Progress in dementia research has been limited, with substantial gaps in our knowledge of targets for prevention, mechanisms for disease progression, and disease-modifying treatments. The growing availability of multimodal datasets opens possibilities for the application of machine learning and artificial intelligence (AI) to help answer key questions in the field. We provide an overview of the state of the science, highlighting current challenges and opportunities for utilisation of AI approaches to move the field forward in the areas of genetics, experimental medicine, drug discovery and trials optimisation, imaging, and prevention. Machine learning methods can enhance results of genetic studies, help determine biological effects and facilitate the identification of drug targets based on genetic and transcriptomic information. The use of unsupervised learning for understanding disease mechanisms for drug discovery is promising, while analysis of multimodal datasets to characterise and quantify disease severity and subtype are also beginning to contribute to optimisation of clinical trial recruitment. Data-driven experimental medicine is needed to analyse data across modalities and develop novel algorithms to translate insights from animal models to human disease biology. AI methods in neuroimaging outperform traditional approaches for diagnostic classification and although challenges around validation and translation remain, there is optimism for their meaningful integration to clinical practice in the near future. AI-based models can also clarify our understanding of the causality and commonality of dementia risk factors, informing and improving risk prediction models along with the development of preventative interventions. The complexity and heterogeneity of dementia requires an alternative approach beyond traditional design and analytical approaches. Although not yet widely used in dementia research, machine learning and AI have the potential to unlock current challenges and advance precision dementia medicine.

With the increase in large multimodal cohorts and high-throughput technologies, the potential for discovering novel biomarkers is no longer limited by data set size. Artificial intelligence (AI) and machine learning approaches have been developed to detect novel biomarkers and interactions in complex data sets. We discuss exemplar uses and evaluate current applications and limitations of AI to discover novel biomarkers. Remaining challenges include a lack of diversity in the data sets available, the sheer complexity of investigating interactions, the invasiveness and cost of some biomarkers, and poor reporting in some studies. Overcoming these challenges will involve collecting data from underrepresented populations, developing more powerful AI approaches, validating the use of noninvasive biomarkers, and adhering to reporting guidelines. By harnessing rich multimodal data through AI approaches and international collaborative innovation, we are well positioned to identify clinically useful biomarkers that are accurate, generalizable, unbiased, and acceptable in clinical practice. HIGHLIGHTS: Artificial intelligence and machine learning approaches may accelerate dementia biomarker discovery. Remaining challenges include data set suitability due to size and bias in cohort selection. Multimodal data, diverse data sets, improved machine learning approaches, real-world validation, and interdisciplinary collaboration are required.

Drug discovery and clinical trial design for dementia have historically been challenging. In part these challenges have arisen from patient heterogeneity, length of disease course, and the tractability of a target for the brain. Applying big data analytics and machine learning tools for drug discovery and utilizing them to inform successful clinical trial design has the potential to accelerate progress. Opportunities arise at multiple stages in the therapy pipeline and the growing availability of large medical data sets opens possibilities for big data analyses to answer key questions in clinical and therapeutic challenges. However, before this goal is reached, several challenges need to be overcome and only a multi-disciplinary approach can promote data-driven decision-making to its full potential. Herein we review the current state of machine learning applications to clinical trial design and drug discovery, while presenting opportunities and recommendations that can break down the barriers to implementation.

Genetics and omics studies of Alzheimer's disease and other dementia subtypes enhance our understanding of underlying mechanisms and pathways that can be targeted. We identified key remaining challenges: First, can we enhance genetic studies to address missing heritability? can we identify reproducible omics signatures that differentiate between dementia subtypes? can high-dimensional omics data identify improved biomarkers? How can genetics inform our understanding of causal status of dementia risk factors? and which biological processes are altered by dementia-related genetic variation? Artificial intelligence (AI) and machine learning approaches give us powerful new tools in helping us to tackle these challenges, and we review possible solutions and examples of best practice. However, their limitations also need to be considered, as well as the need for coordinated multidisciplinary research and diverse deeply phenotyped cohorts. Ultimately AI approaches improve our ability to interrogate genetics and omics data for precision dementia medicine. Highlights: We have identified five key challenges in dementia genetics and omics studies. AI can enable detection of undiscovered patterns in dementia genetics and omics data. Enhanced and more diverse genetics and omics datasets are still needed. Multidisciplinary collaborative efforts using AI can boost dementia research.

