Workshop | The Future of Population-Based Studies in Alzheimer’s Disease and Related Dementia Research: Setting Future Scientific Priorities

Theme 1: Diversifying Populations

With the continued increase in diversity in the United States, AD/ADRD studies need to recruit members of minority groups to adequately represent the U.S. population. Because the amyloid, tau, neurodegeneration (ATN) framework is often not translatable to minority populations, AD/ADRD research needs to employ a more complex disease framework that accounts for additional types of data, such as gene and protein expression as well as an individual’s exposome, including environmental exposures and behavioral and social factors. In support of AD/ADRD population study diversification, various current studies, such as the AD Sequencing Project (ADSP) have prioritized recruitment of individuals from more diverse backgrounds; future studies will emphasize participant diversity and collect more detailed exposome data. In addition, future population studies will need to consider within-population heterogeneity and intersectional identities.

Presenters during this session outlined the current landscape, challenges, and ideal future of AD/ADRD research in select priority populations: (1) individuals with early onset sporadic AD, (2) centenarians and their offspring as a proxy for AD/ADRD resistance and resilience, (3) individuals with Down Syndrome and autism spectrum disorder (ASD), (4) sexual and gender minorities, and (5) indigenous populations (specifically, members of the Tsimane and Moseten communities in South America).

During the Theme 1 breakout discussion session, meeting participants outlined research priorities to increase AD/ADRD study population diversity, some of which were specific to particular priority populations. Meeting participants generally agreed that diverse population studies require design and analysis optimization to maximize utility and efficiency in data collection and minimize participant burden. In addition, existing assessment tools for AD/ADRD have inherent biases, necessitating updates and possibly development of new assessments to improve cultural and language inclusivity and minimize bias. The AD/ADRD population research community, including NIA, can also improve their recruitment efforts through collaborative efforts. Participants also identified additional priority populations that were not included in the session presentations: rural populations, occupational cohorts, veterans, immigrants, international populations, Native Hawaiian and Pacific Islanders, Asian-Americans, and individuals with comorbid conditions. Ultimately, the state of AD/ADRD research in each priority population differs from one another, and therefore, to address research gaps, the scientific community needs to implement population-specific research priorities and strategies.

Theme 2: Infrastructure Needs for Current and Future Population Studies

Federally funded scientific data collection increasingly relies on public databases and open access data, creating opportunities for innovative secondary analyses of existing AD/ADRD datasets. While open access databases can inform the complex characteristics of late life cognitive decline pathologies, data platform designers must ensure that data is reasonably accessible to the broader research community. Researchers requesting data from established AD/ADRD databases often face barriers to access in the form of complex data request forms, restrictive data use limitations, and interoperability issues between large databases.

NIH and other federal institutions rely on several initiatives to facilitate data sharing, including data coordination centers like the AD Knowledge Portal and the Exceptional Longevity Translational Resources Portal. NIH repository models support effective and sustainable data sharing platforms with pre-harmonized data that can be queried and flexibly filtered across studies. Accessible databases pave the way for innovative data analysis techniques that can extract patterns from large scale data, including (1) linking AD research protocols with non-AD legacy study, or (2) computational models linking AD/ADRD multiomics data with existing electronic health records.

During the Theme 2 breakout session, meeting participants assessed existing gaps in data accessibility and discussed future initiatives to expand publicly available data for researchers and the public. Participants emphasized a need for federal entities to pioneer interoperable data infrastructure that can bridge existing data siloes. Attendees also highlighted the importance of developing user-friendly infrastructure to remove barriers on currently available data. Data users are often hindered by massive datasets that incur significant time costs to navigate, and principal investigators (PIs) frequently spend additional time teaching trainees how to use their data. Participants acknowledged that any efforts at standardizing data elements for user and interoperability will need continual updating, as relevant measures for AD/ADRD research will evolve over time. Future expansions of population study infrastructure will need to (1) engage international data harmonization, (2) increase data accessibility and awareness on data resources for non-researchers, (3) incorporate community perspectives on subject data privacy in public datasets, and (4) conscientiously apply artificial intelligence tools with open science datasets to accelerate further data discovery.

Theme 3: Life Course Environmental Influences on AD Risk

Measuring environmental exposure has large discrepancies compared to measuring genetics; unlike an individual’s genetic code, the environment changes over time, affecting an individual’s length and dosage of total exposure. Researchers should consider the accuracy and etiological-relevant timing of exposure measures and consider assessing both cumulative profiles that represent average lifespan exposure and longitudinal profiles that can identify high-risk exposure events at specific timepoints. Some key exposures that are associated with disease and impact health outcomes are climate change and education. Health conditions associated with extreme weather include post-traumatic stress disorder (PTSD), stroke, and dementia, while educational experiences are associated with cognitive changes. However, the study of the exposome does not solely include measures of environmental influence and exposure on health outcomes, but also the corresponding biological and physiological responses. The exposome represents the totality of environmental exposures encountered by an individual throughout their lifespan (Wild 2005, Wild 2012). Relationships between exogenous chemicals and endogenous metabolites identified from blood, urine, and bone samples can help researchers understand biological mechanisms underlying the exposome.

During the Theme 3 breakout session, participants outlined several exposure types and measures that should be prioritized in future populations studies on AD/ADRD, discussed key unanswered questions and next steps in measuring and assessing priority exposures and their relationship to aging and AD/ADRD, and identified opportunities for collaboration to explore priority exposure types and address key unanswered questions within the field of exposomics. Participants emphasized the importance of community engagement during the research process, such as making study results available to the public and collaborating with community leaders to establish trust and garner feedback on study design. In addition, participants suggested establishing working groups to help prioritize exposure measurement practices, develop best research practices, improve data documentation, harmonize existing datasets, and identify common data elements (CDE) to standardize data collection.

Meeting Summary - Day 1

Welcome and Meeting Logistics

Damali Martin, PhD, MPH, Chief, Population Studies and Genetics Branch, National Institute on Aging (NIA)

Dr. Damali Martin began the Future of Population Studies for Alzheimer’s Disease (AD) and Related Dementias (ADRD) Research meeting by outlining key meeting objectives: (1) to foster the development of research collaborations among investigators involved in population-based AD/ADRD studies, (2) to identify and leverage resources to enhance AD/ADRD population-based research, (3) to identify avenues to enhance existing data and supplement the collection of new data for AD/ADRD population-based research, and (4) to encourage ongoing discussions regarding the development of scientific priorities focused on informing future precision-based intervention and prevention strategies for AD/ADRD among U.S. populations. She also provided background information on the National Institute on Aging (NIA) and the Division of Neuroscience (DN) as well as a summary of current population studies of AD/ADRD.

The Role for the National Institute on Aging in Alzheimer’s Disease Research

NIA leads the broad scientific effort to understand the nature of aging and to extend healthy active years of life. It is the primary institute conducting research on AD/ADRD and now the third largest federal research institute in terms of budget. NIA’s growth is due to an increase in government appropriations for AD/ADRD research, while its general payline remains stable.

DN aims to foster and support extramural research and training to further understand dementias of old age as well as neural and behavioral processes associated with aging in the brain, with a special emphasis on brain behavior relationships. The depth and breadth of DN enables strong translational and precision medicine approaches to inform potential prevention, intervention, and treatment studies for AD/ADRD.

NIA’s bench-to-bedside approach to AD/ADRD research focuses on understanding the multifactorial etiology and pathology of AD/ADRD as shaped by genetic and environmental factors. Genetics accounts for approximately 70 percent of variance in dementia risk that increases with age, and the exposome plays an important role in influencing overall risk for AD/ADRD. Dementia also has multiple prodromal phenotypes and progression trajectories, necessitating precision medicine approaches to deliver the correct interventions at optimal doses at the appropriate stages of the disease for each patient.

Current National Institute on Aging Efforts in Alzheimer’s Disease Research

NIA embraces the core principles of data sharing, transparency, and scientific rigor to support various population-based discovery studies. In support of these efforts, NIA hosts Alzheimer’s Research Summits every three years, which have identified research gaps and opportunities that serve as the basis for research milestones. These milestones detail specific steps and success criteria toward the development of effective treatments and prevention strategies for AD/ADRD. Some population-based studies-related milestones include endophenotyping at-risk and resistant cohorts, quantifying the exposome, diversifying study cohorts, developing state-of-the-art research protocols, expanding AD profiling efforts, including non-AD cohorts in downstream analyses, performing longitudinal immunological profiling, accessing and integrating electronic health records (EHRs), conducting exposome and health disparities studies, using artificial intelligence-based approaches, and assessing the contributions of pollution and climate change to AD/ADRD risk. Current NIA-supported research projects that address these milestones are summarized below.

Diversification of Study Cohorts

The United States is becoming increasingly racially and ethnically diverse. Therefore, these racial and ethnic groups should be well represented in both epidemiological and clinical studies. Age-standardized incidence rates differ both across racial and ethnic groups and within racial groups. For example, while the age-standardized dementia incidence rate among Asian-American Kaiser Permanente Northern California members from 2000-2013 was lower than White members, incidence varied across Asian-American subpopulations. Further complicating AD/ADRD research in diverse populations, the amyloid, tau, neurodegeneration (ATN) framework does not replicate well in minority populations. Therefore, AD/ADRD research needs a more complex framework that accounts for omics as well as environmental, behavioral, and social factors.

Studies of minoritized populations not only provide information on the burden of AD in these groups but also will help identify AD risk and resilience factors. Current longitudinal and epidemiological studies of aging and AD/ADRD in racial and ethnic minority populations include the Minority Aging Research Study (MARS) (African American), Religious Orders Study (ROS) Latino Core, Asian Cohort for AD (ACAD), Brain Health and Ethnic Disparities in ADRD Risk: The Case of Arab Americans, Strong Heart Study (American Indian/Alaska Native [AI/AN]), Washington Heights/Inwood Columbia Aging Project (WHICAP) and WHICAP-Offspring studies (African American and Hispanic/Latino), Risk Factors for Future Cognitive Decline and AD in Older African Americans (African American), Indigenous Samoan Partnership to Initiate Research Excellence (INSPIRE) Program, Stress and Resilience in Dementia (STRIDE) Study (African American, AI/AN), and Health and Aging Brain Study-Health Disparities (HABS-HD) (African American and Hispanic/Latino). Dr. Martin outlined some current research efforts that address AD/ADRD milestones and focus on increasing diversity in study populations.

Genomic Profiling of AD

The AD Sequencing Project (ADSP) started in 2012 and aims to increase study enrollment diversity. By conducting additional AD genomic profiling and harmonizing existing data across other studies for minority populations, ADSP continues to improve understanding of genetic variants associated with AD/ADRD. Moving forward, researchers will use these identified variants to perform functional genomics studies and, in combination with artificial intelligence and machine learning (AI/ML) tools, develop new risk classification strategies for AD/ADRD.

Quantification of the Exposome

Although longitudinal cohort projects collect extensive information on participant demographics, family history, genetic screening, and lifestyle, environmental exposure data are required to fully characterize the participant’s exposome. NIA DN has been developing a precision environmental health approach that leverages NIA investments in population studies, genetics, epigenomics, metabolomics, systems biology, and data infrastructure as well as infrastructure and expertise from the National Institute of Environmental Health Sciences (NIEHS). With these resources, NIA aims to expand the existing multi-omics systems to build predictive models of wellness and disease that reflect gene-environment interactions. Incorporating new technologies such as AI/ML into these efforts may yield fruitful results.

Relevant Funding Opportunities

NIA has developed four funding opportunities to support population-based studies focused on the exposome. DN and the Division of Behavioral and Social Research released a joint Notice of Special Interest (NOSI): Administrative Supplements to Support Research Infrastructure on Exposome Studies in AD/ADRD ( NOT-AG-22-022 ). DN is also accepting applications for three different requests for application (RFAs): Understanding Gene-Environment Interactions in Brain Aging and AD/ADRD ( RFA-AG-24-021 ), Quantifying the Impact of Environmental Toxicants on AD/ADRD in Cohort Studies ( RFA-AG-24-022 ), and Preclinical Studies to Characterize the Impact of Toxicants on Brain Aging and AD/ADRD ( RFA-AG-24-023 ).

