jpad journal

AND option

OR option



D. Appiah1, G. Ashworth2,3, A. Boles3, N. Nair4


1. Department of Public Health, Texas Tech University Health Sciences Center, Lubbock, TX, USA; 2. Department of Pharmacology and Neuroscience, Texas Tech University Health Sciences Center, Lubbock, TX, USA; 3. Garrison Institute on Aging. Texas Tech University Health Sciences Center, Lubbock, TX, USA; 4. School of Medicine. Texas Tech University Health Sciences Center. Lubbock, TX., USA

Corresponding Author: Duke Appiah, Department of Public Health, Texas Tech University Health Sciences Center, 3601 4th Street, STOP 9430 Lubbock, TX 79430. Phone: 806-743-9472. Email:

J Prev Alz Dis 2022;
Published online January 19, 2022,



BACKGROUND: Cardiovascular disease (CVD) and Alzheimer’s disease and related dementias (ADRD) disproportionately affect rural communities. Identifying strategies to effectively communicate CVD risk to prevent these conditions remains a high priority.
OBJECTIVE: We assessed the relation between predicted heart/vascular age (PHA), an easily communicated metric of CVD risk, and mild cognitive impairment (MCI), an early manifestation of ADRD.
DESIGN, SETTING, PARTICIPANTS: Data were from 967 rural West Texas residents aged ≥40 years without CVD at baseline (2009-2012) enrolled in Project FRONTIER, an ongoing, multi-ethnic cohort study on cognitive aging.
MEASUREMENTS: MCI was diagnosed using the standardized consensus review criteria. PHA was calculated using the Framingham CVD risk equation. High excess PHA (HEPHA) was defined as the difference between PHA and chronological age >5 years. Logistic regression models were used to calculate odds ratios (OR) and 95% confidence intervals (CI).
RESULTS: At baseline, the mean age of participants (70% women and 64% Hispanics) was 55 years. Almost 13% had MCI and 65% had HEPHA. After adjusting for socio-demographic and health factors, HEPHA was positively associated with MCI (OR=2.98; 95%CI: 1.72-5.15). Among participants without MCI at baseline who returned for follow-up exam after three years (n=238), a three-year negative change in PHA was seemingly associated with reduced odds for MCI (OR=0.98; 95%CI: 0.96-1.01).
CONCLUSIONS: In this study, PHA was positively associated with MCI, with improvement in CVD risk profile seemingly related to reduced odds for MCI. PHA may provide a low-cost means of communicating CVD risk in rural settings to prevent both CVD and ADRD.

Key words: Cognitive impairment, cognition, cardiovascular disease, epidemiology, rural health, prevention.



Several lines of evidence confirm that Alzheimer’s disease-related dementias (ADRD share several pathophysiological pathways with cardiovascular disease (CVD), such as inflammation, increased oxidative stress, and changes to nitric oxide bioavailability (1). Thus, metabolic and vascular damage may play important roles in the etiology of cognitive impairment leading to dementia (1-4). Accordingly, up to a third of all ADRD cases are attributed to adverse CVD risk factors (1), suggesting that CVD prevention efforts may also be effective in slowing preclinical manifestations of ADRD such as mild cognitive impairment (MCI).
Rural populations are disproportionately affected by chronic diseases, such as CVD and ADRD (5, 6). Additionally, the prevalence of low socioeconomic status, cigarette smoking, hypertension, obesity, leisure-time physical inactivity, and low health literacy is higher among rural than urban populations (5, 7). In rural communities, most population-wide CVD educational prevention strategies targeted at improving heart health knowledge and habits do not consistently achieve success, partly due to the low education and health literacy in these communities, as well as limited access to health education/outreach programs (8-10). With more than 20% of rural populations in the U.S. already over the age of 65 years and bearing the greatest burden of CVD and ADRD risk factors (6) coupled with limited access to healthcare, there is an urgent need to identify and implement cost-effective evidence-based population-wide interventions that have the potential to advance cardiovascular health equity and in turn reduce ADRD, among rural communities.
The inability to understand the concept of risk has been identified as a major reason for the low rate of adoption of healthy behavioral habits among individuals (11, 12), especially those in rural and resource–limited settings (13, 14). An intuitive means of quantifying and communicating CVD risk is the concept of predicted heart/vascular age (PHA), which is the predicted age of a person’s cardiovascular system based on their CVD risk factor profile (15). The comparison of PHA to chronological age has been shown to provide a simple, meaningful, and easily understood measure of CVD risk (16-18), and it elicits more emotional impact to adopt healthy lifestyles especially among younger adults at higher levels of CVD risk than the 10-year absolute risk (12, 19).
Although the PHA is an intuitive and impactful metric of CVD risk in lower-resourced settings, no studies to date have investigated the relation of PHA with MCI in rural settings, especially among ethnic minority populations like Hispanics who bear a greater burden of ADRD among all ethnic groups in the U.S. (20). The first step in informing public health policy makers to adopting a low-cost measure to prevent MCI is to show that a low-cost tool such as predicted heart/vascular age (PHA) can predict MCI. This study aimed to investigate the association of PHA with MCI and evaluated the influence of sociodemographic and lifestyle-related factors on PHA in a multiethnic rural cohort of predominantly Hispanic middle-aged men and women.



Study population

Data for this study were from the Project FRONTIER (Facing Rural Obstacles to healthcare Now Through Intervention, Education & Research), an ongoing cohort study exploring the natural course of chronic disease development and its impact on cognitive, physical, social, and interpersonal functioning using a community-based participatory research approach in a multi-ethnic sample of adults living in rural communities of West Texas. The National Institute of Environmental Health Sciences recommends the use of community-based participatory research for rural research and underserved communities who may not respond well to classic research approaches such as recruitment by random digit dialing or obtaining information by mailed surveys (21). Details of the design and procedures for Project FRONTIER are provided elsewhere (22, 23). Briefly, before participant recruitment, partnerships with the local hospitals and clinics (who conduct the medical examinations and clinical labs as well as provide office space) and community organizations were created. Community recruiters and research personnel recruited potential participants through door-to-door solicitation, flyers, and at community events, churches, and food banks. To be eligible for the study, which begun in 2009, participants had to be 40 years or older and be currently living in four counties in West Texas namely Cochran, Bailey, Parmer, and Hockley. These counties encompass 2,482.07 square miles in total, and all of them have per-capita incomes that are below the national and Texas averages of per-capita income. Demographic characteristics of participants recruited for Project FRONTIER are similar to the demographic characteristics of the four counties from which participants were recruited (23). Two follow-up, exams 3 years apart, have occurred since baseline. All participants provided written informed consent with data collection protocols approved by the Institutional Review Board of Texas Tech University Health Sciences Center (IRB #: L06-028). After informed consent process was obtained, participants completed a detailed interview (in English or Spanish at the discretion of the participant) that included questions regarding demographic information, medical history, and neuropsychological assessments as well as assessment for depressive symptoms. The present analysis was restricted to 1093 participants aged 30–74 years at baseline (the age limits for the Framingham risk score equation). Of this number, the following exclusions were made: participants with prevalent CVD (n=85); participants with dementia (n=5), participants with insufficient data to diagnose MCI due to insufficient or incomplete information from cognitive tests (n=13), and participants with missing values for covariates used in calculating PHA (n=23). This resulted in an analytic sample of 967 men and women.


At every visit, cognitive function was assessed using a full cognitive testing battery. These included the Clinical Dementia Rating Scale; the Mini Mental State Exam (MMSE) to assess global cognitive function; the Trails Making Test (TMTA and TMTB) to measure attention, processing speed and mental flexibility; and detailed neuropsychological testing (e.g. Wechsler Memory Scale Logical Memory and FAS, Animal Naming, Clock Drawing, Boston Naming Test, Exit Interview (EXIT25) (24, 25) and Repeatable Battery for the Assessment of Neuropsychological Status (23)). Diagnosis of MCI was determined using standardized criteria by consensus review consisting of physicians and neuropsychologists who reviewed and analyzed participants’ performance across all tests (24, 26, 27). MCI is often considered as a transitional stage between normal cognition and dementia, during which a person is not demented but has measurable cognitive deficits in some form (28).


PHA was calculated based on Framingham risk score equation for general CVD developed by D’Agostino et al (15). The algorithm included age, sex, systolic blood pressure, treatment for hypertension, current smoking status, diabetes status, HDL cholesterol and total cholesterol. PHA facilitates easier understanding than the concept of risk. In other words, PHA is a person’s CVD risk transformed into the units of age rather than risk. For example, if a 61-year-old woman with risk factors above normal levels has a PHA score of 73, she has a predicted heart age/vascular age equivalent to a 73-year-old woman with normal risk factors (15). Thus, the heart of the 61-year-old woman is predicted to be 12 years older than her chronological age. Excess PHA was calculated as the difference between PHA and chronological age, with a positive value indicating higher CVD risk and a greater burden of CVD risk factors. High excess PHA (HEPHA) was defined as the differences between PHA and chronological age > 5 years (19).


Age, sex, race and ethnicity, education, household income, health insurance, marital status, smoking status, general health, and intake of alcohol were all self-reported. BMI was calculated by dividing participants weight (kg) by height (m2). Blood pressure was measured three times with participants seated after 5 minutes of rest. The average of these values was used for the analysis. Participants were asked to fast before each exam. Fasting glucose, total cholesterol and high-density lipoprotein (HDL) cholesterol were all determined by enzymatic methods. Diabetes was defined by self-report, current treatment for diabetes, fasting blood glucose level ≥126 mg/dL or hemoglobin A1c ≥ 6.5. Other medical conditions/history were evaluated using portions of the CDC Behavioral Risk Factor Surveillance System (BRFSS) questionnaire capturing information on physician diagnosis of hypertension and cardiovascular disease (29).

Statistical analysis

Characteristics of the sample stratified by HEPHA were evaluated using chi-square and t-tests or Wilcoxon rank sum test for categorical and continuous variables, respectively. Logistic regression analyses were used to calculate odds ratios (OR) and 95% confidence intervals (CI) for the association of HEPHA with MCI at baseline (cross-sectional analysis) and three-year change in PHA and incident MCI (longitudinal analysis). For the longitudinal analyses, the sample was limited to participants without MCI at baseline who returned for follow-up exam after three years and had information for estimating PHA (n=238). All models were adjusted for ethnicity, education, income, marital status, health insurance status, alcohol use and general health. Interaction between ethnicity (Hispanic or non-Hispanic participants) and HEPHA (yes or no) on the odds of MCI was tested and found not to be statistically significant. A two-tailed probability value less than 0.05 was considered statistically significant. All analyses were performed using SAS software version 9.4 (SAS institute, Cary, NC) and R software version 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria).



At baseline, the mean age of the 967 participants was 55 (standard deviation: 9.6) years, with 70% of them being women. A greater proportion of participants reported Hispanic ethnicity (64%) compared to 34% reporting being non-Hispanic White and 2% being non-Hispanic Black. Overall, 12.6% of participants had MCI, and the prevalence of HEPHA was 64.8%. The prevalence of MCI was higher among participants who were Hispanic compared to non-Hispanic adults (15.0 vs. 8.5%), higher among participants with no college education compared to those with college education or higher (15.0 vs. 6.3%), higher among adults who were either separated/divorced/widowed (19.5%) or single (18.4%) compared to married participants (10.0%), and higher among participants with total household income less than $10,000 (21.6%) compared to those with household incomes of $10,000 to $40,000 (11.9%) or those with household income of more than $40,000 (6.5%). Characteristics of participants by HEPHA status are presented in Table 1. A greater proportion of HEPHA occurred among Hispanic or non-Hispanic Black participants, as well as participants with lower household income or poor general health. Table 2 shows the distribution of variables (and other CVD risk factors) included in the algorithm for estimating PHA. As expected, participants who had HEPHA were older, had higher systolic blood pressures and higher body mass index, with a greater proportion of them being current smokers, current users of anti-hypertensive medication and having diabetes. There was however, no difference in total cholesterol levels by HEPHA status. Although several differences were found in scores for the various cognitive tests between participants with HEPHA and those without HEPHA, there was no difference in family history of Alzheimer’s disease or the prevalence of physiologic depression by HEPHA status (Table 3). The prevalence of MCI was more than three times higher among participants with HEPHA than those with no HEPHA.

Table 1. The distribution of sociodemographic and lifestyle/behavior factors among adults aged 30-74 years enrolled in project FRONTIER at baseline

Table 2. The distribution of components of the Framingham risk equation among adults aged 30-74 years enrolled in project FRONTIER

* Values are mean (standard deviation) or median (interquartile range) for continuous variables and percentages for categorical variables. HEPHA: High excess predicted heart/vascular age (defined as the differences between PHA and chronological age >5 years)

Table 3. The distribution of cognitive factors among adults aged 30-74 years enrolled in project FRONTIER

CLOX: clock drawing task, MMSE: The Mini Mental State Exam, RBANS: Repeatable Battery for the Assessment of Neuropsychological Status. HEPHA: High excess predicted heart/vascular age (defined as the differences between PHA and chronological age >5 years)


Several sociodemographic and socioeconomic factors were associated with the odds of HEPHA (Figure 1). In multivariable adjusted models (Table 4), a one-year increment in excess PHA was associated with 3% elevated odds for MCI (OR=1.03, 95%CI: 1.01-1.05) while HEPHA was associated with almost threefold elevated risk for MCI (OR=2.98; 95%CI: 1.72-5.15). Although this association was higher among Hispanic participants than non-Hispanic participants, this difference was not statistically significant (OR= 3.29, 95%CI: 1.72- 6.27 vs. OR=2.26, 95% CI:0.82- 6.26, p=0.542). At the first follow-up exam (3 years from baseline), excess PHA was still significantly associated with MCI (OR=1.04, 95%CI: 1.01-1.08) but the elevated odds of MCI associated with HEPHA did not attain statistical significance (OR=1.73, 95%CI: 0.68-4.42). A three-year negative change, that is an improvement in cardiovascular health, was associated with reduced odds of MCI (or=0.98, 95%CI: 0.96-1.01), however, this did not attain statistical significance in the small sample of participants who returned for the follow-up visit.

Figure 1. Adjusted estimates for the association of sociodemographic and socioeconomic factors with high excess predicted heart/vascular age at baseline, Project FRONTIER

Table 4. The association between predicted heart age and mild cognitive impairment among adults aged 30-74 years enrolled in project FRONTIER

* All models were adjusted for ethnicity, education, income, marital status, health insurance, alcohol use and general health.



In this study of community-dwelling rural adults, more than two-thirds of participants were found to have adverse cardiovascular risk profiles as indicated by HEPHA. The prevalence of MCI was almost twice as high in Hispanic participants compared to non-Hispanic participants. HEPHA was associated with almost three-fold elevated odds for MCI, an association that was consistent across ethnicities. Improvement in cardiovascular risk profile over time among a subsample of the cohort appeared to be associated with reduced odds for MCI, however, this did not attain statistical significance.
The first step in informing public health policy makers in adopting this low-cost measure to prevent MCI is to show that PHA is related to MCI. To our knowledge this is the first study to evaluate the relation of PHA with MCI in rural settings, especially among ethnic minority populations like Hispanic adults who bear a greater burden of ADRD among all ethnic groups in the U.S. (20). The prevalence of CVD and ADRD is increasing in the US resulting in a decline in quality of life and considerable mortality. CVD remains the leading cause of mortality in the U.S. with 121.5 million American adults living with CVD (30). Currently, about 5.3 million cases of ADRD are known to occur in the U.S. and this number has been projected to increase to 8.4 million cases by 2030 (31). The increasing prevalence of CVD and ADRD are largely due to aging. By 2030, over 20% of the U.S. population is estimated to be at least 65 years old; the age at which the incidence of CVD and ADRD begins to increase rapidly (32). Rural populations are disproportionately affected by chronic diseases such as CVD and ADRD (5, 6). With more than 20% of rural populations in the U.S. already over the age of 65 years and bearing the greatest burden of CVD and ADRD risk factors (6) coupled with limited access to healthcare, there is a critical and an urgent need to identify and implement cost-effective evidence-based interventions that have the potential to advance cardiovascular health equity and in turn reduce ADRD among rural communities.
Several lines of evidence confirm that ADRD share several pathophysiological pathways with CVD, such as inflammation, increased oxidative stress, and changes to nitric oxide bioavailability (1). Thus, metabolic and vascular damage play important roles in the etiology of cognitive impairment leading to dementia (1-4). Accordingly, up to a third of all ADRD cases are attributed to adverse CVD risk factors (1), suggesting that CVD prevention efforts may also be effective in slowing early manifestations of ADRD, such as MCI. Disturbingly, in rural communities, most population-level CVD educational prevention strategies to improve knowledge, heart health habits, and related skills do not achieve consistent success, partly due to low education and health literacy (8-10). The inability to understand the concept of risk has been identified as a major reason for the low rate of adoption of healthy behavioral habits among individuals (11, 12), especially those in rural and resource–limited settings (13, 14).
A major factor that influences motivation to adapt healthy lifestyles to improve cardiovascular health is the individual’s perception of CVD risk (12, 19). Many CVD risk prediction algorithms which estimate absolute risk are not easily comprehended by individuals who are non-scientists resulting in underestimation of risk (19, 33, 34). Prior studies have demonstrated that PHA, a low-cost means of estimating and communicating CVD risk reduction, is superior in eliciting more emotional impact to adopt healthy lifestyles especially among adults at higher risk of CVD than the 10-year absolute risk (12, 19). Additionally, the comparison of PHA to chronological age has been shown to provide a simple, meaningful and easily understood measure of CVD risk compared to the commonly used 10-year absolute risk (16-18). In the current study, we observed that HEPHA was associated with substantially elevated odds for MCI, a finding that confirm that CVD and ADRD share similar pathways which offers opportunities for prevention of both conditions.
Although correlated with late-life CVD risk factor burden, several reports show that early life CVD risk factors are associated with greater late-life cognitive decline (20, 35, 36). Unlike the current study that looked holistically at the burden of CVD risk, most prior studies looked at the influence of individual CVD risk factors on cognitive function. While a negative change in PHA has been observed to result in favorable changes in cardiovascular health (12), evidence for change in PHA or changes in individual CVD risk burden influencing future cognitive function is limited. A recent report from the Prevention of Renal and Vascular End-Stage Disease cohort evaluated the relation of treatable vascular risk which was defined based on components of the Framingham Risk Score for cardiovascular disease that are amenable to treatment (diabetes mellitus, current smoker status, systolic blood pressure, total cholesterol, HDL cholesterol and use of blood pressure lowering drugs) with changes in cognitive performance over 5.5 years (37). In that large community-based cohort of 3,572 participants, aged 35–82 years, van Eersel et al (37) observed that poor cognitive performance was associated with high treatable vascular risk scores.
In the current study of predominantly midlife adults, favorable changes in CVD burden over three years was associated with reduced odds for MCI. Although this finding did not meet statistical significance in part due to limited statistical power as a result of the small sample with the necessary follow-up data, they do not understate the importance of CVD risk factor burden on future cognitive function. Accordingly, due to the ability of PHA to communicate CVD risk to influences adaption of healthy lifestyles and to improve cardiovascular health especially among young adults, some medical organizations such as the Canadian Cardiovascular Society (38) and the Joint British Societies recommend the use of PHA to communicate lifetime risk and in some cases, inform pharmacological therapy decision making (39). Taken together, the findings from the current study and those of other studies suggest that young adulthood and midlife offer a window of opportunity to provide public health education regarding the importance of cardiovascular health and brain health as well as a time to implement interventions to reduce CVD risk factor burden, and eventually ADRD (35).
This study has notable strengths that include the use of a multi-ethnic cohort of rural community-dwelling adults with extensive assessment of CVD risk factors and cognitive function. The primary limitation of this study is that the sample were obtained from a geographically limited area and therefore the findings of this study may or may not be generalizable to other rural communities in the U.S. Although the sample was multi-ethnic, a greater proportion of them were non-Hispanic White and Hispanic participants. Finally, due to the high attrition of participants during follow-up, the presence of selection bias cannot be ruled out. However, in cross-sectional analyses at the follow-up exam, a positive association was observed between PHA and MCI, similar to those observed at baseline. Therefore, the potential for selection bias influencing the results of the current study, if present, may be minimal.
In conclusion, findings of this study show that PHA is positively associated with MCI, with improvement in CVD risk profile seemingly related to reduced odds for MCI. This study adds valuable information to the expanding evidence of midlife CVD risk factor burden influencing cognitive dysfunction. Therefore, the use of PHA in understanding MCI holds promise of finding a cost-effective and health literacy-appropriate means of communicating risk to prevent CVD and ADRD in rural communities and resource-limited settings. Future short-term, small-scale prevention trials with individuals randomized to receive regular app-based information of their PHA compared to conventional CVD risk communication on cognitive function among rural community-dwelling adults are warranted. Because the burden of CVD, ADRD and their risk factors are common in rural communities where 20% of the U.S. population (60 million persons) are reported to live (5), the potential public health impact of reducing these risk factors will be significant on the health of Americans.


Funding: Project FRONTIER is supported by the Garrison Institute of Aging at Texas Tech University Health Science Center, Lubbock, TX, USA. Dr. Appiah was supported by the Collaborative Seed Grant Program in Aging from the Garrison Institute of Aging. The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript

Conflict of Interest: None.

Ethical Standards: All participants provided written informed consent with data collection protocols approved by the Institutional Review Board of Texas Tech University Health Sciences Center (IRB #: L06-028)



1. Kulshreshtha A, Saini J, German T, Alonso A. Association of Cardiovascular Health and Cognition. Current Epidemiology Reports. 2019;6(3):347-63. doi: 10.1007/s40471-019-00210-8.
2. Samieri C, Perier MC, Gaye B, Proust-Lima C, Helmer C, Dartigues JF, et al. Association of Cardiovascular Health Level in Older Age With Cognitive Decline and Incident Dementia. JAMA : the journal of the American Medical Association. 2018;320(7):657-64. doi: 10.1001/jama.2018.11499.
3. Stampfer MJ. Cardiovascular disease and Alzheimer’s disease: common links. Journal of internal medicine. 2006;260(3):211-23. doi: 10.1111/j.1365-2796.2006.01687.x.
4. Leritz EC, McGlinchey RE, Kellison I, Rudolph JL, Milberg WP. Cardiovascular Disease Risk Factors and Cognition in the Elderly. Curr Cardiovasc Risk Rep. 2011;5(5):407-12. doi: 10.1007/s12170-011-0189-x.
5. Garcia MC, Rossen LM, Bastian B, Faul M, Dowling NF, Thomas CC, et al. Potentially Excess Deaths from the Five Leading Causes of Death in Metropolitan and Nonmetropolitan Counties – United States, 2010-2017. Morbidity and mortality weekly report Surveillance summaries. 2019;68(10):1-11. doi: 10.15585/mmwr.ss6810a1.
6. Abner EL, Jicha GA, Christian WJ, Schreurs BG. Rural-Urban Differences in Alzheimer’s Disease and Related Disorders Diagnostic Prevalence in Kentucky and West Virginia. J Rural Health. 2016;32(3):314-20. doi: 10.1111/jrh.12155.
7. Zahnd WE, Scaife SL, Francis ML. Health literacy skills in rural and urban populations. Am J Health Behav. 2009;33(5):550-7. doi: 10.5993/ajhb.33.5.8.
8. Abbott LS, Slate EH. Improving Cardiovascular Disease Knowledge among Rural Participants: The Results of a Cluster Randomized Trial. Healthcare (Basel). 2018;6(3). doi: 10.3390/healthcare6030071.
9. Ruiz-Perez I, Bastos A, Serrano-Ripoll MJ, Ricci-Cabello I. Effectiveness of interventions to improve cardiovascular healthcare in rural areas: a systematic literature review of clinical trials. Prev Med. 2019;119:132-44. doi: 10.1016/j.ypmed.2018.12.012.
10. Record NB, Onion DK, Prior RE, Dixon DC, Record SS, Fowler FL, et al. Community-wide cardiovascular disease prevention programs and health outcomes in a rural county, 1970-2010. JAMA : the journal of the American Medical Association. 2015;313(2):147-55. doi: 10.1001/jama.2014.16969.
11. Kreuter MW, Strecher VJ. Changing inaccurate perceptions of health risk: results from a randomized trial. Health Psychol. 1995;14(1):56-63.
12. Lopez-Gonzalez AA, Aguilo A, Frontera M, Bennasar-Veny M, Campos I, Vicente-Herrero T, et al. Effectiveness of the Heart Age tool for improving modifiable cardiovascular risk factors in a Southern European population: a randomized trial. European journal of preventive cardiology. 2015;22(3):389-96. doi: 10.1177/2047487313518479.
13. Homko CJ, Santamore WP, Zamora L, Shirk G, Gaughan J, Cross R, et al. Cardiovascular disease knowledge and risk perception among underserved individuals at increased risk of cardiovascular disease. J Cardiovasc Nurs. 2008;23(4):332-7. doi: 10.1097/01.JCN.0000317432.44586.aa.
14. Wiese LK, Williams CL, Tappen RM, Newman D. An updated measure for investigating basic knowledge of Alzheimer’s disease in underserved rural settings. Aging Ment Health. 2019:1-8. doi: 10.1080/13607863.2019.1584880.
15. D’Agostino RB, Sr., Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117(6):743-53. doi: 10.1161/CIRCULATIONAHA.107.699579.
16. Yang Q, Zhong Y, Ritchey M, Cobain M, Gillespie C, Merritt R, et al. Vital Signs: Predicted Heart Age and Racial Disparities in Heart Age Among U.S. Adults at the State Level. MMWR Morbidity and mortality weekly report. 2015;64(34):950-8. doi: 10.15585/mmwr.mm6434a6.
17. Wells S, Kerr A, Eadie S, Wiltshire C, Jackson R. ‘Your Heart Forecast’: a new approach for describing and communicating cardiovascular risk? Heart. 2010;96(9):708-13. doi: 10.1136/hrt.2009.191320.
18. Groenewegen KA, den Ruijter HM, Pasterkamp G, Polak JF, Bots ML, Peters SA. Vascular age to determine cardiovascular disease risk: A systematic review of its concepts, definitions, and clinical applications. European journal of preventive cardiology. 2016;23(3):264-74. doi: 10.1177/2047487314566999.
19. Appiah D, Capistrant BD. Cardiovascular Disease Risk Assessment in the United States and Low- and Middle-Income Countries Using Predicted Heart/Vascular Age. Sci Rep. 2017;7(1):16673. doi: 10.1038/s41598-017-16901-5.
20. Lamar M, Durazo-Arvizu RA, Sachdeva S, Pirzada A, Perreira KM, Rundek T, et al. Cardiovascular disease risk factor burden and cognition: Implications of ethnic diversity within the Hispanic Community Health Study/Study of Latinos. PloS one. 2019;14(4):e0215378. doi: 10.1371/journal.pone.0215378.
21. O’Fallon LR, Dearry A. Commitment of the National Institute of Environmental Health Sciences to community-based participatory research for rural health. Environ Health Perspect. 2001;109 Suppl 3:469-73. doi: 10.1289/ehp.109-1240567.
22. O’Bryant S, Zhang Y, Owen D. The Cochran County aging study: methodology and descriptive statistics. Texas Public Health Journal. 2009;61:5-7.
23. O’Bryant SE, Edwards M, Menon CV, Gong G, Barber R. Long-term low-level arsenic exposure is associated with poorer neuropsychological functioning: a Project FRONTIER study. Int J Environ Res Public Health. 2011;8(3):861-74. doi: 10.3390/ijerph8030861.
24. O’Bryant SE, Johnson L, Reisch J, Edwards M, Hall J, Barber R, et al. Risk factors for mild cognitive impairment among Mexican Americans. Alzheimers Dement. 2013;9(6):622-31 e1. doi: 10.1016/j.jalz.2012.12.007.
25. Johnson LA, Mauer C, Jahn D, Song M, Wyshywaniuk L, Hall JR, et al. Cognitive differences among depressed and non-depressed MCI participants: a project FRONTIER study. Int J Geriatr Psychiatry. 2013;28(4):377-82. doi: 10.1002/gps.3835.
26. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology. 1984;34(7):939-44. doi: 10.1212/wnl.34.7.939.
27. Mild cognitive impairment: Aging to Alzheimer’s disease. Mild cognitive impairment: Aging to Alzheimer’s disease. New York, NY, US: Oxford University Press; 2003.
28. Huang C, Mattis P, Perrine K, Brown N, Dhawan V, Eidelberg D. Metabolic abnormalities associated with mild cognitive impairment in Parkinson disease. Neurology. 2008;70(16 Pt 2):1470-7. doi: 10.1212/01.wnl.0000304050.05332.9c.
29. Menon C, Westervelt HJ, Jahn DR, Dressel JA, O’Bryant SE. Normative performance on the Brief Smell Identification Test (BSIT) in a multi-ethnic bilingual cohort: a Project FRONTIER study. Clin Neuropsychol. 2013;27(6):946-61. doi: 10.1080/13854046.2013.796406.
30. Benjamin EJ, Muntner P, Alonso A, Bittencourt MS, Callaway CW, Carson AP, et al. Heart Disease and Stroke Statistics-2019 Update: A Report From the American Heart Association. Circulation. 2019;139(10):e56-e528. doi: 10.1161/CIR.0000000000000659.
31. Alzheimer’s A. 2015 Alzheimer’s disease facts and figures. Alzheimers Dement. 2015;11(3):332-84. doi: 10.1016/j.jalz.2015.02.003.
32. Ortman JM, Velkoff VA, Hogan H. An Aging Nation: The Older Population in the United States, Current Population Reports, P25-1140. Washington, DC: U.S. Census Bureau; 2014.
33. Robertson RM. Women and cardiovascular disease: the risks of misperception and the need for action. Circulation. 2001;103(19):2318-20.
34. Webster R, Heeley E. Perceptions of risk: understanding cardiovascular disease. Risk Manag Healthc Policy. 2010;3:49-60. doi: 10.2147/RMHP.S8288.
35. Yaffe K, Vittinghoff E, Hoang T, Matthews K, Golden SH, Zeki Al Hazzouri A. Cardiovascular Risk Factors Across the Life Course and Cognitive Decline: A Pooled Cohort Study. Neurology. 2021;96(17):e2212-e9. doi: 10.1212/WNL.0000000000011747.
36. Vintimilla R, Balasubramanian K, Hall J, Johnson L, O’Bryant S. Cardiovascular Risk Factors, Cognitive Dysfunction, and Mild Cognitive Impairment. Dement Geriatr Cogn Dis Extra. 2020;10(3):154-62. doi: 10.1159/000511103.
37. van Eersel MEA, Joosten H, Gansevoort RT, Slaets JPJ, Izaks GJ. Treatable Vascular Risk and Cognitive Performance in Persons Aged 35 Years or Older: Longitudinal Study of Six Years. J Prev Alzheimers Dis. 2019;6(1):42-9. doi: 10.14283/jpad.2018.47.
38. Anderson TJ, Gregoire J, Pearson GJ, Barry AR, Couture P, Dawes M, et al. 2016 Canadian Cardiovascular Society Guidelines for the Management of Dyslipidemia for the Prevention of Cardiovascular Disease in the Adult. Can J Cardiol. 2016;32(11):1263-82. doi: 10.1016/j.cjca.2016.07.510.
39. Board JBS. Joint British Societies’ consensus recommendations for the prevention of cardiovascular disease (JBS3). Heart. 2014;100 Suppl 2:ii1-ii67. doi: 10.1136/heartjnl-2014-305693.



C.K. Lewis1, O.M. Bernstein2, J.D. Grill3,4,5, D.L. Gillen2,3, D.L. Sultzer3,5


1. University of California – Irvine, School of Medicine, Irvine, USA; 2. Department of Statistics, University of California – Irvine, Donald Bren School of Information and Computer Sciences, Irvine, USA; 3. University of California – Irvine, Institute for Memory Impairments and Neurological Disorders (UCI MIND), Irvine, USA; 4. Department of Neurobiology and Behavior, University of California – Irvine, School of Biological Sciences, Irvine, USA; 5. Department of Psychiatry and Human Behavior, University of California – Irvine, School of Medicine, Irvine, USA.

Corresponding Author: Catriona Lewis, University of California – Irvine, School of Medicine, 1001 Health Sciences Rd, Irvine, CA 92617, USA,

J Prev Alz Dis 2022;
Published online January 18, 2022,



Background: There is evidence of relationships between behavioral symptoms and increased risk for Alzheimer’s Disease and/or Alzheimer’s Disease biomarkers. However, the nature of this relationship is currently unknown.
Objectives: To evaluate the relationship between anxiety and depressive symptoms and amyloid-β deposition in cognitively unimpaired older adults, and to assess mediating effects of either objective or subjective cognitive skills.
Design: Cross-sectional analysis of screening data from participants enrolled in the Anti-Amyloid Treatment in Asymptomatic Alzheimer Disease (A4) Study ( Identifier: NCT02008357)
Setting: Data analysis
Participants: 4492 cognitively unimpaired adults, age 65-85, enrolled in the Anti-Amyloid Treatment in Asymptomatic Alzheimer Disease (A4) Study
Measurements: We used linear regression to estimate the associations between amyloid-β standard uptake value ratio (SUVR) and Geriatric Depression Scale (GDS) and State Trait Anxiety Inventory (STAI) scores while adjusting for potential confounding factors as well as for Cognitive Function Index (CFI) or Preclinical Alzheimer’s Cognitive Composite (PACC) scores as possible mediational variables.
Results: 4399 subjects with complete covariates were included (mean age: 71.3, 59% female), GDS ranged 0-13 (mean: 1.0), and STAI ranged 6-24 (mean: 9.9). Amyloid-β SUVR was modestly associated with STAI; mean STAI score was estimated to be 0.275 points higher (95% CI: 0.038, 0.526; p-value = 0.023) for each 0.5-point increase in cortical amyloid-β SUVR. Subjective cognitive decline (CFI) attenuated the relationship between SUVR and STAI, while objective cognitive function (PACC) did not. No statistically significant relationship between SUVR and GDS was observed (p = 0.326).
Conclusions: In cognitively unimpaired adults with low levels of depression and anxiety, cortical amyloid-β deposition is associated with anxiety but not depressive symptoms. Attenuation of this relationship by subjective cognitive difficulties suggests that anxiety may be partly due to such a perception resulting from cortical amyloid-β deposition.

Key words: Amyloid-β, depression, anxiety.



