Abstract Body

Aims

The FINGER trial is a Finnish two-year intervention program to prevent cognitive decline by improving several risk factors simultaneously (nutrition, physical exercise, cognitive training, metabolic and vascular factors). The aim of this study was to investigate 2-year longitudinal changes in cognitive decline and its association with blood metabolites using data from the trial, to shed light on related metabolic pathways.

Methods

We acquired metabolomics data of blood plasma at two time points (baseline and follow-up after 2 years) for a subset of 1089 participants (547 intervention and 542 control). LC-MS data consists of 10,825 metabolomic features, with 6000 currently putatively annotated. 1H NMR was used to quantify 40 small molecules and 112 lipoprotein fractions. Linear mixed models were used to evaluate the influence of metabolite levels at baseline on the cognition at 2-years follow-up.

Results

We found a robust pattern of association between baseline plasma metabolites and increased levels of cognition after two years. This association was positive for carotenoids and omega-3 fatty acid side chains across multiple lipid classes (fatty acids, phosphatidylcholines, phosphatidylethanolamines, acyl carnitines, triglycerides, among others), while negative for arachidonic acid moieties as well as saturated acyl carnitines and ceramides.

Conclusions

Using large-scale comprehensive metabolomic platforms and cognition tests from the FINGER trial, our findings suggest that inflammation-related pathways play a role in cognitive decline. We found a pro/anti inflammation-related metabolite profile in baseline plasma samples predicting the decrease/increase of a global cognition score in the participants after two years.

Aims

The Oral Glucose Tolerance Test (OGTT) is widely used in clinical practice for identifying glucose metabolism disorders. However, few longitudinal studies have explored the association between OGTT markers and both cognitive and brain changes over time. This study investigated OGTT and insulin resistance markers in association with cognitive and neuroimaging changes over 2 years in the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER).

Methods

1.025 FINGER participants without previously diagnosed diabetes underwent the OGTT. Brain MRI scans were available for 132 participants, and PiB-PET and FDG-PET scans for 47 participants. Cognition was assessed using a composite Z-score (modified Neuropsychological Test Battery, mNTB). Neuroimaging measures included global GM volume, hippocampal volume, WML volume, cortical thickness in AD-signature areas and FDG-PET and PiB-PET composite scores. Mixed effect regression models were used for cognition and linear regressions for neuroimaging analysis.

Results

Higher baseline area-under-curve (AUC)-OGTT (estimate [95%CI] -0.02 [-0.03–-0.00]; p<0.001) and 2h-post-load plasma glucose (-0.02 [-0.03–-0.01]; p<0.001) were associated with cognitive worsening in mNTB total score. Similar associations were found with memory (-0.03 [-0.05–-0.01]; p<0.005) and processing speed (-0.02 [-0.04–-0.01]; p<0.01). Baseline AUC-OGTT (standardised β coefficient=-0.28; p=0.01), 2h-OGTT (β=-0.26; p=0.03) and fasting plasma glucose (β=-0.26; p=0.03) were inversely linked to 2-year change in hippocampal volume. Higher glucose-AUC was associated with more decline in FDG-PET composite (β=-0.38; p=0.03). Serum fasting insulin and HOMA-IR were not associated with cognition and neuroimaging measures.

Conclusions

In older individuals at-risk for dementia, baseline OGTT measures were associated with greater cognitive decline and reduced hippocampal volume, irrespective of randomisation group and adjustments for abnormal glucose metabolism. This findings suggest that OGTT may be more sensitive in detecting glucose metabolism abnormalities impacting long-term cognition and brain structure.

Abstract Body

Empirical evidence suggests a declining incidence of dementia in higher-income countries, attributed to healthier lifestyles. This trend underscores the potential for dementia prevention by implementing personalized strategies to counter its rising prevalence due to increasing life expectancy. The European task force for dementia brain health services (dBHS) drafted recommendations for the implementation of dBHS. These are a new health offer devoted to evidence-based and ethical dementia prevention in at-risk individuals. The initiative falls under the scope of the Brain Health Plan of the European Academy of Neurology.

The foundational interventions of dBHS encompass the four key activities of risk assessment, risk communication, risk reduction, and cognitive enhancement. A few of pilot dBHS have been set up in Europe in the past three years, with variable structure and health offers but similar aim and mission. Risk reduction interventions can be categorised into:

- Vascular Risk Reduction Programs, focusing on controlling vascular risk factors, comprising lifestyle elements such as the Mediterranean diet, physical exercise and cognitive training;

- Alzheimer’s Risk Reduction Programs, focusing on innovative approaches such as transcranial alternating current stimulation, multisensory stimulation, probiotics, melatonin, the LipiDiDiet regimen, and cognitive training.

An overview will be given of the programs and their early results.

Aims

We aim to predict conversion from subjective cognitive decline (SCD) to mild cognitive impairment (MCI) and Alzheimer’s Disease (AD) by means of a machine learning (ML) algorithm, trained on features derived from demographic, neuropsychological and genetic assessment.

Methods

We included 112 patients who met the criteria for SCD and were followed up for at least 10 years in cases where AD did not develop. We excluded patients with other causes possibly associated with cognitive decline. All the patients underwent clinical and neuropsychological examination (NPS) and APOE genotype analysis at baseline. Follow-up NPS were conducted every two years for all patients. We considered six demographic, 24 neuropsychological, and one genetic variable as input features to train a random forest algorithm with leave-one-out cross-validation.

