Citations

With increasing number of people living with dementia, the problem of late diagnosis significantly impacts a person’s quality of life while early signs of dementia may provide useful insights to facilitate better treatment plans. With time, this progressive neurodegenerative syndrome could progress from mild cognitive impairment to dementia. A pattern of health conditions can be characterized in unsupervised manner to help predict this progress. As a significant extension to our previous work with streaming clustering model, we consider additional information for predicting dementia onset. With empirical observations, we discover the importance of examining sex and age to predict dementia onset. To this end, we propose a sex-specific model with age-constraint for predicting dementia onset and validate the effectiveness of our models using data from Mayo Clinic Study of Aging (MCSA). The proposed sex-specific models for older adult populations (>=65 years of age) outperformed the previous models with F-score of 77% and 78% for male-specific and female-specific models, respectively. Our experiments of sex-specific temporal clustering of features in older adults demonstrate the potential of more personalized models for early alerts of dementia.

Abstract BACKGROUND We compared the ability of several plasma biomarkers versus amyloid positron emission tomography (PET) to predict rates of memory decline among cognitively unimpaired individuals. METHODS We studied 645 Mayo Clinic Study of Aging participants. Predictor variables were age, sex, education, apolipoprotein E ( APOE ) ε 4 genotype, amyloid PET, and plasma amyloid beta (Aβ)42/40, phosphorylated tau (p‐tau)181, neurofilament light (NfL), glial fibrillary acidic protein (GFAP), and p‐tau217. The outcome was a change in a memory composite measure. RESULTS All plasma biomarkers, except NfL, were associated with mean memory decline in models with individual biomarkers. However, amyloid PET and plasma p‐tau217, along with age, were key variables independently associated with mean memory decline in models combining all predictors. Confidence intervals were narrow for estimates of population mean prediction, but person‐level prediction intervals were wide. DISCUSSION Plasma p‐tau217 and amyloid PET provide useful information about predicting rates of future cognitive decline in cognitively unimpaired individuals at the population mean level, but not at the individual person level.

Genetic variation at the transmembrane protein 106B gene ( TMEM106B) has been linked to risk of frontotemporal lobar degeneration with TDP-43 inclusions (FTLD-TDP) through an unknown mechanism. We found that presence of the TMEM106B rs3173615 protective genotype was associated with longer survival after symptom onset in a postmortem FTLD-TDP cohort, suggesting a slower disease course. The seminal discovery that filaments derived from TMEM106B is a common feature in aging and, across a range of neurodegenerative disorders, suggests that genetic variants in TMEM106B could modulate disease risk and progression through modulating TMEM106B aggregation. To explore this possibility and assess the pathological relevance of TMEM106B accumulation, we generated a new antibody targeting the TMEM106B filament core sequence. Analysis of postmortem samples revealed that the TMEM106B rs3173615 risk allele was associated with higher TMEM106B core accumulation in patients with FTLD-TDP. In contrast, minimal TMEM106B core deposition was detected in carriers of the protective allele. Although the abundance of monomeric full-length TMEM106B was unchanged, carriers of the protective genotype exhibited an increase in dimeric full-length TMEM106B. Increased TMEM106B core deposition was also associated with enhanced TDP-43 dysfunction, and interactome data suggested a role for TMEM106B core filaments in impaired RNA transport, local translation, and endolysosomal function in FTLD-TDP. Overall, these findings suggest that prevention of TMEM106B core accumulation is central to the mechanism by which the TMEM106B protective haplotype reduces disease risk and slows progression.

Abstract Disruption of the default mode network is a hallmark of Alzheimer’s disease, which has not been extensively examined in atypical phenotypes. We investigated cross-sectional and 1-year longitudinal changes in default mode network sub-systems in the visual and language variants of Alzheimer’s disease, in relation to age and tau. Sixty-one amyloid-positive Alzheimer’s disease participants diagnosed with posterior cortical atrophy (n = 33) or logopenic progressive aphasia (n = 28) underwent structural MRI, resting-state functional MRI and [18F]flortaucipir PET. One-hundred and twenty-two amyloid-negative cognitively unimpaired individuals and 60 amyloid-positive individuals diagnosed with amnestic Alzheimer’s disease were included as controls and as a comparison group, respectively, and had structural and resting-state functional MRI. Forty-one atypical Alzheimer’s disease participants, 26 amnestic Alzheimer’s disease participants and 40 cognitively unimpaired individuals had one follow-up functional MRI ∼1–2 years after the baseline scan. Default mode network connectivity was calculated using the dual regression method for posterior, ventral, anterior ventral and anterior dorsal sub-systems derived from independent component analysis. A global measure of default mode network connectivity, the network failure quotient, was also calculated. Linear mixed-effects models and voxel-based analyses were computed for each connectivity measure. Both atypical and amnestic Alzheimer’s disease participants had lower cross-sectional posterior and ventral and higher anterior dorsal connectivity and network failure quotient relative to cognitively unimpaired individuals. Age had opposite effects on connectivity in Alzheimer’s disease participants and cognitively unimpaired individuals. While connectivity declined with age in cognitively unimpaired individuals, younger Alzheimer’s disease participants had lower connectivity than the older ones, particularly in the ventral default mode network. Greater baseline tau-PET uptake was associated with lower ventral and anterior ventral default mode network connectivity in atypical Alzheimer’s disease. Connectivity in the ventral default mode network declined over time in atypical Alzheimer’s disease, particularly in older participants, with lower tau burden. Voxel-based analyses validated the findings of higher anterior dorsal default mode network connectivity, lower posterior and ventral default mode network connectivity and decline in ventral default mode network connectivity over time in atypical Alzheimer’s disease. Visuospatial symptoms were associated with default mode network connectivity disruption. In summary, default mode connectivity disruption was similar between atypical and amnestic Alzheimer’s disease variants, and discriminated Alzheimer’s disease from cognitively unimpaired individuals, with decreased posterior and increased anterior connectivity and with disruption more pronounced in younger participants. The ventral default mode network declined over time in atypical Alzheimer’s disease, suggesting a shift in default mode network connectivity likely related to tau pathology.

AbstractINTRODUCTIONThe implications of positive tau positron emission tomography (T) with negative beta amyloid positron emission tomography (A) are not well understood. We investigated cognitive performance in participants who were T+ but A‐.METHODSWe evaluated 98 participants from the Mayo Clinic who were T+ and A‐. Participants were matched 2:1 to A‐ and T‐ cognitively unimpaired (CU) controls. Cognitive test scores were compared between different groups.RESULTSThe 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 (AVLT) delayed recall (p = 0.004), category fluency (animals p = 0.005; vegetables p = 0.021), Trail Making Test A and B (p < 0.001), and others. There were no significant differences in demographic features or apolipoprotein E (APOE) e4 genotype between CU A‐T+ and CI A‐T+.DISCUSSIONA‐T+ participants show an association with lower cognitive performance.

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.