Alzheimer’s disease (AD) features a long preclinical stage in which the pathophysiological process progresses without overt decline in cognitive or functional skills. This preclinical phase is defined by increased amyloid-β burden, which can be detected by positron emission tomography (PET) imaging (1). While symptoms of episodic memory decline have been studied most in early AD, neuropsychiatric symptoms (NPS) such as apathy, depression, and anxiety may also represent the initial clinical presentation of the AD process (2, 3). The Mild Behavioral Impairment syndrome was developed to define the psychiatric and non-cognitive behavioral symptoms that may occur before the onset of memory and function impairment (2).
There is evidence of relationships between behavioral symptoms and increased risk for AD and/or AD biomarkers (4–11). The existing literature seems to support two main hypotheses: 1. early-life depression and anxiety are risk factors for later life neurodegeneration, AD (6, 8, 12) or late-life depression and 2. anxiety and depressive symptoms are an early clinical expression of neurodegeneration (4, 7, 11). It is still debated, however, whether depressive or anxiety symptoms are risk factors or a prodrome of dementia, if these processes are related at all. Singh-Manoux et. al., analyzed data from a 28-year longitudinal cohort study and found no evidence for late-life depression as a risk factor for dementia, concluding that associations between late-life depression and dementia are due to shared etiologies or risks (13). Other evidence suggests that depression may interact with amyloid-β or neurofibrillary tangle burden to promote more rapid cognitive decline in patients with AD (7, 9, 10). Currently, the relationship between anxiety and depressive symptoms and AD pathology is unclear. A better understanding of the relationship between underlying neuropathology and individual psychiatric symptoms could help define the sequence of events in the clinical expression of AD, with implications for improving preventive treatment strategies for AD or late-life psychiatric symptoms.
To better understand the relationship between anxiety and depressive symptoms and preclinical AD, marked by amyloid-β deposition, we utilized screening data from the Anti-Amyloid Treatment in Asymptomatic Alzheimer’s Disease (A4) Study. The A4 study is a randomized placebo-controlled trial designed to test whether treatment with solanezumab, a monoclonal anti-amyloid antibody, can reduce cognitive decline in older adults with preclinical AD. We investigated the hypothesis that the extent of cortical amyloid-β deposition is positively correlated with severity of depressive and anxiety symptoms in these cognitively unimpaired older adults, prior to randomized treatment. We further investigated the hypothesis that, if present, the above associations are mediated by perceived and objective cognitive decline. Finally, because previous studies have shown that white matter hyperintensities and hippocampal volume on MR images may be associated with mood symptoms or the subsequent development of depressive symptoms (4, 6, 14) we also examined the effect of these measures on the observed relationships in A4 participants with elevated cortical amyloid-β.



A4 Study and Participants

The A4 study ( Identifier: NCT02008357) tested the ability of solanezumab treatment to slow the course of cognitive decline in cognitively unimpaired older adults with preclinical AD, as demonstrated by elevated cortical amyloid-β binding on PET imaging. The treatment phase is ongoing for this multisite, randomized, placebo-controlled, 240-week treatment study (15). The study design and measures have been described previously (15, 16). In brief, participants recruited to A4 were between age 65 and 85 inclusive and living independently. The study excluded those with dementia, unstable medical conditions, or substantial depression or anxiety posing possible risk with amyloid-β imaging disclosure, although there were no eligibility cutoff scores on symptom rating instruments (17).
Investigators utilized six screening visits to assess eligibility and collect participant background information. Assessments completed during the screening phase included demographic information, medical history, apolipoprotein ε4 genotype (APOE4), and clinical symptom measures described below. The primary A4 study treatment outcome was the Preclinical Alzheimer Cognitive Composite (PACC) (18), which represents the sum of normalized scores on four component tests: the Mini-Mental State Exam (MMSE), the Digit Symbol Substitution Test, the Logical Memory Delayed Recall Test (LMDR IIa), and the Free and Cued Selective Reminding Test (sum of free and total cued recall; FCSRT). To be eligible for amyloid-β imaging, participants had to have a Clinical Dementia Rating Scale global score of 0, a MMSE score of 25-30, and a LMDR IIa score of 6 to 18. To enroll participants with greater likelihood of elevated cortical amyloid-β and future cognitive decline, the study excluded those with LMDR IIa scores >1.5 SD above age-adjusted norms.
Next, participants underwent PET imaging with 18F-florbetapir to measure cortical amyloid-β binding, as described previously (15, 19). Those with elevated amyloid-β, defined as a standard uptake value ratio SUVR > 1.15 or having SUVR 1.0-1.15 with confirmed expert visual read of elevated amyloid-β, proceeded with MRI and study randomization.
The current work focuses on a subset of 4492 participants in the A4 study who underwent screening 18F-florbetapir PET imaging and had cortical SUVR values. We downloaded the analytic dataset from the LONI site on July 14, 2020. We removed 93 participants with missing covariate values for a final sample size of 4399 in our analyses. Most of the missingness was confined to race/ethnicity (N=61) and comorbidities (N=10).

Study Assessments

Clinical measures

Clinical measures included the PACC cognitive composite score and the Cognitive Function Index (CFI), a 15-item self-rated assessment of perceived decline in cognitive skills over the past year (20). Each CFI item is rated “yes,” “no,” or “maybe” (scored as 1, 0, and 0.5, respectively; possible total score: 0-15). Depressive symptoms were measured with the Geriatric Depression Scale (GDS) (21), a 15-item measure of mood symptoms experienced over the past week and self-rated as “yes” or “no” (scored as 1 or 0, respectively; possible total score: 0-15). Anxiety symptoms were measured using the state items from the State-Trait Anxiety Scale (STAI) (22), a self-assessment of six current anxiety symptoms, each rated “not at all,” “somewhat,” “moderately so,” or “very much” (scored as 1, 2, 3, or 4, respectively; possible total score: 6-24). Participants completed all clinical assessments prior to amyloid-β PET imaging.


18F-florbetapir PET was used to assess mean cortical amyloid-β SUVR in an AD composite that included six regions: frontal cortex, temporal cortex, precuneus, parietal cortex, anterior cingulate, and posterior cingulate (1). SUVR calculations were referenced to the mean activity in the cerebellum. To confirm study eligibility and to provide baseline volumetric data, participants who demonstrated elevated amyloid-β on 18F-florbetapir PET imaging underwent MR imaging. This subset (1238 of the 4399 overall participants with 18F-florbetapir SUVR PET values) was eligible for randomization to solanezumab or placebo.

Statistical Analysis

Descriptive statistics included all 4492 participants who underwent screening and 18F-florbetapir imaging. We summarized continuous covariates by mean (standard deviation) and categorical covariates by count (percent). We stratified the descriptive statistics by low (< 1.15) and high (≥ 1.15) SUVR. Additionally, we examined violin plots of GDS and STAI scores split by SUVR strata of size 0.2. Finally, we fit the following models in the subset of participants (N=4399) with fully observed covariates.
We used linear regression to assess the relationship between amyloid-β deposition SUVR as a continuous variable predictor of interest and GDS and STAI scores as responses. In both models we adjusted for the a priori specified potential confounders of race, ethnicity, gender, age, employment, housing situation, marital status, education level, heavy alcohol use, any smoking use, medical morbidity score, hours of exercise per week, hours of sleep per night, and history of neurological disease. The medical morbidity score was defined as the sum of scores for individual medical illnesses, where we scored individual illnesses as 1, 2, or 3, representing mild, moderate, or severe morbidity, respectively. Heavy alcohol use was defined as drinking an average of 3 or more alcoholic beverages per day. All presented results utilize the robust variance estimator to account for potential deviations from the assumption of homoscedastic errors (23). We presented Wald-based confidence intervals and corresponding p-values for each association of interest.
In our secondary analysis we addressed substantial mood or anxiety symptoms by transforming the continuous responses to binary indicators to assess the difference in the odds of having high versus low test scores. Past literature suggests that a GDS score ≥ 5 is indicative of clinically meaningful depression so we a priori created an indicator for scores of 5 or above (21, 24). There is no standard cut point for the STAI, so we used the sample 75th percentile as the cut-point and created an indicator of a score greater than 12.
To assess if objective (PACC score) or subjective (CFI score) cognitive abilities mediated any association between amyloid-β deposition and depressive or anxiety symptoms, we repeated the previous models while additionally adjusting for these measures and evaluated any changes in the magnitude of the relationship between cortical amyloid-β SUVR and GDS or STAI score.
In exploratory analyses, we investigated if APOE4 status impacted the relationship between cortical amyloid-β SUVR, CFI, and depressive and anxiety symptoms. We fit all models used to assess the relationship with APOE4 with the subset of participants who had non-missing APOE4 values (N=4355). We compared the estimated association between SUVR and GDS and STAI when only adjusting for potential confounding variables, when additionally adjusting for CFI, when additionally adjusting for APOE4, and when adjusting for both CFI and APOE4.
We additionally assessed whether the extent of small-vessel cerebrovascular disease, represented as the total volume of WMH, or hippocampal volume, represented as the hippocampal occupancy score (HOC), mediated the relationship between cortical amyloid-β SUVR and GDS and STAI. The present study utilized MRI measures of hippocampal volume, reflected in the HOC (25), and volume of white matter hypointensities (WMH), as measured on T1- weighted images obtained only in the subset of participants with elevated cortical amyloid-β binding (N=1238), to assess their effects on relationships between 18F-florbetapir SUVR and depression and anxiety symptoms, in an exploratory analysis. We averaged the left and right hemisphere volumes of WMH to obtain a global score. We compared the association between amyloid-β SUVR and GDS and STAI when only adjusting for confounders, also adjusting for WMH, HOC, or adjusting for both WMH and HOC.
As a sensitivity analysis, we repeated the primary analysis of relationships between GDS and STAI scores and amyloid-β SUVR, using amyloid-β SUVR values in two a priori specified cortical regions that are considered most likely related to depressive or anxiety symptoms: the anterior cingulate and the medial orbitofrontal cortex. Additionally, we repeated the linear regression model between SUVR and GDS but omitted the memory item from the total score of the GDS to assess if subjective cognitive decline was driving a relationship between amyloid-β deposition and depression. All analyses were performed using R version 4.0.3.



Descriptive statistics stratified by SUVR are reported in Table 1. The mean age of participants was 71.3 years. 88% of participants were non-Hispanic (NH) white, 71% were college educated, and 59% were female. Mean GDS score was 1.03 (SD = 1.47; range: 0-13) and mean STAI score was 9.94 (SD = 3.11, range: 6-24). 147/4492 (3.3%) of participants had a GDS score of at least 5 and 842/4488 (18.8%) participants had an STAI score above 12. Mean cortical amyloid-β SUVR was 1.09; 27.4% of participants had elevated cortical amyloid-β, with SUVR ≥ 1.15. Participants with cortical amyloid-β SUVR ≥ 1.15 were observed to have higher CFI scores (mean of 2.41 versus 1.96) and lower PACC scores (-0.49 versus 0.19). They also were more likely to be older and NH white compared to participants with SUVR < 1.15. Mean GDS scores for those with amyloid-β SUVR ≥ 1.15 and with SUVR < 1.15 were 1.05 and 1.03, respectively; mean STAI scores were 10.06 and 9.90, respectively.

Table 1. Baseline demographics stratified by SUVR. Continuous covariates are summarized by mean (standard deviation; N missing) and categorical variables are summarized by N (%)


The estimated coefficients for the linear regression between SUVR and GDS are shown in Table 2. There was no statistically significant relationship between cortical amyloid-β SUVR and GDS score in either the unadjusted (p-value = 0.236) or the adjusted model (p-value = 0.326). The estimated regression coefficients for models assessing the association between SUVR and STAI are presented in Table 3. The mean STAI score was estimated to be 0.275 points higher (95% CI: 0.038, 0.512; p-value = 0.023) for each 0.5-unit difference in SUVR, when controlling for potential confounding factors. Violin plots of GDS and STAI scores by SUVR are presented in Figure 1, and estimated coefficients in three strata of SUVR levels are included in Tables 2 and 3. There were no substantial changes to the results in the sensitivity analysis that used amyloid-β SUVR values from the anterior cingulate and frontal cortex subregions as predictors of either GDS or STAI scores. The association between SUVR and the GDS score excluding the memory item was also not statistically significant.

Table 2. Coefficient estimates for the linear regression between SUVR and GDS with adjustment variables

Table 3. Coefficient estimates for the linear regression between SUVR and STAI with adjustment variables


Figure 1. Violin plots of Geriatric Depression Scale (GDS) scores and the state portion of State-Trait Anxiety Inventory (STAI) scores by Standard Uptake Value Ratio (SUVR) strata

The median is marked with a circle and the first and third quartiles are shown with a line inside each violin plot.


The estimates from the logistic regression models of the association between SUVR and having a GDS ≥ 5 and between SUVR and an STAI > 12 were not statistically significant. The odds of having a GDS ≥ 5 were estimated to be 30% lower (95% CI: 0.43, 1.13; p-value = 0.143) for each 0.5-unit difference in SUVR when controlling for potential confounders. The odds of having STAI > 12 were estimated to be 11.4% higher (95% CI: 0.91, 1.36; p-value = 0.287) for each 0.5-unit difference in SUVR when adjusting for confounders. Acknowledging that there is little empiric support for specific GDS and STAI cut off scores, we repeated the logistic regression analyses using a lower threshold for the presence of symptoms, GDS ≥ 3 and STAI ≥ 8. In this post-hoc exploratory analysis, the odds of having a GDS ≥ 3 are estimated to be 14% higher (95% CI: 0.90, 1.45; p-value = 0.285) for each 0.5 unit difference in SUVR when controlling for potential confounders. The odds of having STAI ≥ 8 are estimated to be 32% higher (95% CI: 1.09, 1.59; p-value = 0.004) for each 0.5 unit difference in SUVR when adjusting for confounders.
Additional adjustment for CFI, but not PACC, attenuated the relationship between SUVR and STAI (Figure 2A). The estimate of the SUVR effect on STAI decreased from 0.275 points higher when adjusting for potential confounders compared to 0.05 point higher when also adjusting for CFI. Although the association between SUVR and GDS was not statistically significant, there was a small negative association after adjusting for CFI. Adjusting for PACC did not substantially alter the coefficient estimate in either the GDS or STAI model.

Figure 2. 2A. Forest plots of the estimated association between the standardized uptake value ratio (SUVR) and the Geriatric Depression Scale (GDS) or the state portion of State-Trait Anxiety Inventory (STAI) when including different adjustment variables to assess possible mediation of cognition; 2B. Forest plots of the estimated associations between the standardized uptake value ratio (SUVR) and the Geriatric Depression Scale (GDS) or the state portion of State-Trait Anxiety Inventory (STAI) when adjusting for CFI, apolipoprotein ε4 (APOE4), neither, or both. All models are fit with the 4355 participants who had APOE4 status collected


To assess if APOE4 status changed relationships among SUVR, CFI, GDS and STAI, we compared models additionally adjusting for APOE4. When studying the 4355 participants who had APOE4 results, adding APOE4 into the models did not qualitatively alter the relationship between SUVR and GDS or STAI whether or not CFI was included (Figure 2B).
When comparing models fit on the subset of 1238 participants with elevated cortical amyloid-β, including HOC and/or WMH did not impact the relationship between SUVR and STAI or between SUVR and GDS. In the subset of elevated amyloid-β participants, the mean STAI score was estimated to be 0.47 points higher (95% CI: 0.01, 0.92; p-value = 0.04) for each 0.5-unit difference in SUVR, when controlling for potential confounding factors. A trend relationship was found between GDS and SUVR where the mean GDS score was estimated to be 0.21 points higher (95% Cl: -0.01, 0.42; p = 0.06) for each 0.5-unit difference in SUVR, when controlling for potential confounding factors.



We evaluated the relationships between the extent of cortical amyloid-β deposition and depressive and anxiety symptoms in cognitively unimpaired older adults with low levels of depression and anxiety. Increased amyloid-β burden was modestly associated with increased STAI scores. This finding is consistent with other studies that demonstrated a relationship between elevated amyloid-β levels and increased anxiety symptoms (4, 7, 11) and supports the MBI concept, with anxiety as a potential early correlate of cortical amyloid-β deposition. The range of possible STAI values is from 6 to 24 (sample mean: 9.9); an STAI score of 6 represents no anxiety symptoms. Although the magnitude of the regression coefficient linking amyloid-β to STAI score is small (0.275), it corresponds to 7% of the mean STAI score of the sample, when adjusted for the minimum possible STAI score. The magnitude of the relationship between STAI and SUVR is slightly larger (0.47) in the subset of participants with elevated amyloid-β. The extent of association overall in our study was similar to that seen by Krell-Roesch et al. in a recent study that included both cognitively unimpaired and MCI participants, using PIB PET amyloid imaging and Beck Anxiety Inventory total scores, although in that study the relationship in the cognitively unimpaired subsample was not significant (26). To provide additional context from our study, the association is similar in magnitude to the association between a one-hour decline of sleep per night and STAI score in this study (0.281), but less than the effect of sex (0.685). Having a history of smoking or a past neurological diagnosis also has an association of similar magnitude. The effect demonstrated here did not extend to those with clinically substantial STAI scores, given that no relationship was found in our logistic regression model with STAI > 12 defining clinically meaningful anxiety symptoms. Prior studies have also indicated that associations between amyloid-β and anxiety are modest, and also have noted low depression and anxiety levels in their community samples (11). When we reduced the STAI cutoff score to >8 for the presence of anxiety symptoms (indicating scores of “somewhat” or more on at least two STAI items, such as “I felt upset” or “I was worried”), higher amyloid SUVR was associated with an increased likelihood of anxiety symptoms. This supports our finding from the analysis with anxiety included as a continuous variable, indicates that mild anxiety symptoms are associated with higher cortical amyloid, and suggests that cognitively unimpaired older adults with mild anxiety symptoms may represent an enriched group in the screening process to identify those with preclinical AD.
Our analysis found that participants with higher cortical amyloid-β had higher CFI scores (Table 1) as seen previously (15, 27) and additionally demonstrated that CFI score attenuated the relationship between amyloid-β and anxiety symptoms, suggesting that anxiety symptoms might be partly due to concern for perceived cognitive decline or a direct consequence of cortical amyloid-β deposition. Alternatively, anxiety symptoms may contribute to the perception of cognitive decline. In contrast, objective cognitive performance, assessed by PACC score, did not impact the relationship between amyloid-β and STAI score. Notably, Pietrazk et. al. found that healthy older adults with elevated amyloid-β and elevated anxiety symptoms experienced greater cognitive decline over time compared to their counterparts without anxiety symptoms, suggesting that anxiety interacts with cortical amyloid-β, accelerating the decline in cognitive function (7). Our study did not assess change in cognitive functioning over time and cannot discern if anxiety symptoms are an expression of subjective cognitive complaints resulting from cortical amyloid-β, if amyloid-β deposition independently drives both anxiety symptoms and subjective cognitive impairment, or if anxiety symptoms are promoting amyloid-β deposition in AD.
Our results also suggest that the relationship between amyloid-β and anxiety symptoms are independent of APOE4 genotype. Presence of one or more APOE4 alleles is an important risk factor for the early development of AD and carrier status is associated with elevated amyloid-β deposition in the preclinical state and earlier age of onset of memory decline (15, 28, 29). The results of the present study indicate that the link between cortical amyloid-β deposition and anxiety symptoms is not mediated by APOE4 allele status.
In the subset of participants with elevated amyloid-β who subsequently underwent MRI imaging in the A4 screening process, neither hippocampal volume nor the extent of subcortical white matter hypointensities on T1-weighted images impacted the observed relationship between amyloid-β and anxiety symptoms. However, the extent of T1 white matter hypointensities in this sample, thought to represent small-vessel cerebrovascular disease, was generally mild. It is also possible that the low levels of anxiety and depression in this sample partially masked an effect of cerebrovascular disease or hippocampal atrophy on the relationship between cortical amyloid-β and anxiety symptoms. These findings, however, suggest that while microvascular disease or hippocampal atrophy may contribute to or be a consequence of depression or anxiety over the lifespan, they do not appear to be a significant driving factor linking the AD process to the expression of anxiety symptoms.
Our analysis did not find a significant relationship between cortical amyloid-β deposition and depressive symptoms. Interestingly, in this subset of participants with elevated amyloid-β, we did observe a trend relationship between GDS and SUVR, but it did not reach the threshold of significance and was small in clinical magnitude. This is consistent with existing literature demonstrating either non-significant (4, 7), or small (11, 30, 31) relationships between depression and AD biomarkers. The lack of an observed relationship between depressive symptoms and amyloid-β may be due to low GDS scores among this self-selecting study population of older adults with unique willingness to participate in clinical trial therapy. Additionally, the GDS scale was developed as a screening tool for clinical depression and may not be adequately sensitive to mild depressive symptoms. Furthermore, the study excluded participants with a history of major depressive disorder within the past two years, possibly further contributing to the low level of depression in the sample. Surprisingly, after adjusting for CFI there was a small negative association between amyloid-β deposition and depressive symptoms. We would not expect amyloid-β to be protective against depression for groups of participants with similar subjective memory decline. Such a finding warrants replication in future studies for confirmation and to further develop its basis.
While this analysis benefited from a large sample with carefully defined inclusion/exclusion criteria and detailed assessments, there are limitations. Our analysis is observational in nature and cross-sectional. Therefore, we cannot conclude a causal relationship between amyloid-β and anxiety symptoms nor how this relationship may change over AD progression. As additional findings from the A4 trial emerge, however, future analyses may be able to explore these questions. We have adjusted for variables identified a priori as potential confounders, but we could not account for unmeasured potential confounding factors such as income level, psychological characteristics, or history of cerebrovascular disease, major depressive disorder, or major psychiatric conditions in our analysis. Moreover, we were unable to assess relationships between cortical amyloid-β deposition and other important NPS such as apathy or irritability that may occur early in AD. In addition, the goal of this study was to address relationships between two current individual neuropsychiatric symptoms, depression and anxiety, and cortical amyloid-β deposition, rather than a broad range of more-enduring psychiatric symptoms such as those included in the MBI construct and the MBI-Checklist. Studies evaluating relationships between individual neuropsychiatric symptoms and AD biomarkers interrogate brain-behavior relationships differently from studies evaluating broader symptom clusters over time. The MBI-Checklist can address overall neuropsychiatric symptoms or five symptom classes using the domain subscores and can identify the overall MBI syndrome. However, measures of individual symptoms can help define more specific relationships cross-sectionally or over time, and the MBI-Checklist cannot distinguish some individual symptoms such as anxiety and depression because they are scored in the same domain. Finally, the generalizability of our results is limited by inclusion of a relatively homogenous sample of participants with a unique willingness to participate in a treatment clinical trial and relatively low rates of depression compared to the larger population.
Despite these limitations, this study demonstrates in a large sample of cognitively healthy older adults that amyloid-β deposition is associated with increased anxiety symptoms, and that this relationship is attenuated by subjective cognitive difficulties. The study contributes to a growing understanding of NPS in early AD and the interacting pathophysiological pathways that may underlie their expression. Further studies investigating the progression of AD and NPS in this population will further elucidate the complex relationships among amyloid-β deposition, other specific pathologies, NPS, and cognitive decline in the AD process.


Author Contributions: Catriona Lewis, Olivia Bernstein, and David Sultzer conceived of the presented idea and designed the study. Olivia Bernstein and Daniel Gillen designed and executed the statistical analysis. All authors analyzed the data and contributed to the interpretation of the results. Catriona Lewis and Olivia Bernstein wrote the manuscript and designed the figures with significant input and feedback from Joshua Grill, Daniel Gillen, and David Sultzer.

Funding: Catriona Lewis was supported by University of California – Irvine, MIND & University of California – Irvine, School of Medicine Summer Research Mentorship Program. Olivia Bernstein was supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-183928 and the ARCS foundation. Joshua Grill, Daniel Gillen, and David Sultzer are supported by NIA AG066519. The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review of approval of the manuscript. The A4 Study is a secondary prevention trial in preclinical Alzheimer’s disease, aiming to slow cognitive decline associated with brain amyloid-β accumulation in clinically normal older individuals. The A4 Study is funded by a public-private-philanthropic partnership, including funding from the National Institutes of Health-National Institute on Aging, Eli Lilly and Company, Alzheimer’s Association, Accelerating Medicines Partnership, GHR Foundation, an anonymous foundation and additional private donors, with in-kind support from Avid and Cogstate. The companion observational Longitudinal Evaluation of Amyloid Risk and Neurodegeneration (LEARN) Study is funded by the Alzheimer’s Association and GHR Foundation. The A4 and LEARN Studies are led by Dr. Reisa Sperling at Brigham and Women’s Hospital, Harvard Medical School and Dr. Paul Aisen at the Alzheimer’s Therapeutic Research Institute (ATRI), University of Southern California. The A4 and LEARN Studies are coordinated by ATRI at the University of Southern California, and the data are made available through the Laboratory for Neuro Imaging at the University of Southern California. The participants screening for the A4 Study provided permission to share their de-identified data in order to advance the quest to find a successful treatment for Alzheimer’s disease.

Acknowledgments: We would like to acknowledge the dedication of all the participants, the site personnel, and all of the partnership team members who continue to make the A4 and LEARN Studies possible. The complete A4 Study Team list is available at:

Disclosures/Conflict of Interest: Dr. Sultzer has received research support from NIH and Eisai, has participated as a paid member of a DSMB or adjudication committee with Acadia, Avanir, Janssen, and Otsuka, and has received consulting fees from Avanir. Dr. Grill reports research support from NIH and is site Co-Investigator of the ongoing A4 study at University of California – Irvine and has received consulting fees from SiteRx, Cogniciti, and Flint Rehab in the last 36 months. Dr. Gillen reports research support from the Alzheimer’s Disease Research Center, University of California – Irvine and from the NIH. Dr. Gillen has also received consulting fees from Eli Lilly, ChemoCyntrix, FibroGen, GlaxoSmithKline, ProventionBio, and Biom’Up and participated as a paid member of a DSMB or advisory board with Pfizer, Biomarin, Novo Nordisk, Novartis, Amgen, Celgene, CRISPR, AstraZeneca, Merck Serano, Array, Seattle Genetics, Genentech/Roche, UCB, Acerta, Juno Therapeutics, Medivation. Other authors report no potential conflicts with any product mentioned or concept discussed in this article.



1. Johnson KA, Sperling RA, Gidicsin CM, et al. Florbetapir (F18-AV-45) PET to assess amyloid burden in Alzheimer’s disease dementia, mild cognitive impairment, and normal aging. Alzheimers Dement. 2013;9(5S):S72-S83. doi:10.1016/j.jalz.2012.10.007
2. Ismail Z, Smith EE, Geda Y, et al. Neuropsychiatric symptoms as early manifestations of emergent dementia: Provisional diagnostic criteria for mild behavioral impairment. Alzheimers Dement. 2016;12(2):195-202. doi:10.1016/j.jalz.2015.05.017
3. Creese B, Brooker H, Ismail Z, et al. Mild behavioral impairment as a marker of cognitive decline in cognitively normal older adults. Am J Geriatr Psychiatry. 2019;27(8):823-834. doi:10.1016/j.jagp.2019.01.215
4. Banning LCP, Ramakers IHGB, Köhler S, et al. The association between biomarkers and neuropsychiatric symptoms across the Alzheimer’s disease spectrum. Am J Geriatr Psychiatry. 2020;28(7):735-744. doi:10.1016/j.jagp.2020.01.012
5. Zhao Q-F, Tan L, Wang H-F, et al. The prevalence of neuropsychiatric symptoms in Alzheimer’s disease: Systematic review and meta-analysis. J Affect Disord. 2016;190:264-271. doi:10.1016/j.jad.2015.09.069
6. Bennett S, Thomas AJ. Depression and dementia: cause, consequence or coincidence? Maturitas. 2014;79(2):184-190. doi:10.1016/j.maturitas.2014.05.009
7. Pietrzak RH, Lim YY, Neumeister A, et al. Amyloid-β, anxiety, and cognitive decline in preclinical Alzheimer Disease: A multicenter, prospective cohort study. JAMA Psychiatry. 2015;72(3):284. doi:10.1001/jamapsychiatry.2014.2476
8. Kim HK, Nunes PV, Oliveira KC, Young LT, Lafer B. Neuropathological relationship between major depression and dementia: a hypothetical model and review. Prog Neuropsychopharmacol Biol Psychiatry. 2016;67:51-57. doi:10.1016/j.pnpbp.2016.01.008
9. Babulal GM, Chen S, Williams MM, et al. Depression and Alzheimer’s disease biomarkers predict driving decline. J Alzheimers Dis. 2018;66(3):1213-1221. doi:10.3233/JAD-180564
10. Ehrenberg AJ, Suemoto CK, França Resende E de P, et al. Neuropathologic correlates of psychiatric symptoms in Alzheimer’s disease. J Alzheimers Dis. 2018;66(1):115-126. doi:10.3233/JAD-180688
11. Krell-Roesch J, Lowe VJ, Neureiter J, et al. Depressive and anxiety symptoms and cortical amyloid deposition among cognitively normal elderly persons: The Mayo Clinic Study of Aging. Int Psychogeriatr. 2018;30(2):245-251. doi:10.1017/S1041610217002368
12. Ownby RL, Crocco E, Acevedo A, John V, Loewenstein D. Depression and risk for Alzheimer disease: Systematic review, meta-analysis, and metaregression analysis. Arch Gen Psychiatry. 2006;63(5):530. doi:10.1001/archpsyc.63.5.530
13. Singh-Manoux A, Dugravot A, Fournier A, et al. Trajectories of depressive symptoms before diagnosis of dementia. JAMA Psychiatry. 2017;74(7):712-718. doi:10.1001/jamapsychiatry.2017.0660
14. Teodorczuk A, O’Brien JT, Firbank MJ, et al. White matter changes and late-life depressive symptoms: Longitudinal study. Br J Psychiatry. 2007;191(3):212-217. doi:10.1192/bjp.bp.107.036756
15. Sperling RA, Donohue MC, Raman R, et al. Association of factors with elevated amyloid burden in clinically normal older individuals. JAMA Neurol. 2020;77(6):735. doi:10.1001/jamaneurol.2020.0387
16. Sperling RA, Rentz DM, Johnson KA, et al. The A4 study: Stopping AD before symptoms begin? Sci Transl Med. 2014;6(228):228fs13. doi:10.1126/scitranslmed.3007941
17. Grill JD, Raman R, Ernstrom K, et al. Short-term psychological outcomes of disclosing amyloid imaging results to research participants who do not have cognitive impairment. JAMA Neurol. 2020;77(12):1504. doi:10.1001/jamaneurol.2020.2734
18. Mormino EC, Papp KV, Rentz DM, et al. Early and late change on the preclinical Alzheimer’s cognitive composite in clinically normal older individuals with elevated amyloid β. Alzheimers Dement J Alzheimers Assoc. 2017;13(9):1004-1012. doi:10.1016/j.jalz.2017.01.018
19. Pontecorvo MJ, Arora AK, Devine M, et al. Quantitation of PET signal as an adjunct to visual interpretation of florbetapir imaging. Eur J Nucl Med Mol Imaging. 2017;44(5):825-837. doi:10.1007/s00259-016-3601-4
20. Amariglio RE, Donohue MC, Marshall GA, et al. Tracking early decline in cognitive function in older individuals at risk for Alzheimer disease dementia: The Alzheimer’s disease cooperative study cognitive function instrument. JAMA Neurol. 2015;72(4):446. doi:10.1001/jamaneurol.2014.3375
21. Yesavage JA, Sheikh JI. Geriatric depression scale (GDS): recent evidence and development of a shorter version. Clin Gerontol. 1986;5(1-2):165-173. doi:10.1300/J018v05n01_09
22. Marteau TM, Bekker H. The development of a six-item short-form of the state scale of the Spielberger State—Trait Anxiety Inventory (STAI). Br J Clin Psychol. 1992;31(3):301-306. doi:
23. White H. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica. 1980;48(4):817-838. doi:10.2307/1912934
24. Lyness JM. Screening for depression in elderly primary care patients: A comparison of the Center for Epidemiologic Studies-Depression Scale and the Geriatric Depression Scale. Arch Intern Med. 1997;157(4):449-454. doi:10.1001/archinte.157.4.449
25. Heister D, Brewer JB, Magda S, Blennow K, McEvoy LK. Predicting MCI outcome with clinically available MRI and CSF biomarkers. Neurology. 2011;77(17):1619-1628. doi:10.1212/WNL.0b013e3182343314
26. Krell-Roesch J, Syrjanen JA, Rakusa M, et al. Association of cortical and subcortical β-amyloid with standardized measures of depressive and anxiety symptoms in adults without dementia. J Neuropsychiatry Clin Neurosci. 2021;33(1):64-71. doi:10.1176/appi.neuropsych.20050103
27. Amariglio RE, Sikkes SAM, Marshall GA, et al. Item-level investigation of participant and study partner report on the cognitive function index from the A4 Study screening data. J Prev Alzheimers Dis. Published online 2021:1-6. doi:10.14283/jpad.2021.8
28. Corder EH, Saunders AM, Strittmatter WJ, et al. Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer’s disease in late onset families. Science. 1993;261(5123):921-923. doi:10.1126/science.8346443
29. Lim YY, Kalinowski P, Pietrzak RH, et al. Association of β-amyloid and apolipoprotein E ε4 with memory decline in preclinical Alzheimer disease. JAMA Neurol. 2018;75(4):488. doi:10.1001/jamaneurol.2017.4325
30. Harrington KD, Gould E, Lim YY, et al. Amyloid burden and incident depressive symptoms in cognitively normal older adults: Amyloid burden and depressive symptoms. Int J Geriatr Psychiatry. 2017;32(4):455-463. doi:10.1002/gps.4489
31. Krell-Roesch J, Syrjanen JA, Mielke MM, et al. Association between neuropsychiatric symptoms and functional change in older non-demented adults: Mayo Clinic Study of Aging. J Alzheimers Dis JAD. 2020;78(3):911-917. doi:10.3233/JAD-200764



A. Shah1, O. Ysea-Hill2, A. Torres-Morales3, C.J. Gomez2, A. Castellanos2, J.G. Ruiz2,3,4


1. Memorial Healthcare System, Hollywood, Florida, USA; 2. Miami VA Healthcare System Geriatric Research, Education and Clinical Center (GRECC), Miami, Florida, USA; 3. University of Miami / Jackson Health System, Miami, Florida, USA; 4. University of Miami Miller School of Medicine, Miami, Florida, USA

Corresponding Author: Jorge G. Ruiz, MD, Associate Director for Clinical Affairs, Miami VA Healthcare System, Geriatric Research, Education and Clinical Center (GRECC), GRECC (11GRC), Bruce W. Carter Miami VAMC, 1201 NW 16th Street, Miami, Florida 33125, USA, Telephone: (305) 575-3388 / Fax: (305) 575-3365, E-mail:, ORCID: 0000-0003-3069-8502

J Prev Alz Dis 2022;
Published online January 18, 2022,



Evidence suggests that dementia can be prevented. Patients with frailty may be particularly at risk for cognitive impairment (CI). The aim of this study was to determine dementia risk in older Veterans and whether the risk varies according to frailty status. We also evaluated the feasibility of mailed dementia risk screening. Participants were mailed a questionnaire and the Self-Administered Gerocognitive Examination (SAGE). High dementia risk was defined as having mild cognitive impairment (MCI) on SAGE or a CAIDE score ≥6. Out of 5,432 mailed surveys, we obtained a response rate of 19.75%. Most responders completed the questionnaire items. We identified a total of 689 (75.9%) subjects to be at high risk for dementia. Individuals with frailty were at a greater risk for dementia when compared to robust individuals OR:1.921 (95%CI:1.259-2.930), p=.002. The mailed screening represents a convenient, alternative and scalable approach to screen for dementia risk.