INTRODUCTION: the use of applied modeling in dementia risk prediction, diagnosis, and prognostics will have substantial public health benefits, particularly as "deep phenotyping" cohorts with multi-omics health data become available. METHODS: This narrative review synthesizes understanding of applied models and digital health technologies, in terms of dementia risk prediction, diagnostic discrimination, prognosis, and progression. Machine learning approaches show evidence of improved predictive power compared to standard clinical risk scores in predicting dementia, and the potential to decompose large numbers of variables into relatively few critical predictors. RESULTS: This review focuses on key areas of emerging promise including: emphasis on easier, more transparent data sharing and cohort access; integration of high-throughput biomarker and electronic health record data into modeling; and progressing beyond the primary prediction of dementia to secondary outcomes, for example, treatment response and physical health. DISCUSSION: Such approaches will benefit also from improvements in remote data measurement, whether cognitive (e.g. online), or naturalistic (e.g. watch-based accelerometry).

INTRODUCTION: Artificial intelligence (AI) and neuroimaging offer new opportunities for diagnosis and prognosis of dementia. METHODS: We systematically reviewed studies reporting AI for neuroimaging in diagnosis and/or prognosis of cognitive neurodegenerative diseases. RESULTS: a total of 255 studies were identified. Most studies relied on the Alzheimer's Disease Neuroimaging Initiative dataset. Algorithmic classifiers were the most commonly used AI method (48%) and discriminative models performed best for differentiating Alzheimer's disease from controls. The accuracy of algorithms varied with the patient cohort, imaging modalities, and stratifiers used. Few studies performed validation in an independent cohort. DISCUSSION: the literature has several methodological limitations including lack of sufficient algorithm development descriptions and standard definitions. We make recommendations to improve model validation including addressing key clinical questions, providing sufficient description of AI methods and validating findings in independent datasets. Collaborative approaches between experts in AI and medicine will help achieve the promising potential of AI tools in practice. HIGHLIGHTS: There has been a rapid expansion in the use of machine learning for diagnosis and prognosis in neurodegenerative disease Most studies (71%) relied on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset with no other individual dataset used more than five times There has been a recent rise in the use of more complex discriminative models (e.g. neural networks) that performed better than other classifiers for classification of AD vs healthy controls We make recommendations to address methodological considerations, addressing key clinical questions, and validation We also make recommendations for the field more broadly to standardize outcome measures, address gaps in the literature, and monitor sources of bias.

INTRODUCTION: Experimental models are essential tools in neurodegenerative disease research. However, the translation of insights and drugs discovered in model systems has proven immensely challenging, marred by high failure rates in human clinical trials. METHODS: Here we review the application of artificial intelligence (AI) and machine learning (ML) in experimental medicine for dementia research. RESULTS: Considering the specific challenges of reproducibility and translation between other species or model systems and human biology in preclinical dementia research, we highlight best practices and resources that can be leveraged to quantify and evaluate translatability. We then evaluate how AI and ML approaches could be applied to enhance both cross-model reproducibility and translation to human biology, while sustaining biological interpretability. DISCUSSION: AI and ML approaches in experimental medicine remain in their infancy. However, they have great potential to strengthen preclinical research and translation if based upon adequate, robust, and reproducible experimental data. HIGHLIGHTS: There are increasing applications of AI in experimental medicine. We identified issues in reproducibility, cross-species translation, and data curation in the field. Our review highlights data resources and AI approaches as solutions. Multi-omics analysis with AI offers exciting future possibilities in drug discovery.

BACKGROUND: Observational studies have found an association between attention-deficit/hyperactivity disorder (ADHD) and ischemic stroke. AIMS: the purpose of this study was to investigate whether genetic liability to ADHD has a causal effect on ischemic stroke and its subtypes. METHODS: in this two-sample Mendelian randomization (MR) study, genetic variants (nine single-nucleotide polymorphisms; P 

BACKGROUND: the identification of effective dementia prevention strategies is a major public health priority, due to the enormous and growing societal cost of this condition. Consumption of a Mediterranean diet (MedDiet) has been proposed to reduce dementia risk. However, current evidence is inconclusive and is typically derived from small cohorts with limited dementia cases. Additionally, few studies have explored the interaction between diet and genetic risk of dementia. METHODS: We used Cox proportional hazard regression models to explore the associations between MedDiet adherence, defined using two different scores (Mediterranean Diet Adherence Screener [MEDAS] continuous and Mediterranean diet Pyramid [PYRAMID] scores), and incident all-cause dementia risk in 60,298 participants from UK Biobank, followed for an average 9.1 years. The interaction between diet and polygenic risk for dementia was also tested. RESULTS: Higher MedDiet adherence was associated with lower dementia risk (MEDAS continuous: HR = 0.77, 95% CI = 0.65-0.91; PYRAMID: HR = 0.86, 95% CI = 0.73-1.02 for highest versus lowest tertiles). There was no significant interaction between MedDiet adherence defined by the MEDAS continuous and PYRAMID scores and polygenic risk for dementia. CONCLUSIONS: Higher adherence to a MedDiet was associated with lower dementia risk, independent of genetic risk, underlining the importance of diet in dementia prevention interventions.