Emerging Areas of Interest and Recent Activities in Alzheimer’s Disease Research

Dr. Martin outlined other emerging research areas of interest and recent activities relevant to AD/ADRD, including the relationship between oral health and AD, expansion of AD research into low- and middle-income countries (LMICs), and a National Academies of Sciences, Engineering, and Medicine (NASEM)-sponsored public session aimed at assessing the current AD/ADRD landscape.

Epidemiological Relationship Between Oral Health and Alzheimer’s Disease

NIA hosted the Oral Health and AD/ADRD Workshop during June 2023, which convened researchers in the areas of dentistry and neuroscience. The workshop included an overview of oral health in older populations as well as the epidemiological relationship between oral health and AD. Participants discussed the oral microbiome and potential biological mechanisms linking it to AD; this is an area of research of great interest to NIA.

Global Expansion of Alzheimer’s Disease Research

Populations in LMICs have a higher prevalence of potentially modifiable risk factors compared to those in high-income countries (HICs). Expanding AD/ADRD research to include LMIC populations can provide unique insights into AD/ADRD risk and resilience. To support research in LMICs, particularly in the continent of Africa, NIA will offer multiple funding opportunities, including the Building Neuroscience Research Infrastructure for AD/ADRD in Africa ( RFA-24-027 ) and the Small Research Grant Program for the Next Generation of Researchers in LMICs for AD/ADRD ( PAR-23-179 ). Typical K mechanism grants for early career researchers require 75 percent effort committed to research, which is often not possible for researchers in LMICs. Therefore, the small research grant program supporting early career researchers uses an R03 mechanism to allow researchers to designate an appropriate level of effort required for research project completion. Grantees can also use these funds to support additional staff to help achieve research goals.

Upcoming National Academies of Sciences, Engineering, and Medicine, Session to Assess the Current Research Landscape

NASEM will host the Research Priorities for Preventing and Treating AD/ADRD Public Session on November 8, 2023 to examine and assess the current state of research aimed at preventing and treating AD/ADRD along the research and development pipeline. Session participants will: (1) assess the evidence on non-pharmacological interventions aimed at preventing and treating AD/ADRD, (2) identify key barriers and opportunities to advancing AD/ADRD prevention and treatment strategies, and (3) review the most promising research areas in preventing and treating AD/ADRD.

Theme 1: Diversifying Populations

Moderator: Damali Martin, PhD, MPH, Chief, Population Studies and Genetics Branch, NIA

Plenary: ADRD study populations, needs and gaps moving forward

Rachel Whitmer, PhD, University of California-Davis

Dr. Rachel Whitmer outlined the current state of ADRD population-based studies, including current gaps and needs to address to improve the utility of this research across diverse populations. Because dementia incidence rates vary across racial and ethnic groups, research findings from primarily White study populations are not broadly applicable. The National Institutes of Health (NIH) has previously defined priority populations, which include racial and ethnic minorities: Hispanic/Latino, AI/AN, Black/African American, Asian American, and Native Hawaiian and other Pacific Islander (NHPI). Other underserved populations (e.g., rural, disabled, sexual and gender minority [SGM], etc.) have higher exposures to risk factors for poorer brain health across the lifespan. For example, birthplace has enduring consequences on dementia risk in Northern California Kaiser Permanente members, with those born in high stroke mortality states being at higher risk for dementia compared to those born in lower stroke mortality states. These consequences remained even in those who moved to low stroke mortality states.

The National Institute on Minority Health and Health Disparities (NIMHD) released a framework to help researchers address the differences in dementia incidence, etiology, and pathology across diverse populations. The framework considers multiple levels of influence that can impact people at the individual, interpersonal, community, and societal levels as well as domains of influence—biological, behavioral, physical and built environment, sociocultural environment, and health care system. Notably, these domains of influence can affect people across different levels of influence (e.g., individual, interpersonal, community, societal), and these influences can change over a person’s lifetime. The exposome framework can also be used to systematically capture both the unique set of initial or early-life characteristics related to AD/ADRD risk (e.g., endogenous exposures, epigenetic and genetic changes, physiology) as well as environments and exposures that can impact AD/ADRD risk across the lifespan. Several critical life stages have been identified for which some exposures may have a greater impact compared to other stages with respect to future disease. However, these critical life stages, risk magnitudes, and meaningful exposure thresholds likely differ across different populations. In addition, because brain changes accumulate over decades, the impacts of risk factors may change across the lifespan; for example, the association between hypertension and dementia differs at midlife versus late life.

Common Population-Based Study Designs and Strategies to Improve Study Feasibility in Diverse Populations

While life course approaches are essential for addressing NIA priorities, population-based studies are expensive in terms of time, effort, and funds, and may not always be feasible. For longitudinal studies in older individuals, enrollment at birth may not be required to access relevant early-life data. A person enrolled at midlife can be linked with their corresponding EHR to access earlier medical records. In addition, the participant can provide information about where they grew up, attended school, and worked, which, with geocoding, can be retrospectively linked to different exposures (e.g., air pollution, toxic chemicals, weather patterns, crime, green space availability, food density, area deprivation index).

Overall, EHRs present an untapped opportunity to evaluate dangerous and rare exposures, as well as special subgroups with comorbid diseases, as they provide data for many people. For example, while hypoglycemia in type I diabetes is rare, review of EHRs for hemoglobin A1c (HbA1c) levels showed that those with very high or very low HbA1c levels were at higher risk for dementia than those with intermediate levels. Although review of EHRs grants more equal access across racial and ethnic groups, this data source relies on a person’s touchpoints with the health care system. In addition, this data source contains an inherent lack of precision for health outcomes.

Dr. Whitmer outlined two population-based study examples that rely on obtaining antecedent information from both EHRs and participant self-reporting.

Life After 90 Study

The Life After 90 Study , which began in 2018, is a prospective, diverse cohort embedded in the Kaiser Northern California health care delivery system with antecedent retrospective information collection. Participants self-report retrospective information on social and behavioral factors, as well as midlife health factors; midlife health information is also supplemented by EHRs. Self-report information dates back to 1964, while EHR data dates back to 1996. Notably, this cohort has cardiovascular risk factors that vary in magnitude across racial and ethnic groups. Because this study focuses on those aged 90 and older, researchers will also consider various resilience factors.

The Adult Changes in Thought Study

The Adult Changes in Thought Study consists of a longitudinal cohort embedded within a health care system that includes access to data from medical records (starting in 1947) and EHRs (starting in 1977), interviews, biological samples, physical measurements, brain imaging data, life course exposures, and autopsy. This study began enrolling dementia-free adults aged 65 and older in 1994 and continues to enroll new participants, particularly those from underrepresented minority populations. Data collection on a biannual basis enables a life course approach to identify dementia risk factors, and autopsy data enables confirmation of AD/ADRD diagnoses based on quantitative neuropathology.

Importance of Studying Heterogeneity Within Priority Populations

Dr. Whitmer emphasized the value of single ethnoracial group cohorts for providing sufficient statistical power and richness to address research questions related to intersectionality and heterogeneity within groups. Examples of these single ethnoracial group cohorts include the Vietnamese Insights into Cognitive Aging Program (VIP) , HABS-HD , and the Study of Health and Aging in African Americans . VIP aims to characterize longitudinal cognitive function and AD/ADRD risk in a community-based cohort of older Vietnamese Americans in Santa Clara and Sacramento counties in California. This study is particularly well positioned to examine the influences of adversity, trauma, and sociocontext on AD/ADRD risk, as well as how cardiovascular risk factors potentially mediate these influences and affect cognitive outcomes. HABS-HD is focused on Hispanic and African American populations and collects various data types, including imaging, clinical exams and laboratory findings, and cognitive assessments, as well as information on sociocultural, environmental, and behavioral factors. The Study of Health and Aging in African Americans is currently examining the relationship between midlife cardiovascular risk factors and late life cognition, as well as the effects of school desegregation on late life cognition. Importantly, researchers should consider how the timing of desegregation may impact studies of the relationship between education and AD/ADRD.

Best Practices for Inclusion of Diverse Study Populations

Dr. Whitmer outlined some best practices in recruitment and retention of diverse study populations while avoiding overly coercive actions. Researchers need to build trust with local minority communities, in part by contextualizing their research, and ensure a true return on investment both through individual compensation and community benefits. Ensuring these benefits requires community engagement beginning at early stages of the design process. Protocols should be designed to maximize participant access and the value of data being collected while minimizing the overall participant burden, which could involve reducing the amount of time required to participate as well as providing options for community-based and home-based visits. Increasing access to research studies can also include offering study assessments in multiple languages.

Additional strategies can both increase diversity and maximize the utility of existing study cohorts. Through harmonization approaches, researchers can leverage and link existing datasets for larger comparisons across and within populations. These within-group analyses enable studies that address heterogeneity and intersectionality within populations. AD/ADRD researchers can also collaborate with other population-based research projects that enroll midlife and young adults to leverage their successful recruitment efforts and collect data relevant to AD/ADRD risk from these populations as they age.

Population-based studies will always have to contend with selection and survival biases in enrollment. Because AD/ADRD usually affects older adults, many may not survive long enough for a dementia diagnosis. Therefore, risk calculations are likely underestimates of true risks of dementia, particularly when related to other health issues, such as hypertension. To mitigate these biases, researchers should account for differential survival across populations related to common health issues that may be related to AD/ADRD. They can also study larger cohorts and use inverse probability weighting.

Discussion

One participant noted that studies focused on single ethno-racial groups may still benefit from having a non-Hispanic White comparison group to ensure matched controls and facilitate data harmonization and cross-dataset comparisons. However, inclusion of participants outside of the single ethno-racial group may reduce overall study power when analyzing within-group heterogeneity. Others have challenged the idea that non-Hispanic White should serve as the control group. Lisa Barnes has noted this design perpetuates a bias toward this group being a "gold standard' of sorts, so investigators should utilize caution when determining appropriate comparison groups.

Another participant emphasized the importance of developing new methods to maximize study participant inclusion. Using the same methods with known biases simply perpetuates those same biases across diverse populations. Instead, the participant suggested viewing these communities from their own perspective—what are their skills, talents, and capabilities? Rather than impose an HIC research framework, NIA should empower LMICs to design and lead their own life course study approaches.

Early (Young) Onset Sporadic AD: A rare and underserved segment of the AD population

Brad Dickerson, MD, Harvard Medical School

Early-onset sporadic AD (EOAD) occurs prior to 65 years of age and typically affects people between their late 40s through early 60s. EOAD is not simply an earlier onset of typical AD; instead, symptoms are usually more aggressive and disabling and have genetic underpinnings different from late-onset AD (LOAD). Those diagnosed with EOAD often have a higher tau burden at presentation as well as greater involvement of non-memory cognitive domains and less involvement of brain networks typically associated with LOAD. EOAD is under-detected and often initially misdiagnosed as a psychiatric illness or frontotemporal dementia (FTD). Many with EOAD are not Medicare beneficiaries, complicating EOAD assessment data access.

Longitudinal Early-Onset Alzheimer’s Disease Study

The Longitudinal Early-Onset Alzheimer’s Disease Study (LEADS) is one of the largest EOAD cohorts, consisting of 600 cognitively impaired and 100 cognitively unimpaired individuals ages 40 to 64 years old. LEADS participants are evaluated annually via clinical, medical, cognitive, behavioral, and fluid and imaging biomarker assessments. By pairing these annual data with participants’ genetic information and eventual autopsy data, this study aims to (1) define EOAD and phenotypic variants to enable more efficient and timely diagnoses, (2) understand the progression of biological disease and clinical illness in EOAD, and (3) develop an optimal population of individuals with specific EOAD pathology who are motivated to participate in future interventional trials (i.e., Trials Unit). With further understanding of EOAD progression, researchers can begin identifying factors that confer vulnerability or resilience to clinical symptoms.