Key words: Dementia risk, mild cognitive impairment, dementia.



Most types of dementia are secondary to neurodegenerative disorders which are generally deemed progressive and incurable (1). However, growing evidence suggests the possibility that dementia may be prevented by targeting modifiable risk factors (2-6). A group at particular risk for dementia is individuals with baseline frailty, a state of vulnerability to stressors due to multisystemic dysfunction, which is associated with poor clinical outcomes (7). Epidemiological evidence demonstrates that these conditions often coexist with over half of patients with frailty having concurrent cognitive impairment. In longitudinal studies the onset of frailty may lead over time to dementia (7). Identifying frailty may lead to a more targeted screening of these patients for dementia risk.
Screening for dementia risk will lead to the early identification of those patients at higher risk, allowing clinicians to implement early interventions addressing modifiable risk factors. Practical approaches to increase dementia risk screening are needed to ensure successful population-based implementation. The study aim was to determine the risk of dementia in community-dwelling older Veterans and whether the risk varies according to frailty status. A secondary aim was to ascertain the feasibility of administering dementia screening using a mailed questionnaire.



Study Design, Setting and Participants

This cross-sectional study was conducted at a government-run, US Veterans Health Administration (VHA) Medical Center, an integrated healthcare organization in the US Southeast. Participants included 5,432 community-dwelling Veterans 50 years and older who had at least 2 visits from July 2018-June 2019. Patients with baseline diagnosis of cognitive impairment (MCI or dementia) were excluded. A battery of questionnaires, including socio-demographic, education and exercise surveys, and the SAGE test (Self-Administered Gerocognitive Examination) were mailed to the participants’ addresses on file from July, 2019 to April, 2020. This study was reviewed and approved as a quality improvement project by the Miami VA Healthcare System IRB, therefore, informed consent was not required.

Outcomes and Measurements

Dementia Risk

For the purposes of this study, we defined dementia risk as either having a high score in the CAIDE or mild cognitive impairment (MCI) in the SAGE.

Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE)

This validated risk score predicts the risk of late-life dementia. The CAIDE Dementia Risk Score was validated in a multiethnic American population, similar to our participants, and has been used for participant selection in several multi-domain lifestyle trials for preventing cognitive decline (8, 9). The CAIDE variables include age, hypertension, hypercholesterolemia, physical inactivity, obesity and educational level (8). The information regarding education and physical activity was collected from the mailed questionnaires whereas age, gender, systolic blood pressure (SBP), body mass index (BMI) and total cholesterol were retrieved from the electronic health record (EHR) within one year of their second visit. The scores range from 0 to 15. In this study, a CAIDE score of ≥6 indicated a high risk for dementia. In a large population-based, cohort study, a CAIDE score ≥6, demonstrated a higher probability of dementia in 20 years (8).

Mild Cognitive Impairment

The Self-Administered Gerocognitive Exam (SAGE) is a validated instrument that evaluates the cognitive domains of orientation, language, memory, executive function, calculation, abstraction, and visuospatial abilities. The SAGE has been used in community mild cognitive impairment and dementia screening in the USA. It is self-administered, brief and has good sensitivity (79%) and specificity (95%) at detecting MCI. The SAGE is a 12-item, self-administered test that takes 10-15 minutes to complete and whose scores range between 0 and 22 points (10). There are four alternative versions of the SAGE designed to avoid testing effects (11). In this study, we used the SAGE form 1. SAGE scores between 17 and 22 indicate normal cognition, 16 and 15 MCI, and 14-or-less suggest the presence of dementia [10]. The SAGE has a 79% sensitivity and 95% specificity in determining cognitive impairment (10).


We used the 31-item VA-FI (Table 1) which is based on a deficit accumulation conceptual framework. The items belong to five major domains: morbidity, function, sensory loss, cognition and mood, and other (miscellaneous items). The VA-FI is calculated by adding up the number of deficits obtained from the patient and dividing by 30 (total number of health deficits). For each variable, the presence of the deficit was scored as 1 whereas its absence was scored as 0. Participants were then categorized as robust (<0.10), pre-frail (0.10-0.20) and frail (≥0.21) based on previously published cut-off points (12). For each patient, the frailty index was automatically generated from EHR data on the date of the geriatric primary care visit between September 1, 2019 and May 31, 2020.

Table 1. Participant Characteristics

SD= standard deviation; n= number of participants; BMI= body mass index; VA-FI= VA Frailty Index; MCI= mild cognitive impairment. The assigned superscript letters (a, b or c) are representative of the low risk, high risk (CAIDE ≥6) and high risk (MCI) groups. If a pair of values is significantly different, the values have different subscript letters assigned to them. If a pair of values are not significantly different, the values will have the same superscript letters assigned to them. Data without superscripts is not significantly different between dementia risk groups. Significant differences are in bold (p<.05)


Research Procedure

We mailed 10,152 packages in 2 waves. During wave 1, 5,432 packages were mailed from July 8th, 2019 through January 17th, 2020. Those participants failing to respond within 2 months of the first correspondence (n=4,720), received a second package between October 10th, 2019 through April 4th, 2020. Each package included: 1) Cover letter; 2) SAGE; 3) Questions on physical activity and education; and 4) Stamped self-addressed envelope. All responses received on or before May 31st, 2020 were included in the study. The SAGE was manually scored, and the information was recorded on the dataset. Patients with SAGE scores were ≤ 14 (consistent with dementia) were excluded.

Data Analysis

We reported descriptive sociodemographic characteristics. Categorical variables were presented as frequency (percent), and continuous variables as mean ± SD. Normality of the distribution was checked using the Kolmogorov-Smirnov test. For categorical and continuous variables, Pearson chi-square and Mann-Whitney tests were used, respectively. In addition, we used binomial logistic regression to calculate the association between frailty (independent variable) and the risk of developing dementia (dependent variable), after adjusting for race, ethnicity, marital status, smoking history, alcohol and substance abuse, obstructive sleep apnea and use of anticholinergic drugs. We determined feasibility by calculating the number and proportion of participants who completed all the items in the demographic survey and the SAGE questionnaire. Any questions left unanswered on SAGE were considered incorrect as per test instructions. All analyses were performed using SPSS 26.0 for Windows (SPSS, Inc., Chicago, Illinois). All statistical tests were two-tailed and statistical significance was assumed for a p-value <0.05.



Participant Characteristics

We received 1073 responses (19.8%), 165 had a positive screen for dementia (SAGE score of ≤ 14). The remaining participants were 908 Veterans, 95.5% males, mean age 68.27 (SD=8.36, range 51-96) years, 59.9% Caucasian, 68.5% non-Hispanic and 54.4% married (Table 1). As compared with non-responding individuals, responders were more likely to be older, Caucasian, Hispanic, and married. There were no differences in terms of gender or frailty status between responders and non-responders (Supplementary Table 1).

Dementia Risk

Most participants were at high risk for dementia (n = 689, 75.9%): 127 (18.4%) had MCI, and 562 (81.6%) had a CAIDE score of ≥6. Other than a higher mean VA-FI score in those with MCI (0.22, SD=0.11) as compared to the CAIDE score ≥6 subgroup (0.19, SD=0.10), p<001, the high dementia risk subgroups were not significantly different in their baseline characteristics. When compared to participants at low risk for dementia, patients in the high-risk group had lower levels of education, higher BMI and were more likely to be frail. (Table 1).

Dementia Risk and Frailty

The proportions of robust, pre-frail and frail Veterans were 25.9% (n=235), 38.7% (n=351) and 35.5% (n=322), respectively. The proportion of robust patients at high risk for dementia (66.4%, n=156) was significantly lower than the pre-frail (76.9%, n=270) and frail groups (81.7%, n=263), p<.001. Pre-frailty and frailty were significantly associated with an increased risk of dementia when compared to the robust status, adjusted OR:1.586 (95%CI:1.083-2.322), p=.018 and OR:1.921 (95%CI:1.259-2.930), p=.002, respectively (Table 2).

Table 2. Adjusted* Odd Ratios of Dementia Risk among Veterans according to Frailty Status

*Covariates: Race, ethnicity, marital status, smoking history, alcohol and substance abuse, obstructive sleep apnea and use of anticholinergic drugs



Most respondents completed the mailed survey required items: questions about physical activity (94.4%, n=1013), years of education (95.4%, n=1024) and the SAGE (96.6%, n=1036) items.



We found a high prevalence of dementia risk in this sample of community dwelling Veterans. As expected, frailty was associated with a higher risk for dementia. The use of a mailed screening was feasible and convenient, and may serve as an alternative, and scalable approach to identify individuals at risk for dementia.
Most studies on the prevalence of dementia risk have used community-based samples as part of cross-sectional or cohort studies where participants were screened on site. The high prevalence of dementia risk in our study is consistent with previous studies using a CAIDE cut-off score of ≥ 6 in Japanese American men and Finish samples of all genders ranging from 75% to 99% (8, 13, 14). However, there are no previous studies on the prevalence of dementia risk in the US Veteran population. The high prevalence of dementia risk in our study has several explanations. Forty percent of our sample consisted of minority groups, 95% were males, 43% were obese, 38.1% were sedentary, and over one third frail, groups that in most studies have shown the highest risk for cognitive impairment. Furthermore, studies in US male veterans demonstrate a high prevalence of cardiovascular risk factors which are often associated with a higher risk for dementia (15). In addition to the high prevalence of cardiovascular risk, veterans show higher rates of medical multimorbidity (16) and mental illness (17) that may further contribute to a higher incidence of cognitive impairment. Unlike community samples, veterans may be more representative of patients in health care settings. Most of these patients may be identified as at risk for dementia when seeking routine medical attention at health care facilities. However, many individuals may not regularly receive on-site care at health care institutions. The use of postal screening may serve as a convenient, scalable, and alternative population-based strategy for outreach dementia risk screening of high-risk individuals who would not otherwise actively seek health care.
Unlike previous studies including volunteer participants in cohorts or randomized controlled trials, this study investigated patients seeking medical care at an integrated healthcare system. Population studies estimations suggest that up to 30% of all cases of dementia can be attributed to modifiable risk factors (3). Early identification of those individuals at greater risk for dementia may lead to the implementation of targeted interventions that may alter the course of cognitive decline. The association of frailty would indicate that efforts at screening patients for dementia risk should especially focus on individuals with frailty as the yield is most likely to be higher.
Strengths of this study are the large number of participants receiving medical care at an integrated healthcare system, large number of minorities, data available from electronic health records and the use of validated instruments. Limitations include the low response rate, a predominantly male sample at one VA medical center, whose ethnic, racial, educational, and socio-economic composition may be different from other community settings. The low survey response rate is in line with previous studies in Veteran populations (18). We used several evidence-based strategies to improve response rates including a short questionnaire, a personalized letter of introduction, a self-addressed envelope and a second letter, “reminder” letter for non-respondents (19). However, other effective techniques may consist of prior notifications, monetary incentives, and format changes such as a larger envelopes or double-sided questionnaires (ours was single sided to minimize bulk) (18-20). Additional study limitations include reliance on self-reported levels of physical activity, which is susceptible to recall and response bias, lack of measurement data of potential confounders, and cross-sectional study design where exposure and outcome are concurrently assessed, and the temporal association between exposure and outcome cannot be definitively established.
There are two major clinical implications of this study. First, the possibility that older adults from certain groups may be more likely to be at high risk for dementia, because they may have multiple risk factors, which if not recognized early, could increase the likelihood of developing dementia. Early detection of patients at risk for dementia can lead to early interventions on modifiable factors and improved prognosis (6, 21). Second, this study shows an alternative population-based approach to properly screen individuals for dementia risk in health care settings. This strategy represents an attempt to reach out to individuals at risk for dementia who do not routinely seek medical care. These postal-based strategies extend existing health services by providing awareness, prevention, and screening, for dementia risk to a larger segment of the population.
In conclusion, this study identified a high prevalence of dementia risk in this sample of community dwelling Veterans. Individuals with frailty were at higher risk for dementia. Mailed screening was feasible and convenient and may serve as an alternative, and scalable approach to identify individuals at risk for dementia.


Funding: This material is the result of work supported with resources and the use of facilities at the Miami VA Healthcare System GRECC. The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.

Conflict of interest statement: Aakashi Shah, Otoniel Ysea-Hill, Angelica Torres-Morales, Christian J. Gomez, Alejandro Castellanos, and Jorge G. Ruiz declare that they have no known competing financial interests or personal relationships, which have or could be perceived to have influenced the work reported in this article.

Ethics declaration: A protocol of this study was submitted to and approved by Miami VA Healthcare System Institutional Review Board (IRB) and was exempted from informed consent.





1. Yaffe, K., Modifiable Risk Factors and Prevention of Dementia: What Is the Latest Evidence? JAMA Intern Med, 2018. 178(2): p. 281-282; DOI: 10.1001jamainternmed.2017.7299.
2. Di Marco, L.Y., et al., Modifiable lifestyle factors in dementia: a systematic review of longitudinal observational cohort studies. J Alzheimers Dis, 2014. 42(1): p. 119-35; DOI: 10.3233/JAD-132225.
3. Norton, S., et al., Potential for primary prevention of Alzheimer’s disease: an analysis of population-based data. Lancet Neurol, 2014. 13(8): p. 788-94; DOI: 10.1016/S1474-4422(14)70136-X.
4. Sprint Mind Investigators for the SPRINT Research Group, et al., Effect of Intensive vs Standard Blood Pressure Control on Probable Dementia: A Randomized Clinical Trial. JAMA, 2019. 321(6): p. 553-561; DOI:10.1001/jama.2018.21442.
5. Livingston, G., et al., Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet, 2020. 396(10248): p. 413-446; DOI:
6. Isaacson, R. and N. Saif, A Missed Opportunity for Dementia Prevention? Current Challenges for Early Detection and Modern-Day Solutions. J Prev Alzheimers Dis, 2020. 7(4): p. 291-293; DOI: 10.14283/jpad.2020.23.
7. Chu, N.M., et al., Hierarchical Development of Frailty and Cognitive Impairment: Clues Into Etiological Pathways. J Gerontol A Biol Sci Med Sci, 2019. 74(11): p. 1761-1770; DOI: 10.1093/gerona/glz134.
8. Kivipelto, M., et al., Risk score for the prediction of dementia risk in 20 years among middle aged people: a longitudinal, population-based study. Lancet Neurol, 2006. 5(9): p. 735-41; DOI: 10.1016/S1474-4422(06)70537-3.
9. Rosenberg, A., et al., Multidomain Interventions to Prevent Cognitive Impairment, Alzheimer’s Disease, and Dementia: From FINGER to World-Wide FINGERS. J Prev Alzheimers Dis, 2020. 7(1): p. 29-36; DOI: 10.14283/jpad.2019.41.
10. Scharre, D.W., et al., Self-administered Gerocognitive Examination (SAGE): a brief cognitive assessment Instrument for mild cognitive impairment (MCI) and early dementia. Alzheimer Dis Assoc Disord, 2010. 24(1): p. 64-71; DOI: 10.1097/WAD.0b013e3181b03277.
11. Scharre, D.W., et al., Community Cognitive Screening Using the Self-Administered Gerocognitive Examination (SAGE). J Neuropsychiatry Clin Neurosci, 2014. 26(4): p. 369-75; DOI: 10.1176/appi.neuropsych.13060145.
12. Orkaby, A.R., et al., The Burden of Frailty Among U.S. Veterans and Its Association With Mortality, 2002-2012. J Gerontol A Biol Sci Med Sci, 2019. 74(8): p. 1257-1264; DOI: 10.1093/gerona/gly232.
13. Chosy, E.J., et al., The CAIDE Dementia Risk Score and the Honolulu-Asia Aging Study. Dement Geriatr Cogn Disord, 2019. 48(3-4): p. 164-171; DOI: 10.1159/000504801.
14. Vire, J., et al., Secular changes in dementia risk indices among 70-year-olds: a comparison of two Finnish cohorts born 20 years apart. Aging Clin Exp Res, 2020. 32(2): p. 323-327; DOI: 10.1007/s40520-019-01204-8.
15. Fryar, C.D., et al., Cardiovascular Disease Risk Factors Among Male Veterans, U.S., 2009-2012. Am J Prev Med, 2016. 50(1): p. 101-105; DOI: 10.1016/j.amepre.2015.06.011.
16. Steinman, M.A., et al., Patterns of multimorbidity in elderly veterans. J Am Geriatr Soc, 2012. 60(10): p. 1872-80; DOI: 10.1111/j.1532-5415.2012.04158.x.
17. Maynard, C., et al., The Burden of Mental Illness Among Veterans: Use of VHA Health Care Services by Those With Service-connected Conditions. Med Care, 2017. 55(11): p. 965-969; DOI: 10.1097/MLR.0000000000000806.
18. Sack, D.I., et al., Evaluation of Three Postal Invitational Strategies to Increase Survey Response Rates in a Combat-Injured U.S. Military Population: Findings From the Wounded Warrior Recovery Project. Mil Med, 2019. 184(Suppl 1): p. 521-528; DOI: 10.1093/milmed/usy374.
19. Edwards, P., et al., Increasing response rates to postal questionnaires: systematic review. BMJ, 2002. 324(7347): p. 1183; DOI: 10.1136/bmj.324.7347.1183.
20. Puffer, S., et al., Increasing response rates to postal questionnaires: a randomised trial of variations in design. J Health Serv Res Policy, 2004. 9(4): p. 213-7; DOI: 10.1258/1355819042250159.
21. Zhang, X.X., et al., The Epidemiology of Alzheimer’s Disease Modifiable Risk Factors and Prevention. J Prev Alzheimers Dis, 2021. 8(3): p. 313-321; DOI: 10.14283/jpad.2021.15.



H. Shimada1, S. Lee1, K. Harada1, S. Bae1, K. Makino1, I. Chiba1, O. Katayama1, H. Arai2


1. Department of Preventive Gerontology, Centre for Gerontology and Social Science, National Centre for Geriatrics and Gerontology, Obu, Aichi, Japan; 2. National Centre for Geriatrics and Gerontology, Obu, Aichi, Japan

Corresponding Author: Prof. Hiroyuki Shimada, National Centre for Geriatrics and Gerontology, 7-430 Morioka-cho, Obu, Aichi 474-8511, Japan, Tel: +81-562-44-5651 (ext. 5680), Fax: +81 562-46-8294, E-mail:

J Prev Alz Dis 2022
Published online January 14, 2022,



BACKGROUND: Several technical devices are available to monitor and promote changes in behavior toward higher activity. In particular, smartphones are becoming the primary platform for recognizing human activity. However, the effects of behavior change techniques that promote physical, cognitive, and social activities on incident dementia in older adults remain unknown.
Objectives: This randomized controlled trial aims to examine the effects of behavior change techniques on the prevention of dementia among community-dwelling older adults using a smartphone as a behavior change tool.
Design: A randomized controlled trial.
Setting: Community in Japan.
Participants: The study cohort comprises 3,498 individuals, aged ≥60 years, randomized into two groups: the smartphone group (n = 1,749) and the control group (n = 1,749).
Intervention. The smartphone group will be asked to use smartphone applications for at least 30 minutes daily to self-manage and improve their physical, cognitive, and social activities. The smartphone group will perform 60-minute group walking sessions using application-linked Nordic walking poles with cognitive stimulation twice a week during the intervention period. The walking poles are a dual-task exercise tool that works with a smartphone to perform cognitive tasks while walking, and the poles are equipped with switches to answer questions for simple calculation and memory tasks. The smartphone and control groups will receive lectures about general health that will be provided during the baseline and follow-up assessments.
Measurements: Incident dementia will be detected using cognitive tests (at baseline, after 15 months, and after 30 months) and by preparing diagnostic monthly reports based on data from the Japanese Health Insurance System. Participants without dementia at baseline who will be diagnosed with dementia over the 30-month follow-up period will be considered to have incident dementia.
Conclusions: This study has the potential to provide the first evidence of the effectiveness of information communication technology and Internet of Things in incident dementia. If our trial results show a delayed dementia onset for self-determination interventions, the study protocol will provide a cost-effective and safe method for maintaining healthy cognitive aging.

Key words: Dementia, geriatric medicine, protocols and guidelines, preventive medicine.



Dementia is a serious social problem in an aging society. The aging rate in Japan has exceeded 28%, and dementia prevention efforts have been implemented as a national strategy. Nonpharmacological interventions that address cognitive function and its effect on daily living are widely studied for preventing cognitive decline. A meta-analysis of longitudinal studies demonstrated low-to-moderate inverse associations between physical activity (PA) and cognitive decline and dementia, with overall relative risk estimates of 0.65 (95% confidence interval [CI] 0.55–0.76) and 0.86 (95% CI 0.76–0.97), respectively (1).
Despite the established PA benefits, 30% of the world’s population fail to reach the recommended PA levels (2). Adherence with PA recommendations can be affected by behavioral (motivation and personal beliefs) and environmental factors (availability of public transport and exercise venues) (3). The most commonly reported behavior change techniques to improve PA are self-monitoring, goal setting, tailoring, relapse prevention training, feedback, and strategies for increasing motivation and self-efficacy (4). Currently, several technical devices (smartphones, pedometers, heart rate monitors, and smartphone applications) are available to monitor and promote changes in behavior toward higher PA (5). They can be used separately or together with a computer, smartphone, or personal tablet computer when self-monitoring PA. Smartphones are ubiquitous and are changing the landscape of the daily lives of all generations. Many smartphones are equipped with various sensors, including GPS, accelerometers, gyroscopes, and barometers. These sensors are rich data sources for measuring the various aspects of the users’ daily lives. Due to their unobtrusiveness, low installation cost, and ease of use, smartphones are becoming the main platform for human activity recognition. Although smartphones are likely to be beneficial in improving the health of older people (6), it is unclear whether smartphone use is effective for preventing dementia in older adults.
We previously reported a significantly lower probability of dementia in individuals who engage in active lifestyle behaviors—including daily conversation, driving a car, shopping, fieldwork, and gardening using longitudinal observational data (7). Moreover, the participation in these activities was associated with reversibility of mild cognitive impairment (MCI) to normal cognition. Participants with MCI were more likely to revert if they continued driving a car; used maps to travel; read newspapers or books; and participated in cultural lessons, community meetings, sports, hobbies, gardening, and fieldwork (8). However, the effects of lifestyle activities on cognitive function improvement and the incidence of dementia in older adults remain unclear.
Several large-scale and nonpharmacological multimodal intervention studies for dementia prevention have failed to provide clear evidence of delayed onset of dementia. The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER), which is a two-year multicenter randomized controlled trial (RCT) conducted in Finland, investigated the effects of a multidomain intervention on cognitive impairment and disability delay in older adults who are at risk of cognitive decline (9). At the two-year interim analysis, FINGER found that the multidomain intervention improved cognitive function in older adults with cognitive impairments. The U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk (U.S. POINTER) by the Alzheimer’s Association sought to replicate the Finnish trial results in the United States (10). The U.S. POINTER was a 2-year clinical trial that evaluated whether lifestyle interventions that simultaneously target many risk factors protect cognitive function in older adults who are at increased risk for cognitive decline. Although these multimodal interventions may have strong effects, they are expensive and difficult to implement. Additionally, evidence from RCTs is limited, and existing evidence does not support that nonpharmacological interventions reduce the risk of developing dementia (11). There is also limited evidence on the effectiveness of nonpharmacological approaches, and the cost-effectiveness of such approaches may be negative. A major reason for not being able to articulate the effectiveness of nonpharmacological interventions in preventing dementia may be the reduction in statistical power due to the short duration of the intervention and the small sample size. Future RCTs with a prolonged intervention phase and larger sample size are required to identify effective strategies for preventing dementia in older adults.
We developed a self-monitoring smartphone application and walking pole to facilitate daily physical, cognitive, and social activities. These information and communication technology and Internet of Things tools will enable RCTs with a prolonged intervention phase and larger sample size. We designed an RCT to determine the effect of dementia prevention by behavior change techniques—including the effects of daily smartphone application use on activities of daily living, and the effects of a pole walking program on cognitive function—in community-dwelling older adults. We also need to identify older adults with limited activities to prevent the spread of coronavirus disease 2019 (COVID-19) (12) by encouraging specific activities. In this study, the intervention will be conducted using a smartphone, which allows the participants to work by themselves and promotes non-face-to-face communication. In addition, group exercise can be conducted outdoors with sufficient interpersonal distance, thus, the program can be continued while managing COVID-19 infection precautions.



The Self-Management Activity Programme for the Older (SMAFO) study is funded by the Ministry of Health, Labour and Welfare in Japan (Subsidy of Long-Term Care Insurance [large-scale demonstration project]). This proof-of-concept trial aims to determine the effects of behavior change techniques on dementia prevention via smartphone use. This study may generate high-quality, long-term population-based data to support dementia prevention strategies (e.g., cost-effective smartphone-based SMAFO). Given that major cities in Japan provide similar services, the results can be used to develop a national strategy. This is one of the largest nonpharmacological intervention studies that focus on preventing dementia in older adults. Considering that SMAFO is a low-cost intervention that is adaptable to many older people, the information that we generate may inform future strategies in countries with similar service structures.

Study design

The SMAFO study is a single-center, assessor-blind RCT, parallel-group superiority trial with a 1:1 allocation based on the National Centre for Geriatrics and Gerontology–Study of Geriatric Syndromes (NCGG-SGS), which is a population-based cohort study that began in 2011. The overall goal of NCGG-SGS is to establish a screening system for geriatric syndromes and validate evidence-based interventions for preventing geriatric syndromes. Eligible NCGG-SGS participants are aged ≥60 years at recruitment and live in the survey site, which includes Obu, Nagoya, Takahama, Tokai, Toyoake, and Chita Cities. Individuals with certificated Japanese long-term care insurance (LTCI) are excluded. Japan implemented the national social LTCI system on 1 April 2000 to publicly address the issue of care. Every Japanese aged ≥65 years is eligible for the benefits strictly on the basis of physical and mental frailty or disability (13). The LTCI categorizes a person as “Support Level 1 or 2” or “Care Levels 1, 2, 3, 4, or 5” if he or she needs support for daily activities or continuous care, respectively. A trained local government official evaluates nursing care needs by using a questionnaire on current physical and mental status (73 items) and on the use of medical procedures (12 items). The results are used to calculate the applicant’s standardized scores for the seven dimensions of physical and mental status, estimate the time of care, and assign a care needs level based on the total estimated care minutes. The Nursing Care Needs Certification Board, which comprises physicians, nurses, and other experts in health and social services, is appointed by a mayor. The board determines whether the initial assessment is appropriate based on the applicant’s primary care physician’s statement and home visit assessor’s notes to make a final decision about LTCI certification.
This study will be performed in the three NCGG-SGS subcohorts (Chita City, Takahama City, and Midori Ward in Nagoya City). The first surveys in Chita City (2021), Takahama City (2015) and Midori Ward in Nagoya City (2013) have been completed. The NCGG-SGS participants were 6,367 adults aged ≥65 years in Chita City, 4,167 adults aged ≥60 years in Takahama City, and 5,257 older adults aged ≥70 years in Nagoya City. We will recruit candidates for intervention research for these NCGG-SGS participants. Our research group plans to conduct a survey to recruit potential participants in other areas if the required number of cases is not reached in these three cities.

Eligibility criteria

Inclusion criteria

The initial eligibility criteria for the study are as follows: (1) NCGG-SGS survey participants in Chita City, Takahama City, or Midori Ward in Nagoya City, and (2) living in any of the three cities. All potential participants meeting the eligibility criteria will be recruited via direct mail from older adults enrolled in the NCGG-SGS or non-NCGG-SGS participants aged ≥60 years as of the briefing session date living in the cohort survey site or via public relations. The participants will be asked to attend a SMAFO study briefing session and will be asked to provide a signed written consent before participation to indicate that they understand and agree with the purpose and content of SMAFO.

Exclusion criteria

Participants will be excluded if they are unwilling or cannot access trial tasks on the SMAFO, have difficulty in participating in intervention sessions or in attending sessions due to severe disability or restricted exercises by the doctor or for health problems, have dementia, are LTCI certified; have cognitive function assessment deficit, or are unable to write/converse in Japanese. Participants will undergo the Mini-Mental State Examination (MMSE) to exclude older adults with severe cognitive impairment. The MMSE scores will be set at an absolute of <23, <24, or <26 for persons with <12, 12–15, or ≥16 years of education, respectively (14).


The SMAFO group participants are asked to use a smartphone daily and to visit one of the 9 research fields twice weekly, where the 60-minute exercise program will be performed for 30 months in a total of 200 sessions. The SMAFO participants will use the online community salon application called Online Kayoinoba App, developed to enhance physical, cognitive, and social activities by using self-monitoring techniques. This application promotes outdoor activities by automatically creating walking courses and by posting photos of favorite places. The application also has a cognitive training game to improve memory, attention, and executive function, and has functions for home exercise, health check-up, communication, information of Kayoinoba, dietary control, and self-management (Table 1). Furthermore, SMAFO participants will be encouraged to communicate by using small group chats via a simple notification service and check whether they eat a well-balanced meal every day; these functions are included in the application developed for the SMAFO study. The smartphone group will be asked to use smartphone applications for at least 30 minutes daily to self-manage and improve physical, cognitive, and social activities.
Two trained facilitators will check the attendance of the participants at group sessions, provide instructions on how to use the smartphone, and facilitate communication between the participants at each SMAFO site. The group session will consist of 60 minutes of walking in the park. During the first month, the intervention group will perform simple walks, and the facilitator will teach them how to use their smartphones and applications. In the second month of the intervention, the participants will walk using the Nordic walking poles. From the third month of the intervention, the participants will perform the walk under multitasking conditions that include physical and cognitive tasks; we call this the combination training “Cognicise” (15). The SMAFO participants will use an application-linked Nordic walking pole with cognitive stimulation called Cognipole. A switch is equipped to the Cognipole handgrip to answer yes/no (right-hand/left-hand button) to cognitive tasks such as simple calculation and memory presented by the smartphone while walking. The cognitive tasks have a system of gradually increasing the difficulty level according to the correct answer rate, and the results are immediately provided to participants. The Cognipole was developed for this study and is not for sale at this time, but will be available from medical device manufacturers in the future.

Table 1. Functional details of the Online Kayoinoba Application


The participants’ adherence to the interventions is expected to improve with free smartphone use, as well as group chatting and group sessions, during study periods. Adherence will be monitored via the application, Cognipole, and supervised walking program. If a participant in the smartphone group fails to participate, the study facilitator will contact the person via chat.
Participants in both groups will attend three 90-minute health education classes that focus on health promotion during the 30-month trial period. The instructors will provide the participants with information regarding the aging process and geriatric syndromes. The topics in this lecture are planned so that they do not promote physical, cognitive, and social activities or significantly influence the study outcomes. Additional contacts will be made with the control participants over the study period to promote participation adherence in the follow-up examinations. Participants will be permitted to continue taking their normal medications and to perform daily activities, including sports, walking, and use of smartphone.

Outcome measures

Primary outcome: incidence of dementia

In Japan, all adults aged >65 years have public health insurance, including health insurance for employed individuals (Employees’ Health Insurance), national health insurance for unemployed and self-employed individuals aged 65–74 years (Japanese National Health Insurance), or health care for individuals aged >75 years (Later-Stage Medical Care) (16). In this study, participants will be tracked monthly for new incident dementia (Alzheimer’s disease or other dementia subtypes) during the follow-up period, as recorded by the Japanese National Health Insurance and Later-Stage Medical Care systems. Dementia will be diagnosed by medical doctors according to the International Classification of Diseases, 10th Revision (17). Participants without dementia at baseline but who are later diagnosed with dementia over a 30-month follow-up period will be considered incident dementia patients.
MMSE will be performed 15 months after the baseline survey and after the intervention to identify delayed dementia diagnosis for participants’ failure to visit the hospital. Participants with any less than the standard MMSE scores will be encouraged to visit a hospital: absolute scores of <23, <24, and <26 for persons with <12, 12–15, or ≥16 years of education, respectively, which are based on normative data and previous work of the New Haven Epidemiologic Catchment Area study (14). Participants who do not undergo MMSE will be confirmed by telephonic survey, and a suspicion of dementia or certification as requiring long-term care will lead to hospital consultation recommendation. Figure 1 shows the follow-up survey flow chart for the onset of dementia.

Figure 1. Flow chart of the follow-up survey for the onset of dementia

MMSE, Mini-Mental State Examination


Secondary measurements

Before the intervention and at the 15- and 30-month intervention and control periods, secondary measurements (neuropsychological testing, completion of questionnaires, and assessments of physical function, blood test, neuroimaging assessment, incident of disability, and mortality) will be performed.

Neuropsychological testing

We will use the National Centre for Geriatrics and Gerontology–Functional Assessment Tool (NCGG-FAT), which is an iPad application, to conduct cognitive screening (18, 19). The NCGG-FAT includes the following domains: 1) memory (wordlist memory-I [immediate recognition] and II [delayed recall]), 2) attention (a tablet version of the Trail Making Test part A), 3) executive function (a tablet version of the Trail Making Test part B), 4) processing speed (a tablet version of the Digit Symbol Substitution Test), 5) working memory (digit span memory test), and 6) global cognition. The NCGG-FAT has high test–retest reliability, moderate-to-high criterion-related validity (18), and predictive validity (19) among community-dwelling older persons. Well-trained study assistants will conduct cognitive functioning assessments in community facilities, including community halls. Before the study begins, all staff will receive training from the authors regarding the protocols for administering the assessments. For all cognitive tests, we will use established standardized thresholds for each corresponding domain to define impairment in population-based cohorts comprising community-dwelling older persons (scores > 1.5 standard deviations [SDs] below age- and education-specific means). We will also assess MMSE (20) to measure the global cognitive function.


Participants will be evaluated for basic activity of daily living (ADL), which includes five items, to assess basic self-care abilities, feeding, grooming, walking, bathing, and walking up stairs. The scores range from zero (complete independence; not applicable to all the items) to five (complete dependence). Participants will also be assessed for instrumental ADL by using the National Centre for Geriatrics and Gerontology-ADLs scale (21), which includes 13 items, including going out alone, taking the bus and train, handling medication, managing money, and making a meal. The scores range from 0 (low function) to 13 (high function). Self-reported PA will be assessed using the International Physical Activity Questionnaire-Short Form, which includes the frequency (days/week) and duration (minutes/day) of walking and moderate and vigorous intensity PA in the past 7 days (22, 23). Participants will be evaluated for depressive symptoms by using the 15-item Geriatric Depression Scale (GDS) (24), with scores ranging from 0 to 15; higher scores indicate a more depressed mood. The participants’ level of life satisfaction will be evaluated using the 13-item life satisfaction scale, where the scores range from 13 to 52 (higher scores indicate higher overall satisfaction) (25). Loneliness will be assessed using the 20-item version of the University of California Los Angeles Loneliness Scale version 3, with higher scores reflecting a greater degree of loneliness (26). Participants will be assessed for their dietary variety by using a dietary variety score comprising 10 food-based components; the scores range from 0 to 10, with a higher score indicating greater dietary variety (27). Participants will be assessed for social contact by using the National Centre for Geriatrics and Gerontology–Social Network Scale (NCGG-SNS), where the scores range from 0 to 64, with a higher score indicating better social network (28).