BACKGROUND: Individual cardiometabolic disorders and genetic factors are associated with an increased dementia risk; however, the relationship between dementia and cardiometabolic multimorbidity is unclear. We investigated whether cardiometabolic multimorbidity increases the risk of dementia, regardless of genetic risk, and examined for associated brain structural changes. METHODS: We examined health and genetic data from 203 038 UK Biobank participants of European ancestry, aged 60 years or older without dementia at baseline assessment (2006-10) and followed up until March 31, 2021, in England and Scotland and Feb 28, 2018, in Wales, as well as brain structural data in a nested imaging subsample of 12 236 participants. A cardiometabolic multimorbidity index comprising stroke, diabetes, and myocardial infarction (one point for each), and a polygenic risk score for dementia (with low, intermediate, and high risk groups) were calculated for each participant. The main outcome measures were incident all-cause dementia and brain structural metrics. FINDINGS: the dementia risk associated with high cardiometabolic multimorbidity was three times greater than that associated with high genetic risk (hazard ratio [HR] 5·55, 95% CI 3·39-9·08, p

Sporadic Creutzfeldt-Jakob disease, the most common human prion disease, typically presents as a rapidly progressive dementia and has a highly variable prognosis. Despite this heterogeneity, clinicians need to give timely advice on likely prognosis and care needs. No prognostic models have been developed that predict survival or time to increased care status from the point of diagnosis. We aimed to develop clinically useful prognostic models with data from a large prospective observational cohort study. Five hundred and thirty-seven patients were visited by mobile teams of doctors and nurses from the National Health Service National Prion Clinic within 5 days of notification of a suspected diagnosis of sporadic Creutzfeldt-Jakob disease, enrolled to the study between October 2008 and March 2020, and followed up until November 2020. Prediction of survival over 10-, 30- and 100-day periods was the main outcome. Escalation of care status over the same time periods was a secondary outcome for a subsample of 113 patients with low care status at initial assessment. Two hundred and eighty (52.1%) patients were female and the median age was 67.2 (interquartile range 10.5) years. Median survival from initial assessment was 24 days (range 0-1633); 414 patients died within 100 days (77%). Ten variables were included in the final prediction models: sex; days since symptom onset; baseline care status; PRNP codon 129 genotype; Medical Research Council Prion Disease Rating Scale, Motor and Cognitive Examination Scales; count of MRI abnormalities; Mini-Mental State Examination score and categorical disease phenotype. The strongest predictor was PRNP codon 129 genotype (odds ratio 6.65 for methionine homozygous compared with methionine-valine heterozygous; 95% confidence interval 3.02-14.68 for 30-day mortality). of 113 patients with lower care status at initial assessment, 88 (78%) had escalated care status within 100 days, with a median of 35 days. Area under the curve for models predicting outcomes within 10, 30 and 100 days was 0.94, 0.92 and 0.91 for survival, and 0.87, 0.87 and 0.95 for care status escalation, respectively. Models without PRNP codon 129 genotype, which is not immediately available at initial assessment, were also highly accurate. We have developed a model that can accurately predict survival and care status escalation in sporadic Creutzfeldt-Jakob disease patients using clinical, imaging and genetic data routinely available in a specialist national referral service. The utility and generalizability of these models to other settings could be prospectively evaluated when recruiting to clinical trials and providing clinical care.