Study Population Characteristics

The LEADS study has had difficulty in recruiting from diverse populations, as official study sites are specialty clinics and academic institutions. Researchers do recruit from local communities, although due to the rarity of EOAD, this strategy is difficult and often inefficient.

The LEADS study population consists of approximately one-third classified with mild cognitive impairment (MCI) and the other two-thirds with mild dementia. Because MCI occurs so early in the life course of EOAD patients, recruiting these individuals requires the additional challenge of screening much younger people. A large proportion of this study population also the carries apolipoprotein E4 (APOE4) genetic variant. Most study participants experience a cortical-based amnestic phenotype with executive dysfunction that differs from the medial temporal lobe phenotype. The study population also includes individuals with atypical syndromes, such as posterior cortical atrophy, primary progressive aphasia, and predominant dysexecutive function.

Study Findings

Based on imaging data, most study participants with EOAD experienced sparing of anterior and medial temporal lobe damage and more prominently affected lateral temporal and lateral parietal and medial parietal regions compared to LOAD. This imaging signature of EOAD correlates with clinical severity as measured by Clinical Dementia Rating Scale Sum of Boxes (CDR-SB), MMSE, amyloid-PET, and tau-PET. Notably, study site researchers are better at distinguishing EOAD at the mild dementia stage compared to the MCI stage. Approximately 39 percent of individuals with suspected EOAD at the MCI stage based on clinical information actually had early-onset amyloid-negative/non-AD cognitively impaired (EOnonAD) based on imaging.

Patient Support Activities and Future Directions

Supplemental funds for LEADS have provided additional support to study participants and their families, including a social worker employed at each study site as well as a patient- and family-led conference organized by the Alzheimer’s Association.

As the enrollment for LEADS nears completion, researchers are now comparing EOAD to LOAD disease signatures and creating a Trials Unit to support future disease-modifying therapy (DMT) clinical trials. Moving forward, future studies of EOAD should include blood biomarkers to enable earlier and more rapid screening, as well as consider initiating brain health screenings at age 50 to improve enrollment. In addition, further public and provider education and outreach efforts can help increase the diversity of EOAD cohorts.

Discussion

LEADS excluded individuals with two or more first-degree relatives with a diagnosis of dementia, regardless of type, to maintain the study focus on sporadic EOAD. Six study participants had simple dominant mutations in either presenilin (PSEN) genes or amyloid precursor protein (APP), only one of whom had a family member with the same mutation. At least one of the participants with a PSEN1 mutation had long-lived family members without the mutation, suggesting a de novo origin.

Resilience to Alzheimer’s and Related Dementias as a Window into Disease Mechanisms

Stacy Andersen, PhD, Boston University

Dr. Stacy Andersen outlined multiple centenarian studies that are providing crucial insights into the mechanisms of resilience and resistance to AD/ADRD.

Deep Phenotyping in the New England Centenarian Study

The N ew England Centenarian Study has assessed 1,800 centenarians (including 150 over 110 years of age) and 500 of their offspring, aged 63 to 101 years of age as a model of healthy aging. Many of these centenarians continued living independently into their 90s and experienced delayed onset of age-associated illnesses, not just by postponing morbidity but also by compressing it into a much smaller percentage of their end of life compared to non-centenarians. The genetic backgrounds of these centenarians were also enriched for variants related to lower risks for cardiovascular disease and heart disease, as well as promotion of immunity and metabolism. The incidence rates of the APOE4 risk allele were much lower and APOE2 protective allele much higher in centenarians compared to non-centenarians. In addition, when compared to offspring from non-centenarians, centenarian offspring experienced lower mortality; reduced prevalence and incidence of cardiovascular disease, hypertension, and diabetes; and were less likely to be cognitively impaired at and 8 years post-study enrollment.

Two additional studies—the Longevity Consortium Centenarian Project and the Integrative Longevity Omics Project —are performing deep phenotyping of the New England Centenarian Study cohort to better understand healthy aging. Data collection can occur remotely with activity trackers and tablets where the clinician can help the participant interact with the technology. Biological samples collected at clinical sites are used to collect omics data and generate induced pluripotent stem cell (iPSC) lines for subsequent neuronal cell studies.

Long Life Family Study

The Long Life Family Study collects information from the children and grandchildren of long-lived people in the United States and Denmark to identify factors associated with familial longevity. Families are selected for enrollment based on the Family Longevity Selection Score (FLoSS), which is used to identify individuals living longer than expected compared to others in their birth cohorts. The Long Life Family Study is very selective—when applying the study’s FLoSS criteria to those enrolled in the Framingham Heart Study, only one percent of families met study eligibility requirements. Those enrolled in the Long Life Family Study participate in in-home assessments and follow-up phone appointments to collect various clinical, cognitive, and biomarker measures. Current findings from this study indicate that, compared to their spouses, children of long-lived people have a lower prevalence of AD-type cognitive impairment, better episodic memory and processing speed, and slower decline in processing speed.

Resilience/Resistance Against Alzheimer’s Disease in Centenarians and Offspring

The Resilience/Resistance Against Alzheimer’s Disease in Centenarians and Offspring (RADCO) leverages existing centenarian cohorts to identify cognitive “super-agers” (i.e., those with episodic memory performance at or above people 30 years younger) to (1) gauge their cognitive resilience, (2) understand the underlying protective biology, and (3) translate findings into development of therapeutic targets. This approach is based on the hypothesis that centenarian cohorts will be enriched for cognitive super-agers because they have a delay or entire avoidance of AD. Using biomarkers and neuropathology, RADCO researchers will characterize resistance (i.e., avoidance of AD pathology) and resilience (i.e., presence of AD pathology without functional phenotypes) endophenotypes to identify protective factors.

Down Syndrome and Autism: Current knowledge and opportunities

Ben Handen, PhD, University of Pittsburgh

While more than 30 years of research in Down Syndrome (DS) has identified biomarkers and genetic risk factors for AD, researchers have only recently begun examining aging and AD in autism spectrum disorder [ASD]. The health care system needs strategies to manage accessibility challenges to ensure those with DS and ASD receive adequate health. Dr. Ben Handen outlined the current state of research on the relationships between DS and AD as well as ASD and AD.

Down Syndrome and AD/ADRD

DS occurs in 1 in 750 live births and the lifespan of those with DS has increased considerably due to improvements in health care services and strategies to address comorbid medical disorders. With increased life expectancy, adults with DS are at significant risk for EOAD. Due to their additional copy of chromosome 21, adults with DS have an extra copy of APP and, starting from infancy, generate more amyloid compared to the general population. Nearly all adults with DS exhibit neuropathological changes consistent with AD by 40 years of age and most develop dementia in their late 60s.

To better understand the relationship between DS and AD, the Alzheimer’s Biomarker Consortium-Down Syndrome (ABC-DS) measured the rate of amyloid-PET positivity in adults with DS. Between the ages of 40 and 44, 50 percent of participants were amyloid-positive; between 45 and 49 years of age, 80 percent were amyloid-positive, with the rate approaching 100 percent later in the lifespan. Through imaging comparisons between this DS cohort and an autosomal dominant AD cohort, researchers found a significantly shorter gap (17 years) between amyloid positivity and subsequent AD diagnosis in those with DS compared to the autosomal dominant AD cohort (23 years). Similarly, adults with DS exhibit a shorter gap between amyloid positivity and tau deposition compared to adults without DS. Based on memory performance, amyloid positivity alone in adults with DS does not significantly impact memory, but both amyloid and tau positivity combined does significantly impact memory. Notably, adults with DS also have higher levels of GFAP compared to amyloid status matched adults without DS.

Dr. Handen further explained how cohort study results can help inform effective prevention trial study design. With amyloid positivity seen in adults with DS as early as mid-30s and average age of AD diagnosis at age 55, prevention trials using monoclonal antibodies should enroll adults with DS between the ages of 35 and 55 to ensure the presence of the antibody target (i.e., amyloid) and lack of AD diagnosis. Drug prevention trials in adults with DS will begin in 2024, and Dr. Handen highlighted the unique opportunities for additional lifestyle and health interventions for AD prevention that also require additional study.

Autism and AD/ADRD

Based on data from the National Vital Statistics System, dementia as cause of death occurs at a lower rate in those with ASD (6 percent) compared to individuals without ASD (11 percent). However, based on Medicaid Analytic eXtract files from 2008 to 2012, the prevalence of EOAD in those with ASD is much higher, particularly for those with ASD and intellectual disability.

Biomarker studies indicate APP dysregulation in brains and plasma of those with ASD. In younger individuals with ASD, brain levels of amyloid are lower than their non-ASD counterparts; however, the effect of lower amyloid levels in early life on dementia risk remains unclear. Interestingly, young people with Fragile X syndrome, a genetic disorder strongly associated with ASD, had higher levels of amyloid in the brain compared to non-ASD counterparts without Fragile-X.

The Toward Healthy Aging in Adults with Autism: A Longitudinal Clinical and Multimodal Brain Imaging Study is currently examining how aging-related differences contribute to different outcomes in adults with ASD. Additional opportunities for research on ASD and AD include examination of candidate biomarkers, studies of individuals with ASD older than 65 years of age, and examination of genetic and lifestyle risk factors.

Discussion

Dr. Caterina Rosano suggested looking at other biomarkers in individuals with DS that could explain the shorter gap between amyloid positivity and dementia diagnosis.

When asked whether DS and ASD participants’ parents and grandparents are examined in AD/ADRD studies, Dr. Handen explained that while he is unaware of any research in ASD, the ABC-DS study was awarded an administrative supplement to support assessment of modification of AD risk in ABC-DS participant parents.

Increasing representation of sexual and gender minorities in Alzheimer’s disease and related dementias research

Jason Flatt, PhD, University of Nevada-Las Vegas

In 2016, NIMHD designated SGM as a health disparities population for research purposes, and the lesbian, gay, bisexual, transgender, queer/questioning, and intersex (LGBTQI+) community is growing, with approximately 1.1 million LGBTQI+ individuals aged 65 and older. Historically, SGMs have faced discrimination and pathologizing along with a lack of protections for housing, employment, and health care. This discrimination is further compounded for those with intersecting marginalized identities. SGM individuals currently face barriers to health care access and use, particularly as recent legislation restricting youth access to gender affirming care has led to some health care providers being fearful of providing said care to older adults. These barriers make SGM individuals less likely to seek out health care than their cisgender, heterosexual counterparts. SGMs are also less likely to be married or have children and are two-to-three times as likely to live alone with little to no caregiver support. Lower rates of marriage for SGM populations are in part due to a lack of historical legal marriage protections. Therefore, SGMs are more likely to rely on care from chosen families who lack the legal protections and common practices that support biological and spousal caregivers.

Current Research on AD/ADRD in Sexual and Gender Minorities

Most studies of AD/ADRD in SGM are those of convenience. Results of these studies showed higher subjective cognitive decline among SGM compared to non-SGM, with even greater subjective decline among transgender adults from AI/AN, Asian, Black, Hispanic/Latino, and multiracial backgrounds. Based on Medicare beneficiary data, transgender individuals had 6 to 8 percent higher rates of dementia compared to their cisgender counterparts. Overall, national data indicate 80 percent higher odds of cognitive impairment among sexual minorities compared to heterosexual individuals.

Collection of Sexual and Gender Minority Data

Dr. Jason Flatt emphasized the importance of uniform collection of SGM information and shared recommendations for SGM data collection based on NACC Uniform Data Set (UDS) version 4. For the collection of sexual orientation data, the survey instrument should allow participants to select more than one option. For the collection of gender identity, the survey should use a two-step method that first asks for sex assigned at birth followed by a question asking the participant to describe their current gender identity and allows for selection of multiple answers. AI/AN participants should also have the option to select “two-spirit” for SGM questions. NACC UDS version 4 also asks participants about intersex status and difference of sex development (DSD).