Physical function

The walking speed will be measured in seconds by using a stopwatch. The participants will walk five times on a flat and straight surface at a comfortable walking speed, and the average walking speed will be calculated. Dominant-hand grip strengths will be measured using a standard digital hand grip dynamometer (Takei Scientific Instruments Co., Ltd., Niigata City, Japan) in standing position with shoulder adducted and neutrally rotated and elbow in full extension (29). For all participants, the determination of physical frailty was based on the frailty phenotype proposed by Fried et al. (30) in the Cardiovascular Health Study. This phenotype consists of shrinking, weakness, slowness, self-reported exhaustion, and low PA.

Neuroimaging assessment

We will measure the brain structure (T1-weighted image, T2-weighted image, fluid attenuated inversion recovery image, diffusion tensor image, and phase difference enhanced image) and the resting state brain activity function (T2* BOLD-weighted image) by using 3-T magnetic resonance imaging.

Mortality and LTCI certification

We requested for new LTCI certification and mortality reports during the study periods from the administrative organization at Chita, Takahama, and Nagoya Cities.

Potential confounding factors

Numerous interrelated factors affecting the occurrence of dementia (including demographic variables, chronic medical conditions, psychosocial factors, and physical performance) are associated with dementia incidence in older persons (31, 32). We will assess outcome measurements and potential confounding factors to evaluate independent intervention effects. In this study, all multivariate models will include the following covariates unless otherwise specified: age at enrolment, sex, educational level, current smoking/drinking status, chronic medical illnesses, use of digital tools in participants’ usual life, and apolipoprotein E4 (ApoE4) carrier status. The presence of the following self-reported chronic medical illnesses will be entered into the models: history of stroke, heart disease, respiratory disease, hypertension, hyperlipidemia, and diabetes mellitus.


Figure 2 shows the predicted number of participants that will be included at each stage of the recruitment process. We would contact potential participants from the existing database via standard post. A two-stage screening process would be used to determine eligibility. Firstly, potential participants will be screened for major inclusion and exclusion criteria via recruitment letter. Potential participants will then be invited to attend a baseline assessment to measure secondary outcomes and confounding factors by a well-trained tester to confirm eligibility. Suitable participants will provide written informed consent prior to the collection of baseline measures.

Flow chart of participants and estimated number of participants

MCI, mild cognitive impairment;


Blinding and data collection

After baseline measures have been obtained, participants will be randomized with a 1:1 allocation to one of two interventions: i) smartphone group or ii) control group. The individuals will be assigned to each study group by using a stratified randomization protocol. The participants will be stratified according to residence and cognitive status. Established standardized thresholds will be used for all cognitive tests conducted in this study to define impairment in the corresponding domain for a population-based cohort of community-dwelling older persons (scores > 1.5 SDs below age- and education-specific means). Major cognitive impairment will be characterized by deficits in more than one of the four NCGG-FAT domains. The randomization schedule will be generated by an investigator who is not involved in determining eligibility procedures (IC). Allocation concealment will be maintained by this same investigator (IC), who will only reveal group allocation to an investigator (SL) once baseline measures have been obtained.
All outcome measures will be collected by a research assistant and analyzed by a primary investigator (HS) by using the data without personally identifiable information. Due to the nature of the interventions, neither the primary investigator providing the interventions nor the participants can be blinded to treatment group allocation. In the case of an adverse event requiring the consideration of participant withdrawal, investigators who are not involved in data management and analysis will be consulted.

Study power and sample size

Power analysis has been performed and was based on the proportion of incident dementia, which was determined via meta-analysis, to estimate the incidence of dementia in community settings (33). The meta-analysis yielded a pooled incidence rate of dementia per 1,000 person-years of 17.18 (95% confidence interval [CI]: 13.90–21.23). By assuming a 40% reduction in the incidence of dementia due to this study intervention, the prevalence rates of dementia during the 30-month follow-up period will be 2.58% and 4.30% in the intervention and control groups, respectively. The 40% between-group difference is deemed realistic on the basis of previously published data (34). For instance, a meta-analysis that was conducted to assess the effects of exercise on the onset of dementia revealed that the incidence rates of dementia in exercisers and controls were 3.7% (n = 949) and 6.1% (n = 1,017), respectively (11). The meta-analysis found 39% between-group differences due to exercise interventions. When the α level is 0.05 and the power is 0.80, the required number of cases will be 2,972 (1,486 on each side). Considering a dropout rate of 15%, the required number of participants can be met by setting the number of registered cases to 3,498.

Data management and monitoring

All participant data will be coded by a research assistant, with no group identifier to maintain the blinding of the investigator responsible for data analysis (HS). All cases of nonadherence and nonretention will be electronically documented on a master spreadsheet to ensure the appropriate handling and interpretation of results. All electronic data were deidentified and stored on a password-protected computer. All hard copy data were deidentified and kept in a locked filing cabinet in a secured office. Paper-completed outcome measures will undergo double data entry by the blinded assistant into electronic spreadsheets with embedded range checkers for data value.

Statistical analyses

Student’s t-test and Pearson’s chi-squared test will be used to calculate intergroup differences in baseline characteristics. We will calculate the cumulative dementia incidence during follow-up and estimate intergroup differences by using the log-rank test. Cox proportional hazards regression models will be used to examine the intervention effects of SMAFO on dementia incidence. After adjusting for age and sex in the first model, we will use a multiple adjustment model adjusted for demographic variables, chronic medical diseases, lifestyle, psychological and physical performance, and cognitive function variables. Adjusted hazard ratios for dementia incidence and their 95% CIs will be estimated. The primary analysis will be conducted according to the intention-to-treat principle to determine whether there is a significantly reduced decline in primary and secondary outcome measures in participants in the combined program compared with those in the control group. To account for possible bias introduced by participants without 30-month data, multiple imputation analyses will be performed. The Markov chain Monte Carlo method, which is based on multivariate normality and is appropriate for measurement outcomes in this study, will be used to generate 50 imputed datasets based on age, sex, education history, ApoE4 carrier status, rating on the GDS, MCI subtype, and prior observations of outcome variables.
To evaluate differences in secondary outcome changes from the baseline and at 30-month follow-up between the SMAFO and control groups, the mixed-effects model for repeated measures with an unstructured covariance structure will be conducted. The differences between pre- and post-data for each outcome will be the dependent variables adjusted for age, sex, educational level, depressive symptoms, ApoE4 genotype carrier status, and MCI as covariates in the multivariable models. For secondary categorical variables, logistic regression analyses or chi-squared tests will be conducted as appropriate. Given that we hypothesized that the intervention effects may be affected by the baseline cognitive deficit status, we will also perform the above analyses separately for the MCI and non-MCI subgroups. All analyses will be performed using SPSS Statistics for Mac version 25.0 (IBM Corp., Armonk, NY, USA). Statistical significance threshold will be set at P <.05.

Patient and public involvement

Residents in the research field assisted in the development of this research protocol.

Study limitations

This study has several limitations. Considering that the SMAFO study participants are Japanese seniors, it is unclear if this intervention can be performed outside Japan. Despite these uncertainties, the SMAFO is relatively inexpensive to implement and is readily disseminated to other countries. Second, there is a possible selection bias for participants in this study. Participants in the study are seniors who have or are willing to use smartphones. Such seniors may have an unusually high understanding of information communication technology and high cognitive function (35). Therefore, our results may underestimate dementia onset. Third, the smartphone-based SMAFO cannot determine the effects of individual interventions. Future studies will be needed to determine the effects of individual interventions on reducing the incident dementia. Fourth, there is a risk of overestimating the onset of dementia because using MMSE results for excluding dementia may not identify very early dementia. Nevertheless, the impact on the intervention effect is expected to be negligible since participants will be randomly assigned.

Ethics and Dissemination

The Human Research Ethics Committee of the National Centre for Geriatrics and Gerontology approved this study (No. 1335). The study is registered with the University Hospital Medical Information Network Clinical Trials Registry (identification number: UMIN000041926). The annual expense for controls to obtain similar tools, including the smartphone and Cognipole, remains equal (at €900); therefore, this does not exclude them from pensioners’ privileges within the community. The study researchers will publish the articles according to the International Committee of Medical Journal Editors recommendations (36). According to the publication and data policy of the NCGG, all research information produced by this project will be made available for the scientific community and society as a whole. New results will be presented to the general public via NCGG web pages. When applicable, research results will be published in peer-reviewed, high-quality scientific journals that perform reliable scientific review processes for research papers. Anonymized data are planned to be dispensed to the scientific community in an appropriate open-access data repository.


Funding: This work was supported by the Japan Agency for Medical Research and Development, grant numbers 16dk0110021h0001 and 17le0110004h0001. The funding source had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.

Conflict of Interest: None.

Patient consent for publication: Not required.

Ethics approval: The SMAFO studies have been reviewed and approved by the Ethics Committee of the National Centre for Geriatrics and Gerontology. All regulations and measures of confidentiality are handled in accordance with the Declaration of Helsinki.

Acknowledgments: The authors appreciate the assistance of Dr. Takehiko Doi, Dr. Kota Tsutsumimoto, Dr. Sho Nakakubo, Dr. Satoshi Kurita, and Dr. Hideaki Ishii for their professional help with the intervention protocol.

Trial registration number: Clinical Trials Registry (identification number: UMIN000041926).



1. Blondell SJ, Hammersley-Mather R, Veerman JL. Does physical activity prevent cognitive decline and dementia?: A systematic review and meta-analysis of longitudinal studies. BMC Public Health 2014;14:510. doi: 10.1186/1471-2458-14-510
2. Hallal PC, Andersen LB, Bull FC, et al. Global physical activity levels: surveillance progress, pitfalls, and prospects. Lancet 2012;380:247-257. doi:10.1016/S0140-6736(12)60646-1
3. Franco MR, Tong A, Howard K, et al. Older people’s perspectives on participation in physical activity: a systematic review and thematic synthesis of qualitative literature. Br J Sports Med. 2015;49:1268-1276. doi: 10.1136/bjsports-2014-094015
4. van Achterberg T, Huisman-de Waal GG, Ketelaar NA, Oostendorp RA, Jacobs JE, Wollersheim HC. How to promote healthy behaviours in patients? An overview of evidence for behaviour change techniques. Health Promot Int 2011;26:148-162. doi:10.1093/heapro/daq050
5. Bravata DM, Smith-Spangler C, Sundaram V, et al. Using pedometers to increase physical activity and improve health: a systematic review. JAMA 2007;298:2296-2304. doi: 10.1001/jama.298.19.2296
6. Joe J, Demiris G. Older adults and mobile phones for health: a review. J Biomed Inform 2013;46:947-954. doi:10.1016/j.jbi.2013.06.008
7. Shimada H, Makizako H, Lee S, Doi T, Lee S. Lifestyle activities and the risk of dementia in older Japanese adults. Geriatr Gerontol Int 2018;18:1491-1496. doi: 10.1111/ggi.13504
8. Shimada H, Doi T, Lee S, Makizako H. Reversible predictors of reversion from mild cognitive impairment to normal cognition: a 4-year longitudinal study. Alzheimers Res Ther 2019;11:24. doi:10.1186/s13195-019-0480-5
9. Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet 2015;385:2255-2263. doi:10.1016/S0140-6736(15)60461-5
10. Espeland MA, Baker LD, Carrillo MC, et al. Increasing adherence in a centre-based trial of lifestyle intervention in older individuals: U.S. POINTER Trial. Innov Aging 2018;2:809. doi: 10.1093/geroni/igy023.3011
11. de Souto Barreto P, Demougeot L, Vellas B, Rolland Y. Exercise training for preventing dementia, mild cognitive impairment, and clinically meaningful cognitive decline: a systematic review and meta-analysis. J Gerontol A Biol Sci Med Sci 2018;73:1504-1511. doi:10.1093/gerona/glx234
12. Tison GH, Avram R, Kuhar P, et al. Worldwide effect of COVID-19 on physical activity: a descriptive study. Ann Intern Med 2020;173:767-770. doi:10.7326/M20-2665
13. Tsutsui T, Muramatsu N. Care-needs certification in the long-term care insurance system of Japan. J Am Geriatr Soc 2005;53:522-527.doi:10.1111/j.1532-5415.2005.53175.x
14. Satizabal CL, Beiser AS, Chouraki V, Chene G, Dufouil C, Seshadri S. Incidence of dementia over three decades in the Framingham Heart Study. N Engl J Med 2016;374:523-532. doi:10.1056/NEJMoa1504327
15. Shimada H, Makizako H, Doi T, et al. Effects of combined physical and ognitive exercises on cognition and mobility in patients with mild cognitive impairment: a randomized clinical trial. J Am Med Dir Assoc 2018;19:584-591. doi:10.1016/j.jamda.2017.09.019
16. Ministry of Health Labour, and Welfare of Japan. Annual Health, Labour, and Welfare Report 2011-2012. 2012. Accessed 1 January 2017.
17. World Health Organization. International Classification of Diseases, 10th Revision (ICD-10). 1992. World Health Organization, Geneva.
18. Makizako H, Shimada H, Park H, et al. Evaluation of multidimensional neurocognitive function using a tablet personal computer: test-retest reliability and validity in community-dwelling older adults. Geriatr Gerontol Int 2013;13:860-866. doi:10.1111/ggi.12014
19. Shimada H, Makizako H, Park H, Doi T, Lee S. Validity of the National Center for Geriatrics and Gerontology-Functional Assessment Tool and Mini-Mental State Examination for detecting the incidence of dementia in older Japanese adults. Geriatr Gerontol Int 2017;17:2383-2388. doi:10.1111/ggi.13079
20. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12:189-198. doi:10.1016/0022-3956(75)90026-6
21. Makino K, Lee S, Bae S, Shinkai Y, Chiba I, Shimada H. Predictive validity of a new Instrumental Activities of Daily Living Scale for detecting the incidence of functional disability among community-dwelling older Japanese adults: a prospective cohort study. Int J Environ Res Public Health 2020;172291. doi:10.3390/ijerph17072291
22. Booth M. Assessment of physical activity: an international perspective. Res Q Exerc Sport 2000;71 Suppl 2:114-120. doi:10.1080/02701367.2000.11082794
23. Tomioka K, Iwamoto J, Saeki K, Okamoto N. Reliability and validity of the International Physical Activity Questionnaire (IPAQ) in elderly adults: the Fujiwara-kyo Study. J Epidemiol 2011;21:459-465. doi:10.2188/jea.je20110003
24. Yesavage JA. Geriatric Depression Scale. Psychopharmacol Bull 1988;24:709-711.
25. Shimada H, Lee S, Bae S, Hotta R. A new Life Satisfaction Scale predicts depressive symptoms in a national cohort of older Japanese adults. Front Psychiatry 2020;11:625. doi:10.3389/fpsyt.2020.00625
26. Russell DW. UCLA Loneliness Scale (Version 3): reliability, validity, and factor structure. J Pers Assess 1996;66:20-40. doi:10.1207/s15327752jpa6601_2
27. Kumagai S, Watanabe S, Shibata H, et al. [Effects of dietary variety on declines in high-level functional capacity in elderly people living in a community]. Nihon Koshu Eisei Zasshi 2003;50:1117-1124.
28. Bae S, Harada K, Chiba I, et al. A New Social Network Scale for Detecting Depressive Symptoms in Older Japanese Adults. Int J Environ Res Public Health 2020;17:8874. doi: 10.3390/ijerph17238874
29. Watanabe T, Owashi K, Kanauchi Y, Mura N, Takahara M, Ogino T. The short-term reliability of grip strength measurement and the effects of posture and grip span. J Hand Surg Am 2005;30:603-609. doi:10.1016/j.jhsa.2004.12.007
30. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001;56:M146-M156. doi:10.1093/gerona/56.3.m146
31. Verghese J, Lipton RB, Katz MJ, et al. Leisure activities and the risk of dementia in the elderly. N Engl J Med 2003;348:2508-2516. doi:10.1056/NEJMoa022252
32. Morley JE, Morris JC, Berg-Weger M, et al. Brain health: the importance of recognizing cognitive impairment: an IAGG consensus conference. J Am Med Dir Assoc 2015;16:731-739. doi:10.1016/j.jamda.2015.06.017
33. Fiest KM, Jette N, Roberts JI, et al. The prevalence and incidence of dementia: a systematic review and meta-analysis. Can J Neurol Sci 2016;43 Suppl 1:S3-S50. doi:10.1017/cjn.2016.18
34. Forette F, Seux ML, Staessen JA, et al. Prevention of dementia in randomised double-blind placebo-controlled Systolic Hypertension in Europe (Syst-Eur) trial. Lancet 1998;352:1347-1351. doi:10.1016/s0140-6736(98)03086-4
35. Gordon ML, Gatys L, Guestrin C, Bigham JP, Trister A, Patel K. App Usage Predicts Cognitive Ability in Older Adults. Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. 2019.
36. International Committee of Medical Journal Editors. Defining the role of authors and contributors 2020. Accessed 29 October 2020.



G. Price1, C. Udeh-Momoh1, M. Kivipelto1,3, L.T. Middleton1,2


1. Ageing Epidemiology Research Unit, School of Public Health, Imperial College London, London, UK; 2. Department of Primary Care and Public Health, Imperial College Healthcare NHS Trust, London, UK; 3. Division of Clinical Geriatrics, Department of Neurobiology, Care Sciences and Society, Karolinska Institute, Sweden.

Corresponding Author: Prof. Lefkos Middleton, Ageing Epidemiology Research Unit, School of Public Health, Faculty of Medicine, Imperial College London, London W6 8RP, UK. E-mail:; Tel: +44 20 3311 0216; Fax: +44 20 3311 0216.

J Prev Alz Dis 2022;
Published online January 14, 2022,


An ageing society, whilst representing a major success of humanity in recent decades, brings a significant increase in the numbers of dementia sufferers, world-wide. There is evidence to suggest that successive cohorts remain cognitively healthy for longer than their forebears and reach an older age before developing dementia. This has been attributed to a combination of life-course factors and testifies to the multifactorial aetiology of Alzheimer’s disease (AD) and other forms of late-onset dementia (LOD). But as longevity increases, more people live long enough to develop LOD, and survive longer with the condition. Consequently, despite (or because of) improvements in population health, LOD is becoming an increasing burden on society. It is a public health imperative to prevent or delay the onset or clinical progression of this multifactorial disorder. The manuscripts in this issue address a range of aetiological risk factors for LOD, novel technologies and study designs, with implications for preventative strategies.
Age-specific incidence rates for dementia have fallen by an estimated 13% per decade since 1988 in the United States and Europe (1). Both dementia-free life expectancy and years of survival with dementia have increased (2). Taking into account declining age-specific incidence and increasing life expectancy, the number of cases of dementia is projected to continue to increase over coming decades (3). Reducing the burden of dementia is identified as a key global health priority (4).
The aetiologies of AD and other LOD forms are evidently multifactorial, with risk factors identified throughout the life course. The heritability of dementia varies according to type, with that of late-life Alzheimer’s disease reaching highly significant levels (h2 ranging between 0.58 and 0.79) (5). Fetal, perinatal and early childhood factors have been proposed as potentially relevant to risk of late-life common and complex diseases, such as dementia or cardiovascular and metabolic conditions (6). Low educational attainment, mid- and late-life hypertension, diabetes, obesity, physical inactivity, excessive alcohol consumption, smoking, depression, hearing impairment, brain injury, social isolation and air pollution may account collectively for 40% of dementia cases worldwide (7). Potential explanatory factors which may have contributed to the reduction in age-adjusted dementia incidence across successive cohorts include higher educational attainment, reductions in smoking and more effective management of cardiovascular diseases and diabetes (8).
With a complex and multifactorial disorder, the identification of aetiological pathways or risk factors should broaden intervention efforts rather than narrowing them. Dementia may in some cases be prevented by interventions which target individual risk factors, as has been demonstrated in the case of risk reduction of cardiovascular diseases via antihypertensive medication (9). But other single-factor strategies have, to date, been less successful. Trials of pharmaceutical interventions targeting beta-amyloid as potential treatments for AD have yielded predominantly negative results. Some non-pharmaceutical interventions, such as diet, physical exercise and cognitive training, have demonstrated beneficial effects on aspects of cognition though not prevention of dementia (9). Multifactorial aetiology calls for a multidomain approach to intervention, whereby multiple risk factors are addressed concurrently, such as the Finnish Geriatric Intervention Study to prevent cognitive impairment and disability (FINGER) study (10) and the Vascular Care Intervention to Prevent Dementia (preDIVA) study (11). The natural extension of this approach will be the strategy of personalised precision medicine, whereby individuals’ personal profiles of risk factors are identified through careful assessment and targeted through bespoke intervention packages delivered to at-risk individuals.
The manuscripts included in this edition relate, in different and complementary ways, to this ambition (12-20). An essential requirement for successful trialling of preventative interventions is the identification and recruitment of at-risk individuals. Mourzi and colleagues present an example of developments in the genetic prediction of the risk and clinical manifestation of dementia in their investigation of the genetics of frailty. Ford et al. discuss the value of neuroimaging in risk prediction and early detection, the pragmatic challenges which limit its widespread adoption at large scale, and the potential for new and emerging technologies to provide solutions to these challenges. Udeh-Momoh et al. present an analysis of the utility and economic viability of plasma amyloid assays in pre-screening for AD prevention trials.
Another requirement is the identification and measurement of modifiable risk factors or exposures potentially amenable to intervention. Drawing on preclinical animal models, distinct mechanisms underpinning commonly cited behavioural and non-pharmacological dementia prevention factors with significant potential for AD risk reduction are described comprehensively by Alanko et al. (21). The study by Zheng and colleagues takes forward current knowledge on type 2 diabetes as a risk factor for dementia and identifies a range of comorbidities whose prevention and effective management could contribute to maintaining cognitive health in this at-risk population. The review by Abbott et al. discusses the challenges in the measurement of dietary exposures, and the scope for novel -omics technologies to enhance its accuracy and reliability in investigations of the effects of diet on cognitive health.
The trial protocols in this issue stand as examples of the diversity of approaches, by different means and at different stages of progression to disease, adopted in trials of interventions to prevent or treat dementia. Avgerinos et al. describe a randomized control trial to evaluate the effects of induced ketosis, through the use of ketone monoester, on brain metabolism, cognitive performance and AD pathogenic cascades, in cognitively healthy adults with metabolic syndrome – thus at higher risk for AD. Xu and colleagues present the Singaporean member of the Worldwide FINGERS international network of multidomain lifestyle interventions to delay cognitive decline in those at risk of dementia. Sindi et al. present the pilot MIND-ADMINI randomized trial to extend the FINGER concept to individuals with prodromal AD, with and without medical food supplement. For patients already at the early stages of dementia, Gonzales and colleagues describe a pilot trial of senolytic therapy as a potential modulator of clinical progression.
The manuscripts in this special issue thus argue for diversity – of technologies, risk indicators, therapeutic targets and interventions, both multi-domain and targeted, using a precision medicine model approach – as a fitting and necessary feature of global efforts to reduce the burden of Alzheimer’s disease and other forms of late onset dementia.



1. Wolters FJ, Chibnik LB, Waziry R et al. Twenty-seven-year time trends in dementia incidence in Europe and the United States. Neurology 2020; 95 (5) e519-e531. doi: 10.1212/WNL.0000000000010022
2. Grasset L, Pérès K, Joly P et al. Secular trends of mortality and dementia-free life expectancy over a 10-year period in France. Eur J Epidemiol 2019; 34, 115–123. doi: 10.1007/s10654-019-00482-w
3. Ahmadi-Abhari S, Guzman-Castillo M, Bandosz P et al. Temporal trend in dementia incidence since 2002 and projections for prevalence in England and Wales to 2040: modelling study. BMJ 2017; 358:j2856. doi: 10.1136/bmj.j2856
4. Shah H, Albanese E, Duggan C et al. Research priorities to reduce the global burden of dementia by 2025. Lancet Neurology 2016; 15 (12), 1285-1294. doi: 10.1016/S1474-4422(16)30235-6
5. Gatz M, Reynolds CA, Fratiglioni L et al. Role of genes and environments for explaining Alzheimer disease. Arch Gen Psychiatry 2006; 63 (2),168-174. doi:10.1001/archpsyc.63.2.168
6. Whalley LJ, Dick FD, & McNeill G. A life-course approach to the aetiology of late-onset dementias. Lancet Neurology 2006; 5 (1), 87-96. doi: 10.1016/S1474-4422(05)70286-6
7. Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 2020; 396: 413–446. doi: 10.1016/S0140-6736(20)30367-6
8. Wu Y-T, Beiser AS, Breteler MMB, et al. The changing prevalence and incidence of dementia over time – current evidence. Nat Rev Neurol 2017; 13 (6): 327–339 doi: 10.1038/nrneurol.2017.63
9. Livingston G, Sommerlad A., Orgeta V et al. Dementia prevention, intervention, and care. Lancet 2017; 390: 2673–734. doi: 10.1016/S0140-6736(17)31363-6
10. Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet 2015; 385: 2255–2263 doi: 10.1016/S0140-6736(15)60461-5
11. Moll van Charante EP, Richard E, Eurelings LS et al. Effectiveness of a 6-year multidomain vascular care intervention to prevent dementia (preDIVA): a cluster-randomised controlled trial. Lancet 2016; 388: 797-805, doi: 10.1016/S0140-6736(16)30950-3
12. Udeh-Momoh C, Zheng B, Sandebring-Matton A et al. Blood Derived Amyloid Biomarkers for Alzheimer’s Disease Prevention. J Prev Alz Dis 2022. doi: 10.14283/jpad.2021.70
13. Gonzales MM, Garbarino VR, Marques Zilli E et al. Senolytic Therapy to Modulate the Progression of Alzheimer’s Disease (SToMP-AD): A Pilot Clinical Trial. J Prev Alz Dis 2022. doi: 10.14283/jpad.2021.62
14. Sindi S, Thunborg C, Rosenberg A et al. Multimodal Preventive Trial for Alzheimer’s Disease: MIND-ADmini Pilot Trial Study Design and Progress. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.4
15. Xu X, Chew KA, Wong ZX et al. The SINgapore GERiatric Intervention Study to Reduce Cognitive Decline and Physical Frailty (SINGER): Study Design and Protocol. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.5
16. Abbott KA, Posma JM, Garcia-Perez I et al. Evidence-Based Tools for Dietary Assessments in Nutrition Epidemiology Studies for Dementia Prevention. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.6
17. Avgerinos KI, Mullins RJ, Egan JM et al. Ketone Ester Effects on Biomarkers of Brain Metabolism and Cognitive Performance in Cognitively Intact Adults ≥ 55 Years Old. A Study Protocol for a Double-Blinded Randomized Controlled Clinical Trial. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.3
18. Ford J, Kafetsouli D, Wilson H et al. At a Glance: An Update on Neuroimaging and Retinal Imaging in Alzheimer’s Disease and Related Research. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.7
19. Mourtzi N, Hatzimanolis A, Xiromerisiou G et al. Association between 9p21-23 Locus and Frailty in a Community-Dwelling Greek Population: Results from the Hellenic Longitudinal Investigation of Ageing and Diet. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.2
20. Zheng B, Su B, Udeh-Momoh C et al. Associations of Cardiovascular and Non-Cardiovascular Comorbidities with Dementia Risk in Patients with Diabetes: Results from a Large UK Cohort Study. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.8
21. Alanko V, Udeh-Momoh C, Kivipelto M et al. Mechanisms Underlying Non-Pharmacological Dementia Prevention Strategies: A Translational Perspective. J Prev Alz Dis 2022. doi: 10.14283/jpad.2022.9



T. Daly


Sorbonne Université, Sciences Norms Democracy, UMR 8011, Paris, France.

Corresponding Author: Timothy Daly, Sorbonne Université, Sciences Norms Democracy, UMR 8011, Paris, France,

J Prev Alz Dis 2022;
Published online January 11, 2022,


Dear Editor,

As Rosenberg and colleagues describe in the journal (1), the Worldwide FINGERS is “a landmark initiative” to develop “Multidomain Interventions to Prevent Cognitive Impairment, Alzheimer’s Disease, and Dementia” across the globe (p. 35, ibid). It demonstrates the research community’s investment to move beyond pharmacological-only approaches for these conditions, and to find actionable solutions across the high/low income country divide “with shared core methodology, but local culture and context-specific adaptations” (p. 29, ibid). There are four steps that larger society, including governments and public health bodies, could take now to bolster dementia risk reduction efforts.
Firstly, help individuals and communities carry out a FINGER-inspired lifestyle that involves physical, mental, and social stimulation, nutritional changes, and management of metabolic and vascular risk factors. Existing government campaigns have recommended lifestyle changes (2), but each recommendation ought to be accompanied by investment in making it feasible, e.g. by “ensuring people have access to affordable, healthy food, rather than just encouraging them to eat healthily” (3).
Secondly, take action against social determinants of health. Rosenberg and colleagues recognise the importance of economic development for dementia. The wealth–brain health link is a double-edged sword: socioeconomic deprivation not only increases the risk of dementia but also reduces participation in FINGER-style interventions (4). It is therefore important that research with these interventions is complemented by action against the social determinants of health such as socioeconomic deprivation, barriers to education, and social isolation as three particularly relevant examples.
Thirdly, not be moralistic about participation in risk-reduction activities. The data on the impact of these interventions on individuals is still not definitive. In a pluralistic society, it is important to respect those who do not take part and to warn against victim-blaming where people with dementia are seen as complicit for not having done more to prevent their condition (2).
Finally, design a risk-reducing society. Living and working spaces ought to be made more conducive to encouraging participation in risk-reduction activities. This means making those environments not only safe and accessible, but also as physically, mentally, and socially stimulating as possible across society (5).
In conclusion, the vital Worldwide FINGERS initiative would be greatly helped along by the further collaboration of broader society to build a positive and democratic view of dementia in which risk reduction is seen as a shared challenge and responsibility.


Acknowledgements: Timothy Daly thanks the Fondation Médéric Alzheimer for the doctoral bursary that funds his studies.

Conlict of interest: The author declares no conflicts of interest.



1. A. Rosenberg, F. Mangialasche, T. Ngandu, A. Solomon, M. Kivipelto, Multidomain Interventions to Prevent Cognitive Impairment, Alzheimer’s Disease, and Dementia: From FINGER to World-Wide FINGERS. J Prev Alzheimers Dis 2020;7, 29-36, doi: 10.14283/jpad.2019.41.
2. D. Horstkötter, K. Deckers, S. Köhler, Primary Prevention of Dementia: An Ethical Review. J Alzheimers Dis 2021;79, 467-476, doi: 10.3233/JAD-201104.
3. The Lancet Neurology, Amid competing priorities, dementia must not be forgotten. Lancet Neurol 2021;20, 685, doi: 10.1016/S1474-4422(21)00258-1.
4. N. Coley et al., Disparities in the participation and adherence of older adults in lifestyle-based multidomain dementia prevention and the motivational role of perceived disease risk and intervention benefits: an observational ancillary study to a randomised controlled trial. Alzheimers Res Ther 2021;13, 157, doi: 10.1186/s13195-021-00904-6.
5. T. Daly, Giving a fairer face to urban space: Progress on the long road to dementia prevention. Int J Geriatr Psychiatry, 2021, doi: 10.1002/gps.5657.




B. Zheng1, B. Su2, C. Udeh-Momoh1, G. Price1, I. Tzoulaki2,3,4,5, E.P Vamos6, A. Majeed6,7, E. Riboli2,7, S. Ahmadi-Abhari1, L.T. Middleton1,7


1. Ageing Epidemiology Research Unit, School of Public Health, Imperial College London, London, UK; 2. Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, UK; 3. BHF Centre of Excellence, Imperial College London, London, UK; 4. Dementia Research Institute, Imperial College London, London, UK; 5. Hygiene and Epidemiology, University of Ioannina, Ioannina, Greece; 6. Department of Primary Care and Public Health, School of Public Health, Imperial College London, London, UK; 7. Public Health Directorate, Imperial College NHS Healthcare Trust, London, UK

Corresponding Author: Prof. Lefkos Middleton, Ageing Epidemiology Research Unit, School of Public Health, Faculty of Medicine, Imperial College London, London W6 8RP, UK. E-mail:; Tel: +44 20 3311 0216; Fax: +44 20 3311 0216.

J Prev Alz Dis 2022;
Published online January 11, 2022,



Background: Type 2 diabetes (T2D) is an established risk factor for dementia. However, it remains unclear whether the presence of comorbidities could further increase dementia risk in diabetes patients.
Objectives: To examine the associations between cardiovascular and non-cardiovascular comorbidities and dementia risk in T2D patients.
Design: Population-based cohort study.
Setting: The UK Clinical Practice Research Datalink (CPRD).
Participants: 489,205 T2D patients aged over 50 years in the UK CPRD.
Measurements: Major cardiovascular and non-cardiovascular comorbidities were extracted as time-varying exposure variables. The outcome event was dementia incidence based on dementia diagnosis or dementia-specific drug prescription.
Results: During a median of six years follow-up, 33,773 (6.9%) incident dementia cases were observed. Time-varying Cox regressions showed T2D patients with stroke, peripheral vascular disease, atrial fibrillation, heart failure or hypertension were at higher risk of dementia compared to those without such comorbidities (HR [95% CI] = 1.64 [1.59-1.68], 1.37 [1.34-1.41], 1.26 [1.22-1.30], 1.15 [1.11-1.20] or 1.10 [1.03-1.18], respectively). Presence of chronic obstructive pulmonary disease or chronic kidney disease was also associated with increased dementia risk (HR [95% CI] = 1.05 [1.01-1.10] or 1.11 [1.07-1.14]).
Conclusions: A range of cardiovascular and non-cardiovascular comorbidities were associated with further increases of dementia risk in T2D patients. Prevention and effective management of these comorbidities may play a significant role in maintaining cognitive health in T2D patients.

Key words: Type 2 diabetes, dementia, comorbidity, cohort.