OBJECTIVE: to optimise dementia prevention strategies, we must understand the complex relationships between lifestyle behaviours, frailty and genetics. METHODS: We explored relationships between frailty index, healthy lifestyle and polygenic risk scores (all assessed at study entry) and incident all-cause dementia as recorded on hospital admission records and death register data. RESULTS: the analytical sample had a mean age of 64.1 years at baseline (SD=2.9) and 53% were women. Incident dementia was detected in 1762 participants (median follow-up time=8.0 years). High frailty was associated with increased dementia risk independently of genetic risk (HR 3.68, 95% CI 3.11 to 4.35). Frailty mediated 44% of the relationship between healthy lifestyle behaviours and dementia risk (indirect effect HR 0.95, 95% CI 0.95 to 0.96). Participants at high genetic risk and with high frailty had 5.8 times greater risk of incident dementia compared with those at low genetic risk and with low frailty (HR 5.81, 95% CI 4.01 to 8.42). Higher genetic risk was most influential in those with low frailty (HR 1.31, 95% CI 1.22 to 1.40) but not influential in those with high frailty (HR 1.09, 95% CI 0.92 to 1.28). CONCLUSION: Frailty is strongly associated with dementia risk and affects the risk attributable to genetic factors. Frailty should be considered an important modifiable risk factor for dementia and a target for dementia prevention strategies, even among people at high genetic risk.

GWASs have identified numerous genetic variants associated with a wide variety of diseases, yet despite the wide availability of genetic testing the insights that would enhance the interpretability of these results are not widely available to members of the public. As a proof of concept and demonstration of technological feasibility, we developed PAGEANT (Personal Access to Genome & Analysis of Natural Traits), usable through Graphical User Interface or command line-based version, aiming to serve as a protocol and prototype that guides the overarching design of genetic reporting tools. PAGEANT is structured across five core modules, summarized by five Qs: (i) quality assurance of the genetic data; (ii) qualitative assessment of genetic characteristics; (iii) quantitative assessment of health risk susceptibility based on polygenic risk scores and population reference; (iv) query of third-party variant databases (e.g. ClinVAR and PharmGKB) and (v) quick Response code of genetic variants of interest. Literature review was conducted to compare PAGEANT with academic and industry tools. For 2504 genomes made publicly available through the 1000 Genomes Project, we derived their genomic characteristics for a suite of qualitative and quantitative traits. One exemplary trait is susceptibility to COVID-19, based on the most up-to-date scientific findings reported.

Dementia has a devastating impact on the quality of life of patients and families and comes with a huge cost to society. Dementia prevention is considered a public health priority by the World Health Organization. Delaying the onset of dementia by treating associated risk factors will bring huge individual and societal benefit. Empirical evidence suggests that, in higher-income countries, dementia incidence is decreasing as a result of healthier lifestyles. This observation supports the notion that preventing dementia is possible and that a certain degree of prevention is already in action. Further reduction of dementia incidence through deliberate prevention plans is needed to counteract its growing prevalence due to increasing life expectancy. An increasing number of individuals with normal cognitive performance seek help in the current memory clinics asking an evaluation of their dementia risk, preventive interventions, or interventions to ameliorate their cognitive performance. Consistent evidence suggests that some of these individuals are indeed at increased risk of dementia. This new health demand asks for a shift of target population, from patients with cognitive impairment to worried but cognitively unimpaired individuals. However, current memory clinics do not have the programs and protocols in place to deal with this new population. We envision the development of new services, henceforth called Brain Health Services, devoted to respond to demands from cognitively unimpaired individuals concerned about their risk of dementia. The missions of Brain Health Services will be (i) dementia risk profiling, (ii) dementia risk communication, (iii) dementia risk reduction, and (iv) cognitive enhancement. In this paper, we present the organizational and structural challenges associated with the set-up of Brain Health Services.

We evaluated whether the association between cigarette smoking and dementia risk is modified by genetic predisposition including apolipoprotein E (APOE) genotype and polygenic risk (excluding the APOE region). We included 193,198 UK Biobank participants aged 60-73 years without dementia at baseline. of non-APOE-ε4 carriers, 0.89% (95% CI 0.73-1.08%) current smokers developed dementia compared with 0.49% (95% CI 0.44-0.55%) of never smokers (adjusted HR 1.78; 95% CI 1.39-2.29). In contrast, of one APOE-ε4 allele carriers, 1.69% (95% CI 1.31-2.12%) current smokers developed dementia compared with 1.40% (95% CI 1.25-1.55%) of never smokers (adjusted HR 1.06; 95% CI 0.77-1.45); of two APOE-ε4 alleles carriers, 4.90% (95% CI 2.92-7.61%) current smokers developed dementia compared with 3.87% (95% CI 3.11-4.74%) of never smokers (adjusted HR 0.94; 95% CI 0.49-1.79). of participants with high polygenic risk, 1.77% (95% CI 1.35-2.27%) current smokers developed dementia compared with 1.05% (95% CI 0.91-1.21%) of never smokers (adjusted HR 1.63; 95% CI 1.16-2.28). A significant interaction was found between APOE genotype and smoking status (P = 0.002) while no significant interaction was identified between polygenic risk and smoking status (P = 0.25). APOE genotype but not polygenic risk modified the effect of smoking on dementia risk.