SGM Recruitment Support and Future Directions

AD/ADRD researchers interested in increasing SGM representation in their population-based cohort studies can contact the Research Inclusion Supports Equity (RISE) Registry — a registry for LGBTQI+ people with memory loss and their caregivers—for assistance with recruitment.

Dr. Flatt received a new R01, Enhancing Measurement and Characterization of Roles and Experiences of SGM Caregivers of Persons Living with AD/ADRD ( R01AG083177 ), which aims to (1) identify and explore domains of AD/ADRD caregiving for SGM caregivers, (2) develop and refine new measures of AD/ADRD caregiving for SGM caregivers, and (3) test the new and existing measures of AD/ADRD caregiving with SGM and non-SGM caregivers.

Discussion

Principal investigators and research staff who have implemented SGM questions in their population-based studies have not encountered issues with participants refusing to answer these questions. To help with implementation, Dr. Flatt recommended training and practice for those tasked with asking these questions as well as only asking about SGM in demographics survey sections, not when asking about entitlements and service needs.

Dr. Rosano commented that Dr. Flatt’s studies that separate cisgender, heterosexual individuals from SGM may find that the control group differs from other studies that did not stratify by sexual orientation and gender identity.

When asked about potential biological mechanisms underlying the differences of AD/ADRD in SGM, Dr. Flatt explained that these biological mechanisms are likely linked to common health disparities seen in other marginalized groups. In addition, other biological mechanisms could include depression and traumatic brain injury (TBI). SGM populations experience high rates of depression, and late life depression is linked to a two-fold increased risk for dementia. Transgender individuals are also disproportionately the targets of interpersonal violence, resulting in TBIs which increase dementia risk. However, Dr. Flatt noted that the link between higher education and lower dementia risk may not be translatable to SGM populations, as those with higher levels of education who also identify as SGM do not necessarily have lower risks for dementia.

Inspiring conundrums posed by research with indigenous populations

Hillard Kaplan, PhD, Chapman University

Historically, hunters and gatherers—assuming they survived childhood—typically still had relatively long lifespans with an average age of death of 70 years. However, the exposomes and lifestyles in which an increasing proportion of the world’s population live are outside the range of those for which human biology has evolved and adapted. Three important lifestyle and exposure differences between today’s society and historic hunter-gatherers are: (1) lower adiposity relative to caloric intake in the past; (2) higher physical exertion in past societies; and (3) higher pathogen exposure and burden consisting of parasitic, fungal, bacterial, and viral infections in past societies compared to the current lower, viral infection-dominant burden. Studies of current hunter-gatherer societies aim to improve understanding of how these lifestyle and exposure changes impact AD/ADRD risk and resilience.

The Tsimane-Moseten Health and Aging Project

The Tsimane-Moseten Health and Aging Project is a multi-disciplinary effort to investigate the impacts of the exposome, lifestyle, and evolved biology on diseases of aging, including atherosclerosis, heart disease, brain aging, and AD/ADRD. The Tsimane are a Native South American forager-farmer population with high rates of physical activity, high pathogen burden, and low prevalence of obesity. The Moseten are a mixed ethnicity farming population in South America with intermediate physical activity and obesity as well as a high pathogen burden. Compared to those living in the United States, the Tsimane and Moseten people had lower rates of hypertension, total cholesterol, low-density lipoprotein (LDL) cholesterol, and blood glucose. In addition, the Tsimane have high rates of parasitic intestinal infections, poor oral health, and tuberculosis.

Consistent with higher pathogen burden, Tsimane individuals have higher levels of cytokines and leukocytes compared to people living in the United States, indicative of higher inflammation. Based on current understanding of the link between inflammation and atherosclerosis in the United States, the Tsimane people should have high rates of atherosclerosis. However, the Tsimane have far lower rates of coronary calcium plaques (less than 10 percent of those aged 80) compared to those in the United States (55 percent of those aged 80). Based on these results, different types of inflammation may have differential effects on atherosclerosis.

Similarly, heightened inflammation in Tsimane people would suggest high levels of AD/ADRD, based on current understanding of the link between inflammatory processes and AD/ADRD. However, the prevalence of dementia in Tsimane and Moseten individuals is nearly zero before the age of 80. In addition, the rate of age-related brain volume shrinkage is decreased in Tsimane compared to Moseten people and decreased in Moseten people compared to those living in the United States. Notably, Australian Aboriginal people who live a more modern lifestyle do not exhibit a similar, very low prevalence of dementia, suggesting a link between lifestyle and lower rates of dementia rather than simply aboriginal status.

The Tsimane and Moseten cohorts are now helping researchers to understand how different types of inflammation impact disease risk pathways for AD/ADRD and atherosclerosis, as well as the roles that physical activity, low adiposity, and diverse pathogen burden play in promoting disease resilience. Data from these cohorts can also be used to understand certain anomalies, such as differential calcification of coronary and carotid arteries, differences in regional intracranial calcifications in Tsimane and Moseten individuals, and how these anomalies impact AD/ADRD and atherosclerosis risks. To better characterize biological mechanisms that may contribute to dementia resistance in Tsimane and Moseten individuals, the Tsimane-Moseten Health and Aging Project is taking a biological network approach to determine whether biological pathways behave differently in these populations compared to people living in the United States. Dr. Kaplan outlined previous work from Dr. Ken Buetow (Arizona State University) that identified multiple biological networks that enable researchers to distinguish between people with MCI or dementia from normal controls. Dr. Buetow identified an AD network of 169 genes that discriminated between people with MCI/dementia and healthy controls. Roles for these genes include mitochondrial oxidative phosphorylation as well as calcium and insulin signaling.

Future Directions

Dr. Kaplan outlined potential future directions in population-based AD/ADRD research that can leverage populations with unique lifestyles. To better understand the effects of physical activity and infection burden, Dr. Kaplan proposed comparing Tsimane and Moseten populations with subsets of the U.S. population who engage in comparable levels of physical activity or have similarly high infection burdens. To understand how biological systems interact more generally with exposomes and lifestyles and their overall impact on AD/ADRD risk requires comparative, multi-omic datasets collected across the lifespan. Probing these complex datasets will require a mix of theory-guided and discovery-based strategies.

Further diversification of populations in AD/ADRD research, both domestically and abroad, requires respect for data sovereignty as well as community-level investment in return for data and knowledge. To address the issues of data sovereignty, the research community needs to rethink models for data sharing as well as balance the responsibility for protecting indigenous data with open science.

Discussion

Meeting participants discussed the use of existing cognitive and neuropsychological assessments in indigenous populations. Simply translating existing assessments is insufficient, as these assessments are culturally and educationally biased, particularly against those who do not have formal education. However, people in these communities are still learning and addressing cognitive challenges throughout their lives. Dr. Kaplan explained that his study team adjusted these assessments for cultural relevance and the assessment results were scored independently by research teams in Bolivia and the United States. Overall, the diagnosis of dementia was much easier for both research groups to agree upon than the diagnosis of MCI. Other aging studies in rural Chinese and Indian populations have implemented changes to the MMSE to also create more culturally relevant assessments.

Q&A and Discussion

Moderator: Damali Martin, PhD, MPH, Chief, Population Studies and Genetics Branch, NIA

Dr. Martin facilitated a broader discussion about enrolling more diverse populations in AD/ADRD population-based studies. Because current cognitive and neuropsychological tests are biased, Dr. Preeti Pushpalata Zanwar asked how researchers should assess cognition in indigenous and other diverse populations. Notably, common neuropsychological assessments intended to measure a single cognitive domain in fact require performance across multiple cognitive domains. Thus, although hunter-gatherer societies are using their cognitive capabilities, the common neuropsychological assessments may not accurately capture this cognitive activity.

Meeting participants also discussed the relative contributions of exercise, diet, and genetics to AD/ADRD resilience. Dr. Dallas Anderson cited a Lancet Commission Report on modifiable lifestyle factors that can reduce AD, and explained that exercise, followed by diet, are likely larger contributors to AD/ADRD resilience than genetics.

Theme 2: Infrastructure Needs for Current and Future Population Studies

Moderator: Dallas Anderson, PhD, Program Director, Population Studies and Genetics Branch, NIA

Plenary: Whose data is it anyway? Data and resource sharing to advance mechanistic insights into the etiology of Alzheimer’s disease and Alzheimer’s disease related dementias (AD/ADRD)

David Bennett, MD, Rush University

The 1980 Bayh-Dole Act protected ownership rights for federal funding recipients with patents while ensuring that their inventions would be licensed for commercial use in the public interest. As a result, the scientific community has historically considered data collected during federally funded research as property of the scientist who collected the data and the organization that oversaw the data collection. However, the culture of scientific data collection increasingly relies on public databases and open access data, making the question of data ownership significantly less clear. Several federal funding programs now operate under open science principles, including NIH R01 Research Project Grants. Dr. Bennett presented several publicly accessible databases supported by the NIA and other institutions, most of which only require a data use agreement (DUA) to access.

Complexities in Characterizing AD/ADRD

Cognitive decline occurs on a continuum, meaning that discrete labels of whether AD/ADRD are present often fail to accurately capture a patient’s condition. Individuals with AD/ADRD often present highly individualized biomarker expression in their pathologies of cognitive decline (e.g., atherosclerosis, infarcts, amyloid-beta loads)—only 43 percent of variance in cognitive decline can be explained by pathology. The remaining residual cognitive decline results from a spectrum of resilience that can be linked to protein expression. Taken together, the interactions between various pathologies and individual resilience as they contribute to cognitive decline are highly complex.

Open access databases can inform the complex characteristics of late life cognitive decline. In one such case, researchers utilized massive, enriched datasets like the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to develop a gene expression contrastive trajectory inference (GE-cTI) model, estimating timecourse trajectories from non-pathologic to pathologic expression of cognitive decline.

New Open Access Resources for Researchers

Dr. Bennett presented two new data resources available to AD/ADRD researchers in an effort to meet the need for accessible AD/ADRD data: (1) a novel model of human cell lines taken from the Religious Order Study and Memory and Aging Project (ROSMAP), and (2) the Pathology Alzheimer’s and Related Dementias Study (PARDoS), an array of autopsy samples from a massive, diverse sample of individuals across the state of Sao Paulo, Brazil. ROSMAP’s 53 induced pluripotent stem cell (iPSC) lines generated from participants link with previous data from each participant to assist researchers with genetic sequencing and disease modeling. Through PARDoS, researchers will also be able to access autopsy data from 2,516 Latinx individuals (ages 65 – 108), including quantifiable measures of exposomes like pollution and heavy metal accumulation through bone and tissue samples; of note, 43 percent of PARDoS patients are people of color, making it a valuable resource for data on AD/ADRD characterization in non-White populations.
Both datasets will be accessible through the Rush Alzheimer’s Disease Center (RADC).

Recommendations

Dr. Bennett encouraged researchers to expedite discoveries in AD/ADRD therapy by participating in open science efforts; noting that NIA investments in biobanks and repositories are funded by members of the public, Dr. Bennett recommended that the NIA’s Notice of Award (NOA) for future grants stipulate that any data generated using these resources be publicly available as well.

Getting behind the smoke and mirrors on data sharing: Lessons from Framingham and other legacy studies

Rhoda Au, PhD, Boston University

Despite several established databases dedicated to being publicly available resources for AD/ADRD research, the actual accessibility of these databases is often limited. In an assessment of data availability for five large scale legacy cohort studies—the Framingham Heart Study (FHS), Brain Aging Program, ADNI, Washington Heights Inwood Columbia Project (WHICAP), and Atherosclerosis in Risk Communities (ARIC) study—Dr. Au noted two key factors contributing to inaccessible data: (1) lengthy approval processes for data release, and (2) usability issues of data that is often too large in quantity and difficult to filter. Datasets like FHS and ARIC require individuals requesting data to complete a lengthy approval process. While this process is meant to verify that requesters are sufficiently qualified, Dr. Au cautioned that using criteria to determine a researcher’s qualifications might hinder useful research by individuals outside the AD/ADRD field. Even when granted access to datasets, Dr. Au demonstrated examples of conglomerated data spanning multiple pages, emphasizing the time costs of trying to extract useful data from such expansive datasets.

