Preclinical Alzheimer’s disease Cognitive Composite (Mayo‐PACC) and Global‐z, and (3) remote MTD comprised of the Stricker Learning Span (SLS) and the Symbols test, which are combined into an MTD composite. Multiple regressions adjusted for age, sex, and education queried associations between imaging biomarkers [amyloid meta‐region of interest (ROI), tau meta‐ROI, and tau entorhinal cortex (EC)‐ROI)] and composite and subtest scores from remote and in‐person measures. Davidson‐Mackinnon J‐test compared R‐squared values of models with remote versus in‐person composites.ResultLower performances on MTD composite and Mayo‐PACC were associated with higher amyloid meta‐ROI (p’s<.05; Table 1). MTD composite was also negatively associated with tau meta‐ROI and EC‐ROI (p’s<.05). For memory measures, SLS sum of trials was negatively associated with tau meta‐ROI (p = .01) and EC‐ROI (p<.001) with no associations between AVLT sum of trials and PET (p’s≥.06; Table 2). Among processing speed/executive function measures, worse Trails B performance was associated with higher amyloid meta‐ROI (p<.001). Symbols and Digit Symbol Coding were not associated with PET (p’s≥.29). A multivariable model with MTD explained significantly less variance (p = .04) for amyloid meta‐ROI (adjusted R‐squared = 0.165) than a multivariable model with Mayo‐PACC (adjusted R‐squared = 0.176); remote and in‐person composite measures were comparable for other associations (p’s>.05).ConclusionMTD is associated with continuous measures of amyloid and tau PET biomarkers of AD, demonstrating criterion validity for MTD. Several associations were comparable to relationships between gold‐standard in‐person neuropsychological tests and imaging biomarkers.

AbstractBackgroundBrief self‐administered cognitive assessments deployed remotely and combined with plasma biomarkers of Alzheimer’s disease (AD) are well suited to assess clinical symptoms and underlying biology in large‐scale decentralized studies. We investigated whether associations of self‐administered cognitive measures and plasma biomarkers were comparable to associations of in‐person cognitive measures and plasma biomarkers.MethodParticipants included 327 adults from the Mayo Clinic Study of Aging (mean age = 70.8; 48% female; mean education = 16.0 years; 96.9% white). Most (94.5%) were cognitively unimpaired; 5.5% had consensus diagnosis of mild cognitive impairment. Participants completed Mayo Test Drive (MTD; Stricker et al., 2022), a remote, self‐administered, multi‐device compatible web‐based cognitive assessment comprised of the Stricker Learning Span (SLS) and Symbols test that are combined into a MTD composite. Participants also completed in‐person neuropsychological assessment (global‐z; Mayo Preclinical Alzheimer’s disease Cognitive Composite [Mayo‐PACC]) and fasting venipuncture. Plasma was assayed via HD‐X simoa platform for amyloid‐beta (Aβ)‐42 and Aβ‐40 to create an Aβ42/40 ratio, glial fibrillary acidic protein (GFAP), and neurofilament light chain (NfL) using the N4PE Quanterix kit as well as phosphorylated‐tau protein 181 (p‐tau181). P‐tau181, GFAP, and NfL were log10‐transformed. Multiple regression analyses adjusting for age, sex, and education examined associations between plasma biomarkers and composite and subtest scores from remote and in‐person measures. Davidson‐Mackinnon J‐test compared R‐squared values of models with remote versus in‐person composites.ResultLower performances on MTD, global‐z, and Mayo‐PACC (Table 1) were associated with higher GFAP (p’s≤.02), NfL (p’s≤.003), and p‐tau181 (p’s = .04). For memory, worse SLS and AVLT sum of trials were associated with higher GFAP (p’s≤.01) and NfL (p’s≤.03). For processing speed/executive function, worse Symbols and Trails B were associated with higher p‐tau181 (p’s≤.02). Worse Trails B and Digit Symbol Coding were associated with higher NfL (p’s≤.02), and worse Digit Symbol Coding was associated with lower Aβ42/40 (p = 0.03; Table 2). A multivariable model with MTD explained significantly more variance (p = .04) for NfL (adjusted R‐squared = 0.461) than a multivariable model with Mayo‐PACC (adjusted R‐squared = 0.447); remote and in‐person composite measures were comparable for other associations (p’s>.05).ConclusionRemote and in‐person composite measures showed comparable associations with AD‐related plasma biomarkers, supporting MTD’s criterion validity.