We envisage the development of new Brain Health Services to achieve primary and secondary dementia prevention. These services will complement existing memory clinics by targeting cognitively unimpaired individuals, where the focus is on risk profiling and personalized risk reduction interventions rather than diagnosing and treating late-stage disease. In this article, we review key potentially modifiable risk factors and genetic risk factors and discuss assessment of risk factors as well as additional fluid and imaging biomarkers that may enhance risk profiling. We then outline multidomain measures and risk profiling and provide practical guidelines for Brain Health Services, with consideration of outstanding uncertainties and challenges. Users of Brain Health Services should undergo risk profiling tailored to their age, level of risk, and availability of local resources. Initial risk assessment should incorporate a multidomain risk profiling measure. For users aged 39-64, we recommend the Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE) Dementia Risk Score, whereas for users aged 65 and older, we recommend the Brief Dementia Screening Indicator (BDSI) and the Australian National University Alzheimer's Disease Risk Index (ANU-ADRI). The initial assessment should also include potentially modifiable risk factors including sociodemographic, lifestyle, and health factors. If resources allow, apolipoprotein E ɛ4 status testing and structural magnetic resonance imaging should be conducted. If this initial assessment indicates a low dementia risk, then low intensity interventions can be implemented. If the user has a high dementia risk, additional investigations should be considered if local resources allow. Common variant polygenic risk of late-onset AD can be tested in middle-aged or older adults. Rare variants should only be investigated in users with a family history of early-onset dementia in a first degree relative. Advanced imaging with 18-fluorodeoxyglucose positron emission tomography (FDG-PET) or amyloid PET may be informative in high risk users to clarify the nature and burden of their underlying pathologies. Cerebrospinal fluid biomarkers are not recommended for this setting, and blood-based biomarkers need further validation before clinical use. As new technologies become available, advances in artificial intelligence are likely to improve our ability to combine diverse data to further enhance risk profiling. Ultimately, Brain Health Services have the potential to reduce the future burden of dementia through risk profiling, risk communication, personalized risk reduction, and cognitive enhancement interventions.

Although prevention of dementia and late-life cognitive decline is a major public health priority, there are currently no generally established prevention strategies or operational models for implementing such strategies into practice. This article is a narrative review of available evidence from multidomain dementia prevention trials targeting several risk factors and disease mechanisms simultaneously, in individuals without dementia at baseline. Based on the findings, we formulate recommendations for implementing precision risk reduction strategies into new services called Brain Health Services. A literature search was conducted using medical databases (MEDLINE via PubMed and SCOPUS) to select relevant studies: non-pharmacological multidomain interventions (i.e. combining two or more intervention domains), target population including individuals without dementia, and primary outcomes including cognitive/functional performance changes and/or incident cognitive impairment or dementia. Further literature searches covered the following topics: sub-group analyses assessing potential modifiers for the intervention effect on cognition in the multidomain prevention trials, dementia risk scores used as surrogate outcomes in multidomain prevention trials, dementia risk scores in relation to brain pathology markers, and cardiovascular risk scores in relation to dementia. Multidomain intervention studies conducted so far appear to have mixed results and substantial variability in target populations, format and intensity of interventions, choice of control conditions, and outcome measures. Most trials were conducted in high-income countries. The differences in design between the larger, longer-term trials that met vs. did not meet their primary outcomes suggest that multidomain intervention effectiveness may be dependent on a precision prevention approach, i.e. successfully identifying the at-risk groups who are most likely to benefit. One such successful trial has already developed an operational model for implementing the intervention into practice. Evidence on the efficacy of risk reduction interventions is promising, but not yet conclusive. More long-term multidomain randomized controlled trials are needed to fill the current evidence gaps, especially concerning low- and middle-income countries and integration of dementia prevention with existing cerebrovascular prevention programs. A precision risk reduction approach may be most effective for dementia prevention. Such an approach could be implemented in Brain Health Services.