Other issues in data sharing for AD/ADRD research stem from interoperability issues. Dr. Au presented one limitation of existing use case agreements for FHS data: due to use case restrictions, FHS researchers assessing phenotypic data are not permitted to concurrently assess genetic data, restricting the focus of their research topics. In contrast, the Database of Genotypes and Phenotypes (dbGaP) requires researchers using their FHS phenotypic dataset to have a genetics or genomics oriented research focus. Therefore, FHS phenotype data researchers are effectively blocked from accessing the necessary phenotype data, as the only alternative source of FHS phenotype data, the Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC), only stores a portion of the dataset. Dr. Au noted further issues with data sharing as follows:

• Research embargoes have poorly defined parameters on when data must be released to the public, which can be taken advantage of to extend embargoes.
• The use of digital data opens up several new avenues for collecting enriched datasets; however, digital data entails a lengthy de-identification process, and the mechanisms for sharing that data are limited
• Current efforts to harmonize larger databases often shift data from smaller research siloes into bigger ones, illustrating the need for intentional data sharing structures designed for interoperability from the onset.

Discussion

In response to a question raised by Dr. Amy Kind, Dr. Au agreed that databases should be formatted so that information is accessible and easily understood not just by researchers or members of the AD/ADRD field, but also by community stakeholders. Dr. Anna Greenwood noted one digital health repository currently available to the public, though Dr. Au noted these platforms came with unique complexities.

Opportunities and challenges for AD/ADRD research in traditional non-AD/ADRD cohorts

José Luchsinger, MD, Columbia University

Using an AD/ADRD study nested within the fourth phase of the longitudinal Diabetes Prevention Program Outcomes Study (DPPOS), Dr. Luchsinger and colleagues have leveraged extensive phenotypic data, biospecimens, and research infrastructures for diabetic or prediabetic individuals to identify risk factors and determinants of AD/ADRD in this high-risk population. The DPPOS study amasses over 25 years of data on a diverse cohort of over 1,900 participants of an average age of 73 years, including data on peripheral cardiometabolic mediators of disease progression and metformin drug exposure that can be linked with data on typical neuropathologies. The DPPOS AD/ADRD Study includes three major AD/ADRD outcome categories: (1) a fully implemented National Alzheimer’s Coordinating Center (NACC) version 3 Uniform Data Set (UDS) form, (2) AD plasma biomarkers of amyloid, neurodegeneration, and inflammation, and (3) the continuation of legacy neuropsychological tests from Phase 3 of the DPPOS. A subset of participants are additionally undergoing MRI and PET scanning to collect data on brain structure and amyloid burden. Additionally, insulin signaling will be characterized in neuronal extracellular vesicles.

While the DPPOS AD/ADRD Study offers novel applications of non-AD research data, Dr. Luchsinger emphasized that the differences between typical AD research protocols and non-AD research protocols required thoughtful and actionable planning for training and protocol implementation. Existing DPPOS study staff and researchers received additional training and familiarization with AD/ADRD specific research considerations, such as shifting from short yearly visits with participants to longer biyearly visits with occasional neuroimaging. The NACC-UDS neurologic exams were also adapted to better suit the needs of participants and project coordinators—project coordinators across the 25 sites could choose to either read a prepared script or present a prerecorded video with exam instructions, and the neurologic exam was recorded. Videos were uploaded for central reading. Dr. Luchsinger’s team developed an additional mechanism for central adjudicators to note any abnormal findings during the neurologic exam that were actionable. 90 percent of the administered prerecorded neurological exams assessed so far are of excellent quality, and while the team is working on developing a validation study for the neurologic exam, there are plans to publish a paper and manual on this protocol so other researchers can use their DPPOS AD/ADRD prerecorded neurological exam in their own studies.

AD Knowledge Portal

Anna Greenwood, PhD, Sage Bionetworks

To bridge the gap between advances in basic health science research and their clinical application, data sharing requires sufficient curation, standardization, maintenance, and interoperability. However, this is not the reality for the majority of open data—in one assessment, 90 percent of authors who received a data sharing request either declined or did not respond. NIH has historically relied on data coordination centers (DCCs) to facilitate data sharing, including NIA funded DCCs like the AD Knowledge Portal and the Exceptional Longevity Translational Resources (ELITE) Portal. These DCCs coordinate translational consortia and projects through the NIA Division of Neuroscience and Division of Geriatrics and Clinical Gerontology, respectively. Portal resources include data on human specimens, data from preclinical animal models, multiomics data, and imaging data at various processing levels. The AD Knowledge Portal has been cited in over 653 publications and demonstrates use from both consortium and non-affiliated researchers, illustrating its efficacy as a data sharing resource.

Opportunities and Challenges of DCCs

While there are multiple AD/ADRD data repositories, Dr. Greenwood emphasized that interoperability of these repositories was dependent on users finding easily accessible, compatible data across different database systems that could be analyzed within a common environment. The NIH Cloud Platform Interoperability Effort (NCPI) currently seeks to address this by funding projects to establish use cases and strategies for connecting NIH repositories. NCPI identified key technical standards for effective interoperability, including (1) authentication and authorization, (2) data discovery and access, (3) data formats and modalities, (4) data models, and (5) computation. Dr. Greenwood noted additional initiatives aimed at addressing other barriers to effective data sharing, such as directed funding opportunities and efforts to standardize nomenclature. To balance respect for patient privacy with data accessibility needs, the AD Knowledge Portal is adding data use ontology to datasets to provide greater transparency on their use permissions. The portal is also considering using researcher passports to determine accessible datasets or AI-generated synthetic data deemed safe for electronic health records (EHR). While these efforts are commendable, Dr. Greenwood noted that further action is needed to incentivize researchers to share data and to identify key stakeholders who can pioneer efforts to shift research culture towards open science.

Discussion

In response to a question from Dr. Au, Dr. Greenwood noted that Sage Bionetworks is currently creating workflows that could be implemented internationally for linking datasets across countries. Attendees also raised concerns on how to effectively prioritize data harmonization efforts—while new data management sharing policies and dedicated funds hold potential, members were unconvinced that making data harmonization a requirement in NIH RFAs would lead to large scale change. Dr. Chatterjee noted that harmonization networks have worked in other NIH divisions and could be implemented at NIA as well.

Leveraging electronic medical records data and machine learning to better understand Alzheimer’s Disease

Marina Sirota, PhD, University of California-San Francisco

The rise in publicly available data, paired with the lowered costs for multiomics profiling, have created a novel opportunity for innovative new analyses of existing data that can be linked to EHR datasets. Dr. Sirota presented two real world applications for data sharing in drug discovery and disease prediction modeling.

Drug Discovery

Drug development is a cost-intensive process with variable rates of success. Repurposing drugs offers a lower cost solution that also reduces the risks of failure associate with testing. Dr. Sirota’s team developed a computational pipeline comparing gene expression in disease signatures to gene expression assays in a drug database. Using signatures for AD pathology with APOE4/4 status, the team found significant reversal of pathological gene expression in patients treated with bumetanide, a loop diuretic used to treat edema and hypertension by decreasing sodium reabsorption by the kidneys. The team was further able to demonstrate the efficacy of bumetanide treatment by linking it to (1) rescued learning and memory behavior in AD mouse models, (2) EMR data where patients on bumetanide had lower prevalence of AD, and (3) further computational analyses of patient disease measures, such as comorbidity networks, associated diagnoses, and dimensional analysis.

Disease Prediction

Running computational models on massive, enriched datasets has also allowed teams like Dr. Sirota’s to extract features of AD other methods could not identify, such as hyperlipidemia, dizziness, vitamin D deficiency, and osteoporosis. The team used the Scalable Precision Medicine Open Knowledge Engine (SPOKE), a database comprising several other databases, to map clinical phenotypes to associated biological interpretations. By linking research databases for use in cloud computing models, computational tools can inform more significant advances for precision medicine.

NIH Biomedical Research Informatics Computing System (BRICS)

Matthew McAuliffe, PhD, NIH

BRICS is an ongoing effort to make data findable, accessible, interoperable, and reusable by utilizing key desirable characteristics for data from federally funded research. Dr. McAuliffe outlined the major capabilities of the BRICS repository model to support an effective and sustainable data sharing platform:

• Secure and authorized access to data,
• Scalability to deploy on both local servers and the Cloud,
• Implementation of a data dictionary that identifies CDEs, common data models, and any unique data elements,
• Privacy preserving record linkage to aggregate de-identified subject data across time and studies,
• Support for multiple types of processed data,
• Automated access support for data and biospecimen review,
• Core Trust Seal certification to ensure a trustworthy repository,
• Tools for digital data collection, and
• Translation of tool source data to BRICS instance data, keeping consistent with CDEs.

Rather than retroactively harmonizing unstructured research data, BRICS aims to standardize and clean data prior to database submission, thereby simplifying future analyses. The system utilizes several data collection efforts linked by prerequisite requirements for data to be harmonized to CDEs; this saves time and costs during later analyses. BRICS studies also contain descriptive metadata to capture unique notes from study adjudicators that would otherwise be lost during data harmonization. While BRICS supports the integration of multimodal data, Dr. McAuliffe noted that raw genomics data is not yet compatible with the platform.

An ongoing NIA BRICS Pilot Study has demonstrated success with this system with a 1-year data sample of 16,613 participants across four longitudinal aging studies. Collaborative research data from the pilot study can be queried across studies and flexibly filtered, and harmonized data from multiple studies can be downloaded along with a complete data dictionary. Future work on the pilot study will evaluate additional data years from the aging studies.

Q&A and Discussion

Moderator: Magdalena Beran, PhD Candidate, Maastricht University

Magdalena Beran facilitated a general discussion on critical resources to support ongoing data sharing efforts for AD/ADRD research.

Data Infrastructure Needs

Members reiterated that a core concern in addressing data infrastructure needs for AD/ADRD studies was the disconnect between data availability and data accessibility. Dr. Abu Mosa suggested that the issue was primarily cultural, as barriers to access may be perpetuated by overexerting governance structures—for example, a task may need to be completed twice for two separate institutions due to differing governance systems. Dr. Natascha Merten recommended aligning institutions with similar governance structures on data sharing. Dr. Brittany Dugger added that data sharing responsibilities often fall on specific individuals such as PIs, and future dialogue should explore how to better delegate those responsibilities. To relieve the burden on PIs, Dr. Merten proposed that PIs develop restricted subsets of study data with only core study variables for public use, streamlining the amount of documentation required. Ms. Beran suggested that additional staff resources and training could increase the number of individuals qualified to release data, further removing the full responsibility of data documentation from PIs.

Members noted that data literacy issues contribute to barriers in data sharing, where the demand for data is often low when data are difficult to interpret. Drs. Dugger and Merten proposed increased efforts to streamline quality control for data cleaning and data completeness to ensure that shareable data is also usable. Members acknowledged that efforts to standardize and identify relevant CDEs may be limited, as researchers would be unable to fully predict what variables would be prioritized in future research.

Dr. Dugger proposed that database designers (e.g., people who develop EMRs), data collectors, and database users would be key stakeholders to involve in infrastructure development.

Maximizing Use of Existing Data

Members emphasized that data sharing efforts should focus more on utilizing the large amounts of existing unanalyzed data that. Dr. Dugger commended efforts like the DPPOS AD/ADRD study for leveraging existing longitudinal studies to inform data infrastructure; however, she noted that existing tools should also be adapted to sufficiently capture AD/ADRD features of interest. To make existing data more user friendly when being restructured, Dr. Dugger recommended researchers streamline question items when incorporating them into instruments. She additionally proposed that databases adopt models similar to commercial websites, where researchers using one dataset could receive AI-based recommendations for related datasets.