Alzheimer’s disease (AD) and other forms of late-onset dementia (LOD) are fast becoming major healthcare and socio-economic challenges across the globe, in parallel with increases in life expectancy and population ageing (1, 2). Type 2 diabetes (T2D) is an established risk factor for dementia and is associated with a 53%-73% higher risk of LOD or AD (3-5). We previously showed that high and unstable glycaemia, and acute or microvascular diabetic complications are associated with increased dementia risk among older T2D patients (6), highlighting the importance of effective diabetes management in maintaining cognitive health.
The presence of comorbidities is another clinical challenge in diabetes management. Comorbidities, especially cardiovascular diseases (7), are common in T2D patients. Based on a cross-sectional study of 9832 T2D patients across 13 countries, the authors reported that over 30% of patients had prevalent cardiovascular diseases (7). The aim of the current study was to explore and quantify potential effects of the presence of common, chronic diseases of ageing on dementia risk, in T2D patients.
In this regard, we have comprehensively evaluated the longitudinal associations between a diverse range of major cardiovascular and non-cardiovascular comorbidities and the risk of incident dementia in a cohort of T2D patients, aged over 50 years, leveraging electronic health record (EHR) data from the UK Clinical Practice Research Datalink (CPRD) (8).



Data sources

The UK CPRD GOLD database (8) is a longitudinal national primary care database that includes EHR data of over 17 million individuals, currently or previously registered with over 700 general practitioner (GP) practices in the UK. CPRD has also been linked to secondary care data (Hospital Episode Statistics, HES), mortality data from the Office for National Statistics (ONS) and area-based data on measures of social deprivation. The CPRD population has the same profile regarding age, sex and ethnicity as the general population of the UK (8).

Study population

Participants were included if they were aged 50 years or over at any point during their CPRD registration period between 1987 and 2018 and had a diagnosis of diabetes, based on relevant CPRD Medcode or a prescription of anti-diabetes drugs (oral hypoglycaemic agents or insulin) (6). In addition, eligible patients should have been registered in CPRD for at least one year before diabetes onset to allow time for baseline information to be recorded and to ensure that the date of newly diagnosed diabetes was captured. Patients with a diagnosis of type 1 diabetes, or those who had a diagnosis of diabetes or initiation of anti-diabetic treatment before the age of 30 were excluded. Patients who had a diagnosis of dementia before cohort entry were also excluded. A total of 489,205 individuals were included in the analysis.

Exposure assessment

Records of six cardiovascular comorbidities (coronary heart disease, stroke, atrial fibrillation, heart failure, peripheral vascular disease and hypertension) and three major non-cardiovascular comorbidities (chronic kidney disease, chronic obstructive pulmonary disease [COPD] and cancer) were extracted using the corresponding CPRD Medcode and Enttype code. These comorbidities were selected as they are common chronic diseases and leading causes of death and disability in older adults (9). To comprehensively identify hypertensive cases, we additionally used blood pressure recordings (systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg). The date of onset of a specific type of comorbidity was defined according to its first relevant health record. The date of onset of overall cardiovascular comorbidities was defined as the earliest date of developing one of the six comorbidities mentioned above.
In addition, information on the following covariates at cohort entry was extracted: age, sex, calendar year of cohort entry, region in the UK, quintiles of index of multiple deprivation (IMD, a proxy of socio-economic status), body mass index (BMI), smoking status, duration of diabetes (based on the first clinical record of diabetes diagnosis), history of prescription of anti-diabetes medications and recorded diabetic complications.

Outcome ascertainment

The outcome event was dementia incidence. Patients were considered to have dementia if they had: 1) a dementia diagnosis based on Medcode in CPRD; 2) a dementia diagnosis based on ICD codes in linked HES or ONS records; or 3) at least one dementia-specific drug prescription (donepezil, galantamine, rivastigmine or memantine) (6). Among the extracted dementia cases, 96% were based on diagnosis codes and 4% were based on dementia-specific drug record. The outcome event date was defined as the date of the first dementia diagnosis or the first prescription date of dementia-specific drugs, whichever occurred earlier.

Statistical analyses

Distributions of baseline characteristics were summarised and compared between patients with and without cardiovascular disease (except hypertension) at baseline. Time-varying Cox proportional hazards models, with age as the time-scale, were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) of dementia associated with comorbidities. The presence of six cardiovascular comorbidities in aggregate and by type, as well as three non-cardiovascular comorbidities, were treated as time-varying variables and examined in separate Cox models. Patients with newly developed comorbidities during follow-up contributed person-years to the no comorbidity group up until the comorbidity diagnosis date and then contributed person-years to the comorbidity group. For each patient, time of cohort entry was defined as the date of diabetes onset, aged 50 or January 1, 1987, whichever was the latest. The end of follow-up was defined as the date of dementia incidence, death, transfer out date, last data collection date of the GP practice or May 1, 2018, whichever occurred first.
To account for potential confounding biases, three sequential models with increasing levels of covariate adjustment were created for all analyses: Model 1 only adjusted for age, sex, calendar year and region; Model 2 additionally adjusted for IMD (quintiles), smoking status (non-smoker, current smoker, ex-smoker or missing), BMI category (<25, 25 to <30, ≥30 kg/m2 or missing), and history of comorbidities (for mutual adjustment); Model 3 additionally adjusted for diabetes-related factors, including the duration of diabetes, HbA1c level, presence of diabetic complications (including hypoglycaemia) and prescription of anti-diabetes drugs (no drug, only oral hypoglycaemic drug, or insulin). Covariates that could change over time (e.g., HbA1c level and status of diabetic complications) were also modelled as time-varying variables and updated at comorbidity diagnosis date during follow-up. Missing values of BMI category and smoking status during follow-up were imputed with the last observation carried forward.
The statistical analyses were conducted using Stata (version 15, Stata). All statistical tests were two-sided, and the significance level was P < 0.05.



Baseline characteristics of study population

Of the 489,205 patients with type 2 diabetes, 52.0% were male; the mean baseline age was 65.2 (SD = 11.2) years (Table 1). At cohort entry, 41.4% of patients had obesity (BMI ≥30 kg/m2) and 19.1% or 35.1% were self-reported current or ex-smokers; 17.4% had been prescribed anti-diabetes medications, and the mean baseline HbA1c level was 7.4%. Before baseline, 84,702 (17.3%), 35,808 (7.3%), 415,197 (84.9%), 28,270 (5.8%), 20,069 (4.1%) and 56,918 (11.6%) patients already had coronary heart disease, stroke, hypertension, atrial fibrillation, heart failure and peripheral vascular disease, respectively. There were 20,324 (4.2%), 30,881 (6.3%) and 48,940 (10.0%) patients with COPD, chronic kidney disease and cancer before baseline.
Compared to T2D patients without cardiovascular disease (not including hypertension) at baseline, those with pre-existing cardiovascular disease were slightly older, more socio-economically deprived, and more likely to be male, current or ex-smokers and have more non-cardiovascular comorbidities (P < 0.05).

Table 1. Baseline characteristics of diabetes cohort by presence of cardiovascular disease

Note: IMD = index of multiple deprivation; BMI = body mass index; CKD = chronic kidney disease; COPD = chronic obstructive pulmonary disease. * The presence of cardiovascular disease was defined here as the presence of any of coronary heart disease, stroke, atrial fibrillation, heart failure and peripheral vascular disease before baseline. IMD, BMI, smoking status and baseline HbA1c had 6.6%, 12.2%, 18.9% and 35.6% missing values, respectively. The statistics for these variables presented in this table are based on complete cases.


Cardiovascular comorbidities in associations with dementia risk

During a median of 6 years follow-up (ranging from 0-31 years) of the 489,205 T2D patients, 33,773 (6.9%) incident dementia cases were recorded. During the follow-up, 38,148 (7.8%), 27,232 (5.6%), 44,620 (9.1%), 27,615 (5.6%), 23,394(4.8%) and 75,258 (15.4%) patients developed coronary heart disease, stroke, hypertension, atrial fibrillation, heart failure and peripheral vascular disease, respectively, before dementia diagnosis or censoring.
After adjusting for a full set of covariates (Model 3), there were significant associations of stroke, peripheral vascular disease, atrial fibrillation, heart failure and hypertension with higher risk of dementia incidence (HR [95% CI] = 1.64 [1.59-1.68], 1.37 [1.34-1.41], 1.26 [1.22-1.30], 1.15 [1.11-1.20] and 1.10 [1.03-1.18], respectively; Table 2). The HR estimates were even higher in Model 1 and Model 2 with fewer covariates (P < 0.05). In contrast, the presence of coronary heart disease was not robustly associated with dementia risk in Model 3 (P = 0.085), while Model 1 and 2 revealed a modest association (Table 2). Having any cardiovascular comorbidity was associated with an additional 25% increase of dementia risk in T2D patients (95% CI: 15%-36%; Table 2).

Table 2. Associations between presence of cardiovascular and non-cardiovascular comorbidities and risk of dementia among 489,205 diabetes patients

Note: Results are based on time-varying Cox regressions, with no-comorbidity group as the reference group. Model 1 adjusted for age, sex, calendar year and region; Model 2 further adjusted for IMD, smoking status, BMI category and history of comorbidities where applicable (coronary heart disease, stroke, hypertension, heart failure, atrial fibrillation, peripheral vascular disease, chronic kidney disease, chronic obstructive pulmonary disease and cancer); Model 3 additionally adjusted for duration of diabetes, prescriptions of anti-diabetes drugs, HbA1c level and diabetic complications.


Non-cardiovascular comorbidities in associations with dementia risk

Among the non-cardiovascular comorbidities evaluated, 18,357 (3.8%), 70,852 (14.5%) and 52,056 (10.6%) patients developed COPD, chronic kidney disease and cancer during follow-up, respectively. The presence of COPD or chronic kidney disease in T2D patients resulted in a 5% (95% CI: 1%-10%) or 11% (95% CI: 7%-14%) increased risk of dementia incidence, respectively, in the fully adjusted model (Table 2); the magnitudes of these associations were higher in Model 1 and Model 2 with fewer covariates (P < 0.05). In contrast, the presence of cancer was not associated with dementia risk in T2D patients (HR = 1.00, 95% CI: 0.98-1.03).



This study is the largest cohort study to date (489,205 T2D patients and 33,773 incident dementia cases) that has attempted to explore and quantify the associations of cardiovascular and non-cardiovascular comorbidities with dementia risk, in T2D patients. After taking into account the time-varying nature of the presence of comorbidities and adjusting for a large set of potential confounding factors in the time-varying Cox regressions, we found that a range of cardiovascular and non-cardiovascular comorbidities in T2D patients were associated with further increases of dementia risk, among which stroke and peripheral vascular disease are the strongest risk factors.
Our results showing increased dementia risk in T2D patients with cardiovascular comorbidities are in line with previous evidence (10). Exalto et al. (10) reported a summary risk score for prediction of 10-year dementia risk in type 2 diabetes patients based on data from two cohorts (n = 29,961 and 2413). In their prediction model, the presence of cerebrovascular disease (HR = 1.65, 95% CI: 1.50-1.82) and the presence of cardiovascular disease (HR = 1.21, 95% CI: 1.13-1.29) were among the strongest predictors for dementia in T2D patients. Our study provided more comprehensive evidence by investigating the independent associations of different subtypes of cardiovascular comorbidities. The strong association between stroke and dementia risk may be attributed, at least partially, to vascular dementia (11). Stroke may also contribute to the pathogenesis of other LOD forms such as AD (12, 13); furthermore, it is well established that the majority of LOD patients, over the age of 75, harbour mixed cerebral pathologies (13, 14). Unlike stroke, the presence of peripheral vascular disease, atrial fibrillation and heart failure is likely to merely represent peripheral markers of cerebrovascular pathology (12, 13). Our finding of diminished association between coronary heart disease and dementia after adjusting for more covariates implies that the association is not independent and could be driven by other risk factors.
This study also showed that hypertension was associated with 10% increased dementia risk in T2D patients. In fact, the 2020 Report of the Lancet Commission for dementia prevention, intervention and care (15) has identified hypertension and diabetes amongst 12 modifiable risk factors for dementia that could be directly implicated in the neurodegenerative process. The Lancet report has highlighted the potential of prevention strategies for dementia. On the other hand, a pioneering randomised control trial (RCT) of multidomain lifestyle intervention, the “Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability” (FINGER) trial (16), showed that a 2-year multidomain intervention of diet, physical activity, cognitive training and management/monitoring of vascular/metabolic risk had efficacy in improving memory and cognition in at-risk older adults.
In contrast, non-cardiovascular comorbidities in T2D patients have drawn less attention with respect to dementia risk. The underlying mechanisms linking COPD and chronic kidney disease with increased dementia risk (e.g., microvascular pathologies or chronic inflammation (17)) warrant further investigation. Future well-powered clinical studies, with deep phenotyping and biomarker-based characterisation of dementia patients, may help to elucidate the precise contributions of these comorbidities in the pathogenesis of LOD. Previous studies showed that stroke and hypertension were associated with increased risk for dementia (HR for stroke = 1.69, 95% CI: 1.49-1.92 (18); relative risk, RR for hypertension = 1.20, 95% CI: 1.06-1.35 (19)) in the general population regardless of the presence of diabetes, whereas evidence for COPD and chronic kidney diseases was scarce. Whether these diseases and T2D influence dementia risk independently or interactively (e.g., through a synergistic effect) warrants further research.
The incidence rate of dementia among T2D patients in our study (mean age 65.2 years) was 9 cases per 1000 person-years, which is higher than other studies in the general population that also used data from primary care electronic health records. For example, a study of the UK adults aged over 60 years in The Health Improvement Network database reported a dementia incidence rate of 3-4/1000 person-years between 1997 and 2007 (20); another study using the same database (median age 59.5 years) reported an incidence rate of 5.65/1000 person-years in 2015 (21). This reflected the fact that T2D patients are more likely to develop dementia (22, 23). A recent review on T2D and cognitive dysfunction (24) pointed out that the screening and diagnosis of cognitive impairment and dementia is important for older adults with T2D, and suggested that future clinical research is warranted to evaluate the role of comorbidities associated with diabetes in increasing risk of cognitive dysfunction. The findings of our study provide evidence on the role of comorbidities in further increasing the risk of cognitive decline and dementia in T2D patients. Their appropriate management, including lifestyle modification, might have a preventative effect on dementia risk in clinical practice.
There are several limitations of this study to be considered in interpreting our results. Firstly, dementia cases could be underreported or under-diagnosed in the CPRD database. Nevertheless, we maximised the dementia case detection using the linked data from HES and ONS databases (secondary diagnosis and cause of death) and dementia-specific drug records. We have also controlled for calendar year in all analyses to account for the increasing diagnosis rate of dementia and changes in dementia diagnostic criteria over time (25). In addition, we did not distinguish by specific types of LOD, as such granular level of data is variably registered in CPRD and the precise coding boundaries and heterogeneity of these dementia types still remain poorly defined (13, 14). Although we aimed to capture a broad age range of patients with diabetes (26) to increase the representativeness of this study, future studies focused on older elderly populations with higher dementia incidence rate and possibly different comorbidity profiles are needed. Furthermore, the possibility of residual confounding cannot be excluded. For example, there is little information on education level and on physical and social activities in the CPRD database, which are known risk factors for dementia (15). Finally, although we identified several strong and robust associations, causality cannot be established due to the observational nature of this study.
In conclusion, this large-scale cohort study showed that the evaluated cardiovascular and non-cardiovascular comorbidities, with the exception of coronary heart disease and cancer, are independent risk factors for dementia incidence in patients with T2D. In the absence of disease-modifying therapies for late-onset dementia (27) and given the need to mitigate the increased risk for dementia in people with diabetes, the prevention and effective management of these comorbidities may have a significant role in maintaining cognitive health and reducing the dementia burden among older adults with diabetes.


Funding: This report is based on independent research supported by Diabetes UK (Grant No: 18/0005851) and the National Institute for Health Research (NIHR) Applied Research Collaboration (ARC) Northwest London. The views expressed in this publication are those of the authors and not necessarily those of Diabetes UK and the National Institute for Health Research or the Department of Health and Social Care. The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.

Data availability: This study is based on data from the Clinical Practice Research Datalink obtained under license from the UK Medicines and Healthcare Products Regulatory Agency (protocol approved by the Independent Scientific Advisory Committee: No. 19_065R). According to the UK Data Protection Act, information governance restrictions (to protect patient confidentiality) prevent data sharing via public deposition. Data extracts can be requested by applying to the Clinical Practice Research Datalink (

Conflict of Interest Disclosure: L T Middleton has received research funding (to Institution) from Janssen, Merck (USA), Gates, Takeda/ Millenium, Novartis, Invincro and EIT Health, outside the submitted work. All other authors declare no conflict of interest.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.



1. Alzheimer’s Association. 2020 Alzheimer’s disease facts and figures. Alzheimers Dement. 2020;16:391-460. DOI:
2. Ahmadi-Abhari S, Guzman-Castillo M, Bandosz P, et al. Temporal trend in dementia incidence since 2002 and projections for prevalence in England and Wales to 2040: modelling study. BMJ. 2017;358:j2856. DOI:
3. Gudala K, Bansal D, Schifano F, Bhansali A. Diabetes mellitus and risk of dementia: A meta-analysis of prospective observational studies. J Diabetes Investig. 2013;4(6):640-650. DOI:
4. Zhang J, Chen C, Hua S, et al. An updated meta-analysis of cohort studies: Diabetes and risk of Alzheimer’s disease. Diabetes Res Clin Pract. 2017;124:41-47. DOI:
5. Chatterjee S, Peters SAE, Woodward M, et al. Type 2 Diabetes as a Risk Factor for Dementia in Women Compared With Men: A Pooled Analysis of 2.3 Million People Comprising More Than 100,000 Cases of Dementia. Diabetes Care. 2016;39(2):300-307. DOI:
6. Zheng B, Su B, Price G, Tzoulaki I, Ahmadi-Abhari S, Middleton L. Glycemic Control, Diabetic Complications, and Risk of Dementia in Patients With Diabetes: Results From a Large U.K. Cohort Study. Diabetes Care. 2021;dc202850. DOI:
7. Mosenzon O, Alguwaihes A, Leon JLA, et al. CAPTURE: a multinational, cross-sectional study of cardiovascular disease prevalence in adults with type 2 diabetes across 13 countries. Cardiovasc Diabetol. 2021;20(1):154. DOI:
8. Herrett E, Gallagher AM, Bhaskaran K, et al. Data Resource Profile: Clinical Practice Research Datalink (CPRD). Int J Epidemiol. 2015;44(3):827-836. DOI:
9. National Center for Chronic Disease Prevention and Health Promotion. Chronic Diseases in America. Accessed 6 September 2021.
10. Exalto LG, Biessels GJ, Karter AJ, et al. Risk score for prediction of 10 year dementia risk in individuals with type 2 diabetes: a cohort study. Lancet Diabetes Endocrinol. 2013;1(3):183-190. DOI:
11. Ott A, Stolk RP, van Harskamp F, Pols HA, Hofman A, Breteler MM. Diabetes mellitus and the risk of dementia: The Rotterdam Study. Neurology. 1999;53(9):1937-1942. DOI:
12. Sonnen JA, Larson EB, Brickell K, et al. Different Patterns of Cerebral Injury in Dementia With or Without Diabetes. Arch Neurol. 2009;66(3):315-322. DOI:
13. Toledo JB, Arnold SE, Raible K, et al. Contribution of cerebrovascular disease in autopsy confirmed neurodegenerative disease cases in the National Alzheimer’s Coordinating Centre. Brain. 2013;136:2697-2706. DOI:
14. Schneider JA, Arvanitakis Z, Bang W, Bennett DA. Mixed brain pathologies account for most dementia cases in community-dwelling older persons. Neurology. 2007;69(24):2197-2204. DOI:
15. Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 2020;396(10248):413-446. DOI:
16. Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet. 2015;385(9984):2255-2263. DOI:
17. Walker KA, Ficek BN, Westbrook R. Understanding the Role of Systemic Inflammation in Alzheimer’s Disease. ACS Chem Neurosci. 2019;10(8):3340-3342. DOI:
18. Kuzma E, Lourida I, Moore SF, Levine DA, Ukoumunne OC, Llewellyn DJ. Stroke and dementia risk: A systematic review and meta-analysis. Alzheimers Dement. 2018;14(11):1416-1426. DOI:
19. Ou YN, Tan CC, Shen XN, et al. Blood Pressure and Risks of Cognitive Impairment and Dementia: A Systematic Review and Meta-Analysis of 209 Prospective Studies. Hypertension. 2020;76(1):217-225. DOI:
20. Rait G, Walters K, Bottomley C, et al. Survival of people with clinical diagnosis of dementia in primary care: cohort study. BMJ. 2010;341:c3584. DOI:
21. Pham TM, Petersen I, Walters K, et al. Trends in dementia diagnosis rates in UK ethnic groups: analysis of UK primary care data. Clin Epidemiol. 2018;10:949-960. DOI:
22. Rochoy M, Rivas V, Chazard E, et al. Factors Associated with Alzheimer’s Disease: An Overview of Reviews. J Prev Alzheimers Dis. 2019;6(2):121-134. DOI:
23. Ng TP, Lee TS, Lim WS, et al. Development, Validation and Field Evaluation of the Singapore Longitudinal Ageing Study (SLAS) Risk Index for Prediction of Mild Cognitive Impairment and Dementia. J Prev Alzheimers Dis. 2021;8(3):335-344. DOI:
24. Srikanth V, Sinclair AJ, Hill-Briggs F, Moran C, Biessels GJ. Type 2 diabetes and cognitive dysfunction-towards effective management of both comorbidities. Lancet Diabetes Endocrinol. 2020;8(6):535-545. DOI:
25. Donegan K, Fox N, Black N, Livingston G, Banerjee S, Burns A. Trends in diagnosis and treatment for people with dementia in the UK from 2005 to 2015: a longitudinal retrospective cohort study. Lancet Public Health. 2017;2(3):e149-e156. DOI:
26. Wang K, Liu H. Early-Onset Subgroup of Type 2 Diabetes and Risk of Dementia, Alzheimer’s disease and Stroke: A Cohort Study. J Prev Alzheimers Dis. 2021;4(8):442-447. DOI:
27. Gauthier S, Albert M, Fox N, et al. Why has therapy development for dementia failed in the last two decades? Alzheimers Dement. 2016;12(1):60-64. DOI:



J. Ford1, D. Kafetsouli1, H. Wilson2, C. Udeh-Momoh1, M. Politis2, S. AhmadiAbhari1, I. Rabiner3, L.T. Middleton1,4


1. Age & Epidemiology Research Unit, School of Public Health, Faculty of Medicine, Imperial College London, UK; 2. Neurodegeneration Imaging Group, University of Exeter Medical School, Exeter, UK; 3. Invicro, Centre for Imaging Sciences, Hammersmith Hospital, London, UK; 4. Primary Care and Public Health Directorate, Charing Cross Hospital, Imperial College Healthcare NHS Trust, UK

Corresponding Author: Lefkos T Middleton, Ageing and Epidemiology research Unit, School of Public Health, Charing Cross Hospital, East Wing, St Dunstan’s Rd, London W6 8RP, United Kingdom, Tel: +44 (0)20 33117290. Email:

J Prev Alz Dis 2022;
Published online January 11, 2022,



Neuroimaging serves a variety of purposes in Alzheimer’s disease (AD) and related dementias (ADRD) research – from measuring microscale neural activity at the subcellular level, to broad topological patterns seen across macroscale-brain networks, and everything in between. In vivo imaging provides insight into the brain’s structure, function, and molecular architecture across numerous scales of resolution; allowing examination of the morphological, functional, and pathological changes that occurs in patients across different AD stages (1). AD is a complex and potentially heterogenous disease, with no proven cure and no single risk factor to isolate and measure, whilst known risk factors do not fully account for the risk of developing this disease (2).
Since the 1990’s, technological advancements in neuroimaging have allowed us to visualise the wide organisational structure of the brain (3) and later developments led to capturing information of brain ‘functionality’, as well as the visualisation and measurement of the aggregation and accumulation of AD-related pathology. Thus, in vivo brain imaging has and will continue to be an instrumental tool in clinical research, mainly in the pre-clinical disease stages, aimed at elucidating the biological complex processes and interactions underpinning the onset and progression of cognitive decline and dementia.
The growing societal burden of AD/ADRD means that there has never been a greater need, nor a better time, to use such powerful and sensitive tools to aid our understanding of this undoubtedly complex disease. It is by consolidating and reflecting on these imaging advancements and developing long-term strategies across different disciplines, that we can move closer to our goal of dementia prevention. This short commentary will outline recent developments in neuroimaging in the field of AD and dementia by first describing the historical context of AD classification and the introduction of AD imaging biomarkers, followed by some examples of significant recent developments in neuroimaging methods and technologies.

Key words: Alzheimer’s disease, dementia, imaging, biomarkers.




In 1984, the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS) and the Alzheimer’s Disease and Related Disorders Association (ADRDA) developed the first set of criteria in the attempts to describe cases of “probable AD” (4). Defined, at the time, as the presence of an acquired amnestic disorder in which at least two cognitive domains (including memory) were measured to be impaired, and this impairment negatively impacted the individual’s day-to-day life (4). In 2011, the term “preclinical AD” was introduced as an early stage, along the AD continuum, that acknowledges increased risk among older adults who display no overt clinical symptoms (5-6). The preclinical AD stage is a long phase prior to the onset of measurable mild cognitive decline that warrants a diagnosis of mild cognitive impairment (MCI) due to AD (5). This presents a “long window of opportunity” for targeted secondary prevention measures and interventions (7).
In 2018, the National Institute of Aging / Alzheimer’s Association (NIA/AA) expert group proposed the Research Framework biological definition of AD, for research purposes only, based on the in vivo AT[N] triad biomarkers, reflecting the disease pathology of abnormal burden of amyloid (A), tau (T) and neurodegeneration (N) (8). Individuals were classified using a dichotomous construct of positive (+) and negative (-) to symbolise above and below threshold values of these biomarkers (8). This binary classification has subsequently met with criticism pertaining to clinical and practical limitations (9). Asymptomatic (thus cognitively healthy) older adults but still demonstrating biomarkers above a pre-defined threshold – would, under the Research Framework criteria, be categorised as falling along the disease continuum. A concern that was echoed by a recent set of recommendations from the International Working Group (IWG) stating that defining a disease state is not conceivable in the absence of clinical manifestations and that AD-related biomarkers fall along a continuous scale and, as such, a binary classification “does not reflect the reality of amyloid-β and tau pathology” (10). Our stance on the biological classification of AD to consider carefully the “grey” peri-threshold areas can help to disentangle the complexity of interactions of various risk factors (9).
Non-imaging risk-based assessments measuring various facets of lifestyle (11), genetics (12), and more recently novel blood-based approaches (13) are methods that have proven utility in measuring older adults’ increased risk of developing AD during early, preclinical, disease stages. On-the-other-hand neuroimaging methods are costly and, as yet, not widely available. Nevertheless, the benefits in bringing neuroimaging into the clinical trial space provides invaluable insight into the brains of live human subjects of older adults, across the AD continuum.
The next sections will describe recent developments both in commonly applied neuroimaging modalities, as well as areas that have only until recently been introduced and, still, evaluated. This commentary by no means provides an exhaustive review of all the available methods to explore the brain’s structure and function in vivo, but an attempt to provide an update on recent developments and an induction to support the understanding of the many contributions that neuroimaging can offer in AD/ADRD research.


Magnetic Resonance Imaging (MRI)

Structural MRI (sMRI) is a non-invasive imaging modality that can visualise brain anatomy and measure volumetric and other structural pathologies across white and grey matter, by utilising specialised MR sequences. Evidence of loss of volume and neurodegeneration [N] is one of the triad hallmarks of AD of the NIA/AA Research Framework and sMRI was acknowledged as a key [N] biomarker tool (14), and those who are suspected to have dementia often receive an MRI scan to ascertain possible signs of neurodegeneration and the extent to which it may be present, as well to exclude unrelated pathologies.
Measuring differences in cortical thickness is one of the several metrices of neurodegeneration based on sMRI. Recent research has highlighted the asymmetric nature of cortical thinning, commonly seen in AD patients (15). Yet, the asymmetrical distribution of cortical thickness is often overlooked in AD research – despite it being an “important feature of normal brain aging that is both shared by and accelerated in neurodegenerative AD” (15). Patients with AD were found to exhibit steeper reductions in cortical thickness within the left hemisphere, compared to the right, when measured against healthy controls (HC). Suggesting the patterns of neurodegeneration are not symmetrical – particularly in some ‘at risk’ areas, including the frontal and temporal brain regions. These regions are particularly vulnerable to asymmetrical loss in healthy aging but can exhibit further (accelerated) changes in AD patients (15).
Volumetric measures, calculated from 3D T1-weighted images, have been shown to identify several “brain atrophy subtypes” that can be used to assess neurodegeneration and for AD categorisation (16). MRI scans at baseline have been used to define and differentiate between older adults with (a) “limbic predominant atrophy”, (b) “hippocampal sparing atrophy”, (c) “typical/diffuse atrophy” and (d) “no evidence of brain atrophy”. Those with “typical/diffuse atrophy” and “limbic predominant atrophy” were at increased risk of developing subsequent dementia. These four brain atrophy subtypes were found to be more informative (or sensitive in predicting future dementia incidence) than less specific cortical volume measures – in which sub-categorisation of atrophy patterns could be a potential new biomarker for future studies (16).
The assessment of structural features using MRI also presents valuable opportunities for disease categorisation. For example, cognitive and MRI measures (alongside machine learning [ML] approaches [– covered in more detail below]), were used to aid the classification of participants with MCI, into AD subcategories (stable [sMCI] and converters to AD [cAD]) (17). The researchers successfully differentiated between sub-groups with the inclusion of MRI features, whereby additions to the models allowed for the measurement of changes in the entorhinal cortex and hippocampal volume. This “mixed-effects derived features” approach in combining cognitive test results and structural MRI was successfully used to measure the long-term trajectories of change within the brain and for the effective prediction and classification of multiple sub-groups. Therefore, the inclusion of imaging biomarkers (in this case MRI) is potentially complimentary with psychometric testing and allows for (a) the detection of early changes in memory function and (b) provided valuable insight into the role of entorhinal cortex as a potential future biomarker in AD imaging studies.
Unlike sMRI, functional MRI (fMRI) measures brain activity. Task-based fMRI paradigms involve exposing participants to stimuli or asking them to engage in cognitive tasks and measuring the elicited neural responses. Alternatively, ‘task-free’ paradigms collect resting-state fMRI data with the participant simply lying in the MRI scanner without such engagement with stimuli or tasks. For both task and task-free paradigms, blood oxygen level dependent response (BOLD) contrasts are used to measure the changes in blood oxygenation and serving as a proxy for neural activity.
Resting state fMRI has recently been used to investigate potential “compensatory mechanisms” among patients with “MCI due to AD” and HC (18). Using connectivity metrics such as degree centrality – i.e., the extent to which a node (e.g., a brain region) is connected to all other nodes within a network, the researchers found evidence of “compensatory” regions of interest (ROI). These included the right superior parietal gyrus, the right- and the left precentral gyri, and the right middle frontal gyrus. In these ROI, an increased degree centrality was observed, suggesting “more robust connectivity, despite regional atrophy”. Increased degree centrality among some ROI was associated with increased cognitive performance, despite localised aggregation of amyloid and neurodegeneration. These findings suggest that neural atrophy and brain functional decline are not necessarily co-dependent as previously thought, and that resting state functional connectivity measures may be useful for studies on neural compensation (18).
Both task-based and task-free fMRI data can measure functional connectivity (assessing for similarities in BOLD signal fluctuations between brain voxels or ROI), and recent research has attempted to use functional connectivity approaches to predict AD-related pathology (19). The authors adopted connectome-based predictive modelling (CPM) to predict CSF p-tau/Aβ42 (termed the “PATH-fc” model), derived from measures of functional connectivity in those with MCI and AD at a macro-scale spatial resolution. Whilst successfully predicting pathology, this approach also predicted rates of cognitive decline and captured alterations in well-cited resting state networks (e.g., the default mode network, among others). Thus, a whole-brain model, derived from measures of alterations in functional connectivity, can successfully help predict two major outcomes of risk factors of AD (pathology and cognitive decline).
Increasingly the field is becoming aware of the significant contributions that vascular-related risk factors have on AD/ADRD development (20). fMRI data can be utilised to investigate changes in vascularity and is emerging as a potential new biomarker to measure AD risk. 4D flow MRI can quantify blood velocities across the whole brain, and to measure changes in cardiac pulse pressure and speed (21). This is particularly useful for measuring arterial stiffness and overall “vascular compliance” (21). Arterial spin labelling-MRI (ASL-MRI) uses the hydrogen atoms present in water as a proxy for blood flow (21). Advanced imaging protocols used in mouse models e.g., multi-time echo ASL-MRI, can identify microscopic changes in blood-brain barrier permeability using a dynamic contrast-enhanced MRI when comparing age-related water permeability changes (22).
Vascular markers can be visualised using a multitude of imaging modalities. Lacunes, white matter hyperintensities (WMH), and microbleeds are visible using T2*-weighted imaging, susceptibility weighted images, and fluid-attenuated inversion recovery (FLAIR; (23)). FLAIR can measure evidence of cerebrovascular pathology and image processing techniques in this domain has continued to evolve. Recently, when testing the diagnostic accuracy of multiple image segmentation algorithms, estimating WMH when excluding more diffuse, smaller “lower-intensity” volumes led to a higher correlation between WMH and eventual AD clinical diagnosis and reduced cognitive performance, in the presence of abnormal amyloid burden (24).
Furthermore, improvements in imaging protocols e.g., fast BOLD scans to estimate variability in cardiorespiratory frequencies among AD patients (25). These fast-imaging sequences allow for the detection of frequency oscillations attributed to cardiac, respiratory, and other physiological based sources. As an example, variances in frequency oscillations were observed in AD patients (25). The authors hypothesised that these may be driven by cerebral hypoperfusion, alongside evidence of small vessel disease pathology (e.g., cerebral amyloid angiopathy) that may provide a novel measure in this population, in addition to small vessel reactivity and blood flow velocity in individual penetrating arteries (25).
Whilst the main emphasis of imaging studies in AD/ADRD research focuses on grey matter, diffusion weighted imaging (DWI) is the only non-invasive, albeit highly informative neuroimaging modality, that can provide valuable information pertaining to the microstructural connectivity of white matter (WM) tracts of the brain. Achieved by measuring the movement (diffusion) of water molecules within tissues (26). The introduction of DWI followed from a seminal paper on the topic in 1994 that proved the anisotropic diffusion of water can be exploited to understand WM fibre tract orientation (27). Now, there is increasing understanding in how alterations in diffusion signal between AD stages suggestive of changes in the microstructural integrity of WM tracts (26).
However, DWI’s main limitation resides in its low specificity, as “cross-networks” of fibres are complex, overlapping structures that cannot be detected using conventional DWI methods. An exciting development in diffusion MRI is the development of fixel-based analysis (FBA). Like a voxel as a volume element (3D), a fixel is a fibre element, and therefore FBA has the potential to expand on the field of tractography extensively, allowing to measure axon density, intra- and extra-cellular water, and evidence of demyelination and inflammation. Recently a two-year longitudinal study was conducted using participants from the Alzheimer’s disease neuroimaging initiative (ADNI) study (28). The authors identified specific fascicles associated with AD development and utilised two primary measures of free-water fraction (FW; a neuroinflammation proxy) and “apparent fibre density” (AFD). Significant increases in FW were found, within multiple ROIs, to be associated with disease’s progression in AD patients but not among MCI and HC individuals. On the other hand, AFD was found to decrease within multiple ROI in both AD and MCI patients (by 7-8 and 3-5% respectively)– suggesting a shared feature between these stages. These new findings exemplify the potential utility of FBA in exploring changes in white matter microstructural integrity between AD stages (28).
Thus, MRI is a powerful tool in that it can measure changes in both grey and white matter depending on its calibration. Moreover, multimodal MR approaches combining imaging modalities aimed at exploring changes in both grey and white matter may aid in unravelling some of the pathogenic mechanisms underlying AD development.
The dysregulation of neural iron content is emerging as another risk-factor associated with AD development that can be visualised using MRI. An imbalance in iron homeostasis is implicated in the development of AD proteinopathies, e.g., Aβ plaques and neurofibrillary tangle (NFT) accumulation (29). Several methods have been shown to measure iron dysregulation, such as “quantitative susceptibility mapping” (QSM). QSM is a non-invasive technique that can be used as a proxy for iron content (30). QSM signal in the inferior temporal gyrus (ITG) has been associated with the degree of cognitive impairment, with the ITG showing to be preferentially affected by tau deposits, suggesting a relationship between iron dysregulation and tau accumulation (30). Importantly, not just iron dysregulation has been implicated in AD but also copper, aluminium, zinc (and other metal ions) that interact with amyloid and tau in complex ways (31).