IMPORTANCE: Machine learning algorithms could be used as the basis for clinical decision-making aids to enhance clinical practice. OBJECTIVE: to assess the ability of machine learning algorithms to predict dementia incidence within 2 years compared with existing models and determine the optimal analytic approach and number of variables required. DESIGN, SETTING, AND PARTICIPANTS: This prognostic study used data from a prospective cohort of 15 307 participants without dementia at baseline to perform a secondary analysis of factors that could be used to predict dementia incidence. Participants attended National Alzheimer Coordinating Center memory clinics across the United States between 2005 and 2015. Analyses were conducted from March to May 2021. EXPOSURES: 258 variables spanning domains of dementia-related clinical measures and risk factors. MAIN OUTCOMES AND MEASURES: the main outcome was incident all-cause dementia diagnosed within 2 years of baseline assessment. RESULTS: in a sample of 15 307 participants (mean [SD] age, 72.3 [9.8] years; 9129 [60%] women and 6178 [40%] men) without dementia at baseline, 1568 (10%) received a diagnosis of dementia within 2 years of their initial assessment. Compared with 2 existing models for dementia risk prediction (ie, Cardiovascular Risk Factors, Aging, and Incidence of Dementia Risk Score, and the Brief Dementia Screening Indicator), machine learning algorithms were superior in predicting incident all-cause dementia within 2 years. The gradient-boosted trees algorithm had a mean (SD) overall accuracy of 92% (1%), sensitivity of 0.45 (0.05), specificity of 0.97 (0.01), and area under the curve of 0.92 (0.01) using all 258 variables. Analysis of variable importance showed that only 6 variables were required for machine learning algorithms to achieve an accuracy of 91% and area under the curve of at least 0.89. Machine learning algorithms also identified up to 84% of participants who received an initial dementia diagnosis that was subsequently reversed to mild cognitive impairment or cognitively unimpaired, suggesting possible misdiagnosis. CONCLUSIONS AND RELEVANCE: These findings suggest that machine learning algorithms could accurately predict incident dementia within 2 years in patients receiving care at memory clinics using only 6 variables. These findings could be used to inform the development and validation of decision-making aids in memory clinics.

BACKGROUND: Brief cognitive assessments can result in false-positive and false-negative dementia misclassification. We aimed to identify predictors of misclassification by 3 brief cognitive assessments; the Mini-Mental State Examination (MMSE), Memory Impairment Screen (MIS) and animal naming (AN). METHODS: Participants were 824 older adults in the population-based US Aging, Demographics and Memory Study with adjudicated dementia diagnosis (DSM-III-R and DSM-IV criteria) as the reference standard. Predictors of false-negative, false-positive and overall misclassification by the MMSE (cut-point

Earlier diagnosis of dementia is increasingly being recognized as a public health priority. As screening is not generally recommended, case-finding in clinical practice is encouraged as an alternative dementia identification strategy. The approaches of screening and case-finding are often confused, with uncertainty about what case-finding should involve and under what circumstances it is appropriate. We propose a formal definition of dementia case-finding with a clear distinction from screening. We critically examine case-finding policy and practice and propose evidence requirements for implementation in clinical practice. Finally, we present a case-finding pathway and discuss the available evidence for best practice at each stage, with recommendations for research and practice. In conclusion, dementia case-finding is a promising strategy but currently not appropriate due to the substantial gaps in the evidence base for several components of this approach.

BACKGROUND: Vitamin D deficiency has been linked with dementia risk, cognitive decline, and executive dysfunction. However, the association with memory remains largely unknown. OBJECTIVE: to investigate whether low serum 25-hydroxyvitamin D (25(OH)D) concentrations are associated with memory decline. METHODS: We used data on 1,291 participants from the US Cardiovascular Health Study (CHS) and 915 participants from the Dutch Longitudinal Aging Study Amsterdam (LASA) who were dementia-free at baseline, had valid vitamin D measurements, and follow-up memory assessments. The Benton Visual Retention Test (in the CHS) and Rey's Auditory Verbal Learning Test (in the LASA) were used to assess visual and verbal memory, respectively. RESULTS: in the CHS, those moderately and severely deficient in serum 25(OH)D changed -0.03 SD (95% CI: -0.06 to 0.01) and -0.10 SD (95% CI: -0.19 to -0.02) per year respectively in visual memory compared to those sufficient (p = 0.02). In the LASA, moderate and severe deficiency in serum 25(OH)D was associated with a mean change of 0.01 SD (95% CI: -0.01 to 0.02) and -0.01 SD (95% CI: -0.04 to 0.02) per year respectively in verbal memory compared to sufficiency (p = 0.34). CONCLUSIONS: Our findings suggest an association between severe vitamin D deficiency and visual memory decline but no association with verbal memory decline. They warrant further investigation in prospective studies assessing different memory subtypes.