Meeting Summary | Day 2

Richard Kwok, PhD, Program Director, Population Studies and Genetics Branch, NIA

The themes for Day 2 focused on how diverse environmental exposures (i.e., physical, chemical, social, psychosocial, and economic) throughout the life course, and how the timing, duration, and intensity of those exposures influence health outcomes and well-being in older adults. NIA recognizes that additional research is needed to capture how a comprehensive set of exposures, known as the exposome, affects risk and resilience associated with age-related diseases and conditions, including AD/ADRD. Speakers presented on topics including the effects of climate change on aging and age-related cognitive decline, education as a proxy for environmental exposure, exposome relations to biological mechanisms and other omics measures, and linking data from brain banks and biorepositories to the social exposome.

Theme 3: Life Course Environmental Influences on AD Risk

Moderator: Richard Kwok, PhD, Program Director, Population Studies and Genetics Branch, NIA

Plenary: Assessing environmental exposures across the lifespan for epidemiologic studies

Marc Weisskopf, PhD, Harvard University

Measuring environmental exposure has large discrepancies compared to measuring genetics; unlike an individual’s genetics, the environment changes over time, which affects an individual’s length of exposure. In addition, the environment can change an individual’s physiology, which in turn can affect their behavior and change their environmental exposure. The effect of behavioral change on exposure is known as reverse causation and often confounds exposure measures. For example, geographical pollutant estimates may capture the overall population affected by pollutants, but do not account for changes in an individual’s personal behaviors (e.g., time spent outdoors, lifestyle activities) that can significantly affect the magnitude of their exposure. Researchers should thus consider the accuracy and etiological-relevant timing of exposure measures and consider assessing both cumulative profiles that represent average lifespan exposure and longitudinal profiles that can identify high-risk exposure events at specific timepoints.

Dr. Weisskopf presented a conceptual model for exposure-related disease, which outlined the flow of external factors (e.g., emission, media, micro-environment, personal exposure) to internal factors (e.g., absorbed dose, target tissue dose, altered structure and function) that ultimately lead to disease. Along the continuum from external to internal disease factors, the accuracy of personal exposure measurements increases, but this dynamic can come with increased risk of confounding by personal factors that can be hard to measure. Sometimes, more proxy (less individual) exposure measures are more advantageous when used in an epidemiologic exposure-outcome analysis. In addition, biomarkers of exposure, including blood, urine, toenails, bone, and teeth, represent varying exposure timeframes. If an individual has already been diagnosed with a disease, biomarkers and toxicant measures from blood and urine may not provide insights into exposure-disease relationships, whereas bone and teeth can reflect exposure from many years prior to the diagnosis. The Developmental Origins of Health and Disease study found that early life lead exposure was associated with cortical changes in β-amyloid precursor protein 695 (APP695) in older rats; teeth collection from humans can potentially identify similar associations. However, Dr. Weisskopf emphasized the importance of also understanding disease progression, as many physiological changes occur before the time of diagnosis. In this case, a biomarker collected after disease onset could still be useful. Dr. Weisskopf’s group is currently developing a grant proposal to collect opinions from the general public on teeth donation. Findings from this study could help inform methods for increasing teeth donations to measure early-life exposures.

Dr. Weisskopf presented an example of an exposure model for air pollution that can estimate particulate matter (PM) less than 2.5 micrometers in diameter (PM¬2.5) by integrating geographical factors (e.g., distance to roadways, population density) and meteorological data (e.g., wind speed). Such models are less prone to differential errors because they are independent of any individual’s case status. In contrast, individual biomarkers are sensitive to individual disease status and so can be more prone to differential errors. Researchers can decrease the chance of differential errors by implementing prospective study designs with multiple internal measurements over time. However, Dr. Weisskopf emphasized the importance of appropriately matching exposure assessments to the study design and hypothesis.

While prospective studies aim to identify causal relationships between exposure and disease, artificial intelligence (AI) can predict disease risk and progression from large exposure datasets to help researchers understand who is likely to develop a disease or possibly where an individual is in the disease course. However, researchers should exercise caution in inferring causality of any specific factors that go into an AI model to predict disease or disease course, as that is typically not what AI models are designed to do. Furthermore, researchers should train AI on datasets that include a range of races, ethnicities, sex, and gender to avoid introducing biases in AI models.

Considering climate adaptation strategies to slow age-related cognitive decline

Carina Gronlund, PhD, University of Michigan, NIA Climate Change Scholar

Climate change has been described by the World Health Organization as “the biggest health threat facing humanity.” The effects of climate and climate change are location specific; coastal regions experience hurricanes and floods that cause housing displacement, North America experiences wildfires that spread PM, and the Artic experiences flooding, loss of permafrost, and melting sea ice. Dr. Gronlund outlined health conditions associated with extreme weather, including post-traumatic stress disorder (PTSD), social isolation, stroke, and pollutant exposure, all of which can influence the development of dementia later in life. Furthermore, individuals with cognitive impairments are more affected by climate change; for example, they have more difficulty moving to safe locations in a natural disaster compared to healthy individuals.

Dr. Gronlund’s group conducted a study to identify associations between changes in precipitation and cognitive trajectory. Although individuals exposed to high levels of precipitation showed higher baseline cognitive scores, the rate of cognitive decline in these individuals was faster than those exposed to low levels of precipitation. In another study, researchers found that higher levels of extreme heat exposure are correlated with faster rates of cognitive decline among Black individuals compared to White individuals. Although disaggregating data by race and ethnicity is important for understanding the impacts of extreme weather, Dr. Gonlund emphasized the need to understand the factors underlying racial disparities; for example, neighborhood safety can influence whether windows are left open at night for cooling, which may provide an advantage in warm climates. Climate change mitigation creates opportunities to change homes and promote healthy lifestyles and aging.

Dr. Gronlund shared an outline of interventions that can help improve health outcomes in residential areas, such as improving insulation, upgrading appliances, and educating the community on energy efficiency. In addition, neighborhood recreation centers, performing arts theaters, museums, and social organizations can enhance resiliency in extreme weather. Dr. Gronlund’s group aims to expand the understanding of these interventions by investigating whether changing indoor environments can improve sleep and reduce psychosocial stress and depression. Dr. Gronlund suggested that researchers build trust with diverse and underserved communities of interest by incorporating short-term solutions and appropriate compensation, sharing study progress and results, and conducting community-based participatory research.

Aging in a changing climate: Harnessing the exposome to understand social and environmental determinants

Joan Casey, PhD, University of Washington

Studying the exposome can help researchers leverage big data to characterize exposure-response relationships. Certain populations have socioeconomic disadvantages and disproportionate environmental exposures, such as individuals forced to move due to wildfires and who may subsequently experience accelerated cognitive decline. In addition to harmful exposures, researchers should also consider environmental exposures that are protective against disease, such as early-life greenspace exposure. Many exposures are measured as social determinants of health (SDOHs) that are included in exposome models, such as the Public Health Exposome. However, these models rarely reflect how multiple social identities (e.g., race, sexual identity, religion, gender, disability) intersect with each other and with environmental exposures to affect health outcomes.

Health disparities exist within both the context of ADRD and environmental exposure. ADRD incidence is 40 to 100 percent higher in Black Americans compared to non-Hispanic White Americans, and Black Americans are exposed to higher-than-average levels of NO2 in most US states. In addition, Dr. Casey’s group found that as the proportion of Black residents in an area increases, nighttime noise levels also increase, which can affect mental and cardiac health and sleep quality.

Dr. Casey’s group measures historical exposures to better understand how early-life exposures may influence later-life health disparities. The group is working to associate early-life traffic air pollution (TRAP) from the 1940s with data from Health and Retirement Study (HRS) participants. The group developed a novel TRAP metric called “roadiness,” in which the number of street networks and road lengths are summed to estimate street network density; this can then be compared against redlining grids to estimate geographical gasoline consumption as a proxy for TRAP exposure. Preliminary analysis reveals that higher levels of redlining and segregation across the United States in the 1940s is correlated with higher gasoline consumption and potentially higher levels of TRAP exposure.

Wildfires increasingly contribute to PM exposure and are exemplars of how climate change can exacerbate existing exposure disparities and co-occur with other hazards (e.g., housing instability, population displacement, power outages). Dr. Casey’s group has measured multiple metrics of long-term exposure to wildfire PM2.5 and found elevated AD odds ratios for several of these metrics. Dr. Casey suggested that researchers develop indices of overall exposure that capture patterns of hazards co-occurrent with wildfires and that account for a dataset’s spatio-temporal dependencies. Currently, Dr. Casey’s group is measuring wildfire burn perimeters in co-occurrence with public safety power shutoffs to understand how this co-occurrence may affect ADRD individuals’ healthcare utilization.

Proxies for environmental exposures (education as an exposure)

Jennifer Manly, PhD, Columbia University

Any exposome research framework that includes historical and contemporary systems of oppression, such as structural racism, sexism, and xenophobia, must also include education measures. A 2018 study of population-attributable risk found that low education is a contributor to ADRD, especially in non-Hispanic Black and AI/AN populations. Dr. Manly shared multiple studies that associated educational attainment levels with various health disparities, such as disparities in verbal episodic memory among Hispanic individuals, baseline cognition among individuals attending desegrated schools in the Southern United States, cognitive reserve in Black and Hispanic older adults, and later-life memory function in India.

Dr. Manly suggested that researchers assess aspects of educational experiences beyond attainment that drive health outcomes, such as educational quality, educational timing and interruptions, and exposure to racism during schooling. A recent study assessed educational quality based on measures of student-teacher ratio, attendance rate, and term length, and found that lower educational quality was associated with a higher incidence of dementia over a 23-year follow-up period. Another recent study identified dimensions of schooling that were correlated with reduced ADRD risk, including higher numbers of teachers with graduate training and higher overall school quality.

Dr. Manly’s group leverages data from multiple large cohort studies, including the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study, the Project Talent Aging Study, and the National Longitudinal Study of Adolescent to Adult Health (Add Health). Dr. Manly’s group used data from these studies to identify associations between elementary school investments and later-life cognitive impairment risks, numbers of teachers with graduate degrees and ADRD incidence, and achievement of higher SES and overall health. In addition, the Education Study for Healthy Aging Research (EdSHARe) leverages data from New England Centenarian Study (NECS) cohorts and is currently assessing various means of characterizing aspects of educational exposure in high schools, including collecting data on classes offered within schools, proportions of graduates attending college, standardized test scores, and individuals’ grades. Future research on education within the exposomal framework should harmonize exposure measures, recognize the influence of educational policies on later life brain health, include population-feasible biomarkers to determine relative contributions to health outcomes, and embrace changes to cognitive intercept as a path to increasing cognitive life expectancy.

Exposomics: omic-scale analysis of the environment for Alzheimer’s disease

Gary Miller, PhD, Columbia University

The exposome does not solely include measures of environmental influence and exposure on health outcomes, but also corresponding biological and physiological responses throughout the lifespan. A study that assessed polygenic risk scores for genetics in relation to exposome composite scores found that the exposome predicted all-cause mortality more accurately than genetics. If non-genetic factors play a critical role in predicting mortality, then researchers should focus on collecting systematic data on factors that drive the exposome at an individual level.

Dr. Miller’s group focuses on understanding molecular mechanisms of biological change due to environmental exposures. The group uses high-resolution mass spectrometry methods on blood samples to detect exogenous chemicals and endogenous metabolites associated with disease outcomes. For example, the group collected blood samples from individuals with liver disease and found elevations in exogenous chemicals that potentially disrupt endogenous metabolites involved in several biological pathways. The group has also detected elevated metabolites of DDE, a breakdown product of the pesticide DDT, in AD patients and, in the WHICAP cohort, found correlations between elevated levels of the glutathione intermediate cysteinylgylcine and higher PM2.5 exposure in AD patients. These associations allow researchers to focus only on those exogenous chemicals that have effects on disease and health outcomes. Furthermore, researchers can study chemicals associated with disease in animal models to generate additional mechanistic hypotheses.