Positron Emission Tomography (PET)

MRI has demonstrated its flexibility as an in vivo and non-invasive imaging modality in AD/ADRD research – crucial for assessing for evidence of structural atrophy (neurodegeneration; [N] within the AT[N] framework), as well as changes in brain function. Unlike MRI, Positron emission tomography (PET) provides a non-invasive means to visualise and measure the density of biologically relevant molecular targets at sub-nanomolar concentrations, such as amyloid (A) and tau (T). In tandem, both MRI and PET are instrumental in vivo biomarker tools, in AD/ADRD research trials, reflecting AD pathology. During PET studies, radioactive ligands are injected into the bloodstream and bind to the molecular target in question, and their kinetics at the target can be quantified via the emitted gamma radiation captured using the PET scanner. PET uses various metrics to quantify the specific molecular target within ROI e.g., the standard uptake value ratio (SUV) – which is regional concentration of ligand, controlling for the individual’s weight and the dose injected (32). A refinement of this approach designed to partially address potential differences in peripheral kinetics of the radioligand, is the standardized uptake value ratio (SUVR), where the SUV in the target region is normalised to a reference region (33). While most PET ligands commonly used in AD research do not have a true reference region, the use of a pseudo-reference region, which is known to remain relatively stable, has provided significant improvements over the SUV.
Quantification of misfolded proteins such as amyloid and tau is particularly challenging in the context of typical multi-site studies conducted by a large number of imaging centres which require appropriate control of site/scanner qualification/set-up, acquisition, and harmonising analysis. Such studies demand robust analysis processes and pipelines to ensure adequate signal-to-noise and sufficient statistical power for PET studies that may be relatively expensive to conduct. The centiloid scale has recently been introduced, and widely adopted for both amyloid and tau radioligands, that has proven incredibly useful in standardising and harmonizing measurements between research studies and across different radioligands (34). Centiloid values are calculated by linearly scaling the tracer measurement to give a value ranging from 0 to 100 (30). A value of “0” refers to the average uptake of the measured tracer comparable to a “young control”, and on the other end of the scale, a value of “100” is representative of the average uptake found among “typical AD patients at the dementia stage” (34). Another significant advance has been the development of the AmyloidIQ and TauIQ methods that provide a generalised metric such as “amyloid load” (AβL) and have been demonstrated to provide significantly greater power than standard analytical methods (35-37).
PET is a widely used imaging modality in AD/ADRD research, that has proven its flexibility in targeting specific molecules. Two well-known types are fluorodeoxyglucose- (FDG) and amyloid-PET that have become ubiquitous research tools in the field and the subsequent focus of hundreds (if not thousands) of publications over the years. One of the first radio-ligands, FDG-PET has a wide range of clinical applications in multiple disease areas, as it allows the measurement of regional alterations in glucose metabolism (both hyper- and hypo-metabolism). On-the-other-hand, amyloid-PET provides quantitative data on the regional amyloid-beta (Aβ) plaque burden (38). Several amyloid tracers have recently been developed, with significantly increased half-lives, allowing for their industrial (vs site radiochemistry) production; indeed, three such radioligands are currently, widely available commercially: 18F-Florbetapir, 18F-Florbetaben, and 18F-Flutemetamol (38). There is no doubt that both FDG and amyloid-PET have vastly increased our knowledge and understanding of AD progression (and has been widely reported as such). As to not directly echo previous articles on the topic, this article will be instead focusing on tau-PET, a comparatively more novel approach. Tau-PET holds incredible promise in broadening our understanding of AD/ADRD even further, whilst also complimenting FDG- and amyloid-PET in multimodal imaging studies.
Tau-PET involves a tau-labelled radiotracer, allowing the visualisation of cerebral tau load. The existing tau ligands provide a good signal of 3R tau isoform, that features in AD, but have poor affinity for the 4R isoform that is more commonly present in tauopathies, such as progressive supranuclear palsy (PSP), and in frontotemporal dementia (FTD). The spread of tau pathology was until recently believed to cascade throughout the brain in a predictable or stereotypical pattern outlined by the Braak’s tau staging system. However, this has been questioned recently, with four distinct tau trajectory phenotypes proposed (39). Using from the first-generation tau ligand 18F-flortaucipir, the following spatiotemporal subtypes were proposed (a) limbic, (b) medial temporal load (MTL)-sparing, (c) posterior, and (d) lateral temporal. These features may broaden our understanding in explaining interindividual differences among patients across the AD continuum. The “limbic” pattern subtype was found to occur later and was the most frequently found among patients, demonstrating characteristics typically associated with AD (e.g., amnestic symptoms) and a greater proportion were Apolipoprotein ε4 (APOE ε4) carriers. By comparison, the patients with the “MTL-sparing” phenotype had a younger onset, the “posterior” pattern was associated with comparatively slower cognitive decline, and the “lateral temporal” subtype was associated with a more rapid clinical progression in multiple cognitive domains (39).
It has also recently been reported that the global signal intensity of tau-PET (but not Aβ–PET) predicted the level and spatial distribution of cortical atrophy over a one-year period (36). There appeared to be a sequential relationship between aggregated tau and so-called “downstream” neural degeneration leading the transition from MCI to AD (40). A clustering-based approach was used to identify three AD atrophy subtypes (a) “hippocampal-sparing (frontoparietal predominant)”, (b) “limbic-predominant (medial temporal lobe predominant)” and (c) “typical (temporal predominant)”. Using the 18F-Flortaucipir radiotracer, in MCI and AD patients, the tau burden was greatest among those displaying the hippocampal-sparing and “typical” atrophy subtypes, with the former patients showing the most rapid cognitive decline (one-year post-PET), and the latter, the most pronounced WMH volumes of the atrophy subtypes. Whereas, among those with “limbic-predominant” atrophy, tau burden was especially present within the entorhinal cortex. The sub-classification of atrophic subtypes may, thus, allow to disentangle disease heterogeneity, improve diagnosis capabilities, and support future clinical development (41).
The above outlined the utility of tau-PET to inform our understanding of the distribution of cortical atrophy, but tau-PET can also make valuable contributions alongside brain function in multimodal imaging studies. A notable recent example was the combinational use of resting state fMRI and tau PET data (42). Higher rates of tau accumulation were associated with increased resting-state functional connectivity and brain regions with higher-tau accumulation rates were preferentially connected to other regions that showed high tau accumulation (as with lower tau accumulation and low tau areas) suggesting a spatial affinity for its propagation between functionally connected regions (42).
As tau-PET is becoming increasingly recognised as a valuable tool, new “second generation” radiotracers are being developed, with higher affinity for tau aggregates, such as 18F-MK-6240, 18F-PI2620, and 18F-APN1607. These radiotracers were shown to also address the liability of first-generation ligands for offsite binding (presumed to be MAO-B). 18F-MK-6240 was found to accurately discriminate between ‘Aβ negative’ without cognitive impairment and those on the “AD continuum”, defined here as ‘Aβ positive’ with or without cognitive impairment. These results are promising in 18F-MK-6240 being utilised to broaden our understanding of AD progression. The authors concluded that longitudinal studies with larger sample sizes would be needed to address the large variance in the results of their study (43).
Another recently developed tau radiotracer that has promising applications in AD research is 18F-PI-2620. It has shown a high binding affinity to aggregated tau in the brain that bolstered a high signal-to-noise ratio in both AD and HC participants with good tolerance (44).
Ascertaining the precise aetio-pathogenic mechanisms of AD/ADRD presents an undoubtedly complex problem but imaging protocols, using multiple tracers may be used during a study protocol to better understand the nature of the spatial aggregation of AD pathology and it’s sequencing along numerous clinical pathways. Recent evidence suggests that, when measuring tau in the cerebrospinal fluid (CSF) and amyloid- and tau-PET, CSF measuring phosphorylated tau (p-CSF) and CSF total-tau started to increase before the threshold for amyloid PET “positivity” (45), whereas other studies suggest that tau-PET start to progress after amyloid PET positivity. The effects of amyloid PET on tau-PET may be mediated by phosphorylated tau species, with high p-CSF levels predicting increases in tau-PET uptake levels (45). This expands on Braak’s staging of neurofibrillary pathology in that tauopathy – cytoskeletal alterations within neurons – (a) follows a “predictable sequencing pattern” that tends to stem from the transentorhinal region (and the basal temporal neocortex), and (b) these intraneuronal changes (e.g., NFTs) precede the aggregation of insoluble amyloid deposits (29). The aggregation and accumulation of NFTs and Aβ are not mutually exclusive events but the understanding of the pathological temporal sequencing of AD across multiple stages is undoubtedly important for developing new treatments and interventions for AD.
As the field is increasingly recognising the role of vascular risk factors in AD development, multimodal imaging studies have begun to explore the synergistic relationships between AD pathological markers and vascular risk. In a study involving preclinical older adults, a significant interaction was found between higher measures of vascular risk and higher levels of Aβ burden in most regions (bar the entorhinal cortex) that was then associated with increased tau levels (46). These results were gathered using a combination of the “Framingham Heart Study cardiovascular disease risk” to measure “vascular risk” (measures of body mass index, history of diabetes, smoking behaviours, among others), 18F-Flortaucipir (tau), and 11C-Pittsburgh Compound-B (PiB) PET (Aβ), that raises the interesting possibility of increased vascular risk inducing a “second hit” – “that further potentiates the spread of Aβ-related neocortical tau pathology” (46). Elevated vascular risk may, thus, present an important avenue in future AD intervention studies and/or to attenuate the impact of AD-related pathology in older adults.
It may be tempting to focus research and development efforts primarily on tau, given the history of unsuccessful drug development trials focusing on anti-amyloid targeting medication – that is, until the recent FDA “accelerated” approval of aducanumab (47). However, several recent studies emphasise the significant role that Aβ seems to play in the temporal sequencing of events leading to a clinical diagnosis of AD. A recent study on the histopathological interactions between global amyloid and medial temporal neuroinflammation (48) reported that, in relation to tau spread, amyloid appeared to interact with transentorhinal neuroinflammation which, in turn, triggered the spread of tau across the neocortex (Braak stages II-III), followed by a subsequent spread of tau across Braak stages III-IV and the associated brain regions. A sentiment that was supported a year later, among a group of researchers who concluded that a “moderate” Aβ level is required for tau within the neocortex to be detectable, and that >40 centiloid of global Aβ burden is required to induce an accelerated spread of tau (49).
Another potential PET target is neuroinflammation. Targeted PET radiotracers can be used to visualise the presence of activated microglia that respond to inflammation and injury, within the central nervous system (50). A recent immunofluorescence study, using the AD PDAPPJ20 mice model, found evidence of “autophagic impairment”; as the mice aged, the hippocampus showed lower phagocytic activity when exposed to Aβ, indicative of an impairment to effectively promote Aβ clearance from the brain. “Impaired autophagy and lysosome dysfunction” in AD mice may provide a novel research avenue for future studies (51).
The initial radiotracers aimed at visualizing microglial activation and neuroinflammation targeted the translocator protein 18kDa (TSPO). The first TSPO tracer was 11C-PK11195 followed by 11C-PBR28, 18F-DPA-714 and several others.
In attempting to gain insight into the relationship between neuroinflammation and AT[N] biomarkers, a recent longitudinal study used 11C-PK11195, 11C-PiB PET (amyloid) and 18F-Flortaucipir (tau) in MCI patients, over a two years’ period (52). The authors reported that among MCI patients displaying with low Aβ burden at baseline, but subsequently rising levels of such, showed higher microglial activation that were then found to decline as the individuals approached Aβ loads at “AD levels”. As tau tangles form later along the temporal pathological sequencing in Aβ+ individuals, increased tau aggregation was then associated with higher levels of neuroinflammation. The “two peak hypothesis” was therefore proposed, stipulating that the first peak is driven by Aβ aggregation, and the second peak being driven by the accumulation of tau tangles (52).
Non-TSPO neuroinflammation radiotracers have also been developed. For example, the PET tracer 11C-BU99008, targeting imidazoline 2 binding sites (I2-BS; found mainly in the mitochondria) was used recently to explore astrocyte reactivity among MCI due to AD or AD patients, compared to healthy controls (53). The authors found a relationship between 11C-BU99008 and 18F-Florbetaben uptake (in support of previous research finding a link between increased astrocytic activity and above threshold Aβ deposition). A further key finding was the observed increased 11C-BU99008 uptake among a significant portion of those classified as “Aβ-negative”, suggesting that astrocyte reactivity may still occur at below-threshold Aβ burden. These findings may add credence to the potential value of neuroinflammation as a target for therapeutic intervention.
The development and application of other novel radioligands allow for the exploration of other potential mechanistic pathways that can lead to AD development. A specific example are novel PET tracers that attempt to elucidate the complex relationships between the dysregulation of mitochondria in older adults in connection with alterations in energy metabolism and oxidative stress. Recently, two PET tracers have been developed for in vivo human studies of mitochondrial function: 18F-BCPP-EF and 11C-SA4503 (54). The mitochondrial complex-1 (MC-1) is the first enzyme in the electron transport chain (ETC) critical for the generation of mitochondrial adenosine triphosphate (ATP), and a major source of reactive oxygen species in the cell. [18F]BCPP-EF binding reflects electron transport chain (ETC)-related mitochondrial function. The σ-1 receptor (σ1R) is a chaperone protein enhancing Ca2+ influx from the endoplasmic reticulum into the mitochondria, thus modulating mitochondrial (ATP) production (54). These tracers demonstrated high reliability in the quantification measurements of mitochondrial function in the brains of healthy human participants and have been applied to AD (and other neurodegenerative diseases) research over the last few years.
A notable example is with a recent study aimed to assess for mitochondrial and glycolytic impairments in patients among those with early to moderate AD (55). The core aim was to explore whether there are regional differences of glycolysis (the breakdown of glucose and measured via [18F]FDG) and of mitochondrial oxidative activity ([18F]BCPP-EF) among AD patients. The authors found reductions in [18F]BCPP-EF binding within various regions of the temporal (both medial and lateral) cortex; the parahippocampus appeared particularly susceptible to alterations in [18F]BCPP-EF availability, whereas the same region was least affected by [11C]PiB. Thus, in the context of the temporal sequencing of AD, parahippocampal mitochondrial dysfunction may precede hypometabolism (at least in this brain region).
Finally, to further illustrate the flexibility of PET as an in vivo tool for AD and dementia research, the [11C]UCB-J radiotracer has been developed to target synaptic vesicle glycoprotein 2 (SV2; a presynaptic membrane glycoprotein expressed in almost all synapses) that can be used to measure the distribution of synaptic density across the brain (56). [11C]UCB-J was used to identify evidence of synaptic pathology among older adults with a diagnosis of amnestic MCI (aMCI; (56). The research was conducted in response to prior results finding localised reductions in SV2A binding within the hippocampus among those with aMCI and mild AD (56, 57), found an inverse correlation between global Aβ deposition ([11C] PiB) and hippocampal synaptic density ([11C]UCB-J) among those with aMCI but not with dementia. The authors concluded that this may be explained by Aβ reaching a “relative plateau as a pool of primarily insoluble fibrillar Aβ, a point at which Aβ may uncouple from neurodegenerative processes including synaptic loss”. However, it is worth emphasising that 11C radiotracers (e.g., those exemplified in this commentary) are often limited by their incredibly short half-lives, and therefore, require an on-site cyclotron which is not feasible for most clinical research centres.


Retinal Imaging (RI)

The neurosensory retina, described as ‘a window to the brain’, given its direct connection to the central nervous system via the retinal ganglion cells (RGC) and ease-of-access, is emerging as a potentially valuable medium for studying AD in vivo (58). Given the eye-brain link, it is not surprising that the retina manifests similar pathological attributes that are evident in the neurodegenerative process. Indeed, AD patients present with retinal degenerative abnormalities such as RGC preferential loss, thinning of the optic nerve and vascular changes have been found in the retina of patients with MCI and AD (59). Increasing evidence also confirmed the presence of canonical AD pathologies (Aβ and tau) within the retina and optic nerve (60, 61).
Posited as low-cost and non-invasive, retinal imaging technologies are emerging as potentially useful AD biomarker tools in AD research. One such example is Optical Coherence Tomography (OCT), an imaging technique widely used in ophthalmology to generate high-resolution 2D/3D images of retinal anatomy over a wide field of view. Notably, recent OCT studies in MCI and AD patients have shown evidence of retinal fibre layer (RNFL) thinning, RGC loss and micro-vascular abnormalities (62, 63). Furthermore, in a prospective longitudinal setting, lower RNFL thickness at baseline was associated with increased risk of subsequent cognitive decline and dementia (64, 65). However, lack of a standardization in OCT studies is a major limitation in its wider use as a clinical research tool (66).
Confocal scanning laser ophthalmoscopy (cSLO), allowing for retinal fluorescence scanning, represents another promising new RI technology in AD research, as it has been shown to directly visualize amyloid deposits in the retina of MCI and mild to moderate AD patients, previously given oral Curcumin (67). Indeed, the food additive curcumin has natural fluorescent properties and binds to fibrillary Aβ thus enabling the identification of plaques with non-invasive retinal fluorescent imaging. Recently, Ngolab et al. (68) used this technique to compare the retinal (curcumin binded) with brain Aβ burden, as detected by PET, in a small sample of asymptomatic participants above and below amyloid threshold; they observed a distinct discriminatory capacity for distinguishing cognitively healthy older adults with high cerebral amyloid load from those with below-threshold amyloid load. Recent evidence suggests that amyloid retinal accumulation may be detected without the use of Curcumin or another reagent, through blue-light autofluorescence scanning (69). This new technique is currently being adopted in our on-going, industry funded, CHARIOT PRO study at Imperial College London (70) and the recently initiated UK-wide “Deep and Frequent Phenotyping (DFP)” study. Assuming that the above preliminary findings are further validated in these and other large-scale prospective longitudinal studies, retinal fluorescence scanning, and OCT may become important low-cost and non-invasive RI techniques allowing for risk assessment-for-AD in cognitively healthy individuals and in pre-selecting potential RCT participants for the far more expensive and more invasive PET and/or CSF studies for determining AT[N] status.


Expanding role of Machine Learning and Artificial Intelligence in Imaging analysis

Given the mammoth amount of information collected during the acquisition process, imaging is fundamentally a “big data” problem. Analysing such high-dimensional data collected from multiple participants’ brain slices, often over several sessions, and even against the data of other imaging methods may be daunting but by being cognisant of the tools available to handle such a challenge is key; and one such approach is via ML.
Amongst many ML techniques, a recent study combining convolutional neural networks (CNN) and ensemble learning (EL) found CNN as an effective tool in automated feature learning with the use of a variety of multilayer perceptrons”, whilst EL effectively integrates multiple models (71). The authors adopted a CNN-EL method alongside structural MRI data to classify participants between (a) those with AD vs. HC, (b) MCI-converters and HC, and (c) MCI-converters vs. MCI-nonconverters. This data-driven approach proved more accurate in distinguishing the AD, MCI-converters, and HC control groups (but not MCI-converters vs. MCI-nonconverters) compared to other similar methodologies.
Furthermore, ‘deep learning’ provides an effective means to analyse data collected from multiple sources, in this case imaging (MR), genetic (single nucleotide polymorphisms), and clinical testing data with the aim to support AD staging analysis (72). Deep learning algorithms are advantageous over “shallow” methods due to their ability to learn which features are most predictive for a particular outcome (in this case AD staging using multimodal data). In short, the researchers found that integrating data from multiple sources improved the predictive accuracy between different AD stages (cognitively normal, MCI, and AD). Supporting deep learning models in future AD trials to improve the predictive accuracy in staging categorisation (72).


Future Directions

Despite years of exhaustive research and thousands of publications, we have still not fully elucidated the biological mechanisms and processes underpinning the development of AD and ADRDs, nor the extent and mechanisms underlying its heterogeneity. A paradigm shift going beyond the exclusive examination of AT[N] biomarkers is warranted to reflect on the strategic long-term considerations in how best to use neuroimaging in AD/ADRD research in the future to support our goals in dementia prevention, and as such we suggest a three-pronged approach:
1. Consolidating the wealth of new information that neuroimaging studies continue to give us and by using emerging imaging methods to promote the discovery of new biomarkers that may provide important insight into AD’s underlying causes and reasons behind its heterogeneity within studies with long follow-ups and appropriate sample sizes.
2. It is becoming increasingly clear that multimodal neuroimaging studies, possibly coupled with genetic and other fluid biomarkers studies, may be necessary to untangle the effects of multiple risk factors on AD development. Novel imaging protocols may prove instrumental in uncovering the underlying neural mechanisms behind AD that cannot be achieved using singular imaging methods alone. With emerging neuro-imaging and other multi-modal biomarker discovery and validation work, novel analytic ML/artificial intelligence (and other cutting-edge methodologies) are urgently required and are, indeed, beginning to be applied to AD/ADRD research.
In dementia prevention, whereas MRI and PET methodologies remain instrumental tools in randomised clinical trial (RCT) protocols, in pre-clinical AD stages, their use for risk status assessment in the context of preventative public health strategies in the general population is neither feasible nor clinically relevant. Further research is warranted towards the development and validation of low-cost and non- invasive technologies, such as retinal imaging, plasma biomarkers and genetics, that are amenable to a wider implementation. Such technologies may also be of value in the pre-screening phase of RCTs for selection of candidates for the more expensive and/or invasive neuroimaging or CSF studies.
3. Neuroimaging is by no means a standardised practice at this point in time (73). Given the sheer magnitude of decisions made at various stages, whether it is calibrating the technology or analysing the vast quantities of data, these decisions can significantly affect the outcomes and results. Consequently, often poor reproducibility between research teams remains an ongoing issue. It is only by coming together to determine ‘best practice’ procedures in how imaging data is collected, recorded, and analysed, can then be disseminated with confidence.

In conclusion, whereas singular neuroimaging techniques may effectively address hypothesis specific objectives, the wealth of options and emerging broad capabilities of neuroimaging makes it invaluable in clinical research to enhance our understanding of this complex and potentially highly heterogeneous disease. The choice is dependent upon the desired outcomes of the clinical research, and the right “tool” for the job depends on the targeted hypothesis and explicit outcome measure(s). The fast pace of the field of neuroimaging in new modalities, more effective data acquisition protocols, and integrative approaches means that brain imaging will continue to play an important part in AD/ADRD research. By reflecting on recent developments in this field, we can make a unified effort to consolidate and branch into new research avenues in the attempts to disentangle the complexity of this relentlessly progressive and devastating disease.


Conflict of Interest: Prof. Lefkos Middleton has received research funding (to Institution) from Janssen , Merck (USA), Gates, Takeda/ Millenium, Novartis, EIT Health, and Invincro, outside the submitted work. Prof M Politis has received grants from Michael J Fox Foundation , City Electrical Factors, Roche Pharmaceuticals, Life Molecular Imaging, Invicro , personal fees from Roche Pharmaceuticals and from Movement Disorder Society, outside the submitted work. Dr E A Rabiner is a full time employee of Invicro, a company conducting commercial studies to support industry and academic research.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.



1. Leuzy A, Heurling K, Ashton NJ, Schöll M, Zimmer ER. In vivo Detection of Alzheimer’s Disease. Yale J Biol Med. 2018 Sep;91(3):291–300.
2. Rajasekhar K, Govindaraju T. Current progress, challenges and future prospects of diagnostic and therapeutic interventions in Alzheimer’s disease. RSC Adv. 2018;8(42):23780–804.
3. Raichle ME. A brief history of human brain mapping. Trends in Neurosciences. 2009 Feb;32(2):118–26.
4. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group* under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology. 1984 Jul 1;34(7):939–939.
5. Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, Fagan AM, et al. Toward defining the preclinical stages of Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimer’s & Dementia. 2011 May;7(3):280–92.
6. McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Kawas CH, et al. The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimer’s & Dementia. 2011 May;7(3):263–9.
7. Isaacson, R. and Saif, N., 2020. A missed opportunity for dementia prevention? current challenges for early detection and modern-day solutions. The Journal of Prevention of Alzheimer’s Disease, pp.1-3.
8. Jack CR, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimer’s & Dementia. 2018, Apr;14(4):535–62.
9. McRae-McKee K, Udeh-Momoh CT, Price G, Bajaj S, de Jager CA, Scott D, et al. Perspective: Clinical relevance of the dichotomous classification of Alzheimer’s disease biomarkers: Should there be a “gray zone”? Alzheimer’s & Dementia. 2019 Oct;15(10):1348–56.
10. Dubois B, Villain N, Frisoni GB, Rabinovici GD, Sabbagh M, Cappa S, et al. Clinical diagnosis of Alzheimer’s disease: recommendations of the International Working Group. The Lancet Neurology. 2021 Jun;20(6):484–96.
11. Zhang, X., Tian, Y., Wang, Z., Ma, Y., Tan, L. and Yu, J., 2021. The Epidemiology of Alzheimer’s Disease Modifiable Risk Factors and Prevention. The Journal of Prevention of Alzheimer’s Disease, pp.1-9.
12. Daunt, P., Ballard, C., Creese, B., Davidson, G., Hardy, J., Oshota, O., Pither, R. and Gibson, A., 2020. Polygenic Risk Scoring is an Effective Approach to Predict Those Individuals Most Likely to Decline Cognitively Due to Alzheimer’s Disease. The Journal of Prevention of Alzheimer’s Disease, pp.1-6.
13. Keshavan, A., 2020. Population-based blood screening for preclinical Alzheimer’s disease in a British birth cohort at age 70. Brain, 144(2), pp.434–449.
14. Jack CR, Bennett DA, Blennow K, Carrillo MC, Feldman HH, Frisoni GB, et al. A/T/N: An unbiased descriptive classification scheme for Alzheimer disease biomarkers. Neurology. 2016 Aug 2;87(5):539–47.
15. The Australian Imaging Biomarkers and Lifestyle Flagship Study of Ageing, Roe JM, Vidal-Piñeiro D, Sørensen Ø, Brandmaier AM, Düzel S, et al. Asymmetric thinning of the cerebral cortex across the adult lifespan is accelerated in Alzheimer’s disease. Nat Commun. 2021 Dec;12(1):721.
16. Planche V, Bouteloup V, Mangin J, Dubois B, Habert M, Tison F, et al. Alzheimer’s disease brain atrophy subtypes are associated with incident dementia in subjective cognitive complaint and mild cognitive impairment: Neuroimaging / Optimal neuroimaging measures for early detection. Alzheimer’s &amp; Dementia [Internet]. 2020 Dec [cited 2021 May 30];16(S5). Available from:
17. Mofrad S, Lundervold A, Vik A, Lundervold A. Cognitive and MRI trajectories for prediction of Alzheimer’s disease. Scientific Reports. 2021;11(1).
18. Behfar Q, Behfar SK, von Reutern B, Richter N, Dronse J, Fassbender R, et al. Graph Theory Analysis Reveals Resting-State Compensatory Mechanisms in Healthy Aging and Prodromal Alzheimer’s Disease. Front Aging Neurosci. 2020 Oct 22;12:576627.
19. Prakash R, McKenna M, Gbadeyan O, Shankar A, Andridge R, Scharre D. A Whole-Brain Functional Connectivity Model of Alzheimer’s Disease Pathology. 2021;.
20. Cheng Y, Chiu M, Chen Y, Cheng T, Lai Y, Chen T. The contribution of vascular risk factors in neurodegenerative disorders: from mild cognitive impairment to Alzheimer’s disease. Alzheimer’s Research & Therapy. 2020;12(1).
21. Rivera-Rivera LA, Cody KA, Eisenmenger L, Cary P, Rowley HA, Carlsson CM, et al. Assessment of vascular stiffness in the internal carotid artery proximal to the carotid canal in Alzheimer’s disease using pulse wave velocity from low rank reconstructed 4D flow MRI. J Cereb Blood Flow Metab. 2021 Feb;41(2):298–311.
22. Ohene Y, Harrison IF, Evans PG, Thomas DL, Lythgoe MF, Wells JA. Increased blood–brain barrier permeability to water in the aging brain detected using noninvasive multi-TE ASL MRI. Magn Reson Med. 2021 Jan;85(1):326–33.
23. Badji A, Westman E. Cerebrovascular pathology in Alzheimer’s disease: Hopes and gaps. Psychiatry Research: Neuroimaging. 2020 Dec;306:111184.
24. Tubi MA, Feingold FW, Kothapalli D, Hare ET, King KS, Thompson PM, et al. White matter hyperintensities and their relationship to cognition: Effects of segmentation algorithm. NeuroImage. 2020 Feb;206:116327.
25. Tuovinen T, Kananen J, Rajna Z, Lieslehto J, Korhonen V, Rytty R, et al. The variability of functional MRI brain signal increases in Alzheimer’s disease at cardiorespiratory frequencies. Sci Rep. 2020 Dec;10(1):21559.
26. Vogt NM, Hunt JF, Adluru N, Dean DC, Johnson SC, Asthana S, et al. Cortical Microstructural Alterations in Mild Cognitive Impairment and Alzheimer’s Disease Dementia. Cereb Cortex. 2020 May 14;30(5):2948–60.
27. Basser P, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophysical Journal. 1994;66(1):259-267.
28. Roy M, Dumont M, Houde J, Morency F, Descoteaux M. Longitudinal assessment of neuroinflammation and axonal loss in white matter tracts in Alzheimer’s disease. Alzheimer’s & Dementia. 2020;16(S5).
29. Braak H, Del Tredici K. Alzheimer’s disease: intraneuronal alterations precede insoluble amyloid-β formation. Neurobiology of Aging. 2004 Jul;25(6):713–8.
30. Wilson H, Politis M, Rabiner EA, Middleton LT. Novel PET Biomarkers to Disentangle Molecular Pathways across Age-Related Neurodegenerative Diseases. Cells. 2020 Dec 2;9(12):2581.
31. Liu Y, Nguyen M, Robert A, Meunier B. Metal Ions in Alzheimer’s Disease: A Key Role or Not? Acc Chem Res. 2019 Jul 16;52(7):2026–35.
32. Innis R, Cunningham V, Delforge J, Fujita M, Gjedde A, Gunn R et al. Consensus Nomenclature for in vivo Imaging of Reversibly Binding Radioligands. Journal of Cerebral Blood Flow & Metabolism. 2007;27(9):1533-1539.
33. Gunderson, L & Tepper, J. Clinical Radiation Oncology 4th Edition. 2015.
34. for the Alzheimer’s Disease Neuroimaging Initiative, for the ALFA Study, Salvadó G, Molinuevo JL, Brugulat-Serrat A, Falcon C, Grau-Rivera O, et al. Centiloid cut-off values for optimal agreement between PET and CSF core AD biomarkers. Alz Res Therapy. 2019 Dec;11(1):27.
35. Whittington A, Sharp D, Gunn R. Spatiotemporal Distribution of β-Amyloid in Alzheimer Disease Is the Result of Heterogeneous Regional Carrying Capacities. Journal of Nuclear Medicine. 2017;59(5):822-827.
36. Whittington A, Gunn R. Amyloid Load: A More Sensitive Biomarker for Amyloid Imaging. Journal of Nuclear Medicine. 2018;60(4):536-540.
37. Whittington A, Gunn R. TauIQ: A Canonical Image Based Algorithm to Quantify Tau PET Scans. Journal of Nuclear Medicine. 2021;62(9):1292-1300.
38. Chételat G, Arbizu J, Barthel H, Garibotto V, Law I, Morbelli S, et al. Amyloid-PET and 18F-FDG-PET in the diagnostic investigation of Alzheimer’s disease and other dementias. The Lancet Neurology. 2020 Nov;19(11):951–62.
39. Vogel JW, Young AL, Oxtoby NP, Smith R, Ossenkoppele R, Strandberg OT, et al. Characterizing the spatiotemporal variability of Alzheimer’s disease pathology [Internet]. Neurology; 2020 Aug [cited 2021 Jul 11]. Available from:
40. Ossenkoppele R, Lyoo CH, Sudre CH, Westen D, Cho H, Ryu YH, et al. Distinct tau PET patterns in atrophy-defined subtypes of Alzheimer’s disease. Alzheimer’s & Dementia. 2020 Feb;16(2):335–44.
41. La Joie R, Visani AV, Baker SL, Brown JA, Bourakova V, Cha J, et al. Prospective longitudinal atrophy in Alzheimer’s disease correlates with the intensity and topography of baseline tau-PET. Sci Transl Med. 2020 Jan 1;12(524).
42. Alzheimer’s Disease Neuroimaging Initiative (ADNI), Franzmeier N, Neitzel J, Rubinski A, Smith R, Strandberg O, et al. Functional brain architecture is associated with the rate of tau accumulation in Alzheimer’s disease. Nat Commun. 2020 Dec;11(1):347.
43. Krishnadas, N., Dore, V., Mulligan, R., Tyrrell, R., Bozinovski, S., Huang, K., Lamb, F., Burnham, S., Villemagne, V., & Rowe, C. 18F-MK6240 longitudinal tau PET in ageing and Alzheimer’s disease. 2021 Retrieved from:
44. Mueller A, Bullich S, Barret O, Madonia J, Berndt M, Papin C, et al. Tau PET imaging with 18F-PI-2620 in Patients with Alzheimer Disease and Healthy Controls: A First-in-Humans Study. J Nucl Med. 2020 Jun;61(6):911–9.
45. Mattsson-Carlgren N, Andersson E, Janelidze S, Ossenkoppele R, Insel P, Strandberg O, et al. Aβ deposition is associated with increases in soluble and phosphorylated tau that precede a positive Tau PET in Alzheimer’s disease. Sci Adv. 2020 Apr;6(16):eaaz2387.
46. Rabin J, Yang H, Schultz A, Hanseeuw B, Hedden T, Viswanathan A et al. Vascular Risk and β-Amyloid Are Synergistically Associated with Cortical Tau. Annals of Neurology. 2019;85(2):272-279.
47. Cummings J, Aisen P, Lemere C, Atri A, Sabbagh M, Salloway S. Aducanumab produced a clinically meaningful benefit in association with amyloid lowering. Alzheimer’s Research & Therapy. 2021;13(1).
48. Pascoal T, Kang M, Therriault J, Benedet A, Chamoun M, Lussier F et al. Medial temporal neuroinflammation unleashes tau spreading over the neocortex. Alzheimer’s & Dementia. 2020;16(S4).
49. Doré V, Krishnadas N, Bourgeat P, Huang K, Li S, Burnham S et al. Relationship between amyloid and tau levels and its impact on tau spreading. European Journal of Nuclear Medicine and Molecular Imaging. 2021;.
50. Eme-Scolan E, Dando SJ. Tools and Approaches for Studying Microglia In vivo. Front Immunol. 2020 Oct 7;11:583647.
51. Pomilio C, Gorojod RM, Riudavets M, Vinuesa A, Presa J, Gregosa A, et al. Microglial autophagy is impaired by prolonged exposure to β-amyloid peptides: evidence from experimental models and Alzheimer’s disease patients. GeroScience. 2020 Apr;42(2):613–32.
52. Ismail R, Parbo P, Madsen LS, Hansen AK, Hansen KV, Schaldemose JL, et al. The relationships between neuroinflammation, beta-amyloid and tau deposition in Alzheimer’s disease: a longitudinal PET study. J Neuroinflammation. 2020 Dec;17(1):151.
53. Calsolaro V, Matthews P, Donat C, Livingston N, Femminella G, Guedes S et al. Astrocyte reactivity with late-onset cognitive impairment assessed in vivo using 11C-BU99008 PET and its relationship with amyloid load. Molecular Psychiatry. 2021;.
54. Mansur, A., Rabiner, E., Comley, R., Lewis, Y., Middleton, L., Huiban, M., Passchier, J., Tsukada, H. and Gunn, R., 2019. Characterization of 3 PET Tracers for Quantification of Mitochondrial and Synaptic Function in Healthy Human Brain: 18F-BCPP-EF, 11C-SA-4503, and 11C-UCB-J. Journal of Nuclear Medicine, 61(1), pp.96-103.
55. Terada T, Obi T, Bunai T, Matsudaira T, Yoshikawa E, Ando I et al. In vivo mitochondrial and glycolytic impairments in patients with Alzheimer disease. Neurology. 2020;94(15):e1592-e1604.
56. O’Dell R, Mecca A, Chen M, Naganawa M, Toyonaga T, Lu Y et al. Association of Aβ deposition and regional synaptic density in early Alzheimer’s disease: a PET imaging study with [11C]UCB-J. Alzheimer’s Research & Therapy. 2021;13(1).
57. Chen M, Mecca A, Naganawa M, Finnema S, Toyonaga T, Lin S et al. Assessing Synaptic Density in Alzheimer Disease With Synaptic Vesicle Glycoprotein 2A Positron Emission Tomographic Imaging. JAMA Neurology. 2018;75(10):1215.
58. Koronyo-Hamaoui M, Koronyo Y, Ljubimov A, Miller C, Ko M, Black K et al. Identification of amyloid plaques in retinas from Alzheimer’s patients and noninvasive in vivo optical imaging of retinal plaques in a mouse model. NeuroImage. 2011;54:S204-S217.
59. Blanks J, Schmidt S, Torigoe Y, Porrello K, Hinton D, Blanks R. Retinal pathology in Alzheimer’s disease. II. Regional neuron loss and glial changes in GCL. Neurobiology of Aging. 1996;17(3):385-395.
60. Liao H, Zhu Z, Peng Y. Potential Utility of Retinal Imaging for Alzheimer’s Disease: A Review. Frontiers in Aging Neuroscience. 2018;10.
61. Chiasseu M, Alarcon-Martinez L, Belforte N, Quintero H, Dotigny F, Destroismaisons L et al. Tau accumulation in the retina promotes early neuronal dysfunction and precedes brain pathology in a mouse model of Alzheimer’s disease. Molecular Neurodegeneration. 2017;12(1).
62. Mahajan D, Votruba M. Can the retina be used to diagnose and plot the progression of Alzheimer’s disease?. Acta Ophthalmologica. 2017;95(8):768-777.
63. Chan V, Sun Z, Tang S, Chen L, Wong A, Tham C et al. Spectral-Domain OCT Measurements in Alzheimer’s Disease. Ophthalmology. 2019;126(4):497-510.
64. Ko F, Muthy Z, Gallacher J, Sudlow C, Rees G, Yang Q et al. Association of Retinal Nerve Fiber Layer Thinning With Current and Future Cognitive Decline. JAMA Neurology. 2018;75(10):1198.
65. Mutlu U, Colijn J, Ikram M, Bonnemaijer P, Licher S, Wolters F et al. Association of Retinal Neurodegeneration on Optical Coherence Tomography With Dementia. JAMA Neurology. 2018;75(10):1256.
66. Mukherjee C, Al-Fahad Q, Elsherbiny S. The role of optical coherence tomography in therapeutics and conditions, which primarily have systemic manifestations: a narrative review. Therapeutic Advances in Ophthalmology. 2019;11:251584141983115.
67. Dumitrascu O, Lyden P, Torbati T, Sheyn J, Sherzai A, Sherzai D et al. Sectoral segmentation of retinal amyloid imaging in subjects with cognitive decline. Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring. 2020;12(1).
68. Ngolab J, Donohue M, Belsha A, Salazar J, Cohen P, Jaiswal S et al. Feasibility study for detection of retinal amyloid in clinical trials: The Anti-Amyloid Treatment in Asymptomatic Alzheimer’s Disease (A4) trial. Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring. 2021;13(1).
69. Snyder P, Johnson L, Lim Y, Santos C, Alber J, Maruff P et al. Nonvascular retinal imaging markers of preclinical Alzheimer’s disease. Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring. 2016;4(1):169-178.
70. Udeh-Momoh C, Watermeyer T, Price G, de Jager Loots C, Reglinska-Matveyev N, Ropacki M et al. Protocol of the Cognitive Health in Ageing Register: Investigational, Observational and Trial Studies in Dementia Research (CHARIOT): Prospective Readiness cOhort (PRO) SubStudy. BMJ Open. 2021;11(6):e043114.
71. Pan D, Zeng A, Jia L, Huang Y, Frizzell T, Song X. Early Detection of Alzheimer’s Disease Using Magnetic Resonance Imaging: A Novel Approach Combining Convolutional Neural Networks and Ensemble Learning. Front Neurosci. 2020 May 13;14:259.
72. Venugopalan J, Tong L, Hassanzadeh HR, Wang MD. Multimodal deep learning models for early detection of Alzheimer’s disease stage. Sci Rep. 2021 Dec;11(1):3254.
73. Knudsen G, Ganz M, Appelhoff S, Boellaard R, Bormans G, Carson R et al. Guidelines for the content and format of PET brain data in publications and archives: A consensus paper. Journal of Cerebral Blood Flow & Metabolism. 2020;40(8):1576-1585.