Results

During the follow-up period, 63 (56.2%) patients progressed to MCI and were classified as p-SCD. Among these, 22 (19.6%) further progressed to AD (14.7%) and were classified as c-SCD. Forty-seven (42.0%) patients did not experience objective cognitive decline and were classified as stable SCD (s-SCD). Two patients (one who had a stroke and another who was diagnosed with amyloid angiopathy) were excluded from the analysis. Regarding the progression from SCD to MCI, the ML algorithm accuracy was 77.6%, correctly classifying 80.9% of p-SCD patients and 68.1% of s-SCD patients. Concerning the progression from SCD to AD, the ML algorithm accuracy was 80.8%, correctly classifying 72.7% of the 22 c-SCD patients and 84.1% of s-SCD patients. The importance of each feature is reported in Figure 1.

Conclusions

This is the first study that demonstrated that ML algorithms based on easily accessible features are a promising tool to predict AD in patients with SCD.

Aims

Self-reported measures have emerged as potential indicators for early detection and prediction of dementia and mortality. We aimed to investigate the predictive value of different self-reported measures, including subjective cognitive decline (SCD) based on memory complaints and general cognitive abilities complaints, subjective olfactory impairment (SOI), subjective taste impairment (STI), and self-reported general health, in order to determine the risk of progressing to Alzheimer’s Dementia (AD), Parkinson’s Dementia (PD) or any-other-cause dementia

Methods

A total of 6028 cognitively unimpaired individuals from the 8^thwave of the English Longitudinal Study of Ageing (ELSA) as baseline sample and 5297 individuals from the 9^thwave as two-year follow-up sample were included in this study. Self-rated complaints were assessed using different questions from the ELSA structured interview. Three logistic regresion models were fitted to predict different dementia outcomes.

Results

SCD based on memory complaints (Exp(B) = 11.145; p < 0.001) and older age (Exp(B) = 1.108, p < 0.001) significantly predicted the progresion to AD dementia at two-year follow-up. SOI (Exp(B) = 7.440; p <0.001) and older age (Exp(B) = 1.065, p = 0.035) significantly predicted the progression to PD dementia at two-year follow-up. SCD based on memory complaints (Exp(B) = 4.448; p < 0.001) jointly with non-memory complaints (Exp(B) = 6.662; p < 0.001) and older age (Exp(B) = 1.147, p < 0.001) significantly predicted the progression to dementia of any other cause (i.e. non-AD, non-PD).

Conclusions

Our study demonstrates that different self-reported measures are a usefull and cost-efficient tool when screening for individuals at risk of dementia in the general population, being specifically associated with different dementia outcomes. Clinical assessments may enrich by the inclusion of self-reported measures questionaires, covering amnestic and non-amnestic cognitive complaints and olfactory complaints.

Aims

The advancement of screening tools accessible to the general population for the early detection of Alzheimer’s disease (AD) and the prediction of its progression is of significant interest in facilitating timely therapeutic interventions and conducting decentralized clinical trials. This study delves into the application of Machine Learning (ML) techniques by leveraging paralinguistic features extracted directly from a brief spontaneous speech (SS) protocol. Firstly, we explored the capability of ML techniques to differentiate various AD stages based on SS. Secondly, and for the first time, we examined the relationship between SS and a broad and standardized neuropsychological battery.

Methods

We extracted physical-acoustic features from voice recordings of patients from a Memory Unit (n = 1,500) while undergoing a SS protocol. Based on this information, we implemented several ML models evaluated via cross-validation to distinguish different AD stages. In addition, we established models capable of predicting cognitive domain performance from SS-derived information, using the neuropsychological battery from Fundació ACE (NBACE) as reference.

Results

Our findings showed that, based on a paralinguistic analysis of the sound, it is possible to identify individuals with dementia caused by AD (ADD) (F1-score = 0.92) and patients with mild cognitive impairment (MCI) (F1-score = 0.84). Furthermore, our models predicted cognitive performance from SS data, with correlations exceeding 0.5 relative to actual values across domains such as attention, memory, executive functions, language, and visuospatial functions.

Conclusions

Here, we show the potential of a fast and cost-effective protocol to infer changes in different cognitive functions within the AD continuum. Our results demonstrate a high correspondence with protocols traditionally used to assess cognitive function. Overall, it opens up novel prospects for developing screening tools and remote disease monitoring.

Aims

Multidomain interventions that comprehensively manage modifiable risk factors of dementia should be adjusted to specific geographical and cultural contexts and their efficacy tested. We aimed to investigate the effects of a multidomain intervention on cognition in mild cognitive impairment (MCI).

Methods

In a multicenter, outcome assessor-blind, randomized controlled trial, participants with mild cognitive impairment (MCI) and with one or more modifiable dementia risk factors, aged 60-85 years, were randomly assigned in a 1:1 ratio to the multidomain intervention (MI) group or the control group receiving general health advice. The 24-week intervention comprised vascular risk management, cognitive training, social activity, physical exercise, nutrition guidance, and motivational enhancement using Tablet PC app. Face-to-face intervention was conducted at a facility once every 1-2 weeks, and intervention via Zoom was conducted 2-3 times a week. The primary end point was the change from baseline at 24 weeks in total scale index score of Repeatable Battery for the Assessment of Neuropsychological Status (RBANS).

Results

Three hundred participants were randomly assigned to the MI (n=148) and control (n=152) groups. In the MI and control group, the retention rates were 87.8% and 80.9%, respectively. The total adherence to the intervention was 87.5% and the adherence in each intervention domain was greater than 85%. The adjusted least-squares mean change from baseline at 24 weeks was 8.50 in the MI group and 4.18 in the control group (difference, 4.17; 95% confidence interval [CI], 1.92 to 6.42; P<0.001). Compared to the control group, depression, quality of Life, dietary habits evaluated by Nutrition Quotient for the Elderly, physical performance, and motivation were significantly improved in the MI group.

Conclusions

Multidomain interventions to modifiable risk factors with high adherence could improve cognitive function in MCI.