Dr. Miller shared multiple studies that demonstrate how biomarkers and metabolite measures, in conjunction with subjective clinical measures, can provide insights into mechanisms of disease. The Estudio Familiar de Influencia Genetica en Alzheimer (EFIGA) project has identified differences in metabolite correlations between individuals categorized by AD clinical diagnosis or AD biomarkers. Next, Dr. Miller’s group found changes in several caffeine metabolites that showed high correlation with self-reported caffeine consumption, which is known to be protective against the development of Parkinson’s disease. However, Dr. Miller cautioned that studies measuring metabolomics from blood samples cannot determine when changes in metabolites occur from exposure to exogenous chemicals. Longitudinal studies that collect blood samples at multiple timepoints can help mitigate this issue.

The Neighborhoods Study: Linking brain banks (and other AD biorepositories) to the social exposome

Amy Kind, PhD, University of Wisconsin-Madison

In alignment with traditional multi-level mechanistic disparities theories, such as the social ecological theory or the fundamental causes theory, many social science disciplines leverage quantifiable measures of the social exposome. One of the most commonly leveraged of these is the area deprivation index (ADI), a validated measure of small area multi-domain social determinants of health demonstrated to independently impact health over and above individual-level measures. Social exposome metrics translate across the research and application spectrum with additional use in analyses, recruitment, assessment of generalizability, and policy. Studying the social exposome can help researchers to better understand and intervene upon structural inequities that affect neighborhoods and lead to health disparities.

The ADI uses 17 measures of SDOH across small, population-sensitive areas to produce rank-based indices for every neighborhood in the US including Puerto Rico. These measures are time concordant, strongly linked to health, and can be used with data dating back many decades. Dr. Kind’s group curates the ADI for the US, updating it annually and sharing it freely through the Neighborhood Atlas data democratization tool. The Atlas is designed so all persons regardless of technical background can use and make meaning of the ADI, often through the customized mapping and geo-overlays available on the Atlas. Because of this wide-accessibility and the metric’s underlying rigor, the ADI social exposome data available on the Neighborhood Atlas have been used by tens of thousands of groups, including traditional academic and industry research programs, community groups, the federal government, Congress, professional medical societies, private insurers, state Medicaid programs, and CMS. Milwaukee Water Works has also implemented the Neighborhood Atlas’ ADI to focus their lead water line replacement efforts to ensure an equitable and fair approach to this scarce resource allocation towards the mitigation of lead exposure and improved brain health in Milwaukee’s children. The linkage between living in a neighborhood with high ADI and poorer brain health has now been demonstrated across dozens of studies.

Dr. Kind’s group has conducted several studies that link rich social exposome data to ADRD brain bank tissue data, which is very difficult given the near total lack of social factors available to these deeply biologically phenotyped biorepository data. Dr. Kind leads the Neighborhood Study, a 23-site study which collects, constructs, curates, and analyses life course data and residential history from living individuals and decedents (brain donors) involved with AD Research Centers and HABS-HD. For decedents, Dr. Kind’s group builds residential history frameworks across the lifespan by searching public archives and records to identify important aspects of the decedents’ lives and their critical windows of exposures. A major challenge in constructing residential history is collecting the addresses and other protected health information (PHI) across multiple sites; each state has their own privacy laws, requiring extensive customized data use agreements for each site and enhanced cybersecurity protocols and infrastructure. To address this challenge, the group has developed and is expanding a PHI-capable exposome data core to store PHI data from the Neighborhood Study and subsequently de-identify and link the data to non-PHI NIA core research resources.

Report Back from Breakout Groups and Discussions

Moderator: Damali Martin, PhD, MPH, Chief, Population Studies and Genetics Branch, NIA

Theme 1: Diversifying Populations

Breakout group notetakers summarized discussions that occurred during the Theme 1 Breakout Session. Questions used to guide these discussions are outlined below.

1. What are the current AD/ADRD research priorities for each study population presented during Session 1? Which of these are achievable in the short term versus long term?
2. What are some collaborative opportunities with other research entities (e.g., research consortia, institutes, centers, and funding agencies) to address these research priorities? What unique collaborative opportunities should we consider to enhance cross-cutting research in the field?
3. What are some other important populations for studying aging that were not discussed during Session 1? What are some research priorities corresponding to these populations? What about timeline and collaborative opportunities to address these priorities?

Research Priorities Across All Study Populations

Optimization of Study Design

Design of research studies that involve diverse and large populations should take into consideration the following: (1) the effects of using different recruitment methods for different populations on study results, (2) how to measure diverse experiences of trauma in ways that enable cross-population comparisons, and (3) overall cost-effectiveness. Research studies focused on single ethnoracial groups should consist of research questions tailored to the specific population and based on needs of the community, as well as strategies for exploring within-population heterogeneity.

Increasing study cohort diversity requires additional investment in lower resource areas to improve accessibility to research assessment tools, such as MRI and PET. Accessibility improvements may result in reduced data missingness. Simultaneously, study designs should use a minimally viable protocol that includes only digital and blood-based biomarkers, which are more accessible than other measurement types.

When assessing grant proposals for research on structural racism and AD/ADRD, study sections need to include epidemiologists and sociologists with the expertise necessary to properly evaluate study design.

Assessment Tool Improvement

Existing assessment tools need to be improved to increase cultural and language inclusivity and relevance, and to minimize bias. Notably, straightforward language translations are insufficient, as cultural and language nuances differ across different dialects within the same language. Meeting participants also noted the need for standardized protein-based biomarkers, improved digital assessment tools for sleep and activity measures, and inclusion of environmental pollutant measures.

Meeting participants indicated interest in an NIH toolbox that includes harmonized approaches to measure SDOH and demographic information. The level of prescriptiveness should balance minimization of study participant burden and sufficient capture of information to support U.S.-based and international population-based studies. One strategy to balance burden and sufficient data collection would involve compiling a minimum number of assessment questions with additional optional questions.

Improvement of Recruitment Strategies

To improve study cohort diversity, researchers should seek out collaborations with other successful population-based studies and programs (e.g., Systolic Blood Pressure Intervention Trial), smaller universities, and other organizations (e.g., cancer centers, Clinical and Translational Science Institute) that have already built relationships with underserved communities. NIA could consider developing a platform that enables researchers to quickly find organizations that work with specific underserved populations. Social scientists who already engage with specific populations of interest can also assist study teams with forming connections with communities of interest. In addition, researchers can utilize geographic information system (GIS) tools to identify catchment areas for additional recruitment and outreach efforts.

Researchers should also consider collaborating with current birth cohort studies (e.g., NIEHS Environmental Influences on Child Health Outcomes [ECHO]) that have already recruited from diverse populations. Data from these birth cohort studies can provide crucial information on early life exposures that may be relevant for AD/ADRD later in life. In addition, using a minimal list of questions for continued engagement between the sunsetting of a birth cohort study and the initiation of an aging cohort study can retain study participants throughout their lifespans. Early enrollment and tracking from birth also enable tracking of study withdrawals and developing subsequent strategies to mitigate biases that arise due to differential dropout rates across different populations. Finally, educational outreach strategies should focus on the importance of donating brains for research.

Notably, many community-based recruitment strategies are not scalable outside of the initial community. Instead, a meeting participant suggested using social media to engage people at a larger scale and communicate the overall societal benefits of research study participation. Social media engagement could include precision marketing tools and AI, but use of these tools would raise ethical concerns.

Analysis Method Improvement

Due to the complexity of life course approaches, analysis methods need to account for survivor bias as well as competing risk and hazard ratios that can switch at different periods of life (e.g., after age 80, Black people have longer life expectancies compared to White people). In addition, life course analyses should account for medication and supplement burdens that differ across cultures and change over a person’s lifetime.

Some meeting participants expressed concern about the reductionistic nature of combining datasets and how these actions may remove certain nuances of individual studies. The effects of data harmonization need to be carefully considered when performing cross-study analyses.

Understanding Data Source Caveats

Leveraging existing data sources, such as EHRs, can help reduce costs of conducting large, prospective, population-based studies. However, EHRs often have racial and ethnic inequalities, so any potential AD/ADRD risk factors identified from EHR analyses require follow-up studies. In addition, the participants recruited to prospective studies via smartphone or the internet are likely quite different from those who visit a medical center and have up-to-date EHRs. This difference in self-selection may impact overall study results. Lastly, data linkages to information from other federal agencies, such as the Centers for Medicare and Medicaid Services (CMS), may be incomplete due to a lack of information on certain groups of people (e.g., undocumented immigrants).

Data Accessibility

With the increase in single ethnoracial cohort studies, data platforms need to be both interoperable and capable of siloing datasets from one another. This interoperability capacity should also extend beyond NIH datasets. Researchers can also consider creating an open access, limited subset of data to promote citizen science.

Obtaining Informed Consent in the Age of Artificial Intelligence

In the age of AI/ML, study participant data is more easily re-identifiable and used for unintended purposes. Therefore, researchers need to continue to honor study participant protection and wishes about how their data is used. Researchers should also consider granting data ownership to the groups being studied to honor data sovereignty.

Relevant and Proportionate Compensation

Study participant compensation needs to appropriately reflect the time commitment while not amounting to coercion. Researchers should consider other compensation strategies related to the community of interest’s values, such as scholarships.

Outreach and Communication

NIA can initiate multiple outreach and communication activities to help researchers increase diversity in their population-based studies and cross-study analyses. For example, NIA can publish a blog post outlining ongoing studies of diverse populations to educate researchers on cohort availability. In addition, NIA can help identify the health needs of specific populations by assisting study teams with community outreach and collaborating across federal social programs. Building these community relationships requires significant effort, and t, grants funding population-based studies should account for the time and resources needed to recruit truly diverse populations.

Population-Specific Research Priorities and Potential Collaborations

Meeting participants outlined research priorities for populations presented during Session 1 presentations (i.e., indigenous, resilient, DS, ASD, SGM, and EOAD populations) as well as priorities for other important populations—rural, occupational, veteran, immigrant, international, NHPI, Asian American, and comorbid condition populations. Research priorities and potential collaborations to address these priorities are briefly summarized below.

Indigenous Populations

• Assess potential unique drivers of AD/ADRD, such as rapid culture change, restriction to reservation areas, and overall culture of despair.
• Leverage existing studies to capture cognitive measures from indigenous participants.
• Collaborate with existing indigenous population cohort studies.
• Collaborate with indigenous people living in urban areas to enhance community outreach.
• Engage with indigenous populations to repair broken trust.

Resilient Populations

• Use AD/ADRD biomarkers to screen for additional resilient populations.
• Leverage potential funding from CTSI grants to study biomarkers and general health in resilient populations.

Down Syndrome and Autism Spectrum Disorders

• Study individuals with DS who have slowed dementia progression compared to others with DS.
• Assess AD/ADRD risk and resilience in ASD subpopulations.
• Leverage collaborations with the National Institute of Child Health and Human Development (NICHD) and the National Institute of Mental Health (NIMH).

Sexual and Gender Minorities

• Harmonize SGM information across existing datasets.
• Train research staff in protections for participant privacy.
• Grow the overall culture of acceptance and inclusivity of SGM.
• Consider how viewing sex as a nonbinary attribute affects normalizations that quantify relative risk of AD/ADRD.

Early Onset AD

• Use existing data, including real-world observational data, to slow disease progression as well as improve care and functional outcomes. The Patient-Centered Clinical Research (PCOR) Network has been successful in using real-world data to inform clinical care practices.
• Collaborate with industry partners to design and execute a pilot nutritional index study in sporadic EOAD.