AbstractBackgroundThe cardiac and respiratory cycles have been suggested as major drivers of cerebrospinal fluid(CSF) flow. The glymphatic system, one of the mechanisms for the clearance of amyloid‐beta and tau proteins, relies heavily on the movement(flow) of the CSF, and thus, investigating the forces that modulate the CSF flow is essential for understanding the glymphatic flux. Because breathing is trainable through exercise and self‐directed methods, various breathing methods may enhance the CSF flow.MethodReal‐time phase‐contrast MRI(pc‐MRI) was used to measure the CSF flow at the foramen magnum on a 3T. Fast resting‐state fMRI(rsfMRI;TR400ms) was conducted on a compact 3T. Under an IRB‐approved protocol, participants were instructed to breathe slowly and deeply, and subsequently switch to their spontaneous breathing. For rsfMRI data preprocessing, we performed motion correction, skull stripping, spatial smoothing(4mm), temporal filtering(0.01 to 0.1Hz), and linear and quadratic temporal trends removal. fMRI time courses were then normalized to Z‐score at each voxel before being averaged. On the pcMRI, the phase was summed in a CSF region‐of‐interest and plotted vs time. Additionally, we confirmed for possible confounding effects such as head motion or motion‐induced susceptibility changes.ResultrsfMRI findings showed that deep breathing promoted large oscillations in the CSF flow in the ventricles, while spontaneous breathing did not(Figure 1). Additionally, deep breathing resulted in large oscillations in lobar cortical regions. Similar results were observed in the pcMRI, with larger CSF velocity through the foramen magnum during the deep breathing compared to the spontaneous breathing(Figure 2A‐B). The cumulative summation result showed that the volume of CSF flow was increased during deep breathing(Figure 2C).ConclusionThis study provides new evidence that slow and deep breathing promotes CSF flow and global brain BOLD signal, which may be useful in promoting circulation in the brain. The study’s findings also open new avenues for research in the field of neurovascular coupling and mechanisms associated with glymphatic clearance. As hampered CSF flow and/or worsen glymphatic clearance have been shown to associate with pathological aging and dementia, including Alzheimer’s disease and Parkinson’s disease, future research is needed explore the potential of breathing training methods as a non‐pharmacological intervention to support brain health.

AbstractBackgroundThe language and visual variants are two common clinical phenotypes of atypical Alzheimer’s disease (AD). There are reports of overlap between these clinical phenotypes, however, the extent and frequency of overlap, and its neuroanatomical underpinnings remain unclear.MethodEighty‐two patients with biomarker‐confirmed AD who presented with either predominant language (n = 34) or visuospatial/perceptual (n = 48) deficits were recruited by the Neurodegenerative Research Group and underwent neurological and neuropsychological testing, MRI and [18F]flortaucipir‐PET. The clinical examination included testing for simultanagnosia, Gerstmann’s syndrome, ideomotor apraxia, visuospatial‐visuoperceptual function, confrontational naming, sentence repetition and letter and animal fluency. Cut‐points for abnormality were defined for each test and used to classify the following groups: language‐pure (no visual‐function deficit), language‐plus (impaired on one visual‐function), language‐mixed (impaired on two visual‐functions), visual‐pure (no language‐function deficit), visual‐plus (impaired on one language‐function) and visual‐mixed (impaired on two language‐functions). Patterns of relative volume loss and flortaucipir uptake were assessed using SPM12.ResultOf the 82 patients, 11 (13%) were language‐pure, 16 (20%) language‐plus, 7 (8%) language‐mixed, 10 (12%) visual‐pure, 17 (21%) visual‐plus and 21 (26%) visual‐mixed, with no difference in disease duration across groups. Visual deficits in the language‐plus group consisted of mild simultanagnosia in 29%, Gerstmann’s syndrome in 24%, and mild visuospatial deficits. In the language‐mixed group 86% had simultanagnosia and 43% had Gerstmann’s syndrome, with visuospatial‐visuoperceptual deficits comparable to the visual‐pure group. Language deficits in the visual‐plus group consisted of mild anomia and reduced letter and animal fluency. The visual‐mixed group reported naming and repetition deficits, and reduced fluency, comparable to the language‐predominant groups. All three language groups showed left temporal volume loss and flortaucipir uptake, with greater involvement of the parieto‐occipital and frontal lobes in language‐plus and language‐mixed groups. All three visual groups showed occipital volume loss and flortaucipir uptake, with greater involvement of the parieto‐temporal and frontal lobes in visual‐plus and visual‐mixed groups.ConclusionOnly 25% of our cohort showed a pure language or visual presentation, highlighting the high frequency of syndromic overlap in atypical AD and the diagnostic challenge of categorical phenotyping. Syndromic overlap was associated with neuroanatomical overlap and more widespread patterns of atrophy and tau deposition.