K.A. Abbott1, J.M. Posma2, I. Garcia-Perez2, C. Udeh-Momoh1, S. Ahmadi-Abhari1, L. Middleton1,3, G. Frost2


1. Age & Epidemiology Research Unit, School of Public Health, Faculty of Medicine, Imperial College London, UK; 2. Department of Metabolism, Digestion and Reproduction, Faculty of Medicine, Imperial College London, UK; 3. Primary Care and Public Health Directorate, Charing Cross Hospital, Imperial College Healthcare NHS Trust, UK

Corresponding Author: Prof Gary Frost, Department of Metabolism, Digestion and Reproduction, Faculty of Medicine, Imperial College London Hammersmith Campus, Commonwealth Building, Du Cane Road, London W12 ONN, United Kingdom, Tel: +44 (0)20 7594 0959. Fax: +44 (0208383 8320, Email :

J Prev Alz Dis 2022;
Published online January 11, 2022,



Increasing evidence proposes diet as a notable modifiable factor and viable target for the reduction of Alzheimer’s Disease risk and age-related cognitive decline. However, assessment of dietary exposures is challenged by dietary capture methods that are prone to misreporting and measurement errors. The utility of -omics technologies for the evaluation of dietary exposures has the potential to improve reliability and offer new insights to pre-disease indicators and preventive targets in cognitive aging and dementia. In this review, we present a focused overview of metabolomics as a validation tool and framework for investigating the immediate or cumulative effects of diet on cognitive health.

Key words: Metabolomics, dietary assessments, nutrition, nutriome, Alzheimer’s disease, dementia, evidence-based tools, precision nutrition.



A steady but certain shift towards the integration of evidence-based tools to enhance traditional methods of dietary assessment is currently emerging in the field of nutrition science. This is noteworthy, because much of our understanding of dietary associations in relation to Alzheimer’s disease (AD) and other late onset dementia (LOD) forms has derived from self-administered dietary assessments which are prone to biases and reporting errors.
Specifically, the integration of -omics technologies into the evaluation of dietary exposures heralds a new era of research that can be more robust with potential to improve reliability and offer novel insights to pre-disease indicators and preventive targets in cognitive aging and dementia.
Recent years saw a plethora of publications from non-pharmacological interventional studies such as the FINGER randomized controlled trial (RCT) providing evidence that diet, as part of a multi-domain lifestyle intervention, can contribute to significantly mitigate decline in cognitive performance and potentially delay the onset or progression of AD and other types of dementia (1). These promising results, based on simultaneous management of several vascular and lifestyle related risk factors, in which the intervention group were assigned a multidomain intervention consisting of diet, exercise, cognitive training, vascular risk monitoring versus the control group assigned general health advice, showed efficacy after a 2 year period in improving cognitive performance in at-risk, but cognitively unimpaired and/or mildly impaired individuals (2). This was pivotal because interest in lifestyle modifications to reduce disease prevalence was affirmed and personalised strategies/optimal timing became discussion points.
The increasing awareness that pathological changes in AD-LODs develop many years prior to clinical disease onset and the results of several observational and interventional studies has led the field to acknowledge that the pre-clinical or early clinical stages are the optimal time points for intervention; thus, attention has shifted towards health-promoting behaviours such as healthy diet, maintaining an optimal weight, physical and mental activities and social interactions, in addition to potential pharmacological therapies, if and when novel disease modifying medicines of proven efficacy and cost effectiveness become available.
Diet and nutrition are viable targets for strategies aimed at preserving cognitive health in older adults and have become a focus in dementia related studies. Moreover, there has been a gradual shift from single nutrient analyses towards dietary pattern analyses, reflecting trends in nutritional epidemiology, in which synergistic effects of food combinations and possible nutrient interactions are deemed more informative (3).
Despite accruing evidence from a wealth of epidemiological studies, showing that adherence to healthy dietary patterns, such as the Mediterranean, Nordic, DASH, MIND or anti-inflammatory diets may lend neuroprotective effects, and more recently, the ketogenic diet, in its capacity to alter brain metabolism (4-7), these findings have not translated uniformly into dietary guidelines, to improve cognitive health and disease burden reduction in older adults, due to inconsistencies in the literature.
In order to understand the precise effects of dietary intake on cognitive function and their mechanism of action, dietary assessment methodologies must be able to measure dietary intake as accurately as possible. As it stands, current methods of dietary assessment typically collect self-reported data through food-frequency questionnaire (FFQ), 24-hour recalls, or food diaries which are dependent on subject recall and cognitive functioning; all self-reported methodologies have considerable scope for measurement error and inherent bias, with under-reporting biased towards unhealthy foods and dietary energy intake and over-reporting towards healthier foods (8).
Also, there is potential error in estimating portion sizes and the limitation of the number of foods and dishes that have been directly analysed. Reporting is known to deteriorate further in the obese and likely to include significant inaccuracies in older populations. Prevalence of misreporting has been estimated as between 30-88% in epidemiological surveys (9). In large scale studies, use of FFQ’s are commonly applied, but since these are tailored to suit the population they serve, there is wide variation in number and range of items appearing on food lists.
Another limiting factor with traditional dietary assessment models is the limited number of nutrients that have been measured accurately in food. Evidence suggests that bioactive molecules in food such as polyphenols may play a role, but these molecules are not measured in most nutritional data sets. This limits the scope of understanding between the chemical composition of food and cognitive associations.
Reliability of dietary assessment is further compromised by analytical methods applying either a priori methodologies using constructed scores based on an underlying hypothesis and dietary guidelines which do not reflect entire dietary intake or a posteriori methods, applying data reduction techniques, such as principal component analysis, factor analysis, or cluster analysis to categorise on intercorrelations which provides insight to shared characteristics within a population but has limited comparability and reproducibility in other population samples. Both approaches are limited by pre-defined selections of food and nutrient groupings resulting in varying interpretations of dietary exposures and disease risk.
In recent years, the advent of online dietary assessment tools has made data collection and analysis much easier, albeit not mitigating the biases, as outlined above; also including technological competence in older adults, recently noted as a possible confounder for reporting accuracy; a feasibility study on use of online dietary recalls among older adults indicated that participants who completed multiple recalls reported higher self-confidence with technology and a higher technology readiness score than those who did not complete any recalls (10). There is, still, a need for evidence-based tools to be evaluated for validation and reliability in study outcomes and clarity in mechanisms that might be protective against dementia. We therefore propose a focus on applying metabolomics as a validation tool and framework for investigating the immediate or cumulative effects of diet on cognitive status and decline.


Metabolomics as a validation tool

The implementation of high-throughput -omics technologies in dietary assessments holds promise for evidence-based data, by providing objective measures of dietary intake in targeted or untargeted analyses, thus mitigating the risks of bias and subjectivity of self-administered data collecting methods.
Such technologies include Nuclear Magnetic Resonance (NMR) spectroscopy and Mass Spectroscopy (MS) based techniques using Liquid Chromatography (LC-MS) and Gas Chromatography (GC-MS) for improved separation.
A novel approach, developed to assess dietary intake against metabolic profiles, has been the adoption of NMR spectroscopy in urinary analyses to detect concentrations of small molecule metabolites which reflect “actual” rather than “estimate” measures of food intake. In this way, distinguishable metabolites can be used to validate dietary intake of FFQs and dietary patterns (8).
For example, a higher intake of fish consumption or protein in a FFQ can be validated by protein related metabolites in the metabolomic profile or other metabolites such as proline betaine, a marker of citrus fruit intake (11). Established dietary biomarkers, such as urinary sodium or nitrogen balance can track intake of specific nutrients, however metabolic profiling goes further in providing insight to metabolomic response of overall dietary intake following digestion, absorption and metabolism, providing insight to functional relationships between diet and health outcomes (12).
In addition to the feasibility of using metabolic profiles to validate dietary patterns, merging -omics with dietary assessments provides an opportunity to monitor and objectively assess dietary intake against healthy eating targets using urine composition; thus, enabling the quantification and monitoring of the potential effect of the adherence to or changes in dietary pattern in response to risk reduction strategies.
Several nutrition studies have explored how metabolomics can establish accurate associations between diet and disease risk to predict health outcomes. In a randomised, controlled trial, four dietary patterns, administered in controlled feeding conditions, revealed diet-discriminatory metabolomic profiles associated with different degrees of non-communicable disease risk, based on compliance to the WHO recommended healthy diets (8). The model was validated using internal and external cohort data with 24-hour recalls and found to associate with predicted scores of dietary profiles derived from the urinary metabolic profiles.
Applying the same methodology, a recent study demonstrated agreement between urinary metabolic biomarkers and self-reported data built on a model of 46 urinary metabolites, paired with 24-hour dietary recalls from 1,848 US individuals to accurately differentiate between healthy and unhealthy dietary patterns (13). Additionally, in a RCT of subjects with hypertension, the presence of specific metabolic markers suggested a mediating response to pre-disease markers in the gut microbiome (14).
Furthermore, an integrated metabolomic approach across platforms (NMR, LC-MS and GC-MS) may allow to accurately assess thousands of molecules associated with food intake that are not captured in current food tables.
The integration of nutritional metabolomics into dementia research invites new possibilities; improved reliability in dietary assessments might prove valuable but may only be the tip of the iceberg. If we consider that the ‘functional nutriome’, used to describe “chemically-defined diet-derived molecular species” (13) and expression of metabolic phenotype, may affect diet and disease risk, this has relevance in the context of AD risk and precision nutrition. We know that multiple processes leading to cognitive decline and dementia begin many years prior to the onset of cognitive decline; in determining whether there are specific metabolomic markers present in the earliest stage of cognitive decline which are observed in the food-derived metabolome, would be an important step in identifying predictive signatures of disease risk for earlier diagnosis in AD and other LODs. Furthermore, metabolite profiles can differentiate dietary patterns, so theoretically diet can be interrogated without dependency on self-reported dietary assessments.

Figure 1. From urine collection to dietary assessment pipeline

1.Urine collection with timepoint 2. Sample analysis 3.Generation of one metabolic profile per urine sample 4. Mathematical modelling generates score classifying dietary profile in relation to the adherence to dietary recommendations


Biofluid used for metabolomic dietary assessment

The vast majority of current metabolomic-based dietary assessment use either plasma or urine, however evidence on comparability of the two biofluids in relation to diet is limited. For urine, protocols are beginning to emerge which give guidance on the use of spot samples. Evidence suggests the first void urine gives the most relevant information (15). Moreover, the application of metabolomics to measurement molecular profile of food allows molecules to be traced from consumption to metabolism endpoints.
However, a major limitation of current metabolomic dietary assessments is that it is not possible to assess carbohydrate intake fully, due to the lack of relevant biomarkers. Given that carbohydrates are a major source of energy in most diets this is a limitation and an area of active research. Another limitation relates to the time-frame of the biofluid sample report; since it has been suggested that the first void urine relates to the previous day’s dietary intake (16).

Table 1. NMR Metabolomics validation studies of self-reported dietary pattern intake


Dementia-related metabolomic studies

The National Institute on Aging–Alzheimer’s Association (NIA-AA) research guidelines for AD and cognitive decline due to AD pathology, emphasize the need for the implementation of biochemical markers and validation measures to unify findings across the variation of diverse methodologies (17), under the revised NIA-AA Research Framework biomarker-based AD criteria for research purposes of amyloid and tau abnormal accumulation and neurodegeneration AT(N) (18).
The advancement of AD biomarkers has gained momentum in recent years in determining the AT(N) status in vivo, through the use of positron emission tomography (PET), cerebrospinal fluid (CSF), plasma-based assays and magnetic resonance imaging (MRI) studies, as core indicators of disease pathology (19). Moreover, metabolites in different biofluids (serum, plasma, CSF, urine) are proving to be promising indicators of alterations in lipids, amino acids, hormones and other circulating metabolites and their potential associations with cognitive performance change (20), (21) and as precision medicine tools contributing to the classification of patients into cognitive status subgroups based on metabolite signatures (22).
However, the literature is lacking examples of diet-dementia related studies and reference models applying nutritional metabolomics in cognitive and dietary pattern associations. Results from one case-control study linked a baseline serum signature of 22 metabolites with subsequent cognitive decline over 12 years and suggested specific foods (coffee, cocoa and fish) may be protective (23). Despite the novelty of the findings, the study was not validated in other cohorts, so reproducibility is still unclear, delineating the need for additional high-quality studies that apply complementary metabolomic platforms and approaches to identity predictive signatures of AD risk and preventive targets in cognitive decline and dementia.
To conclude, the emergence of improved measurements of ageing and dietary biomarkers represents a potentially exciting new development in dementia research. The enrichment of dietary assessments through the application of novel -omics technologies provides an opportunity for greater confidence and reliability in self-reported measures, whilst allowing to better understand the metabolomic responses in relation to dietary and cognitive associations. As Alzheimer’s and other late onset dementia forms are multi-faceted diseases, evidence-based and multi-pronged approaches are required in disentangling the respective roles of modifiable factors in disease development. It is our view that future research aimed at exploring the role of diet in brain health and dementia prevention should leverage emerging innovative high-throughput technologies, such as metabolomics, that may more sensitively and accurately inform on the functional nutriome.


Conflict of Interest: L T Middleton has received research funding (to Institution) from Janssen, Merck (USA), Gates, Takeda/ Millenium, Novartis, Invincro and EIT Health, outside the submitted work. S Ahmadi Abhari and L T Middleton were supported by the European Institute of Innovation and Technology (EIT)-Health for the brain ageing PhD school. J M Posma received a Medical Research Council (MRC) funded HDR UK Rutherford Fund Fellowship (MR/S004033/1). Isabel Garcia-Perez received a NIHR Career Development Research Fellowships (NIHR-CDF-2017-10-032). J M Posma, I Garcia-Perez and G Frost hold a patent in preparation (GB2111739.5) on NMR spectrum data fitting. G Frost – has received research funding (to Imperial College) from Nestle, Unilever, Quorn and Heptaris. G Frost and I Garcia-Perez are directors of Melico Ltd a dietary metabolomic spinout company.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.



1. Kivipelto M, Mangialasche F, Ngandu T. Lifestyle interventions to prevent cognitive impairment, dementia and Alzheimer disease. Nature reviews Neurology. 2018;14(11):653.
2. Ngandu T, Lehtisalo J, Solomon A, Levälahti E, Ahtiluoto S, Antikainen R, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. The Lancet. 2015;385(9984):2255-63.
3. Fb H. Dietary pattern analysis: a new direction in nutritional epidemiology. Current opinion in lipidology. 2002;13(1):3-9.
4. Samadi M, Moradi S, Moradinazar M, Mostafai R, Pasdar Y. Dietary pattern in relation to the risk of Alzheimer’s disease: a systematic review. Neurological Sciences. 2019;40(10):2031-43.
5. Chen X, Maguire B, Brodaty H, O’Leary F. Dietary Patterns and Cognitive Health in Older Adults: A Systematic Review. J Alzheimers Dis. 2019;67(2):583-619.
6. Davis JJ, Fournakis N, Ellison J. Ketogenic Diet for the Treatment and Prevention of Dementia: A Review. Journal of Geriatric Psychiatry and Neurology. 2021;34(1):3-10.
7. Behnaz S, Debora R, Susanna CL, Gerd F-I, Laura F, Wei-Li X. The Nordic Prudent Diet Reduces Risk of Cognitive Decline in the Swedish Older Adults: A Population-Based Cohort Study. Nutrients. 2018;10(2):229.
8. Garcia-Perez I, Posma JM, Gibson R, Chambers ES, Hansen TH, Vestergaard H, et al. Objective assessment of dietary patterns by use of metabolic phenotyping: a randomised, controlled, crossover trial. The Lancet Diabetes & Endocrinology. 2017;5(3):184-95.
9. Rennie KL, Coward A, Jebb SA. Estimating under-reporting of energy intake in dietary surveys using an individualised method. Br J Nutr. 2007;97(6):1169-76.
10. Ward HA, McLellan H, Udeh-Momoh C, Giannakopoulou P, Robb C, Wark PA, et al. Use of Online Dietary Recalls among Older UK Adults: A Feasibility Study of an Online Dietary Assessment Tool. Nutrients. 2019;11(7):1451.
11. Heinzmann SS, Brown IJ, Chan Q, Bictash M, Dumas M-E, Kochhar S, et al. Metabolic profiling strategy for discovery of nutritional biomarkers: Proline betaine as a marker of citrus consumption. The American journal of clinical nutrition. 2010;92(2):436-43.
12. Guasch-Ferré M, Bhupathiraju SN, Hu FB. Use of Metabolomics in Improving Assessment of Dietary Intake. Clinical chemistry. 2018;64(1):82.
13. Posma JM, Garcia-Perez I, Frost G, Aljuraiban GS, Chan Q, Van Horn L, et al. Nutriome-metabolome relationships provide insights into dietary intake and metabolism. Nat Food. 2020;1(7):426-36.
14. Loo RL, Zou X, Appel LJ, Nicholson JK, Holmes E. Characterization of metabolic responses to healthy diets and association with blood pressure: application to the Optimal Macronutrient Intake Trial for Heart Health (OmniHeart), a randomized controlled study. The American Journal of Clinical Nutrition. 2018;107(3):323-34.
15. Wilson T, Garcia-Perez I, Posma JM, Lloyd AJ, Chambers ES, Tailliart K, et al. Spot and Cumulative Urine Samples Are Suitable Replacements for 24-Hour Urine Collections for Objective Measures of Dietary Exposure in Adults Using Metabolite Biomarkers. J Nutr. 2019;149(10):1692-700.
16. Beckmann M, Wilson T, Lloyd AJ, Torres D, Goios A, Willis ND, et al. Challenges Associated With the Design and Deployment of Food Intake Urine Biomarker Technology for Assessment of Habitual Diet in Free-Living Individuals and Populations-A Perspective. Frontiers in nutrition (Lausanne). 2020;7:602515.
17. Jack CR, Therneau TM, Weigand SD, Wiste HJ, Knopman DS, Vemuri P, et al. Prevalence of Biologically vs Clinically Defined Alzheimer Spectrum Entities Using the National Institute on Aging-Alzheimer’s Association Research Framework. JAMA neurology. 2019;76(10).
18. Jack CR, Jr., Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 2018;14(4):535-62.
19. Ahmed RM, Paterson RW, Warren JD, Zetterberg H, O’Brien JT, Fox NC, et al. Biomarkers in dementia: clinical utility and new directions. Journal of neurology, neurosurgery and psychiatry. 2014;85(12):1426-34.
20. van der Lee SJ, Teunissen CE, Pool R, Shipley MJ, Teumer A, Chouraki V, et al. Circulating metabolites and general cognitive ability and dementia: Evidence from 11 cohort studies. Alzheimers Dement. 2018;14(6):707-22.
21. Jiang Y, Zhu Z, Shi J, An Y, Zhang K, Wang Y, et al. Metabolomics in the Development and Progression of Dementia: A Systematic Review. Frontiers in neuroscience. 2019;13:343-.
22. Kurbatova N, Garg M, Whiley L, Chekmeneva E, Jiménez B, Gómez-Romero M, et al. Urinary metabolic phenotyping for Alzheimer’s disease. Sci Rep. 2020;10(1):21745.
23. Low DY, Lefèvre-Arbogast S, González-Domínguez R, Urpi-Sarda M, Micheau P, Petera M, et al. Diet-Related Metabolites Associated with Cognitive Decline Revealed by Untargeted Metabolomics in a Prospective Cohort. Molecular Nutrition & Food Research. 2019;63(18):n/a-n/a.



V. Alanko1,2, C. Udeh-Momoh1,3, M. Kivipelto1,3,4,5, A. Sandebring-Matton1,2,3,#


1. Division of Clinical Geriatrics, Center for Alzheimer Research, NVS, Karolinska Institutet, Stockholm, Sweden; 2. Division of Neurogeriatrics, Center for Alzheimer Research, NVS, Karolinska Institutet, Stockholm, Sweden; 3. Ageing Epidemiology (AGE) Research Unit, School of Public Health, Imperial College London, London, United Kingdom; 4. Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finland; 5. Theme Inflammation and Aging, Karolinska University Hospital, Solna, Sweden

Corresponding Author: Anna Sandebring-Matton, Division of Clinical Geriatrics, Center for Alzheimer Research, NVS, Karolinska Institutet, Stockholm, Sweden,, telephone +46-8-524 800 00, fax +46-8-31 11 01

J Prev Alz Dis 2022;
Published online January 11, 2022,



Since developing an effective treatment for Alzheimer’s disease (AD) has been encountered as a challenging task, attempts to prevent cognitive decline by lifestyle modifications have become increasingly appealing. Physical exercise, healthy diet, and cognitive training are all modifiable, non-pharmacological lifestyle factors considered to influence cognitive health. Implementing lifestyle modifications on animal models of AD and cognitive impairment may reveal underlying mechanisms of action by which healthy lifestyle contribute to brain health. In mice, different types of lifestyle interventions have been shown to improve cognitive abilities, alleviate AD-related pathology and neuroinflammation, restore mitochondrial function, and have a positive impact on neurogenesis and cell survival. Different proteins and pathways have been identified to mediate some of the responses, amongst them BDNF, Akt–GSK3β, JNK, and ROCK pathway. Although some important pathways have been identified as mediating improvements in brain health, more research is needed to confirm these mechanisms of action and to improve the understanding of their interplay. Moreover, multidomain lifestyle interventions targeting multiple risk factors simultaneously may be a promising avenue in future dementia prevention strategies. Therefore, future work is needed to better understand the synergistic impact of combinatory lifestyle strategies on cellular mechanisms and brain health.

Key words: Alzheimer’s disease, dementia, prevention, mouse models, multidomain, lifestyle.



The high global prevalence of the leading neurodegenerative disease-causing dementia, Alzheimer’s disease (AD) has been linked to exponential increases in ageing populations world-wide (1), though trends of attenuated incidence in high-income countries have been observed (2). Due to the slow progress in AD-dementia drug development, dementia prevention has become increasingly important and has drawn significant attention during the past few decades. Recent reports have been disseminated in the last few years underpinning the possibilities and importance of considering lifestyle factors for dementia prevention (3). Livingston et al.’s seminal report (2020) suggested that as much as 40% of dementia risk is attributable to lifestyle factors (4). Around the same time, the World Health Organization published guidelines for lifestyle modification to reduce the risk of cognitive decline and dementia (5).
Dementia prevention has been investigated both by observational studies and randomized controlled trials (RCT), both with their respective benefits and limitations (6). Still, there is an incomplete understanding of the underlying mechanisms of action. Although brain changes and functions can be assessed during preventive interventions using neuroimaging techniques and fluid biomarkers, such methods will not reveal the complete mechanisms underlying the alterations. Identifying these mechanisms could result in new treatment strategies. To better understand probable underlying mechanisms, it is necessary to conduct prevention studies in pre-clinical AD models.
While several modifiable risk factors have been identified, in this review we will discuss the effects of lifestyle interventions rather than pharmacological strategies that may directly impact risk factors such as hypertension and diabetes. We consider physical exercise, diet, and cognitive training as the main preventive factors, as these may either directly or indirectly impact some of the other risk factors, including hypertension, obesity, and diabetes (7–9). We will further describe current evidence from non-pharmacological pre-clinical prevention studies in mouse models of mainly AD and ageing, and how they may be translated to clinical applications.


Physical exercise

AD-pathology, inflammation, and glial alterations

Extensive studies performed on dementia mouse models exposed to exercise training report positive outcomes in cognitive tests (10–18). Many of these studies report an increase of synaptic markers indicating preservation of synapses and their function in the brains of exercising mice (10–14, 19). Additionally, desired alterations in AD-related pathology after a period of exercise in AD transgenic mice have been observed in several studies (10–12, 15, 20). Physical training can induce the phosphorylated Akt/phosphorylated GSK3β system, which in turn may reduce tau pathology (11, 14). Exercise attenuates pro-inflammatory markers, microglia and astrocyte activation, and increases anti-inflammatory markers in mice; such changes have been observed in 3xTg-AD mice as a result of different exercise paradigms (11, 12, 21). In APP/PS1 mice, exercise reduced the count and intensity of plaque-associated astrocytes (20) while in physically active 5xFAD mice they were increased (13). The reduced neuroinflammation is suggested to be mediated via inhibition of the c-Jun N-terminal kinase (JNK) pathway (11) or activation of the microRNA miR-129–5p (16).
White matter is also affected early in AD progression (22). Total white matter volume decreases in mouse models of AD (17) but can be restored to the levels of wild type (WT) mice (17) or significantly increased compared to sedentary counterparts (23) with exercise. Decreased myelin in old WT mice can be improved by exercise (24), yet with conflicting results (25). Indeed, exercise seems to promote differentiation of oligodendrocyte progenitor cells and increase the number of mature oligodendrocytes in mice (26). Axon growth can be inhibited by myelin-associated factors via RhoA/Rho kinase (ROCK) pathway (24, 27), but also differentiation of oligodendrocytes is inhibited by the same pathway (28). Physical activity downregulates the ROCK pathway, which is suggested to mediate the increased myelination. Exercise further improves vasculature in the brain, particularly in the white matter, which otherwise is compromised in APP/PS1 mice (17).

Neurogenesis & anti-apoptotic pathways

One common finding in several pre-clinical studies is the increased levels of the brain-derived neurotrophic factor (BDNF) as a result of physical activity (10, 13, 14, 19, 29–31). BDNF expression is regulated to a great extent by other muscle-derived myokines, such as FNDC5 (fibronectin type III domain containing 5) and its cleavage product Irisin (32). FNDC5 expression is regulated by peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) (32). Expression of the mitochondrial biogenesis regulator PGC-1α may be induced by exercise (11, 12), although conflicting results have been reported (21). Furthermore, BDNF has a precursor form (proBDNF) and a mature form (hereafter referred to as BDNF), where proBDNF binds the receptor p75 and its co-receptor, Sortilin, activating for instance apoptotic signalling while BDNF mediates its effects via tropomyosin-related kinase B (TrkB) receptor signalling (33). Extracellular cleavage of proBDNF is executed by several different proteases into the mature form (34). There are at least two upstream regulators of BDNF processing, tissue-type plasminogen activator (tPA) and urokinase-type plasminogen activator (uPA), which process plasminogen to plasmin that in turn cleaves proBDNF to BDNF (35,36). Interestingly, tPA is activated by physical exercise (36). In the hippocampus, BDNF is a facilitator of long-term potentiation (LTP) and thus enhances memory and learning, neurogenesis, and synaptic plasticity. FNDC5/Irisin is involved in maintaining synaptic plasticity by stimulating BDNF expression, and levels of these proteins are reduced in the AD brain and CSF (18). When these are downregulated, LTP is impaired. The levels can, however, be restored by exercising, and thus memory can be improved or maintained (18). Additionally, exercise activates other neurotrophic factors involved in neurogenesis, as Kim and colleagues (12) did observe an increase in TrkB expression while BDNF levels were unchanged.
In 3xTg-AD mice, physical activity restored the levels of BrdU/NeuN-positive cells to the WT levels (12) or at least significantly increased compared to sedentary controls (14). In 5xFAD mice such effects have not been reported (13). An increase in newly formed neuronal precursors have also been shown in active WT mice (25, 37). Horowitz et al. (30) showed that not only was neurogenesis increased in exercising WT mice, but even in inactive aged mice when administered with plasma from the exercised mice. Similar results were reported when 3xTg-AD mice were administered plasma from young, exercised mice (19). This denotes that there is a clear axis between the training-induced systemic changes and the brain. Horowitz et al.’s study (30) specifically investigated the role of Gpld1 (glycosylphosphatidylinositol-specific phospholipase D1) and connected it to the increased BDNF levels and neurogenesis. The liver produces increased levels of Gpld1 in exercised mice, but also in physically active persons (30). However, the compound mainly remains in the periphery (30). Reflecting on the report, the effect on BDNF could be attributed to a Gpld1–uPA–plasminogen cascade. Finally, these effects were observed as an improved cognitive performance of the mice (19, 30).
Complementary to neurogenesis, exercise might also alleviate cell death in the brain (14, 15, 19, 24). This effect has been attributed to a decrease in pro-apoptotic markers like caspases, cytochrome c, and Bax, and additionally to an increase in anti-apoptotic markers like Bcl-2 (14, 15, 24, 31). Downregulation of the ROCK system is a plausible pathway reducing apoptosis (24), and as a downstream target of receptor p75 (38) the effect could be attributed either to a reduction in proBDNF or increase in BDNF. In exercised AD mice (14, 15) or in AD mice administered plasma from exercised mice (19) the levels of these different proteins are not necessarily restored to a WT level, but still significantly improved compared to sedentary mice. Intriguingly, resistance training did not affect Bax and Bcl-2 levels (11).