Rural Populations

• Use existing frameworks for recruiting rural populations, such as Practice-Based Research in Pain Through Rural Clinics, to address the lack of academic medical centers in rural areas.
• Recruit ethnoracially diverse populations from different geographic areas to capture intersectionality within rural populations as well as diverse chemical exposures.
• Compare social and environmental factors related to AD/ADRD risk between rural and urban populations.
• Collaborate with the National Heart, Lung, and Blood Institute (NHLBI) to investigate the relationship between farmers’ chemical exposures and their AD/ADRD risk.
• Collaborate with institutions with prior experience engaging rural populations in research.

Occupational Cohorts

• Collaborate with the National Institute for Occupational Safety & Health (NIOSH) to study the effects of occupational hazards—such as chemical exposures encountered by auto- and metalworkers—on AD/ADRD risk.
• Link occupational cohorts to other data sources such as Medicare and the National Death Index.
• Investigate the impact of workplace racial diversity on AD/ADRD risk and resilience.

Veterans

• Investigate the potential link between adverse military conditions and exposures (including PTSD, head trauma, toxins, etc.) and AD/ADRD risk.
• Collaborate with the Department of Veterans Affairs (VA) and other veteran groups to recruit study participants and leverage existing data.
• Leverage novel infrastructures to link DOD records to existing deeply phenotyped biorepositories.

Immigrants

• Investigate how changes in immigrant way of life pre- and post-immigration affects AD/ADRD risk.
• Stratify population-based study data based on wave of immigration, country of origin, and whether the person is a first- or second-generation immigrant.

International Populations

• Address the lack of AD/ADRD research in the country of India by identifying people with MCI who may benefit from future interventions.
• Empower local investigators and build infrastructure within LMICs.
• Determine how to manage data access issues that may arise due to differing data protections in different countries.

Native Hawaiian and Pacific Islanders

• Increase the currently very low enrollment of NHPI individuals in AD/ADRD population-based studies.
• Investigate the effects of NHPI trauma and mental health on AD/ADRD risk.
• Collaborate with VA to access existing data on NHPI individuals.

Asian American Populations

• Increase enrollment of East Asian American communities in AD/ADRD population-based studies.
• Further investigate the potential trauma and depression link to AD/ADRD risk in Vietnamese Americans.

Comorbid Condition Populations

• Study subgroups of people with specific conditions that are often comorbid with AD/ADRD to better understand disease heterogeneity and improve the applicability of population-based study results.
• Collaborate with All of Us to assess hypertension and type II diabetes as risk factors for AD/ADRD.
• Investigate potential AD/ADRD risk and resilience factors, including loss of housing and food insecurity, in people with severe mental illness.
• Study the psychiatric symptoms experienced by people with dementia and a previous history of psychiatric symptoms.

Theme 2: Infrastructure Needs for Current and Future Population Studies

Breakout group notetakers summarized discussions that occurred during the Theme 2 Breakout Session. Questions used to guide these discussions are outlined below.

1. What are the most important data infrastructure needs for population studies of AD/ADRD that are not currently met with existing resources?
2. Are there any existing data resources that can be repurposed to help satisfy these data infrastructure needs, or that are underutilized and could be leveraged more fully?
3. Who are the key stakeholders (e.g., who should be involved in infrastructure development)?

Federal Initiatives towards Interoperability

Meeting participants expressed a need for interoperability between NIH databases, especially for data on legacy population studies, to prevent accumulated research siloes. NIH can collaborate with the NHLBI or the CDC to link federal databases with several other platforms, including census data. Funding opportunities for novel approaches to data linkage should be explored.

Data Standardization

Participants identified several approaches for NIH to better align datasets and increase user accessibility:

• Centralizing data analysis tools such as biostatistics and epidemiology,
• Standardizing nonclinical real-world data,
• Providing guidance to researchers on CDEs, common data sets, and data structures being employed in similar population studies,
• Standardizing AD/ADRD instruments and research protocols,
• Working with insurance companies to prevent cost overlap of for-profit data aggregators,
• Developing a centralized resource with updates on available or soon-to-be available population-based study data,
• Eliminating single-site IRB requirements that often contribute to administrative barriers,
• Standardizing materials transfer agreements across institutions,
• Revising current AD/ADRD data collection protocols to increase data collection timeframes, and
• Collaborating with industry partners to develop guidelines on data sharing technologies and documentation for data processing.

Participants acknowledged that efforts to update CDEs will need to be continually addressed, as the relevance of specific elements will change as new data emerge.

International Data Harmonization

Meeting participants emphasized the need for international harmonization efforts to maintain a long-term focus on data sharing on a global scale. Existing data infrastructure will need to adapt to key considerations for global data sharing, including (1) infrastructure support for foreign investigations, (2) authorization of international visas for data access requests, and (3) integration of data collected in languages other than English. Several participants acknowledged that global data harmonization efforts would be limited by the equal priority of test validity; cognitive tests are often culturally specific and efforts to generalize cognitive testing instruments could impact data quality.

Applications of AI in Data Sharing Infrastructure

AI-driven data sharing informed by statisticians and data managers could resolve the need for investigators to spend a significant amount of time assisting with secondary analyses of their massive datasets. While AI can be a powerful tool for data harmonization, the implementation of AI in data sharing infrastructure needs to be both knowledge-based and data driven. Participants suggested that NIA develop competitive funding opportunities, such as hackathons based on guided questions on AI research, to accelerate data discovery.

Data Accessibility for Non-Researchers

Meeting participants emphasized the need for community and patient oriented data engagement. NIH should take further steps to understand community perspectives on publicly available data, especially creating opportunities for community members to inform the development of research data policy. One participant highlighted demographic considerations for data privacy protections, such as Indigenous Data Sovereignty. Future endeavors to increase community engagement with publicly available data should also focus on scalability, leveraging tools such as online media to scale outreach initiatives.

Updating Existing Data Infrastructure

Rather than develop new systems for data accessibility, participants encouraged revising existing infrastructure to reduce barriers to public data. Proposals for restructuring existing resources included:

• Collaborating with state and local governments to link shareable data,
• Updating data creation and usage considerations to be more user oriented,
• Incentivizing data sharing through trainee grants, competitive funding opportunities, and data analysis mentorships,
• Increasing grant funding for secondary analyses of existing datasets,
• Collaborating with industry partners to inform best practices for inclusive infrastructure, and
• Creating data infrastructure supported offline to assist users in rural or low-connectivity regions.
• Enhancing infrastructure specialized for securely managing protected health information for purposes of future linkage and exposome study.

Theme 3: Life Course Environmental Influences on AD Risk

Breakout group notetakers summarized discussions that occurred during the Theme 3 Breakout Session. Questions used to guide these discussions are outlined below.

1. Which exposures or exposure types should be prioritized in future population studies of AD/ADRD?
2. What are the key unanswered questions about these exposures or exposure types?
3. What are some potential collaborative opportunities to address these research questions?

Exposure Prioritizations for Future Populations Studies

Participants outlined several exposure types and measures that should be prioritized in future populations studies on AD/ADRD, including:

• Indoor and outdoor air pollution, as many lower-income countries’ cooking methods produce harmful pollutants,
• Energy and security as a social construct associated with AD,
• The effect of hurricanes on the trajectory of ADRD progression,
• Noise and light pollution, as well as emerging detection devices,
• Structural racism and sexism as a macro exposure,
• Factors of an overall healthy diet (e.g., nutrition, food preparation, effects on the microbiome),
• Toxicant constituents in water,
• Native language and reasons for bilingualism (e.g., culture, relocation, profession),
• Individual exposure level measures, such as blood, urine, hair, bone, and teeth,
• Neighborhood level exposure measures, such as small level area-based metrics of disadvantage across the life course,
• Other factors within the occupational and social exposome, including veteran-focused exposures,
• Education and the distinction between academic and work or environmental education, and
• Exposure types that are amenable to intervention.

Several participants emphasized the difficulty of prioritizing exposure types and measures because they are all inter-related and their individual and relative contributions to AD and ADRD remain unknown.

Next Steps for Priority Exposures

The integration of Alzheimer's disease (AD) outcome measures and their precursors into well-characterized environmental cohorts, such as birth cohorts, holds significant promise for advancing our understanding of the complex interplay between environmental factors and AD risk. By expanding data collection to include measures of upstream environmental and socioeconomic factors, these cohorts can provide valuable insights into the early-life determinants of AD and related dementias (e.g. developmental origins of health and disease [DOHaD)). Leveraging existing investments in well-characterized cohorts—particularly in understudied populations and those with archived biospecimens and environmental samples collected during key windows of susceptibility—will strengthen scientific collaboration in the environmental health sciences to elucidate complex diseases, such as mild cognitive impairment and dementia. This initiative aligns with the concept of precision environmental health, which focuses on the comprehensive characterization of environmental exposures and their impact on health outcomes, including neurodegenerative diseases like AD. There is a need to work with other partners (such as NIEHS) to develop state-of-the-art tools and technologies to capture relevant environmental exposure data and biological samples, and to foster open-science principles and broad sharing of cohort data and specimens. Furthermore, it is crucial to include diverse and understudied populations in these cohort studies to address environmental health disparities and enhance the overall representativeness of research findings. By leveraging innovative designs and approaches, these well-characterized environmental cohorts can contribute to a more comprehensive understanding of the environmental determinants of AD, ultimately informing strategies for disease prevention and intervention.

Participants discussed key unanswered questions and next steps in measuring and assessing priority exposures and their relationship to aging and AD/ADRD, including:

• Determining whether populations drink directly from the tap or bottled water,
• Defining exposure and conceptualizing culture and environment,
• Determining why individuals are bilingual (e.g., culture, profession),
• Understanding education determinants within schools, families, and friends,
• Exploring relationships between social exposure and resilience,
• Determining when and how interventions can reduce the impact of exposure,
• Establishing standardized tools and harmonizing exposure data assessments,
• Establishing a publicly accessible omics database,
• Providing transparent documentation of detail data generation,
• Integrating genetics to better understand biological mechanisms underlying chemical exposures,
• Utilizing multi-aging cohorts to study the influence of the microbiome, and
• Framing study goals to align with the health priorities of the population of interest.
• Ensuring that modifiable social factors are a focus in exposome study, including but not limited to neighborhood and occupational exposome factors.
• Investing in studies that intervene on the exposome either through direct study intervention or via study of naturally occurring policy changes or social events.
• Conducting studies to determine how, when and whereby resilience to adverse exposome occurs and could be cultivated as part of future interventions.

Opportunities for Collaboration

Participants discussed opportunities for collaboration to explore priority exposure types and address key unanswered questions within the field of exposomics. Several participants suggested establishing working groups to help prioritize exposure measurement practices, develop best research practices, improve data documentation, harmonize existing datasets, and identify CDEs to standardize data collection. These working groups can partner with federal agencies, such as the CDC, NIA, the Agency for Healthcare Research and Quality (AHRQ), and the Environmental Protection Agency (EPA), to establish linkages across public databases and aggregated map data. Next, participants discussed the need to collaborate with junior faculty and early investigators as senior faculty often collaborate within niche research networks. Fostering a research culture of inclusion through transparent grant application processes and mentoring systems can increase equitable research practices and diversify the scientific workforce. Finally, participants emphasized the importance of community engagement during the research process, such as making study results available to the public and collaborating with community leaders to establish trust and garner feedback on study design. Furthermore, improving community engagement can help establish a communication infrastructure and increase study recruitment and participation. The need for investment in intervention and resilience studies was also emphasized.

Other suggested opportunities for collaboration in exposome research include:

• Developing NIH grants that are available across multiple institutes (e.g., NIA and the National Cancer Institute [NCI] to explore the effects of exposures on cancer and aging and with the National Institute of Environmental Health Sciences (NIEHS) on the exposome and other environmental factors and aging and other ICs),
• Partnering with WHO to discuss standards for global communication,
• Establishing international collaborations to help harmonize education levels across countries and diversify the populations under study,
• Fostering collaboration through promoting secure PHI data sharing,
• Increasing funding opportunities to support working groups and community collaborations, and
• Funding collaborations to explore associations between the social exposome and biological variables.