Metabolic factors & mitochondrial function

Glucose hypometabolism is a known feature of AD (39) and many mouse models with memory impairment demonstrate a deficiency in glucose uptake (10,40). Exercise can induce an increase in brain expression of glucose transporters GLUT1 and GLUT3 in APP/PS1 double-transgenic mice (10) as well as GLUT1 in NSE/APPswe mice (31). In addition to improved brain glucose homeostasis, physical activity improves peripheral glucose metabolism in 3xTg-AD mice (21). In accord-ance with improved glucose metabolism, exercise alleviated insulin resistance in high-fat-diet fed WT mice by activating insulin receptors substrate (IRS) and its downstream pathways PI3K–PDK-1–Akt–GSK3β (37).
An additional effect that physical activity appears to have on the brain is maintenance of mitochondrial function. Kim and colleagues (12) detected restoration of mitochondrial length and enhanced mitochondrial biogenesis in exercised AD mice. Likewise, in APP/PS1 mice, exercise recovered mitochondrial integrity and partly the ATP levels (10). When 3xTg-AD mice were administered with plasma from young, exercised mice, it resulted in a better capacity to maintain calcium homeostasis and in reduction of reactive oxygen species when compared to the other AD mice – yet the mitochondrial function was not returned to the WT levels (19). A similar effect was evident in 3xTg-AD mice undergoing treadmill exercise combined with light flickering (14). One possible mechanism of these effects could be the restoration of brain iron dyshomeostasis that contributes to oxidative stress (15).



Olive oils and polyphenols

There have been several attempts mimicking the Mediterranean diet (MeDi) in animal studies. As composing a full MeDi for mice is largely impossible, studies have investigated the effects of single nutrients or groups of nutrients that are thought to be the main health-promoting components of MeDi. Many of these nutrients are found in olive products and fish oils. Extra virgin olive oil-enriched diet improved working and spatial memory and increased synaptophysin expression in 3xTg-AD mice (41). Furthermore, both Aβ and tau pathology were diminished in the treatment group, active microglia alleviated, and autophagy seemed to be induced (41). In the APP mutant TgSwDI mice, Aβ pathology was ameliorated and some phosphorylated tau species were reduced by extra virgin olive oil consumption (42). The reduction of Aβ plaques and vascular depositions of Aβ was coupled to an increased clearance of the protein but even to a favourable APP processing pattern (42). Improved clearance was to some extent a result of increased expression of transporter proteins ApoE and ABCA1 (Apolipoprotein E and ATP-binding cassette transporter, respectively) (42).
Moreover, many studies have investigated the role of individual bioactive components of olive oil. Particular interest has been in studying the relationship between polyphenols found in olive oil and mitochondrial dysfunction and oxidative stress since polyphenols possess antioxidant activities (43). In APP/PS1 mice, hydroxytyrosol diet did not cause any major cognitive improvement nor did it affect Aβ pathology (44). Nonetheless, hydroxytyrosol reversed oxidative stress in the brains and restored mitochondrial protein levels, reduced the levels of the cleaved caspase 3, and alleviated neuroinflammation through inhibition of JNK pathway (44). Impacts on mitochondria have further been studied in aged NMRI mice fed with oleocanthal or ligstroside (45). Neither of the supplements affected the cognitive abilities of the mice, though those were not extensively studied. However, the life expectancy and cerebral ATP levels improved due to ligstroside diet, and a rationale that olive polyphenols improve mitochondrial respiration is supported by in vitro studies (45). In NMRI mice, a cocktail of purified secoiridoid polyphenols, including for example hydroxytyrosol, elevated ATP levels and moreover restored cognitive abilities (46). With oleuropein aglycone diet the AD mouse model TgCRND8 improved cognitive performance (47) as well as reduced Aβ load in different brain areas (47,48). The diet further induced autophagy and histone acetylation (47, 48).

Fish oils and Fortasyn Connect

Fortasyn Connect (FC) is a multi-nutrient supplement consisting of docosahexaenoic acid (DHA), eicosapentaenoic acid (EPA), uridine monophosphate, choline, vitamins B12, B6, C, E, and folic acid, phospholipids, and selenium (49). FC diet has been shown to partly reverse cognitive deficits in APP/PS1 mice (50, 51). Yet, in another study with the same mouse model and diet, no such findings were reported (52), even though all three research settings have applied Morris Water Maze (MWM) tasks as the testing paradigm. ApoE4 and ApoE knock-out (KO) mice – models featuring the main genetic risk factor of sporadic AD – did not either improve learning nor memory in this particular cognitive test after FC diet (53). Although APP/PS1 mice improved learning after ingesting FC, the dietary impacts on Aβ burden, inflammation, and reactive oxygen species were very minor overall, as were the effects with fish oils or a combination of fish oils and plant sterols (50). However, when the dietary intervention was initiated at an earlier time-point (compare 3 months of age (54) versus 5 months of age (50)), even if the intervention was of shorter duration, the Aβ pathology was profoundly alleviated in the APP/PS1 mice (54). These results highlight the importance of early nutritional intervention for an impact on Aβ neuropathology. Moreover, while the FC diet seems to have a positive effect on the Aβ burden at an early stage of disease progression, this was not the case for a diet only enriched with DHA and uridine monophosphate, nor with either of them used alone (54).
Intriguingly, Jansen et al. (52) observed a significant increase in doublecortin-positive cells with the FC diet in APP/PS1 mice, indicating that the FC enhances neurogenesis. The level of these cells was restored to the WT level. On the contrary, a diet enriched only with DHA, EPA, and uridine monophosphate did not affect the number of doublecortin-positive cells (52), and the FC diet did not significantly increase the number of these cells in ApoE4 or ApoE KO mice (53). In APP/PS1 mice, the FC diet had also a positive impact on the degenerative burden of the cortex (54). Both of the diets did still result in a change in brain fatty acid composition, for example, by increasing the omega-3 to omega-6 ratio significantly and altering levels of brain oxysterols (52–54).


Environmental enrichment and cognitive training

Cognitive training

Cognitive training per se and its effects on cognitive abilities has not been extensively studied in mice – particularly in settings where the retention time is long, i.e. the time between training and testing. Nevertheless, the lack of research with long retention time was addressed in a recent study investigated APP/PS1 mice in different cognitive training set-ups (55). In the first setting (referred to as “trained”) mice were trained with MWM at the age of three months, and in the second setting (“overtrained”) at the age of two months and again at three months. Importantly, training was scheduled before or at the time of plaque formation. However, the study did not include control groups that would not receive training. In both set-ups, mice were tested at age of seven months with a similar setting as during the training periods, but with an additional six-day-long reversal task. At seven months, “trained” transgenic mice showed rescued memory, although there was a significant group difference compared to WT littermates that underwent the same training program. Such difference was not observed between the groups of “overtrained” mice. In further analyses, “overtrained” APP/PS1 mice performed overall at the same level as WT mice and better compared to “trained” APP/PS1 mice. No molecular nor cellular changes were assessed in the study.
In an earlier study, Tg2576 mice were trained with MWM (56). The researchers included non-training control groups, both transgenic and WT littermates. The day after the water maze tasks, mice were subjected to Contextual Fear Conditioning where trained mice performed significantly better than non-trained transgenic mice. Furthermore, training enhanced LTP and restored dendritic complexity in the hippocampus to a WT level. Training also increased levels of some postsynaptic proteins – an effect that was coupled to activation of calcium/calmodulin-dependent protein kinase II (CaMKII) – and ameliorated AD-related pathology.
A large-scale study by Billings et al. (57), investigated the effects of water maze training throughout the lifetime of 3xTg-AD mice, with training sessions once every three months beginning at the age of two months. Until approximately the age of 12 months, transgenic mice had a clear benefit of training and were performing at a WT level, yet at 15 months of age, this effect started to diminish. In addition to the learning and memory improvements, the harmful Aβ load was attenuated in 12-month-old trained mice, and tau phosphorylation was reduced possibly through reduced activity of GSK3β. Still, the favourable effects of cognitive training were most prominent when the training began before the emergence of neuropathology, although a later start of the training paradigm also improved the mice’s abilities. On the contrary, in another study, even if 16-month-old PDAPP mice demonstrated faster forgetting than WT counterparts already seven days post-training, their memory could be retrieved with a short retraining period at seven weeks after the initial training period (58).
MWM is, however, stressful for mice. Although the improved memory in both Tg2576 (56) and 3xTg-AD (57) mice can be attributed to the cognitive training, as the control groups exposed to swimming only did not perform on the same level during cognitive testing, stress is an important component affecting physiology and learning. Certain amount of stress enhances learning, yet highly stressed individuals may encounter an opposing effect (59). Future research should aim to implement less stressful cognitive training paradigms to measure a more accurate effect on cognitive performance.

Environmental enrichment

An alternative to cognitive training is utilising environmental enrichment as a cognitive stimulator in pre-clinical research (60). Environmental enrichment can compose of different factors and components, but generally, it includes larger cages compared to standard housing together with various objects, such as tunnels, ladders, nesting material, and items of different sizes and colours (60). These all contribute not only to enhanced cognitive stimulus but even to sensory and motor stimulations (60). Moreover, cages may also have running wheels and house a greater number of animals simultaneously. For this review, we want to separate the impacts of environmental enrichment and exercise, and therefore only studies implementing environmental enrichment paradigms excluding running wheels are discussed.
Independent of the time point of intervention start, enriched environment restored recognition memory in Tg2576 mice (61). However, spatial memory, as measured by MWM, was rescued only in those mice that lived in the enriched environment before the emergence of neuropathology. Enriched housing additionally counteracted the build-up of Aβ pathology in these mice. APP23 mice housed in an enriched environment have demonstrated improvements both in MWM and Novel Object Recognition (62). Moreover, environmental enrichment induced both hippocampal and cortical BDNF expression and reduction in Aβ plaque formation (62). On the contrary, in PDAPP-J20 mice that underwent enrichment intervention, a significant reduction in Aβ1–40 and Aβ1–42 peptides were observed, while the plaque load was unchanged (63). The volume of plaque-associated GFAP-positive astrocytes was also unaltered. Still, the volume of non-plaque-associated GFAP-positive cells was restored to a WT level due to environmental enrichment in these mice that otherwise have a significantly reduced number of astrocytes. Furthermore, the non-plaque-associated astrocytes in mice from enriched housing did not have as complex morphology as in standard housed mice.
Additionally, effects between exercise and environmental enrichment have been compared (64). In APP23 transgenic mice that were housed either in enriched cages or in a standard cages equipped with running wheels no effects on Aβ load were observed (64). Still, the mice living in an enriched environment demonstrated improved learning. Based on analysis of BrdU- and doublecortin-positive cells, hippocampal neurogenesis was not induced in either intervention group. However, a marker for a postmitotic phase, Calretinin, was increased in mice living in enriched cages. Levels of neurotrophin and BDNF expression were also increased in these mice compared to controls while such effects were not evident in the exercised mice.


Discussion & future directions

There is a high demand for strategies to cure and prevent dementia. Several of the herein discussed pre-clinical studies have reported clear benefits induced by singular lifestyle domain in mice (summarised in Fig. 1). When studying the effects in humans, some clinical trials investigating the effects from single-domain lifestyle changes have reported positive results, but likewise non-efficacious studies have been published (reviewed in 65).

Figure 1. Summary figure of intervention effects

The figure gives an overview of the the effects observed in lifestyle intervention paradigms in AD and WT mouse models. The light green arrows indicate an increase/amelioration in adjacent factors whereas the dark green arrows indicate a decrease/alleviation in adjacent factors. Abbreviations: Aβ, Amyloid beta; GLUT, glucose transporter; p-Tau, phosphorylated tau; WT, wild type.


Contradictions in efficacy between clinical and pre-clinical lifestyle intervention studies may be explainable by various considerations. First and foremost, mice are not men, and although there is great homology between brains, the significant inter-species differences in, for instance, cell types (66) may direct how the brain is influenced by various interventions. The issue is similar regarding biological homogeneity between mice with the same genetic background. In mice, it is near impossible to replicate the huge heterogeneity of different genetic and environmental risk factors present in humans. Considering AD, the transgenic mouse models developed for AD research replicate mainly familial AD instead of the more common sporadic AD. Thus, it is dubious to translate findings from these mice to sporadic AD. In attempts to translate findings, the great majority of drugs developed for AD have failed to reach their primary clinical outcomes (67). Additionally, mice lack the variation in living conditions experienced by humans. This results in mice adhering to the intervention naturally more stringently than humans. A factor that however remains to be explored is how the synergy of different lifestyle interventions affects the brain on a cellular and molecular level. Such questions could be investigated in mice exposed to multimodal lifestyle intervention strategies even if the discrepancies between mice and men are acknowledged.
Exercise seems to be the most studied lifestyle factor regarding molecular and cellular mechanisms in mice. Increased BDNF levels are likely to be one of the primary mediating players of different positive outcomes when exercising. BDNF impacts several downstream pathways, for example, reducing levels of Bax and boosting expression of Bcl-2 (as in (14)), thus promoting cell survival, inducing expression of glucose transporters (as in (10)), activating Akt for phosphorylation of GSK3β (68), hence suppressing tau phosphorylation, and mediating neurogenesis (30,37). Physical training promotes the expression of PGC-1α (11,12) and FNDC5/irisin (18) in the brain; factors that further contribute to the upregulation of BDNF expression. After translation, proBDNF can be further processed into mature BDNF. tPA levels increase due to exercise and hence increase the BDNF/proBDNF ratio (36). The exercise-induced BDNF levels could likely be attributed also to elevated liver-derived Gpld1 levels that in turn promote plasminogen processing through uPA (30). BDNF levels were mainly measured in exercise-related studies, but also a couple of studies report elevated brain BDNF levels as a result of environmental enrichment (62, 64).
GSK3β is considered to have a significant role in AD pathogenesis (69) and has been studied in several of the pre-clinical studies discussed in this review. Cognitive training (57) and physical exercise (11, 14, 37) have been found to increase the inactive, phosphorylated form of GSK3β and/or reduce the levels of the active form. These alterations have been accompanied by reductions in levels of its substrate phospho-tau (11, 14, 57). Nonetheless, even if infusion of plasma from young, exercised mice into AD mice had ameliorated several pathological features, it did not impact GSK3β activation and accordingly phospho-tau levels were also unchanged (19). Phosphorylation of GSK3β can be attributed to enhanced insulin signalling and activation of IRS (37).
Different lifestyle modifications furthermore have an impact on markers of neuroinflammation and oxidative stress. Microglia and astrocytes reduce in number, reactivity, and size or complexity (11, 12, 20, 41), but also some favourable alternations in cytokine levels have been observed (11, 21, 44). Some studies have shown that plaque-associated astrocytes decrease in number or complexity (20, 63), while others observe them as being more active around the plaques (13). The ameliorated neuroinflammation may be attributed to activation of the miR-129–5p (16) or inhibition of the JNK pathway and its down-stream targets (11, 44). Since JNK pathway is a down-stream target of receptor p75 (38), JNK pathway mediated reduction in neuroinflammation could be linked to changes in BDNF levels. A healthy lifestyle also maintains mitochondrial integrity and structure (10, 12), as well as mitochondrial function (10, 14, 45, 46). Regarding mitochondrial function, PGC-1α not only indirectly promotes BDNF expression (32), but serves as a marker of mitochondrial biogenesis (12).
To incorporate this multitude of beneficial mechanisms that promote brain health, multidomain interventions targeting several risk factors and mechanisms are likely to be the most beneficial approach, given the heterogeneity and multifactorial nature of AD (70, 71). Results from the pioneering multidomain lifestyle RCT – The Finnish Geriatric Intervention Study to Prevent Cog-nitive Impairment and Disability (FINGER) – supports the rationale for implementing multifactorial lifestyle changes, particularly in at-risk populations (72). Still, it should be mentioned that other multidomain lifestyle trials (e.g. preDIVA (Prevention of Dementia by Intensive Vascular Care) and MAPT (Multidomain Alzheimer Prevention Trial)) did not have measurable cognitive effects (73, 74).
Increased incidence of AD in global populations may be attributable to increased longevity, thus there is great need for promoting healthy ageing, especially in the context of brain health. The interventions reviewed here e.g. physical activity have also been shown to impact development of neuropathology, at least from studies in animal models. Further research is needed in earlier timepoints such as mid-life to evaluate the effect of lifestyle interventions on neurodegenerative brain changes in late-life. The diverse lifestyle-related interventions discussed in this paper have some clear benefits that are worth highlighting. First, they are all relatively easily available and do not necessarily require major financial input. Physical activity, a healthy diet, and cognitive challenges are further factors individuals may modify independently, in contrast to factors such as air pollution, traumatic head injury, and depression, which are all listed in Livingston et al.’s report as modifiable risk factors (4). Next, as encountered in the FINGER trial (72), the most common adverse event from such interventions is musculoskeletal pain. Thus, the disadvantages of healthy lifestyle changes are minor. Finally, an important aspect of such changes is the pervasive benefit of overall health. Not only was the CAIDE dementia risk score reduced (75), and cognitive functioning maintained or improved (72) in the FINGER trial, but additionally the risk of other chronic diseases and conditions related to lifestyle were diminished (76). Elucidating the mechanisms underpinning dementia prevention could yield precision medicine biomarkers and plausibly inform the discovery of novel therapeutic targets. Non-hypothesis-driven investigations to decipher the spectrum of mechanistic and cellular alterations in the brain as a result of lifestyle interventions could be a fruitful future avenue.


Funding: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.

Acknowledgments: This study was supported by the Swedish Research Council, Center for Innovative Medicine (CIMED) at Flemingsberg Campus, Stiftelsen Stockholms sjukhem, Sweden, Knut and Alice Wallenberg Foundation, Gun och Bertil Stohnes Stiftelse and Demensfonden.

Conflict of interest: The authors declare no disclosures.

Consent for publication: All authors have read the final version of the manuscript and have given their consent for publication.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.



1. 2021 Alzheimer’s disease facts and figures. Alzheimer’s Dement 2021;17:327–406.
2. Ahmadi-Abhari S, Guzman-Castillo M, Bandosz P, et al. Temporal trend in dementia incidence since 2002 and projections for prevalence in England and Wales to 2040: Modelling study. BMJ 2017;358:j2856.
3. Kivipelto M, Mangialasche F, Ngandu T. Lifestyle interventions to prevent cognitive impairment, dementia and Alzheimer disease. Nat Rev Neurol 2018;14:653–66.
4. Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 2020;396:413–46.
5. World Health Organization. Risk reduction of cognitive decline and dementia: WHO guidelines. Geneva: CC BY-NC-SA 3.0 IGO; 2019.
6. Yu JT, Xu W, Tan CC, et al. Evidence-based prevention of Alzheimer’s disease: Systematic review and meta-analysis of 243 observational prospective studies and 153 randomised controlled trials. J Neurol Neurosurg Psychiatry 2020;91:1201–9.
7. Ozemek C, Tiwari S, Sabbahi A, Carbone S, Lavie CJ. Impact of therapeutic lifestyle changes in resistant hypertension. Prog Cardiovasc Dis 2020;63:4–9.
8. Blüher M. Obesity: global epidemiology and pathogenesis. Nat Rev Endocrinol 2019;15:288–98.
9. Zheng Y, Ley SH, Hu FB. Global aetiology and epidemiology of type 2 diabetes mellitus and its complications. Nat Rev Endocrinol 2018;14:88–98.
10. Pang R, Wang X, Pei F, et al. Regular Exercise Enhances Cognitive Function and Intracephalic GLUT Expression in Alzheimer’s Disease Model Mice. J Alzheimer’s Dis 2019;72:83–96.
11. Liu Y, Chu JMT, Yan T, et al. Short-term resistance exercise inhibits neuroinflammation and attenuates neuropathological changes in 3xTg Alzheimer’s disease mice. J Neuroinflammation 2020;17:4.
12. Kim D, Cho J, Kang H. Protective effect of exercise training against the progression of Alzheimer’s disease in 3xTg-AD mice. Behav Brain Res 2019;374:112105.
13. Belaya I, Ivanova M, Sorvari A, et al. Astrocyte remodeling in the beneficial effects of long-term voluntary exercise in Alzheimer’s disease. J Neuroinflammation 2020;17:271.
14. Park S-S, Park H-S, Kim C-J, et al. Physical exercise during exposure to 40-Hz light flicker improves cognitive functions in the 3xTg mouse model of Alzheimer’s disease. Alzheimers Res Ther 2020;12:62.
15. Choi DH, Kwon KC, Hwang DJ, et al. Treadmill Exercise Alleviates Brain Iron Dyshomeostasis Accelerating Neuronal Amyloid-β Production, Neuronal Cell Death, and Cognitive Impairment in Transgenic Mice Model of Alzheimer’s Disease. Mol Neurobiol 2021;58:3208–23.
16. Li Z, Chen Q, Liu J, Du Y. Physical Exercise Ameliorates the Cognitive Function and Attenuates the Neuroinflammation of Alzheimer’s Disease via miR-129-5p. Dement Geriatr Cogn Disord 2020;49:163–9.
17. Zhang Y, Chao FL, Zhou CN, et al. Effects of exercise on capillaries in the white matter of transgenic AD mice. Oncotarget 2017;8:65860–75.
18. Lourenco M V., Frozza RL, de Freitas GB, et al. Exercise-linked FNDC5/irisin rescues synaptic plasticity and memory defects in Alzheimer’s models. Nat Med 2019;25:165–75.
19. Kim TW, Park SS, Park JY, Park HS. Infusion of plasma from exercised mice ameliorates cognitive dysfunction by increasing hippocampal neuroplasticity and mitochondrial functions in 3xtg-ad mice. Int J Mol Sci 2020;21:3291.
20. Zhang J, Guo Y, Wang Y, et al. Long-term treadmill exercise attenuates Aβ burdens and astrocyte activation in APP/PS1 mouse model of Alzheimer’s disease. Neurosci Lett 2018;666:70–7.
21. Do K, Laing BT, Landry T, et al. The effects of exercise on hypothalamic neurodegeneration of Alzheimer’s disease mouse model. PLoS One 2018;13:e0190205.
22. Bartzokis G, Cummings JL, Sultzer D, et al. White matter structural integrity in healthy aging adults and patients with Alzheimer disease: A magnetic resonance imaging study. Arch Neurol 2003;60:393–8.
23. Zhou CN, Chao FL, Zhang Y, et al. Sex differences in the white matter and myelinated fibers of APP/PS1 mice and the effects of running exercise on the sex differences of AD mice. Front Aging Neurosci 2018;10:243.
24. Bao C, He C, Shu B, et al. Aerobic exercise training decreases cognitive impairment caused by demyelination by regulating ROCK signaling pathway in aging mice. Brain Res Bull 2021;168:52–62.
25. Sack M, Lenz JN, Jakovcevski M, et al. Early effects of a high-caloric diet and physical exercise on brain volumetry and behavior: a combined MRI and histology study in mice. Brain Imaging Behav 2017;11:1385–96.
26. McKenzie IA, Ohayon D, Li H, et al. Motor skill learning requires active central myelination. Science (80- ) 2014;346:318–22.
27. Fujita Y, Yamashita T. Axon growth inhibition by RhoA/ROCK in the central nervous system. Front Neurosci 2014;8:338.
28. Baer AS, Syed YA, Kang SU, et al. Myelin-mediated inhibition of oligodendrocyte precursor differentiation can be overcome by pharmacological modulation of Fyn-RhoA and protein kinase C signalling. Brain 2009;132:465–81.
29. Bashiri H, Enayati M, Bashiri A, Salari AA. Swimming exercise improves cognitive and behavioral disorders in male NMRI mice with sporadic Alzheimer-like disease. Physiol Behav 2020;223:113003.
30. Horowitz AM, Fan X, Bieri G, et al. Blood factors transfer beneficial effects of exercise on neurogenesis and cognition to the aged brain. Science (80- ) 2020;369:167–73.
31. Um HS, Kang EB, Leem YH, et al. Exercise training acts as a therapeutic strategy for reduction of the pathogenic phenotypes for Alzheimer’s disease in an NSE/APPsw-transgenic model. Int J Mol Med 2008;22:529–39.
32. Wrann CD, White JP, Salogiannnis J, et al. Exercise induces hippocampal BDNF through a PGC-1α/FNDC5 pathway. Cell Metab 2013;18:649–59.
33. Teng HK, Teng KK, Lee R, et al. ProBDNF induces neuronal apoptosis via activation of a receptor complex of p75NTR and sortilin. J Neurosci 2005;25:5455–63.
34. Wang M, Xie Y, Qin D. Proteolytic cleavage of proBDNF to mBDNF in neuropsychiatric and neurodegenerative diseases. Brain Res Bull 2021;166:172–84.
35. Reuland CJ, Church FC. Synergy between plasminogen activator inhibitor-1, α-synuclein, and neuroinflammation in Parkinson’s disease. Med Hypotheses 2020;138:109602.
36. Ding Q, Ying Z, Gómez-Pinilla F. Exercise influences hippocampal plasticity by modulating brain-derived neurotrophic factor processing. Neuroscience 2011;192:773–80.
37. Park HS, Park SS, Kim CJ, Kim TW, Kim TW. Exercise alleviates cognitive functions by enhancing hippocampal insulin signaling and neuroplasticity in high-fat diet-induced obesity. Nutrients 2019;11:1603.
38. Houlton J, Abumaria N, Hinkley SFR, Clarkson AN. Therapeutic potential of neurotrophins for repair after brain injury: A helping hand from biomaterials. Front Genet 2019;10.
39. Levin F, Ferreira D, Lange C, et al. Data-driven FDG-PET subtypes of Alzheimer’s disease-related neurodegeneration. Alzheimer’s Res Ther 2021;13:49.
40. Ismail M-A-M, Mateos L, Maioli S, et al. 27-Hydroxycholesterol impairs neuronal glucose uptake through an IRAP/GLUT4 system dysregulation. J Exp Med 2017;214:699–717.
41. Lauretti E, Iuliano L, Praticò D. Extra-virgin olive oil ameliorates cognition and neuropathology of the 3xTg mice: role of autophagy. Ann Clin Transl Neurol 2017;4:564–74.
42. Qosa H, Mohamed LA, Batarseh YS, et al. Extra-virgin olive oil attenuates amyloid-β and tau pathologies in the brains of TgSwDI mice. J Nutr Biochem 2015;26:1479–90.
43. Pandey KB, Rizvi SI. Plant polyphenols as dietary antioxidants in human health and disease. Oxid Med Cell Longev 2009;2:270–8.
44. Peng Y, Hou C, Yang Z, et al. Hydroxytyrosol mildly improve cognitive function independent of APP processing in APP/PS1 mice. Mol Nutr Food Res 2016;60:2331–42.
45. Grewal R, Reutzel M, Dilberger B, et al. Purified oleocanthal and ligstroside protect against mitochondrial dysfunction in models of early Alzheimer’s disease and brain ageing. Exp Neurol 2020;328:113248.
46. Reutzel M, Grewal R, Silaidos C, et al. Effects of long-term treatment with a blend of highly purified olive secoiridoids on cognition and brain ATP levels in aged NMRI mice. Oxid Med Cell Longev 2018;2018:4070935.
47. Grossi C, Rigacci S, Ambrosini S, et al. The Polyphenol Oleuropein Aglycone Protects TgCRND8 Mice against Aß Plaque Pathology. PLoS One 2013;8:e71702.
48. Luccarini I, Grossi C, Rigacci S, et al. Oleuropein aglycone protects against pyroglutamylated-3 amyloid-ß toxicity: Biochemical, epigenetic and functional correlates. Neurobiol Aging 2015;36:648–63.
49. Van Wijk N, Broersen LM, De Wilde MC, et al. Targeting synaptic dysfunction in Alzheimer’s disease by administering a specific nutrient combination. J Alzheimer’s Dis 2014;38:459–79.
50. Koivisto H, Grimm MO, Rothhaar TL, et al. Special lipid-based diets alleviate cognitive deficits in the APPswe/PS1dE9 transgenic mouse model of Alzheimer’s disease independent of brain amyloid deposition. J Nutr Biochem 2014;25:157–69.
51. Wiesmann M, Jansen D, Zerbi V, et al. Improved spatial learning strategy and memory in aged Alzheimer AβPPswe/PS1dE9 mice on a multi-nutrient diet. J Alzheimer’s Dis 2013;37:233–45.
52. Jansen D, Zerbi V, Arnoldussen IAC, et al. Effects of Specific Multi-Nutrient Enriched Diets on Cerebral Metabolism, Cognition and Neuropathology in AβPPswe-PS1dE9 Mice. PLoS One 2013;8:e75393.
53. Jansen D, Zerbi V, Janssen CIF, et al. Impact of a multi-nutrient diet on cognition, brain metabolism, hemodynamics, and plasticity in apoE4 carrier and apoE knockout mice. Brain Struct Funct 2014;219:1841–68.
54. Broersen LM, Kuipers AAM, Balvers M, et al. A Specific multi-nutrient diet reduces alzheimer-like pathology in young adult aβPPswe/PS1dE9 Mice. J Alzheimer’s Dis 2013;33:177–90.
55. Rai SP, Krohn M, Pahnke J. Early Cognitive Training Rescues Remote Spatial Memory but Reduces Cognitive Flexibility in Alzheimer’s Disease Mice. J Alzheimer’s Dis 2020;75:1105–34.
56. Jiang X, Chai GS, Wang ZH, et al. Spatial training preserves associative memory capacity with augmentation of dendrite ramification and spine generation in Tg2576 mice. Sci Rep 2015;5:9488.
57. Billings LM, Green KN, McGaugh JL, LaFerla FM. Learning decreases Aβ*56 and tau pathology and ameliorates behavioral decline in 3xTg-AD mice. J Neurosci 2007;27:751–61.
58. Daumas S, Sandin J, Chen KS, et al. Faster forgetting contributes to impaired spatial memory in the PDAPP mouse: Deficit in memory retrieval associated with increased sensitivity to interference? Learn Mem 2008;15:625–32.
59. Harrison FE, Hosseini AH, McDonald MP. Endogenous anxiety and stress responses in water maze and Barnes maze spatial memory tasks. Behav Brain Res 2009;198:247–51.
60. Nithianantharajah J, Hannan AJ. Enriched environments, experience-dependent plasticity and disorders of the nervous system. Nat Rev Neurosci 2006;7:697–709.
61. Verret L, Krezymon A, Halley H, et al. Transient enriched housing before amyloidosis onset sustains cognitive improvement in Tg2576 mice. Neurobiol Aging 2013;34:211–25.
62. Polito L, Chierchia A, Tunesi M, et al. Environmental enrichment lessens cognitive decline in APP23 mice without affecting brain sirtuin expression. J Alzheimer’s Dis 2014;42:851–64.
63. Beauquis J, Pavía P, Pomilio C, et al. Environmental enrichment prevents astroglial pathological changes in the hippocampus of APP transgenic mice, model of Alzheimer’s disease. Exp Neurol 2013;239:28–37.
64. Wolf SA, Kronenberg G, Lehmann K, et al. Cognitive and Physical Activity Differently Modulate Disease Progression in the Amyloid Precursor Protein (APP)-23 Model of Alzheimer’s Disease. Biol Psychiatry 2006;60:1314–23.
65. Andrieu S, Coley N, Lovestone S, Aisen PS, Vellas B. Prevention of sporadic Alzheimer’s disease: Lessons learned from clinical trials and future directions. Lancet Neurol 2015;14:926–44.
66. Hodge RD, Bakken TE, Miller JA, et al. Conserved cell types with divergent features in human versus mouse cortex. Nature 2019;573:61–8.
67. Scearce-Levie K, Sanchez PE, Lewcock JW. Leveraging preclinical models for the development of Alzheimer disease therapeutics. Nat Rev Drug Discov 2020;19:447–62.
68. Gupta V, Chitranshi N, You Y, et al. Brain derived neurotrophic factor is involved in the regulation of glycogen synthase kinase 3β (GSK3β) signalling. Biochem Biophys Res Commun 2014;454:381–6.
69. Llorens-Martín M, Jurado J, Hernández F, Ávila J. GSK-3β, a pivotal kinase in Alzheimer disease. Front Mol Neurosci 2014;7:46.
70. Boyle PA, Yu L, Leurgans SE, et al. Attributable risk of Alzheimer’s dementia attributed to age-related neuropathologies. Ann Neurol 2019;85:114–24.
71. Tijms BM, Gobom J, Reus L, et al. Pathophysiological subtypes of Alzheimer’s disease based on cerebrospinal fluid proteomics. Brain 2020;143:3776–92.
72. Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): A randomised controlled trial. Lancet 2015;385:2255–63.
73. Andrieu S, Guyonnet S, Coley N, et al. Effect of long-term omega 3 polyunsaturated fatty acid supplementation with or without multidomain intervention on cognitive function in elderly adults with memory complaints (MAPT): a randomised, placebo-controlled trial. Lancet Neurol 2017;16:377–89.
74. van Charante EPM, Richard E, Eurelings LS, et al. Effectiveness of a 6-year multidomain vascular care intervention to prevent dementia (preDIVA): a cluster-randomised controlled trial. Lancet 2016;388:797–805.
75. Solomon A, Handels R, Wimo A, et al. Effect of a multidomain lifestyle intervention on estimated dementia risk. J Alzheimer’s Dis 2021;82:1461–6.
76. Marengoni A, Rizzuto D, Fratiglioni L, et al. The Effect of a 2-Year Intervention Consisting of Diet, Physical Exercise, Cognitive Training, and Monitoring of Vascular Risk on Chronic Morbidity—the FINGER Randomized Controlled Trial. J Am Med Dir Assoc 2018;19:355-360.e1.